Lipid Binding Ridge on Loops 2 and 3 of the C2A Domain of Synaptotagmin I as Revealed by NMR Spectroscopy*

The C2A domain of synaptotagmin I, which binds Ca2+ and anionic phospholipids, serves as a Ca2+ sensor during excitation-secretion coupling. We have used multidimensional NMR to locate the region of C2A from rat synaptotagmin I that interacts, in the presence of Ca2+, with phosphatidylserine. Untagged, recombinant C2A was double-labeled with 13C and15N, and triple-resonance NMR data were collected from C2A samples containing either Ca2+ alone or Ca2+ plus 6:0 phosphatidylserine. Phospholipid binding led to changes in chemical shifts of backbone atoms in residues Arg233 and Phe234 of loop 3 (a loop that also binds Ca2+) and His198, Val205, and Phe206 of loop 2. These residues lie along a straight line on a surface ridge of the C2A domain. The only other residue that exhibited appreciable chemical shift changes upon adding lipid was His254; however, because His254 is located on the other side of the molecule from the phospholipid docking site defined by the other residues, its shifts may result from nonspecific interactions. The results show that the “docking ridge” responsible for Ca2+-dependent membrane association is localized on the opposite side of the C2A domain from the transmembrane and C2B domains of synaptotagmin.

The C 2 A domain of synaptotagmin I, which binds Ca 2؉ and anionic phospholipids, serves as a Ca 2؉ sensor during excitation-secretion coupling. We have used multidimensional NMR to locate the region of C 2 A from rat synaptotagmin I that interacts, in the presence of Ca 2؉ , with phosphatidylserine. Untagged, recombinant C 2 A was double-labeled with 13 C and 15 N, and triple-resonance NMR data were collected from C 2 A samples containing either Ca 2؉ alone or Ca 2؉ plus 6:0 phosphatidylserine. Phospholipid binding led to changes in chemical shifts of backbone atoms in residues Arg 233 and Phe 234 of loop 3 (a loop that also binds Ca 2؉ ) and His 198 , Val 205 , and Phe 206 of loop 2. These residues lie along a straight line on a surface ridge of the C 2 A domain. The only other residue that exhibited appreciable chemical shift changes upon adding lipid was His 254 ; however, because His 254 is located on the other side of the molecule from the phospholipid docking site defined by the other residues, its shifts may result from nonspecific interactions. The results show that the "docking ridge" responsible for Ca 2؉ -dependent membrane association is localized on the opposite side of the C 2 A domain from the transmembrane and C 2 B domains of synaptotagmin.
Neuronal exocytosis is strictly regulated by Ca 2ϩ ions (1). Ca 2ϩ triggers the fusion of synaptic vesicles with the presynaptic plasma membrane on the submillisecond time scale; this sets an aggregate upper limit for the t1 ⁄2 values of protein conformational changes that couple Ca 2ϩ binding to exocytotic membrane fusion (2, 3). Although several proteins that function in exocytosis have been identified (4 -6), little is known concerning the Ca 2ϩ -driven conformational changes or proteinlipid interactions that catalyze lipid bilayer fusion. The key to unraveling this mechanism lies in the Ca 2ϩ sensor for exocytosis. Recent gene disruption experiments (7-10) have established that the Ca 2ϩ -binding synaptic vesicle protein, synaptotagmin I (11)(12)(13), is essential for rapid and efficient Ca 2ϩtriggered release of neurotransmitters. Thus, synaptotagmin has been proposed to function as the major Ca 2ϩ sensor of regulated exocytosis.
The 12 known members of the synaptotagmin gene family (14 -16) span the vesicle membrane once, have a short aminoterminal intravesicular domain and a large cytoplasmic region that contains two C 2 domains, designated "C 2 A" 1 and "C 2 B", which interact with a variety of distinct molecules (14). The C 2 A domain mediates the interaction of synaptotagmin with anionic phospholipids (17,18), but both C 2 domains are required for high affinity binding to plasma membrane proteins syntaxin (19) and SNAP-25. 2 These interactions are regulated by Ca 2ϩ (20) at calcium ion concentrations similar to those required for neuronal exocytosis (21). Syntaxin and SNAP-25 form a complex with the synaptic vesicle protein synaptobrevin that can catalyze membrane fusion in vitro (22) and in vivo (23).
C 2 domains are found in more than 50 distinct proteins, including lipid and serine/threonine kinases, phospholipases, GTPase-activating proteins, proteins involved in ubiquitin-mediated protein degradation and cytolic pore formation, and a number of proteins involved in membrane traffic (24). Despite the growing number of proteins that possess C 2 domains, relatively little is known concerning the structural elements that mediate C 2 domain-effector interactions. The crystal structures of three distinct C 2 domains have been determined (25)(26)(27), and the fold of each is an eight-stranded ␤-sandwich. In each, three flexible loops (designated loops 1-3) protrude from one end of the domain. Loops 1 and 3 contain the metal-binding ligands that serve to bind a maximum of two or three Ca 2ϩ ions.
A number of C 2 domains bind membranes in a Ca 2ϩ -dependent manner (17, 18, 28 -35). Because the Ca 2ϩ -dependent interaction of the C 2 A domain of synaptotagmin with membranes is likely to serve as an important step in excitation-secretion coupling, we have used multidimensional, multinuclear NMR spectroscopy to investigate the structural basis for this inter- action. We report that short acyl chain phosphatidylserine interacts exclusively with residues in loops 2 and 3 of C 2 A. Loops 2 and 3 become fully ordered only after phospholipid is bound, not simply upon binding Ca 2ϩ .

EXPERIMENTAL PROCEDURES
Plasmid Construction, Protein Expression, and Purification-cDNA encoding rat synaptotagmin I (12) was kindly provided by T. C. Sü dhof (Dallas, TX). The region of the gene coding for the C 2 A domain (residues 140 -267) of synaptotagmin I was amplified by PCR, creating NdeI (5Ј) and BamHI (3Ј) sites. This PCR product was digested with NdeI and BamHI and inserted into a pET9a vector (Novagen, Madison, WI), which had been cut previously with the same two enzymes. The resulting plasmid, named pET9a/C 2 A, was transformed into BL21(DE3) containing pLysS. A single colony picked from a plate was used to inoculate 50 ml of minimal (LB) medium; following overnight growth, this was used to inoculate 1 liter of LB medium. The cells were induced when A 600 (absorbance at 600 nm) reached 1.0 -1.3 and harvested by centrifugation 3 h later. Protein labeled uniformly with 13 C and 15 N was produced from 1 liter of minimal medium containing [ 13 C]glucose and [ 15 N]ammonium chloride (Isotec, Miamisburg, OH). The harvested cells were resuspended in 10 mM Tris buffer at pH 8.5 and subjected to a freeze-and-thaw step to break the cell walls followed by a sonication step. The cell lysate was centrifuged at 30,000 ϫ g for 15 min, and the supernatant was retained, filtered, and loaded onto a Mono Q column (Amersham Pharmacia Biotech) on an FPLC system (Amersham Pharmacia Biotech). C 2 A eluted around 150 mM NaCl, and the fractions containing C 2 A were pooled, concentrated, and loaded onto a Superdex 75 (Amersham Pharmacia Biotech) column. C 2 A eluted around 96 ml. The purity was tested by SDS-polyacrylamide gel electrophoresis and Coomassie Blue staining. The typical yield of C 2 A was around 40 mg/liter of culture.
NMR Spectroscopy and Data Processing-All data were collected on a Bruker (Billerica, MA) DMX-500 NMR spectrometer at 303 K. For both samples, three-dimensional HNCA (36, 37) and HNCOCA (38) spectra were collected as 1024*( 1 H) ϫ 64*( 13 C) ϫ 48*( 15 N) and FIG. 1. Two-dimensional ( 1 H-15 N) HSQC spectrum (at 500 MHz 1 H) of the isolated C 2 A domain of rat synaptotagmin I (residues 140 -267) at pH 7.5. The spectrum (thick lines) of the ternary complex, C 2 A⅐(Ca 2ϩ ) x ⅐(6PS) y , is overlaid on top of that (thin lines) of the binary complex, C 2 A⅐(Ca 2ϩ ) x . Resonances exhibiting appreciable shift changes are labeled with amino acid type and residue number. 1024*( 1 H) ϫ 46*( 13 C) ϫ 48*( 15 N) matrices, respectively, where the number followed by an asterisk indicates the number of complex data points. Raw data were converted and processed with FELIX (Molecular Simulations, Inc., San Diego, CA) software. All data were digitally filtered to remove negative data points. The convolution difference solvent suppression routine in FELIX was applied, and the resulting data were apodized by a square-cosine window function, Fourier-transformed, and phase-corrected. The size of the data was halved by remov-ing the right half of the spectrum. To data in the 13 C dimension, a square-cosine window function was applied; data in this dimension were zero-filled to 256 points, Fourier-transformed, and phased. To data in the 15  University of California, San Francisco (http://www.cgl.ucsf.edu/Research/Sparky.html) was used for peak-picking.

RESULTS AND DISCUSSION
Sequence-specific Resonance Assignments-Sequential assignments were determined from the combined interpretation of HNCA (36,37) and HNCOCA (38) spectra. Resonances from residues Met 173 , Gly 174 , Asp 188 , Lys 213 , Arg 233 , and Phe 234 were not observed in spectra of the binary C 2 A⅐(Ca 2ϩ ) x complex, probably because of line broadening resulting either from hydrogen exchange with solvent water or from dynamic disorder (25). Signals from these residues were identified, however, in NMR spectra of the C 2 A⅐(Ca 2ϩ ) x ⅐(6PS) y ternary complex, suggesting that loops 2 and 3 of C 2 A are more ordered in the ternary complex than in the binary complex.
6PS Versus PS/PC Liposomes-The C 2 A domain of synaptotagmin exhibits preferential interactions with liposomes that contain anionic phospholipids, such as phosphatidylserine (17,18). 6PS was used here because of its high solubility, which made it possible to saturate the lipid binding sites of C 2 A while keeping the free phospholipid concentration below its critical micellar concentration. The relative concentrations used here (C 2 A:Ca 2ϩ :6PS ϭ 1:5:5) were chosen on the basis of results suggesting that the maximum number of metal ions bound is three (26) and that the number of phospholipid molecules is limited (29). Lipid binding was followed by following chemical shifts changes in the two-dimensional 1 H-15 N HSQC spectrum of [ 15 N]C 2 A as a function of added 6PS (Fig. 1).
As a test of the relevance of 6PS as a model for longer chain phosphatidylserine, it was determined 3 that, in the presence of Ca 2ϩ , 6PS induces the same level of resistance to limited proteolysis as PS/PC liposomes (made with PS derived from brain with acyl chain composition: ϳ40% 18:0, ϳ31% 18:8, ϳ8.5% 22:6) (17). This suggests that, in the presence of Ca 2ϩ , 6PS induces the same kind of structural rearrangement (or protection from proteolysis) produced by the PS/PC liposomes.
Comparison of the Binary and Ternary Complexes-To maximize the chemical shift comparisons (Fig. 2), two gaps in the data at pH 7.5 (Arg 233 and Phe 234 , whose signals were not seen in the binary complex but sharpen up in the ternary complex) were filled by chemical shifts for the binary complex at pH 5.0 where the peaks are sharper (39). Because the referencing system used by these authors (39) was different from that used here (40), their data set was first brought into register by minimizing the shift differences of signals common to both data sets.
Of the residues (Fig. 2) showing the largest shifts upon addition of 6PS (His 198 , Val 205 , Phe 206 , Arg 233 , Phe 234 , His 254 ), all but His 254 are on loops 2 and 3. Because His 254 is located on the opposite side of the molecule from loops 2 and 3, its shift likely results from a nonspecific interaction with 6PS or possibly a 6PS-mediated protein-protein interaction. Consistent with this interpretation, a fluorescent reporter (tryptophan) placed at residue 153, which lies near His 254 in the threedimensional structure of C 2 A, was found not to interact with lipid bilayers (41). Evidence for the involvement of Ca 2ϩ -binding loop 3 in lipid binding is consistent with the fluorescence study of Chapman and Davis (41), which indicated that residue 234 interacts directly with liposomes. The residues on loops 2 and 3 implicated in lipid binding (His 198 3. Molscript (44) representations of x-ray structures of C 2 domains highlighting residues of the synaptotagmin C 2 A domain (or their homologues in the other two structures; see Table I) whose backbone atoms exhibited large 1 H, 13 C, or 15 N NMR chemical shift changes upon addition of semisynthetic 6:0 phosphatidylserine in the presence of Ca 2؉ . A, x-ray structure of the rat synaptotagmin I C 2 A⅐Ca 2ϩ complex (25). B, comparison of three C 2 domain structures: top, C 2 A domain of rat synaptotagmin I with one bound Ca 2ϩ ions (25); middle, C 2 domain of human cytosolic phospholipase A 2 with two bound Ca 2ϩ ions (27); bottom, C 2 domain of rat phospholipase C ␦1 with two bound Ca 2ϩ ions (26).
Phe 234 ) are oriented along a straight line on a surface ridge of C 2 A (Fig. 3A). An important finding of the present work is that this "docking ridge" is located on the opposite side of the C 2 A domain from the transmembrane and C 2 B domains of synaptotagmin I.
Results reported here show that the phospholipid binding site(s) of C 2 A are localized to loops 2 and 3. In contrast, the largest shifts that occurred upon the binding of C 2 A to a fragment of syntaxin were localized to Ca 2ϩ -binding loops 1 and 3, with only subtle shifts in the backbone of loop 2 (42). Asp 238 , a Ca 2ϩ ligand in loop 3 and the residue of C 2 A that exhibited the largest shift perturbation on forming the C 2 A⅐syntaxin-fragment complex, did not shift significantly upon binding 6PS (Fig. 2). It is localized to a region of loop 3 distinct from the proposed membrane docking site (Fig. 3A). Thus, loop 3 interacts with both lipids and syntaxin, whereas loops 1 and 2 interact preferentially with syntaxin and lipids, respectively. Future studies will determine whether C 2 A can engage both lipids and syntaxin simultaneously.
Localization of the phospholipid binding site adjacent to the calcium binding sites helps explain the positive cooperativity of Ca 2ϩ and phospholipid binding to C 2 A (17). Through sequence alignments, residues analogous to amino acids 198, 205, 206, 233, 234, and 254 in the C 2 A domain of rat synaptotagmin I were determined for the other two C 2 domains whose structures are known, human cytosolic phospholipase A 2 (27) and rat phospholipase C ␦1 (26) ( Table I). In all three cases, these residues (excluding residue 254) lie along a straight line on the protein surface, are adjacent to known Ca 2ϩ binding sites, and involve loops 2 and 3 of the C 2 domain (Fig. 3B). The net charge on these residues at physiological pH is consistent with what is known about headgroup preferences for these domains: synaptotagmin (net charge for these residues of ϩ1 or ϩ2) has a preference for negatively charged phospholipids such as PS, whereas cytosolic phospholipase A 2 (net charge zero for these residues) has a preference for positively charged phospholipids, such as phosphatidylcholine. The lipid preference for the C 2 domain of phospholipase C ␦1 is unknown. Given the proximity of the residues that are perturbed by 6PS binding to the calcium binding sites (Fig. 3), it appears probable that the serine headgroup interacts directly with Ca 2ϩ , displacing water in open coordination sites. This kind of direct interaction would explain how PS binding increases the affinity of C 2 A for Ca 2ϩ .
Genetic and biochemical studies have provided strong evidence that synaptotagmin I serves as a major Ca 2ϩ sensor for neuronal exocytosis (14). A major challenge is to determine how interactions of synaptotagmin with membranes and components of the membrane fusion complex (22,23) serve to trigger fusion. In this study we have begun to elucidate the molecular mechanism by which the C 2 A domain of synaptotagmin docks onto membranes. Further structural studies should shed additional light on the conformational changes that couple Ca 2ϩ influx to exocytosis.