Introduction

L-glutamate, the most common excitatory neurotransmitter in the mammalian central nervous system (CNS), primarily regulates presynaptic and postsynaptic receptors1–3. Abnormal transmission of glutamate can cause neurological diseases such as communication dysfunction, cognitive impairments, schizophrenia, Parkinson's disease, stroke and epilepsy4,5. Glutamate transformation in excitatory and neurovirulent processes has been studied for decades6–11. However, activity in the nervous system includes a complex combination of biochemical and electrical events in space and time12–14. Local field potential (LFP) integrates predominantly synaptic input signals from a large population of neurons, whereas spikes or action potentials are output signals of a single neuron15–17. Thus, investigating the extracellular electrophysiological signals corresponding to these activities may enable better understanding of glutamatergic processes16,18.

Several methods have been designed to conduct dual-mode (electric and chemical) neural information recording. Using separate electrodes to examine glutamate and electrophysiological signals within the same preparation has been a commonly employed method19–22; however, it is problematic when the separate microelectrodes are not located within the same microenvironment. Several reports delivered external glutamate or drugs that could affect glutamate release into the rat brain and recorded only electrophysiological signal changes23–28; however, this method can serve in only a limited capacity for glutamate change is not quantitative. Studying dual-mode activity by randomly dividing rats into groups for electrophysiological and neurotransmitter monitoring is another method that has been pursued29. When using this method, it is important to consider the sample individual difference. Consequently, the development of a novel electrode capable of simultaneously detecting electrochemical and electrophysiological signals in vivo is of great significance and should enable the monitoring of glutamatergic activity in living systems.

For this purpose, an implantable silicon microelectrode array (MEA) probe incorporating both Pt electrophysiological and glutamate recording sites was fabricated using micro-electromechanical systems (MEMS) method. In this approach, the microelectrode arrangement can be flexibly designed and precisely controlled for use in dual-mode in vivo recording30–33. We modified the MEA probe with Pt nanoparticles and 1,3-phenylenediamine (mPD) to improve electrical performance. We then implanted the MEA probe into the striatum of anaesthetised rats and, following intraperitoneal (i.p.) administration of KCl, examined extracellular glutamate, LFP and spike activity. The results demonstrated the ability of the implantable MEA to simultaneously record dynamic changes in L-glutamate and neural electrical activity in vivo.

Materials and Methods

Reagents and apparatus

Glutamate oxidase (GluOx) was obtained from Yamasa Corporation, Japan. Ascorbic acid (AA, ≥99%), 3,4-dihydroxyphenylacetic acid (DOPAC, ≥98%) and 5-hydroxytryptamine (5-HT, 99%) were obtained from Alfa Aesar Corporation, USA. Dopamine (DA, ≥99%) was obtained from Acros Organics, Belgium. Saline (0.9% NaCl) was purchased from the Shuanghe Company, China. Urethane (≥98%) was obtained from Sinopharm Chemical Reagent Co., Ltd, China. mPD (≥99%) was purchased from Aldrich, USA. Bovine serum albumin (BSA, ≥99%) was purchased from Amresco, USA. Glutaraldehyde solution (GA, 25%) and L-glutamate sodium (>98%) were purchased from Shanghai Chemical Reagent Company, China. Phosphate-buffered saline (PBS, 0.1 M, Na2HPO4-NaH2PO4-KCl, pH 7.4) was prepared from a PBS tablet (Sigma) with deionised water. Water was purified through a Michem ultrapure water apparatus, China (resistivity >18 MΩ).

The KCl solution (100 mM, 1 mL kg−1) used for i.p. administration was prepared in normal saline and sonicated for 5 min at 37 °C to ensure complete dissolution. The remaining solutions were prepared in PBS (0.1 mM, pH 7.4). DOPAC, DA, 5-HT and AA solutions were prepared just before use because of their propensity to decompose over time.

All electrochemical measurements were performed on a Gamry electrochemical workstation (Gamry Reference 600, Gamry Instruments, USA). Electrophysiological signals were recorded using an integrated 16-channel filter amplifier and data acquisition system (USB-ME16-FAI-System, MultiChannel Systems, Germany).

MEA probe fabrication and modification

An implantable MEA probe was created using silicon-on-insulator substrates (SOI, 30 μm Si/2 μm SiO2/600 μm Si) by MEMS method using the following steps. (i) The front side of the SOI (30 μm Si) was coated with 0.5 µm of silicon dioxide by wet oxidation. (ii) The electrophysiology and glutamate recording sites, lead wires and bonding pads in the MEA probe were patterned by photolithography (positive photoresist AZ1500, 1 μm thick), Pt/Ti (250 nm/30 nm thick) sputtering and lift-off methods. (iii) A final silicon nitride (0.8 µm) layer was deposited via plasma enhanced chemical vapour deposition (PECVD, 300 °C) to provide insulation. Windows to the recording points and bonding pads were opened using reactive-ion etching (RIE). (iv) The MEA probe shape was defined by deep RIE from the top of the 30 μm thick silicon layer down to the buried oxide. (v) A thick picein wax (Kunlun 80#, Oil Refining Chemical General Factory Chinese Oil Yumen Oilfield Company, China) was spun on the wafer front side for protection, and KOH (30%) was used to wet etch the full 600 μm span of backside silicon down to the buried oxide. In this way, the individual MEA probe was released from the underlying silicon substrate. The picein wax was washed out using negative photoresist developer and fuming nitric acid. The imbedded oxide was smashed and removed by ultrasonic cleaning in pure water with a power of 40 W for 2 min.

The MEA probe consisted of a single 7 mm-long shank and was 30 μm × 343 μm in cross section. Along the shank, there was a linear array of 14 microelectrode sites that alternated between being round (diameter = 15 μm) and rectangular (60 μm × 125 μm) for transduction of electrophysiological and amperometric signals, respectively (Figure 1a). The centre-to-centre site separation was 170 μm between two adjacent electrochemical sites, 80 μm between two adjacent electrophysiological sites, and 150 μm between an electrophysiological site and an adjacent electrochemical site. Maintaining a 50–200 μm distance between recording sites was reasonable as a shorter inter-electrode separation could result in possible cross-talk artefacts, whereas a larger separation would miss cross-correlations between neuron pairs18,34. The MEA probe was wire bonded to a printed circuit board (PCB) holder for handling and connecting to the recording equipment (Figure 1b). The bonding pads, bonding wires and metal lines on the PCB were embedded in silicone rubber (Nanda 705#, Liyang Kangda Chemical Co. Ltd., China) for the purpose of mechanically protecting and electrically isolating the connections. The assembled MEA probe could then be implanted into a rat's brain (Figure 1c).

Figure 1
figure 1

(a) Photomicrograph of the MEA probe tip. Along the shank, the microelectrodes alternated between being round (diameter = 15 μm) and rectangular (60 μm × 125 μm) in shape to enable the recording of electrophysiological and amperometric signals, respectively. Pt nanoparticles were deposited on the electrophysiological electrodes. (b) The MEA probe was assembled into a PCB holder. (c) The assembled MEA probe implanted in a freely moving rat.

Full size image

To reduce the impedance of the electrophysiological electrode, Pt nanoparticles were electrodeposited on the round microelectrodes using chloroplatinic acid (48 mM) and lead acetate (4.2 mM) in a 1:1 mixed solution at −1.0 V opposite a Pt electrode for 60 s (Figure 1a). The adsorption and underpotential deposition of lead ions in the electrodeposited platinum led to changes in morphology and crystal size of the Pt nanostructures35.

Cross-linked covalent binding through glutaraldehyde was employed to immobilise the GluOx in the Pt electrochemical recording site. Using a three-dimensional micro-operator, the MEA probe was inserted into a glass capillary filled with a mixed solution of GluOx (1%), BSA (1%) and GA (0.125%), and the probe insertion length was precisely controlled. Under microscopic guidance, when the first glutamate recording site was covered with GluOx solution, the probe was drawn out of the solution. Thus, only the first site was exposed to GluOx (Supplementary Figure S1).

MPD was electrochemically polymerised on the eight electrochemical electrodes (every area equalled 60 × 125 μm2) at a potential of +0.6 V versus an Ag|AgCl reference electrode for 15 min in a 5 mM mPD solution prepared in PBS.

Electrochemical characterisation

The impedance spectrum of the electrophysiological microelectrodes with and without Pt nanoparticles was evaluated using electrochemical impedance spectroscopy (EIS) in an electrochemical workstation. The MEA probe and an Ag|AgCl reference electrode were placed in a PBS solution. The impedance was assessed over a frequency range of 0.1 Hz to 100 kHz at a voltage level of +0.02 V.

Electrochemical detection of glutamate by the sensor occurs as follows. First, glutamate is converted by GluOx to hydrogen peroxide (H2O2), α-ketoglutarate and NH3. Next, H2O2 is electrochemically oxidised on the surface of the Pt electrode, thereby providing a current signal proportional to the glutamate concentration. Calibrations were carried out in PBS buffer with a two-electrode configuration, the applied potential was held at +0.7 V vs. Ag|AgCl; this value is typically used for H2O2 detection. A 5 × 2 mm2 stir bar was added to the PBS and the solution was stirred slowly to prevent the formation of a vortex in the solution. The sensitivity and detection limit were determined by a stepwise addition of 5 μM to 30 μM glutamate into the buffer. There are several common neurotransmitters that can cause interference in brain extracellular fluid (ECF), including DOPAC (≈20 μM in ECF), AA (250–500 μM in ECF), DA (≤100 nM in ECF) and 5-HT (≤1 μM in ECF)36. To investigate the selectivity of the electrode, the responses of the GluOx-coated microelectrodes and mPD–GluOx-modified microelectrodes to the above-listed interferences were compared. Stock solutions of the interfering molecules were diluted into buffer to produce the following solutions: 20 μM DOPAC, 1 μM DA, 1 μM 5-HT and 250 μM AA.

In vivo experiments

For in vivo testing, male Sprague–Dawley rats (270 g) were individually anaesthetised with urethane (1.4 g kg−1, i.p.), and an MEA probe was implanted into the striatum of each animal (AP: +1.0 mm, ML: −2.5 mm, DV: −5.0 mm). A homemade Ag|AgCl reference electrode (an Ag wire on which AgCl was electrodeposited) was placed into the cortex. An earth wire, used as a ground, was attached to one of the support screws during recording.

The implanted MEA probe was connected to the electrophysiological recording system and the electrochemical workstation. The six-channel electrophysiological signals were sampled simultaneously at a rate of 25 kHz. A low pass filter was applied at 100 Hz to view the LFP, and a high pass filter was applied at 500 Hz to view the neural spikes. Such broadband recordings allow for the simultaneous investigation of spikes and LFP. Power spectral densities of recorded LFPs were calculated in MATLAB using the correlation function and Fourier transform function (Hamming window, 2 s window, 1 s steps). The glutamate signals were recorded using constant amperometry responses. The working potential was held at +0.7 V vs. the Ag|AgCl reference wire. Glutamate recordings were not made until at least 30 min had elapsed after probe insertion to allow for glutamate levels to equilibrate, and signals were recorded at 0.1 s intervals. The entire experiment was conducted inside a Faraday cage. All of the procedures detailed above complied with the guidelines of the State Scientific and Technological Commission for the care and use of laboratory animals.

Results and Discussion

Impedance test

Figure 2 details the electrochemical impedance spectrum of the electrophysiological microelectrodes before and after Pt nanoparticles were deposited on them. The nanoparticles possessed large surface areas37 and therefore increased the geometrical surface area of the electrode, reducing electrode impedance. In using Pt nanoparticles, the mean impedance of the microelectrodes at 1 kHz decreased from 0.33 to 0.018 MΩ (n = 6) (Figure 2), which is a desirable range for electrical neural recording.

Figure 2
figure 2

Impedance spectra of electrophysiological microelectrodes before and after Pt nanoparticles were deposited on them. The frequency range was 0.1–100 kHz, and the voltage level was +0.02 V vs. Ag|AgCl reference electrode.

Full size image

Glutamate calibration

Prior to the surgical procedures, each GluOx–mPD-modified recording site was characterised for its sensitivity to glutamate. The microelectrode exhibited a sensitivity of 56 pA µM−1 to L-glutamate and had a relative coefficient of 0.995 over a range of 5–30 µM (Figure 3a and b). The detection limit was 0.5 µM when the signal-to-noise ratio was greater than 3. These are viable conditions for in vivo recording as basal glutamate concentrations of SD rats ranges between 1.2 μM and 8.1 μM4,38,39. As shown in Figure 3c, the response time of the electrode was shorter than 8 s after the addition of 5 µM L-glutamate. When the glutamate level was changed to a lower concentration, the response time was also shorter than 8 s (Supplementary Figure S2). Figure 3d illustrates the typical amperometric responses to glutamate, DOPAC, DA, 5-HT and AA of the GluOx-coated electrode (with and without mPD electropolymerisation) in the on-line detection system at a working potential of +0.7 V. The results demonstrated that the interference conditions could not produce measurable responses in the mPD–GluOx-modified channel, which convincingly indicates that measurements of glutamate were essentially interference-free with respect to these electroactive species.

Figure 3
figure 3

Performance of the glutamate microelectrode, the applied potential was held at +0.7 V vs. Ag|AgCl. (a) mPD–GluOx microelectrode response to varying concentrations of glutamate. (b) Plot of current vs. glutamate concentration (n = 3). The sensitivity was 56 pA µM−1, R = 0.995. (c) Enlargement of selected section from curve (a). (d) Performance of GluOx-coated microelectrodes without (top trace) and with (bottom trace) mPD electropolymerisation. Arrows indicate the additions of various substances: L-glutamate (3 µM, three times), DOPAC (20 μM), DA (1 μM), 5-HT (1 μM), AA (250 μM), and another addition of L-glutamate (3 μM), numbers in the parentheses are the final concentration.

Full size image

The stability of the glutamate microelectrode can be illustrated by following observations: the sensitivity to glutamate decreased to 97.60 ± 3.08% (n = 3) after 2 h of recording in vivo, and it decreased to 46.32 ± 5.08% (n = 3) after 9 h of recording in vivo. The coefficient of variation for glutamate measurements acquired under the same in vivo conditions was 2.92% (n = 3); measurements of reproducibility were performed at different time points using a single rat. The glutamate microelectrode maintained 93.39 ± 2.71% (n = 3) of its original sensitivity after being stored at 4 °C for 26 days.

The cross-linked enzyme (GluOx) layer that was coated on the microelectrode had a thickness of 140–180 nm, as measured by a step profiler, and there was evidence that the subsequent electropolymerisation of mPD produced a thin, self-sealing, insulating polymer film with thickness of ~15 nm40 on the microelectrode surface (Supplementary Figure S3). Poly (mPD) film possesses excellent interference-rejection characteristic as it can prevent larger molecules such as AA, DA, 5-HT and DOPAC from reaching the microelectrode surface41. Smaller molecules, such as H2O2, are still able to pass through the film. The use of a poly (mPD) film also serves as a method of immobilising GluOx and protecting the microelectrode surface from fouling. L-glutamate is not able to penetrate a poly (mPD) film; however, when it comes into contact with the GluOx molecules located at the polymer|electrolyte interface, it can be catalysed to produce H2O2. The resultant H2O2 can pass through the poly (mPD) film and is oxidised on the microelectrode surface. Thus, the microelectrode current signal that is produced is proportional to L-glutamate concentration.

Concurrent recordings of striatum glutamate overflow and electrophysiological changes

An MEA probe was implanted into the striatum of an anaesthetised rat. To explore the capability of the MEA probe at detecting dynamic changes in glutamate and electrophysiological signals, KCl (100 mM, 1 mL kg−1) was administered to the rat through i.p. injection.

Figure 4a shows the influence of KCl on striatum glutamate levels. The basal striatum glutamate concentration was 3.01 ± 1.27 μM before KCl injection. After 1.23 ks of KCl injection, the glutamate increased to 4.67 ± 0.82 μM, and this increased concentration lasted for 0.75 ks (Figure 4a). Following this, there was a gradual return to baseline values. The high concentration of K+ caused massive depolarisation of neurons and consequently led to the release of glutamate from vesicular, cytosolic pools and astrocytes38. The average striatum glutamate release stimulated by K+ was calculated to be ~1.66 µM.

Figure 4
figure 4

I.p. injection of KCl (100 mM, 1 mL kg−1) evokes changes in glutamate concentration and electrophysiological activity, which were simultaneously recorded along an MEA probe implanted into the striatum of an anaesthetised rat. The left panel shows the marked recording sites in the MEA probe. (a) Variation in glutamate concentration elicited by i.p. injection of KCl (indicated by an arrow). The power spectrogram of LFP recorded in Channel 1 (Ch. 1). (b) from 0 to 100 Hz and (c) from 12 to 18 Hz. (d) The corresponding mean firing rates of different type spikes. The same time epochs were used in (a), (b), (c) and (d). (e) LFPs (upper) and spike trains (lower) recorded at 0.26 ks before (T0) and 2.08 ks after (T1) KCl injection. (f) Power spectral densities of the LFPs from Ch. 1 and Ch. 6 in (e). (g) Spike waveforms recorded in each recording channel before (T0) and after (T1) KCl injection. The spike colours are the same as in (d), (e) and (f). T0 and T1 indicate the times at which the LFPs and spike traces were obtained.

Full size image

Prior to KCl stimulation, the striatum LFP exhibited slow-wave activity (0.1–4 Hz) (Figure 4b), and the spike exhibited low firing rates and non-bursting activity (Figure 4d and e). These conditions are characteristics of urethane anaesthesia. The extracellular recorded spikes were classified into two different types according to their waveforms (Figure 4g). The type 1 spike, which appeared in microelectrode channels 1, 2, 4 and 5, was fired with a 0.4 ms positive protrusion before the 0.4 ms negative phase and was followed with a positive phase of 0.6 ms. The type 2 spike, which appeared in microelectrode channels 3 and 6, was initially fired with a negative phase of 0.4 ms and was followed with a positive phase of 0.6 ms.

After KCl stimulation, the released glutamate led to a decrease in rheobase and depolarisation voltage and compressed the range of recruitment threshold current42, thereby exciting the glutamate-exposed neurons. Therefore, the firing rates of both types of spike and the LFP low-beta (12–18 Hz) activity (Figure 4c), which represents the potentiation of excitatory neurotransmission between co-activated neuronal networks, were increased by glutamate augmentation. In the case of the type 1 spike, the firing rate increased from 7.24 ± 3.06 spikes s−1 at 0.26 ks before KCl injection (T0) to 19.60 ± 8.07 spikes s−1 at 2.08 ks after of KCl injection (T1). The firing rate of the type 2 spike increased from 2.72 ± 0.35 spikes s−1 at T0 to 9.01 ± 3.10 spikes s−1 at T1. The duration of the switch in LFP power and spike firing rates paralleled the time course of KCl-induced extracellular glutamate fluctuation. To validate the dual-mode results, glutamate and electrophysiological signals were obtained from the brain of a dead rat, which should not produce any biochemical or bioelectrical signals (Supplementary Figure S4).

Multichannel signals reflect the spatial composition distribution in a recorded brain region. Figure 4f shows that the LFP power spectral density obviously changed with respect to the positioning of recording sites, which indicates that the striatum has inputs from different encephalic regions43. In Figure 4g, two types of spike waveform are indicated, revealing that two classes of neurons were recorded for each neuronal type generated identical action potentials44. These results provided evidence that different glutamate input and output pathways exist in the striatum.

In the present study, the potential of using a dual-mode implantable MEA as a research tool in vivo was explored. The MEA probe presents several advantages over alternative neural recording methods. First, due to its dual-mode design, a lesser degree of brain damage is caused when using this probe, which may result in the detection of more physiologically relevant extracellular glutamate pools and electrophysiological signals. Second, the sensor is able to simultaneously detect second-to-second changes in glutamate and electrophysiological signals. Finally, the microelectrode sites have precise spatial definition, which can be an advantage when studying layered brain structures. With the real-time multichannel dual-mode signals that can be acquired by the MEA probe, the spatiotemporal relationships between neurons that are embedded in different regions of brain tissue can be thoroughly analysed.

Conclusion

Using MEMS, we created a novel implantable MEA probe that can produce simultaneous measurements of glutamate, LFP and spike activities across multiple spatial locations in the rat brain. Recordings from the MEA probe that were taken in the striatum of urethane-anaesthetised rats revealed spatiotemporal K+-evoked dual-mode signal changes, which indicates that the MEA probe is useful for investigating neural dynamics in vivo. In the future, we intend to apply this biosensor technology to analyse neurologic dysfunction in a rat model of Parkinson's disease. Footnote 1