Alternative Splicing Generates a Novel Truncated Cav1.2 Channel in Neonatal Rat Heart*

Background: L-type Cav1.2 Ca2+ channel undergoes extensive alternative splicing, generating functionally different channels. Results: Neonatal rat heart expresses a higher level of a truncated Cav1.2 Ca2+ channel from alternative splicing. Conclusion: The truncated channel can alter electrophysiological properties of a wild type channel. Significance: Although aberrantly spliced Cav1.2 channels may not conduct Ca2+ ions, they can affect functional channels. L-type Cav1.2 Ca2+ channel undergoes extensive alternative splicing, generating functionally different channels. Alternatively spliced Cav1.2 Ca2+ channels have been found to be expressed in a tissue-specific manner or under pathological conditions. To provide a more comprehensive understanding of alternative splicing in Cav1.2 channel, we systematically investigated the splicing patterns in the neonatal and adult rat hearts. The neonatal heart expresses a novel 104-bp exon 33L at the IVS3-4 linker that is generated by the use of an alternative acceptor site. Inclusion of exon 33L causes frameshift and C-terminal truncation. Whole-cell electrophysiological recordings of Cav1.233L channels expressed in HEK 293 cells did not detect any current. However, when co-expressed with wild type Cav1.2 channels, Cav1.233L channels reduced the current density and altered the electrophysiological properties of the wild type Cav1.2 channels. Interestingly, the truncated 3.5-domain Cav1.233L channels also yielded a dominant negative effect on Cav1.3 channels, but not on Cav3.2 channels, suggesting that Cavβ subunits is required for Cav1.233L regulation. A biochemical study provided evidence that Cav1.233L channels enhanced protein degradation of wild type channels via the ubiquitin-proteasome system. Although the physiological significance of the Cav1.233L channels in neonatal heart is still unknown, our report demonstrates the ability of this novel truncated channel to modulate the activity of the functional Cav1.2 channels. Moreover, the human Cav1.2 channel also contains exon 33L that is developmentally regulated in heart. Unexpectedly, human exon 33L has a one-nucleotide insertion that allowed in-frame translation of a full Cav1.2 channel. An electrophysiological study showed that human Cav1.233L channel is a functional channel but conducts Ca2+ ions at a much lower level.

Voltage-gated calcium channels govern the depolarizationinduced Ca 2ϩ entry into cardiac muscles. The channel consists of a pore-forming ␣ 1 subunit associated with ␤, ␣ 2 ␦, and/or ␥ auxiliary subunits to form an oligomeric complex. In mammalian myocardium, excitation-contraction coupling is characterized by a transient increase in cytosolic Ca 2ϩ . The influx of Ca 2ϩ through voltage-gated calcium channels subsequently induces Ca 2ϩ release from sarcoplasmic reticulum in a process known as Ca 2ϩ -induced Ca 2ϩ release. This process is prominent in the adult heart. The contraction of neonatal hearts relies more directly on the influx of Ca 2ϩ through L-type voltagegated calcium channels, in particular the Ca v 1.2 channels (1,2). L-type Ca v 1.2 channels are high voltage-activated channels and are expressed widely in cardiovascular and nervous systems (3). The human Ca v 1.2 gene, CACNA1C, contains 55 exons, of which Ͼ19 exons are subjected to alternative splicing (4). Novel splice variants of Ca v 1.2 channels were uncovered (5,6) as of the first systemic screening of cardiac Ca v 1.2 channels (7). Alternative splicing of the Ca v 1.2 channels is coupled to the generation of splice variants with altered electrophysiological and pharmacological properties (4, 8 -11). They also show tissue-specific expressions (8,9,12) and can be altered under pathological conditions (13,14).
In this report we screened the major alternatively spliced loci of Ca v 1.2 channels and identified a novel exon 33L in neonatal rat hearts that has not been reported previously. Inclusion of exon 33L generates a truncated channel Ca v 1.2 33L that does not conduct Ca 2ϩ . However, Ca v 1.2 33L produced a dominant-negative effect when co-expressed with functional Ca v 1.2 channels.

MATERIALS AND METHODS
RT-PCR and Single-cell RT-PCR-Young adult male Wistar rats (2 months, 150 -200 g) were sacrificed by CO 2 and subsequent cervical dislocation. Neonatal hearts were harvested from postnatal day 1 rats of either sex. This study was approved and performed in accordance with the guidelines of the Institutional Animal Care and Use Committee of the University. Fetal human heart cDNA was purchased from Clontech. Adult human heart tissues were obtained from donors for heart transplantation in the National Heart Center of Singapore with the approval of the institutional review board. The work was approved by the Institutional Review Board committees of the National Heart Center of Singapore, Singapore General Hospital. Total RNA was isolated from the left ventricles. cDNA was generated from total RNA by incubating with reverse transcriptase and 18-mer oligo(dT). The PCR protocol includes: a denaturation step at 95°C for 5 min; 35 cycles of 95°C for 30 s, 52°C or 55°C for 45 s, and 72°C for 1 min; a final extension step at 72°C for 10 min. The primers used for the rat Ca v 1.2 exon 33L were: 5Ј-GCCTCTTCACGGTGGAG-3Ј (forward) and 5Ј-TCC-CAATCACTGCATAGATAA-3Ј (reverse). The primers used for human Ca v 1.2 exon 33L are: 5Ј-CCAAGACCTAGAATAC-CGGG-3Ј (forward) and 5Ј-CTACCACAGGGTGTTTC-ACC-3Ј (reverse). The primers used for amplifying human Ca v 1.2 exons 18 to 24 as a control were: 5Ј-ATGAGGATAA-GAGCCCCTACCC-3Ј (forward) and 5Ј-ACTCGCAAGAT-CTTCACGACATTG-3Ј (reverse). For single-cell RT-PCR, after electrophysiological characterization by the patch clamp method, the cells were collected into a vial containing reverse transcription reaction mix for first strand cDNA synthesis using the reverse primer. Single-cell PCR protocol was similar to RT-PCR, except for a 40-cycle amplification.
Colony Screening and Restriction Enzyme Digestion-This method was described previously to identify novel splice variations (7,14). In brief, the PCR products were cloned into pGEM-T Easy vector (Promega). After being transformed into DH10B Escherichia coli cells, each transformant was selected and grown in a single well in a 96-well plate. Colony PCR was performed with the same set of primers and conditions to identify the component of exons in each colony. Usually 96 colonies were selected for each sample. The genotype of each PCR product amplified from a colony was predicted by size and randomly confirmed by DNA sequencing. To differentiate mutually exclusive exons that are of same size, restriction enzyme digestions were used to identify the expression of individual exons. The digestion was carried out in a 30-l reaction volume at 37°C for 5 h. To control for complete digestion, the same amount of PCR product of a positive control DNA was subjected to similar digestion conditions. Sequencing, Analysis, and Cloning of 33L Channels-Rat exon 33L was verified by DNA sequencing. The sequences were compared with rat genome sequence from the Human Genome Sequencing Center, Baylor College of Medicine. To characterize the functions of exon 33L containing channels, a fragment containing exon 33L was cloned into a wild type rat Ca v 1.2 channel with AvrII and NheI sites. This wild type Ca v 1.2 channel contains the exons predominantly expressed in rat cardiac tissues.
Generation of Polyclonal Antibody against Exon 33L Polypeptide-Rat exon 33L was cloned in-frame into pGEX-4T-1. Amino acid sequences from adjacent exons were not included. GST-fused protein was purified with glutathioneagarose (Sigma). Purified protein was used to immunize female New Zealand White rabbit once a month. Complete Freund's adjuvant was first mixed with GST-33L for immunization, and incomplete Freund's adjuvant was used in subsequent injections once a month. Serum collected after immunization was preabsorbed with GST protein first to remove GST antibodies, and 33L polyclonal antibody was affinity-purified from immobilized 33L protein with an IgG elution buffer (Pierce). The antibody concentration is 1 g/l. Serum from rabbit before immunization was used as preimmune control.
Immunofluorescent Staining-Animals were sacrificed and perfused with saline and subsequently 4% paraformaldehyde. The heats were then collected and post-fixed with 4% paraformaldehyde for 2 h. Dehydration was subsequently carried out by immersing the hearts into 15% sucrose followed by 30% sucrose. Next, the rat heart was sectioned at 20 m of thickness. After washing with 0.2% Triton X-100 phosphate-buffered saline (PBST), 100 l of blocking serum (10% goat serum and 1% bovine serum albumin in 0.2% PBST) was added onto the sections for 1 h. The sections were then incubated with primary antibodies overnight at 4°C. Monoclonal antibodies anticonnexin 43 (MAB 3068) was purchased from Chemicon International. On the following day, tissue sections were washed 3 times with TNT wash buffer (0.1 M Tris-HCl buffer, pH 7.5, containing 0.15 M NaCl and 0.05% Tween 20). The slides were incubated with FITC-conjugated or Texas red-labeled secondary antibodies for 1 h at room temperature. After washing three times of wash buffer, the slides were mounted with Fluor-SaveTM reagent (Merck). The results were visualized using laser scanning confocal microscope system (Fluoview BX61, Olympus). The negative control was done in an identical procedure except for primary antibody incubation, and no positive signal was identified.
The rat Ca v 3.2 channel and Ca v 1.3 channel were described earlier (15,16). I Ca was recorded with a whole-cell patch clamp technique at room temperature (ϳ25°C) 48 -72 h post-transfection. Patch electrodes were pulled using a Flaming/Brown micropipette puller (Sutter Instrument) and polished with a microforge. The external solution contained 140 mmol/liter tetraethylammonium methanesulfonate, 10 mmol/liter HEPES, and 1.8 mmol/liter CaCl 2 (the pH was adjusted to 7.4 with CsOH and osmolarity to 300 -310 mosM with glucose). For I Ba recording, 1.8 mM Ca 2ϩ was replaced by 5 mM Ba 2ϩ . The internal solution (pipette solution) contained 138 mmol/liter Cs-MeSO 3 , 5 mmol/liter CsCl 2 , 0.5 mmol/liter ethylene glycol tetraacetic acid, 10 mmol/liter HEPES, 1 mmol/liter MgCl 2 , and 2 mg/ml MgATP, pH 7.3 (adjusted with CsOH). The osmolarity was adjusted to between 290 and 300 mosM with glucose. Under voltage clamp and using an Axopatch 200B amplifier (Axon Instruments), whole-cell currents were filtered at 1-5 kHz and sampled at 5-50 kHz. The series resistance was normally Ͻ5 megaohms after compensation. The capacity transient was compensated using an online P/4 protocol. The steady-state inactivation curves were obtained from experiments by stepping from a holding potential of Ϫ90 mV to a 30-ms normalizing pulse to 10 mV followed by a family of 15-s-long prepulses from Ϫ80 to ϩ20 mV. A 104-ms test pulse to ϩ10 mV was recorded finally. Each test pulse was normalized to the maximal current amplitude of the normalizing pulse. The steady-state inactivation (SSI) 3 data were fitted with a single Boltzmann equation, where I relative is the normalized tail current, V1 ⁄ 2 is the potential for half-inactivation, and k is the slope value. G-V curves were obtained from a tail activation protocol. The cells were activated by a 20-ms test pulse of variable voltage family from Ϫ60 to 100 mV, and tail currents were measured after repolarization to Ϫ50 mV for 10 ms. The tail currents were normalized to the peak currents before fitting with a dual Boltzmann equation.
where G is the tail current, and G max is the peak tail current, F low is the fraction of low threshold component, and 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, and V1 ⁄ 2act was calculated when G ϭ 0.5G max .

RESULTS
Exon 31 and 32 Are Developmentally Regulated-Ca v 1.2 channel undergoes extensive alternative splicing, which affects its tissue distribution, pharmacology, and electrophysiological properties (4,10). Although up to 20 exons are alternatively spliced (17), only exons 31 and 32 have been reported to be developmentally regulated in the heart (18). Mutually exclusive exons 31 and 32 encode the IVS3 transmembrane segment and part of the IVS2-S3 intracellular linker, whereas another alternatively spliced exon 33 encodes the IVS3-S4 extracellular linker. There exist extensive alternative splicings within this region (7) involved in many cardiovascular diseases (6,13,14,19).
RT-PCR across exons 30 -35 will generate a 382-bp amplicon containing exon 33 and one of the mutually exclusive exons 31 or 32. Enzyme digestion on exon 32 by NsiI produces two smaller fragments with the sizes of 68 and 314 bp, respectively (Fig. 1A). In the presence of exon 31, only one band exists (382 bp), as no digestion occurs. Deletion of the cassette exon 33 (33 nt in length) has rarely been found in rat hearts (8). When we digested RT-PCR products from the neonatal and adult rat hearts, partial digestions were observed in both samples, indicating that neonatal and adult hearts contain a mixture of exons 31 and 32. However, the extent of digestion by NsiI was different, as more PCR products from the adult hearts were digested than that from the neonatal hearts. Therefore, exon 32 expression is higher in adult rat hearts than in neonatal rat hearts (Fig.  1B). Such developmental changes of exon 31 and 32 have been reported by Diebold et al. (18), and our experiment supports the previous finding.
Identification of Exon 33L from Neonatal Rat Heart-Besides mutually exclusive exons 31, 32, and cassette exon 33, there exists an additional 66-nt extension at the 5Ј end of exon 34 with different acceptor sites, producing exon 34a, an isoform of exon 34 (6). To characterize the detailed expression of exon 33 and 34a, we used the colony screening method that is able to detect exons of lower expression (8). The inclusion or exclusion of exons 33 and/or 34a could be easily identified by differences in size of the PCR products, whereas mutually exclusive exons 31 and 32 cannot, as they are of the same length. Colony PCR screening showed that the size of the majority of colonies was ϳ382 bp, indicating that the splice variants with exon 33 deletion or exon 34a inclusion are not predominant in neonatal and adult rat hearts (Fig. 2, A and B). DNA sequencing confirmed that the colonies with smaller size contained only exon 31 or 32 but with the exon 33 deletion. However, some colonies of a Digestion at exon 32 with NsiI will generate two fragments of 68 and 314 bp. B, partial digestion with NsiI was observed in both NH and AH. Exon 32 is expressed higher in AH as more PCR produce was digested. Equal amounts of DNA from both AH and NH were digested under same conditions in parallel.
larger size have been found to appear more frequently in the neonatal heart than in the adult heart (indicated by arrows in Fig. 2, A and B). Subsequent DNA sequencing confirmed that these colonies did not contain exon 34a, but expressed a novel exon, which is named exon 33L here. We repeated RT-PCRbased colony screening experiments and confirmed that neonatal hearts (NH) express 9.7% of exon 33L, which represents a Ͼ2-fold increase to the 4.3% in adult hearts (AH) (Fig. 2C, p ϭ 0.0127, NH (n ϭ 5), AH (n ϭ 5), Student's t test).
Exon 33L contains an additional 71-nt 5Ј to exon 33 (Fig. 3A). To understand the nature of this 71-nt fragment, we searched the rat genome sequence from the Human Genome Sequencing Center from Baylor College of Medicine. The mechanism for the generation of the additional 71-nt 5Ј of exon 33 can be rationalized by the inspection of the genomic sequence of the rat Ca v 1.2 channel gene. These 71 nt locate at chromosome 4 and 5Ј to exon 33 of the rat Ca v 1.2 gene. No genomic sequence was found between the 71 nt and exon 33. More importantly, the canonical -ag-acceptor dinucleotides were found 5Ј to the 71 nt. Therefore, the inclusion of the 71 nt could be explained by the use of an alternate acceptor site of exon 33. However, the inclusion of 71 nt will result in a frameshift of the coding sequences and produce a premature stop codon (Fig. 3, A and B). As the premature stop codon appeared in the IVS3-4 linker, it will generate a truncated channel lacking the last three transmembrane segments of domain IV and the entire C terminus (Fig. 3C). Thus, there exists two isoforms of exon 33. One is the original 33-nt exon 33 and the other is the new 104-nt exon 33L (71 nt ϩ 33 nt). We further compared the DNA sequencing results and found that exon 33L can be linked either with exon 31 or exon 32.
Translational Expression of Exon 33L in Heart-To determine whether the aberrant splice variants are translated into proteins, a polyclonal antibody was generated against the 22amino acid polypeptides encoded by exon 33L (Fig. 3A). The 22-amino acid polypeptides were expressed as a fusion protein with GST (Fig. 4A) and were injected into rabbits to produce exon 33L specific antibody ␣-33L. No adjacent amino acids were included for antibody production. Using ␣-33L, we detected by Western blot a band of 160 kDa in neonatal and adult hearts, having the same molecular mass of the truncated 33L containing Ca v 1.2 channel expressed in HEK 293 cells (Fig.  4B).
Therefore, neonatal rat hearts expressed higher levels of Ca v 1.2 33L channel than adult hearts. To prove the specificity of ␣-33L antibody, HEK cells were transfected with Ca v 1.2 33L channel or the reference channel Ca v 1.2 WT , which does not contain exon 33L. ␣-33L antibody stained Ca v 1.2 33 -transfected cells but not Ca v 1.2 WT -transfected cells (Fig. 4C). Next, we used an immunohistological method to examine the cellular expression of the Ca v 1.2 33L channel in rat hearts during development. Connexin 43 was labeled for the gap junctions between adjacent cardiac muscle cells. Again, ␣-33L staining showed a stronger fluorescent signal in neonatal heart than in adult heart, supporting a higher expression of Ca v 1.2 33L channel in neonatal rat heart (Fig. 4D). It should be noted that Ca v 1.2 33L did not colocalize with Connexin 43 in both neonatal and adult hearts, suggesting that the majority of Ca v 1.2 33L channels may not traffic to the plasma membrane.
Presence of Ca v 1.2 33L Channels Reduces Currents of Functional Ca v 1.2 Channels-The functional significance of the inclusion of exon 33L in Ca v 1.2 channels has not been reported. We sought to assess exon 33L functions by whole-cell patch clamp recording. Exon 33L was cloned into a reference rat cardiac-form Ca v 1.2 channel that contained exon 1a and 8a, with the absence of exon 9* (8). To control for transfection efficiency and the conditions of the HEK 293 cells, we performed the experiments in parallel by transfecting the HEK 293 cells on the same day and did the electrophysiological experiments alter-  nately using the same solutions on the same day. The presence of successful transfection of Ca v 1.2 33L channels was evidenced by the single cell RT-PCR. All the cells patched with green fluorescence (␤ 2a is tagged with GFP) contained Ca v 1.2 33L channels. The cells were maintained at a Ϫ90 mV holding potential before a test pulse of 0 mV for 900 ms was applied. No currents were recorded (Fig. 5A, n ϭ 18) for Ca v 1.2 33L channel, indicating that the presence of exon 33L generates a novel Ca v 1.2 channel that is impermeable to Ca 2ϩ ions. To investigate whether the presence of the 33L channels affects the reference cardiac channels, we co-transfected the 33L channels and the reference channels into HEK cells. The exon 33L expression in neonatal heart and adult heart is 9.7% and 4.3%, respectively (Fig. 2C). Therefore, we transfected Ca v 1.2 33L channels into HEK 293 cells at 5 and 10% to the reference wild type channels. Tail currents were recorded in 5 mM Ba 2ϩ to exclude Ca 2ϩrelated effects. A slightly higher current density was observed in 5% Ca v 1.2 33L channels. However, there was no statistical significance between 5 and 10% Ca v 1.2 33L channels (Fig. 5B).
It is possible that the effect of Ca v 1.2 33L channels on the reference channels is masked by the variability of the data recorded. We decided to increase the 33L channel ratio by transfecting equal amounts of both Ca v 1.2 33L channels and the reference channels into HEK 293 cells for electrophysiological recordings. The control experiments were carried out by transfecting the same amount of empty pcDNA3 vector together with the reference channels. Single cell RT-PCR was performed to confirm that both the 33L channels and the reference channels were equally expressed (Fig. 6A). The data collected from cells with unequal expression were discarded. Similar to the dominant-negative effect observed by Ebihara et al. (20) in the three-domain Ca v 1.2 channels, the presence of Ca v 1.2 33L channels reduced the tail current density of the reference Ca v 1.2 channels (Fig. 6B) in 1.8 mM Ca 2ϩ bath solution, mimicking a physiological condition. The tail current density at ϩ100 mV was 41.6 Ϯ 13.2 pA/pF (n ϭ 7) when control pcDNA3 vectors were co-transfected with the reference channels, whereas the tail current density was 11.3 Ϯ 2.3 pA/pF (n ϭ 9) when Ca v 1.2 33L channels were co-transfected, representing a 73% decrease in the presence of Ca v 1.2 33L channels (p ϭ 0.025, Student's t test). Measuring the ON-gating charge (Q ON ) at the reversal potential (typically ϩ50 mV, Fig. 6C) provides a convenient measure of the number of channels at the membrane (21). Therefore, we measured the Q ON of the Ca v 1.2 33L -and pcDNA3-transfected cells. Interestingly, 33L transfection greatly reduced the Q ON as compared with pcDNA3-transfected cells: 33L Q ON , 0.8374 Ϯ 0.2296 femtocoulombs/pF, n ϭ 7; pcDNA3 Q ON , 2.499 Ϯ 0.4639 femtocoulombs/pF, n ϭ 6; p ϭ 0.0063, Student's t test. The reduction of 66.5% in Q ON was similar to the 73% reduction in current density.
Ca v 1.2 33L Channels Exert a Dominant Negative Effect on Ca v 1.3 Channels but Not on Ca v 3.2 Channels-To examine whether Ca v 1.2 33L channels down-regulate other voltage-gated calcium channels, we cotransfected Ca v 1.2 33L channels with the paralogous L-type Ca v 1.3 channels or with the T-type Ca v 3.2 channels at a ratio of 1:1. Again, pcDNA3 vector was used as a control. One difference between Ca v 1.3 channels and Ca v 3.2 channels is that Ca v 1.3 channels require the ␤ subunit for functional expression, whereas Ca v 3.2 channels do not need the ␤ subunit. Current densities from I-V curves were measured in 1.8 mM Ca 2ϩ solution. Interestingly, the dominant negative effect of Ca v 1.2 33L channels is prominent in Ca v 1.3-expressed cells but not in Ca v 3.2-expressed cells (Fig. 7). Compared to the pcDNA3 control, the presence of Ca v 1.2 33L channels reduced Ca v 1.3 current densities at Ϫ10 mV from Ϫ16.16 Ϯ 2.205 pA/pF (n ϭ 8) to Ϫ7.952 Ϯ 1.250 pA/pF (n ϭ 15); p ϭ 0.0021, Student's t test (Fig. 7A). However, in Ca v 3.2expressed cells, the current densities were similar in pcDNA3 and Ca v 1.2 33L -transfected cells (Fig. 7B). At Ϫ20 mV, the current densities of Ca v 1.2 33L -expressed cells were Ϫ16.93 Ϯ 4.158

channels.
A, single-cell RT-PCR was performed after each patching experiment. The upper bands represent exon 33L, whereas lower bands represent the reference channel. Data from cells with unequal transfection were discarded. B, tail current density (pA/pF) was calculated in 1.8 mM Ca 2ϩ from cells expressing the reference cardiac Ca v 1.2 channels together with the Ca v 1.2 33L channels (•, n ϭ 7) or pcDNA3 vector (E, n ϭ 9). C, a sample trace at ϩ50-mV test pulse illustrates the measurement of gating current, indicated by an arrow. D, summary of the gating currents (femtocoulombs/pF (fC/pF)) measured from 33L channel or pcDNA3 transfected cells. *, p ϭ 0.0063, Student's t test.

FIGURE 7. Ca v 1.2 33L channels yield dominant negative effect on Ca v 1.3 channels but not on Ca v 3.2 channels.
Ca v 1.2 33L channels were cotransfected with Ca v 1.3 or Ca v 3.2 channels at a ratio of 1:1. I-V curves were obtained in a bath solution containing 1.8 mM Ca 2ϩ . As a control, pcDNA3 vectors were transfected at the same molar ratio. A, in Ca v 1.3 channels, which require ␤ subunits, Ca v 1.2 33L channels greatly inhibited Ca v 1.3 currents. At Ϫ10 mV, the reduction of current density from Ϫ16.16 Ϯ 2.205 pA/pF (n ϭ 8) to Ϫ7.952 Ϯ 1.250 pA/pF (n ϭ 15), p ϭ 0.0021, Student's t test, represents an almost 50% decrease. B, Ca v 1.2 33L cotransfection had no effect on Ca v 3.2 channels, which do not require ␤ subunits.
Presence of Ca v 1.2 33L Channels Changes Electrophysiological Properties of Functional Ca v 1.2 Channels-To investigate whether the electrophysiological properties of the reference cardiac channel is altered with the presence of Ca v 1.2 33L channels, we used a recording protocol for the SSI and a tail protocol for the activation properties, which were described previously (8,9). Interestingly, the presence of Ca v 1.2 33L channels negatively shifted both the SSI and the activation properties of the reference channels (Fig. 8). The V1 ⁄ 2 of the SSI was hyperpolarize-shifted for Ϫ3.54 mV (p ϭ 0.0174, Student's t test), V1 ⁄ 2SSIpcDNA (Ϫ22.96 Ϯ 0.86 mV (n ϭ 6)) versus V1 ⁄ 2SSI33L (Ϫ26.5 Ϯ 0.9 mV (n ϭ 6)), whereas the V1 ⁄ 2 of the activation curve was hyperpolarize-shifted for Ϫ5.73 mV (p ϭ 0.0036, Student's t test), V1 ⁄ 2ActpcDNA (18.43 Ϯ 1.05 mV (n ϭ 9)) versus V1 ⁄ 2Act33L (12.7 Ϯ 1.28 mV (n ϭ 7)). Therefore, the calcium window current of the reference cardiac channel was shifted to a more negative potential in the presence of Ca v 1.2 33L channels.
Ca v 1.2 33L Channels Down-regulate Functional Ca v 1.2 Channels by Enhancing Protein Degradation via the Ubiquitin-Proteasome System-To investigate the mechanism for the dominant negative effect of Ca v 1.2 33L , the channels were first co-transfected with the reference channels at 1:1 ratio. A Western blot showed that the total protein level (membrane and cytosol) of the reference Ca v 1.2 channels was reduced by the presence of Ca v 1.2 33L (Fig. 9A). Treatment with 3 M MG132, a proteasome inhibitor, successfully increased the expression of the reference Ca v 1.2 channels, suggesting that Ca v 1.2 33L might enhance proteasomal degradation. A similar down-regulation effect of Ca v 1.2 33L was observed on Ca v 1.3 channels (Fig. 9B) but with a smaller reduction. Next, we examined protein ubiquitination using immunoprecipitation and subsequent Western blot. Interestingly, transfection with Ca v 1.2 33L greatly enhanced wild type Ca v 1.2 channel ubiquitination. The ubiquitinated Ca v 1.2 channels were further increased with the treatment of 3 M MG132 (Fig. 9C). Again, Ca v 1.3 channel ubiquitination was also increased with Ca v 1.2 33L co-expression (Fig.  9D).
Human Exon 33L-containing Channel Conducts Ca 2ϩ Ions-When we compared the exon 33L sequences of mouse, rat, and human (Fig. 10A), we found that the acceptor for exon 33L is preserved across the three species, indicating that exon 33L may exist in all these species. The mouse and the rat shared exactly the same sequences of exon 33L, whereas the human owns a different sequence. Interestingly, the human genomic sequence had an additional base within exon 33L compared to mouse or rat. Therefore, no frameshift was generated (Fig. 10  B), and the human Ca v 1.2 33L channel may conduct Ca 2ϩ ions. To prove this, we first performed RT-PCR on human heart samples using primers specific for exon 33L. The sequences of the primers are described under "Materials and Methods." A band with a size of 502 bp was found in the left and right ven-  tricles of a human adult heart and in the left ventricle of a fetal human heart (Fig. 10C). DNA sequencing confirmed that the bands contained human exon 33L and exon 33 sequences, as predicted in Fig. 10A. The human fetal heart expressed a higher level of exon 33L as compared with adult hearts, which is similar to the rat exon 33L expression during cardiac development. Next, we cloned the human exon 33L into a reference Ca v 1.2 channel and performed patch clamp experiments. Fig. 10D shows that the human Ca v 1.2 33L channel conducted Ca 2ϩ ions when it was cotransfected with ␤ 2a and ␣ 2␦ subunits. Compared to the wild type reference Ca v 1.2 channel, the Ca v 1.2 33L channel had a lower current density. For the Ca v 1.2 33L channel, current density at 0 mV was Ϫ4.095 Ϯ 0.8472, n ϭ 15, whereas in wild type channel the current density was Ϫ10.85 Ϯ 1.668; n ϭ 15; p ϭ 0.0012, Student's t test. Conductance was further calcu-lated using the equation G ϭ I/(V m Ϫ E Ca ) and normalized by the maximum conductance value and fit with the Boltzmann equation (Fig. 10E), where E Ca is reverse potential. The V1 ⁄ 2 of human Ca v 1.2 33L channel was Ϫ9.17 Ϯ 1.24 mV, which is significantly higher than the V1 ⁄ 2 of the wild type channel (Ϫ13.3 Ϯ 0.6 mV, p ϭ 0.0056).

DISCUSSION
Truncated Na ϩ and Ca 2ϩ channels have been found in brain and muscles (22,23). In rabbit heart aberrant splicing in the II-III loop of Ca v 1.2 channels produces two truncated channels that contain only domains I and II (23). The truncation is caused by the deletion of exons 17 and 18 or the deletion of exon 19, which results in a frameshift leading to premature termination of translation. In contrary, alternative splicing of exon 33L FIGURE 10. Human Ca v 1.2 33L channels conduct Ca 2؉ ions and are developmentally regulated in heart. A, alignment of the exon 33L DNA sequences from Rattus norvegicus, Mus musculus, and Homo sapiens. The canonical -ag-acceptors (indicated in the gray box) is preserved in all three species. The mouse and the rat share the same exon 33L sequence. Human exon 33L is predicted to contain an additional nucleotides. All the differences in DNA sequences are highlighted by boxes. B, alignment of the amino acid sequences of exon 33L predicted from R. norvegicus, M. musculus, and H. sapiens. The differences in amino acid sequences are labeled by boxes. No stop codon is generated by the inclusion of human exon 33L. C, using human exon 33L-specific primers, a band of 502 bp was found in human adult left ventricle (LV), adult right ventricle (RV), and fetal left ventricle (FH). M, marker. DNA sequencing confirmed the sequence to be human exon 33L. PCR using primers across exon 18 -exon 24 was illustrated as the loading control. As a water control (-), cDNA was replaced by water to exclude possible contaminations. D, human exon 33L was cloned into a Ca v 1.2 reference channel. I-V curves were obtained in a bath solution containing 1.8 mM Ca 2ϩ . In HEK cells, human Ca v 1.2 33L channels indeed conduct Ca 2ϩ ions with a smaller current density. E, conductance was calculated as G ϭ I/(V m Ϫ E Ca ) and normalized by the maximum conductance value and fit with Boltzmann equation, where E Ca is reverse potential. of the Ca v 1.2 channels we report here yielded 2 significant findings. 1) The functional modification is species-specific in that rat exon 33L results in a truncated channel that likely competes for ␤ subunit binding to wild type channels. However, the human exon 33L has a single nucleotide insertion resulting in a longer functional full-length channel that, however, produced reduced current density. 2) The truncated rat Ca v 1.2 33L variant not only competes for ␤ subunit binding to expose the wild type Ca v 1.2 channel to presumably ERAD degradation (24), but it also similarly down-regulated the expression of another ␤ subunit requiring L-type Ca v 1.3 channel but not the T-type Ca v 3.2 channel.
The presence of rat exon 33L shifts the coding frame and leads to a deletion of the last three transmembrane segments of the domain IV and the downstream C terminus. Therefore, the predicted topography of this rat Ca v 1.2 33L channels is larger than the two domain channels reported previously (23).
Rat Ca v 1.2 33L channels do not conduct Ca 2ϩ ions, indicating that the last three transmembrane segments within domain IV are critical in forming a complete functional channel. The pore of the channel is likely to be disrupted in particular with the deletion of IVS4 containing positively charged amino acids. Another result from this study is that the presence of Ca v 1.2 33L channels could down-regulate the current flux through the reference wild type channels. The inhibitory effect is not prominent when Ca v 1.2 33L channels are expressed at 5% or 10%, possibly due to data variability when recording small changes in current density. However, when Ca v 1.2 33L was expressed in equal amount to the reference channels, the dominant negative effect is clear. We found a 73% current density reduction, largely in correlation with a 66.5% decrease in gating current, indicating that the current density reduction is mainly due to a decrease of the functional Ca v 1.2 channels on the cell membrane. The slight difference between the changes in current density and gating current (ϳ6.5%) suggests that a small number of channels on the surface membrane may not conduct Ca 2ϩ ions. This small population of channels is likely the truncated Ca v 1.2 33L channels. Immunostaining of Ca v 1.2 33L channels in rat neonatal and adult hearts suggests that the majority of Ca v 1.2 33L channels were not expressed on the plasma membrane. Therefore, only very few Ca v 1.2 33L channels may traffic to the membrane, contributing to a small portion of the gating currents.
The mechanism for the inhibitory effect of Ca v 1.2 33L channels could be explained in that Ca v 1.2 33L channels compete with the reference channels, presumably for ␤ subunit for expression on the plasma membrane. It is well known that the ␤ subunit is required to target the functional L-type Ca 2ϩ channels to the cell surface (25), and the binding requires a conserved motif in the domain I-II cytoplasmic loop of the Ca v 1.2 channels (26). As the I-II loop is well preserved in Ca v 1.2 33L channels, the expression of this truncated channel may scavenge ␤ subunits and, therefore, down-regulate the surface expression of the reference channels. This mechanism is supported by a drastic reduction of gating currents caused by Ca v 1.2 33L channels. Another possible mechanism is that the truncated channels may prevent the correct folding of wild type channels and inhibit channel synthesis or stability. This hypothesis was shown by Dolphin and co-workers (27) in two domain truncated Ca v 2.2 channels, in which the dominant negative effect is independent on ␤ subunit. Therefore, such modulation is generally channel-specific, seldom affecting other channels with a different protein structure. Our results from experiments performed on Ca v 1.3 and Ca v 3.2 channels suggest that the second mechanism may not explain the effect Ca v 1.2 33L has on wild type Ca v 1.2 and Ca v 1.3 channels. The dominant negative effect of Ca v 1.2 33L channels was exerted as well on Ca v 1.3 channels, which require ␤ subunits, but not on Ca v 3.2 channels, which do not need ␤ subunits, suggesting that ␤ subunits are critical for Ca v 1.2 33L suppression. The binding of ␤ subunits has been demonstrated by the Zamponi and coworkers (24) to increase membrane expression of Ca v 1.2 channels by preventing endoplasmic reticulum-associated protein degradation. Without ␤ subunits, Ca v 1.2 channels undergo robust ubiquitination. Therefore, if Ca v 1.2 33L channels compete with wild type Ca v 1.2 channels for ␤ subunits, wild type channels without ␤ subunit association will be degraded. Our biochemical results support this hypothesis. In the presence of Ca v 1.2 33L channels, wild type Ca v 1.2 channel ubiquitination was enhanced for degradation, and thus total protein level was reduced.
However, we cannot exclude the mechanism proposed by Dolphin completely. If the number of intracellular ␤ subunits is fixed, the expression of Ca v 1.2 33L channels should correlate with the inhibitory effect on the reference channels, which is an ϳ50% decrease when the Ca v 1.2 33L channel and the reference channel are expressed at a 1:1 ratio. However, the current density is reduced by 73%, more than the predicted 50%, indicating the presence of the second mechanism. Furthermore, the Ca v 1.2 33L channel suppressed the current density in Ca v 1.3 channels from Ϫ16.16 pA/pF to Ϫ7.952 pA/pF, representing an ϳ50% reduction, strongly correlating with the molar ratio of Ca v 1.2 33L channels to Ca v 1.3 channels, which is 1:1. As Ca v 1.3 channels have a different structure from Ca v 1.2 channels, the second mechanism is unlikely to apply to Ca v 1.3 channels. Therefore, we would hypothesize that the inhibitory effect of Ca v 1.2 33L channels on Ca v 1.3 channels is solely via competing for ␤ subunits.
Ca v 1.2 33L channels are also different from another three-domain Ca v 1.2 channel that was reported earlier (20). In that truncated channel, the N terminus and the first domain as well as the entire I-II loop are deleted. Thus, there will be no competition for ␤ subunits from this three-domain channel, which could explain why the inhibitory effect was not found on other channels. Interestingly, the three-domain Ca v 1.2 channel exhibits a more prominent dominant-negative effect on wild type channels compared with the Ca v 1.2 33L channel. As the C terminus is well preserved in the three-domain Ca v 1.2 channel and the rest of the channel is the same as in the Ca v 1.2 33L channel, the potent inhibitory effect is likely due to the intact C terminus. Zamponi and co-workers (24) showed that the endoplasmic reticulum retention motif of Ca v 1.2 channel locates at the C terminus and likely interferes with the ␤ subunits and the I-II linker sterically or allosterically in a properly folded channel. The three-domain Ca v 1.2 channel with a complete C ter-minus may interrupt such interactions, yielding a dominant negative effect. Additional experiments are, therefore, needed.
Our results further showed that the presence of Ca v 1.2 33L channels could alter the electrophysiological properties of the reference cardiac Ca v 1.2 channels. The negative shifting of SSI and activation curves produced a calcium window current that peaks at more hyperpolarized membrane potential. The exact mechanism of such alterations is not fully understood. In a previous study on Ca v 1.1 channel (␣1S), the two-domain (I and II)-containing hemichannels were expressed on the cell membrane, and the gating characteristics were modified significantly, presumably by interactions with domains III and IV (28). Thus, it is possible that the interaction between the intact I and II domains of 33L channels with the reference Ca v 1.2 channels may change the electrophysiological properties of those reference channels. The deletion of the C terminus in 33L channels may also cause the electrophysiological property changes in reference channels. The C terminus of Ca v 1.2 channels is known to regulate Ca 2ϩ -dependent and voltage-dependent channel inactivation (29). Therefore, the interaction with the truncated 33L channels may alter the regulatory functions of C terminus in the reference channel. Additional experimentations are needed to elucidate the regulatory role of the 33L channels on functional Ca v 1.2 channels.
The truncated Ca v 1.2 33L channel may not have a major physiological impact on neonatal heart function, as the expression level is only at 9.7% and even lower in adult hearts. However, during the search for more altered exonal expression in neonatal hearts, at least one more site was identified in the IIIS2 region involving exons 21 and 22 in forming an aberrant channel (data not shown). This aberrant channel again loses the ability for Ca 2ϩ influx but exhibits a dominant-negative effect on functional channels. Therefore, we believe that aberrant channels in neonatal hearts may have some functions during cardiac development. More experiments with gene manipulation can help to identify the physiology of these aberrant channels.
When we searched human genomic sequence of Ca v 1.2 channel, we identified a similar acceptor site for human exon 33L. Unexpectedly, human exon 33L sequence contained an additional nucleotide and was unlikely to generate a stop codon that is found in rat exon 33L. We successfully cloned human exon 33L into a reference Ca v 1.2channel, and electrophysiology studies on these channels eventually showed that they are functional in conducting Ca 2ϩ ions, albeit with a smaller current density. At 0 mV, the current density of human Ca v 1.2 33L channels is only ϳ40% that in wild type channels. The mechanism for the smaller current remains unknown. Exon 33L codes for the extracellular loop between IVS3 and IVS4, which has been shown to affect channel activation property (7). It was proposed that a longer IVS3-S4 loop shifts the activation potential to a more positive potential. This trend was previously shown for alternatively spliced exon 33, which is of 33 bp in length. Deletion of exon 33 in smooth muscles shifts the activation and inactivation properties of Ca v 1.2 channels to a more negative potential. Such structural variations are believed to contribute to the electrophysiological differences between cardiac and smooth muscle Ca v 1.2 channels (8). With the inclusion of human exon 33L, which encodes an additional 24 amino acids, the loop between IVS3 and IVS4 is much longer and may change the channel properties as well. Indeed, when we compared the activation properties of the human Ca v 1.2 33L channel with the wild type channel, the V1 ⁄ 2 of the human Ca v 1.2 33L channel is significantly more positive than that of wild type channel.
Interestingly, human exon 33L is also developmentally regulated in heart, with a similar trend as seen in rat exon 33L, suggesting that the mechanism for regulating exon 33L expression during cardiac development is conserved in human and rodents. In the brain, two exons, 9* and 33, of Ca v 1.2 channel are developmentally regulated (30). Fox proteins 1 and 2 regulate the expression of the two exons during embryonic mouse brain development. In this study, the longer exon 33L is formed by adopting an alternate acceptor site. As there is no intron between exon 33L and the downstream exon 33, the splicing mechanism of 33L is likely regulated by Fox proteins as well. The increase of Fox proteins during embryonic development causes more inclusion of exon 33 in cerebral Ca v 1.2 channel (30). Therefore, the lower level of Fox proteins in fetal rats may affect the selection of the acceptor sites for exon 33 or exon 33L. Further experiments in changing the Fox binding elements locating within the intron after exon 33 will help to elucidate the underlying mechanism.
To the best of our knowledge, only one report showed an alteration of the structure of IVS3 segment (encoded by alternatively spliced exons 31/32) of Ca v 1.2 channels during cardiac development (18). Interestingly, subsequent studies revealed a reappearance of the fetal Ca v 1.2 channel in hearts of rat myocardial infarction and human heart failure (13,19), indicating an alteration of Ca v 1.2 channel in response to cardiac malfunctions.
In conclusion, we report here for the first time that the neonatal rat heart expresses a higher level of an aberrant truncated Ca v 1.2 33L channel. Albeit not permeable to Ca 2ϩ ions, Ca v 1.2 33L channels affect the functional reference channels, mainly via competing for ␤ subunits. Due to a genomic difference, human Ca v 1.2 33L channels are complete and functional channels that conduct Ca 2ϩ ions. Importantly, human exon 33L is also developmentally regulated in heart. Neonatal Ca v 1.2 splice variants have been shown to re-emerge in chronic heart diseases (13,19). Therefore, it is important to understand whether these neonatal channels appear and function in cardiac diseases.