A novel mechanism for Ca2+/calmodulin-dependent protein kinase II targeting to L-type Ca2+ channels that initiates long-range signaling to the nucleus

Neuronal excitation can induce new mRNA transcription, a phenomenon called excitation–transcription (E-T) coupling. Among several pathways implicated in E-T coupling, activation of voltage-gated L-type Ca2+ channels (LTCCs) in the plasma membrane can initiate a signaling pathway that ultimately increases nuclear CREB phosphorylation and, in most cases, expression of immediate early genes. Initiation of this long-range pathway has been shown to require recruitment of Ca2+-sensitive enzymes to a nanodomain in the immediate vicinity of the LTCC by an unknown mechanism. Here, we show that activated Ca2+/calmodulin-dependent protein kinase II (CaMKII) strongly interacts with a novel binding motif in the N-terminal domain of CaV1 LTCC α1 subunits that is not conserved in CaV2 or CaV3 voltage-gated Ca2+ channel subunits. Mutations in the CaV1.3 α1 subunit N-terminal domain or in the CaMKII catalytic domain that largely prevent the in vitro interaction also disrupt CaMKII association with intact LTCC complexes isolated by immunoprecipitation. Furthermore, these same mutations interfere with E-T coupling in cultured hippocampal neurons. Taken together, our findings define a novel molecular interaction with the neuronal LTCC that is required for the initiation of a long-range signal to the nucleus that is critical for learning and memory.

Neuronal excitation can lead to activity-dependent gene expression, a process known as excitation-transcription (E-T) 4 coupling (1,2). E-T coupling is thought to be essential for learning and memory consolidation (3). In particular, phosphorylation of CREB transcription factor at Ser 133 is a key regulatory step in the transcription of several immediate early genes encoding proteins, such as c-Fos, BDNF, Arc, and Homer1a (4), and is important for synaptic plasticity and long-term memory (5,6). Multiple signaling mechanisms have been implicated in CREB phosphorylation and E-T coupling, and their relative roles appear to depend on the cell type, the specific stimulating signal, and the strength and duration of stimulation (7)(8)(9).
One major pathway to trigger CREB phosphorylation at Ser 133 is the activation of L-type Ca 2ϩ channels (LTCCs) (10 -12). LTCC activation can induce global increases of neuronal Ca 2ϩ concentrations. However, at least under some conditions, the initiation of LTCC signaling to trigger nuclear CREB phosphorylation appears to require increased Ca 2ϩ concentrations only within a nanodomain in the immediate vicinity of the channel. In this paradigm, Ca 2ϩ binds to the ubiquitous Ca 2ϩ sensor, calmodulin, within the LTCC nanodomain, and Ca 2ϩ / calmodulin then translocates to the nucleus to induce CREB phosphorylation (13,14).
Molecular and pharmacological studies have provided robust evidence that Ca 2ϩ /calmodulin-dependent protein kinase II (CaMKII) is required for LTCC-dependent E-T coupling. CaMKII is specifically recruited to neuronal LTCCs during E-T coupling (11,15). In fact, recent studies indicate that E-T coupling requires precisely coordinated recruitment and activation of two CaMKII holoenzymes within the LTCC nanodomain (10). CaMKII has been reported to directly interact with multiple proteins within LTCC complexes, including the pore-forming ␣1 (Ca V 1.2 or Ca V 1.3) subunits, auxiliary ␤1 or ␤2 subunits, and associated scaffolding proteins, such as densin (16 -20). CaMKII interactions with ␤2 and densin play a role in modulating Ca 2ϩ -dependent facilitation of Ca V 1.2 and Ca V 1.3 LTCCs, respectively (16,19). However, the roles, if any, of these interactions in E-T coupling are unclear.
Here we identify a novel direct interaction between CaMKII and the N-terminal domain (NTD) of LTCC ␣1 subunits. Residues in the Ca V 1.3 NTD and in CaMKII that are critical for the interaction were identified by site-directed mutagenesis. Finally, these mutated proteins were used to show that CaMKII-NTD interaction is important for the association of CaMKII with Ca V 1. 3

Ca V 1.3 NTD directly binds activated CaMKII
CaMKII has been suggested to interact with the pore-forming ␣1 subunits of Ca V 1.2 and Ca V 1.3 LTCCs (17)(18)(19)(20), although there are conflicting data about the specific domains involved. To address this question, we expressed and purified a family of GST fusion proteins containing each intracellular domain of the Ca V 1.3 ␣1 subunit (Fig. 1A). Despite extensive efforts to further optimize the experimental conditions, some of these proteins were partially degraded; however, full-length proteins were readily detected (asterisks in Fig. 1, B and C). Because direct interactions of CaMKII with several neuronal proteins are differentially modulated by CaMKII activation (21)(22)(23), we tested for direct binding of these GST fusion proteins to purified mouse CaMKII␣ following preincubation to induce different conformations. There was no consistently detectable binding of inactive CaMKII to any of the intracellular domains above the level of GST control, but the Ca V 1.3 NTD directly and specifically interacts with activated CaMKII conformations induced by pre-autophosphorylation at Thr-286 (Fig. 1B) or by the binding of Ca 2ϩ /calmodulin and ADP (Fig. 1C). The fact that binding of Ca 2ϩ /calmodulin and ADP to CaMKII is sufficient to induce interaction with the NTD shows that Thr-286 phosphorylation is not necessary for binding; rather, an open, activated conformation of CaMKII is required. Thus, these data show that activated CaMKII directly interacts with the NTD of Ca V 1.3 with very high selectivity.

CaMKII specifically binds to the LTCC NTDs
To investigate the specificity of this novel CaMKII interaction, we compared the amino acid sequences of NTDs from all 10 human VGCC ␣1 subunits. The NTDs are quite divergent in the initial membrane-distal sections but become more conserved in the membrane-proximal region ( Fig. 2A). A similar conservation pattern holds true for mouse and rat VGCCs. To test binding specificity, we expressed and purified GST-tagged NTDs from Ca V 1.2, Ca V 2.2, and Ca V 3.2. As noted above, preactivated purified CaMKII robustly interacts with the Ca V 1.3 NTD, and there was a slightly weaker interaction with the Ca V 1.2 NTD. However, interaction of preactivated CaMKII with the Ca V 2.2 and Ca V 3.2 NTDs was barely detected above the GST negative control (Fig. 2, B and C). These data show that activated CaMKII selectively interacts with NTDs of the LTCCs.

Molecular determinants for Ca V 1.3 NTD interaction with CaMKII
Previous studies indicate that the Ca V 1.2 and Ca V 1.3 NTDs contain conserved binding sites for calmodulin (residues Ser 52 -Lys 64 in Ca V 1.3), termed NSCaTE (24,25), and for CaMKII (Lys 110 -Trp 123 in Ca V 1.2) (20) (Fig. 3A). To investigate the potential roles of these domains in the CaMKII binding detected here, we mapped the site of direct CaMKII interaction in the Ca V 1.3 NTD. There was no detectable interaction between preactivated CaMKII and the membrane-distal fragment (NT-A; amino acids 1-68) containing the NSCaTE domain (Fig. 3B), but the membrane-proximal fragment (NT-B; amino acids 69 -126) robustly interacts with preactivated CaMKII. Further dissection of NT-B revealed that preactivated CaMKII interacts with a GST-tagged fragment contain-  (24)) (purple box)-binding and CaMKII (20) (white box)-binding domains in the NTD, the ␣ subunit interaction domain (AID, for ␤ subunit interaction) in the I/II linker (50) (blue box), and overlapping calmodulin-and CaMKIIbinding sites in the CTD (17) (green box). B, glutathione-agarose co-sedimentation assays show that there is no reliably detectable interaction of inactive (non-autophosphorylated) conformations of CaMKII␣ with any of the Ca V 1.3 intracellular domains but that activated (pre-autophosphorylated) CaMKII␣ specifically binds to the NTD. C, activation of CaMKII␣ by binding of Ca 2ϩ /calmodulin and Mg-ADP is sufficient for interaction with the Ca V 1.3 NTD. The immunoblots shown are representative of three independent experiments.
The amino acid sequence of Ca V 1.3 residues 69 -93 shares little identifiable similarity with known CaMKII-binding domains in other proteins (16,21,26). However, we identified three basic amino acids in Ca V 1.3 (Arg 83 -Lys 84 -Arg 85 ) that are largely conserved in NTDs of Ca V 1.2 and other LTCC ␣1 subunits, but not in the NTDs of Ca V 2.2 or Ca V 3.2 (which do not bind CaMKII) or of other ␣1 subunits ( Fig. 2A). Replacement of this RKR motif with three alanines in the Ca V 1.3 NTD almost completely abrogated binding of preactivated CaMKII (Fig.  3D). These data identify three amino acids in the Ca V 1.3 NTD that are required for strong and direct in vitro interactions with preactivated CaMKII.
Preferential interactions of preactivated CaMKII with several other CaMKII-associated proteins (CaMKAPs) are mediated by the catalytic domain. Therefore, to identify CaMKII␣ residues critical for binding to the Ca V 1.3 NTD, we screened previously characterized as well as novel CaMKII mutations in the catalytic domain (Fig. 4A) using a fluorescence-based 96-well plate binding assay (see "Experimental procedures"). An I205K mutation, previously shown to disrupt binding to GluN2B and the densin-IN domain (26,27), also reduced binding to the Ca V 1.3 NTD by ϳ80% (Fig. 4B). We identified two additional CaMKII mutations (V102E and E109K) that also significantly interfere with binding to the Ca V 1.3 NTD (Fig. 4B), whereas another mutation (Y210E) had no significant impact. Strikingly, the CaMKII␣-V102E mutation had no significant effect on binding to the ␤2a subunit of VGCCs, to the densin-IN or -CTD domains (Fig. 4C), or to GluN2B (not shown). In combination, these data suggest that the mechanism underlying binding of activated conformations of CaMKII to the Ca V 1.3 NTD is partially distinct from the mechanisms for binding to other known CaMKAPs.

The Ca V 1.3 NTD is important for CaMKII association with LTCC complexes
To begin to address the importance of the Ca V 1.3 NTD in CaMKII targeting to LTCC complexes, we first directly com-

Nuclear signaling by L-type Ca 2؉ channel and CaMKII
pared CaMKII binding to GST fusion proteins containing the Ca V 1.3 NT-B1 fragment or previously defined minimal CaMKII-binding domains in the VGCC ␤2 subunit (residues 485-505) (16) and a Ca V 1.2 CTD fragment (residues 1639 -1660) that is fully conserved in Ca V 1.3 (17). Although we did not detect CaMKII binding to the full-length Ca V 1.3 CTD (Fig.  1, B and C), we rationalized that the full-length CTD may adopt a conformation that prevents CaMKII interaction with this previously defined domain, perhaps due to binding of a C-terminal modulatory domain to the calmodulin-binding IQ domain (28). Similar levels of preactivated CaMKII bound to the Ca V 1.3 NT-B1 and ␤2 (residues 485-505) fragments, but we could not detect an interaction with the Ca V 1.3 CTD (residues 1639 -1660) fragment under these conditions (Fig. 5A). Nevertheless, these data indicate that CaMKII can directly interact with multiple components of native LTCC complexes.
The most abundant VGCC auxiliary ␤ subunit in the brain appears to be ␤3 (29). Therefore, we investigated whether the NTD is important for CaMKII targeting to Ca V 1.3 LTCC complexes containing the ␤3 subunit. CaMKII␣, ␤3, and the ␣2␦ subunit were co-expressed in HEK293T cells with the HAtagged WT Ca V 1.3 ␣1 subunit or with chimeric ␣1 subunits in which the Ca V 1.3 NTD was replaced with NTDs from either Ca V 2.2 or Ca V 3.2 NTD (Fig. 5B). Antibodies to the HA tag were then used to immunoprecipitate ␣1 subunits from aliquots of the same cell lysate in the presence of EDTA or following the addition of excess Ca 2ϩ /CaM and Mg 2ϩ -ATP to activate CaMKII␣ (see "Experimental procedures"). Immunoblotting revealed that the HA immune complexes isolated in the presence of Ca 2ϩ /calmodulin/Mg 2ϩ -ATP contained significantly more HA-Ca V 1.3 ␣1 subunit and CaMKII␣, relative to immune complexes isolated in parallel in the presence of excess EDTA. Similar data were obtained from two independent sets of experiments (Fig. 5, C and D). Combining the quantitative analysis of these two data sets revealed that Ca 2ϩ /CaM/Mg 2ϩ -ATP increased the levels of HA-Ca V 1.3 and CaMKII␣ by 2.5 Ϯ 0.3and 12.8 Ϯ 2.6-fold, respectively (mean Ϯ S.E., n ϭ 7; p Ͻ 0.01 for both, one-sample t test compared with a theoretical value of 1.00 indicating no change). Therefore, the addition of Ca 2ϩ / calmodulin/Mg 2ϩ /ATP significantly increases the ratio of CaMKII to HA-Ca V 1.3 in the immune complexes by 6.7 Ϯ 2.3fold (n ϭ 7; p Ͻ 0.001).
To explore the mechanism underlying these changes, we first found that the addition of Ca 2ϩ /calmodulin/Mg 2ϩ /ATP failed to increase the levels of immunoprecipitated HA-Ca V 1.3 in the absence of co-expressed CaMKII␣ (data not shown). Moreover, replacement of the entire Ca V 1.3 NTD with corresponding NTDs from Ca V 2.2 or Ca V 3.2 ( Fig. 5B) abrogated Ca 2ϩ /calmodulin/Mg 2ϩ /ATP-induced increases in the levels of both HA-tagged channels and CaMKII␣ in the HA immune complexes, as well as in the CaMKII␣/HA-Ca V 1.3-chimera ratio (Fig. 5C). Similarly, significant Ca 2ϩ /calmodulin/Mg 2ϩ /ATPinduced increases in levels of CaMKII and HA-Ca V 1.3 and in the CaMKII/HA-Ca V 1.3 ratio in HA immune complexes were prevented by the deletion of residues 69 -93 or by mutation of 83 RKR 85 to AAA within the Ca V 1.3 NTD (Fig. 5D). Finally, the CaMKII␣-V102E mutation also prevented Ca 2ϩ /calmodulin/ Mg 2ϩ /ATP-induced increases in levels of HA-Ca V 1.3 and  (51)). Right, a single CaMKII␦ subunit in an activated conformation (displaced regulatory domain) with a bound inhibitor (SU6656, yellow) (PDB entry 2WEL (52)). The catalytic and regulatory domains are shown in gray and pink, respectively. For clarity of presentation, C-terminal holoenzyme association domains are not shown, and the displaced regulatory domain with bound Ca 2ϩ /calmodulin is not shown in PDB entry 2WEL. Thr 286 and Thr 305 (green) are two regulatory autophosphorylation sites. Mutation of Ile 205/206 (orange) to Lys disrupts CaMKII interaction with GluN2B and densin-IN (22,26), whereas mutation of Asp 238/239 (cyan) to Arg disrupts GluN2B binding but spares interactions with densin-IN (22). A naturally occurring de novo Glu 183 (purple) to Val mutation in CaMKII␣ is linked to autism spectrum disorder and disrupts CaMKII interaction with multiple CaMKAPs (48,53). B, a 96-well glutathione plate assay to screen activated mApple-tagged CaMKII␣ mutants for interactions with GSTtagged Ca V 1.3 NTD. C, binding of activated mApple-tagged WT and V102E-CaMKII␣ to multiple GST-CaMKAP proteins in the 96-well plate assay. A V102E mutation selectively disrupts CaMKII␣ binding to the Ca V 1.3 NTD; Val 102/103 is highlighted in red in A. Data from three independent experiments were analyzed by one-way ANOVA (for B) and two-way ANOVA followed by Sidak's multiple-comparison test (for C), respectively. ***, p Ͻ 0.001; ns, not significant (p Ͼ 0.05). Error bars, S.E.

Nuclear signaling by L-type Ca 2؉ channel and CaMKII
CaMKII␣ and in the CaMKII␣/HA-Ca V 1.3 ratio in HA immune complexes (Fig. 5D). Because all of the molecular changes tested here disrupt the CaMKII-NTD interaction in vitro, these data collectively indicate that CaMKII␣ interaction with the NTD is required for activity-dependent association of CaMKII␣ with intact HA-Ca V 1.3 complexes as well as for a more modest increase in the levels of immunoprecipitated HA-Ca V 1.3.

Figure 5. The NTD is important for CaMKII association with LTCC complexes.
A, preactivated CaMKII␣ robustly interacts with the minimal CaMKII-binding sites from the Ca V 1.3 NTD and the ␤2 auxiliary subunit, but not with a previously reported minimal CaMKII-binding site in the Ca V 1.2 CTD that is identical in Ca V 1.3. B, a schematic diagram showing the structure of chimeric Ca V x.x NTD-Ca V 1.3 channels in which the Ca V 1.3 NTD was substituted by NTDs from Ca V 2.2 or Ca V 3.2. C, equal aliquots of lysates from cells expressing CaMKII␣ with WT or NTD chimeric HA-tagged Ca V 1.3s were immunoprecipitated using anti-HA antibodies without (EDTA) or with the addition of excess Ca 2ϩ /calmodulin/Mg 2ϩ -ATP. C2 plots levels of immunoprecipitated HA-Ca V 1.3 proteins (black) and CaMKII (purple) in the presence of Ca 2ϩ /calmodulin/Mg 2ϩ /ATP normalized to levels isolated in the presence of EDTA in each experiment. C3 compares levels of immunoprecipitated CaMKII␣ normalized to immunoprecipitated HA proteins in the presence of EDTA and Ca 2ϩ /calmodulin/Mg 2ϩ -ATP. D, similar analysis of the co-immunoprecipitation of WT or V102E-CaMKII␣ with WT, ⌬69 -93, or RKR-AAA HA-Ca V 1.3 in the presence of EDTA or Ca 2ϩ /calmodulin/Mg 2ϩ /ATP. Levels of immunoprecipitated HA-Ca V 1.3 and CaMKII␣ are compared in D2, and normalized CaMKII␣/HA-Ca V 1.3 ratios are shown in D3. Data are from 3-4 independent experiments and analyzed by two-way ANOVA followed by Sidak's multiple-comparison test. *, p Ͻ 0.05; ***, p Ͻ 0.001; ns, not significant (p Ͼ 0.05). Error bars, S.E.

Ca V 1.3 NTD is required for LTCC-and CaMKII-mediated nuclear signaling
We then tested whether the NTD-CaMKII interaction is important for LTCC-mediated downstream signaling to increase Ser 133 phosphorylation of the CREB transcription factor in primary cultures of hippocampal neurons. We first established a stimulation paradigm to induce LTCC-dependent increases of Ca 2ϩ concentrations in neurons based on Fura-2based Ca 2ϩ imaging. Neurons were preincubated in 5 mM K ϩ Tyrode's solution containing APV and CNQX to block the activation of NMDA-and AMPA-type glutamate receptors and with tetrodotoxin (TTX) to inhibit voltage-dependent sodium channels. Neuronal depolarization by replacing the solution with 40 mM K ϩ Tyrode's solution in the presence of APV, CNQX, and TTX induced a significant increase in intracellular (somatic) Ca 2ϩ , which is reduced by ϳ80% in the presence of 10 M nimodipine, a highly selective LTCC antagonist (Fig. 7A). In parallel, we showed that depolarization with 40 mM KCl in the presence of APV, CNQX, and TTX for 90 s induces a robust increase of nuclear staining using a phospho-Ser 133 -specific CREB antibody (pCREB staining) that can be completely blocked by 10 M nimodipine (Fig. 7, B and C), consistent with previous findings (15,30,31).
We then used a pharmacological knock-in approach (32) to compare the E-T coupling efficiency of wild-type Ca V 1.3 with Ca V 1.3-⌬69 -93, which compromises CaMKII binding. We expressed an HA-tagged nimodipine-resistant Ca V 1.3 mutant (T1033Y, designated as Ca V 1.3 DHPR ) to allow for activation of exogenous Ca V 1.3 channels while using nimodipine to block all of the endogenous Ca V 1.2 and Ca V 1.3 LTCCs. The expression of Ca V 1.3 DHPR almost completely rescued the increase of pCREB staining in the presence of nimodipine (transfected neurons were identified by staining for the HA epitope), but this rescue of pCREB signaling was disrupted by the deletion of NTD residues 69 -93 from Ca V 1.3 DHPR (Fig. 7C), which prevents CaMKII binding.
To complement these findings, we examined the impact of the CaMKII␣-V102E mutation using an shRNA knockdown and rescue strategy that has been used previously to demonstrate a key role for CaMKII␣/␤ (15). We first verified that shRNA knockdown of CaMKII␣ and CaMKII␤ expression had no significant effect on LTCC-dependent somatic Ca 2ϩ responses to stimulation with 40 mM K ϩ Tyrode's solution. Moreover, re-expression of shRNA-resistant wild-type CaMKII␣ (CaMKII␣ R -WT) or CaMKII␣ R -V102E also did not alter the amplitude or kinetics of Ca 2ϩ responses (Fig. 8A). However, CaMKII␣/␤ knockdown significantly attenuated the increase of pCREB staining induced by 40 mM KCl, and the effect of this knockdown was largely rescued by re-expression of CaMKII␣ R -WT but not by CaMKII␣ R -K42R (a kinase-dead mutant). Notably, CaMKII␣ R -V102E, which cannot bind to the Ca V 1.3 NTD but retains full kinase activity (data not shown), was also unable to rescue the pCREB staining (Fig. 8B). In contrast, pilot experiments using two independent neuronal cultures found that increases of nuclear pCREB staining elicited using a longer (5 min) and more robust (identical except lacking CNQX) stimulation paradigm were partially independent of LTCCs (5 mM KCl control: 0.00 Ϯ 0.02, n ϭ 56 cells; 40 mM KCl: 1.00 Ϯ 0.05, n ϭ 68; 40 mM KCl plus 10 M nimodipine: 0.51 Ϯ 0.05, n ϭ 71; mean Ϯ S.E., p Ͻ 0.01, one-way ANOVA followed by Tukey's multiple-comparison test) and were independent of CaMKII␣/␤ (control/untransfected cells: 1.00 Ϯ 0.10, n ϭ 50; CaMKII␣/␤ shRNA: 0.74 Ϯ 0.27, n ϭ 15; mean Ϯ S.E., p ϭ 0.25, unpaired Student's t test). Taken together, data obtained by expressing nimodipine-resistant channels and using CaMKII knockdown/rescue approaches suggest that the NTD-CaMKII interaction is essential for LTCC-mediated E-T coupling.   A, CaMKII␣/␤ knockdown or re-expression/rescue has no effect on high K ϩ -induced increases of somatic Ca 2ϩ . Cultured hippocampal neurons were transfected with mApple only (n ϭ 17), mApple with CaMKII␣/␤ shRNA (n ϭ 23), mApple/CaMKII␣/␤ shRNA with shRNA-resistant CaMKII␣ R -WT (n ϭ 26), or CaMKII␣ R -V102E cDNA (n ϭ 17), respectively. Transfected neurons were then monitored for Ca 2ϩ influx in response to 40 mM K ϩ -induced depolarization (black arrow; see Fig. 7A). B, the CaMKII␣-V102E mutant does not support nuclear signaling. Expression of CaMKII␣/CaMKII␤ shRNAs significantly reduces nuclear CREB phosphorylation following 40 mM KCl treatment. This reduction in CREB phosphorylation is rescued by co-expression of shRNA-resistant CaMKII␣ R -WT, but not CaMKII␣ R -K42R (kinase-dead) or CaMKII␣ R -V102E (deficient in Ca V 1.3 NTD binding). Each data point represents analysis of a single cell collected from 3-4 independent neuronal cultures/transfections. Pooled data were analyzed by one-way ANOVA followed by Tukey's multiple-comparison test. ***, p Ͻ 0.001; ns, not significant (p Ͼ 0.05). All confocal images show a 40 ϫ 40-m area. Error bars, S.E.

Discussion
Ca 2ϩ influx into neurons via ligand-and voltage-gated Ca 2ϩ channels plays a key role in a variety of processes, including synaptic plasticity and long-term memory formation. Elucidation of the molecular mechanisms that underlie the specificity and efficiency of signaling downstream of the different channels is critical to understanding their biological roles. Our data identify a novel CaMKII binding site in the NTD of LTCCs that is important for the coupling of LTCCs to a nuclear response. The specificity of this interaction for Ca V 1 LTCCs over Ca V 2 and Ca V 3 VGCCs presumably contributes to their preferential role in E-T coupling (30).
We systematically compared CaMKII binding to all intracellular domains of the Ca V 1.3 ␣1 subunit, expressed as GST fusion proteins. Activated CaMKII bound with high specificity to the Ca V 1.3 NTD. This interaction was conserved with Ca V 1.2 NTDs, but not with NTDs from Ca V 2.2 or Ca V 3.2 VGCCs. The fact that we failed to detect CaMKII binding to the full-length Ca V 1.3 CTD or to the previously defined minimal CTD CaMKII-binding domain that is 100% conserved between Ca V 1.2 and Ca V 1.3 was somewhat surprising, because it was previously reported that CaMKII binding to the CTD is important for Ca 2ϩ -dependent facilitation (17). Another prior study detected a putative CaMKII-binding site in the membraneproximal region of the Ca V 1.2 NTD (residues 110 -123) that is conserved in Ca V 1.3 based on co-immunoprecipitations from cell lysates and found that mutation of this motif disrupted membrane trafficking yet enhanced Ca 2ϩ influx (20). However, this study provided no evidence for a direct interaction of CaMKII with this domain (e.g. with purified proteins). Moreover, neither of these prior studies established that the domains identified are important for CaMKII association with intact LTCC complexes. Reasons for the discrepancies between the present studies and these prior studies are unclear, but it is possible that the conditions used here favor detection of more specific, high-affinity, direct interactions.

Comparison of NTD with previously identified CaMKII-binding domains
Like several other CaMKAPs, the LTCC NTD preferentially interacts with activated CaMKII. However, the molecular bases for these interactions appear to be distinct. CaMKII-binding domains in the NMDA receptor GluN2B subunit and ␤1/␤2 subunits of VGCCs share sequence similarity with the CaMKII autoregulatory domain, including the presence of a (auto)phosphorylation site (16). However, these domains share no sequence similarity with the internal CaMKII-binding domain in densin, which resembles CaMKIIN, a naturally occurring CaMKII inhibitor protein (22). Moreover, neither of these classes of CaMKII-binding domain shares noticeable sequence similarity with CaMKII-binding domains in the Ca V 1.3 and Ca V 1.2 NTDs identified here. Furthermore, we identified a V102E mutation in the CaMKII␣ catalytic domain that substantially reduced interactions with the Ca V 1.3 NTD, without significantly affecting interactions with the ␤1/␤2 subunits, densin or GluN2B. In contrast, an I205K mutation in the CaMKII␣ catalytic domain (I206K in CaMKII␤, -␥, and -␦) pre-viously shown to interfere with binding to GluN2B (26) and densin (22), also disrupts CaMKII binding to ␤1/␤2 subunits and the Ca V 1.3 NTD (Fig. 4B). Taken together, despite some overlap in the CaMKII residues required for binding, our data suggest that the newly identified Ca V 1.3 NTD CaMKII-binding domain may represent a new class of CaMKAP.

Roles of the NTD and other CaMKAPs in LTCC complexes
It is well-established that CaMKII associates with LTCCs in cardiomyocytes and neurons, as revealed by co-immunoprecipitation and/or by co-localization (17,18,30), but the molecular basis for this interaction is unclear. In one series of studies, CaMKII was shown to bind directly to ␤1 and ␤2 LTCC auxiliary subunits, but not to ␤3 or ␤4 (16), and co-immunoprecipitated from brain extracts with ␤1 subunits, but not ␤4 subunits (18). Moreover, mutation of the CaMKII-binding domain in the ␤2 subunit reduces CaMKII co-immunoprecipitation with Ca V 1.2 channels in heterologous cells (18) as well as Ca 2ϩ -dependent facilitation of Ca V 1.2 (33). In contrast, another CaMKAP, the synaptic scaffolding protein densin, forms ternary complexes with CaMKII and Ca V 1.3 LTCCs in the brain and is necessary for CaMKII-and Ca 2ϩ -dependent facilitation of Ca V 1.3 (19). Here, we found that the ␣1 subunit NTD is important for CaMKII association with intact Ca V 1.3 LTCC complexes by co-immunoprecipitation of activated CaMKII with HA-tagged Ca V 1.3 LTCCs. Co-immunoprecipitation of CaMKII was substantially reduced by replacement of the Ca V 1.3 NTD with NTDs from Ca V 2.2 or Ca V 3.2, which do not significantly bind CaMKII in vitro. Similarly, co-immunoprecipitation was substantially reduced by either deletion of residues 69 -93 or mutation of 83 RKR 85 to AAA in the Ca V 1.3 NTD or by the CaMKII␣-V102E mutation. Thus, the present findings demonstrate the importance of a direct CaMKII interaction with a novel CaMKII-binding domain in the Ca V 1.3 ␣1 subunit NTD, significantly extending our understanding of biochemical mechanisms involved in LTCC signaling.
It is important to note that our heterologous cell studies were conducted using HA-tagged Ca V 1.3 and the ␤3 auxiliary subunit, which does not directly interact with CaMKII (16). ␤3 is thought to be the most abundant in brain (29), but the other three ␤ subunits also are expressed in neurons. It seems likely that the association of ␤ subunit variants with the I-II linker domains of VGCC ␣1 subunits is determined in part by their relative expression levels. However, VGCCs may exhibit selectivity for the ␤ subunits in cells, consistent with data showing that overexpressed ␤ subunit variants are differentially localized in cultured hippocampal neurons (34). Nevertheless, we posit that neurons contain multiple subpopulations of Ca V 1.3 LTCC complexes associated with different ␤ subunit variants. Ca V 1.3 LTCCs containing ␤3 or ␤4 may rely only on the NTD for CaMKII association, whereas those containing ␤1 or ␤2 have a second interaction site. Indeed, although this novel CaMKII-binding domain is highly conserved, Ca V 1.2 NTD binding to CaMKII is somewhat weaker than that of Ca V 1.3 NTD (Fig. 2C), and we previously reported that CaMKII association with Ca V 1.2 channel complexes depends in part on interaction with ␤2 subunits (18).

Nuclear signaling by L-type Ca 2؉ channel and CaMKII
Neuronal Ca 2ϩ channels are often part of larger complexes containing other proteins. For example, a canonical PDZ domain-binding motif at the C terminus of the Ca V 1.3 ␣1 subunit interacts with PDZ domains in synaptic scaffolding proteins like densin or Shank3 (19,35). CaMKII interactions with such scaffolding proteins may represent an additional mechanism for targeting CaMKII to certain subpopulations of neuronal Ca V 1.3 LTCCs (in addition to ␣1 subunit NTDs and ␤1/2 subunits). Indeed, the key role for densin in targeting CaMKII to promote Ca 2ϩ -dependent facilitation of Ca V 1.3 LTCCs was noted above (19). Taken together, these observations indicate that several "flavors" of neuronal Ca V 1.3 LTCC complexes with distinct protein compositions may associate with CaMKII in different ways, perhaps conferring distinct roles for CaMKII in regulating LTCCs and/or downstream signaling.

Role of CaMKII binding to the Ca V 1.3 NTD in excitationtranscription coupling
As noted above, the precise regulation of CREB phosphorylation at Ser 133 is critical for the regulation of gene expression during normal learning and memory consolidation. Although Ser 133 phosphorylation may not be sufficient for gene expression under all conditions (36), it is frequently used as a readout for E-T coupling to CREB, as in the current studies. It is wellestablished that the selective activation of several different receptors and ion channels can engage diverse signal transduction pathways (e.g. cyclic AMP, Ca 2ϩ , and MAPKs) to increase neuronal CREB phosphorylation at Ser 133 . Moreover, distinct Ca 2ϩ -dependent pathways can be engaged to increase Ser 133 phosphorylation, depending on the specific channel that generates the Ca 2ϩ signal. Among the known mechanisms, selective LTCC activation is sufficient for immediate early gene expression in vivo (37) and for CREB Ser 133 phosphorylation in cultured neurons (30,38). The stimulation of CREB phosphorylation can be driven by increased nuclear Ca 2ϩ concentrations, which can be induced using some stimulation paradigms (e.g. NMDA receptor activation) (39). However, the initiation of E-T coupling to CREB by moderate LTCC activation seems to be independent of increases in nuclear Ca 2ϩ and only requires increased Ca 2ϩ concentrations in the immediate vicinity of the channel itself (12,30). Stronger, more prolonged, stimulation paradigms may overcome this requirement for Ca 2ϩ signaling within the LTCC nanodomain by recruiting additional mechanisms. This may be evident in superior cervical ganglion neurons (30), where the global increase of Ca 2ϩ in response to modest stimulation (40 mM KCl) involves similar contributions from Ca V 1 and Ca V 2 channels, but the resulting increase of CREB phosphorylation at Ser 133 is preferentially coupled to Ca V 1 LTCCs, correlating with the co-localization of CaMKII with Ca V 1, but not Ca V 2, channels under these conditions. Ca V 2 channel-dependent increases of global Ca 2ϩ in superior cervical ganglion neurons are shaped by mitochondria and the endoplasmic reticulum, but more robust stimulation induces Ca V 2-dependent increases of CREB phosphorylation by a mechanism that is only partially dependent on CaMKII (30). Sufficiently robust Ca V 2-dependent increases in global Ca 2ϩ may also increase nuclear Ca 2ϩ , potentially contributing to the stimulation of CREB phosphorylation (39).
In this study, we focused on an E-T coupling paradigm that has been shown to depend on signaling within the LTCC nanodomain (12,15). We found that mutations in the Ca V 1.3 NTD that disrupt CaMKII binding in vitro interfere with E-T coupling in cultured neurons. Similarly, mutation of the CaMKII catalytic domain to selectively disrupt binding to the NTD interferes with E-T coupling. Notably, although CaMKII␣/␤ knockdown significantly interfered with LTCC-dependent E-T coupling, there was no significant effect on the Ca 2ϩ signal detected in the soma. Thus, the data presented herein show that a novel and unique CaMKII binding site in the N-terminal domain of Ca V 1.3 LTCCs plays an important role in E-T coupling, apparently by initiating a local mechanism within the LTCC nanodomain rather than modulating the global Ca 2ϩ signal.
Our current data do not preclude additional roles for CaM-KII binding to other components of LTCC complexes (see above) in E-T coupling. This may be particularly germane in light of a recent study indicating that Ca 2ϩ influx drives activation and Thr 286/287 autophosphorylation of CaMKII␣/␤ holoenzymes within the LTCC nanodomain, which in turn trans-autophosphorylate a CaMKII␥ holoenzyme at Thr 287 , trapping bound calmodulin to be shuttled to the nucleus (10). Given that trans-holoenzyme autophosphorylation is very inefficient in solution (40), it is tempting to speculate that this process is facilitated by simultaneous targeting of CaMKII␣/␤ and CaMKII␥ holoenzymes within the LTCC nanodomain. Our data indicate that CaMKII␣/␤ holoenzymes can dock to the NTD, and it is possible that a CaMKII␥ holoenzyme interacts with other components of the larger LTCC complex (see above). Interestingly, CREB phosphorylation was reported to be preferentially coupled to Ca V 1.3 over Ca V 1.2 LTCCs (31). Notably, Shank3 selectively associates with Ca V 1.3 LTCCs and has been shown to play an important role in E-T coupling (35). Moreover, Shank3 was identified as an abundant component of synaptic CaMKII complexes in a recent proteomics study (41). Multiple CaMKII docking sites in the LTCC complex also may be linked to voltage-dependent conformational changes that appear to be required to initiate E-T coupling (11). Such changes may be required to facilitate docking of CaMKII holenzyme(s) and appropriately position CaMKII relative to other components of the nanodomain. Additional studies are clearly required to explore these ideas and provide further insight into the molecular mechanisms underlying the initiation of E-T coupling.

Animals
Timed pregnant Sprague-Dawley rats were purchased from Charles River Laboratories. Embryonic day 18.5 pregnant rats were euthanized in a CO 2 chamber before embryos were removed from the uterus. All animal experiments were approved by the Vanderbilt University institutional animal care and use committee and were carried out following the National Institutes of Health Guide for the Care and Use of Laboratory Animals.  (17)) and the rat Ca 2ϩ channel ␤2a subunit (His 485 -Glu 505 ) (16) were also generated. All cDNAs were inserted into pGEX-4T1 using traditional ligation or sequence-and ligation-independent cloning (42).

Nuclear signaling by L-type Ca 2؉ channel and CaMKII
A plasmid encoding Ca V 1.3 with an N-terminal HA-tag (pCGNH-Ca V 1.3, for co-immunoprecipitation) was made by inserting rat Ca V 1.3 cDNA into pCGN vector (a gift from Dr. Winship Herr, Université de Lausanne, Switzerland, Addgene plasmid ID 53308). Ca V 1.3 chimeric constructs were made in the following way: pCGNH-Ca V 1.3 was used as a template to delete the N-terminal domain of Ca V 1.3, leaving the XbaI site intact (pCGNH-Ca V 1.3-⌬NTD); cDNAs encoding the Ca V 2.2 and Ca V 3.2 NTDs were then inserted into the XbaI site using sequence-and ligation-independent cloning. Plasmid encoding Ca V 1.3 with an extracellular HA tag was generated by first removing the sequence encoding the N-terminal HA in pCGNH-Ca V 1.3 and then inserting the sequence encoding an HA tag flanked by flexible linkers on both ends between Gln 693 and Lys 694 (pCGN0-Ca V 1.3-sHA). The inserted amino acid sequence is TRHYPYDVPDYAVTFDEMQ, where the HA sequence is in boldface type (43). The nimodipine-resistant Ca V 1.3 was then generated by mutagenesis of pCGN0-Ca V 1.3-sHA to generate a T1033Y mutant (32,44). The region encoding Met 69 -Leu 93 was deleted from pCGN0-Ca V 1.3-sHA plasmid to remove the CaMKII-binding domain. Site-directed mutagenesis, epitope insertions, and all deletions were done following the one-step mutagenesis protocol described by Liu and Naismith (45).
CaMKII shRNA constructs for pCREB staining were expressed with GFP using a pLL3.7 plasmid (a gift from the Luk Van Parijs laboratory, Massachusetts Institute of Technology, Cambridge, MA) that was modified to replace the CMV promoter with a 0.4-kb fragment of the mouse CaMKII␣ promoter (designated as pLLCK) that is primarily active only in excitatory neurons (46). The shRNA sequences were designed following Ma et al. (10). The shRNA-targeted sequence in the mouse CaMKII␣ cDNA contains two mismatches from the corresponding rat sequence, rendering it resistant to the shRNA. Knockdown and shRNA resistance were confirmed by Western blotting and immunostaining. All constructs were confirmed by DNA sequencing.

Recombinant mouse CaMKII␣ and GST-tagged protein purification
Expression and purification of recombinant mouse CaMKII␣ has been described previously (47). pGEX-4T1 plasmids were transformed into BL21(DE3) bacteria cells to express GSTtagged proteins. Cells were grown in LB medium at 37°C to reach an OD of ϳ0.6. Isopropyl 1-thio-␤-D-galactopyranoside (0.2 mM) was then added to induce the protein expression at room temperature for 2 h. We found that the Ca V 1.2 and Ca V 1.3 full-length C-terminal domain fragments do nottk;1 express well in BL21(DE3) cells. We identified several rare codons in the cDNAs encoding both CTDs and found that their expression was substantially improved in Rosetta 2(DE3) BL21 cells engineered to contain rare tRNAs (EMD Millipore, catalog no. 71400). Expressed proteins were purified using Pierce glutathione-agarose beads (catalog no. 16101) following the manufacturer's instructions. Eluted proteins were then dialyzed in 10 mM HEPES, pH 7.5, 25 M PMSF, 62.5 M benzamidine, 62.5 M EDTA, 0.1% Triton X-100 overnight with one buffer change.

CaMKII autophosphorylation and GST pulldown
Purified mouse CaMKII␣ was incubated with 50 mM HEPES, pH 7.5, 10 mM Mg(CH 3 COO) 2 , 0.5 mM CaCl 2 , 2.5 M calmodulin, 40 M ATP on ice for 90 s before the addition of EDTA (20 mM final) to terminate phosphorylation by chelation of Mg 2ϩ and Ca 2ϩ . The reaction was then diluted 10-fold using 1ϫ GST pulldown buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% (v/v) Triton X-100). A final protein concentration of 125 nM was used for both CaMKII␣ and GST-tagged proteins. An aliquot (5%) of each incubation was saved as input followed by the addition of 5 l of prewashed glutathione magnetic beads (Pierce, catalog no. 88821, 25% (v/v)). After incubating at 4°C for 1 h, beads were separated magnetically and washed three times with GST pulldown buffer. GST protein complexes were eluted by incubation with 40 l of 20 mM glutathione (pH 8.0) in GST pulldown buffer at 4°C for 10 min.

Electrophysiology
HEK293T cells in 35-mm dishes were transfected with 2 g of Ca V 1.3 WT or ⌬69 -93 pcDNAs together with 1 g of ␤3, 1 g of ␣2␦, and 0.05 g of enhanced green fluorescent protein pcDNAs. Cells were split into new dishes 36 h after transfection, and whole-cell Ca 2ϩ currents were recorded at room temperature 48 h after transfection. Data were collected through an Axopatch 200B amplifier and pCLAMP10 software (Molecular Devices). Pipette resistance was 4 -6 megaohms when loaded with the intracellular solution and immersed in the extracellular solution. Series resistance and membrane capacitance were compensated up to 80%. The intracellular solution contained 132 mM CsCl, 10 mM tetraethylammonium chloride, 10 mM EGTA, 1 mM MgCl 2 , 3 mM Mg-ATP, 5 mM HEPES, pH 7.3 adjusted by CsOH. The external solution contained 112 mM NaCl, 20 mM tetraethylammonium chloride, 10 mM CaCl 2 , 5 mM CsCl, 1 mM MgCl 2 , 10 mM HEPES, 5 mM glucose, pH 7.3 adjusted by NaOH. The osmolarity is 300 mosM for the intracellular solution and 305 mosM for the extracellular solution. For current-voltage protocols, the membrane voltage was depolarized in 50-ms steps from Ϫ70 mV to various voltages in 10-s intervals. A P/4 protocol was used for leak subtraction.

Neuronal pCREB imaging and quantification
Images were collected using an Olympus FV-1000 inverted confocal microscope with a 40ϫ/1.30 numeric aperture Plan-Neofluar oil lens. The binocular lens was used to identify transfected neurons based on the Alexa 546 signal from the HA staining or enhanced green fluorescent protein from the shRNA construct. The DAPI channel was then used to focus on the z plane that yielded the highest DAPI signal (one that presumably runs across the nuclei) for image acquisition. Images were then collected in all of the channels, and MetaMorph Microscope Automation and Image Analysis Software (Molecular Devices) was used to quantify the pCREB signal. Briefly, nuclei were identified by thresholding the DAPI channel to create and select the nuclear regions of interest. The regions of interest were then transferred to other channels to measure the average pCREB intensity.

Nuclear signaling by L-type Ca 2؉ channel and CaMKII
The relative pCREB intensity was computed as (pCREB x -pCREB 5K )/(pCREB 40K -pCREB 5K ), where pCREB x is the pCREB signal being calculated, and pCREB 5K and pCREB 40K are the average signals of the 5K and 40 K conditions in that batch of cultured neurons, respectively. Data shown were collected from images of the indicated total number of neurons from 3-4 independent cultures.

Neuronal Ca 2؉ imaging
Dissociated rat hippocampal neurons cultured in coated 29-mm glass bottom dishes (Cellvis, catalog no. D29-10-1.5-N) were transfected with a total of 2 g of DNA/dish after 8 DIV. All neurons (nontransfected for Fig. 7A, transfected with shRNA vectors for Fig. 8A) were imaged on DIV 13-14. Because GFP interferes with Fura-2 imaging, CaMKII shRNA constructs lacking the CaMKII promoter and GFP were co-expressed with mApple to label transfected cells. Cells were incubated at 37°C for 20 min in culture medium (neural basal medium with 2% B27, 0.25% glutamax, and 1% penicillin-streptomycin) supplemented with 2 M Fura-2 acetoxymethyl ester (Thermo Fisher Scientific, catalog no. F1221), washed twice with 5K Tyrode's solution, and then incubated at 37°C for 15 min in 5K Tyrode's solution with TTX, APV, and NBQX (as above). For nimodipine-treated groups, this medium was replaced with 5K Tyrode's solution containing 10 M nimodipine (in addition to TTX, APV, and NBQX) ϳ5 min before imaging. Fura-2 fluorescence images were collected using a Nikon Eclipse TE2000-U microscope equipped with an epifluorescence illuminator (Sutter Instrument Co.) and an HQ2 CCD camera (PhotoMetrics Inc.). Baseline Ca 2ϩ was recorded for 30 s in 5K Tyrode's solution before replacing by 40K Tyrode's solution. Cell somas were selected as regions of interest using Nikon Elements software; transfected neurons were selected based on mApple fluorescence. The ratios of emitted fluorescence (505 nm) intensities at excitation wavelengths of 340 and 380 nm (F 340 /F 380 ) were measured every 5 s. Responses of individual cells at each time point were quantified as the change in fluorescence ratio above baseline (⌬F ϭ (340/380 value)/(baseline 340/380 value)). The peak change in fluorescence ratio was used to compare responses between cells within each group (⌬F ϭ (maximum 340/380 value)/(baseline 340/380 value)), and outlier cells were excluded based on a ROUT outliers test (Q ϭ 1%). The numbers of excluded outliers were as follow: