The Cytosolic Phospholipase A2 Catalytic Domain Modulates Association and Residence Time at Golgi Membranes*

Cytosolic phospholipase A2 (cPLA2) catalyzes release of arachidonic acid from membranes following translocation to Golgi and endoplasmic reticulum. In response to an intracellular calcium concentration ([Ca2+]i) increase, the C2 domain binds Ca2+ and brings the catalytic domain into proximity with its phospholipid substrate. Because membrane residence is important in the regulation of cPLA2 activity, we explored the contributions of the C2 and catalytic domains in mediating membrane residence using an imaging approach in live cells with fluorescent protein chimeras of cPLA2. The isolated cPLA2 C2 domain associated with Golgi membranes rapidly in proportion to the [Ca2+]i, allowing for its use as a [Ca2+]i indicator. cPLA2 association with Golgi was slower than the isolated C2 domain in response to a [Ca2+]i increase. After [Ca2+]i decrease, cPLA2 remained associated with membrane in a Ca2+-independent fashion whereas C2 domain rapidly dissociated. Ca2+-independent membrane association was greatly reduced by mutation of Trp464, located at the membrane-exposed face of the catalytic domain, to Gly or Ala. Mutation of Trp464 to Phe supported Ca2+-independent association similar to wild type. These results demonstrate a role for the cPLA2 catalytic domain in regulating membrane association and membrane residence time.

cPLA 2 1 plays a central role as regulator of arachidonic acid (AA) release from membranes in response to [Ca 2ϩ ] i mobilizing agents (1,2). The lipid second messenger AA functions directly to regulate a number of cellular processes and is the precursor for a variety of other lipid mediators, including prostaglandins and leukotrienes (3). Because cPLA 2 substrates are membrane phospholipids, a fundamental regulatory mechanism of cPLA 2 lies in controlling its access to membrane.
Ca 2ϩ sensing and membrane targeting of cPLA 2 are functions of the N-terminal calcium-dependent lipid-binding C2 domain (4 -9). These modules are found in many signaling proteins that function at membrane surfaces, including protein kinase Cs (PKC) (10,11), phospholipase C␦ (12), and synaptotagmins (13). C2 domains are structurally similar and are characterized by a ␤ sandwich composed of two four-stranded ␤ sheets and loops connecting the strands (14). In response to increases in [Ca 2ϩ ] i , three loops at one end of the structure coordinate the binding of 2 or 3 Ca 2ϩ and together define the membrane-interacting surface of the C2 domain. In the case of cPLA 2 , Ca 2ϩ binding neutralizes the negative charge at the tips of the Ca 2ϩ binding loops, allowing hydrophobic residues in the C2 domain to insert into the lipid surface and stabilize membrane association (5,6,(15)(16)(17). Disruption of either the Ca 2ϩ binding residues or the hydrophobic residues responsible for membrane penetration results in the loss of membrane binding in vitro (5,6) and the failure of membrane association in vivo (9,18). Deletion of the C2 domain also prevents membrane association in vivo (8,18).
Because of the central roles of Ca 2ϩ and the C2 domain in membrane binding, there have been several studies examining [Ca 2ϩ ] i requirements in vitro. Two early reports investigating full-length and isolated C2 domain binding have reported Ca 2ϩ requirements in the range of 100 -300 nM for cPLA 2 membrane association in vitro (4,19). However, most in vitro studies have used the isolated C2 domain and have reported a minimum [Ca 2ϩ ] of between ϳ2 and 11 M for membrane binding (6, 16, 20 -22). Diverse experimental conditions make comparisons between the in vitro studies difficult to interpret, but the magnitude of the [Ca 2ϩ ] predicted for binding from most of the in vitro studies is higher than would be expected in vivo for a cytosolic protein that responds to agonist-mediated [Ca 2ϩ ] i changes. An investigation of membrane binding in vivo reported that the [Ca 2ϩ ] i requirements for full-length and isolated C2 domain membrane association to be ϳ120 nM for Golgi and ϳ250 nM for ER membranes (8).
Investigation of full-length cPLA 2 binding in vivo has demonstrated that, following Ca 2ϩ -dependent membrane association, the full-length cPLA 2 remains associated with membranes in a Ca 2ϩ -independent manner (8,23). This suggests a possible role for the catalytic domain, or cooperativity between the C2 and catalytic domains in prolonging membrane residence time. Although studies have evaluated dissociation kinetics of the cPLA 2 C2 domain in vitro (20,21), there are few in vitro studies to date that have evaluated the contribution of the catalytic domain in Ca 2ϩ -independent association with membrane (24). We have used imaging of fluorescent protein-cPLA 2 chimeras to investigate further the role of the cPLA 2 catalytic domain in regulating membrane association and residence time in living cells.

MATERIALS AND METHODS
Fluorescent Protein Constructs-For imaging studies, DNAs encoding wild-type human cPLA 2 and the cPLA 2 C2 domain (residues 1-749 and 17-148, respectively; GenBank TM accession no. M72393) were cloned into the vector pEGFP-C3 (Clontech) to create pEGFP-cPLA 2 , and pEGFP-cPLA 2 C2, as previously described (8). pECFP-cPLA 2 W464G, pECFP-cPLA 2 W464A, and pECFP-cPLA 2 W464F were made from pECFP-cPLA 2 by site-directed mutagenesis (Stratagene). pECFP-GT, a Golgi marker encoding the N-terminal 82 residues of ␤-1,4-galactosyltransferase, was purchased from Clontech. The pEYFP plasmid used in this study was constructed from pEYFP-C1 (Clontech) by introducing a Q70M mutation by site directed mutagenesis (Stratagene) to improve its qualities (25). Different color fluorescent protein constructs were produced by interchanging NheI/BsrGI fragments encoding the fluorescent protein-encoding region between constructs. Expression of FP-tagged wild-type and mutant proteins at the predicted molecular weights was confirmed by immunoblot of protein lysates from MDCK cells transfected with each construct. A band at ϳ135 kDa was observed in protein lysates from cells expressing pECFP-cPLA 2 W464A, pECFP-cPLA 2 W464G, and pECFP-cPLA 2 W464F, which migrated at the same molecular weight as pEGFP-cPLA 2 (data not shown) (8). We have previously demonstrated that pEGFP-cPLA 2 C2 is expressed at the predicted molecular weight in MDCK cells (8). All constructs were confirmed by sequencing.
Microscopy of Fluorescent Proteins-Transfected MDCK cells grown on glass-bottomed culture dishes (MatTek) were washed with and incubated in Hanks' balanced salt solution additionally buffered with 25 mM HEPES, pH 7.4 (HHBSS). Cells were imaged using an Olympus inverted microscope equipped with a 60ϫ, 1.25 numeric aperture oil immersion objective, ECFP and EYFP emission filters (Chroma) in a Sutter filter wheel, a ECFP/EYFP dichroic mirror, and a TILL Imago CCD camera (TILL Photonics). Excitation light of 430 and 510 nm for ECFP and EYFP, respectively, was provided using a Polychrome IV monochromator (TILL Photonics). TILLvisION software was used for acquisition and analysis. Final images were produced using Adobe PhotoShop. Contamination of ECFP fluorescence into the EYFP channel was Ͻ0.2%, and Ͻ0.35% for EYFP into the ECFP channel.
FP fluorescence at Golgi (FP Golgi ) with respect to time was determined for a region of interest (ROI) corresponding to an area of Golgi membrane (ROI Golgi ) by the following method. Background-corrected average pixel values were determined for ROI Golgi at each time point. Because of photobleaching of the FPs during the course of the experiments, bleach correction factors were calculated for each time point by first determining the background-corrected average pixel values for an ROI corresponding to the entire cell (ROI cell ) and then normalizing those values to the initial value at t ϭ 0. The background-and bleachcorrected fluorescence value, F t , was calculated by dividing the background-corrected average pixel value for ROI Golgi at each time point by the bleach correction factor at the corresponding time point. The normalized fluorescence at Golgi was calculated by dividing F t by the t ϭ 0 fluorescence value, F 0 , to yield F t /F 0 . Because the F t /F 0 values are equivalent to the integrated fluorescence from the ROI Golgi divided by the integrated fluorescence from the ROI cell , the F t /F 0 values are proportional to the fraction of the total cellular fluorescence in the area of the Golgi. In addition, because ROI Golgi and ROI cell were identical for ECFP and EYFP images, direct comparisons of ECFP Golgi and EYFP Golgi with respect to time in the same cell could be made.
Ca 2ϩ Clamp Conditions-Transfected cells in glass-bottomed dishes were treated with buffer containing 3 mM EGTA in HHBSS (EH buffer) for Ͼ45 min to deplete internal Ca 2ϩ stores, and then made permeable to [Ca 2ϩ ] i with 10 M ionomycin in EH buffer. Cells exhibiting translo-cation after ionomycin addition were excluded from study. For the experiments to determine membrane association at static [Ca 2ϩ ] i (Figs. 1 and 8), a small volume of 100 mM CaCl 2 (ϳ20 -23 l, depending on the experiment) was added to 1 ml of the ionomycin/EH buffer covering the cells, the buffer was mixed, and the [Ca 2ϩ ] i allowed to establish an equilibrium with [Ca 2ϩ ] e for Ͼ5 min before imaging. The process of CaCl 2 addition, in increments of 2-3 l, mixing, equilibration, and imaging was repeated for each [Ca 2ϩ ] e step. As previously reported (8), this treatment results in incremental increases in [Ca 2ϩ ] i in the physiological range of ϳ10 -600 nM that remain stable for Ͼ10 min; however, the [Ca 2ϩ ] i was lower than the [Ca 2ϩ ] e at equilibrium.
For time-lapse experiments to determine kinetics of membrane association and residence (Figs. 2, 3, 4 (A and B), and 7), cells were incubated in EH buffer and permeabilized with ionomycin as above.
[Ca 2ϩ ] i increases were stimulated by exchanging the buffer over the cells with 1 ml of EH buffer containing 2.5 mM CaCl 2 and 10 M ionomycin at the times indicated. To decrease [Ca 2ϩ ] i , 1 ml of EH buffer was added at the times indicated.
For some time-lapse experiments (Figs. 4 (C and D) and 9), cells were not pre-incubated in EH buffer and the internal Ca 2ϩ stores remained intact. In these experiments, [Ca 2ϩ ] i was increased by addition of 100 l of 10 M ionomycin in EH to 900 l of EH buffer over the cells, which resulted in Ca 2ϩ release from internal stores. The [Ca 2ϩ ] i was subsequently decreased by chelation of [Ca 2ϩ ] i by the EH buffer in the bath. This second treatment resulted in a faster, higher amplitude [Ca 2ϩ ] i increase, but the results obtained were not different from the experiments where exogenous CaCl 2 was added to store-depleted cells.
Calcium Imaging-MDCK cells grown on glass-bottomed dishes were washed with HHBSS containing 1 mM probenecid and incubated with 5 M Fura2-AM (Calbiochem) in HHBSS, 1 mM probenecid, and 0.1% Me 2 SO for 45 min at 37°C. Cells were then washed with HHBSS containing 1 mM probenecid and imaged after a 30-min incubation for de-esterification of the Fura2-AM. Single-cell imaging was performed on the Olympus system described above, but using a 40ϫ, 1.35 numeric aperture oil immersion objective and a Fura2/EGFP dichroic mirror and emission filter (Chroma). Fura2/EGFP image sets were illuminated at 340, 380, and 465 nm at the time intervals specified. The [Ca 2ϩ ] i increase is expressed as the ratio of the background-corrected Fura2 fluorescence at 340 and 380 nm (26).
Structural Models-The cPLA 2 models used the Protein Data Bank identifier 1CJY (7). Structures were visualized with the SwissPDB Viewer 3.7 program by GlaxoSmithKline.

Expression and Localization of cPLA 2 in MDCK Cells-To
determine the intracellular membrane to which cPLA 2 translocates in response to increases in [Ca 2ϩ ] i in live cells, EYFP-cPLA 2 C2 was co-expressed with ECFP-GT, which localizes to medial and trans-Golgi cisternae (27), and imaged in timelapse in response to [Ca 2ϩ ] i mobilization. In response to 5 M ATP, EYFP-cPLA 2 C2 moved from the cytosol to juxtanuclear membranes (Fig. 1A), and co-localized with ECFP-GT (Fig. 1, B and C). This shows that cPLA 2 associates prominently with the Golgi in living cells and extends previous immunofluorescence findings in fixed and permeabilized MDCK cells (8).
cPLA 2 and the cPLA 2 C2 Domain Have Equivalent Ca 2ϩ Sensitivities for Membrane Association-In previous experiments, we correlated [Ca 2ϩ ] i with the extent of EGFP-tagged cPLA 2 (EGFP-cPLA 2 ) or EGFP-tagged cPLA 2 C2 domain (EGFP-cPLA 2 C2) association with Golgi and ER membranes and found the [Ca 2ϩ ] i threshold for association with Golgi to be ϳ120 and ϳ250 nM for ER and nuclear membrane (8). The subcellular localization of the C2 and full-length cPLA 2 constructs also appeared similar. However, we were unable to directly compare membrane association of both constructs simultaneously at any given [Ca 2ϩ ] i or directly compare targeting in the same cell. Here, we directly compared membrane association of EYFP-cPLA 2 C2 and ECFP-cPLA 2 co-expressed in cells under calcium clamp conditions, where [Ca 2ϩ ] e was increased in discrete steps and the chimeras imaged. At [Ca 2ϩ ] i less than 50 nM (resting level), fluorescence from both constructs was distributed in the cytoplasm (Fig. 1, D and H). EYFP-cPLA 2 C2 is also evident in the nucleoplasm, whereas ECFP-cPLA 2 is excluded because of differences in protein size, as we have previously shown (8).
At each step increase in [Ca 2ϩ ] e , roughly equivalent fractions of EYFP-cPLA 2 C2 and ECFP-cPLA 2 associated with Golgi ( Fig.  1, panels E and F and panels I and J). Because cPLA 2 associates with ER and nuclear membranes at [Ca 2ϩ ] i greater than ϳ250 nM, the [Ca 2ϩ ] i in images in panels E and F and panels I and J were judged to be below 250 nM. In Fig. 1 (G and K), [Ca 2ϩ ] i is above 250 nM, as evidenced by fluorescence from both constructs on the nuclear envelope (white arrowhead) as well as Golgi (black arrowhead). Although the ER is not resolved in panels G and K as it is in panel A, we have shown previously that isolated C2 and full-length cPLA 2 translocate to the nuclear membrane and ER at the same [Ca 2ϩ ] i (8). EYFP-cPLA 2 C2 and ECFP-cPLA 2 localized to the same intracellular membranes as shown in the merged image (Fig. 1L). These observations using dual imaging of cells co-expressing full-length cPLA 2 and the isolated C2 domain directly confirm that the C2 domain provides targeting specificity and that the isolated C2 domain and full-length cPLA 2 have equivalent Ca 2ϩ sensitivities.   (28). Based on our earlier work, translocation to Golgi, but not ER, occurred at [Ca 2ϩ ] i of 125-250 nM (8). By titrating the [Ca 2ϩ ] e in the extracellular buffer against the extent of EYFP-cPLA 2 C2 translocation (Fig. 1), we determined the [Ca 2ϩ ] e that reproducibly elicited association of cPLA 2 C2 with Golgi, but not ER (i.e. produced [Ca 2ϩ ] i of 125-250 nM). Cells expressing EGFP-cPLA 2 C2 and loaded with Fura2 were placed in calcium clamp conditions, and exogenous CaCl 2 was added at 7.5 s after initiation of imaging (between the first and second image set) to increase [Ca 2ϩ ] i , and EGTA was added at 247.5 s to decrease [Ca 2ϩ ] i . Representative images from a timelapse recording show transient association of EGFP-cPLA 2 C2 at the area of the Golgi ( Fig. 2A and Supplemental Movie 2A (available in the on-line version of this article)). The images show that a fraction of the cellular EGFP-cPLA 2 C2 associated with Golgi and that the association was limited primarily to Golgi, indicating that the [Ca 2ϩ ] i did not exceed ϳ250 nM in the cell. Graphical analysis of EGFP-cPLA 2 C2 and ratioed Fura2 fluorescence in the area of the Golgi (EGFP-cPLA 2 C2 Golgi and Fura2 Golgi ) with respect to time (Fig. 2B) shows that after the [Ca 2ϩ ] i exceeded the threshold for membrane association, EGFP-cPLA 2 C2 Golgi increased at a similar rate as the Fura2 Golgi . After addition of EGTA, both EGFP-cPLA 2 C2 Golgi and Fura2 Golgi decreased at similar rates until initial values were re-established. These results demonstrate that cPLA 2 C2 domain association with Golgi membranes is coincident with changes in [Ca 2ϩ ] i .
Contribution of the cPLA 2 Catalytic Domain in Membrane Association and Residence Time-The results above show that the isolated cPLA 2 C2 domain functions well as a [Ca 2ϩ ] i indicator in a physiological [Ca 2ϩ ] i range. This allowed for imaging studies to examine the relative contributions of the C2 and catalytic domains in regulating membrane association and residence time by co-expressing the cPLA 2 C2 domain and fulllength cPLA 2 , and directly comparing association and dissociation using the C2 domain as a monitor of [Ca 2ϩ ] i changes. For these experiments, it was necessary to utilize a system where [Ca 2ϩ ] i changes in cells were induced by sequential addition of exogenous Ca 2ϩ and EGTA to cells depleted of Ca 2ϩ stores and permeabilized with ionomycin. This method resulted in a relatively slow [Ca 2ϩ ] i increase and decrease, and allowed us to stimulate translocation of full-length cPLA 2 and the isolated C2 domain to Golgi, but not ER, by titrating exogenous Ca 2ϩ . In contrast, in response to ATP and thapsigargin, which both mobilize intracellular Ca 2ϩ , the cPLA 2 C2 domain associates first with both Golgi and ER, because of the high local [Ca 2ϩ ] i surrounding the ER immediately after stimulation, then dissociates from ER and remains associated with Golgi exclusively (8). However, unlike the cPLA 2 C2 domain, the full-length cPLA 2 associates primarily with Golgi in response to ATP or thapsigargin (8). The quantitative method we developed to compare translocation rates requires that full-length cPLA 2 and the isolated C2 domain associate with membranes in the same area of the cell, and so precluded use of ATP or thapsigargin as Ca 2ϩ mobilizing agonists.
Cells expressing ECFP-cPLA 2 and EYFP-cPLA 2 C2 were placed in calcium clamp conditions, where intracellular Ca 2ϩ stores were depleted and cells were made permeant to [Ca 2ϩ ] e with ionomycin. Representative frames from a time-lapse recording of such a cell shows association of ECFP-cPLA 2 and EYFP-cPLA 2 C2 with Golgi at the times indicated in response to addition of CaCl 2 to the bath at 7.5 s after initiation of imaging (Fig. 3 (A and B) and Supplemental Movie 3AB (available in the on-line version of this article)). Addition of EGTA at 217.5 s resulted in the reduction of EYFP-cPLA 2 C2 fluorescence at Golgi, but had little effect on the retention of ECFP-cPLA 2 at Golgi. Image sequences of another experiment in which the EGTA was added at 255 s, but where the C2 domain translocated to ER and Golgi, is provided in Supplemental Movie 1 (available in the on-line version of this article).
Graphical analysis shows that, in response to the [Ca 2ϩ ] i increase, EYFP-cPLA 2 C2 fluorescence at Golgi (EYFP-cPLA 2 C2 Golgi ) increased more rapidly than ECFP-cPLA 2 fluorescence at Golgi (ECFP-cPLA 2Golgi ) (Fig. 3C). Detailed analyses of increases in EYFP-cPLA 2 C2 Golgi and ECFP-cPLA 2Golgi in response to a buffered [Ca 2ϩ ] i increase reveals that the rate of association of EYFP-cPLA 2 C2 with Golgi is initially linear and slows after 2.5 min, whereas association of ECFP-cPLA 2 with Golgi continues to increase up to 4 min (Fig. 3C). Because ECFP-cPLA 2 and EYFP-cPLA 2 C2 are imaged in the same cell, and the F t /F 0 is directly proportional to the fraction of total cellular fluorescent protein at Golgi (see "Materials and Methods"), ECFP-cPLA 2Golgi and EYFP-cPLA 2 C2 Golgi may be compared directly (Fig. 3D). A comparison of the initial rates of EYFP-cPLA 2 C2 Golgi and ECFP-cPLA 2Golgi graphed on the same y axis shows a greater initial rate of association for EYFP-cPLA 2 C2 than for ECFP-cPLA 2 . We do not believe that the presence of isolated C2 domain is responsible for the relatively slower association of full-length cPLA 2 either by chelating available Ca 2ϩ or by out competing full-length for membrane binding sites for the following reasons. First, membrane association of both full-length and C2 cPLA 2 reach a plateau, as shown in the calcium clamp experiments (Fig. 1) and earlier work (8), so the extent of their translocation is proportional to [Ca 2ϩ ] i and equivalent fractions of both full-length and C2 cPLA 2 associate with Golgi at a given [Ca 2ϩ ] i (Fig. 1). The earlier plateau of the C2 domain suggests that it establishes an equilibrium with [Ca 2ϩ ] i before full-length cPLA 2 , and the con- tinued association of full-length with Golgi, after the slowing of isolated C2, suggests that Ca 2ϩ availability is not a limiting factor. Second, the relative rates of full-length and C2 cPLA 2 association with membrane remain constant for the first 2 min (Fig. 3D) and full-length cPLA 2 continues to associate with Golgi after C2 association rate slows, suggesting that available binding sites are also not a limiting factor. EYFP-cPLA 2 C2 Golgi returned to base-line levels ϳ90 s after addition of EGTA (Fig. 3C). In contrast, ECFP-cPLA 2Golgi remained elevated and was at ϳ75% of the peak value 8 min after EGTA addition, indicating that prolonged membrane retention mediated by the catalytic domain is Ca 2ϩ -independent. Ca 2ϩindependent membrane retention of full-length cPLA 2 was consistently observed in all experiments. Often, full-length cPLA 2 was clearly observed to be associated with Golgi membranes for Ͼ30 min after dissociation of the isolated C2 domain. Because of cell-specific differences in translocation resulting from differences between rates of [Ca 2ϩ ] i increase (after Ca 2ϩ addition) and decrease (after EGTA addition) and between the final [Ca 2ϩ ] i values reached before EGTA addition, we were not able to derive meaningful errors for fluorescence intensity change for translocation. Additional analyses of two cells treated as above are provided in Supplemental Fig. 1 (available in the on-line version of this article).
Differences in association and dissociation rates of the fulllength cPLA 2 and the C2 domain were consistently observed, even under conditions when initial association rates differed as a result of differences in the rate of [Ca 2ϩ ] i increase. For example, in Ca 2ϩ store-depleted cells (Fig. 4A), the rates of initial membrane association after Ca 2ϩ addition (Fig. 4B) were much slower than the rates of membrane association in Ca 2ϩ storereplete cells treated with ionomycin (Fig. 4, C and D). In the latter, the rate of full-length cPLA 2 association with Golgi was initially rapid, but slowed considerably after a few seconds and was slightly slower than that of the isolated C2 domain (Fig.  4D). The very rapid initial association of full-length cPLA 2 is most likely the result of a short-lived, high magnitude local [Ca 2ϩ ] i increase caused by ionomycin-mediated release of Ca 2ϩ from the intact stores, highlighting the role of local [Ca 2ϩ ] i in translocation of cPLA 2 , as we have previously noted (8).
Similar results to those above were obtained in control experiments where the fluorescent tags on the FP-cPLA 2 and FP-cPLA 2 C2 chimeras were switched (data not shown). In other control experiments using cells co-expressing ECFP-and EYFP-tagged C2 domain, rates of membrane association and dissociation were similar (data not shown). Rapid dissociation of C2 domain and slow dissociation of full-length cPLA 2 was also observed in experiments where EYFP-cPLA 2 C2 and ECFP-cPLA 2 fluorescence were associated with both Golgi and ER after Ca 2ϩ addition, indicating that the [Ca 2ϩ ] i was increased to greater than 250 nM (Supplemental Movie 1 and Supplemental Fig. 1 (A and B), available on-line).
These results demonstrate that the catalytic domain both retards Ca 2ϩ -mediated association of cPLA 2 with membrane and mediates a prolonged, Ca 2ϩ -independent membrane residence.

Membrane Residence in Response to Oscillations in [Ca 2ϩ ] i -Repetitive increases in [Ca 2ϩ
] i (Ca 2ϩ oscillations) may represent a normal physiological state in cells, and many cell responses and enzymes are sensitive to variations in the frequency of Ca 2ϩ oscillations (29 -32). To investigate the effect of Ca 2ϩ oscillations on the cPLA 2 C2 domain, cells expressing EGFP-cPLA 2 C2 were loaded with Fura2 and stimulated with 5 M ATP, and the EGFP-cPLA 2 C2 Golgi and Fura2 Golgi monitored over time. ATP elicited a sharp spike in Fura2 Golgi , which was followed by oscillations of diminishing magnitude that persisted for ϳ5 min (Fig. 5A). EGFP-cPLA 2 C2 Golgi temporally followed the Fura2 Golgi and was proportional to the magnitude of the Fura2 Golgi . These results demonstrate that the cPLA 2 C2 domain association with Golgi is sensitive to rapid changes in [Ca 2ϩ ] i and proportional to [Ca 2ϩ ] i .

FIG. 5. Effect of oscillations in [Ca 2؉ ] i on membrane residence.
A, cells expressing EGFP-cPLA 2 C2 were loaded with Fura2 and stimulated with 5 M ATP in HHBSS at 3 s after initiation of imaging. Fura2 Golgi and EGFP Golgi were monitored and graphed as in Fig. 2. B, cells expressing ECFP-cPLA 2 and EYFP-cPLA 2 C2 were stimulated with 5 M ATP in HHBSS at 3 s. ECFP-cPLA 2Golgi and EYFP-cPLA 2 C2 Golgi were monitored and graphed with respect to time as in Fig. 3. ECFP-cPLA 2Golgi and EYFP-cPLA 2 C2 Golgi are graphed on different axes to highlight the difference in time courses. C, analysis of the association of ECFP-cPLA 2Golgi and EYFP-cPLA 2 C2 Golgi on the same axis to directly compare rates of translocation between 0 and 20 s. From the results above, we expected that full-length cPLA 2 would accumulate at Golgi in response to [Ca 2ϩ ] i oscillations because of its slow dissociation rate. To investigate this hypothesis, cells co-expressing ECFP-cPLA 2 and EYFP-cPLA 2 C2 were stimulated with 5 M ATP and the ECFP and EYFP fluorescence at Golgi monitored. As shown in Fig. 5B, both EYFP-cPLA 2 C2 Golgi and ECFP-cPLA 2Golgi increased in response to [Ca 2ϩ ] i increase. Because of the different scaling of the y axis in Fig. 5B, ECFP-cPLA 2Golgi appears to increase before EYFP-cPLA 2 C2 Golgi in response to an ATP-induced [Ca 2ϩ ] i increase. However, when compared directly, EYFP-cPLA 2 C2 Golgi increases at a faster rate than ECFP-cPLA 2Golgi (Fig. 5C), as is observed in response to an ionomycin-induced [Ca 2ϩ ] i increase (Figs. 3 and 4).
Whereas the EYFP-cPLA 2 C2 Golgi declined to base-line levels between the initial [Ca 2ϩ ] i increase and the first Ca 2ϩ oscillation, the ECFP-cPLA 2Golgi remained elevated (Fig. 5B), although there was a slight decrease in fluorescence as we have noted in response to ATP and thapsigargin previously (8). The period between the initial increase and the first oscillation (Fig.  5B) is greater than that observed in Fig. 5A because of the cell-to-cell heterogeneity observed in [Ca 2ϩ ] i mobilization responses to ATP (33). To illustrate the differences between the behavior between the isolated C2 domain and the full-length protein, we chose an example where the fluorescence of the isolated C2 domain returned to base-line level between oscillations.
In response to the [Ca 2ϩ ] i increase associated with the first oscillation, EYFP-cPLA 2 C2 Golgi increased to less than one half of the level observed in response to the initial [Ca 2ϩ ] i increase (Fig. 5B). However, ECFP-cPLA 2Golgi increased to nearly the same level as in response to the initial [Ca 2ϩ ] i increase. This result highlights the effect of the catalytic domain on prolonging membrane residence in response to physiological [Ca 2ϩ ] i signals.
Trp 464 Mediates Ca 2ϩ -independent Membrane Residence of cPLA 2 -To investigate the mechanism for Ca 2ϩ -independent membrane residence of full-length cPLA 2 , we next examined its structural characteristics. The crystal structure of cPLA 2 reveals a helical region on the membrane binding face of the catalytic domain, previously identified as helix Fd, which constitutes part of the catalytic domain "cap" structure ( Fig.  6A) (7). Trp 464 orients outward at the top of helix Fd (Fig. 6B). The residues associated with this helical structure and the placement of the tryptophan are evolutionarily conserved in all sequenced cPLA 2 proteins (Fig. 6C). Because Trp residues are known to play critical roles in the interfacial binding of other PLA 2 isoforms (34), we investigated the role of Trp 464 in Ca 2ϩ -independent membrane residence by treating cells expressing cPLA 2 C2 and a cPLA 2 W464G mutant in calcium clamp conditions with CaCl 2 and EGTA, using the same protocol as in Fig. 3. Representative images from time-lapse imaging (Fig. 7, A and B) show that, in response to CaCl 2 addition, ECFP-cPLA 2 W464G and EYFP-cPLA 2 C2 both associated with Golgi. However, in contrast to full-length cPLA 2 , ECFP-cPLA 2 W464G rapidly dissociated from Golgi and ER following EGTA addition at 217.5 s (Fig. 7B). Graphical analysis shows that following addition of EGTA, ECFP-cPLA 2 W464G Golgi declined at a faster rate than EYFP-cPLA 2 C2 Golgi , but remained slightly elevated after ϳ300 s, whereas EYFP-cPLA 2 C2 Golgi returned to base-line levels (Fig. 7C). Similar results were observed using a Trp to Ala mutant, ECFP-cPLA 2 W464A (data not shown). These results suggest that Trp 464 in the catalytic domain is important in mediating Ca 2ϩ -independent membrane association.
Shorter residence time of the cPLA 2 W464G mutant on Golgi than the C2 domain was surprising and suggested that the [Ca 2ϩ ] i requirement for W464G association was greater than for the isolated C2 domain. To test this possibility, cells ex-pressing ECFP-cPLA 2 W464G and EYFP-cPLA 2 C2 were placed in calcium clamp conditions and subjected to step increases in [Ca 2ϩ ] i , as in Fig. 1 (D-K). Images show that at low [Ca 2ϩ ] i , which is comparable with resting [Ca 2ϩ ] i or lower, the distributions of ECFP-cPLA 2 W464G and EYFP-cPLA 2 C2 were similar except for the nucleoplasmic distribution of the C2 domain (Fig. 8, A and D). Following step increases in [Ca 2ϩ ] i , EYFP-cPLA 2 C2 associated with Golgi and ER membranes (Fig. 8B), indicating that the [Ca 2ϩ ] i exceeded 250 nM, but ECFP-cPLA 2 W464G remained cytoplasmic (Fig. 8E). This suggests that the cPLA 2 W464G mutant requires a higher [Ca 2ϩ ] i for membrane association (Ͼ250 nM) than does wild-type cPLA 2 (Ͼ120 nM). At still higher [Ca 2ϩ ] i (Fig. 8, C and F), ECFP-cPLA 2 W464G associated with Golgi and ER. The merged image of ECFP-cPLA 2 W464G and EYFP-cPLA 2 C2 at the highest [Ca 2ϩ ] i shows similar distributions of fluorescence (Fig. 8G), indicating that targeting of ECFP-cPLA 2 W464G is not altered.
Phe Can Substitute for Trp in Mediating Ca 2ϩ -independent Membrane Residence-To determine whether another aromatic residue, Phe, can substitute for Trp in mediating Ca 2ϩ -independent membrane residence, a cPLA 2 W464F mutant was constructed and assayed. Representative images from time-lapse analysis of cells, expressing ECFP-cPLA 2 W464F and EYFP- FIG. 7. Mutation of Trp 464 diminishes Ca 2؉ -independent membrane residence. Cells co-expressing ECFP-cPLA 2 W464G (A) and EYFP-cPLA 2 C2 (B) were placed in calcium clamp conditions, and CaCl 2 was added at 7.5 s after initiation of imaging. EGTA was added at 217.5 s. C, graphical analysis of ECFP-cPLA 2 W464G fluorescence from the area of the Golgi (ECFP-cPLA 2 W464G Golgi and EYFP-cPLA 2 C2 Golgi , inset, black line). cPLA 2 C2 and treated with ionomycin, show that both constructs associated with Golgi, but, although the EYFP-cPLA 2 C2 dissociated, the ECFP-cPLA 2 W464F remained with Golgi (Fig. 9, A and B), similar to what is observed using wild-type cPLA 2 (Fig. 3A). Graphical analysis of ECFP-cPLA 2 W464F Golgi and EYFP-cPLA 2 C2 Golgi (Fig. 9C) showed that the W464F mutant displayed a Ca 2ϩ -independent association with Golgi, similar to what was observed with wild-type cPLA 2 (Fig. 4C). This result suggests that Phe can substitute for Trp in mediating Ca 2ϩ -independent membrane residence. DISCUSSION cPLA 2 is a critical enzyme in the release of AA, an important lipid second messenger, and the precursor for biologically active oxygenated metabolites. Spatial and temporal regulation of cPLA 2 has been ascribed to the properties of the C2 domain, which functions as the targeting and calcium-sensing module of the protein, and we have confirmed these roles for the C2 domain in the work presented here. The surprising result is the unexpected role of the catalytic domain in regulating cPLA 2 residence time on membrane, in addition to its role in catalysis of membrane phospholipids.
For signaling proteins that associate with membranes for function, such as the C2 domain-containing PKCs (␣, ␤I, ␤II, and ␥), PLC␦, and cPLA 2 , regulation of membrane residence is crucial. A recent analysis of C2 domain translocation characteristics identified two discrete modes of signaling based on the persistence of C2 domain membrane residence as a function of temporal Ca 2ϩ signals (28). Transient association of C2 domain on membrane in response to Ca 2ϩ release resulted in what was termed a "differentiating" signaling mode, and sustained association in response to Ca 2ϩ influx was termed an "integrating" signaling mode. Our present work shows that, in the case of cPLA 2 , the inherently differentiating response of the C2 domain in response to transient [Ca 2ϩ ] i increase (transient association with membranes) is modified to an integrating response by the Ca 2ϩ -independent membrane association mediated by Trp 464 in the catalytic domain. Modification of membrane residence by adjoining domains is not unique to cPLA 2 ; PKC membrane residence mediated by the C2 domain is modified by C1 domain association with diacylglycerol in the membrane (35), although the effect of the PKC C1 domain mediating Ca 2ϩ -independent membrane association is more subtle than that observed here by the cPLA 2 catalytic domain. Because of the role of the catalytic domain in maintaining membrane residence revealed here, membrane-binding properties of fulllength cPLA 2 cannot be implied by the properties of the isolated C2 domain (36). This is exemplified by the experiments investigating membrane binding in response to Ca 2ϩ oscillations, where the isolated C2 domain returned to base-line levels during the period between [Ca 2ϩ ] i spikes, but the full-length cPLA 2 did not.
In the case of cPLA 2 , Ca 2ϩ -independent membrane residence in MDCK cells does not lead to an integrated catalytic response. In vivo experiments temporally correlating cPLA 2 on membrane with AA release have shown that release ceases within minutes after stimulation (8), although cPLA 2 remains associated with membranes, suggesting that the enzyme inactivates after a period of catalysis. In vitro experiments have also demonstrated this apparent "inactivation" of cPLA 2 (24). cPLA 2 also becomes "trapped" on negatively charged lipid bilayers and trapping may be caused in part by product accumulation (37). It is also possible that trapping may be the result of a Ca 2ϩ -independent association of the catalytic domain with anionic membranes (24). It has been suggested that trapping of cPLA 2 may be responsible for this apparent inactivation because of local substrate depletion. Experiments with a catalytically inert cPLA 2 S228A mutant or in the presence of pyrrolidine-1, a specific cPLA 2 inhibitor that blocks AA release in MDCK cells (38), showed no effect on membrane residence in our system (data not shown), suggesting that catalytic activity or product formation does not play a role in Ca 2ϩ -independent membrane residence. Thus, trapping in vitro and the Ca 2ϩindependent membrane residence in vivo may result from different mechanisms. We have also shown that Ca 2ϩ -mediated association of cPLA 2 with membrane is not in itself sufficient for optimal catalytic activity (39). Although the mechanism is unknown, these studies demonstrate an additional level of catalytic regulation aside from membrane residence. The results from the experiment investigating membrane residence in response to Ca 2ϩ oscillations elicited by low concentrations of ATP also suggest that additional factors may be important in cPLA 2 regulation. Should cells in tissues exhibit physiological Ca 2ϩ oscillations, a fraction of cPLA 2 may be constitutively associated with membranes, but the small amount of dissociation/reassociation may lead to a more prolonged low level AA release.
Phosphorylation is a major regulator of cPLA 2 activity (1, 2), and phosphorylation of cPLA 2 on Ser 505 by mitogen-activated protein kinases, including p42/p44 extracellular signal-regulated kinases and p38, and on Ser 727 by mitogen-activated protein kinase-interacting kinase I are important in regulating cPLA 2 -mediated AA release in vivo (40 -42). However, the mechanism by which phosphorylation of these sites regulates cPLA 2 is not clear. We have shown that the rates of association with Golgi of cPLA 2 S505A and S727A mutants are equivalent to wild type and that the mutants target the same membranes as wild type in MDCK cells (39). We found that cPLA 2 S505A and S727A mutations also exhibited Ca 2ϩ -independent membrane residence (data not shown). Likewise, incubation with the MEK1 inhibitor U0126, which results in down-regulation of extracellular signal-regulated kinase activity and inhibition of AA release, failed to affect membrane residence in MDCK cells (data not shown).
The imaging approach utilized here allowed us to investigate, in a physiological setting, the contributions of the C2 and catalytic domains in membrane association and residence. Direct comparison of full-length cPLA 2 and the isolated C2 domain in the same cell allowed us to control for the otherwise confounding problems of cell-to-cell heterogeneity and for cellto-cell differences in the rate and magnitude of [Ca 2ϩ ] i increase, which both influence membrane association. Using the C2 domain as a [Ca 2ϩ ] i indicator was critical in evaluating membrane residence because we found considerable heterogeneity in the response of [Ca 2ϩ ] i to EGTA in the calcium clamp In the co-expression experiments presented here, the slower association of cPLA 2 with Golgi is readily observed. This slower association of cPLA 2 , compared with the isolated C2 domain, may account for our previous observation that full-length cPLA 2 translocation is limited to Golgi in response to stimulation by agonists that elicit a rapid [Ca 2ϩ ] i transient, such as ATP or thapsigargin, but the isolated C2 domain translocates to ER, Golgi, and nuclear membrane in response to the same agonists (8). Release of Ca 2ϩ from internal stores results in a much faster rate of full-length association than does extracellular influx in store-depleted cells, suggesting that the local high [Ca 2ϩ ] i may be important in mediating translocation, as we have previously suggested (8). Interestingly, in response to agonists stimulating such local [Ca 2ϩ ] i increases, such as ATP and thapsigargin, a fraction of full-length cPLA 2 dissociates from Golgi in concert with the initial few seconds of [Ca 2ϩ ] i decrease, suggesting that there may be a certain residence time required for the Ca 2ϩ -independent membrane association, as has been previously suggested (23).
Many reports have demonstrated that hydrophobic residues in the calcium binding loops of the cPLA 2 C2 domain partially or fully insert into lipid membranes (5, 6, 15, 16, 43) (Fig. 10). Our results suggest that Trp 464 on the interfacial binding surface of the catalytic domain may also penetrate into the lipid bilayer when brought into proximity by the binding of the C2 domain while [Ca 2ϩ ] i is elevated, and that this penetration may be largely responsible for Ca 2ϩ -independent residence after [Ca 2ϩ ] i declines (Fig. 10). Membrane association via the C2 domain is perhaps critical in the correct orienting of the catalytic domain and in providing the energy for Trp 464 insertion as the cPLA 2 catalytic domain by itself fails to bind membrane in live cells (8,18). Generally, Trp plays an important role in interfacial binding of proteins to membranes as a result of its high affinity for the membrane-water interface (44). More specifically, Trp on the interfacial binding surfaces of other PLA 2 isoforms is important in binding and activity (34). Mutagenesis studies of cobra venom PLA 2 demonstrate that aromatic residues on the binding surface are critical for monolayer penetration and activity (45). Similarly, a single Trp in the membranebinding surface of human group V PLA 2 is critical for penetration and activity (46). In both cobra venom and group V PLA 2 proteins, the Trp determines enzyme preference for zwitterionic phospholipids such as phosphatidylcholine (PC) (45,46). The importance of interfacial Trp in PC preference and in membrane binding is exemplified by experiments where addition of a Trp residue in human group IIa PLA 2 greatly increased affinity for and hydrolysis of PC vesicles and cell membranes (47). Interfacial Trp has also been implicated in binding of bacterial phosphoinositide-specific phospholipase C to membranes (48). Although the C2 domain is responsible for targeting and also shows preference for zwitterionic phospholipids (6,15,17,49), the interfacial binding of the catalytic domain may provide a secondary mechanism of PC preference.
The W464G mutant required a higher [Ca 2ϩ ] i than the isolated C2 domain for membrane association. This may be the result of the presence of other residues in the catalytic domain that decrease the overall binding provided by the C2 domain. Possibly surface-exposed glutamic acid residues in the catalytic domain "lid" that covers the cPLA 2 active site (7) (Fig. 10) provide electrostatic repulsive forces that repel the catalytic domain from the membrane surface when Trp 464 is mutated. Mutation of glutamic acid residues in the lid increases the affinity of cPLA 2 for anionic membranes, and it has been proposed that repulsive forces between the lid and the membrane surface are required for moving the lid aside to expose the active site (50). Possibly tethering of the catalytic domain by Trp 464 acts a fulcrum for this proposed electrostatic repulsion that exposes the active site.