Calcium-Stimulated Adenylyl Cyclase Activity Is Critical for Hippocampus-Dependent Long-Term Memory and Late Phase LTP

Neuron, Vol. 23, 787–798, August, 1999, Copyright 1999 by Cell Press

Calcium-Stimulated Adenylyl Cyclase Activity Is Critical for Hippocampus-Dependent Long-Term Memory and Late Phase LTP Scott T. Wong,*§ Jaime Athos,*§ Xavier A. Figueroa,* Victor V. Pineda,* Michele L. Schaefer,† Charles C. Chavkin,* Louis J. Muglia,† and Daniel R. Storm*‡ * Department of Pharmacology University of Washington School of Medicine Seattle, Washington 98195 † Departments of Pediatrics and Molecular Biology and Pharmacology Washington University School of Medicine St. Louis, Missouri 63130

Summary It is hypothesized that Ca21 stimulation of calmodulin (CaM)–activated adenylyl cyclases (AC1 or AC8) generates cAMP signals critical for late phase LTP (L-LTP) and long-term memory (LTM). However, mice lacking either AC1 or AC8 exhibit normal L-LTP and LTM. Here, we report that mice lacking both enzymes (DKO) do not exhibit L-LTP or LTM. To determine if these defects are due to a loss of cAMP increases in the hippocampus, DKO mice were unilaterally cannulated to deliver forskolin. Administration of forskolin to area CA1 before training restored normal LTM. We conclude that Ca21-stimulated adenylyl cyclase activity is essential for L-LTP and LTM and that AC1 or AC8 can produce the necessary cAMP signal. Introduction The central nervous system has the extraordinary capacity to process and store enormous amounts of information. Consequently, there is intense interest in the mechanisms that underlie learning and memory in vertebrates (Dubnau and Tully, 1998) as well as the relationships between synaptic plasticity and memory (Collingridge and Bliss, 1995; Stevens, 1998). Since the introduction of genetic analysis to cognitive neuroscience, attempts to distill memory to its molecular components have focused heavily on the signal transduction pathways that mediate synaptic plasticity in the hippocampus and other areas of brain. Although a number of signal transduction pathways are implicated in neuronal plasticity, there is increasing evidence that cross-talk between the Ca21, Erk/MAP kinase, and cAMP regulatory pathways may play a pivotal role for some forms of synaptic plasticity and memory formation (Wu et al., 1995; Xia and Storm, 1997; Atkins et al., 1998; Impey et al., 1998a; Storm et al., 1998). The importance of the cAMP signal transduction system for learning and memory was first demonstrated in invertebrate models for associative learning. Characterization of Drosophila memory mutants revealed that four ‡ To whom correspondence should be addressed (e-mail: dstorm@

u.washington.edu). § These authors contributed equally to this work.

of them are due to defects in the cAMP pathway: amnesiac, a putative adenylyl cyclase–activating peptide (Feany and Quinn, 1995); dunce, a cAMP phosphodiesterase (Chen et al., 1986); rutabaga, a Ca21-stimulated adenylyl cyclase (Livingston et al., 1984); and DCO, a cAMP-dependent protein kinase (Foster et al., 1984). In addition, induction of a dominant-negative form of the cAMP response element (CRE)–binding protein (CREB) in Drosophila blocks memory formation (Yin et al., 1994), whereas induced expression of an activator isoform of dCREB2 enhances LTM formation (Yin et al., 1995). Furthermore, cAMP-dependent protein kinase (PKA) and apCREB contribute to both short- and long-term presynaptic facilitation in mechanosensory neurons of Aplysia, another invertebrate model for learning and memory (Bartsch et al., 1998). cAMP signaling is also critical for some types of LTM and synaptic plasticity in vertebrates including LTP (for reviews, see Silva et al., 1998; Tully, 1998). LTP in the hippocampus is a usage-dependent increase in synaptic efficacy that may be an appropriate electrophysiological model for learning and memory (for reviews, see Malenka, 1994; Stevens, 1998). LTP can be subdivided into two temporally and mechanistically distinct forms: an early phase LTP (E-LTP), independent of increased gene transcription, and a late phase LTP (L-LTP) that requires cAMP and de novo gene transcription (Frey et al., 1991, 1993). Memory can also be divided into at least two mechanistically distinct forms: short-term memory (STM), lasting minutes to hours, and LTM, persisting for days or longer (Goelet et al., 1986; Tully et al., 1994). LTM depends upon activation of Ca21 and/or cAMP signaling (Livingston et al., 1984; Silva et al., 1992a; Wu et al., 1995) as well gene expression (Squire and Barondes, 1973; Tully et al., 1994). A role for cAMP in L-LTP and hippocampus-dependent memory is strongly supported by recent transgenic mice studies that showed that a reduction in PKA activity causes defects in L-LTP, spatial memory, and long-term contextual fear conditioning (Abel et al., 1997). One of the transcriptional pathways implicated in L-LTP and LTM is the CREB/CRE transcriptional pathway. For example, the stimulus paradigm that induces L-LTP and training for contextual or passive avoidance learning both stimulate CRE-mediated transcription and CREB phosphorylation in the hippocampus of mice (Impey et al., 1996, 1998b). Since cAMP increases are obligatory for Ca21 stimulation of CREmediated transcription in hippocampal neurons (Ginty et al., 1991; Impey et al., 1998a), PKA activation may play a key role for LTP and LTM by contributing to the expression of a family of genes regulated through the CREB/CRE transcriptional pathway. Activation of NMDA receptors or stimulus paradigms that generate LTP increase cAMP in area CA1 of the hippocampus (Chetkovich et al., 1991). How does activation of glutamate receptors in the hippocampus lead to increased cAMP? To date, genes for ten adenylyl cyclases have been cloned, nine of which are expressed in the hippocampus. These enzymes exhibit remarkable regulatory diversity and provide several mechanisms for

Neuron 788

Figure 1. Ca21-Stimulated Adenylyl Cyclase Activity Is Ablated in Brains of DKO Mice (A and B) Membrane preparations from the hippocampus (A) or cerebellum (B) of wild-type (WT, closed circles; n 5 9), AC1 (ACI-M, open triangles; n 5 5), AC8 (AC8-M, closed diamonds; n 5 5), or DKO mutant mice (open circles; n 5 15) were assayed for Ca21-stimulated adenylyl cyclase activity as described in the Experimental Procedures. (C) Forskolin-stimulated adenylyl cyclase from hippocampal (shaded bars) and cerebellar (dark, hatched bars) membranes prepared from wild-type and DKO mice. (D) Ca21-stimulated adenylyl cyclase activity in whole DKO brain preparations in the presence (hatched bars) or absence (dark, shaded bars) of 320 nM free Ca21.

glutamate receptor stimulation of cAMP. For example, Ca21 activates AC2, AC3, and AC5 through protein kinase C (Choi et al., 1993; Jacobowitz et al., 1993). Furthermore, the release of G protein bg subunits upon activation of metabotropic glutamate receptors stimulates AC2 and AC4 (Tang and Gilman, 1991; Gereau and Conn, 1994), both of which are expressed in the hippocampus (Baker et al., 1999). The most direct mechanism for coupling NMDA receptor activation to cAMP increases is through Ca21 activation of one of the calmodulin (CaM)–stimulated adenylyl cyclases, AC1 or AC8. However, transgenic mice lacking AC1 (Wu et al., 1995) or AC8 both show L-LTP in area CA1 of the hippocampus, suggesting that CaM-stimulated adenylyl cyclase activity is unnecessary for L-LTP. To assess the importance of the CaM-stimulated adenylyl cyclases for L-LTP and LTM, we made a transgenic mouse strain in which the genes for AC1 and AC8 were both disrupted (DKO mice). Results AC1 and AC8 Are the Only CaM-Stimulated Adenylyl Cyclases in Brain Membrane preparations isolated from whole brain or specific areas of brain including the hippocampus and cerebellum express Ca21-stimulated adenylyl cyclase

activity that is dependent upon CaM (Westcott et al., 1979; Villacres et al., 1995). Of the ten known adenylyl cyclases, only AC1 and AC8 are stimulated by Ca21/ CaM, although there may be other CaM-stimulated adenylyl cyclases that have not yet been discovered. To evaluate the contribution of AC1 and AC8 to Ca21-stimulated adenylyl cyclase in mouse brain, adenylyl cyclase activities in membranes from AC1, AC8, and DKO mutant mice were compared to wild-type mice. Ca21-stimulated adenylyl cyclase activity in the hippocampus from AC1 or AC8 single mutants was depressed 50% and 30%, respectively (Figure 1A). In the cerebellum, Ca21-stimulated activity was depressed approximately 50% in each single mutant (Figure 1B). In contrast, DKO mutants had no measurable Ca21-stimulated adenylyl cyclase activity in the hippocampus or cerebellum at any concentration of free Ca21 examined. There was, however, considerable forskolin-stimulated adenylyl cyclase activity in the cerebellum and hippocampus of DKO mice, consistent with the expression of other adenylyl cyclases in these regions of brain (Figure 1C). Membranes prepared from whole brain extracts of DKO mice were also devoid of Ca21-stimulated adenylyl cyclase activity (Figure 1D). This leads us to conclude that there are no other Ca21stimulated adenylyl cyclases expressed in brain, and no compensating mechanisms to restore this activity when the genes for AC1 and AC8 are both disrupted.

Ca21-Stimulated Adenylyl Cyclases in LTM and LTP 789

Figure 2. Hippocampal Morphology of WildType and DKO Mice (A and B) Giemsa staining for cell bodies in the hippocampal formation of wild-type (A) and DKO (B) mice. Areas CA1 and CA3 as well as the dentate gyrus, DG, are indicated (A). (C–F) Timm’s staining for mossy fiber processes in the DKO mice (D and F) and wildtype mice (C and E). Representative 43 (C and D) and 103 (E and F) magnified sections are displayed. PoDG, polyform layer of the dentate gyrus; MF, mossy fiber tract. Scale bar, 1 mm.

Morphology of the Hippocampus in DKO Mice Is Normal To determine if the ablation of Ca21-stimulated adenylyl cyclase activity causes neuroanatomical abnormalities in the hippocampus, we compared coronal sections from DKO and wild-type brains using two histological dyes: Giemsa, to stain cell bodies, and Neo-Timm’s, for staining zinc containing processes. Giemsa staining patterns in wild-type mice (Figure 2A) were indistinguishable from DKO mice (Figure 2B) at 43 magnification. Hippocampal morphology in the DKO hippocampal sections were intact and lacked any obvious malformations. The Neo-Timm’s stain was used to evaluate the integrity of some of the finer detail within the hippocampus; it effectively stains the polyform layer of the dentate gyrus (PoDG) as well as the mossy fiber track (MF) that synapses with the CA3 region of the hippocampus. The Timm’s stained sections of DKO mice (Figures 2D and 2F) demonstrate that the mossy fiber projections are intact and extend properly to the CA3 region, in a similar pattern as wild-type mice (Figures 2C and 2E). Higher magnification of Giemsa and Neo-Timm’s stained sections showed comparable cell densities in the dentate gyrus and pyramidal cell layers of DKO and wild-type sections, with no obvious abnormalities in the mossy fiber density (data not shown). Electrophysiological measurements described below also indicated that there are no major developmental defects in the hippocampus of DKO mutant mice. L-LTP Is Disrupted in DKO Mice Paired pulse facilitation and input–output relationships in wild-type and DKO hippocampal slices were compared to determine if basal synaptic transmission in

DKO mutant mice is intact. Paired pulse facilitation in DKO mice was indistinguishable from wild-type mice at all time intervals measured (Figure 3A). This suggests that the initial probabilities of transmitter release in DKO and wild-type mice are comparable. An input–output curve was constructed by taking the ratio of the field excitatory postsynaptic potential (fEPSP) and the postsynaptic fiber volley (fv) over a range of stimulus strengths (Figure 3B). A constant ratio through a set of stimulus strengths generally indicates a direct relationship between input stimulus and output response. The fEPSP/fv ratios for wild-type (1.683 6 0.17 s21) and DKO mice (1.585 6 0.063 s21) slices were comparable, indicating that the DKO and wild-type hippocampal slices behave similarly to input current. AC1 or AC8 mutant mice exhibited L-LTP with an increase in percentage field EPSP slope that persisted for 3 hr or longer following tetanus (Figure 3C). Although the average increase in percentage field EPSP slope appeared somewhat lower in AC8 mutant animals compared to wild type (166% 6 21% versus 188% 6 11%), the difference was not statistically significant. We were unable to generate sustained L-LTP in hippocampal slices from any of the DKO mice and a significant decrement in fEPSP slope was observed at all times after tetanus (p , 0.001). LTP from DKO brain slices dropped within 80 min. Interestingly, DKO mice exhibited E-LTP; only transcriptionally dependent L-LTP was severely impaired. The loss of L-LTP in DKO mice, but not in the single AC1 or AC8 mutants, suggests that Ca21-stimulated adenylyl cyclase is obligatory for L-LTP and that either AC1 or AC8 can generate the necessary cAMP signal. To characterize the loss in L-LTP, we examined the

Neuron 790

Figure 3. DKO Mice Have a Defect in L-LTP (A) Paired pulse facilitation in wild-type and DKO mutant mice. Percent facilitation, measured by taking the ratio of the second fEPSP slope over the first fEPSP slope between 30 ms to 150 ms, was quantitated. No significant differences in paired pulse facilitation were observed between DKO mice (open circles; n 5 8 mice, 13 slices) and wild-type mice (closed circles; n 5 5 mice, 10 slices) samples (at 30 ms and 50–150 ms, p . 0.5). (B) The ratio of field EPSP slope to fiber volley (fv) amplitude, acquired over a range of stimulus strengths, was used as another measure of basal synaptic transmission. The rate of stimulus was one stimulus per 120 s. The difference between wild-type (closed bar; n 5 6 mice, 13 slices) and DKO (open bar; n 5 5 mice, 10 slices) ratios (p . 0.5) was not significant. (C) L-LTP was induced in AC1 mutant (AC1-M, shaded triangles; n 5 9 mice, 17 slices) and AC8 mutant mice (AC8-M, open diamonds; n 5 13 mice, 25 slices) slices by four tetanic trains of 100 Hz (200 ms, 6 s apart) stimulus. There was no statistically significant difference between the mean fEPSP of AC1-M and AC8-M mutant mice 180 min after tetanization (p . 0.28). Representative fEPSP traces taken from area CA1 in AC1-M and AC8-M slices at 1 min and 180 min after tetani are shown in the insets. Each trace is superimposed over a baseline trace taken 2 min before tetani for ease of comparison. Scale bar, 1 mV, 10 ms. (D) L-LTP could not be evoked in DKO mice. Wild-type mice (closed circles; n 5 13 mice, 26 slices) preparations gave an L- LTP lasting up to 180 min, whereas potentiation in the DKO (open circles; n 5 8 mice, 19 slices) preparations declined to near baseline values within 80 min. Representative fEPSP traces taken from area CA1 in wild-type and DKO brain slices at 1 min and 180 min after tetani are shown in insets. The dashed lines (C and D) represent an arbitrary cutoff (120%), below which potentiation was considered to be near baseline. Scale bar, 1 mV, 10 ms.

effect of forskolin on synaptic activity using hippocampal slices from wild-type and DKO mutant mice. Application of forskolin causes a persistent increase in fEPSP slope that mimics L-LTP (Huang et al., 1994; Weisskopf et al., 1994; Impey et al., 1996) and occludes further increases in field EPSP slope caused by electrophysiological stimulation. Bath application of forskolin led to stable increases in fEPSPs in both wild-type and DKO hippocampal slices (153% 6 24% and 149% 6 19%, respectively) that lasted for at least 2 hr (Figure 4). Wildtype and DKO synaptic potentiations induced by forskolin were comparable in magnitude and induction kinetics. Presumably, forskolin treatment bypasses the requirement for Ca21-dependent initiation of LTP by robustly activating other isoforms of adenylyl cyclase in DKO hippocampal neurons (Figure 1C). These data indicate that the disruption of both the AC1 and AC8 genes does not lead to indirect developmental defects in signal transduction components downstream from the cAMP signal and are consistent with the observations that DKO

mice show normal hippocampal morphology and basal synaptic transmission. Ca21-Stimulated Adenylyl Cyclase Activity Is Required for Hippocampus-Dependent LTM DKO mutant mice were analyzed for several forms of hippocampus-dependent LTM including passive avoidance (Stubley Weatherly et al., 1996) and contextual memory (Logue et al., 1997). Since these learning paradigms all depend upon the ability of the mouse to detect a foot shock, which serves as an aversive cue, it was important to determine if DKO mice differ from wildtype mice in their sensitivities to foot shocks of varying intensities (Figure 5A). The stimulus thresholds for flinching, jumping, and vocalizing for the wild type and DKO mutants were statistically indistinguishable. Therefore, any differences in LTM between wild-type and DKO mice using a foot shock as a cue are not due to altered shock sensitivity. During passive avoidance training, mice were trained

Ca21-Stimulated Adenylyl Cyclases in LTM and LTP 791

Figure 4. Phenotypic Rescue of L-LTP in DKO Brain Slices by Forskolin Forskolin induced potentiation in hippocampal slices from wild type (closed circle; n 5 4 mice, 8 slices) and DKO mutants (open circle; n 5 3 mice, 7 slices). Forskolin (50 mM) was perfused in ACSF for 25 min, as indicated by the horizontal bar. DKO and wild-type slices showed comparable levels of potentiation after forskolin application.

to refrain from moving from the lighted to the darkened section of a conditioning chamber by delivering a foot shock when they entered the darkened side. Learning and memory was assessed by comparing the latency of “step through” in a training trial to that in a test trial carried out after training. Following one training session, wild type, AC1 mutants, and AC8 mutants learned to associate stepping through to the darkened chamber with the foot shock (Figures 5B and 5C). This was reflected as a greatly enhanced step through latency measured 30 min, 24 hr, or 8 days later. DKO mice also exhibited passive avoidance learning but a memory deficit that was evident as early as 30 min after training (Figure 5B). This memory defect for passive avoidance learning was also evident when DKO mice were tested 24 hr and 8 days after training (1.4 6 0.41 min and 1.9 6 0.50 min) compared to wild-type mice (4.7 6 0.31 min and 5.2 6 0.29 min). Since the DKO mice learn normally, performance factors such as perception and pain sensitivity are intact. The DKO latencies at 24 hr and 8 days were significantly lower than the wild type or the single mutants (analysis of variance [ANOVA], p , 0.001). These data indicate that Ca21-stimulated adenylyl cyclase activity is dispensable for passive avoidance learning but is required for normal passive avoidance LTM. Apparently either AC1 or AC8 can generate the required cAMP signal. The small amount of residual LTM for passive avoidance learning at 24 hr and 8 days after training in DKO mice may reflect contributions from other adenylyl cyclases or a component of memory that does not depend upon cAMP. Nevertheless, it is clear from the data that Ca21-stimulated adenylyl cyclases play the major role in generating cAMP signals required for LTM. To ascertain if DKO mice are deficient in other forms of hippocampus-dependent memory, they were examined for contextual learning. Contextual learning is characterized by the association of a normally innocuous context with an aversive stimulus, in this case, a foot shock (Kim and Fanselow, 1992; Logue et al., 1997). Mice that learn to associate the conditioning chamber with the shock show increased freezing during subsequent exposures to that chamber. Wild type, AC1 mutants, and AC8 mutants all exhibited strong contextual learning and memory that persisted for at least 8 days (Figure 6A). Although DKO mice learned to associate the conditioning context with the foot shock and remembered the training

context for 24 hr, they demonstrated a marked decrease (z43%) in freezing behavior compared to wild-type mice after 8 days. This indicates that DKO mice are not able to remember contextual learning as well as wild-type mice. Although passive avoidance and contextual memories both depend upon the hippocampus, the differences in the severity of memory loss for the two forms of learning in DKO mutant mice suggest that there are underlying mechanistic differences. To determine if DKO mice are generally deficient in learning, they were analyzed for cued learning and memory that depends upon the amygdala but not the hippocampus (Kim and Fanselow, 1992; Paylor et al., 1994). In this paradigm, mice were conditioned to fear an otherwise innocuous tone by pairing it with a foot shock. When presented with the tone in a novel environment, wild-type cue-trained mice spent a significantly higher percentage of time freezing in response to the tone than untrained mice (Figure 6B). The AC1, AC8, and DKO mice all showed normal cued learning and memory that persisted for at least 8 days. These data suggest that the DKO mice show a selective defect in memory for hippocampus-dependent learning and are not generally defective in memory. Restoration of LTM for Passive Avoidance Learning by Administration of Forskolin to Area CA1 of DKO Mice We cannulated DKO mice to deliver forskolin to area CA1 of the hippocampus during training to determine if the defect in passive avoidance learning is due to the loss of a cAMP signal required for learning. Cannulation of wild-type or DKO mice, without drug administration, did not alter passive avoidance learning. Verification of the cannula placement and forskolin penetration to area CA1 was done by infusing bodipy forskolin, a fluorescent-labeled forskolin, into the brains of DKO mice. Bodipy forskolin penetrated to area CA1 only on the cannulated side of the brain (Figure 7A). To determine if forskolin delivered by cannulation is sufficient to activate the CREB/CRE transcriptional pathway, a CRE– lacZ reporter mouse strain (Impey et al., 1996) was also cannulated. CRE-mediated gene transcription was used to confirm that forskolin increased cAMP in area CA1. Forskolin caused a significant increase in CRE-mediated transcription in the CA1 pyramidal cell body layer compared to vehicle controls (Figures 7B and 7C). Forskolin

Neuron 792

Figure 5. DKO Mice Are Deficient in LTM for Passive Avoidance Learning (A) Sensitivity of wild-type (closed bars, n 5 10) and DKO mice (open bars, n 5 10) to foot shock was evaluated to determine the threshold current required to evoke the following stereotypical behaviors: flinch, jump, and jump 1 vocalization. (B) Wild-type (closed bars, n 5 9) and DKO (open bars, n 5 14) mutant mice were submitted to passive avoidance training as described in the Experimental Procedures and analyzed for short-term memory at 5 min and 30 min. DKO mice exhibited comparable crossover latencies at 5 min, whereas at 30 min DKO mutant mice exhibit a statistically significant deficit (p , 0.002). (C) Wild-type (n 5 23), ACI (n 5 21), AC8 (n 5 26), and DKO (n 5 19) mutant mice were then submitted to passive avoidance. Pretrained mice (closed bars), 24 hr–naive (nonshocked controls; open, hatched bars), 24 hr–trained (shocked animals; open, stippled bars), and 8 day–trained (shocked animals; open bars). DKO mice showed marked differences in cross-over latencies at 24 hr (F3,85 5 23.35, p , 0.001) and 8 days (F3,85 5 21.77, p , 0.001) compared to wild type or the single mutants.

infusion also increased adenylyl cyclase activity relative to a vehicle control in both wild-type and DKO mice 30% 6 2% and 43% 6 12%, respectively (Figure 7D). Since the entire hippocampus was assayed for adenylyl cyclase activity, the local cAMP increase in area CA1 was likely greater than 30%–40%. Forskolin or vehicle was administered to DKO mice 15 min before passive avoidance training, and the mice were tested 24 hr and 8 days later (Figure 7E). DKO animals treated with forskolin demonstrated a significantly higher cross-over latency at both 24 hr (4.3 6 0.47 min, p , 0.001) and 8 days (4.77 6 0.24 min, p , 0.05) compared to vehicle-treated mice (1.28 6 0.51 and 2.2 6 0.18 min). Forskolin-treated DKO mice showed memory for passive avoidance learning at 24 hr and 8 days that was comparable to wild-type mice (4.7 6 0.34 and 5.2 6 0.29). Passive avoidance learning and memory

Figure 6. DKO Mice Are Deficient in LTM for Contextual Learning but Exhibit Normal Cued Learning and Memory (A) Wild type (n 5 15), AC1 (n 5 19), AC8 (n 5 19), and DKO (n 5 15) mutants were trained for contextual learning and tested 24 hr and 8 days after training. DKO mice showed a significant impairment in memory for contextual learning compared to wild-type mice after 8 days (F3,68 5 6.45; p , 0.001) but not after 24 hr. ANOVA analysis revealed a significant difference in the freezing of DKO mice measured at 24 hr and 8 days (F7,130 5 3.74; p , 0.001). (B) Cued fear-conditioning training revealed no significant differences between DKO and wild-type mice, AC1 mutants, or AC8 mutants at 24 hr (F3,34 5 1.33; p . 0.27) or 8 days after training (F3,34 5 0.47; p . 0.70).

in the vehicle-treated DKO controls were indistinguishable from uncannulated DKO mice. These data indicate that the passive avoidance memory defect in DKO mice can be overcome pharmacologically by increasing cAMP in area of CA1 during training. We conclude that the DKO mice are not generally deficient in other signaling components important for passive avoidance learning and that the primary defect is the loss of cAMP signaling required for LTM. These data also indicate that the LTP and memory defects exhibited by DKO mice are not due to subtle differences in genetic background between the mutants and wild-type mice. Discussion Although the relationships between L-LTP and LTM have not been fully defined, both depend on de novo

Ca21-Stimulated Adenylyl Cyclases in LTM and LTP 793

Figure 7. Administration of Forskolin to Area CA1 of DKO Mice Restores LTM for Passive Avoidance Learning (A) Bodipy forskolin (2 mM, 0.5 ml) infusion into the CA1 region of the hippocampus was visualized by confocal microscopy. Forskolin diffused from the site of injection (white arrows) to the CA1 region of the cannulated half of the hippocampus. Presented here is a coronal section from a representative animal (43 magnification; scale bar, 1 mm). CA1, CA1 hippocampus; CA3, CA3 hippocampus; DG, dentate gyrus; and CTx, cerebral cortex. (B and C) CRE–lacZ mice were cannulated and used to monitor the effect of forskolin administration on CRE-mediated transcription in area CA1. Magnified images (203) of CA1 pyramidal neurons from animals infused with either forskolin (B) or vehicle (C) are presented. (D) Adenylyl cyclase assays were performed on membranes prepared from hippocampi excised from either wild-type or DKO mice to determine if forskolin administered by cannulation increases adenylyl cyclase activity in the hippocampus at 15 min. Animals were sacrificed immediately after forskolin infusion, their brains were isolated, and membranes were prepared from the hippocampus and cerebellum. Forskolin infusion (hatched bars) significantly stimulated adenylyl cyclase activity in both wild-type (p , 0.001) and DKO (p , 0.001) hippocampi compared to vehicle controls, 3% DMSO (shaded bars). (E) The effect of forskolin infusion on passive avoidance learning using DKO mice. Passive avoidance training was carried out in the presence of 240 mM forskolin (n 5 10) or control vehicle (3% DMSO, n 5 4). Forskolin or vehicle (3% DMSO) was administered to the mice 15 min before passive avoidance training, and mice were tested 24 hr and 8 days later.

transcription and the cAMP regulatory system, suggesting that there may be a common underlying mechanism. A mechanistic relationship between L-LTP in area CA1 and LTM is supported by the data reported in this study and other research showing that defects in L-LTP correlate with loss of LTM (reviewed in Milner et al., 1998). A key unanswered question is the mechanism for generation of cAMP signals required for modulation of synaptic plasticity and memory formation. In principle, inhibition of cAMP phosphodiesterases, stimulation of adenylyl cyclases, or a combination of both could provide the cAMP increase. Since L-LTP and LTM are initiated by an intracellular Ca21 increase, Ca21-inhibited phosphodiesterases or Ca21-activated adenylyl cyclases are possible candidates for generation of cAMP transients needed for synaptic plasticity. In this study, we explored the hypothesis that Ca21 stimulation of adenylyl cyclases provides cAMP increases necessary for L-LTP and LTM. Although there are several mechanisms for Ca21 activation of adenylyl cyclases, the most direct mechanism is through the CaM-activated adenylyl cyclases (Westcott et al., 1979). The absence of any detectable Ca21-stimulated adenylyl cyclase activity in the

brains of DKO mice made it possible to unambiguously evaluate the role of this enzyme activity for L-LTP and LTM. DKO Mice Have a Selective Defect in L-LTP and Memory for HippocampusDependent Learning Ablation of the gene for either AC1 or AC8 did not seriously alter L-LTP at the Schaffer collateral/CA1 synapse, LTM for passive avoidance learning, or contextual learning. DKO mice, however, showed no L-LTP and were deficient in memory for passive avoidance and contextual learning. The defect in passive avoidance LTM was the most pronounced; memory impairment was evident 30 min after training. Deficiencies in contextual memory were not evident at 24 hr but were detected 8 days after training. The difference in kinetics for memory loss when DKO mice were trained for passive avoidance and contextual learning suggests that these two forms of learning use different neural pathways. This is consistent with lesioning studies that show that disruption of the entorhinal cortex interferes with passive avoidance learning but not contextual learning (Phillips and LeDoux, 1995;

Neuron 794

Good and Honey, 1997). In addition, training for contextual learning only stimulates CRE-mediated transcription in areas CA1 and CA3 (Impey et al., 1998b). In contrast, passive avoidance learning activates CREmediated transcription in areas CA1, CA3, and the DG, suggesting that signaling through all three areas is required for passive avoidance memory. Although we do not know why DKO mice exhibit much faster loss in memory for passive avoidance learning compared to contextual learning, it may be because signaling through the perforant pathway is required for passive avoidance memory. We hypothesize that lack of L-LTP and the memory defects are due to the loss of Ca21 stimulation of adenylyl cyclase in area CA1 of the hippocampus. This interpretation is supported by experiments that showed that administration of forskolin to area CA1 restored normal L-LTP and passive avoidance memory to DKO mice. In these experiments, forskolin increased cAMP levels in area CA1 on the cannulated-side of the hippocampus because of the presence of other forskolin-stimulated adenylyl cyclases in DKO mice. The recovery of L-LTP and LTM in DKO mice by administration of forskolin to area CA1 suggests that cAMP can play a permissive role for neuronal plasticity and that activation of other pathways (e.g., the Erk/MAP kinase pathway) may provide the anatomic specificity required for L-LTP and LTM. This is consistent with recent studies implicating the Erk/MAP kinase pathway for Ca21 activation of the CREB/CRE transcriptional pathway, L-LTP, and LTM (reviewed in Impey et al., 1999). Treatment by forskolin alone can rescue L-LTP and LTM in the DKO mice because cAMP activates Erk/MAP kinase in neurons (Martin et al., 1997; Impey et al., 1998a). The results from our study, however, do not preclude mechanisms wherein L-LTP and LTM arise from coactivation of the cAMP and Erk/MAP kinase pathways at specific synapses. It seems likely that the cAMP and Erk/MAP kinase pathways are costimulated in a synaptic specific manner by Ca21 during L-LTP and LTM. It is interesting that LTM for passive avoidance learning was restored to DKO mice when forskolin was administered by unilateral cannulation of the hippocampus. We believe that cAMP was increased in only one half of the hippocampus in these experiments because administration of fluorescent-labeled forskolin showed no diffusion of the drug to the noncannulated side of the brain. Furthermore, CRE-mediated transcription was only increased by forskolin right below the cannulation site. This suggests that the mice remember passive avoidance training when cAMP increases are generated in only one-half of the hippocampus, providing that other signal transduction pathways are intact. These data imply that only half the hippocampus is required for consolidation of passive avoidance memory. This is consistent with other studies showing that rats subjected to unilateral hippocampal lesions display normal T maze alternation (Douglas, 1975) and only a minor deficiency in the Morris water maze (van Praag et al., 1998). Similarly, animals with unilateral lesions of the fimbria fornix are able to learn in a variety of spatial paradigms including the Morris water maze (Spruijt et al., 1990) and the eight arm radial maze (Sara et al., 1992). Since the fimbria

fornix provides the major connections of the hippocampus with subcortical forebrain structures, these lesion studies provide further evidence for a bilateral functional redundancy in the hippocampal system.

The Role of cAMP for L-LTP and LTM Why are cAMP increases and PKA activation required for L-LTP and LTM? Even though transcriptional events underlying L-LTP and LTM have not been fully defined, the CREB/CRE transcriptional pathway has several properties that implicate it in L-LTP and LTM. Coactivation of Ca21 and cAMP signaling synergistically increases CRE-mediated gene expression in neurons (Sheng et al., 1990; Impey et al., 1998a). In addition, CRE-mediated transcription is increased following L-LTP (Impey et al., 1996), as well as contextual and passive avoidance learning (Impey et al., 1998b). Furthermore, LTM, but not STM, is deficient in mice that lack a- and d-CREB (Bourtchuladze et al., 1994; Kogan et al., 1997). Recent evidence suggests that activation of the Erk/MAP kinase cascade is obligatory for Ca21 stimulation of CRE-mediated transcription in hippocampal neurons and that the sequential activation of Erk and Rsk2 by Ca21 leads to the phosphorylation and transactivation of CREB (Impey et al., 1998a). Moreover, inhibition of the Erk/MAP kinase pathway by MEK inhibitors blocks both L-LTP (Impey et al., 1998a) and LTM (Atkins et al., 1998). Interestingly, the Ca21-induced translocation of ERK to the nucleus requires activation of PKA. This may be one of the major reasons why PKA activity is obligatory for the induction of CREBdependent transcription by Ca21 in neurons and why cAMP signaling is required for L-LTP and LTM.

Ca21-Stimulated Adenylyl Cyclase Activity Is Required for Several Types of LTP Most forms of LTP have a common requirement for increased intracellular Ca21 initiated postsynaptically through NMDA receptors or presynaptically through voltage-sensitive Ca21 channels. Although the mechanisms by which Ca21 enhances synaptic efficacy have not been fully defined, Ca21 activates several protein kinases including CaM-dependent protein kinases, Erk/ MAP kinase, and PKA. Furthermore, several forms of synaptic plasticity including LTP at Schaffer collateral, mossy fiber, and the medial perforant pathways are positively regulated by cAMP. This has lead to the general hypothesis that Ca21-stimulated adenylyl cyclases may be pivotal for specific types of LTP in the hippocampus and other areas of brain (Xia and Storm, 1997). In support of this hypothesis, AC1 is required for full expression of mossy fiber D-LTP (Villacres et al., 1998), an NMDA-independent, presynaptic form of LTP. The mechanism by which PKA activation causes a persistent increase in glutamate release in mossy fibers is not known, but PKA phosphorylation of Rab3A may contribute to PKA-mediated neurotransmitter release (Castillo et al., 1997). This mechanism may not be limited to LTP in the hippocampus since cerebellar parallel fibers exhibit an LTP with properties similar to hippocampal mossy fiber LTP (Salin et al., 1996). Cerebellar LTP is

Ca21-Stimulated Adenylyl Cyclases in LTM and LTP 795

independent of NMDA receptors but dependent on extracellular Ca21 and adenylyl cyclase activation. Interestingly, AC1 mice completely lack parallel fiber/Purkinje cell LTP measured in dissociated neuron cultures (Storm et al., 1998). Collectively, these studies suggest that Ca21 stimulation of AC1 or AC8 may be a general mechanism for regulation of synaptic plasticity in different areas of brain. Nevertheless, there are distinct mechanisms for regulation of synaptic plasticity by cAMP including enhancement of neurotransmitter release presynaptically, as in mossy fiber LTP, or activation of the CREB/CRE transcriptional pathway postsynaptically in area CA1. Furthermore, there are other forms of cAMPdependent LTP that do not depend upon AC1 or AC8. Amygdala LTP depends upon PKA (Huang and Kandel, 1998), but AC1 and AC8 are expressed at very low levels in the amygdala. Consequently, other adenylyl cyclases may provide the necessary cAMP signal for amygdala LTP. In summary, this study demonstrates that Ca21-stimulated adenylyl cyclase activity is required for L-LTP and specific forms of hippocampus-dependent LTM. Also, these studies reveal the novel observation that a memory defect in a transgenic mutant mouse strain can be systematically overridden by administration of a drug to a specific region of brain. These data indicate that either AC1 or AC8 may be useful drug target sites to modulate synaptic plasticity. In this respect, AC1 is particularly interesting since it is neurospecific (Xia et al., 1991, 1993) and provides a pharmacological window of opportunity to specifically increase cAMP in brain without perturbing cAMP-regulated processes in other organs. Experimental Procedures Generation of AC1 and AC8 Double Knockout Mice AC1 mutant mice were prepared as previously described (Wu et al., 1995). The AC8 knockout allele was generated by deletion of exon 1 and 2 kb of the 59-flanking region (Muglia et al., 1999) and results in complete absence of AC8 mRNA (M. L. S. et al., unpublished data). Disruption of the AC8 gene was directly confirmed by assay of Ca21-stimulated adenylyl cyclase activity in the hippocampus and cerebellum (Figure 1). In addition, Ca21-stimulated adenylyl cyclase activity was also assayed in the hypothalmus, which expresses AC8 but not AC1. AC8 mutant mice were completely devoid of Ca21stimulated adenylyl cyclase activity in the hypothalmus (data not shown). ACI and AC8 homozygote mice were mated to obtain doubly heterozygotic F1 progeny. F1 mice were then backcrossed to ACI homozygous mice to afford an F2 progeny with a 1:1 (heterozygote:homozygote) distribution at the ACI allele and a 1:1 (wild type:heterozygote) distribution at the AC8 allele. F2 mice heterozygous for the AC8 allele and homozygous for the ACI allele were backcrossed to afford 25% doubly homozygous mice (DKO mice). Control mice were derived by backcrossing F2 mice that were heterozygous for the ACI allele and wild type for the AC8 allele. Offspring from those matings that were wild type for both alleles were used as the control strain. The mice used in this study were in a mixed C57BL/6 3 129/SV background with generation of the mice with mutations in both AC1 and AC8 derived from heterozygous offspring from each of the initial chimeras. This ensured that significant background drift between mutant allele backgrounds had not occurred and provided wild-type controls in a comparable genetic background. In addition, several lines of wild-type and DKO mice were used to minimize effects due to strain variation. The loss of memory for passive avoidance learning and absence of L-LTP was seen in three independent lines. Furthermore, the LTP and memory defects associated with DKO mice were completely overridden by

forskolin, indicating that the phenotypic differences associated with DKO mice are due primarily to loss of Ca21-sensitive adenylyl cyclases activity and not to genetic background differences. Since it was not clear a priori that mice individually heterozygous for each allele would be identical behaviorally or electrophysiologically to wild-type mice, we used wild-type controls rather than heterozygote controls. Furthermore, since AC8 and AC1 have redundant functions, animals doubly heterozygous could exhibit significant phenotypic differences resulting specifically from decreased calcium-sensitive function. Mice doubly homozygous for the ACI and AC8 alleles were viable and fertile with an average litter size of 8–12 pups. These mice were well groomed and showed no overt sign of disease or obvious morphological abnormalities. DKO gait and movement were normal with no indication of tremor, seizure, or ataxia. Histology Freshly dissected brains from wild-type or DKO mice were fixed in phosphate-buffered, 4% paraformaldehyde (10 hr), and then cryoprotected in 30% sucrose (24 hr). Subsequently, sections (40 mm) were stained with Giemsa dye for neuron cell counting. Neo-Timm’s staining was performed by the published method (Holm and Geneser, 1991). Digital images were obtained using a Sony DKC5000 3CCD digital camera coupled to a Nikon eclipse E800 microscope. Images were acquired directly into Photoshop. Adenylyl Cyclase Assay Excised brain regions from wild-type and mutant mice were assayed for adenylyl cyclase activity as previously described (Storm et al., 1998). Free Ca21 concentrations were calculated using the Bound and Determined computer algorithm (Brooks and Storey, 1992). Forskolin (50 mM) was dissolved in ethanol and diluted 1:2000 in buffer A prior to treatment. Adenylyl cyclase activity levels are the means of triplicate determinations. Protein concentrations in the cell membranes were determined as described (Hill and Straka, 1988). Electrophysiology Freshly dissected mouse brains were immediately placed in cold (48C) artificial cerebrospinal fluid (ACSF; 120 mM NaCl, 3.5 mM KCl, 1.3 mM MgCl2, 2.5 mM CaCl2, 1.25 mM NaH2PO4, 25.6 mM NaHCO3, 10 mM glucose) that was aerated with a 95% O2/5% CO2 mixture. Transverse hippocampal slices (400 mm thick) from wild-type and mutant mice (male or female, 8–12 weeks old) were transferred to a submerged recording chamber containing ACSF (308C). ACSF was continuously perfused through the chamber at a rate of 1–2 ml/ min. Recordings were conducted after a 1 hr recovery period. A concentric bipolar electrode (100 mm diameter, David Kopf) was used to stimulate the Schaffer collateral afferent fibers in the stratum radiatum of the hippocampal CA1 region. Field EPSP synaptic responses were sampled at 0.02 Hz using a test pulse with an intensity that produced a half-maximal response. Responses were recorded with low resistance glass microelectrodes (1–2 mm) filled with 3.0 M NaCl. L-LTP was induced by giving four tetanic trains (100 Hz, 200 ms duration) of 20 pulses with an interpulse interval of 6 s (Silva et al., 1992b). Potentiation was determined by normalizing the fEPSP slope recorded after tetanic stimulation by the average of the baseline fEPSP slope. Traces were sampled and digitized using a Digidata 1200 (Axon Instruments) and collected using Axoscope 1.1 (Axon Instruments). Forskolin was stored in a concentrated stock (50 mM in EtOH) and diluted 1000-fold when applied to the perfusion solution. Contextual and Cued Fear-Conditioning Studies Sensitivity of DKO and wild-type animals to foot shock was measured as published (Bourtchuladze et al., 1994). Contextual fear conditioning was performed as described previously (Impey et al., 1998b). Briefly, context-trained mice were placed into a conditioning chamber, consisting of a glass enclosure and an electrified floor. On the training day, mice were placed in the chamber and allowed to roam for 2 min. During this time, baseline freezing behavior was monitored for each mouse. The mice were then given a mild foot shock (2 s, 0.7 mA). Mice were left in the training chamber an additional 1 min to recover and returned to their home cages. During testing, mice were returned to the conditioning chamber. Freezing

Neuron 796

behavior was defined as immobility at the sampling time and was measured at 10 s intervals over 2 min. Data was expressed as the percentage of time freezing over the sampling period. Cued fear-conditioning experiments were conducted in a manner similar to contextual with the exception that on training day, a 20 s auditory cue was given immediately prior to the foot shock. On the day of testing, 24 hr or 8 days after training, the mice were placed into a novel chamber (10 3 19 in, plastic enclosure, scented with dilute peppermint oil). Baseline freezing in the novel context (pretone) was monitored for 2 min immediately prior to delivering the cue. Freezing was measured over 2 min in the continuous presence of the cue (tone).

Statistical Analysis For statistical analysis of behavioral experiments, data from male and female mice were collapsed, since ANOVA describing gender versus either genotype or shock sensitivity revealed no significant differences. Experiments were analyzed with one-way ANOVA and two-factor ANOVA with replication. Posthoc analysis was conducted with the Tukey’s test for significance. The Student’s t test was used for analyzing electrophysiological data and adenylyl cyclase data. All error values are expressed as mean 6 SEM. In all studies, the investigator was blind to the genotype of the mice.

Passive Avoidance Learning Passive avoidance learning experiments were conducted in the conditioning chamber described for contextual fear-conditioning section except that enclosure was divided into two sections by a metal partition with a sliding trap door. One side of the box was shielded from light and the other exposed to the light. During testing, 5 min, 30 min, 24 hr, or 8 days later, mice were placed into the illuminated side of the box and allowed to roam freely to the darkened half. Step through latencies, defined as the time required for the mouse to cross completely to the darkened half, were recorded for the mice. The mouse was given a foot shock (2 s, 0.7 mA) after crossing into the darkened half of the chamber. The mouse was allowed to recover for 10 s and returned to its home cage. During testing, the latency of step through from the illuminated half of the box to the darkened half was monitored. Naive animals are animals that were processed under identical conditions as the trained animals; however, they were not shocked.

This study was supported by National Institutes of Health grants NS 37056 and NS 20498 to D. R. S. as well as the Howard Hughes Medical Institutes and a Burroughs Wellcome Fund Career Development Award to L. J. M. Support to J. A. was given by PHS NRSA T32 GM07270 from NIGMS.

Cannula Implantation and Forskolin Infusion Adult DKO mice (8 weeks old; 20–30 g) were anesthetized intraperitoneally with a mixture of ketamine (6.5 mg/ml) and xylazine (0.44 mg/ml) dissolved in bacteriostatic 0.9% saline. Once anesthetized, mice were placed in a sterotaxic device (David Kopf Instruments) and cannulae (24 gauge, stainless steel) were implanted into the left hippocampus of each animal, 1 mm dorsal to area CA1. Sterotaxically, cannulae resided 1.5 mm lateral, 1.5 mm posterior, and 0.7 mm ventral to bregma (Slotnick and Leonard, 1975). Cannulae were affixed with dental cement and fitted with wire plugs (30 gauge) to maintain cannula patency. After at least 1 week of postoperative recovery, mice were used for behavioral assays. Accuracy of cannula placement was verified by postmortem measurements on coronal sections. The drug infusion apparatus consisted of a Hamilton syringe (10 ml), connected to an injector needle (30 gauge) by a thin polyethylene tube, and motorized syringe pump. The needle was inserted into the cannulae and either forskolin (0.24 mM, Calbiochem) or vehicle (20% DMSO in filter-sterilized phosphate buffered saline [pH 7.4]) were infused into the mice at a rate of 0.1 ml/min over a 5 min period. Passive avoidance experiments were conducted 15 min after infusion of forskolin. Immunohistochemistry and Confocal Microscopy Transgenic mice harboring a CRE–lacZ reporter (Impey et al., 1996) were unilaterally cannulated and infused with either vehicle or forskolin, as described above. After an 8 hr induction period, the brain tissue was fixed in 6% paraformaldehyde, then cryoprotected in 30% sucrose overnight. Coronal sections (40 mm) were prepared and stained for b-galactosidase expression as described previously (Impey et al., 1996). Stained sections were visualized on a Bio-Rad MRC-600 confocal microscope, and image preparation was carried out using Photoshop 5.0. Bodipy Forskolin Infusion Bodipy-conjugated forskolin (2 mM, 0.5 ml) or control vehicle was infused into wild-type mice as described above. After 15 min, brains were rapidly removed and fixed in 6% paraformaldehyde for 12–14 hr. Coronal sections (50 mm) from cryoprotected tissue were taken and immediately mounted on glass slides. Slides were air dried then dipped briefly in propidium iodide (4% aqueous) to counterstain nuclei and rinsed in distilled water. Sections were visualized by confocal microscopy as described above.

Acknowledgments

Received April 27, 1999; revised July 1, 1999. References Abel, T., Nguyen, P.V., Barad, M., Deuel, T.A.S., and Kandel, E.R. (1997). Genetic demonstration of a role for PKA in the late phase of LTP and in hippocampus-based long-term memory. Cell 88, 615–626. Atkins, C.M., Selcher, J.C., Petraitis, J.J., Trzaskos, J.M., and Sweatt, J.D. (1998). The MAPK cascade is required for mammalian associative learning. Nat. Neurosci. 1, 602–609. Baker, L.P., Nielsen, M.D., Impey, S., Hacker, B.M., Poser, S.W., Chan, M.Y.M., and Storm, D.R. (1999). Regulation and immunohistochemical localization of bg-stimulated adenylyl cyclases in mouse hippocampus. J. Neurosci. 19, 180–192. Bartsch, D., Casadio, A., Karl, K.A., Serodio, P., and Kandel, E.R. (1998). CREB1 encodes a nuclear activator, a repressor, and a cytoplasmic modulator that form a regulatory unit critical for long-term facilitation. Cell 95, 211–223. Bourtchuladze, R., Frenguelli, B., Blendy, J., Cioffi, D., Schutz, G., and Silva, A.J. (1994). Deficient long-term memory in mice with a targeted mutation of the cAMP-responsive element-binding protein. Cell 79, 59–68. Brooks, S.P., and Storey, K.B. (1992). Bound and Determined: a computer program for making buffers of defined ion concentrations. Anal. Biochem. 201, 119–126. Castillo, P.E., Janz, R., Sudhof, T.C., Tzounopoulos, T., Malenka, R.C., and Nicoll, R.A. (1997). Rab3A is essential for mossy fibre longterm potentiation in the hippocampus. Nature 388, 590–593. Chen, C.N., Denome, S., and Davis, R.L. (1986). Molecular analysis of cDNA clones and the corresponding genomic coding sequences of the Drosophila dunce1 gene, the structural gene for cAMP phosphodiesterase. Proc. Natl. Acad. Sci. USA 83, 9313–9317. Chetkovich, D.M., Gray, R., and Sweatt, J.D.D. (1991). N-methyl-Daspartate receptor activation increases cAMP levels and voltage gated Ca21 channel activity in area CA1 of hippocampus. Proc. Natl. Acad. Sci. USA 88, 6467–6471. Choi, E.J., Wong, S.T., Dittman, A.H., and Storm, D.R. (1993). Phorbol ester stimulation of the type I and type III adenylyl cyclases in whole cells. Biochemistry 32, 1891–1894. Collingridge, G.L., and Bliss, T.V. (1995). Memories of NMDA receptors and LTP. Trends Neurosci. 18, 54. Douglas, R.J. (1975). The development of hippocampal function: implications for theory and for therapy. In The Hippocampus, R.L. Isaacson and H.P. Karl, eds. (New York: Plenum Press). Dubnau, J., and Tully, T. (1998). Gene discovery in Drosophila: new insights for learning and memory. Annu. Rev. Neurosci. 21, 407–444. Feany, M.B., and Quinn, W.G. (1995). A neuropeptide gene defined by the Drosophila memory mutant amnesiac. Science 268, 869–873. Foster, J.L., Guttman, J.J., Hall, L.M., and Rosen, O.M. (1984).

Ca21-Stimulated Adenylyl Cyclases in LTM and LTP 797

Drosophila cAMP-dependent protein kinase. J. Biol. Chem. 259, 13049–13055. Frey, U., Reymann, K.G., and Matthies, H. (1991). The effect of dopaminergic D1 receptor blockade during tetanization on the expression of long-term potentiation in the rat CA1 region in vitro. Neurosci. Lett. 129, 111–114. Frey, U., Huang, Y.Y., and Kandel, E.R. (1993). Effects of cAMP simulate a late stage of LTP in hippocampal CA1 neurons. Science 260, 1661–1664. Gereau, R.W., and Conn, P.J. (1994). A cyclic AMP-dependent form of associative synaptic plasticity induced by coactivation of betaadrenergic receptors and metabotropic glutamate receptors in rat hippocampus. J. Neurosci. 14, 3310–3318. Ginty, D.D., Glowacka, D., Bader, D.S., Hidaka, H., and Wagner, J.A. (1991). Induction of immediate early genes by Ca21 influx requires cAMP-dependent protein kinase in PC12 cells. J. Biol. Chem. 266, 17454–17458. Goelet, P., Castellucci, V.F., Schacher, S., and Kandel, E.R. (1986). The long and short of long-term memory—a molecular framework. Nature 322, 419–422. Good, M., and Honey, R.C. (1997). Dissociable effects of selective lesions to hippocampal subsystems on exploratory behavior, contextual learning, and spatial learning. Behav. Neurosci. 111, 487–493. Hill, H.D., and Straka, J.G. (1988). Protein determination using bicinchoninic acid in the presence of sulfhydryl reagents. Anal. Biochem. 170, 203–208. Holm, I.A., and Geneser, F.A. (1991). Histochemical demonstration of zinc in the hippocampal region of the domestic pig: the dentate area. J. Comp. Neurol. 308, 409–417. Huang, Y.Y., and Kandel, E.R. (1998). Postsynaptic induction and PKA-dependent expression of LTP in the lateral amygdala. Neuron 21, 169–178. Huang, Y.Y., Li, X.C., and Kandel, E.R. (1994). cAMP contributes to mossy fiber LTP by initiating both a covalently mediated early phase and macromolecular synthesis-dependent late phase. Cell 79, 69–79. Impey, S., Mark, M., Villacres, E.C., Poser, S., Chavkin, C., and Storm, D.R. (1996). Induction of CRE-mediated gene expression by stimuli that generate long-lasting LTP in area CA1 of the hippocampus. Neuron 16, 973–982. Impey, S., Obrietan, K., Wong, S.T., Poser, S., Yano, S., Wayman, G., Deloulme, J.C., Chan, G., and Storm, D.R. (1998a). Cross talk between ERK and PKA is required for Ca21 stimulation of CREBdependent transcription and ERK nuclear translocation. Neuron 21, 869–883. Impey, S., Smith, D.M., Obrietan, K., Donahue, R., Wade, C., and Storm, D.R. (1998b). Stimulation of cAMP response element (CRE)mediated transcription during contextual learning. Nat. Neurosci. 1, 595–601. Impey, S., Obrietan, K., and Storm, D.R. (1999). Making new connections: role of Erk/MAP kinase signaling in neuronal plasticity. Neuron 23, 11–14. Jacobowitz, O., Chen, J., Premont, R.T., and Iyengar, R. (1993). Stimulation of specific types of Gs-stimulated adenylyl cyclases by phorbol ester treatment. J. Biol. Chem. 268, 3829–3832. Kim, J.J., and Fanselow, M.S. (1992). Modality-specific retrograde amnesia of fear. Science 256, 675–677. Kogan, J.H., Frankland, P.W., Blendy, J.A., Coblentz, J., Marowitz, Z., Schutz, G., and Silva, A.J. (1997). Spaced training induces normal long-term memory in CREB mutant mice. Curr. Biol. 7, 1–11.

H.X., and Kandel, E.R. (1997). MAP kinase translocates into the nucleus of the presynaptic cell and is required for long-term facilitation in Aplysia. Neuron 18, 899–912. Milner, B., Squire, L.R., and Kandel, E.R. (1998). Cognitive neuroscience and the study of memory. Neuron 20, 445–468. Muglia, L.M., Schaefer, M.L., Vogt, S.K., Gurtner, G., Imamura, A., and Muglia, L.J. (1999). The 59-flanking region of the mouse adenylyl cyclase type VIII gene imparts tissue-specific expression in transgenic mice. J. Neurosci. 19, 2051–2058. Paylor, R., Tracy, R., Wehner, J., and Rudy, J.W. (1994). DBA/2 and C57BL/6 mice differ in contextual fear but not auditory fear conditioning. Behav. Neurosci. 108, 810–817. Phillips, R.G., and LeDoux, J.E. (1995). Lesions of the fornix but not the entorhinal or perirhinal cortex interfere with contextual fear conditioning. J. Neurosci. 15, 5308–5315. Salin, P.A., Malenka, R.C., and Nicoll, R.A. (1996). Cyclic AMP mediates a presynaptic form of LTP at cerebellar parallel fiber synapses. Neuron 16, 797–803. Sara, S.J., Dyon Laurent, C., Guibert, B., and Leviel, V. (1992). Noradrenergic hyperactivity after partial fornix section: role in cholinergic dependent memory performance. Exp. Brain Res. 89, 125–132. Sheng, M., McFadden, G., and Greenberg, M.E. (1990). Membrane depolarization and calcium induce c-fos transcription via phosphorylation of transcription factor CREB. Neuron 4, 571–582. Silva, A.J., Paylor, R., Wehner, J.M., and Tonegawa, S. (1992a). Impaired spatial learning in alpha calcium calmodulin kinase II mutant mice. Science 257, 206–210. Silva, A.J., Stevens, C.F., Tonegawa, S., and Wang, Y. (1992b). Deficient hippocampal long-term potentiation in alpha-calmodulin kinase II mutant mice. Science 257, 201–206. Silva, A.J., Kogan, J.H., Frankland, P.W., and Kida, S. (1998). CREB and memory. Annu. Rev. Neurosci. 21, 127–148. Slotnick, B.M., and Leonard, C.M. (1975). A stereotaxic atlas of the albino mouse forebrain, Volume DHEW-75–100, P.D. MacLean, ed. (Washington, D.C.: US Government Printing Office). Spruijt, B., Pitsikas, N., Algeri, S., and Gispen, W.H. (1990). Org2766 improves performance of rats with unilateral lesions in the fimbria fornix in a spatial learning task. Brain Res. 527, 192–197. Squire, L.R., and Barondes, S.H. (1973). Memory impairment during prolonged training in mice given inhibitors of cerebral protein synthesis. Brain Res. 56, 215–225. Stevens, C.F. (1998). A million dollar question: does LTP 5 memory? Neuron 20, 1–2. Storm, D.R., Hansel, C., Hacker, B., Parent, A., and Linden, D.J. (1998). Impaired cerebellar long-term potentiation in type I adenylyl cyclase mutant mice. Neuron 20, 1199–1210. Stubley Weatherly, L., Harding, J.W., and Wright, J.W. (1996). Effects of discrete kainic acid-induced hippocampal lesions on spatial and contextual learning and memory in rats. Brain Res. 716, 1–2. Tang, W.J., and Gilman, G.A. (1991). Type specific regulation of adenylyl cyclase by G protein bg subunits. Science 254, 1500–1503. Tully, T. (1998). Toward a molecular biology of memory: the light’s coming on! Nat. Neurosci. 1, 543–545. Tully, T., Preat, T., Boynton, S.C., and Del Vecchio, M. (1994). Genetic dissection of consolidated memory in Drosophila. Cell 79, 35–47. van Praag, H., Qu, P.M., Elliott, R.C., Wu, H., Dreyfus, C.F., and Black, I.B. (1998). Unilateral hippocampal lesions in newborn and adult rats: effects on spatial memory and BDNF gene expression. Brain Res. 92, 21–30.

Livingston, M.S., Sziber, P.P., and Quinn, W.G. (1984). Loss of calcium calmodulin responsiveness in adenylyl cyclase of rutabaga, a Drosophila learning mutant. Cell 37, 205–215.

Villacres, E.C., Wu, Z., Hua, W., Nielsen, M.D., Watters, J.J., Yan, C., Beavo, J., and Storm, D.R. (1995). Developmentally expressed calcium-sensitive adenylyl cyclase activity is disrupted in the brains of AC1 mutant mice. J. Biol. Chem. 270, 14352–14357.

Logue, S.F., Paylor, R., and Wehner, J.M. (1997). Hippocampal lesions cause learning deficits in inbred mice in the Morris water maze and conditioned-fear task. Behav. Neurosci. 111, 104–113.

Villacres, E.C., Wong, S.T., Chavkin, C., and Storm, D.R. (1998). AC1 mutant mice have impaired mossy fiber long-term potentiation. J. Neurosci. 18, 3186–3194.

Malenka, R.C. (1994). Synaptic plasticity in the hippocampus: LTP and LTD. Cell 78, 535–538.

Weisskopf, M.G., Castillo, P.E., Zalutsky, R.A., and Nicoll, R.A. (1994). Mediation of hippocampal mossy fiber long-term potentiation by cyclic AMP. Science 23, 1878–1882.

Martin, K.C., Michael, D., Rose, J.C., Barad, M., Casadio, A., Zhu,

Neuron 798

Westcott, K.R., LaPorte, D.C., and Storm, D.R. (1979). Resolution of adenylyl cyclase sensitive and insensitive to Ca21 and CDR by CDR-Sepharose affinity chromatography. Proc. Nat. Acad. Sci. USA 76, 204–228. Wu, Z.L., Thomas, S.A., Villacres, E.C., Xia, Z., Simmons, M.L., Chavkin, C., Palmiter, R.D., and Storm, D.R. (1995). Altered behavior and long-term potentiation in type I adenylyl cyclase mutant mice. Proc. Natl. Acad. Sci. USA 92, 220–224. Xia, Z.G., and Storm, D.R. (1997). Calmodulin-regulated adenylyl cyclases and neuromodulation. Curr. Opin. Neurobiol. 7, 391–396. Xia, Z.G., Refsdal, C.D., Merchant, K.M., Dorsa, D.M., and Storm, D.R. (1991). Distribution of mRNA for the calmodulin-sensitive adenylate cyclase in rat brain: expression in areas associated with learning and memory. Neuron 6, 431–443. Xia, Z., Choi, E.J., Wang, F., Blazynski, C., and Storm, D.R. (1993). Type I calmodulin-sensitive adenylyl cyclase is neural specific. J. Neurochem. 60, 305–311. Yin, J.C.P., Wallach, J.S., Del Vecchio, M., Wilder, E.L., Zhou, H., Quinn, W.G., and Tully, T. (1994). Induction of a dominant negative CREB transgene specifically blocks long-term memory in Drosophila. Cell 79, 49–58. Yin, J.C., Del Vecchio, M., Zhou, H., and Tully, T. (1995). CREB as a memory modulator: induced expression of a dCREB2 activator isoform enhances long-term memory in Drosophila. Cell 81, 107–115.