HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in ISI Web of Science Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (85) Citing Articles via Google Scholar Google Scholar Articles by Roth, M. G. Search for Related Content PubMed PubMed Citation Articles by Roth, M. G.

Phosphoinositides in Constitutive Membrane Traffic

Department of Biochemistry, University of Texas Southwestern Medical Center at Dallas, Dallas, Texas

ABSTRACT

I. INTRODUCTION

II. ENZYMES THAT MODIFY PHOSPHOINOSITIDES ARE IMPORTANT FOR MEMBRANE TRAFFIC

A. PI 3-Kinases and PtdIns 3-Kinases

1. PtdIns3K and Vps34p

2. PtdIns3KIIC2{alpha}

B. PtdIns 4-Kinases

1. PtdIns4KIII{beta}/PIK1

2. PtdIns4KII/PIK2

C. PIP 5-Kinases

1. PIP5KI{alpha}, PIP5KI{beta}, PIP5KI{gamma}, and MSS4p

2. Fab1 and PIKfyve

D. PIP 4-Kinases

E. Phosphatases Acting on Phosphoinositides

1. Sac family phosphatases

2. Other 5-phosphatases

3. 3-Phosphatases

III. PHOSPHOINOSITIDE-BINDING MODULES THAT FUNCTION IN CONSTITUTIVE MEMBRANE TRAFFIC

A. PH Domains

B. PX Domains

C. FYVE Domains

D. ENTH/ANTH Domains

E. Basic Sequences and Other Motifs That Bind Phosphoinositides

IV. INTRACELLULAR DISTRIBUTION AND FUNCTION OF PHOSPHOINOSITIDES FOR MEMBRANE TRAFFIC

A. PtdIns(4,5)P2 at the Plasma Membrane

B. PtdIns(3)P on Endosomes

C. PtdIns(3,5)P2 on MVEs/Vacuoles

D. PtdIns(3)P on the Golgi

E. PtdIns(4)P and PtdIns(4,5)P2 on the Golgi

F. Phosphoinositides on the ER

V. SUMMARY AND CONCLUSIONS

REFERENCES

	ABSTRACT

Top Next References

	I. INTRODUCTION

Top Previous Next References

 
The organelles of the secretory and endocytic pathways in eukaryotic cells have distinct functions, molecular composition, and luminal environment. To achieve this diversity of structure and purpose, the membranes of these organelles must be kept separate for the most part, or they would rapidly fuse and become homogeneous (98, 263). Constitutive membrane traffic is the process by which membrane lipids, integral membrane proteins, and the soluble protein content of membrane organelles are moved from the endoplasmic reticulum (ER) where they are synthesized to the sites where they function. This process requires collecting proteins and lipids that should move from those that should remain behind and packaging them into a transport intermediate (usually thought of as a small vesicle, but this might not be the case in every instance). The transport intermediate fissions from its source membrane and is actively moved to its destination membrane where it fuses and delivers cargo. The process of forming transport intermediates is now understood in some detail (331). Constitutive membrane traffic is very active, and in mammalian cells, many membrane proteins constantly cycle through multiple organelles on a time scale of tens of minutes. In addition to this continual traffic, there are many instances where the rates and direction of movement of membrane proteins and lipids are rapidly changed in response to signals from the extracellular environment. This "regulated" membrane traffic uses much the same machinery and principles as constitutive traffic. The history of research into the mechanisms of constitutive membrane traffic is that of continual discovery of unexpected levels of complexity and regulation (228). There are many more proteins that function for any one step in membrane traffic and many more levels of regulation than we would have thought necessary. In this aspect, research into membrane traffic resembles research into mechanisms of signal transduction. This is more than coincidence. Membrane traffic is controlled through intracellular signal transduction mechanisms that probably work by the same general principles as those that regulate gene expression in response to environmental cues.

Phosphoinositides were first recognized to be important as intermediates in signal transduction cascades where they serve as second messengers and signal integrators. Subsequently mutations that interfered with membrane traffic were mapped to genes encoding kinases and phosphatases that act on phosphatidylinositol (PtdIns) and phosphoinositides (the phosphorylated derivatives of PtdIns), and the role that these lipids play in membrane traffic began to be appreciated (reviewed in Refs. 53, 64, 66, 70, 74, 217, 219, 294, 295).¹ Phosphoinositides (PIPs) are found in unicellular organisms and thus must have appeared quite early in evolutionary history. Whether the phosphorylation of PtdIns to multiple species first arose as part of a mechanism by which cells sensed their external environment, or as part of the machinery for moving membrane proteins between compartments, the molecular strategies for their use are probably similar. Thus it is likely that in membrane traffic PIPs contribute to a complex web of feedback pathways that generate a combinatorial control mechanism, just as they do in signal transduction. This ensures that actions such as vesicle budding or fusion do not occur unless multiple conditions are satisfied, and vesicles do not normally form unless they contain cargo and do not fuse unless they have reached the correct destination.

Currently there are abundant data indicating that phosphatidylinositol 3-phosphate [PtdIns(3)P], phosphatidylinositol 4-phosphate [PtdIns(4)P], phosphatidylinositol 3,5-bisphosphate [PtdIns(3,5)P₂], and phosphatidylinositol 4,5-bisphosphate [PtdIns(4,5)P₂] play important roles in constitutive membrane traffic. However, the number and diversity of roles that any lipid plays in any step in membrane traffic are not yet known precisely. Proteins known to be essential for membrane traffic bind to PIPs at defined locations in the cell, suggesting that one of the functions of these lipids is to establish membrane identity. The activity of some proteins is modified when they bind a particular PIP, indicating that these lipids can act as allosteric regulators. PIPs are also generated and consumed at locations where lipid bilayers are being sharply curved in the processes of membrane fission and fusion. It has been proposed that the changes in lipid shape that occur as PIPs are phosphorylated or dephosphorylated contribute to this process (43). To complicate matters, PIP species can be interconverted by kinases and phosphatases and can even act as regulators of their own production. Determining which, if any, of these processes is important for a particular step in membrane traffic has been a challenge. In addition, PIPs affect the membrane activity of a number of proteins that may have no direct impact on membrane traffic but are found on the organelles where membrane traffic occurs. There currently is little understanding of how potentially mobile lipids such as PIPs can be restricted to membrane subdomains or how competition among various cytosolic PIP-binding proteins is controlled.

Most of our current knowledge of the role of PIPs in constitutive membrane traffic is based on experiments that show that a particular enzyme activity is required (directly or indirectly) for a specific step in membrane traffic or on the discovery that a protein implicated as functioning in membrane traffic can bind PIPs. Relatively less is known about how the lipids themselves function in this process. Thus I will begin by summarizing what is known about the role in constitutive membrane traffic for enzymes that modify PIPs followed by a summary of the proteins known to be involved in membrane traffic that bind each lipid. I will summarize what is known about the distribution of PIPs in cells and finish with more speculative comments on the possible roles of the lipids. I will borrow concepts from our understanding of signal transduction processes to propose a hypothesis to explain how local production of PIPs might contribute to the generation of a transient membrane microdomain and how this might function in the process of constitutive membrane transport.

Superimposed on constitutive membrane traffic are many examples of regulated membrane transport in which PIPs probably serve multiple functions in both membrane traffic and signal transduction. The task of separating direct functions of PIPs in membrane traffic from indirect consequences of downstream signaling cascades is exceedingly difficult. For example, there is an extensive and interesting literature on the effects on membrane traffic that occur when peptide hormone or growth factor receptors activate phosphatidylinositol 3-kinases (63, 71), but it is still unclear if the lipid products of the hormone-activated 3-kinases have a direct effect on endocytosis. For the sake of simplicity, this review is focused on the discussion of aspects of membrane traffic not acutely controlled by extracellular signals. In addition, many genetic studies have identified mutations in proteins that make or bind to phosphoinositides that impact membrane traffic, as well as having pleotropic effects on cytoskeleton or other aspects of cell metabolism. Many of the proteins are undoubtedly interesting, but this review is limited to those for which there is some evidence that they directly affect membrane traffic.

	II. ENZYMES THAT MODIFY PHOSPHOINOSITIDES ARE IMPORTANT FOR MEMBRANE TRAFFIC

Top Previous Next References

 
Although it has been appreciated for 50 years that secretion stimulated by hormones is accompanied by changes in intracellular PIP pools (147), the realization that PIPs were also important for constitutive membrane traffic arose when a mammalian PI 3-kinase was cloned (140, 312) and found to have strong sequence homology to Vps34p, a yeast protein known to be essential for the delivery of proteins to the vacuole in Saccharomyces cerevisiae (137, 334). Subsequently, the PI 3-kinase inhibitor wortmannin was discovered to affect multiple steps in membrane traffic (reviewed in Refs. 36, 295). Although wortmannin inhibits multiple enzymes (68, 250, 253), complicating the interpretation of many experiments, its use pointed the way to more sophisticated investigations employing forward and reverse genetics. It is now clear that a number of enzymes that modify PIPs are required for constitutive membrane traffic and that others probably influence membrane traffic indirectly by modifying the actin cytoskeleton (Table 1). Most recently, some enzymes that participate in signal transduction events have been discovered to also participate in the rapid removal of hormone receptors by endocytosis.

The kinases that phosphorylate PtdIns or PIPs on the D-3 position can be organized into three classes according to amino acid sequence relationships (82, 106). Although class I PI3Ks can phosphorylate PtdIns, PtdIns(4)P, PtdIns(5)P, and PtdIns(4,5)P₂ in vitro, agonists that stimulate their activity mainly generate PtdIns(3,4)P₂ and phosphatidylinositol 3,4,5-trisphosphate [PtdIns(3,4,5)P₃], two lipids present in very low amounts in quiescent cells. Thus the class I enzymes probably phosphorylate mainly PtdIns(4)P and PtdIns(4,5)P₂ in vivo and are important for signal transduction. Upon activation, platelet-derived growth factor (PDGF) receptors bind PI3KI enzymes. Both PI3KI binding and enzyme activity are required to sort the receptors into the degradative pathway after internalization (165). However, it is not clear whether this is a downstream effect of signaling through PtdIns(3,4,5)P₃ produced at the plasma membrane or a consequence of generating PIPs on endosomes. Low concentrations of wortmannin and LY294002 inhibit the endocytic traffic of transferrin receptors (220, 321, 329). Because these concentrations of inhibitors inhibit PI3KIs but not the yeast Vps34p, it was proposed that a class I enzyme might be involved in endocytosis. However, mammalian PI3KIII is inhibited by wortmannin at low nanomolar concentrations (367) and probably was responsible for some of the effects observed. PI3KI isoforms bind to regulatory subunits, and one of these, p85, was identified as part of a protein complex required for the budding of vesicles containing the polyimmunoglobulin receptor from isolated Golgi membranes (170). This budding reaction was inhibited by micromolar wortmannin, a concentration much higher than required to inhibit known class I PI3Ks, and more consistent with inhibition of PI3KII. The identity of the PI3K that might interact with p85 on the Golgi is currently unknown. Thus there is currently no clear evidence for a direct role of PI3KI enzymes in constitutive membrane traffic.

Class II PI3Ks can phosphorylate PtdIns, PtdIns(4)P, and PtdIns(4,5)P₂ in vitro. Relatively little is known about the lipids that class II enzymes produce in vivo. However, PtdIns3KIIC2 is found in clathrin-coated pits at the plasma membrane and on the Golgi (80) and may play a role in membrane traffic (112). The majority of PtdIns(3)P in mammalian cells is produced by class III PI3K (363), the mammalian counterpart of Vps34p (367). PI3KIII phosphorylates only PtdIns to PtdIns(3)P and is more properly called PtdIns3K. This enzyme is required for membrane traffic in the endocytic pathway and probably does not play a role in signal transduction.

1. PtdIns3K and Vps34p

Vps34p associates with a serine/threonine protein kinase, Vps15p, as a heterodimer (334). Binding to Vps15p is required for Vps34p to associate with membranes. Human PtdIns3K binds a 150-kDa homolog of Vps15p and in this heterodimeric form associates with, and is activated by, phosphatidylinositol transfer protein (PITP) (264). PITP plays a role in a number of different membrane traffic events (52, 179), perhaps by transferring substrates to the kinases. Vps34p was first identified as a protein required for the sorting of vacuolar enzymes into the pathway leading from the Golgi to the vacuole in S. cerevisiae (137, 312, 334). It was assumed that the site of action of the enzyme was at the Golgi, since in the absence of Vps34p vacuolar proteins were not sorted correctly and were secreted. Additional evidence supporting a Golgi location for Vps34p/PtdIns3K was provided by treating mammalian cells with wortmannin. Wortmannin inhibited the delivery of cathepsin D to lysosomes (28, 72) apparently through an effect on the sorting of cargo into vesicles rather than through an inhibition of vesicle budding. Wortmannin had only a minor effect on the production of small vesicles from trans-Golgi network (TGN) membranes but prevented the mannose 6-phosphate receptor from entering into vesicles (110). However, yeast Vps34p or mammalian PtdIns3K have not been shown to be located on Golgi or TGN membranes. As is described in more detail below, the proteins that have been discovered to bind PtdIns(3)P are all located on endocytic membranes. In fact, the phenotypes observed when PtdIns3K activity is inhibited do not require that the enzyme act at the TGN. If PtdIns(3)P generated on endosomes was necessary for the recycling of sorting receptors from endosomes to the Golgi, then loss of PtdIns3K function would cause the observed phenotypes at the Golgi as the sorting receptors became trapped in endosomes.

2. PtdIns3KIIC2

PtdIns3KIIC2 has recently been reported to localize in clathrin-coated structures at the plasma membrane and TGN (80, 112). This enzyme binds to clathrin through amino-terminal sequences, and this derepresses enzymatic activity (112). Overexpression of PtdIns3KIIC2 in COS cells inhibited internalization of transferrin receptors and prevented accumulation of mannose 6-phosphate receptors in the TGN, presumably by inhibiting the uncoating of clathrin-coated vesicles (112). Previously, it was observed that PtdIns(3)P and, to a lesser extent, PtdIns(3,4)P₂ enhanced the binding of AP2 adaptors to peptides containing internalization signals. Clathrin binding enhanced the affinity of AP2 for internalization signals to a similar extent, but the effect of clathrin and PtdIns(3)P were not additive (286). However, when fluorescent proteins that bind PtdIns(3)P are expressed in cells, they label endosomes and not the plasma membrane (32, 92, 318), suggesting that there is little free PtdIns(3)P at the cell surface. There is a recent report that PtdIns3KIIC2 localizes to the nucleus rather than the plasma membrane (77). More work is required to determine how PtdIns3KIIC2 functions in membrane traffic and to what extent PtdIns(3)P is involved in membrane traffic from the plasma membrane.

B. PtdIns 4-Kinases

Two distinct classes of kinases phosphorylate PtdIns at the D-4 position. These are called type II and type III (type I was discovered to be a PtdIns 3-kinase). Type III enzymes are homologous to two yeast PtdIns4Ks, PIK1 (102) and STT4 (405). The mammalian ortholog of STT4 is PtdIns4KIII (251, 391) and the ortholog of PIK1 is PtdIns4KIII (12, 233, 250). In yeast, STT4 and PIK1 do not compensate for each other in deletion studies, with STT4 functioning in regulation of the actin cytoskeleton and PIK1 essential for membrane traffic (9, 129, 370). PtdIns4KIII has been localized to the Golgi, a location consistent with the work on yeast PIK1 (120, 392). STT4 was identified in a screen for mutations in S. cerevisiae that prevent aminophospholipid transport to the Golgi (358), and its mammalian counterpart, PtdIns4KIII, is reported to localize at the ER (251). However, currently there is no evidence that the effect of STT4 on lipid traffic is direct. By extension, it is possible that mammalian PtdIns4KIII is not involved in membrane traffic.

Type II PtdIns4K and - were only recently cloned (17, 237). These enzymes do not have the signature PIK domain and belong to a family of lipid kinases distinct from the other PtdIns3/PtdIns4 kinases. PtdIns4KII is a major contributor to cellular PtdIns(4)P levels and is located on the Golgi where it plays a role in membrane traffic from the TGN (372). PtdIns4KII is a cytosolic protein that is recruited to plasma membrane and activated by Rac1 and may have no role in membrane traffic (377). In S. cerevisiae, Lsb6p is the ortholog of mammalian PtdIns4KII enzymes (130, 315).

1. PtdIns4KIII/PIK1

Pik1p is an essential 125-kDa enzyme of S. cerevisiae (102) found both in the nucleus and on Golgi membranes (370). The phenotype of pik1^ts yeast grown at nonpermissive temperature is similar to the phenotype produced by a conditional loss of ARF function (9). Overexpression of Pik1p suppresses the defect in secretion in yeast expressing a temperature-sensitive allele of Sec14, the yeast PITP (129). The mammalian Pik1p counterpart, PtdIns4KIII, is a 90-kDa enzyme that has been localized to Golgi membranes (120, 392). PtdIns4KIII enzyme activity is stimulated in vitro by ARF1 (120), suggesting that it is an effector of ARF for membrane traffic. Overexpression of mammalian frequenin, the homolog of an activator of Pik1p in S. cerevisiae (136), stimulates the delivery of a reporter protein to the apical surface in polarized cells (379). Both the PtdIns4KIII and Pik1p behave as soluble proteins (102, 392), suggesting that their association with membranes must be dynamic and regulated. PtdIns4KIII is inhibited by 50–100 nM wortmannin (233), but Pik1p is highly resistant to this inhibitor.

PtdIns4K activity is found on chromaffin granules and on small synaptic vesicles and is required for vesicle fusion (386, 387). PtdIns4K activity is also found on vesicles containing the Glut4 glucose transporter isolated from muscle (190). The identity of the enzymes responsible for these activities is not known.

2. PtdIns4KII/PIK2

Biochemical studies indicate that type II PtdIns4K is responsible for much of the PtdIns 4-kinase activity in response to extracellular signal transduction (106). There are two isoforms of the mammalian enzyme (17, 237, 372, 377). PtdIns4KII is located on perinuclear membranes that include the Golgi and on synaptic membranes (123, 372). Overexpression of PtdIns4KII in fibroblasts has no effect on transport of a reporter from the ER to the Golgi but does stimulate transport from the TGN to plasma membrane (372). Knockout of PtdIns4KII by small interfering RNA oligonucleotides causes AP1 clathrin coats to be released from Golgi membranes, and the Golgi to fragment. Export of a viral glycoprotein from the TGN is also inhibited. The binding of AP1 to Golgi in these cells can be rescued by replacing PtdIns(4)P but not by replacing PtdIns(4,5)P₂. However, the glycoprotein transport defect is rescued by both lipids. Thus the effect of PtdIns4KII appears to be to produce PtdIns(4)P, which is directly recognized by coat proteins at the TGN and also serves as a precursor of the PtdIns(4,5)P₂ required for membrane transport to the plasma membrane (372).

A PtdIns4KII activity coprecipitates with CD63, a tetraspannin protein found mainly on lysosomes (50, 231), and is also found in plasma membrane lipid rafts, although it is not enriched in rafts that contain caveolin (376). The single yeast homolog of mammalian PtdIns4KIIs is LSB6, now renamed PIK2. This gene is nonessential (130, 315), indicating that it is not required for the functioning of the secretory pathway, although it still might have a more specialized role in secretory processes. The functions of endosomes and the vacuole are not essential for yeast to grow in the laboratory, so a role of PIK2 in endocytic processes is still possible.

C. PIP 5-Kinases

PIP 5-kinases were originally purified as activities that phosphorylated PtdIns(4)P to PtdIns (4,5)P₂ and subsequently cDNAs for six enzymes have been cloned (25, 37, 157, 159, 209). Based on sequence relationships, the PIP5Ks were grouped into two families called types I and II. Subsequently, it was realized that type II kinases phosphorylated the D-4 position and not D-5 and that both type I and II enzymes also phosphorylated the D-3 position (283, 409). Thus the PIP5KII proteins are more accurately called PIP4Ks. The - and -type I kinases were shown to phosphorylate PtdIns to PtdIns(5)P in vitro (357) and could properly be called PI5Ks. However, it is not clear if they produce PtdIns(5)P in vivo. Due to this uncertainty and because the type I enzymes are most often referred to as PIP5Ks in the literature, that term will be used in this review. Three human PIP5K enzymes have been identified, and all are related to the MSS4p kinase of S. cerevisiae. As major producers of PtdIns(4,5)P₂, the PIP5Ks play important roles in membrane traffic; however, the precise roles played by the various isoforms of these enzymes are not yet clear. S. cerevisiae has a second PI 5-kinase, FAB1, that has recently been shown to phosphorylate PtdIns(3)P to produce PtdIns(3,5)P₂ (57, 115). The mouse ortholog of this enzyme has been cloned and named PIKfyve (306, 320). PtdIns(3,5)P₂ is necessary for proper membrane traffic to the yeast vacuole; thus PIKfyve is very likely to play a similar role in mammalian cells.

1. PIP5KI, PIP5KI, PIP5KI, and MSS4p

The mammalian PIP5Ks - and -isoforms were cloned independently from human or mouse by two laboratories and named in a reciprocal manner (157, 209). A third isoform of PIP5KI, , has also been cloned from both species (158). Unfortunately, the reciprocal nomenclature for isoforms and has been perpetuated in the genome sequence databases for human and mouse. This complication of the nomenclature means that one must be careful to note the species of enzyme used in any study. To simplify the process of specifying which gene product has a particular function, for the purpose of this review I will adopt the human genomic nomenclature where PIP5KIA indicates human PIP5KI and murine PIP5K, PIP5KIB is human PIP5KI and murine PIP5KI, and PIP5KIC indicates PIP5KI.

PIP5KIA and B are 61-kDa proteins, and C is larger and has two forms, 87 and 90 kDa, due to alternative splicing. The kinase domain of PIP5KIs is comprised of 400 amino acids located centrally and is 80% identical among all three proteins. Phosphatidic acid (PA) has been shown to stimulate these enzymes (164, 243), and all three of the PIP5Ks are activated 10-fold by this lipid (158). A major source of PA is through the hydrolysis of phosphatidylcholine by phospholipases D₁ and D₂, enzymes that are themselves activated by PtdIns(4,5)P₂ (326). In vitro the PIP5KIs can phosphorylate phosphoinositides other than PtdIns(4)P and will convert PtdIns(3,4)P₂ to PtdIns(3,4,5)P₃ and PtdIns(3)P to both PtdIns(3,5)P₂ and PtdIns(3,4,5)P₃ (106, 356, 409).

The regulation of PIP5KIs is complicated. These enzymes can be precipitated in complexes that contain the small G proteins Rho and Rac (46, 289, 355), although it is not known if this is a direct interaction or involves additional proteins. PtdIns(4,5)P₂ is an important regulator of the actin cytoskeleton, and overexpression of PIP5KIB causes massive actin polymerization that is not prevented by coexpression of a dominant negative form of RhoA, suggesting that the lipid kinase is downstream of Rho (316). Recently an activator of PIP5KIB was purified and found to be the small G protein Arf1 (151). Arf1 has many activities that are important for membrane traffic (293). Purified Arf1, Arf5, and Arf6, but not RhoA or Rac1, could stimulate purified PIP5KIB in vitro, and this stimulation required PA. However, only Arf6 colocalized with PIP5KIB at the plasma membrane in vivo and is likely to be the regulator of PIP5KIB relevant to regulation of the actin cytoskeleton (27, 39, 151). Although an activated Rac1 allele stimulated membrane ruffles and colocalized with Arf6 and PIP5KIA, membrane ruffles were prevented when a dominant negative form of Arf6 was expressed (151). Thus current data suggest that Arf6 may be downstream of RhoA and Rac1 and interact directly with PIP5KIB at the plasma membrane.

In addition to regulation by small G proteins, PIP5KIB is regulated by phosphorylation (266). PIP5K isolated from Saccharomyces pombe membranes is also regulated by phosphorylation by Cki1, a casein kinase I ortholog (362). All three mammalian PIP5Ks will autophosphorylate in vitro, and this is stimulated by PtdIns but not other phosphoinositides (160). In all cases documented so far, phosphorylation suppressed PIP5K activity.

PtdIns(4,5)P₂ is important for many aspects of membrane traffic including endocytosis, synaptic vesicle fusion and recycling, regulated exocytosis, phagocytosis, and vesicle formation at the Golgi (53, 66, 119, 171, 217). In most cases, systematic comparisons of the PIP5K and PIP4K isoforms that might be responsible for producing PtdIns(4,5)P₂ for these activities have not been performed. Overexpression of wild-type PIP5KIA, but not PIP5KIB, stimulates internalization of the EGF receptor, and overexpression of a kinase dead mutant blocks EGFR endocytosis (14). However, overexpression of PIP5KIB and to a lesser extent PIP5KIA enhances endocytosis of the transferrin receptor in a different cell type (262). Knock-down of PIP5KIB by siRNA, but not knock down of PIP5KIA or PIP5KIC, inhibits endocytosis of transferrin. Interestingly, expression of the C isoform was increased when transcription of either A or B was inhibited, but this did not result in rescue of endocytosis rates or a change in total cellular PtdIns(4,5)P₂ levels, suggesting that activity of the enzyme is regulated at a posttranscriptional level (262). PIP5KIB can be recruited on to Golgi membranes where it is directly activated by Arf1 (168). PIP5KIC is the major PIP5K in synapses and is an important regulator of the recycling of synaptic vesicles (78, 382).

MSS4 encodes the PIP5K of S. cerevisiae (76, 150) and is an essential gene (388). Acute loss of MSS4p function causes alterations in the actin cytoskeleton but not in secretion. Currently there are no data supporting a role for PtdIns(4,5)P₂ for secretion in S. cerevisiae, in contrast to a requirement for PtdIns(4)P (129, 370). There is indirect evidence for a role for PtdIns(4,5)P₂ in the internalization step in endocytosis in yeast (154), and by extension, a role for Mss4p in that process.

2. Fab1 and PIKfyve

Fab1p is the PI3P 5-kinase in S. cerevisiae and is required for maintenance of the vacuole, although not for membrane traffic to the vacuole (57, 115, 399). A mouse homolog of FAB1 has been cloned (306). Sequence homology searches reveal that the human Fab1 ortholog PIKfyve is encoded by a single gene on chromosome 2. Both PIKfyve and Fab1p contain a domain called a FYVE domain (see below) that binds to PtdIns(3)P, and in the case of PIKfyve, is required for the enzyme to locate on endosomes (308). PIKfyve will phosphorylate PtdIns or PtdIns(3)P to PtdIns(5)P or PtdIns(3,5)P₂, respectively, and also has intrinsic protein kinase activity (307). Like the PI5Ks, PIKfyve phosphorylates itself, and this inhibits its lipid kinase activity (307). Overexpression of a mutant PIKfyve able to produce PtdIns(5)P but unable to produce PtdIns(3,5)P₂ causes extensive vacuolation of cells that is rescued by microinjecting them with PtdIns(3,5)P₂ but not PtdIns(5)P (156). Thus PtdIns(3,5)P₂ appears to be a product of PIKfyve in vivo required for endosome function.

D. PIP 4-Kinases

Three 48-kDa kinases have been identified that phophorylate phosphatidylinositol 5-phosphate. PIP4K was originally purified as an activity that phosphorylated a commercial PtdIns(4)P preparation to PtdIns(4,5)P₂ (18, 208) and was named PIP5K type II. After this and a related enzyme, PIP5K type II, were cloned (25, 37, 79), it was discovered that the enzymes actually phosphorylated the D-4 position (283). The actual substrate in the bovine brain PtdIns(4)P that was used for the original purification of the enzymes was a previously unidentified lipid, PtdIns(5)P. Thus these enzymes should be called PIP4Ks. PIP4K can phosphorylate PtdIns(3)P and PtdIns(5)P but does not make PtdIns(3,4,5)P₃ (283, 409) and will not phosphorylate PtdIns (106). The PIP4Ks are also not stimulated by PA (164). Currently little is known about the roles of these enzymes in cells or if they impact membrane traffic in any way. PIP4K partially localizes to the nucleus (49). A third member of this family, PIP4K, has been identified as resident in the ER (159).

E. Phosphatases Acting on Phosphoinositides

As one would expect, cells not only generate PIPs through phosphorylation but also consume or interconvert them through the action of phosphatases. Many different enzymes have been identified that specifically remove phosphate from one or more of the positions on the inositol ring (155, 213, 240). There is currently evidence for a role in membrane traffic for enzymes that remove phosphate from the D-5 position of PtdIns(4,5)P₂, the D-5 position of PtdIns(3,5)P₂, the D-3 position of PtdIns(3,5)P and that remove the phosphate from both PtdIns(4)P and PtdIns(3)P (Table 2). It is likely that enzymes that specifically hydrolyze phosphate from PtdIns(3,4,5)P₃ or PtdIns(3,4)P₂ play roles in signal transduction, but not directly in membrane traffic. There is currently not much information about enzymes that remove phosphate only from the D-4 position of PtdIns(4)P, and S. cerevisiae does not have homologs to the currently identified mammalian enzymes. It is likely that the important function of degrading PtdIns(4)P in S. cerevisiae is performed by Sac1p.

The first lipid phosphatases shown to have a role in membrane traffic contain a domain originally recognized in the yeast protein Sac1p. One group of these enzymes contains only a NH₂-terminal Sac domain, and the second class contains an additional, 5-phosphatase domain in the center of the protein followed by various other domains (155). Sac1p was identified in a screen for mutations that relieved the block in secretion caused by loss of activity of the S. cerevisiae PITP, Sec14p (51, 385), and was subsequently discovered to have PIP phosphatase activity (47, 124). In S. cerevisiae there are two such proteins, Sac1p and Fig4p, and in humans three, KIAA0274, KIAA0966, and KIAA0851. Sac1p localizes to the ER and Golgi (311, 385), and sac1 cells contain 10-, 2.5-, and 2-fold increases in PtdIns(4)P, PtdIns(3,5)P₂, and PtdIns(3)P, respectively (124). In the ER, Sac1p is necessary for ATP import into the lumen (224), but it also has important phosphatase function antagonizing the PtdIns 4-kinase Pik1p in the Golgi. Fig4p is induced by pheromone and required for actin polarization during mating (96). KIAA0274 is probably the human ortholog to yeast Sac1p, and KIAA0851 is probably the Fig4p ortholog. KIAA0966 is a large protein of unknown function. The catalytic Sac domains of all these proteins except KIAA0966 hydrolyze PtdIns(3)P, PtdIns(4)P, and PtdIns(3,5)P₂, but not PtdIns(4,5)P₂ (155). KIAA0966 is a 5-phosphatase with preference for PtdIns(4,5)P₂ over PtdIns(3,4,5)P₃ (236).

The second group of Sac domain phosphatases includes the yeast proteins Ins51p (Slj1p), Inp52p (Slj2p), and Inp53p (Slj3p) and mammalian proteins synaptojanin 1 and 2. The phenotypes of cells with null mutations in either INP51, INP52, or INP53 are relatively normal, but deletion of all three is lethal (332, 343). Ins51p is a 5-phosphatase specific for PtdIns(4,5)P₂, and its Sac domain is inactive (124). INS51 has genetic interactions with PAN1, a protein required for endocytosis and regulation of actin (380). Inp52p and Inp53p have both 5-phosphatase activity as well as Sac domain phosphatase activity and are involved in regulating actin patches at the plasma membrane (PM) (260, 338). Disruption of INP53, but not INP51 or INP52, causes sorting defects at the TGN and increases the rate of transport of reporter proteins to the vacuole (19, 125). This is due to a defect in the pathway between the TGN and early endosomes that requires AP1/clathrin (126). This defect is more severe in an inp52 inp53 mutant and is complemented by a mutant inp52 lacking Sac phosphatase activity, indicating that it is the excess PtdIns(4,5)P₂ that is the primary problem in these cells (338). Double mutants inp51 inp52 or inp52 inp53, but not inp51 inp53, have defects in endocytosis and disruption of the actin cytoskeleton (327, 332, 343). This suggests that Inp52p has overlapping function with the other two enzymes. In inp51 inp53 inp52^ts cells at nonpermissive temperature, PtdIns(4,5)P₂ is detected on internal membranes where it is normally not detected (338). Thus these enzymes control the location of PtdIns(4,5)P₂ in S. cerevisiae.

Synaptojanin 1 and 2 are dual-function mammalian phosphatases that contain a 5-phosphatase activity as well as a Sac domain and convert PtdIns(4,5)P₂ to PtdIns (124). This ability to decrease PtdIns(4,5)P₂ without producing the potentially active intermediate PtdIns(4)P may be important for maintaining spatial segregation of PIPs (discussed below). Synaptojanin 1 is found in nerve terminals associated with membranes coated with clathrin (127, 227). Mice deficient in synaptojanin 1 (67, 181) and mutants in Unc26, the C. elegans synaptojanin ortholog (134), have neurological defects and nerve endings that accumulate coated vesicles, suggesting that there is a defect in vesicle uncoating. Synaptojanin 2 has a broader tissue distribution and differs from synaptojanin 1 at the COOH terminus (254). Synaptojanin 2 is reported to be an effector for the small GTPase Rac1, and overexpression of activated Rac1 or a synaptojanin 2 targeted to membranes inhibits endocytosis (214). Depletion of synaptojanin 2, but not synaptojanin 1, by small interfering RNA decreased internalization of epidermal growth factor (EGF) and reduced numbers of coated pits in lung carcinoma cells (299).

2. Other 5-phosphatases

Inp54p, the fourth 5-phosphatase enzyme in S. cerevisiae, lacks a sac domain and has a single 5-phosphatase domain. It localizes to the ER, and its deletion increases the rate of secretion of a reporter, indicating that it functions directly or indirectly in secretion (389). In mammals there are a number of other 5-phosphatases, most of which have been investigated for roles in signal transduction or apoptosis, but not membrane traffic. However, the OCRL gene product that is defective in the human disease Lowe syndrome is a 5-phosphatase that localizes to the TGN (87, 344, 408). Kidney cells from patients with Lowe syndrome have elevated PtdIns(4,5)P₂ levels (407) as well as elevated serum levels of lysosomal enzymes (360), raising the possibility that there is a defect in sorting proteins at the TGN.

3. 3-Phosphatases

PTEN/MMAC1, myotubularin, and myotubularin-related proteins are phosphatases specific for the D-3 position of phosphoinositols (212). PTEN hydrolyzes PtdIns(3,4,5)P₃ and is a negative regulator of signaling. It probably has no direct role in membrane traffic. Myotubularin (MTM1) was identified as the defective gene in X-linked myotubular myopathy, a defect in muscle development. Mutations in a related gene, MTMR2, cause type 4B Charcot-Marie-Tooth syndrome. There are eight myotubularin-related genes in humans and one in S. cerevisiae (YMR1). Recombinant MTM1, MTMR2, MTMR3 (KIAA0371), MTMR6, and Ymr1p proteins are active PtdIns(3)P phosphatases that also hydrolyze PtdIns(3,5)P₂ (309, 350). The activity of MTM1 toward PtdIns(5)P in vitro is 200-fold less than for PtdIns(3)P, and other phosphoinositides are significantly poorer substrates (350). The physiological function of MTM family members is not known. However, MTMR3 contains a COOH-terminal FYVE domain for binding to PtdIns(3)P, as do a number of proteins that bind to endosomes, and may be a candidate for an enzyme responsible for controlling the amount and/or location of that lipid on endosomal membranes.

	III. PHOSPHOINOSITIDE-BINDING MODULES THAT FUNCTION IN CONSTITUTIVE MEMBRANE TRAFFIC

Top Previous Next References

 
The second discovery suggesting that phosphoinositides might be important for regulating constitutive membrane traffic was the realization that some proteins known to be required for membrane traffic contained a motif, the pleckstrin homology (PH) domain, known to bind PtdIns(4,5)P₂. Subsequently, three additional motifs, the FYVE, ENTH, and PX domains, were identified in many proteins including some that had been identified through screens for mutations that affected membrane traffic in S. cerevisiae. These modules were subsequently shown to bind phosphoinositides. In certain proteins, these lipid-binding domains show great specificity for one phosphoinositide, but in most cases, they will bind to more than one species. A current unresolved problem is that in vitro many of these modules will show higher affinity for PtdIns(3,4,5)P₃ than other PIPs. However, it is unclear if this difference in affinity is enough to be meaningful in the cell, where PtdIns(3)P, PtdIns(4)P, and PtdIns(4,5)P₂ are 50–1,000 times more abundant than is PtdIns(3,4,5)P₃.

A. PH Domains

The PH domain was first identified as a sequence motif of 100–120 amino acids that was found in many signaling proteins as well as in the cytoskeletal protein spectrin (135, 223, 246). In phosphoinositide-specific phospholipase C-1 (246), the PH domain was located in the region previously shown to bind to PtdIns(4,5)P₂ (48). Subsequently the PH domains from several proteins were found to bind specifically to liposomes containing PtdIns(4,5)P₂ (133). Primary sequence conservation among different PH domains is only 7–30%, but the structures that have been solved are quite similar. PH domains have a seven-stranded, antiparallel -sheet that is twisted to fold back on itself as an orthogonal sandwich (100, 204, 288, 303). Most PH domains have strong charge polarity with one edge of the curved sheet much more positive than the other. The positive side interacts with the negative head group of the lipid. The affinity and specificity with which different PH domains bind phosphoinositides vary greatly. Most have only weak affinity (K_D of 30–40 µM) (133, 303), but some bind 10–1,000 times more tightly, especially those specific for PtdIns(3,4,5)P₃ (200). The concentration of PtdIns(4,5)P₂ in neutrophil plasma membranes has been estimated as 3–5 mM and that of PtdIns(3,4,5)P₃ at 5 µM (basal) to 200 µM (after stimulation) (341). Thus, unless a PH domain exhibits 25- to 1,000-fold greater affinity for the PtdIns(3,4,5)P₃ than PtdIns(4,5)P₂, it is unlikely to bind selectively in the cell. A K_D in the range of 30 µM means that stable binding to membranes by low-affinity PH domains should require additional interactions and experimental evidence supports this. In many if not most cases, stable membrane binding of proteins that contain PH domains involves interaction with other segments of the same protein, or interaction with additional proteins (184, 196–198). Binding to phosphoinositides has not been observed to alter the conformation of the PH domain (133), although it may change the interaction between the PH domain and another domain in the same protein (23).

Of the 251 PH domains identified in the human proteome, 20 or so have a conserved motif that allows high-affinity binding to PtdIns(3,4,5)P₃ or PtdIns(3,4)P₂ (100, 204) and probably function in signal transduction (200). Many of the remaining proteins do not yet have known functions, and some of these may contribute to membrane traffic. The proteins known to function in membrane traffic that contain PH domains include the three dynamin GTPases, some of the guanine nucleotide exchange proteins for Arf family members, some of the GTPase activating proteins for Arf, two kinesin motor proteins, and several lipid modifying enzymes (Table 3). Additional information on PH domains and the proteins that contain them can be obtained from recent review articles (23, 200, 278, 303).

PX, or phox (phagocyte oxidase), homology domains are found in two subunits of NADPH-oxidase, in PI3KC2 and in a family of small proteins called sorting nexins (SNX) (128, 277). They are also found in certain other yeast proteins known to be required for protein traffic between the Golgi and endosomes (91, 153, 368) and in phospholipase D₁ (PLD₁) and phospholipase D₂ (PLD₂) (105) (Table 4). The SNX proteins were key to understanding the function of the PX domain. The first sorting nexin family member, SNX1, had been identified as a protein that bound to the EGF receptor and that had sequence homology to a yeast protein known to function in membrane traffic to the vacuole (192). Subsequently, a number of SNX proteins have been shown to associate with specific membrane receptor proteins as part of an oligomeric complex that regulates sorting of receptors between recycling and degradation pathways (128, 153, 342). SNX proteins appear to associate into complexes with other SNX proteins, as well as with adaptor proteins that bind to receptors and to clathrin (193, 206, 211, 267). The fraction of the genome devoted to SNX proteins in S. cerevisiae is threefold greater than it is in the human genome. Therefore, most SNX proteins probably function in evolutionarily conserved processes common to most cell types. Evidence for the functions of SNX proteins currently is dominated by results of experiments in which wild-type or mutant forms are overexpressed as dominant negative inhibitors of endocytosis. Because overexpression of one member of a family of proteins can affect processes unrelated to the normal function of that protein, more work is needed to determine where each human SNX protein functions.

C. FYVE Domains

The FYVE domain, named as a acronym for the first four proteins in which it was recognized (340), is a type of zinc finger that binds to PtdIns(3)P (32, 117, 269). The signature of this domain is a defined spacing of cysteines that coordinate zinc and three additional blocks of residues that participate in lipid binding. Stenmark et al. (339) have identified 27 FYVE domain proteins in humans, 15 in C. elegans, and 5 in S. cerevisiae (339). Automated sequence analysis tools report larger numbers of FYVE domains in these organisms. EEA1, one of the four proteins in which the FYVE domain was first recognized, binds to early endosomes. The FYVE domain of EEA1 is necessary, but not sufficient, for this binding (35, 194). The COOH-terminal portion of EEA1 contains a coiled-coil domain and terminates in the FYVE domain. The crystal structure of this COOH-terminal fragment reveals a dimer held together through interactions between the two coiled coil domains and between one edge of each of the two FYVE domains. Together, the FYVE domains form a flat surface orthogonal to the coiled coil domain. Residues that contact the lipid head group are on the face of the protein opposite the coiled coil, and four hydrophobic residues near the phosphate binding residues form a loop that would dip down into the hydrophobic core of the bilayer (88). Thus, in EEA1, specificity of the head group-binding pocket is supplemented by nonspecific hydrophobic interactions as well as strengthened by being bivalent. A similar mode of binding has been reported for the FYVE domains of Vps27p and Hrs (335), and it is likely that FYVE domains, like PH and PX domains, use multiple contacts to achieve proper membrane location.

Proteins containing FYVE domains have been shown to function in endocytosis, in growth factor signaling, and in regulation of the actin cytoskeleton (62, 339, 396) (Table 5). Proteins with the first two functions are found on endosomes, and those of the last class, which also contain multiple PH domains, are found on the plasma membrane. The arrays of other domains found in proteins that contain FYVE domains suggest that many of them will assemble into multiprotein complexes. Table 5 lists proteins known or thought to function in membrane traffic that contain FYVE domains.

The ENTH (epsin 1 NH₂-terminal homology) domain was originally recognized as an NH₂-terminal sequence in epsin 1 and several other proteins that had known or suspected roles in endocytosis (41, 178). This region of one of these proteins, AP180, had previously been shown to bind to inositol hexakisphosphate (403), and the ENTH domain was found to be a phosphatidylinositol binding module preferring PtdIns(4,5)P₂ (104, 161). However, structural data showed that proteins related to AP180 and CALM have a NH₂-terminal domain with a distinct fold (ANTH domain). The ANTH domain binds the phosphoinositide head group via solvent-exposed basic residues (104), whereas the ENTH domain binds the lipid in a pocket that contacts the head group as well as the attached glycerol (103). Binding of PtdIns(4,5)P₂ by the ENTH domain causes the NH₂-terminal helix of the domain to penetrate into the lipid bilayer, slowing membrane dissociation of the domain and inducing membrane curvature (103, 336). Proteins that contain ENTH/ANTH domains also contain multiple recognition motifs for other types of protein-protein interaction modules and probably function as scaffold proteins that assemble protein complexes on membranes (73).

The ANTH domain proteins AP180 and CALM bind to clathrin and AP2 adaptors and are proposed to nucleate the polymerization of the clathrin coat (172). Epsin1 binds to the AP2 - and -ear domains, and another family member, EpsinR, binds to the AP1 -ear domain and to GGA1–3 at the TGN (144, 173, 235, 373). EpsinR contains an acidic phenylalanine motif found in two yeast proteins, Ent3p and Ent5p, that is necessary for them to bind to AP1 and Gga2p (89, 90, 235). EpsinR shows preference for binding to PtdIns(4)P and PtdIns(5)P in vitro (144, 235), which is consistent with the finding that PtdIns(4)P is abundant on Golgi membranes (372). Epsin and EpsinR contain a ubiquitin binding domain just COOH terminal to the ENTH domain and are themselves monoubiquitylated (146, 182, 259, 276). Ubiquitylation of hormone receptors (202, 210), as well as certain other membrane proteins (108), is necessary for sorting them in early endosomes from the recycling to the degradative pathway. It is less clear what role ubiquitylation would have for a protein that functions in TGN to endosome sorting, such as EpsinR (144, 235). In yeast, monoubiquitylation also serves as a signal for internalization from the plasma membrane (353). Internalization of the mammalian growth hormone receptor apparently requires that it be recognized by ubiquitylation machinery, but the actual ligation with ubiquitin is not necessary for internalization (300). The proteins that recognize ubiquitylated receptors on endosomal membranes, such as Hrs, are themselves monoubiquitylated (276, 280, 317). Although detailed knowledge of the precise interactions is lacking, it appears that the ENTH domain of the epsins links a membrane recognition event regulated by PtdIns(4,5)P₂ and perhaps PtdIns(4)P to another membrane recognition event regulated by PtdIns(3)P through the FYVE domain of Hrs.

Two other proteins that contain ENTH domains, HIP1 and HIP1R, and their counterpart in S. cerevisiae, Sla2p, interact with the actin cytoskeleton as well as play a role in endocytosis (95, 232, 238, 369, 383, 401). Phosphatidylinositides play a major role in regulating actin (404), and the ENTH domains of HIP1 and HIP1R may function to coordinate the assembly of endocytic coat proteins with changes in the actin cytoskeleton. A list of human proteins containing ENTH domains and thought to function in membrane traffic is presented with their yeast orthologs in Table 6.

In addition to the modular phosphoinositide binding motifs, a number of binding sites have been identified that have in common clusters of basic residues interspersed with hydrophobic residues and little other primary sequence conservation. These have been found in many cytoskeletal proteins (reviewed in Refs. 218, 404), clathrin adaptor proteins (55, 111, 132), C2 domains (38, 60), including in synaptotagmin (310) and in PLD (314, 410). PLD₁ and PLD₂ contain both PX and PH domains, but the site that binds PtdIns(4,5)P₂ that regulates enzyme activity is a conserved sequence of the basic/hydrophobic type found in PLD₁, PLD₂, yeast Spo14, and less conserved in plant PLDs. The lesson from this example is that the observation that a protein is regulated by a phosphoinositide and contains a known phosphoinositide-binding module does not necessarily lead to the conclusion that the module regulates protein activity. The number of examples of proteins with multiple phosphoinositide binding sites is increasing rapidly.

PHD domains are orphan zinc-finger domains found in a large number of nuclear proteins (1). ING2, a protein that associates with histone acetyltransferase and histone deacetylase complexes (99), was identified as a protein that bound to a phosphoinositide resin (121). ING2 bound preferentially to PtdIns(5)P and PtdIns(3)P through its PHD domain, as did several other proteins with PHD domains. The PHD domain is structurally similar to a FYVE domain (239, 268), and basic residues in the ING2 PHD domain predicted to be on a surface analogous to the PtdIns(3)P binding surface of the FYVE domain were found to be necessary for binding PtdIns(5)P (121). No protein known to be important for membrane traffic has been identified that has a PHD domain.

	IV. INTRACELLULAR DISTRIBUTION AND FUNCTION OF PHOSPHOINOSITIDES FOR MEMBRANE TRAFFIC

Top Previous Next References

 
It is likely that the PIPs important for constitutive membrane traffic are the more abundant ones (see Fig. 1). An exception to this may be PtdIns(3,5)P₂, which has been shown to be important for function of the vacuole in yeast, but about which very little quantitative data exist. Together, PtdIns and the PIPs are <10% of cellular lipids. Of this total, 5% is PtdIns(4)P (0.5% total lipids) and 5% is PtdIns(4,5)P₂ (282). Because all PIPs are confined to the inner leaflet, PtdIns(4)P and PtdIns(4,5)P₂ would each be 1% of inner leaflet lipids. Depending on the cell type, the plasma membrane can be from 2 to 10% of total membranes and the Golgi 7–12% (5, 122). Thus, if the major pool of PtdIns(4,5)P₂ is in the plasma membrane and PtdIns(4)P is on the Golgi (see below), these two lipids could be 5% or more of the cytoplasmic leaflet of the organelle where they are concentrated. Because these lipids are not uniformly distributed in the membranes that contain them, their local concentration is likely to be higher. Less than 0.25% of total inositol lipid is phosphorylated on D-3 (this equals 0.05% of inner leaflet lipid) (282). In murine fibroblasts, the ratio of PtdIns(3)P to PtdIns(3,5)P₂ is 4:1 (384), and in unstimulated neutrophils, the ratio of PtdIns(3)P₂ to PtdIns(3,4,5)P₃ and PtdIns(3,4)P₂ is 8:1 (341). Thus PtdIns(3)P is 0.04% of total membrane phospholipid. Endosomes and lysosomes are 2% of cellular membranes, and if most of the PtdIns(3)P is present in the endosomal pathway, then PtdIns(3)P in the cytoplasmic leaflet of endosomes is 4% of phospholipid and as abundant as PtdIns(4,5)P₂ at the plasma membrane or PtdIns(4)P at the Golgi.

View larger version (38K): [in this window] [in a new window]   FIG. 1. The major phosphoinositide species are concentrated at distinct sites in intracellular membrane traffic pathways and may serve as organelle markers. The major concentration of phosphatidylinositol 4-phosphate [PtdIns(4)P] (blue) is at the Golgi complex, and very little free PtdIns(4)P is detected at the plasma membrane or on endosomes. PtdIns(3)P (green) is concentrated on early endosomes. The majority of phosphatidylinositol 4,5-bisphosphate [PtdIns(4,5)P2] (red) is at the plasma membrane at steady state. PtdIns(3,5)P2 (orange) is found on multivesicular endosomes and lysosomes. Some phosphoinositides are found in the endoplasmic reticulum and in the nucleus, but probably do not play major roles in membrane traffic.

A. PtdIns(4,5)P₂ at the Plasma Membrane

Most of the PtdIns(4,5)P₂ in the cell is in the plasma membrane (11, 375). PtdIns(4,5)P₂ is enriched in detergent-resistant membranes (152), and PtdIns(4,5)P₂ that is turned over in response to hormone signaling may be enriched in caveolae or lipid raft membranes (273). However, recent studies by light or electron microscopy suggest that PtdIns(4,5)P₂ is not especially concentrated in caveolae at steady-state (364, 375). Thus the degree of subcompartmentalization of PtdIns(4,5)P₂ in the plasma membrane is currently uncertain. PtdIns(4,5)P₂ is generated during the process of fusion of regulated secretory vesicles or synaptic vesicles with the plasma membrane (66, 219), at the locations where actin rearrangements occur (59, 67, 404), and is required for clathrin-mediated endocytosis (67, 104, 161, 171, 262). Although required for vesicle fusion at the plasma membrane and for homotypic fusion of yeast vacuoles in vitro (222), the role of PtdIns(4,5)P₂ in vesicle fusion is not known precisely. In the vacuole fusion reaction, PtdIns(4,5)P₂ is required at the priming step where SNARE protein complexes are dissociated and also after the docking of the vesicle but prior to fusion (222). During dense-core vesicle secretion in mammalian cells, PtdIns(4,5)P₂ is detected mainly on the plasma membrane rather than on the vesicles (149, 219). In PC12 cells, the small GTP-binding protein Arf6 regulates the pool of PtdIns(4,5)P₂ required for dense-core vesicle fusion by activating a PIP5KI enzyme in a calcium-dependent manner (4). The calcium dependence may be in part a protein kinase C (PKC)-regulated dephosphorylation of the PIP5KI, which activates it (4, 266, 382). In the same cell type vesicle fusion is inhibited in cracked cells incubated with recombinant C2A domains from various synaptotagmins, and this inhibition correlates with the ability of the domain to bind to PtdIns(4,5)P₂ (359). Therefore, a synaptotagmin is likely to be one of the effectors for PtdIns(4,5)P₂ for membrane fusion at the plasma membrane, at least for regulated secretion.

In the regulation of actin and the formation of clathrin-coated pits, it is clear that PtdIns(4,5)P₂ increases the affinity of many mutually interacting proteins for membranes. As such, it may participate in defining the membrane location for the events that require these proteins. The AP2 adaptor, which binds membrane protein cargo to the clathrin lattice, contains two binding sites for PIPs. In the crystal structure of soluble AP2 core complex, these sites are orthogonal to each other and could not simultaneously bind to the membrane (55). In this structure, the binding pocket for internalization signals on receptors is occluded. Owen and colleagues (55) have proposed that binding to PIPs facilitates a conformational change that opens the signal-binding pocket and allows both PI binding sites to face the membrane. This change may be regulated by phosphorylation and the open conformation stabilized by binding to PIPs. Mutation of the PIP binding site on the AP2 subunit inhibits membrane binding, and mutation of the binding site on the AP2µ subunit prevents binding to cargo, demonstrating the importance of the interaction with PIPs (111, 291). AP2 can bind multiple PIPs, and in vitro those with D-3 phosphate increase the affinity of AP2 complexes for peptides that have internalization signals (286). This may be relevant to the observations that PtdIns3KIIC2 localizes in clathrin-coated pits (112) and that PtdIns(3,4,5)P₃ is important for recruitment of AP2 to activated -adrenergic receptors (248). However, given the relative scarcity of PtdIns(3,4,5)P₃, PtdIns(3,4)P₂, and PtdIns(3)P at the plasma membrane, it is likely that PtdIns(4,5)P₂ is the major regulator of AP2 for constitutive endocytosis. The less abundant PIPs may be generated as part of the signaling of hormone receptors and function locally to accelerate the internalization of those receptors. Expression of PH domains specific for PtdIns(4,5)P₂ inhibits both early and late stages of clathrin-coated vesicle formation (171), and increased or decreased expression of PIP5KIB causes more or less AP2 to bind to membranes (262). In addition to AP2, a number of other proteins important for endocytosis either have been shown to bind PIPs or are recruited to membranes in response to proteins that bind PIPs. The number of clathrin-coated pits at the plasma membrane and internalization of transferrin receptors increases or decreases in direct relation to the production of PtdIns(4,5)P₂ in several cell types (262). This shows that the rates of constitutive endocytosis not only require PIPs, but also might be regulated through the control of PIP5KI activity. In HeLa cells, PIP5KIB is responsible for most of the cellular PtdIns(4,5)P₂ with a minor contribution by PIP5KIA and no detectable contribution of PIP5KIC long or short isoforms (262). However, inhibition of PIPK5IA or PIP5KIC by small interfering RNA increases transcription of both of the remaining two isoforms, indicating that although isoforms A and C do not impact total cell PtdIns(4,5)P₂ levels very much, their presence is sensed by the cell and all three lipid kinases are coordinately regulated (262). PIP5KIA has been shown to bind to the EGF receptor and accelerate EGF receptor internalization (14). PIP5KIC is the major producer of PtdIns(4,5)P₂ in the synapse and probably responsible for the PtdIns(4,5)P₂ required for endocytosis of synaptic vesicle components (382). Thus different PIP5KI enzymes may potentially regulate endocytic rates of different types of cargo.

There is a correlation in both time and space between the formation of endocytic vesicles and the reorganization of the actin cytoskeleton (107, 214, 327, 380, 382). In S. cerevisiae, where clathrin is not required for endocytosis, actin plays a crucial role (245). In mammalian cells, clathrin, AP2, and dynamin bind to proteins that also bind to actin (154, 180, 199, 229, 244, 279, 390). The functional consequence of these two events is not understood at present. It is possible that cortical actin is a barrier to the budding of a clathrin coat and must be reorganized so that a coated vesicle can move away from the cell surface. However, rather than depolymerize, actin polymerizes as coated vesicles move away from the membrane (229). In an early study of the distribution of clathrin coats on membranes of primary human fibroblasts, Anderson et al. (6) observed that clathrin-coated pits that contained low-density lipoprotein (LDL) receptors lined up over stress fibers. A more recent study by light microscopy of GFP-clathrin in live cells confirms that clathrin coats appear to organize in relation to cortical actin (302). However, a caveat to these experiments is that clathrin forms two structures on plasma membranes, flat lattices and curved pits. It has been assumed that flat clathrin is the precursor to curved clathrin pits, but this has never been proven. Recently, clathrin has been shown to interact dynamically, binding and releasing from the clathrin lattice (394). Thus the binding and release of GFP-clathrin from membranes cannot necessarily be equated with the formation and fission of clathrin-coated pits.

Overexpression of PIP5KIB causes actin tails to form on vesicles that contain lipid rafts (297). These actin structures require activation of Arp2/3 and contain dynamin 2 (261, 297). Although the actin "comets" produced in cells overexpressing PIP5KIB are exaggerated structures, a similar polymerization of actin has been observed on endosomes in Xenopus eggs and in cultured mast cells (230, 349). The polymerization of actin appears to move the organelles on which it is bound, but it is not clear if this process is used to move vesicles in cells or actin polymerization plays some other role.

Experiments in which the 5-phosphatase synaptojanin is deleted prove that there must be turnover of the PtdIns(4,5)P₂ on clathrin-coated vesicles for normal rates of vesicle uncoating to occur. As mentioned previously, synaptojanins are dual function phosphatases capable of converting PtdIns(4,5)P₂ to PtdIns without producing PtdIns(4)P. In fact, at steady-state, a GFP-oxysterol binding protein PH domain probe or anti-PtdIns(4)P antibody does not label the plasma membrane, suggesting that very little free PtdIns(4)P₂ is present there (372). Because PtdIns(4)P is the precursor to PtdIns(4,5)P₂, either it is generated from PtdIns by a plasma membrane PtdIns4K, such as PtdIns4KII (377), directly at the site where it is converted to PtdIns(4,5)P₂, or it might be generated at the Golgi but be rapidly converted to PtdIns(4,5)P₂ upon arrival at the plasma membrane.