Identification of a Novel Human Subfamily of Mitochondrial Carriers with Calcium-binding Domains*

Aralar1 and citrin were identified as calcium binding aspartate/glutamate carriers (AGC) in mitochondria. The presence of calcium binding motifs facing the extramitochondrial space allows the regulation of the transport activity of these carriers by cytosolic calcium and provides a new mechanism to transduce calcium signals in mitochondria without the requirement of calcium entry in the organelle. We now report the complete characterization of a second subfamily of human calcium binding mitochondrial carriers named SCaMC (short calcium-binding mitochondrial carriers). We have identified three SCaMC genes in the human genome. All code for highly conserved proteins (about 70–80% identity), of about 500 amino acids with a characteristic mitochondrial carrier domain at the C terminus, and an N-termi-nal extension harboring four EF-hand binding motifs with high similarity to calmodulin. All SCaMC proteins were found to be located exclusively in mitochondria, and their N-terminal extensions were dispensable for the correct mitochondrial targeting of the polypeptides. SCaMC-1 is the human orthologue of the rabbit Efinal protein, which was reported to be located in peroxisomes, and SCaMC-2 is the human orthologue of the rat of the yield from first strand cDNA synthesis used as template for sub- sequent PCR amplifications using Taq DNA polymerase (PerkinElmer and last cycle of elongation for 10 min. Blunt-ended fragments were purified and ligated into the PCR-cloning vector pST-Blue-1 (Novagen). The nucleotide sequences determined using ABI prism dye terminator cycle sequencing kit (PerkinElmer Life Sciences). fragment of pUC18-SCaMC-2 to produce pUC18-SCaMC-2b (accession number AJ619990) and pUC18-SCaMC-2c (accession number AJ619991). SCaMC-3— A partial human SCaMC-3 sequence (nucleotides from 1 to 1329) was obtained after sequencing human EST AI073878 (1435 bp, obtained from ATCC). SCaMC-3 3 (cid:2) -end sequence was amplified by RT-PCR from total RNA of HEK-293T cells. Reverse transcription was performed with oligo(dT) (Roche Applied Science) as primer. For PCR reactions a forward primer (5 (cid:2) -AGCAGTACAGCCACGACTCG-3 (cid:2) , corresponding to nucleotides 1202–1220 of AI073878) and a reverse primer (5 (cid:2) -GGGATCCTGTGGTTGGATCA-3 (cid:2) , derived from human EST BF920631 sequence) were used. The 394-bp product obtained was cloned and sequenced. Finally, after digestion at a KpnI site, located in the overlapping regions of AI073878 and the PCR product 3 (cid:2) -end, and flanking EcoRI sites, a full-length coding region was assembled into pBluescript-SK, thus obtaining a final cDNA of 1594 bp (accession number AJ619988).

Metabolite transport across the inner mitochondrial membrane is mediated by structurally related proteins belonging to the mitochondrial carrier (MC) 1 family, which exist exclusively in eukaryotes (for reviews see Refs. [1][2][3][4]. A large number of proteins with the characteristic features of the MC family have emerged from genome sequencing of various organisms (4,6,7). MC members are integral proteins of the mitochondrial inner membrane, and function in the shuttling of metabolites, nucleotides, and cofactors between the cytosol and mitochondria. These proteins have a molecular mass of about 35 kDa, contain six transmembrane spanning segments, and have a tripartite structure (2,8). Members of this family are responsible for the transport of ADP/ATP, phosphate, citrate, fumarate/succinate, carnitine/acylcarnitine, aspartate/glutamate, flavine adenine dinucleotide (FAD), and other substrates (9, 10, reviewed in Ref. 4). It has been reported that two MC proteins are located in peroxisomes (11), and one of them was shown to transport ATP in exchange for AMP (12,13), whereas other MC members are located in hydrogenosomes (14).
We have recently identified a subfamily of calcium-binding mitochondrial carriers (CaMCs) (15,16) with a bipartite structure: the N-terminal half harbors EF-hand calcium binding motifs, whereas the C-terminal half has the characteristic features of the MC family. The CaMC subfamily is made up of two groups of proteins that differ in their MC homology sequences and in the length of their polypeptides (16). The longer CaMCs are aralar1 and citrin, two isoforms of aspartate/glutamate carrier (AGC) important in the malate-aspartate NADH shuttle and the urea cycle (15)(16)(17)(18). Overexpression of AGC isoforms increases the supply of reducing equivalents to mitochondria and mitochondrial ATP production in cells stimulated with Ca 2ϩ -mobilizing agonists (18,19). The AGC isoforms are activated by calcium on the external face of the inner mitochondrial membrane, suggesting a novel mechanism whereby cytoplasmic Ca 2ϩ signals are decoded by mitochondria into activation of mitochondrial metabolism (18,19).
A second group of CaMCs, related to the product of yeast YNL083w gene, have shorter N-terminal regions, hence the name of SCaMCs (for short CaMC) used to design the members of this group (16,20,21). Only two proteins with similarity to the product of yeast YNL083w have been reported, the peroxisomal carrier Efinal in rabbit (11) and a similar protein identified in rat liver by mRNA differential display (22).
Our working hypothesis is that SCaMCs may be calciumregulated transporters in mitochondria important in calcium sensing by these organelles. The calcium uniporter is the main pathway whereby mitochondria decode cytosolic calcium signals into mitochondrial activation events, after allowing the entry of calcium in the organelle. In addition, there are other calcium sensors in mitochondria, such as that involved in facilitation of the calcium uniporter, that are responsible for its sustained activation (23), the activation of the permeability transition (24), or the activation of the Ca 2ϩ -regulated ATP-Mg/P i carrier (25,26). These await identification. In the present work, we report the complete characterization of the human SCaMC genes. We have identified three SCaMC genes in the human genome. All code for highly conserved proteins with a characteristic MC domain and an N-terminal extension very similar to calmodulin, and all were found to be located exclusively in mitochondria. One of the SCaMC genes, SCaMC-2, has four variants generated by alternative splicing, making SCaMC one of most complex subfamilies of MCs.

Sequence Analysis
The BLAST homology searches were performed with the available web-based programs of the National Center for Biotechnology Information at www.ncbi.nlm.nih.gov. DNA and protein sequence alignments were also carried out using the ClustalW program of the European Bioinformatics Institute (EBI, www.ebi.ac.uk). The percentages of identical and similar amino acids were calculated using the program Blastp. Chromosomal location, intron-exon boundaries, and homology scores were assigned by comparing cDNA sequences with those deposited in GenBank TM using BLASTs programs. Oryza sativa sequences were obtained of the TIGR Rice Genome Project. Multiple alignment outputs were colored with BOXSHADE 3.21 (www.ch.embnet.org) or the Belvu program (www.sanger.ac.uk/Software/Pfam/help/belvu_setup.shtml).

RNA Isolation and RT-PCR Analysis
Total RNAs were extracted from different cell lines by the guanidinium thiocyanate method. First strand cDNA was synthesized by using 5 g of total RNA obtained from different cell lines as template, 100 ng of oligo(dT) primer (Roche Applied Science), or 20 pmol of specific reverse oligonucleotides (Isogen Bioscience BV) and avian myeloblastosis virus reverse transcriptase (Promega). Usually, 2-5% of the yield from first strand cDNA synthesis was used as template for subsequent PCR amplifications using TaqDNA polymerase (PerkinElmer Life Sciences). Unless otherwise stated, conditions for PCR amplification were: 35 cycles with 1 min at 95°C, 1 min at 55°C, and 1 min 30 s at 72°C, and a last cycle of elongation at 72°C for 10 min. Blunt-ended fragments were purified and ligated into the PCR-cloning vector pST-Blue-1 (Novagen). The nucleotide sequences were determined using ABI prism dye terminator cycle sequencing kit (PerkinElmer Life Sciences).

Isolation of Human SCaMC Family Sequences
SCaMC-1-A partial SCaMC-1 sequence was obtained from human EST AA001086 (purchased from ATCC). To obtain 5Ј-end sequences, a human heart LambdaZap cDNA library (Stratagene, La Jolla, CA) was screened using as probe the 2.3-kb NotI-EcoRI insert contained in AA001086 as described previously (15). Only one positive clone, p3, containing sequences upstream of the AA001086 5Ј-end was purified and sequenced (encompassing nucleotides 687-1357 of final assembled SCaMC-1 sequence). Complete 5Ј-end extension was carried out by RT-PCR from HEK-293T total RNA. cDNA synthesis was performed with oligo(dT) (Roche Applied Science), and PCR amplification was realized using as forward primer, 5Ј-TCTGGGACCATGTTGCGCTG-3Ј (designed according to the sequence of XM_001285), and as reverse primer, 5Ј-TGCCTGCATAAGGTATGATACC-3Ј (complementary to nucleotides 401-380 of p3). The PCR product, of 1087 bp, was analyzed and joined to p3 and AA001086 sequences to assemble a full-length SCaMC-1 cDNA of 3290 bp (accession number AJ619987).
The SCaMC-2 5Ј-end variants were also obtained by RT-PCR. For SCaMC-2b, reverse transcription was performed with antisense primer 42-9. Then, for PCR amplification, the reverse primer used was 42-7 and the forward one was 42b (5Ј-CCTGTGTGTCAACGACCTGG-3Ј, based on EST AL523261 sequence). The 5Ј-end of exon 1b was amplified using 0.5 g of human genomic DNA as template. The PCR reaction was performed in 10% Me 2 SO to amplify CG-rich sequences. The primers used were forward (5Ј-CCGATGGTGAGCAGTGTGTTG-3Ј) and reverse (5Ј-TCGGTGCGGTGCAGTCCCAG-3Ј, from EST AL523261). The fragments obtained of 426 and 251 bp, respectively, were digested in an overlapping NarI site, purified, and fused together into pUC18 cloning vector. To obtain the 5Ј-end of SCaMC-2c, human astrocytoma U87-MG total RNA was used for reverse transcription and 42-9 as reverse primer. The primers for PCR reactions were antisense 42-7 primer and sense 42c (5Ј-GACCGTGATGTTGCAGATGC-3Ј, according to human KIAA1896 protein sequence, accession number XM_0277668). The clones containing specific 5Ј-end isoforms were double-digested in an internal AvaI site (located in the common SCaMC-2 sequence) and in EcoRI flanking sites, the inserts obtained were ligated with AvaI-EcoRI fragment of pUC18-SCaMC-2 to produce pUC18-SCaMC-2b (accession number AJ619990) and pUC18-SCaMC-2c (accession number AJ619991).
SCaMC-3-A partial human SCaMC-3 sequence (nucleotides from 1 to 1329) was obtained after sequencing human EST AI073878 (1435 bp, obtained from ATCC). SCaMC-3 3Ј-end sequence was amplified by RT-PCR from total RNA of HEK-293T cells. Reverse transcription was performed with oligo(dT) (Roche Applied Science) as primer. For PCR reactions a forward primer (5Ј-AGCAGTACAGCCACGACTCG-3Ј, corresponding to nucleotides 1202-1220 of AI073878) and a reverse primer (5Ј-GGGATCCTGTGGTTGGATCA-3Ј, derived from human EST BF920631 sequence) were used. The 394-bp product obtained was cloned and sequenced. Finally, after digestion at a KpnI site, located in the overlapping regions of AI073878 and the PCR product 3Ј-end, and flanking EcoRI sites, a full-length coding region was assembled into pBluescript-SK, thus obtaining a final cDNA of 1594 bp (accession number AJ619988).

Expression Analysis
A human multiple tissue Northern blot containing 2 g of poly(A) ϩ RNA per lane was purchased from Clontech. Northern blot analysis was done using nick-translated [␣-32 P]dCTP-labeled specific-probes for human SCaMC-1 gene (2.3-kb NotI-EcoRI insert contained in AA001086), SCaMC-2 (a 2.15-kb NotI-EcoRI fragment from H42187, which detects all SCaMC-2 variants), and SCaMC-3 (a 1.4-kb fragment NotI-EcoRI from AI073878). An EcoRI-PstI fragment specific for SCaMC-2c, containing exclusively exon 1c sequences (corresponding to nucleotides 1-211 of SCaMC-2c) was also used. Hybridization was carried out for 16 h in a buffer containing 5ϫ SSC, 5ϫ Denhardt's solution, 50% formamide, 0.1% SDS, and 200 g/ml denatured salmon sperm DNA at 42°C. The blot was washed under conditions of high stringency, at 65°C in a buffer containing 1ϫ SSC, 0.2% SDS for 30 min and with 0.1ϫ SSC, 0.1% SDS for 30 min. It was then stripped in 0.1% SDS at 100°C for 30 min and consecutively rehybridized, under identical con-ditions, with each SCaMC probe as well as with a rat ␣-actin as control.

Plasmids Construction
An 8-amino acid FLAG epitope tag (peptide DYKDDDDK) was introduced at the C terminus immediately preceding the termination codon as described (15,16). These modified 3Ј-fragments encoding C terminus FLAG-tagged forms of human SCaMC variants were confirmed by sequencing and inserted into assembled cDNAs by means of the use of appropriate endonucleases restriction sites. The complete cDNAs modified at the C terminus were cloned into the bicistronic expression vector pIRES1 (Clontech) BamHI-digested to obtain pIRES-SCaMC-2a FLAG (also -2b FLAG , -2c FLAG , and -2d FLAG ) and pIRES-SCaMC-3 FLAG . For SCaMC-1 expression a partial cDNA encoding amino acids 8 -464 was obtained. PCR amplification was performed using as forward primer, 5Ј-AACATGGGCTTCGTGCTGCCCAC-CGCG-3Ј (nucleotides 28 -48 of SCaMC-1) containing an additional upstream ATG codon, and as reverse primer, 5Ј-TATTTTCATAACCA-CATAACTG-3Ј (complementary to nucleotides 1411-1389). Both primers were designed according to the AF123303 sequence. The amplified partial cDNA was tagged at the C terminus with FLAG epitope obtaining finally pIRES-SCaMC-1 (8 -464)FLAG .

Cell Culture and Plasmid Transfection
COS-7, HEK-293T, SH-SY-5Y (neuroblastoma), U87-MG (astrocytoma), clone 9 (a rat liver cell line), and INS-1 (rat insulinoma) cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% (or 5% in COS-7 and HEK-293) inactivated fetal bovine serum (Invitrogen) at 37°C in a 7% CO 2 atmosphere. RAW-264 cell line was maintained in RPMI medium 1640 supplemented with 5% fetal bovine serum. For transfections, COS-7 and HEK-293T cells were grown in 6-cm dishes and transiently transfected using the LipofectAMINE reagent (3 l/g of DNA) (Invitrogen) according to the manufacturer's recommendation. 6 g of reporter plasmid were used per dish. Cells were incubated with LipofectAMINE-DNA complexes for 6 h, and then the media was replaced with fresh Dulbecco's modified Eagle's medium containing 5% fetal bovine serum. After 16 -36 h of incubation to allow expression, cells were harvested for Western analysis. For immunoflu-orescence assays, cells were grown over coverslips, transfected, and, after 36 h, fixed and processed for fluorescence microscopy.

Immunofluorescence Analysis and Mitochondria-specific Staining
For the study of mitochondrial location, living cells were incubated with MitoTracker Red CMXRos (200 nM, 30 min, Molecular Probes) at 37°C, and washed in prewarmed PBS. MitoTracker-loaded cells were fixed in 2% paraformaldehyde (4 min, 4°C) and 100% methanol (Ϫ20°C, 3 min) rinsed with phosphate-buffered saline (PBS) and incubated with NaBH 4 1 mg/ml in PBS for 10 min. Cells were permeabilized with 0.2% Triton X-100 in PBS for 10 min and then blocked in PBS/1% BSA for 1 h. Incubation with primary antibody (anti-FLAG M2 antibody diluted 1:100, Sigma) in blocking medium was carried out at room temperature for 1 h. After washing with PBS, cells were incubated for 30 min with secondary antibody, FITC-conjugated anti-mouse IgG (1: 100 dilution in blocking medium, Sigma). To detect by immunofluorescence the endogenous SCaMC-1 protein, cells were processed as described above, and the primary and secondary antibodies used were an antibody against the N terminus of human SCaMC-1 (1:200), and an FITC-conjugated anti-rabbit (Alexa) antibody (1:500) both in PBS/1% BSA. Cells were washed three times with PBS, and the coverslips were mounted using Mowiol. Fluorescence microscopy was performed using an Axiovert epifluorescence microscope (Carl Zeiss) at a nominal magnification of ϫ100.

Immunoblot Analysis
Mitochondrial-enriched fractions were obtained from cells as described (28) and analyzed by Western blotting using and enhanced chemiluminescence (ECL) kit (Amersham Biosciences). Anti-FLAG M2 antibody was used at 1:3,000 and secondary horseradish peroxidasecoupled anti-mouse antibody at 1:1,000 (Bio-Rad). Rabbit serum against the N terminus of human SCaMC-1 was used at a dilution of 1:2,000, and antibodies against the N terminus of SCaMC-2b and SCaMC-3 were used at 1:1,000. To control for the amount of mitochondrial protein loaded, blots were stripped and incubated with an antibody to the mitochondrial protein ␤-F 1 ATPase (a gift from the J. M. Cuezva laboratory, Centro de Biología Molecular Severo Ochoa) at a dilution of 1:5,000. Goat peroxidase-coupled anti-rabbit antibody, 1:10,000, was from Bio-Rad.

Identification of Human SCaMC
Proteins-With the purpose of finding human SCaMC proteins, we performed TBLASTN searches in public databases using Efinal protein as query, a protein of 475 amino acids previously described in rabbit (11, GenBank TM accession number AAB69156). By means of this approach we have identified several human ESTs homologous to the rabbit sequence. These sequences did not display complete identity, suggesting the existence of, at least, three human proteins homologous to Efinal. Fig. 1 shows the strategy used for cloning these human cDNAs (Fig. 1A), and the homology among the human ESTs found and between these and Efinal (Fig. 1B). These new genes have been named SCaMC-1, -2a, and -3.
SCaMC-1-We found two EST clones, AA001086 and H42187, with homology of about 70 -80% at the nucleotide level, and coding for amino acid sequences 96 and 67% identical to Efinal (11), respectively, indicating that they encoded different Efinal-related proteins. These ESTs contained partial cDNAs sequences, both providing the 3Ј-ends and poly(A ϩ ) tails, with predicted sequences matching the C-terminal 200 amino acids of Efinal (Fig. 1A). To obtain the complete cDNAs, the inserts from both ESTs were used as probes in a hybridization screening using a human heart cDNA library (Zap, Clontech). A single clone was obtained, corresponding to AA001086 with upstream sequences going up to amino acid at position 227 of Efinal. A further inspection of databases revealed partial cDNAs that contained frame-shifts but provided the 5Ј-region, including the start codon (accession number AF123303 and XM_001285) of the cDNA. This 5Ј-end sequence was verified by RT-PCR using total RNA from human HEK-293T cells as template and sequencing of the PCR product (see "Materials and Methods" for primers used). The sequences were assembled in a full-length cDNA of 3290 bp (accession number AJ619987). This sequence belongs to UniGene cluster Hs.24713 and has been named SCaMC-1. The 477-amino acid long protein encoded by SCaMC-1 presents a remarkable homology (97% similarity and 95% identity), with the Oryctolagus cuniculus Efinal protein, suggesting that SCaMC-1 is the true human Efinal orthologue (see Fig. 1C).
SCaMC-2a-SCaMC-2a was obtained as the cDNA corresponding to human EST H42187. To obtain the 5Ј-sequences of this cDNA, BLAST searches in the EST division of GenBank TM were carried out using H42187 and assembled SCaMC-1 sequences as query. Although we did not identify any EST overlapping the 5Ј-sequence of H42187, we initially found a human EST, AW614226, that showed relevant homology, but not identity, to sequences at the most 5Ј-end of SCaMC-1, suggesting that it could belong to the same transcript as H42187. To test whether this clone represented the true 5Ј-region of H42187, RT-PCR assays were performed with primer pairs derived from internal sequences from H42187 and AW614226 (see "Materials and Methods"). The analysis of amplified products obtained confirmed that both EST sequences belong to the same transcript. Also, this approach has allowed us to obtain a full-length of 3240 bp, which has been named SCaMC-2a (accession number AJ619989). The lengths of the open reading frames of SCaMC-2a and SCaMC-1 are slightly different due to the variable 5Ј-initial sequence. SCaMC-2a belongs to UniGene cluster Hs.24713 and encodes a putative protein of 469 amino acids. The corresponding protein shows lower homology with Efinal than SCaMC-1 (67% identity and 82% of homology), and the first 50 amino acids do not align with the N terminus sequences of Efinal or SCaMC-1 (Fig. 1C). Recently, a rat orthologue for SCaMC-2a has been identified by differential display as an up-regulated transcript in pancreatic dexamethasone-treated AR2J cells (accession number NM_145677 (22)), and a partial sequence of human SCaMC-2a was also reported as an mRNA present in human retina (29).
SCaMC-3-The third human cDNA homologue to Efinal was obtained from db EST AI073878. Except for its coding 3Ј-end, this clone has high similarity to the SCaMCs characterized so far. Because the 3Ј-end was not related to the previously characterized cDNAs, a further search was carried out into db ESTs and genomic databases to detect the correct 3Ј-sequences. Thus, we found ESTs whose putative translated sequence showed high homology to SCaMC-1 and -2 proteins (accession numbers BF953775, AW379976, and BF920631). Equivalent ESTs were found in mouse (BI145091 and BF142980). The full-length SCaMC-3 sequence (accession number AJ619988) was confirmed by RT-PCR with specific primers. SCaMC-3 encodes a putative protein of 468 amino acids, with 78% of similarity and 61% of identity to Efinal (Fig. 1C).
The Human SCaMC Genes Exhibit Identical Genomic Organization-The intron-exon organization of the SCaMC-1 gene was obtained from the sequences of two bacterial artificial chromosomes (BAC) (GenBank TM accession numbers AC013627 and AL390036) mapping to chromosome 1p36.13, available from the Human Genome Project. The genomic structure of the SCaMC-2 gene was obtained from BAC clone AL590708 mapping to chromosome 9q34.13. For SCaMC-3, the genomic organization was obtained from BAC clone AC010503 mapping to chromosome 19p13.3. The human SCaMC genes have identical genomic structure (Tables I-III). Their coding regions are formed by 10 exons and 9 introns with highly conserved exon sizes and splicing junctions (see Tables I-III). All exon-intron splice junctions follow the gt/ag rule, each intron beginning with GT at the 5Ј-splice donor site and ending with a AG at the 3Ј-splice acceptor site. Other consensus elements associated with splice junctions are also found with a high frequency. Thus, except for SCaMC-1 exon 4, all exons show 5Ј-splice sites that end with an AG (seven of them) or a G. The major difference among the genes lies in the size of introns (Tables I-III), so that SCaMC-1 spans ϳ65 kb, SCaMC-2a, about 10 kb, and SCaMC-3, about 24 kb of genomic DNA.
Mouse SCaMCs genes show a genomic organization identical    (Fig. 2A). Equivalent mouse db ESTs were also found (BF783865 and BF784790 for the first group and BB652192 and BI738380 for the second one). In addition, a new related transcript was detected in the non-redundant data base (accession number XM_027668, putative protein KIAA1896) differing from the other two from the same point suggesting that a third alternative 5Ј-coding sequence would exist (Fig. 2A). These sequences have been confirmed by RT-PCR amplification using total RNA from different human cells lines as template and with specific forward primers for each alternative 5Ј-end sequence (described under "Materials and Materials, " Fig. 2B). The alignment of amplified sequences confirmed the existence of, at least, three transcripts that present different 5Ј-ends probably arising from the use of alternative promoters. A multiple alignment with human genomic sequences upstream to the former exon 1 (contained into BAC clone AL360268) confirmed the existence of two distant forms of exon 1 (see Fig. 2C). These distal alternative exons 1 are located at 32 and 9 kb from the common exon 2. The new transcripts, different from SCaMC-2a, have been named SCaMC-2b for sequences containing the 5Ј-sequence of EST AL559911, and SCaMC-2c for transcripts containing the 5Ј-sequence of XM_027668 (accession numbers AJ619990 and AJ619991, respectively). A diagram with the relative positions of SCaMC-2a first exons 1a, 1b, and 1c is shown in Fig. 2C. The XM_027668 sequence differs from that of the SCaMC-2c transcript, in that it shows an insertion of 36 nucleotides, which we have been unable to amplify.
Interestingly, the use of alternative transcription start sites results in proteins that differ in their N terminus (Fig. 2D). The ORFs of SCaMC-2 variants differ due to the variable sequence and length of their corresponding exons 1, which include the ATG start codon and encode putative proteins of 503 and 488 amino acids, corresponding to SCaMC-2b and SCaMC-2c, respectively (Fig. 2, C and D). Moreover, the amino acid sequences encoded by alternative exons 1 are quite different (see alignment in Fig. 2D). Mouse SCaMC-2 displays identical alternative initiation sites, and we have also found mouse ESTs homologous to SCaMC-2c (BQ179791 and BU709066). In addition, the organization of the 5Ј-region of murine SCaMC-2 perfectly matches that of the human gene. The relative positions and sizes of exons 1a, 1b, and 1c are coincident in the rat and mouse genome (not shown).
In addition to variants 2a, 2b, and 2c, a new  SCaMC-2 gene was cloned from mRNA derived from HEK-293 cells. This variant, named SCaMC-2d, was obtained during the cloning of amplified SCaMC-2a by RT-PCR, where clones with a fragment longer than that expected were obtained. The fragment was analyzed, and its sequence was compared with that of SCaMC-2a as well as with human BAC AL360268. It contained an insertion of 43 bp between the 3Ј-end of exon 1a and the 5Ј-end of exon 2 matching 3Ј-flanking sequences of exon 1a (Fig. 3). This indicates that the SCaMC-2d transcript (accession number AJ619992) was probably generated by the use of an alternative 5Ј-splicing donor site 43 bp downstream in intron 1. The 43-bp insertion generates a frameshift in the mRNA and is followed by various stop codons. However, a shorter protein could still be generated if a downstream ATG is used as translational start codon. The first downstream in-frame ATG, corresponding to methionine at position 104 of SCaMC-2a, would originate a protein of 366 amino acids, and a second one (position 126 of SCaMC-2a) a protein of 344 amino acids. Although sequences corresponding to SCaMC-2d transcript have not been found so far in EST databases, we have also obtained SCaMC-2d sequences when SCaMC-2a transcripts were cloned from a human lung cDNA library (Fig. 3) ruling out that this isoform is solely an artifact of HEK-293 cells. Moreover, a similar protein is also expressed in Drosophila (see below). Nevertheless, the new variant was only detectable by hybridization with a specific primer derived from the insertion indicating that SCaMC-2d mRNA has a low expression (Fig. 3).
SCaMC Proteins Form a Conserved Family with Calciumbinding Domains Related to Calmodulin-The three human SCaMC proteins described so far, as the calcium binding AGC isoforms aralar1 and citrin (15,16,18), have a bipartite structure with a MC domain of about 300 amino acids at the C terminus and a long N terminus harboring EF-hand calcium binding motifs. However, the predicted size for SCaMC proteins, around 480 amino acids, is 200 residues shorter than AGC proteins. In fact, even though both AGCs and SCaMCs contain EF-hand motifs, the sequences of their N-terminal halves and MC domains are quite different. This suggests that SCaMC proteins represent a new subfamily of mitochondrial carriers with transporter activity and calcium regulation properties both different from that of the AGCs.
As for human AGCs, SCaMC isoforms show the highest similarity at their carboxyl-terminal halves, that contain the MC homology region (about 68 -83% identity at amino acid level). The MC domain is responsible for the transport activity of the protein, suggesting that all SCaMCs may have the same transport properties. Comparison between MC homology regions of SCaMCs and other members of the human MC family indicates that the highest similarity is obtained with the Grave's disease protein (33-35% identity and 53% similarity), a carrier recently related with mitochondrial coenzyme A accumulation (30), followed by the ADP/ATP translocase 1 (26 -30% identity and 44 -48% of similarity).
SCaMC proteins are widely conserved, and we have identified orthologues in most eukaryotic organisms, although the number of related genes is variable for each organism. A multiple alignment of SCaMC-related proteins from several species is shown (Fig. 4A). A phylogenetic dendogram is shown in Fig.  4B. A single SCaMC-related isoform was found in Saccharomyces cerevisiae (YNL083w, P48233), Schizosaccharomyces pombe (NP_595952), or Neurospora crassa (CAC18152). Other species such as Caenorhabditis elegans or Arabidopsis thaliana showed several isoforms. Two were also found in C. elegans (Q20799 and Q19529) and O. sativa (1953.m00141 and 4272.m00154, contained in genomic clones AP003629 and AP004869, respectively) and three in A. thaliana (CAB87921, BAB10081, and BAB08751). In addition, except for the S. pombe orthologue, all SCaMC-related proteins display four EF-hand motifs at equivalent positions. In Drosophila melanogaster two proteins of 520 and 363 amino acids were found (accession numbers AY122132 and AY119650, respectively), probably generated by alternative promotor usage. The shorter protein shows a good alignment with SCaMC-2d, the human variant with a single calcium-binding domain.
Similarity searches indicate that the N terminus of SCaMCs has high homology to calmodulin (CaM). The SCaMC-1 N terminus shows 25% identity and 53% similarity to human CaM, with four EF-hand motifs at conserved positions, and SCaMC-2 and -3 have close values (25% identity and 56% and 48% similarity, respectively). Indeed, Efinal and rat SCaMC-2 have been shown to bind calcium (11,22). An alignment between SCaMCs and human CaM is shown in Fig. 5. Important differences between SCaMCs and CaM include a lack of canonical EF-hand 4 (see Fig. 5, the non-canonical threonine is indicated) and the fact that the central helix that connects each pair of EF-hand motifs is six amino acids shorter than CaM and consists of one and half ␣-helical turn (Fig. 5). SCaMCs also contain a lower percentage of methionine (2-2.5%) than CaM (6.6%), a percentage similar to that found in recoverin (2.5%), another four-EF-hand calcium-binding protein (31).
The N termini of the SCaMC isoforms also display variations among each other (Fig. 5). These variations are largely confined to EF-hand 1 and the N terminus (Fig. 5). Remarkably, EFhand 1 was absent in both SCaMC-2a and SCaMC-2c. In contrast, SCaMC-2c and -2b variants showed long N-terminal extensions (Fig. 2D). In SCaMC-2b, this N terminus is rich in Ala and Ser residues, as found in mitochondrial presequences (32,33). However, the most relevant differences are found in SCaMC-2d, in which the predicted shortened ORF lacked EFhands 1-3 and contained only the non-canonical EF-hand 4 (Fig. 5).
Expression Pattern of Human SCaMC Isoforms-The expression pattern was studied by Northern blot analysis with specific probes for each human SCaMC isoform (Fig. 6A). For SCaMC-2, the probe corresponded to a 3Ј-fragment of cDNA common for all SCaMC-2 variants. Except for SCaMC-2, a single mRNA of around 3.4 kb is detected for the three SCaMC isoforms, in agreement with the size of the corresponding cDNAs. Most human tissues, pancreas, kidney, skeletal muscle, lung, brain, and heart, express the three SCaMC isoforms. Except for skeletal muscle, liver, and brain, in most tissues SCaMC-1 had the highest expression levels, followed by SCaMC-2 and -3. SCaMC-2 expression was prominent in skeletal muscle, and it was the major liver isoform. SCaMC-2 and -3 were the major brain isoforms.
Two SCaMC-2 mRNAs, a 3.6-kb band in addition to the common 3.4-kb transcript, are detected in brain using as probe a 3Ј-fragment of cDNA common for all SCaMC-2 variants (Fig.  6A). To investigate the origin of this 3.6-kb band, we re-hybridized the mRNA blot with specific probes for SCaMC-2c and 2b, which have longer cDNAs. As is shown in Fig. 5A, a single band is detected with a SCaMC-2c-specific probe, and it corresponds to the upper hybridization signal (arrow in Fig. 6A). Thus, the 3.6-kb brain-specific band represents the SCaMC-2c messenger FIG. 6. Expression of human SCaMC isoforms. A, Northern blot analysis of SCaMC isoform expression in human tissues. A human multiple-tissue Northern blot (from Clontech) was hybridized with 32 P-labeled probes specific for each human isoform. The blot was stripped and subsequently reprobed under identical conditions with each probe indicated. For specific hybridization probes for SCaMC-1, SCaMC-2, and SCaMC-3 were inserts that contained human ESTs AA001086, H42187, and AI07387, respectively. The PCR-amplified sequence of exon 1c was used as probe to detect the SCaMC-2c transcript (indicated by arrows). ␤-Actin was used as an internal control (lower panel). The molecular weight markers (in kb) are indicated. B, expression analysis of SCaMC-2b and SCaMC-2 variants in human tissues. Equivalent aliquots of a multiple human cDNA panel, consisting of heart, brain, placenta, lung, skeletal muscle, kidney, and pancreas (Clontech), were used as templates. Shown are the PCR products obtained with primers 42b and 42-7, specific for the SCaMC-2b transcript, and the product was amplified with primers 42-15 and 42-7, corresponding to SCaMC-2 transcript. The amplified fragments were sequenced to verify their exonic compositions. Human ␤-actin sequences were amplified for semiquantitative comparison. The results indicate a more restricted distribution of SCaMC-2b sequences with respect to the ubiquitous expression of SCaMC-2. (Fig. 2). In fact, we only have successfully amplified SCaMC-2c-specific sequences using RNA from neural cell lines as template, and all mouse and human ESTs belonging to SCaMC-2c have been obtained from neural tissues. This indicates that the expression of the SCaMC-2c variant is restricted to brain.
When samples were hybridized with a specific probe for SCaMC-2b, a new band of 4.7 kb is observed in all samples (not shown), which was shown to represent cross-hybridization with residual 28 S rRNA present in poly(A ϩ ) multiple tissue blot through G/C-rich sequences of exon 1b (80% rich in G/C content), as described for other probes enriched in G/C sequences (34). Thus, we analyzed SCaMC-2a and -2b expression patterns by PCR analysis using a multitissue cDNA library (Clontech) and specific forward primers for exons 1a and 1b (Fig.  6B). SCaMC-2a was expressed in all the tissues tested. On the other hand, SCaMC-2b shows a more restricted distribution with presence in kidney and in low amounts in lung (Fig. 6B). Thus, SCaMC-2a, which contains exon 1a, is widely expressed, whereas the transcripts generated from the most distal promoters, SCaMC-2b and -2c, show a more restricted distribution.
Mitochondrial Localization of SCaMCs Proteins-Having established the expression pattern of the SCaMC proteins, we studied the intracellular location of these proteins. To this end, SCaMCs were transiently overexpressed in COS-7 cells, and their intracellular distribution was analyzed by immunofluorescence. cDNA sequences encoding for SCaMC-2a, -2b, -2c, and -3 isoforms were fused to the FLAG epitope at the C terminus immediately preceding the stop codon and cloned into the expression vector pIRES-1. For SCaMC-1, a shortened cDNA, corresponding to amino acids 8 -464, was fused to the FLAG sequence. The FLAG-tagged proteins are visualized with a FLAG-specific antibody. We detected FLAG-positive signals only in intracellular organelles with the expected morphology of mitochondria (Fig. 7). Moreover, the staining pattern of SCaMCs completely overlapped that of MitoTracker Red CMXRos (Fig. 7), a mitochondrial-specific dye that labels mitochondria in vivo (15). No other structures than mitochondria were labeled in FLAG-positive cells, ruling out a peroxisomal localization of the overexpressed products. Identical subcellular distribution is obtained when SCaMCs were transfected in a HEK-293T cell line (not shown). A mitochondrial distribution pattern has also been recently reported for the rat orthologue of SCaMC-2a (22).
As mentioned above, the SCaMC-2d cDNA is predicted to generate shorter proteins than any of the SCaMC-2 variants by using a translational start codon downstream of that used in SCaMC-2a. The first downstream in-frame ATG, corresponding to methionine at position 104 of SCaMC-2a, would originate a protein of 366 amino acids. To determine the size and intracellular localization of SCaMC-2d, a full-length FLAG-tagged SCaMC-2d was transfected in COS-7 cells. The distribution of FLAG epitopes completely overlapped that of MitoTracker (Fig.  7B) indicating that SCaMC-2d FLAG is also localized to mitochondria and that this variant is indeed translated from an internal ATG.
The size of expressed SCaMC-2d protein has been studied after expression in HEK-293T cells. Mitochondria-enriched fractions from SCaMC-2d FLAG -transfected cells were subjected to SDS-PAGE and immunoblotting with anti-FLAG antibody. Unexpectedly, we detected a doublet, with equal intensities for the two bands (Fig. 7C). The apparent molecular mass of the upper band, around 40 kDa, agrees with the use of the first ATG (nucleotides 356 -358) in the predicted SCaMC-2d ORF (Fig. 6C). N-terminal processing or the use of an alternative internal start codon could generate the lower band. In fact, SCaMC-2d has two close in-frame ATGs, corresponding to methionines 23 and 29 (Fig. 6C). To study whether any of these ATG were used as translation start sites, a new construct lacking the initial 400 nucleotides, including the first ATG (named SCaMC-2⌬ FLAG ), was generated and transfected into HEK-293T cells. As shown in Fig. 6C, a single immunoreactive band coincident with the doublet lower one was obtained, confirming that a second downstream ATG is really used as the start codon and that the smaller protein is not a processed product. This second ATG was probably methionine 23, because it is flanked by sequences that fit Kozac's rules. The localization of endogenous SCaMCs was also verified by immunofluorescence with isoform-specific SCaMC antibodies. Fig. 7B shows the labeling pattern of SCaMC-1 in COS-7 and HEK-293T, which clearly matches that of MitoTracker, demonstrating that endogenous SCaMC-1 is also exclusively targeted to mitochondria (Fig. 8B). Identical results are obtained in other cells lines with high SCaMC-1 protein levels as mouse 3T3-L1 or human fibroblasts (not shown). Equivalent assays with anti-SCaMC-2 or anti-SCaMC-3 failed to detect both proteins probably due to their lower expression levels with respect to that of SCaMC-1.

Endogenous SCaMCs Are Also Localized in Mitochon
N-terminal Extensions Are Not Necessary for Mitochondrial Targeting-Unlike other mitochondrial proteins that carry cleavable N-terminal presequences for mitochondrial targeting, MC family proteins contain several targeting signals distributed over the entire length of the protein (35). Because SCaMC members are clearly different from other MCs, and some (SCaMC-2b) have regions of homology with mitochondrial presequences in their N-terminal extensions, we tested whether mitochondrial targeting signals were present exclusively in the C-terminal half of the mature proteins and not at the N-terminal region. To this end, COS-7 cells were transiently transfected with FLAG-tagged carboxyl-truncated variants, CT-SCaMCs, that contained the complete MC homology sequence and lacked their N-terminal extensions. As shown in   9, MitoTracker staining totally overlapped FLAG distribution indicating that truncated SCaMC proteins localize to mitochondria exactly as the full-length proteins. This result shows that, as with AGC isoforms (16,36), the C-terminal half of SCaMC proteins contains sufficient information for import and assembly into mitochondria. DISCUSSION We report here the complete characterization of a new group of human mitochondrial carriers with N-terminal extensions harboring calcium-binding domains. The SCaMCs have three human isoforms and, together with the AGCs, make up the subfamily of calcium-binding mitochondrial carriers. The first MC protein belonging to this SCaMC subfamily was predicted from the study of S. cerevisiae genome (YNL083w (5, 20)). Weber et al. (11), then found Efinal, the rabbit SCaMC-1 orthologue, showing that it was localized both to peroxisomes and mitochondria. In contrast with this, we have now found that all SCaMC members, as the AGC proteins, are mitochondrial proteins. This was shown by studying the distribution of FLAGtagged SCaMC proteins in transfected cell lines of endogenous SCaMC proteins with isoform-specific antibodies and by using Western blot analysis of mitochondrial-enriched extracts. Furthermore, rat SCaMC-2a has been shown to localize to mitochondria (22). In addition, SCaMC-1 has been identified by a proteomic approach as a mitochondrial protein in IMR-32, a neuroblastoma cell line (37), and in human heart (38). Indeed, SCaMCs proteins show a higher degree of similarity to classic MCs such as ADP/ATP translocase or Grave's disease protein than with peroxisomal carriers as the Candida boidinii protein PMP47 or its human and yeast homologues, PMP34, recently characterized as a peroxisomal ATP carriers (12,13,39).
The new subgroup of SCaMCs is among the most complex ones described so far. It is composed of three isoforms that show a high degree of conservation (around 70% of identity at the amino acid level), one of which has additional N-terminal variants arising from alternative splicing and promotor usage. Among mitochondrial carriers, only the ADP/ATP translocase is represented by three isoforms in the human genome. Other MCs, such as aspartate/glutamate carriers (AGC), ornithine, or glutamate carriers, have two isoforms with 63-69% (glutamate carrier and AGC) and 87% (ornithine) identity at the amino acid level, similar to that among SCaMC isoforms (16,40,41). As for other MC isoforms, SCaMCs differ in their expression levels and in their tissue distribution. Human SCaMCs are largely ubiquitous, with higher levels of SCaMC-1 than SCaMC-2 or SCaMC-3 expression in all tissues (except brain and liver).
All SCaMC homologues, excluding that of S. pombe, show EF-hands motifs at equivalent positions of their N-terminal extensions, and both SCaMC-1 and -2a have been shown to bind calcium (11,22). This suggests that Ca 2ϩ binding could be an important and conserved mechanism for regulation of their function. Unlike AGCs, the N-terminal extensions of SCaMC proteins are remarkably similar to calmodulin (CaM) and calmodulin-related proteins such as recoverin (15,16). CaM and CaM-related proteins with two EF-hand pairs typically undergo a calcium-dependent conformational change, which opens a target binding site (42)(43)(44). However, SCaMC proteins differ with respect to CaM-related proteins in a number of ways (45). First, EF-hand 4 is non-canonic in all three SCaMCs. Second, SCaMCs have a lower percentage of methionine residues than CaM. It is believed that methionine residues, very abundant in CaM, are important in the binding of target peptides, because methionine side chains are flexible and the sulfur atom has a larger polarizability than carbon, resulting in stronger van der Waals interactions (44). Third, the central ␣-helix between EF-hand 2 and EF-hand 3, which is involved in interaction with target peptides and undergoes major conformational changes upon Ca 2ϩ binding (31,43,45), is six or five amino acids shorter than CaM or recoverin, respectively (Fig.  5), i.e. about one turn and a half shorter. Thus, taken together, these results suggest that SCaMCs may undergo smaller Ca 2ϩdependent conformational changes than typical four EF-hand proteins.
The most variable region in the SCaMC isoforms is the N-terminal half. Moreover, the splice variants of SCaMC-2 give rise to further variation in this region of the protein, with EF-hand numbers varying from three or four to only one, for SCaMC-2d. Data on alternative splicing of MCs genes is scarce. However it has been reported that UCP5 and phosphate carrier isoforms are generated by alternative splicing of internal exons (46,47). For both carriers, the splice isoforms differ in their functional characteristics (46,47). Although the lack of EFhand 1 in the rat orthologue of SCaMC-2a does not prevent calcium binding (22), it is possible that inactivation or absence of specific EF-hand motifs may provide an mechanism for Ca 2ϩ signaling diversification. Thus, the variable number of EFhand motifs in the SCaMC-2 variants arising from tissuespecific promotor usage, may provide additional mechanisms to modulate their sensitivity to Ca 2ϩ .
The presence of a SCaMC member virtually in every cell type and their high degree of conservation from S. cerevisiae to humans suggest that the function of these proteins is an important one. For the AGCs, sequence conservation between yeast and humans is limited to the MC domain, i.e. the cata- lytic part of the molecule. Thus, the yeast AGC orthologue (Agc1p) lacks calcium binding motifs in its N-terminal extension, suggesting that calcium regulation of aspartate/glutamate exchange was acquired later in evolution (48). However, EF-hand motifs are already present in the yeast representative of SCaMC, YNL083p (20,21).
In the early 1980s, June Aprille and coworkers reported the existence of a Ca 2ϩ -activated ATP-Mg/P i mitochondrial carrier that catalyzes the electroneutral exchange of ATP-Mg for P i and is important for regulating the mitochondrial adenine nucleotide pool size (25, 26, 49 -51). By virtue of this role, it could control the mitochondrial permeability transition (52). The molecular identity of this carrier remains unresolved to this day. The following observations suggest that the SCaMCs may represent the ATP-Mg/P i carrier or a related carrier: 1) The MC with highest similarity to the SCaMCs are the ADP/ATP translocases and MCs involved in the transport of nucleotides (5,7,32). Indeed, the closest relatives to SCaMCs or their yeast orthologue, YNL083p, include the ADP/ATP translocases, the yeast orthologue of the Grave's disease protein that is involved in coenzyme A transport (30), the thiamine pyrophosphate carrier and deoxynucleotide carriers (10,53), and the peroxisomal ATP/ADP carrier (12,13). This suggests that the SCaMCs may also function in nucleotide transport, as the ATP-Mg/P i carrier. 2) As the AGC (18), the ATP-Mg/P i carrier is activated by extramitochondrial Ca 2ϩ (26). The very conserved topology of MCs strongly supports the notion that the Ca 2ϩ binding domains of SCaMCs, similar to those of the AGCs (18), face the extramitochondrial space, suggesting that cytosolic Ca 2ϩ rather than mitochondrial Ca 2ϩ , regulates the activity of the carriers. Nosek et al. (26) noted that the Ca 2ϩregulated site of the ATP-Mg/P i carrier had characteristics in common with calmodulin, because it was sensitive to CaM antagonists. The striking similarity between SCaMCs and CaM strongly suggests that SCaMCs could be sensitive to CaM antagonists.
In conclusion, SCaMCs make up one of the most complex subfamilies within the mitochondrial carriers, in terms of number of isoforms and splice variants. They are distributed essentially in all cell types and their degree of conservation from lower eukaryotes to humans is remarkably high. From their structural characteristics, it is temping to suggest that SCaMCs may function in nucleotide transport in mitochondria, such as ATP-Mg/P i exchange or related transport systems, in a calcium-regulated mode. In addition to the ATP-Mg/P i carrier, which is well represented in liver, other tissues such as the heart and kidney have alternative and/or additional transport systems that accomplish unidirectional nucleotide transport with varying Ca 2ϩ sensibility (54). The high abundance of SCaMC isoforms, the presence of small differences in their relatedness with the adenine nucleotide translocase, and the differences in their expression patterns, could possibly explain the tissue-specific differences in unidirectional transporters of nucleotides.