Evolutionarily Conserved Multiple C2 Domain Proteins with Two Transmembrane Regions (MCTPs) and Unusual Ca2+ Binding Properties*

C2 domains are primarily found in signal transduction proteins such as protein kinase C, which generally contain a single C2 domain, and in membrane trafficking proteins such as synaptotagmins, which generally contain multiple C2 domains. In both classes of proteins, C2 domains usually regulate the respective protein's function by forming Ca2+-dependent or Ca2+-independent phospholipid complexes. We now describe MCTPs (multiple C2 domain and transmembrane region proteins), a novel family of evolutionarily conserved C2 domain proteins with unusual Ca2+-dependent properties. MCTPs are composed of a variable N-terminal sequence, three C2 domains, two transmembrane regions, and a short C-terminal sequence. The invertebrate organisms Caenorhabditis elegans and Drosophila melanogaster express a single MCTP gene, whereas vertebrates express two MCTP genes (MCTP1 and MCTP2) whose primary transcripts are extensively alternatively spliced. Most of the MCTP sequences, in particular the C2 domains, are highly conserved. All MCTP C2 domains except for the second C2 domain of MCTP2 include a perfect Ca2+/phospholipid-binding consensus sequence. To determine whether the C2 domains of MCTPs actually function as Ca2+/phospholipid-binding modules, we analyzed their Ca2+ and phospholipid binding properties. Surprisingly, we found that none of the three MCTP1 C2 domains interacted with negatively charged or neutral phospholipids in the presence or absence of Ca2+. However, Ca2+ titrations monitored via intrinsic tryptophan fluorescence revealed that all three C2 domains bound Ca2+ in the absence of phospholipids with a high apparent affinity (EC50 of ∼1.3-2.3 μm). Our data thus reveal that MCTPs are evolutionarily conserved C2 domain proteins that are unusual in that the C2 domains are anchored in the membrane by two closely spaced transmembrane regions and represent Ca2+-binding but not phospholipid-binding modules.

C 2 domains are primarily found in signal transduction proteins such as protein kinase C, which generally contain a single C 2 domain, and in membrane trafficking proteins such as synaptotagmins, which generally contain multiple C 2 domains. In both classes of proteins, C 2 domains usually regulate the respective protein's function by forming Ca 2؉ -dependent or Ca 2؉ -independent phospholipid complexes. We now describe MCTPs (multiple C 2 domain and transmembrane region proteins), a novel family of evolutionarily conserved C 2 domain proteins with unusual Ca 2؉ -dependent properties. MCTPs are composed of a variable N-terminal sequence, three C 2 domains, two transmembrane regions, and a short C-terminal sequence. The invertebrate organisms Caenorhabditis elegans and Drosophila melanogaster express a single MCTP gene, whereas vertebrates express two MCTP genes (MCTP1 and MCTP2) whose primary transcripts are extensively alternatively spliced. Most of the MCTP sequences, in particular the C 2 domains, are highly conserved. All MCTP C 2 domains except for the second C 2 domain of MCTP2 include a perfect Ca 2؉ /phospholipid-binding consensus sequence. To determine whether the C 2 domains of MCTPs actually function as Ca 2؉ /phospholipid-binding modules, we analyzed their Ca 2؉ and phospholipid binding properties. Surprisingly, we found that none of the three MCTP1 C 2 domains interacted with negatively charged or neutral phospholipids in the presence or absence of Ca 2؉ . However, Ca 2؉ titrations monitored via intrinsic tryptophan fluorescence revealed that all three C 2 domains bound Ca 2؉ in the absence of phospholipids with a high apparent affinity (EC 50 of ϳ1.3-2.3 M). Our data thus reveal that MCTPs are evolutionarily conserved C 2 domain proteins that are unusual in that the C 2 domains are anchored in the membrane by two closely spaced transmembrane regions and represent Ca 2؉ -binding but not phospholipid-binding modules.
The C 2 domain is defined as a sequence motif in a comparison of the primary structures of different protein kinase C isoforms and named in an unbiased fashion as the "second constant sequence" of protein kinase C isozymes (1). Later studies revealed that a large number of proteins include C 2 domains, with Ͼ200 such proteins in the human genome alone (2). We observed that in synaptotagmin 1, a synaptic vesicle protein that binds Ca 2ϩ and contains two C 2 domains (3), C 2 domains are autonomously folded Ca 2ϩ -binding modules (4,5). Subsequently, most C 2 domain proteins were found to form Ca 2ϩ -dependent phospholipid complexes, although some appear to bind to phospholipids in the absence of Ca 2ϩ (e.g. PTEN) (6) and others constitute protein-interaction domains instead of binding to either Ca 2ϩ or phospholipids (e.g. the N-terminal C 2 domain of Munc13-1 (7) or the C-terminal RIM1␣ C 2 domain (8)). Although the Ca 2ϩ -binding properties of many C 2 domain proteins remain to be examined, the large number of C 2 domain proteins in the vertebrate genome makes it likely that this domain represents the second most common Ca 2ϩ binding motif after the EF-hand motif.
Most C 2 domain proteins are either signal transduction enzymes, such as protein kinase C, or membrane trafficking proteins, such as synaptotagmin 1. At least some isoforms of all major signal transduction enzymes, from ubiquitin ligases to kinases to various phospholipases, contain a C 2 domain. Without exception, these proteins are soluble cytosolic enzymes that include a single C 2 domain. In contrast, membrane trafficking proteins generally include at least two C 2 domains, although a few proteins such as the ␥-RIM isoforms (9) and some splice variants of piccolo/aczonin and intersectin (10 -12) contain only a single C 2 domain. In membrane trafficking proteins the different C 2 domains often feature conserved sequence differences, indicating that the C 2 domains are functionally specialized. For example, although in synaptotagmin 1 the C 2 A and C 2 B domains both bind Ca 2ϩ and phospholipids with similar affinities (13), the C 2 B domain contains an additional "bottom" ␣-helix that is absent from the C 2 A domain but is conserved in all of the C 2 B domains of synaptotagmins (14). Membrane trafficking proteins with multiple C 2 domains either have no TMR 1 or a single TMR either at the N terminus (e.g. synaptotagmins; see Fig. 1) (15) or the C terminus (e.g. ferlins) (16).
Work over the last few decades established Ca 2ϩ as the major intracellular second messenger in eukaryotic cells, with specificity achieved by the spatial segregation of Ca 2ϩ signals. In vertebrate genomes, proteins containing EF-hand Ca 2ϩbinding sites are more common than C 2 domain proteins (2). However, Ca 2ϩ binding in EF-hand proteins often does not serve a direct regulatory function but instead acts in Ca 2ϩ buffering (e.g. parvalbumin and calbindin) or subserves a structural role (reviewed in Refs. 17 and 18). Among the EF-hand proteins with a regulatory Ca 2ϩ -binding site, one particular * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AY656715, AY656716, and AY656717.
‡ To whom correspondence should be addressed. protein, calmodulin, appears to mediate more Ca 2ϩ -dependent regulatory actions than all other EF-hand proteins combined and is in fact expressed as an identical sequence from multiple independent genes (reviewed in Refs. 19 and 20).
This situation seems to be completely different for C 2 domain proteins. All functionally characterized proteins containing a C 2 domain that binds Ca 2ϩ also act as Ca 2ϩ sensors, and no single C 2 domain protein dominates the Ca 2ϩ -dependent regulation of a cell. Thus, to fully understand the targets of Ca 2ϩsignaling in cells it is essential to characterize all principal C 2 domain proteins. Only a complete overview of the Ca 2ϩ -dependent properties of different C 2 domain proteins will provide insight into how Ca 2ϩ signaling works. In view of these considerations, we have searched for conserved C 2 domain proteins that might function as widely distributed Ca 2ϩ sensors and have focused on membrane-anchored C 2 domain proteins because these are most likely involved in membrane traffic.

MATERIALS AND METHODS
Sequence Analyses and Data Bank Searches-Various data banks of the National Center for Biotechnology Information (NCBI) and Celera Genomics were searched for multiple C 2 domain proteins, expressed sequence tags, splice variants in reported full-length sequences, and the genes of MCTPs using programs available on the NCBI web site with default settings. The MCTP cDNA sequences were submitted to Gen-Bank TM (accession numbers AY656715, AY656716, and AY656717).
Expression and Purification of Recombinant Proteins-The DNAs encoding the human MCTP1L-C 2 A (residues 243-391), MCTP1L-C 2 B (residues 451-598), MCTP1L-C 2 C (residue 608 -755), and MCTP2-C 2 C (residues 491-638) domains were PCR amplified, subcloned into pGEX-KG vector, and verified by DNA sequencing. All recombinant proteins were produced as bacterial GST fusion proteins, purified essentially as described (21), additionally treated with Benzonase, and washed with 20 mM CaCl 2 and high salt buffers to remove the bacterial contaminants that stick to C 2 domains and alter their properties (22).

Construction and Expression of Vectors Encoding Various EYFP Fusion Proteins of MCTP2 Fragments to Test the Functionality of Putative
Transmembrane Regions-pCMV-EYFP, pCMV-EYFP-MCTP-TM1 (MCTP2 residue 641-750; does not correspond to a physiological splice variant), pCMV-EYFP-MCTP-TM2 (MCTP2 residue 641-878 with residue 696 -736 deletion; junction sequence: STLRSTIAFA-deleted sequence-ESTDIDDEEDE, precisely corresponding to a splice variant observed in random cDNA sequences in GenBank TM ; see Fig. 2B), and pCMV-EYFP-MCTP-TM1-TM2 (MCTP2 residue 641-878) were generated by standard procedures and transfected into COS-7 cells using DEAE-dextran. After 2 days, cells were washed three times using phosphate-buffered saline, and 0.5 ml of buffer (50 mM HEPES-NaOH, pH 7.2, 100 mM NaCl, 4 mM sodium EGTA, 2 mM MgCl 2 , 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and a protease inhibitor mixture from Roche Diagnostics) was added to each 100-mm plate. Scraped cells were passed through the 27-gauge, 0.5-inch needle (BD Biosciences) five times, and the cell suspension was centrifuged at 500 ϫ g for 10 min to precipitate unbroken cells and nuclei. The supernatant was additionally centrifuged at 100,000 ϫ g for 1 h, and total membrane and cytosol fractions were separated. All fractions were adjusted to the same volume, and Western blotting using a polyclonal antibody against GFP (T3743) was performed.
Centrifugation phospholipid binding assays were carried out with purified soluble GST fusion proteins in buffer A (50 mM HEPES-NaOH, pH 6.8, 0.1 M NaCl, and 4 mM sodium EGTA). The C 2 domain GST fusion proteins were incubated with liposomes of defined phospholipid composition in buffer A containing variable amounts of CaCl 2 , SrCl 2 , or BaCl 2 to provide defined concentrations of free Ca 2ϩ , Sr 2ϩ , or Ba 2ϩ , respectively (calculated using EqCal for Windows software from Biosoft, Ferguson, MO). For liposome generation, dried phospholipids (obtained from Avanti Co.) were resuspended in buffer A containing 0.5 M sucrose and sonicated. The resulting "heavy" liposomes were then isolated by centrifugation. After the incubations of C 2 domains with liposomes, liposomes with bound C 2 domains were re-isolated by centrifugation essentially as described (14,23), and bound proteins were precipitated, resuspended in 30 l of 2ϫ SDS sample buffer, and analyzed by SDS-PAGE and Coomassie Blue staining.
Fluorescence Spectroscopy-Purified GST-C 2 domain proteins (3 mg in 1 ml of buffer B (40 mM Tris-HCl, pH 8, 0.1 M NaCl, and 0.5 mM sodium EGTA)) were cleaved with 10 units of thrombin (Amersham Biosciences) and reconstituted in buffer B by rotation for 18 h at room temperature, and complete cleavage was confirmed by SDS-PAGE. GST-protein was removed by three sequential incubations with 100 l of glutathione-Sepharose 4B (Amersham Biosciences) for 1 h at 4°C, and the absence of GST was confirmed by SDS-PAGE. C 2 domain proteins without GST were used without further purification for recordings of fluorescence spectra after ϳ1:1,000 dilution using buffer B. Fluorescence emission spectra were acquired in an LS55 luminescence spectrometer (PerkinElmer Life Sciences) with an excitation wavelength of 282 nm. For Ca 2ϩ titrations, aliquots of Ca 2ϩ were added to the cuvette from a concentrated stock solution, and fluorescence was monitored at the stated emission maximum wavelengths. The concentration of free Ca 2ϩ was calculated with the EqCal program, and the data were analyzed using GraphPad software. Reversibility of the Ca 2ϩ -induced increases in tryptophan fluorescence was tested by the addition of an excess of EGTA.
Cell Culture and Confocal Imaging-HEK293 cells were plated on polylysine-coated (1 mg/ml in 0.1 M borate buffer; Sigma) 18-mm diameter, 1.5 coverglasses (VWR Scientific), and maintained in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum at 37°C and 5% CO 2 . One day after plating, cells were transfected with the EYFP-MCTP2 fusion vectors using FuGENE-6 (Roche Applied Science). Two days after transfection, cover glasses with cells were incubated in phosphate-buffered saline containing 10 M FM5-95 (Molecular Probes) for 5 min on ice, and then mounted in an imaging chamber (Warner Instruments) for observation. Confocal images were acquired on a Leica TCS2 laser-scanning confocal microscope using a 100ϫ oil objective lens (numerical aperture, 1.3).
Cracked PC12 cell secretion assays were carried out with freeze-thaw permeabilization of PC 12 cells (13,23). 70% confluent PC12 cells were loaded with [ 3 H]norepinephrine for 24 h, washed with physiological saline (145 mM NaCl, 5.6 mM KCl, 2.2 mM CaCl 2 , 0.5 mM MgCl 2 , 5.6 mM glucose, and 15 mM HEPES-NaOH, pH 7.4), harvested by pipetting a stream of Ca 2ϩ -free ice-cold buffer C (120 mM potassium glutamate, 20 mM potassium acetate, 2 mM EGTA, and 20 mM HEPES-NaOH, pH 7.2), and washed twice with same buffer. Cell "ghosts" were prepared by freezing cells overnight at Ϫ80°C, thawing on ice for 2 h, and washing the ghosts 3ϫ with 6 ml of buffer C containing 1% bovine serum albumin. Standard secretion reactions (ϳ20 reactions per 100-mm plate; 0.1 ml of total volume in 1.5-ml tubes) contained washed cell ghosts, 2 mM ATP, 2 mM MgCl 2 , 10 l of rat brain cytosol (10 g/liter) in buffer C, and 6 M of recombinant C 2 domain protein with various concentrations of Ca 2ϩ to obtain the indicated free concentrations (calculated by EqCal). Reactions were incubated for 30 min at 30°C and terminated on ice, and the samples were centrifuged at 4°C for 3 min at 20,800 ϫ g. Supernatants and pellets solubilized in 1% Triton X-100 were analyzed by liquid scintillation counting.

RESULTS
Characterization of MCTP Sequences-In searching for proteins containing C 2 domains and transmembrane regions, we identified four classes of evolutionarily conserved proteins as follows: (i) synaptotagmins, which are expressed in at least 15 isoforms and are defined by the presence of a single N-terminal TMR and two C-terminal C 2 domains with characteristic sequence motifs (reviewed in Ref. 15); (ii) ferlins, which contain 3-6 C 2 domains and a C-terminal TMR (reviewed in Ref. 16); and (iii) two novel protein families that comprise 3-6 C 2 domains with either N-or C-terminal TMRs (see schematic drawing in Fig. 1). We refer to the proteins containing an N-terminal TMR and multiple cytoplasmic C 2 domains as E-Syts (for "extended synaptotagmins"), because the topology of these proteins resembles that of synaptotagmins. One member of the E-Syt family was described previously as an unnamed plasma membrane protein of adipocytes (26); these proteins will be examined in a later study. The present report will focus on the second class of novel C 2 domain proteins, namely the proteins referred to here as MCTPs (for multiple C 2 domain and TMR proteins"), because at least some splice variants of these proteins contain multiple TMRs in addition to the C 2 domains ( Fig.  1). We focused on MCTPs because RNA interference experiments in Caenorhabditis elegans revealed that the C. elegans MCTP homolog (1H206) is an essential gene whose ablation leads to early embryonic lethality (27), and we undertook a molecular characterization of MCTPs as a first step toward understanding their essential functions.
We assembled full-length sequences and cDNA clones for human MCTPs using commercially available expressed sequence tag clones and PCR on human cDNA. Data bank searches revealed that all vertebrates contain two MCTP genes with the same overall architecture and a high degree of sequence identity, whereas invertebrate animals (C. elegans and Drosophila melanogaster) contain a single MCTP gene. For comparison of MCTPs, we aligned the human MCTP1 and MCTP2 and the Drosophila and C. elegans MCTP sequences ( Fig. 2A). The Drosophila sequences were assembled from two separate predicted transcription units in the data bank (CG33148 and CG33146). Although Drosophila MCTP is predicted from the genome sequence to represent two separate genes, the precise colinearity of the Drosophila sequence with the human and C. elegans sequences ( Fig. 2A) and the corresponding Anopheles gambiae sequence (data not shown) indicates that this represents a single transcription unit.
The sequence alignments show that the N termini of MCTPs are highly variable; furthermore, both human MCTP1 and C. elegans MCTP are expressed with two alternative 5Ј-sequences (referred to as MCTP1L and MCTP1S and as MCTPL and MCTPS, respectively), possibly because the genes contain two independent promoters. Motif searches demonstrated that all MCTPs contain three C 2 domains (referred to as C 2 A, C 2 B, and C 2 C domains; shown in red in Fig. 2A) followed by two transmembrane regions (shown in gray). Only short linkers separate the three C 2 domains, whereas a longer region connects the C 2 domains to the TMRs (presumptive cytoplasmic non-C 2 domain sequences are shown in yellow in Fig. 2A). The sequence that connects the two TMRs is short and highly charged. All of the MCTP sequences except for the N-terminal region are highly conserved, with the highest degree of conservation of MCTPs being in the C 2 C domain, the second TMR, and the short C-terminal cytoplasmic sequence. The latter includes a 12residue stretch that is identical in all species and isoforms (NNELLDFLSRVP; Fig. 2A).
MCTPs Contain Two Functional TMRs-The proposed transmembrane topography for MCTPs ( Fig. 1) is based on the absence of an N-terminal signal peptide in the MCTP sequences and the fact that all known C 2 domains are cytoplasmic. With this topography, MCTPs are the only C 2 domain proteins that include more than one TMR. Interestingly, analysis of the MCTP sequences from expressed sequence tags reveals evidence of at least two sites of alternative splicing. First, the linker that separates the C 2 A and C 2 B domain varies in size dependent on alternative splicing. Second, we observed in multiple independent clones alternative splicing of the first TMR (Fig. 2B). As a result of this alternative splicing, some mRNAs for both MCTP1 and MCTP2 only encode the second TMR. If correct, this alternative splicing would convert the conserved C-terminal cytoplasmic sequence of MCTPs into a non-cytoplasmic sequence, i.e. turn the topography of the C terminus around (Fig. 1).
Although the predicted TMRs of MCTPs have all the features of classical TMRs, including a high degree of hydrophobicity, prediction of TMRs is not always accurate. To test whether each TMR is individually capable of anchoring MCTPs into a membrane, we generated constructs that encode EYFP fusion proteins of MCTP2 (Fig. 3). In these proteins, EYFP is fused to the C-terminal part of MCTP2 containing either both TMRs or only the first or second TMR. Of these proteins, the one carrying a deletion of the first TMR corresponded precisely to the sequence of the expressed sequence tag clones containing alternatively spliced MCTP variants that lack the first TMR (see Fig. 2B).
We first transfected the EYFP-MCTP2 fusion constructs into COS cells and examined whether the presence of either TMR would render the protein particulate as expected for a TMR protein (Fig. 3). Indeed, we found that EYFP itself was soluble, but when fused to MCTP2 fragments containing either one of the two TMRs it became particulate. We next examined the localization of the transfected EYFP-MCTP2 fusion proteins in transfected HEK293 cells. All three proteins were localized to intracellular vesicular structures, suggesting that each TMR by itself is competent to anchor the protein to membranes and indicating that MCTP2 may normally be a vesicular protein (Fig. 4).
Structure of the MCTP Genes-Using public and the Celera Genomics databases, we determined the organization and chromosomal localizations of the human MCTP genes. MCTP1 is encoded by a large gene (ϳ600 kb) on chromosome 5q15, whereas MCTP2 is encoded by a smaller gene (ϳ200 kb) on chromosome 15q26 (Fig. 5). The C-terminal half of the MCTPs, which exhibits the most sequence similarity, contains the same exon-intron organization ( Fig. 2A). Small differences in the number and placement of introns interrupting homologous se-FIG. 1. Structures of proteins containing multiple C 2 domains anchored to membranes by a transmembrane region. Four classes of proteins were identified in data bank searches, namely synaptotagmins (Syts), extended synaptotagmins (E-Syts; to be described in the future), ferlins, and multiple C 2 domain and TMR proteins or MCTPs. Proteins are depicted embedded in a membrane (green) by transmembrane regions (labeled TM) with presumptive non-cytoplasmic sequences colored blue and cytoplasmic sequences colored red (for the C 2 domains) and yellow (for all other cytoplasmic sequences). N and C termini are indicated. In the case of the MCTPs, two major Cterminal splice variants result in the same sequence (shown in purple), being either cytoplasmic (when two TMRs are present) or non-cytoplasmic (when one of the two TMRs is spliced out). C 2 domains were assigned based on the conserved domain data base of the NCBI; some proteins, especially ferlins, may have additional unpredicted C 2 domains that do not precisely fit the consensus sequence as well as alternative transcripts with fewer C 2 domains. and yellow letters on a purple background, C-terminal sequence. Sequences are numbered on the right; note that the C. elegans sequence includes a 66-residue insertion at amino acid number 535 that is missing from the other species. Alternatively spliced sequences in the linker between the C 2 A and C 2 B domains are underlined. B, alternative splicing of the C-terminal TMRs of MCTPs. The full-length variant and two variants for MCTP1 (GenBank TM accession number AK057694 and AK058012) and one variant for MCTP2 (GenBank TM accession number R98750) are aligned. Note that the splice variants truncate TMR1 but leave TMR2 intact, thereby causing an inversion of the topology of the C terminus (see Fig. 1).
quences are observed in the N-terminal half of the proteins (Tables I and II). The different N-terminal sequences observed in MCTP1L and MCTP1S are encoded by distinct exons (referred to as 1a and 1b) that are separated by 200 kb in the genome (Fig. 5). Most introns disrupt the coding sequence of the MCTPs at the same position, most often the "0" position that lies precisely between individual codons. The gene organization explains the alternative splicing that we observed. The alternative splicing of sequences separating the C2A and C2B domains is due to the presence or absence of exon 7 in MCTP1 and MCTP2, and that of the first TMR is due to the presence or absence of exon 18 in MCTP1 and exon 17 in MCTP2.
Tissue Distribution of MCTPs-Data bank analysis revealed that the expressed sequence tags encoding MCTP1 and 2 were isolated from a large number of tissues such as brain, bone marrow, pancreas, spleen, thymus, placenta, blood vessel, and kidney. To test if the protein encoded by the MCTP mRNAs is actually synthesized and determine which tissues contain the highest steady-state levels of these proteins, we produced two antibodies against recombinant proteins derived from MCTP1 and MCTP2 (Fig. 6A). Immunoblotting of various rat tissues revealed that MCTP1 is highly expressed in skeletal muscle and to a lesser degree in heart muscle, whereas MCTP2 was primarily detectable in heart muscle and testis (Fig. 6B). The immunoblotting results for MCTP1 are likely to be an accurate reflection of the protein levels because two different antibodies, raised to non-overlapping MCTP1 epitopes, gave the same results, whereas only one of the two MCTP2 antibodies was usable. MCTP1 in heart muscle appears to be slightly smaller than in skeletal muscle, possibly because of alternative splicing. Because of the limited sensitivity of the antibodies, the Constructs encoding the proteins schematically diagrammed on the left were transfected into HEK293 cells and analyzed by subcellular fractionation and immunoblotting as shown on the right. In the constructs containing only a single TMR of MCTP2, the junctions were designed to correspond precisely to alternatively spliced variants that could arise by the inclusion or exclusion of in-frame exons (see gene structures described below). Such variants were detected in expressed sequence tag data banks for MCTP1 and MCTP2 lacking TM1 (see Fig. 2B). Transfected cells were lysed and separated into soluble and particulate fractions by centrifugation and analyzed by immunoblotting for EYFPs as shown on the right, using EYFP alone as a control. Fig. 3 were expressed in HEK293 cells, and the cells were labeled with FM5-95 to selectively stain the plasma membrane. Images shown were obtained in a confocal microscope to visualize the FM5-95 fluorescence (left column, red) and EYFP fluorescence (middle column, green). Two examples are shown for each construct. Scale bar at the bottom of the right column (Merged) (2 m) applies to all sections. results do not exclude the possibility of a low amount of expression of MCTP1 in non-muscle tissues. Nevertheless, the results clearly demonstrate that by far the highest levels of MCTPs are present in excitable muscle cells and that MCTPs are not enriched in brain, another tissue with a large number of excitable cells.

FIG. 4. Localizations of EYFP-MCTP2 fusion proteins in transfected HEK293 cells. EYFP-MCTP2 fusion proteins encoded by the constructs described in
Structure of the MCTP C 2 Domains-Because the three closely spaced C 2 domains are the major feature of the cytoplasmic sequences of MCTPs, these domains presumably determine the function of these proteins. RNA interference experiments in C. elegans revealed that the MCTP homolog (1H206) is an essential gene and that its ablation leads to early the numbers above each arrow designate the location in the codon where the intron disrupts the coding sequence (0, intron is exactly between codons; 1, intron is after first nucleotide of a codon; 2, intron is after the second nucleotide of a codon). Note that most exons in the MCTP genes are flanked by introns that disrupt the codons in the same frame; these exons can, in principle, be alternatively spliced without a loss of reading frame. Precise locations of exons and sequences of exon/ intron junctions are shown in Tables I  and II. aa, amino acids.   (27). Most, but not all, C 2 domains function as Ca 2ϩ -binding modules (reviewed in Ref. 28). As a first approach toward understanding the function of MCTPs, we therefore studied its C 2 domains. Specifically, we investigated the two major activities associated with C 2 domains, namely Ca 2ϩ binding and interaction with phospholipid membranes. Alignment of the C 2 domain sequences of MCTPs reveals that all three C 2 domains belong to class 2 C 2 domains. Class 1 and 2 C 2 domains have similar structures composed of an eight-stranded ␤-sandwich but differ in strand topology. In class 2 C 2 domains, the ␤-strand corresponding to the first ␤-strand of class 1 C 2 domains is transplanted to the end of the C 2 domain, making this the last ␤-strand (referred to as ␤1Ј; see Fig.  5). As a result, the topology of class 2 C 2 domains represents a circular permutation of the topology of class 1 C 2 domains (28), with the N-and C termini being on the top of the domain in class 1 C 2 domains and at the bottom of the domain in class 2 C 2 domains. Although all MCTP C 2 domains belong to class 2, they are otherwise not very similar (Fig. 7). The C 2 A, C 2 B, and C 2 C domains are well conserved between various MCTP isoforms, but the degree of identity between C 2 domains is low.
Most C 2 domains bind Ca 2ϩ with a low intrinsic binding affinity in the absence of phospholipid membranes but with a high apparent affinity in the presence of phospholipid membranes. This behavior probably results from the fact that the coordination spheres for bound Ca 2ϩ ions are incomplete in these C 2 domains in the absence of phospholipids but are completed by the phospholipid head groups. The Ca 2ϩ binding mode of C 2 domains has been best characterized for the C 2 domain from protein kinase C (29,30), the C 2 A and C 2 B domains of synaptotagmin 1 (5,14,22), and the C 2 domain from phospholipase C␦ (31). In all of the C 2 domains studied, Ca 2ϩ is bound exclusively by the loops emerging from the top of the ␤-sandwich. Ca 2ϩ is coordinated by five aspartate, glutamate,  1  9807022-9807486  465  5Ј-UTR ѧѧѧѧѧATG,GAT,CTG,GAT,-GAA,GAG,CCA,GAG,gtgagaataggg  2  9823038-9823100  63  ttctttgcagAAG,CTA,TGT,GGA,-GAA,GAG,CCA,GAG,gtgagtggcatt  3  9824285-9824393  109  atctgtgcagGTA,CCG,GGG, used as a loading control). Positions of molecular mass markers are shown on the left. Note that the two independent MCTP1 antibodies result in exactly the same pattern of reactivity on the multitissue immunoblots despite being raised to distinct epitopes, suggesting that MCTP1 is primarily expressed in skeletal muscle and, to a lesser extent, in heart. The second MCTP2 antibody was not successful (not shown).
or asparagine residues that are present in the top loops. All of these residues are conserved in the MCTP C 2 domains, including the Drosophila and C. elegans MCTP C 2 domains, except for the C 2 B domain of MCTP2, which lacks two of the five Ca 2ϩbinding residues (Fig. 7). In addition, the other conserved residues of the Ca 2ϩ -binding loops of C 2 domains are also retained in the MCTP domains (e.g. the typical GXSD sequence of the first top loop that is also present in all synaptotagmins) (28), indicating that these domains have the expected features of Ca 2ϩ /phospholipid binding domains. However, recent results (32) revealed that the Ca 2ϩ binding properties of C 2 domains cannot be predicted from sequence analyses because the C 2 B domain of synaptotagmin 4, despite a perfect Ca 2ϩ -binding consensus sequence, exhibits no intrinsic or phospholipid-dependent Ca 2ϩ -binding. This raises the question of whether MCTPs indeed function as Ca 2ϩ -binding molecules.
Phospholipid Binding Properties of the MCTP C 2 Domains-We first tested whether the MCTP C 2 domains bind to phospholipids similarly as other C 2 domains. We produced all three C 2 domains of MCTP1 and the C 2 C domain of MCTP2 as recombinant GST fusion proteins. We then tested the binding of these domains to liposomes with five different phospholipid compositions that included neutral phospholipids, phosphatidylinositol phosphates, and negatively charged phospholipids (Fig. 8). Binding was examined in the presence of EGTA, Ca 2ϩ , Sr 2ϩ , and Ba 2ϩ using a sensitive centrifugation assay that detects Ca 2ϩ -dependent phospholipid binding by measuring the amount of C 2 domain protein that can be isolated with liposomes in the presence of Ca 2ϩ but not EGTA (23). This binding assay is important in assessing the ability, or lack thereof, of C 2 domains to interact with phospholipids, because the pull-down assay that we originally developed (4) and has been widely adopted does not detect weaker interactions such as the binding of the synaptotagmin C 2 B domain (14). Surprisingly, using the centrifugation assay we detected no significant Ca 2ϩ -dependent binding of any MCTP C 2 domain to any of the phospholipid membranes (Fig. 8). Some weak binding was observed, especially for the C 2 C domain of MCTP1, but this binding never reached the level observed with typical phospholipid-binding C 2 domains. This result indicates that MCTPs are not phospholipid-binding molecules for either charged or neutral phospholipids, raising the question of whether they are at all involved in Ca 2ϩ -binding (32).

Intrinsic Ca 2ϩ
Binding to MCTP C 2 Domains-To monitor Ca 2ϩ binding to MCTP1 C 2 domains, we purified the recombinant C 2 domains as GST fusion proteins, cleaved them off the GST moiety with thrombin, and examined their intrinsic tryptophan fluorescence as a function of Ca 2ϩ (Fig. 9). This experiment was prompted by the observation that all MCTP C 2 domains contain a tryptophan in the middle of ␤-strand 5 (see Fig. 7) and that the C 2 B domains additionally contain a tryptophan in top loop 3 and in the bottom N-terminal sequence.
The tryptophan fluorescence spectra of the C 2 domains exhibited characteristic differences in the number of fluorescence maxima and their emission wavelength. In every C 2 domain, however, the addition of a saturating concentration of Ca 2ϩ increased the intrinsic tryptophan fluorescence in a manner that was fully reversible upon the addition of excess EGTA (Fig. 9). The Ca 2ϩ -induced fluorescence increase was quite large (Ͼ10%) for the C 2 A and the C 2 B domains and smaller for the MCTP1 C 2 C domain. For this reason we also studied the C 2 C domain from MCTP2, because the C 2 C domain of this isoform contains an extra tryptophan (Fig. 7), and we included this domain in all other assays as well. Indeed, the MCTP2 C 2 C domain does exhibit a significantly larger Ca 2ϩ -dependent fluorescence increase (Fig. 9). Ca 2ϩ induced no shift in emission maxima, and Mg 2ϩ had no effect on any tryptophan fluorescence property of the C 2 domains (data not shown).
We next exploited the tryptophan fluorescence changes to titrate Ca 2ϩ binding (Fig. 10). Saturable Ca 2ϩ -dependent increases in tryptophan fluorescence were observed for all C 2 domains, with half-maximal changes between 1.3 and 2.3 M free Ca 2ϩ . The signal-to-noise ratio was robust for the C 2 A and C 2 B domains, but the changes in the C 2 C domains were rather small (Fig. 10). However, multiple independent experiments provided similar results, indicating that the changes observed correspond to a reliable Ca 2ϩ binding event.
Effect of Purified MCTP C 2 Domains on Ca 2ϩ -triggered Secretion from Permeabilized PC12 Cells-The properties of the MCTP C 2 domains, high affinity intrinsic Ca 2ϩ binding and no phospholipid binding, are unexpected. The apparent Ca 2ϩ affinity of the C 2 domains approximately corresponds to that of secretion from PC12 cells. In previous studies we have shown that C 2 domains with an apparent high Ca 2ϩ affinity in phospholipid complexes potently inhibit Ca 2ϩ -induced exocytosis from permeabilized PC12 cells (23). However, for traditional C 2 Residues that are shared among at least three of the six sequences are highlighted with a color code that reflects the predicted secondary structure of the C 2 domains as follows: blue, ␤-strands (labeled on top; note that because all of the C 2 domains are class 2 C 2 domains, the numbering starts with ␤2 and ends with ␤1Ј); yellow, top loops; green, bottom loops. In addition, presumptive Ca 2ϩ -binding ligands are marked by a white typeface on a black background, and tryptophans that contribute to the intrinsic fluorescence of the domains are shown on a red background.
domains, altering phospholipid binding always leads to a change in apparent Ca 2ϩ affinity, because phospholipid binding and Ca 2ϩ binding are interdependent (33). Because the MCTP C 2 domains have the requisite Ca 2ϩ affinity to interfere with PC12 cell exocytosis but do not bind phospholipids, they may still function in Ca 2ϩ -dependent exocytosis. To address this possibility, we compared the effect of purified MCTP C 2 domains with those of synaptotagmin 7 C 2 domains (which are potent inhibitors of exocytosis) on Ca 2ϩ -induced secretion in permeabilized PC12 cells (Fig. 11). We observed a strong inhibitory effect of the synaptotagmin 7 C 2 domains on secretion but no effect by the MCTP C 2 domains. These results indicate that consistent with the tissue distribution of MCTPs (Fig. 6), MCTPs are not components of the secretory machinery. Furthermore, these results show that for inhibition in the permeabilized PC12 cell assay, a high Ca 2ϩ binding affinity is not sufficient for an effect. DISCUSSION In the present study we describe a novel family of C 2 domain proteins that exhibit unusual properties indicative of a role in FIG. 8. Lack of Ca 2؉ -dependent phospholipid binding by MCTP C 2 domains. Purified recombinant C 2 domain GST fusion proteins were tested with liposomes of the composition described below each set of gel images (PE, phosphatidylethanolamine; PC, phosphatidylcholine; PIP 2 , phosphatidylinositol bisphosphate; PIP, phosphatidylinositol phosphate; PS, phosphatidylserine). C 2 domains were incubated with the liposomes in the presence of the concentration of divalent cations indicated at the top. Afterward, liposomes were isolated by centrifugation through a sucrose cushion, and proteins bound to the liposomes were examined by SDS-PAGE and Coomassie Blue staining. Images shown are Coomassie Blue-stained gels from a representative experiment repeated three times. Note that even very sticky, highly charged liposomes (e.g. liposomes containing 5% PIP2) were unable to bind MCTP C 2 domains.
FIG. 9. Fluorescence spectra of purified recombinant C 2 domains from MCTPs and Ca 2؉ -dependent changes. Intrinsic fluorescence spectra were recorded of 3 M purified C 2 domains lacking the GST moiety in 40 mM Tris-HCl, pH 8, 0.1 M NaCl, and 0.5 mM sodium EGTA at an excitation wavelength of 282 nm. After the initial spectra were measured, a second spectrum of the same sample was obtained upon the addition of 100 M free Ca 2ϩ , and a third spectrum was recorded upon the further addition of 1 mM sodium EGTA. Data shown are from a representative experiments independently repeated multiple times. Ca 2ϩ signaling. These properties, which differentiate MCTPs from other previously studied C 2 domain proteins, are described in the next two paragraphs.
First, the architecture of the MCTPs is unique in that they are composed of a large, presumptively cytoplasmic sequence primarily composed of three class 2 C 2 domains and two TMRs. With the MCTPs there are now four families of membraneanchored multiple C 2 domain proteins (Fig. 1). MCTPs are the only membrane-bound C 2 domain proteins that contain two functional TMRs (Figs. 3 and 4).
Second, the functional properties of MCTPs are unique in that they bind Ca 2ϩ but not phospholipids. We only showed this for MCTP1 and for the C 2 C domain of MCTP2, but the C 2 A domain of MCTP2 is highly homologous to that of MCTP1 and thus is likely to also bind Ca 2ϩ . The C 2 B domain of MCTP2, however, lacks the consensus sequence for C 2 domain Ca 2ϩbinding sites (Fig. 7) and is unlikely to bind Ca 2ϩ . Invertebrate MCTPs from C. elegans and D. melanogaster also contain all Ca 2ϩ -binding site sequences ( Fig. 2A), consistent with the notion that Ca 2ϩ binding is evolutionarily conserved and that the C 2 B domain of MCTP2 is the only C 2 domain of an MCTP that does not bind Ca 2ϩ .
With these properties, MCTPs are veritable Ca 2ϩ -binding machines in which three Ca 2ϩ -binding C 2 domains are attached to the membrane. The high affinity of all three MCTP1 C 2 domains for Ca 2ϩ without phospholipid binding was unexpected. We recently showed that the C 2 B domain of synaptotagmin 4 contains a perfect predicted Ca 2ϩ -binding site with a high degree of sequence identity to that of the C 2 B domain of synaptotagmin 1, but, nevertheless, it does not bind Ca 2ϩ (32); thus, the presence of a predicted Ca 2ϩ -binding site is not sufficient to deduce the Ca 2ϩ binding properties of a C 2 domain. In extension of this conclusion, the current data indicate that the presence of a consensus Ca 2ϩ -binding site in a C 2 domain also does not allow the prediction of the mode of Ca 2ϩ binding. The standard mode for C 2 domains is to bind Ca 2ϩ in a complex with phospholipids in which the phospholipid head groups are essential to complete the coordination spheres of the bound Ca 2ϩ ions (33). The present data show that this mode of Ca 2ϩ binding is not universally true in that the MCTP Ca 2ϩ -binding sites must have reasonably complete coordination spheres for Ca 2ϩ in the absence of phospholipids. The sequences of the MCTP C 2 domains, however, provide no clue as to how this might work. It is possible that in the top loops of the C 2 domains, negatively charged residues, in addition to the canonical five aspartates, contact the Ca 2ϩ ions. However, no conserved, additional negatively charged residues are present that could fulfill this role. Alternatively, it is possible that Ca 2ϩ ions bind between two C 2 domains, with the coordination sphere formed by residues from the top loops of two different C 2 domains.
The Ca 2ϩ binding properties and structures of MCTPs demonstrate that these proteins function in Ca 2ϩ signaling at the membrane. Elucidating the nature of this function will require a genetic approach that builds on the observation that in C. elegans MCTP is an essential gene whose ablation causes embryonic lethality (27). The essential nature of MCTP is consistent with an important role in Ca 2ϩ signaling that could, in FIG. 10. Ca 2؉ titration of the tryptophan fluorescence of purified recombinant C 2 domains from MCTPs. Intrinsic fluorescence of the indicated purified C 2 domains from MCTPs (3 M) was monitored as function of the free Ca 2ϩ concentration (excitation, 282 nm; emission of the MCTP1 C 2 A domain, 332 nm; emission of the MCTP1 C 2 B domain, 344 nm; emission of the MCTP1 C 2 C domain, 328 nm; emission of the MCTP2 C 2 C domain, 328 nm). Ca 2ϩ concentrations were clamped with Ca 2ϩ -EGTA buffers (see "Materials and Methods"). Data shown are means Ϯ S.D. (n ϭ 3 for C 2 A and C 2 B domains of MCTP1; n ϭ 7 for C 2 C domains from MCTP1 and MCTP2; each experiment was performed in triplicate). The numbers indicated display the Ca 2ϩ concentration that caused a half-maximal fluorescence change, as calculated by curve fitting using GraphPad PRISM version 3.02 software.
FIG. 11. Effect of MCTP C 2 domains on Ca 2؉ -induced exocytosis from permeabilized PC12 cells. PC12 cells were loaded with 3 H-labeled norepinephrine, cracked, and preincubated with 6 M purified C 2 domain GST fusion proteins. Exocytosis was measured as the amounts of norepinephrine released during a 30-min incubation period after the addition of a buffer containing no Ca 2ϩ or 10 M Ca 2ϩ . Data shown are means Ϯ S.D. (n ϭ 3; each experiment was performed in duplicate).
principle, consist of a Ca 2ϩ -controlled regulatory function or an activity as a Ca 2ϩ buffer. Although a Ca 2ϩ buffer function cannot be excluded, a Ca 2ϩ regulatory function appears more likely. With the generally fast Ca 2ϩ binding, the relatively low Ca 2ϩ affinities, and the obligatory formation of phospholipid complexes by C 2 domains (reviewed in Ref. 28), C 2 domain proteins are more suited than EF-hand proteins for some Ca 2ϩ regulatory functions and less suited for Ca 2ϩ buffering functions. In the case of MCTPs, however, this assumption may not apply because of their relatively high Ca 2ϩ affinity in the absence of phospholipids. In addition, MCTPs are exceptional among C 2 domain proteins because they contain two TMRs, one of which (TMR1) may be spliced in and out in both isoforms (Fig. 2B). Alternative splicing that removes the first TMR would retain the membrane-bound nature of MCTPs, because the second TMR is sufficient to anchor the cytoplasmic C 2 domains to the membrane (see Figs. 3 and 4) but would convert the normally cytoplasmic C-terminal sequence of MCTP into an extracellular sequence (see Fig. 1).
It is interesting that of the four families of proteins that contain both C 2 domains and a TMR, the two families that have been functionally studied (synaptotagmins and ferlins) are involved in membrane fusion (reviewed in Refs. 15 and 16). Members of the third family, referred to here as extended synaptotagmins or E-Syts because of their structural similarity to synaptotagmins (Fig. 1), have not been studied beyond the cloning of one of their isoforms, and the fourth family consists of the MCTPs examined here. The fact that synaptotagmins and ferlins are involved in membrane traffic strongly supports the notion that MCTPs might also act in membrane traffic. Furthermore, the presence of two evolutionarily conserved TMRs, when one would have been sufficient to anchor the C 2 domains to the membrane, and the presence of three C 2 domains with distinct conserved sequences argue against a Ca 2ϩ buffer function. Again, genetic experiments will have to address this important issue.