The pleckstrin homology domain of phosphoinositide-specific phospholipase Cdelta4 is not a critical determinant of the membrane localization of the enzyme.

The inositol lipid and phosphate binding properties and the cellular localization of phospholipase Cdelta(4) (PLCdelta(4)) and its isolated pleckstrin homology (PH) domain were analyzed in comparison with the similar features of the PLCdelta(1) protein. The isolated PH domains of both proteins showed plasma membrane localization when expressed in the form of a green fluorescent protein fusion construct in various cells, although a significantly lower proportion of the PLCdelta(4) PH domain was membrane-bound than in the case of PLCdelta(1)PH-GFP. Both PH domains selectively recognized phosphatidylinositol 4,5-bisphosphate (PI(4,5)P(2)), but a lower binding of PLCdelta(4)PH to lipid vesicles containing PI(4,5)P(2) was observed. Also, higher concentrations of inositol 1,4,5-trisphosphate (Ins(1,4,5)P(3)) were required to displace the PLCdelta(4)PH from the lipid vesicles, and a lower Ins(1,4,5)P(3) affinity of PLCdelta(4)PH was found in direct Ins(1,4,5)P(3) binding assays. In sharp contrast to the localization of its PH domain, the full-length PLCdelta(4) protein localized primarily to intracellular membranes mostly to the endoplasmic reticulum (ER). This ER localization was in striking contrast to the well documented PH domain-dependent plasma membrane localization of PLCdelta(1). A truncated PLCdelta(4) protein lacking the entire PH domain still showed the same ER localization as the full-length protein, indicating that the PH domain is not a critical determinant of the localization of this protein. Most important, the full-length PLCdelta(4) enzyme still showed binding to PI(4,5)P(2)-containing micelles, but Ins(1,4,5)P(3) was significantly less potent in displacing the enzyme from the lipid than with the PLCdelta(1) protein. These data suggest that although structurally related, PLCdelta(1) and PLCdelta(4) are probably differentially regulated in distinct cellular compartments by PI(4,5)P(2) and that the PH domain of PLCdelta(4) does not act as a localization signal.

regulator of several cellular processes. PI(4,5)P 2 is a precursor of important second messengers, such as the water-soluble InsP 3 , which regulates Ca 2ϩ release from intracellular Ca 2ϩ stores, and the hydrophobic diacylglycerol, a potent activator of protein kinase C (1,2). PI(4,5)P 2 has been shown to regulate proteins, by interacting with their lipid recognition domains, and to participate in membrane remodeling events, including the fusion of secretory vesicles with the plasma membrane (3), and at several steps along the endocytic pathway (4 -6).
The best known regulators of PI(4,5)P 2 levels are the phospholipase C (PLC) enzymes that hydrolyze the phosphodiester group between the diacylglycerol backbone and the phosphate group linking the inositol ring within the PI(4,5)P 2 molecule. Several isoforms of PLC have been described that can be classified into five major groups, PLC␤, PLC␥, PLC␦, PLC⑀, and PLC (7,8). PLC␤ isoforms are regulated by the ␣or ␤␥subunits of heterotrimeric G proteins, whereas the two PLC␥ isoforms are activated by receptor tyrosine kinases as well as by the lipid products of PI 3-kinases. PLC⑀ is a recently identified enzyme that is associated with the small GTP-binding protein, Ras (9,10), and PLC is a novel sperm-specific PLC isoform that is responsible for the initiation of Ca 2ϩ oscillations following fertilization (11). PLC␦ is the isoform that evolutionarily is the most conserved, its homologue already appearing in yeast, yet the regulation of this enzyme is the least understood. As with all other PLCs, PLC␦ is activated by Ca 2ϩ ions, and the cytosolic Ca 2ϩ increase is believed to be the primary means by which this enzyme is regulated. One important and distinctive feature of PLC␦ 1 has been its high affinity binding to InsP 3 , and the ability of InsP 3 to inhibit the catalytic activity of the enzyme (12,13). The part of the molecule responsible for InsP 3 binding is the pleckstrin homology (PH) domain (12), a conserved motif first described in pleckstrin (14), and one which is also present in all ␤, ␥, and ␦ PLC isoforms (8), as well as in a number of other regulatory molecules (15). The isolated PH domain of PLC␦ 1 has been shown to specifically bind PI(4,5)P 2 both in vitro (16) and in vivo (17,18), and its crystal structure was solved with a bound Ins(1,4,5)P 3 molecule (19). Although the PH domains found in the other PLCs are also capable of binding certain isomers of inositol lipids, they do not display a * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ Supported in part by the Hungarian National Science Foundation Grant OTKA, T-034606.
‡ ‡ To whom correspondence should be addressed: National Institutes of Health, Bldg. 49, Rm. 6A35, 49 Convent Dr., Bethesda, MD 20892-4510. Tel.: 301-496-2136; Fax: 301-480-8010; E-mail: tambal@box-t.nih.gov. 1 The abbreviations used are: PI(4,5)P 2  similarly high affinity to InsP 3 . Most intriguing, a PLC␦ 1 homologue with similar InsP 3 binding properties but without PLC enzymatic activity, termed p130, has been isolated and characterized (20). Additional genes encoding isoforms of PLC␦ have been described recently (21)(22)(23)(24). Among these, PLC␦ 4 was found to be critical in sperm to induce the acrosome reaction in the zona pellucida (25). PLC␦ 4 is also unique in that it has three splice variants, one of which has been reported recently to be a potent negative regulator of the PLC␦ enzymes (24). This latter study also showed that the PH domain of PLC␦ 4 binds InsP 3 very poorly but still binds the phospholipid, PI(4,5)P 2 , a feature crucial to the inhibitory effect of the unique splice variant. The finding of a PH domain with the ability to bind PI(4,5)P 2 and not being influenced by Ins(1,4,5)P 3 would greatly aid studies in which such PH domains are fused to GFP in order to report PI(4,5)P 2 distribution and dynamics in living cells without being ''contaminated'' by the effects of Ins(1,4,5)P 3 binding (18, 26). The unique features reported for the PLC␦ 4 PH domain prompted us to compare the inositol phosphate and inositol lipid binding properties of the PH domains of PLC␦ 1 and -␦ 4 in vitro and to study the localization and dynamics of these domains when expressed in cells as GFP fusion proteins in vivo.
Our data suggest that the PH domain of PLC␦ 4 has a lower affinity to both InsP 3 and PI(4,5)P 2 than the similar domain of PLC␦ 1 and, therefore, shows less prominent plasma membrane localization in intact cells than the PLC␦ 1 PH domain. Nevertheless, the PLC␦ 4 PH domain is still capable of reporting PLC activation. Most surprising, the localization of the full-length PLC␦ 4 protein is significantly different from that of its PH domain, and both the full-length protein and its truncated form missing the PH domain show primarily ER localization. These data suggest that in the case of the PLC␦ 4 protein, the PH domain is not a major localization signal but still could confer PI(4,5)P 2 regulation to the protein.
DNA Constructs-The PH domain of human PLC-␦ 1 (NM_006225) (residues 1-170) and its mutant R40L has been described previously (18). The PH domain of the PLC␦ 4 (NM_080688) (residues 1-163) was amplified from the cDNA clone described previously (23), using primers with KpnI/SmaI restriction sites. After digestion with the restriction enzymes, the PCR product was ligated into the pEGFP-N1 plasmid (Clontech) cut with the same two restriction enzymes. A shorter variant of this domain, containing residues 1-127, was also constructed, which showed very similar results in localization studies. The full-length PLC␦ 4 as well as the truncated version (⌬1-128) were subcloned into a pCDNA3.1 plasmid with an HA epitope placed at the C termini of the proteins. All constructs were sequenced with dideoxy sequencing. The integrity and expression levels of the fusion proteins were assessed by Western blot analysis from cells lysates prepared from COS-7 or NIH 3T3 cells transfected with the constructs, using a polyclonal antibody against GFP (Clontech). The same HA-tagged constructs were subcloned into the pET23b plasmid (Novagen) for expression in Escherichia coli.
Transfection of Cells for Confocal Microscopy-Cells were plated onto polylysine-coated 30-mm diameter circular glass coverslips at a density of 5 ϫ 10 4 cells/35-mm dish and cultured for 3 days before transfection with plasmid DNAs (1 g/ml) using the LipofectAMINE reagent (10 g/ml, Invitrogen) and Opti-MEM (Invitrogen) or calcium phosphate (for N1E-115 neuroblastoma cells). Thirty six hours after transfection, cells were washed twice with a modified Krebs-Ringer buffer (120 mM NaCl, 4.7 mM KCl, 1.2 mM CaCl 2 , 0.7 mM MgSO 4 , 10 mM glucose, Na-Hepes 10 mM, pH 7.4), and the coverslip was placed into a chamber that was mounted on a heated stage with the medium temperature kept at 33°C unless stated otherwise (in the case of N1E-115 neuroblastoma cells). Cells were incubated in 1 ml of the Krebs-Ringer buffer, and stimuli were added in 0.5 ml of prewarmed buffer after removing 0.5 ml of medium from the cells. Cells were examined in an inverted microscope under a 40ϫ oil-immersion objective (Nikon) and a Bio-Rad laser confocal microscope system (MRC-1024) with the Lasersharp acquisition software (Bio-Rad). N1E-115 neuroblastoma cells were examined with an inverted Leica TCS-SPII confocal system (27).
Immunocytochemistry-Cells were plated, cultured, and transfected on glass coverslips as described above. Cells were fixed with fresh 2% paraformaldehyde in PBS, pH 7.4, for 10 min at room temperature. After three washes with PBS (5 min each), fixed cells were incubated in blocking solution (10% FBS in PBS) for 10 min to decrease nonspecific binding of the antibodies. This blocking solution was complemented with 0.2% saponin for diluting the primary antibody (anti-HA 1:500), and cells were incubated for 1 h at room temperature. After three washes, cells were incubated in the same buffer with a fluorescent secondary antibody (1:1000) for 1 h at room temperature. After a final washing step (three times for 5 min with blocking solution), the cells were rinsed with PBS, air-dried, and mounted on glass slides using Aqua PolyMount mounting medium (Polysciences Inc.).
Fluorescence Recovery after Photobleaching (FRAP)-N1E-115 neuroblastoma cells grown on coverslips were mounted on an inverted Leica TCS-SPII confocal system and maintained in Hepes/bicarbonatebuffered saline at 37°C. The beam of an external 25-milliwatt ArKr laser was coupled into the backfocal plane of the objective via an additional 30/70 beamsplitter using a home-built adapter that allowed simultaneous imaging and spot bleaching. Spots of ϳ1.3 m (full-width, halfmaximum) were bleached in the basal membrane using a single 30-ms pulse, and data were continuously collected with the confocal microscope in-line scan mode at 400 Hz. Data were corrected for slight (Ͻ10%) background bleaching using the intensity of a reference area well away from the bleach spot and fitting with a single exponential (27).
Recombinant Proteins and InsPx Binding Assays-For bacterial expression of the GFP-fused protein domains, the coding sequences were amplified from the GFP plasmids (see above) and were subcloned into the pET-23b bacterial expression vector (Novagen) using the NdeI/ EcoRI restriction sites. The PLC␦ 4 PH was also subcloned into the pGEX6P3 expression plasmid (Amersham Biosciences) for its expression as a GST fusion protein. The resulting plasmids were used to transform the BL-21-DE3 strain of E. coli (Novagen). Bacterial cells were grown to A 600 as follows: 0.6 at 37°C and induced with 300 M isopropyl-1-thio-␤-D-galactopyranoside at 18 -20°C for 7 h. Bacterial lysates were prepared by sonication in lysis buffer (20 mM NaCl and 20 mM Tris, pH 8.0) followed by centrifugation at 10,000 ϫ g for 30 min at 4°C. The supernatant was incubated with Ni 2ϩ -nitrilotriacetic acidagarose beads (Qiagen) in the presence of 5 mM imidazole for 1 h at 4°C. The beads were washed three times with lysis buffer, and the recombinant proteins were eluted with the same buffer containing 1 M imidazole. GST fusion proteins were isolated from bacterial lysates on glutathione-Sepharose columns (Amersham Biosciences) following standard procedures. Protein samples were concentrated and stored in PBS containing 5 mM dithiothreitol and 20% glycerol at Ϫ20°C.
For bacterial expression of PLC␦ 4 and its truncated form lacking the PH domain, the C-terminally HA-tagged proteins were expressed from the pET23b plasmid and purified using an anti-HA antibody and protein A/G plus-Sepharose (Calbiochem). Induction of the expression was achieved by overnight incubation in the presence of 100 M isopropyl-1-thio-␤-D-galactopyranoside at 12°C, because induction at higher temperatures resulted in the production of mostly insoluble full-length protein. The concentration of recombinant proteins were assessed by quantifying the bands of Coomassie Blue-stained SDS gel containing the recombinant proteins and bovine serum albumin as a standard.
The incubation buffer of the in vitro InsP 3 binding assay contained 50 mM Na-Hepes, pH 7.4, 50 mM KCl, 0.5 mM MgCl 2 , and 0.01 mM CaCl 2 . About 0.2 g of soluble recombinant proteins were incubated in 50 l of this buffer with 0.74 kBq (0.5 nM) [ 3 H]Ins(1,4,5)P 3 and the various unlabeled inositol phosphates or short side-chained inositol lipids for 10 min on ice. The binding reaction was terminated by adding 5 l of human ␥-globulin (10 mg/ml) and 50 l of polyethylene glycol 6000 (30%) (28). Tubes were left on ice for 5 min and were centrifuged at 10,000 ϫ g for 10 min. The precipitates were dissolved in 0.1 ml of 2% SDS, and the radioactivity was counted in a liquid scintillation counter.
Measurements of Binding to PI(4,5)P 2 -containing Lipid Vesicles-Phospholipid binding was performed with mixed lipid vesicles as described recently (29). Briefly, 110 g of PI(4,5)P 2 (Roche Applied Sci- 1. A, amino acid sequence alignment between the PH domains of PLC␦ 1 (human) and PLC␦ 4 (rat). The positions of the ␤-sheets and ␣-helices are indicated above the sequences. Conserved residues are labeled as dots and are highlighted with a dark background, and hyphens indicate gaps. Residues participating in Ins(1,4,5)P 3 binding in PLC␦ 1 PH (19) are marked with black dots. B, lipid binding specificity of the PLC␦ 4 PH domain expressed as a GST fusion protein based on binding to PIP strips (31). PE, phosphatidylethanolamine; PC, phosphatidylcholine; PA, phosphatidic ence) and 1.4 mg of PE (bovine liver; Avanti) were mixed and dried under a nitrogen stream followed by high vacuum, and the dried mixtures were suspended to a final total lipid concentration of 2 mM in 20 mM Hepes, pH 7.2, 100 mM NaCl, 2 mM EGTA, 0.1 g/ml bovine serum albumin by bath sonication. 5 l of the purified GFP fusion protein (1 g) and 5 l of inositol 1,4,5-trisphosphate stock solution were added to 90 l of phospholipid vesicles. As a precaution, proteins were subjected to ultracentrifugation (85,000 rpm for 20 min at 4°C) prior to the assays to remove possible protein aggregates, although protein preparations were used fresh when they had no significant aggregation. The reaction mixture was incubated at 30°C for 10 min and followed by ultracentrifugation at 85,000 rpm for 20 min at 4°C. The 100-l supernatant was mixed with 30 l of 5ϫ Laemmli buffer, and the pellet was resuspended in 100 l of incubation buffer followed by the addition of 30 l of 5ϫ Laemmli buffer. After vortexing, 40 l of each fraction was loaded onto a 12% Tris/glycine gel without boiling and separated by SDS-PAGE at 4°C. After electrophoresis, gels were analyzed in a Storm 860 (Amersham Biosciences) PhosphorImager using blue fluorescence screening for quantitation of the GFP fusion protein band on the gel. Western blot analysis was also performed on parallel samples by using the purified polyclonal antibody against GFP (Clontech).
For binding of recombinant proteins to lipids on PIP strip membranes (30), 100 pmol of recombinant protein was incubated in 5 ml of binding buffer (10 mM Tris, pH 7.5, 150 mM NaCl, 3% bovine serum albumin (lipid-free), 2 mM sodium pyrophosphate, and 0.1% Tween 20) overnight at 4°C, after blocking the strips with the same buffer for 90 min at room temperature. After washing, GFP was visualized by Western blotting using the polyclonal anti-GFP antibody from Clontech essentially as described previously (31). (24) that the isolated PH domain of PLC␦ 4 is unable to bind InsP 3 despite its ability to bind the lipid PI(4,5)P 2 . Because inositol lipids and their watersoluble inositol phosphate counterparts usually compete for the same binding pocket within the PH domain, we wanted to explore further this unique difference between the binding of lipids versus inositol phosphates by comparing the PH domains of PLC␦ 1 and PLC␦ 4 , which show very significant sequence homology within this domain (Fig. 1A). Both PH domains were created as GFP fusion proteins so that their in vitro binding properties could be compared with the cellular distribution of a similar protein construct expressed in mammalian cells. As shown in Fig. 1B, the PLC␦ 4 PH showed the same selectivity for binding only PI(4,5)P 2 as described for the PLC␦ 1 PH (e.g. Ref. 29), and both recombinant PH-GFPs showed PI(4,5)P 2 -dependent association with lipid vesicles. A larger fraction of the PLC␦ 1 PH domain was associated with lipids than of the PLC␦ 4 PH domain (87 Ϯ 2.9 versus 72 Ϯ 5.7, mean Ϯ S.E., n ϭ 6) (Fig.  1C). When InsP 3 was added to displace the protein from the lipids, higher concentrations were needed to displace the PH domain of PLC␦ 4 (IC 50 , 8 M) compared with that of PLC␦ 1 (IC 50 , 1.2 M) (Fig. 1D). These data showed that the PLC␦ 4 PH domain displays high specificity for PI(4,5)P 2 , but it binds both the lipid and the soluble Ins(1,4,5)P 3 with lower affinity than the PLC␦ 1 PH domain.

Binding Properties of Isolated PH Domain-GFP Fusion Proteins-It has been reported recently
Next, the ability of the domains to bind the water-soluble Ins(1,4,5)P 3 was investigated using [ 3 H]Ins(1,4,5)P 3 as a tracer. In these assays, there was a clearly measurable specific binding of [ 3 H]Ins(1,4,5)P 3 to the PLC␦ 4 PH domain, although this was only about 50 -60% of that bound to equal amounts of the PLC␦ 1 PH-GFP protein. Displacement curves showed a significantly lower affinity of the PLC␦ 4 PH domain to the soluble Ins(1,4,5)P 3 (IC 50 , 50 and 15 nM for PLC␦ 4 and PLC␦ 1 PH domains, respectively) ( Fig. 2A), and this difference was also observed when Ins(1,3,4,5)P 4 was used as a displacer (IC 50 , 950 acid; PS, phosphatidylserine. C, binding of recombinant PLC␦ 1 PH-GFP and PLC␦ 4 PH-GFP proteins to lipid vesicles in the presence of increasing concentrations of Ins(1,4,5)P 3 . Lipid vesicles containing PI(4,5)P 2 and phosphatidylethanolamine (or only the latter) were made by sonication and incubated with the recombinant proteins as described under ''Experimental Procedures.'' The fraction of proteins bound to the lipid vesicles were separated by ultracentrifugation, and both the pellets (P) and the supernatants (S) were dissolved in Laemmli buffer and subjected to PAGE analysis. Wet gels were analyzed and quantitated by a PhosphorImager, using the blue laser for excitation of the GFP molecule. D, displacement of the PH domains from the lipid vesicles by Ins(1,4,5)P 3 . The combined results from six experiments are shown where the initial localization (87 Ϯ 2.9 and 72 Ϯ 5.7, percent (mean Ϯ S.E.) of the total for PLC␦ 1 PH-GFP and PLC␦ 4 PH-GFP, respectively) was taken as 100%. PLC␦ 1 PH-GFP, filled circles; PLC␦ 4 PH-GFP, open circles. The faint band migrating below the PLC␦ 1 PH-GFP protein is a minor proteolytic fragment that is invariably present in this preparation. and 160 nM, respectively) (Fig. 2B). These data together suggested that the PH domain of PLC␦ 4 displays lower affinity to both the inositol lipid, PI(4,5)P 2 , and its soluble counterpart, Ins(1,4,5)P 3 , but did not indicate a large degree of discrepancy between binding of the lipid versus the inositol phosphate.
In Vivo Localization and Agonist-induced Responses of GFPfused PH Domains-To determine whether the lower affinity of the PLC␦ 4 PH domain alters its inositide recognition properties within intact cells, these constructs were also expressed in COS-7, NIH 3T3, or HEK 293 cells stably expressing the Ca 2ϩmobilizing AT 1 -angiotensin II receptor. The distribution of the two fluorescent proteins showed significant differences as follows: although the PLC␦ 1 PH domain showed its characteristic plasma membrane localization, the plasma membrane localization of the PLC␦ 4 PH domain was less prominent, and a significant fraction of the signal was found in the cytosol. More strikingly, PLC␦ 4 PH-GFP was also found in the nucleus, with bright dot-like structures present within the nucleoli (Fig. 3A). Addition of ionomycin to elevate cytosolic Ca 2ϩ concentration (to activate endogenous PLC enzymes) caused the translocation of the PLC␦ 4 PH domain from the plasma membrane to the cytosol, and this was reversed upon chelation of Ca 2ϩ (Fig. 3B). These changes were very similar to those observed previously with the PLC␦ 1 PH domain (18), and indicated that the ␦ 4 PH domain was able to report on plasma membrane PI(4,5)P 2 levels, despite its lower PI(4,5)P 2 affinity. There was no obvious change in the distribution of the nuclear signal in response to the Ca 2ϩ increases (Fig. 3A, right).
To compare the dynamic properties of the two constructs within the cells, FRAP analysis was performed in N1E-115 neuroblastoma cells where such properties of the PLC␦ 1 PH-GFP had been characterized previously (27). These experiments showed a significantly faster recovery of the fluorescence in the bleached area in the case of the PLC␦ 4 PH-GFP than with the PLC␦ 1 PH-GFP construct (Fig. 4). Because FRAP of PLC␦ 1 PH-GFP is already much faster than that of a membrane-delimited fluorescent construct (27), these data indicated that the weak membrane localization of PLC␦ 4 PH-GFP is a result of a dynamic equilibrium in which the PH domain rapidly dissociates and re-associates with the membrane PI(4,5)P 2 . Most interesting, FRAP was very slow ( Ͼ100 s) in the nuclear spots, suggesting that the association of PLC␦ 4 PH-GFP with nuclear structures is of a completely different nature than that with the plasma membrane.
Next, the kinetics of changes of PH domain translocation during more physiological conditions was studied. HEK 293 cells stably expressing the AT 1 receptor were transfected with the PH domain GFP constructs, and the distribution of the fluorescent protein between the plasma membrane and the cytosol was followed after agonist addition. Similar experiments were performed in N1E-115 neuroblastoma cells stimulated via endogenous bradykinin receptors. As shown in Fig. 5, angiotensin II stimulation caused a rapid and full translocation of both PH domains from the membrane to the cytosol in HEK 293 cells. When the fluorescence ratios (I membrane /I cytosol ) were used as an index of localization, the translocation responses could be plotted as a function of time following stimulation (Fig.  6). These data showed that although the translocation response of PLC␦ 4 PH-GFP was somewhat more transient than that of PLC␦ 1 PH-GFP in angiotensin II-stimulated HEK cells, the opposite was true in N1E-115 neuroblastoma cells. These small kinetic differences were not statistically significant, but in both cases, the amplitudes of changes were smaller in the case of the PLC␦ 4 PH domain, simply because of its weaker association with the plasma membrane (Fig. 6, A and C). These data showed that, despite its lower lipid affinity, the PLC␦ 4 PH domain is still able to detect PI(4,5)P 2 changes after stimula-tion with a Ca 2ϩ -mobilizing agonist with no significant difference in the kinetics compared with PLC␦ 1 PH.
Studies on the Full-length PLC␦ 4 Protein-Whereas the interaction of the PH domains of PLC␦ 4 and PLC␦ 1 with the inositol lipid, PI(4,5)P 2 , showed many similarities and only quantitative differences, we wanted to determine whether the PH domain of PLC␦ 4 has a similarly important role in the localization of the enzyme as has been widely documented in the case of PLC␦ 1 (32). For this, the full-length enzyme was HA epitope-tagged in its C terminus, and its cellular localization was determined by immunocytochemistry after transfection into COS-7 cells. As shown in Fig. 7, the enzyme was mostly localized to intracellular membranes, with prominent staining in the nuclear membrane, the pericentriolar area, and the endoplasmic reticulum. There was some signal detectable also over the plasma membrane, but clearly, most of the signal was over the ER structures. Because the localization in fixed cells is often slightly different from that observed in live cells, we determined the localization of the PLC␦ 1 protein as well as both isolated PH domains as GFP fusion proteins under similar conditions in the fixed cells for a comparison. As shown in Fig.  7, PLC␦ 1 and both isolated PH domains showed the prominent plasma membrane localization, and only a small signal was present in the pericentriolar area. These data already indicated that the PLC␦ 4 PH domain does not make an important contribution to the localization of the enzyme. To test this conclusion further, we created a truncated PLC␦ 4 protein that lacked the entire PH domain, and the localization of this truncated version was found indistinguishable from that of the full-length protein (Fig. 7). Moreover, increasing the cytoplasmic Ca 2ϩ concentration by ionomycin treatment did not make a significant change in the distribution of the protein other than what was because of the fragmentation of the ER (not shown).
Next we determined whether the PH domains still can display their lipid and Ins(1,4,5)P 3 binding within the full-length enzymes. Therefore, we examined the lipid associations of the full-length proteins and their displacement with Ins(1,4,5)P 3 in a similar manner as with their isolated PH domains. As shown in Fig. 8, although both enzymes were able to bind to the lipid vesicles, higher concentrations of Ins(1,4,5)P 3 were required to displace the full-length proteins from the lipid vesicles than their respective isolated PH domains (IC 50 , 10 and 210 M for PLC␦ 1 and PLC␦ 4 , respectively). Moreover, a bigger difference was observed between the Ins(1,4,5)P 3 sensitivities of the fulllength ␦ 1 and ␦ 4 enzymes (ϳ20-fold) than between their PH domains (ϳ6-fold) (Fig. 8). It is important to note that neither the full-length nor the PH domain-deleted PLC␦ 4 enzyme showed measurable association with PI(4,5)P 2 in the PIP strip assays, and instead, both proteins showed a weak association with the PI(3)P, PI(5)P, and PI(4)P but not with PI(4)P of animal origin (not shown). No specific binding to PI(4,5)P 2containing lipid vesicles was demonstrable with the PLC␦ 4truncated protein (not shown).

DISCUSSION
The PH domain is an important determinant of the mode of operation of PLC␦ enzymes. In the case of PLC␦ 1 , it has been well documented that its relatively high affinity to PI(4,5)P 2 allows the enzyme to be tethered to the plasma membrane so that it can effectively catalyze the hydrolysis of many PI(4,5)P 2 molecules without being released from the membrane. In this model of processive catalysis (33), one of the reaction products, Ins(1,4,5)P 3 , serves as a negative regulator by binding to the PH domain and competing for PI(4,5)P 2 binding and hence disrupting localization of the enzyme. Although all PLC isoforms contain a PH domain, their binding specificity differs from that of PLC␦ 1 , and they do not show the high affinity binding and inhibitory effect of Ins(1,4,5)P 3 , described above (8).  C and D). Cells were transfected, stimulated, and examined by confocal microscopy as described under ''Experimental Procedures.'' A, the ratio of fluorescence intensities between the plasma membrane and the cytosol was calculated for a number of cells in each frame, and these values are plotted as a function of time following stimulation. The same data are expressed as % of maximal (initial) localization in B. For N1E-115 cells (C), the quantification of the ratio of I membrane /I cytosol was done by using post-acquisition automated image analysis as described previously (27). Briefly, for each image, a mask outlining the cell was made based on fluorescence intensity by a thresholding step. Regions of interest corresponding to the "membrane" and "cytosol" areas were then assigned by sequential erosion steps, and the mean intensity in membrane and cytosol area were plotted as a ratio versus time. Again, the same data are expressed as percent of maximal localization on D.
Recent studies performed with PLC␦ 4 already indicated that this enzyme may not be regulated in the same manner as PLC␦ 1 , because it was shown that the ␦ 4 isoform does not show a great sensitivity to inhibition by Ins(1,4,5)P 3 (24). Our studies with the isolated PLC␦ 4 PH domain indicate that although it also binds PI(4,5)P 2 , it does so with a lower affinity than the PLC␦ 1 PH domain. The lower lipid affinity is also paralleled with a comparably lower affinity to Ins(1,4,5)P 3 , although, contrary to the conclusion of a previous report (24), we found that this domain is still capable of binding Ins(1,4,5)P 3 . We cannot explain the difference between our results and those of Nagano et al. (24), but these authors used the PLC␦ 4 PH domain as a GST fusion protein, whereas our studies utilized the PH domain in its natural N-terminal position fused to GFP. Most intriguing, in a recent study (34) molecular modeling of the PH domains of various PLC proteins predicted that the PI(4,5)P 2 binding affinity of PLC␦ 4 PH would be lower than that of PLC ␦ 1 PH.
In agreement with the in vitro binding data, the PLC␦ 4 PH-GFP construct showed only moderate plasma membrane recruitment when expressed in mammalian cells. The faster recovery of PLC␦ 4 PH-GFP than PLC␦ 1 PH-GFP after photobleaching is also consistent with a larger fraction of the protein being cytosolic, causing a faster association rate, resembling that observed with the PLC␦ 1 PH-GFP construct after a partial PLC activation (27). Even with its limited plasma membrane binding, the PLC␦ 4 PH domain shows translocation from the membrane to the cytosol when phospholipase C is activated by a Ca 2ϩ ionophore or by a Ca 2ϩ -mobilizing agonist, indicating that it can also detect the PI(4,5)P 2 changes within the cell. Comparison of the kinetics of translocation obtained by the two PH domains after stimulation with the Ca 2ϩ -mobilizing hormones in two different cell lines and different laboratories revealed only subtle (if any) differences between the two domains. These data together suggest that the two PH domains show lots of similarities and only quantitative differences in their lipid and Ins(1,4,5)P 3 binding properties. A unique and intriguing feature of the PLC␦ 4 PH domain was its prominent presence in the nucleus tightly associated with dot-like structures within the nucleoli. These dots are reminiscent of the so-called ''fibrillar centers,'' the unique locations of tandemly repeated ribosomal genes (35), and seem to be different from the nuclear ''speckles'' that had been shown to contain PIP kinases (36) and diacylglycerol kinase isoforms (37). The significance of this localization is not clear at present, but the lack of similar localization of the full-length protein casts doubts whether this localization would reflect upon an important regulatory aspect of PLC function.
The contribution of the PH domain to the cellular localization of the two proteins is quite different. Whereas PLC␦ 1 localizes primarily to the plasma membrane (8), our data show that PLC␦ 4 mostly associates with the ER. Also, the PH domain of PLC␦ 1 is a critical determinant of the plasma membrane localization of the protein. In contrast, a truncated PLC␦ 4 lack-ing the PH domain showed the same cellular distribution as the wild-type enzyme, suggesting that the PH domain does not make a significant contribution to the localization of this protein, and most likely the C2 domain or even the EF-hand domain is more important in this regard. It is noteworthy that we did not detect nuclear localization of the full-length PLC␦ 4 protein, even though one of its splice variants was shown to localize within the nucleus (38). The primarily ER localization of PLC␦ 4 raises the question of whether this enzyme is regulated by the PI(4,5)P 2 and hydrolyzes the same lipid only at the plasma membrane where a small fraction of the enzyme is probably present, or the ER contains sufficient amounts of PI(4,5)P 2 to regulate the enzyme in that membrane compartment. Alternatively, the enzyme might act on an different inositide substrate, because the ER is not believed to contain high levels of PI(4,5)P 2 (39). Unfortunately, in vitro PLC assays cannot answer these questions, because under in vitro conditions PLC enzymes can usually hydrolyze PI, PI(4)P, and PI(4,5)P 2 (8).
In the present study, the concentrations of Ins(1,4,5)P 3 required to inhibit the binding of full-length PLC␦ 1 or -␦ 4 to lipid vesicles have been one order of magnitude higher than those found with the isolated PH domains. Such a difference has also been documented in the literature based on the reported affinity values of Ins(1,4,5)P 3 binding to isolated PH domains (16,26,29) and the inhibitory concentrations of Ins(1,4,5)P 3 on lipid binding (40,41) or PLC activity (40,41) of the full-length molecule. One explanation for this difference could be simply that the high affinity binding site of the PH domain is not fully exposed in the full-length protein. It is not unreasonable to assume that the PH domain interacts with the rest of the molecule, and Ins(1,4,5)P 3 (or PI(4,5)P 2 ) binding to the PH domain would alter the intramolecular interaction so that the access or affinity of the catalytic site to the substrate lipid is changed. The potent PH domain-mediated inhibitory effect of the PLC␦ 4 splice variant, AltIII, on PLC␦ (but not -␤ or -␥) catalytic activity (24) could be also interpreted as an indication of an inhibitory interaction between the PH domain and the rest of the molecules affecting catalysis. It has been suggested recently (42) that the PH domain could serve as an allosteric modulator of the catalytic activity of PLC␦ 1 via an intramolecular interaction. Although there is no structural information available to evaluate this hypothesis (the PH domain and the rest of PLCd 1 have been crystallized separately), several functional data reported in the literature and some of our present findings could be consistent with such a cooperative interaction between the PH domain and other parts of the molecule.
In summary, the present study shows that the PH domains of PLC␦ 1 and PLC␦ 4 show major differences in their Ins(1,4,5)P 3 and PI(4,5)P 2 affinities but not in their lipid binding specificities. Consistent with its lower PI(4,5)P 2 affinity, the PLC␦ 4 PH-GFP shows only limited plasma membrane localization but still reports very similar Ca 2ϩ and agonist-induced lipid changes as the PLC␦ 1 PH-GFP. Most intriguing, despite its lipid binding, the PH domain of PLC␦ 4 is not an important factor in the localization of the protein, which is primarily associated with the ER. Therefore, the PH domain of PLC␦ 4 may act as a regulator of the enzyme rather than a localization signal via possible intramolecular interactions. More studies are needed to explore the intriguing question of whether the PH domains could serve as lipid-mediated allosteric regulators of the catalytic activity of the PLC␦ enzymes.