Sequential ionic and conformational signaling by calcium channels drives neuronal gene expression

Abstract

Voltage-gated Ca_V1.2 channels (L-type calcium channel α1C subunits) are critical mediators of transcription-dependent neural plasticity. Whether these channels signal via the influx of calcium ion (Ca²⁺), voltage-dependent conformational change (VΔC), or a combination of the two has thus far been equivocal. We fused Ca_V1.2 to a ligand-gated Ca²⁺-permeable channel, enabling independent control of localized Ca²⁺ and VΔC signals. This revealed an unexpected dual requirement: Ca²⁺ must first mobilize actin-bound Ca²⁺/calmodulin-dependent protein kinase II, freeing it for subsequent VΔC-mediated accumulation. Neither signal alone sufficed to activate transcription. Signal order was crucial: Efficiency peaked when Ca²⁺ preceded VΔC by 10 to 20 seconds. Ca_V1.2 VΔC synergistically augmented signaling by N-methyl-D-aspartate receptors. Furthermore, VΔC mistuning correlated with autistic symptoms in Timothy syndrome. Thus, nonionic VΔC signaling is vital to the function of Ca_V1.2 in synaptic and neuropsychiatric processes.

Voltage-gated CaV1.2 channels (L-type calcium channel α1C subunits) play an important role in transcription-dependent forms of synaptic and homeostatic plasticity (1–6), and Ca_V1.2 alterations have been linked to severe neuropathologies (7, 8). The influx of Ca²⁺ is required in Ca_V1.2-mediated transcription (4), but it remains unclear whether voltage-dependent conformational change (VΔC) provides a necessary additional signal. There is precedence for VΔC signaling: Ca_V1.1 uses only VΔC to initiate skeletal-muscle contraction (9, 10). However, a signaling role for VΔC has been difficult to establish in excitation-transcription coupling, because eliminating voltage-dependent opening of the channel also prevents Ca²⁺ influx through the Ca_V [but see (11)].

We fused Ca_V1.2 to an adenosine triphosphate (ATP)–gated, Ca²⁺-permeable, tandem-trimeric P2X2 channel (ttP2X) (Fig. 1A). The ttP2X was used to provide Ca²⁺ influx independently of whether Ca_V1.2 was open or closed; the tethering between the channels (Fig. 1A) localized ttP2X influx to the Ca_V1.2 nanodomain (12, 13).

Fig. 1. Decoupling Ca_V1.2 Ca²⁺ and VΔC signals reveals dual requirement for CREB activation.

(A) The engineered chimeric construct, showing transmembrane domains and intra- and extracellular loops. (B) Ca²⁺ currents mediated by the chimeric channel in HEK cells. (C) Changes in voltage (top, current clamp recording) and Ca²⁺ influx (bottom, fura-2 imaging) under key stimulation conditions (N = 6 cells; fig. S2, A and B, shows SEM). (D and E) Nuclear pCREB in cultured cortical neurons assayed after 3 min of stimulation (conditions are indicated on the left; the cartoons represent the ttP2X-Ca_V1.2 chimera, with plus and minus signs indicating the presence and absence of the signals, respectively). pCREB was elevated only with dual Ca²⁺ and VΔC stimuli [second row in (D) and first and fourth rows in (E)]. Elimination of either Ca²⁺ [third row in (D)] or VΔC [second and third rows in (E)] abrogated signaling. Black bars show means ± SEM (normalized to no stimulation) of 50 cells from three independent cultures. Scale bar, 5 μm. **P < 0.01, determined by one-way analysis of variance (ANOVA) followed by Fisher’s least significant difference test.

We confirmed the integrity of the chimeric protein and the functionality of its components (fig. S1). In human embryonic kidney (HEK) 293 cells, the Ca²⁺ current appeared only at depolarized potentials in the absence of ATP (Fig. 1B, black u-shaped trace), as expected for a voltage-gated Ca_V1.2. On addition of ATP, the ttP2X portion of the chimera also supported Ca²⁺ influx, evident as an additional inward current at negative potentials (Fig. 1B, gray trace). Cd²⁺, which blocks the Ca_V1.2 pore (14) but not the opening of ttP2X, did not prevent the ttP2X-mediated entry of Ca²⁺ (Fig. 1B, blue trace). We further validated the function of the chimeric components in cultured neurons, rendering the Ca_V1.2 portion dihydropyridine-insensitive (DHPi) to enable its distinction from endogenous channels (6) and confirming that ATP had no effect on untransfected neurons (fig. S1). Whereas ttP2X on its own distributed uniformly over the somatodendritic surface, ttP2X fused to Ca_V1.2 formed puncta and signaled more potently (fig. S1), consistent with localization to signaling hotspots.

These control experiments framed critical tests of the roles of Ca²⁺ and VΔC in signaling to nuclear CREB (cyclic adenosine monophosphate response–element binding protein), a transcription factor that is critical in many forms of learning and memory (Fig. 1, D and E) (1, 2, 15). Providing Ca²⁺ and VΔC in combination via the depolarization of neurons (3 min of exposure to 40 mM K⁺ (40K⁺)] (Fig. 1C, black traces) increased the phosphorylation of nuclear CREB (pCREB) at Ser¹³³ threefold (Fig. 1D, second row) (3–5). The pCREB response was abolished by Cd²⁺ (Fig. 1D, third row), which prevents Ca_V1.2 from conducting Ca²⁺ without affecting depolarization (Fig. 1C, gray traces) or voltage-dependent gating (14); this confirms the known requirement for Ca²⁺ influx in signaling to pCREB (4, 6, 13). Next, we rerouted Ca²⁺ through the neighboring ttP2X by blocking the Ca_V1.2 pore with Cd²⁺ and opening the ttP2X portion of the chimera with ATP. This generated depolarizations and increases in bulk Ca²⁺ that were nearly identical to those achieved with 40K⁺ (Fig. 1C, compare blue with black traces, and fig. S2, A and B) and increased pCREB to the same degree as 40K⁺ (compare Fig. 1E, top row, with Fig. 1D, second row), confirming the utility of the chimeric channel approach.

We thus could use the chimera to determine whether localized Ca²⁺ influx can drive CREB activation in the absence of VΔC. In a first test, we provided the localized influx of Ca²⁺ by means of ATP activation of the chimeric ttP2X, but we attenuated VΔC signals through hyperpolarization, which was achieved by coexpressing and activating an ivermectin-gated chloride channel (GlyIVR) (16). This manipulation prevented neuronal depolarization (fig. S2C) and thus increased Ca²⁺ flux through the chimeric ttP2X (Fig. 1B, blue). Nonetheless, this manipulation inhibited signaling to CREB (Fig. 1E, second row). The attenuation of CREB signaling required a combination of GlyIVR expression and ivermectin (fig. S2D) and could not be attributed to increased Ca²⁺ influx (fig. S2E).

In a second test, we inhibited Ca_V1.2 conformational opening with nimodipine. This Ca_V1-selective agent (17, 18) blocked ATP-mediated CREB signaling (Fig. 1E, third row) without affecting depolarization or Ca²⁺ influx (Fig. 1C, compare orange with blue). Nimodipine did not block the pCREB response when the chimeric Ca_V1.2 was rendered nimodipine-insensitive (Fig. 1E, bottom row), excluding potential off-target effects of the drug. Similar results were obtained with CREB-dependent expression of the immediate early gene c-fos (fig. S3). Thus, Ca_V1.2 signaling to pCREB and gene expression requires two distinct messages: The Ca²⁺ signal works in conjunction with an equally indispensable VΔC signal arising from Ca_V1.2.

Previous work has implicated Ca²⁺/calmodulin-dependent protein kinase II (CaMKII) in mediating the signaling from Ca_V1.2 to nuclear CREB (3–5, 19). CaMKIIs are activated and recruited into puncta near Ca_V1.2 upon depolarization (4, 5), where they play a critical role in dispatching a signal to the nucleus (19). We confirmed the critical role of α- and βCaMKII by means of pharmacology (Fig. 2A and fig. S4A) and selective knockdowns of either isoform with small hairpin RNA (shRNA) (Fig. 2, B and C, and fig. S4, B and C). In contrast, signaling remained intact when protein kinase C, protein kinase A, or mitogen-activated protein kinase pathways (3, 6, 20) were blocked (fig. S4A). We further established that αCaMKII forms an activity-dependent complex with Ca_V1.2: Immunoprecipitation with a Ca_V1.2 antibody pulled down αCaMKII, and the degree of the coimmunoprecipitation was increased by stimulation with 40K⁺ (Fig. 2D).

Fig. 2. Redistribution of CaMKII to Ca_V1.2 puncta requires dual Ca²⁺ and VΔC signals.

(A to C) pCREB signaling (3 min of 40K⁺ or ATP-plus-Cd²⁺ stimulation) was blocked by the CaMKII inhibitor KN93 (KN92 is an inactive analog) (A), or by shRNA against either αCaMKII (B) or βCaMKII (C). (D) Co-immunoprecipitation with antibodies against Ca_V1.2 indicated that the association between Ca_V1.2 and αCaMKII increased after stimulation (IB, immunoblotting; IP, immunoprecipitation). (E, G, and I) pCaMKII (E), αCaMKII (G), and βCaMKII (I) in dendrites of cultured cortical neurons after stimulation (ATP plus Cd²⁺ for 3 min). Their staining is punctate (green) and colocalizes with surface ttP2X-Ca_V1.2 channels (red). Scale bar, 10 mm. (F, H, and J) Quantification of puncta intensity for pCaMKII (F), αCaMKII (H), and βCaMKII (J) (Nim, nimodipine). Puncta intensity increases with dual Ca²⁺ and VΔC signals (second and fourth rows) but not with Ca²⁺ or VΔC signals in isolation (third and fifth rows). The bars show means ± SEM (normalized to no stimulation) of 50 cells from three independent cultures. **P < 0.01, determind by one-way ANOVA followed by Fisher’s least significant difference test.

We next assessed the contributions of Ca²⁺ and VΔC signals in mediating the spatial recruitment of CaMKII to the Ca_V1.2 channel. Staining with a phospho-specific CaMKII antibody revealed the formation of intense phospho-CaMKII (pCaMKII) puncta near surface ttP2X-Ca_V1.2 channels after dual Ca²⁺ and VΔC stimulation (Fig. 2, E and F), but not after stimulation by Ca²⁺ or VΔC alone (Fig. 2F). A similar pattern was observed for isoform-specific antibodies against αCaMKII (Fig. 2, G and H, and fig. S4, D and E) or βCaMKII (Fig. 2, I and J, and fig. S4E), which is consistent with CaMKII isomers relocating as heteromultimers (21, 22). Thus, Ca_V1.2 communication to multiple isoforms of CaMKII follows the signaling logic observed in Ca_V1.2-mediated CREB signaling; all demand a conjunction of Ca²⁺ and VΔC signals.

We looked for dynamic changes of CaMKII produced by Ca²⁺ or VΔC signals in isolation. Whereas VΔC-only stimuli had no effect, Ca²⁺-only stimuli (10 s) mobilized βCaMKII [Fig. 3, A (green trace) and B (top two images)], which is consistent with its known dissociation from F-actin upon Ca²⁺ stimulation (21, 22). The mobilization of βCaMKII (Fig. 3A, green trace) outlasted bulk Ca²⁺ elevation (Fig. 3A, blue trace) but returned to the initial distribution within 60 s, which is consistent with the kinetics of calmodulin trapping (23). Evidently, mobilized βCaMKII reverted to its cytoskeleton-bound state in the absence of VΔC signaling (Fig. 3B). To find out when VΔC signaling was most effective, we varied the timing of brief (10-s) pulses of Ca²⁺ and VΔC inputs (Fig. 3C), quantifying efficacy by downstream CREB activation (Fig. 3D). Synchronous stimuli (difference in time, Δt = 0) were suboptimal. Instead, VΔC potency peaked at Δt = 10 to 20 s after Ca²⁺ influx and persisted at Δt = 40 s (Fig. 3D, black bars). Reversing the order of stimuli (VΔC first) resulted in little signaling to CREB (Fig. 3D; Δt = −10 s), a pronounced temporal asymmetry. The dynamics of VΔC potency (Fig. 3D, black bars) and of βCaMKII mobilization (Fig. 3D, green trace from Fig. 3A) were similar. This suggests that Ca²⁺ mobilization of βCaMKII [or α/βCaMKII heteromultimers (21)] is a prerequisite for their subsequent VΔC-mediated accumulation at Ca_V1.2 channels and for downstream signaling to pCREB (fig. S5) (19).

Fig. 3. Temporal dissection of Ca²⁺ and VΔC signals reveals a slow coincidence detection scheme.

(A) The protocol for the 10-s Ca²⁺-only pulse produced with ATP plus Nim (top) and the resulting bulk Ca²⁺ elevation (bottom, blue waveform; ΔF/F₀, ratio of fluorescence difference to basal value). Cells were fixed at various time points (black arrows) and βCaMKII mobilization (green waveform) was determined by a variance-over-mean (var/mean) analysis. Blue shading and green bars show means ± SEM of 25 cells from three cortical cultures. (B) Sample images of βCaMKII distribution in response to different stimuli. Ca²⁺ alone transiently mobilizes βCaMKII (top three images) but does not form βCaMKII puncta (compare with the bottom image). Scale bar, 10 μm. (C) The protocol for temporally shifted 10-s pulses of Ca²⁺ and VΔC. (D) Nuclear pCREB signaling in response to stimuli in (C) (black bars), overlaid with waveforms from (A). The bars show means ± SEM of 120 cells from four independent cortical cultures.

To test whether VΔC signals interact with other Ca²⁺ sources, such as N-methyl-D-aspartate receptors (NMDARs) (21, 22), we examined CaMKII puncta formation in response to NMDAR activity in isolation (Fig. 4A, right; nimodipine present) or in combination with Ca_V1.2 VΔC (Fig. 4A, left; nimodipine absent and Ca_V1.2 pore blocked with Cd²⁺). VΔC inclusion enhanced the amplitude of the NMDAR-induced signal (CaMKII puncta formation) (Fig. 4, B and C). Thus, VΔC signals not only interact with Ca²⁺ entry from the home Ca_V1.2 (Fig. 3) but also synergistically modulate the signaling potency of other Ca²⁺ sources (Fig. 4).

Fig. 4. Ca_V1.2 VΔC synergistically augments NMDAR signaling and is mistuned in TS mutations.

(A) Scheme for examining the functional crosstalk of NMDAR and Ca_V1.2 VΔC. (B and C) pCaMKII (B) and αCaMKII (C) puncta intensity analysis in neurons after 1 min of stimulation. Shown are means (points) ± SEM (bars) of 50 cells for each condition. *P < 0.05 and **P < 0.01, determined by a two-tailed independent Student’s t test. (D) Approach to examining the VΔC potency of Ca_V1.2 variants by means of chimeras (MUT, mutation). (E) Signaling to both CaMKII and CREB is substantially elevated (~30 to 70%) in G406R relative to either WT or G402S Ca_V1.2. Bars show means ± SEM of 50 cells from three cultures. *P < 0.05 and **P < 0.01, determined by one-way ANOVA followed by Fisher’s least significant difference test. (F) Surface expression of TS mutants is indistinguishable from WT Ca_V1.2 (live-cell surface stain).

With regard to disease, Timothy syndrome (TS) (7) arises from either of two point mutations in helix IS6 of Ca_V1.2 (fig. S6A). Both the G406R and G402S TS variants cause a prolonged Ca²⁺ current (arising from the slowing of Ca_V1.2 inactivation), and both variants produce a long-QT cardiac arrhythmia (fig. S6B) (7, 8). Whereas G406R produces autism spectrum disorder with >60% penetrance, G402S patients are neurologically intact, suggesting that Ca²⁺ influx is not what determines the neurological phenotype (fig. S6B, black symbols) (8, 24, 25). To test the role of VΔC in TS, we made TS variants in the Ca_V1.2 portion of the chimera while holding the Ca²⁺ flux fixed via the fused ttP2X (Fig. 4D). Mutant-specific effects on Ca_V1.2 Ca²⁺ flux (7, 8, 24–27) were suppressed with Cd²⁺, and nimodipine eliminated endogenous Ca_V1 VΔC contributions, whereas chimeric channels were nimodipine-insensitive (Fig. 4D). We found that G406R Ca_V1.2 exhibited gain-of-function VΔC signaling: α/βCaMKII and CREB signaling were substantially elevated (~30 to 70%). However, chimera surface expression was no different than in wild-type (WT) Ca_V1.2 (Fig. 4, E and F). In contrast, G402S Ca_V1.2 did not differ at all from WT Ca_V1.2 (Fig. 4E). Thus, mistuning of VΔC is associated with the G406R (28) but not with the G402S variant of TS (fig. S6B, colored symbols), in correlation with the autistic symptoms in TS.

Since the classic discovery of flux-independent Ca_V1.1 signaling (9, 10), other nonionic modes of channel signaling have been reported to operate independently from ionic signaling (11, 28–30). In this study, we found that conformational and ionic signals from the same channel can work in concert to regulate transcription. Such dual-mode signaling offers enhanced specificity: Maximal Ca_V1.2-mediated CREB activity requires a sequence of Ca²⁺ influx followed ~10 to 20 s later by VΔC signaling, a form of coincidence detection that is reminiscent of spike timing–dependent plasticity (31) but more than a thousand times slower (fig. S5). A second benefit is an increase in the voltage dependence of signaling. Voltage-dependent contributions from Ca²⁺ and VΔC signals render Ca_V1.2-mediated transcription more sensitive to small changes in depolarization (4). Taken together, these two advantages—heightened temporal specificity and voltage sensitivity—would make suboptimal patterns of activity less effective, while allowing temporally optimal stimuli of even small magnitudes to signal strongly. This may explain why synaptic plasticity that is dependent on Ca_V1-mediated transcription typically requires prolonged bouts of activity or multiple bursts spread over tens of seconds (2, 32).

Yet another advantage of dual-mode signaling is the possibility of interaction between VΔC signals and recent Ca²⁺ entry from other sources. We found that Ca_V1.2 VΔC synergizes with NMDAR signaling (Fig. 4), raising the possibility of interactions with Ca²⁺ signals derived from other sources, such as internal Ca²⁺ stores or Ca^2+-permeable AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) receptors. Looking beyond Ca_V1.2, our chimeric-channel approach may provide a generalizable strategy to investigate nonconducting roles of NMDARs (29) and of other Ca²⁺-permeant channels such as Ca_V1.3. Ca_V1.3 variants that produce autism (33) consistently display negative shifts in voltage dependence that would be expected to enhance VΔC signaling, much as we found for Ca_V1.2 G406R (Fig. 4E).