I. INTRODUCTION

Top Previous Next References

Phosphoinositide-specific phospholipase C (PLC) isozymes found in eukaryotes comprise a related group of proteins that cleave the polar head group from inositol phospholipids. Under the control of cell surface receptors, these enzymes hydrolyze the highly phosphorylated lipid phosphatidylinositol 4,5-bisphosphate [PI(4,5)P₂], generating two intracellular products: inositol 1,4,5-trisphosphate (InsP₃), a universal calcium-mobilizing second messenger, and diacylglycerol (DAG), an activator of protein kinase C.

Historically, the PLC isozymes have been studied since the 1950s. Early observations by the Hokins (142), and later by Michell (243) and others, led to the recognition of PLC as a key enzyme in agonist-stimulated phosphoinositide metabolism and calcium signaling. The direct link between PLC and the release of intracellular calcium stores was finally forged with the publication of a seminal paper in 1983 by Streb et al. (351) describing the second messenger properties of InsP₃ (reviewed in Ref. 25).

In the late 1980s and early 1990s, three mammalian PLC subtypes, , , and , were isolated and their corresponding cDNA sequences determined (300). Paralleling this work, a number of PLC regulators were identified, especially the GTP-binding (G) _q related subunits (349) (reviewed in Ref. 91) and protein tyrosine kinases (reviewed in Ref. 45). Isolation and identification of these components allowed investigators to test whether already recognized regulatory mechanisms controlled the PLC subtypes. Their results gave rise to the current G protein and tyrosine kinase models of PLC regulation.

PLCs are soluble multidomain proteins ranging in molecular masses from 85 to 150 kDa. Four -, two -, four -isoforms, and numerous spliced variants have been described in mammals. Those found in yeasts, slime molds, filamentous fungi, and plants closely resemble mammalian . Comparisons of their DNA sequences suggest an evolutionary relationship in which PLC- appeared first in primitive single-celled eukaryotes. The PLC- and - subtypes arose later, after the split between fungi or plants and animals, but before the parazoan-eumetazoan split, about 940 million years ago (198); their delayed appearance coincides with the diversification of other signaling components, such as G subunits and protein kinase C. Later duplications of each PLC subtype led to the appearance of the numerous isoforms in animals.

At present, many of the players in phosphoinositide/calcium signaling are identified, some with three-dimensional pictures. On a cellular level, questions of which PLC isozymes go with which regulators are mostly answered. Despite this progress, our understanding of how and where PLC isozymes work in living cells is limited. New information suggests a higher level of organization than is implied by the current regulatory schemes, giving rise to a number of questions: Are these freely diffusing effector proteins or part of a highly organized network? Do these enzymes only act at the plasma membrane? Might they act in concert? Where is their substrate localized and how is it supplied? Finally, what are the physiological functions of the many isoforms and how is their expression controlled? In our review we attempt to address these questions (for other recent reviews see Refs. 91, 182, 299, 335).

	II. STRUCTURE AND CATALYTIC FUNCTION

Top Previous Next References

The sequences of the eukaryotic PLC contain a string of modular domains organized around a catalytic /-barrel formed from the characteristic X- and Y-box regions (392). They include a pleckstrin homology (PH) domain, EF-hand motifs, and a single C2 domain that immediately follows the Y-box region (see Fig. 1). Additional regulatory motifs are present in the - and -subtypes, but absent in PLC-. To simplify the discussion of common domains, we draw comparisons to ₁, the only eukaryotic PLC for which the three-dimensional structure is known (Fig. 1).

View larger version (42K): [in this window] [in a new window]   Fig. 1. Crystal structure of phospholipase C (PLC)-1. Structures of the enzyme (88, 98) lacking the PH domain and the 1 PH domain were solved separately. Four Ca2+ bind PLC-1, three to the C2 domain, and one to the catalytic TIM barrel. Box: linear representation of different domains of 1 and their binding partners. IP3, inositol 1,4,5-trisphosphate; PIP2, phosphatidylinositol 4,5-bisphosphate.

In the crystallographic structure of PLC-₁, the catalytic domain is formed from the X and Y regions, 147 and 118 residues, respectively (88, 90). The domain is comprised of alternating -helices and -strands and resembles an incomplete triose phosphate isomerase (TIM), /-barrel. Like similar structures, the catalytic residues of PLC-₁ are located at one end of the barrel. In this case, the site which is partly rimmed by hydrophobic residues is formed by a shallow cavity at the carboxy-terminal end. The unfinished lip of the barrel forms a spoutlike structure that may allow entry and egress of substrate or product at the membrane surface. The intervening sequence joining the X and Y halves of the barrel (43 residues) is highly disordered and not an integral part of the structure, although it may have an important regulatory function (see discussion below).

Eukaryotic and prokaryotic forms of PLC catalyze hydrolysis of the O-P bond connecting phosphoinositol to DAG. The requirement for inositol is absolute, because the substrate, through the 2-position hydroxyl, participates in nucleophilic attack on the phosphorous, resulting in a cyclic intermediate. Catalysis proceeds by an in-line sequential mechanism involving the cyclic 1,2-phosphodiester intermediate, which can be further hydrolyzed to myo-inositol 1-phosphomonoester (41, 146, 147, 221, 378). Eukaryotic forms readily hydrolyze this intermediate, although the relative amounts of the cyclic and noncyclic product depend on the particular isozyme, substrate, pH, and calcium concentration (188).

Various features of the polar head group affect substrate preference. Although the prokaryotic forms of PLC prefer phosphatidylinositol (PI) and PI-glycans, the eukaryotic enzymes have an order of preference that is generally PI(4,5)P₂ > phosphatidylinositol 4-phosphate [PI(4)P] > PI. Unlike their secreted prokaryotic counterparts, they are incapable of cleaving the polar head group of PI-glycan anchors. Neither forms are capable of hydrolyzing the 3-phosphorylated phosphoinositides.

Within the PLC-₁ catalytic site, a network of hydrogen bonds and salt-bridges ligate the inositol ring substituents and generally account for the observed substrate preference. Lys-438 and Lys-440 of the first half of the /-barrel, and Ser-522 and Arg-549 of the second half, ligate the phosphomonoesters at positions 4 and 5 of the PI(4,5)P₂ polar head group. These are conserved in the - and -isozymes. Interestingly, single amino acid substitutions of Arg-549 do not abolish catalytic activity but switch substrate preference from PI(4,5)P₂ to PI (52, 383).

The catalytic residues, conserved in all eukaryotic PLCs, include His-311, His-356, Glu-341, Asp-343, and Glu-390 (89). A single calcium ion is bound to the active site coordinated by the side chains of Asn-312, Glu-341, Asp-343, and Glu-390 of PLC-₁. The 2-position hydroxyl of the inositol ring and the exocyclic phosphodiester oxygen also appear to contact this metal which plays an essential role in catalysis, lowering the pK_a of the attacking hydroxyl and the negative charge of the transition state. In the current model of the reaction (89), an active site base, possibly Glu-390 in a charge relay system with His-392, strips a proton from the 2-position hydroxyl of the inositol ring, promoting intramolecular attack on the phosphorous and cyclization. His-311 is too far removed to abstract a proton but instead stabilizes the developing charge on the initial pentacovalent transition state. His-356 participates in general acid/base catalysis, protonating the DAG leaving group during the formation of the cyclic 1,2-phosphoinositol intermediate. Acting as a general base, this residue then abstracts a proton from water which attacks the cyclic phosphodiester intermediate. Consistent with this model, amino acid substitutions of His-311 and His-356, as well as the calcium binding residue Glu-341, have been shown to reduce or abolish catalytic activity (52, 85).

Although other domains in PLC have the potential to bind calcium, the single catalytic calcium ion seems to be the only essential metal. This is supported by studies of a PLC-₁ mutant missing other calcium binding sites located in the C2 domain (121). This mutated enzyme which has the same activation constant (K_act) for calcium (~1.4 µM) as the wild-type PLC. The in vitro results also agree with calcium activation constants obtained in permeabilized cells (2). Interestingly, the dissociation constant (K_d) for calcium binding to the catalytic site of PLC-₁, measured by isothermal titration calorimetry, is ~ 30-50 µM, in the absence of phospholipid (121). Although this is substantially greater than its K_act, the crystallographic structure shows that the PI(4,5)P₂ polar head group helps coordinate the metal ion, accounting, at least in part, for the weak affinity measured in the absence of substrate. The affinities of the - and -catalytic sites for calcium have yet to be determined, but their K_act is generally less than those reported for  (see Refs. 165 and 379 for examples). Thus the - and -, but not the -isoforms, are expected to be active at resting cytoplasmic calcium concentrations.

Surrounding the active site is a ridge of hydrophobic residues, Leu-320, Tyr-358, Phe-360, Leu-529, and Trp-555 (88); a similar ridge or rim is found in the prokaryotic forms (130). Such a ridge could insert into the membrane surface in a process required for full enzymatic activity. This proposal is based on studies of PLC-₁, -₂, -₁, and -₁ in which raising the surface pressure of phospholipid monolayers to levels equivalent to, or slightly beyond, the packing densities found in membrane bilayers profoundly inhibits catalytic activity (33, 160, 161, 294). One notable exception is the PLC- isoform found in turkey erythrocytes, which exhibits a pressure optimum that is nearly equivalent to bilayer packing density.

Inhibition by lateral pressures implies that the enzyme must do work to penetrate the membrane surface, bringing the substrate into register with catalytic residues. The presence of hydrophobic residues surrounding the active site further suggests that this ridge inserts into the acyl-chain region. Mutagenesis of the rim, involving replacement of bulky nonpolar residues with alanine, reduces the effects of increased surface pressure in monolayers, without affecting PI(4,5)P₂ hydrolysis in detergent-mixed micelles (84). These results point to an important interaction between the hydrophobic rim and the membrane bilayer and are consistent with a hydrophobic insertion model (a smaller area of penetration reduces the negative slope of the pressure/activity relation). Nevertheless, the identity of the protein sequence and the depth to which it penetrates are unknown. Moreover, these enzymes need only dip between the polar head groups to effectively engage substrate.¹ At this depth they would still experience the lateral pressures exerted in the monolayer experiments.

Among the PLC subtypes, sequences linking the X- and Y-box regions are poorly conserved and are not required for catalysis (39, 83, 318), suggesting a role in subtype-specific regulation. These regions are also susceptible to proteolysis (59, 83, 318), consistent with the idea that they are highly flexible.

Unlike the - and -subtypes, the X/Y-spanning polypeptide in PLC- (also known as the Z region) is extensive, consisting of multiple adaptor domains (Table 3). The -sequences contain two Src homology (SH)2, an SH3, and a single PH domain that engage both protein and lipid binding partners. Although these domains are critical to extrinsic regulation of the -isoforms, they also exert an intrinsic control on catalytic activity (see sect. IV). In contrast, the comparable sequences of PLC- and - subtypes lack any identifiable regulatory motifs. Nonetheless, the relatively short sequences in the - and -subtypes also appear to exert an intrinsic control over the catalytic core (318), raising the possibility that Z-region sequences are key to a general mechanism for controlling PLC catalytic activity.

The PH domain was originally described as a novel protein motif of ~100-amino acid residues, repeated twice in the protein, pleckstrin (platelet and leukocyte C kinase substrate) (128, 239). These motifs have now been identified in >100 other proteins (reviewed in Refs. 219, 296). Most can be grouped by function into a few classes: Ser/Thr protein kinases, Tyr protein kinases, small G protein regulators, endocytic GTPases, phosphoinositide-metabolizing enzymes, and cytoskeletal-associated proteins. Many feature a catalytic site (e.g., protein kinase) and additional adaptor domains. Because PH domains lack any obvious catalytic properties and are found in proteins associated in some way with the membrane, it was suggested that these domains function as adaptors or tethers, linking their host proteins to the membrane surface (99). Principal binding partners are phosphoinositides and the -subunits of heterotrimeric G proteins.

Most eukaryotic PLCs contain a single PH domain of ~130 residues located in the amino-terminal region. An additional PH motif, found in the Z region of PLC-₁ and -₂, is split by two SH2 and a single SH3 domain. PH sequences are not well conserved among the PLCs, suggesting their association with subtype-specific regulation. Interestingly, some PLCs lack any PH domain (Tables 1 and 2; see PLC in higher plants and PLC-₄ spliced variants).

View this table: [in this window] [in a new window]   Table 1. Phospholipase C- isozymes

View this table: [in this window] [in a new window]   Table 2. Phospholipase C- isozymes

The  PH domain binds the polar head group of PI(4,5)P₂ (112, 220) and is archetypical. The domain is required for PLC-₁ to processively hydrolyze its substrate (59, 228, 402), suggesting that it tethers the enzyme to the membrane surface during catalysis.² This is further supported by equilibrium binding measurements (52, 58, 280, 295). Photolabeling studies of the whole protein also point to a single high-affinity PI(4,5)P₂ binding site (180, 361).

The crystallographic structure of the PLC-₁ PH domain, bound to Ins(1,4,5)P₃ (Fig. 1), provides a molecular view of this high-affinity site (98). The whole structure is highly dipolar, with the positively charged surface surrounding the binding cavity where nine residues ligate the 4 and 5 position phosphomonoesters through hydrogen bonds and salt bridges.

This remarkable specificity for PI(4,5)P₂ is also found in the related, but noncatalytic, InsP₃/PI(4,5)P₂ binding protein (IP3BP130) (179, 413) and the PH domain of PLC-₄ (254). Comparable sequences of the primitive -isoforms (in Saccharomyces cerevisiae, Schizosaccharomyces pombe, and Dictyostelium discoidium) bear little resemblance to their mammalian counterparts.

Instead of PI(4,5)P₂, PLC- isoforms bind the higher order polyphosphoinositide, phosphatidylinositol 3,4,5-trisphosphate [PI(3,4,5)P₃], in vitro (10, 94) and in living cells (94), forging a direct link between PI 3-kinase activation and recruitment of PLC-. The K_d for binding the amino-terminal PH domain of PLC-₁ to this lipid is ~1 µM (94). Mutating a sequence localized to the putative loop between -strands 3 and 4 of the ₁ PH domain, a region also important in the binding of PI(4,5)P₂ by PLC-₁, blocks binding.

Although the PH domains of the -isoforms may also serve as membrane tethers, they are not polyphosphoinositide specific. PLC-₁ and -₂ bind strongly to membranes regardless of the presence of these lipids (165, 309). Moreover, PLC-₁, -₂, and -₃ fail to bind InsP₃ or other inositol polyphosphate or polyphosphoinositide analogs, with measurable affinity (361) (see point 1 in NOTE ADDED IN PROOF). The isolated PH domains, like the full-length enzymes, bind with moderate affinity to artificial membrane bilayers, but with relatively little specificity for the phospholipid head group (387). These nonspecific lipid interactions can account, at least in part, for the membrane binding of the intact enzyme (see also sect. IIG).

PH domains of PLC and other host proteins also bind to so-called WD-40 proteins, which fold into -propeller structures whose surfaces harbor binding sites for other proteins (326). One of these, the G- subunit, has special relevance for the PLC-₂ and -₃ isoforms that are activated by G heterodimers. Although the affinities of most PH domains for these G subunits are low (296), their colocalization with PLC at the membrane surface could translate relatively weak binding into specific lateral interactions. Using resonance energy transfer to measure these interactions, we obtained evidence of specificity in the binding of G subunits to the PH domains of several PLC isotypes (387). The order of affinity is PLC-₂ > ₁ and ₁. Moreover, the ₂ PH domain is sufficient for G binding and enzyme activation, since exchanging the PH domain of PLC-₁, a G-insensitive subtype for the corresponding domain of ₂, results in a chimera that is highly activated by G subunits (385). These data are consistent with an earlier study that mapped the sequence for G activation to the amino-terminal two-thirds of the PLC-₂ isoform (395). Although the PH domain seems sufficient, other G interaction sites have also been considered. One promising candidate encompasses the sequence 580-641 of PLC-₂, located within the well-conserved Y half of the catalytic /-barrel (202, 313, 406). These observations support the view that PLC- isoforms engage G subunits through multiple sites, but the critical determinants reside in the PH domain.

PLC isoforms have up to four EF-hand motifs, each consisting of a helix-loop-helix structure. In PLC-₁, as in other EF-hand proteins such as calmodulin and tropinin C, the motifs are divided into pairwise lobes (88). This striking similarity to calmodulin extends to their main chain conformations that are nearly superimposable. As originally noted (88), the conformation of the second lobe and the EF-hand/C2 interface correspond closely to the calcium-saturated form of calmodulin bound to its target polypeptide. The interesting juxtaposition to the catalytic /-barrel suggests the second lobe is not part of a calcium switch, but is instead an integral part of the enzyme's core structure. Indeed, deletions in this region completely inactivate mammalian PLC-₁ (257).

The first two EF-hands present in mammalian, yeast (22, 107, 283, 337, 411), and D. discoideum isoforms (79) possess residues that would appear capable of binding calcium or magnesium ions, whereas EF-hands 3 and 4 do not. There is no evidence, however, that the first two motifs actually bind metal ions, since none is found in crystals of the enzyme soaked in calcium or its analogs (90); calcium binding to these motifs is also not discernible in solution, whereas a single calcium binding site, corresponding to the catalytic /-barrel, is readily detected (121). Furthermore, EF-hands 1 and 2 do not influence calcium sensitivity, since substitution of the putative binding residues in PLC- from D. discoideum is without effect (78). This rule may extend to other PLC subtypes as well, since - and -isoforms retain the four helix-loop-helix motifs but lack residues critical to metal binding.

Although the EF-hand region may have an important regulatory function, it has yet to be identified. In fact, the first two EF-hands, as well as the amino-terminal PH domain, have been dispensed with entirely in higher plants. Of four Arabidopsis PLC sequences, two are also missing a portion of the third EF-hand (125, 139). Similarly, two spliced variants of PLC-₄ from mammalian retina lack the PH domain and first EF-hand motif (102).

C2 motifs, ~120 residues in length, have been identified in numerous proteins, many of which function in lipid-signaling pathways, including PLC (reviewed in Refs. 258, 303). The C2 domain from PLC-₁ (88, 90, 120), like that of synaptotagmin-I (SytIA) (325, 357), consists of eight antiparallel -strands arranged as a sandwich; their main chains can be superimposed. In the ₁-domain, three loops at one end of the -sandwich form the binding sites for up to three calcium ions (90). The calcium binding regions of the ₁ C2 domain, designated (CBR) 1(643-653), 2 (675-680), and 3 (706-714), are well conserved in the various -isoforms found in organisms ranging from yeast to humans, suggesting that some important function has been retained.

Because each coordination complex is completed by water, the binding affinity is assumed to be weak in the absence of membranes, in agreement with the low calcium affinity measured by isothermal calorimetric titration (121). Moreover, disruption of these calcium sites in PLC-₁ fails to affect the calcium-dependent hydrolysis of PI(4,5)P₂ in detergent micelles and phosphatidylcholine bilayers. Despite this evidence, additional ligands (lipid or protein) could contribute to the coordination of calcium in living cells, forming a stable complex that modulates PLC catalytic activity at lower calcium concentrations.³

More recent results indicate that PLC-₁ forms a functional ternary complex with phosphatidylserine (PS) and calcium (EC₅₀ ~ 8 µM) through its C2 domain, in vitro (227). The inability of phosphatidic acid (PA) to substitute suggests the involvement of specific phosphoserine determinants. Importantly, the complex is highly activating, but only when the membrane concentration of PI(4,5)P₂ is limiting (~1 mol%). These results support the tether and fix model (88), wherein the PH domain bound to PI(4,5)P₂ tethers the enzyme, whereas the low-affinity binding of the C2 domain orients and fixes the catalytic core to the membrane surface. Thus PS and calcium bound to the rigid C2 domain could enhance surface sampling by the tethered catalytic core, facilitating processive substrate hydrolysis when the density of substrate is on the order of 1 mol% or less. While attractive, this idea remains to be tested.

Both - and -subtypes also contain C2 domain motifs, yet the key residues involved in calcium ligation are not conserved, a situation reminiscent of some PKC and synaptotagmin subtypes whose C2 domains are also unable to bind calcium. Although these domains may have been retained as an integral part of the PLC catalytic core, they could also function in recognition of other regulatory lipids and proteins. The later possibility is consistent with our recent finding that the PLC-₁ C2 domain binds specifically to GTP-charged _q, its physiological activator (386). As discussed below, the C2 domain of PLC- appears to operate in concert with the carboxy-terminal extension to effectively engage this protein. Whether comparable determinants are present in the C2 domains of the -isoforms is not known.

A single C2 domain and short peptide cap the carboxy-terminal ends of - and -isozymes, whereas -subtypes have extensions of ~400-amino acid residues that contain sequences important to membrane binding, nuclear localization, and their activation by G protein subunits (165, 185, 274, 393). Deletion of this entire carboxy-terminal region from PLC-₁ does not destroy catalytic activity but abolishes activation by G_q and related proteins in vitro (274) and in living cells (393) (see sect. III). Deletions of the carboxy-terminal extensions of PLC-₁ or -₂ or portions thereof, also block binding to acidic phospholipids in vitro (165), association of PLC-₁ with the cell's particulate fraction (185, 393), and transfer to the nucleus (185). The carboxy-terminal region is also required to stimulate the intrinsic GTPase activity of G_q/11 and to link the -isoforms to membrane-associated scaffolding proteins (see sect. IIIE).

	III. THE PHOSPHOLIPASE C- ISOZYMES

Top Previous Next References

Four -isotypes and additional spliced variants have been identified in mammals. -Homologs also have been found in turkey (PLC-_tk)(382), Drosophila (NorpA and PLC-21)(32, 191, 330, 421), Xenopus (105), sponge, and hydra (198) (Table 1); each isoform has distinctive sequences outside the canonical X and Y regions.

PLC- isoforms are regulated by heterotrimeric GTP-binding proteins in a manner that generally fits the adenylyl cyclase control paradigm (see Refs. 91, 92, 335 for previous reviews), yet the coupling of G proteins to PLC involves some features of note. Of special interest is the high GTPase stimulating (GAP) activity of PLC itself, which appears to require an opposing activity of agonist-bound receptor to generate an InsP₃/calcium signal (see sect. IIIB5).

Although PLC regulation has been intensively studied, a number of questions remain concerning the determinants of specificity in the coupling of different receptor subtypes to individual -isoforms and the lateral organization of these signaling components on the plasma membrane surface. Their presence in the nucleus and participation in the nuclear PI cycle further reflects our limited understanding of what they do. While transgenic animal experiments have provided some insights, they have also raised many questions concerning the biological role of each isoform. Especially relevant are recent studies of NorpA in Drosophila phototransduction and the phenotypes of transgenic mice lacking particular PLC isoforms or their regulatory G protein subunits. The various signaling and developmental phenotypes associated with their disruption suggest a few well-circumscribed functions for each isoform.

Mammalian PLC -isoforms are differentially distributed, with each pattern of expression reflecting, to some degree, the functions identified in transgenic work. PLC-₁ is most widely expressed, with the highest concentrations found in specific regions of the brain (114, 145, 246, 302, 353). This PLC is prominent in the pyramidal cells of the hippocampus and, to a lesser extent, in the granule cells of the dentate gyrus, the reticular, mediodorsal, and anteromedial thalamic nuclei (307). PLC-₁ mRNA levels are highest in the cerebellar Purkinje and granule cells, frontal and pyriform cortex, hippocampus, and dentate gyrus; hindbrain structures have relatively low levels of this isoform.

PLC-₁ exists as alternatively spliced variants _1a and _1b (14). The _1b-variant replaces 75 carboxy-terminal residues of the original PLC-₁ cDNA with a unique 32-amino acid sequence. Both are abundant in brain, although the _1a-variant more so. Neither variant is detected in kidney or stomach. Interestingly, most cell lines fail to express any detectable PLC-₁; exceptions include C6Bu-1, PC-12, and NIH-3T3 cells. One of these, the rat C6Bu-1 glioma cell line, expresses both spliced variants (15). Whether the variants are coexpressed in specific neuronal tracts is not yet known.

PLC-₂, first isolated from an HL-60 cDNA library (200, 272), is expressed at highest levels in cells of hematopoeitic origin (211). This pattern of expression is consistent with the part PLC-₂ plays in leukocyte signaling and host defenses (see sect. IIIG).

PLC-₃ protein, originally isolated from rat brain, is widely expressed, with the highest concentrations found in brain, liver, and parotid gland (166). In brain, its mRNA is discretely distributed, with the highest levels found in cerebellar Purkinje and granule cells, and the pituitary gland (362).

PLC-₄ was first isolated from cerebellum (244, 245) and retina (173, 210). Its mRNA is highly concentrated in cerebellar Purkinje and granule cells (308, 362), the median geniculate body (308), whose axons terminate in the auditory cortex, and the lateral geniculate nucleus, where most retinal axons terminate in a visuotopic representation of each half of the visual field. This pattern of expression may be highly relevant to the phenotypes of PLC-₄ null animals (see sect. IIIG).

Several alternatively spliced variants of PLC-₄ have been identified. One PLC-₄ protein, designated the "b" form, is missing the carboxy-terminal 162 amino acids (190). The sequence is replaced by a unique 10-residue peptide. This isoform, which is also found in brain, is recovered exclusively in the cytoplasmic fraction, unlike the "a" variant.

Additional variants have been described in the retina (101, 102). They are divided into two groups: PLC-₄ class "I" and "II," each containing a and b variants that lack the PH domain and first EF-hand motif. These sequences are replaced by a unique amino-terminal region in class II forms. The Ib and IIb variants have an additional 12-amino acid insert within the sequence connecting the X and Y halves of the /-barrel, the result of alternative splicing. More recently, another retinal variant has been described that contains 14 unique amino-terminal residues (3); we term this form variant III (see Table 1).

The Drosophilia PLC-₄ homolog, NorpA, was originally described as an eye-specific gene product (32), but hybridization with cRNA probes in Northern blots showed that the gene encodes at least four transcripts ranging in size from 5 to 7.5 kb (421). These transcripts are expressed in adult body and early stages of development. The various size transcripts of the NorpA gene are accounted for by alternative splicing of two forms of exon 4 which encode slightly different sequences between residues 130 and 155 at the carboxy-terminal boundary of the PH domain motif (191). Termed subtypes 1 and 2, NorpA subtype 1 is eye specific, whereas type 2 is diffusely present in brain and leg and is at high levels in thorax and abdomen. Subtype 1 is required for normal photoreceptor cell development, as well as the normal phototransduction process itself, whereas specific functions have yet to be assigned to the second subtype.

Drosophila PLC-21 differs from NorpA (330) and is expressed as two alternatively spliced variants. More recently, -homologs have been identified in sponge (PLC-S) and hydra (PLC-H₁ and H₂) (198). Although their domain organization seems identical to the mammalian and Drosophila isoforms, there is currently no information on their distribution or function.

PLC- isoforms function as effector enzymes for receptors belonging to the rhodopsin superfamily of transmembrane proteins that contain seven transmembrane spanning (heptahelical) segments (169). They are activated by a wide range of stimuli, from photons and tiny odorant molecules, to full-sized proteins and require specific combinations of G and G subunits to couple to their effectors. In the standard G protein model of PLC activation, binding of agonist triggers receptor-catalyzed exchange of GTP for bound GDP on the -component of the heterotrimer. The GTP-charged subunit then dissociates in the plane of the membrane, and either the -subunit monomer, the -heterodimer, or both bind to PLC-, increasing its catalytic activity and thereby amplifying the initial receptor stimulus (Fig. 2). Because the evidence supporting this model has been extensively reviewed before, we focus on some of the unsettled issues and some of the newer developments that have led to a more complete view of how these enzymes participate in signaling events.

View larger version (18K): [in this window] [in a new window]   Fig. 2. Regulation of PLC- by G protein-coupled receptors (GPCR). Both Gq and Gi/o proteins regulate the function of PLC-. The q subunit of Gq activates PLC directly, whereas -subunits typically released from Gi/o also activate PLC. Hydrolysis of phosphatidylinositol 4,5-bisphosphate (PIP2) generates diacylglycerol (DAG) and IP3, which in turn releases calcium from internal stores. RGS proteins act as receptor-specific down-modulators of the G subunits. Recent evidence also shows that PLC- isozymes act as GTPase-activating proteins. Hence, the continued receptor-catalyzed charging of q with GTP is required to maintain the activated state of PLC-.

Before the identification of specific PLC isotypes or the G proteins involved, both pertussis toxin (PTX)-sensitive and -insensitive G proteins were implicated in phosphoinositide/calcium signaling (36, 93). Differential sensitivity to PTX suggested that G protein-activated forms of PLC might be heterogeneous, which was substantiated by the identification of the four different mammalian -isotypes. Concurrent with their identification, the G_q subfamily of PTX-insensitive -subunits (_q, ₁₁, ₁₄, ₁₅, and ₁₆) were isolated and characterized. _q and ₁₁ are found in nearly all tissues (350), whereas _14,15, and ₁₆ are generally restricted to cells of hematopoeitic origin (390).

When reconstituted into artificial vesicles, these G subunits activate PLC- isoforms but fail to stimulate PLC-₁ or ₁ (91, 334); comparable results are obtained in cotransfection experiments. Nonetheless, the -isoforms are not uniformly responsive to these subunits, with PLC-₂ being considerably less sensitive in vitro (134, 166, 310, 343). Coexpression of PLC- isotypes 1, 2, or 3 and various _q subfamily members produces a similar pattern (173, 199, 208, 395). Interestingly, a recently identified ₄-variant, which is missing a portion of the carboxy-terminal region (G box), is insensitive to _q stimulation (190). Hence, PLC-₁, -₃, -_tk, and most PLC-₄ variants seem to be controlled by _q-related proteins, whereas the less sensitive ₂-isoform is not. Nonetheless, this simple state of affairs is unlikely to pertain when these proteins are expressed in their natural setting, as discussed below.

Although the PLC- isoforms are differentially sensitive to G_q-related subunits, the subunits themselves are nearly interchangeable in their activation of individual isoforms, whether examined by reconstitution in artificial membranes (134, 208, 255, 394) or in cotransfection experiments (173, 199, 208, 397). This promiscuity is also observed, with few exceptions in the coupling of various receptor types to PLC through _q-related subunits (399).

-Heterodimers are now recognized as regulators of many effectors, including selective potassium and calcium ion channels, several isotypes of adenylyl cyclase, and PLC (reviewed in Ref. 61). The realization that these heterodimers activate PLC helps explain the disruption of phosphoinositide/calcium signaling by PTX and integrates toxin-sensitive G proteins of the G_i subfamily into the overall scheme of PLC regulation.

When reconstituted into artificial and biological membranes, G subunits strongly activate mammalian PLC-_2, -₃ (42, 43, 273), and -_tk (37, 382). PLC-₁ is weakly stimulated (42), whereas PLC-₄ is completely insensitive (209). Both ₃ and _tk can be stimulated by _q in the presence of saturating amounts of free G, suggesting the formation of a ternary complex (343, 382). Although the dual activation of these PLC isoforms resembles the regulation of adenylyl cyclase, the effector types differ significantly since the G-sensitive cyclase isoforms, AC II and IV, require concurrent activation by both G and GTP-charged _s (363).

Although PLC-₃ is strongly activated by G in vitro, this isoform is only weakly stimulated in coexpression experiments in COS cells (170). The reason for this discrepancy is unclear, but other work demonstrates that some receptors are coupled through PTX-sensitive pathways to this isoform (250-252). In smooth muscle, adenosine A1, M₂ muscarinic, somatostatin, and µ-, -, and k-opioid receptors can couple through G_i and G_o to PLC, although the extent varies, with some receptors coupling through G_q related proteins as well. These cells contain the PLC- isoforms 1, 2, and 3. Interestingly, the G_i/G_o-mediated component is blocked by antibodies against PLC-₃, but not ₂ or ₁; the extent of inhibition correlates with the extent of PTX sensitivity. Presumably, high concentrations of free -subunits arise from the normally abundant G_i heterotrimers, thereby stimulating PLC-₃. If true, then PLC-₂, which is also present, should have been stimulated, but was not. These results imply that PLC-₂ may be activated by G_q or its associated G subunits, but not the G arising from G_i/o. This seems to contradict the trend determined in artificial and biological reconstitution assays. How these differences arise, when the PLCs are expressed in their natural settings, is unknown.

There are at least 16 distinct G, 6 G, and 12 G subunits, yielding more than 1,300 different heterotrimer combinations. Although not all combinations are possible, the number is huge, giving support to the notion of "right" -subunit combinations. The possibility that specific -combinations are required to activate particular PLC- isoforms has been investigated. Of the limited number of - and -subunits tested, however, most are completely interchangeable when reconstituted with PLC in artificial vesicles (35, 368) or coexpressed in living cells (389). The few exceptions are G_5, which is most effective in stimulating PLC-₂ when cotransfected with the G₂ subunit (389), and retinal G, which is less effective than other subunit combinations in stimulating PLC (368). The relevance of the G₅ activity is questionable, since tissue expression of this subunit and PLC-₂ do not coincide. Moreover, weak stimulation by the retinal heterodimer appears to be caused by the attachment of a farnesyl rather than a geranylgeranyl group to the retinal -subunit. Thus when normal patterns of expression and lipid modification are considered, there is little evidence of effector selectivity for -subunit combinations; rather, this specificity appears at the level of receptor/G protein coupling. This raises the question of specificity in the pathway from receptor to PLC, and the origin of the -subunits. What are the actual concentrations of -subunits liberated by activated receptors in living cells? Are other factors necessary to enhance the potency of -subunits? Are the right -subunit combinations really necessary? Are these subunits sequestered with, or directed to, a particular isoform? These questions remain unresolved.

Like the PLC- isoforms, G protein-coupled receptors also discriminate poorly among G_q subfamily members and various -subunit combinations when reconstituted into artificial membranes or overexpressed in cultured cells (399). Nonetheless, evidence of combinatorial specificity in receptor/effector coupling can be obtained in living cells expressing their normal receptor/G protein complement. The evidence is based mainly on antisense RNA experiments involving the suppression of individual G subunits or their combinations. In the case of the M₁ muscarinic receptor coupling to PLC, several preferred combinations have been identified (G- _q or ₁₁, ₁ or ₄, and ₄) (76). Comparable results have been obtained for other signaling pathways. For example, efficient coupling of M₄ muscarinic and somatostatin receptors to the inhibition of voltage-gated calcium channels requires two completely different heterotrimers, _o134 and _o213, respectively (195).

What confers this remarkable specificity? Clearly, determinants intrinsic to receptor and effector are insufficient, since these bestow selectivity only among -subunit classes. Determinants found in other factors, such as the RGS proteins (see sect. IIIB5), also fail most tests of specificity, although combining different determinants (receptor/G protein/RGS/effector) may help in the selection process. Alternatively, a physical sorting mechanism may sequester specific G-subunit combinations with their cognate receptors. The findings that receptors, G and subunits, PLC isoforms and their substrates, are laterally organized lend support to this idea (see sect. IIIE). Another plausible mechanism would involve unique phosphorylation states of the receptor itself, modulating its ability to couple through distinct heterotrimers and PLC- isoforms. Evidence for this mechanism was obtained in an examination of the PKA-mediated phosphorylation of the ₂-adrenergic receptor and its coupling to G_s and G_i (reviewed in Ref. 218).

Although regions of PLC- that are important to their interactions with G proteins have been identified (see sect. II), the molecular basis of their activation is still unknown. G subunits do not serve as membrane tethering devices for the -isozymes (165, 309), nor do they affect the penetration of the membrane by PLC (249), nor do they increase its sensitivity to calcium (28, 30). The same appears to be true for G subunits. This would suggest that the mechanisms involve acceleration of some step in the catalytic cycle itself, yet attempts to measure the effects of G-subunit activation on substrate or product affinities or the catalytic rate constants have been inconclusive. Part of the problem lies in deriving mechanistic inferences from the kinetics of lipid-hydrolyzing enzymes. These catalysts operate at a membrane/solution interface, forcing the investigator to deal with artificial membrane binding and exchange steps that may be independent of the affinities of the enzyme for substrate and product. Fortunately, water-soluble glycero-1-phosphoinositol 4,5-bisphosphate and similar compounds are now known to be hydrolyzed by PLC (398). This should remove some of the complications, permitting a mechanistic understanding of G protein stimulation.

Members of the G_q subfamily have a slow intrinsic GTPase activity in vitro (~0.8 min¹) (103). This leisurely rate is increased dramatically by the so-called regulators of G protein signaling (RGS) proteins (reviewed in Ref. 24), and the PLC- isoforms themselves. First discovered in yeast (Sst2), at least 19 different RGS proteins have been identified in mammals. Of those tested, RGS2 (137), RGS3 (259), and RGS4 (133, 149) interact most effectively with G_q and block activation of PLC-₁. RGS4 activates _q GTPase by ~25-fold, whereas G-interacting protein (GAIP), which interacts strongly with _i-related subunits, stimulates only about 2-fold at the equivalent RGS4 concentrations (133). In artificial vesicles, RGS2, which is widely expressed (49), is more selective than RGS4 for _q, and is 10-fold more potent at inhibiting PLC-₁ activation in vitro (137). Like RGS4, RGS2 is also an effective GTPase activating protein (GAP) for _i and suppresses G_i-dependent signaling in living cells (157). Thus RGS proteins that exclusively recognize the _q-class have not emerged.

Nonetheless, some RGS proteins are able to suppress calcium signaling in an agonist-selective manner, suggesting they do discriminate among receptor/G_q complexes. In permeabilized pancreatic acinar cells, RGS4 inhibits PLC activation and calcium release from internal stores, but the amounts required to suppress carbachol, bombesin, and CCK differ by more than 10-fold (400); similar results are obtained in intact cells where RGS4 decreases the frequency and amplitude of calcium oscillations and raises the threshold for stimulation. While RGS1 and RGS16 are also potent inhibitors of muscarinic receptor signaling, they weakly suppress CCK-induced calcium release. In this case, the degree of selectivity for muscarinic compared with CCK receptor is 100- to 1,000 fold. In contrast, RGS2 is equivalent in its suppression. This differential sensitivity is not due to the expression of different levels of receptor or G_q-related proteins.

In addition to RGS proteins, _q GTPase is stimulated by PLC-₁, a property that extends to all PLC- isoforms (E. Ross, personal communication). This activity was first demonstrated by reconstitution of M₁ muscarinic receptors, G_q, and PLC-₁ (26). In these artificial membranes, muscarinic agonist stimulates exchange of GDP for GTP on G_q , whereas PLC and agonist increase the steady-state GTPase activity up to 20-fold. Thus the GAP activity of PLC-₁ is balanced by receptor-catalyzed GDP/GTP exchange activity, yielding a population of G_q-GTP that increases the steady-state PLC activity by 90-fold (28) which explains the old observation that hydrolysis-resistant analogs, but not GTP itself, supports PLC/receptor coupling in biological membranes. Interestingly, G subunits also play an important role, suppressing the GAP activities of both PLC and RGS (54). This would further expand the number of points where G subunits could mediate or regulate PLC/receptor coupling.

The remarkable activities of PLC and RGS proteins, and the opposing exchange activity of the receptor, have important implications for the kinetics of signal generation and termination, the stimulus threshold, and the specificity of receptor/G protein coupling to PLC. Their GAP activities suggest PLC and RGS proteins rapidly dissipate the amplification cascade, and control the "noise" of agonist-independent activation, raising the threshold for stimulus-response coupling. Beyond their capability as noise suppressors, RGS proteins have been found to act on specific receptor/_q complexes rather than the isolated  subunits, helping to account for the agonist-specific nature of the PI calcium signal. Moreover, differential RGS expression could enforce a tissue-specific response pattern to a given set of agonists. Depending on the complement of receptors and RGS proteins, multiple different receptors could generate temporally distinct calcium signals in the same cell, while drawing on a common pool of G protein subunits and effector enzymes. The spatial distribution and sensitivity of elementary calcium release and refilling events could further magnify these differences.

Phosphorylation of PLC- by cAMP-dependent protein kinase (PKA) and protein kinase C (PKC) link the -isoforms to heterologous and homologous receptor pathways that modulate phosphoinositide/calcium signals. Reciprocal and synergistic links between cAMP and InsP₃/calcium paths have long been recognized. Most notable are the findings that calcium stimulates some forms of adenylyl cyclase (356), permitting a wide range of cellular responses to agonists that engage these pathways.

The results of most studies show that increasing cAMP suppresses PLC activation, although a few have reported potentiation (reviewed in Ref. 106). The ability of cAMP to inhibit PLC is thought to occur at the level of PLC- phosphorylation. For example, PKA directly phosphorylates PLC-₂, in vitro and in cotransfection experiments, thereby inhibiting its activation by G subunits (224). One of the putative phosphorylation sites, Ser-954, is located in the carboxy-terminal P-box region. Similar results have been reported for PLC-₃ (1). Interestingly, phosphorylation by PKA uncouples receptors that activate PLC-₃ through G_i/o, while preserving the activation by receptors that utilize G_q. The results are entirely consistent with the independent and simultaneous activation of this PLC isoform by G and _q, previously observed in vitro (343). On the other hand, PKA-mediated phosphorylation has been reported to partially block the activation of PLC-₃ by G_q (416). Here the PLC-₃ phosphorylation has been mapped to a single site, Ser-1105, located in the G-box region. Mutation of this Ser to Ala confers resistance to PKA inhibition. Whether this might be related to activation by G subunits derived from G_q, rather than the G_q itself, is unclear.

Various heptahelical receptors linked to PLC can be downmodulated by PKC through the generation of DAG and the rise in cytoplasmic calcium (106). This rapid desensitization may occur at the levels of receptor, G protein, or effector, or some combination thereof. Importantly, the PLC- isoforms themselves are substrates for PKC. Studies of PC12, C6Bu1, and NIH-3T3 cells, which contain PLC-_1, -₁, and -₁, show that treatment with tetradecanoylphorbol 13-acetate (TPA) stimulates phosphorylation of PLC-₁, but not the other subtypes (312). Phosphorylation of PLC-₁ by PKC in vitro results in a stoichiometric incorporation of phosphate at Ser-887, but without any measurable effect on PLC activity. This result was surprising since PKC-mediated desensitization seems to correlate with the level of PLC- phosphorylation. More recently, significant inhibition of the partially purified enzyme was reported, but G_q stimulation was unaffected (223). Suppression of this calcium-stimulated G_q independent activity requires an as yet unidentified cofactor. Comparable results have been reported for PLC-_tk reconstituted in erythrocyte membranes (104). Here the stimulation by P_2Y agonist and guanosine 5'-O-(3-thiotriphosphate) (GTPS) is partly suppressed by phosphorylation. On artificial surfaces, however, activation of the purified enzyme by G_q is hardly affected and stimulation by G subunits is completely insensitive. These results would suggest that other factors, such as the receptor and RGS proteins, are necessary to observe the PKC-mediated inhibition.

PLC-₂, which is closely related to _tk, is not phosphorylated when coexpressed with PKC isoforms, which themselves fail to inhibit PLC activation by G (224). On the other hand, PKC-dependent phosphorylation of PLC-₃ has been observed. Phosphorylation correlates with uncoupling from platelet-activating factor receptors, which are linked to G_q, but not formyl-Met-Leu-Phe (FMLP) receptors, which are linked to G_i/o. These results suggest the same covalent modifications of PLC produce functionally distinct consequences that depend on the particular receptor and its cognate G protein. This might appear to contradict the generally held idea of heterologous desensitization, in which engagement of second messenger-activated protein kinases by one receptor should desensitize other receptors, assuming a common set of consensus phosphorylation sites. Yet heptahelical receptor desensitization is mediated by many different protein players and can occur at many steps on the path to PLC activation, some of which are receptor specific. Differential phosphorylation of the receptors or associated proteins (RGS proteins, -arrestins), could affect how receptor-specific signaling components are uncoupled from their effector, laterally segregated, and internalized (218).

During the sustained phases of receptor activation of PLC- isoforms, the mass of InsP₃ produced often exceeds the fall in cellular PI(4,5)P₂ levels by severalfold. In some cases, levels of this lipid fail to decrease or even rise. The entire agonist-sensi