Analysis of the role of calmodulin binding and sequestration in neuromodulin (GAP-43) function.

We demonstrated previously that forced expression of the neuronal phosphoprotein neuromodulin (also known as GAP-43, F1, B-50, and p57) in mouse anterior pituitary AtT-20 cells enhances depolarization-mediated secretion and alters cellular morphology. Here we analyze the role of calmodulin binding by neuromodulin in these responses. In cells expressing wild-type neuromodulin, a complex with calmodulin that is sensitive to intracellular calcium and phosphorylation is localized to the plasma membrane. Transfection of several mutant forms of neuromodulin shows that the effects of this protein on secretion are dependent on both calmodulin binding and association with the plasma membrane. In contrast, the morphological changes depend only on membrane association. Thus, the multitude of effects of neuromodulin noted in previous studies may result from divergent properties of this protein.

However, a line of PC12 cells in which neuromodulin expression is nearly undetectable is nevertheless capable of robust neurite elongation in response to nerve growth factor (29). Additionally, cultured neurons derived from mice in which the neuromodulin gene has been disrupted by gene targeting extend axons to the same extent as cells that express this protein.
Further analysis of the neuromodulin (Ϫ) embryos revealed that retinal axons appear incapable of crossing the midline decision point in the optic chiasm, implying that their growth cones fail to respond to environmental guidance cues (30).
These results suggest that neuromodulin is perhaps not essential for axon elongation, but rather might function as a mediator of signal transduction pathways in the growth cone that serve to modulate the rate, extent, and trajectory of axonal growth.
In the adult nervous system, neuromodulin expression persists in pre-synaptic terminals in those regions where the synaptic modifications associated with learning and memory are thought to occur (31)(32)(33)(34)(35)(36). Additionally, the correlation of PKC 1mediated neuromodulin phosphorylation with long term potentiation of synaptic transmission in the hippocampus (37)(38)(39)(40) and neurotransmitter release in vitro (41) suggest a role for neuromodulin in neuronal plasticity, possibly via modulation of neurotransmitter release. Support for this notion is derived from the demonstration that antibodies to neuromodulin that prevent its phosphorylation inhibit evoked neurotransmitter release in permeabilized synaptosomes (42)(43)(44). Furthermore, the levels of neuromodulin mRNA expression and neurotransmitter release are well correlated in NG108 -15 cells (25). Finally, inhibition of neuromodulin expression in PC12 cells by antisense RNA markedly diminishes depolarization-mediated secretion of noradrenaline (45,46).
Neuromodulin appears to interact with several second messenger systems in the growth cone and nerve terminal. The N-terminal domain of neuromodulin has been shown to stimulate GTP/GDP exchange on the ␣ subunit of the GTP-binding proteins G o or G i (47)(48)(49). Injection of neuromodulin or neuromodulin N-terminal peptides in cultured dorsal root ganglion neurons or Xenopus laevis oocytes enhanced G-protein-mediated intracellular signaling (50,51). In PC12 cells neuromodulin also appears to be modulate the function of L-type calcium channels, thereby affecting their electrical excitability (52). Additionally, as an inhibitor of the phosphatidylinositol 4-phosphate kinase, neuromodulin influences the metabolism of phosphoinositides and may act therefore as a feedback inhibitor of calcium mobilization and PKC activity (53). Neuromodulin was also shown recently to be ADP-ribosylated (54); however, the functional significance of this is unknown.
Neuromodulin has also been shown to bind to calmodulin, with a higher affinity in the absence of Ca 2ϩ than is its presence. Calmodulin binding is also disrupted by phosphorylation of neuromodulin by the Ca 2ϩ /phospholipid-dependent PKC (55). These results have led to the hypothesis that neuromodulin serves to sequester calmodulin at the inner face of the plasma membrane in close proximity to calmodulin-dependent enzymes. Elevation of intracellular calcium would promote dissociation of calmodulin from neuromodulin, allowing the rapid * This work was supported by National Institutes of Health Grant NS26806 and National Science Foundation Grant IBN-9409721. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ activation of these substrates (56 -58). The abundance of neuromodulin in the nerve terminal, together with its relative affinity for calmodulin, suggests that neuromodulin might be a major neuronal calmodulin-binding protein. While studies have demonstrated that neuromodulin in synaptic plasma membranes binds exogenous radiolabeled calmodulin, the functional significance of the neuromodulin/calmodulin interaction in the intact cell remains to be explored.
We reported recently that forced expression of neuromodulin in mouse anterior pituitary AtT-20 cells enhanced K ϩ -evoked hormone secretion and induced changes in cellular morphology (59). These results suggested that AtT-20 cells would be a useful experimental system to explore the molecular mechanism(s) of neuromodulin action. Here, we describe the results of studies designed to test the hypothesis that neuromodulin modulates K ϩ -evoked secretion from AtT-20 cells by sequestering CaM at the plasma membrane and releasing it upon elevation of intracellular Ca 2ϩ .

EXPERIMENTAL PROCEDURES
Plasmid Construction-The cDNAs encoding [Asp 41 ]rat neuromodulin (55) and [Gly 3 ,Gly 4 ] neuromodulin (60) were the generous gift of Dr. Dan Storm (University of Washington, Seattle). The expression vectors for these mutant forms of rat neuromodulin were produced by ligating a 750-base pair HindIII/EcoRI restriction fragment containing the mutated rat neuromodulin coding sequence into the HindIII and EcoRI sites of the pRc/RSV vector (Invitrogen) with the Rous sarcoma virus (RSV) promoter (61) driving expression of the cDNA.
Cell Culture and Transfections-All cell culture reagents were from Life Technologies, Inc. Monolayer cultures of D16 cells and transfected cell lines were maintained in 95% OptiMEM-I, 5% fetal bovine serum. The original AtT-20 cells were cultured in 85% OptiMEM-I, 10% equine serum, 5% fetal bovine serum; medium for routine culture of the transfected cells contained 200 g/ml G418. All cell lines were incubated in humidified 95% air, 5% CO 2 at 37°C. The AtT-20/D16 cells were generously provided by Dr. Lee Limbird, Department of Pharmacology, Vanderbilt University, and the original AtT-20 cells were obtained from the American Type Culture Collection (CCL89).
The original AtT-20 cells were transfected using the LipofectAMINE reagent, according to the manufacturer's instructions as described previously (59). Permanent transfected cell lines were selected in medium containing 400 g/ml G418. Clones were then isolated by limiting dilution and expanded in culture.
Primary cultures of rat hippocampal neurons, generously provided by Dr. Randy Hall and Andrès Barria-Roman, were prepared from rat pups 24 -48 h post-natal. Hippocampi were dissected into room temperature osmotically balanced saline solution (SS: 137 mM NaCl, 5.3 mM KCl, 0.17 mM Na 2 HPO 4 , 0.22 mM KH 2 PO 4 , 10 mM Hepes, 33 mM glucose, 44 mM sucrose, 0.024 g/liter phenol red, pH 7.3, 325 mosm), cut into 5-6 pieces and incubated for 1 h with 10 ml of a 20 units/ml papain (Worthington) solution in SS. The papain solution was removed and the hippocampal fragments washed once in minimal essential medium supplemented with 10% heat-inactivated fetal calf serum, 1 l/ml serum extender (Collaborative Research), 21 mM glucose, 10 g/ml 5-fluoro-2Ј-deoxyuridine, 25 g/ml uridine (Sigma) (complete medium). Cells were then dissociated into fresh complete medium via 15-20 passes through a Pasteur pipette and plated into sterile, poly-L-lysinecoated 35-mm culture plates at an approximate density of 2 hippocampi/dish. Cultures were fed on post-culture days 1, 4, and 7 and used for experiments on culture day 8. Cultures prepared via this method are of extremely high density (roughly 1-2 million cells/35-mm plate) and have a very high neuron-to-glia ratio since they are cultured in the presence of a mitotic inhibitor (5-fluoro-2Ј-deoxyuridine) from the first day.
Intact Cell Chemical Cross-linking Experiments-Cells were washed free of medium twice with Krebs-Ringer-Hepes (KRH) buffer (125 mM NaCl, 4.8 mM KCl, 2.6 mM CaCl 2 , 1.2 mM MgSO 4 , 25 mM Hepes, 5.6 mM glucose, pH 7.4) and incubated at 20°C for the indicated time with the thiol-cleavable homobifunctional cross-linker dithiobis(succinimidyl propionate) (DSP, Pierce) at 1 mM diluted in KRH from a 25 mM stock solution in Me 2 SO (cross-linked cells) or solvent alone (non-cross-linked cells). At the end of the incubation period, cells were transferred on ice, washed three times with ice-cold KRH, scraped off the dish, and cell pellets were frozen at Ϫ70°C until the immunoblot analysis was performed. To study the effects of calcium on the association of neuromodu-lin and CaM, cells were incubated for 5 min in KRH alone or KRH containing 0.1% dimethyl sulfoxide (Me 2 SO, vehicle for A23187) as controls, or KRH containing either 10 M calcium ionophore A23187 (Sigma), or 56 mM KCl (in which the NaCl concentration was decreased to maintain iso-osmolarity) prior to treating the cells with DSP. To study the effects of PKC phosphorylation on the association of neuromodulin and CaM, cells were incubated for 5 min in KRH containing either 1 M 12-O-tetradecanoylphorbol-13-acetate (TPA, Sigma) or 0.1% Me 2 SO (vehicle for TPA) as a control, prior to treating the cells with DSP.
Immunoprecipitation and Immunoblot Analysis-Total cellular protein and subcellular fractionation were performed as described (59). Five hundred g of protein were immunoprecipitated with an anti-rat neuromodulin polyclonal antibody (9) for 1 h on ice. Immune complexes were collected on protein A-Sepharose CL4B beads (Sigma) and washed once with 1 ml of 1 M NaCl, 1% Triton X-100; twice with 1 ml of radioimmune precipitation buffer, 1 M urea; and once with 1 ml of 10 mM Tris, 1 mM EDTA, pH 7.5. The material bound via the cross-linking agent to the neuromodulin immune complex was released by incubating the washed beads for 30 min at 37°C with elution buffer (10 mM Tris, pH 7.5, 5 mM EDTA, 0.2 M dithiothreitol), followed by a brief centrifugation. The eluate containing the material released from the immune complex by the reducing agent was analyzed by immunoblot analysis with an anti-calmodulin antibody.
Immunoblot analysis was performed as described (59), using an anti-GAP-43 monoclonal antibody (clone GAP-7B10, Sigma,), or an anti-calmodulin monoclonal antibody (Upstate Biotechnology, Inc.), followed by a peroxidase-conjugated secondary antibody (Sigma,). Bound antibodies were detected by enhanced chemiluminescence (DuPont NEN) and exposure to x-ray film (X-Omat AR, Eastman Kodak Co.). Quantitation was done by scanning exposed films with the SigmaGel analysis software (Jandel Scientific Sofware).
Secretion Studies-Secretion from the different cell lines was analyzed as described (59). Secreted and cellular hormones were measured by radioimmunoassay. Net secretion (corticotrophin-releasing factor or K ϩ -stimulated minus basal) was expressed as the percent of total cellular stores of ␤-endorphin released during the incubation period.
Radioimmunoassay-␤-endorphin immunoassays were performed as described previously (62), using an antiserum that is specific for ␤-endorphin residues 15-26. Synthetic acetyl-␤-endorphin 1-27 was used as tracer and standard, and a 12-point standard curve was assayed with each group of samples. The unpaired t test was used to determine the statistical significance of the results.

Identification of a Protein Cross-linked to Neuromodulin-
While neuromodulin has been shown to interact with CaM in vitro (64 -66) and in purified synaptosomal plasma membranes (67), this interaction has yet to be demonstrated in vivo. Because the affinity of CaM for neuromodulin has been reported to be relatively low (68), we chose to use the approach of protein cross-linking to determine if CaM binds to neuromodulin in the intact cell. For these studies we used AtT-20/D16 (D16) cells, which express high levels of neuromodulin (59), and the thiolcleavable homobifunctional cross-linker DSP, which due to its lipophilic nature penetrates into the cytoplasm of living cells (69).
Cultures of AtT-20/D16 cells were treated with 1 mM DSP for different times, and proteins were extracted and analyzed by Western blot with an anti-neuromodulin monoclonal antibody or with an anti-CaM monoclonal antibody. With both antibodies, a prominent 70-kDa band is detected and increases in intensity with increased time of cross-linking (Fig. 1A). Other neuromodulin immunoreactive bands are detected, with the 50-kDa band corresponding to that reported for native neuromodulin in a variety of studies (reviewed in Ref. 1); the higher molecular bands (Ͼ100 kDa) may correspond to neuromodulin complexed with actin, as these two proteins have also been shown to interact (70). Anti-CaM antibodies also detect a 20-kDa band corresponding to native CaM. The 70-kDa band is the result of cross-linking since it appears only after crosslinking ( Fig. 1, B, lane 4; and C, lane 3) and is absent from membrane fractions from non cross-linked cells (Fig. 1, B, lane 2; and C, lane 1), or after reduction of cross-linked samples with ␤-mercaptoethanol ( Fig. 1, B, lane 3; and C, lane 4). These results provide support for the notion that neuromodulin and CaM are associated in the intact AtT-20 cell. Quantitation by scanning densitometry indicates that as much as 70% of the CaM in the particulate fraction of AtT-20/D16 cells is present in this 70-kDa complex (Fig. 1C, lane 3).
To analyze this association further, the original AtT-20 cells, which lack detectable expression of neuromodulin (59), and the D16 cells were cross-linked as described previously, and neuromodulin and CaM were detected in particulate and soluble cellular fractions by immunoblot analysis with anti-neuromodulin or anti-CaM antibodies, respectively. Neuromodulin is detectable only in the particulate fraction of D16 cells (Fig. 2, lane 1) and is absent from the soluble fraction of the D16 cells and either fraction of the original AtT-20 cells (Fig. 2, lanes  2-4). Both cell lines have similar levels of CaM, which is mostly a soluble protein (Fig. 2, lanes 6 and 8), but some CaM is also present in membrane fractions (Fig. 2, lanes 5 and 7). In the particulate fraction from cross-linked D16 cells, which express neuromodulin, 70% of CaM is present in a 70-kDa complex (Fig. 2, lane 5). This complex is not detected in the absence of neuromodulin, either in the soluble fraction of D16 cells (Fig. 2, lane 6) or in the original AtT-20 cells (Fig. 2, lanes 7 and 8), suggesting that the 70-kDa complex results from the association of neuromodulin and CaM.
Finally, immunoprecipitation of solubilized membrane proteins from cross-linked D16 cells with a polyclonal anti-neuromodulin antibody, followed by immunoblot analysis with a monoclonal anti-CaM antibody of the material eluted from the immune complex, provides additional evidence that CaM is associated with neuromodulin in the intact AtT-20 cells (Fig. 3).

Intracellular Calcium Elevation or PKC Phosphorylation of Neuromodulin Disrupt the Association between Neuromodulin and CaM-Extensive in vitro analysis of the neuromodulin-
CaM interaction has demonstrated that the affinity of neuromodulin for CaM is higher in the absence of Ca 2ϩ than in the presence of 1 mM Ca 2ϩ (68), leading to the suggestion that an elevation of intracellular calcium would promote dissociation of calmodulin from neuromodulin (58,68). To determine if this is in fact occurring in the intact cell, we performed cross-linking and immunoblot experiments with D16 cells cultured in conditions that induce elevation of intracellular Ca 2ϩ . D16 cells were treated with the Ca 2ϩ ionophore A23187 (10 M) for 5 min at 20°C or with 0.1% Me 2 SO (control) and then cross-linked with 1 mM DSP for 5 min at 20°C, and membrane proteins were analyzed by immunoblot analysis with anti-neuromodulin and anti-CaM antibodies (Fig. 4A). In the absence of the Ca 2ϩ ionophore, a prominent 70-kDa band corresponding to the neuromodulin-CaM complex was detectable with both antibodies (Fig. 4A, lanes 2 and 4). Treatment of the cells with the Ca 2ϩ ionophore caused a marked reduction in the intensity of this 70-kDa band (approximately 90% reduction, as measured by scanning densitometry ; Fig. 4A, lanes 1 and 3). Membrane depolarization is an alternative method for inducing an elevation of intracellular Ca 2ϩ in AtT-20 cells (59). Cultures of D16 cells were therefore also incubated for 5 min in KRH solution containing 56 mM KCl, prior to cross-linking with DSP, followed by analysis of membrane extracts by immunoblot analysis (Fig. 4B). The transient elevation of intracellular Ca 2ϩ produced by depolarization also caused a significant reduction of the complex between neuromodulin and CaM (Fig. 4B, lanes  2 and 4).
Phosphorylation of neuromodulin by PKC on Ser 41 , a residue in the CaM binding domain, has also been shown to interfere with the binding of CaM to neuromodulin (55). D16 cells were treated for 5 min with 1 M TPA, an activator of PKC, prior to cross-linking with DSP, followed by analysis of membrane extracts by immunoblot analysis (Fig. 4C). Phosphorylation of neuromodulin by PKC also caused a significant 50% reduction in the relative abundance of the neuromodulin-CaM complex  ( Fig. 4C, lanes 2 and 4).
In Primary Cultures of Hippocampal Neurons, the Association of Neuromodulin and CaM Is Also Regulated by Calcium and Phosphorylation-In the mature nervous system, neuromodulin is expressed in nerve terminals of regions associated with learning and memory (31-36). We decided therefore to investigate neuromodulin binding to CaM in isolated hippocampal neurons. Cultures (8 days after initiation) were treated in the same conditions that have been shown to disrupt the complex between neuromodulin and CaM in AtT-20/D16 cells: 10 M A23187, 1 M TPA or 56 mM KCl for 5 min at 20°C or with 0.1% Me 2 SO or KRH alone (controls), and then crosslinking and immunoblot analysis were performed as described.
In control cultures, a prominent 70-kDa band corresponding to the neuromodulin-CaM complex is detectable with anti-neuromodulin antibodies (Fig. 5A, lanes 1 and 6) and with anti-CaM antibodies (Fig. 5B, lanes 1 and 6). Treatment of the cultures with the Ca 2ϩ ionophore A23187, TPA, or elevated potassium caused a marked reduction in the intensity of this 70-kDa band (approximately 85% reduction, Fig. 5, A and B, lanes 2-5), similar to the results obtained in the AtT-20/D16 cells.
Requirement of CaM for Secretion-The above results verify that neuromodulin is associated with CaM in the intact cell, and that this interaction is regulated by intracellular Ca 2ϩ concentration and phosphorylation. We therefore proceeded to determine if CaM is necessary for depolarization-evoked secretion in the AtT-20 cells. To block the function of CaM, we used three different cell permeable CaM antagonists: calmidazolium chloride (71), trifluoperazine dimaleate (72), and W7 (73). D16 cells were treated for 30 min with increasing concentrations of the CaM antagonist, prior to stimulation with KRH containing 56 mM KCl (K ϩ -evoked release) or KRH (basal release). All three CaM antagonists inhibit K ϩ -evoked secretion of ␤-endorphin in a dose-dependent fashion. The IC 50 was 4 M for calmidazolium chloride, 8.5 M for trifluoperazine dimaleate, and 20 M for W7 (data not shown).
Transfection of AtT-20 Cells with Mutants of Neuromodulin-After verifying that CaM is necessary for K ϩ evoked secretion in AtT-20 cells, we investigated whether neuromodulinmediated concentration of CaM at the plasma membrane is also required. In neurons, neuromodulin is mostly associated

FIG. 4. Intracellular calcium elevation or PKC phosphorylation of GAP-43 disrupt the association between GAP-43 and
CaM. D16 cells were treated with the indicated agonist for 5 min, followed by cross-linking with 1 mM DSP for 5 min. Subcellular fractionation was performed as described under "Experimental Procedures" and 20 g of membrane proteins were analyzed on a 12% acrylamide-SDS gel in non-reducing conditions, followed by immunoblot analysis with an anti-GAP-43 monoclonal antibody (lanes 1 and 2) or an anti-CaM monoclonal antibody (lanes 3 and 4). Panel A, treatment with 10 M Ca 2ϩ ionophore A23187 (lanes 1 and 3) or Me 2 SO (vehicle for A23187; lanes 2 and 4). Panel B, treatment with KRH (control, lanes 1 and 3) or 56 mM KCl (lanes 2 and 4). Panel C, treatment with Me 2 SO (vehicle for TPA; lanes 1 and 3) or 1 M TPA (lanes 2 and 4). with the plasma membrane (1,2,12,13,15,74), and when the neuromodulin cDNA is transfected into non-neuronal cells, the protein shows a similar subcellular distribution (19,(75)(76)(77). The original AtT-20 cells, which lack neuromodulin (59), were transfected with plasmids in which expression of the [Asp 41 ]rat neuromodulin mutant cDNA or the [Gly 3 ,Gly 4 ]rat neuromodulin mutant cDNAs were driven by the RSV promoter. Binding of CaM to neuromodulin has been shown to be abolished by mutation of Ser 41 , a residue in the CaM-binding domain, to aspartate (55). Mutation of the 2 cysteine residues Cys 3 and Cys 4 in the N-terminal domain of neuromodulin to glycine has been shown to prevent neuromodulin association with the plasma membrane (60,78).
Permanently transfected cells were selected with G418, and five independent neuromodulin-expressing cell lines were obtained for each transfection. Subcellular fractionation was performed, and expression of neuromodulin protein in these cell lines was then analyzed by immunoblot (Fig. 6) 6A). However, [Gly 3 ,Gly 4 ]rat neuromodulin is present only in the soluble fraction of the cell lines transfected with this mutant (Fig. 6B).
Investigation of the binding of CaM to [Asp 41 ]neuromodulin by cross-linking and immunoblot analysis as described previously shows that, although both neuromodulin and CaM are readily detectable in the cell lines expressing this mutant, the 70-kDa species representing the neuromodulin-CaM complex is undetectable with either antibody, indicating that CaM cannot form a complex with [Asp 41 ]neuromodulin (Fig. 7A). In contrast, the [Gly 3 ,Gly 4 ]neuromodulin mutant does bind to CaM, as demonstrated by the presence of a 70-kDa band, detected with both anti-neuromodulin and anti-CaM antibodies, in the soluble fraction of the cell lines expressing this mutant form of neuromodulin (Fig. 7B).
Effect of Mutants of Neuromodulin on Secretion-The transfected AtT-20 cell lines were stimulated for 30 min with 100 nM corticotrophin-releasing factor or for 5 min with 56 mM KCl and secretion of ␤-endorphin was measured as described. These studies revealed that net corticotrophin-releasing factorevoked hormone secretion in the transfected cell lines expressing the mutant forms of neuromodulin (average of 8% of total cellular stores in 30 min, Table I) was similar to that in the parental AtT-20 and D16 cells (59). All of the essential components of the secretory machinery are thus present and functional in all cell lines. However, K ϩ -evoked ␤-endorphin secretion in the AtT-20:[Asp 41 ]rat neuromodulin and in the AtT-20: [Gly 3 ,Gly 4 ]rat neuromodulin cell lines was low (on the average 2.6% of total cellular stores in 5 min, Table I) and not significantly different from that in the parental AtT-20 cell line or AtT-20 cell lines obtained from transfection with the pRc/RSV vector alone (59). Hence, the mutant forms of neuromodulin that cannot bind CaM or associate with the plasma membrane fail to enhance K ϩ -evoked release in the transfected cells.
In control experiments, we have verified by spectrofluorimetric analysis with the dye fura2/AM that depolarization-mediated Ca 2ϩ influx is not impaired in these cell lines (Fig. 8). We also have verified by immunoblot analysis with an anti-CaM antibody that the levels of CaM expression in the transfected cell lines are not significantly different from those in the control cell lines (data not shown).
Effects of Mutant Forms of Neuromodulin on Cellular Morphology-As neuromodulin has been shown to induce process outgrowth when transfected into non-neuronal cells (19,20,26,27,52,79), and we have shown that transfection of neuromodulin in AtT-20 cells caused the cells to flatten and extend processes (59) (Fig. 9D), similar to the AtT-20 cells expressing wild-type neuromodulin (Fig. 9B). These morphological changes were not observed for the AtT-20:[Gly 3 ,Gly 4 ]rat neuromodulin (Fig. 9C) or control AtT-20:pRc/RSV CC1 cell lines cultured in the same conditions (Fig. 9A). DISCUSSION We have shown previously that forced expression of the neuronal phosphoprotein neuromodulin in mouse anterior pituitary AtT-20 cells enhances K ϩ -evoked ␤-endorphin secretion and causes morphological changes in these cells (59). Here we have used the AtT-20 cells to investigate the molecular mechanism of action of neuromodulin in these responses. We show that in the intact cell, neuromodulin is associated with CaM and that this interaction is regulated by intracellular Ca 2ϩ concentration and phosphorylation. Mutant forms of neuromodulin, which cannot bind CaM or associate with the plasma membrane, fail to enhance K ϩ -evoked secretion from transfected cells. In contrast, the morphological changes caused by neuromodulin depend only on the association of this protein with the plasma membrane.
Neuromodulin Binding to CaM and Regulation of This Interaction by Ca 2ϩ and Phosphorylation in Vivo-The interaction between CaM and neuromodulin has been characterized extensively via in vitro studies, which have demonstrated that the affinity of CaM for neuromodulin is higher in the absence of Ca 2ϩ that in its presence and that association of the two proteins decreased by phosphorylation of neuromodulin by PKC on Ser 41 , a residue in the CaM-binding domain. These results have led to the hypothesis that neuromodulin might act to sequester CaM at the plasma membrane in the unstimulated cell (57). In this scenario, upon elevation of intracellular Ca 2ϩ , CaM would be released and in a position to activate rapidly membrane-associated Ca 2ϩ /CaM-dependent enzymes. Despite the appeal of this hypothesis, the interaction between CaM and neuromodulin in the intact cell has only been demonstrated in preliminary studies using the yeast two-hybrid system (80), and the precise physiological significance of this interaction remains to be established.
The intact cell cross-linking studies described here have resulted in the identification of a 70-kDa immunoreactive band that reacts with both anti-neuromodulin and anti-CaM anti-  bodies. As this 70-kDa complex is present only where neuromodulin is expressed, it is likely that this molecular species represents a complex formed between neuromodulin and CaM. Additional evidence for this association comes from the coimmunoprecipitation of CaM with an anti-neuromodulin antibody from extracts of cross-linked cells. Previous investigations considering the abundance of these two proteins and the relative affinity of their interaction have suggested that a majority of the CaM in the intact cell would be bound to neuromodulin. Our studies have provided some support for this notion, as approximately 70% of the membrane CaM is associated with neuromodulin, and the efficacy of the cross-linking is less than complete. Our findings also complement those of Gispen and colleagues (67), who showed that in synaptosomal plasma membrane binding of neuromodulin to exogenous radiolabeled CaM can be demonstrated after addition of a cross-linking agent. Finally, our demonstration that the interaction between neuromodulin and CaM is regulated by intracellular Ca 2ϩ and PKC-mediated neuromodulin phosphorylation provides additional support for the hypothesis that the function of neuromodulin is to concentrate CaM at the plasma membrane and release it upon a stimulus-evoked influx of Ca 2ϩ .
Experiments with primary cultures of hippocampal neurons show that here CaM and neuromodulin also formed a 70-kDa complex similar to that observed in the AtT-20 cells. Since hippocampal neurons express higher levels of neuromodulin than the transfected AtT-20 cells, a correspondingly larger fraction of cellular CaM is bound to neuromodulin. Furthermore, the association of CaM and neuromodulin is also regulated by intracellular Ca 2ϩ concentration and phosphorylation in the primary cultures. Neuromodulin is expressed only in discrete regions of the adult brain (31)(32)(33)(34)(35)(36), including the hippocampus, where it is thought to be involved in synaptic plasticity and where its phosphorylation has been correlated with long term potentiation (37)(38)(39)(40). Therefore, sequestration of CaM may represent an vital component of the role of neuromodulin in synaptic plasticity in the mature nervous system. Essential Role of CaM Binding and Association with the Plasma Membrane in Neuromodulin-mediated Enhancement of Hormone Release-The physiological significance of the neuromodulin-CaM interaction at the plasma membrane in AtT-20 cells was confirmed by showing that, although forced expression of wild-type neuromodulin markedly increases depolarization-mediated hormone secretion, transfection of neuromodulin mutants that cannot bind CaM (Asp 41 ) or associate with the plasma membrane (Gly 3 ,Gly 4 ) fails to produce a significant enhancement of secretion.
In presynaptic nerve terminals, synaptic vesicles are docked in clusters at specific sites of the plasma membrane termed active zones, the proximity of which to Ca 2ϩ channels permits rapid fusion of the vesicles with the plasma membrane upon an influx of Ca 2ϩ (reviewed in Refs. 81 and 82). Similar to this specific localization of synaptic vesicles, the association of the CaM/neuromodulin complex to the plasma membrane may also function to permit the rapid secretion of neurotransmitter in response to massive stimulation, possibly via modulation of the fusion of synaptic vesicles with the plasma membrane.
Requirement for Membrane Association in Neuromodulinmediated Morphological Changes-Expression of wild-type neuromodulin in AtT-20 cells causes the cells to flatten and extend processes that are stable for several days in culture. Similarly, expression of the Asp 41 mutant neuromodulin also causes these morphological changes, indicating that CaM binding is not required for this effect. In contrast, AtT-20 cells expressing the (Gly 3 ,Gly 4 ) mutant neuromodulin resemble control cell lines morphologically, suggesting that membrane as-sociation may be necessary for the role of neuromodulin in process extension. These results are similar to those of other studies, which have shown that the morphological alterations resulting from forced expression of neuromodulin in several cell lines required association of this protein with the membrane skeleton and the plasma membrane (26,83).
Neuromodulin has been shown to be localized in clusters at the plasma membrane and to interact with the membrane skeleton, a prominent component of which is filamentous actin (8,74,84). Neuromodulin has been shown to bind to actin in vitro (70), possibly through a C-terminal domain homologous to a portion of neurofilament NF-L (85). Furthermore, in primary sensory neurons depleted of neuromodulin by antisense neuromodulin oligonucleotides, lamellar extensions of the growth cone lacked local F-actin concentrations and showed poor adhesion (28), suggesting that interaction of neuromodulin with the actin-based membrane skeleton may be important for the function of neuromodulin in the growth cone, either in the direct regulation of process extension or in the response of the growth cone to extracellular guidance cues (30).