The Calcium Binding Loops of the Cytosolic Phospholipase A₂ C2 Domain Specify Targeting to Golgi and ER in Live Cells

Abstract

Translocation of cytosolic phospholipase A₂ (cPLA₂) to Golgi and ER in response to intracellular calcium mobilization is regulated by its calcium-dependent lipid-binding, or C2, domain. Although well studied in vitro, the biochemical characteristics of the cPLA₂C2 domain offer no predictive value in determining its intracellular targeting. To understand the molecular basis for cPLA₂C2 targeting in vivo, the intracellular targets of the synaptotagmin 1 C2A (Syt1C2A) and protein kinase Cα C2 (PKCαC2) domains were identified in Madin-Darby canine kidney cells and compared with that of hybrid C2 domains containing the calcium binding loops from cPLA₂C2 on Syt1C2A and PKCαC2 domain backbones. In response to an intracellular calcium increase, PKCαC2 targeted plasma membrane regions rich in phosphatidylinositol-4,5-bisphosphate, and Syt1C2A displayed a biphasic targeting pattern, first targeting phosphatidylinositol-4,5-bisphosphate-rich regions in the plasma membrane and then the trans-Golgi network. In contrast, the Syt1C2A/cPLA₂C2 and PKCαC2/cPLA₂C2 hybrids targeted Golgi/ER and colocalized with cPLA₂C2. The electrostatic properties of these hybrids suggested that the membrane binding mechanism was similar to cPLA₂C2, but not PKCαC2 or Syt1C2A. These results suggest that primarily calcium binding loops 1 and 3 encode structural information specifying Golgi/ER targeting of cPLA₂C2 and the hybrid domains.

INTRODUCTION

Correct spatiotemporal targeting of proteins to specific cellular sites is critical in the regulation of cell signaling (Teruel and Meyer, 2000), and many cytoplasmic signaling proteins that function at membrane surfaces rely on modular domains, such as C2, pleckstrin homology (PH), Phox, FYVE, and ENTH domains (Rizo and Sudhof, 1998; Itoh and Takenawa, 2002) for their targeting. Of these, C2 domains differ by responding to intracellular Ca²⁺ signals, thus providing a link between receptor-mediated Ca²⁺ signals and interaction of host proteins with membrane (Oancea and Meyer, 1998; Teruel and Meyer, 2002).

In response to intracellular Ca²⁺ signals, cPLA₂ translocates to Golgi and ER where it catalyzes the release of arachidonic acid (AA) from phospholipids (Glover et al., 1995; Schievella et al., 1995; Hirabayashi et al., 1999; Perisic et al., 1999; Choukroun et al., 2000; Evans et al., 2001). Targeting specificity and sensitivity to Ca²⁺ is provided to the intact enzyme by its N-terminal C2 domain (Nalefski et al., 1994; Perisic et al., 1998; Dessen et al., 1999; Gijón et al., 1999; Hirabayashi et al., 1999; Perisic et al., 1999; Evans et al., 2001), which is similar to C2 domains from other signaling proteins, such as PKCα and Syt1 (Sutton et al., 1995; Nalefski and Falke, 1996; Shao et al., 1998; Verdaguer et al., 1999) (Figure 1, A and B). C2 domains are composed of ∼130 amino acids that fold into a β sandwich structure and are found in two topologies (types I and II) due to a circular permutation in the amino acid sequence (Nalefski and Falke, 1996). Most of the homology among C2 domains is found in the conserved core region that supports the surface-exposed calcium binding loops (CBLs) (Sutton et al., 1995). Less homology is found among the CBLs themselves, with cPLA₂C2 having a unique α-helical region in the first CBL.

Figure 1.

Structural and sequence comparisons of cPLA₂C2, PKCαC2, and Syt1C2A domains. (A) Ribbon diagrams of cPLA₂C2, PKCαC2, and Syt1C2A (Protein Data Bank entries 1RLW, 1DSY, and 1BYN, respectively) show CBLs 1-3 and β strands 2 and 3 of cPLA₂C2 (homologous to β strands 3 and 4 of PKCαC2 and Syt1C2A). (B) Amino acid sequence alignment of cPLA₂C2 (residues 17-138, AAB00789), PKCαC2 (residues 158-277, P17252), and Syt1C2A domains (residues 140-262, P21707). The residues included in the CBLs are shaded in gray and the residues in the conserved core are under the black line.

For many C2 domains, including PKCαC2 and synaptotagmin 1 C2A (Syt1C2A), Ca²⁺ binding results in the negative electrostatic potential in the area of the CBLs becoming positive, thereby promoting electrostatic interactions with negatively charged membranes (Shao et al., 1997; Rizo and Sudhof, 1998; Murray and Honig, 2002). As a result, Syt1C2A and PKCαC2 preferentially bind to anionic phospholipids, such as phosphatidylserine (PS) and phosphatidylinositolpolyphosphates, but not to phospholipids with neutral headgroups, such as phosphatidylcholine (PC) (Schiavo et al., 1996; Arcaro et al., 1998; Davletov et al., 1998; Zhang et al., 1998; Corbalán-García et al., 1999; Verdaguer et al., 1999; Gerber et al., 2001; Nalefski et al., 2001; Kohout et al., 2002; Ochoa et al., 2002). In contrast to the Syt1C2A and PKCα C2 domains, hydrophobic interactions predominate in the interaction of the cPLA₂C2 domain with membranes, and Ca²⁺ binding neutralizes the negative electrostatic potential surrounding hydrophobic residues located at the tips of the cPLA₂C2 CBLs 1 and 3, facilitating their penetration into the membrane surface (Davletov et al., 1998; Perisic et al., 1998; Xu et al., 1998; Bittova et al., 1999; Perisic et al., 1999; Gerber et al., 2001; Nalefski et al., 2001). Also, unlike Syt1C2A and PKCαC2, cPLA₂C2 preferentially binds to lipid vesicles composed entirely of PC (Nalefski et al., 1997; Davletov et al., 1998; Hixon et al., 1998; Nalefski and Falke, 1998; Perisic et al., 1999; Gerber et al., 2001) but can also bind to membranes composed partially, but not entirely, of anionic phospholipids (Davletov et al., 1998; Gerber et al., 2001; Nalefski et al., 2001; Stahelin et al., 2003). The importance of the CBLs in determining membrane-binding properties is exemplified in experiments where the phospholipid preference of Syt1C2A was switched to that of cPLA₂C2 by grafting the CBLs of cPLA₂C2 onto Syt1C2A (Gerber et al., 2001).

Although much has been learned about the structural characteristics, electrostatic properties, and lipid preferences of C2 domains, the mechanisms by which they target specific membranes in living cells remains unclear. Here, we use an imaging approach to correlate structural features of the cPLA₂C2 domain with targeting to Golgi and ER in response to Ca²⁺ mobilization. Using fluorescent protein (FP)-labeled C2 domains having diverse biochemical properties and FP-labeled lipid-specific and organelle-specific probes, we first characterized the spatiotemporal characteristics of cPLA₂C2, Syt1C2A, and PKCαC2 domain translocation in response to intracellular calcium ([Ca²⁺]_i) increase. We then directly compared targeting between these wild-type C2 domains and hybrid Syt1C2A/cPLA₂C2 and PKCαC2/cPLA₂C2 domains containing cPLA₂C2 CBLs.

MATERIALS AND METHODS

Fluorescent Protein Constructs

pEGFP-cPLA₂C2 encoding the human cPLA₂ C2 domain (residues 17-148, accession M72393) fused to enhanced green fluorescent protein (EGFP) was cloned as described previously (Evans et al., 2001). Rat Syt1C2A fused to enhanced yellow fluorescent protein (EYFP) (residues 140-267, pEYFP-Syt1C2A) was constructed by polymerase chain reaction (PCR) amplification of the C2 domain from pGEX65-4 (Nishiki et al., 1994) by using primers Syt1C2Afwd (CGCGGATCCGAAAACTGGGAAAGCTCCAATATTCACTG) and Syt1C2Arev (CCCAAGCTTTTTCTCAGCGCTCTGGAGATCGCGCCACTC), and subcloning the PCR product into the BglII/HindIII site of EYFP(C1). A construct encoding rat Syt1C2A and the three calcium binding loops from rat cPLA₂C2 fused to EYFP (pEYFP-Syt1C2A_cPLA₂C2_L1.2.3) was constructed by ligating the BamHI/HindIII fragment from pGEXSyt1C2A_cPLA₂αC2_L1.2.3 (Gerber et al., 2001) into the BglII/HindIII site of EYFP(C1). Human PKCα (accession X52479) fused to EGFP (pEGFP-PKCα) was constructed by inserting the EcoR1 fragment from pHACE-PKCα, a gift from I. Weinstein (Columbia University, New York, NY), into the EcoR1 site of pEGFP(C2). A construct encoding human PKCα C2 domain (residues 158-286) fused to ECFP (pECFP-PKCαC2) was constructed by PCR amplification of pEGFP-PKCα by using primers PKCαC2fwd (GTCAAGCTTAAGAGGGGGCGGATTTACCTA) and PKCαC2rev (CGTCGACTGGTAGTACTCACCTTCTTCTTG), and cloning the PCR product into the HindIII/SalI site of ECFP(C3). pECFP-PKCαC2/N189F and pECFP-PKCαC2/N189F/T250Y were produced by site-directed mutagenesis (Stratagene, La Jolla, CA). Constructs encoding the human PKCα C2 domain with human cPLA₂C2 calcium binding loops 1 and 3 or 1, 2, and 3 (pECFP-PKCαC2_cPLA₂C2_L1.3 and pECFP-PKCαC2_cPLA₂C2_L1.2.3, respectively) were assembled from overlapping PCR fragments obtained by amplification of pECFP-PKCαC2 with PKCαC2fwd, PKCαC2rev primers and primers containing the cPLA₂C2 loop sequences (PKCαC2_cPLA₂C2_loop1rev = TGGAGTATCAAGCATGTCACCAAAGGCCCCCTTTGTTAGATTTTTTGCATCTCGTACTGTGACAT GGAGC; PKCαC2_cPLA₂C2_loop1fwd = ACAAAGGGGGCCTTTGGTG-ACA TGCTTGATACTCCAGATCCTTATGTGAAGCTGAAACTTATTCCTGAT CCC; PKCαC2_cPLA₂C2_loop2rev = GTTTATGTCATTATTGAAGGTT-TTGG TTTTTTGCTTGC; PKCαC2_cPLA₂C2_loop2fwd = TTCAATAATGA-CA TAAACCCGCAGTGGAATGAGTCC; PKCαC2_cPLA₂C2_loop3rev = TTCATCCATGACATAATTGGCGTCCCAGATTTCTACAGACAGTCG; PK CαC2_cPLA₂C2_loop3fwd = GCCAATTATGTCATGGATGAATTCATGGGATCCCTTTCCTTTGG). The fragments were assembled by PCR and cloned into the HindIII/SmaI site of ECFP(C3). pECFP-GT, which contains the N terminal 82 amino acids of β-1,4-galactosyltransferase fused to enhanced cyan fluorescent protein (ECFP), was purchased from BD Biosciences Clontech (Palo Alto, CA). TGN38-ECFP and TGN38-EYFP were gifts from K. Simons (Max Planck Institute, Dresden, Germany). EGFP-PLCδ₁PH domain was a gift from M. Lemmon (University of Pennsylvania School of Medicine, Philadelphia, PA). ECFP-Rab5 was a gift from A. Sorkin (University of Colorado Health Sciences Center, Denver, CO). The pEYFP plasmid used in this study was constructed from pEYFP purchased from BD Biosciences Clontech by introducing a Q70M mutation by site-directed mutagenesis (Stratagene) to improve its qualities (Griesbeck et al., 2001). Different fluorescent protein constructs were produced by substituting NheI/BsrGI fragments encoding the fluorescent protein-encoding region from pECFP and pEYFP. All constructs were confirmed by sequencing.

Cell Culture

MDCK cells obtained from American Type Culture Collection (Manassas, VA) were cultured in DMEM containing 10% fetal bovine serum, 100 U/ml penicillin, 100 μg/ml streptomycin, 0.292 mg/ml glutamine in 5% CO₂ at 37°C. Subconfluent cells (1 × 10⁴ cells/cm²) were transfected with the relevant plasmid(s) by using FuGENE-6 (Roche Diagnostics, Indianapolis, IN) in DMEM containing 0.2% bovine serum albumin, 100 U/ml penicillin, 100 μg/ml streptomycin, 0.292 mg/ml glutamine following the manufacturer's protocol.

Microscopy of Fluorescent Proteins

Transfected Madin-Darby canine kidney (MDCK) cells grown on MatTek plates were washed with and incubated in Hanks' balanced salt solution additionally buffered with 25 mM HEPES, pH 7.4 (HEPES-Hanks' balanced salt solution). Cells were imaged using an Olympus inverted microscope equipped with a 60×, 1.25 numerical aperture oil immersion objective, cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP) emission filters (Chroma Technology, Brattleboro, VT) in a Sutter filter wheel, a CFP/YFP dichroic mirror, and a TILL Imago charge-coupled device camera (TILL Photonics, Gräfelfing, Germany). Excitation light of 430 and 510 nm for CFP and YFP, 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.

FP fluorescence with respect to time was determined for regions of interest (ROI) corresponding to the plasma membrane or to trans-Golgi network (TGN) by the following method. Background-corrected average pixel values were determined for ROI_TGN at each time point. Due to photobleaching 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 = 0. The background- and bleach-corrected fluorescence value, F_t, was calculated by dividing the background-corrected average pixel value for each ROI at each time point by the bleach correction factor at the corresponding time point. The normalized fluorescence at the plasma membrane or TGN was calculated by dividing F_t by the t = 0 fluorescence value, F₀, to yield F_t/F₀.

Structural Models

The Protein Data Bank (Berman et al., 2000) identifiers for the experimentally determined C2 domain structures used in the calculations are as follows: cPLA₂C2 (1RLW; Perisic et al., 1998), Syt1C2A (1BYN; Shao et al., 1998), and PKCαC2 (1DSY; Verdaguer et al., 1999). It was assumed that the C2 domains bound the number of calcium ions specified by the structural studies. Specifically, the calcium-bound forms of cPLA₂C2, Syt1C2A, and PKCαC2 used in the electrostatic calculations contained two, three, and two calcium ions, respectively. The hybrid C2 domain models illustrated in Figures 4, 9, and 10 were constructed by placing the respective loop regions from cPLA₂C2 onto the structural cores of Syt1C2A and PKCαC2. The structures of both Syt1C2A and PKCαC2 were superimposed onto the cPLA₂C2 structure by using the program Combinatorial Extension (Shindyalov and Bourne, 1998). Once structurally aligned, the appropriate portions of the cPLA₂C2 Protein Data Bank file replaced regions of the Syt1C2A and PKCαC2 Protein Data Bank files as dictated by the sequence alignments depicted in Figures 1, 4, and 9. The hybrid C2 domain models were energy minimized in the program Modeler (Sali and Blundell, 1993) to fix any mismatches between the various structural segments. Because the calcium coordination schemes of the hybrid C2 models are closest to that of cPLA₂C2, it was assumed that the hybrid C2 models bind two calcium ions in the same manner as cPLA₂C2. The calcium ions were docked onto the hybrid models by structurally superimposing the models onto the experimentally determined structure of cPLA₂C2 (1RLW) and transferring the coordinates of the calcium ions from the cPLA₂C2 coordinate file to the coordinate files for the models. A model for the N189F/T250Y mutant of PKCαC2 was obtained from the original Protein Data Bank coordinates for PKCαC2 by building in the appropriate side chains at positions 189 and 250 by using the program CHARMM (Brooks et al., 1983). Hydrogen atoms were added to the heavy atoms in the structures and models with CHARMM. The structures with hydrogens were subjected to conjugate gradient minimization with a harmonic restraint force of 50 kcal/mol/Ų applied to the heavy atoms located at the original crystallographic coordinates.

Figure 4.

PKCαC2 domains containing hydrophobic residues target plasma and nuclear membranes, but not Golgi and ER. Images of a cell coexpressing (A) EYFP-PKCαC2 and (B) ECFP-PKCαC2/N189F/T250Y at 150 s after treatment with 10 μM IONO are shown. Insets in A and B show an enlarged area of the cell, including the nuclear membrane. Images of a cell coexpressing (C) EYFP-cPLA₂C2 and (D) ECFP-PKCαC2/N189F/T250Y at 240 s after treatment with 10 μM IONO are shown. Arrowheads in C point to Golgi and, in D, to the same area as in C. Images are representative of at least five experiments performed on different days.

Figure 9.

Hybrid PKCαC2/cPLA₂C2 domains containing cPLA₂C2 Ca²⁺-binding loops colocalize with cPLA₂C2. (A) Ribbon diagram of the hybrid C2 domain with Ca²⁺-binding loops from cPLA₂C2 (red) on a PKCαC2 backbone (white). (B) The amino acid sequence of the hybrid shows PKCαC2 sequence (black, K158-L183, D193-T214, P221-D246, F255-Y285) interrupted by insertion of cPLA₂C2 loop sequences (red, T31-P42, F63-I67, A94-E100). Images of cells coexpressing (C) ECFP-cPLA₂C2 and (D) EYFP-PKCαC2 at 45 s after treatment with 10 μM IONO are shown. (E) A merged image of C and D. Images of cells coexpressing (F) EYFP-cPLA₂C2 and (G) ECFP-PKCαC2_cPLA₂C2_L1.2.3 at 90 s after treatment with 10 μM IONO are shown. (H) A merged image shows the overlap of F and G. Images of cells coexpressing (I) EYFP-cPLA₂C2 and (J) ECFP-PKCαC2_cPLA₂C2_L1.3 in medium supplemented with 5 mM CaCl₂ at 42 s after treatment 10 μM IONO are shown. (K) A merged image shows the overlap of I and J.

Figure 10.

Electrostatic properties of C2 domains of known structure, mutants, and hybrid models. In each panel, the structure or model is represented as a Cα backbone worm (white), and calcium ions are represented as yellow spheres. The electrostatic potentials were calculated and visualized in GRASP (Nicholls et al., 1991) for 0.1 M KCl. The red and blue meshes represent, respectively, the -25 and +25 mV equipotential profiles. (A) PKCαC2. (B) Syt1C2A. (C) cPLA₂C2. (D) PKCαC2/N189F/T250Y; F189 and Y250 are denoted by green arrows. (E)PKCαC2_cPLA₂C2_L1.2.3.(F)Syt1C2A_cPLA₂ C2_L1.2.3. The structures in A, B, and C were taken from the Protein Data Bank, and the hybrid models in D, E, and F were constructed as described in MATERIALS AND METHODS.

Membrane Models

A structural model for the 2:1 PC/PS bilayer was built as described previously (Ben-Tal et al., 1996). Because the purpose of the calculations in Table 1 is to quantify the change in the electrostatic free energy of interaction of the C2 domains with charged membrane surfaces, it is assumed that the lipids change neither structure nor position upon interaction with the protein. These approximations are reasonable for the C2 domains from Syt1 and PKCα because both are highly charged and do not penetrate the membrane interface to a large degree. A similar treatment of C2 domains was applied in previous computational work (Murray and Honig, 2002).

Table 1.

The calculated electrostatic free energy of interaction (ΔG_el) of Ca²⁺-bound C2 domains with a membrane containing 33 mole percent acidic lipid in 0.1 M KCl. See MATERIALS AND METHODS for details.

C2 domain	(ΔGel), kcal/mol
PKCαC2	-4.5
Syt1C2A	-5.6
cPLA2C2	+0.6
Syt1C2A_cPLA2C2_L1.2.3	0.0
PKCαC2_cPLA2C2_L1.2.3	-0.5

Electrostatic Calculations

In Figure 10, the electrostatic properties of the C2 domains of known structure, the PKCaC2/N189F/T250Y mutant, and Syt1C2A/cPLA₂C2 and PKCαC2/cPLA₂C2 hybrids were calculated for 0.1 M KCl, and visualized in GRASP (Nicholls et al., 1991).

For the calculations depicted in Table 1, the electrostatic potentials and free energies were obtained as described previously (Murray and Honig, 2002). The finite difference Poisson-Boltzmann (FDPB) method has previously been shown to yield excellent agreement with experimental measurements of the binding of peptides and proteins to charged membranes (Ben-Tal et al., 1996, 1997; Murray et al., 1998; Murray and Honig, 2002). Electrostatic free energies are obtained from the calculated potentials (Sharp and Honig, 1990), and the electrostatic free energies of interaction are determined as the difference between the electrostatic free energy of a C2 domain in a specific orientation with respect to the membrane surface, G_el(P·M), and the electrostatic free energies of the C2 domain, G_el(P), and membrane, G_el(M), infinitely far apart, i.e., taken separately:

The membrane-associated orientations of the calcium-bound Syt1C2A and PKCαC2 domains are the orientations of minimum electrostatic free energy. These were determined by comparing the electrostatic free energies of interaction of different orientations sampled about an orientation obtained by visual inspection of the electrostatic potential profiles in GRASP. Previous work has established that, for peripheral association, both the relative binding free energies and the electrostatic contribution to the absolute binding free energies are well described by consideration of minimum free energy orientation alone (Ben-Tal et al., 1996, 1997; Murray et al., 1998). The membrane-associated orientations of cPLA₂C2 and the hybrid models are similar to those of the Syt1C2A and PKCαC2 domains for comparison purposes.

RESULTS

We first performed experiments to establish the spatiotemporal targeting patterns of three, biochemically diverse, C2 domains: cPLA₂C2, PKCαC2, and Syt1C2A.

Intracellular Targeting of cPLA₂C2 Domain

We have previously shown that FP fusions of cPLA₂ (FP-cPLA₂) and the cPLA₂C2 domain (FP-cPLA₂C2) colocalize completely after Ca²⁺-mediated translocation, demonstrating that the C2 domain is entirely responsible for targeting of the enzyme. We have also demonstrated previously that FP-cPLA₂ colocalizes with golgin 97, a cis-Golgi cisternae marker and that FP-cPLA₂C2 colocalizes with ECFP-GT, which is a marker for medial and trans-Golgi cisternae, but not TGN (Sciaky et al., 1997; Evans et al., 2001, 2003). To better define cPLA₂ targeting, we compared colocalization of FP-cPLA₂C2 with FP-TGN38, a TGN marker (Keller et al., 2001), and with FP-GT. MDCK cells coexpressing FP-cPLA₂C2 and FP-GT were stimulated with 10 μM ionomycin (IONO) and imaged by time-lapse microscopy. Images demonstrate colocalization of FP-cPLA₂C2 with FP-GT at Golgi (Figure 2A-C and Fig 2C video) and at ER and nuclear membranes. These images also show dynamic, nuclear invaginations looking like intranuclear tubules to which cPLA₂C2 translocates (Ellenberg et al., 1997; Fricker et al., 1997; Evans et al., 2001). To determine if Golgi cisternae could be differentiated from TGN, MDCK cells coexpressing FP-GT and FP-TGN38 were imaged. The FP-GT and FP-TGN38 signals were juxtaposed, but displayed little overlap (Figure 2D-F and Fig 2F video). Most overlap was observed near the nucleus, where the cell is thickest, suggesting that the fluorescence overlap may be due to superpositioning of Golgi cisternae and TGN instead of colocalization of the markers on the same membranes. We would expect, based on these observations, that cPLA₂C2 targeting to be distinct from TGN. In IONO-treated MDCK cells coexpressing FP-cPLA₂C2 and FP-TGN38, the FP-cPLA₂C2 and FP-TGN38 signals were juxtaposed in a manner similar to TGN38 and GT (Figure 2, G-I), but failed to overlap well except very near the nucleus. These results demonstrate that the C2 domain of cPLA₂ preferentially targets ER and Golgi cisternae, although it is possible that a small fraction of cPLA₂C2 also localizes to the TGN.

Figure 2.

The cPLA₂ C2 domain targets Golgi and ER. Images of a cell coexpressing (A) EYFP-cPLA₂C2 and (B) ECFP-GT at 27 s after treatment with 10 μM IONO are shown. A merged image (C) shows the region of the Golgi from A and B. Images of a cell coexpressing (D) ECFP-GT and (E) TGN38-EYFP, and (F) a merged image of D and E in the region of TGN and Golgi are shown. Images of cells coexpressing (G) ECFP-cPLA₂C2 and (H) TGN38-EYFP at 30 s after treatment with 10 μM IONO are shown. A merged image (I) shows the region of the TGN. Images are representative of a minimum of five experiments.

Intracellular Targeting of PKCαC2 Domain

Many reports have shown that PKCα and its C2 domain translocate to the plasma membrane in response to an [Ca²⁺]_i increase (Mineo et al., 1998; Almholt et al., 1999; Corbalán-García et al., 1999; Maasch et al., 2000; Bolsover et al., 2003). Besides its C2 domain, PKCα also has a C1 domain that targets diacylglycerol in plasma membrane and can promote translocation of PKCα to plasma membrane in the absence of a Ca²⁺ signal (Oancea and Meyer, 1998). Thus, it is possible that the C1 or other domains may modify Ca²⁺- and C2 domain-dependent targeting of PKCα. To determine whether the PKCαC2 domain is solely responsible for Ca²⁺-dependent membrane targeting, as is the case for cPLA₂, FP fusions of full-length PKCα and the isolated C2 domain were coexpressed in MDCK cells, stimulated with IONO, and imaged by time-lapse microscopy. In unstimulated cells, PKCα was primarily restricted to the cytoplasm, whereas the C2 domain was found also in the nucleoplasm (our unpublished data). PKCα (Figure 3A) and PKCαC2 (Figure 3B) rapidly moved from the cytoplasm to areas on the plasma membrane in response to IONO, where they overlapped completely (Figure 3C). This result demonstrates that the PKCαC2 domain is responsible for targeting of the holoenzyme in response to increases in [Ca²⁺]_i. The distribution of PKCα and PKCαC2 on the plasma membrane was not uniform, but exhibited a punctate distribution on the dorsal cell surface (Figure 3D) and mottled appearance at the ventral surface (Figure 3G and Fig 3G video), similar to previous reports (Maasch et al., 2000) and similar to the reported distribution of the PLCδ₁PH domain (Stauffer et al., 1998; Varnai and Balla, 1998). Indeed, in response to IONO, FP-PKCαC2 (Figure 3, D and G) and FL-PLCδ₁PH fluorescence (Figure 3, E and H) overlapped at both the dorsal and ventral cell surfaces (Figure 3, F and I, respectively). However, unlike PKCαC2, PLCδ₁PH loses its punctate and irregular distribution within tens of seconds as the [Ca²⁺]_i increases and becomes homogeneous in the cytosol (Fig 3H video), as has been described previously (Varnai and Balla, 1998). Because PLCδ₁PH specifically recognizes phosphatidylinositol-4,5-bisphosphate (PIP₂) in plasma membranes (Stauffer et al., 1998; Varnai and Balla, 1998; Balla and Varnai, 2002), these results suggest that PKCαC2 specifically targets areas in the plasma membrane enriched in PIP₂, consistent with observations that the PKCαC2 prefers anionic membrane phospholipids (Corbalán-García et al., 1999; Verdaguer et al., 1999; Kohout et al., 2002; Ochoa et al., 2002) and that the β 3 and 4 strands of PKCαC2 are involved in specific binding of PIP₂ (Corbalán-García et al., 2003).

Figure 3.

PKCα C2 domain is responsible for PKCα targeting and targets PIP₂-rich regions in the plasma membrane. Images of a cell coexpressing (A) ECFP-PKCα and (B) EYFP-PKCαC2 at 15 s after treatment with 10 μM IONO are shown. (C) A merged image of A and B. Images of a cell coexpressing (D) EYFP-PKCαC2 and (E) ECFP-PLCδ₁PH at 10 s after treatment with 10 μM IONO are shown. Focusing is on the dorsal surface of the cell. Insets in D and E show an enlarged area of the cell. A merged image (F) shows a smaller region of the cell. Images of a cell coexpressing (G) EYFP-PKCαC2 and (H) ECFP-PLCδ₁PH at 21 s after treatment with 10 μM IONO are shown. Focusing is on the ventral surface of the cell. A merged image (I) shows overlap of images G and H. Images are representative of a minimum of five experiments.

Intracellular Targeting of PKCαC2 Mutants

The tips of CBLs 1 and 3 of cPLA₂C2 contain hydrophobic residues that insert into the hydrophobic core of the membrane bilayer. In particular, F35 on loop 1 and Y96 on loop 3 have been shown to penetrate well into the membrane (Nalefski and Falke, 1998; Perisic et al., 1998, 1999; Bittova et al., 1999; Frazier et al., 2002; Stahelin et al., 2003). To determine whether addition of hydrophobic residues on PKCαC2 would result in ER/Golgi targeting, a FP-PKCαC2/N189F/T250Y mutant was constructed and coexpressed in cells with FP-PKCαC2. PKCαC2 N189 is located at the tip of CBL1, and PKCαC2 T250 is located at a structurally similar site as cPLA₂C2 Y96, both in CBL3. FP-PKCαC2/N189F/T250Y (Figure 4B) targeted plasma membrane, as did FP-PKCαC2 (Figure 4A); however, the mutant also targeted nuclear membrane, unlike the wild-type PKCαC2 domain. As seen in Figure 4C and in previous work, cPLA₂C2 targets Golgi (arrowheads), ER, and nuclear membrane, including the nuclear membrane invaginations. In cells coexpressing FP-PKCαC2/N189F/T250Y and FP-cPLA₂C2 and treated with IONO, the mutant was observed at plasma membrane and colocalized with FP-cPLA₂C2 on the nuclear membrane and its invaginations, but it failed to colocalize at the ER or Golgi (Figure 4, C and D, arrowheads, and Fig 4D video). Similar results were obtained with FP-PKCαC2/N189F (our unpublished data). Comparison of the electrostatic profile models of PKCαC2 and PKCαC2/N189F/T250Y (Figure 10, A and D, respectively) show that they are very similar. Therefore, if electrostatic determinants are important in membrane targeting, then PKCαC2 and PKCαC2/N189F/T250Y would exhibit similar targeting. Together, these results demonstrate that adding hydrophobic residues to PKCαC2 adds the nuclear membrane to its targeting pattern but fails to confer ER/Golgi targeting similar to cPLA₂C2 or change the electrostatic profile.

Intracellular Targeting of Syt1C2A Domain

Although Syt1 is an integral membrane protein and its C2A domain mediates membrane binding but not translocation (Davletov and Sudhof, 1993; Chapman and Jahn, 1994; Geppert et al., 1994; Sudhof and Rizo, 1996; Zhang et al., 1998), analyzing its intracellular targeting pattern can help to reveal the properties of C2 domains that confer intracellular targeting. MDCK cells expressing FP fused to Syt1C2A (FP-Syt1C2A) were treated with IONO and imaged by time-lapse microscopy. In contrast to cPLA₂C2 and PKCαC2, translocation of Syt1C2A required that the medium be supplemented with 5 mM CaCl₂ (final extracellular calcium ([Ca²⁺]_e) = 6.3 mM), which reflects the lower affinity of Syt1C2A for Ca²⁺ relative to cPLA₂C2 and PKCαC2 (Nalefski and Newton, 2001; Kohout et al., 2002). In unstimulated MDCK cells, FP-Syt1C2A is distributed homogeneously in the cytoplasm and nucleoplasm. In response to IONO, FP-Syt1C2A exhibited a biphasic targeting pattern and first targeted the plasma membrane region in a similar pattern to FP-PKCαC2 (Figure 5, A and B, and Fig 5 video). However, unlike FP-PKCαC2, FP-Syt1C2A partially dissociated from plasma membrane and moved to a juxtanuclear region several seconds after IONO treatment (Figure 5C and Fig 5 video). Quantitative analysis of FP-Syt1C2A translocation (Figure 5D) shows the biphasic nature of Syt1C2A targeting. Although the timing varied between experiments, the pattern of FP-Syt1C2A first moving to the plasma membrane and then to the juxtanuclear area was consistently observed. To identify specific membrane domains and organelles targeted by FP-Syt1C2A, experiments similar to those described above were performed. Cells coexpressing FP-Syt1C2A and FP-PLCδ₁PH were treated with IONO and imaged by time-lapse microscopy. The distribution of FP-PLCδ₁PH looks punctate and irregular, as noted previously (Figure 6A). In the earliest times of FP-Syt1C2A translocation (9 s; Figure 6B), there is a partial overlap of the FP-PLCδ₁PH and FP-Syt1C2A signals at the margins of the cell (Figure 6C). These results suggest that, initially, FP-Syt1C2A specifically targets areas of plasma membrane enriched in PIP₂ in a manner similar to FP-PKCαC2. It was not possible to image colocalization of FP-PLCδ₁PH and FP-Syt1C2A at later times due to rapid dissociation of FP-PLCδ₁PH from the plasma membrane. Because the juxtanuclear targeting of FP-Syt1C2A seen later is reminiscent of Golgi cisternae or TGN (Figure 2), probes were used to identify the structure targeted in the later phase of Syt1C2A translocation. The majority of the FP-TGN38 fluorescence in the cell is located in a juxtanuclear position representing TGN; however, there was a small fraction of fluorescence located at small, motile structures, presumably endosomes, and at the plasma membrane (Figure 6D). After IONO treatment, FP-Syt1C2A (Figure 6E) overlapped with the TGN fraction of the FP-TGN38 fluorescence, but not with the endosomal or plasma membrane fractions (Figure 6F). Conversely, when FP-Syt1C2A was coimaged with FP-GT (Figure 6, H and G, respectively), or with FP-Rab5 (Figure 6, K and J, respectively), a marker for early endosomes (Novick and Zerial, 1997), there was only a small region of overlap of the fluorescence (Figure 6, I and L). The spatial relationship between FP-Syt1C2A and FP-Rab5 (Figure 6L) was similar to that between FP-Syt1C2A and the endosomes identified by FP-TGN38 (Figure 6F). Similar experiments using FP-Rab8, a marker for late endosomes, showed no overlap of fluorescence with Syt1C2A (our unpublished data). TGN was further identified as the site of Syt1C2A translocation by coimaging cells expressing FP-cPLA₂C2 and FP-Syt1C2A in response to [Ca²⁺]_i mobilization after incubation for 1.5 h in 100 μM brefeldin A (BFA), which disrupts Golgi cisternae but not TGN (Lippincott-Schwartz et al., 1989; Chege and Pfeffer, 1990). In response to an increase in [Ca²⁺]_i, no clear cPLA₂C2 targeting of Golgi was observed (compare Figure 6M with Figures 3, 8, and 9), demonstrating that BFA had disrupted targeting to Golgi as we have shown previously (Evans et al., 2001). However, BFA had no effect on Syt1C2A translocation (Figure 6N). These results suggest that the TGN is the primary target of the later phase of Syt1C2A translocation.

Figure 5.

The Syt1C2A domain exhibits a biphasic pattern of translocation. Images of a cell expressing EYFP-Syt1C2A in medium supplemented with 5 mM CaCl₂ treated with 10 μM IONO show distribution of Syt1C2A (A) before, (B) 40 s and (C) 90 s after addition of IONO. (D) Quantification of fluorescence change at areas corresponding to the plasma membrane (PM, solid line) and the juxtanuclear (JN, dotted line) regions of the cell in images A-C. Images and graph are representative of six independent experiments.

Figure 6.

Syt1C2A targets TGN and PIP₂-rich regions in the plasma membrane. Images of a cell coexpressing (A) ECFP-PLCδ₁PH and (B) EYFP-Syt1C2A in medium supplemented with 5 mM CaCl₂ at 9 s after treatment with 10 μM IONO are shown. A merged image (C) shows a region of the cell membrane. Images of a cell coexpressing (D) TGN38-ECFP and (E) EYFP-Syt1C2A in medium supplemented with 5 mM CaCl₂ at 56 s after treatment with 10 μM IONO are shown. A merged image (F) shows the region of the TGN. Images of a cell coexpressing (G) ECFP-GT and (H) EYFP-Syt1C2A in medium supplemented with 5 mM CaCl₂ at 20 s after treatment with 10 μM IONO are shown. A merged image (I) shows the region of TGN and Golgi. Images of a cell coexpressing (J) ECFP-Rab5 and (K) EYFP-Syt1C2A in medium supplemented with 5 mM CaCl₂ at 20 s after treatment with 10 μM IONO are shown. A merged image (L) shows the region of early endosomes and TGN. Cells coexpressing (M) ECFP-cPLA₂C2 and (N) EYFP-Syt1C2A were treated for 1.5 h with 100 μM BFA at 37°C before stimulation with 10 μM IONO in medium supplemented with 5 mM CaCl_2. Images are 20 s after stimulation. Images are representative of a minimum of five experiments.

Figure 8.

A hybrid Syt1C2A/cPLA₂C2 domain containing cPLA₂C2 Ca²⁺-binding loops colocalizes with cPLA₂C2. (A) Ribbon diagram of the hybrid C2 domain with Ca²⁺-binding loops from cPLA₂C2 (red) on a Syt1C2A backbone (white). (B) The amino acid sequence of the hybrid shows Syt1C2A sequence (black, E140-A165, D178-T195, N203-T223, G241-K267) interrupted by insertion of cPLA₂C2 loop sequences (red, T28-P42, R61-I67, L87-I102). Images of cells coexpressing ECFP-cPLA₂C2 (C) and EYFP-Syt1C2A (D) in medium supplemented with 5 mM CaCl₂ at 55 s after treatment with 10 μM IONO are shown. (E) A merged image of C and D. Images of cells coexpressing (F) EYFP-cPLA₂C2 and (G) ECFP-Syt1C2A_cPLA₂C2_L1.2.3 in medium supplemented with 5 mM CaCl₂ at 70 s after treatment with 10 μM IONO are shown. (H) A merged image of F and G.

The biochemical and electrostatic characteristics of Syt1C2A and PKCαC2 membrane binding are very similar (Nalefski et al., 2001; Kohout et al., 2002; Murray and Honig, 2002) and it was unexpected that the two C2 domains would exhibit different targeting profiles. Because Syt1C2A required increased [Ca²⁺]_e for translocation, we examined whether PKCαC2 also targeted TGN and displayed a biphasic targeting pattern in response to an increased [Ca²⁺]_e. Cells coexpressing FP-PKCαC2 and FP-Syt1C2A were stimulated with IONO in medium supplemented with 5 mM CaCl₂ and imaged by time-lapse microscopy. No difference was observed between the targeting of PKCαC2 in medium containing ∼1.3 mM (Figure 3B) or ∼6.3 mM (Figure 7A) CaCl₂. Thus, the difference between the targeting patterns of Syt1C2A and PKCαC2 (Figure 7, A-C) is not an artifact of increased [Ca²⁺]_e.

Figure 7.

Syt1C2A and PKCαC2 target different subcellular organelles in response to a [Ca²⁺]_i increase. Cells coexpressing (A) ECFP-PKCαC2 and (B) EYFP-Syt1C2A in medium supplemented with 5 mM CaCl₂ at 55 s after treatment with 10 μM IONO are shown. (C) A merged image of A and B. Images are representative of a minimum of five experiments.

Intracellular Targeting of a Syt1C2A/cPLA₂C2 Hybrid C2 Domain

Having established the targeting patterns of the individual C2 domains, we proceeded to investigate the importance of the cPLA₂C2 CBLs in determining intracellular targeting of cPLA₂C2. A FP fusion of a hybrid C2 domain consisting of cPLA₂C2 CBLs substituted into Syt1C2A was used (Figure 8, A and B) (Gerber et al., 2001). In liposome binding assays, the hybrid Syt1C2A/cPLA₂C2 domain displayed lipid specificity similar to cPLA₂C2 (Gerber et al., 2001). Because cPLA₂C2 and Syt1C2A display Golgi/ER and plasma membrane/TGN targeting, respectively, we expected to observe little similarity between the Ca²⁺-dependent targeting patterns of cPLA₂C2 and Syt1C2A in the same cell. In response to IONO in medium supplemented with 5 mM CaCl₂, cPLA₂C2 displayed the usual targeting of ER, Golgi, and nuclear membrane (Figure 8C). In contrast, Syt1C2A was observed primarily at the TGN (Figure 8D). Consistent with our earlier observations, FP-cPLA₂C2 and FP-Syt1C2A fluorescence signals are primarily juxtaposed and only partial overlap in the area of the Golgi (Figure 8E). Of particular interest was the observation that Syt1C2A failed to target nuclear membrane invaginations or ER; in fact, FP-Syt1C2A fluorescence was excluded from the area of the ER. In contrast, a FP-tagged, hybrid C2 domain containing CBLs1, 2, and 3 from cPLA₂ on a Syt1C2A backbone (Syt1C2A_cPLA₂C2_L1.2.3; Figure 8G and Fig 8G video) displayed an intracellular targeting pattern that was indistinguishable from cPLA₂C2, although translocation of the hybrid required a higher (6.3 mM) [Ca²⁺]_e. This Syt1C2A/cPLA₂C2 hybrid targeted Golgi, ER, and the nuclear membrane invaginations, but not plasma membrane or TGN. This result demonstrates that the CBL regions of cPLA₂C2 confer targeting to Golgi and ER and that substituting cPLA₂C2 CBLs results in a complete switch in the targeting pattern. However, this Syt1C2A/cPLA₂C2 hybrid domain also included portions of cPLA₂C2 β strands that are below the Ca²⁺-binding regions (Figure 8, A and B), which may have influenced targeting.

Intracellular Targeting of PKCαC2/cPLA₂C2 Hybrid C2 Domains

To investigate to role of specific cPLA₂C2 CBLs in more detail, FP-tagged PKCαC2/cPLA₂C2 hybrids were constructed that contained cPLA₂C2 CBLs 1, 2, and 3, or 1 and 3 on a PKCαC2 backbone. The CBLs in the PKCαC2/cPLA₂C2 hybrids contained fewer residues than the CBLs in the Syt1C2A/cPLA₂C2 hybrid (Figure 9, A and B). To verify the different targeting patterns of cPLA₂C2 and PKCαC2, cells coexpressing FP-cPLA₂C2 and FP-PKCαC2 were treated with IONO and imaged. FP-cPLA₂C2 (Figure 9A) and FP-PKCαC2 (Figure 9B) translocated to very different areas of the cell (Figure 9E), as observed above. In contrast to FP-PKCαC2 targeting, a PKCαC2/cPLA₂C2 hybrid construct containing all three cPLA₂C2 CBLs (PKCαC2_cPLA₂C2_L1.2.3; Figure 9G and Fig 9G video) translocated to the Golgi, ER, and nuclear membrane, in a pattern similar to cPLA₂C2 (Figure 9, F and H). To narrow down further the structures required for Golgi/ER targeting, similar experiments were performed with a PKCαC2/cPLA₂C2 hybrid construct containing cPLA₂C2 CBLs 1 and 3 (PKCαC2_cPLA₂C2_L1.3; Figure 9J and Fig 9J video). This construct was observed to translocate to Golgi in a pattern similar to cPLA₂C2 (Figure 9F). However, PKCαC2_cPLA₂C2_L1.3 required a higher [Ca²⁺]_e than PKCαC2 (6.3 mM) and was not observed at the ER or nuclear envelope (Figure 9K). Targeting was observed at ER and nuclear membrane, but only when cells were treated with extremely high [Ca²⁺]_e (>25 mM), which also caused vesiculation of the ER and blebbing at the plasma membrane (our unpublished data). We have shown previously that cPLA₂C2 translocates to Golgi at lower [Ca²⁺]_i than it does to ER and nuclear membrane (Evans et al., 2001), thus the observation of preferential Golgi targeting of the PKCαC2_cPLA₂C2_L1.3 construct was not unexpected. Together, these results suggest that cPLA₂ CBLs 1 and 3 confer Golgi/ER targeting and that inclusion of CBL2 improves Ca²⁺ binding properties of the hybrid.

Electrostatic Ca²⁺ Binding of Syt1C2A/cPLA₂C2 and PKCαC2/cPLA₂C2 Hybrids

A recent computational study of the association of C2 domains with phospholipid membranes by using the FDPB method provides insight into the physical basis of C2 domain/membrane interactions. For example, the electrostatic free energy of interaction of Ca²⁺-bound Syt1C2A with a membrane containing 33 mol% acidic lipid (2:1 PC/PS) was predicted to be -5.6 kcal/mol and highly dependent on ionic strength of the solution (Table 1, Figure 10B), which is in good agreement with experimental measurements (Gerber et al., 2001; Zhang et al., 1998; Nalefski et al., 2001) and suggest that electrostatic interactions are sufficient to mediate significant membrane binding. However, Ca²⁺-bound cPLA₂C2, in a similar orientation, is predicated to have a positive electrostatic free energy of interaction (+0.6 kcal/mol) with 2:1 PC/PS due to electrostatic repulsions (Table 1, + 0.6 kcal, Figure 10C). The favorable electrostatic free energy of interaction for Ca²⁺-bound Syt1C2A contrasts greatly with the unfavorable electrostatic free energy for Ca²⁺-bound cPLA₂C2 at the membrane surface and highlights the mechanistic differences between Syt1C2A (primarily favorable electrostatic) and cPLA₂C2 (primarily hydrophobic) interactions with membrane. To determine the electrostatic contribution to the membrane association of other C2 domains considered in this study, FDPB calculations were performed with molecular models for PKCαC2 and the hybrid C2 domains, and a 2:1 PC/PS membrane in 0.1 M KCl (Table 1, Figures 10A, 10E, and 10F). The electrostatic free energy of membrane interaction for Ca²⁺-bound PKCαC2 was -4.5 kcal/mol, which is similar to that of Ca²⁺-bound Syt1C2A (-5.6 kcal/mol) (Murray and Honig, 2002). In contrast, the electrostatic free energies of membrane interaction for the PKCαC2/cPLA₂C2 and Syt1C2A/cPLA₂C2 hybrids containing all three CBLs were -0.5 and 0.0 kcal/mol, respectively, and reminiscent of the repulsive electrostatic interactions experienced by cPLA₂C2. Thus, unlike Syt1C2A or PKCαC2, but similar to cPLA₂C2, the contribution of electrostatic attraction to membrane binding for the hybrids is negligible and membrane binding is most probably due to hydrophobic interactions (Table 1).

DISCUSSION

By comparing the spatiotemporal patterns of translocation between wild-type and mutant or hybrid C2 domains, we were able to correlate structural features of the cPLA₂C2 domain with Ca²⁺-dependent targeting to Golgi and ER. Interestingly, topological differences between the cPLA₂C2 and Syt1C2A or PKCαC2 domains were found to be relatively unimportant, because we were able to freely exchange homologous structures between types I (Syt1C2A and PKCαC2) and II (cPLA₂C2) C2 domains. The coimaging approach used here allowed us to directly compare translocation of any two domains under identical conditions in a physiological setting, thus providing “built-in” controls for cell-to-cell heterogeneity in organelle morphology and in the amplitude and kinetics of the [Ca²⁺]_i increases. The dual imaging approach also allowed us to readily visualize subtle or gross changes in targeting of mutant and hybrid domains.

Targeting of C2 Domains to the Plasma Membrane

The inner leaflet of the plasma membrane serves as a platform on which a multitude of signaling complexes form in response to different cellular signals, including increases in [Ca²⁺]_i. We demonstrate here that the C2 domain specifically targets PKCα to regions of the plasma membrane enriched in PIP₂, which is consistent with recent reports that lysine residues in the β3 and 4 strands of PKCαC2 specifically bind PIP₂ in vitro (Corbalán-García et al., 2003). A primary role for the β 3-4 strand region in targeting, however, is questionable, because although PKCαC2/cPLA₂C2 hybrid domains have intact PKCαC2 β 3-4 strands, they fail to target PIP₂-rich areas of the plasma membrane. Because many reports have shown that specific binding of PKCαC2 to PS is important in membrane interaction (Verdaguer et al., 1999; Conesa-Zamora et al., 2001; Ochoa et al., 2002), further work will be required to differentiate between the roles of nonspecific electrostatics and specific interactions between PKCαC2 and inner leaflet PS or PIP₂ in membrane targeting.

In contrast to the PKCαC2 domain, the Syt1C2A domain serves to facilitate the fusion of vesicles with plasma membrane in response to [Ca²⁺]_i signals. The observation here that the isolated Syt1C2A localizes to regions rich in PIP₂ in the plasma membrane is consistent with observations that Syt1C2A binds to liposomes containing PIP₂ and nonspecifically binds membranes containing anionic phospholipids (Zhang et al., 1998; Murray and Honig, 2002). On the other hand, our finding that Syt1C2A exhibits a biphasic spatiotemporal targeting pattern and that the secondary target is the TGN, but not closely related membrane domains, suggests membrane-specific determinants, either lipid or protein, aid in Syt1C2A-membrane binding. It is possible that targeting of TGN is due to protein-protein interactions (Chapman et al., 1995; Shao et al., 1997; Sugita and Sudhof, 2000). For example, Syt1C2A binds to the N-terminal region of syntaxin 1a in a Ca²⁺-dependent manner (Chapman et al., 1995; Shao et al., 1997) and syntaxin 6, which exhibits homology to the N-terminal region of syntaxin 1a (Misura et al., 2002), is found at the TGN (Bock et al., 1997). Alternatively, Syt1C2A targeting to TGN may involve a more complex interplay of specific phospholipid interaction, electrostatics, and hydrophobic forces. For example, phosphatidylinositol-4-phosphate has been identified as a constituent of the TGN in MDCK cells by using the phosphatidylinositol-4-phosphate adaptor protein 1 PH domain as a probe (Evans, unpublished observations; Dowler et al., 2000; Balla and Varnai, 2002), and, although Syt1C2A association with membrane is primarily electrostatic in nature, several reports have demonstrated that CBL 3 of Syt1C2A can penetrate into the lipid core of membranes, although not as deeply as cPLA₂ (Chae et al., 1998; Chapman and Davis, 1998; Frazier et al., 2003). Perhaps both electrostatic and hydrophobic forces regulate Syt1C2A interaction with membrane. Given the biochemical and electrostatic similarities between PKCαC2 and Syt1C2A, it was surprising to observe the difference in targeting patterns, which may represent a fundamental difference between the modes of membrane interaction between these domains.

Targeting C2 Domains to Golgi and ER

Of those C2 domains whose intracellular targeting in response to [Ca²⁺]_i signals has been studied, the cPLA₂C2 domain is unique in that it targets intracellular membranes exclusively. This is not entirely unexpected, because targeting ER and nuclear membranes may promote coupling of cPLA₂ to cyclooxygenases and 5-lipoxygenase, enzymes that metabolize AA produced by cPLA₂ that are located at ER and nuclear membranes, respectively (Funk, 2001).

Targeting of cPLA₂ to Golgi, ER, and nuclear membranes is very specific. We have reported previously that cPLA₂ translocation to Golgi is sensitive to BFA treatment, translocation is seen at Golgi “ministacks” in response to nocodazole treatment and that cPLA₂ colocalizes with golgin 97, a cis-Golgi marker and with GT, a medial- and trans-Golgi marker (Evans et al., 2001, 2003). We have extended these observations to report here that cPLA₂C2 fails to colocalize substantially with TGN (Figure 2, D-I). The juxtaposed localization of TGN38 to cPLA₂C2 was very similar to the localization of TGN38 to GalNAc transferase II, a Golgi cisternae marker, in PtK₂ cells (Keller et al., 2001). If cPLA₂ targeting requires specific protein or lipid determinants, it is possible that those determinants are not present in TGN membranes.

The observation that the Syt1C2A/cPLA₂C2 hybrid faithfully recapitulates cPLA₂C2 targeting points to the important role of the CBLs in Golgi/ER targeting. The higher [Ca²⁺]_e required for hybrid translocation, compared with cPLA₂C2, is consistent with the significantly lower Ca²⁺ affinity of the Syt1C2A/cPLA₂C2 hybrid compared with cPLA₂C2 in binding to liposomes, probably due to incorrect geometry of the Ca²⁺-coordinating residues (Gerber et al., 2001). The PKCαC2 domain was chosen for making the hybrids because, like cPLA₂C2, it binds two Ca²⁺ ions and has significant homology to cPLA₂C2 in the core region. These hybrids contained swapped loops that were two to nine residues smaller than similar swaps in the Syt1C2A/cPLA₂C2 hybrid (Figures 7B and 8B). A PKCαC2/cPLA₂C2 hybrid with only a swapped CBL1 remained cytoplasmic even when the [Ca²⁺]_e was raised to very high levels (our unpublished data), possibly due to an inability to bind Ca²⁺, as was the case for Syt1C2A/cPLA₂C2 hybrids containing only CBL3 or CBLs1 and 2 (Gerber et al., 2001). However, addition of cPLA₂C2 CBLs 1 and 3 to PKCαC2 resulted in Golgi targeting, but also required a higher [Ca²⁺]_e. This lower affinity for Ca²⁺ of the CBL 1.3 hybrid is probably due to the lack of the Ca²⁺ coordinating residue, N65, located in CBL2 of cPLA₂C2 (Perisic et al., 1998; Xu et al., 1998). The CBL 1.3 hybrid could be driven to ER at high [Ca²⁺]_e, consistent with our results that demonstrated that cPLA₂C2 preferentially translocates to Golgi at lower [Ca²⁺]_i (Evans et al., 2001). The CBL 1.2.3 hybrid translocated to Golgi and ER without increasing the [Ca²⁺]_e, again suggesting that CBL 2 is important for [Ca²⁺]_i binding, but is not involved in targeting Golgi and ER membranes. These results suggest that the CBLs are sufficient for Golgi/ER targeting. A reciprocal hybrid C2 domain, where the PKCαC2 CBLs are fused to a cPLA₂C2 backbone, fails to target Golgi/ER, demonstrating that the cPLA₂C2 CBLs are both necessary and sufficient and that there are no other Golgi/ER targeting determinants on the rest of the cPLA₂C2 domain (Evans, unpublished observations). Similarly, we showed that simply substituting hydrophobic residues at the tips of PKCαC2 CBLs 1 and 3 fails to change the targeting pattern of PKCα to one resembling cPLA₂C2.

The electrostatic equipotential profiles qualitatively demonstrate that the binding of Ca²⁺ to PKCαC2 and Syt1C2A relies on electrostatic forces, and the FDPB calculations show that the electrostatic free energy of interaction favors electrostatic interaction of PKCαC2 and Syt1C2A with membranes containing anionic lipids (Murray and Honig, 2002). This contrasts greatly with the FDPB results obtained for cPLA₂C2 and the Syt1C2A/cPLA₂C2 and PKCαC2/cPLA₂C2 hybrids, which show that electrostatic free energy of interaction between these domains and membrane was unfavorable in all conditions simulated (Murray and Honig, 2002). Thus, binding of cPLA₂C2 and the hybrid C2 domains to membrane most likely relies on hydrophobic interactions facilitated by the binding of Ca²⁺, which reduces the free energy of interaction as a function of membrane penetration by neutralizing the negatively charged Ca²⁺-binding residues (Figure 10C), thus allowing membrane penetration.

The results of this study contribute to the understanding of the structural elements and electrostatic changes required for targeting to Golgi and ER membranes. It has been suggested that translocation of cPLA₂C2 to interior membranes is due to the distribution of PC (Stahelin et al., 2003). However, a study on lipid content in different membrane compartments in CHO cells has shown that PC is equally abundant, as a percentage of total phospholipids, in Golgi and plasma membrane (Gkantiragas et al., 2001). We find that targeting of cPLA₂C2 in Chinese hamster ovary cells is identical to that observed in MDCK cells (Evans, unpublished observations). These observations suggest that other factors in these membranes may play a role in cPLA₂C2 targeting, which is supported by the observations that the [Ca²⁺]_i required for cPLA₂C2 membrane association in live cells is an order of magnitude lower than what has been reported for binding to PC vesicles in vitro. Previous reports have suggested that PIP₂ plays a role in activation of cPLA₂ (Leslie and Channon, 1990; Mosior et al., 1998), and a recent report suggests that residues in the cPLA₂ catalytic domain bind PIP₂ (Six and Dennis, 2003). In our results here, however, we find no evidence for the C2 domain targeting areas rich PIP₂ as determined by FP-tagged PLCδ₁PH domain localization. However, a high-resolution study with glutathione S-transferase-tagged PLCδ₁PH domain has demonstrated only a small fraction of PIP₂ at the Golgi stacks and ER (Watt et al., 2002), suggesting that small amounts of PIP₂ may have a role in cPLA₂ activation, but not targeting. Ceramide, a constituent of the Golgi, has been shown to bind the cPLA₂ C2 domain and increase the fraction of cPLA₂ bound to membrane in response to an increase in Ca²⁺, possibly by modifying the phospholipid organization in membranes (Sato et al., 1999; Kitatani et al., 2000; Huwiler et al., 2001). Several proteins have been shown to interact with cPLA₂C2 in vitro, including annexins (Buckland and Wilton, 1998; Solito et al., 1998), vimentin (Nakatani et al., 2000), and PLIP (Sheridan et al., 2001), but it is not clear as to their role in targeting. Further investigation will aim to identify the lipid or protein constituents of Golgi that aid in cPLA₂ targeting.

Besides its role in generation of AA, an intriguing role for cPLA₂ may be in membrane remodeling that is important for membrane trafficking. Many recent studies have implicated PLA₂ activity in constitutive membrane trafficking (Choukroun et al., 2000; Brown et al., 2003). Most reports, however, implicate calcium-independent PLA₂ isoforms and because cPLA₂ is only on membranes after an increase in [Ca²⁺]_i, it would not be a good candidate for regulating constitutive trafficking. On the other hand, cPLA₂ may play a role in trafficking in response to [Ca²⁺]_i signals, although there have been no studies investigating this possibility.