The top loops of the C(2) domains from synaptotagmin and phospholipase A(2) control functional specificity.

The phospholipid-binding specificities of C(2) domains, widely distributed Ca(2+)-binding modules, differ greatly despite similar three-dimensional structures. To understand the molecular basis for this specificity, we have examined the synaptotagmin 1 C(2)A domain, which interacts in a primarily electrostatic, Ca(2+)-dependent reaction with negatively charged phospholipids, and the cytosolic phospholipase A(2) (cPLA(2)) C(2) domain, which interacts by a primarily hydrophobic Ca(2+)-dependent mechanism with neutral phospholipids. We show that grafting the short Ca(2+)-binding loops from the tip of the cPLA(2) C(2) domain onto the top of the synaptotagmin 1 C(2)A domain confers onto the synaptotagmin 1 C(2)A domain the phospholipid binding specificity of the cPLA(2) C(2) domain, indicating that the functional specificity of C(2) domains is determined by their short top loops.

The phospholipid-binding specificities of C 2 domains, widely distributed Ca 2؉ -binding modules, differ greatly despite similar three-dimensional structures. To understand the molecular basis for this specificity, we have examined the synaptotagmin 1 C 2 A domain, which interacts in a primarily electrostatic, Ca 2؉ -dependent reaction with negatively charged phospholipids, and the cytosolic phospholipase A 2 (cPLA 2 ) C 2 domain, which interacts by a primarily hydrophobic Ca 2؉ -dependent mechanism with neutral phospholipids. We show that grafting the short Ca 2؉ -binding loops from the tip of the cPLA 2 C 2 domain onto the top of the synaptotagmin 1 C 2 A domain confers onto the synaptotagmin 1 C 2 A domain the phospholipid binding specificity of the cPLA 2 C 2 domain, indicating that the functional specificity of C 2 domains is determined by their short top loops. C 2 domains are widely distributed, independently folding domains that are present in many signal transduction and membrane trafficking proteins, such as phospholipases, protein kinase C, and synaptotagmins (reviewed in Refs. [1][2][3][4]. Most C 2 domains bind Ca 2ϩ and interact with phospholipid membranes upon Ca 2ϩ binding. The three-dimensional structures of several C 2 domains have been elucidated at atomic resolution (5)(6)(7)(8)(9)(10)(11). These structures revealed that C 2 domains, despite a low overall sequence identity, are composed of similar ␤-sandwiches with flexible loops at the top and the bottom. However, the ␤-strands in the ␤-sandwiches exhibit two distinct topologies, resulting in type 1 (e.g. the C 2 A domain of synaptotagmin 1) and type 2 C 2 domains (e.g. the C 2 domain of cytosolic phospholipase A 2 (cPLA 2 ) 1 (5,11). The two topologies are circular permutations of each other, such that the first ␤-strand in type 1 C 2 domains is the eighth ␤-strand of type 2 C 2 domains, and the eighth ␤-strand of type 1 C 2 domains corresponds to the seventh ␤-strand of type 2 C 2 domains as illustrated in the sequence alignment in Fig. 1A. In C 2 domains, Ca 2ϩ binds exclusively to the top loops, which coordi-nate 2-3 Ca 2ϩ ions primarily via multidentate aspartate residues (7,(11)(12)(13)(14)(15). The Ca 2ϩ -binding sites are similar among C 2 domains with relatively low but variable intrinsic binding affinities (7,13,14,16). The best characterized function of C 2 domains consists of their Ca 2ϩ -dependent binding to phospholipid membranes, although not all C 2 domains appear to share this function. Phospholipid binding mediates the Ca 2ϩ -stimulated recruitment of C 2 domain proteins to membranes upon stimulation (1)(2)(3)(4). Phospholipids dramatically increase the overall Ca 2ϩ affinity of C 2 domains, most likely because the phospholipid head groups provide additional coordination sites for Ca 2ϩ ions at the tip of the C 2 domains (15,(17)(18)(19).
At least two different phospholipid-binding modes have been observed in C 2 domains, as best described for the C 2 A domain of synaptotagmin 1 and the C 2 domain of cPLA 2 (Table I). In both C 2 domains, the top loops bind the Ca 2ϩ ions, but the architecture and stoichiometry of the Ca 2ϩ -binding site differ ( Fig. 1A; Refs. 13 and 14). The most important difference between the two C 2 domains, however, lies in their binding specificity. The synaptotagmin C 2 A domain binds to negatively charged phospholipids in a primarily electrostatic interaction (17,22). In contrast, the cPLA 2 C 2 domain interacts with neutral phospholipids in a hydrophobic reaction (Table I; Refs. 16, 22, and 23). Since Ca 2ϩ and phospholipids bind to the top loops of C 2 domains, it is plausible to hypothesize that the top loops may be instrumental in determining their respective properties, but this has never been demonstrated. The top loops are relatively short, the cPLA 2 and synaptotagmin C 2 domains exhibit a low overall sequence similarity and contain different ␤-strand topologies ( Fig. 1), suggesting that the loops may not be sufficient in specifying the properties of a C 2 domain. To address this question, we have now tested if replacing the top loops of the synaptotagmin 1 C 2 A domain with those of the cPLA 2 C 2 domain will "reprogram" the phospholipid-binding mode of the synaptotagmin C 2 A domain.

Construction of Expression Vectors and Protein Expression-pGEX-
cPLA2-C2 domain encoding the human cPLA 2 C 2 domain (residues 1-141) fused to GST was constructed by PCR amplification of the C 2 domain (primer sequences: A ϭ CCGGAATTCTAATGTCTTTCATAG-ATCCTTATCAG; B ϭ CCCAAGCTTCTAGCAGCTACAAACTTCAAG-AGACATTTC) and subcloning of the PCR product into the EcoRI/Hind-III sites of pGEX-KG (24). The synaptotagmin 1 C 2 A domain expression vector (pGEX65-4; residues 140 -267) was described previously (20). The synaptotagmin 1/cPLA 2 Loop-Swap C 2 domain expression vectors (pGEXSytI-C2A-cPLA2-loops1.2, -loops3, and -loops1. 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.
TCCAGTCTTCAATGAAC; H ϭ CATTACGTAGTTTGCGTCCATTAG-TGTTATCTCGAGTGTTTTGCCACCTAATTCCG; I ϭ CTAATGGACG-CAAACTACGTAATGGACGAAACACTAGGAGAGTTCAAAGTTCC; J ϭ GCGAAGCTTATTTCTCAGCGCTCTGGAG). The resulting fragments were assembled by PCR and cloned into the BamHI/HindIII sites of pGEX-KG. All plasmids were verified by sequencing. Recombinant GST fusion proteins were purified on glutathione-agarose by standard procedures (24) and used for phospholipid binding measurements with GST fusion proteins immobilized on glutathione-agarose (18). Amounts, purity, and integrity of proteins were standardized by SDS-PAGE and Coomassie Blue staining.
Phospholipid Binding Assays-Phospholipids (1.75 mg total; Avanti Polar Lipids, Alabaster AL) were solubilized in chloroform, mixed in the indicated weight ratios with a trace of 3 H-labeled PC (Ͻ0.01% of total; Amersham Pharmacia Biotech), and dried under nitrogen. Dried lipids were resuspended in 10 ml of 50 mM HEPES-NaOH, pH 7.4, 0.1 M NaCl by vortexing (1 min), sonicated (5 min) in a waterbath sonicator (model G112PIG; Laboratory Supply, Co. Inc., Hicksville, NJ; output: 80 kc, 80 watts), and centrifuged (15 min) at ϳ5,000 ϫ g to remove aggregates. Beads containing ϳ25 g of recombinant protein (1 g/liter wet glutathione beads) were equilibrated in 0.1 ml of the respective binding buffers (50 mM HEPES-NaOH, pH 6.8, 0.1 M NaCl (if not indicated differently), 4 mM Na 2 EGTA, 8.75 g of phospholipids with 0.025 Ci of 3 H-labeled PC, and either no additions or respective divalent cations as indicated). For Ca 2ϩ titrations, the binding buffers contained Ca 2ϩ / EGTA concentrations calculated using a commercial software (EqCal for Windows, Biosoft, Ferguson, MO). The mixture was incubated for 10 min at room temperature with vigorous shaking in an Eppendorf shaker, briefly centrifuged, and washed three times with 800 l of the respective binding buffers. Phospholipid binding was quantified by scintillation counting of the beads (LS6000SC; Beckman Instruments, Inc., Fullerton, CA). All buffers were made in high resistance MilliQ water using a 1 M Ca 2ϩ standard solution (Fluka Chemical Corp., Rankonkoma, NY). EC 50 and Hill coeff were calculated from the binding data with the GraphPad Prism program (GraphPad Software Inc., San Diego, CA).
Miscellaneous Procedures-SDS-PAGE was performed as described previously (20). Protein concentrations were determined by comparison of samples run on SDS-PAGE with known amounts of bovine serum albumin standards analyzed on the same gels.

RESULTS AND DISCUSSION
We produced GST fusion proteins of the C 2 A domain of rat synaptotagmin 1, the single C 2 domain of human cPLA 2 , and chimeric C 2 A domains in which the synaptotagmin Ca 2ϩ -binding loops were replaced with those from the cPLA 2 C 2 domain ("loop swaps"; Fig. 1, A and B). The five C 2 domain proteins expressed well in bacteria and were isolated in an electrophoretically pure form on glutathione-agarose (data not shown). Using these proteins, we first tested if the loop-swap C 2 A domains were still capable of Ca 2ϩ -dependent phospholipid binding.
As described previously (17,18,20,21,25,26), we found that the cPLA 2 C 2 domain bound equally well to liposomes composed of 100% PC, 50% PC/50% PE, 30% PS/70% PC, or 50% PI/50% PC, whereas the synaptotagmin 1 C 2 A domain only bound to liposomes that contained the negatively charged phospholipids PS or PI (Fig. 2 and data not shown). All binding was Ca 2ϩ -dependent. The chimeric loop-swap C 2 A domains that contained either only loops 1 and 2 or loop 3 exhibited no Ca 2ϩ -dependent binding of 100% PC liposomes and little binding to liposomes composed of 30% PS/70% PC. Very high Ca 2ϩ concentrations were tested in these experiments (10 mM) to exclude the possibility that these partial loop swaps may simply have a very low Ca 2ϩ affinity. By contrast, the complete loop-swap C 2 domain bound to phospholipids as a function of Ca 2ϩ and thus was functional (Fig. 2). Strikingly, although the loop-swap C 2 A domain contains only the top loops of the cPLA 2 C 2 domain and is otherwise identical with the synaptotagmin C 2 A domain, it exhibited similar binding to neutral phospholipids (PC and PE) as the cPLA 2 C 2 domain (Fig. 2 and data not shown). Quantitations revealed that the loop-swap C 2 A domain bound as much neutral liposomes as the cPLA 2 C 2 domain, but was less capable of binding negatively charged liposomes than the cPLA 2 C 2 domain (data not shown).
FIG. 1. Structure-based sequence alignment of the C 2 domain from human cPLA 2 ) and the C 2 A domain of rat synaptotagmin 1 (Syt1) and schematic diagram of the C 2 domain constructs used in the current study. A, sequences of the cPLA 2 and synaptotagmin 1 C 2 domains aligned for maximal homology, with the locations of ␤-strands and top Ca 2ϩ -binding loops based on the atomic structures of the domains (5, 6, 10, 11) indicated above and below the sequences. Vertical lines identify the positions at which the top loops from the cPLA 2 C 2 domain were grafted onto the synaptotagmin 1 C 2 A domain. Residues shared between the synaptotagmin 1 and cPLA 2 C 2 domains are shaded, and residues involved in Ca 2ϩ binding in the top loops are shown in white on a black background. Note the divergence of the Ca 2ϩ -binding sites between the two C 2 domains where the synaptotagmin 1 C 2 A domain binds three Ca 2ϩ ions via five aspartate and one serine residues, whereas the cPLA 2 C 2 domain binds two Ca 2ϩ ions via three aspartate, two asparagine, and one threonine residues (13, 14) (see Table I for a description of the distinct properties of these C 2 domains). B, schematic overview of the chimeric C 2 domain constructs used. We next examined Ca 2ϩ -dependent binding of the cPLA 2 C 2 domain, the synaptotagmin C 2 A domain, and the complete loop-swap C 2 domain to PC and PS liposomes in the presence of increasing concentrations of NaCl to test whether phospholipid binding is primarily electrostatic or hydrophobic (Fig. 3). The rationale of this experiment is that high NaCl should stabilize hydrophobic interactions but disrupt electrostatic interactions. In agreement with previous studies (18,22), hydrophobic bind- ing of the cPLA 2 C 2 domain to phospholipids was not impaired by high salt. In contrast, the primarily electrostatic binding of the synaptotagmin C 2 A domain to phospholipids was highly salt-sensitive and was abolished by 0.6 M NaCl. In the same experiments, the loop-swap C 2 A domain behaved like the cPLA 2 C 2 domain, with phospholipid binding resistant to interference by NaCl at concentrations of up to 1.0 M (Fig. 3).
The functionality of the loop-swap C 2 A domain composed of the synaptotagmin C 2 A domain with only short Ca 2ϩ -binding loops from the cPLA 2 C 2 domain is surprising considering the fact that the cPLA 2 and synaptotagmin C 2 domains have divergent properties and distinct ␤-strand topologies. Studies with different divalent cations showed that in all three C 2 domains, Mg 2ϩ was unable to stimulate phospholipid binding at high concentrations, whereas Ba 2ϩ and Sr 2ϩ at least partly substituted for Ca 2ϩ , albeit with lower affinity (data not shown). The cPLA 2 C 2 domain and the loop-swap C 2 domain were also similar with respect to divalent cation binding in that Ba 2ϩ and Sr 2ϩ were slightly more effective in replacing Ca 2ϩ in these C 2 domains than in the synaptotagmin C 2 A domain (data not shown). To test more accurately if the precise conformation of the ␤-strand sandwich contributes to the Ca 2ϩ binding properties of the overall C 2 domains, we examined their apparent Ca 2ϩ affinities in the presence of phospholipid membranes. Ca 2ϩ /EGTA buffers were used in these experiments to measure the Ca 2ϩ dependence of phospholipid binding to the various immobilized C 2 domains. Since Ca 2ϩ -dependent phospholipid binding by the synaptotagmin C 2 A domain is primarily electrostatic, the apparent Ca 2ϩ affinity of the synaptotagmin C 2 A domain depends on the phospholipid composition of the liposomes used for such measurements (18). Therefore we measured the apparent Ca 2ϩ affinity of all three C 2 domains with four types of liposomes containing an increasing abundance of negatively charged phospholipids (Table II).
As shown in an exemplary experiment in Fig. 4, the cPLA 2 C 2 domain exhibited a slightly higher Ca 2ϩ affinity than the synaptotagmin C 2 A domain, as also shown in earlier studies (16,18,20,22). The complete loop-swap C 2 A domain displayed a significantly lower Ca 2ϩ affinity (5-15-fold depending on phospholipid composition) than either the cPLA 2 C 2 domain or the synaptotagmin C 2 A domain ( Fig. 4 and Table II). As predicted, the Ca 2ϩ affinity of the synaptotagmin C 2 A domain was dependent on phospholipid composition, and two to three times higher in 45% PS/55% PC than in 22.5% PS/77.5% PC. By contrast, the cPLA 2 C 2 domain exhibited similar affinities in the presence of liposomes independent of the surface charge (Table II). Unexpectedly, the apparent Ca 2ϩ affinity of the loop-swap C 2 domain was dependent on the phospholipid composition with the opposite relationship as the synaptotagmin C 2 A domain. This affinity decreased almost 5-fold with increasingly negatively charged liposomes (Table II). Together these data suggest that the precise conformation of the underlying ␤-sandwich has little influence on the phospholipid binding specificity or on the hydrophobic versus electrostatic binding mechanism mediated by the top Ca 2ϩ -binding loops, but has a marked effect on the overall Ca 2ϩ affinity of the C 2 domain (Table II).
Summary-The C 2 A domain of synaptotagmin 1 and the C 2 domain of cPLA 2 are composed of ␤-sandwiches that contain flexible Ca 2ϩ -binding loops on top, which were previously hypothesized to contribute to the Ca 2ϩ -dependent interactions of these domains with phospholipid membranes. However, the two C 2 domains have different ␤-strand topologies and bind phospholipids with distinct Ca 2ϩ -dependent mechanisms and  Table II. TABLE II Apparent Ca 2ϩ -binding affinities of C 2 domains Apparent Ca 2ϩ binding affinities of the C 2 domain from cPLA 2 , the complete loop-swap C 2 domain, and the synaptotagmin 1 C 2 domain were determined using Ca 2ϩ titrations of phospholipid binding with Ca 2ϩ /EGTA buffers (Fig. 4). Apparent Ca 2ϩ affinities were determined by binding of radiolabeled liposomes with the phospholipid composition indicated on the left. Data shown are means Ϯ S.E. specificities, suggesting that the top loops may not be sufficient for the specific properties of the various C 2 domains. To test this directly, we have now grafted the top loops of the cPLA 2 C 2 domain onto the tip of the synaptotagmin C 2 A domain. We find that grafting all of the Ca 2ϩ -binding loops of the cPLA 2 C 2 domain enables the synaptotagmin C 2 A domain to bind to neutral phospholipids in a salt-insensitive reaction similar to the cPLA 2 C 2 domain, whereas grafting only one or two of the three loops is insufficient. Furthermore, the loop-swap C 2 domain exhibited a lower Ca 2ϩ affinity, indicative of a contribution of the backbone of the C 2 domain ␤-sandwich to the precise positioning of the Ca 2ϩ -binding loops. These data demonstrate that the functional features and phospholipid binding modes of the cPLA 2 and synaptotagmin C 2 domains are largely determined by their top loops, a surprising result considering the shortness of these sequences in the various C 2 domains.