**Electrochemical Evaluation of a Multi-Site Clinical Depth Recording Electrode for Monitoring Cerebral Tissue Oxygen**

#### **Ana Ledo 1,2,3,\*, Eliana Fernandes 1, Jorge E. Quintero 4, Greg A. Gerhardt 4,5 and Rui M. Barbosa 1,2,3**


Received: 3 June 2020; Accepted: 25 June 2020; Published: 28 June 2020

**Abstract:** The intracranial measurement of local cerebral tissue oxygen levels—PbtO2—has become a useful tool for the critical care unit to investigate severe trauma and ischemia injury in patients. Our preliminary work in animal models supports the hypothesis that multi-site depth electrode recording of PbtO2 may give surgeons and critical care providers needed information about brain viability and the capacity for better recovery. Here, we present a surface morphology characterization and an electrochemical evaluation of the analytical properties toward oxygen detection of an FDA-approved, commercially available, clinical grade depth recording electrode comprising 12 Pt recording contacts. We found that the surface of the recording sites is composed of a thin film of smooth Pt and that the electrochemical behavior evaluated by cyclic voltammetry in acidic and neutral electrolyte is typical of polycrystalline Pt surface. The smoothness of the Pt surface was further corroborated by determination of the electrochemical active surface, confirming a roughness factor of 0.9. At an optimal working potential of −0.6 V vs. Ag/AgCl, the sensor displayed suitable values of sensitivity and limit of detection for in vivo PbtO2 measurements. Based on the reported catalytical properties of Pt toward the electroreduction reaction of O2, we propose that these probes could be repurposed for multisite monitoring of PbtO2 in vivo in the human brain.

**Keywords:** brain tissue oxygen; in vivo monitoring; multi-site clinical depth electrode

#### **1. Introduction**

Monitoring local cerebral tissue oxygen levels—PbtO2—is increasingly used in neurological intensive care units to guide therapeutic strategies aimed at maintaining O2 levels above threshold, namely in patients suffering of severe acute brain conditions such as traumatic brain injury (TBI) and aneurysmal subarachnoid hemorrhage (aSAH), and during certain neurosurgery procedures [1–3]. The rationale for this comes from studies showing improved outcome in patients with brain O2 monitoring and PbtO2 targeted therapeutic approaches [4,5]. Monitoring of PbtO2 has allowed the determination of the normal range of PbtO2 in the healthy brain tissue to be 25–30 mmHg [6–9], while values below 15 mmHg are typically associated with hypoxia and ischemia [10,11].

Due to the dependency of neuronal activity over oxidative metabolism for energy supply, monitoring PbtO2 can be a surrogate signal of neurovascular response and metabolic activity, both of

which can become severely compromised in situations of acute brain injury [12,13]. In addition, cerebral ischemia is linked to the onset of secondary brain injury by predisposing brain tissue to an energetic crisis as well as contributing to the initiation of spreading cortical depolarizations [14,15]. These are all-or-none tissue level events characterized as near-complete breakdown of neuronal membrane potential which, in injured tissue, can initiate cell death cascades [16]. Finally, besides being a key metabolite in energy metabolism, O2 is also a substrate for multiple enzymes within cells and can act as a signal for genetic adaptation to situations of hypoxia, regulating gene expression via hypoxia-inducible factor (HIF) dependent pathways [17].

Currently approved methods for clinical monitoring of PbtO2 include invasive techniques such as amperometric sensors and optical sensors, and non-invasive techniques such as those using near infrared spectroscopy (NIRS) [18]. NIRS does not directly measure O2, but rather the level of hemoglobin saturation within a given tissue volume [19]. Among the invasive techniques, the amperometric LICOX® probe by Integra® LifeScience is considered by some to be the gold standard for PbtO2 monitoring [20]. This is a Clark-type sensor comprising cathode and anode electrodes encased in an 80 μm-thick polyethylene membrane [21] across which tissue O2 diffuses to then be detected in an electrochemical reduction reaction, producing an analytical redox current signal. Optical probes such as Neurovent®-PTO by Raumedics® are based on fluorescence quenching by O2 of a probe [22]. Each one of these approaches has advantages and limitations, as expected with any other sensing technology. The amperometric and optical probes for focal PbtO2 and NIRS probes allow real-time monitoring of brain oxygenation in patients with variable temporal and spatial resolution. Both focal probes are invasive and are single site recording devices, and thus positioning of the probe limits the information obtained. On the other hand, NIRS probes are non-invasive and can assess several regions simultaneously but can experience contamination of the signal due to extracerebral circulation. Finally, focal PbtO2 probes are considered the most effective bedside method for detection of cerebral ischemia, which is not true for NIRS probes due to lack of standardization between commercial devices and undetermined threshold for ischemia [23].

Platinum recording surfaces display excellent properties for both stimulation and recording electrodes [24]. Clinical grade intracranial recording electrodes used for invasive monitoring of brain tissue electrical activity are typically composed of Pt on the recording sites [25]. This includes strip and grid subdural multi-electrode devices as well as intracranial multi-site recording electrodes for depth recordings. Considering that Pt displays electrocatalytic behavior toward the electroreduction of O2 [26,27], we propose that clinical grade recording electrodes can be used for amperometric monitoring of PbtO2, an application that has not, to the best of our knowledge, been explored.

In the current work, we have characterized the electrochemical properties of an AuragenTM depth electrode (Integra® LifeScience, Princeton, NJ, USA) approved for brain mapping comprised of 12 cylindrical Pt recording sites. Furthermore, we investigated the analytical performance properties toward the reduction of O2. This is a critical first step towards the scaling up of our previous work in rodent models aimed at establishing fast sampling amperometry coupled to multi-site electrodes as a tool for concurrent electrophysiology and electrochemistry in clinical settings such as the neurocritical care unit and neurosurgery.

#### **2. Materials and Methods**

Reagents and Solutions: All reagents used were analytical grade and obtained from Merck (Algés, Portugal). Unless otherwise stated, all in vitro electrode evaluations were performed in PBS Lite solution, 0.05 M, pH 7.4 with the following composition: 10 mM Na2HPO4, 40 mM NaH2PO4, and 100 mM NaCl. Saturated O2 solutions for electrode calibration were prepared by bubbling PBS with 100% O2 (Air Liquide, Algés, Portugal) for 20 min, resulting in an O2 solution of 1.3 mM concentration at 22 ◦C [28]. Removal of O2 from solutions was achieved by purging with N2 (Air Liquide) for at least 20 min.

AuragenTM Depth Electrode: In the current study we used a clinical grade flexible AuragenTM depth electrode (ref. AU12D5L25) comprising 12 cylindrical Pt recording contacts with 2.5 mm in length and 5 mm spacing between consecutive recording sites (Figure S1) and gold connector contacts. The geometrical area of each lead was calculated to be 0.094 cm2 based on the measured diameter of 1.2 mm. The probe was used with no surface modifications or treatments.

Scanning Electron Microscopy and Elemental Composition Analysis: High-resolution scanning electron microscopy (SEM) was performed using a field emission scanning electron microscope coupled with energy dispersive X-ray spectroscopy (EDS) (Zeiss Merlin coupled to a GEMINI II column). The elemental composition was obtained from backscattered electron detection using EDS at 10 keV (X-Max, Oxford Instruments, High Wycombe, UK). Conductive carbon adhesive tabs were used to ground the electrode surface and secure the sample onto the specimen holder.

Electrochemical Instrumentation: Electrochemical characterization was performed on aMultiPalmSens4 Potentiostat equipped with a MUX8-R2 Multiplexer (PalmSens BV, Houten, The Netherlands) and controlled by MultiTrace software (PalmSens BV, The Netherlands). We used a three-electrode electrochemical cell comprising the depth electrode as working electrode, Ag/AgCl in 3 M NaCl as reference electrode (RE-5B, BAS Inc, West Lafayette, IN, USA) and a Pt wire as auxiliary electrode.

Electrode Calibration: The depth electrodes were calibrated to assess analytical performance toward O2 response. Calibrations were performed in 0.05 M PBS Lite pH 7.4 (20 mL) at room temperature (22 ◦C) with continuous stirring at low speed (240 rpm). Oxygen was removed by purging the solution with N2 for a minimum period of 20 min, after which the needle was removed from the solution and kept above the surface to decrease O2 back-diffusion. Once a stable baseline was obtained, 8.25 μM aliquots of the O2 saturated solution were added in 7 consecutive repetitions (concentration range 0–57.75 μM). The mean recording display frequency was set at 4 Hz.

Data Analyses: Data analyses were performed usingMultiTrace (PalmSens BV, Houten, The Netherlands), OriginPro 2016 (OriginLab, Northampton, MA, USA) and GraphPad 5.0 (GraphPad Software, San Diego, CA, USA). Values are given as the mean ± SD. The number of repetitions is indicated in each individual determination. The sensitivity of depth electrode sites towards O2 reduction was determined by linear regression analysis in the range 0–60 μM. The limit of detection (LOD) was defined as the concentration that corresponds to a signal-to-noise ratio of 3, calculated using the expression:

$$\text{LOD} = 3 \times \text{SD/m}\_{\text{\textdegree}}$$

where SD is the standard deviation of the baseline (20 s interval) and m is the slope of the calibration curve obtained [29].

#### **3. Results**

#### *3.1. Characterization of the Electrode Surface—Morphology and Chemical Analysis*

To evaluate the morphology of the Pt surface of the depth electrode, we obtained SEM micrographs of recording sites. As shown in Figure 1A and B, the Pt surface or the recording site appears to be smooth. Higher amplification revealed that the Pt coverage is not completely uniform throughout the surface (Figure 1C). The elemental composition of the surface of the recording sites was analyzed by energy dispersive X-ray spectroscopy (EDS). As shown in Figure 1D, the active surface is primarily composed of Pt (approx. 80%), although C, O, and Al were also found to be present in lower proportions (approx. 9, 7, and 5%, respectively). To further investigate the composition of the surface, we determined the elemental composition of different regions of the surface, as indicated in Figure 1E. The lighter regions of the SEM (labeled "Spectrum 10" in Figure 1E) are composed primarily of Pt (approx. 93%) with some C contamination (Figure 1F), while the darker regions (labeled "Spectrum 9" in Figure 1E) are composed of Al and O (50 and 44%, respectively, Figure 1G). The pseudo-color map of the relative distribution of different elements on the EDS-analyzed surface (Figure 2) revealed that C is uniformly distributed over the surface and is likely a contaminant. Furthermore, Pt distribution is complimentary

of that of Al and O, which overlap, suggesting that the Pt has been deposited over an aluminum oxide surface (most likely Al2O3), although small regions are not coated with Pt. Comparison of the surface morphology and elemental composition before and after electrochemical evaluation revealed no significant differences, indicating a stable Pt surface (Supplementary Figure S2).

#### *3.2. Electrochemical Active Surface Area*

A standard electrochemical redox couple was used to determine the electrochemical behavior of the Pt surface of the recording sites of the depth electrode. Cyclic voltammetry was carried out in 5.0 mM hexaamineruthenium (III) chloride (Ru(III)(NH3)6) in 0.5 M KCl solution at scan rates from 25 to 200 mV s−1. As shown in Figure 2, the cyclic voltammograms revealed a conventional cyclic voltammetry behavior with well-defined symmetrical oxidation and reduction peaks appearing at 25 mV s<sup>−</sup>1. In addition, both the anodic and cathodic peak currents (*I*p,a and *I*p,c, respectively) varied linearly with the square root of the scan rate (Figure 3 inset; R<sup>2</sup> values of 0.999 for both *I*p,a and *I*p,c) indicating that the process was diffusion-controlled. The average *I*p,a/*I*p,c ratio was 0.8 ± 0.2 (*n* = 12), which is close to the theoretical value of 1 for a totally reversible reaction [30]. The E1/<sup>2</sup> and Epa – Epc values were determined to be 182 ± 3 mV and 72 ± 2 mV (*n* = 12), respectively.

**Figure 1.** (**A**) General view of a recording site on the AuragenTM Probe; (**B**) and (**C**) High-resolution micrographs of the recording surface; (**D**) EDS elemental analysis of the surface (ROI indicated in inset with orange rectangle); (**E**) Different regions of the recording surface were targeted for EDS elemental analysis, indicated with red and yellow "+" signs, respectively; (**F**) and (**G**) show the elemental composition spectrum of the lighter (conductive) regions (predominantly Pt) and of the darker (non-conductive) region (predominantly Al and O), respectively.

**Figure 2.** Pseudo-color map of the relative distribution of different elements on the EDS analyzed surface. Blue—platinum; green—aluminum; red—carbon, yellow—oxygen. Pt and Al distribution are complementary. Note that Al and O overlap, suggesting Al-O as a substrate for the Pt overcoat. Uniform distribution of C is in line with contamination.

**Figure 3.** Reversible redox reaction of Ru(III)(NH3)6 in 0.5 M KCl at increasing scan rates (from 20 mV s−<sup>1</sup> in black to 200 mV s−<sup>1</sup> in green) and respective I*p* vs. ν1/<sup>2</sup> plot for determination of the electrochemical surface area of the depth electrode recording site.

The electrochemically active surface area of the Pt recording sites was estimated using the Randles–Sevick equation for a reversible oxidation-reduction reaction considering a diffusion coefficient of *<sup>D</sup>* <sup>=</sup> 7.1 <sup>×</sup> <sup>10</sup>−<sup>6</sup> cm<sup>2</sup> <sup>s</sup>−<sup>1</sup> [31]. The calculated surface area was 8.5 <sup>×</sup> <sup>10</sup>−<sup>2</sup> <sup>±</sup> 1.0 <sup>×</sup> <sup>10</sup>−<sup>2</sup> cm<sup>2</sup> corresponding to a surface roughness of 0.90 ± 0.1 (*n* = 12). This is in line with the smooth Pt surface observed in the SEM micrographs, as well as the Pt coverage determined from the EDS elemental analysis, which showed that the Pt coverage of the recording site is roughly 90% of the analyzed area, with Al and O making up most of the remaining area.

#### *3.3. Electrochemical Behavior in Acidic Electrolyte and in Neutral PBS*

The well-known characteristic cyclic voltammogram of Pt in acid solution was used to further examine the electrochemical behavior of the Pt recording site of the depth electrode. For this purpose, the probe was characterized by cyclic voltammetry in N2-purged H2SO4 (0.1 M). Figure 4A shows cyclic voltammograms recorded between <sup>−</sup>0.4 and 1.4 V *vs* Ag/AgCl at increasing scan rates (50–1000 mV s<sup>−</sup>1) of a single recording site. The typical cyclic voltammogram exhibited redox peaks at −0.08 and <sup>−</sup>0.2 V. Furthermore, the presence of the three distinct peaks for H<sup>+</sup> desorption was clearly observed. An oxidation wave was observed for *E* > 0.5 V due to the formation of Pt oxide species Pt-O and Pt-OH, and there is a strong reduction peak at about 0.52 V corresponding to oxide reduction.

**Figure 4.** Electrochemical behavior in acidic and neutral electrolyte media. (**A**) Successive cyclic voltammograms (25th scan) at increasing scan rates (50−1000 mV s<sup>−</sup>1) obtained in N2 saturated 0.1 M H2SO4, detailing the typical Pt oxide formation and reduction, proton adsorption (2 peaks) and desorption (3 peaks), and double layer zones. (**B**) Comparative CV plots (0.2 V s<sup>−</sup>1) recorded in N2-saturated 0.05 M, pH 7.4 PBS (black line) and N2-saturated 0.1 M, H2SO4 (red line) highlighting the positive shift in hydrogen evolution potential and increasing currents for Pt-oxide formation and reduction at lower pH on the Pt surface of the Integra Probe recording site.

We further characterized the electrode behavior in a neutral physiological-like media (0.05 M PBS Lite at pH 7.4), which simulates brain extracellular fluid. As shown in Figure 4B, increasing the electrolyte pH resulted in the expected negative shift in hydrogen adsorption/desorption and Pt-O formation/reduction peaks as well as a decrease in peak current width. In both electrolytes, the potential window—that is, the potential range between molecular hydrogen evolution and the evolution of molecular oxygen—is approximately 1.5 V.

#### *3.4. Electrochemical Impedance Spectroscopy*

Electrochemical impedance spectroscopy (EIS) allows the study of the physical and interfacial properties of electrochemical systems. Spectra were recorded in a N2-purged solution containing 5 mM K4[Fe(II)(CN)6] and 5 mM K3[Fe(III)(CN)6] in KCl 0.5 M at room temperature by applying a sinusoidal wave of amplitude 10 mV between 100 kHz and 0.1 Hz (10 frequencies per decade) at the OCP (+0.24 V vs. Ag/AgCl). Before recording each spectrum, the electrode was held at this applied potential for 5 minutes.

The Bode plot (Figure 5A) and complex plane plot (Figure 5B) display the expected profile for a single step charge transfer process with diffusion of the reactants to the electrode surface. The complex plane plot shows the typical capacitive arc at high frequencies followed by a straight line (45◦) at lower frequencies. The data were fitted to the Randles circuit [32] shown in the inset of Figure 5B, and consisting of the cell resistance (R1) in series with a combination of a constant phase element (Q1) in parallel with the series combination of a charge transfer resistance (*R*2) and a Warburg impedance element (*W*). The latter accounts for mass transfer limitations imposed by diffusion, which appear at lower frequencies. The values for the charge transfer resistance, Warburg coefficient and double layer capacitance from fitting to the equivalent electrical circuit are presented in Table 1.

**Table 1.** Summary of fitted parameter results for impedance spectroscopy measurements (*N* = 10) a.


<sup>a</sup> Values shown with \* are normalized by calculated surface area.

The value of *Z* at 1 kHz is typically reported for impedance measurements on electrodes. In this work, the recording sites of the depth electrode showed a *Z* value of 5.4 <sup>±</sup> 1.3 <sup>Ω</sup>·cm<sup>2</sup> (77.4 <sup>±</sup> 20.6 <sup>Ω</sup> before area normalization) at 1 kHz.

#### *3.5. Oxygen Reduction Reaction at the Platinum Surface*

To determine the most suitable working potential for monitoring O2 in vivo, we performed calibrations of the depth electrode at applied potentials ranging from 0.0 to −0.8 V vs. Ag/AgCl. The slopes corrected for electrochemical surface area are plotted in Figure 6A, revealing an increase in sensitivity as the applied potential decreased from 0.0 to −0.8 V. Although it seems tempting to use the highest value of −0.8 V, as seen in Figure 6B, the increase in sensitivity is accompanied by an increase in the baseline current. Considering the expected low values of brain tissue O2 and variations in the μM range, to optimize resolution and LOD, we chose −0.6 V vs. Ag/AgCl as the optimal applied potential. In Figure 6C, we show a representative calibration recording and respective calibration curve (inset). We observed linearity for the concentration range 0–50 μM and mean sensitivity of <sup>−</sup>1.2 <sup>±</sup> 0.2 A M−1·cm−<sup>2</sup> and mean LOD of 0.4 <sup>±</sup> 0.1 <sup>μ</sup>M (*<sup>n</sup>* <sup>=</sup> 4).

**Figure 5.** Electrochemicalimpedance spectroscopymeasurements obtainedin K4[Fe(II) (CN)6]/K3[Fe(III)(CN)6] (5 mM) in KCl 0.5 M. (**A**) Impedance−frequency plot (Bode plot). Filled circles represent |*Z*| values, and open circles are those obtained for the phase shift. The red circle highlights the |*Z*| value at 1 kHz. (**B**) Complex plane electrochemical impedance spectrum (Nyquist plot) of experimental data (open circles). Red line shows fitting to the electrical equivalent circuit shown in the inset. R1, solution resistance; R2, electron or charge transfer resistance; W, Warburg impedance element; Q, constant phase element.

**Figure 6.** *Cont.*

**Figure 6.** Electrochemical behavior of O2 reduction at the depth probe surface. (**A**) Average sensitivity as a function of the applied potential, obtained from calibration of 4 sites in PBS. (**B**) Average baseline current as a function of the applied potential, obtained in N2-purged PBS. (**C**) Representative calibration obtained at −0.6 V vs. Ag/AgCl and the calibration curve (inset) of a single recording site of the Integra probe. *n* = 4.

#### **4. Discussion**

Platinum surfaces are of paramount importance in applications such as those involving neural interfacing and design of electrochemical sensors and biosensors [24]. In the present study, we have investigated the electrochemical properties of a clinical depth recording electrode, in particular its suitability for monitoring *p*O2 in brain tissue.

It is commonly accepted that PbtO2 reflects changes in metabolism and cerebrovascular response and monitoring this parameter has become increasingly standardized in neurocritical care [33]. Currently available probes for monitoring PbtO2 in situ in patients include the Clark-type amperometric Licox® probe and the Neurovent®-PTO optical probe, both of which are single site recording devices [21,22]. Based on our previous experience with multisite Pt microelectrode arrays, we propose that clinical multisite recording electrodes such as the depth probe used here may be suitable for monitoring PbtO2 from multiple sites in the brain, offering a more integrated vision of brain oxygenation.

The morphological evaluation of the recording surface of the depth electrode revealed a smooth surface of Pt over what appears to be an Al-oxide substrate, most likely Al2O3. The Pt coverage was shown not to be complete, accounting for roughly 90% of the surface area. This was further corroborated by determination of the electroactive surface area, which indicated a roughness factor of 0.9.

The electrochemical characterization of the Pt surface performed by cyclic voltammetry in acidic media revealed redox peaks at −0.08 and −0.2 V, which could be attributed to strong and weak proton adsorption on Pt surfaces with (100) and (110) basal planes of a polycrystalline structure, respectively [34,35]. Furthermore, the presence of the three distinct peaks for H<sup>+</sup> desorption indicates a high-quality Pt surface [35,36]. An oxidation wave is observed for *E* > 0.5 V due to the formation of Pt oxide species Pt-O and Pt-OH, and there is a strong reduction peak at about 0.52 V corresponding to oxide reduction. Furthermore, the potential window of 1.5V did not widen in buffered aqueous solution corroborating the smooth structure of the Pt surface as nanostructure surfaces tend to show an increase in the potential width in neutral vs. acid media [35].

The electrochemical impedance spectroscopy revealed that the value of *Z* was 77.4 ± 20.6 Ω at 1 kHz, a value that is slightly lower than that reported for a Pt/Ir depth electrode with similar recording surface area [37], and much lower than that reported in our previous study (0.2 MΩ) for a smooth thin-film Pt microelectrode array [26]. As impedance is inversely proportional to recording site size but should be low as to decrease noise in recording [25,38]. In microelectrodes, where high impedance due to size can become a limiting factor, surface area is typically increased through roughening or functionalization [39,40].

The evaluation of the analytical performance toward the oxygen reduction reaction revealed that the optimal working potential for monitoring O2 is −0.6V vs. Ag/AgCl, in line with our previous observations. However, we found that the sensitivity was slightly lower than our previously reported value for thin-film Pt microelectrode arrays (3.2 <sup>±</sup> 0.5 A M−<sup>1</sup> cm<sup>−</sup>2), while the LOD is in a similar range as previously reported value of 0.3 ± 0.2 μM [26]. This supports the suitability of the present probe to monitor changes in PbtO2 in vivo in brain tissue, where basal values have been found to be approx. 30 μM [41]. We have previously shown that Pt electrode surfaces are optimal for the electrochemical reduction of O2, allowing direct and real-time monitoring of PbtO2 in brain tissue in rodent models [26]. Further validation of PbtO2 monitoring in using this depth recording electrode will be carried out in a swine model.

Considering that brain cerebrovascular function and neurometabolism are coupled to neuronal activity, there is great interest in simultaneous monitoring of electrical activity and PbtO2, as changes in PbtO2 result from both cerebral blood flow and oxidative metabolism [42–44]. Currently available invasive technology requires that two separate probes be implanted in distinct brain regions [45], which greatly hinders the establishment of correlations between changes in tissue oxygenation and neural activity. In line with previous observations [46–48], we demonstrated that high frequency amperometry can concurrently report PbtO2 and local field potential related currents in vivo in a rodent model of seizures [49]. Combined with other studies using enzyme-based biosensors and the same methodological approach, strong evidence supports that the high frequency component (>1 Hz) of an amperometric signal is a surrogate signal for local field potential [47,48,50]. This is an attractive approach toward simultaneous monitoring of electrical and chemical activity in vivo in brain tissue using a seamless methodology (amperometry) and a single probe that may have multiple recording sites [51].

#### **5. Conclusions**

We present an electrochemical evaluation of the analytical properties toward oxygen detection of a clinical grade depth recording electrode comprising 12 Pt recording contacts. Based on the reported catalytical properties of Pt toward the electroreduction reaction of O2, we propose that these probes could be repurposed for multisite monitoring of PbtO2 in vivo in the human brain. We found that the surface of the recording sites is composed of a thin film of smooth Pt and that the electrochemical behavior evaluated by cyclic voltammetry in acidic and neutral electrolytes is typical of a polycrystalline Pt surface. The smoothness of the Pt surface was further corroborated by determination of the electrochemical active surface, confirming a roughness factor of 0.9. At an optimal working potential of −0.6 V vs. Ag/AgCl, the sensor displayed suitable values of sensitivity and LOD, supporting its capability for monitoring PbtO2 in vivo in the brain. The repurposing of these probes from electrophysiology to electrochemical detection of O2 will allow seamless multisite monitoring of PbtO2 in clinical setting, which holds promise in the context of multimodal monitoring in neurocritical care where PbtO2 has become increasingly standardized following evidence of improved patient outcomes when PbtO2 targeted therapeutic approaches are used.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-666X/11/7/632/s1. Figure S1: Photograph and detail of the Aurogen® probe used in the present study; Figure S2: SEM and EDS analysis of depth probe before and after electrochemical evaluation

**Author Contributions:** Conceptualization, A.L., G.A.G., and R.M.B.; methodology, A.L. and R.M.B.; formal analysis, A.L. and E.F.; investigation, A.L. and E.F.; resources J.E.Q. and G.A.G.; writing—original draft preparation, A.L.; writing—review and editing, J.E.Q., G.A.G., and R.M.B.; supervision, A.L. and R.M.B.; funding acquisition, R.M.B. and G.A.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was financed by the European Regional Development Fund (ERDF) through the COMPETE 2020 – Operational Programme for Competitiveness and Internationalization and Portuguese national funds via FCT – Fundação para a Ciência e Tecnologia, under projects POCI-01-0145-FEDER-028261 and UIDB/04539/2020. **Acknowledgments:** Additional support was provided through the University of Kentucky Brain Restoration Center and the Integra electrodes were provided in-kind by Integra LifeSciences.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Mechanically Robust, Softening Shape Memory Polymer Probes for Intracortical Recording**

**Allison M. Stiller 1,\*, Joshua O. Usoro 1, Jennifer Lawson 1, Betsiti Araya 1, María Alejandra González-González 2, Vindhya R. Danda 3, Walter E. Voit 1,3,4, Bryan J. Black <sup>1</sup> and Joseph J. Pancrazio <sup>1</sup>**


Received: 21 May 2020; Accepted: 23 June 2020; Published: 25 June 2020

**Abstract:** While intracortical microelectrode arrays (MEAs) may be useful in a variety of basic and clinical scenarios, their implementation is hindered by a variety of factors, many of which are related to the stiff material composition of the device. MEAs are often fabricated from high modulus materials such as silicon, leaving devices vulnerable to brittle fracture and thus complicating device fabrication and handling. For this reason, polymer-based devices are being heavily investigated; however, their implementation is often difficult due to mechanical instability that requires insertion aids during implantation. In this study, we design and fabricate intracortical MEAs from a shape memory polymer (SMP) substrate that remains stiff at room temperature but softens to 20 MPa after implantation, therefore allowing the device to be implanted without aids. We demonstrate chronic recordings and electrochemical measurements for 16 weeks in rat cortex and show that the devices are robust to physical deformation, therefore making them advantageous for surgical implementation.

**Keywords:** intracortical microelectrode arrays; shape memory polymer; softening; robust

#### **1. Introduction**

Intracortical microelectrode arrays (MEAs) are devices that can be implanted in brain tissue to stimulate or record electrical activity from surrounding neural populations [1], making them important tools for investigating the function of the nervous system [2]. Additionally, intracortical recording MEAs have been used as critical components in brain–machine interfaces (BMIs) [3–5], which may be used to restore or replace loss of motor function in patients suffering from paralysis, limb loss, or neurodegenerative disorders. Despite the promise of MEA technology, the clinical adoption of these devices has been limited for many reasons, several of which are associated with stiff material composition. Commercially available devices are often fabricated from high modulus materials, such as silicon, to leverage reproducible photolithography techniques. However, this material choice also results in structures that are brittle due to the small device dimensions necessary to mitigate a severe chronic neuroinflammatory response, which may contribute to behavioral deficits [6]. This fragility renders devices vulnerable to fracture during fabrication [7] and handling [8,9]. Even after implantation, the device may be susceptible to breakage or cracking [10] due to the tethering forces caused by constant micromotion of the brain [11]. This effect may be exacerbated by the high degree of mechanical mismatch between the implanted device and the surrounding tissue, which creates a constant source of mechanical strain at the brain–device interface [12–14]. Because of this, many groups are investigating the potential use of robust, yet softer, polymer materials for intracortical device fabrication.

Unfortunately, there are many inherent issues associated with fabricating functional devices from soft materials. From a mechanical standpoint, polymer-based devices may lack the structural rigidity required to successfully implant in brain tissue if they are fabricated with small cross-sectional dimensions. To alleviate this issue, there have been demonstrations of the use of insertion guides to aid in implantation [15], but implementation of these aids may complicate surgical use. Others have addressed structural instability with stiffening coatings that dissolve after implantation [16–18], but these may increase the footprint of the device during implantation, thus increasing the risk of disrupting vasculature. Additionally, while many polymers such as SU-8 and Parylene C exhibit an elastic modulus of 1–5 GPa, approximately 2–3 orders of magnitude lower than that of materials commonly used for intracortical device fabrication, this is still 5–6 orders of magnitude higher than the estimated modulus of brain tissue (~0.5–10 kPa [19,20]). With these drawbacks in mind, there is motivation to design a robust polymer-based device that can be handled and implanted using existing surgical techniques, but that exhibits material properties that more effectively bridge mechanical mismatch with the brain.

Our group has previously investigated the use of shape memory polymers (SMPs) for fabrication of intracortical devices [21]. SMPs are a class of material that are unique in their ability to undergo dynamic changes in material properties in response to stimuli from the environment [22–24]. Specifically, we used a thiol-ene/acrylate formulation softened from 2 GPa in dry, room temperature conditions to ~300 MPa in wet, body-temperature conditions. This range ensured that devices could maintain structural stability during implantation, but soften by an order of magnitude in a few minutes after making contact with the brain tissue. This thiol-ene/acrylate SMP softens primarily due to an increase in temperature and minimally due to water absorption (<3% by weight). In this study, we used photolithography to create reproducible SMP based devices with the ability to soften to 20 MPa after implantation, an order of magnitude softer than our previously reported functional softening biolectronics. We demonstrate that these devices are capable of intracortical recording, show consistent electrical performance after substantial mechanical perturbation, and induce a tissue response consistent with similarly sized silicon shanks. Furthermore, we provide pilot data indicative of a reduced behavioral deficit associated with the SMP based devices.

#### **2. Materials and Methods**

#### *2.1. Polymer Preparation and Device Fabrication*

The SMP substrate was prepared as previously described [25] and material properties were determined using dynamic mechanical analysis (DMA), a methodology that applies cyclic, tensile loading to samples while changing the temperature in dry or aqueous conditions. DMA results indicated that the SMP exhibited an elastic modulus of 2 GPa in dry, room-temperature conditions and 20 MPa in soaked, body-temperature conditions. The pre-polymer solution was prepared using 1,3,5-triallyl-1,3,5-triazine-2,4,6 (1H,3H,5H)-trione (TATATO), trimethylolpropane tris(3-mercaptopropionate) (TMTMP), and (2-(3-mercaptopropionyloxy) ethyl) isocyanurate (TMICN) at molar ratios of 0.5:0.45:0.05, respectively, with 0.1 weight percent 2,2-dimethoxy-2-phenylacetophenone (DMPA), a photoinitiator. The solution was spun onto 4 inch silicon wafers (test grade, laser scribed, P-type, boron-doped wafers) at 600 rpm for 30 s and subsequently cured under UV light for 3 min at 254 nm and 1 h at 365 nm. The polymerized samples were then transferred to an oven for an overnight vacuum bake at 120 ◦C. This spin coating process yielded a 29 μm thick SMP layer on the silicon wafers. All processing, with the exception of SMP spin-coating and device release, was done in a Class 10,000 cleanroom at the University of Texas at Dallas (UTD). All photomasks needed for device fabrication were designed using AutoCAD and fabricated in the above cleanroom using a Heidelberg DWL66 laser lithography system (Heidelberg, Germany). The SMP layer was cleaned with deionized water followed by a nitrogen dry and dehydration bake at 115 ◦C in a vacuum oven for 15 min to promote adhesion. A 1 μm layer of Parylene C was then deposited on the bottom SMP layer to act as an insulator, using a SCS PDS 2010 coater (Specialty Coating Systems, Indianapolis, IN, USA). Gold interconnects (5 μm wide) were patterned by liftoff lithography using nLOF photoresist and low stress silicon nitride (900 mT, 150 ◦C, 100 W with SiH4, NH3, H3, and N2 gases at 280, 4600, and 200 sccm, respectively). A 400 nm gold thin film was electro-beam evaporated on the wafer. A secondary 1 μm layer of Parylene C was deposited, and the wafers were patterned using standard lithography techniques and etched using oxygen plasma to form a coaxial layer around the gold traces. The wafers then went through a dehydration bake at 115 ◦C for 15 min and immediately received a 7 μm thick top coat of the SMP using spin parameters of 1200 rpm for 2 min. They were then returned to the cleanroom for additional processing. Electrode sites, bond pads, and the device outline were patterned using standard photolithography methods. A hard mask of ~200 nm low stress silicon nitride was used to perform the plasma etching of SMP to open the electrode sites and bond pads and create the device outline. After etching, the hard mask was stripped off using hydrofluoric acid, and electrode sites (18 × 10 μm) were patterned with liftoff lithography using AZ400K solvent at room temperature. Next, the electrode sites received sputtered iridium oxide film (SIROF) to decrease electrode site impedance. First, a 20–50 μm layer of titanium was deposited (200 V, 4 Torr, with Ar flow at 50 sccm) to act as an adhesion layer for the SIROF. Next, a 160–180 μm layer of iridium oxide was sputtered onto the electrode sites (100 V, 30 Torr, with Ar, O2, and 2% H2/98% Ar at 15, 20, and 10 sccm, respectively). After the fabrication was complete, wafers were soaked in DI water in a 37 ◦C oven until individual devices released from the wafers. Released devices were assembled with packaging with Omnetics connectors (A79040-001, Omnetics Connectors Corporation, Minneapolis, MN, USA) by Qualia Labs, Inc. (Figure 1a). A thicker variation of the SMP devices were also fabricated for the purposes of mechanical testing. In this case, all fabrication steps were identical with the exception of a 29 μm thick top SMP layer in place of the 7 μm layer (Figure 1b). Size-matched non-functional silicon devices were also fabricated in UTD cleanroom facilities using standard photolithography techniques (Figure 1c).

**Figure 1.** SMP and silicon device design. (**a**) Optical image of a packaged SMP device with the red inset depicting 16 electrode sites on an SMP device (the electrodes appear to be black due to the SIROF coating). The Omnetics connector is sealed off with epoxy (grey encapsulation material). (**b**) Representative cross-sectional view of a gold trace and coaxial Parylene C insulation in both device variations. (**c**) A size-matched non-functional silicon device mounted in a zero insertion force (ZIF) connector.

#### *2.2. Surgical Implantation*

All animal handling, housing, and surgical procedures were approved by the University of Texas Institutional Animal Care and Use Committee. Male, Long Evans rats (n = 8, Charles River) weighing 600–800 g received bilateral implants of functional SMP devices and non-functional silicon shanks.

Surgical methods followed those previously reported [21]. Briefly, animals were anesthetized with an intraperitoneal injection of a ketamine cocktail (ketamine (65 mg/kg), xylazine (13.33 mg/kg), and acepromazine (1.5 mg/kg)) followed by an intramuscular injection of atropine sulfate (0.05 mg/kg) to counteract depressive cardiac effects of the ketamine and a subcutaneous injection of dexamethasone (2 mg/kg) to help mitigate post-operative inflammation. When unconscious, the animal's head was shaved and then mounted in a stereotaxic frame with a nose cone delivering 2% isoflurane in 100% oxygen for maintenance of the anesthetic plane. Ophthalmic ointment was applied to the eyes to retain moisture and prevent drying during the surgery. The scalp was cleaned alternating with 10% iodine solution and 70% ethanol wipes three times and then injected with 0.4 mL of 0.5% lidocaine to numb the area. An incision was made using a surgical blade down the midline of the scalp, and hemostatic forceps were used to hold the skin away from the skull. The surface of the skull was thoroughly cleaned and roughened to promote adhesion of the head cap.

First, a surgical drill was used to create three small holes in the skull for positioning of stainless steel screws. These screws provided mechanical support for the head cap but also served as ground and reference connections for external wiring on the functional SMP device. Next, 1–2 mm craniotomies were drilled over each motor cortex, centered approximately 2.5 mm rostral and 2.5 mm lateral from bregma (Figure 2a). The dura was resected over one craniotomy and a device was positioned over the center (Figure 2b). The device was implanted at a speed of 1 mm/s using a hydraulic micropositioner (Kopf Instruments, Tujunga, CA, USA) to a depth of 2 mm. The dura was then replaced with collagen-based grafts (Biodesign Dural Graft, Cook Medical, Bloomington, IN, USA). After the grafts were rehydrated by the surrounding cerebrospinal fluid, a layer of Gluture (World Precision Instruments, Sarasota, FL, USA), a biocompatible cyanoacrylate, was used to seal off the craniotomy. After implantation of the SMP device, ground and reference wires were wrapped around stainless steel screws while waiting for the cyanoacrylate to dry. Finally, a small amount of dental cement was applied around the device base in order to stabilize it during implantation of the second device. This process was then repeated for the second device in the contralateral cortex. Following both implantations, the head cap was built up to encapsulate the devices, leaving only the Omnetics connector visible. The skin behind the implantation sites was closed using surgical staples and left in place for 10 days following the operation. The animal then received 2–3 mL of sterile saline for hydration, a subcutaneous injection of sustained release buprenorphine SR LAB (ZooPharm, Windsor, CO, USA), and an intramuscular injection of cefazolin antibiotic (5 mg/kg). The animal received a follow-up shot of buprenorphine 72 h after surgery.

**Figure 2.** Device implantation. (**a**) Schematic depicting bilateral implant locations (red crosses) in rat brain. The circled plus signs indicate placement of stainless steel screws. (**b**) SMP shank held over a craniotomy with resected dura immediately before implantation. Scale bar = 1 mm.

#### *2.3. Electrophysiological Recordings and Electrochemistry*

We performed weekly electrophysiological recordings and electrochemistry on anesthetized animals (1–2% isoflurane) starting immediately after surgery, at Week 0, for 16 weeks. Spontaneous wideband data were recorded for 10 min at 40 kHz with functional SMP devices using a 16-channel Plexon headstage and Omniplex acquisition system (Plexon, Inc., Dallas, TX, USA). A four-pole Butterworth high pass filter with a cutoff frequency of 250 Hz was applied to the wideband data, and spikes were manually sorted using 2D principal component space after identifying potential spikes that crossed a threshold of –4 σ (based on the root mean square (RMS) of the filtered signal). Units were only considered for analysis if they contained at least 100 spike waveforms. A custom MATLAB code was used to extract mean peak-to-peak voltage (Vpp), RMS noise, and signal to noise ratio (SNR) calculated by dividing Vpp by two times the RMS noise of the associated channel.

Electrochemical impedance spectroscopy (EIS) was performed using a model 604E series electrochemical analyzer/workstation (CH Instruments, Inc., Austin, TX, USA). An 18-pin dual strip Nano-D female connector (NSD-18-WD-18.0-C-GS, Omnetics Connector Corporation, Minneapolis, MN, USA) was attached to the SMP Omnetics connector to interface with the system. The device wires wrapped around the stainless steel screws were shorted together to act as a reference electrode and the SIROF-coated electrodes served as the working electrodes. A 10 mV RMS sinusoidal signal ranging from 100 kHz to 1 Hz was sent to the device, and current was recorded 3 times per decade of frequency. CH Instruments software calculated impedance magnitude, and a custom MATLAB code extracted real impedance values at each point.

#### *2.4. Behavioral Testing*

In an effort to compare the potential behavioral changes associated with implantation of SMP and silicon shanks, we conducted pilot experiments using the cylinder test. The cylinder test takes advantage of rodents' natural affinity towards exploring novel environments to determine changes in limb-use asymmetry and overall exploratory behavior over time [26,27]. Rats were placed in a transparent plastic cylinder and were allowed to freely explore for 3 min while being filmed from underneath through a clear acrylic sheet (Figure 3a). The number of times the left paw, right paw, or both paws, were used to initially reach out and touch the walls of the cylinder was manually counted from the video (Figure 3b). Initial paw preference was determined by baseline measurements that were taken prior to surgical implantation. Cylinder testing was performed weekly beginning one week post-implantation. For analysis, rats were grouped based on initial paw preference and the type of

device that was implanted contralateral to that paw. The percent use of each paw was then tracked over the 16-week study.

**Figure 3.** Cylinder test setup. (**a**) The rat is filmed from underneath as it rears up. (**b**) The rat rears up to explore its surroundings. The yellow circle indicates a paw touch against the cylinder.

#### *2.5. Immunohistochemistry and Analysis*

After 16 weeks, animals were sacrificed with an intraperitoneal injection of sodium pentobarbital (200 mg/kg). Tissue fixation was performed by transcardial perfusion with 4% paraformaldehyde (PFA). Whole brains were dissected, sectioned into hemispheres, and stored in PBS + 0.5% sodium azide at 4 ◦C. For cryoprotection, brains were incubated overnight in 15% sucrose (Sigma-Aldrich, St. Louis, MO, USA) solution in PBS and the following night in a 30% sucrose solution. The samples were then transferred to cryomolds filled with optimal cutting temperature compound (OCT, Sekura Finetek, Torrance, CA, USA). Flash freezing was then achieved by submerging cryomolds containing individual hemispheres into dry ice plus isopropyl alcohol. Once the OCT-embedded samples were frozen, they were placed in a −20 ◦C freezer until sectioned. Individual brain hemispheres were horizontally sectioned into 50 μm slices using a Leica CM3050 S cryostat and adsorbed onto charged slides (Fisher Scientific, Waltham, MA, USA) for staining.

Cryosections were permeabilized using 0.5% Triton X-100 (Sigma-Aldrich, USA) in 4% normal goat serum (NGS) (Abcam, Cambridge, UK) and 1× PBS and then blocked in a solution of 0.5% TritonX, 4% NGS, and 1× PBS. Sections were then incubated with mouse anti-neuronal nuclei (NeuN, 1:200, Abcam, ab104224), chicken anti-glial fibrillary acidic protein (GFAP, 1:200, Abcam, ab4674), and rabbit anti-CD-68 (1:200, Abcam, ab125212) diluted in a solution of 4% NGS and PBS for 24 h on a rocker at 4 ◦C. Following primary incubation, tissue was washed 3 times for 30 min using PBS + 0.5% Triton X-100 solution and then washed with 4% NGS + PBS solution for three hours. Sections were then incubated overnight on a rocker at 4 ◦C with secondary antibodies: goat anti-chicken Alexa Fluor 647 (1:200, Abcam, ab150171), goat anti-rabbit Alexa Fluor 488 (1:200, Abcam, ab150077), and goat anti-mouse Alexa Fluor 555 (1:200, Abcam, ab150118), along with 0.06% concentration DAPI (0.06%, Abcam, ab228549). Sections were then washed 3 times for 10 min in PBS solution. The PBS was replaced with drops of Image-iT (Invitrogen, Camarillo, CA, USA), and then sections were sealed with coverslips using clear nail polish.

Stained slides were imaged at 10× magnification on an inverted Nikon eclipse Ti (Nikon, Tokyo, Japan) scanning confocal microscope. With the implantation site centered in the field of view, 4-color z-stack confocal images were collected across the entire tissue slice thickness in 5 μm axial steps. Care was taken to adjust laser line and sensory gains such that no saturating pixels were

detected. Each laser line was scanned individually to reduce signal bleed-over. Z-stack images were then converted to a single maximum intensity projection image for analysis. Normalized GFAP and CD68 fluorescence intensities and normalized NeuN+ cell counts were carried out using boutique ImageJ [28,29] macros, as previously described [30] with minor alterations. Briefly, any regions of the image that were dark due to absence of tissue (including tears or edges) were manually excluded and converted to "not a number" (NaN) values. Shank sites were then manually defined to enable analysis within concentric contours from the site. Average mean intensity values were then automatically calculated within concentric bands of 50 μm up to 400 μm from the shank site. Lastly, based on user-defined noise tolerances, a 3 pixel Gaussian blur was applied to the NeuN channel image and cell counts were automatically collected based on local maxima detection. GFAP intensity measurements and NeuN+ cell counts were then normalized to the furthest concentric band (350–400 μm).

#### *2.6. Device Physical Robustness*

We tested physical robustness of functional SMP devices compared to non-functional silicon devices using electrochemical endpoints after physical deformation. For these experiments, we tested an updated version of the device that featured a thickness of 58 μm but kept all other parameters constant. These devices were also tested in a subset of animals to demonstrate improved mechanical stability during implantation as compared to the thinner design. Before device bending, we performed EIS on functional SMP devices in 1× PBS, pH 7.4, at room temperature using methods identical to those described in Section 2.3. After testing, the devices were dipped in deionized water to remove any salts that accumulated on the electrode sites during testing and allowed to dry for 24 h.

Next, non-functional silicon devices were mounted on a stage such that the shank was parallel to the ground (Figure 4). The tip of the shank was then deflected at a speed of 1 mm/s using a custom attachment mounted to a NeuralGlider cortical neural implant inserter (Actuated Medical, Inc., Bellefonte, PA, USA). As soon as the shank broke, the system was manually stopped, and the displacement distance was recorded. This was repeated for n = 3 samples, and the average distance was recorded and used as an endpoint for deflection of the SMP devices. After drying completely, the SMP devices were mounted in an identical fashion to the silicon devices. The shank tips were displaced to the distance at which the silicon shanks broke at a speed of 1 mm/s. The SMP devices then underwent another round of EIS testing to allow for comparison of impedance magnitude values before and after deformation. Impedance magnitudes before and after physical deformation were then compared. Electrodes with an impedance magnitude greater than 1 MΩ at 1 kHz (approximately double the expected impedance) before deflection testing were excluded from analysis.

**Figure 4.** Device robustness testing setup. The device is mounted via the connector as a cantilevered beam on a stage with the entire length of the shank suspended parallel to the ground. The tip of the device is deflected downward at a speed of 1 mm/s (indicated by downward arrow) and the shank deforms accordingly (dashed line). In the case of the silicon shank, fracture occurs after a certain displacement (d). This displacement is then transiently applied in the same manner to the SMP devices.

#### *2.7. Statistics*

All statistics were performed in MATLAB R2018b (The MathWorks, Inc., Natick, MA, USA) and OriginPro 2020b (Origin Lab, Northampton, MA, USA). Significant increases/decreases (ρ < 0.05) in electrophysiological and electrochemical data over time were determined using analysis of variance (ANOVA) tests on residuals. ANOVA tests were performed to determine significant differences (ρ < 0.05) in immunohistological outcomes between brain slices containing silicon and SMP probes. A paired t-test was used to determine significant difference (ρ < 0.05) between EIS measurements taken before and after SMP probe deflection during robustness testing.

#### **3. Results**

#### *3.1. Single Unit Recordings and In Vivo Electrochemistry*

We performed weekly electrophysiological recordings from functional SMP devices in rat cortex and tracked single-unit amplitude, active electrode yield percentage (AEY), RMS noise, and signal-to-noise ratio (SNR) over 16 weeks (Figure 5). Results showed stable measurements for single-unit amplitude (Figure 5c), a decrease in RMS noise (Figure 5e), and an increase in signal-to-noise ratio over time (Figure 5f). However, the percentage of active electrodes, while relatively stable for the first 12–13 weeks, showed a decline in the last month of recording (Figure 5d). While one device failed to record single units during every recording session, 7 out of 8 devices exhibited active electrodes during each recording session.

**Figure 5.** Chronic electrophysiological recordings. (**a**) Representative filtered data from three electrodes on one device. (**b**) Representative single unit activity on one electrode. We extracted measurements for single-unit amplitude (**c**), active electrode yield (**d**), RMS noise (**e**), and signal-to-noise ratio (**f**) from filtered wideband data. Linear regression analysis showed no change in single-unit amplitude over time. There was a decrease in active electrode yield (R2 = 0.03, ρ = 0.03) and RMS noise (R<sup>2</sup> = 0.09, <sup>ρ</sup> <sup>&</sup>lt; 0.01) and an increase in SNR (R<sup>2</sup> <sup>=</sup> 0.02, <sup>ρ</sup> <sup>&</sup>lt; 0.01) over time. Data are represented as mean <sup>±</sup> SEM.

We also performed weekly EIS measurements to assess electrode electrochemical stability. Results showed a slight decrease in impedance magnitude 1 kHz over 16 weeks with average values around 600 kΩ on Week 0 and 500 kΩ on Week 16 (Figure 6). Because of the ultra-softening nature of this SMP, we did not perform in vitro EIS on devices before implantation to premature

material deformation. However, EIS measurements were taken on a small cohort of devices from the same fabrication batch to ensure impedance magnitudes were within an acceptable range (100–900 kΩ) on a majority of electrodes. In vitro measurements yielded impedance magnitudes (at 1 kHz) of ~100–250 kΩ, indicating that electrode impedance increased immediately after implantation.

**Figure 6.** Impedance magnitude at 1 kHz over 16 weeks. Linear regression analysis shows a slight decrease in impedance over time (R2 <sup>=</sup> 0.03, <sup>ρ</sup> <sup>&</sup>lt; 0.01). Data are represented as mean <sup>±</sup> SEM.

#### *3.2. Device Physical Robustness*

We subjected non-functional silicon shanks to deflection until fracture and then applied the same single displacement of approximately 1.92 mm to functional SMP devices. The mechanical deformation caused a minor, yet statistically significant change in the impedance magnitude at 1 kHz across the microelectrode sites of all tested devices. The mean impedance magnitude before deformation was 539.6 ± 39.0 kΩ, rising to 583.8 ± 46.9 kΩ following deformation, a 7.6% increase (n = 31 microelectrode sites, ρ = 0.013). Similar results were observed at 10 kHz, which showed 7.5 ± 1.2% (ρ < 0.001) increases after deformation. Increases in impedance at 100 Hz, 6.9 ± 3.9% were not statistically significant. Regardless, the electrochemical characteristics of the SMP microelectrode sites exhibited tolerance to mechanical deformation (Figure 7).

**Figure 7.** Representative EIS results from a single SMP device before and after deformation show minimal changes in impedance magnitude at 100 Hz, 1 kHz, and 10 kHz due to device deformation. The electrodes on this device exhibited an average decrease in impedance magnitude of −0.05 ± 4.2% (mean ± SEM) at 100 Hz (n = 16, ρ = 0.84), and an average increase in impedance magnitude of 2.8 ± 2.4% at 1 kHz (n = 16, ρ = 0.46) and 7.2 ± 1.2% at 10 kHz (n = 16, ρ < 0.001). The electrode layout was such that electrode 1 was located at the tip of the shank and electrode 16 was furthest up the shank.

#### *3.3. Immunohistochemistry*

To assess the extent of neuroinflammation after 16 weeks, we performed immunohistochemistry to stain for markers associated with three cell types commonly implicated in the chronic neuroinflammatory response: GFAP for astrocytes, NeuN for neuronal nuclei, and CD68 for activated microglia/macrophages (Figure 8). We also measured the size of the hole left behind by the explanted device in each slice. Histological results were quantified for silicon and SMP devices at three depths along the device shank. Intensity or density values for each band were normalized to the band 350–400 μm from the perimeter of the device. ANOVA results indicated no significant differences in GFAP intensity, neuronal density, and CD68 intensity between silicon and SMP shanks at any comparison point (band or depth). There was also no significant difference in hole size between slices with silicon and SMP shanks at each depth; however, the hole size at the top layer (for both device types) was significantly larger than that found at the middle or deep slices. This trend was expected due to the tapered geometry of the shanks. Similarly, for both GFAP and CD68 intensity, we observed a trend of decreased astrogliosis and microglia/macrophage activation deeper along the shank, likely also in response to decreasing cross-section of the device. There appeared to be no loss of neuronal density in any band at any depth. Separate brain slices were also stained for immunoglobulin G (IgG), a marker associated with blood–brain-barrier leakage, but ANOVA results indicated no significant differences at any comparison point.

**Figure 8.** Representative immunofluorescent images for GFAP, NeuN, CD68, and DAPI stain in a middle brain slice containing a silicon device (top row) and an SMP device (bottom row). Scale bar = 100 μm.

#### *3.4. Pilot Behavioral Deficit Data*

All animals were assessed weekly for motor deficit through use of the cylinder test. Their initial preferred paw was determined by majority percentage of paw contacts from baseline data taken a week before device implantation. Figure 9 demonstrates percentage of paw contacts as a function of device type implanted contralateral to the preferred paw. Unfortunately, due to randomized placement of devices, there was only one animal with a silicon device implanted contralateral to its preferred paw. However, these preliminary data suggest that animals with the SMP device implanted contralateral to their preferred paw continued using this paw, whereas the opposite was true for the animal with a silicon shank implanted contralateral to the preferred paw.

**Figure 9.** Percent of paw contacts as a function of device type implanted contralateral to the preferred paw (based on baseline data). SMP probe data are represented as mean ± SEM.

#### **4. Discussion**

Functional 16-channel SMP devices were implanted in rat motor cortexes for 16 weeks and their performance was tracked with weekly electrophysiological and electrochemical measurements. Within the first few weeks of implantation, active electrode yield decreased and impedance magnitude at 1 kHz increased; however, both of these trends began to reverse by week 4. Similar results were seen in our prior work [21] and may suggest onset and subsequent resolution of the acute neuroinflammatory response. While impedance magnitude remained stable for the remainder of the study, active electrode yield began dropping off in the last month of recordings. Though stable impedance values indicate that the integrity of the recording sites was likely maintained, it is possible that some degree of polymer breakdown could have contributed to decreased active electrode yield. Previous in vitro testing has validated that this variation of SMP is not hydrolytically stable and may experience degradation after chronic soaking [31]. Because of this, we are investigating a novel SMP formulation that is hydrolytically stable [31] for future chronic studies.

These updated devices were also used to demonstrate mechanical robustness in bend tests. SMP devices were able to withstand physical deformations that resulted in brittle fracture of silicon counterparts, a common issue that can make them difficult to handle during fabrication, packaging, and sterilization, and during surgical implantation. Polymers provide a clear advantage in that they may undergo large deformations but maintain integrity of electrical contacts. Furthermore, SMP-based devices may exhibit an even greater robustness to physical deformation when in the softened state due to the increase in flexibility, or decrease in stiffness, of the device. For example, the approximate stiffness of all discrete device components (gold electrode traces, Parylene C encapsulation, and SMP substrate) can be calculated using the modulus, area moment of inertia, and length of the individual components. SMP in the dry state exhibits a stiffness of approximately 1 N/m, but decreases to 0.01 N/m after softening. For comparison, an individual gold trace and coaxial Parylene C insulation exhibits a stiffness of approximately 10−<sup>7</sup> (N/m) and 10−<sup>6</sup> (N/m), respectively, meaning that neither the gold nor the Parylene C should inhibit robustness to physical deformation. The improved flexibility of this SMP formulation compared to that which was previously investigated (0.01 N/m versus 0.1 N/m, respectively) [21] is notable, considering the role that device stiffness may play in chronic functionality. The brain experiences continuous micromotion [11], thus warranting the need for a device that can

endure constant mechanical disturbances. This need extends even further to brain macromotion, which may require even more flexibility of devices, especially in larger animal models or humans.

Additionally, given that SMP devices elicit a minimal neuroinflammatory response that is no worse than size-matched silicon devices, they may be advantageous alternatives devices for chronic use. Finally, preliminary behavioral outcomes indicated that softer devices might mitigate motor defects associated with implantation of stiff, silicon-based devices. More comprehensive analyses are required to examine the benefit of softening devices relative to behavioral measures.

#### **5. Conclusions**

In this work, we demonstrated fabrication and implementation of an SMP-based, 16-channel intracortical device that soften from ~2 GPa to ~20 MPa after implantation. We demonstrated an active electrode yield of 20–30% within the first 3 months of recordings with a final active electrode yield of ~10%. Weekly EIS measurements showed electrochemical stability over 16 weeks. Additionally, chronic immunohistological outcomes showed that SMP devices elicited a response no worse than bilaterally implanted silicon devices, making them good candidates for surgical implementation. Moreover, deformation testing indicated that SMP devices are mechanically robust as compared to silicon counterparts that are vulnerable to brittle fracture, further indicating that SMP offers material advantages for fabrication and handling.

**Author Contributions:** Conceptualization, A.M.S., B.J.B., W.E.V., and J.J.P.; methodology, A.S., J.O.U., J.L., M.A.G.-G., V.R.D., and B.J.B.; software, A.M.S., J.O.U., and B.J.B.; validation, A.M.S. and B.J.B.; formal analysis, A.M.S., J.O.U., and B.J.B.; investigation, A.M.S., J.O.U., B.A., J.L., and B.J.B.; resources, J.J.P.; data curation, A.M.S., J.O.U., B.A., J.L., and B.J.B.; writing—original draft preparation, A.M.S.; writing—review and editing, A.M.S., J.O.U., J.L., B.A., M.A.G.-G., V.R.D., B.J.B., J.J.P., and W.E.V.; visualization, A.M.S., J.O.U., and B.J.B.; supervision, J.J.P.; project administration, A.M.S. and J.J.P.; funding acquisition, J.J.P. and W.E.V. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Office of the Assistant Secretary of Defense for Health Affairs through the Peer-Reviewed Medical Research Program, grant number W81XWH-15-1-0607, DARPA Phase II SBIR, grant number D17PC0011, and the Eugene McDermott Graduate Fellowship Program, grant number 201606 at The University of Texas at Dallas.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Microelectrode Array based Functional Testing of Pancreatic Islet Cells**

**Ahmad Alassaf 1,2,3, Matthew Ishahak 1,2, Annie Bowles 1,2 and Ashutosh Agarwal 1,2,\***


Received: 11 February 2020; Accepted: 15 May 2020; Published: 17 May 2020

**Abstract:** Electrophysiological techniques to characterize the functionality of islets of Langerhans have been limited to short-term, one-time recordings such as a patch clamp recording. We describe the use of microelectrode arrays (MEAs) to better understand the electrophysiology of dissociated islet cells in response to glucose in a real-time, non-invasive method over prolonged culture periods. Human islets were dissociated into singular cells and seeded onto MEA, which were cultured for up to 7 days. Immunofluorescent imaging revealed that several cellular subtypes of islets; β, δ, and γ cells were present after dissociation. At days 1, 3, 5, and 7 of culture, MEA recordings captured higher electrical activities of islet cells under 16.7 mM glucose (high glucose) than 1.1 mM glucose (low glucose) conditions. The fraction of the plateau phase (FOPP), which is the fraction of time with spiking activity recorded using the MEA, consistently showed distinguishably greater percentages of spiking activity with high glucose compared to the low glucose for all culture days. In parallel, glucose stimulated insulin secretion was measured revealing a diminished insulin response after day 3 of culture. Additionally, MEA spiking profiles were similar to the time course of insulin response when glucose concentration is switched from 1.1 to 16.7 mM. Our analyses suggest that extracellular recordings of dissociated islet cells using MEA is an effective approach to rapidly assess islet functionality, and could supplement standard assays such as glucose stimulate insulin response.

**Keywords:** islets of Langerhans; insulin secretion; microelectrode array (MEA); glucose stimulated insulin response; electrochemical transduction

#### **1. Introduction**

Islets of Langerhans are three-dimensional (3D) multicellular clusters that range from 50 to 500 μm in diameter [1,2]. As the functional unit of the pancreas, islets maintain glucose homeostasis through the interdependent secretion of hormones from α, β, δ, and γ cells [3]. Dysfunction of the insulin-secreting β cells, arising from autoimmune destruction or insulin-resistance, results in the development of diabetes mellitus [4]. To elucidate the pathophysiology of diabetes, new approaches are being employed to study islets at the cellular level.

The electrophysiology of β cells has been investigated using a patch-clamp technique [5–7]. However, the disadvantages to performing the patch clamp technique are the invasiveness to the sample, technical complexity, and limited recording time (hours) [8]. Microelectrode arrays (MEAs), on the other hand, have been employed to collect electrical activities of islets [9–13]. The advantages of MEAs are the non-invasiveness to the sample, ease of execution, and higher duration of recording time (days) [13,14]. Pfeiffer et al. performed extracellular recordings of whole islets using MEA, a glass holding pipette angled at 30◦, and a micromanipulator to control the islet location on top of the

recording electrode. They were able to show continuous bursts of spikes at high glucose concentration (15 mM) and concluded that the length of the bursts correlated with the amount of insulin released. The fraction of a plateau phase (FOPP), which is the fraction of time with spiking activity recorded using the MEAs was developed as a metric for beta-cell function [9]. Schonecker et al. and Brouwer et al. also used comparable methods on whole islets to confirm that extracellular recordings correlated with intracellular electrical recordings. Phelps et al. reported a new method for culturing dissociated islet cells on glass coverslips, where they were able to perform detailed imaging studies by super-resolution and live cell microscopy. More importantly, cells in the dissociated human and rat islet cell monolayers (α, β, δ, and γ) were in proportions similar to native 3D islets [15].

Herein, we report MEA recordings of dissociated islet cells as an innovative method to capture the islet function while circumventing the limitations of the previously used techniques. Moreover, standard functional tests, i.e., glucose stimulated insulin secretion (GSIS) assays, were performed concomitantly. Our data correlated measurable parameters of electrical activities by the MEA to the functional secretory response of islet cells at the early time points of culture. Moreover, we were able to determine that detection of electrical activities of the islet cells in response to the high glucose stimulation were sustained throughout the long-term culture whereas insulin responses from GSIS were only detectable at the early time points. Together, this evidence supports the utility of MEA for measuring islet function in a highly sensitive, non-invasive, and real-time manner.

#### **2. Materials and Methods**

#### *2.1. Islets Dissociation and Culture*

Human pancreatic islets were procured from organ donors at the Human Islet Cell Processing Facility at the Diabetes Research Institute (University of Miami, Miller School of Medicine, Miami, FL, USA), under Institutional Review Board (IRB) approval for use of human tissue for research. Human islets are from approved cadaveric organ donors from which at least one other organ has been approved for transplantation. Since the donors are brain dead, the IRB's from the institutions that isolate the islets consider the tissue as "Exempt" from Human Studies Approval. In this study, islets were obtained from two normal non-diabetic donors, a 51 year old male and a 44 year old female, with body mass indices of 29.5 and 32.8 kg/m2, respectively.

Human collagen IV stock solution (1 mg/mL, Sigma Aldrich, St. Louis, MO, USA) was prepared and diluted to 50 μg/mL into Hanks' Balanced Salt Solution (HBSS) with Ca2+/Mg2<sup>+</sup> (Life Technologies, Carlsbad, CA, USA). Prior to collagen coating, all MEAs were placed inside a UV ozone cleaner (Jelight) for 8 min in order to sterilize and activate the MEA surface for protein coating. A 100 μL drop of the diluted collagen was then added to each of the UV ozoned MEAs. MEAs were then incubated with collagen IV overnight in 37 ◦C and washed 3 times with HBSS with Ca2+/Mg2<sup>+</sup> right before cells seeding.

For dissociation, islets were collected in a conical tube and centrifuged for 2 min at 800 rpm. 400 μL of warmed 0.05% trypsin (Gibco, Waltham, MA, USA) was used for dissociation after three washes with phosphate buffer saline (PBS) were completed. While islets were suspended in trypsin in a cryogenic vial, gentle agitation was applied to the vial in a 37 ◦C beads bath for 3 min to help with dissociation. After trypsinization for 3 min, 15 mL of neuronal medium was added to deactivate the trypsin and centrifuged for 6 min at 1400 rpm. Approximately 200 cells/mm2 were added to each collagen IV coated MEA.

Neuronal culture medium was prepared by supplementing minimum essential medium (MEM, Life Technologies, Carlsbad, CA, USA) with 5% fetal bovine serum (FBS), 2% B-27 (50×, Life Technologies), 1% Penicillin-Streptomycin (100×, Life Technologies), 1% HEPES (1 M, Life Technologies), 1% Glutamax (100×, Life Technologies), 1% Na-pyruvate (100 mM, Life Technologies), and final glucose (Life Technologies) concentration of either 1.1 mM glucose (low glucose media) or 16.7 mM glucose (high glucose media).

#### *2.2. Glucose-Stimulated Insulin Secretion (GSIS) Assay*

Insulin secretion was assessed by GSIS of dissociated islets under static incubation. Briefly, dissociated islets cultured on each MEA were incubated for one hour at 37 ◦C in low glucose media (1.1 mM) followed by a one hour incubation in high glucose media (16.7 mM). After each incubation period, a 500 μL sample of media was collected and insulin concentrations were measured using a human insulin enzyme-linked immunosorbent assay (ELISA) kit (Mercodia, Uppsala, Sweden) after diluting the samples 1:500 in deionized water to ensure measurements were within the range of the ELISA kit.

#### *2.3. Electrophysiological Recordings*

Electrical activity was recorded from the dissociated islets during GSIS assays using MEA2100 system (Multi Channel Systems MCS GmbH, Reutlingen, Germany). Dissociated islets were cultured on a standard microelectrode array chip (60MEA200/30IR–TI–GR, Multi Channel Systems) that fits inside the MEA2100 system, which was connected with a temperature (37 ◦C) controller and an interface board that linked the whole system to a PC computer (Multi Channel Systems). Recordings were done on days 1, 3, 5, and 7 post seeding for 15 min using low glucose media first and then high glucose media for each MEA chip. Multichannel experimenter and analyzer programs were used to do on-line recordings and off-line analysis of data, respectively. Electrical signal from each recording was filtered with a high pass filter (200 Hz) and then a low pass filter (4000 Hz), adapted from a previous study [13], and sampled at 25 kHz. A threshold of 10 times the standard deviation of the noise was set to determine a spike. We submit that we were, in fact, more stringent about the threshold compared to some studies that used 5–6 times the standard deviation of the average noise amplitude [13,16]. For spiking profile plots, a bin size of 10 s was used for the total 30 min of recording, where the total number of spikes in each of these 10 s windows was calculated (30 min recording = 180 windows of 10 s) and used to plot the spiking profile.

#### *2.4. Immunofluorescent Staining*

Dissociated islets from were cultured for 7 days on 18 mm glass coverslips (Electron Microscopy Sciences). On day 7, each coverslip was incubated for 10 min at room temperature with 4% ice cold paraformaldehyde (PFA) solution after a quick rinse with warm PBS. Three washes with PBS were followed, where each wash was for 5 min with gentle shaking. Next, 0.1% Triton X-100 with gentle shaking was applied for 10 min at room temperature to permeabilize the cells. Three washes with PBS were followed, where each wash was for 5 min with gentle shaking. Using PBS 10% and 1% donkey serum were prepared and used as a blocking buffer. Samples were incubated at room temperature in a dark place for 1 hour after a 200 μL drop of 10% blocking buffer was applied to each sample. Primary antibodies (rabbit anti-insulin antibody, rat anti-somatostatin antibody, and goat anti-pancreatic polypeptide antibody, all purchased from Abcam) solution was prepared as 1:200 dilution with 1% donkey serum and 0.25% Triton-X100. After incubation for 1 hour with the blocking buffer, 200 μL drop of the primary antibodies solution was applied to each sample and then placed inside dark 4 ◦C fridge overnight.

On the second day, three washes with PBS and 0.01% Triton-X100 were followed, where each wash was for 5 min with gentle shaking. Secondary antibodies (Alexa Fluor 488 donkey anti-rabbit (Life Technologies), Alexa Fluor 555 donkey anti-rat (Abcam), and Alexa Fluor 594 donkey anti-goat (Abcam) solution was prepared as 1:500 dilution in addition to 4 ,6-diamidino-2-phenylindole (DAPI, 1:200), where all were mixed with 1% donkey serum and 0.25% Triton-X100. 200 μL drop of the secondary antibodies solution was applied to each sample and incubated for 1 hour at room temperature in a dark place. Coverslips were then rinsed with PBS three times and mounted onto glass slides with ProLong Gold Anti-Fade Reagent (Life Technologies) and sealed with nail polish after curing of the mountant. Stained dissociated islets were imaged on a Nikon Eclipse Ti inverted fluorescent microscope with an Andor Zyla sCMOS camera using a 60× oil immersion objective.

#### *2.5. Statistical Analyses*

All statistical analyses were performed on Prism v8 software (GraphPad, San Diego, CA, USA). Paired student *t*-tests were used for statistical comparisons between the low and high glucose conditions for the different days. All values were reported as the mean ± standard error of the mean unless reported otherwise, and *p* < 0.05 was considered statistically significant.

#### **3. Results**

#### *3.1. Dissociation and Culture of Islets*

Extracellular recordings of intact islets using MEA platform (Figure 1A) and classical MEA chips require proper contact and adhesion between cells and electrodes. Given that islets are large multicellular spheroids, limited contact area with the recording electrodes of planar MEA precludes MEA recording for functional evaluation of islets. Thus, we first enzymatically dissociated islets into single cells (Figure 1B,C) and cultured as adherent cells (hours; Figure 1D) on MEA electrodes to improve cell contacts and accuracy of recorded electrical activity. Fluorescence images on day 7 post seeding on the substrate (Figure 1E) showed that this technique dissociated and retained islet cell types, e.g., β, δ, and γ cells. After dissociation, extracellular recordings were performed the following day (day 1) and until the end culture (day 7).

**Figure 1.** Dissociated islets on a microelectrode array (MEA). (**A**) MEA system setup contains five components; an MEA chip (A1), a two head-stages of MEA2100 system (A2), an interface board (A3), a temperature controller (A4), and a computer (A5). (**B**) Schematic illustration of clusters of whole islets in digestion solution (B1) dissociated into separated islets using a 40 μm cell strainer (B2), and finally seeding these dissociated islets on MEA coated with collagen IV (B3). (**C**) Bright field images of whole islet and dissociated islets on MEA. (**D**) Bright field images of MEA before seeding, and after four hours of seeding the dissociated islets. (**E**) Fluorescent images on day 7 for dissociated islets on glass coverslip showing successful separation of different cell population within the islet. Green color represents insulin (indicating β-cells), red color represents somatostatin (indicating δ-cells), magenta color represents pancreatic polypeptide (indicating γ-cells), and blue color represents DNA (indicating the cell nucleus).

#### *3.2. Extracellular Recordings of Dissociated Islets*

Pancreatic β cells show oscillatory electrical activity known as slow waves in response to glucose [5,17]. Extracellular recordings using MEA were previously performed and compared to intracellular measurements obtained by traditional techniques with intact islets to demonstrate an alternate detection method to interrogate the electrical activity of islets [9]. MEA recordings performed using the dissociated islet cells that were seeded on four separate MEA chips (MEA1, MEA2, MEA3, and MEA4). The recordings were for a duration of 30 min on subsequent days, and they showed that high glucose induced longer electrical activity (minutes) with higher amplitudes (mV) compared to the low glucose recordings, which showed shorter electrical activity (seconds) with lower amplitudes (μV; Figure 2). MEA1 on day 3 and MEA2 on day 5, however, showed almost no electrical activity when the high glucose was introduced to the culture. Furthermore, the total number of spikes of the 30-minute recording from each MEA was quantified and the spiking profile plot was generated after binning the number of spikes every 10 s (Figure 3). Interestingly, the time course of binned spike profiles reveal electrophysiological activity that might be associated with the first (immediate) and second (sustained plateau) phase of insulin secretion that is usually seen during dynamic GSIS measurements [18].

**Figure 2.** Electrical activity of dissociated human islets. Representative MEA recordings showing electrical activity of islet cells induced by switching from 1.1 to 16.7 mM glucose for different batches of human islets across 7 days.

**Figure 3.** *Cont.*

**Figure 3.** Total spiking with a bin size of 10 s for a total of 30 min recording. Each MEA was recorded with low glucose (1.1 mM) media for 900 s followed by another 900 s recording with high glucose (16.7 mM) media starting from day 1 post seeding until day 7. The number of spikes were summed every 10 s until the end of the recording.

#### *3.3. Insulin Secretion and FOPP*

β cells release insulin in response to varying blood glucose levels in vivo, thus measuring insulin concentrations upon exposure to varying levels of glucose in vitro is used as a method to correlate islet function. Standard functional testing using GSIS assays were performed simultaneously to the MEA recordings. Measured insulin concentrations under low (2989 ± 260 μg/L) and high (3275 ± 350 μg/L) glucose at day 1 were comparable to low (2683 ± 95 μg/L) and high (3177 ± 225 μg/L) glucose at day 3 (Figure 4A). However, insulin secretion decreased by day 5 (low G: 3132 ± 284 μg/L, and high G: 2930 ± 147 μg/L) and day 7 (low G: 3178 ± 215 μg/L, and high G: 2618 ± 226 μg/L) of culture. Figure S1 shows the insulin concentrations measurements for each MEA. Interestingly, functional testing using MEA showed FOPP measurements with continuous and high spiking activity with high glucose relative to the low glucose for all culture days. The FOPP calculations were low G: 6% ± 1.3% and high G: 46% ± 4.7% for day 1, low G: 33% ± 5.9% and high G: 40% ± 14.9% for day 3, low G: 17% ± 4.0% and high G: 50% ± 18% for day 5, and low G: 12% ± 1.8% and high G: 44% ± 13.8% for day 7 (Figure 4B). Even though the significant difference between the high and low glucose groups was only on day 1, but the other days showed distinguishable difference. Figure S2 shows the FOPP calculations for each MEA separately.

**Figure 4.** Assessment of insulin secretion and fraction of the plateau phase (FOPP). (**A**) Mean insulin concentration under static incubation of 4 released by dissociated islets on MEAs in low (1.1 mM) and high (16.7 mM) glucose media for different culture days (*n* = 4 MEAs for each condition). (**B**) Quantification of the mean FOPP measured by MEAs with dissociated islets in response to low (1.1 mM) and high (16.7 mM) glucose media for the different culture days (*n* = 4 MEAs for each condition). \* *p* < 0.05.

#### **4. Discussion**

Diabetes mellitus is estimated to affect over 400 million people worldwide by 2030 making it one of the most common and costly chronic diseases [3]. Diabetes is characterized by hyperglycemia related to autoimmune destruction of insulin-secreting β cells (type 1) or insulin resistance (type 2) [19,20]. Islet transplantation is a therapeutic alternative for β-cell replacement, which restores glycemic control in type 1 diabetes patients [21,22]. Islet function can be investigated by traditional assay such as GSIS [18], or by utilizing emerging MEA technology [11,23], which provides information about the islet electrophysiology to test the islet function.

Elucidating the complex and dynamic physiologic processes of healthy islets is imperative prior to transplantation. In general, varying blood glucose levels lead to changes in the membrane potential of β cells inducing an electrochemical mechanism resulting in the release of insulin [12,24–26]. More specifically, increased blood glucose concentration fuels glucose metabolism within β cells, and the product of glycolysis is adenosine triphosphate (ATP). The produced ATP reduces the resting membrane potential, which leads to membrane depolarization (electrical activity). After membrane depolarization, the voltage-gated Ca2<sup>+</sup> channels open, increasing intracellular Ca2<sup>+</sup> concentrations, which trigger fusion of vesicles containing insulin with the cell membrane, and subsequent exocytosis

of insulin. Insulin is then released into the blood to allow all cells of the body to utilize glucose for energy [5,12].

MEA recordings using human islets were previously measured and electrical activity was detected with high glucose concentration [23,27]. Schonecker et al. showed no electrical activity corresponded to 1 mM glucose, while oscillatory activity (60–80 μV) was evoked by 10 mM glucose concentration when recording from whole islets [23]. In our study, moderate electrical activity (μV) was seen with 1.1 mM glucose, while high spiking (mV) was observed under 16.7 mM glucose concentration. The difference in amplitude between our study and the Schonecker et al. study may be inferred by several factors. Firstly, the Schonecker et al. study used the whole islet, which means only a small portion of the islet surface was in contact with the electrode, whereas we used dissociated islet that had more cells surface area in contact with an electrode. While our seeding density was 50,000 cells on the MEA, the Schonecker et al. study recorded from one whole islet. Additionally, where the Schonecker et al study used only a low pass filter of 100 Hz to filter their signal, compared to our study that used a high pass 200 Hz filter followed by a low pass 4000 Hz filter, which was adapted from Raoux et al' study. Lastly, Schonecker et al. used mouse islets, while our studies exclusively utilized human islet cells.

Insulin exocytosis has been studied and known to follow a biphasic time course [24,28–31]. The first phase linked to a rapid transient increase rate of insulin secretion, commonly within 5 min of glucose stimulation. Then, insulin secretion decreased to a plateau (second phase) before it completely stopped with the end of glucose stimulus. This data suggests that the spiking initiation of this first and second phases in some of the MEAs from the extracellular recording when we plotted the total 30 min spiking profile after we used a 10 s bin size of and calculated the total number of spikes in each MEA. This is the first study to date that correlates spiking profiles to islet function during glucose stimulated insulin secretion of dissociated islets demonstrating an attractive use of MEA.

The length of the spiking when using MEA has been correlated with the amount of insulin released during glucose stimulation [9,12]. Therefore, FOPP assessment for all MEAs on the different days always showed higher percentages of FOPP during the high glucose condition compared to the low glucose. This evidence supports using MEA as a highly sensitive and robust tool to measure the function of human islets. Insulin secretion measurements using the conventional GSIS assay, on the other hand, showed the same trend between high glucose and low glucose conditions only on the early time points of culture. This could suggest the need for cell–cell contact, which was lost in the dissociated islets, and the cooperating cells may require direct intercommunication to secrete higher levels of insulin under high glucose concentrations. Together, these spiking profiles that correspond to glucose stimulated insulin secretion validates the use of MEA for examining islet function.

This study provided supportive evidence that extracellular recordings using MEA is non-invasive and a quick approach that could be used to test islet functionality. By dissociating the islets, the individual islet cells can be cultured, monitored, and recorded for an extended period, which was not possible before MEA technology. Future improvement to this platform includes seeding at various densities to determine the correlation with spiking activity and physiological insulin levels from islets cells. The next logical step to this study and previous MEA studies should be recording from MEA-based in vitro disease models (diabetic islets) and comparing that with the normal healthy islets (baseline) was done in this study. Diabetic islets once obtained, the use of diazoxide, tolbutamide, and KCL could be investigated to show their effect on the electrical activity. Furthermore, our methods can be used to screen new drugs as well as evaluate some of intervention strategies that could be performed when islets are chronically challenged by glucolipotoxicity or stress-inducing agents.

#### **5. Conclusions**

In conclusion, we demonstrated that extracellular recordings of dissociated islet cells using MEA is an effective approach to rapidly assess islet functionality, and could supplement standard assays such as glucose stimulate insulin response. Evidence from this study demonstrated islet dissociation, and creating layer of islet cells on planar MEA electrodes is critical for the assay. MEA recordings

showed more electrical activity and FOPP percentages induced by the high glucose compared to the low glucose recordings. Furthermore, spiking profile plots from multiple MEA recordings revealed the electrophysiological activity that precedes the initiation of the first (immediate) phase and second (sustained) phase of insulin secretion usually seen during dynamic GSIS measurements. Our approach of dissociating the islets and utilizing the MEA platform to non-invasively test islet functionality is enabling detailed efforts to study islet physiology and screen potential pharmacological interventions

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-666X/11/5/507/s1, Figure S1: Glucose stimulated insulin secretion for functional assessments. Insulin concentration by dissociated islets on each MEA under low (1.1 mM) and high (16.7 mM) glucose media for the different culture days, Figure S2: Assessment of FOPP. Quantification of FOPP of each MEA for low (1.1 mM) and high (16.7 mM) glucose media for the different culture days.

**Author Contributions:** Conceptualization, A.A. (Ahmad Alassaf) and A.A. (Ashutosh Agarwal); Data curation, A.A. (Ahmad Alassaf) and M.I.; Formal analysis, A.A. (Ahmad Alassaf), M.I. and A.B.; Funding acquisition, A.A. (Ashutosh Agarwal); Investigation, A.A. (Ahmad Alassaf), M.I. and A.B.; Methodology, A.A. (Ahmad Alassaf) and A.A. (Ashutosh Agarwal); Project administration, A.A. (Ashutosh Agarwal); Resources, A.B. and A.A. (Ashutosh Agarwal); Supervision, A.A. (Ashutosh Agarwal); Writing—original draft, A.A. (Ahmad Alassaf) and A.A. (Ashutosh Agarwal); Writing—review & editing, A.A. (Ahmad Alassaf), M.I., A.B. and A.A. (Ashutosh Agarwal). All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by NIDDK-supported Human Islet Research Network (HIRN, RRID: SCR\_014393; https://hirnetwork.org; UC4DK104209 to A.A. and UG3DK122638 to A.A.). M.I. is supported by F31DK118860-01A1.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
