Structuring of CPs on the nanoscale can be used to fundamentally alter the performance of the CP electrode coatings. Nanostructuring can take the form of nanoporosity, nanowires or nanotubes (Liu et al., 2008; Ghasemi-Mobarakeh et al., 2009; Xie et al., 2009; Kang et al., 2011). Nanoporosity has been used to increase the surface area of electrodes as well as modulate the drug-release and biosensing properties of CP systems (Luo and Cui, 2009a,b; Guo et al., 2011; Kang et al., 2011; Szultka-Mlynska et al., 2016). Nanoporosity is typically achieved by depositing around a template, such as polystyrene beads, which is removed after deposition. CP nanotubes have increased the CSC and decreased impedance compared to conventional CP coatings. CP nanotubes are typically fabricated by electrospinning or by electrodepositing CP around a nanotube template (Abidian et al., 2006, 2009; Liu et al., 2008; Ghasemi-Mobarakeh et al., 2009; Xie et al., 2009; Prabhakaran et al., 2011). Abidian et al. fabricated PEDOT and PPy nanotubes by electrochemically depositing CP around electrospun fibers of poly(lactic acid) and poly(lactic—co-glycolic acid) (Abidian et al., 2006, 2009, 2010). PEDOT nanotubes were found to have lower impedance and CSC (2.5 ± 1.4 kΩ at 1 kHz and 392 ± 6.2 mC·cm⁻²) compared to iridium (468.8 ± 13.3 kΩ at 1 kHz and 0.1 ± 0.5 mC·cm⁻²). When PEDOT nanotube electrodes were implanted in the barrel cortex of rats, they provided a stable reduction in impedance at 1 kHz over 49 days of implantation from 980 ± 15 kΩ for uncoated gold electrodes down to 521 ± 18 kΩ for PEDOT nanotube coated electrodes (Abidian et al., 2009).

While conducting polymers offer improved electrochemical properties compared to metallic electrodes, their mechanical properties present key limitations in the context of a chronic neuroprosthetic implant device. While conducting polymers are considerably softer than metals with moduli ranging 50 MPa to 8 GPa (Wang and Feng, 2002; Wang et al., 2009; Hassarati et al., 2014), they are still several orders of magnitude stiffer than neural tissue (0.5–1 kPa) and the retina (200–400 kPa). This mechanical mismatch is hypothesized to be a contributing factor to the biological inflammatory response to implantation which results in fibrotic encapsulation, increasing the impedance of the tissue-electrode interface and increasing the distance between the electrode surface and the target cells for electrical stimulation. Furthermore, conventional conducting polymers are brittle and prone to break-up and delamination. Studies have examined the long-term stability of CP coatings, examining accelerated aging, sterilization and long-term stimulation and found that while some CP coatings can be very stable under prolonged use (over 1.2 billion stimulation cycles) others are very dependent on the roughness of the underlying substrate and can fail in as little as 10 days (Boretius et al., 2011; Green et al., 2011, 2012a). This limitation is a considerable challenge to the implementation of CP-based technologies within chronically implanted devices and as such improving the mechanical robustness of CP-based coatings has become a focal point for ongoing research. One approach to overcome these mechanical limitations is the fabrication of CP-based composites, combining the electrochemical functionality CPs with the mechanical properties of softer materials such as hydrogels. Hydrogels are commonly used in tissue engineering because of their mechanical and structural similarity to soft tissue. Conducting hydrogels (CH) can be fabricated by electrochemically depositing a CP throughout a preformed hydrogel matrix, thus creating an interpenetrating network of the two polymer systems. Green et al. has demonstrated that this method can be used to create a composite electrode material that maintains the electrochemical benefits of conventional CP coatings while reducing the modulus from 40 MPa down to 2 MPa (Green et al., 2012b). The encapsulation of CP components within a soft hydrogel matrix also removes the problems of brittle failure and delamination associated with conventional CP coatings. Many CH systems have been reported, a summary is given below in Table 1.

To date, there have been limited reports of the use of conducting polymer electrodes for in vivo stimulation applications. However, early reports have demonstrated that the use of CP-based electrodes greatly improves the performance of stimulating electrodes. A summary of in vivo testing of CP electrodes is given below in Table 2. While in vitro testing of CIL has shown CP-based systems to have a CIL of between 10 and 30 times greater than bare platinum (Green et al., 2012a, 2013a), this has yet to be reliably demonstrated in the in vivo setting.

However, early results demonstrate PEDOT coatings reduces the transient and residual voltage across the electrode under biphasic stimulation, indicating the increase in CIL observed in vitro is carried over to the in vivo setting (Venkatraman et al., 2011; Green et al., 2013a). The increase in CIL conferred by CP coatings may ultimately allow for the reduction of electrode size without compromising device safety and efficacy. Furthermore, the reduced impedance of CP electrodes may allow devices to operate with reduced power requirements. The application of CP coatings to vision prosthesis devices may facilitate the development of arrays with smaller, more densely packed electrodes enabling improved device selectivity and resolution.

CNTs offer several advantages over metal stimulating electrodes. Mainly, the combination of capacitive charge transfer mechanism and high electrochemical surface area, contributing to high charge injection capacity, high specific capacitance, and low interfacial impedance (Gabay et al., 2007; Bareket-Keren and Hanein, 2012; David-Pur et al., 2014). CNTs also provide an effective scaffold for neuronal growth and attachment, and it has been suggested that a mechanical mechanism of neurite entanglement facilitate this strong neuron-CNT affinity (Sorkin et al., 2008; Malarkey and Parpura, 2010; Voge and Stegemann, 2011). CNTs can be readily modified with different bioactive molecules (polymers, peptides, proteins) to improve their biocompatibility (Bekyarova et al., 2005; Bottini et al., 2011). Mattson et al. (2000) were the first to propose the use of CNTs as a substrate for neuronal growth. Since then extensive investigations revealed the suitability of CNTs to support neuronal growth and neurite branching (Mattson et al., 2000; Hu et al., 2004; Matsumoto et al., 2007; Cellot et al., 2009; Chao et al., 2009), to electrically interface with neurons (Gabay et al., 2005; Wang et al., 2006; Mazzatenta et al., 2007; Keefer et al., 2008; Shein et al., 2009; Bareket-Keren and Hanein, 2012), and further toward neural implants (Webster et al., 2003; Gabriel et al., 2009; Fuchsberger et al., 2011; David-Pur et al., 2014).

Recording of APs from RGCs of isolated rabbit retinas using CNT coated Pt electrodes was achieved by Gabriel, et al. (Gabriel et al., 2009). CNT coatings were easily achieved by drop casting CNT suspension onto an electrode and drying. Pt/CNT electrodes exhibited lower interfacial impedance and lower average noise in the recorded signals compared with uncoated Pt electrodes (Gabriel et al., 2009). While using drop casting is simple and useful for proof-of-concept in vitro studies, these coatings may not be suitable for long-term clinical applications due to poor adhesion of CNT to the metal substrate, in particular under sustained electrical stimulation.

Stronger adhesion can be formed by direct growth of CNTs directly onto TiN substrate using chemical vapor deposition (Gabay et al., 2007). Such pristine, high density, 3D CNT microelectrodes (60 and 30 μm in diameter) were used by Shoval et al. (2009) to record and stimulate RGC activity in vitro. Neonatal mouse retinas were mounted onto a CNT multi-electrode array (MEA), and a “Velcro” effect, where CNTs became tightly bound to the tissue was noticeable. Electrical recordings of spontaneous neuronal activity were consistently obtained, exhibiting typical bursting and propagating waves behavior. The SNR of the recorded signals was by up to three times higher compared to that of porous TiN electrodes. Moreover the amplitude of the signals gradually increased over the duration of the experiment (hours), implying an improvement in cell-to-electrode coupling (not observed with TiN arrays) (Shoval et al., 2009). Similar effects, as well as a gradual decrease in stimulation thresholds and an increase in cellular recruitment were observed in a later study by the same group (Eleftheriou et al., 2017). Here the investigators used explants of dystrophic mouse retinas (outer retina degeneration) mimicking the case of epiretinal implantation. The extracted retinas were flattened either on CNT electrodes or on isolated CNT islands and changes to the morphology of the bio-hybrid retina-CNT composite, and to the activity of ganglion cells were examined over a period of up to 3 days (Eleftheriou et al., 2017). Progressive integration of CNT structures into the inner retina without initiating gliosis response was observed. The authors noted that vitreous residues may compromise this effect. It was suggested that digestion of portions of the vitreous and of the internal limiting membrane prior to implantation of the electrodes using glycosidic enzymes could be applied in future bio-hybrid retinal arrays. This strong CNT-retina adhesion could potentially solve the fundamental problem with fixation of epiretinal arrays. Moreover, intracellular activation could be achieved by penetration of CNTs into individual RGCs, thus substantially lowering the amount of charge required for stimulation.

Flexible CNT MEAs, based entirely on CNT embedded in different polymeric supports (medical tape, PDMS, PI and Parylene) were also investigated to support mechanical compatibility with the tissue in contact. David-Pur et al. (2014) realized such flexible CNT arrays through a unique fabrication process based on growing loosely bound high-density CNT patterns on a silicon dioxide substrate, and transferring them directly onto a flexible support through a peel-off process (David-Pur et al., 2013, 2014). Embryonic chick retinas were flattened on these flexible arrays, with RGC layer facing down, as in an epi-retinal implant. Spontaneous activity waves were recorded and stimulation was achieved with a threshold of 4 nC. Validation of the synaptic processes was also demonstrated through silencing of RGC activation using synaptic blockers (David-Pur et al., 2014).

Nanodiamonds were also recently proposed as neuroelectrode material (Hadjinicolaou et al., 2012; Bendali et al., 2015; Ahnood et al., 2017). NCD becomes electrically conductive providing it is doped with nitrogen or trimethyl boron (Garrett et al., 2012; Kiran et al., 2012). Both boron and nitrogen doped diamond coatings show promising biocompatibility and suitability for neural stimulation neurons (Chen et al., 2010; Bendali et al., 2014). To increase the double layer capacitance and reduce the impedance, nitrogen doped ultra-nanocrystalline diamond (N-UNCD) electrodes were coated with Pt or EIROF, or alternatively activated via an electrochemical process (Garrett et al., 2012). For similar effects, boron doped diamond (BDD) electrodes were enhanced through a topographical modification by deposition of the NCD onto vertically aligned CNTs as a template interlayer, resulting with CIC of up to 10 mCcm⁻² (measured in vitro in saline; Hébert et al., 2014; Piret et al., 2015).

Researchers from University of Melbourne have recently reported the development of an epiretinal system based on NCD technology. High-density monolithic integration of conductive N-UNCD directly onto insulating polycrystalline diamond enables construction of hermetic all-diamond arrays (Hadjinicolaou et al., 2012; Ganesan et al., 2014; Ahnood et al., 2017). The all-diamond electrode array consists of 256 microelectrodes (120 × 120 μm, and pitch of 150 μm) directly deposited onto a sheet of polycrystalline diamond (Ahnood et al., 2017). The authors noted that the fabrication process allows the pitch to be reduced by one order of magnitude (up to ~15 μm) potentially permitting thousands of electrodes within a 2 by 2 mm array. The Suitability of N-UNCD microelectrodes for stimulation of retinal neurons was assess in vitro using explanted rat retinas (Hadjinicolaou et al., 2012). APs were recorded from single ganglion cells adjacent to the stimulating channel using patch clamp (Hadjinicolaou et al., 2012). The biocompatibility of various forms of diamond was verified in vitro using cultures of rat cortical neurons (Tong et al., 2014) and in vivo when implanted into the back muscle of guinea pigs for 4–15 weeks (Garrett et al., 2016). Next, the compatibility of N-UNCD with retinal tissue was examined in vivo. Discs of 40 μm thick N-UNCD were subretinally implanted in the eyes of blind rat (model for RP) for 3 months. Fundus imaging and immunohistochemical studies showed that the retina tissue remained intact with no significant difference between sections examined away from the implant (control) and next to it, thus validating that N-UNCD is non-cytotoxic and biochemically stable for the duration of the study.

Increasing the resolution of retinal implants by using BDD/protein electrode coatings that specifically repel the adhesion and growth of glia cells while promoting the attachment of neurons was recently proposed by Bendali et al. (2014). Retinal implants are often in contact with a surface glial layer that may compromise the effective resolution of the device (Bendali et al., 2014). Neuronal electrodes are often end up being in contact with a surface sealing glial layer rather than with the neurons, leading to a gradual loss of function (Rousche and Normann, 1998; Maynard et al., 2000). This is the case for both retinal (specifically, epiretinal, and subretinal) and cortical visual implants. In the study by Bendali et al. (2014), adult retinal cell cultures were grown on either pristine of protein coated BDD electrodes. Glial cells growth was enhanced on the protein coated BDD whereas bipolar cells grew equally well on both pristine and coated BDD. In addition it was possible to engineer the adhesion of the protein to the BDD through different terminations either enhanced (Hydrogen-terminated) or prohibited (oxygen-terminated) protein coating. The authors suggested a design where a protein coated base will promote glial cell growth thus repelling them from the non-coated penetrating electrode tip (Bendali et al., 2014).

Next, flexible NCD electrode arrays were realized through a peel-off process using flexible substrates deposited on a 3D mold. The conducting material (either BDD or Pt) was deposited inside the groves formed by the molding process (Bendali et al., 2015). Three flexible prototypes were examined: polyimide, polyimide/BDD and polyimide/Pt. The implants were then inserted into the subretinal space of P23H rats, a model of retinal degeneration. Degenerated retinas could “mold” themselves into the inside of the wells, thereby isolating bipolar neurons for specific, independent stimulation (Bendali et al., 2015). The investigators further proposed a model for predicting the visual acuity that can be achieved using planar and 3D NCD implants with a distant ground or with a ground grid, thus demonstrating the potential of 3D designs for increasing the resolution of retinal implants (Bendali et al., 2015).

A hybrid optoelectrical retinal prosthesis based on light sensing Si NWs is being developed by Silva et al. (Khraiche et al., 2011, 2013; Ha et al., 2016). The system comprises of an inductive telemetry link, stimulation pulse demodulator, charge-balancing series capacitor, and a subretinal Si NW electrode array. The system requires application of external power delivered by a single wireless inductive link. The NWs are designed to serve each as an electrode penetrating into retina. Two wires are used to connect the electrode array to the stimulator and to the nearby ground electrode. Upon application of voltage bias, the device produces a current pulse that scales with the intensity of light detected by the NWs. This hybrid optoelectronic system provides separate spatial (incidence light) and temporal (electrical bias) control over stimulation, as well as tuneable gain in case of lower light intensity thresholds. A distinct advantage of this architecture compared with other systems for electrical activation of the retina is that enhancing the resolution incurs almost no increase in hardware beyond the density of the electrode array. A proof of concept validation of the system was performed ex vivo by recording the activity from isolated retinal explants from P23H degenerated rats. A significant increase in neural response was observed under application of both bias and light. Pulsed light stimulation by itself (at low bias), or light off conditions with bias on, did not elicit a neural response, validating the feasibility of the proposed system.

An alternative approach to electrical stimulation is optical activation of retinal neurons through photoactive interfaces offering a new rout for wire-free, self-powering autonomous retinal prostheses. Photovoltaic polymers and quantum dots (QDs) were proposed as a photoelectrical retinal interface (Bareket-Keren and Hanein, 2014; Gautam et al., 2014; Antognazza et al., 2016; Maya-Vetencourt et al., 2017). QDs are s semiconducting nanometer-size crystals (typically 2–6 nm in diameter) with size-depended optical and electrical properties due to the quantum confinement effect. QDs were proposed as mediators for optical neural stimulation as a temporary electric dipole moment and localized electric fields can be generated upon optical excitation. QD films (Pappas et al., 2007; Molokanova et al., 2008) or QDs directly attached to the cell membrane via antibodies or peptides (Winter et al., 2001) were utilized for activation of neurons with light. Stimulation of RGCs with a composite QD-CNT films was also reported by Bareket-Keren and Hanein (2014).

The composite films were realized through covalent conjugation of CdSe/CdS core-shell semiconducting nanorods (NRs) to CNTs via a plasma polymerized acrylic acid layer (Bareket et al., 2017). It was reported that the NR geometry supports effective transduction of the light to charge-separation at the NR–CNT interface. The highly porous 3D CNT film enables high loading of the light-sensitive semiconducting nanocrystals. Moreover, CNTs contribute to enhancement of the electrochemical properties of the interface and its coupling to neurons due to their biomimetic structure. The electrical response of the film was typified by a voltage increase during photoexcitation followed by a slow discharge as the light was turned off (Bareket-Keren and Hanein, 2014). This kinetics was attributed to the capacitive nature of the films, charging and slowly discharging upon application and removal of the illumination stimulus. Finally, embryonic chick retinas lacking developed photoreceptors were successfully stimulated using pulsed light at a wavelength of 405 nm and ambient light intensities. The retina explants were placed on the NR–CNT composite electrodes with the RGC layer facing down (as in an epiretinal configuration). The array was used for simultaneous optical stimulation and extracellular recording. Finally, in vitro tests further confirmed biocompatibility and stability of the system for up to 21 days.

These results demonstrate the potential of QD based systems for autonomous photovoltaic retinal implants, and toward mimicking the activity of the damaged photoreceptors. Challenges that need to be overcome for realizing a retinal prosthesis based on such photoelectrical interfaces include the efficiency and speed of the photoelectrical response, and the potential toxicity of the photosensitive QDs (Derfus et al., 2004; Gomez et al., 2005). Previous investigations lined toxic effects such as damaged cell morphology, metabolic activity and cell viability to the presence of QDs (Lovric et al., 2005; Chan et al., 2006). These effects are considered to result from QD composition, surface coating and nanometer size (Kotov et al., 2009).