Multisite electrophysiological recordings by self-assembled loose-patch-like junctions between cultured hippocampal neurons and mushroom-shaped microelectrodes

Substrate integrated planar microelectrode arrays is the “gold standard” method for millisecond-resolution, long-term, large-scale, cell-noninvasive electrophysiological recordings from mammalian neuronal networks. Nevertheless, these devices suffer from drawbacks that are solved by spike-detecting, spike-sorting and signal-averaging techniques which rely on estimated parameters that require user supervision to correct errors, merge clusters and remove outliers. Here we show that primary rat hippocampal neurons grown on micrometer sized gold mushroom-shaped microelectrodes (gMμE) functionalized simply by poly-ethylene-imine/laminin undergo self-assembly processes to form loose patch-like hybrid structures. More than 90% of the hybrids formed in this way record monophasic positive action potentials (APs). Of these, 34.5% record APs with amplitudes above 300 μV and up to 5,085 μV. This self-assembled neuron-gMμE configuration improves the recording quality as compared to planar MEA. This study characterizes and analyzes the electrophysiological signaling repertoire generated by the neurons-gMμE configuration, and discusses prospects to further improve the technology.


between cultured hippocampal neurons and mushroom-shaped microelectrodes
Nava Shmoel 1# , Noha Rabieh 1# , Silviya M. Ojovan 1 , Hadas Erez 1 , Eilon Maydan 1 , and Micha E. Spira 1* Supplementary Figure 3. Simulation of the electrical coupling coefficient of synaptic potentials as a function of the seal-(R s ) and junctional-(R jm ) membrane resistances. The simulation in (a), (b) and (c), was conducted by delivery of 100Hz sine wave depicting synaptic potentials to an analog electrical circuit (text figure 4) with R jm values of 1 (a), 10 (b) and 100 GΩ (e). Assuming that: (i) the amplitudes of the synaptic potentials is 10 mV, (ii) that the noise level of the recording system is 20µV, (iii) that potentials with amplitude 3 times larger than the noise level (≥ 60 µV) can be detected. Then, theoretically coupling coefficient of ≥ 0.006 (60 µV) are sufficient to detect synaptic potentials. The gray areas and dashed lines in (b) and (c), mark the seal resistance values that prohibit recording of the simulated synaptic potentials. All other R s values (white background) permit the detection of 100 Hz sine wave generated by a 10 mV source.
The expected recording quality as a function of the seal resistance is further illustrated by simulations of the "recorded synaptic potentials" (d) and (e). For the simulations, a trace of patch electrode recording of an APs and a barrage of synaptic potential (black traces in (d) and (e)) was fed into a simulation circuit where R jm was set to be 10GΩ and R s values of 5, 10, 25, 50 and 100 MΩ (red traces in d,) or an R jm of 1MΩ and R s values of 5, 10, 25, 50 and 100 MΩ (red traces in e). Black traces (ei) and (di) depicts low gain recordings of an action potential and synaptic potentials. Black traces (dii) and (eii) are high gain recordings of the synaptic potentials only. The red traces depict the simulated recordings for the Rs values marked near the recorded potential. Note that for Rjm=10GΩ and R s values of <50 MΩ (d) the expected synaptic signals are lower than the 20 µV noise level (gray background) and thus cannot be recorded. Nevertheless for Rjm=1GΩ synaptic potentials can be recorded even when R s is as low as 5 MΩ (e). For all simulations shown R ep =10 MΩ, CPE = 25 MΩ at 1 KHz and the amplifier impedance =20 MΩ at 1 KHz.

Supplementary Figure 4.
An example of the approach to define the crosstalk levels between individual gMµE-MEA channels. The level of crosstalk between channels was assessed using 7-10 day old cultures. (a) Simultaneous voltage recording from the cluster of 8 gMµE shown in (b). Electrode E76 (red in (a) and (b)) recorded a >2mV monophasic action potential from a spontaneously and repeatedly firing neuron (the black trace in (a) is the average of five firing events). The average of simultaneously recorded voltage traces from 7 nearby electrodes (labeled green in (a) and (b)) are displayed at a higher gain above trace E76. There was NO crosstalk between the channel that recorded the AP and the cluster of the nearby gMµE channels (the small blip in E75, E86 and E57 is not crosstalk as can be clearly seen when enlarging the figure).

Estimation of gMµE resistance and capacitance for simulation purposes
The impedance of freshly fabricated gMµE was measured using an HP 4284A Precision RLC meter, at 1 KHz, at room temperature, in a 0.9 % NaCl solution (Ojovan et al. 2015) 33 . The average values of a constant phase element (CPE) with a resistance of 3.5 MΩ and a capacitance of 5.1 pF in series were extracted. Impedance spectrum measurements revealed that newly fabricated gMµE have pure capacitive characteristics with an impedance that follows Z ∝ 1/f (where f is the frequency in the range of 1 to 100 kHz). Hence, the complex nature of the CPE could have been neglected for all the relevant frequencies and be presented as a simple passive element.
It is well-established that the impedance of gold electrodes can be reduced in a number of ways. Practically, exposure of gold MEA to plasma oxygen and ionic solutions leads to reduced impedance. This reduction can be depicted as an increased parallel conductance to the CPE (Fig. 4). To estimate the value of the parallel resistance (R ep ) we characterized the properties of the gMµE by applying a calibration voltage square pulse (10 mV, 10 ms) between an Ag/AgCl electrode immersed in the solution and the ground, and measured the voltage read by the MEA amplifiers before and after oxygen plasma and incubation of the gMµE in a phosphate buffer solution for 24 h (Fig. 4, a and b). As a result of these treatments, a substantial increase in the rate of rise and decay of the calibration pulse was recorded (defined by ) and a higher saturation voltage was observed. Since the steady state output voltage is mainly determined by the ratio of the gMµE-R ep to the amplifier input resistance, we concluded that the gMµE resistance was reduced. To estimate the R ep of the gMµE we simulated the calibration pulse output for gMµEs of different R ep in the range of 10-1000 MΩ in parallel to a CPE of 25 MΩ at 1 KHz 33 .The simulation revealed that the observed experimental changes in the gMµE resistance were best simulated by changing the R ep from a value of approximately 1 GΩ to 10 MΩ (Fig. 4,  c).
To estimate the range of errors that can be introduced by using the wrong R ep value for the simulations of action potential amplitudes we examined the frequency response of the system for values of R ep between 10 MΩ and 100 MΩ (Fig. 4, d). The line (in Fig. 3, d) represents the absolute difference between the two curves. Whereas at low frequencies (100-400 Hz) the error introduced is significant, at 1,000 Hz (representing action potentials) the difference is 200 µV or . We assume that the margin of inaccuracy of our estimate of R ep is smaller than one order of magnitude (10 MΩ compared to 100 MΩ), thus making our simulations accurate enough to estimate the properties of the gMµE.