The C2A-C2B Linker Defines the High Affinity Ca2+ Binding Mode of Rabphilin-3A*

The Ca2+ binding properties of C2 domains are essential for the function of their host proteins. We present here the first crystal structures showing an unexpected Ca2+ binding mode of the C2B domain of rabphilin-3A in atomic detail. Acidic residues from the linker region between the C2A and C2B domains of rabphilin-3A interact with the Ca2+-binding region of the C2B domain. Because of these interactions, the coordination sphere of the two bound Ca2+ ions is almost complete. Mutation of these acidic residues to alanine resulted in a 10-fold decrease in the intrinsic Ca2+ binding affinity of the C2B domain. Using NMR spectroscopy, we show that this interaction occurred only in the Ca2+-bound state of the C2B domain. In addition, this Ca2+ binding mode was maintained in the C2 domain tandem fragment. In NMR-based liposome binding assays, the linker was not released upon phospholipid binding. Therefore, this unprecedented Ca2+ binding mode not only shows how a C2 domain increases its intrinsic Ca2+ affinity, but also provides the structural base for an atypical protein-Ca2+-phospholipid binding mode of rabphilin-3A.

Rabphilin-3A belongs to a structurally diverse synaptic protein family whose role is to support and regulate the Ca 2ϩ -dependent neurotransmitter release process. A significant number of these proteins share a common structural feature, a tandem of C2 domains. Although specific roles have been attributed to some members of this family, the exact function of rabphilin-3A is still not clear at present. Rabphilin-3A has been shown to interfere with the exocytotic/endocytotic machinery in different secretory systems (1)(2)(3)(4). Many studies suggest that rabphilin-3A functions in the synaptic vesicle trafficking pro-cesses preceding and/or following the vesicle fusion itself. In PC12 cells, the C2B domain of rabphilin-3A was shown to interact with SNAP25 (synaptosome-associated protein of 25 kDa), regulating the docking step of dense core vesicles (5). Rabphilin knock-out studies carried out in mouse (6) and Caenorhabditis elegans (7) have shown that this protein is largely dispensable for the neurotransmitter release process. Nevertheless, in C. elegans, rabphilin-3A is able to potentiate the SNARE (soluble N-ethylmaleimide-sensitive factor attachment protein receptor) function (7). Very recently, it was shown that, in mouse neurons, rabphilin-3A alters, via its C2B domain, the recovery of synaptic responses after high frequency stimulation (8). With the precise role of rabphilin-3A remaining to be clarified, these most recent findings suggest a fine-tuning function for this protein in the vesicle fusion machinery.
Functionally, two regions of rabphilin-3A have been defined so far: the N-terminal Rab-binding domain, responsible for its interaction with the small GTPases Rab3 and Rab27 (9,10), and the C-terminal C2 domain tandem fragment, involved in the Ca 2ϩ -dependent membrane binding property of the protein (11) as well as in several protein-protein interactions.
The tandem C2 domains of synaptic vesicle proteins function as membrane-targeting modules, activated upon an increase in the local Ca 2ϩ concentration. More specifically, the fast synaptic vesicle fusion requires Ca 2ϩ concentrations in the range of tens to hundreds micromolar. In binding assays with negatively charged membranes, this concentration range corresponds to the relative Ca 2ϩ affinity of the two C2 domains of synaptotagmin 1, the hypothesized Ca 2ϩ sensor for this process. In solution, the Ca 2ϩ binding affinity of these C2 domains is significantly lower compared with their relative affinity in the presence of membranes (12). This difference has been explained by a Ca 2ϩ -bridging mechanism underlying these interactions. According to this hypothesis, the incomplete Ca 2ϩ coordination sphere in the C2 domains is filled by phosphate groups of the phospholipids, resulting in complete coordination spheres of the C2 domain-bound Ca 2ϩ ions and thus in a tremendous increase in the apparent Ca 2ϩ affinity (13).
The C2 core domain is defined by the typical eight-stranded ␤-sandwich fold. Many C2 core domains contain also a Ca 2ϩbinding region (CBR) 5 with a conserved consensus sequence of acidic residues, but this motif alone does not reflect the large variations in the Ca 2ϩ and Ca 2ϩ -mediated membrane binding affinities among these domains (14 -16). These differences require subtle (14) or major (17) structural rearrangements. So far, such structural changes have been identified only within the C2 core domains of these proteins.
The Ca 2ϩ binding mode of the C2B domain of rabphilin-3A has not been studied so far at atomic resolution. Its intrinsic Ca 2ϩ binding affinity in solution has been shown to be in the 8 -11 M range (18). Based on the NMR structure of the C2B core domain, this unusually high Ca 2ϩ affinity in comparison with C2 domains with a similar fold but much lower intrinsic Ca 2ϩ affinity could not be explained. We took advantage of complementary x-ray and NMR techniques to probe the structural basis of this fundamental feature. The structural work presented here shows that Ca 2ϩ binding of the C2B domain of rabphilin-3A is not mediated only by the classical CBR. Unexpectedly, a sequence upstream of the C2 core domain that belongs to the C2A-C2B tandem linker participates in the Ca 2ϩ binding. This has important implications for the high intrinsic Ca 2ϩ binding affinity of the C2B domain, the spatial proximity of the C2 domains upon Ca 2ϩ signaling, and the way rabphilin-3A may interact with biological membranes.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The cDNAs encoding fragments 371-510, 528 -684, 519 -684, and 371-684 of rat rabphilin-3A were cloned into the expression vector pGEX2T (Amersham Biosciences). The expression constructs encoding N-terminal glutathione S-transferase fusion proteins of these rabphilin-3A fragments were expressed and purified as described (19). Briefly, protein expression was carried out overnight in LB medium at 17°C. The glutathione S-transferase fusion protein was purified on glutathione-Sepharose resin. Cleavage of the fusion protein was performed with thrombin on a column. The fragment released was further purified on a HiTrap SP XL cation exchange column (Amersham Biosciences). Fragment 519 -684, used for crystallization experiments, was further purified by gel filtration on a Sephadex S-75 16/60 column (Amersham Biosciences). Selenomethionine (SeMet)-labeled protein was expressed in minimal medium supplemented with SeMet according to the EMBL Protein Expression Group (embl-heidelberg.de) and purified according to the established protocol.
Crystallization and Structure Determination-For crystallization, rabphilin-3A fragment 519 -684 was concentrated to 20 mg/ml. Crystals were obtained at 12 and 20°C using vapor diffusion by mixing equal volumes of protein and reservoir solution (0.1 M HEPES (pH 8.5), and 20% polyethylene glycol monomethyl ether 2000 or 20% polyethylene glycol 8000). They grew within 1 week to a final size of ϳ25 ϫ 25 ϫ 100 m 3 and were used for streak seeding to obtain crystals of SeMet protein as well as crystals of native protein with reservoir solutions containing 200 mM Ca 2ϩ . For data collection, all crystals were harvested and flash-frozen in liquid nitrogen in a cryosolution containing the reservoir solution supplemented with 10% glycerol. Data were collected at Swiss Light Source beamline PX6 (Mar225 CCD detector) and processed using the program HKL2000 (20). Statistics for data collection and processing are summarized in Table 1. The structure of crystals grown without adding Ca 2ϩ was solved from SeMet crystals (space group P2 1 2 1 2) by multiwavelength anomalous dispersion phasing using the programs SHELXC, SHELXD, and SHELXE (21,22). The initial electron density map was excellent (supplemental Fig. 1) and allowed automatic model tracing, alternated with structure refinement by ARP/wARP (23), resulting in the modeling of 131 residues. Refinement was performed by manual model building with COOT (24), followed by positional and B-factor refinement with REFMAC5 (25) coupled with ARP solvent building. For placing hydrogen atoms and refining occupancies due to radiation damage, the final refinement was done with SHELXL (26). The final model contains 143 of 165 residues, two Ca 2ϩ ions, and 68 water molecules. Residues at the N terminus (positions 519 -523) and C terminus (positions 578 -584) and in two loop regions (positions 536 -538 and 587-588) are disordered. The structure of crystals grown in Ca 2ϩ (space group P2 1 ) was solved by molecular replacement with the program PHASER (27) using the SeMet structure as the search model. There are two monomers in the asymmetric unit. After initial rigid body and restrained B-factor refinement with REFMAC5 (25), manual model building was performed with COOT (24). The final refinement was done with SHELXL (26). The model contains 149 residues for monomer A, 147 residues for monomer B, one phosphate group, two Ca 2ϩ ions for each monomer, and a total of 77 water molecules. In both monomers, residues at the N and C termini are disordered (monomer A, residues 519 -523 and 679 -684; and monomer B, residues 519 -524 and 678 -684). In monomer A, loop residues 536 -538 and 587-588 are also disordered, whereas in monomer B, only loop residues 533-538 are disordered. The phosphate group forms a hydrogen bond with the backbone NH of Phe 612 in monomer B and with Gly 652 of a symmetry-related monomer B.
For both crystal structures, residues with completely or partially disordered side chains are listed in supplemental Tables 1 and 2. The statistics of the Ramachandran plots and of the refinement are shown in Table 1.

RESULTS
Overall Structure-The fragment chosen for crystallizing rabphilin-3A C2B domain fragment 519 -684 was N-terminally extended by 5 residues compared with the fragment that had been used to solve the NMR structure of this C2B domain (fragment 524 -684) (18). Using synchrotron radiation, native and SeMet crystals grown without the addition of Ca 2ϩ diffracted up to 1.58-and 1.28-Å resolution, and native and SeMet crystals grown in the presence of 200 mM Ca 2ϩ diffracted up to 1.91-and 1.85-Å resolution, respectively. Therefore, the structure of the crystals obtained without the addition of Ca 2ϩ (low Ca 2ϩ structure) was solved by multiwavelength anomalous dispersion using the SeMet crystals. This structure was used as a search model to obtain the structure of the native crystals grown with 200 mM Ca 2ϩ in the reservoir (high Ca 2ϩ structure) by molecular replacement. Under low Ca 2ϩ conditions, the C2B domain fragment crystallized in space group P2 1 2 1 2 with one monomer in the asymmetric unit, whereas the presence of 200 mM Ca 2ϩ in the reservoir resulted in crystals in space group P2 1 with two monomers in the asymmetric unit. As already described for its solution structure (18), the core of the C2B domain folds as a typical type I antiparallel eight-stranded ␤-sandwich with a short helix between strands 5 and 6 and a longer helix between strands 7 and 8 in both crystal structures (Fig. 1, A and B; and supplemental Fig. 2). It strongly resembles the structure in solution (root mean square deviations of 1.2909 Å with the low Ca 2ϩ structure and 1.2386 and 1.2603 Å with monomers A and B of the high Ca 2ϩ structure, respectively) (supplemental Fig. 3). Two Ca 2ϩ ions are bound in the low as c R free is the same as R, expect for a 5% subset of all reflections.
High Affinity Ca 2؉ Binding Mode of Rabphilin-3A FEBRUARY 16, 2007 • VOLUME 282 • NUMBER 7 well as high Ca 2ϩ structure. This is in accordance with the surprisingly high Ca 2ϩ binding affinity described for this C2 domain in solution (18). An unexpected feature is found in both crystal structures: several residues from the region upstream of the core domain (residues 519 -549) interact with the CBR of the C2 domain (Fig. 1, C and D). In rabphilin-3A, this region is part of the linker between the C2A and C2B domains. More specifically, a stretch of 6 residues (Leu 526 -Gln 531 ) docks to the region between Ca 2ϩ -binding loop (CBL) 1 and CBL3, thus forming a lid above the Ca 2ϩ -binding sites. The secondary structure of this short stretch differs between the two crystal structures. In the low Ca 2ϩ structure, it forms an extended loop, whereas in the high Ca 2ϩ structure, it forms a short helix. Despite differences in hydrogen bonding to symmetry-related molecules (supplemental Tables 3  and 4), the two structures obviously represent two possible conformations in which the N-terminal acidic region may contribute to Ca 2ϩ binding. The loop connecting this N-terminal part of the linker region to ␤-strand 1 of the C2 core domain is partially disordered (Fig. 1, A and B; and supplemental Fig. 2).
The C2A-C2B Linker Is Involved in Ca 2ϩ Binding-In both crystal structures, a glutamic acid residue establishes essential contacts with CBL3 and with both Ca 2ϩ ions. This residue is Glu 529 in the low Ca 2ϩ structure and Glu 530 in the high Ca 2ϩ structure. Despite the different secondary structures in the N-terminal stretch, in both structures, these glutamic acid residues are in a nearly identical position and orientation relative to the Ca 2ϩbinding sites and to CBL3 (Figs. 1, C and D; and 2B). They interact with the Ca 2ϩ in binding site 2 through their main chain carbonyl groups and contact the Ca 2ϩ in binding site 1 through water-mediated hydrogen bonding of their O-⑀2. Otherwise, both Ca 2ϩ ions interact with the acidic residues found in the classical consensus sequence (13) of C2 domains (Fig. 2, A and B).  vided by the only Ca 2ϩ -coordinating water molecule, which hydrogen bonds with the side chain O-⑀2 of Glu 529 /Glu 530 (Fig.  2, A and B). Thus, both Ca 2ϩ ions are heptacoordinate with a pentagonal bipyramidal coordination geometry, which recalls the one found in EF-hand proteins (supplemental Fig. 4). The water molecule that participates in the coordination sphere of the Ca 2ϩ in binding site 1 as the seventh ligand interacts extensively with several Ca 2ϩ -binding residues belonging to the consensus sequence, viz. the O-␦2 of Asp 571 , the O-⑀2 of Glu 530 , the O-␦2 of Asp 577, and the O-␦1 of Asp 633 . In both crystal structures, the side chain carboxyl group of Glu 529 /Glu 530 forms additional hydrogen bonds with the backbone NH groups of Ile 634 and Gly 635 , thus securing a strong interaction between the N-terminal linker region and CBL3. Next to the Ca 2ϩbridging Glu 529 /Glu 530 residue, Glu 528 /Glu 529 (in low versus high Ca 2ϩ structures) establishes, through its backbone carbonyl group, a hydrogen bond with the Ala 572 main chain amide, holding CBL1 tightly packed in the CBR. Thus, in both crystal structures, the first of the two acidic patches from the linker (see Fig. 6) connects CBL1 and CBL3 through a hydrogen bond network (Fig. 3, A and B).
Another striking feature concerns the positioning of the Lys 636 side chain, which interacts via hydrogen bonds with two Ca 2ϩ -binding residues, viz. Asp 633 O-␦1 and Asp 639 O-␦2 (Fig.  3, A and B). This has two consequences. First, the motions of the side chains of both Ca 2ϩ -binding residues may be restricted, holding them in an optimal position for binding the Ca 2ϩ ions. Second, the positive charge carried by the side chain amino group of Lys 636 is spatially located at the same position as the third Ca 2ϩ -binding site found in the C2A domain of synaptotagmin 1 (Fig. 3C).
Acidic Residues from the Linker Are Essential for the High Ca 2ϩ Binding Affinity of the C2B Domain-To further investigate the interaction of the C2A-C2B domain linker region with the C2B core domain, we performed a solution NMR study. For this purpose, two C2B domain fragments were produced: a "long" fragment (residues 519 -684), used to crystallize this domain, and a "short" fragment (residues 528 -684), including the first acidic patch (see Fig. 6). Experiments to obtain C2B core domain fragment 541-684 in soluble form failed. Additionally, we also performed NMR measurements on the C2A-C2B tandem (residues 371-684) and C2A domain (residues 371-510) fragments.
Triple-resonance backbone assignment experiments were carried out on C2B domain fragment 519 -684. Of the 22 residues of linker region 519 -540, Arg 522 -Tyr 527 and Glu 533 -Glu 539 could be assigned (supplemental Fig. 5A). The assigned residues belong to the flexible parts of the N-terminal linker, i.e. the disordered (non-core domain-interacting) residues in the crystal structure. To address the flexibility of these residues, we measured their 1 H-15 N heteronuclear nuclear Overhauser effect values. From Met 524 on, the nuclear Overhauser effect intensity ratios of 0.5 (supplemental Fig. 6) indicate restricted motions for the linker. The missing residues, including the first acidic patch (positions 528 -530), are probably involved in chemical exchange on an intermediate time scale. This chemical exchange may be related to the acidic residue switch observed between the low and high Ca 2ϩ crystal structures (Figs. 1, C and D; and 2B).
Because of the spectral broadening of the cross-peaks corresponding to the residues defining the first acidic patch (Glu 528 -Val 532 ), we indirectly observed the localization of the N-terminal linker on the C2B core domain. Analysis of the 1 H-15 N HSQC spectra of fragments 519 -684 and 528 -684 (supplemental Fig. 7, A and B) showed chemical shift differences between both fragments mainly for residues in CBL1-3. Moreover, the peak intensity ratios in the HSQC spectra of fragment 528 -684 compared with fragment 519 -684 illustrate an increase in intensity of the cross-peaks corresponding to the backbone HN of the residues located in CBL3 (positions 632- 638) and of Lys 601 in CBL2. In addition, the shortening of the linker affects the disordered C-terminal tail of the C2B domain as well as the last linker residues (supplemental Fig. 7, A and B). It is noteworthy that the residues preceding the N-terminal ␤-strand of the C2B core domain (the last linker residues) are in close proximity to the residues following the last ␤-strand of the core domain in the crystal structures (Fig. 1, A and B). These observations are topologically in agreement with the orientation of the linker seen in the crystal structures with respect to the CBR.
To confirm the interaction of Glu 529 and Glu 530 with the Ca 2ϩ -binding region of the C2B domain, four mutants (E529A, E530A, E529A/E530A, and E528A/E529A/E530A) were produced. 15 N-1 H HSQC-based Ca 2ϩ titrations performed on these mutants showed that they were still able to bind Ca 2ϩ . Whereas the single mutants were saturating at 0.5 mM Ca 2ϩ , the double and triple mutants required 2 mM Ca 2ϩ to reach the saturation point. In the Ca 2ϩ -bound state, the resonances of residues from all three CBLs were affected in each of these mutants (supplemental Fig. 8). This observation is in agreement with the crystal structures and confirms that the linker binds to the Ca 2ϩ -loaded CBR in solution. In contrast, the overlay of the 1 H-15 N HSQC spectra of these mutants in the presence of 10 mM EGTA with the wild-type Ca 2ϩ -free C2B domain shows no significant perturbation for the core domain residues (supplemental Fig. 9A). This indicates that the linker binds to the CBR only in the presence of Ca 2ϩ .
The chemical shift deviation of cross-peaks from several residues of the double mutant (E529A/E530A) and the triple mutant (E528A/E529A/E530A) can be followed by 1 H-15 N HSQC-based Ca 2ϩ titration. Six residues show chemical shift deviations in the fast exchange regime for these mutants. A simultaneous fit using the Hill equation led to a dissociation constant (K) of 101 Ϯ 25 M with a Hill coefficient (n H ) of 1.45 Ϯ 0.19 for the E529A/E530A mutant and K ϭ 116 Ϯ 22 M with n H ϭ 1.50 Ϯ 0.08 for the E528A/E529A/E530A mutant (supplemental Fig. 10). The overlay of the 1 H-15 N HSQC spectra of both mutants in the Ca 2ϩ -bound state shows identical chemical shifts for cross-peaks of the core domain residues (supplemental Fig. 9B). This indicates that Glu 528 does not interact with the CBR when Glu 529 and Glu 530 are mutated to alanine. Thus, mutating the crucial residues Glu 529 and Glu 530 to alanine leads to a 10-fold decrease in the intrinsic Ca 2ϩ binding affinity of the C2B domain. Although not seen in the crystal structures, the C-terminal residues are strongly affected by the mutations, suggesting an interaction of these residues with the N-terminal linker residues in the CBR-bound state.
The  4A). In the tandem fragment, the relative intensities of the cross-peaks from the linker residues compared with those from the core domain residues are similar to the ratios obtained for C2B domain fragment 519 -684. In addition, well resolved cross-peaks from the residues of the linker show similar chemical shift deviations upon Ca 2ϩ binding compared with C2B domain fragment 519 -684 (supplemental Fig. 11). These observations suggest that the interaction of the linker with the CBR in the isolated C2B domain is maintained in the C2 domain tandem fragment.
We also probed the Ca 2ϩ binding properties of the C2B domain in the C2A-C2B tandem fragment (residues 371-684). 5 mM Ca 2ϩ in the NMR buffer was needed to obtain a fully Ca 2ϩ -loaded C2A domain fragment (residues 371-510) (Fig.  4B), whereas it stayed in the Ca 2ϩ -free state in the presence of 0.1 mM Ca 2ϩ , as verified by comparison with the HSQC spectrum in the presence of 10 mM EGTA. On the other hand, C2B domain fragment 519 -684 was in its Ca 2ϩ -loaded state in the presence of 0.1 and 5 mM Ca 2ϩ . Its intrinsic high Ca 2ϩ binding affinity led to the purification of the Ca 2ϩ -bound form anyway. Therefore, no further chemical shift deviations could be observed in the HSQC spectra of this domain when NMR buffer was supplemented with 0.1 or 5 mM Ca 2ϩ . We then compared the Ca 2ϩ -loaded state of both domains in the C2A-C2B tandem fragment in the presence of these two Ca 2ϩ concentrations. In buffer containing 0.1 mM Ca 2ϩ , the C2B domain in this fragment was in the Ca 2ϩ -bound form, whereas the C2A domain remained Ca 2ϩ -free (Fig. 4B). In the presence of 5 mM Ca 2ϩ , both domains were in the Ca 2ϩ -loaded state. This indicates that the high affinity Ca 2ϩ binding property of the C2B domain is maintained in the C2 domain tandem fragment probably because of the interaction of the acidic linker residues with the C2B domain.
Upon Interaction with Phospholipid Bilayers, the Linker Stays Bound to the CBR-The coordination spheres of the Ca 2ϩ ions in the crystal structures of C2B domain fragment 519 -684 almost completely consist of protein oxygen atoms; there is just one metal-bound water molecule (Fig. 2B). According to the Ca 2ϩ -bridging phospholipid binding model, only one binding site is left for an anionic phospholipid head group if the linker remains on the CBR once the protein is bound to the membrane. This structural feature, unusual for a C2 domain, raises the question of the Ca 2ϩ -dependent phospholipid-binding mechanism of this domain. We used NMR spectroscopy as a tool to assay the binding mode of the C2B domain to soluble phosphatidylserine (PS) and PS-containing liposomes. The PS head group-specific recognition has been addressed using soluble glycerophosphoserine and 1,2-dihexanoyl-sn-glycero-3-phospho-L-serine. Neither a 20-fold excess of glycerophosphoserine nor subcritical micellar concentrations of 1,2-dihexanoyl-sn-glycero-3-phospho-L-serine induced any significant chemical shift perturbation or peak intensity change in the CBR of the C2B domain. Concentrations of 1,2-dihexanoyl-sn-glycero-3-phospho-L-serine above the critical micellar concentration (1.1 mM) induced an overall broadening of all cross-peaks of the C2B domain spectra. Consequently, as for most C2 domains that bind PS in a Ca 2ϩ -dependent fashion, no specific PS head group-binding site could be observed.
Next, binding to phospholipid bilayers was probed by a 1 H-15 N HSQC-based Ca 2ϩ titration starting with the Ca 2ϩfree C2B domain in the presence of 80:20 1,2-dioleoyl-snglycero-3-phosphocholine/1,2-dioleoyl-sn-glycero-3-phospho-L-serine liposomes. Although C2B domain fragments 519 -684 and 528 -684 were expressed and purified without adding Ca 2ϩ to any buffers, the HSQC spectra showed that, for both fragments, the C2B domains were in the Ca 2ϩ -bound form. The removal of the Ca 2ϩ ions by the addition of 10 mM EGTA affected also the chemical shift of most of the residues belonging to the linker (Fig. 5C). It is noteworthy that new highly intense cross-peaks appeared in the Ca 2ϩ -free 1 H-15 N HSQC spectrum of the C2B domain and were localized in a narrow region of the spectrum characteristic for non-structured residues (Fig. 5C). These signals have been assigned (supplemental Fig. 5B) to the N-terminal linker residues that are in chemical exchange at an intermediate time scale in the presence of Ca 2ϩ . Moreover, the heteronuclear 1 H-15 N nuclear Overhauser effect values of the linker residues were reduced in the Ca 2ϩ -free form compared with the Ca 2ϩ -bound form (data not shown). Taken together with the data related to the E529A and E530A mutants, these observations show that the linker is more flexible in the absence of Ca 2ϩ .
The Ca 2ϩ -free C2B domain was not able to bind to liposomes. The addition of increasing amounts of Ca 2ϩ triggered the binding of the C2B domain to the liposomes, as monitored by the broadening of almost all of the 1 H-15 N cross-peaks (Fig.  5A). The only remaining cross-peaks belonged to the flexible residues from the N-terminal linker (Fig. 5A). The chemical shifts of these residues were almost identical to those observed in the Ca 2ϩ -bound form in solution (Fig. 5A). Moreover, the intense cross-peaks observed in the Ca 2ϩ -free C2B domain spectrum, assigned to the acidic residues of the linker (supplemental Fig. 5B), were not seen in the liposomebound form of the C2B domain (Fig. 5, B and C). This suggests that the binding to the liposomes does not trigger the release of the linker. These observations strongly indicate that the structural features previously observed for the CBR of the C2B domain in the presence of Ca 2ϩ are maintained in the membrane-bound state.

DISCUSSION
The C2A-C2B Linker Modulates the Ca 2ϩ Binding Affinity of the C2B Domain-A recent study performed on a rabphilin-3A knock-out mouse attributed, for the first time, a specific function to the C2B domain of this protein in the repriming of vesicles following induced vesicle fusion (8). This process takes place at residual Ca 2ϩ concentrations and probably requires the interaction with SNAP25. The high affinity Ca 2ϩ binding mode is thus essential for the function of rabphilin-3A. Through investigation of this Ca 2ϩ binding mode, we have shown in this work that a number of acidic residues from the C2A-C2B linker region that are not part of the C2B core domain interact with the Ca 2ϩ -binding region of this domain. Because of these interactions, the coordination sphere of both Ca 2ϩ ions is almost completely created by protein residues (Fig. 2, A and B). Most strikingly, these additional linker residues in the coordination sphere of the Ca 2ϩ ions enable the formation of an exhaustive hydrogen bond network stabilizing CBL1 and CBL3 (Fig. 3, A  and B). Two different acidic residues (Glu 529 versus Glu 530 ) are involved in this extensive hydrogen bond network in the two crystal structures (Fig. 3, A and B), but their identical orientation relative to both Ca 2ϩ ions suggests that the architecture of the whole interaction network is energetically highly favorable. It is noteworthy that the Lys 636 side chain amino group establishes two hydrogen bonds with both Ca 2ϩ -coordinating carboxylic groups of Asp 633 and Asp 639 (Fig. 3C). This interaction allows the Asp 639 side chain to participate with the Ca 2ϩ in binding site 2 in our crystal structures, contrary to the equivalent residue Asp 371 in synaptotagmin 1 C2B domain crystal structures (33). In addition, the Lys 636 side chain amino group may well mimic, in terms of charge distribution, Ca 2ϩ -binding site IV found in the C2A domain of synaptotagmin 1 and protein kinase C␦ (34,35). This may be important for the phospholipid binding mode of the C2B domain of rabphilin-3A by constituting an additional localized PS-binding site. Thus, the CBR  of the rabphilin-3A C2B domain structure provides the most optimal Ca 2ϩ -binding site of any C2 domain structurally studied so far.
The double mutation of both glutamic acid residues (E529A/ E530A) involved in the interaction with the CBR lowers the intrinsic Ca 2ϩ binding affinity of C2B by a factor of 10. This shows that the acidic patch of the linker is directly responsible for the high Ca 2ϩ binding affinity of the C2B domain. This unprecedented structural feature defining the high intrinsic Ca 2ϩ binding affinity of the C2B domain suggests a role for this acidic patch in the response of the rabphilin-3A C2B domain to low physiological Ca 2ϩ concentrations. In addition, as shown by NMR, the linker binds to the CBR of the C2B domain in the presence of Ca 2ϩ , whereas it stays flexible in the Ca 2ϩ -free state. This implies a more compact C2 domain tandem in the Ca 2ϩ -bound state, which could be important for modulating the functional interplay between both C2 domains.
The C2A-C2B Linker Is Involved in Phospholipid Binding of the C2B Domain-The standard model for C2 domain-membrane interaction implies that the Ca 2ϩ ions function as a bridge between the protein and the phospholipids. According to this model, the phosphate moiety of the phospholipids and the carboxylic moiety of the PS head group complete the coordination spheres of the Ca 2ϩ ions. The Ca 2ϩ -coordinating water molecules found in the various crystal structures of C2 domains are then displaced by the phospholipids, and no strong specific PS head group-protein interactions are required for the binding (34,36,37). This binding model agrees well with the low intrinsic Ca 2ϩ affinity of these domains in solution compared with their higher relative Ca 2ϩ binding affinity in the presence of membranes.
We could not identify a PS head group-specific binding site on the C2B domain of rabphilin-3A. This is suggestive of a Ca 2ϩ -dependent, multiple ligand-driven binding mode. According to the standard C2 domain-membrane interaction model, the linker region would have to be released from the CBR upon binding to the membrane to partially free Ca 2ϩ coordination sites. The binding of the C2B domain to phosphocholine/PS liposomes did not show any indications for such a switch in the HSQC spectra (Fig. 5). The interpretation of our NMR data rather suggests that the acidic linker residues involved in Ca 2ϩ coordination stay bound to the CBR of the C2B domain upon phospholipid binding. Thus, the standard model for Ca 2ϩ -mediated phospholipid binding by C2 domains is probably not applicable to this specific case. Instead, the unique structural motif arising from the interaction between the acidic linker residues and the CBR may constitute a new type of scaffold for Ca 2ϩ -mediated protein-phospholipid interactions. Such an alternative phospholipid binding mode would probably require the emergence of new PS-binding sites defined by specific structural features of the linker-bound CBR. Further investigations, e.g. using the EPR membrane depth method (38), will be required to obtain a more detailed understanding of the C2B domain-membrane interaction at the molecular level.
What Is the Role of Acidic Motifs for other C2 Domains?-Among the C2 domain tandem proteins, this is, to our knowledge, the first case in which a sequence motif located in the linker connecting two C2 domains interacts with the Ca 2ϩbinding region of one of them. The sequence homology of these linkers is generally low, and their lengths are highly variable. Doc2␥ (double C2 protein) is actually the only protein containing a C2 domain tandem fragment that shows high sequence similarity to rabphilin-3A in the linker region (39). No information on its possible function is available so far. Interestingly, such acidic patches are more generally found upstream of several C2A domains (Fig. 6). In light of our results, these specific sequences may have specific functions for the C2A domains. First, they may significantly affect the intrinsic Ca 2ϩ binding property of these C2 domains. It has been shown that the C2A domains of Doc2␣ and Doc2␤ require a Ca 2ϩ concentration in the low micromolar range to localize at the plasma membrane FIGURE 6. Acidic sequences found upstream of C2 core domains. Shown is the sequence alignment of C2B domains (rabphilin-3A and Doc2␥) and C2A domains (rabphilin-3A; synaptotagmins (SYT) 7, 1, and 2; Doc2␣; and Doc2␤), including sequences upstream of the C2 core domains. All sequences are from Rattus norvegicus. The secondary structure elements correspond to the rabphilin-3A C2B core domain structures. Black arrows represent the ␤-strands, and gray rectangles represent the ␣-helices. The color code for the alignment is as follows: dark green, identical residues; green, similar residues; and pale green, blocks of similar residues. The alignment was performed using Vector NTI. The red box highlights the acidic sequences upstream of the C2 core domains, and the orange boxes show the two acidic patches (first patch, residues 528 -530; and second patch, residues 537-540) in the rabphilin-3A C2A-C2B tandem linker. The consensus sequence is shown at the bottom of the alignment.
in vivo (40). Similarly, synaptotagmin 7 binds phospholipids at low micromolar Ca 2ϩ concentrations (41). Its C2A domain has the highest relative Ca 2ϩ affinity for membrane binding of all synaptotagmin isoforms investigated so far (16). This suggests that the Ca 2ϩ binding mode of these C2A domains may be similar to the Ca 2ϩ binding mode observed for the C2B domain of rabphilin-3A in this study. Second, the binding of the linker to the C2B domain of rabphilin-3A in the presence of Ca 2ϩ probably triggers a closer proximity of both C2 domains. In synaptotagmins, an acidic motif connects their transmembrane part and the first C2 domain. In Doc2␣, such a motif connects the Munc13-1-interacting domain and the C2A domain. In analogy to rabphilin-3A, these acidic patches may be important to restrain the spatial localization of C2 domains to neighboring domains in the presence of Ca 2ϩ . Further investigations will be required to address the functional involvement of such motifs in C2 domain tandem proteins.