Stretchable gold nanowire-based cuff electrodes for low-voltage peripheral nerve stimulation

Bioelectronic medicine [1–3] can be used as a treatment modality in various diseases and disorders by applying electrical stimulation to the central [4] or peripheral nervous system (PNS) [5–7]. The nerves in the PNS consist of interchanging fascicles wrapped in perineurium and enclosed by a soft and stretchable epineurium [8, 9], which ensures that the nerves can withstand ordinarily bodily motions without harm. If a nerve gets injured, the neurons lose their connection to the target tissues, resulting in loss of muscle function and sensibility in the affected area. Severed nerves are treated by suturing the epineurium of the separated nerve ends together with the hope that the nerve fascicles grow back correctly. Bioelectronic medicine has been successfully employed to promote the regeneration of severed nerves, with one example being the application of electrical pulses to the proximal end of a rat femoral nerve transection during surgery, which led to an acceleration of the axonal nerve regeneration as well as to an increase in preferential muscle reinnervation [10]. Electrical stimulation of the PNS can be performed with various stimulation devices, such as hook- and cuff electrodes. Cuff electrodes can be wrapped around peripheral nerves [11, 12] and are well established within the biomedical industry. When a sufficiently high voltage (above the stimulation threshold) is applied between two electrodes in proximity of the nerve, action potentials are evoked in the nerve fibres. The evoked action potentials propagate along the nerve fibres, which in the case of motor fibres lead to muscular activity that can be recorded visually by monitoring muscle movement [13], or electrically via electromyography (EMG) [14]. Common commercial cuff electrodes incorporate rigid electrodes with a high Young's modulus, leading to a large mechanical mismatch between the electrodes and the soft nerve tissue, which can result in a strong immune response and nerve damage from motions [12, 15, 16]. Furthermore, stiff and thick cuff electrodes typically leave some spacing to the nerve due to their rigid nature, which cause current leakage around the nerve. Contrary, a close conformal electrode minimizes leakage current and thereby improves stimulation threshold, stimulation specificity and signal recording quality [17]. A high stimulation voltage can induce tissue damage around the electrodes as well as damage the electrodes over time [18]. Lower stimulation voltages are also beneficial from an energy consumption perspective, as many cuff electrodes are meant to be implanted chronically and therefore operate under limited energy supply (batteries, wireless energy transfer) [19]. Many researchers have acknowledged these problems and developed thin and flexible cuff electrodes based on polymeric substrates of parylene-C [20–23] and polyimide [24–27]. These flexible devices enable better conformability to the nerve, which results in lower stimulation thresholds as most of the current goes into the nerve.

Although flexible cuff electrodes can establish conformal contact to nerves, they cannot follow the natural bodily motions of the nerves which includes bending and stretching [28]. The Young's modulus of a flexible cuff electrode is typically in the GPa range, which is more than three orders of magnitude higher than reported values for peripheral nerves (typically 40 kPa–20 MPa, nerves are however heterogeneous anisotropic structures of a hyperelastic nature) [7, 29, 30]. This means that a flexible cuff will not deform along with the nerve, thereby potentially inducing tissue damage by for example kinking the nerve at the edge of the cuff [31]. Recently, significant efforts have gone into the development of stretchable bioelectronics [32, 33], and these methods have also been applied towards the development of softer and stretchable cuff electrodes that can deform along with the nerve itself. Decataldo et al used microcracked gold tracks on polydimethylsiloxane (PDMS) to record renal sympathetic nerve activity after 14 d of implantation [34]. To reduce the impedance of their devices, Decataldo et al coated the electrode sites with a poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT:PSS) based hydrogel. Liu et al developed soft cuff electrodes based on a PEDOT:PSS hydrogel, which was encapsulated by a photopatternable elastomer [35]. The device could be stretched up to 20%, exhibited a low Young's modulus of 30 kPa, and was used to stimulate the sciatic nerve of rats. In a modified version of this device, a PEDOT:PSS based conductor with glycerol was embedded in a viscoplastic polymer, which showed an increased, although irreversible, stretchability of up to 100% without loss of function. This device could elongate with the growing nerve in chronic implantation and evoke compound action potentials at low stimulation voltages [31]. Stretchable carbon nanotube-PDMS composites [36] and stretchable PEDOT:PSS-polyurethane based cuff electrodes [37] have also been reported for PNS interfacing recently. The above examples demonstrate the potential benefits of soft and stretchable cuff electrodes for nerve interfacing. Electrical stimulation of nerves often requires currents in the 0.1–1 mA, thus, to limit the Ohmic losses and to minimize the applied voltage, it is important that the interconnects and electrodes have low enough impedance. This is a challenge for stretchable interconnects as the sheet resistance of microcracked gold and PEDOT hydrogels are in the order of 10 Ohm sq⁻¹ and ≫100 Ohm sq⁻¹, respectively. Another drawback with PEDOT hydrogels and novel elastomer materials is that they have not been evaluated for long-term clinical applications, while gold, platinum and silicones are well known for their feasibility in bioelectronic implants [7, 19, 38, 39].

To combine high performance with suitable biocompatible materials, here we report on a soft elastically stretchable cuff electrode based on gold, platinum, and PDMS. The device comprises interconnects of gold coated TiO₂ nanowires (AuNWs) embedded in PDMS, with an initial sheet resistance of 0.3 Ohm sq⁻¹ and excellent cyclability up to 50% strain. The electrodes were formed by electroplating a thin layer of platinum on top of partially exposed AuNWs and they showed excellent stability for one million monophasic voltage pulses. The cuff electrodes were made of ≈50 µm thick PDMS exhibiting a low Young's modulus of 800 kPa, which ensured a conformal contact between the nerve and the electrode. The device was designed, with the aid of a computational model, to evoke action potentials across the complete cross section of the peripheral nerve, which is preferential for nerve regeneration applications where the aim is to uniformly stimulate all fascicles within the injured nerve. In a rat sciatic nerve model, we were able to evoke major simultaneous activity in the soleus and the tibialis anterior muscle upon applying low voltage monophasic stimulation pulses of 200 mV (94 µA) for 500 µs and evoked minor EMG signals with voltages as low as 50 mV. This demonstrates how the combination of conformal electrode contact, low resistance interconnects, and low impedance electrodes, can result in minimal voltage loss during stimulation and therefore an optimal voltage stimulation threshold of the nerve. Our stretchable cuff electrode thus constitutes a promising candidate for low-power wireless chronic applications due to its material-based biocompatibility, softness, and elastic stretchability.

2.1. Device fabrication

A suspension of gold nanowires (≈300 nm diameter, 3–20 µm length) was prepared according to a previously reported method [40, 41]. In short, auric chloride solution was reduced by hydroxylamine onto poly(vinylpyrrolidone) stabilized TiO₂-NWs (Novarials, Novawire-TiO-100-RD, diameter ≈100 nm, length <20 µm, 0.72 mg), leading to gold coated TiO₂-NWs (AuNW, diameter ≈300 nm). Figure 1(a) depicts the further fabrication process of the stretchable cuff electrodes. A wax pattern for three devices (area 110 mm²) was printed on a poly(vinylidene difluoride) filter membrane (Merck-Millipore, 0.2 µm pore size) and 3.3 ml of the AuNW suspension was filtered through, resulting in an approximately 4 µm thick AuNWs film deposited on top of the filter membrane (filtration patterning [42]). Polydimethylsiloxane (PDMS, Sylgard 184, Dow Corning, 10:1) was spin coated at 1800 rpm for 30 s on a 2 inch silanized (Trichloro(1H,1H,2H,2H-perfluorooctyl) silane) glass wafer and semi-cured on a hotplate at 70 °C for 10 min until it turned solid but remained sticky. The patterned AuNWs were transferred to the semi-cured PDMS in a hot-press at 200 N force and 75 °C for 20 min. The filter membrane was peeled off the PDMS after wetting with deionised water. The patterned AuNW layer remained on the PDMS and was subsequently partially filled by spin-coating a heptane:PDMS solution (16:1 w w⁻¹) at 6000 rpm for 60 s and cured for 10 min at 80 °C. Additional PDMS was spin-coated at 6000 rpm for 60 s to form an encapsulation layer (≈10 µm) while the contact and electrode sites (6.3 mm long) were protected by pieces of foil (poly(ethylene naphthalate), Teonex, 25 µm thick) that were removed before the final curing step at 80 °C for 4 h, see figure 1(a). The resulting stretchable cuff electrode consisted of two embedded AuNW conductor paths of 700 µm width and 25 mm in length with a spacing of 3 mm in between. The dimensions were chosen to be optimal for complete stimulation of a rat sciatic nerve of 2 mm in diameter, with the overall device dimension being about 1 cm wide and 50 µm thick (60 µm at the conductors).

Zoom In Zoom Out Reset image size

2.2. Electromechanical characterization

2.3. Electrode preparation and characterization

The AuNWs at the electrode sites were partially exposed by reactive ion etching (RIE) for 10 s (250 W, O₂:CF₄ 1:2 ratio, 100 mTorr). The devices were cut out, peeled from the substrate and connected to a flexible flat cable (FFC, 1 mm pitch, six-ways, Wurth E328849) by applying silver epoxy (MG Chemicals, 8330 S) in between the contacts and cured for 1 h at 95 °C under 150 N force. Kapton tape was taped around the device-FFC interconnection to mediate applied stress during handling. Electrochemical experiments were performed on the devices in a three-electrode cell setup (Platinum gauze counter, Alfa Aesar; Ag/AgCl pellet reference, Warner Instruments E210) with a Gamry Potentiostat 1010E. Electrochemical impedance spectroscopy (EIS, 100 kHz–1 Hz, 50 mV) was performed before and after each process step in phosphate buffered saline solution (PBS, 0.01 M, Sigma Aldrich, P4417) [43]. Cyclic Voltammetry (CV, −1–0,6 V, six cycles, 100 mV s⁻¹) was conducted in PBS to clean the AuNWs electrodes, upon which a thin platinum black layer was electroplated by chronoamperometric pulsing (20 Hz, 10% duty cycle, −1.5 V, 90 s) in a chloroplatinic acid solution (0.8% in H₂O, Sigma Aldrich). Strain stability tests were conducted by subjecting the electrodes to 20% and 50% strain consecutively, in a motorized linear stage (X-LSQ300A-E01, Zaber, velocity 2 mm s⁻¹, rest time 1 s) for 1000 cycles. Pulsing stability tests were conducted with a similar protocol as for the in vivo experiments by subjecting the electrodes to 10⁶ monophasic voltage pulses (500 µs, 20 Hz, in PBS, 2 kOhm resistor in series to simulate nerve resistance) of 400 and 800 mV consecutively [43]. After each stability test, EIS (100 kHz–1 Hz, 50 mV) and CV (−0.6–0.6 V, six cycles, 100 mV s⁻¹) was performed consecutively in PBS. Scanning electron microscopy images (SEM, ZEISS Sigma 500 Gemini) and energy dispersive x-ray spectroscopy maps (EDX, BRUKER) were taken of the electrodes before and after RIE, as well as after Pt coating and strain cycling, see figures 2(a)–(d).

Zoom In Zoom Out Reset image size

2.4. Numerical simulations

Simulations were performed in COMSOL Multiphysics 5.4 software using the electric currents module to calculate potential and current distributions in conducting media. The model of the nerve and the cuff electrode was implemented in a 2D axisymmetric geometry with the symmetry axis along the length direction of the nerve, see figure 3(e). The nerve model consisted of a single fascicle of endoneurium (r = 740 µm) encapsulated in perineurium (r = 25 µm) and epineurium (35 µm), with a saline layer of 76.5 µm between the nerve and the cuff. The conductivities were adopted from Zariffa et al for endoneurium (0.571 S m⁻¹ along z, 0.0826 S m⁻¹ along r, θ), epineurium (0.0826 S m⁻¹), perineurium (0.0021 S m⁻¹) and saline (2 S m⁻¹) [44]. The areal capacitance of the perineurium was set to 0.39 µF cm⁻² which is in line with the range provided by Weerasuriya et al [45]. The electrodes were implemented as highly conductive regions with a surface capacitance set to 100 µF cm⁻² to fit the experimental data. Time dependent simulations were performed with identical voltage pulses as in the experimental data. To assess the effect of leakage currents, additional simulations were performed with a saline layer of varying thickness in between the cuff and the nerve, see figure S3.

Zoom In Zoom Out Reset image size

2.5.  In vivo rat sciatic nerve stimulation

Male Sprague Dawley rats (approximately 350 g) were used in the experiments. The rats were anesthetized with a mixture of Isoflurane (Baxter, Deerfield IL, USA) and oxygen, and then placed in a prone position on a heating pad. Anaesthesia was kept constant with isoflurane over a face mask. The hair over the gluteal area on the left side was shaved and cleaned for aseptic procedures. The skin was longitudinally incised over the femoral line at a length of approximately 3 cm. The sciatic nerve was identified through blunt dissection by gently creating a surgical cleavage between the gluteal muscles. Both the sciatic nerve, as well as its division into the tibial and peroneal nerve branches were identified. The stretchable cuff electrode was placed proximal to the division of all muscle branches, where the nerve is approximately 1.6 mm in diameter, see figure 3(b). The stretchable cuff electrodes (n = 5) were connected via a custom made FFC connector to a Gamry Potentiostat 1010E. Cathodic monophasic chronoamperometric pulses were applied from the distal to the proximal electrode with varying frequency (2, 20 Hz), pulse lengths (200, 500 µs) and voltages (50, 75, 100, 150, 200, 400, 600, 800 mV) and the device specific current response was recorded. Two EMG electrodes were placed distal to the nerve in each of the soleus and anterior tibial muscles, and one electrode as ground in the musculature of the back. The EMG electrodes were connected to an IX-BIO-SA biopotential amplifier (iWorx Systems, Inc., 10 kHz sampling rate) which recorded the EMG signals during the experiment. After the experiment, the rats were euthanized without regaining consciousness by immersing them in a closed container containing carbon dioxide. To ensure that the animals were dead, they were thoracotomized. The recorded EMG data from consecutive 2 Hz pulses (n = 4) was integrated and averaged with a custom MatLab script (Mathworks) while excluding EMG recordings with extreme spikes that occurred probably due to static charging of the un-shielded EMG cables. The resulting averaged area values for varying voltages and pulse lengths were then normalized to the maximum value of each EMG recording (stimulation saturation) and the normalized values of each applied voltage (figures 4(e) and (f)) and corresponding current were plotted.

Zoom In Zoom Out Reset image size

3.1. Device design, fabrication, and electromechanical characterization

3.2. Electrode fabrication and characterization

3.3.  In vivo electrical stimulation characteristics

3.4. Muscle EMG response from sciatic nerve stimulation

We have developed a thin and stretchable cuff electrode with the aim of conformally interfacing nerves while minimizing induced mechanical damage to the tissue. The device is based on gold, platinum and PDMS, which are inert and biocompatible materials [19, 38, 39, 49] and therefore well suited for chronic implantation [40]. The stretchable interconnects were based on a highly conductive and stretchable AuNW-PDMS composite, which showed excellent cycling stability at 50% strain. For the patterning of these conductors an unconventional masked filtration patterning approach was employed, which has the advantages that it makes optimal use of the inhouse synthesized AuNWs, allows for cleaning of the deposited nanomaterial, and can easily be adapted to alternative designs with resolutions down to 100 µm [42]. Previously reported photolithography based methods even enable electrode miniaturization down to 50 µm [40]. There are several different aspects to consider when designing stretchable neural stimulation electrodes: (a) the electrode materials should be inert and stable over many stimulation pulses [43]. (b) The electrode should have sufficiently high areal capacitance and low impedance to enable effective stimulation while avoiding excessive charging. (c) The electrode should be electromechanically stable under repeated deformations. The bare AuNWs are not optimal with respect to these conditions as gold has a rather low areal capacitance and can be corroded at higher potentials in chloride-rich environments. To improve the performance of the electrodes, a thin platinum coating was electroplated onto the individually exposed AuNWs. Platinum employs pseudocapacitive charge transfer [18] and has a high specific surface area due to its controllable porosity [50]. Here the Pt coating results in a five-fold reduction in electrode impedance compared to the pure AuNW electrodes at lower frequencies (figure 2(e)). The Pt coating also renders the electrodes stable over one million 800 mV pulses, which is well above the stimulation saturation threshold voltage for the evaluated nerve model. Finally, the electrodes electromechanical stability was evaluated, which showed excellent stability at 50% strain cycling, indicating that the electrodes can withstand large enough deformations to be compatible with the natural movements within the PNS [28]. All-in-all, the performance and stability of the stretchable cuff electrode is excellent and fulfils previously stated criteria.

The equivalent circuit of the cuff electrode-nerve system given in figure 3(a) is useful for discussing the influence of various components on the impedance of the system. The current response to the applied voltage pulses includes a ≈50 µs charging peak followed by a plateau which decrease slowly with time. The current values of the plateau indicate a magnitude of the impedance in the kOhm range, which is much higher than the impedance of the electrodes in PBS buffer, which is below 100 Ohm at >1 kHz frequencies. Thus, the impedance of the system is dominated by the resistive contribution of the nerve, which also is consistent with the low resistance of the stretchable interconnects and the high areal capacitance of the electrodes. To further analyse the system, a numerical model was developed and fitted to the electrical measurement data. The experimentally measured current response was accurately reproduced by the simulation, under the assumptions of ideally capacitive electrodes and realistic conductivity values of the nerve. The fitted areal capacitance in the model was approximately 30 times lower than the experimentally extracted capacitance for the electrodes in PBS buffer. This is not uncommon for electrodes in contact with tissue, as the electrode surface cannot be expected to be clean in this situation. The initial (dis)charging peaks in the measured currents can be attributed to capacitive (dis)charging of the cell layers of the highly resistive perineurium [45]. An exact value for the areal capacitance of the perineurium is not available, but our fitted value (0.39 µF cm⁻²) is within the expected range of previous measurements [45]. One should note that the time constant of our simulated (dis) charging peaks fits the experimental data very well, indicating that both the magnitude and location of the capacitances and resistances within the system are correct. In this study we did not assess the chronic stability of our stretchable cuff electrode, however, previous studies with stretchable cuff electrodes showed improved long term stability compared to non-stretchable cuff electrodes due to reduced inflammatory response and adaptability to the dynamic environment [31, 34, 35].

Based on the muscle EMG measurements (figures 4(e) and (f)), we found that the stimulation onset occurred in the 75–150 mV range (≈55–75 µA) while stimulation saturation appeared around 150–400 mV (≈80–100 µA). Only a minor difference in stimulation voltage threshold between the EMGs of the soleus and the anterior tibialis muscle was detected. This inhomogeneity in stimulation voltage may be related to the location of the corresponding motor fascicles of the muscles, as they are located in different branches of the nerve [9]. A striking feature is that the transition from the stimulation onset to stimulation saturation was rather sharp, typically around a factor of two in voltage. This is consistent with how the electric field is distributed according to the numerical model, as the difference in voltage to reach a certain electric field threshold in the outer and inner parts of the nerve is approximately a factor of two (figure S9). All together this indicates that complete homogeneous stimulation of the nerve can be achieved for as low voltages as 200 mV for the 500 µs pulses. We attribute this to low resistive interconnects, high areal capacitance electrodes in combination with the excellent conformability of the cuff. Less conformal cuff electrodes would suffer from substantial leakage current through the gap in between the nerve and the cuff, as shown in figure S3.

It is challenging to compare stimulation voltage thresholds/saturation between different studies, as the reported values depend heavily on the device geometry, the nerve model, and the characterization methods. Some commercial cuff electrodes show stimulation thresholds of about 300 mV [31]. Localized stimulation requires less current, which correlates with lower voltages. Larger nerves require larger currents to stimulate the whole cross section of the nerve. It is therefore difficult to directly compare the results of this study with recent studies on stretchable cuff electrodes based on conducting polymers which report voltage stimulation thresholds of 50 mV for 200 µs biphasic stimulation pulses [35] and 100 mV for 200 µs monophasic voltage pulses [31]. An alternative approach is to look at the material properties reported in these previous studies and analyse how they would have influenced the performance in the setting used in this study. Both studies by Liu et al [31, 35] employed stretchable, conductive polymer interconnects with sheet resistances of 150 Ohm sq⁻¹ (1.4 µm thick, 47 S cm⁻¹) and 2250 Ohm sq⁻¹ (2 µm thick, 2.2 S cm⁻¹), respectively. The interconnects in our device are ≈60 squares long, which would have resulted in interconnect resistances of 9 and 140 kOhm for respective material. These high resistances would had severely limited device performance and required manyfold higher voltages for generating the same currents as in our devices. We can thus conclude that our cuff electrode would outperform previous stretchable cuff electrodes in our experimental setting. This is especially true in the stretched state, where our interconnects would only increase the total resistance of the system with ≈30% at 50% strain (based on electromechanical characterizations). This is not too surprising, as conducting polymers are rarely used for interconnects due to their lower conductivity compared to metals. From the analysis above, we concluded that in our case the resistance of the nerve is dominating the impedance of the stimulation circuit. This means that our developed electrode is close to optimal from an electrical perspective, as most of the potential drop occurs inside the nerve, which is the ultimate goal of any stimulation cuff electrode design. Low voltage and power operation is of great importance for applications of limited power supply, including implantable systems in general and wireless-powered systems in particular. The developed cuff electrode is thus suitable to combine with such systems to optimize performance within many emerging applications. The versatility of our fabrication methods also allows for adaptation of the technology to other applications. Our current 50 µm thick cuff can conveniently be wrapped around nerves down to 500 µm in diameter. By reducing the cuff thickness and utilizing a softer PDMS formulation, nerve diameters of below 100 µm should be within reach for 30 µm thick cuffs of 200 kPa elastic modulus (based on the same stress within the cuff). The technology is also well-suited for the development of high-resolution multi-electrode designs for selective stimulation of specific nerve fascicles, which is of great interest in many emerging applications.

In this study we present a novel stretchable cuff electrode based on inert and biocompatible materials for peripheral nerve stimulation. The AuNW-PDMS based composite conductors showed an excellent conductivity of ≈8900 S cm⁻¹ and cyclability up to 50% strain. The Pt coated electrodes exhibited an areal capacitance of 3.1 mF cm⁻², an impedance <100 Ohm at >100 Hz, and were stable for one million stimulation pulses. The stretchable cuff electrodes were evaluated in a rat sciatic nerve model and showed excellent conformability to the nerve. The conformal contact and high performance of the electrodes enabled complete homogeneous saturated stimulation at only 200 mV (≈100 µA). By analysing the impedance of the system, we concluded that most of the applied voltage dropped over the nerve, making the stretchable cuff electrode close to optimal for the evaluated application. Thus, the developed stretchable cuff electrode is ideal for low power supply applications like implants and wirelessly powered stimulation. Altogether, we believe the presented stretchable cuff electrode offers many advantages and opportunities within the emerging field of bioelectronic medicine.

The authors wish to thank Dr Aline Renz, Dr Malin Silverå Ejneby and Dr Marie Jakesova for fruitful discussions. This project was financially supported by the Swedish Foundation for Strategic Research, Sweden´s Innovation Agency (VINNOVA), and the Swedish Government Strategic Research Area in Materials Science on Functional Materials at Linköping University (Faculty Grant SFO Mat LiU No 2009 00971). S L, J Z, S F, K T conceived the project. S L performed the material and device related experiments. J Z, S F performed all animal experiments and S L analysed the data. S L wrote the first draft of the manuscript and all authors contributed to the finalization of the paper.

The data that support the findings of this study are available upon reasonable request from the authors.

All animal experiments were approved by the Committee of Laboratory Animal Ethics Committee (5.2.18-19325/17) and the animals were handled in accordance with the Principles of Laboratory Animal Care.