Electrophysiological investigations of the retina tap into the electrical activity of the cells in the retina, particularly the neurons that transmit the signals they receive from the photoreceptors to the brain. This electrical activity is influenced by various signaling pathways and reflects on the retina's developmental stage, circuitry, and viability. The knowledge gained from electrophysiological interrogation of the retina is important for understanding underlying mechanisms of retinal function and connectivity. Non-invasive studies of the electrophysiology of the retina involve techniques such as electroretinography (ERG), which measures the electrical response of the retina to light stimulation, and pattern electroretinography (PERG), which assesses the response of the retina to specific visual patterns (Creel, 2019; Cornish et al., 2021). These tests can be used to evaluate the function of the retina and to diagnose various retinal disorders, such as macular degeneration and retinitis pigmentosa, among others (Moschos and Nitoda, 2018; Menghini et al., 2020). ERG is widely used on animal models in basic and applied research (Vinberg and Kefalov, 2015; Pasmanter and Petersen-Jones, 2020). Despite their many benefits, non-invasive approaches suffer from inherently low resolution.

High-resolution insight into retina electrophysiology can be gained with ex-vivo retinas (retinas that have been surgically removed from an animal and maintained in vitro) as a model system for vision research. There are several benefits regarding ex-vivo retinas as a model system: Ex-vivo retinas retain their in vivo organization and structure, including the layers of cells and their connections, energy metabolism, gradients of electrolytes, and amino acids (Ames and Nesbett, 1981). This relative integrity allows the study of the behavior of retinal cells in a physiologically relevant environment. Ex-vivo retinas can be used to study the structure, function of the retina, and its response to different stimuli, such as light, drugs, and electrical stimulation (Field et al., 2010; Madugula et al., 2022; Shah et al., 2022). Ex-vivo retinas can be used to study retinal diseases and their effects on retinal function and structure, thus providing important information needed in the development of new treatments for retinal degenerative diseases (Chang et al., 2019; Chenais et al., 2019; Tong et al., 2020). However, ex-vivo retinas have limitations as a model system. For example, they may not sustain their full functionality over time, due to environmental changes and the effects of tissue dissection. High-resolution strategies to study the electrophysiology of the intact retina are therefore appealing to explore the differences between the intact and ex-vivo retinas (Vėbraitė and Hanein, 2022).

Although studies on soft neural probes exist, none to the best of our knowledge, documented bi-directional high-resolution electrophysiological investigation of the intact retina. Several studies described flexible electrode arrays that are potentially suitable for bidirectional retina stimulation and recording, presenting preliminary testing with ex-vivo retinal model (not in-vivo) or chronic biostability testing (Rodger et al., 2008; Montes et al., 2019). Others discussed high-density ultra-flexible electrode arrays for chronic cortical recordings (Zhao et al., 2023). Nevertheless, proven high-resolution tools to study the electrophysiology of the intact retina are presently scarce and only a few electrophysiological investigations of the intact retina were described. We refer the interested reader to a mini review on this specific topic (Vėbraitė and Hanein, 2022). In fact, when reviewing the entire expansive literature on soft neural technologies, it is apparent that these technologies manifest either high-quality stimulation or recording (Vėbraitė and Hanein, 2021), not both.

Multi-electrode arrays designed for the bi-directional study of the intact retina demand soft probes with small electrode dimensions that can record and stimulate the retina at high fidelity. The probe properties should include bio-compatibility, durability, and sterilization compatibility. Such retinal probes should not cause damage to the gentle tissue, and for some applications, optical clarity is important (Vėbraitė and Hanein, 2021; Zheng et al., 2021; Oldroyd and Malliaras, 2022). Finally, electrode material has to exhibit stability during stimulation (Schiavone et al., 2020).

In this study, we introduce an innovative method to fulfill these requisites. The focus of this paper is the development of a screen-printed soft multi-electrode array for bi-directional neural interfacing. To showcase the potential of this technology as a neural interface, we illustrate its application in studying the intact retina.

Flexible high-resolution neural probes are conventionally realized using photolithography and thin film technology, which are ideal for high-resolution device fabrication (Herwik et al., 2009; Boehler et al., 2017). However, under physiological conditions, these probes tend to suffer from poor stability, delamination (Boehler et al., 2017), and relatively high rigidity, imposed by substrate material (e.g., polyimide) and compromised electrochemical properties. Moreover, soft neuronal electrodes tend to manifest poor recording performances (Vėbraitė and Hanein, 2021). The approach we describe here builds on stacked screen-printed carbon on soft polyurethane (PU) films. Polyurethane, as a substrate and passivation material, offers low Young's modulus (in the 7–30 MPa range), low water diffusion coefficient (3.18 × 10⁻¹⁰m²s⁻¹), breathability, compatibility with sterilization methods, and adherence to ISO 10993-1 toxicity standards (Rezaei et al., 2010; Vėbraitė and Hanein, 2021; Materials and Technologies, 2022). A screen-printed carbon-based approach resolves several limitations of photolithography-based thin-film probes: Carbon electrodes are characterized by high stability, they are non-Faradaic (within a range allowing safe stimulation), while also having favorable electrochemical properties to enable high-resolution recording.

Our previous work in Vėbraitė and Hanein (2022) demonstrated the feasibility of capturing high-quality spontaneous and light-induced signals from ex-vivo retinal tissue using soft carbon probes, an improvement over traditional titanium nitride (TiN) multi-electrode arrays as shown in Vėbraitė et al. (2021). Expanding on this technological foundation, we have taken a step forward by describing a soft multi-electrode interface with 36 channels, enabling both the recording and stimulation of the intact retina in a two-way manner. The fabrication process relies on the layering of two layers of screen-printed carbon, resulting in stable electrode arrays. To enhance the electrochemical attributes of the electrodes, we incorporated plasma polymerization of EDOT.

The study also encompasses electrochemical measurements, which were conducted to characterize the properties of these electrodes. Furthermore, a feasibility study to showcase the practical application of these devices for the simultaneous recording and stimulation of the intact chick retina is presented. This approach reveals responses within the intact retina, most notably an enhancement in stability over extended periods of stimulation and recording. Finally, our study underscores distinctions between the conditions of an intact retina and an ex-vivo retina, shedding light on the implications of each setting.

Soft electrode arrays (SoftC) were realized using screen printing of carbon ink on soft and thin polyurethane (PU) films. Given the relatively low resolution of screen printing, stacking multiple layers was used to achieve a high electrode count (Figure 1a). Previously, we presented arrays with only eight electrodes (Vėbraitė and Hanein, 2022). In this study, we increased the electrode number to 36 electrodes. We reduced the electrode diameter from 80 to 50 μm and inter-channel distance from 430 to 150 μm, achieving higher spatial resolution and electrode count (Figure 1a). The overall dimensions of the soft array probe are 6 mm × 2.6 cm × 160 μm (width, length, total thickness, respectively) and effective array area of ~ 1.13 mm² incorporating 32 recording-stimulating micro-electrodes, two additional stimulating electrodes, and two big reference electrodes. The final SoftC probe is shown in Figure 1b, consisting of a soft part (with the carbon electrodes), a flexible PCB, and a high-density connector. The electrodes in Layer 2 have a circular shape, as defined by the laser cut hole, and electrodes in Layer 1 have a slightly deformed shape (Figure 1b inset). Electrode's approximate diameter and area were estimated and averaged over five SoftC probes from optical microscopy images (Supplementary Figure 2B), resulting in an overall average diameter of 58.9 μm (SD = 10.7) and an area of 1,475 μm² (SD = 349). Environmental scanning electron microscope (ESEM) imaging was used to examine the electrode surface. Figure 1c shows ESEM images depicting a rough carbon surface, which is important for enhancing an effective electrode surface area. Additionally, a confocal scanning laser microscopy of the carbon on PU film (Supplementary Figure 4) was used to derive a surface roughness of ~1 μm and a height of ~11 μm. Contact angle measurements of the electrode surface (Supplementary Figure 5) demonstrate angles of 32.5 (SD = 2.4) and 81 (SD = 4) for ppEDOT-coated and bare electrodes, respectively. A complete device introduced into an enucleated chick eye is shown in Figure 1d.

Electrochemical measurements were performed to validate electrode performances, including non-Faradaic behavior, specific DC capacitance, impedance, charge injection, and stability (Figure 2). Cyclic voltammetry (CV) of screen-printed electrodes was first used to validate the non-Faradaic nature of the electrodes and to characterize the water window. The low-frequency capacitance of the electrodes was also extracted (Figures 2A, B). In particular, CV measurements reveal the effect of ppEDOT coating. Screen-printed electrodes coated with ppEDOT obtained a specific capacitance of 0.92 mF/cm² (compared with 0.09 mF/cm² for bare electrodes), averaged over five samples each. The high specific capacitance values are important for low-noise electrical recordings at small electrode dimensions. Figure 2A shows that the boundaries of the electrochemical potential window for ppEDOT coated electrode in PBS are −0.5 and +0.5 V (Figure 2A). CV measurements also reveal a capacitive behavior and water window values typical to carbon/ppEDOT electrodes. The cathodal and anodal charge storage capacities (CSC) were calculated by the time integral of the cathodal and anodal current over a potential range determined by a water electrolysis window. Values are presented in the Supplementary Figure 3 for both electrodes. CSC was evaluated to be 2.7 and 2.0 mC/cm² for coated and uncoated carbon electrodes, respectively.

Charge transfer tests were carried out to categorize the stimulating electrodes in terms of their reversibility and maximum charge injection capacity (Figure 2C). The charge injected in the first half-phase is seen to be immediately compensated for by reversing the polarity over the second half. A charge injection test with a specific capacitance of 0.12 mC/cm² (75 μA, 50 ms) shows a potential increase in the 200 mV range at the electrode interface.

To validate the stability of the electrode under electrical stimulation, the potential above the electrode was also measured using a glass-pulled capillary electrode mounted on a motorized stage. Voltage transient V_ts values at different charge densities show an almost linear increase through the 0–2 mC/cm² range. The values of the cathodic peak are similar to those measured at the same distance with commercial 30 μm diameter TiN MEAs (Supplementary Figure 6). In a prior study, screen-printed carbon electrodes on polyurethane films were shown to be compliant with EtOH and ultraviolet (UV) sterilization methods as well as stable in aCSF solution (at 40°C) for over 6 months with no delamination or change in the resistance values (Vėbraitė et al., 2021). Here, the impedance stability of the electrodes in the aCSF solution was validated (Figure 2G) over 120 days. Impedance measurements at the 0.025–5 kHz range on day 66 showed stable values (274 kΩ at 1 kHz).

Electrical activity in the intact retina in response to electrical stimulation was studied and is presented in Figure 3. A developing chick retina at embryonic day 14 (E14) was used as a blind retina model for these studies (N = 5 retinas). Chick retina at E10–E18 has developed and functional ganglion cell layer as well as inner and outer layers but an undeveloped photoreceptor layer with no responses to light until E19 (Mey and Thanos, 2000; Sernagor et al., 2001). The soft probe was placed inside an open blind chick eye (E14) mounted on a custom-made holder (Figure 3A). The eye was kept submerged in oxygenated aCSF (for as long as 3 h).

Electrode impedance measurements were first performed in aCSF and after placement against the retina to validate tissue-electrode coupling (Spira and Hai, 2013; Majdi et al., 2014). The measured impedance increased or remained the same. In six channels (out of 16 measured) impedance values (at 1 kHz) increased by 4–50% from the baseline value measured in aCSF (Figure 3B), indicating coupling. Bi-directional measurements were then used to study retina responses to electrical stimulation. For all SoftC probe stimulations of the intact retina, 300 μs charge balance biphasic pulses were used at either 1 or 10 Hz frequency. Figure 3C illustrates recorded traces contaminated with stimulation artifact, thus it was removed during offline analysis. The filtered and artifact-removed (see Section 2) data reveal a clear retina response to electrical stimulation. Both direct (Figure 3F) and indirect activation (Figures 3C–E) were observed. The first is typified by short, high-amplitude responses that increase with increasing current amplitude. Indirect activation is characterized by low-amplitude and long latency responses. We note that retina responses were clearly visible in those channels with improved coupling (e.g., channels 15, 20, and 8 with 6, 25, and 46% increase in impedance, respectively). These results suggest that despite the curved nature of the intact retina, the soft probe achieved good electrical coupling, especially in those electrodes with high signal-to-noise ratio responses.

Finally, we compare the indirect responses (long latency) in the intact retina with ex-vivo conditions. In the intact retina, responses appear to remain stable upon repetitive stimulation lasting for 10 s (Figures 3D, 4A, B). Similar stimulation in ex-vivo conditions is typified by clear desensitization, especially with higher stimulation frequencies (Jensen and Rizzo, 2007; Chenais et al., 2021; Li et al., 2022), and a strong sensitivity to spontaneous activity, which completely abolish the response to stimulation. Figure 4A presents an example of a spontaneous wave direct inhibition of retina responses to electrical stimulation in the ex-vivo retina, whereas in the intact retina, such an effect was not detected. We observed almost no spontaneous activity [Figure 4A (right)] in the intact retina experiments (N = 5 retinas). This is, to the best of our knowledge, the first direct electrophysiological evidence for a discrepancy between ex-vivo and intact retina conditions.

Soft electrode arrays, engineered for bi-directional electrical interfacing with the retina, were realized and put through testing. The probes were realized by stacking two screen-printed soft layers. We were able to increase the electrode count and realize electrode arrays with good recording and stimulation performances, even when positioned against the intact retina. Various flexible and soft probes were investigated in the past. We briefly discuss a few representative examples. In Graudejus et al. (2009), 11 recording electrodes were realized using photolithography and electrodeposition on 280 μm thick layer of an elastomeric silicone substrate. Silicon nanomembrane transistor arrays on polyimide (few μm thick, Young's modulus 1.3–3 GPa) were used to form dense arrays of thousands of amplified and multiplexed sensors (Viventi et al., 2011). The use of a grid structure was also used to help reduce the rigidity of a recording device as demonstrated in Reference (Khodagholy et al., 2015). These and many other approaches have not demonstrated high-density recording and stimulation with thin and soft support. Transitioning from rigid or flexible substrates to soft ones usually comes with a cost in performance. For example, CNT micro-electrodes (30 μm in diameter) on Si/SiO₂, can reach high specific capacitance in the range of 3–10 mF/cm² (Gabay et al., 2007). Values obtained with state-of-the-art Titanium Nitride (TiN) on glass are in the 2 mF/cm² (Gabay et al., 2007). CNT micro-electrodes on soft support will typically have a specific capacitance in the range of 1–2 mF/cm². ECoG probe using CNTs on PDMS electrodes exhibits specific capacitance of 1.5 mF/cm², low impedance values of approximately 30 kΩ at 1 kHz, and charge storage capacity of 0.35 mC/cm² (Yang et al., 2022). Faradaic silver nanoparticles/PEDOT:PSS [Poly(3,4-ethylenedioxythiophene) Polystyrene Sulfonate] electrodes on polyimide can reach a very low impedance value of 200 Ω (at 1 kHz), exhibit capacitance of 0.4–60.6 mF/cm² and charge storage capacity of 1.083–2.8 C/cm² by increasing PEDOT:PSS coating from 1 to 10 layers (Almaris et al., 2020). The soft and thin electrode arrays we presented here achieve good recording and stimulation performances despite their very simple and straightforward fabrication process and as such represent an advantage compared with alternative approaches.

These pliable probes are particularly adapted to conform to the curvature of the eye's retina, providing a distinctive chance to investigate the retina in its undisturbed state. Using the SoftC technology, we demonstrated, for the first time, electrophysiological differences associated with retina conditions (i.e., ex-vivo vs. intact retina). Differences associated with spontaneous activity were observed thanks to the ability to record directly from the intact retina. Differences associated with evoked responses were facilitated by the bi-directional ability presented here.

Discrepancies between in vitro, ex-vivo, and intact tissues are well documented (Alaylioğlu et al., 2020). Ex-vivo retinas, in particular, are expected to show some structural damage and decreased functional responses compared to intact retinas, owing to mechanical manipulation and the loss of optimal biochemical support. These differences imply that the ex-vivo preparation may not accurately reflect the in vivo state of the retina. Among various expected effects, these differences are expected to profoundly affect retina electrophysiology.

One such effect is spontaneous waves. Retina spontaneous activity is a well-established phenomenon of developing and degenerating retinas and is known to interfere with retina stimulation strategies (Wong et al., 1998; Goo et al., 2016; Haselier et al., 2017). In a previous study (Vėbraitė et al., 2021), we presented bio-potential measurements from intact retinas. In those measurements, spontaneous activity was observed in only 2 out of 11 intact eyes. On those rare appearances, the recorded spontaneous activity was shorter in duration and had a low signal-to-noise ratio compared to that typically observed ex-vivo. In our previous study, the SoftC probes did not have electrical stimulation capability, and the observation of spontaneous retinal activity was an indicator of good probe-retina interface and retina viability. In this study, we report almost no spontaneous activity in the intact retina (N = 5 retinas). Importantly, in the present study, the ability to electrically stimulate retina responses serves as a powerful tool to validate retina viability and to provide greater confidence in asserting the lack of spontaneous activity in intact retina.

Studying retina responses to electrical stimulation is pivotal in the domain of artificial vision (also known as retinal prosthesis or a bionic eye). Electrical stimulation works by using electrodes implanted in the retina to stimulate remaining healthy cells and transmit visual information to the brain (Bloch et al., 2019; Ayton et al., 2020). For individuals with severe vision loss, electrical stimulation of the retina can provide a means of regaining functional vision, allowing them to perform basic daily activities, such as recognizing faces, reading, and navigating the environment (Castaldi et al., 2016; Hallum and Dakin, 2021). Electrical stimulation of the retina has shown promising potential in treating certain forms of visual impairments, particularly retinal degenerative diseases such as retinitis pigmentosa and age-related macular degeneration (Chow et al., 2002; Palanker et al., 2022). There are several lingering challenges related to electrical stimulation of the retina, primarily: Stability, efficacy, safety, and feedback (Ryu et al., 2019). Having demonstrated the intact retina responses to electrical stimulation, it is interesting to discuss these results focusing on stability, efficacy, and safety in comparison to the extensively published ex-vivo retina data.

In this investigation, we were able to observe both direct and indirect responses in the intact retina. Interestingly, the responses observed in the intact retina remain stable, even after extensive stimulation (Figures 3D, 4B). Similar stimulation in ex-vivo retina is typified by clear saturation (Chenais et al., 2021). We hypothesize that the origin of this interesting discrepancy is the physiological state of the retina, which gives rise to better stability of the intact retina under what appears to be better conditions. In the realm of retina stimulation, high efficacy stands for the ability to achieve local stimulation with high temporal resolution. We expect that ganglion and bipolar cell response to stimulation, and in particular, its efficacy will be affected by whether retina circuits are intact or have lost their natural support. These issues will be further discussed and analyzed in a separate publication.

A major concern in electrical neural stimulation is safety. Moreover, the current or charge delivered has to be minimal. Identifying optimal parameters for high-fidelity cell activation is still an active research question (Colodetti et al., 2007; Madugula et al., 2022). Retina activation (direct or indirect) effectiveness is influenced by stimulation parameters such as waveform shape, amplitude, duration, frequency, and polarity (Tong et al., 2020). For example, some studies point out the advantages of monophasic stimulation while others suggest that biphasic stimulation shows a lower activation threshold and shorter response latency. Generally, biphasic stimulation is widely used for retinal implants (Jensen and Rizzo, 2009; Boinagrov et al., 2014; Jalligampala et al., 2017; Celik and Karagoz, 2018; Meng et al., 2018). Another safety concern is oxidative stress. Overproduction of reactive oxygen species (ROS) is known to affect the electrode-tissue interface, which can lead to foreign body response, damage to the cells and tissue, and can also create structural damage, and changes in properties of the implanted device (Takmakov et al., 2015; Ereifej et al., 2018). The retina is particularly sensitive to ROS, which can lead to a range of retinal disorders in vivo, including age-related macular degeneration (AMD), diabetic retinopathy, and retinitis pigmentosa, among others (Wang et al., 2022). Thus, the retina exhibits several protective mechanisms, including the presence of antioxidants such as glutathione, vitamins, and taurine as well as enzymes such as superoxide dismutase and catalase. Additionally, the retina's high metabolic rate and high oxygen consumption are regulated to minimize the production of ROS (Léveillard and Sahel, 2017; Domènech and Marfany, 2020). It is therefore not too surprising that extensive stimulation ex-vivo could lead to various instabilities, some of which may be associated with the electro-chemistry at the electrode interface unless electrode passivation is explicitly used (Fromherz and Stett, 1995; Eickenscheidt et al., 2012). How more efficient stability is offered by in-vivo conditions is an important topic for further investigations.

An additional clear bottleneck in the field of retina stimulation is the lack of informative feedback on the efficacy of a specific stimulation paradigm. Optimized stimulation parameters (e.g., pulse duration, shape, and rate) are important to reduce threshold and energy, provide more selective stimulation (ganglion vs. bipolar cells), reduce fading, and increase localization. In pre-clinical stages, researchers have to rely on animal behavior, evoked potential responses, or data collected from dissociated retinas. These parameters provide only indirect information. For example, the selectivity between ganglion cell, fiber layer, and bipolar cell stimulation has received much attention (Tong et al., 2020), as it is hypothesized that axon fiber activation contributes to reduced localization. Moreover, reports on optimal stimulation parameters are contradictory. Some studies suggested that particularly long (10 ms) pulses can lead to selective activation of bipolar cells (Boinagrov et al., 2014). While Tong et al. (2019), showed that long pulses do not lead to the desired selectivity. One possible explanation for the discrepancy between reports is electrode geometry. An alternative explanation is the different physiological conditions of the retina. Exploring this effect in the intact retina will help in understanding the role of retina physiology in retina stimulation. Desensitization, owing to rapid stimulation, is another major unresolved issue in neural stimulation. In particular, the effect of pulse trains on ganglion cells vs. bipolar cells. The technique reported here can help alleviate the uncertainty associated with electrical stimulation of the retina and will be used in future explorations.

The probes presented in this study can benefit a wide range of neurotechnological applications, far and beyond the study of the retina. Despite the fabrication simplicity, the arrays are stable while also manifesting high-quality performances for bi-directional electrical interfacing. This can be handy in the realm of cortical electrodes, subdermal implants, and much more.

Despite its numerous advantages, the technology presented in its current form is not without drawbacks. One notable limitation is the yield of the manual printing and layer stacking procedures.

An additional challenge is the probe thickness and lack of permeability. Enhancements in probe design, including reducing the thickness of the device, incorporating perforations to increase oxygenation, and maintaining uniformity throughout the fabrication process, coupled with a more extensive validation process both in laboratory settings and within living organisms, will provide stronger support for establishing the showcased technology as a resilient instrument in upcoming research endeavors.

To summarize, this feasibility study demonstrates a novel approach featuring a high-resolution probe for electrophysiological investigation of the intact retina. We conducted electrical stimulation while simultaneously recording both direct and indirect responses with the same device. Our findings revealed low levels of spontaneous activity and stable responses under high-frequency stimulation compared to ex-vivo retinas. These initial observations suggest distinctions between ex-vivo and intact retina conditions. However, further investigations are warranted to comprehensively explore the underlying mechanisms responsible for these differences.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was approved by Institutional Animal Care and Use Committee at Tel Aviv University. The study was conducted in accordance with the local legislation and institutional requirements.

IV: Conceptualization, Investigation, Writing—original draft, Writing—review & editing. CB-H: Investigation, Writing—review & editing. MD-P: Investigation, Conceptualization, Writing—review & editing. YH: Conceptualization, Funding acquisition, Supervision, Writing—original draft, Writing—review & editing.