Prefrontal cortex and hippocampus subserve different components of working memory in rats

Abstract

Both the medial prefrontal cortex (mPFC) and hippocampus are implicated in working memory tasks in rodents. Specifically, it has been hypothesized that the mPFC is primarily engaged in the temporary storage and processing of information lasting from a subsecond to several seconds, while the hippocampal function becomes more critical as the working memory demand extends into longer temporal scales. Although these structures may be engaged in a temporally separable manner, the extent of their contributions in the "informational content" of working memory remains unclear. To investigate this issue, the mPFC and dorsal hippocampus (dHPC) were temporarily inactivated via targeted infusions of the GABA(A) receptor agonist muscimol in rats prior to their performance on a delayed alternation task (DAT), employing an automated figure-eight maze that required the animals to make alternating arm choice responses after 3-, 30-, and 60-sec delays for water reward. We report that inactivation of either the mPFC or dHPC significantly reduced DAT at all delay intervals tested. However, there were key qualitative differences in the behavioral effects. Specifically, mPFC inactivation selectively impaired working memory (i.e., arm choice accuracy) without altering reference memory (i.e., the maze task rule) and arm choice response latencies. In contrast, dHPC inactivation increased both reference memory errors and arm choice response latencies. Moreover, dHPC, but not mPFC, inactivation increased the incidence of successive working memory errors. These results suggest that while both the mPFC and hippocampus are necessarily involved in DAT, they seem to process different informational components associated with the memory task.

Figure 1.

Schema of the automated figure-eight maze. The four gates are raised and lowered (from the maze surface) and water is delivered (at three reward locations) under computer control.

Figure 2.

Illustration of a programmed DAT training procedure on the figure-eight maze. Sequence of gate operation was controlled by the computer based on the animal position on the maze during shaping (A) and testing (B).

Figure 3.

Histological reconstructions of cannula placement site. (Filled circles) Locations where infusion cannula tips were placed within the medial prefrontal cortex (mPFC: prelimbic and infralimbic) (A) and the dorsal hippocampus (dHPC) (B). (Open circles) incorrectly placed cannula tip locations. Numbers indicate the distance in millimeters relative to bregma. Reprinted with permission from Elsevier ©1997, Paxinos and Watson 1997.

Figure 4.

Effects of muscimol infusions on DAT performance. Mean (±SE) percentage of correct trials at 3-, 30-, and 60-sec delay periods when ACSF (black) and MUSC (gray) were microinfused into the mPFC (A) and the dHPC (B). (Horizontal line) Chance level of performance.

Figure 5.

Arm choice reaction time (CRT) as a function of training. (A) CRT was quantified as the latency from the start of a trial (front and side gates down) to when the animal made an arm choice response (side gates up). (B) Mean (±SE) CRT during the first testing session (black) and the criterion testing session (white). (C) Sample movement trajectory maps show decrease in choice response variability (dash box) during the first testing day vs. criterion day.

Figure 6.

Effects of muscimol infusions on CRT. (A) Mean (±SE) CRT for correct (black) and incorrect (gray) trials during testing sessions with no drug infusion (Day 1) at various delays. (B) Mean (±SE) CRT for correct (black) and incorrect (gray) trials in MUSC-mPFC animals. (C) Mean (±SE) CRT for correct (black) and incorrect (gray) trials in MUSC-dHPC animals. (D) Mean (±SE) CRT (correct and incorrect trials combined) with no drug infusion (black), MUSC-mPFC (light gray), and MUSC-dHPC (dark gray) conditions at various delays.

Figure 7.

Effects of muscimol infusions on “back edge errors.” (A) Back edge error was made when the animal entered the opposite edge runway rather than returning to the center holding area. (B) Mean (±SE) percentage of back edge errors made by ACSF(mPFC+dHPC), MUSC-mPFC, and MUSC-dHPC conditions.

Figure 8.

Effects of muscimol infusions on two consecutive incorrect choices. Mean (±SE) percent of incorrect trials immediately following incorrect trials across delays when muscimol was infused into the mPFC (black) and dHPC (white).

Figure 9.

A hypothetical model of neural structures involved in working memory (WM). The model posits that the hippocampus (HPC) is involved in processing spatial memory (SM) and episodic memory (EM), which are used by the PFC in spatial working memory tasks. Accordingly, inactivation of the HPC will lead to WM, SM, and EM deficits, while inactivation of the PFC will produce relatively specific WM impairments.