Merlin Links to the cAMP Neuronal Signaling Pathway by Anchoring the RIβ Subunit of Protein Kinase A*

The cAMP-protein kinase A (PKA) pathway, important in neuronal signaling, is regulated by molecules that bind and target PKA regulatory subunits. Of four regulatory subunits, RIβ is most abundantly expressed in brain. The RIβ knockout mouse has defects in hippocampal synaptic plasticity, suggesting a role for RIβ in learning and memory-related functions. Molecules that interact with or regulate RIβ are still unknown. We identified the neurofibromatosis 2 tumor suppressor protein merlin (schwannomin), a molecule related to the ezrin-radixin-moesin family of membrane-cytoskeleton linker proteins, as a binding partner for RIβ. Merlin and RIβ demonstrated a similar expression pattern in central nervous system neurons and an overlapping subcellular localization in cultured hippocampal neurons and transfected cells. The proteins were coprecipitated from brain lysates by cAMP-agarose and coimmunoprecipited from cellular lysates with specific antibodies. In vitro binding studies verified that the interaction is direct. The interaction appeared to be under conformational regulation and was mediated via the α-helical region of merlin. Sequence comparison between merlin and known PKA anchoring proteins identified a conserved α-helical PKA anchoring protein motif in merlin. These results identify merlin as the first neuronal binding partner for PKA-RIβ and suggest a novel function for merlin in connecting neuronal cytoskeleton to PKA signaling.

The neurofibromatosis 2 (NF2) 1 tumor suppressor protein merlin (schwannomin) is structurally related to ezrin-radixinmoesin (ERM) proteins (1,2), which link the actin-containing cytoskeleton to specific membrane proteins. ERM proteins also interact with cytoplasmic signaling molecules and participate in Rho and PKC signaling (3). In cells merlin is localized to cortical actin structures in patterns that partly overlap with ERM proteins (4 -6). Inter-and intramolecular associations regulate the functions and interactions of merlin with other proteins (7,8). Such interacting molecules include ERM proteins, the transmembrane receptor CD44, the scaffolding protein EBP50/NHE-RF, signaling molecules Rho-GDI and HRS, and cytoskeletal proteins ␤-II-spectrin, actin (3), and paxillin (9). Biallelic inactivation of the NF2 gene, which encodes for merlin, leads to development of schwannomas and meningiomas in the dominantly inherited NF2 disease (10). NF2 mRNA and merlin are widely expressed (1,11,12), but the expression level in most tissues appears low. Human and rodent central nervous system neuronal cells express merlin, and in man merlin is also found in glial cells (11,(13)(14)(15). However, the function of merlin in neuronal cells has not been characterized.
The cyclic AMP-PKA pathway elicits a wide array of metabolic and functional processes including cell growth and division, actin cytoskeleton rearrangements, and gene transcription. Stimulation of adenylate cyclase catalyzes the conversion of cytoplasmic ATP to cAMP. Cyclic AMP exerts its effects through the activation of cAMP-dependent protein kinases. In its inactive form cAMP-dependent protein kinase, PKA, is a tetramer that consists of two regulatory and two catalytic subunits. There are three catalytic subunit isoforms (C␣, C␤, and C␥) and four regulatory subunit isoforms (RI␣, RI␤, RII␣, and RII␤), the expression of which varies among cell types and tissues (16). Cyclic AMP activates PKA by binding to its regulatory subunits, which disassociates the tetrameric complex and the catalytic subunits become free to phosphorylate substrate proteins. The specificity of PKA signaling is achieved by: 1) isoform multiplicity of the tetrameric complex; 2) the expression pattern of various subunits; and 3) compartmentalization of PKA at different subcellular locations through interaction with AKAPs. AKAPs bind PKA via the regulatory subunit of PKA and direct the kinase activity toward specific substrates at distinct intracellular locations. By binding additional signaling molecules, protein kinase A anchoring proteins (AKAPs) may also coordinate multiple signal transduction pathways (17,18).
Of the regulatory subunits, RI␤ is most abundantly expressed in brain (19,20). Hippocampal synaptic plasticity including long term depression and long term potentiation are defective in mice carrying a targeted disruption of the gene encoding for RI␤ (21,22). Multiple characterized protein interactions are involved in the targeting and regulation of RI␣, RII␣, and RII␤, but so far, no such interactions have been described for RI␤. Here, we have identified merlin as the first binding partner for RI␤, characterized a role for merlin as an AKAP for RI␤-containing PKA, and studied the expression and localization of both proteins in the central nervous system and in cultured neuronal cells.
cAMP Pull-down Experiment and Coimmunoprecipitation from Rat Brain Lysates-Five g of freshly prepared rat brain was homogenized in 15 ml of lysis buffer (20 mM Hepes, pH 7.4, 20 mM NaCl, 5 mM EDTA, 5 mM EGTA, 0.5% Triton X-100, 1 mM dithiothreitol, and protease inhibitors). After centrifugation of the homogenate, 20,000 ϫ g for 30 min, the pellet was resuspended in 15 ml of homogenate lysis buffer with 2% Triton X-100. 50 l of 8AHDAA-cAMP-agarose beads (Biolog, Bremen, Germany) were added to the lysate for the cAMP pull-down experiment and incubated at 4°C overnight. For the coimmunoprecipitation experiments PKA-RI␤ antiserum or merlin KF10 mAb were added to the lysate and rotated at 4°C overnight. 50 l of protein A/G-agarose beads (Santa Cruz Biotechnology) were added and incubation was continued for 2 h. Beads were washed twice with high salt buffer (10 mM Hepes, pH 7.4, 500 mM NaCl, 10 mM KCl, 1.5 mM MgCl 2 , 0.1% Nonidet P-40, 1 mM dithiothreitol and protease inhibitors) and four times with low salt buffer (10 mM Hepes, pH 7.4, 10 mM KCl, 1.5 mM MgCl 2 , 0.1% Nonidet P-40, 1 mM dithiothreitol and protease inhibitors). After washing, 50 l of 2ϫ SDS sample buffer was added to the beads. The samples were subjected to SDS-PAGE, transferred to nitrocellulose filters, and immunoblotted with PKA-RI␤ antiserum or merlin KF10 mAb. Bound proteins were detected by enhanced chemiluminescence.
Analysis of Transfected Proteins-COS7 cells were transfected with RI␤-GFP alone or RI␤-GFP and human merlin isoform I cDNA in pcDNA3 vector (Invitrogen, Groningen, the Netherlands) using FuGENE transfection reagent (Hoffmann-La Roche Ltd., Basel, Switzerland). After 60 h cells grown on glass coverslips were fixed in with 3.5% paraformaldehyde and stained with 1398NF2 rabbit antiserum (1:100) followed by Alexa 594 anti-rabbit antibody (Molecular Probes). The remaining cells from a 10-cm plate were lysed in 500 l of ELB buffer (50 mM Hepes, pH 7.4, 150 mM NaCl, 5 mM EDTA), 1% Nonidet P-40, and protease inhibitors, and centrifuged at 15,000 ϫ g for 1 h at 4°C. The supernatant was incubated with 1398NF2 rabbit antiserum, RI␤ rabbit antiserum, preimmune serum, 3C12 mAb or X63, together with protein G-Sepharose beads (Amersham Biosciences) for 4 h at 4°C. Immunoprecipitates were washed with ELB, 0.1% Nonidet P-40 and bound proteins were eluted from the beads by boiling in nonreducing Laemmli sample buffer. Precipitated proteins were detected as above. 293 HEK cells were transfected with RI␤-GFP and human merlin isoform I cDNA in pcDNA3 vector (Invitrogen). Nonconfluent cells were lysed in 500 l of ELB buffer, 1% Nonidet P-40, protease inhibitors, and phosphatase inhibitors, whereas confluent cells were lysed without phos-phatase inhibitors. Coimmunoprecipitations were done with KF10 mAb, RI mAb, GFP mAb, and X63 as above. Precipitated proteins were detected with merlin A-19 rabbit antiserum or GFP mAb.
Affinity Precipitation-GST fusion proteins were purified by absorption to glutathione-agarose beads according to the manufacturer's protocol (Amersham Biosciences). GST-merlin constructs 252-595 and 1-547, GST-AKAP149, or GST alone (1 M) were incubated with human RI␤ subunit (1 M) (27) under rotation in 50 mM Tris, pH 7.4, 300 mM NaCl, 0.1% Triton X-100, 1 mM EDTA, 5 mM dithiothreitol, 5 mM benzamidine and protease inhibitors for 30 min at room temperature. Incubation was continued with glutathione-agarose beads for an additional 2 h at 4°C, after which beads were pelleted and washed three times with the same buffer. After boiling in Laemmli sample buffer, precipitated RI␤ was detected as above.
Autospot Peptide Array and Screening-A 20-mer peptide array encompassing residues 425-480 of merlin was synthesized on cellulose paper by using an Autospot Robot ASP222 (ABiMED, Langenfeld, Germany). The peptide array was screened for RI binding by R-overlay using 32 P-labeled recombinant RI (A98S).
Sequence Alignment-The sequence alignment between merlin and known AKAPs was performed using the Clustal-W program (Lasergene, DNAstar Inc., Madison, WI).

Analysis of Merlin, Ezrin, and RI␤ in the Central Nervous
System-The expression of merlin, ezrin, and RI␤ in various regions of mouse, rat, and human central nervous system was analyzed using Western blotting and immunohistochemistry. Results of this comprehensive analysis will be described in detail elsewhere. 2 In conclusion, Western blot analysis demonstrated widespread expression of merlin, ezrin, and RI␤ in mouse and rat (Fig. 1) as well as human (not shown) central nervous system, as all analyzed regions (cerebrum, cerebellum, medulla, and pons) were positive. Immunohistochemical staining of human brain sections indicated expression of RI␤ and merlin predominantly in cells morphologically identified as neurons. This is exemplified by sections from hippocampus and thalamus. In contrast, ezrin staining was predominantly seen in astrocytes (Fig. 2).

Localization of Merlin and RI␤ in Rat Hippocampal
Cells-In cultured rat hippocampal cells merlin was detected predominantly in neurons where it colocalized with RI␤ ( Fig.  3C). Merlin and RI␤ were colocalized in the cell body and partially along extensions in a punctate pattern. Double staining with the dendritic marker MAP2 indicated the presence of merlin in dendrites but also in axons, because some of the extensions positive for merlin were not positive for MAP2 (Fig.  3F). Weak merlin staining was also seen in astrocytes (Fig. 3G), where it colocalized with ezrin in the cell body and along the extensions. Delicate peripheral astrocyte processes stained strongly for ezrin but not merlin (Fig. 3I).
Coprecipitation of Merlin with RI␤ from Rat Brain by cAMP-Agarose Columns and Coimmunoprecipitation of Merlin and RI␤-The potential merlin-RI␤ interaction was studied with two types of coprecipitation experiments. Both merlin and RI␤ were shown to be present in the cytoskeletal fraction of rat brain lysate (Fig. 4, A and B). RI␤ could be precipitated from rat brain lysate by cAMP-agarose beads as expected. Merlin coprecipitated with the cAMP-agarose⅐RI␤ complex but not with control agarose beads (Fig. 4A). Furthermore, immunoprecipitation of RI␤ coprecipitated merlin, and in a reciprocal experiment, RI␤ was coprecipitated together with merlin. Again, neither merlin nor RI␤ precipitated with control beads (Fig. 4B).
Analysis of the Interaction of Transfected Merlin and RI␤-GFP-The localization and interaction between merlin and RI␤ was further studied by transfection experiments. In COS7 cells, transfected merlin and RI␤-GFP colocalized in regions under the cell membrane (Fig. 5). In RI␤-GFP-transfected cells perinuclear protein aggregates were also seen (Fig. 5B). Merlin immunoreactivity was seen in some of the aggregates (Fig. 5C), a localization not seen in cells transfected with merlin only (not shown).
The interaction was further studied by coimmunoprecipitation of transfected proteins. When COS7 cells, which express endogenous ezrin, were transfected with merlin and RI␤-GFP, merlin could be precipitated with the RI␤ antibody and, as previously shown (7), with the ezrin antibody (Fig. 6A). If COS7 cells were transfected with RI␤-GFP alone, ezrin was not coprecipitated with the RI␤ antibody (Fig. 6B). These results indicate that merlin and RI␤ are present in the same complex in transfected COS7 cells. In addition, ezrin can form a complex with merlin but not with RI␤. Merlin has been demonstrated to exist in three differentially phosphorylated forms with different mobility in SDS-PAGE (34). These forms (hypophosphorylated, phosphorylated, and hyperphosphorylated) can be identified in transfected 293 HEK cells. In subconfluent cells lysed in the presence of phosphatase inhibitors all three forms are present. When cells are grown to confluency the hypophosphorylated form becomes the most prominent (Fig. 6C). Coimmunoprecipitation experiments indicated that RI␤-GFP is specifically associated with the hypophosphorylated form of merlin. The amount of coprecipitated merlin was increased in confluent cells, in which the amount of the hypophosphorylated form was highest. Reciprocal experiments further verified the specific interaction between the proteins (Fig. 6C, bottom panel).
Mapping of the RI␤ Interaction Site in Merlin-The interaction site in merlin was mapped using three methods. Various merlin deletion constructs were expressed as GST fusion proteins and used as an affinity matrix to precipitate recombinant human RI␤. Merlin constructs containing residues 252-595 or 1-547 were able to precipitate RI␤, whereas GST or an irrelevant AKAP, GST-AKAP149, did not precipitate RI␤ (Fig. 7). This indicates a direct interaction between merlin and RI␤, and suggests that the interaction domain resides within the ␣-helical part of merlin. The interaction was further studied with yeast two-hybrid experiments using RI␤, RII␤, merlin, and ezrin constructs. The results demonstrated an interaction between RI␤ and merlin but not between RI␤ and ezrin. The Localization of merlin (A) and RI␤-GFP (B) was studied by immunofluorescence after staining with merlin 1398NF2 rabbit antiserum. Merlin and RI␤-GFP colocalize below the cell membrane. In RI␤-GFP transfected cells perinuclear protein aggregates are also seen. Merlin immunoreactivity is seen in some of the aggregates, a localization not seen in cells transfected with merlin only (C).
interaction was mapped to the ␣Ϫhelical part of merlin (Fig. 8). The S518D mutation mimicking the phosphorylated Ser-518 or the S518A mutation did not bind RI␤. Proline substitution within the potential PKA binding site, A468P, fully abolished the interaction and another substitution, L472P, reduced binding to RI␤ (Fig. 8). These constructs were, however, still active in other interaction assays (interaction tested with merlin 1-595 (wt), merlin 1-339, and ezrin 1-309, data not shown) (7). Full-length merlin did not bind RI␤ and RII␤ did not interact with any of the merlin or ezrin constructs.
The interaction site in merlin was further refined by an overlapping 20-mer peptide spot array encompassing part of the ␣-helical region of merlin. Recombinant radiolabeled RI bound to three overlapping peptides and thereby identified a 14-residue sequence critical for the interaction (Fig. 9).
The regulatory subunit binding regions in AKAPs consist of several conserved residues (28). The binding domain in merlin was aligned with 16 known AKAPs (Fig. 10). This alignment shows that the sequence between amino acids 463 and 480 of merlin contains several of the residues conserved in the R subunit binding domain of AKAPs. The amino-terminal part of the conserved region contains the interaction site identified by proline substitutions and peptide array. Comparison between merlin and D-AKAP1, another protein known to bind type I regulatory subunit, identified an additional six identical or conserved amino acids, which may be associated with the specificity of the interaction between these AKAPs and the regulatory subunits. DISCUSSION Merlin was originally identified as a tumor suppressor involved in development of neurofibromatosis 2-related tumors, schwannomas and meningiomas. Several reports have analyzed the tumor suppressor activity, but so far little is known of the potential of merlin for additional functions. On the other hand, an indispensable role in early development (29) and rather wide tissue expression pattern suggest yet unidentified functions for merlin. Of special interest is the central nervous system, in which high expression levels have been detected during and after organogenesis (11,13). Here we demonstrate that merlin is an interaction partner for the RI␤ regulatory subunit of PKA, a protein involved in neuronal signaling and synaptic plasticity.
The interaction between merlin and RI␤ could be demonstrated by several means. Coprecipitation experiments indicated that merlin and RI␤ are in the same complex in rat brain homogenates and in transfected cells. In addition, in vitro studies using purified proteins, yeast two-hybrid analysis, and peptide blot overlay studies showed that the interaction is direct. Finally, immunolocalization studies of tissue sections, cultured hippocampal cells, and transfected cells demonstrated that merlin and RI␤ are expressed in the same cell types and that their subcellular distribution is overlapping.
The interaction domain in merlin was localized within the ␣-helical part of the protein, in which no interactions have been reported so far. ERM proteins and merlin are regulated by an intramolecular interaction between the amino-and carboxylterminal domains, which results in masking of some protein interaction sites (7, 8, 30 -32). RI␤ does not interact with fulllength merlin in in vitro binding assays, but interacts with truncated proteins that mimic the open conformation of the protein, suggesting that the interaction is subject to conformational regulation. Recently, p21 activated kinase PAK was identified as a molecule that can activate merlin by phosphorylating a carboxyl-terminal Ser-518 (33,34). Merlin S518D did FIG. 10. Alignment of a putative regulatory subunit binding domain in merlin with binding domains in 16 known AKAPs. The alignment was done using the Clustal-W program. The names of the AKAPs and residue numbers of the binding domains are indicated on the right. Reversed text shows conserved residues with functional properties. Underlined residues in merlin and D-AKAP1 represent additional identical or conserved residues between the two AKAPs with capacity to bind RI subunits. In the alignment, the location of these residues is marked by stars. Arrows represent sites of introduced mutations that inhibit the interaction. not bind RI␤ in the yeast two-hybrid experiment, which could indicate that additional forms of activation are needed for this interaction to take place in vivo. Of the three phosphorylated forms of merlin the hypophosphorylated form was associated with RI␤ in coimmunoprecipitation experiments. This is the first report of a merlin interacting protein where the phosphorylated state of merlin is important for the interaction. The amount of hypophosphorylated merlin is increased in confluent cells, whereas the hyperphosphorylated merlin is increased after PAK phosphorylation on serine 518. Interestingly, a cross-talk between PKA and PAK has been demonstrated in some cell types and this cross-talk regulates mitogen-activated protein kinase signaling and cytoskeletal integrity. It will be important to further characterize the role of merlin in coordination of the two signaling pathways.
AKAPs function as multivalent scaffolds that assemble and integrate signals from multiple pathways. An AKAP should contain a sequence motif for binding regulatory subunits, an amphipathic helix with hydrophobic residues aligned along one face of the helix and charged residues along the other (28,35,36). Comparison with known AKAPs identified a sequence between residues 463 and 480 of the ␣-helical domain of merlin as a region, which shares most of the functionally relevant hydrophobic residues. Peptide binding experiments indicated a 14-residue sequence, which contains the amino-terminal part of this homologous region, as sufficient for the interaction. Proline substitution of a conserved residue in this region (Ala-468) abolished the interaction, and substitution of another conserved residue (Leu-472) in the vicinity inhibited binding to RI␤, further demonstrating the importance of this ␣-helical sequence in the interaction. The ERM protein ezrin has previously been shown to bind another regulatory subunit of PKA, RII␣. A signaling complex of ezrin, PKA, and PDZ-containing scaffolding proteins is thought to regulate the activity of ion transporters, indicating that ezrin is also a multivalent scaffold (37)(38)(39). The RII␣-binding region in ezrin was suggested to be a non-canonical 14-amino acid amphipathic ␣-helical within amino acids 373-439, although no further deletion/mutation analysis of the binding domain has been reported (40). The present sequence alignment shows that residues 413-430 of ezrin would best fulfill the criteria of a regulatory binding interphase. Interestingly, the present results show that ezrin does not interact with RI␤ and merlin does not bind to RII␣ (data not shown). Thus, the structurally related proteins with many common interaction partners demonstrate selectivity in their interactions with PKA subunits.
Although the functional significance of the merlin-RI␤ interaction is not clear, interesting possibilities exist. Hippocampal long term depression and potentiation are defective in mice carrying a targeted disruption of the gene encoding for RI␤ implicating a function for RI␤ in hippocampal plasticity (21,22). In contrast to other PKA regulatory subunits expressed in the central nervous system, the RI␤ gene has a characteristic pattern of expression (41) and despite a compensatory increase in total PKA activity, the RI␤ Ϫ/Ϫ mice have impaired hippocampal function suggesting a unique role for the RI␤ subunit (42). The RI␤ subunit may be involved in modulating the responsiveness of cells to a cAMP-mediated signal because PKA holoenzymes containing RI␤ are activated at lower concentrations of cAMP compared with a holoenzyme containing RI␣ (43). As merlin and RI␤ are expressed in the same cell types in the central nervous system, including the hippocampus, and form a complex there, one can speculate that merlin participates in neuronal PKA signaling. Further studies using animals with a targeted disruption of the NF2 gene are needed to further address this question.