Prefrontal cortical microcircuits bind perception to executive control

Abstract

During the perception-to-action cycle, our cerebral cortex mediates the interactions between the environment and the perceptual-executive systems of the brain. At the top of the executive hierarchy, prefrontal cortical microcircuits are assumed to bind perceptual and executive control information to guide goal-driven behavior. Here, we tested this hypothesis by comparing simultaneously recorded neuron firing in prefrontal cortical layers and the caudate-putamen of rhesus monkeys, trained in a spatial-versus-object, rule-based match-to-sample task. We found that during the perception and executive selection phases, cell firing in the localized prefrontal layers and caudate-putamen region exhibited similar location preferences on spatial-trials, but less on object- trials. Then, we facilitated the perceptual-executive circuit by stimulating the prefrontal infra-granular-layers with patterns previously derived from supra-granular-layers, and produced stimulation-induced spatial preference in percent correct performance on spatial trials, similar to neural tuning. These results show that inter-laminar prefrontal microcircuits play causal roles to the perception-to-action cycle.

Figure 1. The perception-to-action cycle with the behavioral paradigm.

(A). The illustration of the perception-to-action cycle. The diagram depicts the flow of spatial and object signals during perceptual and executive selection of target stimuli in a rhesus macaque brain. In visual cortical area V2 visual information splits into dorsal (spatial signals) and ventral (object signals) pathways that send signals to the top of executive hierarchy in prefrontal cortex, and then top-down through the cortico-striatal-thalamo-cortical loops. Blue arrows depict the perceptual flow of information while red arrows indicate the action (executive) signal flow from prefrontal cortical layer 5 to dorsal striatum, with the red dotted arrow indicating the thalamo-cortical projection in the cortico-striatal-thalamo-cortical loop. The two adjacent cortical minicolumns with red and blue filled circles indicate inter-laminar simultaneous recordings, while caudate-putamen recording are shown in green and pink circles. PFC-prefrontal cortex layers L2/3 and L5, and V2-secondary visual cortex region. (B). Behavioral paradigm showing the sequence of events in the rule-based DMS task. Each trial begins with ‘trial start images’ (‘ring’ or ‘box’) to initiate an ‘object’ or ‘spatial’ trial, respectively. Then, presentation of the ‘Sample Target’ image is accompanied by a ‘Sample Response’, followed by a variable ‘Delay’ period of 1–40 sec, with blank screen; followed by presentation of the ‘Match’ screen with Sample image accompanied by 1–6 Non-match (distracter) images, requiring movement of the cursor into the correct Match target determined by ‘trial start’ screen (Spatial trial = same location on the screen, or Object trial = same image-irrespective of position, responded to in the Sample phase) after presentation to receive a juice reward, via an accessible sipper tube. Placement of the cursor into a Non-match target (>0.5 s) caused the screen to blank without reward delivery. Inter-trial interval (ITI) = 10.0 s. (C). Behavioral performance in the DMS task. Behavioral performance (% correct trials) is shown separately for spatial trials (blue) vs. object trials (red) for trials ranging from 2–4 images (F(1,239) = 12.54; p < 0.001) and 1–40 sec delays (F(1,239) = 12.32; p < 0.001). Asterisks: **p < 0.001, ANOVA.

Figure 2. Neural firing in prefrontal cortical layers and striatum on spatial vs. object trials.

(A–F). Example of simultaneous individual activity (individual trial rasters and peri-event histograms) of single neurons recorded in prefrontal cortical layers L2/3 (A, D: blue) and L5 (B, E: red) with the conformal MEA and caudate n. (C, E: green) during Sample (A,B,C) and Match (D,E,F) target presentation on Spatial (left panels) and Object (center panels) trials during a single session (n = 120 trials). The purple marks in the rasters represent the time when the target was reached. Directional tuning plots (A, B, C for perception and C, D, E for executive selection, right panels) depict firing preference, measured by the radial eccentricity (in spike/sec or Hz) in the polygonal contour for the eight different target locations on the screen where images appear. The overlay tuning plots compare firing preferences on Spatial (black arrow) vs. Object (pink arrow) trials for the same cells. The same tuning vectors also show the magnitude of firing for preferred locations during the encoding (left panel) and selection (right panel) phases of the task on Spatial and Object trials. Spatial trials tuning vectors (black) show the same preferred directionality (i.e. 270°) during the encoding and selection phases in both PFC layers and in caudate nucleus, suggesting parallel processing streams/loops through cortical minicolumns and striatum and likely through the entire thalamo-cortical loop. But when processing object information directional preference changes in the three cells tuning plots, suggesting that object information processing does not follow in the same “foot prints” as processing by the same cells on Spatial trials. The radius of polar plots is represented in Hz and tuning amplitude is measured in Hz, as well. Asterisks: **p < 0.001, ANOVA.

Figure 3. Mean firing responses and population tuning of prefrontal cortical and striatal cells during Spatial and Object trials.

(A&B) Spatial trials. Comparison of mean firing rates of neurons during encoding (A) and selection (B) across prefrontal cortical layers (L23 and L5 and Striatum (Caudate nucleus) during “Spatial” trials. Prefrontal cortical L2/3 cells (n = 58) showed elevated firing during encoding and selection on spatial trials. Striatal (Caudate nucleus) cells (n = 52) showed a higher firing rates at the trial start when the spatial rule entered in effect. PFC layer 5 cells (n = 49) displayed moderate involvement in perception and selection. (C&D) Object trials. Comparison of mean firing rates of the same cells during encoding (C) and selection (D) is shown during Object trials. Cells in both prefrontal layers and striatum had much lower firing rates during Object (image) encoding and higher rates during the match, target selection, phase. The F values for (PFC layer 2/3, PFC layer L5, caudate) in (A) Sample-Spatial (F(1,1159) = 21.63, p < 0.001; F(1,979) = 6.73, p < 0.01; F(1,1039) = 7.32, p < 0.01), (B) Match-Spatial (F(1,1159) = 22.47; p < 0.001; F(1,979) = 15.56; p < 0.001; F(1,1039) = 9.13; p < 0.01), (C) Sample-Object (F(1,1159) = 1.46; p > 0.5; F(1,979) = 1.27; p > 0.5; F(1,1039) = 1.23; p > 0.5) and (D) Match-Object (F(1,1159) = 18.67; p < 0.001; F(1,979) = 16.51; p < 0.001; F(1,1039) = 14.31; p < 0.001). (E&F) Selection Phase. Comparison of neural tuning in prefrontal cortical layers and striatum during target selection on Spatial and Object trials. In (F) the arrangement of spatial locations/directions has been rotated so that the highest firing rates for all trials within the session correspond to 0° location/direction for every neuron. Error bars represent SEMs. Asterisks: **p < 0.001 ANOVA.

Figure 4. Distribution of preferred prefrontal-striatal cell firing at each target selection location.

(A–C). Polar plots showing the distribution of preferred firing directions for “Spatial” and “Object” trials in PFC layer 2/3 (A), layer 5 (B) and caudate nucleus (C) recoded simultaneously during the executive selection (match) phase of the DMS task. The average % of cell firing for each cell type tuning vector direction (in Figure 2) is represented by the corresponding target location in a circular histogram. The polar plot measures the percentage of cells with highest firing rates at those locations (tuning vectors) and the asterisks indicate the highest percentage of cells from the total population with firing rates at that particular location/direction. Asterisks: **p < 0.001, Rayleigh test.

Figure 5. Relations of preferred target selection location to stimulation induced enhancement of cognitive performance.

(A). Application of previously employed multi-input multi-output (MIMO) nonlinear model combined with the conformal MEA probe to extract the configuration of electrical (bipolar) stimulation pulses (50 uA and 1 ms) delivered to columnar recording locations in PFC layer 5. (B). Peri-event multigram of a PFC layer 2/3 cell with tuning plot showing preference for the 315° target location. (C). Distribution of MIMO-stimulation facilitated correct performance locations for spatial vs. object trials across multiple sessions. Tuning vectors of percentage of correct responses on Spatial and Object trials show improved performance by the MIMO stimulation was delivered on Spatial trials in which the target was in the same position as shown in Fig. 4, for the preferred firing location (315°) of PFC and caudate neurons on nonstimulation trials. (D). Comparison of the facilitation effect of MIMO stimulation (Stim) with control (no-stim) conditions on Spatial (n = 40 sessions) and Object (n = 50 sessions) locations with the highest performance levels on Stim trials with locations of the highest performance levels on no-stim trials (Facilitated: F(1,319) = 15.34, p < 0.001 on spatial and F(1,399) = 12.68, p < 0.001 on object). These selective changes in performance produced by MIMO stimulation are shown compared with overall changes across all types of trials (ALL: F(1,319) = 6.82, p < 0.01 on spatial and F(1,399) = 9.51, p < 0.01 on object trials) in the same sessions. The differences in tuning reflected as highest % correct performance indicate that MINO stimulation also enhanced the directional preference (spatial tuning bias around 315°) of the PFC layer 2/3 recorded minicolumn (Figure 4). Error bars represent SEMs. Asterisks: **p < 0.001, ANOVA.

Figure 6. Overlap of preferred firing and stimulation induced performance tuning during the perception-to-action cycle.

(A), (B). Polar plots showing the distribution of preferred firing direction for “Spatial” trials in PFC layer 2/3, layer 5 and caudate during the perception phase (A), and the executive selection phase (B) as shown in Figure 4A–C. (C). Distribution of facilitated correct performance for spatial selection during MIMO stimulation sessions (Figure 5D). The red dotted contour of tuned activity of the neurons from PFC layer 5 is overlaid to indicate similar preferred locations for columnar tuning and MIMO-stimulation facilitated performance. (D). Prefrontal-striatal correlation. Normalized cross-correlations (overlay) between n = 54 pairs of cells in PFC layer 5 and Caudate depict synchronized firing during Match target presentation (0, 2 s; red) compared to the pre-Match epoch (−2 s, 0; blue). There was a marked difference between CCHs in Match vs Pre-Match conditions; F(1,107) = 21.82, p < 0.001; ANOVA. (E). Functional diagram showing a representation of the flow of information in the PFC-caudate tuned spatial relationship across brain regions and behavior in the perception-to-action cycle. Same symbols apply as in Fig. 1 A. Asterisks: **p < 0.001, ANOVA.