Epac signaling is required for hippocampus-dependent memory retrieval

Abstract

Previously we uncovered a critical role for norepinephrine and β₁-adrenergic signaling in hippocampus-dependent memory retrieval. Because the β₁ receptor couples to G_s, we examine here whether cAMP is also required for contextual memory retrieval. Using pharmacologic and genetic approaches to manipulate cAMP and downstream signaling, we demonstrate that cAMP and two of its targets, protein kinase A (PKA) and exchange protein activated by cAMP (Epac), are both required for retrieval. These findings demonstrate that cAMP signaling through Epac (as well as PKA) plays an essential role in cognition.

The second messenger cyclic AMP was hypothesized to contribute to cognition, and in particular to learning and memory, nearly 4 decades ago (1). Evidence implicating cAMP in long-lasting synaptic changes and in memory processes was reported shortly thereafter (2, 3). Genetic support for the role of cAMP in learning and memory initially came from the Drosophila learning and memory mutants dunce and rutabaga, which harbor mutations in a cAMP phosphodiesterase (PDE) and a calcium-sensitive adenylyl cyclase (AC), respectively (4–7). Analogously, mutant mice lacking calcium-sensitive AC types I and VIII exhibit impairments in long-term memory caused by a deficit in memory consolidation (8). Further, agents that acutely elevate cAMP often enhance memory consolidation, whereas agents that block cAMP signaling usually impair memory consolidation (9–13).

The best characterized target of cAMP in nonsensory neurons is cAMP-dependent protein kinase A (PKA). Phosphorylation of proteins by PKA provides mechanisms that underlie memory formation mediated by cAMP. For example, short-term memory requires PKA phosphorylation of synaptic proteins such as ion channels, long-term memory requires PKA regulation of gene transcription and translation, and both types of memory may involve presynaptic enhancement of neurotransmitter release mediated by PKA (14, 15). In support of a role for PKA in learning and memory, Drosophila DC0 lines, in which the catalytic subunit of PKA is mutated, exhibit learning and memory impairments (16), and mice expressing a regulatory subunit of PKA that is insensitive to cAMP in forebrain excitatory neurons have deficits in long-term memory (17). Finally, pharmacologic inhibition of PKA suggests that PKA is required for the consolidation of long-term memory (10, 18).

The genetic approaches described above do not distinguish between roles for cAMP and PKA in memory consolidation and retrieval, however. Interestingly, there are only a few studies that address whether signaling by cAMP is required for retrieval, and their results are disparate. For example, when the PDE-resistant cAMP antagonist cyclic adenosine monophosphorothioate (Rp-cAMPS) was infused into the lateral ventricle of mice 30 min before testing contextual fear 1 day after training, no effect on freezing was observed (11). In contrast, bilateral infusion of Rp-cAMPS into the dorsal hippocampus (DH) of rats 10 min before testing inhibitory avoidance 1 day after training reduces step-down latency (19). The opposite results of these two studies could be either because retrieval in the two tasks is mediated by different mechanisms or because technical differences are important.

Elevations in cAMP occur primarily through stimulation of plasma membrane receptors that couple to G_s proteins. These receptors are activated by a diverse set of ligands, including neuromodulators such as norepinephrine (NE), dopamine, and serotonin. While investigating cognitive roles for adrenergic signaling, we uncovered a requirement for NE in the retrieval of hippocampus-dependent memory (20). Interestingly, signaling by NE through the β₁-adrenergic receptor is required for an intermediate term of memory retrieval (from 2 h to ≈4 days after contextual fear conditioning). Surprisingly, NE and epinephrine (E) are not required for consolidation of fear and several other types of memories, as assessed by the same genetic and pharmacologic approaches used to identify the role for NE in retrieval. Those findings provide an opportunity to specifically investigate the molecular mechanisms underlying memory retrieval, for which much less is known relative to those underlying memory consolidation.

Because β₁ receptors couple to G_s, we hypothesized that cAMP signaling would be required for hippocampus-dependent memory retrieval. To test this hypothesis, we infused inhibitory analogs of cAMP into the DH shortly before testing contextual fear memory and found that they impair retrieval. In parallel, we infused activating analogs of cAMP into the DH of NE/E-deficient dopamine β-hydroxylase knockout (Dbh^−/−) mice that are impaired in contextual fear memory retrieval and found that they rescue retrieval. However, when a cAMP analog that selectively activates PKA was infused, no rescue was observed, suggesting that other targets of cAMP are important. A second mediator of signaling by cAMP is the exchange protein activated by cAMP (Epac) 1 and 2, also known as cAMP-guanine nucleotide exchange factor (GEF) I and cAMP-GEFII, which are GEFs for Rap (21, 22). Little is known about the roles of cAMP signaling through Epac in behavior. A role for Epac in mechanical hyperalgesia in the spinal cord has been proposed (23); however, a role for Epac in cognition is lacking, despite prominent expression in the forebrain, especially for Epac2 (22). Here, we show that when a cAMP analog that selectively activates Epac is infused into the DH of Dbh^−/− mice, retrieval is not rescued. However, when low doses of the Epac-selective and PKA-selective activator are infused together, retrieval is rescued. Our results implicate Epac and PKA signaling as being required for DH-dependent memory retrieval.

Results

Role for cAMP in Contextual Fear Memory Retrieval.

Fear conditioning was used as a robust means for assessing contextual memory retrieval. Mice were habituated to handling for 2 days and fear-conditioned the next day (day 0) in a novel context, where a 30-s tone was activated 2 min after introduction of the mouse. The tone coterminated with a 2-s footshock (0.4–1 mA). Mice were tested for fear memory on 1 or 2 subsequent days by reexposure either to the training context or to the training tone in a novel context. Because infusion of the β-adrenergic receptor agonist isoproterenol into the DH rescues contextual fear memory retrieval in Dbh^−/− mice (20, 24), we examined whether bilateral infusion of cAMP analogs into the DH shortly before testing would affect retrieval. Dbh^+/− mice were used in place of wild-type mice in most experiments because the former serve as littermate controls for Dbh^−/− mice and because they have normal levels of NE/E and are phenotypically indistinguishable from Dbh^+/+ mice (25).

As a PDE-resistant cAMP analog that can block activation of PKA and ion channels by cAMP, Rp-cAMPS was tested first because a prior study using it failed to find evidence for a role of cAMP in contextual fear memory retrieval (11). We found that at 5 μg per side, Rp-cAMPS moderately reduced freezing to the training context when infused 30 min before testing 1 day after training with a 1-mA shock (Fig. 1A). A potential difficulty with the use of Rp-cAMPS is that it is not highly membrane-permeable. If Rp-cAMPS acts intracellularly to impair freezing, then analogs with greater membrane permeability should be more potent. Therefore, a more lipophilic Rp-cAMPS prodrug (Rp-2′-O-monobutyryl-cAMPS or Rp-2′-O-MB-cAMPS) that is deesterified intracellularly to generate Rp-cAMPS was tested. Rp-2′-O-MB-cAMPS significantly reduced freezing at 2 and 5 μg per side, with the latter reducing freezing to levels observed in Dbh^−/− mice (Fig. 2). In addition to Rp-cAMPS, esterase cleave of Rp-2′-O-MB-cAMPS generates butyrate. To control for potential effects of butyrate, a second lipophilic cAMP derivative, Rp-8-Br-cAMPS, was tested and gave results nearly identical to those for Rp-2′-O-MB-cAMPS (Fig. 1A).

Fig. 1.

cAMP signaling in the DH is required for intermediate-term contextual fear memory retrieval. Dbh^+/− mice, which possess normal tissue levels of NE/E, were used to examine the requirement for cAMP signaling in retrieval. DH infusions were made 1–5 days after training through chronically implanted cannulae 30 min before testing, with isotonic saline (Sal) as vehicle. Freezing was quantitated as an indicator of fear memory. (A) The PDE-resistant cAMP analog Rp-cAMPS, Rp-2′-O-monobutyryl-cAMPS (Rp-2′-O-MB-cAMPS or Rp-MB, a prodrug for Rp-cAMPS), or Rp-8-Br-cAMPS was infused before testing contextual fear memory. The latter two analogs significantly reduced freezing (n = 7, 5, 5, 8, 6, 6, 7 mice for each group, left to right). (B) Rp-MB (5 μg) was infused before testing cued fear memory and was without effect (n = 7 per group). (C) Rp-MB (5 μg) was infused before testing contextual fear memory 1 day after training (D1). The same mice were tested on day 2 without infusion. Rp-MB produced a transient impairment in freezing on day 1, consistent with a block of retrieval (n = 5 for Sal, 6 for Rp-MB). (D) Rp-MB (5 μg) was infused before testing contextual fear memory in separate groups of mice on day 1–day 5 (n = 6, 8, 5, 7). Significant reductions in freezing occurred on day 1–day 3, whereas day 5 was unaffected, relative to Sal on day 1 (n = 5). Sal on day 1 was used as reference because previous studies indicated that freezing is stable across days for Sal-infused mice (20, 24). Values are mean ± SEM; *, P < 0.05; #, P < 0.001.

Fig. 2.

cAMP signaling in the DH is sufficient to rescue retrieval in NE/E-deficient Dbh^−/− mice. The protocol used is that described for Fig. 1, except Dbh^−/− mice were used. These mice exhibit impaired intermediate-term (day 0–day 4 after training) contextual fear memory retrieval (20). (A) The PDE-resistant cAMP analog Sp-2′-O-monobutyryl-cAMPS (a prodrug for Sp-cAMPS) or Sp-8-Br-cAMPS was infused before testing contextual fear memory. The analogs significantly enhanced freezing in the Dbh^−/− mice (n = 6, 5, 6, 5, 4). (B) Sp-MB (2 μg) was infused before testing cued fear memory and was without effect. To generate cued freezing comparable with that observed for contextual freezing in Dbh^−/− mice, a 0.4-mA (instead of 1-mA) shock was used during training (n = 4 per group). (C) Sp-MB (2 μg) was infused before testing contextual fear memory on day 1. The same mice were tested on day 2 without infusion. Sp-MB produced a transient enhancement in freezing on day 1 (n = 5 per group). Values are mean ± SEM; *, P < 0.05; ⋀, P < 0.01.

Reductions in freezing could be caused by alterations in sensorimotor function, fear, or memory. To distinguish among these possibilities, additional mice were tested for their freezing response to the training tone 1 day after conditioning and 30 min after DH drug infusion. This test was performed because acquisition and expression of cued fear do not depend on the DH, NE/E, or β₁ signaling (20, 26, 27). DH infusion of Rp-2′-O-MB-cAMPS had no effect on freezing to the tone (Fig. 1B), ruling out impaired motor function or fear. Further, the expression of contextual fear does not depend on NE/E or β₁ signaling beginning 5 days after training (24). When Rp-2′-O-MB-cAMPS was infused into the DH 5 days after training, it had no effect on freezing, ruling out impairments in the sensory processing of context. Reduced freezing with Rp-2′-O-MB-cAMPS on days 1–3 mirrors the time course over which NE and β₁ signaling is required for retrieval. Taken together, the results indicate that inhibitory cAMP analogs affect memory for the context when given 1 day after training. To test the possibility that the cAMP analogs affect a late phase of memory consolidation rather than memory retrieval, additional mice were tested 1 day after training, 30 min after DH infusion of Rp-2′-O-MB-cAMPS, and tested again 1 day later without infusion. Freezing was low on day 1 and high on day 2 (Fig. 1C), consistent with a transient impairment in retrieval rather than a lasting effect on consolidation.

Having demonstrated a requirement for DH cAMP signaling in contextual fear memory retrieval, we asked whether stimulating DH cAMP signaling would be sufficient to rescue retrieval in NE/E-deficient mice. Sp-2′-O-MB-cAMPS or Sp-8-Br-cAMPS was infused into the DH of Dbh^−/− mice 30 min before testing retrieval 1 day after training. Both analogs gave nearly full rescue of freezing (Fig. 2A; compare with Dbh^+/− mice in Fig. 1A). To distinguish between effects on memory and nonspecific effects on fear or freezing, Dbh^−/− mice were conditioned by using a 0.4-mA shock, which results in cued freezing 1 day after training that is comparable with contextual freezing after using 1 mA during training (Fig. 2 A and B). DH infusion of Sp-2′-O-MB-cAMPS had no effect on freezing to the cue (Fig. 2B), indicating that its effects are specific for memory. To distinguish between effects on consolidation and retrieval, contextual fear testing on days 1 and 2 was performed, with DH infusion of Sp-2′-O-MB-cAMPS given before testing on day 1. Freezing was high on day 1 and low on day 2 (Fig. 2C), indicating that retrieval rather than consolidation was rescued.

cAMP Signaling via both Epac and PKA Is Required for Retrieval.

Because stimulation of cAMP signaling is sufficient to rescue memory retrieval in NE/E-deficient mice, it was of interest to determine which targets of cAMP (PKA, Epac, and/or ion channels) mediate this effect. Importantly, cAMP analogs highly selective for PKA over Epac and Epac over PKA have been developed (28, 29). Sp-6-Phe-cAMPS and Sp-8-pCPT-2′-O-Me-cAMPS were chosen to selectively stimulate PKA and Epac, respectively. When Sp-6-Phe-cAMPS was infused into the DH 30 min before testing retrieval 1 day after training, no significant rescue of retrieval was observed (Fig. 3A), even though the same dose range (0.5–2 μg) over which the nonselective agonist Sp-8-Br-cAMPS rescues retrieval (Fig. 2A) was used. When Sp-8-pCPT-2′-O-Me-cAMPS was infused into the DH 30 min before testing retrieval 1 day after training, no significant rescue of retrieval was observed except at the highest dose (2 μg) (Fig. 3A). However, when the two analogs were infused together, nearly complete rescue and complete rescue was observed by using 0.2 μg and 0.5 μg of each, respectively, whereas 0.5 μg of each analog alone had no effect (Fig. 3A).

Fig. 3.

Combined treatment with PKA- and Epac-selective agonists in the DH is required to rescue retrieval in NE/E-deficient Dbh^−/− mice. The protocol used is that described for Fig. 1, except Dbh^−/− mice were used. (A) The PDE-resistant PKA-selective agonist Sp-6-Phe-cAMPS (Sp-“PKA”), the PDE-resistant Epac-selective agonist Sp-8-pCPT-2′-O-Me-cAMPS (Sp-“Epac”), or the combination was infused before testing contextual fear. The combination significantly enhanced freezing in the Dbh^−/− mice at doses that had no effect on their own (n = 6, 7, 4, 8, 5, 5, 4, 6, 6, 7, 7). The highest dose (2 μg) of Sp-Epac also significantly enhanced freezing; however, this effect was blocked by the PKA-selective antagonist PKI* (see legend for B). (B) The Epac-selective agonist 8-pCPT-2′-O-Me-cAMP (Epac) was used to replace Sp-Epac from A. Sp-PKA + Epac (0.5 μg each) significantly enhanced freezing in the Dbh^−/− mice, whereas 2 μg of Epac alone had no effect (n = 5, 4, 5). (C) To better detect potential enhancements or reductions in retrieval, a 0.75-mA shock was used during conditioning. The cAMP analogs Sp-PKA, Epac, or the combination (all at 0.5 μg) were infused into the DH of wild-type mice 30 min before testing retrieval. Relative to saline (Sal), none of the infusions had a significant effect on retrieval (n = 5 per group), although the individual agonists tended to enhance freezing modestly. Values are mean ± SEM; *, P < 0.05; ⋀, P < 0.01; #, P < 0.001

The data suggest that stimulation of PKA and Epac together is required to rescue memory retrieval in NE/E-deficient mice, with the caveat that higher doses of the Epac-selective agonist also partially rescue. However, the sulfur-containing Epac-selective agonist may inhibit some PDEs and could lead to elevation of endogenous cAMP, stimulating PKA in the absence of a PKA-selective agonist (30, 31). Indeed, when 8-pCPT-2′-O-Me-cAMP was infused in place of Sp-8-pCPT-2′-O-Me-cAMPS, it did not rescue retrieval at 2 μg alone but fully rescued retrieval at 0.5 μg when combined with Sp-6-Phe-cAMPS (Fig. 3B). Consistent with the above, block of PKA signaling with myr-PKI[14–22]amide (see below) prevented the partial rescue observed with 2 μg of Sp-8-pCPT-2′-O-Me-cAMPS (Fig. 3A), suggesting that this analog indirectly activates PKA at this dose.

cAMP analogs could appear to rescue retrieval in at least two distinct ways. First, they could be stimulating the signaling pathways that fail to get activated during retrieval in the absence of NE/E, i.e., the role of NE in retrieval depends on cAMP. Alternatively, the role of NE in retrieval could be independent of cAMP. In this case, cAMP agonists should have strong effects on memory retrieval when NE/E are present, and these effects would then also be apparent when NE/E are absent. To distinguish between these possibilities, cAMP agonists were infused into the DH of wild-type mice 30 min before testing retrieval 1 day after training. No significant effects on retrieval were observed when the agonists were infused separately or in combination (Fig. 3C). The data suggest that endogenous signaling by Epac and PKA during retrieval is lacking specifically in the NE/E-deficient mice and is sufficient for nearly full memory retrieval in wild-type mice.

The data above strongly suggest that PKA signaling is required for retrieval, but the specificity of Sp-6-Phe-cAMPS for activation of PKA versus activation of cAMP-potentiated ion channels is unknown. To test the specific requirement for PKA in retrieval, two additional, non-cAMP-based inhibitors of PKA were used: 4-cyano-3-methylisoquinoline (CMIQ) and a peptide fragment [14–22] from the highly selective PKA inhibitor protein (PKI), myristoylated to promote membrane permeability (myr-PKI[14–22]amide). When infused into the DH of Dbh^+/− mice 1 day after training, 30 min before testing, each PKA inhibitor dose-dependently impaired freezing (Fig. 4 A and B). The effect of myr-PKI[14–22]amide was specific for memory because it had no effect on cued fear 1 day after training or on contextual fear 5 days after training (Fig. 4 C and E) and was specific for retrieval because memory was normal when mice were tested again 1 day after infusion (Fig. 4D). The effects of myr-PKI[14–22]amide on days 1–4 after training indicate that PKA, in addition to cAMP (Fig. 1D) and NE (20, 24), contributes to the retrieval of intermediate-term contextual memories.

Fig. 4.

PKA signaling in the DH is required for intermediate-term contextual fear memory retrieval. The protocol used is that described for Fig. 1, except that CMIQ infusions were performed 15 min before testing, and 10% dimethylformamide in saline (Sal) was used as vehicle for CMIQ. (A) As an antagonist of the ATP-binding site of PKA, CMIQ was infused before testing contextual fear. CMIQ significantly reduced freezing (n = 5 per group). (B) A myristoylated, amidated peptide fragment [14–22] from the PKI protein (myr-PKI[14–22]amide or PKI*) was infused before testing contextual fear. PKI* significantly reduced freezing at 2 and 5 μg (n = 7, 4, 5, 5, 5). (C) PKI* (2 μg) was infused before testing cued fear and was without effect (n = 7, 5). (D) PKI* (2 μg) was infused before testing contextual fear on day 1. The same mice were tested on day 2 without infusion. PKI* produced a transient impairment in freezing on day 1 (n = 5 per group). (E) PKI* (2 μg) was infused before testing contextual fear in separate groups of mice on day 1–day 5 (n = 5, 8, 8, 5, 5). Significant reductions in freezing occurred on day 1–day 3, a nonsignificant reduction in freezing occurred on day 4, whereas day 5 was unaffected, relative to Sal on day 1 (n = 5). Values are mean ± SEM; *, P < 0.05; ⋀, P < 0.01; #, P < 0.001.

Discussion

Because NE via the β₁ receptor, which couples to G_s, plays an important role in hippocampus-dependent memory retrieval, we hypothesized that cAMP and PKA would have similar roles. Support for such roles was obtained in this work when DH infusion of cAMP or PKA antagonists was found to impair retrieval. Those results appear to conflict with the lack of an effect on retrieval when 45 μg of Rp-cAMPS was infused intracerebroventricularly (i.c.v.) 30 min before testing contextual fear (11). However, our dose–response data for Rp-cAMPS (a partial effect for 5 μg infused into each DH) suggest that 45 μg i.c.v. is too low to antagonize retrieval mediated by the DH.

Interestingly, DH infusion of a PKA-selective agonist failed to rescue retrieval in the Dbh^−/− mice, even though a β-selective agonist rescues (20). Therefore, as a second mediator of signaling by cAMP, a role for Epac was examined. We found that signaling via both Epac and PKA is required to rescue retrieval in Dbh^−/− mice. By using both the PDE-resistant and PDE-sensitive Epac agonists, we were able to rule out two alternative explanations for our results. First, rescue using the PDE-resistant Epac agonist ruled out PDE metabolites of the PDE-sensitive Epac agonist as mediators (30). Second, rescue using the PDE-sensitive Epac agonist ruled out inhibition of PDEs and elevation of endogenous cAMP as the mediator (31). Our data are consistent with the inhibition of PDE at the highest dose of the PDE-resistant Epac agonist used here.

A third class of cAMP effectors is the hyperpolarization- and cyclic nucleotide-activated (HCN) and the cyclic nucleotide-gated (CNG) ion channels. It is formally possible that HCN and/or CNG channels also contribute to retrieval. However, it is highly unlikely that they are activated by the Epac-selective agonists because the cyclic nucleotide-binding domains of HCN and CNG channels have a glutamate (as does PKA but not Epac) that is thought to be essential for high-affinity binding of cAMP via its 2′-OH group (32). Because of methylation at this site, it is predicted that Epac-selective agonists would be much less potent at HCN and CNG channels and at PKA, and this has been confirmed for the latter two (31, 33). In addition, forebrain excitatory neuron-specific loss of HCN1 in mice actually enhances memory, most likely because HCN1 acts to limit dendritic integration of synaptic inputs in CA1 (34). Thus, potentiation of HCN1 by cAMP might impair rather than facilitate retrieval. Finally, despite their existence for at least a decade, no cognitive phenotype has been described for mice genetically lacking CNG channels, including those expressed in the hippocampus (35–38).

Similar to the time course over which NE and β₁ signaling is required for memory retrieval (20), cAMP and PKA signaling is required for ≈4 days after conditioning. This observation is consistent with at least two possibilities. First, retrieval may become independent of the hippocampus after this time because of systems-level consolidation (39, 40). Alternatively, molecular and cellular consolidation events occurring within the hippocampus 3–5 days after conditioning may make retrieval independent of NE, β₁, cAMP, and PKA. The latter is consistent with data from hippocampal lesion studies (26, 41). Additional experiments will be required to distinguish these possibilities.

Overall, our results are consistent with the prominent expression of Epac in the brain and the importance of cAMP in learning and memory and demonstrate that Epac is critically involved in cognition. The results suggest that a role for Epac must be considered whenever cAMP is involved, even if a role for PKA is demonstrated. It is likely that Epac will ultimately have a multitude of roles in the CNS and in cognition. Because Epac is a GEF for Rap, we hypothesize that signaling via Rap is also required for retrieval. Indeed, expression of a dominant-negative Rap construct in the DH interferes with retrieval in a manner identical to antagonists of β₁ receptors, cAMP and PKA (M.O., L.Z., J.J.Z., and S.A.T., unpublished data). Epac signaling may also participate in some of the identified roles for Rap, including glutamate receptor trafficking, membrane excitability, and synaptic plasticity (42–44).

Methods

Animals and Behavior.

Studies were in accordance with National Institutes of Health guidelines and had the approval of IACUC at the University of Pennsylvania. Dbh^−/− mice were rescued prenatally and housed as described in ref. 45. Fear conditioning was performed in the training apparatus (ENV-010MC; Med Associates). Mice were handled for 3 min per day on the 2 days preceding conditioning. Training consisted of placing the animal in the apparatus for 2 min, after which an 84-dB, 4.5-kHz tone was activated for 30 s. Two seconds before the end of the tone, a 2-s, 1-mA footshock was delivered, and the mouse was returned to its home cage 30 s after the shock. Where indicated, 0.4 or 0.75 mA was used for conditioning instead of 1 mA. For testing cued fear, a novel Plexiglas cylinder (21-cm diameter, 24 cm tall) with green wire grid floor and vertical green and white stripes 240° around was used. Each context was cleaned with a distinctly scented solution and was uniquely situated in the room. Cued fear was tested after 2 min by activating the training tone for 3 min. Percentage freezing was estimated by scoring the presence or absence of nonrespiratory movement every 5 s. Scoring was performed blind to experiment and treatment. Because Dbh^−/− mice exhibit ptosis, scoring was not blind to genotype, although no effort was made to identify genotype.

Drugs and Delivery.

DH drug infusions were through chronically implanted cannulas with isotonic saline (with 10% dimethylformamide added for CMIQ) as vehicle. A double-guide cannula (C235 system; Plastics One) was implanted under pentobarbitol anesthesia (≈70 mg/kg i.p.) by using a stereotax (SAS75/EM40M; Cartesian Research), being placed −1.7 mm AP and 1.5 mm bilateral. The guide extended 1.5 mm from the base, and the dummy extended 0.5 mm below the guide. The dual-injection cannula extended 0.9 mm below the guide. One week after recovery, bilateral infusions were made into conscious mice while gently holding the nape of the neck. Drugs were delivered in 0.5–1 μl per side at 0.4 μl/min. The injection cannula was left in place for 30 s before the mouse was returned to its home cage. All cannulations targeted the DH, as confirmed by dye infusion after the experiment. The coordinates and infusion volumes used were nonselective with respect to hippocampal subfield. Occasionally the guide cannula came off the skull before or during infusion, or one side of the injection cannula was found to be blocked immediately after infusion. Those subjects were not included in the data. All cAMP analogs were from BIOLOG. CMIQ and myr-PKI[14–22]amide were from EMD Chemicals.

Data Analysis.

All groups consisted of four to eight mice. Data were analyzed with Statistica 6.0 (StatSoft) with freezing as the dependent variable using ANOVA with infusate and day (when appropriate) as factors. Post hoc comparisons were made by using Duncan's range test or Student's t test, the latter being adjusted for multiple comparisons when appropriate by using Bonferroni's method. Data are reported as mean ± SEM.