Calcium/Calmodulin-dependent Protein Kinase II (CaMKII) Localization Acts in Concert with Substrate Targeting to Create Spatial Restriction for Phosphorylation*□S

Ca /calmodulin-dependent protein kinase II (CaMKII) acts in diverse cell types by phosphorylating proteins with key calcium-dependent functions such as synaptic plasticity, electrical excitability, and neurotransmitter synthesis. CaMKII displays calcium-dependent binding to proteins in vitro and translocation to synaptic sites after glutamatergic activity in neurons. We therefore hypothesized that subcellular targeting of CaMKII can direct its substrate specificity in an activitydependent fashion. Here, we examined whether activitydependent colocalization of CaMKII and its substrates could result in regulation of substrate phosphorylation in cells. We find that substrates localized at cellular membranes required CaMKII translocation to these compartments to achieve effective phosphorylation. Spatial barriers to phosphorylation could be overcome by translocation and anchoring to the substrate itself or to nearby target proteins within the membrane compartment. In contrast, phosphorylation of a cytoplasmic counterpart of the substrate does not require CaMKII translocation or stable protein-protein binding. Cytosolic phosphorylation is more permissive, exhibiting partial calcium-independence. Localization-dependent substrate specificity can also show more graded levels of regulation within signaling microdomains. We find that colocalization of translocated CaMKII and its substrate to lipid rafts in the plasma membrane can modulate the magnitude of phosphorylation. Thus, dynamic regulation of both substrate and kinase localization provides a powerful and nuanced way to regulate CaMKII signal specificity.

of cellular functions including cell division, differentiation, cardiac contraction, and synaptic plasticity (1). CaMKII is required for long term potentiation at CA1 synapses as well as in behavioral learning and memory tasks (2)(3)(4). A number of the prominent CaMKII substrates are membrane proteins. CaMKII phosphorylation of the AMPA receptor (AMPAR), ryanodine receptor, and several ion channels regulate their conductance properties (5)(6)(7). Thus, CaMKII has a wide repertoire of substrates and associated cellular functions, which raises the question of how CaMKII activity can discriminate among its many substrates. A possible point of regulation could be through the complex subcellular targeting of CaMKII.
Several mechanisms of CaMKII subcellular targeting have been characterized as CaMKII splice variants and isoforms can contain targeting sequences such as a nuclear localization sequence and actin binding domains (8,9). Membrane targeting can be mediated by assembly with proteins such as ␣CaMKII association protein (10). Recently, CaMKII has shown calcium/ calmodulin-dependent interactions with membrane proteins such as the NMDA receptor 2B subunit (NR2B) that can affect its distribution in heterologous cells (11,12). Activity-dependent CaMKII translocation in neurons has also been observed. Synaptic calcium influx through NMDA receptors in cultured, dissociated neurons and in vivo can cause rapid accumulation of CaMKII in post-synaptic sites (13,14). Furthermore, induction of long term potentiation results in increased post-synaptic density (PSD) association of autophosphorylated CaMKII that displays calcium-independent activity (15). Disruptions in the synaptic localization of CaMKII have been correlated with behavioral defects in transgenic mice carrying mutated CaMKII or lacking local CaMKII synthesis in dendrites (16,17). A synaptic localization and function is supported by the finding of 28 CaMKII substrates in the PSD, including receptors, cytoskeletal proteins, enzymes, and scaffolds (18) and is consistent with CaMKII regulation of cellular processes such as dendritic spine remodeling and AMPAR exocytosis that span several subcellular compartments (19,20).
The significance of CaMKII targeting for substrate phosphorylation has not been fully examined. Targeting to selected subcellular domains is crucial for effective cAMP-dependent protein kinase, protein kinase C, and MAPK signaling (21)(22)(23). For example, phosphorylation and subsequent regulation of L-type calcium channels by cAMP-dependent protein kinase requires expression of the appropriate cAMP-dependent pro- tein kinase anchoring protein, AKAP79, and targeting of the kinase to the channel (21). Although it is widely speculated that CaMKII translocation might result in similar signal specificity, this has not been directly tested. Furthermore, CaMKII is expressed at extremely high levels in the hippocampal neurons, estimated to be from 1 to 2% of total protein and is enriched in synaptic sites. Since the kinase is at such high concentration for a signaling protein, it is unclear whether dynamic localization would add much specificity since the kinase may be at saturating levels throughout such neurons.
In this work, we tested the hypothesis that access to certain substrates requires a directed translocation of CaMKII. We first determined whether phosphorylation of a substrate requires subcellular targeting of both the kinase and its substrate in a heterologous cell system in which both substrate and kinase concentrations are high. We targeted CaMKII to selected subcellular compartments using the calcium-dependent interaction between CaMKII and the C-terminal domain of NR2B (12). We showed that CaMKII targeting is required for effective phosphorylation of plasma membrane and ER substrates. By contrast, a cytoplasmic substrate does not require anchoring of the kinase for phosphorylation to occur. Furthermore, the cytoplasmic compartment seemed more permissive for phosphorylation at basal calcium levels. We also found that lipid raft microdomains can influence the degree of phosphorylation when both kinase and substrate are targeted to the plasma membrane. These results characterize for the first time how colocalization of CaMKII and its substrates to various cellular compartments regulates substrate phosphorylation.

EXPERIMENTAL PROCEDURES
Gene Construction-All Vim constructs ( Fig. 1) were derived from a PRK-5 vector containing a Myc-tag-labeled vimentin head. To generate the construct Vim-CFP-F, the coding sequence for EGFP was replaced by CFP through NheI and BamHI sites in the construct EGFP-F containing the farnesylation sequence from c-Ha-Ras (Clontech) CFP-F was then subcloned into NheI and BamHI of the cVim vector. CFP contains the A206K mutation to prevent any dimerization between fluorophores (24). mNR2B-Vim was generated by PCR amplifying the Myc-tagged vimentin head sequence with ApaI sites at the 5Ј ends of both sense and antisense primers. PCR products were subcloned into the ApaI site at 4613 of pRK5-NR2B.
cNR2B-Vim was generated by PCR from mNR2B-Vim and subcloned into pRK5. cNR2B was similarly generated by PCR from pRK5-NR2B. lyn-cNR2B was generated by PCR from pRK5-NR2B with the palmitoylation signal sequence from Lyn kinase. Yellow fluorescent protein tagged (YFP-CaMKII) was generated by replacing EGFP with YFP containing the A206K mutation (24) by PCR.
Cell Culture, Transfection, and Stimulation Protocols-HEK293 cells were maintained in Dulbecco's modified Eagle's medium with 10% fetal calf serum and were transfected with plasmids using Lipofectamine Plus TM (Invitrogen) according to the manufacturer's instructions. 24 h after the transfection, the cells were preincubated for 10 min with HBSS, 25 mM Hepes, pH 7.4, alone or 10 M methyl-␤-cyclodextrin (M␤CD, Sigma) and stimulated with 10 M ionomycin (Calbiochem) and 2 mM calcium in the absence or presence of M␤CD. Some were loaded with 5 M BAPTA-AM (Molecular Probes) in HBSS-Hepes for 20 min prior to fixation and immunostain. For GluR1 phosphorylation experiments, cells were preincubated for 30 min in 5 M bisindolyl-maleimide I (Calbiochem).
Immunocytochemistry and Immunoblotting-Cells were fixed with 4% paraformaldehyde in 100 mM phosphate buffer for 10 min followed by treatment with Ϫ20°C methanol for 10 min. They were then incubated with the phosphorylation-specific vimentin antibody MO82 (0.2 g/ml) diluted in phosphate-buffered saline with 2% normal goat serum. The cells were also counterstained with a polyclonal anti-Myc antibody (Santa Cruz Biotechnology) diluted 1:300 to assess expression level of vimentin constructs. The immunoreactivities were visualized by incubation with Alexa Fluor 594-conjugated anti-mouse antibodies and Alexa Fluor 647 anti-rabbit antibodies (Molecular Probes), and the samples were examined under a confocal microscope (Zeiss LSM-510). Images were quantitated on Metamorph by first producing a ratio image by dividing the phospho-vimentin image by the Myc image. All treatment ratio values were then normalized to control ratio values. Thresholding the Myc image to select for cells resulted in a mask that was then applied to the ratio image. For Vim-CFP-F experiments, intensity quantitation was complicated by the variable appearance of intracellular clusters of the Vim-CFP-F construct. Instead, the thresholded ratio image was blindly evaluated in a binary manner for the appearance of a visible plasma membrane ring of phospho-staining at least three times background levels. Immunoblotting was performed as described previously (12), using horseradish peroxidase-conjugated secondary antibodies and the ECL Western blotting detection system (Amersham Biosciences). Anti-Myc antibody was used at 1:750, MO82 was used at 0.2 g/ml, anti-␣CaMKII was used at 1:10,000, anti-phospho-Ser-831 (Upstate Biotechnology) was used at 1:500, anti-GluR1 (Chemicon) was used at 1:1000, and anti-GFP (Molecular Probes) was used at 1:1000. Membranes were stripped between blots with Restore Western blot stripping buffer (Pierce) as per the manufacturer's instructions.

RESULTS
We developed an assay to monitor CaMKII activity in different subcellular compartments. We utilized a protein sequence derived from vimentin to selectively measure CaMKII activity in cells (25). Vimentin is an intermediate filament protein, the assembly of which is regulated by phosphorylation of several protein kinases at multiple sites. One of these sites, Ser-82, has been characterized as a substrate for CaMKII, but not other kinases, including CaMKI, CaMKIV, protein kinase C, and cAMP-dependent protein kinase (25)(26)(27). Using an antibody against the phosphorylated form of Ser-82, ectopically expressed vimentin has proved to be a sensitive probe for CaMKII activity both in heterologous cells and in neurons (25). By attaching an 88-amino-acid segment containing Ser-82 derived from the N terminus of the vimentin head to proteins with known subcellular distribution, we measured CaMKII activity in the vicinity of the fusion protein by immunocytochemistry against phospho-vimentin.
CaMKII has been shown to bind to the NMDA receptor subunit NR2B in a calcium-dependent manner (11,12,28). Such targeting of the kinase may facilitate modulation of AM-PAR nearby at synaptic sites. We therefore designed a probe of CaMKII activity at membrane-targeted NMDA receptors. We inserted a Myc-tagged vimentin head fragment (1-97) at positions 1420 -1421 of NR2B beyond a second CaMKII binding site, one that is not autophosphorylation-dependent (11,12). We expressed the mNR2B-Vim construct together with GFP-CaMKII in HEK293 cells and showed that upon a calcium stimulation induced by ionomycin treatment, GFP-CaMKII translocates to mNR2B-Vim ( Fig. 2A, right panels). The pattern of mNR2B-Vim expression was consistent with previously reported ER membrane retention of wild-type NR2B in the absence of NR1 ( Fig. 2A) (29). We then examined Ser-82 phos- phorylation of the vimentin tag on NR2B and found that it was phosphorylated in a robust and calcium-dependent manner ( Fig. 2A, right panels, and 2B).
Calcium-ionomycin stimulation causes both CaMKII activation and translocation to membrane-localized mNR2B-Vim. To determine whether CaMKII translocation is required for mNR2B-Vim phosphorylation, we utilized a CaMKII point mutant, CaMKII(I205K), that has been previously shown to be defective in binding NR2B in vitro and unable to translocate to NR2B in situ ( Fig. 1A) (12) while retaining calcium-dependent activation and autophosphorylation (30). When GFP-CaMKII(I205K) is cotransfected with mNR2B-Vim, it has the same initial cytosolic distribution as with wild-type GFP-CaMKII (Fig. 3A, left panels). However, upon calcium-ionomycin stimulation, GFP-CaMKII(I205K) does not translocate to mNR2B-Vim, as reported previously (Fig. 3A, right panels) (12). Although the kinase should be activated under this stimulation protocol, immunostain with the Ser-82 phospho-antibody reveals no significant mNR2B-Vim phosphorylation over background levels (Fig. 3A, right panels, and 3B). Similar results were obtained from immunoblot of transfected cell lysates ( Supplementary Fig. 1). Taken together, these data demonstrate that translocation is necessary for phosphorylation of mNR2B-Vim.
Is translocation required for mNR2B-Vim phosphorylation because the substrate is anchored to a membrane? To investigate CaMKII phosphorylation of an equivalent cytosolic substrate, we created a cytosol-targeted mNR2B-Vim substrate (cNR2B-Vim) by generating a truncated construct (residues 1120 -1482) that lacks the membrane domain of NR2B but contains the CaMKII binding site and the vimentin tag (Fig. 1B). We cotransfected cells with cNR2B-Vim and GFP-CaMKII or GFP-I205K. Both wild-type and CaMKII (I205K) remained cytoplasmic in basal conditions and upon calcium-ionomycin stimulation (Fig. 4A,B). We compared cNR2B-Vim phosphorylation between wild-type and CaMKII (I205K) both with and without calcium-ionomycin stimulation and found them to be comparable (Fig. 4C). Since CaMKII (I205K) was equally competent in phosphorylating cNR2B-Vim, although it was unable to stably bind the NR2B tail, we conclude that unlike mNR2B-Vim, phosphorylation of cNR2B-Vim does not require CaMKII anchoring. Thus, when CaMKII and cNR2B-Vim are freely diffusible, phosphorylation does not require anchoring of the kinase to the substrate. Furthermore, these results indicate that the dependence of substrate phosphorylation on translocation demonstrated in Fig. 2 using the CaMKII (I205K) mutant is not due to an intrinsic inability of this mutant to recognize the vimentin substrate. These results suggest that CaMKII translocation can exert extremely tight regulation over substrate phosphorylation, provided that substrates are spatially restricted into subcellular compartments or less accessible conformations.
Phosphorylation of CaMKII substrates in cytosolic and membrane compartments under unstimulated (basal) conditions appears to be differentially regulated. HEK293 cells have low endogenous expression of CaMKII, which results in a barely detectable amount of basal phosphorylation for cNR2B-Vim and no phosphorylation above background for mNR2B-Vim (Fig. 5A). Basal phosphorylation is dependent on CaMKII concentration because both were elevated upon coexpression of GFP-CaMKII (Fig. 5B). However, although both mNR2B-Vim and cNR2B-Vim basal phosphorylation is increased by coexpression of the kinase, the effect is far more dramatic for cNR2B-Vim (Fig. 5, B and D, middle). Basal phosphorylation of cNR2B-Vim in the cytosol is much higher than basal phosphorylation of the mNR2B-Vim at the membrane. To elucidate whether basal phosphorylation was entirely due to continued calcium activation of CaMKII or to calcium-independent CaMKII activity, we determined whether phosphorylation is sensitive to BAPTA-AM treatment, which should eliminate most of the free intracellular calcium. A significant portion of basal cNR2B-Vim phosphorylation was reduced when cells were pretreated for 20 min with 5 M BAPTA-AM, but a significant fraction of the phosphorylation was BAPTA-resistant, perhaps due to a population of autophosphorylated CaMKII with calcium-independent activity (31). Interestingly, mNR2B-Vim phosphorylation did not display any sensitivity to BAPTA (Fig. 5, C and D, right), although it is possible that changes may have been obscured by the smaller magnitude of starting basal phosphorylation for this construct as compared with cNR2B-Vim. Similar results were obtained by the coexpression of vimentin constructs with GFP-CaMKII(I205K), suggesting that basal phosphorylation was not mediated by persistent activity due to CaMKII/NR2B anchoring (data not shown.) Taken together, these data indicated that the cytosolic com- partment can be much more permissive than membrane compartments for submaximal calcium signaling.
To further characterize the nature of localization-restricted phosphorylation, we investigated whether a docking interaction between kinase and substrate was required for membranedelimited phosphorylation or whether mere compartmental colocalization is sufficient. To this end, we constructed a cyan fluorescent protein-tagged farnesylated, plasma membrane-targeted vimentin substrate (Vim-CFP-F) that is not expected to interact with CaMKII or NR2B (Fig. 1B). Indeed, coexpression of Vim-CFP-F and YFP-CaMKII did not result in plasma membrane CaMKII translocation (data not shown). Similarly, coexpression of Vim-CFP-F, YFP-CaMKII, and cNR2B did not allow for plasma membrane CaMKII translocation (Fig. 6A, bottom row). Thus, we could observe whether CaMKII translocated to the plasma membrane via binding interactions with a plasma membrane-targeted NR2B C-tail enhances phosphorylation of an independent plasma membrane substrate that is capable of lateral mobility in the membrane. We tagged a portion of the NR2B C-tail (residues 1120 -1482) that had previously been shown to be sufficient to direct CaMKII subcellular localization (11) with an N-terminal palmitoylation signal from lyn kinase (lyn-cNR2B) to direct this construct to the plasma membrane (Fig. 1B). YFP-CaMKII was used instead of the GFP-CaMKII constructs in previous experiments to allow spectral distinction between YFP-labeled kinase and CFP-labeled vimentin substrate.
In the unstimulated state, YFP-CaMKII is cytosolic, and there is little basal phosphorylation of Vim-CFP-F, as expected based on previous findings with mNR2B-Vim and a cytosolic CaMKII. With lyn-cNR2B coexpressed to provide a membraneanchoring site for CaMKII and upon calcium stimulation, YFP-CaMKII translocates to the plasma membrane and phosphorylates Vim-CFP-F (Fig. 6A, top row). Phosphorylation of Vim-CFP-F required coexpression of a CaMKII plasma membranetargeting construct as both CaMKII translocation and Vim-CFP-F phosphorylation were absent if cytosolic NR2B tail (cNR2B) were substituted for lyn-cNR2B (Fig. 6A, bottom row). Cells assessed as positive for the appearance of plasma membrane phosphorylation demonstrated a clear requirement for lyn-cNR2B (Fig. 6B, see "Experimental Procedures"). These data demonstrate that, as seen for the ER, the plasma membrane also displays a localization requirement for CaMKII substrate phosphorylation. Furthermore, simple colocalization of kinase and substrate to the membrane compartment is sufficient to fulfill localization restrictions placed by membrane localization; direct anchoring interactions of kinase on substrate are not required. These findings have implications for known plasma membrane substrates of CaMKII that do not exhibit direct anchoring to the kinase. In the post-synaptic membrane, one such example is the AMPA receptor subunit GluR1, which can be phosphorylated by CaMKII on Ser-831 (32). To investigate whether a similar translocation-dependent phosphorylation by CaMKII can occur on this natural substrate, we immunoblotted lysates of cells cotransfected with hemagglutinin (HA)-GluR1, GFP-CaMKII, and cNR2B, lyn-

FIG. 3. No effective phosphorylation of mNR2B-Vim by GFP-CaMKII-(I205K) in transfected HEK293 cells. A,
HEK293 cells were cotransfected with mNR2B-Vim and GFP-CaMKII(I205K). Cells were either left in HBSS only or stimulated for 3 min with 10 M ionomycin and 2 mM CaCl in HBSS, fixed, and immunostained as before. GFP-CaMKII(I205K) is unable to translocate to mNR2B-Vim and cannot effectively phosphorylate vimentin upon calcium stimulation. pVim, phospho-vimentin. B, quantification of phospho-vimentin stain was as in Fig. 2. Ionomycin did not significantly increase phosphorylation of mNR2B-Vim as compared with control cells. Cells are representative of at least three independent experiments.

FIG. 4. Cytoplasmic substrate cNR2B-Vim is effectively phosphorylated by both GFP-CaMKII and GFP-CaMKII(I205K). A, cells were cotransfected with the indicated CaMKII and NR2B constructs.
Quantification of phospho-vimentin (pVim) stain was as in Fig. 2. GFP-CaMKII is capable of phosphorylating a cytoplasmic C-terminal NR2B vimentin fusion construct (cNR2B-Vim). B, after ionomycin treatment, GFP-CaMKII(I205K) is equally capable of phosphorylating cNR2B-Vim. C, quantification of experiments comparing GFP-CaMKII and GFP-CaMKII(I205K) phosphorylation of cNR2B-Vim following calcium-ionomycin treatment. Calcium-ionomycin significantly increased phosphorylation of cNR2B-Vim with both mutant and wild-type CaMKII (Student's t test; ***, p Ͻ 0.001). Cells are representative of at least three independent experiments. cNR2B, or full-length NR2B and probed for phosphorylation of Ser-831 after calcium-ionomycin stimulation. As Ser-831 is also phosphorylated by protein kinase C, cells were pretreated for 30 min with 5 M protein kinase C inhibitor bisindolylmaleimide I. Ser-831 phosphorylation was greatly enhanced if either lyn-cNR2B or NR2B was present but not with cNR2B (Supplemental Fig. 2). This suggests that translocation of CaMKII to a membrane anchor facilitates phosphorylation of membrane proteins, consistent with our vimentin reporter construct results.
Within a subcellular compartment, signaling microdomains can further influence access of enzymes to their substrates. Lipid rafts have been found to play important roles in organizing signaling at the plasma membrane, through clus-tering or segregating signaling components (33) Both Vim-CFP-F and lyn-cNR2B target to the plasma membrane through post-translational modifications (farnesylation and palmitoylation, respectively) on targeting sequences previously shown to be sufficient to direct proteins to lipid rafts (24,34). Thus, when CaMKII translocates to lyn-cNR2B, it should also be directed to lipid raft microdomains containing Vim-CFP-F. To determine whether translocation-mediated CaMKII phosphorylation was affected by raft localization of kinase and substrate, we examined phosphorylation of Vim-CFP-F by YFP-CaMKII translocated to lyn-cNR2B when rafts were disrupted by an M␤CD treatment that extracts membrane cholesterol (Fig. 6C). Raft disruption significantly reduced the efficacy of translocation-mediated phosphoryla- FIG. 5. Cytoplasmic compartment displays more permissive phosphorylation. Basal phosphorylation of mNR2B-Vim and cNR2B-Vim differ in degree and BAPTA sensitivity. A, in the absence of GFP-CaMKII coexpression, both constructs show very low basal vimentin phosphorylation. pVim, phospho-vimentin. B, cNR2B-Vim shows a significantly higher basal phosphorylation in the presence of GFP-CaMKII than NR2B-Vim. C, cNR2B-Vim basal phosphorylation is sensitive to, but not completely abolished, by 5 M BAPTA-AM loading. mNR2B-Vim basal phosphorylation is unaffected by BAPTA-AM. D, quantification of experiments comparing mNR2B-Vim and cNR2B-Vim basal phosphorylation. Cytoplasmic cNR2B-Vim is more readily phosphorylated in all conditions (one-way analysis of variance, Student-Newman-Keuls; ***, p Ͻ 0.001). Cells are representative of at least three independent experiments.
FIG. 6. Phosphorylation of a plasma membrane-targeted mNR2B-Vim requires translocation and is regulated by lipid rafts. A, cells were transfected with Vim-CFP-F, YFP-CaMKII, and cNR2B or lyn-cNR2B. Cells were stimulated and fixed as before. pVim, phosphovimentin. B, quantitation of three independent experiments shows significantly greater numbers of phosphorylated cells with translocation (Student's t test, *, p Ͻ 0.05). C, contribution of lipid rafts to translocation-facilitated phosphorylation. Cells were transfected with Vim-CFP-F, YFP-CaMKII, and lyn-cNR2B. Lipid rafts were disrupted with a 10-min 10 mM M␤CD treatment. Cells were stimulated with calcium-ionomycin as before and fixed. Percentage of cells with phosphorylated Vim-CFP-F is significantly affected by M␤CD treatment (Student's t test; *, p Ͻ 0.05). Cells are representative of at least three independent experiments. tion as the number of transfected cells with membrane Vim-CFP-F phosphorylation significantly decreased, but did not completely abrogate phosphorylation. M␤CD did not seem to affect calcium-ionomycin-induced activation of CaMKII or its ability to phosphorylate vimentin as M␤CD treatment did not affect phosphorylation of the cytosolic cNR2B-Vim by YFP-␣CaMKII (data not shown.) Thus, we find that the degree of phosphorylation can be affected by lipid raft localization of CaMKII and its substrate within the plasma membrane. DISCUSSION CaMKII is a prominent kinase in calcium signaling in many cell types and is known to regulate a wide range of substrates. Because CaMKII is involved in so many different signaling pathways, investigating the basis of CaMKII substrate specificity is fundamental to understanding how CaMKII can fulfill its many roles at the appropriate times. In this study, we found a tightly regulated relationship between CaMKII localization and substrate phosphorylation. Basically, phosphorylation of membranedirected substrates required anchoring of the kinase to the membrane domain. Both ER-and plasma membrane-localized substrates required CaMKII translocation to achieve effective phosphorylation, whereas a soluble version did not. These results have implications for several systems in which CaMKII and substrate localization appear to be regulated.
In cardiac myocytes, CaMKII is known to be important in regulation of calcium release from ER internal stores via phosphorylation of the ryanodine receptor and for regulation of the sarcoplasmic reticulum calcium ATPase via phosphorylation of ER protein phospholambin (6,35) and regulation of calcium flux through the L-type calcium channel (36). Our results suggest that CaMKII might need to either translocate to or stably associate with proteins in the ER and plasma membrane to achieve phosphorylation of these substrates. Consistent with this hypothesis, CaMKII has recently been found to complex with the ryanidine receptor in heart tissue (37). It will be of interest to further explore extensions of our findings in this system.
Dynamic translocation of activated CaMKII to post-synaptic sites has been well described in hippocampal neurons, but its specific function in CaMKII signal transduction remains elusive (38). Association of activated CaMKII with in vitro PSD preparations increases phosphorylation of its substrate GluR1 (39), consistent with a role for activity-dependent CaMKII targeting in regulating synaptic substrate phosphorylation. Our results indicate that especially for membrane-bound substrates, localization can regulate substrate phosphorylation with high fidelity so that a link may be reliably drawn between substrate phosphorylation and calcium activation of CaMKII. Furthermore, we find that translocated CaMKII can phosphorylate proteins within the membrane compartment without directly anchoring to them. This greatly expands the reach of translocated CaMKII within a compartment and may give greater flexibility to potential signaling partners even with a small repertoire of actual anchoring partners.
The requirement for CaMKII localization at membranes for effective substrate phosphorylation may result from high phosphatase activities in membrane and scaffold microdomains that would necessitate a higher concentration of CaMKII to counter phosphatase action. In neurons, localization-dependent regulation of dephosphorylation may play a major role in regulating CaMKII activity. PSD-associated CaMKII shows dramatically increased sensitivity to protein phosphatase I dephosphorylation (39), possibly through known neuronal protein phosphatase Itargeting subunits such as spinophilin, neurofilament-L, or neurabin (40 -42). Protein phosphatase I activity has been shown to effect CaMKII calcium ultrasensitivity (43) and has been hy-pothesized to form a switch in synaptic memory processes (44). Autophosphorylation-dependent binding to several membrane proteins, NMDA receptor subunits, and potassium channels has been described (45)(46)(47)(48). Since such binding can have long lasting effects, there may be a higher threshold for CaMKII activation and autophosphorylation at the membrane, designed to suppress spurious binding interactions under basal conditions. This may be achieved by higher phosphatase activity toward both phosphorylated substrates and CaMKII autophosphorylation. Supporting this hypothesis, our results show that basal phosphorylation of membrane substrates is very tightly regulated, unlike cytoplasmic substrate phosphorylation that occurs even under basal conditions.
In addition to acting as an all-or-none regulator of phosphorylation, localization-based specificity can display fine degrees of regulation by utilizing microdomains within subcellular compartments. There is intriguing preliminary evidence that lipid raft signaling may be used by neuronal CaMKII signaling pathways. Recently, CaMKII and many known CaMKII substrates and downstream signaling partners, such as PSD95, AMPAR subunits, synGAP and Ras (49,50), have been found to be present in neuronal dendritic lipid rafts. NMDAR subunits also appear in lipid raft fractions that could allow CaMKII to be dynamically recruited to raft fractions since CaMKII can display reversible translocation to the NR2B and has calcium-dependent binding interactions with NR1 and NR2A as well (45,46). In this study, we have found that lipid raft localization of CaMKII and its substrates can further amplify translocationdependent phosphorylation. In our system, lipid rafts act to enhance substrate phosphorylation after translocation. Because raft disruption did not prevent substrate phosphorylation entirely, there is theoretically room for further suppression of phosphorylation if CaMKII and its substrate were segregated in separate microdomains. It will be of interest to determine whether raft microdomains and their demonstrated effects on CaMKII phosphorylation are used by the neuron to modulate CaMKII signaling in vivo.
In contrast to membrane compartments, we found that the cytosolic compartment was highly permissive for cNR2B-Vim phosphorylation as cNR2B-Vim displayed greater than 2-fold calcium-ionomycin-stimulated phosphorylation over its tethered, full-length mNR2B-Vim counterpart. Furthermore, basal phosphorylation of cNR2B-Vim was 3-fold higher than mNR2B-Vim and was only partially sensitive to BAPTA-AM (Fig. 4C). Our experiments do not reveal whether this is due solely to calcium-independent CaMKII autonomous activity or whether differential phosphatase activity across subcellular compartments also plays a role. However, it is known that CaMKII can display up to ϳ30% autonomous activity in both brain and cultured neurons that is resistant to most activity blockers (51). If neurons maintain a tonic level of CaMKII activity in the cytosol, our results indicate that this may be sufficient to also maintain a significant phosphorylation of cytosolic substrates, without affecting phosphorylation of membrane/tethered substrates.
Consistent with permissive signaling in the cytosolic compartment, cNR2B-Vim displayed marked phosphorylation in unstimulated cells, whereas mNR2B-Vim had little to no basal phosphorylation. This basal phosphorylation was increased by exogenous CaMKII coexpression and decreased by BAPTA application. Thus, diffusible cytosolic proteins are extremely sensitive to both calcium-dependent and -independent forms of basal CaMKII activity. This may arise from highly efficient diffusional access or to relatively reduced phosphatase activity in the cytosolic compartment. The cytosolic compartment may be well suited for sensitive and rapid CaMKII signal propaga-tion. However, a high amount of basal substrate phosphorylation also poses a signal-to-noise problem for cell signaling, which may be resolved by further substrate scaffolding within the cytosol. Perhaps permissive phosphorylation in the cytosol depends on relative free diffusion of substrates. Would cytosolic substrates become more tightly regulated if they were bound in a scaffolded macromolecular complex? Evidence from the MAPK pathway in yeast implicates macromolecular complexes in generating signaling specificity by allowing for relatively sparse signaling molecules to more easily find one another (52). It may be for abundantly expressed signaling molecules such as CaMKII, where cytoplasmic signaling may not be limited by the ability of the enzyme to find its substrate, that macromolecular scaffolds may protect substrates from spurious phosphorylation by soluble CaMKII by adding steric or diffusional constraints. In cells where CaMKII is not so abundantly expressed, cytoplasmic phosphorylation may more closely resemble the MAPK model, in which basal phosphorylation is low and stimulated phosphorylation is greatly enhanced by scaffolding. Indeed, we found that basal phosphorylation of cNR2B-Vim by endogenous CaMKII was very low and the appearance of robust basal phosphorylation required overexpression of CaMKII. It may be possible for the cell to switch between permissive and restrictive cytoplasmic signaling by regulating the level of CaMKII expression.
Our present work demonstrates that indeed, regulated CaMKII localization can act as a permissive gate for substrate phosphorylation in ER and plasma membrane compartments, and furthermore, can show more subtle modulatory regulation within plasma membrane raft microdomains. Cytosolic regulation is tuned to CaMKII and perhaps also substrate levels and may be able to display more than one mode of regulation. This localization-based signaling model combines existing signaling microdomains, substrate scaffolding, and stimulation-dependent localization of CaMKII activity to create substrate specificity, thus providing a link between the unique molecular properties of CaMKII and its effector functions on substrate subsets. Recently, work from several laboratories has demonstrated that CaMKII translocation also results in fundamental changes in the molecular mechanisms of activation and inactivation of CaMKII (12,48,53). It will be of great interest to examine how our findings on pure localization-driven modulation of CaMKII substrate phosphorylation functions in the context of anchoring partner-driven alterations to basic CaMKII biochemical properties.