Interactions between Neurogranin and Calmodulin in Vivo *

Neurogranin is a neural-specific, calmodulin (CaM)-binding protein that is phosphorylated by protein kinase C (PKC) within its IQ domain at serine 36. Since CaM binds to neurogranin through the IQ domain, PKC phosphorylation and CaM binding are mutually exclusive. Consequently, we hypothesize that neurogranin may function to concentrate CaM at specific sites in neurons and release free CaM in response to increased Ca2+ and PKC activation. However, it has not been established that neurogranin interacts with CaM in vivo. In this study, we examined this question using yeast two-hybrid methodology. We also searched for additional proteins that might interact with neurogranin by screening brain cDNA libraries. Our data illustrate that CaM binds to neurogranin in vivo and that CaM is the only neurogranin-interacting protein isolated from brain cDNA libraries. Single amino acid mutagenesis indicated that residues within the IQ domain are important for CaM binding to neurogranin in vivo. The Ile-33 → Gln point mutant completely inhibited and Arg-38 → Gln and Ser-36 → Asp point mutants reduced neurogranin/CaM interactions. These data demonstrate that CaM is the major protein that interacts with neurogranin in vivo and support the hypothesis that phosphorylation of neurogranin at Ser-36 regulates its binding to CaM.

tion within the IQ domain inhibits CaM binding, and conversely, CaM binding inhibits PKC phosphorylation at the serine within the IQ domain (21,22). These observations led to the general hypothesis that neuromodulin and neurogranin may bind and concentrate CaM at specific sites in neurons and that PKC phosphorylation or changes in Ca 2ϩ may result in the release of CaM (16,23).
Although the physiological functions of neurogranin have not been defined, its biochemical properties and postsynaptic localization have implicated it in several signal transduction pathways. For example, both neurogranin and neuromodulin have been shown to regulate CaM-dependent nitric oxide synthase activity through sequestration of CaM (24,25). Conversely, nitric oxide modifies neurogranin, reducing its ability to bind CaM or to be phosphorylated by PKC (26). Neurogranin and neuromodulin are also in vitro substrates for phosphorylase kinase and may interact with membrane phospholipids (27)(28)(29). It has also been hypothesized that neuromodulin and neurogranin may play pivotal roles in LTP by releasing CaM. Free CaM could then activate enzymes, including CaM kinases or adenylyl cyclases, that regulate synaptic plasticity (30,31).
The hypothesis that neurogranin may regulate postsynaptic CaM levels is based upon data showing that the two proteins interact in vitro. However, it has not been demonstrated that neurogranin binds CaM in vivo, and the possibility that neurogranin may interact with other proteins has not been investigated. Here, we utilized yeast two-hybrid technology to detect neurogranin-binding proteins and to characterize neurogranin/ CaM interactions in vivo.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes, T4 DNA ligase, and ␤-agarase were purchased from New England Biolabs. The TA cloning kit was from Invitrogen. The ABI Prism sequencing kit was from Perkin-Elmer. Centri-Sep spin columns were from Princeton Separations Inc. The Quick-change site-directed mutagenesis kit was from Stratagene. Yeast media was purchased from Difco. Sheared DNA and Lex A antibody were from CLONTECH. X-Gal was purchased from 5 Prime 3 3 Prime, Inc., Boulder, CO. Amino acids and other chemicals were from Sigma.
Yeast Media-Yeast were grown on YPAD or YC solid or liquid media containing specific amino acids (32).
Preparation of Neurogranin and Calmodulin Constructs-The coding sequence of rat neurogranin was polymerase chain reaction-amplified with the addition of new BamHI and EcoRI restriction sites using the neurogranin primers 5Ј-CCCCGAATTCATGGACTGCTGC-3Ј and 5Ј-CAATGGATCCTTAATCTCCGCTG-3Ј. The polymerase chain reaction product was cloned into a TA vector and then digested with BamHI and EcoRI. The gel-isolated insert was ligated into the two-hybrid BTM116 Lex A DNA-binding domain vector (33) or pcDNA 3.1 (Invitrogen). The coding sequence of a synthetic mammalian CaM (34) was polymerase chain reaction-amplified with the addition of two flanking BamHI sites using the CaM primers 5Ј-CCCCGGATCCGGATGGCTGACCAACT-CACC-3Ј and 5Ј-CCCCGGATCCTCACTTAGCCGTCATC-3Ј. The insert was cloned into a TA vector, removed by digestion with BamHI, and was ligated into the VP16 activation domain vector (35). Each construct was sequenced to ensure accuracy during cloning.
Yeast Transformations-Small and large scale yeast transformations and two-hybrid screens were performed essentially as described (32). Briefly, for small scale transformations, BTM116-neurogranin wildtype and mutant constructs were transformed into the L40 strain of yeast in combination with VP16-CaM. Positive transformants were grown for 2 days on YC media lacking leucine and tryptophan (ϪLeu/ ϪTrp) and were then streaked to YC media lacking histidine (ϪHis) or again to ϪLeu/ϪTrp. Growth was assayed on the ϪHis plates after 2 days. Filter lifts and ␤-galactosidase assays were performed on the yeast re-streaked to the ϪLeu/ϪTrp media. For large scale transformations and two-hybrid screens, BTM116-neurogranin wild-type and mutant constructs were transformed singly into L40 and were grown on YC media lacking tryptophan (ϪTrp) for 2 days. A single positive transformant colony was grown in 10 ml of liquid ϪTrp YC media overnight. The culture was diluted into 100 ml of ϪTrp YC media for another 24 h and was then diluted to an OD of 0.4 in YPAD media. Large scale transformations were performed using either a rat brain library (CLONTECH) or a mouse brain library (36). Library cDNAs were isolated from colonies that grew in the absence of histidine and had ␤-galactosidase activity. To check for specificity, each cDNA was retested in a mating assay against BTM116 lamin (37) and against the original bait construct. cDNAs that retested as positive were sequenced and identified.
Western Blotting-Western blotting was utilized to demonstrate appropriate expression of wild-type and mutant neurogranin constructs in yeast. The constructs were transformed singly into the L40 strain of yeast, and positive transformants were grown in YC media (ϪTrp) for 2 days. The cultures were pelleted at 1200 ϫ g and were rinsed once in 10 mM Hepes, pH 7.5. Cultures were re-pelleted and resuspended in an ice-cold lysis buffer composed of 10 mM Hepes, pH 7.5, 100 mM NaCl, 2 mM EDTA, 1% Triton X-100, 1 mM dithiothreitol, 2 mM phenylmethylsulfonyl fluoride, 20 g/ml aprotinin, and 20 g/ml leupeptin. Acidwashed beads (Sigma) were added, and cultures were vortexed 5 ϫ 30 s each. Tubes were returned to ice following each vortex. The yeast lysate was separated from the beads by centrifugation at 100 ϫ g. Lysates were re-centrifuged, and the supernatant was saved. The Pierce BCA kit was used to assay protein content in each sample, and SDS-polyacrylamide gel electrophoresis sample buffer was added. Fifteen g of protein were loaded per well and were electrophoresed on 12.5% SDSpolyacrylamide gel electrophoresis. The proteins were transferred to nitrocellulose, and membranes were blocked in 5% non-fat milk. Western blotting was carried out using a Lex A primary antibody (CLON-TECH) at 20 ng/ml for 2 h. Membranes were washed extensively in phosphate-buffered saline plus 0.2% Tween 20. Horseradish peroxidase-conjugated goat anti-mouse secondary antibody was added at a 1:7,500 dilution of a 1 mg/ml stock for 1 h. Membranes were washed again, and proteins were visualized using an ECL kit (Amersham Pharmacia Biotech).
Quantitative ␤-Galactosidase Assay-␤-Galactosidase activity was measured using a fluorometric assay (38,39). Yeast were transformed with wild-type and mutant neurogranin constructs; lysates were prepared as described above for Western blotting, and protein levels were determined. A 15 mg/ml stock solution of the substrate 4-methylumbelliferyl ␤-D-galactosidase was diluted 1:500 into 0.8ϫ Z buffer (32). One hundred l of substrate and 10 l of yeast extract were added to a microtiter plate and were incubated at 37 o C for 15 min. The amount of yeast lysate added was adjusted to yield activities within the linear range. The plate was read on a microfluor microfluorimeter (Dynatech Industries, Inc.). Values were divided by protein content to yield ␤-galactosidase activities. Activity was determined in triplicate for each colony, and four separate colonies were assayed for each condition. cAMP Accumulation Assay-Adenylyl cyclase activity was measured as changes in intracellular cAMP Ϯ stimulus (40). Briefly, HEK-293 cells were plated on 12-well polylysine-coated plates at a density of 2 ϫ 10 5 cells per well. Transfections were carried out for 5-6 h using LipofectAMINE (Life Technologies, Inc.) following the manufacturer's instructions. Each well was transfected with either pCEP-AC1 (0.125 g) or pCEP-AC8 (0.125 g) in combination with green fluorescent protein (control-0.7 g) or pcDNA3-neurogranin (0.7 g). A ␤-galactosidase construct (0.08 g) was transfected and used to normalize transfection efficiency. Cells were loaded with [ 3 H]adenine in Dulbecco's modified Eagle's medium plus 10% fetal bovine serum, 12 h prior to assay. Cells were assayed 48 -72 h following transfection. Radioactive media was aspirated and replaced with Dulbecco's modified Eagle's medium. Each well was pretreated with 1 mM isobutylmethylxanthine for 20 min. Cells were then stimulated with 5 M calcium ionophore, A23187, for 15 min. Reactions were stopped with the addition of 5% trichloroacetic acid plus 1 M cAMP. Proteins were precipitated at 4°C for 1-4 h, and acid-soluble nucleotides were separated by sequential Dowex AG-50W-X4 and neutral alumina chromatography (41). Changes in intracellular cAMP are reported in triplicate as the ratio of cAMP to the complete adenine pool (cAMP ϩ ATP ϩ ADP ϩ AMP). Activity was normalized to ␤-galactosidase activity.

Interaction between Neurogranin and CaM in Vivo-
To determine if neurogranin interacts with CaM in vivo, we utilized yeast two-hybrid technology. In this system, the interaction of two proteins, expressed as fusion constructs, is monitored within the confines of a yeast cellular environment (42). Interactions are measured by read-out from two transcriptional reporters. Activation of one reporter results in the production of histidine, allowing yeast to grow in the absence of this amino acid. Activation of the other reporter construct results in ␤-galactosidase activity which is measured by hydrolysis of an X-Gal substrate that causes yeast to turn blue.
An L40 strain of yeast was transformed with a wild-type neurogranin DNA-binding domain construct and a CaM activation domain construct as described under "Experimental Procedures." Positive transformants were streaked to media lacking histidine and were assayed for growth (Fig. 1). This assay showed that neurogranin interacts strongly with CaM in FIG. 1. Interactions between neurogranin and CaM in vivo. L40 was transformed with BTM116 neurogranin and VP16 CaM, and positive transformants were streaked to media lacking histidine. The plates were incubated for 2 days at 30°C. BTM116 neurogranin (NG) was expressed with VP16-CaM or VP16 alone. Two independent colonies are shown for each condition. Results shown are representative of three independent experiments. vivo but not with the activation domain sequence (VP- 16) alone. We also monitored neurogranin/CaM interactions using ␤-galactosidase activity as a reporter. Positive double transformants were streaked to media lacking leucine and tryptophan. The yeast were transferred to nitrocellulose filters, and ␤-galactosidase activity was assayed. This assay also revealed strong neurogranin/CaM interactions that were detectable within 30 min (data not shown).
Interaction between CaM and Neurogranin IQ Domain Mutants-Site-specific point mutants were created to determine if amino acids within the IQ domain of neurogranin are important for its interaction with CaM in vivo. These point mutants included Ile-33 3 Gln, Arg-38 3 Gln, Ser-36 3 Ala, and Ser-36 3 Asp. To determine whether the mutants were adequately expressed in yeast, Western analysis was performed on yeast lysates as described under "Experimental Procedures." An antibody that recognizes the Lex A DNA-binding domain was used to detect the fusion proteins. All of the neurogranin mutants were expressed and electrophoresed with an apparent mass of 30 kDa, the expected size for the fusion proteins (Fig. 2).
Each of the neurogranin mutants was co-expressed with CaM, and the interaction between the expressed proteins was examined. A range of interaction strengths was observed by growth on media lacking histidine (Fig. 3, A and B). Each mutant was compared with wild-type neurogranin on the same plate. The best indication of strong interactions is the ability of the yeast to form single colonies at the interior of each yeast streak. The Ile-33 3 Gln fusion protein did not interact with CaM, indicating that the isoleucine within the IQ domain is particularly important for neurogranin/CaM interactions. The Arg-38 3 Gln mutant interacted with CaM, but to a lesser degree than wild-type neurogranin (Fig. 3A).
In vitro data indicate that the serine within the IQ domain is a PKC phosphorylation site in neurogranin and other IQ domain proteins (8,15,19,22). When phosphorylated at this site, neurogranin can no longer interact with CaM in vitro. The two-hybrid point mutants at Ser-36 were examined to define the importance of this amino acid for CaM binding and to determine if introduction of negative charge at this site inhibits CaM binding in vivo. The Ser-36 3 Asp mutant was created to mimic neurogranin phosphorylated at this site by introduction of a negative charge. Mutation of the serine to an alanine (Ser-36 3 Ala) removes the putative PKC phosphorylation site. The change from serine to alanine did not affect the neurogra-nin/CaM interaction; yeast growth was comparable to that seen with wild-type neurogranin and CaM (Fig. 3B). Conversion of serine to an aspartic acid reduced but did not completely inhibit the neurogranin/CaM interaction. This is consistent with the hypothesis that introduction of negative charge by phosphorylation at Ser-36 may reduce CaM binding.
Since the results described above using growth on media lacking histidine are qualitative, the ␤-galactosidase reporter was used to provide a more quantitative evaluation of these interactions (Fig. 4). The data obtained with ␤-galactosidase expression are consistent with those reported in Fig. 3. The Ile-33 3 Gln, Arg-38 3 Gln, and Ser-36 3 Asp mutants did not show any interaction with CaM over background (BTM116 alone). Both wild-type neurogranin and Ser-36 3 Ala constructs interacted strongly with CaM. Because of variation between individual colonies and the sensitivity of the assay, it cannot be concluded that the Ser-36 3 Ala mutation has a higher affinity for CaM than native neurogranin. However, the slightly stronger interaction seen with the Ser-36 3 Ala mutant may reflect basal phosphorylation of native neurogranin that lowers the signal somewhat compared with the alanine mutant.
Does Neurogranin Interact with Other Proteins in Vivo?-Although the data described above indicate that neurogranin interacts with CaM in vivo, this does not preclude interactions between neurogranin and other proteins. Therefore, it was of interest to search for other neurogranin-binding proteins using the yeast two-hybrid screen. To address this question, we performed large scale two-hybrid screens using wild-type neurogranin as bait and two different activation domain brain cDNA libraries. A summary of all large scale neurogranin two-hybrid screens performed is reported in Table I. Yeast were sequentially transformed with neurogranin and either a rat brain or a mouse brain cDNA library. Positive transformants were plated to media lacking histidine, and histidine-positive colonies were retested for ␤-galactosidase activity. In the rat brain transformation, 100 histidine-positive colonies were isolated. Of these, all were positive for ␤-galactosidase activity, and the color change occurred within 30 min. For the mouse brain transformation, 500 histidine-positive colonies were retested for ␤-galactosidase activity. All colonies retested as positive, with approximately half of the colonies turning blue within 30 min. cDNAs were isolated from all of the rat brain positives and from a mixture of 150 strong and weaker positive colonies from the mouse brain transformation.
The neurogranin bait construct was segregated from the library cDNA in each L40 colony. Mating assays were used to test the specificity of interaction between the neurogranin and the isolated library cDNAs. The AMR70 strain of yeast was transformed with the wild-type BTM116 neurogranin bait, or with a BTM116-lamin fusion protein. Each L40 colony containing an isolated cDNA was mated to each of the AMR70 bait colonies. Interactions were scored by growth on media lacking histidine and by ␤-galactosidase activity. cDNAs were isolated only from colonies that showed positive growth on (ϪHis) media and ␤-galactosidase activity during the mating. Any cDNA that also interacted with the BTM116-lamin fusion protein was identified as a nonspecific neurogranin binding partner and was discarded.
Following the mating assays, we isolated 98 positive cDNAs from the rat brain transformation and 90 from the mouse brain transformation. Each of these DNAs was digested with Sau3A1, and the restriction patterns were compared. All of the digested DNAs had the same restriction pattern, suggesting that they were identical. Twenty cDNAs from each transformation were sequenced. Each of these cDNAs were identified as either rat or mouse CaM, indicating that CaM is the only neurogranin-binding protein detectable in yeast two-hybrid screens from two different rodent cDNA libraries.
Although the data described above suggests that CaM may be the only protein that interacts with neurogranin in vivo, there may be other neurogranin-binding proteins whose interactions are inhibited by CaM. The high levels of CaM in brain (10 -20 M) may have obscured the detection of other neurogranin-binding proteins. Furthermore, phosphorylated neurogranin may bind to other proteins, which would not have been detected in the original yeast two-hybrid screen. For these reasons, we re-screened the brain cDNA libraries using the Ile-33 3 Gln and Ser-36 3 Asp neurogranin mutants, neither of which interact strongly with CaM in vivo.
The Ile-33 3 Gln mutant was screened against both libraries. Results are shown in Table I. In both screens, none of the isolated cDNAs retested as positive in the mating assay. In addition, each of the positive cDNAs were tested for interactions against wild-type neurogranin and the other neurogranin mutants in mating assays. None of the cDNAs showed positive interactions with any of the baits tested. A subset of these cDNAs were sequenced, and none of the sequences correlated to "in frame" sequences. Similarly, in the Ser-36 3 Asp neurogranin mouse brain screen, none of the isolated cDNAs retested as positive in mating assays or when retransformed in yeast with each of the neurogranin constructs. Sequencing of 12 of the putative positives did not reveal in frame sequences.
Obvious differences were seen between the wild-type and mutant neurogranin two-hybrid screens. All transformations were equally efficient, but many fewer positives were found in the mutant screens. The ␤-galactosidase activity in the wildtype screens appeared within minutes, whereas positives detected in the mutant screens appeared in hours. At each step in the screening process, the number of mutant positives decreased substantially, while very few of the positives in the wild-type screens were lost during successive screens. We were unable to detect any neurogranin-binding proteins by screening the mutant neurogranins suggesting that CaM is the major protein that interacts with neurogranin in vivo.
Neurogranin Regulates Calmodulin-sensitive Targets in Vivo-Previous studies have shown that neurogranin and neuromodulin can regulate calcium/CaM-sensitive enzymes in vitro (24,25). For example, nitric oxide synthase activity decreases in the presence of increasing concentrations of neuromodulin and neurogranin. We used the CaM-stimulated adenylyl cyclases, AC1 and AC8, to determine whether neurogranin could regulate CaM-sensitive targets in vivo. AC1 and AC8 are both stimulated by Ca 2ϩ and CaM, in vivo. In these experiments, 293 cells were transiently co-transfected with neurogranin and either AC1 or AC8. Enzyme activity was measured in response to the calcium ionophore, A23187. Ca 2ϩ stimulation of AC1 or AC8 in vivo was markedly reduced when neurogranin was co-expressed (Fig. 5). These data indicate that neurogranin has the potential to regulate the activity of CaM-regulated enzymes in vivo by complexing and lowering the effective concentration of free CaM.

DISCUSSION
Neurogranin and the related IQ domain protein, neuromodulin, are hypothesized to regulate neuronal CaM levels in response to PKC phosphorylation and increases in intracellular Ca 2ϩ (3,16,23). Therefore, it has been important to determine whether these proteins bind CaM and are phosphorylated by PKC in vivo. Both neurogranin and neuromodulin are phosphorylated by PKC in vivo (8,43,44), and recently it was shown that neuromodulin binds CaM in vivo (45,46). The objectives of this study were to determine if neurogranin binds CaM in vivo through its IQ domain and to determine if other neurograninbinding proteins can be detected by yeast two-hybrid technology. The discovery of other neurogranin-binding proteins would provide mechanistic insights concerning the role of neurogranin in signal transduction and would suggest alternative biochemical mechanisms for neurogranin in neurons.
Our data indicate that neurogranin and CaM are also in vivo binding partners and that the interaction occurs primarily through the IQ domain of neurogranin. The Ile-33 3 Gln mutation completely inhibited the CaM/neurogranin interaction, illustrating the importance of hydrophobic interactions for CaM binding to neurogranin. Reduced binding of CaM to the Ser-36 3 Asp mutant of neurogranin is consistent with the hypothesis that introduction of negative charge at Ser-36 by PKC phosphorylation lowers CaM binding affinity.
CaM was the only neurogranin-interacting protein isolated from the rat and mouse brain two-hybrid screens. In addition, neurogranin-binding proteins were not detected using neurogranin mutant proteins that do not bind CaM. It is interesting that other two-hybrid screens performed on IQ domain proteins have yielded similar results. After extensive testing, CaM was the only protein found to interact with neuromodulin (45), and it was also the only protein isolated using Igloo, a Drosophila neuromodulin homologue, as bait (47). In a screen of IQGAP1, a human putative Ras GTPase-activating protein that contains four IQ domains, CaM was the most frequent interacting protein identified (48). Yeast two-hybrid screens do not necessarily detect all significant interactions. For example, although neurogranin and neuromodulin are PKC substrates, yeast twohybrid screens did not identify PKC as an interacting protein with either neuromodulin or neurogranin. Nevertheless, our data indicate that CaM is the only neurogranin-binding protein detectable by an extensive yeast two-hybrid screen; no other neurogranin-interacting proteins other than PKC have been identified in vitro or in vivo.
If neuromodulin and neurogranin regulate free CaM levels, these proteins should regulate the activity of CaM-stimulated enzymes. In support of this hypothesis, both neurogranin and neuromodulin have been shown to regulate the activity of CaMdependent enzymes (24,25). Addition of neurogranin or neuromodulin to Ca 2ϩ /CaM-activated nitric oxide synthase re- Steps of each two-hybrid screen are outlined to demonstrate screen efficiency. Each bait was screened against a rat and/or mouse brain library. The transformation efficiency is shown as number of transformants. Each library expresses between 1 ϫ 10 6 and 3 ϫ 10 6 independent clones. Histidine-positive transformants grew on media lacking histidine 48 h following transformation. These colonies were retested for ␤-galactosidase (␤-Gal) activity. Strong activity indicates conversion of X-Gal, demonstrated by the color blue, in 30 -60 min. Any activity seen from 4 -24 h was identified as weak. Library and bait plasmids were segregated. L40 containing the library plasmid was then mated to the AMR70 strain transformed with the bait plasmid or lamin. Mated colonies that regrew on media lacking histidine and had ␤-galactosidase activity in the presence of the bait plasmid, but not lamin, were identified as positive after mating. The percent of colonies that retested as positive is shown in the mating column. Proteins found to interact with the neurogranin baits are indicated as isolated clones.  5. Neurogranin inhibits adenylyl cyclase activity in vivo. 293 cells were transiently transfected with pcDNA3-neurogranin and either AC1 (A) or AC8 (B). As a control, when not co-expressed with neurogranin, adenylyl cyclases were expressed with equivalent amounts of a green fluorescent protein. Adenylyl cyclase activity was measured as the change in intracellular cAMP levels Ϯ 5 M A23187 as described under "Experimental Procedures." Enzyme activity is shown as the ratio of cAMP to a total pool of adenine nucleotides. Activities were measured in triplicate and were normalized to ␤-galactosidase expression. Error shown is Ϯ S.D. Data shown are representative of three separate experiments. duces NO synthase activity. This inhibition is not seen when neurogranin or neuromodulin are phosphorylated, or at high Ca 2ϩ concentrations, indicating that sequestration of CaM by neurogranin or neuromodulin regulates the activity of CaMstimulated enzymes. Our results show that neurogranin may also regulate calmodulin-stimulated targets in vivo. When coexpressed with neurogranin, the calcium/calmodulin-stimulated activities of adenylyl cyclases AC1 and AC8 were inhibited. As both AC1 and AC8 are expressed in brain, they may represent physiological targets for neurogranin.
The existence of presynaptic and postsynaptic CaM-binding proteins that may regulate free CaM in neurons has led to the interesting hypothesis that these proteins may play a pivotal role during synaptic plasticity, e.g. long term potentiation (LTP) (30,31). For example, neurogranin and neuromodulin are both phosphorylated by PKC during LTP (49 -53). There are several forms of mechanistically distinct LTP expressed in the hippocampus and other areas of brain, including forms that are predominantly presynaptic and others that are hypothesized to be postsynaptic (54). In either case, initial increases in intracellular Ca 2ϩ arising because of activation of voltage-sensitive Ca 2ϩ channels or glutamate receptors are thought to initiate signal transduction cascades leading to enhanced synaptic efficacy (55)(56)(57). CaM-stimulated enzymes including CaM kinases, NO synthase (58), and adenylyl cyclases (59,60) are thought to play major roles in the initiation, maintenance, and propagation of LTP. Consequently, neuromodulin and neurogranin may be critical components of the molecular machinery used for modulation of synaptic plasticity and the development of learning and memory in vertebrates. Our demonstration that neurogranin can regulate brain adenylyl cyclases through its interactions with calmodulin is consistent with this theory.