Alternative Splicing at C Terminus of CaV1.4 Calcium Channel Modulates Calcium-dependent Inactivation, Activation Potential, and Current Density

Background: Alternative splicing diversifies calcium channel structure to change channel properties. Results: Extensive C-terminal alternative splicing generates channels differing in activation potential and voltage- and calcium-dependent inactivation properties. Conclusion: Diversification of channel function through altered structure is fine-tuned by alternative splicing. Significance: CaV1.4 C-terminal splice variations recapitulate some aspects of native photoreceptor calcium currents. The CaV1.4 voltage-gated calcium channel is predominantly expressed in the retina, and mutations to this channel have been associated with human congenital stationary night blindness type-2. The L-type CaV1.4 channel displays distinct properties such as absence of calcium-dependent inactivation (CDI) and slow voltage-dependent inactivation (VDI) due to the presence of an autoinhibitory domain (inhibitor of CDI) in the distal C terminus. We hypothesized that native CaV1.4 is subjected to extensive alternative splicing, much like the other voltage-gated calcium channels, and employed the transcript scanning method to identify alternatively spliced exons within the CaV1.4 transcripts isolated from the human retina. In total, we identified 19 alternative splice variations, of which 16 variations have not been previously reported. Characterization of the C terminus alternatively spliced exons using whole-cell patch clamp electrophysiology revealed a splice variant that exhibits robust CDI. This splice variant arose from the splicing of a novel alternate exon (43*) that can be found in 13.6% of the full-length transcripts screened. Inclusion of exon 43* inserts a stop codon that truncates half the C terminus. The CaV1.4 43* channel exhibited robust CDI, a larger current density, a hyperpolarized shift in activation potential by ∼10 mV, and a slower VDI. Through deletional experiments, we showed that the inhibitor of CDI was responsible for modulating channel activation and VDI, in addition to CDI. Calcium currents in the photoreceptors were observed to exhibit CDI and are more negatively activated as compared with currents elicited from heterologously expressed full-length CaV1.4. Naturally occurring alternative splice variants may in part contribute to the properties of the native CaV1.4 channels.

the rod photoreceptor and retina bipolar synapses. Mutations in the CACNA1F gene encoding Ca V 1.4 channels have been associated with congenital stationary night blindness type-2 (CSNB2), a condition characterized by various visual impairments in addition to night blindness (1,2). Electroretinography of CSNB2 patients suggests that Ca V 1.4 mediates neurotransmitter release at the synapse (3).
The electrophysiological properties of heterologously expressed Ca V 1.4 included a slow voltage-dependent inactivation (VDI) and a unique absence of calcium-dependent inactivation (CDI) (4 -6). CDI is a property well documented in the other LTCCs as follows: Ca V 1.2, Ca V 1.3, and recently, Ca V 1.1 channels (7). In the Ca V 1.4 pore-forming ␣1-subunit, an autoinhibitory domain (ICDI) was recently identified that resides in the distal end of the C terminus, and it profoundly blunts CDI by interacting with the calcium-sensing and inactivation-coupling machinery in the proximal C terminus (8 -11). The electrophysiological properties of Ca V 1.4 channels are supportive of its physiological role as glutamate release at the photoreceptor synapse occurs in a sustained and graded manner. However, the electrophysiological properties characterized using heterologously expressed Ca V 1.4 channels differed from the properties of the native Ca 2ϩ currents measured at the rod photoreceptors and bipolar cells (12)(13)(14). The heterologously expressed channels activated at membrane potentials that were ϳ30 mV more positive compared with the rod Ca 2ϩ currents. In addition, Ca 2ϩ currents measured in rod photoreceptor and bipolar synapses do exhibit CDI (15)(16)(17). These discrepancies suggest that the endogenous Ca V 1.4 channels could be modulated to activate at more hyperpolarized potentials as well as exhibit CDI. This may occur by mechanisms including cytoplasmic Ca 2ϩ protein buffers or post-transcriptional modifications such as alternative splicing. Interaction with a calciumbinding protein CaBP4 was shown to negatively shift the voltage of activation by 10 mV (18). Additional shifts in activation potential and modulations to other electrophysiological properties may be brought about by other mechanisms. Indeed, in other LTCCs like Ca V 1.2 and Ca V 1.3, alternative splicing has been found to modulate their biophysical and pharmacological characteristics (19 -21). Alternative splicing in voltage-gated calcium channels is extensive throughout the length of calcium channel coding region. In Ca V 1.2, at least 12 exon loci in the gene are alternatively spliced with up to 12 different variations of splicing occurring in one locus (compiled in Ref. 22). Other examples include seven alternatively spliced loci in Ca V 2.1 and six alternatively spliced loci in Ca V 3.1 (23,24). However, so far, only three alternatively spliced loci have been described for Ca V 1.4 (25).
Here, we systematically screened for alternatively spliced exons in Ca V 1.4 transcripts from the human retina using the "transcript-scanning method" (23,24,26) and showed that Ca V 1.4 is extensively alternatively spliced. We characterized the alternatively spliced exons in the C-terminal region and uncovered a population of splice variants that can activate within the photoreceptor operating potentials as well as exhibit CDI. Other electrophysiological properties that were altered include voltage-dependent inactivation (VDI) and current density.

EXPERIMENTAL PROCEDURES
Reference Sequences-The genomic sequence for Ca V 1.4 was from GenBank TM accession number AJ006216. The reference Ca V 1.4 cDNA sequence was from GenBank TM accession number AF201304. This sequence was considered as wild type (WT) in this work. Ca V 1.4 exon positions were determined by aligning the cDNA sequence against the genomic sequence (using the MegAlign module of the Lasergene software suite, DNAS-TAR), and the exons were number in order from 1 to 48. The reference amino acid sequence used was from GenBank TM accession number NP005174. The reference Ca V 1.2 cDNA sequence used was from GenBank TM accession number NM000719.
Nomenclature for Describing Alternatively Spliced Exon Variant-Various suffixes and a prefix applied to an exon number are used to describe the type of alternative splicing that occurred at that exon locus (supplemental Fig. 6). ⌬ denotes a cassette exon that was skipped in the course of alternative splicing. a or d denote the changes to exon length at the acceptor or donor site, respectively. ϩ or Ϫ indicates if the exon is lengthened or shortened, respectively. i suffix denotes retained intron, and x denotes mutually exclusive exons. * denotes a novel exon (not found in public database).
Transcript-scanning Method-To determine the different alternatively spliced exons in the Ca V 1.4 gene, we employed the transcript strategy previously described by Mittman and coworkers (24,26) and Soong et al. (23). In this method, we first designed PCR primer pairs that span at least two exons or four splice boundaries along the length of the Ca V 1.4 gene (supplemental Fig. 1). Sufficient pairs of primers were made such that amplicons form overlapping segments along the entire Ca V 1.4 sequence. The primers, designed using Oligo Primer Analysis Software (Molecular Biology Insights), are given in supplemental Table 7. PCR was performed using human retina Marathon-Ready cDNA (Clontech; catalog no. 7449-1, lot no. 0120657 and 3060598) as template and TaqDNA polymerase (Promega). Each pair of primers produced amplicons of varying sizes, corresponding to different alternative splice variants (or nonspecific products), visualized as multiple bands on an agarose gel (supplemental Fig. 1). Each band was extracted and ligated into pGEM-T Easy vector and transformed into DH10B Escherichia coli. For every band cloned, 8 -30 positive transformants (indicated by blue/white colony selection) were picked and further PCR screened using specific primers. Colonies yielding different sized PCR products were expanded and the plasmid DNAs extracted for DNA sequencing. The DNA sequences were analyzed by comparison with the Ca V 1.4 genomic and cDNA sequences to identify the type of alternative splicing that had occurred and to determine the exact location of the alternative exon-intron splice junctions as well as their adherence to the "gt . . . ag" rule. To transcript scan exon 1, we made use of the Marathon adaptors that were ligated to the ends of Marathon-ready cDNAs. The adaptor primer was provided by the manufacturer, and this was paired with a reverse primer residing on exon 2 or exon 3. To scan exon 48, we used a forward primer from exon 47 or 46 paired with an oligo(dT) that annealed to the poly(A) tail. As one primer in each pair was nonspecific (i.e. adaptor primer, oligo(dT)), the PCRs yielded diffused and multiple bands when separated on agarose gel. These were extracted, cloned, and rigorously screened in the same manner as described above.
Construction of Cloned Full-length Ca V 1.4 Library-Primers that reside in the 5Ј-and 3Ј-untranslated region (UTR) of the Ca V 1.4 gene were employed for long PCR amplification using human retina cDNA as template and the Elongase Enzyme Mix (Invitrogen), a polymerase mixture having proofreading function and capable of long PCRs. The resultant ϳ6-kb amplicons were subcloned into pCR-XL-TOPO vector (Invitrogen) and transformed into DH10B bacteria. Positive transformants were then picked and grown in 96-well plates. Random clones were picked and verified by restriction digest profiling as well as DNA sequencing.
Screening Full-length Library to Determine Abundance of Splice Variants-Pairs of primers were selected that flank exons shown by transcript scanning to have alternatively spliced variants. The clones in the full-length library were then screened using each selected pair of primers by PCR. The clones that produced PCR products and migrated with the expected sizes (WT or alternative splicing) on the agarose gel were counted. Clones were picked at random for sequencing to verify the validity of the screen. The quantity of alternatively spliced exon variants was expressed as a percentage of the total number of clones counted (supplemental Fig. 3).
Generation of Expression Constructs for Electrophysiology Characterization-Ch-WT is a chimera consisting of a Ca V 1.2 backbone and the entire C terminus from Ca V 1.4 cloned in a mammalian expression vector. The Ca V 1.2 fragment was excised using HindIII and BclI restriction sites from the ␣1C77WT clone (kindly provided by Dr. Roger D. Zühlke, University of Bern, Switzerland). The Ca V 1.4 C-terminal fragment was PCR-amplified from the human Ca V 1.4-pCDNA3.1 clone (kindly given by Dr. J. E. McRory, University of British Columbia, Canada), with the addition of a NotI site after the stop codon. The PCR also introduced a silent mutation to remove the second of the two BclI sites. The Ca V 1.2 fragment was ligated to Ca V 1.4 at the remaining BclI site, within a region of sequence identity in the IVS6 domain. The final chimeric construct was assembled in the pCDNA3 vector (Invitrogen) between HindIII and NotI sites.
Subsequent constructs containing the Ca V 1.4 C terminus alternatively spliced exons were generated by appending the C terminus of Ch-WT. DNA fragments of Ca V 1.4 C terminus that contained either ⌬37, 42dϩ, 43*, or 45aϪ splicing were obtained by PCR or restriction digests from Ca V 1.4 clones in the full-length library. These were substituted into Ch-WT using various restriction sites: ⌬37, NsiI/NotI; 42dϩ, SbfI/SbfI; 43* and 45aϪ, BamHI/XhoI. For the deletion constructs, different fragments of the Ca V 1.4 C terminus were first PCR-amplified with the insertion of stop codon to terminate at His-1718, Arg-1755, Ala-1797, Gln-1835, and Leu-1878 for the clones Ch-e43, Ch-e44, Ch-e45, Ch-e46, and Ch-⌬ICDI, respectively. These were substituted into Ch-WT using BamHI and XbaI sites.
The ICDI-mCherry construct consists of the ICDI domain of Ca V 1.4 fused in-frame to a red fluorescent protein, mCherry. ICDI was PCR-amplified from Ca V 1.4-pcDNA3.1 template with the addition of Kozak sequence and start codon. An XbaI restriction site replaced the endogenous stop codon. The mCherry fragment was amplified from the pRSET-B mCherry clone (gift from Dr. Roger Y. Tsien, University of California, San Diego) with the addition of an XbaI site at the 5Ј end and a stop codon at the 3Ј end. These fragments were assembled in a pCDNA3 vector between EcoRI and NotI sites.
The mutant construct Ch-P2A1 contains two mutations P1766A and P1767A, and Ch-P2A2 contains two mutations P1773A and P1774A. Ca V 1.4 C-terminal fragments containing either pair of mutations were first generated using overlap-extension PCR (27) and then substituted into Ch-WT using BamHI and XbaI sites. The integrity of all constructs was validated by DNA sequencing.
Transient Expression of Calcium Channels in HEK-293 Cells-HEK-293 cells were co-transfected with 1.75 g of ␣ (Ch-WT or variants), 1.25 g each of ␤2a and ␣2␦-subunits (kind gifts of Dr. Terry P. Snutch, University of British Columbia, Canada), and 0.2 g of T-antigen plasmids using the calcium phosphate method. The cells were then incubated for at least 36 h in a water-saturated 5% CO 2 incubator at 37°C before whole-cell patch clamp recordings were made.
Whole-cell Patch Clamp Electrophysiology-Whole-cell patch clamp recordings were performed on transfected cells between 36 and 72 h after transfection. The external bath solution was composed of the following (in mM): 10 HEPES, 140 tetraethylammonium methanesulfonate, 5 BaCl 2 or CaCl 2 , pH 7.4, 300 -310 mosM (adjusted with glucose). The pipette internal solution contained (in mM): 138 Cs-MeSO 3 , 5 CsCl, 0.5 EGTA, 10 HEPES, 1 MgCl 2 , 2 mg/ml Mg-ATP, pH 7.3, 290 -300 mosM (adjusted with glucose). The junction potential was determined to be Ϫ11 mV, and the voltages reported here are uncorrected for junction potential. True voltages may be calculated by subtracting 11 mV from the reported values. Wholecell currents were obtained under voltage clamp with either the Axopatch 200A or Axopatch 200B amplifier, operated using pClamp software (Molecular Devices). The signals were filtered at 1-5 kHz and sampled at 5-50 kHz. Series resistance was typically 1.2-1.8 megohms after 80% compensation. A P/4 protocol was applied on line to subtract leak and capacitive transients.
To assess the current-voltage (I-V) relationship of the channels, transfected cells were depolarized to a family of test potentials of Ϫ60 to 60 mV, in steps of 10-mV increments, from a holding potential of Ϫ90 mV. The peak current evoked by each voltage was normalized to the maximal current obtained for each cell recorded and fitted with Equation 1, where G max is the maximum conductance; E rev is the reversible potential; V1 ⁄2 , act is the half-activation potential, and k act is the slope.
For steady-state activation, we analyzed the tail currents (G) obtained at the end of a short depolarizing pulses. Cells were held at Ϫ90 mV before depolarizing to a family of voltages ranging from Ϫ60 to 100 mV, in steps of 10-mV increments, for 20 ms. Following that, a repolarization to Ϫ50 mV for 10 ms evokes the tail currents that are measured. The peak of each tail current was normalized to the maximum obtained for each cell recording and fitted with dual Boltzmann Equation 2, where G is the tail current and G max is the maximum tail current; F low is the fraction of the low threshold component; V1 ⁄2 , low, V1 ⁄2 , high, k low , and k high are the half-activation potentials and slope factors for the low and high threshold components, respectively, and V1 ⁄2 , act may be calculated when G ϭ 0.5 G max .
Steady-state inactivation properties were determined by comparing the test current obtained after a long depolarizing pulse to an initial pre-pulse. For the Ϫ90-mV holding potential, a pre-pulse current was evoked by stepping to 10 mV for 30 ms. This was followed by a family of 15-s-long depolarizing pulses ranging from Ϫ120 to 20 mV. Next, a test pulse current was evoked by stepping the cell to 10 mV for 100 ms. The peak current evoked by each test pulse was divided by the peak current evoked by the pre-pulse to obtain the normalized current. The values were fitted with a single Boltzmann Equation 3, where I relative is the normalized current; V1 ⁄2 ,inact is the potential for half-inactivation, and k inact is the slope value.
The strength of CDI exhibited by the channel is expressed by the f value (28). To determine the f value, first the current amplitude that remained at a given time point (i.e. 30, 50, 100, 20, and 300 ms) after depolarization was measured and normalized against the peak current to obtain the residual current (i.e. r 30 and r 300 for 30 and 300 ms, respectively). Next, the f value was calculated by subtracting the residual I Ba with the residual I Ca . An f value of 0 indicates no CDI, whereas the maximum f value of 1 indicates complete CDI. For comparison between channels, we compared the f values obtained from current evoked by depolarization to 10 mV. The presence of CDI may also be seen in the profile of residual I Ca evoked by a family of voltages. When plotted against voltage, the "U"-shaped dependence of residual I Ca values on voltage is a tell-tale trait of CDI.
For VDI, cells were placed in the external solution with Ba 2ϩ as the charge carrier and depolarized to V max , from a holding potential of Ϫ90 mV. As time elapsed, the current evoked by V max decayed, and the amplitude was measured at different time points. These were subtracted from the peak current evoked to obtain the inactivated current. This was then expressed as a percentage of the peak current (percentage inactivation).
Fractional recovery from inactivation was determined using a two-pulse protocol and following a 2-s depolarizing pre-pulse to V max (determined from the I-V properties of the transfected cell), and a test pulse was applied after a certain time period, ⌬T, has elapsed. The peak I Ba measured during the test pulse was expressed as a fraction of peak I Ba obtained at the beginning of the pre-pulse to obtain the fraction of recovery. This was plotted against ⌬T. The curve was obtained by fitting the values with a double exponential Equation 4, where Y is the fraction of recovery, A f and A s are the maximum values of the fast and slow component, and f and s are their time constants, respectively.
Data and Statistical Analyses-All values presented are means Ϯ S.E. Electrophysiology data were analyzed using Clampfit software (Molecular Devices). Data values were exported to Microsoft Office Excel software (Microsoft Corp.) for calculations. Data plotting, curve fitting, and statistical analyses were performed using GraphPad Prism software (GraphPad Software, Inc.). Statistical significance of the difference between means was determined using Student's t test or one-way analysis of variance. Statistical significance was accepted at p Ͻ 0.05.

Transcript Scanning of Ca V 1.4 from Human Retinal cDNAs-
To determine the different alternatively spliced exons in the Ca V 1.4 gene, we employed the transcript scanning strategy that has been previously reported (23,24,26). As the Ca V 1.4 transcripts are predominantly expressed in the retina (1, 2, 4), we used human retina cDNA in PCRs to amplify across all the splice junctions of the Ca V 1.4 gene; and we generated at least 15 overlapping PCR amplicons (supplemental Fig. 1). Bacterial colonies obtained from plasmids subcloned with each of the 15 PCR amplicons were screened, and the individual clones harboring different size PCR products were subject to automated DNA sequencing (supplemental Fig. 2). The identity and mechanism of the alternative splicing for each splice locus were identified by inspection of the exon-intron boundaries and determined by analysis of the DNA sequences in comparison with the Ca V 1.4 genomic sequence ( Fig. 2; GenBank accession number AJ006216).
Transcript scanning of Ca V 1.4 from human retina cDNA revealed that 16 exons are alternatively spliced, of which some may undergo more than one mode of alternative splicing with some adjacent exons alternatively spliced in tandem (Fig. 2); this results in 19 different exon variants. These include cassette exons that can be fully excluded by alternative splicing and alternate use of splice donor/acceptor junctions that resulted in lengthening or shortening of the exons and mutually exclusive exons. Two of the cassette exons are novel and have not been predicted in the public database. In all cases, the canonical "gt . . . ag " splice junctions were preserved at the intron-exon junctions of the alternative splice sites (47). Not all alternative splicings were productive as some resulted in frameshifts that gave rise to premature stop codons. The relative positions of the alternatively spliced loci along the entire length of the Ca V 1.4 open-reading frame are illustrated in Fig. 1. Alternative splicing at the C terminus produced four splice variants (⌬37, 42dϩ, 43*, and 45aϪ) that contained different lengths of the C-terminal tail. The ⌬37 splice variant is truncated the entire C terminus, although in contrast the use of an alternative donor site in 42dϩ adds an additional 13 amino acids to the region downstream of the IQ motif. The inclusion of the novel alternative exon 43* adds a novel 19-amino acid peptide past exon 42 but contains a stop codon that resulted in premature termination of the C terminus with a truncation of the last 257 amino acids. Conversely, 21 amino acids are removed in the 45aϪ variant by the use of an alternative acceptor site with no frameshift of downstream codons. This occurs in a region of the C terminus that has poor conservation among LTCCs (supplemental Fig.  3). Notably, the splice variants 16aϪ and 17aϩ 1 , ⌬16 and 17aϩ 2 , and ⌬17 and 18a, are pairs of adjacent alternatively spliced exons that when spliced in tandem gave rise to transcripts coding for functional Ca V 1.4 channels. Here, alternative splicing of the first exon generated a frameshift that was corrected by the alternative splicing in the second exon. Had alternative splicing occurred independently in one locus frameshift would result in a premature stop codon. This mode of splicing is interesting and unusual because it manages to alter the gross peptide sequence without relying on a different set of nucleotide sequences.
We also transcript-scanned the bracketing exons (i.e. exon 1 and exon 48) by using Marathon-Ready human retinal cDNA (Clontech) (29) as template in PCRs that contained a primer that targets the sequence in the Marathon adapter and a Ca V 1.4-specific primer. Based on the results from the PCR screen, coupled with DNA sequencing of clones, we found no additional alternate exons to exon 1 or exon 48.
Abundance of Each Alternative Splice Locus within Ca V 1.4 Transcripts-Proteomic variations generated by alternative splicing may alter the electrophysiological properties of Ca V 1.4 channels. The relative abundance of each Ca V 1.4 splice variant would no doubt influence the overall property of the native Ca V 1.4 currents. To determine the levels of occurrence of the splice variants, we first PCR-amplified and cloned full-length Ca V 1.4 cDNAs from human retina to obtain a pool of fulllength Ca V 1.4 clones. Using primer pairs that target alternatively spliced loci, we characterized the full-length Ca V 1.4 clones by PCR screening to identify the combinatorial splicing pattern of each clone. Clones producing PCR product size alterations that tallied with the expected changes were counted and expressed as a percentage of the total population (Fig. 2, left  column). Random clones were selected for DNA sequencing to validate the PCR screen. The most abundant splice variant we have found was 16aϪ (61.7%). Here, the alternative use of an alternative splice acceptor site at exon 16 resulted in a frameshift with a consequent stop codon at IIS6 domain. This variant may give rise to the formation of a hemi-channel. The most abundant productive splice variant we detected was ⌬32 (17.9%). The skipping of exon 32 shortens the IVS3-S4 linker by 7-amino acid residues. Shortening of the IVS3-S4 linker by alternative splicing may be a common mechanism for diversity among LTCC. In Ca V 1.1, alternative splicing at this locus can include 10% of transcripts in adult muscle and more than 66% in regenerating muscle (30). In Ca V 1.2, alternative splicing of exons encoding IVS3 and the IVS3-S4 linker could produce up to 12 variations (22). Although the amino acid content of the IVS3-S4 linker varies greatly between the four LTCC, the length of the linker, however, may be important in modulating channel activation (22).
Electrophysiological Characterizations of Ca V 1.4 C-terminal Splice Variants-Alternatively splicing at four loci produced splice variants ⌬37, 42dϩ, 43*, and 45aϪ that have various lengths of the C terminus. Reports from previous investigations into the biophysical properties of full-length Ca V 1.4 channels have shown that these channels expressed relatively small currents (about three times lower than Cav1.2 (5)), despite the use of high barium concentration in the external solution (15-30 mM), and tend to have a low transfection efficiency (22% (6)). To screen, in a more efficacious manner, the C-terminal splice variants for distinct electrophysiological properties, we chose the chimera approach. We generated a chimeric reference construct (Ch-WT) switching the full-length C terminus of human Ca V 1.4 into the human Ca V 1.2 channel backbone (Fig. 3A). The Ca V 1.2 was selected because it is a well characterized L-type subtype and has an almost identical I-V relationship to Ca V 1.4 (5). Similar chimeras have been successfully used in a robust manner to investigate the effects of the Ca V 1.4 C terminus on channel properties and the mechanisms involved (9 -11).
The entire cytosolic tail is truncated in the ⌬37 splice variant, and this channel may be nonfunctional. However, because a CSNB2 Ca V 1.4 mutant with no cytosolic tail was previously reported to mediate current (4), we attempted to characterize this splice variant as well, but we were not able to evoke any current in the ⌬37 chimeric channel.
43* Splice Variant Inactivated More Rapidly in Ca 2ϩ and Activated in a More Hyperpolarized Potential-To assess the current-voltage (I-V) relationship of the splice variants, transfected cells were depolarized to a family of test potentials from Ϫ60 to 70 mV, in steps of 10-mV increments, from a holding potential of Ϫ90 mV (Fig. 3B). The current trace profile for Ch-43* displayed a much slower inactivating I Ba compared with Ch-WT or the other two functional C-terminal splice variants (Fig. 3C, gray traces), but it showed an early and more pro- nounced inactivation of I Ca (Fig. 3C, black traces). These inactivation properties of the channels in Ba 2ϩ compared with Ca 2ϩ will be further analyzed below.
Ch-43* also activated at a more negative potential with a hyperpolarized shift in voltage for half-maximal activation, V1 ⁄2 , act, by 21.0 mV in Ba 2ϩ (supplemental Table 1) and 19.4 mV in Ca 2ϩ (supplemental Table 2) compared with Ch-WT (Fig. 3, F and J; gray trace; both p Ͻ 0.001; unpaired t test). Such changes in channel activation are slight and less apparent in Ch-42dϩ and Ch-45aϪ. The V1 ⁄2 , act of the Ch-42dϩ hyperpolarized shift by 2.6 mV in Ba 2ϩ and 2.0 mV in Ca 2ϩ (p Ͻ 0.001; unpaired t test), whereas the V1 ⁄2 , act of Ch-45a-depolarized-shift by 1.2 mV in Ba 2ϩ and 0.5 mV in Ca 2ϩ (p Ͻ 0.001; unpaired t test). We then appended the Ca V 1.4 full-length channel with the 43* exon and found that in a similar manner, exon 43* mediated a pronounced hyperpolarized shift in activation (12.5 mV; p Ͻ 0.0001; unpaired t test) in the native channel (Fig. 4B). Also, Ca V 1.4 4.3* mediated a rapidly inactivating I Ca , whereas the WT displayed little inactivation (Fig. 4A).

43* Splice Variant Exhibited a Negative Shift in Window
Current-We have shown that exon 43* caused a pronounced shift of the I-V relationship in the hyperpolarized direction, indicating a more negatively activating channel. To enable a more accurate assessment of the voltage activation of these channels, we analyzed the tail currents (G) obtained at the end of short depolarizing pulses to various potentials (Fig. 5A). Also, to determine the inactivation properties of the channels under steady-state conditions, transfected cells were held at various potentials for 15 s, and currents evoked before and after each inactivating pulse were compared (Fig. 5B).
CDI Is Restored by 43* Splicing-Rapid inactivation of activated channels due to Ca 2ϩ is a phenomenon displayed by many high voltage activated channels. Interaction of Ca 2ϩbound calmodulin with the IQ domain triggers CDI; however, in the Ca V 1.4 channel the ICDI domain at the distal end of the C terminus interacts with the calcium-sensing apparatus, consisting of the EF-hand and IQ motif, to profoundly blunt CDI (8 -11). From the profiles of the exemplar traces in Figs. 3C and 4A, it was clear that the 43* variant still displayed robust CDI. To quantify the degree of CDI exhibited by the channels, the fractions of current that remained at a given time point after depolarization (residual current; i.e. r 30 and r 300 for 30 and 300 ms, respectively; Fig. 6, A-D) to different voltages were determined. The difference between the residual current of Ba 2ϩ and Ca 2ϩ is a measure of CDI strength (f value). Here, we calculated the f value obtained at 10 mV (f 10 ; Fig. 6, E and F). In addition, because the channels exhibit a fair amount of calcium-independent inactivation (VDI), we also used the ratiometric approach to calculate CDI. Here, "netCDI" is obtained by taking the ratio of the remaining current of Ca 2ϩ to Ba 2ϩ mV (netCDI 10 ; Fig. 4, I and J). This method complements the classical f value measurements by avoiding the underestimation of CDI strength in the presence of strong VDI (31-33).  For the Ch-WT channel, the r 30 for Ca 2ϩ was slightly lower than for Ba 2ϩ (Fig. 6A). This became more evident with residual currents measured at a longer depolarization time of 300 ms (Fig. 6B), indicating a small degree of inactivation by Ca 2ϩ (f 10 , 30 ms, 0.05 Ϯ 0.03; 300 ms, 0.18 Ϯ 0.06. netCDI 10 , 30 ms, 0.06 Ϯ 0.03; 300 ms, 0.33 Ϯ 0.14). In the native WT channel, however, FIGURE 5. Activation and steady-state inactivation properties of wild type and 43* splice variant channels. A, voltage protocol used for determining activation properties and an exemplar current trace (WT). Tail currents were measured after 20 ms depolarization to potentials ranging from Ϫ60 to 100 mV, in 10-mV steps. B, voltage protocol used for determining steady-state inactivation, SSI, properties, and an exemplar current trace (WT). Currents evoked at 10-mV test potentials were measure before and after a 15-s depolarization to potentials ranging from Ϫ120 to 20 mV, in 10-mV steps. A holding potential of Ϫ90 mV was maintained for all experiments. C and D, normalized plots for activation and SSI for chimeric channels Ch-WT and Ch-43*, respectively. For SSI, peak currents obtained after the 15-s inactivating pulse were normalized to that obtained before inactivation and plotted against voltage. The curves are fits with the Boltzmann relationship. For the activation plots, the peak of the tail currents (G) were normalized against the largest peak and plotted against voltage. The curves were fitted using the equation given under "Experimental Procedures." The number of cells recorded are given in parentheses. E and F, same as C and D but for native channels Ca V 1.4 WT and Ca V 1.4 43*, respectively. this is less pronounced (Fig. 6E; f 10 , 30 ms, 0.02 Ϯ 0.02; netCDI 10 , 0.02 Ϯ 0.02) and even barely detectable at 300 ms ( Fig. 6F; f 10 , 0.00 Ϯ 0.03; netCDI 10 , 0.00 Ϯ 0.03).
VDI Is Suppressed by 43* Splicing-An appreciable amount of slow inactivation in the absence of Ca 2ϩ is apparent in the current trace profiles of Ch-WT (Fig. 3C) and, to a lesser degree, Ca V 1.4 WT (Fig. 4). We next queried whether VDI may be modulated by alternative splicing in the C terminus. Fig. 6G shows the percentage inactivation of peak I Ba , evoked at V max , as it decays over various time points. Inactivation of I Ba through Ch-WT increased steadily over time; displaying 66.0 Ϯ 5.1% inactivation at 300 ms after depolarization and reaching 85.3 Ϯ 3.6% by the end of 1 s. Truncation of the distal portions of the C terminus by 43* splicing resulted in a channel that inactivated in Ba 2ϩ more slowly compared with Ch-WT. At 300 ms, I Ba inactivation for Ch-43* (21.9 Ϯ 3.2%) was about 3-fold slower than Ch-WT (p Ͻ 0.0001, unpaired t test), and at 1 s the inactivation of Ch-43* was half that of WT (Ch-WT, 85.3 Ϯ 3.6%; Ch-43* 45.7 Ϯ 4.3%; p Ͻ 0.0001, unpaired t test). The native Ca V 1.4 WT displayed a much lower VDI compared with the chimera, exhibiting only 14.3 Ϯ 2.5% inactivation at 1 s. The presence of exon 43*, however, still reduced VDI of the native channel by half (7.1 Ϯ 1.6%). 43* Splicing Mediated an Increase in Current Density-Previous reports of LTCC with C-terminal deletions had described gross increase in current density (Cav1.1 (34), Cav1.2 (35), and Cav1.2 (36)). We therefore examined the currents mediated by the 43* channels. Appending exon 43* to either the chimera or native channel both led to a significant increase in current density (Fig. 7, A and B). The increase in current density, when evoked at 60 mV, was 4.3 times in the chimera (Ch-WT, 35.6 Ϯ 6.7 pA/pF; Ch-43*, 153.2 Ϯ 27.1 pA/pF; p Ͻ 0.01, unpaired t test) and 2.1 times larger in the native channel when evoked at 0 mV (Ca V 1.4 WT, 25.5 Ϯ 4.4 pA/pF; Ca V 1.4 43*, 53.2 Ϯ 11.7 pA/pF; p Ͻ 0.01, unpaired t test). Current density is a function of the number of channels expressed on the plasma membrane, the channel's unitary conductance, and open probability. We next measured the gating charge of Ca V 1.4 WT and 43* channels. As gating charge arises from the voltage-driven movement of charged residues within the activating channel structure, this gives an index of the quantity of channels expressed on the cell membrane (Fig. 7C). Ca V 1.4 43* displayed a gating charge that was more than 6-fold smaller than the WT (Fig. 7D; WT, 1.28 Ϯ 0.14 femtocoulombs/pF; 43*, 0.19 Ϯ 0.08 femtocoulombs/pF), indicating that the larger current density it exhibits is likely due to a larger single channel conductance or open probability or both. Single channel analyses would be required to determine these two properties.
Determinants for Activation and Inactivation in Ca V 1.4 C Terminus-In the 43* splice variant, the distal portion of the C terminus is truncated. This results in the loss of the ICDI, a domain that has been previously demonstrated to be responsible for the suppression of CDI in the Ca V 1.4 channels. Removal of a short length of this ICDI had also been shown to slow VDI (10). It is conceivable that the robust CDI and slow VDI in Ch-43* can be fully attributed to the lack of the ICDI domain. However, as the ICDI is composed of only the last 100 amino acids of the C terminus, we did not rule out the possible participation of the segment between exon 43 and ICDI in channel activation and inactivation.
First, we established whether the changes to channel properties by 43* splicing can be fully reproduced by deleting only the ICDI or whether it requires a greater length of the C terminus to be deleted. We next co-expressed a peptide containing the ICDI together with the Ch-43* to determine whether the ICDI itself can reverse the effects of 43* alternative splicing (i.e. suppression of CDI as well as increased VDI). To these ends, another two chimeric channels were constructed. These consisted of the Ca V 1.2 backbone and truncated C termini of the Ca V 1.4, one terminating after exon 43 and another terminating just before the ICDI (labeled Ch-e43 and Ch-⌬ICDI, respectively; Fig. 7A). The ICDI peptide was also cloned into a mammalian expression vector and fused with an mCherry fluorescent protein (37). These Ca V ␣ 1 clones were each co-expressed, in HEK-293 cells, with the auxiliary subunits and with either the mCherry-fused ICDI peptide (ICDI-mCherry) or mCherry alone as a control. Only cells that were found to express both green and red fluorescence were selected for whole-cell electrophysiological recordings (supplemental Fig. 6).
Hyperpolarized I-V Shift by 43* Splicing Was Due to Loss of ICDI-Analysis of the I-V relationship of Ch-⌬ICDI showed that deletion of the ICDI from the full-length Ch-WT enabled a hyperpolarized shift reminiscent of that mediated by the 43* splicing (compare between Figs. 8D, gray curve, and 3F). This implies that the I-V shift shown by 43* was primarily due to the loss of the ICDI domain. Co-expression of Ch-⌬ICDI with ICDI-mCherry reversed the shift with a 16.3-and 25.4-mV more positive V1 ⁄2 , act in Ba 2ϩ and Ca 2ϩ , respectively (supplemental Tables 4 and 5, p Ͻ 0.001, unpaired t test). The slope of activation, k act , was also decreased by 3.3 and 3.7 in Ba 2ϩ and Ca 2ϩ , respectively (supplemental Tables 4 and 5, p Ͻ 0.001, unpaired t test). In the control experiments, coexpression of mCherry with the Ca V ␣ and auxiliary subunits failed to mediate the I-V shifts. Hence, these data demonstrated that the ICDI domain modulated the activation kinetics of the channel.
Whole-cell electrophysiological recordings of Ch-e43 showed that it displayed an I-V relationship similar to Ch-43* or Ch-⌬ICDI (Fig. 8, B-G). Similarly, co-expression with ICDI-mCherry led to depolarized shifts in V1 ⁄2 , act of 15.4 and 17.1 mV in Ba 2ϩ and Ca 2ϩ , respectively, as well as a decrease in k act by 4.0 and 3.0 in Ba 2ϩ and Ca 2ϩ , respectively (supplemental Tables 4 and 5).
Co-expression of Ch-43* with ICDI-mCherry also triggered a right shift in the I-V curve (Fig. 8, B and E). The presence of mCherry did not affect the channel properties as the I-V curve for the control was similar to that of Ch-43* in the previous experiments (Fig. 3, F and J). The depolarized shift in V1 ⁄2 , act effected by ICDI-mCherry was 16.8 mV in Ba 2ϩ and 19.3 mV in Ca 2ϩ (supplemental Tables 4 and 5; p Ͻ 0.001, unpaired t test). The magnitude of V1 ⁄2 , act shift in Ca 2ϩ reflected a complete reversal of the hyperpolarized shift caused by the 43* splicing. Similarly, co-expression of Ch-43* with ICDI-mCherry restored k act to values similar to Ch-WT (supplemental Tables  1 and 2).
Co-expression with ICDI-mCherry Suppressed CDI and Increased VDI in Ch-43*-The representative traces shown in Fig. 9, A-F, reflect the current profiles obtained when cells in the co-expression studies were subjected to the square-pulse protocol illustrated in Fig. 3B. For Ch-43*, Ch-e43 and Ch-⌬ICDI co-expressed with the mCherry control of the current profile matched that of Ch-43*, whereby I Ca rapidly decayed, and I Ba displayed only modest inactivation (Fig. 9, A, C, and E). However when co-expressed with ICDI-mCherry, the fast inactivation of I Ca was much retarded, and in the case of Ch-e43 and Ch-⌬ICDI, I Ca decayed as slowly as I Ba (Fig. 9, B, D, and F).
Ch-43*, Ch-e43, and Ch-⌬ICDI when co-expressed with mCherry control, exhibited a residual I Ca with a U-shaped dependence on voltage (Fig. 9, G, I, and K) that is indicative of CDI. This U-shape residual I Ca was lost when the channels were co-expressed with ICDI-mCherry and the residual I Ca attained profiles similar to residual I Ba when plotted against voltage (Fig.  9, H, J, and L).
In Fig. 9M, the calculated f 10 as well as netCDI showed that ICDI-mCherry completely eliminated CDI at 30 ms. At 300 ms (Fig. 9N), co-expression with ICDI-mCherry suppressed f 10 from 0.54 Ϯ 0.12 to 0.05 Ϯ 0.15 in Ch-43*, from 0.62 Ϯ 0.08 to 0.04 Ϯ 0.19 in Ch-e43, and from 0.57 Ϯ 0.13 to 0.08 Ϯ 0.16 in Ch-⌬ICDI. Calculated values of netCDI also indicated similar degrees of CDI abolishment. Here, ICDI-mCherry suppressed CDI to levels lower than that measured in Ch-WT. This is perhaps due to the greater abundance of the overexpressed ICDI-mCherry as well as the increased mobility that ICDI has when expressed as a separate peptide.
Taken together, the data so far demonstrated that the hyperpolarized shift in I-V relationship, the restoration of CDI, as well as slow VDI conferred by 43* splicing was predominantly due to the absence of the ICDI domain rather than other portions of the C terminus. This is because the construct with only the ICDI deleted, Ch-⌬ICDI, displayed similar activation and inactivation properties as Ch-43*. Besides, further deletion of the remaining length of C terminus between the ICDI and exon 43 did not significantly alter these properties. Moreover, co-expression with a peptide containing the ICDI was able to restore Ch-WT-like characteristics to all three constructs; however, VDI in Ch-⌬ICDI required a longer depolarization to respond. Therefore, it seems that the ICDI is a multifunction domain capable of modulating activation and inactivation properties of the channel.

DISCUSSION
Voltage-gated calcium channels are extensively and alternatively spliced, and such post-transcriptional modifications diversify the biophysical properties of the channel. In our work, we identified 19 alternative splice variants of Ca V 1.4 in the human retina, 16 of which are completely novel. Importantly, we showed that alternative splicing can restore CDI to the Ca V 1.4 channel that was previously reported to be CDIinsensitive (4,6). Also, the splice variant 43* exhibited a substantial hyperpolarized shift in activation as well as an increase in current density. We had also characterized the recovery of each C-terminal splice variant from inactivation in the chimeric channels, and we showed that 43* dramatically slowed recovery, whereas 45aϪ slowed recovery only at the early phase. The involvement of proline residues within proline-rich exon 45 was also implicated. However, because voltage inactivation was much reduced in the native channel, we were unable to recapitulate these effects here. Nonetheless, the figures for the recovery experiments are given in supplemental Figs. 8 and 9.
Although in previous studies artificially engineered truncations of the C terminus of Ca V 1.4 have determined that removal of ICDI resulted in disinhibition of CDI (9, 10), we now showed that this process may be natively regulated by alternative splicing. This is found in the 43* variant that introduces a termination signal into the C terminus, 137 residues from the IQ motif. The placement of the stop codon by the novel exon cassette is strategic because while maintaining the integrity of the calcium-sensing apparatus, it also generates a channel with a twice larger current density than the original. This increase in current density was not seen in another C-terminal truncated Ca V 1.4 (L1591X mutant) (10)), whereby the entire length of the C terminus immediately after the IQ was removed. These data suggested the importance of the additional sequences retained by the 43* splicing. In support of this, an appreciable degree of conservation across the LTCC isoforms is found in the proximal part of this region (supplemental Fig. 3). In Ca V 1.1 and -1.2, truncation of the C terminus after this region also gave rise to increased current density (Cav1.1 (34), Cav1.2 (35), and Cav1.2 (36)). In addition, a segment of this region (PCRD, supplemental Fig. 3) was also defined here to be the binding FIGURE 9. Current traces, CDI, and VDI of Ch-43* and deletion constructs in co-expression experiments. Channel constructs were co-expressed with either ICDI-mCherry or mCherry (labeled on the top) as described earlier. A-F, representative I Ba (gray) and I Ca (black) traces during a 400-ms depolarization to 10 mV. G-L, residual I Ba (gray) and I Ca (black) for the corresponding co-expression experiments above, measured at 30-and 300-ms as labeled on the axes. M and N, strength of CDI (f-value and netCDI) calculated at 10 mV for r 30 and r 300 , respectively. O, percentage of I Ba inactivation during depolarization to V max in the time course indicated. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001 (compared as indicated; unpaired t test).
target of a distal C-terminal autoinhibitory domain (DCRD) that is responsible for repressing current density. Sequences homologous to the DCRD may also be found within the ICDI of Ca V 1.4. This raises an important question of whether the PCRD and DCRD homologues of Ca V 1.4 interact. This is especially so in the context of abolishment of CDI by ICDI because the requirement for the "A" region, encompassing the PCRD, for ICDI efficacy have been reflected in several previous works (9 -11). Investigating this interaction may mold our understanding on the two mechanisms that have been described for ICDI (8,11).
The increased current density in Ca V 1.4 43* may be due to one or a combination of three causes as follows: an increase in expression of the channel on the plasma membrane, an increase in unitary conductance, and an increase in open probability. We have shown that this was not due to an increased membrane expression. On the contrary, Ca V 1.4 43* membrane expression was six times lower than WT channels. In the Ca V 1.2 truncated channel, there was neither an increase in channel expression, as shown by gating charge analyses, nor an increase in single channel conductance (35). In Ca V 1.3, there was no increase in expression density detected either (38). For Ca V 1.4, it remains to be seen, with single channel analyses, if 43* splicing could alter unitary conductance and open probability.
Our data may be used to provide insights on certain discrepancies between the currents obtained when recording from the rod photoreceptors versus the current characteristics of cloned Ca V 1.4 heterologously expressed in cells. Cloned Ca V 1.4 channels characterized in a heterologous expression system are activated by membrane depolarization to potentials that are more positive compared with the rod photoreceptor dark resting membrane potential of approximately Ϫ40 mV (17,(42)(43)(44). This posed a problem because rod photoreceptors hyperpolarize upon activation by light, apparently implying that in the normal operating range of the rod photoreceptor membrane potential Ca V 1.4 channels are minimally activated or not at all. Also, when native calcium currents in the rod photoreceptor were measured, they could be evoked at ϳ30 mV more hyperpolarized potentials than that obtained from cultured cells expressing cloned full-length Ca V 1.4 channels (12)(13)(14). It is possible that the native currents resulted from the presence of other calcium channel isoforms (i.e. T-type). Taken together, these findings seemingly excluded Ca V 1.4 from the role of inducing neurotransmission at the photoreceptor synapse (45). However given its localization together with its association with CSNB2 etiology (1, 2, 13), Ca V 1.4 activity should play an important role in normal rod photoreceptor function. The possibility of channel modulation by intracellular binding factors was thus explored, and it was found (18) that interaction with CaBP4 was able to shift the activation of the Ca V 1.4 channel by Ϫ10 mV, therefore pushing the channel into the operating range of the rod photoreceptor. In this study, we showed that post-transcriptional modification of the Ca V 1.4 channel by alternative splicing provided another means to shift channel activation potential. Alternative splicing generating the novel exon 43* mediated an ϳ10 mV more hyperpolarized activated Ca 2ϩ current.
Although 43* variants comprised only 13.6% of total Ca V 1.4 transcripts expressed in the human retina, it exhibited a 2-fold larger in current density. Therefore, the current contributed by 43* variants may make up a larger proportion of total endogenous Ca V 1.4 calcium current in the retina.
Another discrepancy between cloned Ca V 1.4 and native currents is the presence of CDI in the latter. This was detected in Ca 2ϩ currents measured from rod photoreceptors and retinal bipolar synapses (15)(16)(17). We likewise propose that the CDI observed here may have been contributed by alternatively spliced variants of Ca V 1.4 like those containing exon 43* that exhibits robust CDI. The main difference between the CDI exhibited by the 43* splicing and the CDI shown in these reports is that the native CDI occurred with slower kinetics. In addition to differences in experimental setup (i.e. levels of charged carriers used, strength of Ca 2ϩ buffering, etc.), the conditions present under the more physiological setting may modulate the rate and degree of CDI. For example, in Ca V 1.2 and Ca V 1.3, the presence of calciumbinding proteins was shown to reduce CDI strength (46). Also, CDI measured from rods and bipolar cells had different inactivation time constants (15), and CDI was stronger, and the rate of recovery from inactivation was pronouncedly slower in rods present in retinal slices than in isolated rod cells (16).