Mutually Exclusive and Cassette Exons Underlie Alternatively Spliced Isoforms of the NdCa Exchanger*

We have analyzed the gene structure that gives rise to tissue-specific isoforms of the NdCa exchanger. Five dis- tinct isoforms of the NdCa exchanger from rabbit brain, kidney, and heart have been identified previously to which we now add a new brain isoform. Reverse-tran-scribed polymerase chain reaction, library screening, and sequence analysis of cDNA coding regions indicate that the only significant alteration of the NdCa ex- changer cDNA in rabbit brain, kidney, and heart iso- forms is located in the carboxyl end of the putative intracellular loop of the protein, a region recently linked to ionic and metabolic regulation of the NdCa exchanger. Additionally, we find that the NdCa exchanger isoforms found in lung and skeletal muscle may arise from among these same six isoforms. Examination of the gene structure of the NdCa exchanger in rabbit indicates how the single gene that encodes for the NdCa exchanger is alternatively spliced to give rise to the five rabbit isoforms. Specifically, sequence analysis of the intron-exon boundaries reveals the presence of two “mutually exclusive” exons in conjunction with four “cassette” exons in the region of the NdCa exchanger gene that codes for the carboxyl end of the predicted intracellular loop region. This unusual arrangement of exons in the NdCa exchanger gene could allow for the generation of up to 32 different NdCa Characterization-A genomic Stratagene with a PCR product from rabbit heart encoding amino acids 510-682 (29). Hybridization and washing conditions were similar for cDNA clone isolation. The positives plaques isolated clonal purity and digested with EcoRI, to a nylon filter by capillary action, and probed with the same PCR product. The fragments that hybridized to this product were then subcloned and sequenced with exon specific oligonucleotides. The sequence of exons and intron-exons boundaries were confirmed in opposite direction by complementary oligonucleotides. The physical map of the genomic frag- ment was performed by cutting the phage DNA with NotZ, followed by

The sodium-calcium (NdCa) exchanger is a transmembrane protein found broadly in animal cells, having been first identified in mammalian heart muscle and squid giant axons (1)(2)(3). This electrogenic transporter couples sodium flux to the countertransport of calcium ions with a stoichiometry of three Na' ions to each Ca2+ ion (4, 5). In heart the NdCa exchanger is responsible for extruding almost all of the calcium that enters the cell during excitation via the sarcolemmal calcium current, Ica (6-8). I n kidney, this exchanger appears to play an important role in regulating calcium re-absorption in the nephron (9, 10). The NdCa exchanger has been found in numerous tissues by immunolocalization and shown to be present in particularly high concentrations in the neuronal synapses (11) and may help to extrude the calcium that enters to activate neurotransmitter release (12). The NdCa exchanger is also found in many * This work was supported by National Institutes of Health Grants other cell types, including astrocytes (12), pancreatic p-cells (13, 141, lung (14), liver (141, and skeletal muscle (14)(15)(16). With the multiplicity of function of these diverse tissues, it might have been expected that a collection of genes would encode for the different transport proteins. However, examinations of the number of genes coding for the NdCa exchanger in human (17,18) a n d rat (14) suggests that a single gene encodes this transporter. This finding is consistent with the remarkable similarity of the cDNA sequences that have been reported. The similarities are found both across tissues (heart, kidney and brain) a n d across species (dog, human, rabbit, rat) (14,17,(19)(20)(21)(22)(23) and raises the possibility that an alternatively spliced single gene product may account for the observed cDNAs.
In the present work focusing particularly on the rabbit, we have investigated the partial gene structure of the N d C a exchanger and have also characterized the principal isoforms of the NdCa exchanger cDNAs found in rabbit heart, kidney, and brain. Six distinct isoforms have been identified including a novel brain isoform reported here for the first time. We have also found that the N d C a exchanger isoforms found in lung and skeletal muscle share common features with these same six isoforms. Our investigation into the structure of the N d C a exchanger gene suggests that all of the isoforms identified by us and others can be explained by two mutually exclusive exons and four cassette exons (24) that together could produce up to 32 distinct isoforms.
EXPERIMENTAL. PROCEDURES Material~-[a-~~P]dCTP, [y-32PldCTP, and [CY-~~SI~ATP were purchased from Amersham Corp. Restriction enzymes and reverse transcriptase were purchased from Life Technologies, Inc. T4 kinase, T4 ligase, and calf intestinal alkaline phosphatase were obtained from Boehringer Mannheim. Chemicals were molecular biology grade acquired from Sigma and Fischer.
Screening of cDNA Library-An adult rabbit brain cDNA library in AglO was purchased from Clontech (Palo Alto, CAI and screened by plaque lifting method (25) using a random-primed 32P-labeled human NdCa exchanger cDNAfragment (nucleotides 1-1830) (14). The hybridizations were carried out at 42 "C in 50% formamide, 6 x standard sodium phosphate EDTA solution (SSPE) (25), 5 x Denhardt's solution, and 0.5% SDS. The final wash of the filters were at 0.2 x SSPE at 50 "C. The positive plaques were resuspended in 50 m~ Tris, pH 7.4, 10 m~ MgSO,, 100 m NaCI, 0.01% gelatin and screened again at lower densities until clonal purity was achieved. Phage lysate DNA was prepared by purification in DEAE-cellulose column (26) and then analyzed by EcoRI restriction digest and electrophoresis in 1% agarose gel. Polymerase Chain Reaction (PCRY-Total RNA was isolated from frozen tissues (27) and reverse transcribed in a reaction mixture containing 20 units of RNasin, deoxynucleotide triphosphates to a 1 m final concentration, 200 units of Moloney murine reverse transcriptase, and 50 pmol of random hexamer or oligo(dT) primer. The reaction was incubated for 1 h and then phenol/chloroform-extracted followed by a The abbreviations used are: PCR, polymerase chain reaction; bp, base pair; kb, kilobase. step of ethanol precipitation and resuspension in HzO. The PCR (28) was performed by adding 1-5 pl of cDNA mixture to a reaction mixture in 50 pl contained 50 m~ KCl, 10 m Tris (pH 9 at 25 "C), 2.5 m MgCl,., 0.1% Triton X-100, 0.2 m~ dATP, dCTP, dGTP, and dl", 2.5 units of Vent DNA polymerase (New England Biolabs), and 50 pmol of each primer. The 5' primer (P1) CTGTGACTCATGTGAGTGAG (sense) recognizes the nucleotides encoding amino acids 511-516 and the 3' primer (P2) l"l'C'lTCAATGATCACTTCCAA(antisense) the amino acids 677-683 in the heart exchanger (29). The amplification was conducted in a thermal cycler (Ericomp, San Diego, CA) under the following conditions: initial denaturation at 94 "C for 5 min, followed by 35 cycles of 1 min at 55 "C, 1 min at 72 "C, and 1 min at 94 "C. The PCR products were then analyzed in a 3% Methaphor agarose (FMC, Rockland, ME) gel.
Genomic DNA Mapping and Characterization-A rabbit genomic library in AFWI vector was purchased from Stratagene (La Jolla, CA) and screened with a PCR product from rabbit heart encoding amino acids 510-682 (29). Hybridization and washing conditions were similar to described above for cDNA clone isolation. The positives plaques were isolated to clonal purity and digested with EcoRI, transferred to a nylon filter by capillary action, and probed with the same PCR product. The fragments that hybridized to this product were then subcloned and sequenced with exon specific oligonucleotides. The sequence of exons and intron-exons boundaries were confirmed in opposite direction by complementary oligonucleotides. The physical map of the genomic fragment was performed by cutting the phage DNA with NotZ, followed by partial digestion with EcoRI, Southern transfer, and probing with T7 and T3 oligonucleotide primers. The distance between exons were determined by a combined approach of restriction digestion and PCR amplification between exons using phage DNA as template.
DNA Sequencing-DNA fragments from the genomic and c D N h libraries were eluted using a DNA purification kit and subcloned in pGEM7Z(+), pBluescript SK(+) or pBluescript KSII(+), and sequenced using the dideoxy chain termination method using T7 DNA polymerase. Sequencing analyses were performed using the PCGene software package (IntelliGenetics, Mountain View, CA).
Nomenclature-The gene designation for the NdCa exchanger is NCXl and is located on the short arm of the human chromosome 2 (18,30). To uniquely identify the expressed isoforms of the NdCa exchanger, we have used the terms NACAl for the cardiac sarcolemmal NdCa exchanger (14,17,19,20,22), NACA2 for isoform isolated from rabbit kidney cortex by Reilly and Shugrue (23), NACM for the new kidney isoform (31), NACA4 and NACA5 for the two brain isoforms isolated very recently from rat brain libraries by Furman et al. (21), and NACA6 for the new brain isoform presented here. The full description of the cardiac isoform is thus NCX1-NACA1.

RESULTS
The first part of the work focuses on the identification of NdCa exchanger isoforms in rabbit with emphasis on brain, heart, and kidney. The second part of the work concentrates on the identification of large fragments of genomic DNA that contain all the isoforms specific exons of the NdCa exchanger so far described, including the new brain isoform presented here. The genomic fragments are then mapped and sequenced so that all introdexon boundaries are characterized.

NalCa Exchanger Isoforms in Rabbit
Brain-Our first goal was to determine the identity and sequence of NdCa exchanger message(s) in brain tissue. Thus, a rabbit brain cDNA library was screened a t moderate stringency with a DNA fragment derived from the human cardiac NdCa exchanger clone (141, resulting in the identification of two full-length clones with deduced open reading frames of 941 and 934 amino acids. Sequence comparison of the two clones showed that the translated protein of both clones would be identical with the exception of an insertion of 21 nucleotides encoding the sequence ALLLNEL (Fig. 1). The deduced polypeptide has a predicted mass of 105 kDa and a characteristic leader peptide at the amino-terminal end. Hydropathy analysis (32) indicates 11 potential membrane-spanning segments and a large hydrophilic domain between membrane-spanning segments 5 and 6, a structure similar to the one proposed for the cardiac NdCa exchange protein (19, 29). The large hydrophilic region is pre-

-D G E T R K I K H L P S P G I y E l & V
Translated amino acid sequence of the rabbit brain Na/Ca exchanger RB11. The numbering of the residues is shown on the left side with the amino acid number 1 designated based on the amino-terminal sequencing of the bovine cardiac sarcolemmal exchanger (42). The putative transmembrane segments are represented by double lines, and the position of the PCR primers are indicated with arrows. The segment absent in clone RB20 is shadowed, and the amino acid changes compared with the rabbit kidney NdCa exchanger (23) are marked with asterisks. sumed to be located on the intracellular side of the membrane and is thought to be involved in the modulatory aspects of the exchanger function (33). With this membrane topology, the 21nucleotide insertion would occur in the carboxyl end of the intracellular loop between amino acids 603-609.
Recently, Furman et al. (21) published two NdCa exchanger isoforms from rat brain libraries. One clone is nearly identical to the -20 clone and is probably the rat homolog of this brain isoform from rabbit. The second one, isolated only from a rat hippocampal cDNA library, was unique as it contained an insertion of 23 amino acids not seen in both rabbit brain fulllength clones (as discussed below). The second rabbit brain isoform that we obtained (RB11) was not was not noted by The two rabbit brain clones were strikingly similar to the rabbit kidney exchanger cloned recently by Reilly and Shugrue (231, with overall 97% identity at amino acid level. Most of the amino acids substitutions between the brain and kidney clones were located in the carboxyl end of the putative intracellular loop as indicated in Fig. 1. PCR Amplification from Distinct Tissues-The results obtained from our rabbit brain library screening suggest that the difference between the rabbit brain NdCa exchanger cDNAs identified above is produced by the insertion or deletion of a small 21-bp exon in the putative intracellular loop. In order to investigate the presence of additional isoforms in this "variable'' region of the NdCa exchanger in brain and other tissues, we performed a reverse-transcribed PCR analysis on first strand of cDNA made from rabbit kidney, heart, brain, skeletal muscle, and lung. The primers were designed to flank the area of sequence divergence between clones RBll and RB20, as shown in Fig. 1, and expected to generate a amplification product of 450 and 429 bp, respectively, with brain rabbit cDNA if these were the only NdCa exchanger isoforms present in brain tissue. The results obtained with the PCR amplification, at high stringency conditions, are shown in Fig. 2B. Amplification using rabbit brain cDNA as template yielded a dominant band of about 420 bp and several other minor bands of larger size. Likewise, parallel amplification with rabbit cardiac and renal XIP the exchanger inhibitory peptide. B , the primers P1 and P2 were used for the amplification on cDNA from rabbit skeletal muscle (diaphragm), lung, heart, kidney, and brain. The PCR products were separated by agarose gel electrophoresis and stained with ethidium bromide. Therepresents the reaction with all components with the exception of the cDNA template. Markers were @X174 digested with Hinff.
cDNA generated bands of 530 and 420 bp, respectively. For the rabbit skeletal muscle cDNA the 530-and 420-bp products were equally intense. These results were consistently seen in two other amplifications with independent cDNA samples.
To further define the diversity of NdCa exchanger message, we proceed to subclone the PCR products and determine the nucleotide and amino acid sequence of several independent clones. The deduced amino acids from the PCR products amplified from rabbit kidney, heart, and brain cDNAs are shown in Fig. 3. All the PCR products derived from heart were identical (six independent clones) and very similar to the canine (98% identity at amino acid level) (19) and human (94% identity at amino acid level) (14) cardiac exchanger. Therefore, the cardiac tissue seems to express only a single NdCa exchanger isoform (NACA1) in agreement with the description of a single cardiac NdCa exchanger cDNA in numerous reports from different species (14,17,19,20,22).
In contrast, the PCR amplification of rabbit brain cDNA yielded two distinct isoforms (NACA4 and NACA6) which were identical in this variable area to the two isoforms obtained from the cDNA library screening and described above. We expect to find additional isoforms in the brain tissue by sequencing additional clones as several faint bands of higher molecular weight can be seen in agarose gels (Fig. 2). Additional diversity was observed by sequencing PCR products from the rabbit kidney cDNA. The isoform described by Reilly and Shugrue (23) (NACA2) and a shorter isoform (NACA3) described by us (31) were both revealed on sequence analysis. The majority of sequenced PCR products (7 out of 10) encoded the NACA3 isoform. This new isoform seemed to be the most abundant isoform in the kidney tissue also on RNase protection assays (31).
The six isoforms of the NdCa exchanger identified to date are presumably derived from a fixed array of exons spliced .-. , lated PCR products from heart, kidney, and brain of rabbit. The   FIG. 3. Comparison of the amino acid sequences f m m trans-PCR products shown in Fig. 2 were digested with EcoRI and cloned into plasmid vectors. The sequences were aligned, a dash indicates a potential deletion. The amino acids conserved in all PCR products are shown with a shadowed background, and the numbering of the NaJCa exchangers isoforms is explained in the text.
together. Another plausible explanation requires the transcription of additional genes, but this seems unlikely given the high degree of sequence conservation noted in the full-length isoforms and also and the presence of single bands in Southern blots of rat genomic DNA digested with several restriction enzymes (14). Additionally, localization of the human NdCa exchanger gene indicates that it resides at a single location on chromosome 2 (18,30).
Genomic Basis for Alternatively Spliced NalCa Exchanger mRNAs-To examine the possibility that alternative splicing gives rise to the diversity of NdCa exchanger isoforms, we began to characterize the structure of the rabbit gene fragments that contained the exons encoding the identified isoforms. We have attempted to determine the minimum number of exons, their position within the genome, and their reading frame. The work described here identifies the genomic organization needed to explain the known six isoforms and provide the foundation for predictions of other possible isoforms. The predictions are based on the sequenced cDNAs and the gene map presented below.
We undertook the isolation of a rabbit genomic fragment harboring the variable segment of the NdCa exchanger gene by plating and screening a rabbit genomic library. The initial screening at high stringency conditions was performed with the heart PCR product (see above) yielding several positive clones. One clone L211 with a total insert size over 15 kb contained nine EcoRI fragments and was further characterized by restriction site analysis. The resulting physical map of the partial NdCa exchanger gene and the relationship with the brain cDNAclone RBll are shown in Fig. 4. This large gene fragment contains only a small fraction of the NdCa exchanger coding sequence with very large intervening intronic areas. However, we cannot discard the possibility that additional exons will be found in these areas upon further sequencing.
The exons encoding the variable part of the NdCa exchanger message were found in this large genomic fragment by Southern blot of the EcoRI fragments and probing with exon specific oligonucleotides. Three EcoRI genomic fragments (1.8-, 2-, and 3.3-kb fragments) were isolated and the sequence of intronexon boundaries determined as shown in Fig. 5. The first two exons (A and B) were found in the 1.8-kb genomic fragment with the exon denoted A of 108 bp encoding amino acids 568-602 of the brain and cardiac exchanger and a second exon B of 105 bp encoding amino acids 568-601 of the kidney isoforms. Both exons are flanked by the characteristics AG donor splicing site in the 5' end and the GT consensus acceptor splice site in the 3' end (Fig. 5). Two interesting features of these exons are that they encode polypeptides with conserved amino acids and that the first amino acid in both exons (lysine) is in frame with the common 5' exon. If both exons were to be spliced together in RG. 6. Schematic representation of the variable region of the NdCa exchanger gene and the splicing pattern found in heart, kidney, and brain. The genomic organization is shown on top for the exons A-E and the splicing patterns for the distinct isoforms found in Merent tissues are shown below. The flanking sequences that are conserved for these tissues are represented by hatched boxes. The isoform NACAl was described in dog (19), human (14,17), rat (22), and bovine (20) heart. The isoforms NACA2 and NACA3 were described in rabbit kidney (23,31). The NACA4 and NACM were first isolated from rat brain (20, whereas the NACA6 isoform has been identified in this study. screening and by reverse-transcribed PCR amplification (Fig.  6). DNA fragments from the genomic clone L211 were subcloned in plasmid vector, and the sequences from the exons and exons-introns boundaries were determined by oligonucleotide-specific priming. The distance between the exons were determined by restriction analysis of the genomic clone and by PCR amplification of the intervening introns. The nucleotide sequences in capital letters represents the exons, whereas the lowercase letters represent the introns. The dots represent segments that were not sequenced. a single message, the deduced polypeptide would be in a different reading frame. Indeed, in all cDNAs encoding the NdCa exchanger, so far described, one and only one of these exons (A or B) is present.
Additionally four exons were found in the 2-and 3.3-kb fragments (Fig. 4). They encode for the amino acids ALLLNEL (exon C), GGETIT (exon D), GKYLY (exon E), GQPVLRKVHA-RDHPWSTVITIA (exon F) and are also flanked by the consensus intron-exon boundary sequences (Fig. 5). The NdCa exchanger has been studied in detail in recent years and is the best characterized antiport mechanistically. The stoichiometry, mechanism of reaction, turnover rate, and intracellular modulation has been investigated with increasingly elegant electrophysiological and biochemical techniques (see Ref. 3). These studies were carried out mainly in two types of preparations: cardiac myocytes and giant cells of invertebrates, primarily due to high density of NdCa exchangers in cardiac membrane and the excellent control of extracellular and intracellular environment in squid giant mons or barnacle muscle fibers. In other cell types, such as mammalian neurons, skeletal muscle fibers, epithelia from distal and proximal nephron, and pancreatic beta cells, the exchanger has been detected but not as well characterized. Biochemical and flux measurements are consistent with the notion that distinct forms of the NdCa exchange proteins are present in the central and peripheral nervous system (34), kidney (351, and in platelets (36).
Our previous work with the cloned human cardiac NdCa exchanger has started to address the issue of localized differences of the NdCa exchanger in diverse tissues (14). We found that transcripts encoding a homologous forms of the exchanger were present in high levels in lung, brain, and kidney when examined in Northern Blots. Furthermore, restriction fragment analysis of genomic DNA suggested that only a single gene encodes this family of plasma membrane transporter (14, 17).
Here we extended these initial observations by first cloning the rabbit brain isofoms of the NdCa exchanger using the human cardiac isoform as a probe. Sequence analysis showed the complete conservation of the rabbit brain NdCa exchanger in comparison with rabbit kidney exchanger (23) with the marked exception of the carboxyl end of the putative intracellular loop where there is only limited homology with the renal exchanger. The extensive amino acid conservation between the brain and cardiac sarcolemmal exchangers are consistent with the immunological studies by Yip et al. (37) reporting the similarity in size for the NdCa exchanger protein in canine heart and in rat brain. In previous reports the NdCa exchanger obtained from synaptic membrane preparations was found to be smaller (38) than reported here. This may be due to p r~t eolysis.
Furthermore, for the first time, we provided evidence that suggests a mechanism by which the mRNA for the specific NdCa exchanger isoforms is generated. The isolation of rabbit genomic clones shows that at least six distinct exons are encoding the variable part of NdCa exchanger message and these exons, some of them as short as 15 bp, are separated by long intronic sequences. Sequence analysis of the intron-exons boundaries reveals the unusual structure of these exons encoding the carboxyl end of the NdCa exchanger intracellular loop.
One pair (exons A and B) appear to be probably mutually exclusive (24). By this we mean that in order to maintain the open reading frame of the coding polypeptide, one member of the pair must always be spliced into the mRNA but that both cannot be spliced together nor can both be skipped. Structurally the polypeptide encoded by both exons is similar with several acid and basic amino acids in homologous positions. The exon B contains an extra arginine which confers a more basic character to this polypeptide.
Even more interesting is the genomic structure of the other 4 exons located downstream from exon A and exon B. The amino acid sequences coded by these small exons are in frame with exon A and B and also in frame with each other. Indeed, these exons could be classified as cassette type exons (24), as individual exons can be included or excluded independently while still maintaining the open reading frame. This intron/ exon arrangement could theoretically allow the generation of up to 32 NdCa exchanger isoforms. In our study in rabbit we isolated five of the six distinct isoforms with distinct distribution among different tissues. The isoform isolated from rat hippocampal library (22) that was not seen in this study can be also explained by the exon-intron arrangement presented here.
This complex genomic organization of the NdCa exchanger gene is strikingly similar to the exon-intron arrangement of troponin-T gene where the hypervariability within the aminoterminal region of troponin-T region is based on the presence of five cassette exons in conjunction with two mutually exclusive exons (39). In this case it has been shown that differences in the splicing pattern in this region of the troponin-T gene have specific effects upon the interaction of troponin with tropomyosin (40).
What is the likely function of these tissue-specific splicing variants in NdCa exchanger intracellular loop? Recently, Matsuoka et al. (33) investigated the role of the cardiac sarcolemmal NdCa exchanger intracellular loop by deletion mutagenesis and giant patch analysis. Mutants with a deletion in almost the whole intracellular loop (amino acids 240-679) still exhibited exchange activity but without the characteristic regulation by intracellular Ca2+ or inhibition by the exchanger inhibitory peptide ("?UP") observed in the wild type exchanger (41). Smaller deletions indicated that the important domain involved in intracellular calcium regulation and exchanger inhibitory peptide interaction was actually between amino acids 562-685. The splicing variants are located in the region corresponding to amino acids 561-645. Hence, the isoform diversity observed in this current study could well be concerned with regulatory aspects of the exchanger function, namely by intracellular calcium and exchanger inhibitory peptide. The in vitro expression of the different brain, kidney, and heart isoforms will allow a direct test of the intriguing possibility that alternative splicing of the NdCa exchanger message generates distinct isoforms with distinct intracellular calcium modulation.
In summary we have presented evidence for a unexpected degree of heterogeneity of the NdCa exchanger message. Several isoforms, differing in the carboxyl end of the intracellular loop, are generated by alternatively splicing of common exons in a tissue-specific manner. The unusual intron-exon arrangement of the NdCa exchanger gene encoding this area of the intracellular loop could potentially lead to as many as 32 isoforms.