Calcium Binding Rigidifies the C2 Domain and the Intradomain Interaction of GIVA Phospholipase A2 as Revealed by Hydrogen/Deuterium Exchange Mass Spectrometry*

The GIVA phospholipase A2 (PLA2) contains two domains: a calcium-binding domain (C2) and a catalytic domain. These domains are linked via a flexible tether. GIVA PLA2 activity is Ca2+-dependent in that calcium binding promotes protein docking to the phospholipid membrane. In addition, the catalytic domain has a lid that covers the active site, presumably regulating GIVA PLA2 activity. We now present studies that explore the dynamics and conformational changes of this enzyme in solution utilizing peptide amide hydrogen/deuterium (H/D) exchange coupled with liquid chromatographymass spectrometry (DXMS) to probe the solvent accessibility and backbone flexibility of the C2 domain, the catalytic domain, and the intact GIVA PLA2. We also analyzed the changes in H/D exchange of the intact GIVA PLA2 upon Ca2+ binding. The DXMS results showed a fast H/D-exchanging lid and a slow exchanging central core. The C2 domain showed two distinct regions: a fast exchanging region facing away from the catalytic domain and a slow exchanging region present in the “cleft” region between the C2 and catalytic domains. The slow exchanging region of the C2 domain is in tight proximity to the catalytic domain. The effects of Ca2+ binding on GIVA PLA2 are localized in the C2 domain and suggest that binding of two distinct Ca2+ ions causes tightening up of the regions that surround the anion hole at the tip of the C2 domain. This conformational change may be the initial step in GIVA PLA2 activation.

The cytosolic group IVA (GIVA) 3 phospholipase A 2 (PLA 2 ), also known as cPLA 2 , was purified and cloned in 1991 (1,2). It is one of the few phospholipases in the phospholipase A 2 super-family shown to be important in lipid mediator biosynthesis (3,4). GIVA PLA 2 hydrolyzes membrane phospholipids at the sn-2 position to release free arachidonic acid, which is the precursor of numerous eicosanoids including the prostaglandins and leukotrienes (5,6) involved in the inflammatory and pain response (7)(8)(9). Understanding the regulation of the catalytic activity of GIVA PLA 2 is crucial for understanding eicosanoid metabolism.
The activity of the 85-kDa GIVA PLA 2 has been suggested to be regulated by several factors including the intracellular Ca 2ϩ concentration, its phosphorylation state, and the binding of various activators. Since the GIVA PLA 2 was discovered in 1986, the activity of this enzyme has been known to be Ca 2ϩdependent (10,11). The Ca 2ϩ -dependent lipid-binding domain (C2 domain) at the N terminus, which is linked to a catalytic domain where phospholipid hydrolysis occurs, was later identified (12). Further studies of the GIVA PLA 2 activity showed up-regulation by p38 protein kinase mainly through Ser-505 phosphorylation (13)(14)(15). Various membrane-associated activators have been shown to bind to GIVA PLA 2 and upregulate its activity. In particular, we have shown that the membrane-associated phosphatidylinositol 4,5-bisphosphate can bind to the "lysine pocket" of GIVA PLA 2 with high affinity and specificity to activate the GIVA PLA 2 independent of Ca 2ϩ (16,17). Another membraneassociated activator, ceramide 1-phosphate, has been shown to bind to the cation groove of the C2 domain to activate GIVA PLA 2 in a Ca 2ϩ -dependent manner (18,19).
Calcium binding is crucial for GIVA PLA 2 activation. Two calcium ions coordinate with the calcium-binding loops, CBL1, CBL2, and CBL3, in the C2 domain, which is illustrated in known C2 domain structures (20 -22). These two Ca 2ϩ ions neutralize the negative charge in the anion hole to facilitate the C2 domain's hydrophobic interaction with phospholipid membranes (21,22). Various models of membrane penetration for the GIVA PLA 2 have been proposed to explain the hydrolysis activity of GIVA PLA 2 (23,24). However, the x-ray crystal structure cannot fit the model perfectly unless there is a slight twist of the catalytic domain. Intradomain interactions may change the position of the C2 domain relative to the catalytic domain from that assumed in the crystal structure interpretation. Also, structural information regarding the calcium-free form of GIVA PLA 2 is limited, because its NMR and x-ray crys-tal structure are not available. It is necessary to analyze the entire GIVA PLA 2 and especially its intradomain interactions to fully understand the regulation of GIVA PLA 2 .
Peptide amide hydrogen/deuterium exchange coupled with liquid chromatography mass spectrometry (DXMS) has been widely used to analyze the interface of protein-protein interactions (25,26), protein conformational changes (27,28), and protein dynamics (29,30). The present study represents the first attempt to use DXMS to study intradomain interactions and the Ca 2ϩ binding effect on the GIVA PLA 2 . This technique should complement what can be learned from NMR and x-ray crystallographic analysis. Our results describe the intradomain interactions between the C2 domain and the catalytic domain and confirm that the major conformational responses to Ca 2ϩ are localized to the calcium-binding loop of the C2 domain. A tightened conformation of the C2 domain appears upon Ca 2ϩ binding. Herein, we explore the resultant effects on the catalytic domain.

EXPERIMENTAL PROCEDURES
Materials-All reagents were analytical reagent grade or better.
Protein Expression and Purification-C-terminal His 6tagged GIVA PLA 2 , the C2 domain, and the catalytic domain were expressed using recombinant baculovirus in a suspension culture of Sf9 insect cells (16). The cell pellet was lysed in 25 mM Tris-HCl, pH 8.0, 150 mM NaCl, 2 mM ␤-mercaptoethanol, and 2 mM EGTA, and then the insoluble portion was removed by centrifugation at 12,000 ϫ g for 30 min. The supernatant was passed through a column comprised of 6 ml of nickel-nitrilotriacetic acid-agarose (Qiagen, Valencia, CA). The protein in the native state was eluted in the "protein buffer" (25 mM Tris-HCl, pH 8.0, 100 mM NaCl, 125 mM imidazole, and 2 mM dithiothreitol). The protein concentration was measured using the Bradford assay, and the activity was assayed using mixed micelles in the modified Dole assay (31,32). Purified GIVA PLA 2 (2 mg/ml) was stored in the protein buffer on ice for DXMS experiments (16).
Preparation of Deuterated Samples-D 2 O buffer contained 50 mM HEPES (pH 6.9), 100 mM NaCl, 2 mM dithiothreitol, (Ϯ) 200 M CaCl 2 . Hydrogen/deuterium exchange experiments were initiated by mixing 20 l of GIVA PLA 2 , the C2 domain, or the catalytic domain (containing 40 g) in protein buffer with 60 l of D 2 O buffer to a final concentration of 70.4% D 2 O at pH 7.0. In calcium binding experiments, the GIVA PLA 2 in protein buffer was preincubated in the presence of 240 M CaCl 2 in a 23°C water bath for 5 min. The D 2 O buffer was then added, and the samples were incubated at 23°C for an additional 10 s, 30 s, 100 s, 300 s, 1000 s, 3000 s, or 200 min. Each time point of the intact GIVA PLA 2 was repeated three independent times. The deuterium exchange was quenched by adding 120 l of ice-cold quench solution (0.96% formic acid, 1.66 M guanidine hydrochloride) that acidified the sample to a final pH 2.5. The samples were placed on ice at the quenched condition for 10 min to partially denature the protein for the purpose of better peptide map coverage and then frozen on dry ice. Fully deuterated samples were incubated in 2.5 M guanidine hydrochloride, 75% D 2 O at 23°C for 2 days. Vials with frozen samples were stored at Ϫ80°C until analysis, usually within 3 days.
Proteolysis Liquid Chromatography-Mass Spectrometry Analysis of Samples-All steps were performed at 0°C as previously described (26,28). The samples were hand-thawed on melting ice and injected onto and passed through a protease column (66-l bed volume) filled with porcine pepsin (Sigma; immobilized on Poros 20 AL medium at 30 mg/ml following the manufacturer's instructions, Applied Biosystems) at a flow rate of 100 l/min with 0.05% trifluoroacetic acid. The eluate from the pepsin column was directly loaded onto a C18 column (Vydac cat. 218MS5150). The peptides were eluted at 50 l/min with a linear gradient of 0.046% trifluoroacetic acid, 6.4% (v/v) acetonitrile to 0.03% trifluoroacetic acid, 38.4% acetonitrile for 30 min. The eluate from the C18 column was directed to a Finnigan Classic LCQ mass spectrometer via its ESI probe operated with a capillary temperature of 200°C as previously described (26,28).
Data Processing-SEQUEST software (Thermo Finnigan Inc.) was used to identify the sequence of the peptide ions. DXMS Explorer (Sierra Analytics Inc, Modesto CA) was used for the analysis of the mass spectra as previously described (26,28). All selected peptides had first passed the quality control threshold of the software and were then manually checked for the mass envelope fitting with the calculated mass envelope for data reduction. The highest signal/noise ion was picked if multiple ionization charges (1, 2, or 3) of a peptide were detected. Normally, the peptide with the lower charge gave a better signal. Incorporated deuteron number was the centroid shift between the non-deuterated and the partially deuterated mass envelope. The deuteration level of each peptide was calculated by the ratio of the incorporated deuteron number to the maximum possible deuteration number.

GIVA PLA 2 Coverage Map of Pepsin Fragmentation-The
protein digestion procedure was optimized to produce a peptide map that yielded the best coverage of GIVA PLA 2 . These optimizations included testing various denaturants and denaturing conditions, denaturation times, amounts of protein, and flow rates for the online pepsin digestion. The final method is described under "Experimental Procedures." The optimized condition gave a 92% coverage of the GIVA PLA 2 sequence with 157 distinct peptic peptides being identified, see Fig. 1. The C2 domain and the catalytic domain alone generated the same digestion pattern as their respective parts did in the intact GIVA PLA 2 . There are 58 unresolved amino acid residues. These are due to regions without identified peptides and the fact that the first two residues of each peptide are rapidly exchanging and do not retain deuterons during the processing. Although we analyzed all 157 peptides, some regions of the protein contained several overlapping peptides. In some cases, comparing the deuterium levels of overlapping peptides will yield information about the deuterium levels of a subsection of the peptide. Thus, one can obtain information about regions of the protein smaller than the peptides covering it. In many cases, however, the orientation of the overlapping peptides is such that no additional information can be obtained. Also, in some cases, the error propagation through these calculations yields large variances that prevent further analysis. Because of these latter two reasons, peptides that could not provide additional information were eliminated. In the end, 59 peptide fragments, which cover 82% of the protein, were employed. In this article, when we use the term "peptide," we are referring to an actual peptide that was identified in the pepsin digestion. When we use the term "region," we are referring to a section of the protein for which the deuterium exchange has been calculated but may or may not correspond to an actual pepsin peptide.
Deuterium On-exchange of GIVA PLA 2 -Full deuteration was also tried under various conditions such as 2 days of exchange in 0.5% formic acid, high temperature exchange for 4 h (50°C), and partial denaturation in 2.5 M guanidine hydrochloride. The resulting low deuteration observed for the ␣/␤ hydrolase region indicated that the protein was not fully deuterated under all conditions. The peptide fragment 266 -279 containing a solvent-exposed loop-helix structure had the highest deuteration level (89%). Hence, this peptide was used as a system back-exchange control. Back-exchange (12%) according to the fully deuterated control and the deuteration level (70.4%) was corrected for the centroid shifts.
Deuterium on-exchange experiments of GIVA PLA 2 (Ϯ) Ca 2ϩ were incubated at pH 7.0, 23°C for seven different time points ranging from 10 s to 200 min and were processed as described above. Under these conditions of temperature, pH, and time, the deuteration levels of the peptides were well distributed between 3 and 90%, and there was a 12% back-exchange. The deuteration level after deuterium on-exchange over the time course is shown in Fig. 2. Fig. 1 shows the actual physical peptides that were isolated, detected, and analyzed via DXMS. Fig. 2 shows the deuterium exchange levels of various regions of the protein, not the peptides per se. The presence of overlapping peptides in Fig. 1 allows one to define the deuteration level of regions of the protein smaller than the peptides by comparing the deuterium levels of several overlapping peptides. Note that the first two N-terminal residues of each pepsin-digested DXMS peptide do not retain deuterium after the pepsin and C18 columns; this fact is also taken into account when calculating the deuteration levels of the protein regions shown in Fig. 2 (28, 33).
The deuteration level after 3000 s on-exchange was plotted onto the GIVA PLA 2 structure (Protein data bank code 1cyj) using a color index to illustrate the H/D exchange results for the structure as shown in Fig. 3. The C2 domain consists of 8 antiparallel ␤-strands that are roughly divided into two domains. The first contains the ␤6-, ␤7-, and part of the ␤8-strands (which includes residues 6 -13, 28 -39, 73-76, and 125-129) and exhibits fast deuterium exchange. The second domain contains ␤1-, ␤2-, and ␤5-strands (which include residues 16 -25, 41-48, and 91-101) and exhibits slow H/D exchange. The fast exchanging regions of the C2 domain tend to face away from the catalytic domain. The slow exchanging regions are on the opposite side of the C2 domain and face the cleft between the C2 and catalytic domains. The crystal structure indicates that there is a significant amount of open space between these two domains, but of course the crystallographic data reflect the crystal packing rather than the solution conformation. This would predict that there would be significant accessibility of water to the internal face of the C2 domain, see Fig. 3. The slow exchange in this region of the C2 domain could be due to the rigidity of the protein structure here, or it could be due to the fact that when free in solution, the C2 domain is folded against the catalytic domain.
There are five regions in the x-ray crystal structure that could not be resolved because of the flexibility of the chains. These are indicated in Fig. 3 as dashed lines. It should be noted, that the DXMS method was able to detect and measure H/D exchange in each of these regions. As expected, these regions showed significant levels of exchange. Two of these regions are the hinges for the catalytic site lid (22).
Deuterium On-exchange of the C2 Domain and the Catalytic Domain of GIVA PLA 2 -To test the hypothesis that there is a dynamic interaction between the C2 domain and the catalytic domain, we expressed and purified His 6 -tagged C2 and His 6tagged catalytic domains individually and conducted DXMS experiments on both under the same conditions as the intact GIVA PLA 2 . The differences in the deuteration level between the C2 domain alone and the intact GIVA PLA 2 are shown in Fig. 4. The C2 domain alone showed greater than a 20% increase of solvent accessibility (residues 41-48, 91-95, and 51-71) on ␤2, ␤3, ␤4, and ␤5 in the cleft. Most regions in ␤1, ␤6, ␤7, and ␤8 outside of the cleft were unchanged, except for region 125-129, which was the only one showing a greater than 10% increase, but it was not significant after 100 s. This result indicates that the catalytic domain either interacts with the C2 domain or it changes the conformation of the C2 domain through the linker region, leading to an increase in its solvent accessibility (Fig. 4).
To find out if the increase in the solvent accessibility of the C2 domain is through the contact with the catalytic domain, we carried out hydrogen/deuterium exchange on the isolated catalytic domain (Fig. 5)   catalytic domain and allowed us to map out the interaction regions between them.
Calcium Binding Effects on GIVA PLA 2 -The Ca 2ϩ -binding site on GIVA PLA 2 consists of the three distinct Ca 2ϩ -binding loops CBL1, CBL2, and CBL3. This region has been shown to be important for lipid interface binding with the CBL1 and CBL3 regions playing a dominant role (23,34,35). We carried out the on-exchange in the presence and absence of 200 M Ca 2ϩ . We first examined the residues that directly interact with Ca 2ϩ . The two Ca 2ϩ ions coordinate with Asp-40, Thr-41, and Asp-43 in CBL1, with Asn-65 in the CBL2 region, and Asp-93, Ala-94, and Asn-95 in CBL3. Thr-41 and Asp-43 in CBL1 exhibit slow exchange in the absence of Ca 2ϩ (Fig. 6). There was only an additional 5% decrease in H/D exchange at 200 min when Ca 2ϩ was bound. Residue Asp-40 in CBL2 is in a gap between two resolved DXMS peptides, and thus we have no H/D exchange information for this residue. Asp-93, Ala-94, and Asn-95 in CBL3 are located in the DXMS peptide 91-101, and this peptide exhibited only a 10% H/D exchange after 200 min without Ca 2ϩ . Its exchange rate did not change with Ca 2ϩ .
Asn-65 in CBL2 is the only residue that is located on a DXMS peptide that showed a 10% decrease in H/D exchange upon Ca 2ϩ binding (see Fig. 6). However, the resolution in this region was not sufficient to assign the H/D exchange changes to Asn-65. There were, however, dramatic changes in other areas of the C2 domain that indicated a conformational change in the protein. Four regions in the C2 domain had a significantly decreased H/D exchange rate; these were residues 26 -39, 51-71, 102-106, and 125-129. The incorporated deuterium number over the time course in these four regions is plotted in Fig. 7. The average change in deuteration level of the FOUR time points from 5-200 min were also mapped onto the GIVA PLA 2 structure in the same figure.
We also examined H/D exchange levels on the rest of the protein and found no significant change in the H/D exchange levels in the catalytic domain. It is worth noting that the linker region (residues 139 -146) showed a 6% solvent accessibility decrease upon calcium binding, which suggests minor interaction or orientation changes between the C2 and the catalytic domain.

DISCUSSION
DXMS results provide a detailed H/D exchange rate profile of the catalytic domain. The crystal structure of the GIVA PLA 2 contains a lid in the cap region that is tethered to the rigid catalytic core by two very flexible peptides and that blocks the catalytic site in the crystal structure. The lid apparently must move out of the way so that substrate can bind to the catalytic site. It has also been suggested that the cap is involved in membrane association (24). Our deuterium exchange results are consistent with these hypotheses. The lid and the tethers on either side of it show fast exchange. Interestingly, the deuterium DXMS method was able to detect and measure the exchange of the two tethers that could not be seen via x-ray crystallography and have confirmed that they have fast exchange. These results are consistent with the hypothesis that the lid is a flexible regulatory structure that can move when the enzyme interacts with lipid interfaces to expose the active site in the rigid catalytic domain.
The GIVA PLA 2 has been shown to translocate to the Golgi and perinuclear regions of cells upon the release of Ca 2ϩ (36 -38). This movement has been suggested to be one important way for the cell to control the action of this enzyme and eicosanoid production. The C2 domain is clearly responsible for this action. The Ca 2ϩ -dependent membrane docking of the C2 domain alone of this and many other proteins has been extensively studied to understand how its structure affects membrane penetration and the sequestration of the enzyme at the lipid surface. It is believed that the C2 domain allows the GIVA PLA 2 to bind to the membrane and that it positions the active site in a way that facilitates the binding of the substrate. Two Ca 2ϩ ions coordinate with Asp and Asn residues in the anion hole on the tip of the C2 domain to neutralize negative charges in this region. The hydrophobic residues on the tip of CBL1 (Phe-35, Met-38, and Leu-39) and CBL3 (Tyr-96, Val-97, and Met-98) interact with the hydrophobic phospholipid headgroup (21). After the binding to the phospholipid bilayer, the dissociation rate of Ca 2ϩ decreases (39,40). Ca 2ϩ binding may cause a conformational change of the C2 domain, which effectively hides the hydrophilic residues and exposes the hydrophobic residues.
One of the most dramatic differences that we observed occurred in the C2 domain. The surface of the C2 domain that faces away from the catalytic domain showed significant H/D exchange, whereas the opposite surface that faces the catalytic domain showed slow exchange. H/D exchange rates are affected by accessibility to water and can reflect the rigidity of the protein secondary structure and the strength of the hydrogen bonds. Both of these regions are composed of antiparallel ␤-sheets, and one might expect that they would have the same structural characteristics, and thus the H/D exchange rates would be similar. The decreased exchange on the surface facing the catalytic domain could be due to the fact that in solution the C2 domain rests up against the catalytic domain, and thus water accessibility is decreased. The slow exchange regions in the cleft between the C2 domain and the catalytic domain had over a 20% increase of the solvent accessibility when we conducted the DXMS experiments on the C2 domain or the catalytic domain alone. This would imply that the cleft delineated in the x-ray structure is in a different orientation of the protein when there is no calcium binding. It should be noted that the C2 and catalytic domains are held together by a single short peptide linker that should be fairly flexible. The interaction or the contact regions between the C2 domain and the catalytic domain can be mapped out by the solvent accessibility difference of the two individual domains and the intact GIVA PLA 2 (Figs. 4 and 5).
To compare the effects of calcium binding on structural perturbations and charge neutralization, the electrostatic potential and the Ca 2ϩ binding effects measured by DXMS were plotted onto the molecular model of GIVA PLA 2 (Fig. 8). The anion hole surrounded by CBL1, CBL2, and CBL3 on the top of the C2 domain is highly negatively charged. The cleft between the C2 and catalytic domains has a low to positive electrostatic potential (Fig. 8a). We also showed low solvent accessibility of the ␤2 and ␤5 sheets (Fig. 3). The linker region between the C2 domain and the catalytic domain is possibly a flexible loop, because the regions 132-136 and 147-156 may be overdigested by pepsin and were not covered in the peptide map (Fig. 1). This evidence suggests that the C2 domain is probably in physical contact with the catalytic domain. Additionally, the two Ca 2ϩ ions FIGURE 8. The electrostatic potential and the decreased deuteration level of the C2 domain in the intact GIVA PLA 2 . a, the electrostatic potential mapped to the molecular surface is calculated by Swiss-Pdb Viewer 3.7, based on the simple coulomb interaction. b, the decreased deuteration level is mapped to the molecular surface with the same orientation as the electrostatic potential.