Cross talk between β subunits, intracellular Ca2+ signaling, and SNAREs in the modulation of CaV2.1 channel steady‐state inactivation

Abstract Modulation of CaV2.1 channel activity plays a key role in interneuronal communication and synaptic plasticity. SNAREs interact with a specific synprint site at the second intracellular loop (LII‐III) of the CaV2.1 pore‐forming α 1A subunit to optimize neurotransmitter release from presynaptic terminals by allowing secretory vesicles docking near the Ca2+ entry pathway, and by modulating the voltage dependence of channel steady‐state inactivation. Ca2+ influx through CaV2.1 also promotes channel inactivation. This process seems to involve Ca2+‐calmodulin interaction with two adjacent sites in the α 1A carboxyl tail (C‐tail) (the IQ‐like motif and the Calmodulin‐Binding Domain (CBD) site), and contributes to long‐term potentiation and spatial learning and memory. Besides, binding of regulatory β subunits to the α interaction domain (AID) at the first intracellular loop (LI‐II) of α 1A determines the degree of channel inactivation by both voltage and Ca2+. Here, we explore the cross talk between β subunits, Ca2+, and syntaxin‐1A‐modulated CaV2.1 inactivation, highlighting the α 1A domains involved in such process. β 3‐containing CaV2.1 channels show syntaxin‐1A‐modulated but no Ca2+‐dependent steady‐state inactivation. Conversely, β 2a‐containing CaV2.1 channels show Ca2+‐dependent but not syntaxin‐1A‐modulated steady‐state inactivation. A LI‐II deletion confers Ca2+‐dependent inactivation and prevents modulation by syntaxin‐1A in β 3‐containing CaV2.1 channels. Mutation of the IQ‐like motif, unlike CBD deletion, abolishes Ca2+‐dependent inactivation and confers modulation by syntaxin‐1A in β 2a‐containing CaV2.1 channels. Altogether, these results suggest that LI‐II structural modifications determine the regulation of CaV2.1 steady‐state inactivation either by Ca2+ or by SNAREs but not by both.


Introduction
Ca 2+ entry through the high-voltage-activated (HVA) Ca V 2.x channels (mainly Ca V 2.1 [P/Q-type] channels) into presynaptic nerve terminals supports a transient Ca 2+ microdomain that is essential for synaptic exocytosis leading to the fast release of classical neurotransmitters (Catterall 2011). To ensure fast and efficient neurotransmitter release, the vesicle-docking/release machinery must be located near the pathway of Ca 2+ entry. In many cases, this close localization is achieved by direct interaction of soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins with the Ca 2+ channel poreforming a 1 subunit, which consists of four repeated domains (I-IV) each containing six transmembrane regions (S1-S6) with a voltage sensor (S1-S4) and a pore region (S5, P-loop, and S6). Indeed, syntaxin-1A/1B, SNAP-25, and synaptotagmin-1 specifically interact with Ca V 2.1 and Ca V 2.2 channels by binding to a synaptic protein interaction site (synprint) located within the intracellular loop connecting domains II and III (LII-III) of the channels (Sheng et al. 1994Rettig et al. 1996;Kim and Catterall 1997;Jarvis et al. 2002) (Fig. 1A). Furthermore, it has been suggested that exocytosis is activated even before Ca 2+ entry, by conformational changes triggered during Ca 2+ binding at the open Ca V channel pore, which are transmitted from the channel to specific residues of Ca V -interacting SNARE proteins, accounting for the rapid time frame of evoked release (Atlas 2013;Bachnoff et al. 2013). Whichever the case, perhaps an equally important consequence of SNARE protein interaction with the Ca 2+ channel is the modulation of presynaptic Ca 2+ channel activity, thus fine-tuning the amount of Ca 2+ that binds to the pore, enters the synaptic terminal, and determines synaptic transmission strength. Specifically, the binding of syntaxin-1A and SNAP-25 to Ca V 2.1 and Ca V 2.2 a 1 subunits shifts the voltage dependence of steadystate inactivation toward more negative membrane potentials following trains of brief depolarizing pulses to reduce channel availability, without affecting channel activation properties (Bezprozvanny et al. 1995;Zhong et al. 1999). Such inhibition is reverted, and channel activity fully restored, by synaptotagmin (Zhong et al. 1999). Thus, Ca V 2.x-SNAREs interaction seems to optimize neurotransmission by favoring Ca 2+ entry through channels presenting docked synaptic vesicles. Accordingly, the disruption of such functional interaction compromises not just vesicle exocytosis in vitro (Mochida et al. 2003;Harkins et al. 2004), but also synaptic transmission and the SNAREmediated inhibitory modulation of Ca V 2.x channels in vivo (Mochida et al. 1996;Rettig et al. 1997;Zamponi 2003;Keith et al. 2007). Moreover, human a 1A mutations impairing the functional interaction between Ca V 2.1 channels and SNARE proteins have clinical relevance in the context of ataxia and the phenotypic expression of both migraine with aura and hemiplegic migraine (Cricchi et al. 2007;Serra et al. 2010;Condliffe et al. 2013).
Beyond the important anchoring function of the synprint site in the SNARE-mediated modulation of Ca V 2.x channel gating, the involvement of other molecular domains has been proposed. Hence, deletions within the LII-III intracellular loop of the Ca V 2.2 a 1B channel subunit that completely eliminate the synprint site reduce but not abolish channel modulation by syntaxin, and syntaxin mutations that have no effect on binding affinity to a 1B -synprint prevent the SNARE-mediated regulation of Ca V 2.2 channel inactivation (Bezprozvanny et al. 2000). Besides, the A454T mutation (placed in the intracellular loop connecting domains I and II (LI-II) of the Ca V 2.1 a 1A channel subunit, and associated to both early-onset progressive ataxia (Cricchi et al. 2007) and the relief of migraine aura symptoms (Serra et al. 2010)) prevents the negative modulation of Ca V 2.1 channels by SNARE proteins and decreases channel coupling to exocytosis, thus revealing the importance of LI-II structural integrity in the Ca V 2.1-SNAREs functional interaction (Serra et al. 2010).
The molecular mechanism by which LI-II influences Ca V 2.1-SNAREs functional interaction is unknown. However, it is well established that LI-II plays a determinant role in the regulation of Ca V 2.1 channel activity (Buraei and Yang 2010). In this sense, it has been suggested that conformational changes induced at LI-II by the binding of functionally different regulatory b subunits not only determine the degree of voltage-dependent inactivation but also the extent of a Ca 2+ -dependent inactivation component (mediated by the binding of Ca 2+ -calmodulin to two adjacent sites in the carboxyl tail [C-tail] of the a 1A subunit: the IQ-like motif and the Calmodulin-Binding Domain [CBD] site) (Lee et al. 2000;DeMaria et al. 2001;Cens et al. 2006) (Fig. 1A). Interestingly, disruption of Ca V 2.1 modulation by calmodulin and related Ca 2+ sensor proteins by mutation of the IQ-like motif has been reported to impair long-term potentiation and spatial learning and memory in mice (Nanoua et al. 2016).
Altogether, it draws a complex scenario in which Ca V 2.1 inactivation is produced by LI-II and modulated by: b channel subunits interacting with LI-II, SNARE proteins binding to the synprint site at the LII-III but requiring the integrity of LI-II, and Ca 2+ -calmodulin attached to the C-tail of the a 1A subunit.
To better understand the role of LI-II in Ca V 2.1-SNAREs functional interaction, we analyzed the modulation of Ca V 2.1 inactivation by syntaxin-1A under intermediate and high Ca 2+ -buffering conditions, in the presence of functionally different regulatory b subunits (b 2a or b 3 ) and distinct human a 1A constructs containing either a LI-II deletion around the A454 residue, mutations in the IQ-like region, or a CBD deletion. Our results reveal a cross talk between different pathways involved in the modulation of Ca V 2.1 inactivation, showing that regulation by syntaxin-1A of the human Ca V 2.1 channel activity requires both the integrity of a 1A LI-II and the lack of a Ca 2+ -dependent component in the channel steady-state inactivation.
Methods cDNA constructs and site-directed mutagenesis cDNA of the human voltage-gated Ca 2+ (Ca V 2.1) channel a 1A subunit (originally cloned into a pCMV vector) was a gift from Professor J. Striessnig (University of Innsbruck, Austria). cDNAs of the rabbit a 2 d and rat b 3 and b 2a regulatory subunits, and syntaxin-1A (subcloned into a pcDNA3 expression vector) were gifts from Dr. L. Birnbaumer (National Institutes of Health, North Carolina, USA) and Dr. J. Blasi (Universitat de Barcelona, Spain). Ca V 2.1 a 1A mutant subunits (DLI-II 451-457 , IM/EE 1964;1965 , DCBD 2020-end ) were generated using site-directed mutagenesis (GenScript Corporation, Piscatway, NJ). All cDNA clones used in this study were sequenced in full to confirm their integrity.
Steady-state inactivation was estimated by measuring peak Ca 2+ currents in response to a 50 ms (or 10 ms, when using the a 1A IM/EE mutant subunit) depolarizing test pulse (to +20 mV) from a holding of À80 mV, following 30-sec steps to various holding potentials (conditioning pulses) between À80 and +20 mV (Fig. 1A). Between the 30-sec conditioning depolarizations and the test pulse we employed a 20-msec interpulse to the holding potential, which does not allow detectable recovery from inactivation of Ca V 2.x channels (Degtiar et al. 2000). This kind of protocol has been reported to detect significant increase in Ca V 2.x steady-state inactivation induced by syntaxin-1A (i.e., a left shift of V 1/2 inact to more negative voltages by~6 mV) (Degtiar et al. 2000). On the contrary, when using shorter (few seconds) conditioning pulses, the influence of syntaxin-1A on Ca V 2.x channel gating was barely detectable (Degtiar et al. 2000). These results are consistent with the action of SNAREs on slow rather than fast channel inactivation (Degtiar et al. 2000). As described in detail previously (Serra et al. 2010), normalized I Ca 2+ persistent currents were fitted to the following Boltzmann equation in order to obtain half-maximal voltage (V 1/2 inact ) and slope factor (k inact ) for steady-state inactivation: (1)

Statistics
Data are presented as the means AE SEM, and n represents the number of cells recorded for each experimental condition. Statistical significance was tested using one-way Analysis of Variance (ANOVA) followed by a Bonferroni post hoc test. Differences were considered significant if P < 0.05. All statistical comparisons were performed using the GraphPad Instat software. All data are sampled from Gaussian (normal) distributions (tested using the method Kolmogorov and Smirnov). . Currents were elicited by 50 ms depolarizing steps to +20 mV applied after 30-sec depolarizing prepulses to the indicated voltages. Amplitudes of currents elicited by test pulses to +20 mV after the different prepulses were normalized to the maximum current amplitude obtained after a 30-sec prepulse to À80 mV in order to generate the corresponding mean steady-state inactivation curves (D), which were fitted to a single Boltzmann function (see Methods, eq. 1) to estimate the half-inactivation potentials (V 1/2 inactivation) (E) for WT Ca V 2.

Results
The impact of the SNARE protein syntaxin-1A on the steady-state inactivation of Ca 2+ currents (I Ca 2+ ) through wild-type (WT) Ca V 2.1 channel containing the regulatory b 3 subunit (WTb 3 ), was measured at intermediate (1 mmol/L EGTA) and high (10 mmol/L BAPTA) intracellular Ca 2+ -buffering conditions to evaluate its calcium dependency. In both conditions, syntaxin-1A expression favored channel steady-state inactivation, as indicated by a significant left shift of V 1/2 inact to more negative voltages (by~5-9 mV) (Fig. 2B-E; Table 1). It must be noted that steady-state inactivation of WTb 3 channels was poorly dependent on intracellular Ca 2+ concentration, as not significant differences were found when comparing  Table 1).
Interestingly, the introduction of a small deletion around the A454 residue at the first intracellular loop of the pore-forming a 1A subunit (DLI-II 451-457 ) (Fig. 1) made the steady-state inactivation of b 3 -containing Ca V 2.1channels (DLI-IIb 3 ) Ca 2+ -dependent. On one hand, I Ca 2+ inactivation was reduced (with a significant~8 mV right shift in the V 1/2 inact ) by increasing the buffering of intracellular Ca 2+ ( Fig. 3A Table 1), when Ca 2+ entry through the channel promotes inactivation. The effect of syntaxin-1A on the inactivation of the DLI-IIb 3 channel was recovered (V 1/2 inact was significantly left-shifted by~6 mV) by increasing intracellular Ca 2+ buffering to abrogate the novel LI-II deletion-induced Ca 2+ -dependent component of inactivation ( Fig. 3B and C [right panel], D; Table 1).
As widely reported before (for a review see Buraei and Yang 2010), Ca V 2.1 inactivation was substantially rightshifted to more depolarized potentials for b 2a -containing than for b 3 -containing channels (Fig. 4 vs. Fig. 2; Table 1). Under this condition, unlike WTb 3 channels, the steady-state inactivation of I Ca 2+ through the b 2a -containing WT Ca V 2.1 channel (WTb 2a ) presented a Ca 2+dependent component (Fig. 4) that is not affected by the deletion in the first intracellular loop of the a 1A subunit (DLI-II 451-457 ) (Fig. 5). Thus, V 1/2 inact was significantly shifted to less negative values for both WTb 2a and DLI-IIb 2a channels (by~9 mV) when increasing intracellular Ca 2+ buffering ( Fig. 4A and B Table 1). Accordingly, such right shift in the voltage dependence of inactivation disappeared once the b 2acontaining Ca V 2.1 channel was rendered insensitive to  Ca 2+ by the introduction of a double mutation (IM to EE) at the calmodulin-binding IQ-like motif ( Fig. 6A Table 1), and the SNARE protein only shifted V 1/2 inact to more negative potentials (by~5 mV) under high intracellular Ca2+ buffering ( Fig. 7B and C (right panel), D; Table 1).

Discussion
Taken together, our results bring to light a functional cross talk between three different signaling pathways regulating Ca V 2.1 channel steady-state inactivation: (1) regulatory b subunits through their interaction with the a interaction domain (AID) located at the first intracellular loop (LI-II) of the Ca V 2.1 pore-forming a 1A subunit (Buraei and Yang 2010), (2) Ca 2+ -calmodulin binding to the IQ-like motif at the a 1A C-tail (DeMaria et al. 2001;Cens et al. 2006; this report), and (3) syntaxin-1A, quite possibly, via its binding to the synprint site at the intracellular loop between domains II and III (LII-III) of a 1A (Sheng et al. 1994Rettig et al. 1996;Kim and Catterall 1997;Jarvis et al. 2002).
As previously reported for fast inactivation (Lee et al. 2000), we observed a substantial Ca 2+ -dependent component in the steady-state inactivation of Ca V 2.1 only in the presence of the palmitoylated, membrane-anchored b 2a subunit (which, contrary to other regulatory b subunits (such as b 1 or b 3 ), reduces voltage-dependent inactivation (Birnbaumer et al. 1998)). In agreement with findings from DeMaria et al. (2001) on Ca V 2.1 fast inactivation, the Ca 2+ -dependent component of the steady-state slow inactivation required the Ca 2+ -calmodulin-binding IQ-like motif, with no detectable role of the previously involved CBD site (Lee et al. 2000). Hence, Ca V 2.1 Ca 2+ -dependent steady-state inactivation was abolished by the introduction of the double mutation IM/EE at the IQ-like motif, but unaffected by a truncation of the a 1A C-tail, downstream the IQ-like motif, that fully removes the CBD site. Besides, the Ca 2+ -dependent component of Ca V 2.1 steady-state inactivation seems to depend also on specific conformational changes induced by the binding of the functionally different b subunits at the a 1A LI-II. Thus, the introduction of a LI-II deletion (DLI-II 451-457 ) downstream the AID, around the A454 residue (of relevance for the modulation of Ca V 2.1 inactivation by b subunits and SNAREs (Serra et al. 2010)), made Ca 2+ -sensitive the steady-state inactivation of b 3 -containing Ca V 2.1 channels. Such Ca V 2.1 LI-II deletion affects a fragment of a poorly conserved LI-II region of 13 amino acids that in the cardiac Ca V 1.2 channel can bind Ca 2+ -calmodulin (Pitt et al. 2001) (Fig. 8). Still, removal of this Ca V 1.2 region (DLI-II 520-532 ) did not abolish the Ca 2+ -dependent component of cardiac channel inactivation (Pitt et al. 2001). This result agrees with our observation that DLI-II 451-457 had no effect on the Ca 2+ -dependent inactivation of Ca V 2.1 channels containing the b 2a subunit.
Interestingly, syntaxin-1A was only able to modulate Ca V 2.1 steady-state inactivation when the Ca 2+ -dependent component was absent either because of the presence of b 3 in a channel formed by a a 1A subunit with unaltered LI-II, or due to the removal of channel Ca 2+ -sensitivity by high intracellular Ca 2+ buffering or by mutation of the IQ-like motif.
Whether the above described functional cross talk is due to a three-dimensional rearrangement of the involved a 1A intracellular domains (i.e., LI-II, LII-III and C-tail), and the subsequent alteration of the interaction pattern between them and/or with their interacting partners (regulatory b subunits, SNARE proteins, and the Ca 2+ -calmodulin complex), remains to be elucidated. However, there is evidence that make this hypothesis plausible since it has been reported that N-tail, intracellular loop between domains III and IV (LIII-IV) and C-tail regions of Ca V 2.x or Ca V 1.2 a 1 subunits modulate channel inactivation through direct and dynamic interactions with LI-II, or indirectly via regulatory b subunits (Geib et al. 2002;Kim et al. 2004;Stotz et al. 2004). To date, there are no structural data regarding the whole Ca V 2.1 channel complex that allow us to confirm these physical interactions between a 1A cytoplasmic domains. Nevertheless, the cryoelectron microscopy (cryo-EM) structure of the rabbit Ca V 1.1 complex, containing the pore-forming a 1S and the   . Currents were elicited by 50-ms depolarizing steps to +20 mV applied after 30-sec depolarizing prepulses to the indicated voltages. Amplitudes of currents elicited by test pulses to +20 mV after the different prepulses were normalized to the maximum current amplitude obtained after a 30-sec prepulse to À80 mV in order to generate the corresponding mean steady-state inactivation curves (C), which were fitted to a single Boltzmann function (see Methods, eq. 1) to estimate the half-inactivation potentials (V  Figure 8. Sequence alignment of intracellular loop between domains I and II (LI-II) of human Ca V 2.1 channel a 1A subunit and rabbit Ca V 1.2 channel a 1C subunit. Alignments were performed with Clustal Omega (www.ebi.ac.uk/Tools/msa/clustalo/). AID sequences are shown in purple, and rabbit (rb) a 1C LI-II site for Ca 2+ -calmodulin binding is highlighted in green. Position of the human (h) a 1A LI-II deletion around A454 (in red) (DLI-II 451-457 ) is shown in orange. "*" identical residues; ":" conservative substitutions (same amino acid group); "." semi-conservative substitution (similar shapes). LI-II residues appear in bold. Figure 9. Sequence alignment of intracellular domains (LI-II, LIII-IV, and C-tail) of human Ca V 2.1 channel a 1A subunit and rabbit Ca V 1.1 channel a 1S subunit. Alignments were performed with Clustal Omega (www.ebi.ac.uk/Tools/msa/clustalo/). AID sequences at LI-II are shown in purple, and the human (h) a 1A LI-II deletion around A454 (in red) (DLI-II 451-457 ) is highlighted in orange. Amino acids involved in the physical interaction between LIII-IV and C-terminal domain (CTD) of rabbit (rb) a 1S , according to cryo-EM structural data (Wu et al. 2016), and the homologous sequences in ha 1A are shown in brown. Sequences of the IQ (rba 1S ) and IQ-like (ha 1A ) motifs are shown in green. ha 1A CBD sequence is depicted in blue. Cytoplasmic segments that were invisible in the cryo-EM structure of Ca V 1.1 rba 1S subunit (Wu et al. 2016) are shown in gray. "*" identical residues; ":" conservative substitutions (same amino acid group); "." semi-conservative substitution (similar shapes). LI-II, LIII-IV, and CTD residues appear in bold.
ª 2018 The Authors. Physiological Reports published by Wiley Periodicals, Inc. on behalf of The Physiological Society and the American Physiological Society regulatory a 2 d 1 , b 1a , and c subunits, has been recently resolved with high, near-atomic (3.6 A) resolution (Wu et al. 2016). The structural analysis provides an atomic model for a potentially inactivated state of the Ca V 1.1 channel. In relation to the three-dimensional arrangement of a 1S intracellular domains, the structural data locate the AID motif at LI-II packed in between the regulatory b 1a subunit and the voltage sensor of a 1S domain II, and shows the formation of a globular helical domain due to the interaction between LIII-IV and the proximal C-tail (upstream the IQ motif) (Wu et al. 2016). The substantial homology between rabbit a 1S and human a 1A subunits with regard to residues involved in such LIII-IV/C-tail physical interaction (Fig. 9) suggests a similar scenario for the Ca V 2.1 channel. Unfortunately, several cytoplasmic segments were not visible in the cryo-EM structure of Ca V 1.1 a 1S subunit, and the structure of LI-II upstream the AID, the whole LII-III and the C-tail after residue D1515 (including the IQ motif) could not be resolved (Wu et al. 2016). Therefore, there are no structural data available neither on the possible interaction of LI-II with either LIII-IV or the C-tail, or on any interaction involving LII-III. Biochemical experiments with recombinant proteins in vitro strongly indicate that the synprint site, located at LII-III of a 1A/B , serves an important anchoring function that may facilitate SNARE's modulation of Ca V 2.1 and Ca V 2.2 gating (Sheng et al. 1994Rettig et al. 1996;Kim and Catterall 1997;Jarvis et al. 2002). Nonetheless, functional studies also suggest that the regulatory action of SNAREs might involve binding to other sites in the poreforming a 1 channel subunit, and LI-II and the C-tail regions have been proposed as candidates (Bezprozvanny et al. 2000;Serra et al. 2010). Supporting this idea, recent findings show that low voltage-activated Ca V 3.x (T-type) a 1 channel subunits, which do not contain the consensus synprint site, biochemically interact with syntaxin-1A and SNAP-25 at the carboxy-terminal domain (Weiss et al. 2012). In particular, syntaxin-1A binding to Ca V 3.x channels potently modulates channel gating in a similar way that found for Ca V 2.x channels (Weiss et al. 2012). Besides, Ca V 3.x-SNAREs interaction also appears essential for T-type channel-triggered low-threshold exocytosis (Weiss et al. 2012), thus providing a molecular mechanism for their coupling to neurotransmitter and hormone release in neurons and neuroendocrine cells near resting conditions or during mild stimulations (Carbone et al. 2014).
In conclusion, our data suggest that conformational modifications of a 1A LI-II (due to the binding of a particular regulatory b subunit, mutation A454T (Serra et al. 2010), or deletion DLI-II 451-457 ) determine the modulation of Ca V 2.1 steady-state inactivation either by Ca 2+ or by SNAREs but not by both.