Prefrontal cortical recordings with biomorphic MEAs reveal complex columnar-laminar microcircuits for BCI/BMI implementation

Abstract

The mammalian prefrontal cortex known as the seat of high brain functions uses a six layer distribution of minicolumnar neurons to coordinate the integration of sensory information and the selection of relevant signals for goal driven behavior. To reveal the complex functionality of these columnar microcircuits we employed simultaneous recordings with several configurations of biomorphic microelectrode arrays (MEAs) within cortical layers in adjacent minicolumns, in four nohuman primates (NHPs) performing a delayed match-to-sample (DMS) visual discrimination task. We examined: (1) the functionality of inter-laminar, and inter-columnar interactions between pairs of cells in the same or different minicolumns by use of normalized cross-correlation histograms (CCH), (2) the modulation of glutamate concentration in layer 2/3, and (3) the potential interactions within these microcircuits. The results demonstrate that neurons in both infra-granular and supra-granular layers interact through inter-laminar loops, as well as through intra-laminar to produce behavioral response signals. These results provide new insights into the manner in which prefrontal cortical microcircuitry integrates sensory stimuli used to provide behaviorally relevant signals that may be implemented in brain computer/machine interfaces (BCI/BMIs) during performance of the task.

Keywords: Columnar processing; Executive control; Glutamate modulation; Microcircuits; Nonhuman primates; Prefrontal cortex.

Figure 1. Simultaneous columnar-laminar recording in primate brain during cognitive tasks

A. Behavioral paradigm showing the sequence of events in the DMS task. The DMS task consisted of (1) presentation of a “Start Ring” to initiate the trial, (2) presentation of the Sample Target image, followed by (3) a variable Delay Period of 1-50 sec, prior to (4) presentation of the Match Target (i.e. Sample) image accompanied by 1-7 other Non-match (distracter) images on the same screen in which movement of the cursor into the correct image (Match response) produced juice reward via a sipper tube placed next to the animals mouth. B. Behavioral performance in the DMS task. Mean percent correct performance (over all animals) in Spatial (blue) and Object (red) trials. C. Site of the recording chamber in the prefrontal cortex (PFC) of NHP. D. Coronal section showing relative location of neuromorphic multielectrode array (MEA) for recording in layers 2/3 and 5.

Figure 2. Example of simultaneous recordings of prefrontal neurons with neuromorphic multi-electrode arrays

A,B,C. Illustration of configuration for three different types of neuromorphic probes W1, W2, W3 used in columnar recordings. D,E,F. Example of simultaneous recordings in the prefrontal cortex. The code color for the neural activity in cortical layers is: layer 2/3 (blue), layer 4 (green) & layer 5 (red). Peri-event histograms (PEHs) of cell activity simultaneously recorded with neuromorphic probes during a single session. Separation distance of the recording pads is shown for each MEA diagram with cells recorded from those locations indicated by different markers.

Figure 3

Peri-event histograms and rasters showing differential firing of two neuron pairs recorded from supra (layer 2/3) and infra-granular (layer 5) layers of the prefrontal cortex. PEHs depict layers 2/3 (blue) & layer 5 (red) in the DMS task on spatial and object trials during sample (A), match (B) and correct vs. error trials (C) in the same session. Figures A & B were adapted from Opris et al, 2013 and Figure C was adapted from Opris et al. 2012a.

Figure 4. Complex minicolumnar firing

Cross-correlation histograms (CCHs) showed inter-laminar (A), inter-columnar (B,C) microcircuit interactions between prefrontal cells in different (or same) cortical layers and different (or same) minicolumns, respectively. To distinguish the time base of the neural interactions we used short-lag (0 to 5 ms) with binsize of 0.1 ms and long lag (0-400ms). with binsize of 5ms. The 99 % confidence intervals are depicted by red lines indicating events with significant covariant firing. Shift predictor was subtracted. This red lines indicate events with significant covariant firing.

Figure 5

A. Example of long-lag cross-correlation histogram (CCH) for an infra-granular cell pair (vertically separated by 100 μm on the MEA) with a broad peak at 50-100 msec lag (binwidth was 5 msec). B. Distribution of the inter-layer (red circle) vs. intra-layer (blue circle) cross-correlation peaks as a function of temporal lag. CCHs show probability of synchronized firing (ratio of extracellular spike occurrences) of layer 5 cells within ± 150.0 ms of individual spike occurrences from the layer 2/3 cell (0 ms in CCH). There were 154 intra-layer pairs and 56 of interlayer pairs. Firing synchrony calculated over entire trials between start ring onset and reward delivery. C. Example of short-range cross-correlation histogram (CCH) for an infra-granular cell pair (horizontally separated by 40μm on the MEA) with a central sharp peak and a broader peak having a 1-2 msec lag (binwidth was 0.1 msec). CCH provides evidence for common input on both cells in the pair and direct input from left to right cell. D: Distribution of the intra-layer (blue circle) cross-correlation peaks as a function of temporal lag. The 99 % confidence intervals are depicted by red lines in A & C. The mean Shift predictor was subtracted from all CCHs.

Figure 6. Glutamate Recording with Biomorphic MEAs on DMS Task

A. Spencer-Gerhardt 2 microelectrode arrays consist of four recording sites (15 x 333 μM) on a 7 cm polyimide shank that were coated with Nafion®, forming an anion exclusion layer. Dorsal recording sites (Sentinel) were coated with BSA + glutaraldehyde. The sentinel site records the current generated from any electroactive interferents that are not excluded by the Nafion® coating. Ventral recording sites were coated with Glutamate oxidase and BSA + glutaraldehyde. The GluOx coating allows the ventral pads to be sensitive to glutamate release through the enzymatic production of H₂O₂. A +0.7V potential applied to the MEA vs. Ag reference electrode oxidizes H₂O₂ resulting in a current that is directly related to the glutamate concentration. B. Coronal section with the MEA inserted for glutamate recording in layer 2/3 of prefrontal cortex in NHPs. C,D. Comparison of mean tonic glutamate concentration over 2 seconds after Match Target presentation, broken down by high/low load in correct/error trials across Spatial vs Object trials. E,F. Comparison of mean phasic glutamate release frequency under the same conditions across Spatial vs. Object trials. G,H. Comparison of mean phasic glutamate release amplitude across Spatial vs. Object trials that were successful (high load) vs. likely errors (low load) determined by assessment of neuron firing (Figure 3). I,J. Scatter plots comparing % change in glutamate concentration (I) and frequency (J) of release events. **p <0.001; ANOVA analyses with post hoc tests.