Smooth Muscle Tissues Express a Major Dominant Negative Splice Variant of the Type 3 Ca2+ Release Channel (Ryanodine Receptor)*

It is well known that the type 3 Ca2+ release channel (ryanodine receptor, RyR3) exhibits strikingly different pharmacological and functional properties depending on the tissues in which it resides. To investigate the molecular basis for this tissue-dependent heterogeneity, we examined the primary structure of RyR3 from various tissues by reverse transcription polymerase chain reaction and DNA sequence analysis. As many as seven alternatively spliced variants of RyR3 were detected. Ribonuclease protection assays revealed that one of these splice variants, RyR3 (AS-8a), which lacks a 29-amino acid fragment (His4406–Lys4434) encompassing a predicted transmembrane helix, was highly expressed in smooth muscle tissues, but not in skeletal muscle, the heart, or the brain. Although the RyR3 (AS-8a) splice variant did not form a functional Ca2+release channel when expressed alone in HEK293 cells, it was able to form functional heteromeric channels with reduced caffeine sensitivity when co-expressed with the wild type RyR3. Interestingly, this RyR3 splice variant was also able to form heteromeric channels with and suppress the activity of the type 2 ryanodine receptor (RyR2). Tissue-specific expression of RyR3 splice variants is therefore likely to account for some of the pharmacological and functional heterogeneities of RyR3. These observations also reveal a novel mechanism by which a splice variant of one RyR isoform (RyR3) can suppress the activity of another RyR isoform (RyR2) via a dominant negative effect.

It is well known that the type 3 Ca 2؉ release channel (ryanodine receptor, RyR3) exhibits strikingly different pharmacological and functional properties depending on the tissues in which it resides. To investigate the molecular basis for this tissue-dependent heterogeneity, we examined the primary structure of RyR3 from various tissues by reverse transcription polymerase chain reaction and DNA sequence analysis. As many as seven alternatively spliced variants of RyR3 were detected. Ribonuclease protection assays revealed that one of these splice variants, RyR3 (AS-8a), which lacks a 29amino acid fragment (His 4406 -Lys 4434 ) encompassing a predicted transmembrane helix, was highly expressed in smooth muscle tissues, but not in skeletal muscle, the heart, or the brain. Although the RyR3 (AS-8a) splice variant did not form a functional Ca 2؉ release channel when expressed alone in HEK293 cells, it was able to form functional heteromeric channels with reduced caffeine sensitivity when co-expressed with the wild type RyR3. Interestingly, this RyR3 splice variant was also able to form heteromeric channels with and suppress the activity of the type 2 ryanodine receptor (RyR2). Tissue-specific expression of RyR3 splice variants is therefore likely to account for some of the pharmacological and functional heterogeneities of RyR3. These observations also reveal a novel mechanism by which a splice variant of one RyR isoform (RyR3) can suppress the activity of another RyR isoform (RyR2) via a dominant negative effect.
Ryanodine receptors (RyRs) 1 were initially described in the sarcoplasmic reticulum of striated muscles. It is now known that there are three mammalian RyR isoforms (RyR1, RyR2, and RyR3) and that they are expressed in a variety of cells and tissues (1)(2)(3). RyR1 is predominantly expressed in skeletal muscle, whereas RyR2 is mainly expressed in heart, brain, and smooth muscles. The expression of RyR3 has been detected in a broad range of tissues, including brain, smooth muscles, and skeletal muscle (4). RyR1 and RyR2 function as Ca 2ϩ release channels and play an essential role in excitation-contraction coupling in striated muscles (5)(6)(7)(8). However, the physiological role and channel properties of RyR3 remain elusive.
Depending on the tissues in which it is expressed, RyR3 exhibits different functional properties. In skeletal muscle, it functions as a caffeine-and ryanodine-sensitive Ca 2ϩ -induced Ca 2ϩ release channel similar to RyR1 and RyR2 (9) and is involved in amplifying the Ca 2ϩ signals generated by RyR1 (10). On the contrary, in smooth muscle cells, RyR3 forms a ryanodine-sensitive but caffeine-insensitive Ca 2ϩ release channel (11,12) and may negatively regulate the activity of RyR2 and/or RyR1 (13). The expression of RyR3 has also been demonstrated in human Jurkat T-lymphocytes and mink lung epithelial cells. Interestingly, these cells also exhibit ryanodinesensitive, caffeine-insensitive Ca 2ϩ release activity (14,15). These observations have led to the notion that RyR3 expressed in smooth muscles and peripheral tissues possesses unique functional properties, although the molecular mechanism underlying this tissue-specific function of RyR3 is unknown.
To account for the pharmacological and functional heterogeneity of RyR3, it has been proposed that smooth muscle and peripheral tissues may express a unique isoform of RyR3 as a result of alternative splicing (16,17). To test this hypothesis, we systematically investigated the existence of alternatively spliced variants of RyR3 expressed in the uterus by amplifying and sequencing the entire ϳ15-kb coding region of RyR3. Our results show that RyR3 expressed in smooth muscle tissues is extensively modified by alternative splicing and that one of the splice variants, RyR3 (AS-8a), is highly and selectively expressed in smooth muscle tissues. Functional characterization reveals that this major RyR3 splice variant does not form a functional channel when expressed alone but is able to form functional heteromeric channels with the wild type RyR3 and RyR2. Our data provide the molecular basis for the tissue-dependent heterogeneity of RyR3 and demonstrate for the first time that a splice variant of RyR3 is able to interact with and suppress the activity of another RyR isoform, RyR2, via the formation of heteromeric channel complexes.
Ribonuclease Protection Assay-A ribonuclease protection assay was carried out using the RPA II kit from Ambion according to the manufacturer's instructions. Briefly, a RyR3 cDNA fragment (231 bp) encompassing the AS-8a splice region (87 bp), a 114-bp 5Ј-flanking, and a 30-bp 3Ј-flanking sequence was generated by PCR and subcloned into pBluescript. The plasmid was linearized and used as template to synthesize 32 P-labeled RNA probe using the T7 RNA polymerase. The 32 P-labeled RNA probe was hybridized with total RNA isolated from various tissues followed by treatment with RNases. The protected probes were analyzed by polyacrylamide gel electrophoresis and autoradiography.
Site-directed Mutagenesis and DNA Transfection-Construction of the full-length RyR3 cDNA and insertion of the c-Myc epitope tag into RyR3 after glutamate 4318 have been described previously (17,19). The HA tag (YPYDVPDYA) was inserted into the same position in RyR3 as the c-Myc tag using the same strategy. A SpeI (12864)-NotI (vector) fragment containing the AS-8a deletion was removed from the PCR8 RT-PCR fragment and was used to replace the corresponding wild type fragment in the full-length RyR3 cDNA. The SpeI (10659)-SpeI (12864) fragment was ligated back to form RyR3 (AS-8a). The KpnI (6313)-AflII (7588) fragment containing the ⌬E2256-T2429 deletion was generated by the overlap extension method (20) and was used to replace the corresponding wild type fragment in the full-length RyR3 to form RyR3 (⌬E2256-T2429). HEK293 cells were transfected with RyR cDNAs using Ca 2ϩ phosphate precipitation (17).
Immunoprecipitation and Immunoblotting Analysis-Cell lysates prepared from transfected HEK293 cells as described previously (21) were incubated with protein G-agarose (30 l) that was prebound with 5-10 g of anti-c-Myc or anti-HA antibodies at 4°C for 17-19 h. The immunocomplexes bound to the agarose beads were solubilized by Laemmli's sample buffer (22) and were separated by 6% SDS-PAGE. The SDS-PAGE resolved proteins were then transferred to nitrocellulose membranes at 45 mV for 18 -20 h at 4°C in the presence of 0.01% SDS according to Towbin et al. (23). The nitrocellulose membrane was blocked for 1 h with a blocking buffer (phosphate-buffered saline containing 0.5% Tween 20 and 5% skim milk powder). The blocked membrane was incubated with primary antibodies, anti-RyR3 (34C), anti-c-Myc, or anti-HA antibodies for 2-4 h and washed for 15 min three times with the blocking buffer. The membrane was then incubated with the secondary anti-mouse IgG (H&L) antibodies conjugated with alkaline phosphatase for 30 -40 min. After washing, the bound antibodies were visualized by the alkaline phosphatase-mediated color reaction using nitro blue tetrazolium and 5-bromo-4-chloro-3-indolyl phosphate as the substrates.
Ca 2ϩ Release Measurements-The free cytosolic Ca 2ϩ concentration in transfected HEK293 cells was measured using the fluorescence Ca 2ϩ indicator dye fluo-3-AM as described previously (17), with some modifications. Cells grown for ϳ18 h after transfection were washed four times with phosphate-buffered saline (137 mM NaCl, 8 mM Na 2 HPO 4 , 1.5 mM KH 2 PO 4 , 2.7 mM KCl) and incubated in KRH buffer (125 mM NaCl, 5 mM KCl, 1.2 mM KH 2 PO 4 , 6 mM glucose, 1.2 mM MgCl 2 , 2 mM CaCl 2 , and 25 mM Hepes, pH 7.4) without MgCl 2 or CaCl 2 at room temperature for 45 min and then at 37°C for 45 min. After being detached from culture dishes by pipetting, the cells were collected by centrifugation at 1,000 rpm for 5 min in a Thermo/EC Centra CL2 centrifuge. The cell pellets were resuspended in Dulbecco's modified Eagle's medium supplemented with 0.1 mM nonessential amino acids, 4 mM L-glutamine, 100 units of penicillin/ml, 100 mg of streptomycin/ml, 4.5 g of glucose/liter, and 10% fetal calf serum and were loaded with 10 M fluo-3-AM at room temperature for 60 min. The fluo-3-loaded cells were washed with KRH buffer three times and resuspended in KRH buffer plus 0.1 mg/ml bovine serum albumin and 250 M sulfinpyrazone. An aliquot of fluo-3-loaded cells was then added to 2 ml (final volume) of KRH buffer in a cuvette, and the fluorescence intensity of fluo-3 at 530 nm was measured in an SLM-Aminco series 2 luminescence spectrometer with 480-nm excitation at 25°C (SLM Instruments, Urbana, IL).
[ 3 H]Ryanodine Binding-Preparation of cell lysates from transfected HEK293 cells was carried out as described previously (21). Equilibrium [ 3 H]ryanodine binding to cell lysate was also performed as described previously with some modifications (21). A binding mixture (300 l) containing 30 l of cell lysate (3-5 mg/ml), 500 mM KCl, 25 mM Tris, 50 mM Hepes, pH 7.4, 5 nM [ 3 H]ryanodine, 0.4 M free Ca 2ϩ , and a protease inhibitor mix containing 1 mM benzamidine, 2 g/ml leupeptin, 2 g/ml pepstatin A, 2 g/ml aprotinin, and 0.5 mM phenylmethylsulfonyl fluoride, was incubated at 37°C for 2.5 h. The binding mix was diluted with 5 ml of ice-cold washing buffer containing 25 mM Tris, pH 8.0, and 250 mM KCl and immediately filtered through Whatman GF/B filters presoaked with 1% polyethylenimine. The filters were washed four times with 5 ml of ice-cold washing buffer, and the radioactivity associated with the filters was determined by liquid scintillation counting. Nonspecific binding was determined by measuring [ 3 H]ryanodine binding in the presence of 50 M unlabeled ryanodine. All of the binding assays were done in duplicate.

Identification of Alternatively Spliced Variants of RyR3-
The entire coding region of RyR3 from rabbit uterus was amplified by using RT-PCR and eight pairs of PCR primers (Fig.  1A). The resulting overlapping cDNA fragments (PCR1 through PCR8) were subcloned into the pBluescript vector. At least 14 individual clones of each RT-PCR fragment were isolated and analyzed by restriction endonuclease digestion. A single digestion pattern was observed when individual clones of the PCR1, PCR3, and PCR4 fragments were digested with multiple restriction endonucleases (data not shown). On the other hand, two or more digestion patterns were detected in clones of PCR2, PCR5, PCR6, PCR7, and PCR8, indicating that each of these later PCR products contain heterogeneous DNA species.
The nature of this heterogeneity was further investigated by DNA sequence analysis. Two or three clones of each PCR fragment representing each unique digestion pattern were sequenced. This analysis revealed that as many as seven regions in RyR3 cDNA were either excluded or included ( Fig. 1B and Table I). Exclusion of regions AS-6, AS-7, AS-8a, and AS-8b led to deletions of 5, 6, 29, and 51 amino acids, respectively, without changing the reading frame. Exclusion of region AS-7 also led to a substitution of Arg for Gly 3724 . On the other hand, exclusion of regions AS-2 and AS-5 resulted in a frameshift. Different from other regions, AS-4 represented an insertion of a single serine residue after valine 2271 as a result of using an alternative splicing acceptor site (Table I) the sequences of these deleted regions (Table I) reveals that each of these regions corresponds to one or more exons of the human RyR3 gene (Table I and Fig. 1B). Thus, the exclusion of these regions is most likely generated by alternative splicing.
Tissue Distribution of Alternatively Spliced Variants of RyR3-To examine whether alternative splicing of the RyR3 transcript also occurs in tissues other than uterus, total RNA from rabbit vas deferens, aorta, stomach, small intestine, heart, brain, and diaphragm were isolated and used for RT-PCR. RyR3-specific primers flanking each splice region identified in uterus (except for AS-4) were used to amplify cDNA fragments both containing and lacking each splice region at the same time. RT-PCR products along with those generated from plasmid clones containing (control 1) or lacking (control 2) the known splice region were analyzed by polyacrylamide gel electrophoresis (Fig. 2). The sequences of RT-PCR fragments obtained from different tissues were confirmed to be identical to RyR3 by direct sequencing. Both the exclusion and inclusion of splice regions AS-2, AS-5, AS-7, AS-8a, and AS-8b were observed in all smooth muscle tissues examined (Fig. 2, a, b, and  d-f, lanes 3-7). However, for the most part, only the inclusion of these regions was detected in heart, brain, and diaphragm ( Fig. 2, a, b, and d-f, lanes 8 -10). Thus, the deletion of these regions seems to be smooth muscle-specific. On the other hand, only the exclusion of region AS-6 was detected in smooth muscle tissues, heart, and diaphragm, whereas both the exclusion and inclusion of this region were observed in brain. Hence, the inclusion of region AS-6 appears to be brain-specific.

The AS-8a Splice Variant Is Highly and Selectively Expressed in Smooth Muscle
Tissues-Of all splice variants detected in smooth muscle tissues, splice variant AS-8a (containing deletion of the AS-8a region) appears to be highly expressed (Fig. 2e). To further quantify the level of the AS-8a splice variant, we performed a ribonuclease protection assay using total RNA isolated from uterus, aorta, heart, brain, and diaphragm and an antisense 32 P-labeled RNA probe that encompasses the AS-8a region. A major band with a size of ϳ114 bases and a minor band of ϳ231 bases were detected in uterus and aorta (Fig. 3). The 231-base band corresponds to transcripts that contain the AS-8a region (AS-8a (ϩ)), whereas the 114-base band represents transcripts that lack the AS-8a region (AS-8a (Ϫ)). The relative level of the 114-base AS-8a (Ϫ) band, as determined by phosphorimaging analysis, was more than 5-fold of that of the 231-base AS-8a (ϩ) band in uterus and aorta. In contrast, the relative level of AS-8a (ϩ) band (231 bases) detected in heart, brain, and diaphragm was about 5-fold of that of the AS-8a (Ϫ) band (114 bases). The AS-8a (Ϫ) transcript was also found to be the major transcript expressed in vas deferens and stomach (data not shown). Hence, consistent with the results of RT-PCR (Fig. 2), these ribonuclease protection assay results indicate that the AS-8a splice variant is highly expressed in smooth muscle tissues but not in the heart, brain, or diaphragm.
Formation of RyR3 (AS-8a)/RyR3 (wt) Heteromeric Complexes-To investigate the functional properties of the highly expressed AS-8a splice variant, we constructed a RyR3 cDNA The nucleotide and amino acid sequences of rabbit RyR3 around the splicing junctions are shown. The boxed nucleotide sequences represent exons of the RyR3 gene. Because the complete intron/exon boundaries of the rabbit RyR3 gene are unknown, each spliced exon is indicated by the corresponding exon number of the human RyR3 gene. Horizontal lines represent splicing that yields the wt RyR3 cDNA, whereas tilted lines represent alternative splicing found in this study in which one or two exons (boxed nucleotide sequences) were excluded, with the exception of AS-4. In the case of AS-4, sequence analysis of the rabbit RyR3 genomic DNA around the AS-4 region revealed two potential acceptor sites, cag, in the intron sequence (italics), leading to two alternatively spliced transcripts with or without the insertion of three nucleotides encoding the amino acid, serine. In the cases of AS-2, AS-5, AS-6, or AS-8a, one entire exon was spliced out. In the case of AS-8b, two exons were removed. In the case of AS-7, alternative splicing resulted in a deletion of an exon and the change of glycine at position 3724 to arginine. The location and length of the deleted cDNA sequences corresponding to one or two exons are indicated.
containing a deletion of the 87-bp AS-8a region and expressed the RyR3 (AS-8a) cDNA in HEK293 cells. Functional characterization revealed that the AS-8a splice variant did not form a functional Ca 2ϩ release channel when expressed alone in HEK293 cells.
Because both the alternatively spliced and unspliced forms of AS-8a transcripts are present in smooth muscle tissues, it is possible that the AS-8a splice variant and the unspliced RyR3 form heteromeric RyR3 channels. To test this possibility, we inserted the c-Myc antibody epitope tag into RyR3 (wt) and the HA tag into RyR3 (AS-8a) and co-expressed them in HEK293 cells. Interactions between RyR3 (wt, c-Myc) and RyR3 (AS-8a, HA) were examined by immunoprecipitation. Fig. 4a 1, bottom panel). These data indicate that the anti-c-Myc and anti-HA antibodies are specific and that RyR3 (AS-8a, HA) when expressed together with RyR3 (wt, c-Myc) can be co-precipitated with RyR3 (wt, c-Myc) by the anti-c-Myc antibody.
We also performed the reciprocal experiment in which the RyR3 (AS-8a, HA) was immunoprecipitated by the anti-HA antibody and the presence of RyR3 (wt, c-Myc) in the anti-HA immunoprecipitates was detected by Western blotting. As shown in Fig. 4b (top panel) shows PCR products using a plasmid clone that contains one of the splice regions as the template, whereas control 2 (lane 2) shows PCR products using a plasmid clone that lacks one of the splice regions as the template. Control 1 for AS-6 (c) is not available, because RyR3 expressed in uterus does not contain the AS-6 region. No signals were detected in RT-PCR in the absence of reverse transcriptase. The sizes of PCR fragments that contain (ϩ) or lack (Ϫ) the corresponding splice region are indicated on the right.

FIG. 3. Splice variant RyR3 (AS-8a) is highly expressed in smooth muscle tissues.
The relative level of the AS-8a splice variant expressed in various tissues was examined by ribonuclease protection assay. An antisense 32 P-labeled RNA probe (307 bases) was hybridized with total RNA isolated from uterus, aorta, heart, brain, and diaphragm. The probes protected from RNases digestion were analyzed by polyacrylamide gel electrophoresis and autoradiography. The 231-base fragment represents RyR3 transcripts containing the AS-8a region, whereas the 114-base fragment represents RyR3 transcripts lacking the AS-8a region. RyR3 (AS-8a) splice variant is able to form heteromeric complexes with the wild type RyR3.

The RyR3 (AS-8a) Splice Variant Is Able to Form Functional Heteromeric Channels with a RyR3
Mutant-The ability of the RyR3 (AS-8a) splice variant to form heteromeric complexes was further assessed by complementation analysis using a RyR3 mutant. The rationale being that if RyR3 (AS-8a) is able to form heteromeric channels with other RyR3 variants, co-expression of RyR3 (AS-8a) with a nonfunctional RyR3 mutant may produce a functional heteromeric channel, because two nonfunctional mutants may complement each other's defects. During deletion analysis of RyR3, we generated a caffeineinsensitive RyR3 mutant, RyR3 (⌬E2256-T2429). To test whether RyR3 (AS-8a) can complement this mutant, HEK293 cells were transfected individually or in combination with RyR3 (AS-8a) and RyR3 (⌬E2256-T2429). As indicated in Fig.  5, HEK293 cells transfected with RyR3 (AS-8a; Fig. 5a) or RyR3 (⌬E2256-T2429; Fig. 5b) alone did not respond to caffeine, whereas co-expression of RyR3 (AS-8a) and RyR3 (⌬E2256-T2429) led to a caffeine-sensitive Ca 2ϩ release (Fig. 5,  c-e). These data suggest that the RyR3 (AS-8a) splice variant is able to form functional heteromeric channels with other RyR3 variants.
Co-expression of RyR3 (AS-8a) Decreases the Sensitivity of RyR3 (wt) to Caffeine Activation-The functional aspect of the RyR3 (AS-8a)/RyR3 (wt) heteromeric complexes was further investigated by examining their caffeine response. HEK293 cells were co-transfected with RyR3 (wt) and pCDNA3 vector cDNA or with RyR3 (wt) and RyR3 (AS-8a). The peaks of Ca 2ϩ release from aliquots of transfected cells induced by different concentrations of caffeine (0.05-20 mM) were measured. As shown in Fig. 6, Ca 2ϩ release from RyR3 (wt)-transfected cells was activated by caffeine with an EC 50 of 0.87 ϩ 0.09 mM (mean Ϯ S.E., n ϭ 3). On the other hand, Ca 2ϩ release from cells co-transfected with RyR3 (wt) and RyR3 (AS-8a) was activated by caffeine with an EC 50 of 2.4 ϩ 0.02 mM (n ϭ 3). Therefore, co-expression of RyR3 (AS-8a) reduces the caffeine sensitivity of RyR3 (wt), suggesting that by forming heteromeric channels, the RyR3 (AS-8a) splice variant can influence the activity of RyR3 (wt).
The RyR3 (AS-8a) Splice Variant Is Capable of Interacting with and Suppressing the Activity of RyR2-We have recently shown that RyR3 (wt) is capable of forming heteromeric channels with RyR2 (wt) in HEK293 cells (24). One would expect that the RyR3 (AS-8a) splice variant would be also able to form heteromeric channels with RyR2 (wt). To directly test this hypothesis, we co-expressed HA-tagged RyR3 (AS-8a, HA) with c-Myc-tagged RyR2 (wt, c-Myc) in HEK293 cells. The association of RyR3 (AS-8a, HA) with RyR2 (wt, c-Myc) was examined by immunoprecipitation followed by immunoblotting. As indicated in Fig. 7A (panel a) . Similarly, in a reciprocal experiment, the anti-HA antibody was able to pull down RyR2 (wt, c-Myc) from cells co-transfected with RyR3 (AS-8a, HA) and RyR2 (wt, c-Myc) (Fig. 7A, panel b, lane 3), but not from cells transfected with RyR2 (wt, c-Myc) alone (Fig. 7A, panel b, lane  1). Taken together, these data indicate that RyR3 (AS-8a, HA) and RyR2 (wt, c-Myc) when expressed together can be coprecipitated either by anti-HA or anti-c-Myc antibody, demonstrating that the RyR3 (AS-8a) splice variant, like RyR3 (wt), can form heteromeric complexes with RyR2 (wt).
To examine whether the physical interaction between RyR3 (AS-8a) and RyR2 (wt) affect the activity of RyR2 (wt), we co-transfected HEK293 cells with RyR2 (wt) and RyR3 (AS-8a) or RyR2 (wt) with vector DNA, pCDNA3. [ 3 H]Ryanodine binding to lysates of co-transfected HEK293 cells were determined. As shown in Fig. 7B, no [ 3 H]ryanodine binding was detected in cells transfected with RyR3 (AS-8a) alone, consistent with the results of Ca 2ϩ release experiments (Fig. 5a) that showed that RyR3 (AS-8a) does not form a functional Ca 2ϩ release channel when expressed alone in HEK293 cells. Furthermore, co-expression of RyR3 (AS-8a) reduced [ 3 H]ryanodine binding to RyR2 (wt) by about 50%. [ 3 H]ryanodine binding has been widely used as a functional assay for RyR channel activities, because ryanodine binds to only the open state of the channel. Thus, these studies indicate that the RyR3 (AS-8a) splice variant is able to suppress not only the activity of RyR3 (wt) (Fig.   FIG. 5. RyR3 (AS-8a) is capable of forming functional heteromeric channels. HEK293 cells were transfected with RyR3 (AS-8a) (16 g) (a), RyR3 (⌬E2256-T2429) (16 g) (b), RyR3 (AS-8a) (4 g) plus RyR3 (⌬E2256-T2429) (12 g) (c), RyR3 (AS-8a) (8 g) plus RyR3 (⌬E2256-T2429) (8 g) (d), or RyR3 (AS-8a) (12 g) plus RyR3 (⌬E2256-T2429) (4 g) (e). The transfected cells were loaded with 10 M fluo-3-AM, and the fluorescent intensity was monitored continuously before and after addition of 5 mM caffeine as indicated by the letter C. A decrease in fluorescence immediately after the addition of caffeine was the result of fluorescence quenching by caffeine. A transient increase, although small, was consistently observed in co-transfected cells (c-e). Similar results were obtained from three separate experiments. 6) but also the activity of RyR2 (wt) (Fig. 7B) through the formation of heteromeric channels.

DISCUSSION
The results of our present study demonstrate that the majority of the RyR3 transcripts expressed in various smooth muscle tissues contain a deletion of an 87-base pair region encoding a 29-amino acid fragment (His 4406 -Lys 4434 ) encompassing a predicted transmembrane segment (25,26) (Fig. 1C), as a result of alternative splicing. This major RyR3 splice variant, RyR3 (AS-8a), when expressed by itself in HEK293 cells, does not form a functional Ca 2ϩ release channel but in combination with the wild type RyR3 is able to form functional heteromeric channels with reduced caffeine sensitivity (Figs. 5 and 6). Furthermore, this major RyR3 splice variant is also able to form heteromeric channels with and suppress the activity of RyR2 (Fig. 7). Together, these observations reveal a novel mechanism of RyR regulation in which the activity of RyR3 and RyR2 can be inhibited by a RyR3 splice variant through a dominant negative effect.
Among the three known RyR isoforms, RyR3 is the most heterogeneous in both function and regulation (3). One of its major heterogeneities is its response to caffeine activation. RyR3 expressed in skeletal muscle and the brain is caffeineand ryanodine-sensitive (9,16), whereas RyR3 expressed in uterine smooth muscle cells is ryanodine-sensitive but caffeineinsensitive (11,12). The molecular basis for this tissue-or cell-dependent RyR3 heterogeneity is unclear. Our findings that the RyR3 (AS-8a) splice variant is highly expressed in smooth muscle tissues but not in skeletal muscle and the brain and that co-expression of this major splice variant with the wild type RyR3 reduces the caffeine sensitivity of RyR3 strongly suggest that tissue-specific expression of RyR3 splice variants, in particular the RyR3 (AS-8a) splice variant, may account for the heterogeneous caffeine response of RyR3 observed in different tissues or cells. In support of this view, the corresponding AS-8a region was found to be deleted in RyR3 from mink lung epithelial cells, which also display ryanodinesensitive but caffeine-insensitive Ca 2ϩ release (14). Ryanodinesensitive but caffeine-insensitive Ca 2ϩ release activity was also observed in human Jurkat T-cells (15), and it will be interesting to see whether the RyR3 (AS-8a) splice variant is expressed in these cells.
The physiological function of the AS-8a splice variant is unknown. Because caffeine activates RyR by sensitizing the channel to Ca 2ϩ activation, altered caffeine response may reflect changes in Ca 2ϩ regulation. In this context, it is of interest to know that RyR3 expressed in nonpregnant myometrial smooth muscle cells does not respond to activation by Ca 2ϩ and caffeine under normal sarcoplasmic reticulum Ca 2ϩ loading but becomes active when sarcoplasmic reticulum Ca 2ϩ loading is increased (27). This observation suggests that RyR3 from these smooth muscle cells has a reduced sensitivity to activation by luminal Ca 2ϩ and that its activity is normally suppressed. It remains to be explored whether the RyR3 (AS-8a) splice variant is highly expressed in these smooth muscle cells and whether this splice variant is involved in luminal Ca 2ϩ regulation.
Another observation that may provide some clues to the physiological role of the RyR3 (AS-8a) splice variant comes from a study using RyR3 knock-out mice. Cerebral artery smooth muscle cells isolated from these mice displayed an increased frequency of Ca 2ϩ sparks and spontaneous transient outward currents and a reduced myogenic tone as compared with the wild type cells (13). These findings led to the suggestion that Ca 2ϩ sparks and spontaneous transient outward currents in arterial vascular smooth muscle cells are negatively regulated by RyR3. The molecular mechanism by which RyR3 inhibits Ca 2ϩ spark frequency is not clear. It has been proposed that RyR1 and RyR2, but not RyR3, are responsible for Ca 2ϩ spark generation and that inhibition of Ca 2ϩ spark frequency may result from RyR3-mediated prolonged Ca 2ϩ release, which may then inactivate RyR1 or RyR2 (13). The findings that RyR3 expressed in myometrial smooth muscle cells is inactive under normal sarcoplasmic reticulum Ca 2ϩ loading and that the RyR3 (AS-8a) splice variant, the major form of RyR3 expressed in various smooth muscle tissues, does not form a functional Ca 2ϩ release channel when expressed alone in HEK293 cells are inconsistent with the idea of RyR3-mediated prolonged Ca 2ϩ release. Alternatively, our observation that the RyR3 (AS-8a) splice variant is able to interact physically with and suppress the activity of RyR2 suggests the interesting possibility that RyR3 expressed in smooth muscle cells as a splice variant may suppress Ca 2ϩ sparks, spontaneous transient outward currents, and myogenic tone by forming heteromeric channel complexes with RyR2. Thus, it appears that the main functional role of RyR3 in smooth muscle cells differs from that of RyR3 in skeletal muscle and the brain. Unlike RyR3 expressed in skeletal muscle, where it functions as a Ca 2ϩ release channel, RyR3 expressed in smooth muscles or epithelial cells in the form of alternatively spliced variants may function largely as a suppressor of Ca 2ϩ release.
The observation that very low levels of RyR3 transcripts that contain the AS-8a region were detected in various smooth muscle tissues as compared with those of the AS-8a splice transcripts is in agreement with this view (Fig. 2). An excess amount of the AS-8a splice variant would ensure that all of the wt RyR3 subunits would be oligomerized with the AS-8a, so that no homomeric wt RyR3 channels or heteromeric wt RyR3/ AS-8a channels with high wt RyR3:AS-8a subunit ratios would be formed to suppress the activity of the wt RyR3 channels. Because AS-8a can also form heteromeric channels with wt RyR2, the AS-8a splice variant may also be involved in oligomerization with wt RyR2 where they are co-expressed, as shown in vascular smooth muscle. It is important to note that an excess amount of AS-8a does not completely suppress the activity of either wt RyR3 or wt RyR2. We were able to detect the activity of RyR3 or RyR2, although at reduced levels, after co-expression of AS-8a with wt RyR3 in a ratio of 15:1 (AS-8a:wt RyR3) or with RyR2 in a 10:1 ratio (AS-8a:wt RyR2) in HEK293 cells (Figs. 6 and 7).
In addition to AS-8a, several other splice regions in RyR3 have also been detected in various smooth muscle tissues (Fig.  2). Of these splice regions, AS-2 and AS-5 are of interest. They are located near the 5Ј-end and in the middle of the RyR3 cDNA, respectively (Fig. 1). Exclusion of each of these regions leads to a frameshift and is predicted to result in the synthesis of truncated RyR3 proteins containing the first NH 2 -terminal ϳ800 and ϳ2900 amino acid residues, respectively. These truncated RyR3 proteins lack the COOH-terminal pore-forming region and, if expressed alone, would be nonfunctional. However, whether they can form heteromeric channel complexes with and thereby affect the channel activity of the full-length RyR3 or RyR2 remains to be assessed. We have previously shown that co-expression of NH 2 -terminal fragments of RyR2 with overlapping COOH-terminal fragments produces functional Ca 2ϩ release channels in HEK293 cells (28). This observation indicates that the NH 2 -terminal region of RyR2 is able to interact functionally with the COOH-terminal region. It will be of interest to determine whether the NH 2 -terminal regions of RyR3, corresponding to the AS-2 and AS-5 splice variants, are able to interact functionally with the COOH-terminal regions of RyR3 and RyR2.
The presence of these frameshifted splice variants raises the possibility that splice variants such as AS-8a, whose splice regions are located downstream of the frameshifted splice regions, are not made into proteins in smooth muscles, despite their existence at the RNA level. Based on the results shown in Fig. 2, it is clear that not all of the RyR3 transcripts are alternatively spliced in the AS-2 or AS-5 regions, because considerable levels of AS-2 (ϩ) and AS-5 (ϩ) transcripts were detected in various smooth muscle tissues. Thus, some RyR3 transcripts without the AS-2 and AS-5 deletions are likely to exist and be translated. Consistent with this view, RyR3 expressed in uterine smooth muscle cells has been shown to function as a Ca 2ϩ release channel with properties different from those of RyR3 expressed in skeletal muscle and the brain (27). Furthermore, RyR3 expressed in vascular myocytes is capable of binding fluorescent ryanodine, suggesting that vascular RyR3 is also functional (29). These observations indicate that, although frameshifted alternatively spliced transcripts are present in smooth muscles, not all of the RyR3 proteins expressed in smooth muscles were in the truncated form and that some of the RyR3 transcripts must have been translated all the way to the 3Ј-end encoding the channel conduction pathway to be functional. The presence of both the alternatively spliced and nonspliced transcripts and the detection of functional RyR3 with altered properties suggest that multiple alternatively spliced variants are co-expressed in smooth muscles at the protein level. To test this possibility, antibodies that recognize the NH 2 -terminal region of RyR3 would be useful in detecting the expression of the AS-2 and AS-5 proteins. Direct amino acid determination of the appropriate protease fragments of RyR3 isolated from smooth muscle tissues that encompasses the AS-8a or other splice regions by the use of mass spectrometry would represent an alternative approach to definitively demonstrate the existence of AS-8a or other splice variants at the protein level. Further investigations are required to characterize the functional properties of other potential RyR3 splice variants and to delineate the physiological roles of RyR3 splice variants in smooth muscles and other cells.