The Ca2+/calmodulin-stimulated adenylyl cyclases, AC1 and AC8, play a critical role in several forms of neuroplasticity, including long-lasting long-term potentiation (L-LTP) and long-term memory (LTM). By coupling neuronal activity and Ca2+ increases to the production of cAMP, AC1 and AC8 activate cAMP-dependent signal transduction and transcriptional pathways critical for L-LTP and LTM.
Adenylyl cyclases catalyze the conversion of ATP to cAMP, an important second messenger with diverse regulatory roles in the nervous system. Ten members of the adenylyl cyclase family have been identified by isolation of specific clones, each with unique regulatory properties and tissue dis-tribution (for review see Ref. 46). In the brain, as in other tissues, adenylyl cyclases can be activated by receptor-coupled stimulatory Gs proteins. For example, noradrenergic neurons of the locus ceruleus project to pyramidal neurons of the hippocampus and activate adenylyl cyclase through Gs protein-coupled β-adrenergic receptors (FIGURE 1⇓). Two of the adenylyl cyclases, AC1 and AC8, are activated by Ca2+ through the Ca2+-binding protein calmodulin (CaM). This review focuses on the CaM-stimulated adenylyl cyclases and their role in neuronal signaling, synaptic plasticity, and memory formation. These enzymes link activity-dependent increases in intracellular Ca2+ to the production of intracellular cAMP.
Isolation and Structure of Adenylyl Cyclases
Brain CaM-sensitive and CaM-insensitive adenylyl cyclases were first separated by using CaM-Sepharose affinity chromatography (41). These two adenylyl cyclase activities were also distin-guishable by a monoclonal antibody raised against the CaM-sensitive adenylyl cyclase (26). A Ca2+/CaM-stimulated adenylyl cyclase activity was purified to near homogeneity by using CaM-Sepharose affinity chromatography (17, 48) and forskolin-Sepharose affinity chromatography (21). Purified adenylyl cyclase protein and amino acid sequencing led to the isolation of a cDNA clone encoding AC1 (12). Additional members of the adenylyl cyclase family were cloned from cDNA libraries using sequence information and probes derived from AC1.
On the basis of hydropathy and sequence analysis, adenylyl cyclases appear to have a similar predicted membrane topology consisting of a short cytoplasmic amino terminus (N1), a six-transmembrane domain (M1), a 40-kDa cytoplasmic loop (C1), another six-transmembrane domain (M2), and a cytoplasmic carboxy terminus of ~35 kDa (C2) (FIGURE 1B⇑). The catalytic domain of adenylyl cyclases is hypothesized to be bipartite and split between C1 and C2 regions. The predicted size of mammalian adenylyl cyclases ranges from 110 to 130 kDa, but these proteins generally have an apparent molecular weight on SDS gels of ~200 kDa due to glycosylation in the extracellular loops of M1 and M2. Although the membrane topology and structure of adenylyl cyclases resemble transporters and ion channels, there is no evidence that these enzymes function as transporters or ion channels. However, cultured neurons express a voltage-sensitive adenylyl cyclase activity (25), suggesting that neurons contain an adenylyl cyclase or regulatory protein that is, like some ion channels, sensitive to membrane potential.
AC1 is a Neurospecific Coincidence Detector
AC1 is the only neurospecific adenylyl cyclase identified thus far. It is expressed in brain, retina, and the adrenal medulla (45). Within the brain, AC1 is expressed in the hippocampus, neocortex, entorhinal cortex, cerebellar cortex, olfactory bulb, and pineal gland (29, 45). AC1 expression in hippocampus increases dramatically during postnatal days 1–16 (31). The tissue specificity and developmentally regulated expression of the AC1 gene may be controlled by a 280-bp region and binary E-box like factor just 5′ to the transcriptional start site (5). AC1 protein is detectable in the mossy fibers and the molecular layer of hippocampus and dentate in the macaque monkey, Macaca nemestrina (13), suggesting that AC1 is localized to axons in neurons of the hippocampus.
Ca2+ and CaM stimulate AC1 enzymatic activity with a half-maximal concentration of 150 nM free Ca2+, slightly above resting concentrations of Ca2+ in neurons (43). CaM interacts with AC1 within the C1 loop region, close to the catalytic domain (43). The CaM-binding site within this region was identified by using peptide competitors and site-specific mutagenesis. For example, a peptide corresponding to amino acids 495–522 blocked Ca2+/CaM stimulation of AC1 (32). In addition, mutagenesis of AC1 residues F503 and K504 decreased Ca2+ sensitivity and stimulation in vivo (43), suggesting that interaction with CaM in this region is required for maximal AC1 activation.
AC1 is not stimulated by Gs-coupled receptors alone. However, when Gs-coupled receptor activation is paired with an increase in intracellular Ca2+, a synergistic level of activation is observed (37). Activation with Ca2+ and β-adrenergic agonists also leads to synergistic stimulation of a CRE-mediated transcription (10). Therefore, AC1 is synergistically activated and functions as a coincidence detector to integrate Ca2+ and Gs-coupled receptor activation. AC1 is inhibited by inhibitory Gi-protein coupled somatostatin and dopamine D2 receptors (20). In addition, AC1 is inhibited by CaM kinase IV in vivo (38). AC1 has two CaM kinase IV consensus phosphorylation sequences, Ser-545 and Ser-552, near its CaM-binding domain. Mutagenesis of either of these serine residues to alanine abolishes CaM kinase IV inhibition of AC1 (38). Inhibitory constraints on AC1 activity may prevent unwanted cAMP production, and, because synaptic plasticity and memory formation depend on optimal cAMP levels, may play an important role by opposing stimulatory mechanisms.
AC8 is a Low-Affinity Ca2+ Detector
AC8 was isolated by using a PCR-based strategy that amplified regions of nucleotide homology between adenylyl cyclase family members (4). AC8 is expressed in brain, lung (19), and parotid gland (36). Within the brain, AC8 is found in hippocampus, olfactory bulb, thalamus, habenula, cerebral cortex, and hypothalamic supraoptic and paraventricular nuclei. This expression pattern can be recapitulated by using a β-galactosidase reporter gene under the transcriptional control of a 10-kb fragment of AC8 promoter (19). AC8 is stimulated by Ca2+/CaM but is approximately five times less sensitive to Ca2+ than AC1, with half-maximal stimulation at 800 nM free Ca2+ (20). The CaM-binding domain of AC8 has been localized to the carboxy-terminal C2 region and has homology to the CaM-binding domain of the α1A-subunit of P/Q-type Ca2+ channels (14). Peptide competitors corresponding to this domain inhibit Ca2+/CaM stimulation of AC8 (8). AC8 is also not stimulated by Gs-coupled receptors; in contrast to AC1, AC8 is not synergistically stimulated by Gs-coupled receptors and Ca2+ (20). Serotonin stimulates AC8 activity in vivo, but this stimulation is mediated by serotonin-induced increases in intracellular Ca2+, not by Gs-coupled stimulation (2). AC8 is not inhibited by Gi-coupled receptors or by CaM kinase IV in vivo. Thus AC8 functions purely as a Ca2+ detector and responds to relatively high concentrations of Ca2+.
AC1 and AC8 are Not Essential for Survival of Mice
The lack of specific and reliable inhibitors of adenylyl cyclases has necessitated the development of homozygous mutant “knockout” mouse strains to study the physiological roles of these genes in vivo. Interestingly, AC1−/− (44), AC8−/−(27), and AC1−/− × AC8−/− double-knockout (DKO) (42) mutant mice are viable and survive into adulthood, suggesting that Ca2+/CaM-stimulated adenylyl cyclase activity is not essential for survival. Consistent with the approximately equal levels of expression of AC1 and AC8 in hippocampus, Ca2+/CaM-stimulated cyclase activity is reduced by ~50% in AC1−/− or AC8 −/− single-knockout strains, indicating that there is no compensation by either enzyme for the other (42). In contrast, no Ca2+/CaM-stimulated adenylyl cyclase activity is observed in membrane preparations from the hippocampus of AC1−/− × AC8 −/− DKO mice (42).
One of the interesting functions of AC1 is in pattern formation in the developing retina and somatosensory cortex. The naturally occurring barrelless strain of mice harbors an inactivating mutation in the AC1 gene (1). These mice do not develop barrel structures, even though the size of their whisker representation is normal (40). Barrel structures are clusters of thalamocortical neurons in the somatosensory cortex that represent the receptive field of a single whisker. AC1−/− mice also fail to develop normal barrel structures (1). Barrelless mice also have deficits in topographical maps of the retina that result from abnormal projection by and pruning of retinal ganglion cells (24). Thus AC1 is important in the development of topographic maps necessary for organization of sensory inputs. AC1 may also be required for the proper trafficking of the GluR1 subunit of the α-amino-3-hydroxy-5-methyl-4-isoxazole proprionic acid (AMPA) receptor because AMPA currents are greatly reduced in thalamocortical slices from barrelless mice (16). However, since it has not been established that the only defect in barrelless mice is the loss of AC1, it cannot be concluded that AC1 is required for proper trafficking of the GluR1 subunit of the AMPA receptor.
Ca2+/CaM-Stimulated Adenylyl Cyclases are Required for Some Forms of Synaptic Plasticity
The role of CaM-stimulated adenylyl cyclases in synaptic plasticity was examined in hippocampal slice preparations from knockout mice. Slices from AC1−/− mice exhibit deficits in long-term potentiation (LTP) at a variety of synapses in brain. For example, in the Schaffer collateral pathway, AC1−/− mice display reductions in the magnitude and rate of increase of excitatory postsynaptic potentials (EPSPs) in the first 20 min following tetanus but were not different from wild-type slices in the level EPSP potentiation at 3 h after tetanus (44). Whereas long-lasting LTP (L-LTP) in AC8−/− slices in the Schaffer collateral pathway was only slightly reduced relative to AC1−/− slices, AC1−/− × AC8−/− DKO slices displayed a complete loss of L-LTP (42). This deficit appears to be restricted to long-term synaptic plasticity; a short-term form of synaptic plasticity, paired-pulse facilitation, and the basal properties of neurotransmission are normal in AC1−/− × AC8−/− DKO mice.
In contrast to Schaffer collateral LTP, which is dependent on N-methyl-d-aspartate (NMDA) receptor activation and on postsynaptic Ca2+, mossy fiber LTP is dependent on presynaptic Ca2+ and PKA activity (49). It has been hypothesized that mossy fiber LTP requires a Ca2+-sensitive adenylyl cyclase activity (39). Indeed, slices from AC1−/− display deficits in mossy fiber LTP (30). This deficit in AC1−/− mossy fiber LTP can be rescued by administration of the general adenylyl cyclase activator forskolin. AC8−/− slices also display deficits in mossy fiber LTP (35). AC8 is targeted to synapses more readily than AC1 in cultured hippocampal neurons, perhaps allowing it to function at mossy fiber synapses despite its lower Ca2+ affinity.
Another presynaptic form of LTP is found in the parallel fiber/Purkinje cell synapses of the cerebellum. AC1 is expressed at high levels in the cerebellum and is targeted to synapses in cultured cerebellar neurons (33, 47). Cerebellar preparations from AC1−/− mice display a complete loss of parallel fiber/Purkinje cell LTP induced by 4–8 Hz stimulation, an observation that may correlate with the poor performance of AC1−/− mice on the accelerating rotarod (15, 28).
Ca2+/CaM-Stimulated Adenylyl Cyclases are Necessary for Long-Term Memory Formation
To examine the role of CaM-stimulated adenylyl cyclases in memory, AC1−/−, AC8−/−, and AC1−/− × AC8−/− DKO mice were tested in a variety of memory tasks. AC1−/− and AC8−/− single-knockout mice exhibit normal long-term memory (LTM) in the passive avoidance and contextual fear conditioning paradigms (42), suggesting that these enzymes are functionally redundant in fear-associated memory formation (FIGURE 2⇓). However, AC1−/− mice do have a deficit in the Morris water maze. Although these mice acquire the hidden platform task normally, they do not show a training quadrant preference during a 24-h memory test (44). These data indicate that AC1 has a role in the recall of spatial memory.
In contrast to the single adenylyl cyclase knockout mice, the AC1−/− × AC8−/− DKO mice display a robust LTM deficit in the passive avoidance memory paradigm. The AC1−/− × AC8−/− DKO mice are significantly impaired when tested for memory at 30 min, 24 h, or 8 days after training (FIGURE 2⇑) (42). These mice perform normally during a 5-min test, indicating that these mice can learn the task normally and that short-term memory is intact. Interestingly, the passive avoidance memory impairment in the AC1−/− × AC8−/− DKO mice can be rescued by unilateral administration of forskolin directly into the CA1 region of the hippocampus 15 min before training, indicating that the loss of Ca2+/CaM-stimulated adenylyl cyclase activity can fully account for this memory phenotype. Activation of the other adenylyl cyclases expressed in the hippocampus of DKO mice by forskolin can compensate for the loss of AC1 and AC8. This does not imply that any drug that raises cAMP can improve memory; general increases of cAMP in the brain of wild-type mice block memory (22). The data reported in Wong et al. (42) suggest that drugs that increase AC1 activity when it is activated by Ca2+ are most likely to enhance memory. For example, it has been recently discovered that genetic increases in AC1 activity in the hippocampus of mice improve memory (34). Together, these behavioral data indicate that Ca2+/CaM-stimulated adenylyl cyclases are required for LTM and that either AC1 or AC8 can produce the cAMP signal required for fear-conditioned memory.
In this review, we have presented evidence indicating that AC1 and AC8 play critical roles in L-LTP and LTM. The necessity for these adenylyl cyclases in L-LTP and LTM is likely due, at least in part, to the activation of cAMP-dependent signal transduction and transcriptional pathways. For example, the Erk/MAPK signal transduction and CREB/CRE transcriptional pathways play central and important roles in plasticity and memory formation (23). The basic elements of these pathways are shown in FIGURE 3⇓.
NMDA receptor activation by glutamate is required for Schaffer collateral L-LTP and for some forms of hippocampus-dependent memory (18). Ca2+ entry via these receptors can stimulate adenylyl cyclase activity and cAMP production (6). cAMP appears to have two major effectors in neurons: PKA and the cAMP-stimulated guanine nucleotide exchange factor, both of which stimulate the Erk/MAPK pathway through Rap-1 (7, 11). PKA is also required for the translocation of Erk/MAPK into the nucleus of neurons (9). Within the nucleus, Erk/MAPK activates rsk2, the major CREB kinase in neurons. Phosphorylation of CREB by rsk2 and recruitment of the CaM kinase IV-phosphorylated transactivator protein CBP promote the transcription of CRE-containing genes. As demonstrated by experiments using inhibitors of transcription, L-LTP and LTM require de novo gene transcription. It is also likely that genes regulated by other signaling pathways and enhancer elements will be important in plasticity and memory. Thus the Ca2+/CaM-stimulated adenylyl cyclases, AC1 and AC8, couple neuronal activity to cAMP production and thereby recruit the signal transduction and transcriptional pathways necessary for L-LTP and LTM.
Ca2+/CaM Adenylyl Cyclases as Targets of Memory-Enhancing Drugs
Memory-enhancing drugs are attractive to pharmaceutical companies because the US population is ageing and concurrently undergoing normal cognitive decline. In addition, individuals are living longer and becoming more likely to develop neurodegenerative disorders such as Alzheimer disease or senile dementia that are accompanied by memory loss. Many of the companies pursuing these drugs are focusing on cAMP and the cAMP pathway because of the wealth of data implicating this pathway in memory and plasticity. For example, rolipram, an older type IV phosphodiesterase inhibitor that leads to increases in cAMP, can promote memory enhancements in aged rats (3). However, rolipram is of limited clinical value because of side effects associated with inflammatory response. The development of rolipram-like drugs with greater specificity and potency are being pursued. AC8 is distributed throughout the body and is therefore not an ideal candidate as a drug target. In contrast, AC1 is an excellent drug target because of its restricted and neurospecific expression pattern. As a proof of principal, our laboratory has developed a transgenic mouse strain overexpressing AC1 in the forebrain. These transgenic mice exhibit increased L-LTP and LTM (34), indicating that an increase in AC1 activity can enhance plasticity and memory formation. Compounds that stimulate AC1 in the presence of Ca2+/CaM may provide the greatest level of specificity and therapeutic value as memory-enhancing drugs.
This work was supported by grants from the National Institutes of Health
- © 2004 Int. Union Physiol. Sci./Am. Physiol. Soc.