Chronically implanted intracortical microelectrodes have been used in invasive brain machine interfaces (BMI) for controlling computers and assistive devices to restore function after paralysis (Hochberg et al., 2006, 2012; Collinger et al., 2013). However, current devices suffer from recording instabilities that lead to loss of cells over time (Dickey et al., 2009). One reason for this is the Young’s modulus mismatch between electrode materials and brain tissue, leading to micromotion-induced trauma (Subbaroyan et al., 2005; Dickey et al., 2009) and a persistent immune response that compromises neural function (Polikov et al., 2005).

One way to mitigate the Young’s modulus (E) mismatch is to produce flexible interfaces for long-term implants (Gilletti and Muthuswamy, 2006). Current flexible electrodes are centered on the use of polymers, such as parylene-C (E = 2.86 GPa), polyimide (E ~ 2.3 GPa), SU-8 (E = 4.02 GPa), and smart materials with a suitable metal for recording, although chronic recording data are lacking. Most of the current flexible devices seem to develop issues, especially for interfacing >1 year. For example, polyimide has been used extensively for flexible electrodes (Rubehn et al., 2009); however, it suffers from pinhole development, which will lead to subsequent in vivo device delamination through fluid infiltration and is also ~3 times stiffer (in the case of PI-2611) than parylene-C-based electrodes (Rubehn and Stieglitz, 2007). Additionally, performing leak impedance measurements after submerging PI-2611 in 40°C saline found sharp declines in impedances after 2–3 days, alluding to insulation failure on probes, which would translate to poor in vivo performance. However, enhanced performance can be obtained, if specific curing measures are used (Ceyssens and Puers, 2015). Smart materials (Harris et al., 2011b; Hess et al., 2011; Ware et al., 2012) that change E when being inserted due to water uptake or a change in temperature (e.g., 5 GPa 12 MPa would be an ideal solution) suffer from poor insulating properties and have only been used as insertion aids for flexible structures or lack major characterization beyond a proof-of-principle demonstration (Ware et al., 2012). Furthermore, chronic implantation after 8 weeks showed a similar distribution of defect nuclei compared with a microwire electrode, therefore conferring no advantage to prolong chronic recording through modulus matching with these materials (Harris et al., 2011a,b). Conductive polymers (Lu et al., 2014) and carbon fibers have also been used; however, these lack in vivo characterization for indwelling periods >1 year (Yoshida Kozai et al., 2012) and seem to confer no advantage over conventional electrodes in terms of gliosis (Canales et al., 2015). Finally, carbon fiber probes, which can be produced at small diameters with E = 241 GPa (Patel et al., 2015), can be more stiff than inflexible silicon probes, although good chronic recording has still been obtained with such probes (Guitchounts et al., 2013).

In a previous study, we were able to record neural activity for >2-year periods, demonstrating the long-term reliability of our microfabricated devices (Sohal et al., 2014). The probe comprised a parylene-C–metal-parylene-C sandwich with design features to minimize the effects of micromotion. This study had an emphasis on the biological impact of such probes to address micromotion-induced trauma, such as high fidelity recording quality for up to 2 years and reduced foreign body response determined through histology, after chronic implantation.

Here, we report the refined design and microfabrication considerations that allowed reliable flexible devices to survive the challenging environment of chronic neural recording (Barrese et al., 2013). A novel use of the probe for acute electrophysiology is also presented. The overall details in this study will aid other groups involved in flexible microfabrication in potentially obtaining reliable devices for chronic interfacing, in a wide variety of biomedical applications.

Compared with the conventional rigid and sharp electrodes, the design had several features to minimize micromotion-induced trauma (Figure 1A). The probe was made out of flexible materials and incorporated a rounded tip with three protruding recording sites, which were anchored within the brain tissue with a polyimide ball. Further, the design incorporated a sinusoidal shaft, instead of a straight shaft. This shaft accommodates movement of the brain, mechanically decoupling the recording end from the surface fixation point and thus allowing the electrode sites and ball anchor to move freely with the tissue. The recording end of the probe was a parylene-C disk of 100 μm diameter, with three protruding, exposed recording sites (96 μm²) to aid single-unit isolation. Three metal tracks (5 μm width and gap) ran within an electrode shaft that was 20 μm deep, 35 μm wide, and had a sinusoidal profile with cycles of 100 μm amplitude and 500 μm period. The probe was 3 mm in length and incorporated an integrated parylene-C-based ribbon cable (3 cm long and 3 mm wide) leading to a standard connector (micro ps1/ps2 series, Omnetics Connector Corporation, USA) (Figures 1B,C). To demonstrate the utility of a sinusoidal shaft, a finite element model (FEM) was constructed (Figure 2). GDS files from mask designs were imported into COMSOL Multiphysics^®, for both straight and sinusoidal shafts with identical dimensions (3 mm length, 20 μm thick, and 35 μm wide) to form a 3D model. The “Solid Mechanics” physics was used, and the top of the electrode shaft was rigid (“fixed constraint”). Parylene-C was used in this simulation with properties of E = 2.86 GPa, density = 1.29 kg/m³, and Poisson’s ratio = 0.4. The shaft itself was free to move. We compared the flexibility of both designs to accommodate displacements in three dimensions. An arbitrary displacement (100 μm) was applied in the orthogonal X, Y, and Z directions of the rounded recording end for both the straight and sinusoidal probe design. Figure 2 shows that a sinusoidal shaft, rather than a straight shaft, will deform more to counter movement of the probe primarily in the Y direction [Average displacement of the probes (±SEM) 85 ± 9 vs. 53 ± 9 μm³, t-test: t = 5.5, P < 0.001] as calculated through COMSOL Multiphysics^® using the “domain probe” feature, which allows the measurement of displacement values for the duration of the simulation. Comparable average displacement values (53.3 μm) were obtained for X and Z directions. The Y-direction has also been shown to be the primary axis of motion (cephalocaudal) for the brain in both rodents and human MR studies (Poncelet et al., 1992; Chen et al., 2010; Köhler et al., 2015).

The simulation allowed the assessment of the Von Mises stress on both the shank types. The von Mises stress is used to predict the yielding of materials under any loading condition from results of simple uniaxial tensile tests. The stress induced by the movement was significantly reduced for a sinusoidal (30.7 MPa) rather than a straight shaft (92.5 MPa), showing that it is also better at minimizing stresses on the shank and at accommodating random brain motion.

Electrode impedances at 1 kHz in saline (IMP-1 meter, Bak Electronics, USA) were characterized for the probes (Figure 8A). A total of 426 recording sites were characterized, and impedance values all fell within the acceptable range for single-unit recordings <5 MΩ (Ludwig et al., 2011). Specifically, impedance values (±SEM) were 770 ± 15 kΩ.

The microfabrication process therefore produced many viable electrodes toward the end of the optimization period with many probes falling in the acceptable impedance range for single-unit recordings. The majority of probes (>83%) had impedances ≤1.2 MΩ, which is classed as good/ideal from data corresponding to the use of acute single shanked microwire probes (Microprobe Inc., USA) to isolate single-unit activity (Soteropoulos and Baker, 2009; Witham and Baker, 2012).

Electrode impedances were also fully characterized over a 180-day indwelling period for 15 electrode-recording sites (Figure 8B). There was no significant difference for explanted devices when comparing the impedance values at 1 kHz in saline (recording day 0 vs. 180, t-test: t = 1.7, P > 0.05). The mean impedance values for recording day 0 and 180 were 1100 ± 150 kΩ (n = 15 sites) and 850 ± 120 kΩ (n = 12 sites), respectively. This showed that the parylene-C-based devices were stable and survived damage, which could have been caused by the long indwelling period.

This probe has registered high fidelity neural activity for over a 2-year period. Consistent waveforms were extracted with stable signal-to-noise ratio (SNR) and mean peak-to-peak amplitude over the recording period. The second probe generation could register neural activity for 140-day period with increasing SNR and peak-to-peak amplitude over the recording period (n = 15 recording sites, five probes) (Sohal et al., 2014). These promising initial recordings highlight the stability of the implanted probes and emphasize the reliable microfabrication process that has been developed, which promotes device survival for long indwelling periods.

As the probe can also be permanently attached to the carrier with two-part epoxy, a novel dual-purpose probe for both acute and chronic recording can be created.

As a demonstration of this, we show that we can acutely record from the mouse cortex with good SNR (Figure 9) following Suner’s method (Suner et al., 2005; Ward et al., 2009). Experiments were approved by the local ethics committee at Newcastle University and appropriate UK Home Office licenses in accordance with the Animals (Scientific Procedures) Act 1986.

From example waveforms extracted (n = 6, Figure 9B), we showed good mean peak-to-peak amplitude (200 ± 50 uV), noise (33 ± 7 uV), and SNR (6.0 ± 0.5). For the signal, the peak-to-peak amplitude (A) of the mean waveform was calculated. For the noise, the mean waveform was subtracted from all waveforms, with the SD calculated from the resulting values. Noise was then calculated as 2× the average SD (ϵ). SNR was then calculated as A/ϵ. For 12 recording sites, there was no difference in pre- and post-implant impedance value post-device cleaning with ethanol and deionized water (t-test: t = 1, P > 0.05; Figure 9C).

The probes are parylene-C-based, so they are easier to clean than conventional silicon probes that have been used for repeated acute recordings due to minimized adhesion of tissue and blood following acute probe explant (Winslow et al., 2010).

The recording parameters and in vivo impedance measurements showed that the probe was reliable during the long indwelling periods, indicating a highly reliable microfabrication process. Current state-of-the-art probes have been reported to lose signal quality over a few weeks implantation period (Jackson and Fetz, 2007; Dickey et al., 2009), with device delamination post-explantation over these short periods (Barrese et al., 2013).

A change from SiO₂ to Al for the sacrificial layer allowed reliable release for the probe, as long as there was good adhesion between the Al and parylene-C. Al has a tendency to develop a thin native oxide over time, and this can interact with the silane-based (A-174) adhesion promoter used for parylene-C. Literature suggests that a greater oxide formation combined with silane will lead to enhanced adhesion between parylene-C and Al (Mitchell et al., 2006). Therefore, to enhance Al adhesion to parylene-C, a native oxide layer should be allowed to form on the Al through a simple delay in microfabrication processing, before parylene-C deposition. This will ensure there is reliable good adhesion. Although some other groups do not use a sacrificial layer with parylene-C or polyimide while demonstrating chronic usage up to 28 days (Kim et al., 2013; Kuo et al., 2013), it was found here that device release was more reliable with it present.

Other groups have baked parylene-C with a nitrogen backflow to increase adhesion between the two layers (Kuo et al., 2013). But a slight over-etch of the metal roughens the first parylene-C layer surface leading to good device adhesion with no delamination to microfabricated, implanted, and explanted devices, as corroborated through impedance measurements and Raman spectroscopy. However, baking should be considered in future, especially with the incorporation of noble metals (e.g., platinum for chronic stimulation) to increase adhesion. Baking the parylene-C-based devices post-microfabrication between 200 and 300°C for 1–2 h may increase longevity of the Parylene-C. For saline-soaked samples at 37°C over a 300-day period, stable DC resistances of samples were observed at 300°C. Furthermore, comparing untreated, 200°C and 300°C samples, it was found that the baseline Yield strength of parylene-C can be enhanced with increasing temperature (77–107 MPa) (Von Metzen and Stieglitz, 2013).

A striking feature is that our devices remained stable for a 2-year indwelling period. Although there have been reports of parylene-C failure on intracortical probes, this is mainly due to deposition onto silicon. For the Utah array, thin (<4 μm) layers are used to insulate the array, and this will eventually break down due to adhesion of the parylene-C to the silicon (Barrese et al., 2013; Xie et al., 2014). Material stresses due to the E mismatch between silicon and parylene-C can also cause cracking (Kozai et al., 2014). No such problems were encountered with our probes when explanted, as similar impedance values were found pre- and postimplantation (Figure 7B). Probes were able to record for 2 years without degradation in the acquired neural signal, emphasizing material stability (Sohal et al., 2014). Literature suggests that parylene-C failure mechanisms include water and salts diffusion (Yasuda et al., 2001; Seymour et al., 2009), interfacial delamination (Barrese et al., 2013), and pinholes (Pierstorff et al., 2008). However, to mitigate this, the thick (although still flexible) parylene-C layers and minimized damage during processing (e.g., enhancing adhesion between layers) used in this study, may have contributed to this improved chronic performance. Interestingly, a reduction in the microglia response was observed around the probes for a 12- and 24-month period (Sohal et al., 2014). Microglia are known to produce toxic chemicals, such as reactive oxygen species, that will degrade electrode material over time (Takmakov et al., 2015). A combination of overall design measures (Figure 1), material choice, and reduced gliosis may have contributed to the enhanced longevity of the probe. Polyimide was found to contain many pinhole defects and therefore reaffirmed the choice for parylene-C use as an insulator.

Here, the acute recording qualities of the sinusoidal probe are demonstrated. The probe is permanently attached to the insulated sharp steel electrode and is further insulated from the effects of electromagnetic cross talk by using a two-part epoxy attachment technique. Good SNR and signal amplitude are observed, which are comparable, and in some cases, better than conventional silicon probes that are used in acute animal preparations (Blanche et al., 2005; Du et al., 2011; Berenyi et al., 2014). The reduced cost and ease of cleaning makes it an attractive alternative to silicon probes for both acute and chronic applications.

In addition to our stable devices, reduced gliosis responses were shown, compared with microwire electrodes (the gold standard in terms of gliosis at the time of the study) at 6-, 12-, and 24-month time points at the electrode tip, with an increase in neural integration around our probe (Sohal et al., 2014). This suggests that the design features incorporated in our probe have reduced the effects of micromotion-induced trauma.

Although these probes were restricted to 3 mm in size and three electrode-recording sites, this can be upscaled from a basic successful proof-of-concept design to allow for differing lengths and recording site arrangements. The design could potentially be adapted for multiple applications and differing species.

We have successfully microfabricated a reliable flexible probe for chronic neural recording, with one of the longest chronic indwelling periods (>2 years), without delamination. We have optimized etching of thick parylene-C layers and WTi metal. We have also developed a reliable device release strategy incorporating an aluminum release layer and TMAH, leading to a high number of functional devices. Further, we have developed a relatively simple and inexpensive device insertion technique that has been reliably used to insert the flexible devices into rabbit brains. This microfabrication process flow can also be applied to construct other flexible electrode types.

HS, AJ, and SB designed the novel electrode. HS performed the microfabrication, implant construction, electrophysiology recording, and subsequent analyses. KV aided in microfabrication techniques and design. HS, AJ, SB, and AO designed the experiments. All authors aided in the writing of this manuscript. AJ, SB, and AO equally contributed to this study.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.