Contributions of narrow- and broad-spiking prefrontal and parietal neurons on working memory tasks

Abstract

Neurons that generate persistent activity in the primate dorsolateral prefrontal and posterior parietal cortex have been shown to be predictive of behavior in working memory tasks, though subtle differences between them have been observed in how information is represented. The role of different neuron types in each of these areas has not been investigated at depth. We thus compared the activity of neurons classified as narrow-spiking, putative interneurons, and broad-spiking, putative pyramidal neurons, recorded from the dorsolateral prefrontal and posterior parietal cortex of male monkeys, to analyze their role in the maintenance of working memory. Our results demonstrate that narrow-spiking neurons are active during a range of tasks and generate persistent activity during the delay period over which stimuli need to be maintained in memory. Furthermore, the activity of narrow-spiking neurons was predictive of the subject's recall no less than that of broad-spiking neurons, which are exclusively projection neurons in the cortex. Our results show that putative interneurons play an active role during the maintenance of working memory and shed light onto the fundamental neural circuits that determine subjects' memories and judgments.

Keywords: monkey; neurophysiology; posterior parietal cortex; prefrontal cortex; working memory.

Figure 1

(A) Sequence of events in the MSNG task. The monkey is required to observe the cue and maintain fixation during the delay period. If a second stimulus appears at the same location as the cue (match), the monkey needs to stay at the fixation point after its color changes; if it deviates (nonmatch), the monkey is required to make an eye movement to the second stimulus once the color of the fixation point changes. (B) Sequence of events in the R1R2 task. In the remember-first task, the fixation point is white, and the monkey is required to make an eye movement to the first stimulus, regardless of the location of a second stimulus, which is a distractor. In the remember-second task, the first stimulus is now a distractor and the monkey is required to make an eye movement to the remembered location of the second stimulus.

Figure 2

Regions of neurophysiological recordings, comprising areas 8 and 46 in the dorsolateral prefrontal cortex (dlPFC) and areas 7a and LIP in the posterior parietal cortex (PPC). IPS, intraparietal sulcus; STS, superior temporal sulcus; AS, arcuate sulcus; PS, principal sulcus.

Figure 3

(A) Example waveform of a single Narrow Spiking unit. Points represent average normalized voltage value across all waveforms of this unit. Line represents fitted generalized additive model through experimental points. Open circle represents the estimated peak of the waveform. (B) Example waveform of a single Broad Spiking unit. (C) Distribution of spike widths among all neurons in the sample, across areas and tasks (sample size of 341 NS units and 2,120 BS units). (D) Firing rate, computed in the fixation interval of the task is plotted against the waveform duration of each unit.

Figure 4

(A) Population PSTH showing mean firing rate from narrow-spiking dlPFC neurons recorded in the MSNG task (n = 47). Inset, mean firing rate of neurons with significantly elevated responses (n = 12). Gray bars indicate time of appearance of the two stimuli in the task (0–0.5 s and 3.5–4 s). Time of −1 s represents the beginning of the fixation period. The solid traces indicate average activity of neurons when the stimulus appeared in location that elicited the best cue response (best cue), while the dotted lines indicate neuronal activity when the stimulus appeared at the most distant location from the preferred location (worst cue). Shaded areas indicate standard error of the mean (SEM). (B) As in A, for broad-spiking dlPFC neurons (n = 530). Inset, mean firing rate of neurons with significantly elevated responses (n = 132). (C) Averaged PSTH of neuronal spike discharges from narrow-spiking PPC neurons recorded in the MSNG task (n = 136). Inset, mean firing rate of neurons with significantly elevated responses (n = 28). (D) As in C, for broad-spiking PPC neurons (n = 723). Inset, mean firing rate of neurons with significantly elevated responses (n = 117).

Figure 5

(A) Population PSTH showing mean firing rate from narrow-spiking dlPFC neurons recorded in the R1R2 task (n = 66). Inset, mean firing rate of neurons with significantly elevated responses (n = 28). Gray bars indicate time of appearance of the two stimuli; right-most dotted line indicates offset of fixation point that signals the beginning of the saccade period (0–0.5 s, 2–2.5 s, and 4–4.5 s). Time − 1 s represents the beginning of the fixation period. Solid lines indicate neuronal activity for the best location (based on mean cue firing rate), while the dotted lines indicate activity when the cue appeared at the location diametric to the best location. Shaded areas indicate standard error of mean (SEM). (B) Same as in A, for broad-spiking dlPFC neurons (n = 357). Inset, mean firing rate of neurons with significantly elevated responses (n = 154). (C) Averaged PSTH of neuronal spike discharges from narrow-spiking PPC neurons recorded in the R1R2 task (n = 92). Inset, mean firing rate of neurons with significantly elevated responses (n = 29). (D) Same as in C, for broad-spiking PPC neurons (n = 510). Inset, mean firing rate of neurons with significantly elevated responses (n = 151).

Figure 6

(A) Averaged area under the ROC curve from NS (red line, n = 41) and BS (blue line, n = 526) neurons recorded from the dlPFC plotted as a function of time for the MSNG task. Each curve represents difference in firing rate distributions between the best and most distant location. Shaded area represents SEM. (B) As in A for NS (red line, n = 136) and BS (blue line, n = 717) neurons recorded from the PPC in the MSNG task. (C) Averaged area under the ROC curve from NS (red line, n = 66) and BS (blue line, n = 352) neurons recorded from the dlPFC in the R1R2 task. (D) As in C, for NS (red line, n = 92) and BS (blue line, n = 504) neurons recorded from the PPC in the R1R2 task.

Figure 7

(A) Averaged area under the ROC curve from NS (red line, n = 65) and BS (blue line, n = 463) neurons recorded from the dlPFC, pooled from both the MSNG and R1R2 tasks. The solid line represents ROC value (choice probability) comparing the distribution of correct and error trials from the preferred cue condition. The shaded area represents SEM. (B) Averaged choice probability value from NS (red line, n = 117) and BS (blue line, n = 685) neurons recorded from the PPC, pooled from both the MSNG and R1R2 tasks.