Choosing memory retrieval strategies: A critical role for inhibition in the dentate gyrus

Abstract

Remembering the location of food is essential for survival. Rodents and humans employ mainly hippocampus-dependent spatial strategies, but when being stressed they shift to striatum-mediated stimulus-based strategies. To investigate underlying brain circuits, we tested mice with a heightened stress susceptibility due to a lack of the GABA-synthetizing enzyme GAD65 (GAD65-/- mice) in a dual solution task. Here, GAD65-/- mice preferred to locate a food reward in an open field via a proximal cue, while their wildtype littermates preferred a spatial strategy. The analysis of cFos co-activation across brain regions and of stress-induced mRNA expression changes of GAD65 pointed towards the hippocampal dorsal dentate gyrus (dDG) as a central structure for mediating stress effects on strategy choices via GAD65. Reducing the GAD65 expression locally in the dDG by a shRNA mediated knock down was sufficient to replicate the phenotype of the global GAD65 knock out and to increase dDG excitability. Using DREADD vectors to specifically interfere with dDG circuit activity during dual solution retrieval but not learning confirmed that the dDG modulates strategy choices and that a balanced excitability of this structure is necessary to establish spatial strategy preference. These data highlight the dDG as a critical hub for choosing between spatial and non-spatial foraging strategies.

Keywords: Cognitive flexibility; Dentate gyrus; GAD65; Inhibition; Spatial learning; Stimulus-based learning.

Graphical abstract

Fig. 1

Strategy preference in GAD65−/− mice. (A) GAD65−/− mice showed elevated corticosterone serum levels already under baseline control conditions that were not further elevated 24 h after a brief restraint stress episode, suggesting altered stress susceptibility. (B) In the dual solution task (scheme on top), GAD65−/− mice showed a good learning performance of the reward location over 6 trials. When tested for their strategy preference 24 h later, they did not display a spatial preference in comparison to their wildtype littermates. (C) When the experimental setup only allowed for a spatial solution in order to learn the reward location, GAD65−/− mice showed a spatial preference for the reward location during retrieval. (D) Similarly, when a cue was present during training but not during retrieval, GAD65−/− mice still displayed a spatial preference similar to their wildtype littermates. All values means ± sem. * significant difference between genotypes, p < 0.05; ***p < 0.001; +++ significant learning effect over trials, p < 0.001; $$ significant difference of target position compared to other positions, p < 0.01; $$$ p < 0.001.

Fig. 2

Interregional activation patterns during dual solution retrieval and GAD65 expression regulation after stress point towards the dDG as a hub for stress-related effects. (A) GAD65−/− mice and their wildtype littermates were trained in the dual solution task and perfused 1 h after retrieval brains for subsequent cFos analysis. Example images are shown for cFos labeling (green: cFos; cyan: DAPI nuclear staining; white arrow head: exemplary cFos-positive cell; scale bar = 100 μm) in the dorsal dentate gyrus. No difference in the number of cFos positive cells was evident in any of the regions analyzed between GAD65 +/+ mice preferring a spatial solution (n = 5) and GAD65−/− mice using cue-based strategy (n = 5). (B) Interregional correlations of cFos activation after DS retrieval and strategy choice demonstrated shifts in co-activation of the dDG together with PL/IL, and CeA in between the groups. (C) The dDG was also susceptible to gene expression changes induced by restrained stress. In (i) GAD65+/+ mice GAD65 mRNA levels were downregulated and mineralocorticoid receptor (MR) upregulated 24 h after brief restrained stress, while (ii) MR expression was unaltered in GAD65−/− mice (n = 5–7 per group). (D) In comparison, no stress-induced expression changes were observed in the striatum of (i) GAD65+/+ and (ii) GAD65−/−mice (n = 5–7 per group). Expression of the isoform GAD67 and of the glucocorticoid receptor (GR) remained unchanged in both regions. All values means ± sem. + significant Pearson's correlation, p < 0.05; ++ p < 0.01; * significant difference between genotypes, p < 0.05. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 3

Local knock down of GAD65 in the dDG leads to loss of spatial preference in a dual solution task. (A) Vector scheme and workflow for local knock down of GAD65 in the dDG before dual solution training and retrieval. (B) The local knock down of GAD65 in the dDG is sufficient to replicate the phenotype of total GAD65−/− mice. (C) Electrophysiological profile of the dDG after local GAD65 knock down (CTR: n = 18 slices from 9 animals; GAD65 KD: n = 12 slices from 8 animals). (i) Scheme for perforant-path induced somatic field postsynaptic potential (fEPSP) and population spike (PS) responses. Input-output (I/O) curves demonstrate an enhanced excitability after GAD65 KD in the dDG evident by (ii) increased fEPSP slope and (iii) PS amplitudes. (iv) Accordingly, the maximum evoked fEPSP slope and (v) PS amplitude is increased in GAD65 KD slices. No differences in the paired pulse ratios were observed for (vi) the fEPSP slopes or (vii) the PS amplitudes. All values means ± sem. +++ significant learning effect over trials, p < 0.001; * significant difference from CTR group, p < 0.05, **p < 0.01, ***p < 0.001. &&& p < 0.001, significant stimulus intensity effect.

Fig. 4

Modulation of dDG excitatory activity during retrieval affects strategy choices. (A) Scheme for chemogenetic activation/inhibition of somatostatin (SST)- and parvalbumin (PV)-positive interneurons during dual solution retrieval. (B, upper panel) Activation or inhibition of dDG SST(+) cells during dual solution retrieval did not affect strategy choice (CTR: n = 9; 3Dq: n = 8; 4Di: n = 8). (B, lower panel) Activation or inhibition of PV(+) interneurons during dual solution retrieval did not affect strategy choice significantly, but spatial strategy preference appeared diminished in the PV-hM3Dq group (not significant; CTR: n = 9; 3Dq: n = 10; 4Di: n = 10). (C) For comparison, chemogenetic inhibtion of excitatory dDG cells was achieved using CamKII-hM4Di constructs. Expression of the marker mCherry is observed in the molecular layer and mossy fibers, containing dendritic and axonal compartments of dDG granule cells. (D) Inhibiting excitatory cell activity during retrieval induced a strategy shift with a loss of spatial preference (n = 14; CTR: n = 13). All groups did not differ in their training performance. All values means ± sem. +++ significant learning effect over trials, p < 0.001; * significant difference to CTR group, p < 0.05.