Cloning of a third mammalian Na+-Ca2+ exchanger, NCX3.

NCX3 is the third isoform of a mammalian Na+-Ca2+ exchanger to be cloned. NCX3 was identified from rat brain cDNA by polymerase chain reaction (PCR) using degenerate primers derived from the sequences of two conserved regions of NCX1 and NCX2. The NCX3 PCR product was used to isolate two overlapping clones totalling 4.8 kilobases (kb) from a rat brain cDNA library. The overlapping clones were sequenced and joined at a unique Bsp106I restriction enzyme site to form a full-length cDNA clone. The NCX3 cDNA clone has an open reading frame of 2.8 kb encoding a protein of 927 amino acids. At the amino acid level, NCX3 shares 73% identity with NCX1 and 75% identity with NCX2 and is predicted to share the same membrane topology as NCX1 and NCX2. Following addition of a poly(A)+ tail to the NCX3 clone, exchanger activity could be expressed in Xenopus oocytes. NCX3 was also expressed in the mammalian BHK cell line. NCX3 transcripts are 6 kb in size and are highly restricted to brain and skeletal muscle. Linkage analysis in the mouse indicated that the NCX family of genes is dispersed, since the NCX1, NCX2, and NCX3 genes mapped to mouse chromosomes 17, 7, and 12, respectively.

Several different isoforms of the Na ϩ -Ca 2ϩ exchanger (NCX) 1 have been cloned. Two mammalian isoforms, NCX1 (Nicoll et al., 1990) and NCX2 (Li et al., 1994), represent the products of distinct genes. NCX1 is predominantly expressed in heart, where it plays a major role in excitation-contraction coupling, but is also abundant in a variety of other tissues. Extensive alternative splicing generates tissue-specific variants of NCX1 (Gabellini et al., 1995;Kofuji et al., 1994;Lee et al., 1994). In contrast, expression of NCX2 seems to be restricted to brain and skeletal muscle (Li et al., 1994). Two nonmammalian NCX isoforms have also been cloned: one from Drosophila (Valdivia et al., 1995) and another from squid (He et al., 1996).
The NCX1 protein has been studied in considerable detail. NCX1 contains 11 putative transmembrane segments (Nicoll et al., 1990) preceded by a cleaved signal peptide (Durkin et al., 1991;Hryshko et al., 1993). There is a long, intracellular loop (sometimes called loop f) between transmembrane segments 5 and 6. NCX1 is inhibited by a peptide called XIP (Li et al., 1991). The amino acid sequence of XIP corresponds to a 20amino acid segment of NCX1 in loop f called the endogenous XIP region. NCX1 has two internal repeats called the ␣and ␤-repeats. 2 The ␣-1 and ␣-2 motifs are in transmembrane segments 2-3 and 8 -9, respectively. The ␣-repeats are involved in the ion binding/translocation reactions of the exchanger . The ␤-1 and ␤-2 motifs are in loop f. NCX1 is regulated by intracellular Ca 2ϩ at a site which is distinct from the Ca 2ϩ transport site (Levitsky et al., 1994;Matsuoka et al., 1993Matsuoka et al., , 1995. The binding site for regulatory Ca 2ϩ has been determined to be in loop f and is comprised of ␤-1 and residues between ␤-1 and ␤-2. Two groups of three acidic amino acid residues in the Ca 2ϩ -binding domain appear to be directly involved in ion binding (Levitsky et al., 1994;Matsuoka et al., 1995).
Additional mammalian NCX-type isoforms were sought by PCR using degenerate oligonucleotide primers derived from two highly conserved regions of NCX1 and NCX2. A unique exchanger clone from a rat brain cDNA library was isolated. This clone is designated as NCX3 and has been sequenced and expressed. We examined the tissue distribution of NCX transcripts and determined the chromosomal localization of the NCX genes. An analysis of the three mammalian NCX protein sequences is also presented.

EXPERIMENTAL PROCEDURES
Cloning NCX3-PCR with degenerate primers was used to identify a fragment of NCX3. The primers were derived from the amino acid sequences of conserved regions of NCX1 and NCX2. The forward primer was a 17-mer with 4-fold degeneracy (5Ј-GCIGCITTYAAYATGTT-3Ј) based on a portion of proposed transmembrane segment 3. The reverse primer was a 19-mer with 16-fold degeneracy (5Ј-GGTAIACRTAYT-TRTARAA-3Ј) based on a portion of the endogenous XIP region (Li et al., 1991). cDNA from rat brain stem was used as template for the PCR reaction. The PCR reaction was carried out for 35 cycles (94°C, 30 s; 42°C, 60 s; 72°C, 120 s), and the PCR product was cloned into the vector pCRII (Invitrogen). After transformation, four plasmids with appropriately sized inserts were sequenced. Three inserts had the sequences of NCX1 or NCX2, and one insert had the previously undescribed sequence of NCX3. In initial experiments, other primer pairs only amplified either NCX1 or NCX2.
Probe synthesized from the cloned PCR product was used to screen a rat brain stem ZAPII cDNA library (Stratagene). After screening about 1.8 ϫ 10 5 plaques, two overlapping clones, p15 and p20 (Fig. 1), were isolated. The full-length clone, pIII, was constructed from p15 and p20 as described under "Results." Expressing NCX3 Activity-Oocytes were prepared, and exchanger activity was measured as described previously (Longoni et al., 1988). The plasmid vector containing NCX3 was linearized with the restriction enzyme EcoRV, and synthesis of cRNA was performed with the T7 mMessage mMachine (Ambion). Unincorporated nucleotides were removed with Chromaspin-100 columns (Clontech). cRNA or water (46 nl) was injected into each oocyte. Twelve days after injection, the oocytes were loaded with Na ϩ in the presence of nystatin, washed, and then incubated in medium containing 45 Ca 2ϩ and either Na ϩ (to measure nonspecific Ca 2ϩ uptake) or K ϩ (to measure Na ϩ -gradient dependent Ca 2ϩ uptake). All solutions were as described previously (Longoni et al., 1988).
To express NCX3 in BHK cells, a 3640-bp NheI/BamHI fragment of NCX3 was subcloned into the SmaI site of the transfer vector pNUT (Palmiter et al., 1987). In this vector, NCX3 is under the control of the metallothionein I promoter. The resultant recombinant plasmid was transfected into BHK cells by the calcium phosphate method (Gorman et al., 1982). Twenty-four h after transfection, methotrexate (2.5 mM) was added to the culture medium and cells were incubated with this selection medium for 10 days. Methotrexate-resistant colonies were picked and screened for exchanger activity by 45 Ca 2ϩ -uptake assay.
The 45 Ca 2ϩ -uptake assay with transfected BHK cells was performed with a protocol similar to those previously described (Condrescu et al., 1995;Li et al., 1992) with modifications. Briefly, cells were grown in 24-well plates to 80 -90% confluence. Zinc sulfate (60 M) was added to the medium 16 h prior to the uptake assay. To load the cells with Na ϩ , cells were incubated with loading buffer containing 140 mM NaCl, 1 mM MgCl 2 , 10 mM MOPS/Tris, pH 7.4, 0.4 mM ouabain, and 25 M nystatin for 20 min at room temperature. Cells were washed three times with washing buffer (140 mM NaCl, 10 mM MOPS/Tris, pH 7.4) and overlaid with 0.5 ml of assay solution (140 mM KCl or NaCl, 10 mM MOPS/Tris pH 7.4, 25 M CaCl 2 , 0.4 mM ouabain, and 10 Ci/ml 45 CaCl 2 ). After the desired time interval, the assay solution was removed by aspiration and 1 ml of cold quenching buffer (10 mM MOPS/Tris, pH 7.4, 100 mM MgCl 2 , 10 mM LaCl 3 ) was added. Cells were washed 3 additional times with quenching buffer and dissolved in 1 ml of NaOH (1 M). Aliquots were subjected to scintillation counting and protein assay. Data are presented with 45 Ca 2ϩ uptake in Na ϩ -containing medium subtracted from 45 Ca 2ϩ uptake in K ϩ -containing medium.
Chromosomal Mapping-Linkage analysis of the NCX1, NCX2, and NCX3 genes was performed using a set of 67 progeny from a (C57BL/ 6J ϫ Mus spretus)F1 ϫ C57BL/6J backcross (Warden et al., 1993). Restriction fragment length variants were identified using Southern analysis following digestion of genomic DNA with various restriction enzymes and hybridization with specific probes corresponding to fragments of NCX1, NCX2, and NCX3 cDNAs. The NCX1 probe consisted of bases 1823-2198 (Nicoll et al., 1990), the NCX2 probe consisted of bases 1455-1907(Li et al., 1994, and the NCX3 probe consisted of bases 1789 -2851 (Fig. 2). No cross-hybridization between NCX genes occurred as judged by the failure to observe common bands following Southern analysis. Linkage was analyzed with the Mapmanager v2.6 program (Manly, 1993).
Northern Blot Analysis-Total RNA was prepared by the method of Chomczynski and Sacchi (1987), and poly (A) ϩ RNA was isolated using the Poly(A)Tract mRNA Isolation System III (Promega, Madison, WI). To increase yield, total RNA was applied twice to the magnetic beads. Poly(A) ϩ RNA (2.5 g) from each tissue was size-fractionated on a 1% (w/v) agarose gel containing 1 ϫ MOPS buffer (20 mM MOPS, 5 mM sodium acetate, 1 mM EDTA, pH 7.0) and 6% formaldehyde. After electrophoresis, the gel was washed 3 ϫ 5 min in diethyl pyrocarbonatetreated H 2 O. RNA was transferred to Hybond-N membrane (Amersham) by capillary diffusion in 20 ϫ SSC (Sambrook et al., 1989) and fixed by UV cross-linking. 32 P-Labeled, single-stranded DNA probes (see below) were hybridized to the membrane overnight at 42°C as described (Li et al., 1994). The blot was washed for 15 min at 42°C with 2 ϫ SSC in 0.1% (w/v) SDS, 2.1 mM EDTA, 11.2 mM tetrasodium pyrophosphate followed by washing steps with increasing stringency up to 0.2 ϫ SSC at 60°C for 10 min. NCX probes were stripped from the membrane by incubation in 0.1 ϫ SSC, 0.2% (w/v) SDS at 95°C for 30 min.

RESULTS
Cloning the NCX3 Isoform of the Na ϩ -Ca 2ϩ Exchanger-We searched for previously undescribed exchanger cDNAs by PCR amplification of rat brain cDNA using degenerate oligonucleotide primers to two highly conserved regions of the exchanger. The primers were designed to hybridize to the nucleotide sequences of transmembrane segment 3 in the ␣-1 motif and to the endogenous XIP motif (Li et al., 1991). A PCR product of the appropriate size (ϳ350 bp) was detected, subcloned into the pCRII vector, and sequenced. The sequence of one clone, which we designate NCX3, was previously undescribed but showed similarity to both NCX1 and NCX2. To obtain the full coding region of NCX3, the NCX3 PCR product was used to probe a rat brain cDNA library. Two overlapping clones, p15 and p20, were isolated. A full-length cDNA clone was constructed by ligating the 1.5-kb SacI-Bsp106I fragment from p20 into the HincII-Bsp106I site of p15 (Fig. 1). The resulting clone, pIII, contained the initiating methionine, the entire coding region, and the 3Ј-UTR from p15 but eliminated 750 bp from the 5Ј-UTR of p20.
NCX3 cDNA Sequence-The combined sequence of clones p20 and p15 (Fig. 2) is 4855 bases long. Both the 5Ј-and 3Ј-UTRs are extensive, with 833 nucleotides in the 5Ј-UTR and 1236 nucleotides in the 3Ј-UTR. The NCX3 cDNA clone, like those reported for NCX1 and NCX2, does not contain a poly-FIG. 1. Construction of NCX3 clones. Overlapping clones p20 and p15 were isolated from a rat brain cDNA library. Clone pIII consists of the SacI-Bsp106I fragment of p20 ligated to the Bsp106I-3Ј-end of p15. pIII/MC has the 3Ј-UTR, including the poly(A) ϩ tail, from the Na ϩglucose cotransporter (Hediger et al., 1987) ligated into the BglII site of pIII. The asterisk indicates the position of the stop codon. adenylation site or poly(A) ϩ tail. However, the NCX3 mRNA is approximately 1000 bases longer than the cDNA clone (see below) and may contain a poly(A) ϩ tail that is not included in the clone.
The NCX3 exchanger clone contains an open reading frame of 2784 nucleotides, starting at base 834 and ending at base 3617, which encodes for a protein of 927 amino acids. The sequence around the initiating methionine conforms well to the consensus sequence described by Kozak (1994) with six identities in 13 bases.
The 5Ј-UTR of NCX3 contains two distinguishing characteristics. First, there are 17 ATG triplets in the region upstream of the initiating methionine (Fig. 3B). None of these is likely to serve as an initiation codon for NCX3, as there are stop codons in each reading frame between the ATG at position 834 and the next nearest upstream ATG at position 672. Also, the longest potential coding region is only 111 bp.
The second characteristic of the 5Ј-UTR is an extremely GC-rich region (Fig. 3A). Between bases 520 and 720, the %GC content averages nearly 80%.
According to the ribosome scanning model of translation initiation, the first ATG in a mRNA is almost always the translation initiation site. The further downstream an ATG, the less likely it is to serve as an initiation site (Kozak, 1994). Also, GC-rich regions are capable of forming secondary structures which could impede the progress of the ribosome. Hence, the 5Ј-UTR of NCX3 is not conducive for translation efficiency. This suggests that translation of NCX3 is a highly regulated event.
We further investigated the 5Ј-UTR of NCX3 by 5Ј-RACE analysis. cDNA synthesis was followed by two rounds of PCR amplification using a nested pair of primers (Fig. 3B). A single RACE product of about 220 bp was observed (Fig. 3C). The PCR product was subcloned into a plasmid vector, and three individual plasmids were sequenced. The 5Ј-ends were all within three bases of nucleotide 734 (Fig. 3B). These results suggest that, in brain tissue, a single 5Ј-UTR exists for NCX3. The 5Ј-RACE products ended just downstream from the site where GC content rises. This may be indicative of strong secondary structure, as mentioned above, although further analysis of the 5Ј-UTR of NCX3 is necessary.
NCX3 Protein Sequence-The translation of the NCX3 open reading frame is shown in Fig. 4. As predicted for NCX1 and NCX2, NCX3 has 12 potential ␣-helical transmembrane segments. The first putative transmembrane segment is designated as a signal peptide by analogy to NCX1, which has been shown to have a cleaved signal peptide sequence (Durkin et al., 1991;Hryshko et al., 1993;Nicoll et al., 1990). A potential signal peptidase recognition site, AEA (Perlman and Halvorson, 1983), is located at amino acid 30. If this site is used in vivo, then cleavage could occur following the second alanine. Thus, the amino terminus is modelled to be extracellular, the carboxyl terminus to be intracellular, and there is a long cytoplasmic loop between putative transmembrane segments 5 and 6.
NCX3 contains 5 potential asparagine-linked glycosylation sites. Two (Asn-45, Asn-67) are in the amino-terminal, putative extracellular loop. Two more (Asn-130, Asn-135) are in the putative intracellular loop between transmembrane segments 1 and 2, and the fifth (Asn-823) is in the putative extracellular loop between transmembrane segments 8 and 9. By analogy to NCX1, where it has been demonstrated that only the first from clones p20 and p15, respectively, used to generate pIII and pIII/MC are in bold. These sequences have been submitted to Gen-Bank TM under accession number U53420.
FIG. 2. Complete nucleotide and predicted amino acid sequences of NCX3. The SacI (GAGCTC) and BglII (AGATCT) sites potential N-linked glycosylation site is used (Hryshko et al., 1993), it is predicted that only the first site in NCX3 is glycosylated.
NCX3 contains four consensus phosphorylation sites. Thr-113 is contained within consensus sites for either Ca 2ϩ /calmodulindependent kinase or cAMP-dependent kinase. Thr-267 is within a consensus site for either cAMP-dependent kinase or protein kinase C. Thr-597 is in a consensus site for protein kinase C, and Tyr-608 is in a potential site for tyrosine kinase.
The region in NCX1 which is involved in binding regulatory Ca 2ϩ is fairly well conserved in NCX3, especially in the vicinity of the two groups of three acidic amino acids which appear to be directly involved in Ca 2ϩ binding. This suggests that NCX3 is also regulated by intracellular Ca 2ϩ .
Schwarz and Benzer 2 have observed that the exchanger proteins have two repeated motifs designated the ␣and ␤-repeats. The ␣-repeats are in the predicted transmembrane segments, and mutational analysis of NCX1  has demonstrated that the ␣-repeats are crucial for ion transport. The ␤-repeats are located in the long intracellular loop between putative transmembrane segments 5 and 6. Using a Pustell protein matrix analysis, we determined that NCX3 also contains the ␣and ␤-repeats as shown in Fig. 5.
Expression of the NCX3 Exchanger-Initial attempts to ex-press NCX3 in oocytes were unsuccessful. We had previously noted that NCX1 expression in oocytes was improved when the injected cRNA contained a poly(A) ϩ tail (Matsuoka et al., 1993). Therefore, we subcloned a poly(A) ϩ -containing fragment of DNA into the BglII-NotI site of pIII (Fig. 1). This replaced most of the original NCX3 3Ј-UTR with the 3Ј-UTR derived from the SGLT1 (Na ϩ -glucose cotransporter) clone (Hediger et al., 1987) as described previously (Li et al., 1994). The resultant clone, pIII/MC, catalyzes the Na ϩ gradient-dependent uptake of 45 Ca 2ϩ into oocytes (Fig. 6A), albeit at a reduced level compared to 45 Ca 2ϩ uptake into oocytes expressing NCX1. The maximum level expressed by NCX3 is about 20% of the level seen for NCX1 expression in oocytes. It also takes much longer for the oocyte to begin expressing NCX3. We normally observe peak levels of NCX1 expression at 3-5 days after cRNA injection. NCX3 expression peaks at 7 or more days post-injection. As further confirmation that NCX3 expresses exchange activity, we also expressed the NCX3 protein in BHK cells. Cells stably transformed with vector containing the NCX3 coding region expressed 12-fold more Na ϩ gradient-dependent Ca 2ϩ uptake compared to cells transformed with vector alone (Fig. 6B).
Tissue Specificity of NCX3-The expression of NCX1, NCX2, and NCX3 was investigated by Northern blot analysis to directly compare transcript sizes and tissue specificity. A Northern blot with poly(A) ϩ RNA from adult rat brain, heart, lung, kidney, spleen, skeletal muscle, stomach, duodenum, jejunum, ileum, liver, and pancreas was sequentially probed for NCX2, FIG. 3. Analysis of the 5-UTR of NCX3. A, GC content of the NCX3 cDNA. The %GC is plotted versus nucleotide sequence. The window size is 100 nucleotides. B, nucleotides in the 5Ј-UTR of NCX3 are in lowercase except the ATG triplets which are in bold, capital letters and underlined. The nucleotides of the coding region are in capital letters with an asterisk to indicate the initiating ATG codon. The 5Ј-RACE end point is shown with an arrow. Primers used for first strand cDNA synthesis and first and second round PCR reactions are boxed and highlighted. Beginning at the 5Ј-end, the boxes correspond to primers GSP-2, GSP-1, and GSP-RT. C, the 220-bp DNA product produced by the 5Ј-RACE reaction.

FIG. 4. Predicted amino acid sequence of NCX3.
A potential signal peptide (SigPep) is underlined, transmembrane segments are double underlined and numbered TMS1-11. Potential N-linked glycosylation sites (CH 2 O), phosphorylation sites, and a potential signal peptidase site (SigPase) are in bold. The potential signal peptidase cleavage site is indicated with an arrow. A potential Ca 2ϩ -regulatory site is highlighted with the conserved acidic residue triplets in bold. CaCamK, calcium/calmodulin-dependent kinase; cAMPK, cAMPdependent protein kinase; PKC, protein kinase C; TyrK, tyrosine kinase. *, site which is conserved in NCX1 and/or NCX2. NCX1, and then NCX3 expression. Each NCX-specific DNA probe was derived from the large intracellular loop of the protein, where the NCX clones differ most.
For NCX1, a 7-kb transcript was detected (Fig. 7). The strongest hybridization signal was observed in heart followed by spleen, with weaker signals in all other tissues except liver and pancreas. After a prolonged exposure time (3 days; not shown), a very weak hybridization signal could be detected in liver but not in pancreas. The transcript size and the tissue distribution of the NCX1 mRNA is in good agreement with previously made observations using multi-tissue Northern blots with RNA from different species (Kofuji et al., 1992;Komuro et al., 1992;Lee et al., 1994). However, the relatively high level of NCX1 transcript in spleen has not previously been noted.
When the Northern blot was hybridized with a probe specific for NCX2, signals were seen in two tissues only. In brain, a transcript of 5 kb could be detected and, in skeletal muscle, a much smaller transcript of about 1.4 kb hybridized strongly with the probe (Fig. 7). After overexposure of the blot, weak signals at 5 kb in duodenum, ileum, and jejunum were also observed. An NCX2-specific hybridization signal of 5 kb in rat brain and skeletal muscle has been described previously (Li et al., 1994). However, no 1.4-kb transcript in skeletal muscle was detected, and hybridization to any tissues other than brain and skeletal muscle was also not reported. The 1.4-kb hybridization signal is not likely to be due to a nonspecific degradation of the mRNA since apparently undegraded 7-and 6-kb transcripts were observed when the blot was probed for NCX1 and NCX3 transcripts, respectively.
After the blot was rehybridized with an NCX3-specific DNA probe, a 6-kb transcript could be detected only in brain and skeletal muscle (Fig. 7). No hybridization signals were seen in the other tissues even after a prolonged exposure. Chromosomal Localization of NCX Genes-The NCX genes were mapped in mouse by linkage analysis of RFLVs. A survey of restriction enzymes using Southern hybridization revealed informative RFLVs for each of the three genes (Table I). In each case, (C57BL/6J ϫ M. spretus)F1 hybrid animals exhibited all of the parental bands. The RFLVs were then scored in backcross progeny and compared with the segregation pattern of over 300 other markers spanning the mouse genome which were typed in this cross (Warden et al., 1993). Linkage analysis revealed significant linkage for each of the three NCX genes (Fig. 8). The NCX1 gene was located in the distal region of chromosome 17, 8.2 cM distal (with respect to the centromere) to marker D17Mit41 and 4.8 cM proximal to D17Ucla2. NCX2 was located in the proximal region of mouse chromosome 7, tightly linked to the gene for apolipoprotein E (Apoe) (zero recombinations out of 66 meioses tested). NCX3 was located near the middle of mouse chromosome 12, tightly linked to the genetic marker D12Ucla3 (zero recombinations out of 65 meioses tested). Each of the above linkages was highly significant, as all lod scores exceeded 10 (a lod score is a statistical measure of whether a gene is linked to a particular genetic marker. A lod score of Ն3 is generally considered to be strong evidence of linkage). No other significant linkages were observed. We conclude that the NCX genes are part of a dispersed gene family.

DISCUSSION
Activity of NCX Proteins-We have described here the cloning and sequencing of a third, distinct, Na ϩ -Ca 2ϩ exchanger, FIG. 5. Pustell matrix plot of NCX3. The NCX3 amino acid sequence was compared against itself with a window size of 20 residues and minimal score of 30%. The ␣and ␤-repeats are circled.
FIG. 6. Expression of NCX3 activity. A, Xenopus oocytes were injected with water or NCX3 cRNA. Twelve days later, 45 Ca 2ϩ uptake was measured in Na ϩ -containing medium to determine the extent of nonspecific 45 Ca 2ϩ uptake or in K ϩ -containing medium to measure the extent of Na ϩ gradient-dependent 45 Ca 2ϩ uptake. B, BHK cells containing the pNUT vector with or without an NCX3 coding sequence were assayed for Na ϩ -dependent uptake of 45 Ca 2ϩ . The approximate sizes of hybridizing fragments are shown using DNA isolated from strain C57BL/6J or M. spretus mice. Segregation of the RFLVs was scored in backcross mice using the underlined fragment.
b Shieh et al., 1992. c From homology to mouse chromosome location.

FIG. 7. Tissue distributions of NCX transcripts.
A Northern blot, with each lane containing 2.5 g of poly(A) ϩ RNA from the indicated adult rat tissues, was sequentially probed with NCX1, NCX2, and NCX3 specific probes. Autoradiographic film was exposed 5-13 h. The weak bands migrating just above the 1.4-kb transcript in the middle panel are due to nonspecific hybridization to 18 S rRNA.
NCX3. The NCX3 cDNA, like those for NCX1 and NCX2, has been expressed in Xenopus oocytes and in BHK cells. A Na ϩ gradient-dependent 45 Ca 2ϩ uptake can be measured, thus verifying that NCX3 encodes a Na ϩ -Ca 2ϩ exchanger. With the relatively low levels of activity expressed by NCX3, complete characterization has not yet been possible. It will be of much interest to determine if NCX3 shares the same regulatory properties which have been described for NCX1 and NCX2.
Sequence Comparison of the NCX Proteins-Overall, the NCXs share a considerable amount of sequence identity. When the putative leader peptides (see below) and potential alternative splice sites are discounted, the percent identity is 68% between NCX1 and NCX2, 73% between NCX1 and NCX3, and 75% between NCX2 and NCX3.
Each of the exchangers can be modelled to contain 12 hydrophobic segments. The first hydrophobic segments in each of NCX2 and NCX3 are likely to be cleaved leader peptides as has been demonstrated for NCX1 (Durkin et al., 1991;Hryshko et al., 1993). The amino acid sequences in this region are highly divergent (Fig. 9). Each of the remaining 11 hydrophobic segments is modelled to be a transmembrane helix. A long intra-cellular loop is located between putative transmembrane segment 5 and 6.
Within the transmembrane segments, the identity between the exchangers exceeds 75% except for transmembrane segment 11. The most highly conserved transmembrane segments are 2, 6, and 8. Transmembrane segments 2 and 8 are part of the ␣-repeats and have been shown to be functionally important in NCX1. The last putative transmembrane segment, 11, is the least conserved with only about 60% identity between the exchangers. This suggests that putative transmembrane segment 11 is not involved in ion transport. Perhaps transmembrane segment 11 plays a structural role in the exchanger.
In NCX1, the intracellular loop has been shown to serve a regulatory role (Levitsky et al., 1994;Matsuoka et al., 1993Matsuoka et al., , 1995. The loop is involved in inactivation of the exchanger and in binding regulatory, but not transported, Ca 2ϩ . The intracellular loops of the exchangers contain several interesting conserved motifs. The first area of sequence conservation is in the aminoterminal end of the loop, in what is designated the endogenous XIP (Fig. 9). The endogenous XIP region may be involved in FIG. 8. Mapping of NCX1, NCX2, and NCX3 to mouse chromosomes. NCX genes were mapped to chromosomes 17 (NCX1), 7 (NCX2), and 12 (NCX3) in an interspecific backcross. Each chromosome is drawn to scale with the centromere at the top and the distance of the most distal marker from the centromere indicated at the bottom (cum, in cM). The ratios of the number of recombinants to the total number of informative mice and the recombination frequencies Ϯ S.E. (in cM), for each pair of loci, are indicated to the left of the chromosome. For pairs of loci that cosegregate, the upper 95% confidence interval is shown in parentheses. Loci are linked with lod scores greater than 4.8. Ucla markers were reported (Warden et al., 1993) or are unpublished data. Ppard was reported in Cohen et al. (1996) and Tcfcoup2 in Welch et al. (1996). Lgals1 and Lgals2 are unpublished. Further information can be obtained from the authors. References for other linked loci can be obtained from the Mouse Genome Data Base (1995). The human gene for NCX1 has been designated SLC8A1 for solute carrier family 8 (sodium calcium exchanger), member 1 (Human Genome Data Base (1990)). To be consistent with the human nomenclature, we used a similar designation for the mouse NCX genes. Therefore, the mouse NCX1, NCX2, and NCX3 genes are designated Slc8a1, Slc8a2, and Slc8a3, respectively. These data have been entered into the Mouse Genome Data base, accession numbers MGD-CREX-572, MGD-CREX-573, and MGD-CREX-574 for NCX1, NCX2, and NCX3, respectively. regulation of the exchanger. It consists of 20 amino acids, and application of a synthetic peptide with the same sequence to the intracellular surface of the exchanger results in inhibition of exchanger activity (Li et al., 1991). Only the first 14 residues of the endogenous XIP are highly conserved. This correlates well with recent mutagenesis studies. Mutation 3 of some conserved basic and aromatic residues in the endogenous XIP region alter the Na ϩ -dependent inactivation (Hilgemann et al., 1992) and rate of Ca 2ϩ regulation of the exchanger. Also, truncation of the amino-terminal residues of the peptide XIP has relatively little effect on inhibitory potency, but substitution of some of the conserved residues reduces the inhibitory effect of the peptide. 4 The second area of sequence conservation in the cytoplasmic loop is at amino acids 354 -382 of NCX1 (Fig. 9). In this stretch of 29 residues, only four amino acids vary, and the amino acid substitutions are chemically similar to the residues in NCX1. No functional role has yet been attributed to this segment.
A third region of conservation is in the ␤-repeats (Fig. 9). Each repeat consists of a stretch of 76 amino acid residues. The ␤-1, ␤-2 repeats of NCX1 are 33% identical to each other. Those of NCX2 and NCX3 are 36% and 32% identical, respectively. The regions between the ␤-repeats are highly variable in both amino acid sequence and length.
Because the ␤-repeats are found in the long intracellular loop, it is likely that they are involved in exchanger regulation. Indeed, the Ca 2ϩ -regulatory site of NCX1 (Levitsky et al., 1994;Matsuoka et al., 1995) contains all of ␤-1 and the stretch of amino acid residues between ␤-1 and ␤-2 (Figs. 4 and 9). Binding regulatory Ca 2ϩ requires the presence of two triads of acidic residues. One of the triads is at the carboxyl end of ␤-1 and the other is just before ␤-2.
Thus, ␤-1 is involved in Ca 2ϩ regulation, and the extent of similarity between ␤-1 and ␤-2 suggests that ␤-2 is also involved in regulation. However, the Ca 2ϩ binding motif of ␤-1 is not fully conserved in ␤-2. Perhaps ␤-2 plays a different regulatory role, for example, via phosphorylation (see below).
Another interesting area in the long intracellular loop is in the variable region of NCX1 where alternative splicing occurs (Kofuji et al., 1994;Lee et al., 1994). In this area, both NCX2 and NCX3 have a deletion of 37 amino acids relative to NCX1.
The fourth conserved area in the cytoplasmic loop is at the carboxyl end of the loop, from residue 708 of NCX1 to the 6th transmembrane segment. Like the Ca 2ϩ -binding domain, this region also contains two groups of acidic residues; one spans residues 710 -718 including four glutamate residues, and one starting at residue 756, where there is a string of at least six consecutive acidic residues. No role has yet been attributed to this conserved region of the exchangers.
Consensus Phosphorylation Sites in the Exchangers-In at least one neural tissue (squid axon (DiPolo and Beauge, 1994)) and in smooth muscle (Iwamoto et al., 1995), there is evidence of regulation of the exchanger by phosphorylation. Also, NCX2 can be inhibited by treatment with a tyrosine kinase inhibitor when expressed in CHO cells (Condrescu et al., 1996).
Each of the exchangers contains several consensus sites for phosphorylation by different protein kinases. The location of some of the phosphorylation consensus sites make them potentially interesting. In NCX3, the residue Thr-113 is located in the intracellular loop connecting transmembrane segments 1 and 2. Transmembrane segment 2 is a part of the ␣-1 repeat and has been shown to be involved in ion transport . Phosphorylation of residue Thr-113 could play a role in modulating ion transport.
Residue Thr-262 in NCX2 could be a substrate for cAMP-dependent kinase, and the analogous residue in NCX3, Thr-267, could be a substrate for either cAMP-dependent kinase or protein kinase C. Both of these residues are located in the endogenous XIP regions. NCX1 does not have a phosphorylation site in this region. The role of the endogenous XIP in regulating the exchanger and the potential for variable types of phosphorylation in this region suggest ways to modify the activity of the different exchangers.
Finally, each of the exchangers contains a potential tyrosine kinase site at an analogous position, Tyr-581 in NCX1, Tyr-601 in NCX2, and Tyr-608 in NCX3. This site is located at a position in the ␤-2 repeat which corresponds to one of the Ca 2ϩbinding acidic triads of ␤-1. Perhaps some type of interaction between Ca 2ϩ and phosphorylation plays a role in modulating FIG. 9. Comparison of NCX1, NCX2, and NCX3 amino acid sequences. The complete amino acid sequence of NCX1 is shown. Identities between NCX1 and NCX2 or NCX3 are indicated with dots and alignment gaps with dashes. Potential transmembrane segments are underlined. The ␣and ␤-repeats and endogenous XIP regions are highlighted. Amino acid numbers are those of NCX1 with the initiating methionine of NCX1 being number 1. one or more of the exchangers.
Tissue Distribution of NCX Gene Products-While NCX transcripts have been detected in nearly every tissue, levels of NCX mRNA are highest in the excitable tissues of brain, heart, and skeletal muscle. Cardiac membranes express the highest levels of NCX transport activity, and the role of Na ϩ -Ca 2ϩ exchange in excitation-contraction coupling has been extensively documented. However, only one NCX gene product, NCX1, is expressed in this tissue.
Brain tissue preparations have the next highest levels of exchanger activity. Significant transcript levels for all three NCX isoforms can be detected in brain. A photoreceptor-like exchanger may also be expressed in brain (Dahan et al., 1991;Lytton et al., 1996). The photoreceptor exchanger catalyzes the exchange of 4 Na ϩ for 1 Ca 2ϩ plus 1 K ϩ . The distribution of the different exchangers in various brain structures remains to be determined. It is likely that the Na ϩ -Ca 2ϩ exchanger has an important role in Ca 2ϩ extrusion following transmitter release at nerve endings (Reuter and Porzig, 1995).
Skeletal muscle has high levels of NCX3 and NCX2 transcripts. The high levels of exchanger transcripts in this tissue are surprising since exchanger transport activities are low in this tissue (Gilbert and Meissner, 1982), and the exchanger does not have an important role in skeletal muscle. In cardiac muscle, the primary role of the exchanger is in extrusion of Ca 2ϩ following excitation. Skeletal muscle does not rely on trans-sarcolemmal Ca 2ϩ influx during excitation, and high levels of a Ca 2ϩ -extrusion mechanism, such as the exchanger, would not be required and have not been documented.
The presence of the 1.4-kb NCX2 transcript in skeletal muscle is puzzling. In previous work, a single transcript at 5 kb was detected in this tissue (Li et al., 1994). In this work, only the 1.4-kb transcript was observed. The different results cannot be ascribed to different probes. In our initial work, a probe encompassing bases 1035-2326 of NCX2 was used. In this work, the probe was shorter and contained within the original probe (bases 1455-1983). Another laboratory 5 has seen similar results with Northern blot analyses of skeletal muscle RNA probed with NCX2. On one blot, a 5-kb transcript was observed, whereas on another blot, only the 1.4-kb transcript was present. The 1.4-kb transcript could be the result of a specific degradation of the 5-kb NCX2 transcript, the product of a different gene with a high degree of similarity to NCX2, or an alternatively spliced product of NCX2. Perhaps the protein product of the small NCX2 transcript has a function other than cation transport in skeletal muscle. There is some precedence for alternatively spliced products of NCX1 with significantly reduced size. Gabellini et al. (1995) have found that when NCX1 is transiently expressed in 293 cells, a truncated clone with a termination codon at base 1844 is produced.
Chromosomal Organization of NCX Genes-Although clearly homologous, the three NCX genes are dispersed in the mouse genome. We localized the mouse homolog of NCX1 to mouse chromosome 17. The NCX1 gene has previously been mapped to human chromosome 2p21-23 (Shieh et al., 1992). These localizations extend a previously observed conserved linkage group between the two species. The NCX2 gene is on mouse chromosome 7, tightly linked to the apolipoprotein E gene, which in humans resides at chromosome 19q13.2. NCX3 is located on mouse chromosome 12 at a position homologous to human chromosome 14q21-31.
We previously reported (Li et al., 1994) that NCX2 was localized to human chromosome 14. That localization was inadvertently performed with an NCX3 probe and not an NCX2 probe. Thus, the present results are consistent with the earlier finding after correction.
None of the NCX genes map near any mutations in mice which are likely to be relevant to the expression of these transporters (Mouse Genome Data Base (1995)). The corresponding human chromosomal regions do contain many known disease genes, such as for neurologic disorders (Human Genome Data Base (1990)), that may involve transporters.