Real-time in vivo thoracic spinal glutamate sensing during myocardial ischemia

Abstract

In the spinal cord, glutamate serves as the primary excitatory neurotransmitter. Monitoring spinal glutamate concentrations offers valuable insights into spinal neural processing. Consequently, spinal glutamate concentration has the potential to emerge as a useful biomarker for conditions characterized by increased spinal neural network activity, especially when uptake systems become dysfunctional. In this study, we developed a multichannel custom-made flexible glutamate-sensing probe for the large-animal model that is capable of measuring extracellular glutamate concentrations in real time and in vivo. We assessed the probe’s sensitivity and specificity through in vitro and ex vivo experiments. Remarkably, this developed probe demonstrates nearly instantaneous glutamate detection and allows continuous monitoring of glutamate concentrations. Furthermore, we evaluated the mechanical and sensing performance of the probe in vivo, within the pig spinal cord. Moreover, we applied the glutamate-sensing method using the flexible probe in the context of myocardial ischemia-reperfusion (I/R) injury. During I/R injury, cardiac sensory neurons in the dorsal root ganglion transmit excitatory signals to the spinal cord, resulting in sympathetic activation that potentially leads to fatal arrhythmias. We have successfully shown that our developed glutamate-sensing method can detect this spinal network excitation during myocardial ischemia. This study illustrates a novel technique for measuring spinal glutamate at different spinal cord levels as a surrogate for the spinal neural network activity during cardiac interventions that engage the cardio-spinal neural pathway.

NEW & NOTEWORTHY In this study, we have developed a new flexible sensing probe to perform an in vivo measurement of spinal glutamate signaling in a large animal model. Our initial investigations involved precise testing of this probe in both in vitro and ex vivo environments. We accurately assessed the sensitivity and specificity of our glutamate-sensing probe and demonstrated its performance. We also evaluated the performance of our developed flexible probe during the insertion and compared it with the stiff probe during animal movement. Subsequently, we used this innovative technique to monitor the spinal glutamate signaling during myocardial ischemia and reperfusion that can cause fatal ventricular arrhythmias. We showed that glutamate concentration increases during the myocardial ischemia, persists during the reperfusion, and is associated with sympathoexcitation and increases in myocardial substrate excitability.

INTRODUCTION

Glutamate serves as the major excitatory neurotransmitter that is released in the spinal cord (1, 2). Extracellular glutamate concentrations are very tightly regulated in the healthy nervous system because of glutamate uptake mechanisms and have not been established to fluctuate in correlation to normal neuron firing patterns (3). However, its concentration can be used as a biomarker for the neural network activation level in the pathological state when uptake systems become dysfunctional (4). Some techniques have been tried to measure glutamate concentration in the spinal cord; for example, microdialysis was used to analyze glutamate concentrations in the pig spinal cord after aortic cross clamping (5). However, a minimum amount of sample is needed for offline assay or HPLC analysis, which takes several minutes to be collected, in this case, one sample every 10 min (6). The resulting poor time resolution means that microdialysis does not provide time resolution sufficient to track the glutamate concentration change during myocardial ischemia-reperfusion (I/R) events. Electrochemical glutamate detection with amperometry allows for subsecond time resolution. The current state-of-the-art glutamate electrochemical sensor is a ceramic microelectrode array consisting of a glutamate-sensitive electrode and a sentinel control electrode (7). These electrodes have well-established selectivity and have been used in rodents to study the healthy brain and disease states, including traumatic brain injury (8–10). We advance this technology by fabricating microelectrode arrays on a flexible SU-8 substrate for superior tissue integration (11). We also demonstrate decreased noise from movement artifacts with the flexible substrate, which is particularly important in a large preclinical animal model. Additionally, probes were fabricated with multiple glutamate-sensing channels for simultaneous measurement across different laminae of the spinal cord. This is an important capability of multichannel-sensing electrodes, as different spinal cord laminae and nuclei process different information and have different functions (processing sensory information, motor neurons that innervate skeletal muscle for movement, preganglionic sympathetic neurons that innervate the paravertebral sympathetic chain to increase the sympathetic tone) (12). Furthermore, the therapeutic approach of spinal cord stimulation during myocardial I/R injury has differential effects on spinal neurons in superficial lamina versus deeper lamina (13). To study the impact of cardiac I/R on the lamina-specific neural networks in the spinal cord or for potential optimization of neuromodulation therapies, it is undeniably essential to investigate the spinal neurotransmitter activity with multichannel-sensing probes.

Myocardial ischemia-induced ventricular arrhythmias are the leading cause of sudden cardiac death (14). The cardiac autonomic nervous system (CANS) plays a key role in regulating cardiac function after myocardial ischemia (15). The autonomic balance between the sympathetic tone and the parasympathetic tone when the heart is not stressed is the critical component of cardiac function regulation. Autonomic imbalance is one of the main causes for the initiation and maintenance of fatal arrhythmias (16–21). When the heart is stressed, for instance during myocardial I/R injury, cardiac sensory neurons in the dorsal root ganglion transmit excitatory signals to the spinal cord. These excitatory signals are processed in the dorsal horn of the spinal cord and are translated to the activation of the sympathetic preganglionic neurons in the intermediolateral nucleus (IML), which results in sympathoexcitation (22, 23). This sympathoexcitation causes an endogenous release of catecholamines at the heart level during phase 1b (∼15–30 min of myocardial ischemia), which is associated with high arrhythmia risk (24, 25). Myocardial ischemia has a different mechanism of action during different phases of ischemia, which results in different sensory inputs to the spinal cord and hence different spinal cord network processing. It is therefore important to understand how the spinal cord is processing the cardiac sensory information at different time points during myocardial ischemia.

In this study, we developed a multichannel custom-made flexible glutamate-sensing probe for use in large animals, and we evaluated the spinal neural network activity in real time during I/R injury with this newly developed probe. The glutamate detection of the developed probe is almost instantaneous and can measure the glutamate concentration continuously. Using this developed continuous spinal glutamate measurement, we have shown that the glutamate concentration increases during the I/R injury, which could be one of the main causes of sympathoexcitation and increases the risk of ventricular arrhythmias. A large-animal porcine model was used in this study because of its high translatability to humans in terms of the cardiovascular system and the CANS.

METHODS

Study Approval

The study protocol was approved by the University of Pittsburgh Institutional Animal Care and Use Committee (IACUC). All experiments were performed in compliance with the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals.

Glutamate Sensor Probe Development

Flexible microelectrode array fabrication.

A 4-in. Si wafer with a 100-µm-thick SiO₂ layer (University Wafer, Boston, MA) was first cleaned with acetone, isopropanol, and deionized (DI) water sequentially. The wafer was then dried with a N₂ spray gun, heated on a hot plate at 150°C for 5 min, and treated with O₂ plasma with a reactive ion etcher (RIE; Trion Phantom III LT) for 2 min at 300-mTorr pressure and 150-W power. The cleaned wafer was spin coated with SU-8 2015 (MicroChemicals, Ulm, Germany) at 5,000 rpm for 1 min and soft baked at 65°C for 3 min and at 95°C for 5 min. Then, the wafer was exposed with a direct-writing maskless aligner (MLA; MLA100, Heidelberg Instruments) with a dose of 350 mJ/cm² to define the bottom insulation layer. After exposure, the wafer was first postbaked at 65°C for 3 min and at 95°C for 3 min, then developed with an SU-8 developer (MicroChemicals) for 1 min, and cleaned with isopropanol and DI water. The patterned SU-8 was subsequently hard baked at 200°C, 180°C, and 150°C for 5 min each and allowed to cool down below 95°C.

After cleaning, the wafer was spin coated with an AZ P4620 photoresist (MicroChemicals) at 5,300 rpm for 1 min and baked at 105°C for 5 min, as previously described (1). After soft baking, the wafer was exposed with the MLA with a dose of 700 mJ/cm², then developed with an AZ400k 1:4 developer (MicroChemicals), cleaned with water, rinsed, and dried by N₂ gas flow. A 10-nm Ti adhesion layer, a 100-nm Au layer, and a 40-nm Pt layer were evaporated on the wafer with an electron-beam evaporator (Plassys MEB550S), and then the metal was lifted off in acetone to define the metal electrodes and interconnections. A top insulation layer of SU-8 2015 was then spin coated at 5,000 rpm for 1 min, soft baked at 65°C for 3 min and at 95°C for 5 min, and photolithography patterned with the MLA with a dose of 350 mJ/cm² to expose the connection pads and to define the top insulation layer. After postbaking and a development procedure with the SU-8 developer, the wafer was cleaned with isopropanol and DI water, hard baked at 200°C, 180°C, and 150°C for 5 min each, and allowed to cool down below 95°C. The microelectrode arrays (MEAs) were lifted off from the wafer with a buffered oxide etchant (1:7) in an acid hood for ∼4 h. Figure 1A shows a schematic of the flexible MEA fabrication. The final MEA design is schematically represented in Fig. 1B.

Figure 1.

Electrode design. A: graphic of the photolithography process. B: schematic of the probe design. C: schematic of how the glutamate sensors work.

Glutamate sensor preparation.

Fabricated electrodes were first characterized by electrochemical impedance spectroscopy (EIS) and cyclic voltammetry (CV) with an Autolab potentiostat (Metrohm, Herisau, Switzerland) with Nova 2.1.4 software to determine baseline functionality. EIS was measured from 10⁵ to 1 Hz, with 10 frequencies recorded per decade with a sine wave oscillation and a 0.01 V root mean square (RMS) voltage. CVs were scanned between −0.6 and 0.8 V versus an Ag/AgCl reference at 0.1 V/s.

Glutamate electrodes were prepared as described previously (7). To prepare the sensors, the Pt electrode sites were first cleaned by immersing them in 0.05 M H₂SO₄ (Fisher Scientific, Pittsburgh, PA) and scanning from −0.35 to 1.5 V at 1 V/s for 100 cycles. A screening layer of m-phenylenediamine (mPD; Sigma-Aldrich) was electrodeposited onto the electrode sites by immersing the electrodes in a 10 mM solution of mPD in 1× phosphate-buffered saline (PBS) and applying 0.7 V for 20 min. Then, 1 µL of an enzyme solution containing 1 U of glutamate oxidase (GluOx) (Cosmo Bio USA, Carlsbad, CA), 13.7 mg of bovine serum albumin (Fisher Scientific), and 6.7 µL of glutaraldehyde (Sigma-Aldrich) in 1 mL of DI water was drop casted onto three of the six electrode sites. These sites became the glutamate-sensitive sensor sites. Figure 1C shows a schematic of the electrode layers. A control solution lacking glutamate oxidase was dropped onto the remaining three electrode sites; these sites became the sentinel control sites.

Glutamate was detected amperometrically at 0.7 V versus Ag/AgCl as the difference between the glutamate sensor sites and the sentinel control sites with a CH Instruments 1430 multichannel potentiostat (CH Instruments, Austin, TX). To evaluate sensor functionality and sensitivity, electrodes were calibrated before and after use in standard concentrations of glutamate. In vivo, three glutamate-sensitive sites and one sentinel site were monitored simultaneously. A calibration curve was constructed from the postcalibration standards and used to convert detected in vivo current into concentration. To prepare reference electrodes for in vivo use, a silver wire was coated with silver chloride by immersing it in a 3 M solution of KCl and applying 4 V for 3 min.

In Vivo Myocardial Ischemia Induction

Animal preparation.

Four male and four female Yorkshire pigs (44 ± 2 kg) were used to evaluate the glutamate-sensing probe performance and investigate the glutamate release probe profile during I/R injury.

A combination of tiletamine-zolazepam (4 mg/kg im) and xylazine (2 mg/kg im) was used to sedate the animals. The animals were given inhaled isoflurane 3–5% via a nose cone. When the animal was completely sedated and anesthetized, the animal was intubated and mechanically ventilated. After intubation, the anesthesia was reduced to 2–4% of inhaled isoflurane. General anesthesia was maintained with the isoflurane until the end of the preparation and then was switched to α-chloralose (50 mg/kg initial bolus followed by 20 mg/kg/h continuous infusion). α-Chloralose limits the impact of anesthesia on the cardiac autonomic nervous system and cardiac myocardial excitability (22, 23, 26, 27). All animals were completely anesthetized during the whole procedure, and the anesthesia level was adjusted throughout the experiment by monitoring corneal reflex, jaw tone, and hemodynamic indexes. Animal temperature was maintained with water heating pads (T/PUMP; Gaymar Industries, Orchard Park, NY).

Heart rate, blood pressure, end-tidal carbon dioxide, and blood oxygen level were monitored continuously during the procedure. Arterial blood gas was tested hourly or more frequently to prevent acid-base disorders, and ventilation rate/volume adjustment or administration of sodium bicarbonate was performed as needed.

Electrocardiogram, systemic blood pressure, and left ventricular pressure were recorded with the Prucka CardioLab recording system (GE Healthcare, United States) and the Millar system (ADInstruments, United States).

Carotid and femoral arteries, as well as external jugular and femoral veins, were catheterized for blood pressure monitoring and drug administration, respectively.

To expose the spinal cord and insert the glutamate-sensing probe in the dorsal horn, laminectomy was performed in these animals in the prone position (Fig. 2A). Upon completion of the laminectomy, the animals were switched to the supine position to perform the sternotomy procedure to expose the heart (Fig. 2B). After sternotomy, the sternum was cauterized to avoid bleeding. The pericardial sac was cut and secured to the sternum with sutures. Once the heart was exposed, a Prolene suture was placed around the left anterior descending (LAD) coronary artery after its second diagonal branch to induce myocardial ischemia (Fig. 2B). The myocardial ischemia induction was confirmed by observing the ST-elevated leads (Fig. 2C). Nylon mesh epicardial sock electrodes were placed around the heart (Fig. 2D) to measure activation recovery interval (ARI) and dispersion of the repolarization (DOR), and a Millar catheter (Millar, United States) was placed in the left ventricle through the carotid artery.

Figure 2.

Simultaneous in vivo spinal glutamate and cardiac electrophysiological recordings. A: after T1–T4 laminectomy, the glutamate-sensing probe was inserted in the dorsal horn region of the spinal cord at the T2–T3 level. B: myocardial ischemia was induced by ligating the suture that was placed around the left anterior descending (LAD) coronary artery. Myocardial ischemia was confirmed by observing the ST-elevated electrograms (EGMs) in the ischemic region. LV/RV, left/right ventricle. C: number of ST-elevated leads increased during myocardial ischemia (n = 6 animals before LAD vs. 6 animals during LAD; *P = 0.031, Wilcoxon test). D: after sternotomy, the 56-channel sock electrode was placed around the heart to record the electrocardiogram. The activation recovery interval (ARI) was measured by subtracting the activation time (AT) from the recovery time (RT). E: representative activation recovery interval shortening during myocardial ischemia from 1 pig. A and B were created with a licensed version of BioRender.com.

The animals were placed in the lateral position to have access to both the heart and spinal cord to perform the myocardial ischemia intervention and the glutamate-sensing probe insertion and recording.

At the end of the data collection protocol, animals were euthanized by inducing ventricular fibrillation via injection of potassium chloride under deep anesthesia.

Experimental protocols.

Glutamate was detected in vivo with a three-electrode setup. The reference electrode was a silver wire coated with silver chloride inserted into the muscle of the pig near the working electrode insertion site (28). A metal retractor was used as a counter electrode. After a 30-min stabilizing period after the glutamate probe insertion, myocardial ischemia was performed for 1 h by ligating the suture that was placed around the LAD coronary artery. Two animals experienced ventricular arrhythmias during the 1-h myocardial ischemia, and the protocol could not be completed in these two animals. The suture was released after the 1-h myocardial ischemia. Heart rate, systemic blood pressure, left ventricular systolic pressure, left ventricular contractility, ARI, DOR, and glutamate concentration were recorded during the baseline, myocardial ischemia, and reperfusion.

Hemodynamic and Cardiac Electrophysiology Measurements

Hemodynamics assessment.

A lead II ECG was used to assess the heart rate during the protocol. Blood pressure was measured with the pressure transducer that was connected to the femoral arterial sheath. Left ventricular pressure was measured via a 5-Fr SPR-350 Millar Mikro-Tip pressure transducer catheter (Millar Instruments, Houston, TX) inserted into the left ventricle and connected to an MPVS Ultra Pressure-Volume Loop System (Millar Instruments). Left ventricular end-systolic pressure as well as left ventricular contractility (dP/dt_max) were calculated with the left ventricular pressure signal. The hemodynamics parameters were analyzed with the Spike2 program (Cambridge Electronic Design) and LabChart software (ADInstruments, Colorado Springs, CO).

Activation recovery interval analysis.

A 56-electrode nylon mesh sock electrode was made with 0.2-mm silver wires, and it was placed around the heart to measure the epicardial unipolar electrograms with a Prucka CardioLab electrophysiology mapping system (GE Healthcare, Fairfield, CT) (Fig. 2D). Unipolar electrograms were filtered (0.05–500 Hz) with the GE CardioLab System (29). Activation and repolarization time were detected with customized software (iScalDyn; University of Utah, Salt Lake City, UT), and the ARI, which is the difference between the repolarization and activation times, was calculated (30). The ARI is a surrogate of local action potential duration (31, 32). Excessive sympathetic outflow shortens the ARIs (Fig. 2E) (33).

Dispersion of repolarization.

DOR is defined as the variance of ARI among all 56 leads. Greater dispersion is associated with higher arrhythmia risk and therefore higher mortality.

Arrhythmia count.

Because the animals with ventricular tachycardia and fibrillation during myocardial ischemia were excluded from this study, for arrhythmia count we only counted the premature ventricular contractions (PVCs).

Statistical Analysis

For the normal paired data, paired t test was performed. When the data did not pass the Shapiro–Wilk normality test, the Wilcoxon test was performed. All figures were created with GraphPad Prism software (version 8, GraphPad Software, San Diego, CA) or MATLAB (2019 version, MathWorks). A P value < 0.05 was considered statistically significant. Data are presented as median (25th, 75th percentiles) or means ± SE. All electrode sites that did not detect baseline glutamate or were found to be no longer sensitive to glutamate in vitro after removal from the spinal cord were excluded from the analysis.

RESULTS

Flexible Microelectrode Array

The flexible MEAs were successfully fabricated on the SU-8 flexible substrate, following a direct-writing maskless photolithography process. SU-8 has demonstrated excellent chronic performance as a neural probe substrate (34–38). The device is composed of a singular shank (140 µm wide and ∼16 µm thick) with 6 × 100-µm-diameter platinum electrode sites, organized into two groups of three electrodes each. The total length of the shank is 6.5 mm to easily target the spinal cord dorsal horn of the pig. The flexible MEAs are connected to the printed circuit board PCB with a zero-insertion force (ZIF) connector, to be interfaced with characterization and recording systems. The PCB is sealed with epoxy to protect the connection from fluids and blood exposure. The final MEA is highly flexible and can be bent significantly without breaking (Fig. 3A).

Figure 3.

In vitro glutamate sensor performance. A: photo demonstrating the flexibility of the probe. B: electrochemical impedance spectroscopy (EIS) and phase angle of all 6 sites on 3 probes with high reproducibility within electrodes and between electrodes. C: cyclic voltammetry (CV) of an electrode. D: sensors in a stirred solution of PBS. Electrodes do not respond to additions of ascorbic acid (AA). Glutamate sites respond to glutamate but not to the addition of AA, whereas sentinel electrode does not respond to either glutamate or AA. All sites respond to hydrogen peroxide. E: calibration curve shows a wide range of activity for the glutamate sites and no activity for the sentinel.

In Vitro and ex Vivo Testing of the Glutamate-Sensing Probe

Electrochemical impedance spectroscopy (EIS) of all six electrode sites on three arrays (for a total of 18 sites) show excellent reproducibility both between sites on a single MEA as well as across multiple MEAs (Fig. 3B). The average impedance at 1 kHz of these sites was 36 ± 3 kΩ (mean ± SE). Similarly, cyclic voltammetry measurements between −0.6 and 0.8 V versus Ag/AgCl, measured at 0.1 V/s, have an average charge storage capacity of 3.5 ± 0.3 µC (Fig. 3C).

Sensitivity of the glutamate electrodes.

The MEAs were modified to be glutamate-sensitive electrodes. The final MEA contains electrode sites that either are sensitive to glutamate (the GluOx sites) or are controls (the sentinel sites). In a GluOx site, GluOx bound to the surface of the electrode reacts with glutamate to produce H₂O₂, which is oxidized at the Pt electrode surface to produce a current. The sentinel sites are not sensitive to glutamate but will react to ambient H₂O₂ in the tissue, along with any other potential interferences. The final glutamate-derived signal is calculated as the difference between the GluOx sites and the sentinel site. The two types of sites are constructed identically, except that the sentinel sites do not have GluOx cross-linked on their surface. Full details of sensor construction are in methods; to briefly summarize the process, a size-exclusion screening layer of m-phenylenediamine (mPD) was first electrochemically deposited onto the surface. Subsequently, either a glutamate oxidase (GluOx)-containing layer or a control sentinel layer was carefully drop casted under a microscope onto each electrode site and allowed to cross-link overnight.

The final electrodes were highly selective to their desired target. Both the GluOx and sentinel sites have no response to the main interfering compound, ascorbic acid. However, the GluOx sites (red, green, and blue traces in Fig. 3D) respond robustly to repeated addition of glutamate to the solution, and, as expected, the sentinel site (purple trace) does not respond. However, as expected, every site responds to H₂O₂. When calibrated, the GluOx sites show sensitivity to glutamate across a wide range of concentrations between 10 µM and 1 mM (blue, green, and red traces in Fig. 3E), whereas the sentinel site shows no response (purple trace). All final in vivo concentrations presented here were calculated with a postcalibration curve constructed by exposing the electrodes to standard concentrations of glutamate after the electrodes were removed from the spinal cord.

To validate the sensing capabilities in tissue, we inserted the electrodes into a dissected piece of pig spinal cord (Fig. 4A). A needle was inserted near the electrode, and a series of injections of vehicle phosphate-buffered saline (PBS) or glutamate were performed. As expected, the electrodes did not respond to the vehicle; however, there was a robust response to the injection of glutamate (Fig. 4B).

Figure 4.

Performance of flexible glutamate probe in ex vivo and in vivo setting. A: photo of electrode inserted into the pig’s spinal cord. B: ex vivo injections of PBS and glutamate into a spinal cord. Three individual glutamate electrodes (red, green, and blue) do not sense the PBS, but the injected glutamate is sensed by the glutamate sensor probe. C, left: photo of the stiff NeuroNexus probe. Right: photo of our developed flexible probe. D: data were recorded during motion artifacts with a stiff acute NeuroNexus probe and our flex probe, both inserted and recorded at the same time. E, left: polyethylene glycol (PEG)-assisted probe. Right: tip of an electrode assembled with sharpened tungsten. F: sharpened shuttle method results in significantly more detected glutamate in vivo because of the PEG covering the electrodes’ surface.

Performance of the Flexible Probe during Insertion and Movement

Because ambient movement due to breathing, heartbeat, etc., is increased in a large-animal model, we fabricated the MEA on a highly flexible SU-8 substrate. Flexible probes have been demonstrated to reduce inflammatory tissue responses caused by micromotion of the brain following implantation in rodents (11). However, in a large-animal model, continuous passive movements caused by a biological process such as breathing are much more dramatic. The spinal cord itself also presents an additional challenge: touching the spinal cord can inadvertently stimulate neuron firing resulting in muscle spasms. These spasms are common during electrode insertion and may damage the tissue and break silicon-based probes that are brittle and fragile, like those we have used previously to study the spinal cord (22, 23). Our previous experience with stiff probes led us to believe that a flexible probe would be more likely to better accommodate the animal’s movement.

To investigate this, we inserted both a flexible and a stiff probe into the spinal cord (Fig. 4C). The stiff probe was a commercially available probe purchased from NeuroNexus with 30-µm platinum electrode sites, whereas the flex probe was a specially fabricated MEA that also had 30-µm platinum electrode sites (compared with the 100-µm sites used in the rest of this work) for an accurate comparison to the stiff probe. Sites from both electrodes were connected to a multichannel potentiostat for simultaneous amperometric measurements at 0.7 V versus Ag/AgCl reference. After a 30-min stabilization time, the spinal cord was lightly brushed with a cotton swab to induce a mild muscle spasm. This movement induced a sudden and dramatic change in the baseline of the stiff probe (red trace in Fig. 4D) but not in the baseline of the flex probe (blue trace). This test was repeated multiple times with a 5-min restabilization period between trials (Supplemental Fig. S1A; all Supplemental Material is available at https://doi.org/10.6084/m9.figshare.22932026). The response of the stiff probe decreased with each subsequent spasm, presumably because of damage and loss of tissue cohesion around the stiff probe.

Optimization of the spinal cord insertion procedure.

Although flexible probes can move with the tissue during micromotions and improve device-tissue integration (11, 39–41), insertion of these flexible devices into the tissue becomes more difficult (42–44). Flexible materials often are too soft to penetrate tissue on their own, so shuttle systems are common to aid insertion (11, 38, 45, 46). An anchor hole was patterned at the shank tip to facilitate the insertion of a sharpened 50-µm tungsten shuttle, to enable the handling and penetration of the flexible device into the spinal cord (Fig. 1B and Fig. 4E). To that end, we compared two insertion methods from the literature to determine the optimal procedure for neurochemical sensing in the spinal cord. In one scheme, a 50-µm-diameter tungsten guide wire was threaded through a 100-µm insertion guide hole at the tip of the SU-8 shank. The wire was then glued to the MEA with 30% polyethylene glycol (PEG) (Fig. 4E). In the second procedure, the hole in the SU-8 shank was 40 µm wide, and a 50-µm SU-8 wire was sharpened by applying 5 V versus Ag/AgCl and repeatedly dipping it into a solution of 5 M potassium hydroxide to etch the surface. The final needlelike wire was threaded through the insertion hole without the PEG glue (Fig. 4E).

When the PEG-assisted and tungsten shuttle insertion methods are compared, the PEG-assisted insertion technique (red in Fig. 4F; n = 8 electrode sites) has a signal several orders of magnitude lower than the sharpened-shuttle method (blue; n = 4 electrode sites). Supplemental Fig. S1B shows the means ± SE of all functional, glutamate-detecting electrodes (n = 12 GluOx sites in n = 6 pigs).

Spinal Cord Hyperactivity during Myocardial Ischemia

We tested the sensing capabilities of our probe in a sham condition and observed a slight decrease in the detected current over 2 h of recordings (Supplemental Fig. S2A). The sham recording was performed after a stabilization period of at least 30 min. We attribute this slow, linear decrease in current to the decrease in sensitivity of our electrode over time, which could be the result of enzyme degradation or biofouling. We have corrected this by fitting a line to the first 30 min of the baseline data and then subtracting the slope of the fitted line to correct the data for the slight decrease in sensitivity over time in vivo (Supplemental Fig. S2, A and B). Supplemental Fig. S2C shows the glutamate concentration changes during the 30-min baseline after the correction of the loss of glutamate sensitivity.

Glutamate signaling was measured as a biomarker for spinal cord neural activity. We measured the glutamate concentration during the 30 min of baseline recording (in 5-min intervals), and we did not see a difference in glutamate concentration during these 30 min (n = 12 electrodes, P = 0.61, Friedman test) (Supplemental Fig. S2C). Also, at different sites we detected different concentrations of glutamate, possibly indicating local spatially resolved concentration differences, and at some locations we detected no glutamate at all despite the postcalibration confirming that the sensor is functional (Supplemental Fig. S2, D and E).

Glutamate percent change from baseline was assessed during the baseline, every 15 min of ischemia, and the 15 min of reperfusion. The percent change in glutamate was not significant during the first 15 min of the LAD ischemia [baseline (BL) −0.03 [−1.85, 1.90]% to LAD15 11.04 [4.75, 17.94]%; P = 0.63] Glutamate concentration started to increase significantly between 15 and 30 min of ischemia (BL −0.03 [−1.85, 1.90]% to LAD30 24.40 [12.81, 259.77]%; P = 0.03) (Fig. 5, A and C). Glutamate augmentation persisted during the rest of the myocardial ischemia during 30–45 min (BL −0.03 [−1.85, 1.90]% to LAD45 53.788 [19.46, 285.46]%; P = 0.0002) and 45–60 min (BL −0.03 [−1.85, 1.90]% to LAD60 69.64 [29.75, 205.63]%; P < 0.0001) of LAD. Glutamate concentration did not decrease during the 15 min of the reperfusion and was still elevated compared with the baseline (BL −0.03 [−1.85, 1.90]% to reperfusion 69.35 [36.87, 159.02]%, P < 0.0001) (Fig. 5, A and C, and Supplemental Fig. S3A).

Figure 5.

Simultaneous in vivo spinal glutamate and cardiac electrophysiological recordings. A: myocardial ischemia-reperfusion injury caused an augmentation in the percent change of glutamate from baseline (Friedman test, ****P < 0.0001). Glutamate did not change during the first 15 min of the left anterior descending (LAD) coronary artery occlusion (n = 12 electrodes pre-LAD vs. 12 electrodes during LAD15, P = 0.63). LAD between 15 and 30 min (n = 12 electrodes pre-LAD vs. 12 electrodes during LAD30, *P = 0.032), LAD between 30 and 45 min (n = 12 electrodes pre-LAD vs. 12 electrodes during LAD45, ***P = 0.0002), LAD between 45 and 60 min (n = 12 electrodes pre-LAD vs. 12 electrodes during LAD60, ****P <0.0001), and reperfusion (Rep) (n = 12 electrodes pre-LAD vs. 12 electrodes during reperfusion, ****P <0.0001) caused an increase in the percent change of glutamate from baseline. For each comparison, Dunn’s multiple comparisons test was used. B: representative glutamate (Glu), activation recovery interval (ARI), and dispersion of repolarization (DOR) response during LAD ischemia. C: glutamate and DOR increased during the LAD ischemia and reperfusion, whereas the ARIs decreased during the ischemia-reperfusion injury.

Shortening of the Activation Recovery Interval during I/R Injury

ARI is a surrogate for the action potential duration, and sympathoexcitation causes the ARI shortening, which could be mitigated by the spinal neuromodulation that decreases the central sympathetic drive (13, 23, 47, 48). In this study, the ARIs were shortened (285.06 ± 15.75 to 256.08 ± 13.88 ms; P = 0.01) during myocardial ischemia (Supplemental Fig. S3B). The ARI shortening occurred within the first 5 min of the ischemia and persisted until the termination of the myocardial ischemia (Fig. 5, B and C). The sympathetic activation was not mitigated during the reperfusion, as the ARI was still shorter during the reperfusion compared with the baseline (285.06 ± 15.75 to 260.29 ± 19.01 ms; P = 0.04) (Supplemental Fig. S3B).

Increased Dispersion of Repolarization during Myocardial Ischemia

DOR is a potential marker for arrhythmogenicity. Increased DOR indicates an increased arrhythmia potential. The DOR started to increase within the first 5 min of the myocardial ischemia (Fig. 5, B and C) and stayed elevated by the end of myocardial ischemia (296.42 ± 42.18 to 2,051.48 ± 337.91 ms²; P = 0.01) (Supplemental Fig. S3C). The DOR was not decreased during the 15-min reperfusion (296.42 ± 42.18 to 2,039.72 ± 435.00 ms²; P = 0.02), and therefore the risk of arrhythmia events was still high during the reperfusion (Supplemental Fig. S3C).

Arrhythmias during Myocardial Ischemia

Premature ventricular contractions were counted during myocardial ischemia. There was an increase in PVC counts during LAD ischemia (baseline 0 ± 0, during LAD ischemia 74 ± 41; P = 0.03) (Supplemental Fig. S3D).

Hemodynamic Changes during Myocardial Ischemia

Heart rate and left ventricular end-systolic pressure were assessed during the myocardial ischemia and 15-min reperfusion. Heart rate increased during the myocardial ischemia (104.76 ± 4.49 to 110.31 ± 4.99 beats/min; P = 0.03) and reperfusion (104.76 ± 4.49 to 118.88 ± 5.67 beats/min; P = 0.04) (Supplemental Fig. S3E). Left ventricular end-systolic pressure decreased during the ischemia (109.23 ± 8.84 to 96.61 ± 7.08 mmHg; P = 0.01). There was not a significant change in left ventricular end-systolic pressure from baseline to reperfusion (109.23 ± 8.84 to 98.15 ± 7.08 mmHg; P = 0.30) (Supplemental Fig. S3F).

DISCUSSION

In this study, we developed a new flexible sensing probe to measure spinal glutamate signaling in vivo in a large-animal model. Our initial investigations involved precise testing of this probe in both in vitro and ex vivo environments. We accurately assessed the sensitivity and specificity of our glutamate-sensing probe and demonstrated its performance. We also evaluated the performance of our developed flexible probe during insertion and compared it with the stiff probe during animal movement. Subsequently, we used this innovative technique to monitor the spinal glutamate signaling during myocardial ischemia and reperfusion, which causes ventricular arrhythmias that are a leading cause of death. We showed that glutamate concentration increases during the myocardial ischemia, persists during the reperfusion, and is associated with sympathoexcitation and increases in myocardial substrate excitability.

Flexible Glutamate-Sensing Probe

A flexible probe was developed in this study to minimize probe insertion trauma to the spinal cord during the glutamate probe insertion. Several studies have evaluated the effect of probe stiffness on neuronal tissue injury, and they support the use of softer and more flexible probes (11, 39). Further studies confirmed the advantage of using flexible probes in terms of their capabilities to move with the neural tissue during micromotion that might be caused by cardiac pulsation or respiration (11, 39, 49, 50). In the present study, we have also shown that the flexible probe is resistant to movement artifacts during larger movements caused by other stimuli. This demonstrates the value of a flexible probe in large-animal models, where motion caused by breathing, heartbeat, etc. is more intense than in a rodent model. Large tissue displacements around a stiff probe can cause damage to both the tissue and the electrode itself; a flexible probe can minimize these effects (51). Another study investigated the effect of probe stiffness in modulating the glial cells, and their results indicate that glial cells respond greater to mechanical signals that are caused by stiff probes (52). The flexible probe design provides a great opportunity for more reliable and robust recordings without causing significant trauma in the neural tissue. Although isolated use of flexible glutamate sensors has been reported as a proof of concept in rodents by a few investigators, we provide a comprehensive testing and in vivo study of the sensors in a large preclinical disease model (53–57). To our knowledge, this is the first report of a flexible glutamate sensor used in a large-animal model. An experiment to analyze the effect of physical motion on the electrode determined that the dramatic motion artifacts created by movement in a stiff probe were almost completely absent in a flexible probe (Fig. 4D).

One of the major drawbacks of flexible sensors in vivo is the difficulty of inserting a probe that may buckle and fail to penetrate the neural tissue. Stiff shuttles or temporary stiffening approaches are often employed to aid in probe insertion into tissue and prevent buckling (11, 37, 38, 40, 58). Here, we compared two common techniques from the literature: a sharpened shuttle “needle and thread” approach and PEG-assisted insertion, where PEG is used as a dissolvable glue to attach the flexible probe to the stiff shuttle. We found that the sharpened shuttle approach resulted in a much more robust detection of glutamate. We hypothesize this to be the result of two phenomena. First, adhering the flexible probe to the shuttle required a very thick layer of PEG that may not be effectively washed away for hours in vivo. The glutamate sensors require glutamate to be able to diffuse to and come into contact with the coating surface to react. Although PEG is very soluble in an aqueous environment, residual PEG may still cover part of the sensing surface, which will decrease the effectiveness of the sensor. Second, the PEG-assisted insertion was much more difficult. In our hands, inserting the electrode with the PEG-assisted method frequently required several attempts. This may have resulted in both damage to the tissue as well as damage to the enzyme coatings on the sensor. The sharpened shuttle approach, therefore, resulted in more robust signals because of the ease of insertion and a more pristine electrode surface.

Ensuring our electrodes’ selectivity for glutamate over other potentially interfering electroactive compounds is critical for in vivo sensing applications. We have incorporated two methods from the literature into the sensor itself to ensure it is selective to glutamate alone (7):

• 1) Electrode sites were coated with a layer of m-phenylenediamine. This is a size-exclusion layer that will only allow H₂O₂ (our electroactive indicator for glutamate) to cross and exclude larger electroactive compounds, such as ascorbic acid.
• 2) The electrode array is self-referencing; glutamate-sensitive electrode sites are referenced relative to “sentinel” control sites on the same array. The sentinel sites do not detect glutamate, as they lack glutamate oxidase in the coating, but their construction is otherwise identical, to detect any potential interfering compounds. The final glutamate-derived signal reported in all figures was the difference between the glutamate-sensitive electrodes and the sentinel electrode, to obtain what we believe to be a “pure” glutamate signal. Small modifications to this technique can be used to construct sensors for other compounds of interest, including glucose, lactate, pyruvate, and acetylcholine (59–62).

Our multielectrode array sensor was fabricated with a total of six channels: three glutamate-sensing sites and three redundant sentinel sites. By performing measurements with a four-channel multipotentiostat, we monitored the three glutamate-sensitive channels and one sentinel channel simultaneously. We attempted to confirm the final position of the electrode postmortem to correlate glutamate concentrations with specific spinal cord structures. However, we encountered some practical barriers that precluded us from performing this analysis for this study. We are currently optimizing methods to reliably determine the electrodes’ positions relative to the spinal cord anatomy for future work. Future experiments to correlate the position of each electrode to the detected glutamate concentration will provide valuable insight into lamina-specific glutamatergic activity in the spinal cord.

Sympathoexcitation and Myocardial Ischemia-Induced Ventricular Arrhythmias

A large-animal porcine myocardial ischemia-reperfusion model was used in this study. This model is well established and studied by our group and others, and it closely mimics the myocardial ischemia-reperfusion condition in humans. The myocardial ischemia was created and confirmed by at least nine (out of 56) ST segment-elevated leads (Fig. 2C). The ischemia-reperfusion injury causes various cellular and ionic changes, which occur in response to reduced oxygen and nutrient supply, as well as the sympathoexcitation, which could be manifested by the shortening of the ARIs (23, 63). I/R also increased the DOR, which is associated with higher myocardial electrical heterogeneity and increased risk of arrhythmias (26, 29, 33, 63). In this study, two pigs (out of 8) experienced ventricular tachyarrhythmias (VTs) during myocardial ischemia (at 5 min and 25 min during ischemia) and needed to be defibrillated and were excluded from the study. The occurrence of VT in these two pigs provides clear evidence of the risk of fatal ventricular arrhythmias during our large-animal myocardial ischemia model. Furthermore, we have shown that the number of PVC counts significantly increased during myocardial ischemia.

Spinal Hyperexcitation during Myocardial Ischemia-Reperfusion Injury

Mechanosensitive, chemosensitive, and multimodal afferent neurons transduce myocardial ischemia. These excitatory sensory signals are transmitted to the dorsal root ganglion and vagal nodose ganglia (64, 65). The dorsal root ganglion neurons transmit this afferent excitatory signal to the spinal dorsal horn neurons, which will be processed in the spinal dorsal horn neurons and will activate the sympathetic preganglionic neurons in the IML region. We have previously shown this spinal hyperexcitation during ischemia by performing direct in vivo spinal neural recording using a multichannel microelectrode array (22, 23, 47). Glutamate is a critical neurotransmitter for these spinal excitatory synaptic transmission (1, 2). There are also multiple pieces of evidence of extrasynaptic glutamate N-methyl-d-aspartate receptors (NMDARs) (66–68), α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs) (69), and metabotropic glutamate receptors (mGluRs) (70) in the spinal cord. Extrasynaptic glutamate signaling is initiated from spillover of synaptic release as well as from astrocytes and microglia (71). The NMDARs have a key role in excitotoxicity (72), and there are key differences between the synaptic NMDA glutamate signaling and the extrasynaptic NMDA glutamate signaling (73). Synaptic NMDA glutamate signaling plays a crucial role in enhancing neuronal health through the initiation of transcriptional changes that promote resilience against diverse traumatic stimuli (74). In contrast, extrasynaptic NMDA glutamate signaling is associated with several signaling pathways that increase neuronal susceptibility to trauma, leading to cell death or vulnerability (74). Extrasynaptic AMPARs play a significant role in the excitatory glutamatergic transmission within the spinal cord by adjusting synaptic efficiency and facilitating activity-dependent plasticity in spinal cord neurons (69). There are also pieces of evidence of the role of the extrasynaptic AMPA in spinal nociceptive signaling (75, 76). Extrasynaptic mGluRs represent one of the most diverse receptor families in the central nervous system (CNS), exhibiting a wide range of locations and functions (71, 77). NMDARs, AMPARs, and mGluRs are also expressed by microglia, astrocytes, and oligodendrocytes (78–84). In this study, we have simultaneously recorded the spinal glutamate signaling and cardiac electrophysiological signals in vivo to provide evidence that the spinal glutamate signaling can be used as a biomarker for acute myocardial ischemia severity detection. We think the cardiac functional effect we are observing is due to elevated synaptic release of the glutamate as well as glutamate spillover causing a complex modulation of the extrasynaptic glutamate receptors. We have shown that the glutamate concentration increases significantly between 15 and 30 min of the myocardial ischemia and it stays elevated even during the reperfusion. In this study we were not able to evaluate the fatal arrhythmias over time because of technical limitations; however, it seems the timing of the significant augmentation of the spinal glutamate might correlate with the second phase of ventricular arrhythmias during myocardial ischemia that has been reported in previous studies (85, 86).

Clinical Implications

The translational large-animal model that was used in this study closely resembles human physiology and pathophysiology, allowing for more accurate predictions of therapeutic outcomes in patients. The current design of this probe necessitates insertion following the laminectomy procedure, which is not feasible for cardiac patients. However, with slight modifications, it could be adapted for implantation using a minimally invasive surgical approach, similar to the placement of the spinal cord stimulation leads.

Real-Time Monitoring of Glutamate Signaling for Cardiac Interventions

The development of a flexible glutamate-sensing multielectrode array offers a novel technique for monitoring thoracic spinal neural activity in real time during cardiac interventions that engage the cardio-spinal neural pathway. Clinicians can use this technology to assess glutamate signaling levels, allowing for a comprehensive evaluation of spinal neural network activation during various stages of cardiac interventions. This real-time monitoring could provide crucial information to guide treatment decisions and optimize interventions for patients undergoing procedures with the potential to impact the sympathoexcitatory feedback to the heart.

Early Detection of Ventricular Arrhythmia Risk

The correlation between elevated glutamate signaling and reduced cardiac ARI, as well as DOR, suggests that increased spinal neural network activity is associated with higher sympathoexcitation and an elevated risk of ventricular arrhythmias. This elevated spinal glutamate signaling that causes sympathoexcitation and increases heart catecholamine level is also occurring during phase 1b of myocardial ischemia, which is associated with the high risk of arrhythmias (24, 25). This insight provides clinicians with a potential biomarker to assess a patient’s risk of arrhythmias during myocardial I/R injury. Identifying patients at a higher risk of arrhythmias at an early stage could allow for timely implementation of preventive measures and targeted interventions to mitigate adverse outcomes.

Potential Therapeutic Targets (Neuromodulation)

By identifying the spinal neural network and glutamate signaling as critical players in the generation of ventricular arrhythmias during I/R injury, this study opens new avenues for the development of therapeutic strategies targeting these pathways. Pharmacological agents or neuromodulation techniques aimed at modulating glutamate signaling in the dorsal horn of the thoracic spinal cord could hold promise in attenuating sympathoexcitation and reducing the incidence of ventricular arrhythmias in high-risk patients. Further research in this area could lead to innovative therapies for enhancing cardiac electrophysiological stability during myocardial I/R injury.

Moreover, although this study was focused on glutamate signaling during a cardiac event, this technique can be used to measure spinal glutamate signaling for other applications. For instance, the high glutamate signaling in the lumbar spinal cord could be a relevant indication of low back pain and can be used as a marker for novel neuromodulation therapies.

Limitations

Our study has some limitations that need to be addressed for clinical translation. First, neural activity can be suppressed in studies performed on anesthetized subjects. Although we used α-chloralose as an anesthetic agent to minimize the effect of the anesthesia on the excitation of neurons, there might still be some suppression effects on the neural network. In this study we have shown that the glutamate increase is correlated with the timing of ARI, DOR, and hemodynamic changes; however, a causal relationship cannot be established with the experimental setup in this study. Our developed sensor records the glutamate neurotransmitter signaling; however, it is not capable of determining the origin (afferent or efferent) of the glutamate it detects. In this study, we have used acute implants for a proof-of-concept study. Future studies are needed to use this technique in a chronically implanted model and evaluate the efficacy of the probes in a chronically implanted model.

SUPPLEMENTAL MATERIAL

Supplemental Figs. S1–S4: https://doi.org/10.6084/m9.figshare.22932026.

GRANTS

The study was funded by National Institutes of Health (NIH) Grants R01 HL136836, R01 NS110564, and R21 DA049592. Author support was received from the Competitive Medical Research Fund from the University of Pittsburgh (to S. Salavatian); NIH Training Grant T32 NS086749 (to E. M. Robbins); and NIH Grants R01 R01NS110564 and R21 DA049592 (to X. T. Cui) and R01 HL136836 and R44 DA049630 (to A. Mahajan).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

S.S., E.M.R., X.T.C., and A.M. conceived and designed research; S.S., E.M.R., Y.K., and E.C. performed experiments; S.S., E.M.R., and Y.K. analyzed data; S.S., E.M.R., E.C., X.T.C., and A.M. interpreted results of experiments; S.S., E.M.R., and Y.K. prepared figures; S.S. and E.M.R. drafted manuscript; S.S., E.M.R., Y.K., E.C., X.T.C., and A.M. edited and revised manuscript; S.S., E.M.R., Y.K., E.C., X.T.C., and A.M. approved final version of manuscript.