I. INTRODUCTION

Glycerophospholipids (hereafter termed phospholipids) are essential components of lipoproteins, surfactants, and cell membranes. They are generally composed of two fatty acids linked via ester bonds to a glycerol backbone containing a polar head-group. Because they are amphiphilic, most phospholipids can spontaneously form bilayers in aqueous environments; such lipid bilayers are major constituents of all cellular membranes. Phospholipid species are categorized according to the nature of their head-group and the length and saturation of their acyl chains. The particular structural features of the constituent membrane phospholipids dictate the viscosity, curvature, and electrostatic charge of the bilayer.

Beyond their role as building blocks of lipid bilayers, phospholipids influence cellular behavior in a variety of ways. They are not just passive structural spectators; phospholipids are also active regulators of key physiological events such as exocytosis, chemotaxis, and cytokinesis. Their comparatively simple chemical structure belies a sophisticated ability to transmit signals. Indeed, the transbilayer distribution of phospholipids, and their ability to undergo hydrolysis or head-group modification encodes a molecular language that guides the recruitment and function of multiple proteins, including transmembrane proteins. Moreover, by forming boundaries and barriers, phospholipid membranes act as the interface separating cells from their surrounding milieu, and organelles from the cytosol, enabling each compartment to maintain a unique biochemical signature. The establishment and maintenance of chemical and electrical gradients across biological membranes are critical for cell survival.

Phospholipids confer plasticity to biological membranes. A prototypical example is provided by the plasma membrane, which continuously undergoes invagination and scission, generating intracellular vesicles or vacuoles. This process is known globally as endocytosis, a hypernym that encompasses several disparate internalization processes, including clathrin-mediated endocytosis, macropinocytosis, and phagocytosis. While the molecular machinery and internalized cargo differ rather drastically between these processes, the resulting endosomes, macropinosomes, and phagosomes are all segregated from the cytosol by a lipid bilayer. Moreover, acute alterations in the phospholipid composition of the plasma membrane accompany and are seemingly required for the scission and subsequent remodeling of the endomembrane compartments as they undergo maturation.

Over the past decade, the emergence of novel molecular tools and imaging modalities has allowed physiologists to study the intricacies of phospholipid signaling during endocytosis.

Visualization of phospholipids in cells is now possible owing to the development of novel fluorescently labeled lipid analogs and phospholipid-specific antibodies. Of particular importance was the design and implementation of genetically encoded probes to detect defined lipid species by fluorescence in intact cells. This approach is based on the generation of chimeric constructs consisting of a fluorescent protein attached to a protein domain that binds selectively to the head-group of a lipid of choice. Thus a construct encompassing the PH domain of phospholipase C (PLC)δ has been used extensively to detect phosphatidylinositol 4,5-bisphosphate [PtdIns(4,5)P₂], while constructs encoding the PX domain of p40^phox have been instrumental in studying phosphatidylinositol 3-phosphate [PtdIns(3)P]. These advances, together with improved microscopic and analytical algorithms, have made it possible to monitor phospholipid dynamics in live cells with subdiffraction resolution. Investigation of the biosynthesis and catabolism of phospholipids, as well as their functional properties, has been facilitated by the introduction of RNA interference, knockout organisms, potent and increasingly specific pharmacological inhibitors, and conditional variants of proteins that are either light, heat, or chemically sensitive.

As a testament to their biological significance, the role of phospholipids in internalization and maturation has been remarkably conserved during evolution among different phyla. In all eukaryotic systems, phospholipids act as membrane identifiers, directing the traffic of individual compartments. Through stereochemical and electrostatic interactions, phospholipids affect the recruitment and activity of proteins that control key tubulation, fission, and fusion events. Some phospholipids also serve as enzymatic substrates, triggering signaling cascades. But phospholipids have protein-independent functions as well. The physical structure of the lipids, whether cylindrical, conical, or inversely conical, can affect the intrinsic curvature of membranes, altering their geometry and fusogenicity. From the physiological vantage point, phospholipids regulate endocytic processes that are important for growth, immune surveillance, synaptic transmission, cell migration, and many other functions. Dysregulation of phospholipid metabolism during endocytosis is associated with chronic infections, metastatic tumors, and neurological disorders. Diseases associated with dysfunction of phospholipid signaling and endocytosis can be grouped into three categories, depending on whether they affect phospholipid synthesis, distribution, or catabolism. An example of the first category is Charcot-Marie-Tooth peripheral neuropathy type 4J, which is associated with reduced PtdIns(3,5)P₂ levels, likely due to decreased synthesis of the phosphoinositide (70). Scott syndrome, a bleeding disorder, is an example of the second category; it is associated with mislocalization of phosphatidylserine (PtdSer), which in the patients fails to redistribute(scramble) to the outer membrane surface of platelets during coagulation (412). Finally, abnormal lipid catabolism is associated with numerous cancers where phosphatidylinositol 3,4,5-trisphosphate [PtdIns(3,4,5)P₃] concentration is elevated due to its ineffective degradation by the tumor suppressor, phosphatase and tensin homolog (54, 408). Clearly, phospholipids are more than just stagehands during endocytic processes; they are lead actors who deserve their share of the (microscope) spotlight.

This review focuses on the classes of phospholipids that are well documented as being implicated in endocytosis: the glycerophospholipids–phosphatidylcholine (PtdCho), phosphatidylethanolamine (PtdEtn), phosphatidylserine (PtdSer), phosphatidylinositol (PtdIns), the polyphosphoinositides, and phosphatidic acid (PtdOH)–and the unconventional phospholipid bis(monoacylglycero)phosphate (BMP), also known as lysobisphosphatidic acid (LBPA). In most biological membranes phospholipids coexist and function in conjunction with cholesterol and sphingolipids. Together with the glycerophospholipids, cholesterol and sphingolipids play essential roles in some forms of endocytosis, especially those involving caveolae and the clathrin-independent carrier/glycosylphosphatidylinositol (GPtdIns)-anchor-enriched endocytic compartment-mediated (CLIC/GEEC) pathway. These specialized pathways have been reviewed recently (92, 166, 319). Readers interested in the unique features and roles of cholesterol and sphingolipids in membrane traffic are referred to References 43, 173, 218, 219.

II. PHOSPHOLIPID GENERATION AND DISPOSITION: A PRIMER

The relative abundance of specific phospholipid classes varies among organelles, across leaflets of a specific bilayer, and even in the lateral plane within a single leaflet. Maintaining these distinct and discrete distributions requires a precise balance of phospholipid synthesis, remodeling, transport, and degradation. Before detailing the specific mechanisms that underpin the regulation of each phospholipid class, we will discuss global features common to all of the classes, with the notable exception of BMP/LBPA, the unconventional phospholipid. The following section briefly summarizes these features and discusses how they influence phospholipid distribution and transmembrane asymmetry. Where pertinent, the role of phospholipids in the pathogenesis of selected diseases is also discussed.

A. Biosynthesis

Excluding the polyphosphoinositides, which can be generated on the cytoplasmic leaflet of various organelles, the bulk of phospholipid synthesis occurs only on the cytoplasmic leaflet of the endoplasmic reticulum (ER) and to a lesser extent in the Golgi complex or, in the case of PtdEtn, inside mitochondria. Most phospholipids contain a saturated fatty acyl chain that has an ester linkage to the sn-1 position of a glycerol moiety, and a polyunsaturated fatty acyl chain attached via an ester linkage to the sn-2 position (369). This combination of two fatty acids and a glycerol backbone creates the diacylglycerol (DAG) moiety. A phospholipid consists of a DAG moiety covalently attached to a phosphate-containing head-group; the type of phosphorylated head-group linked to DAG defines the class of phospholipid.

PtdOH, consisting of a DAG molecule phosphorylated on the sn-3 position, is the simplest phospholipid and the precursor of all other classes (FIGURE 1). Contrary to intuition, however, phosphorylation of DAG is not the primary pathway for PtdOH biosynthesis. In fact, PtdOH is used to produce the bulk of DAG. During phospholipid synthesis, PtdOH is formed by the sequential acylation of glycerol 3-phosphate. Once formed, it is either dephosphorylated into DAG or converted into cytidine diphosphate (CDP)-DAG; both of these resulting species can undergo further linkage to an ionic head group, forming either zwitterionic or anionic lipids (372). Because all phospholipids use PtdOH as a precursor, the regulation of their biosynthesis is inevitably interconnected (158). Nevertheless, knocking out the enzymes responsible for generation of specific phospholipid classes often causes embryonic lethality in mice (19, 368, 394). The interplay between synthesis and interconversion likely serves to fine-tune the individual phospholipid concentrations in organelles.

B. Catabolism

While the majority of phospholipids are synthesized in the secretory pathway and in the case of PtdEtn in mitochondria, their relative concentrations vary greatly in different organelles, including those where synthesis does not occur (367b). This implies that other mechanisms, such as differential degradation, segregation, and selective transport must contribute to the establishment of steady-state concentrations of the lipids in individual compartments. Indeed, a variety of phospholipid catabolic pathways have been identified, and several of these subserve specific signaling pathways; these are often unique to particular classes and will be discussed below in the context of the individual phospholipid classes. In addition, however, there are some generic catabolic pathways. Essentially all glycerophospholipids are subject to continuous remodeling by phospholipase A (PLA) isoforms, which release the fatty acid linked to the sn-1 or sn-2 position of the glycerol moiety. PLA1 removes the 1-acyl chain and PLA2 the 2-acyl chain from the phospholipid backbone. Phospholipase B enzymes release both fatty acids, while diacylglycerol lipases exert an equivalent effect on DAG (266). PLA2 isoforms can be Ca²⁺ dependent or independent. Ca²⁺-sensitive forms such as group IVA cPLA2 show specificity for phospholipids containing arachidonic acid on position sn-2 and are regulated by ion fluxes and often also by kinases including mitogen-activated protein kinase (MAPK) (89). Group IVA cPLA2 is specifically activated by PtdIns(4,5)P₂ and hydrolyzes PtdCho (267), PtdIns, and PtdEtn (89). The arachidonate that is liberated can enter the eicosanoid signaling pathway, acting as a precursor for many inflammatory molecules and thrombogenic mediators that feature prominently in pathological states such as rheumatoid arthritis and atherosclerosis (331). There are also intracellular Ca²⁺-insensitive isoforms like iPLA2 (4); rather than signaling, these enzymes are thought to be important in homeostatic lipid remodeling and degradation (27, 28).

Together with free arachidonate, PLA activity also forms lysophospholipid species. Unlike their diacyl precursors, lysophospholipids are quite water soluble (240, 251) and can undergo spontaneous cytosolic translocation to other organelles. Lysophospholipids can be further deacylated by lysophospholipases for degradation, or reacylated by lysophospholipid acyltransferases for remodeling. Under certain physiological or pathophysiological conditions, lysophospholipids can themselves play a role in signaling. During inflammation, for example, lysophosphatidylcholine is used to synthesize platelet-activating factor, a potent signaling messenger involved in hemostasis (297).

Phospholipase localization is not restricted to the cytosol. Several phospholipase types are found in lysosomes, where they break down lipids delivered by endocytosis or by macroautophagy. These lipases include a lysosomal PLA₂ isoform (160) and also uncharacterized phospholipases of the C, A₁, and possibly D types (122, 179, 243). How the products of lysosomal lipid hydrolysis are handled and recycled to other cellular membranes is incompletely understood (197). Although the lysosomal processing of glycerophospholipids is not causally associated with any known lysosomal storage disorders, other lipids, such as sphingolipids, accumulate in lysosomes when their clearance is deranged and their buildup contributes to such diseases as Gaucher's disease or Niemann-Pick disease. In these cases, the glycerophospholipids may indirectly influence pathogenesis by regulating the trafficking of lysosomal enzymes that in turn affect the phospholipids (181).

C. Transport

Several mechanisms of interorganellar transport of lipids have been postulated. Although their relative contributions have not been defined, the transport processes can be broadly separated into vesicular and nonvesicular subtypes. Vesicular traffic of phospholipids is mediated by bulk membrane transport. For vesicular traffic to alter the relative distribution of phospholipids between compartments, vectorial sorting and concentrating events must occur, allowing for selective lipid delivery or removal. Such class-specific enrichment or depletion may result from intrinsic properties of the lipid, such as microdomain formation, or from lipid binding by extrinsic factors, such as coat proteins. While vesicular lipid transport is unquestionably important, lipid redistribution nevertheless occurs under conditions where secretory membrane traffic is blocked by brefeldin A (371), when the cytoskeleton is disrupted and even following ATP depletion (210), suggesting that membrane bulk flow is not absolutely necessary to mobilize lipids and establish gradients. Such nonvesicular transport complements the redistribution of phospholipids caused by membrane fission and fusion events.

Hypothetically, nonvesicular transport can be mediated by lipid-transfer proteins, by spontaneous diffusion, or can occur at membrane contact points. The hydration shell surrounding lipid membranes makes it unfavorable for individual phospholipid molecules to desorb from membranes. The process is facilitated by lipid-transfer proteins (LTP) that can shield the hydrophobic moieties of lipids while en route from donor to acceptor membrane. Phospholipid-transfer proteins are divided into three broad categories: PtdCho-transfer proteins, PtdIns-transfer proteins, and the nonspecific LTPs, also inaccurately called sterol carrier protein 2 (393). PtdIns-transfer proteins are also called PtdIns/PtdCho-transfer proteins because they can translocate PtdCho as well as PtdIns. While many diverse functions have been attributed to LTPs, it is unclear whether they contribute to the establishment of the distinct composition of different organelles (29). LTPs can in principle act as exchangers, swapping one phospholipid species for another. As such, these passive exchangers would be incapable of generating concentration differences, unless driven by a preexisting gradient or by coupling to metabolic reactions. Such is the case for PtdIns, which is delivered by LTPs to the Golgi complex, where it is continuously phosphorylated into PtdIns(4)P. The ongoing consumption of PtdIns provides a gradient for the net unidirectional delivery of PtdIns to the Golgi by passive exchange (74). This exchange model has yet to be verified in living cells because it is difficult to track the continuous, bidirectional traffic of proteins such as LTPs in the cytosol. Of note, there is evidence that LTPs have exchange-independent functions in vivo. These proteins seemingly act also as sensors of lipid concentration and thereby regulate metabolism, especially lipid biosynthesis. For instance, the yeast PtdIns-transfer protein Sec14 (339) and one of its homologs in mammals, NIR2 (220), regulate the levels of DAG in the Golgi. When these PtdIns-transfer proteins sense that PtdCho levels are elevated, they inhibit the enzyme that produces CDP-choline. Because CDP-choline and DAG are used to form PtdCho via a reversible reaction (366), depletion of CDP-choline increases the ratio of DAG to PtdCho.

Spontaneous diffusion provides another means for lipid transport. Bona fide (i.e., diacylated) phospholipids take hours to days to passively diffuse across an aqueous environment (180), unless ferried by transfer proteins. In contrast, lysophospholipids have been shown to effectively enter and traverse aqueous boundaries both in vitro and in vivo (161, 251, 263). Unshielded (naked) lysophospholipids can detach from bilayers spontaneously, but their solubilization and diffusion can also be aided by fatty acid-binding proteins, which can act as lysophospholipid carriers in the cytosol (75, 376). In view of the increased solubility of lysophospholipids, it is conceivable that ongoing, reversible deacylation of phospholipids may aid in their relocation. Active remodeling of phospholipids by PLAs and reacylation by lysophospholipid acyltransferases (the Lands cycle) may effectively accelerate their diffusive translocation (205, 332). These reactions, however, occur mostly in the ER and mitochondria, and PLA activity seems to be more important for phospholipid remodeling and signaling rather than transport. Indeed, the water solubility of lysophospholipids themselves may be more relevant to signaling than to net lipid transport (247).

The final mode of nonvesicular phospholipid transport is transfer through contact points (collision exchange). This occurs between membranes that, while not fused together, are closely apposed (∼10–30 nm apart) and often connected by bridging proteins (214, 362). Such areas of high phospholipid density that facilitate nonvesicular lipid transport usually form between the profusely-ramified ER, which is involved in producing the bulk of phospholipids, and a second, neighboring organelle. Mitochondria-associated membranes have been the best characterized (153). Mitochondria-associated membranes are proposed to act as contact sites for the exchange of lipids, most pertinently PtdSer and PtdEtn, into and out of mitochondria. Several molecular tethers have been discovered that join the ER and mitochondrion together. These include the mouse protein mitofusin2 (44), the yeast complex ERMES (200), and the Alzheimer's disease-associated protein presenilin 2 (402). In mammalian cells, the interorganellar translocation of lipids is generally ATP dependent, but there is evidence of spontaneous transfer as well, and the exact mechanism of transfer is still debated (362). Tight packing may promote out-of-plane protrusions that make desorption of lipid molecules more favorable. Alternatively, integral membrane proteins or LTPs may transfer the phospholipids. Lastly, lipid delivery across contact points may involve deacylation and spontaneous diffusion of lysophospholipids, as described above for cytosolic translocation. Because contact points are associated with the ER, a rich and dynamic calcium depot, it is tempting to speculate that calcium-dependent lipid remodeling via enzymes like cPLA or PLC may be involved in the translocation process. In this regard, formation of the STIM1-Orai1 complex that transports calcium ions into the cell occurs at plasmalemma-ER contact sites similar to those reported to mediate phospholipid exchange (57, 384). Furthermore, in Drosophila, a transmembrane PtdIns-transfer protein resides at plasmalemma-ER contact sites and regulates PLC activity (364). Similarly, in yeast, several Osh proteins regulate phosphoinositide metabolism, and possibly also translocation (90a), at plasma membrane-ER contact points (347).

It should be apparent from the preceding considerations that, while theories and speculation abound, the mechanisms of lipid transport are not yet well established. Further work is acutely needed to clarify which modes of translocation are involved in the context of specific organelles and species of phospholipids or lysophospholipids, and how these mechanisms ultimately regulate the subcellular distribution of phospholipids.

III. SPECIFIC PHOSPHOLIPID CLASSES AND THEIR ORGANELLAR DISTRIBUTION

The different classes of phospholipid are quite well-conserved across taxa, but several exceptions need to be highlighted. BMP/LBPA is almost exclusively found in mammalian cells; prokaryotes, yeast, nematodes, insects, and plant cells do not contain appreciable levels of this phospholipid. PtdOH, PtdEtn, PtdSer, and PtdCho, on the other hand, are ubiquitous, and their biosynthetic enzymes are found in a variety of organisms, including bacteria (133); nevertheless, these enzymes may be structurally different and catalyze distinct mechanisms. In addition, as highlighted below, the distribution and abundance of these phospholipids can vary drastically across species. PtdIns is found in only a few bacteria, and the polyphosphoinositides are virtually absent from bacteria, except under rare conditions (264). Budding yeast contain several polyphosphoinositides, but PtdIns(3,4,5)P₃ is usually absent (258) and, likewise, it is absent in plant cells. Otherwise, in Dictyostelium discoideum and animal models, such as nematodes, insects, and mammals, the polyphosphoinositides are well-conserved, as would be expected for a crucial signaling pathway essential for endocytosis and other key functions. With the growing application of lipidomics, specifically the LIPID MAPS initiative (351), we will have a clearer picture of each organism's unique phospholipid makeup. For now, we describe the distribution of phospholipid classes in the context of a prototypical mammalian cell and point out interesting phylogenetic differences as necessary.

A. Phosphatidic Acid

Phosphatidic acid (PtdOH), or phosphatidate (its dissociated form prevalent at physiological pH), is a quantitatively minor phospholipid, constituting 1–2% of the total lipids in cells (375). However, it exerts disproportionately large functional effects. Its inverted conical shape can induce membrane curvature (252); in the case of the plasmalemma, accumulation of PtdOH in the cytoplasmic leaflet would promote concave (negative) curvature (129). Moreover, because it undergoes deprotonation at neutral pH, PtdOH is anionic. Since one of its two dissociable groups has a pK_a between 6.5 and 8.0 (53, 262), PtdOH exists as a mixture of the mono- and dianionic species at physiological pH. Its distinctive shape and charge confer onto PtdOH unique functional properties, discussed in more detail below.

1. Biosynthesis and distribution of PtdOH

Its head-group consisting of a single phosphate group, PtdOH is the simplest phospholipid and the precursor of all glycerophospholipids and triglycerides. It is therefore a significant branch-point in lipid metabolism. As a result, mutations in genes necessary for its formation cause severe lipodystrophies (5). For example, mutations in AGPAT2, a gene that catalyzes the production of PtdOH, underlie type 1 congenital generalized lipodystrophy.

PtdOH formation serves two distinct functions: a biosynthetic role and a signaling role. PtdOH is made de novo by sequential, dual acylation of gylcerol-3-phosphate (G3P) using acyl-CoA as the fatty acid donor [G3P → lysophosphatidic acid (LPA or LysoPtdOH) → PtdOH; see FIGURE 1]. This acyl-transferase activity is detected predominantly at the ER. Thus PtdOH is thought to be generated de novo mostly at the ER and to a lesser extent on the mitochondrial outer membrane.

PtdOH can also be formed by metabolic conversion of preexisting lipids during signaling processes. Two such pathways are well established: first, PtdOH can be made from the phosphorylation of DAG by DAG-kinases (DGKs). Up to 10 mammalian isoforms of DGK have been identified (317). These enzymes have diverse subcellular locations and regulatory mechanisms (333). Although DGK activity is often disparaged as a mechanism that merely antagonizes DAG signaling, the consequent formation of PtdOH has direct signaling functions as well. The second pathway involves the cleavage of phospholipids, predominantly PtdCho, into PtdOH by phospholipase D (PLD) (290). Two isoforms of PLD have been identified: PLD1, which is found mainly at the Golgi and perinuclear vesicles but can be induced to translocate to the plasmalemma (47), and PLD2 that primarily localizes to the plasmalemma and endocytic vesicles (79, 95). Both the DGK and PLD pathways are important contributors to PtdOH formation and signaling.

2. Catabolism of PtdOH

Conversion of PtdOH underlies the synthesis of all other glycerophospholipids. After its formation at the ER, PtdOH can be converted into either DAG or CDP-DAG that is used for the synthesis of PtdCho, PtdEtn, PtdSer, and triacylglycerol. To generate DAG, PtdOH is dephosphorylated by phosphatidate phosphohydrolase (PAP).

A second metabolic pathway involves conversion of PtdOH by CDP-DAG synthase (CDS). This reaction occurs at the ER, where the resulting CDP-DAG is used for PtdIns synthesis. Two CDS isoforms (CDS1 and CDS2) have been discovered (146, 174). They are both transmembrane proteins associated with the ER. Mitochondrial CDS activity has also been detected (105, 370), but the responsible enzyme has not yet been identified.

Outside of these metabolic contributions, PtdOH may be deacylated by PLA into LPA. Extracellular LPA is a potent mitogen and survival factor. Although intracellular LPA has been shown to bind PPARγ, it is unclear whether this charged lysophospholipid can transverse into the extracellular fluid, where LPA mediates the bulk of its effects (68). Furthermore, the PtdOH-specific PLA1 acts on extracellular PtdOH (15), which puts into doubt whether LPA is a significant intracellular signaling molecule.

3. Functional role of PtdOH

The distinct shape and charge of PtdOH confers special functional properties. As discussed above, PtdOH can induce membrane curvature due to its unique shape. Conversion of cylindrical lipids, like PtdCho, into PtdOH, which is conical, would dynamically change the conformation of the membrane (198, 199). Although this model is intuitively appealing, it is still hypothetical. Because it constitutes a relatively minor fraction of the total lipids, and because membranes are disordered and relatively fluid, it is not clear that PtdOH can significantly alter curvature, unless confined to a small area of the membrane (268).

In view of its scarcity, it is more likely that PtdOH exerts its biological effects via interactions with proteins, possibly due to its anionic characteristics. PtdOH has been shown to directly bind the polybasic region of Rac (60), a Rho GTPase that contributes to membrane ruffling and chemotaxis. Moreover, PtdOH interacts with DOCK2 (278), a Rac GTP-exchange factor, and helps Rac detach from its guanine nucleotide-dissociation inhibitor either directly or indirectly via PAK1 (2) and also via atypical isoforms of PKC (66). Several other proteins have been shown to interact with and to be regulated by PtdOH, including mTOR (111, 121), PtdInsP5K (265), and sphingosine kinase 1 (88). It is both remarkable and discouraging that, other than polycationic motifs, no conserved PtdOH-binding domains have been identified among these proteins.

Because its more alkaline pK_a is in the neutral range, it has been speculated that PtdOH may serve as a sensor of pH, by altering its protonation and hence its charge (400). It must be borne in mind, however, that the cytosol rarely experiences large excursions in pH, a testament to the effectiveness of pH-regulatory systems that include a number of ion exchangers, co-transporters and pumps (59).

B. Phosphatidylcholine

With the notable exception of Drosophila, where PtdEtn is most prevalent (373), phosphatidylcholine (PtdCho) is usually the most abundant phospholipid in eukaryotic cells, comprising 45–55% of the total lipids (375). PtdCho is a zwitterion, as it contains a quaternary amine in addition to the phosphate moiety. It has a roughly cylindrical shape but the presence of an unsaturated acyl chain on position sn-2 can cause some distortion of its shape. Often, PtdCho is disregarded as a strictly structural component of biological membranes. Nonetheless, it can also function in signal transduction, serving as a substrate of phospholipases that generate signaling messengers. Outside of the cell, PtdCho is an important constituent of bile and lung surfactant, and its dietary intake may impact atherosclerosis and inflammation (389).

1. Biosynthesis and distribution of PtdCho

The ubiquitous pathway of PtdCho biosynthesis utilizes choline phosphotransferases to couple CDP-choline with DAG, releasing CMP (FIGURE 1). In hepatocytes, where it is essential for bile synthesis, PtdCho is also made by an alternative pathway, involving three sequential methylations of PtdEtn by one or more phosphatidylethanolamine N-methyltransferases (84, 369).

PtdCho is found in most cellular membranes. Although it does not discernibly translocate across membranes, PtdCho is somewhat enriched in the outer leaflet of the plasmalemma, compared with the inner leaflet (91). This contrasts with the aminophospholipids PtdSer and PtdEtn that are greatly enriched in the inner leaflet. Indeed, PtdCho enrichment in the outer leaflet may be the result of selective aminophospholipid depletion by the flippases (see below).

2. Catabolism and functional role of PtdCho

PtdCho is a precursor of various signaling molecules. It is cleaved to make DAG, lysophosphatidylcholine, PtdOH, and arachidonic acid. PLD isoforms utilize PtdCho to make PtdOH and choline. This reaction accounts for the intracellular generation of PtdOH by PLD1 and PLD2. In addition, some forms of PLD are secreted, and the PtdOH generated extracellularly may be further processed to produce LPA/lysoPtdOH, a potent mitogen (14). Although phosphoinositides, specifically PtdIns(4,5)P₂, are the major substrate of PLC, an uncharacterized isoform of this enzyme displays activity towards PtdCho, yielding DAG and phosphocholine (330). As discussed in more detail below, DAG is an important signaling molecule that recruits C1 domain-containing proteins, including several protein kinase C (PKC) isoforms. Additionally, PtdCho is hydrolyzed by PLA2 to produce arachidonic acid and lysophosphatidylcholine. As mentioned, arachidonic acid is the precursor of prostaglandins and leukotrienes, while lysophosphatidylcholine is the precursor of platelet-activating factor. Moreover, PtdCho can be hydrolyzed by phospholipase B enzymes, such as neuropathy target esterase, producing one glycerolphosphocholine molecule and two free fatty acids, which are likely important for membrane fluidity, especially in neurons (113).

In addition to these signaling molecules, PtdCho can also be converted into other polar lipids. PtdCho participates in the biosynthesis of sphingomyelin. This process, which is catalyzed by sphingomyelin synthase in the luminal leaflet of the Golgi and exoplasmic leaflet of the plasma membrane (169), entails the transfer of the phosphocholine head-group of PtdCho to ceramide. Finally, PtdSer synthase 1 uses PtdCho as a substrate to generate PtdSer in the ER (373).

Beyond its role as a substrate for the production of important signaling lipids, little is known about the function of PtdCho itself. Few effectors bind specifically to PtdCho or respond to PtdCho levels. Annexin V has been shown to associate with PtdCho, and this interaction may serve an anti-inflammatory role by blocking the metabolism of PtdCho into arachidonic acid and lysophosphatidylcholine (50, 234). Also, as mentioned previously, PtdCho can bind numerous LTPs and can be exchanged for PtdIns. Thus PtdCho can be regarded as structural grout that sustains membrane integrity by filling in defects created when other phospholipids are removed from, or flipped across, a membrane.

C. Phosphatidylethanolamine

Like PtdCho, phosphatidylethanolamine (PtdEtn) functions mainly as a structural component of membranes. It is relatively abundant, constituting 15–25% of the phospholipids (375), and it is very prominent in the mitochondrial inner membrane (273). Since its head-group consists of a primary amine in addition to the phosphate moiety, PtdEtn is zwitterionic. Because its head-group is comparatively small, PtdEtn adopts a moderately conical shape and, like PtdOH, it induces negative curvature (concavity) and has a tendency to form nonlamellar structures (168, 221).

1. Biosynthesis and distribution of PtdEtn

PtdEtn is produced by two alternative pathways: the CDP-ethanolamine pathway and the PtdSer decarboxylation pathway (FIGURE 1). These pathways are nonoverlapping and cannot compensate for each other (127, 346). In cell cultures, PtdSer decarboxylation is the predominant pathway (382), but this could be an artifact caused by the availability of certain substrates in culture conditions, coupled to the differential acyl chain preference between the two synthetic pathways (231, 324).

PtdSer decarboxylation occurs in mitochondria (374). The mitochondrial inner membrane protein, PtdSer decarboxylase (PSD), converts PtdSer into PtdEtn (325). It is a peripheral membrane protein with catalytic activity manifested on the outer aspect of the inner mitochondrial membrane (291, 403). Because PtdSer is made at mitochondria-associated membranes (MAMs), it needs to travel from the ER to reach the mitochondrial decarboxylase. This translocation step, which is poorly understood, limits the rate of PtdEtn formation by this pathway (383).

Although PtdEtn is found in most subcellular compartments, it is enriched in mitochondria presumably because of the mitochondrial location of the PtdSer decarboxylation pathway that generates it (374). PtdEtn is also enriched in the inner leaflet of the plasma membrane compared with the outer leaflet, due to its active translocation across the membrane by aminophospholipid flippases (25, 367a). Despite the ongoing translocation, a measurable amount of PtdEtn is exposed on the outer surface of cells, unlike PtdSer which is virtually undetectable (91).

2. Catabolism and functional role of PtdEtn

PtdEtn is catabolized by a diversity of reactions that provide substrates for other metabolic pathways, rather than generating signals. PtdSer synthase 2 (PSS2) uses PtdEtn to make PtdSer at the ER (202). Although this enzyme is reversible (374), the reverse reaction is probably not a major contributor to PtdEtn synthesis and only acts as a safety valve that limits PtdSer production. PtdEtn also provides the ethanolamine moiety for GPtdIns modification of proteins (233).

The most important (or at least the most publicized) functional role of PtdEtn is in macroautophagy, where it helps target soluble adaptors like p62 and NDP52, to the phagophore membrane, ultimately yielding the autophagosome. This occurs by conjugation of Atg8 (LC3-I) to PtdEtn, thereby forming PtdEtn-LC3 (LC3-II) (360). The source of PtdEtn for macroautophagy is unclear (148), and ER (20, 154), mitochondrial (145), and plasmalemmal (303) sources have been invoked and these may not be mutually exclusive. For more details on the complex machinery underlying macroautophagy, the reader is referred to recent, specialized reviews (213, 261). In the context of lipid catabolism, it is worth mentioning that phospholipids are degraded during macroautophagy; indeed, macroautophagy can affect the lipid content of cells by inducing lipid droplet breakdown (338). There is no evidence, however, that autophagy is a normal regulator of phospholipid content.

From cell death to cell division, PtdEtn has been implicated in numerous cellular functions, yet the mechanisms behind these feats are still obscure. Apart from its partners in macroautophagy, very few effectors that bind PtdEtn have been characterized. To our knowledge, no conserved PtdEtn-binding domain has been identified. Nevertheless, a family of PtdEtn-binding proteins has been discovered (387), and several small antimicrobial peptides, such as cinnamycin and duramycin, have been shown to bind PtdEtn exclusively (it is noteworthy that PtdEtn constitutes 70–80% of the total phospholipids in some bacteria) (87, 232). Even though PtdEtn is externalized during apoptosis, its role is less well characterized than that of PtdSer, which binds a host of cognate receptors and bridging proteins (216, 260). Considering its abundance, PtdEtn likely plays an important role in the morphology and physiology of the mitochondrial inner membrane (346), but its specific function is unknown. Likewise, it is unclear why depleting cellular PtdEtn inhibits cytokinesis (100, 101). Clearly, much remains to be learned about the physiology of PtdEtn.

D. Phosphatidylserine

Phosphatidylserine (PtdSer) makes up 2–10% of the total cellular phospholipid content (375). It is an anionic aminophospholipid by virtue of its phosphoserine head-group.

1. Biosynthesis and distribution of PtdSer

The pathways employed to generate PtdSer vary across species. In yeast and prokaryotes, PtdSer is synthesized by coupling CDP-DAG to serine (277, 374). In contrast, in mammalian cells, PtdSer is generated by a calcium-dependent exchange reaction between PtdEtn or PtdCho with l-serine, catalyzed by the PtdSer-synthases PSS2 or PSS1, respectively (FIGURE 1). Although single knockout mice of PSS1 or PSS2 are, for the most part, normal, double-knockout mice are not viable, highlighting the importance of PtdSer for development and survival (19). PSS1 and PSS2 are integral membrane proteins localized to the ER, specifically the MAMs of the ER (350).

As would be anticipated, PtdSer is present in the ER, where it is synthesized, but it is particularly enriched in the plasma membrane and in organelles of the endocytic pathway (212). In the plasmalemma, PtdSer is almost exclusively found in the cytoplasmic leaflet, a result of continuous and effective inward flipping by aminophospholipid translocases.

2. Catabolism of PtdSer

Most of the pathways for PtdSer degradation have been mentioned previously, in the context of other lipids. To reiterate, the major selective catabolic pathway is PtdSer decarboxylation in mitochondria, which serves to make PtdEtn (see previous section). Furthermore, the reaction catalyzed by PSS2 is reversible, and this serves to siphon off excess PtdSer, converting it back into PtdEtn at the ER. Of note, PtdSer is not thought to serve as a substrate for the known phospholipases, except for the PtdSer-specific PLA1 and some secreted isoforms of PLA2 (16, 274, 337).

3. Functional role of PtdSer

PtdSer has important signaling functions both inside and outside of the cell. Under physiological conditions, it is restricted to the intracellular milieu. During exceptional circumstances, such as apoptosis, injuries leading to blood clotting, or mast cell degranulation, PtdSer is presented on the extracellular surface of the plasmalemma. A calcium-sensitive scramblase is responsible for the abrupt exposure of PtdSer on the outer leaflet (352). PtdSer exposed extracellularly on apoptotic cells acts as an “eat me” signal, being recognized by phagocytic receptors on the surface of macrophages and other professional and nonprofessional phagocytes (104). In platelets, exofacial PtdSer ligates clotting factors, serving as a platform for the coordination of thrombosis (411).

In addition to these well-characterized extracellular functions, PtdSer surely plays multiple, important roles intracellularly. Strikingly, the functions of PtdSer inside the cell, where it is most abundant and constitutively present, have not been sufficiently studied and are hence poorly defined. Nevertheless, we know that intracellular PtdSer interacts with proteins through both selective and nonselective mechanisms. Some proteins, like PKC (172) and synaptotagmin (45), bind PtdSer via their C2 domains, which mediate association with anionic lipids in a calcium-dependent manner. Engagement of C2 domains can target proteins to membranes and/or induce conformational changes that modulate protein function. Other proteins interact with PtdSer via polybasic motifs that are often present near a hydrophobic determinant, sometimes in the form of an amphiphilic alpha-helix. Proteins like K-Ras (149) and some Rho GTPases (405) are thus attracted electrostatically to PtdSer-enriched membranes, notably the plasmalemma. In addition, PtdSer has protein-independent functions. In this regard, it has been postulated to facilitate membrane fusion and fission events (282, 391), although the detailed mechanisms are unknown.

E. Phosphatidylinositol

Phosphatidylinositol (PtdIns), formerly called monophosphoinositide (152), comprises a sizable fraction (10–15%) of the cellular phospholipids (375). PtdIns, an anionic phospholipid, is present in various organelles and, like the aminophospholipids, it is much more abundant in the inner leaflet of the plasmalemma than in the outer leaflet (91). It is unique among phospholipids in that it can undergo transient, reversible changes to its head-group. Specifically, the inositol moiety can be phosphorylated on positions 3, 4, and/or 5, producing seven distinct polyphosphoinositides. The monophosphoinositide PtdIns and its seven polyphosphoinositide derivatives make up the phosphoinositide family; the phosphoinositides have been the subject of intense study because they are quintessential mediators of protein targeting and signal transduction (91a, 310).

1. Biosynthesis and distribution of PtdIns

Unlike the above-mentioned phospholipids that are made from DAG, PtdIns is made from CDP-DAG (FIGURE 1). The enzyme CDP diacylglycerol:inositol-3-phosphatidyltransferase, also known as PtdIns synthase, promotes a condensation reaction between myo-inositol and CDP-DAG (13). PtdIns synthase is a soluble protein that associates with the ER (131, 230) and with highly mobile vesicles (191), where newly synthesized PtdIns is found. Secondary production of PtdIns can occur when the polyphosphoinositides are dephosphorylated. For instance, the dephosphorylation of PtdIns(3)P by myotubularins is a source of PtdIns within the endolysosomal system. Because PtdIns serves as the starting material for all of the polyphosphoinositides, dephosphorylation is a form of recycling, rather than a channel for de novo production.

Like other phospholipids, PtdIns is present in most organellar membranes, but its concentration varies among organelles as a result of the differential abundance and activity of the polyphosphoinositide-forming kinases, PtdIns-forming phosphatases, and PtdIns synthase.

2. Catabolism and functional role of PtdIns

Several important products are derived from PtdIns. First, the lipid moiety of glycosylphosphatidylinositol (GPtdIns)-linked proteins originates from PtdIns (164). Linking otherwise soluble proteins to GPtdIns serves to anchor them to a lipid bilayer, normally the exofacial leaflet of the plasma membrane. GPtdIns-linked proteins are generated by glycosylation of PtdIns, which occurs in the cytosolic side of the ER, followed by insertion of a phosphoethanolamine moiety derived from PtdEtn. This fully formed GPtdIns module is then flipped across the ER membrane to face the lumen, where it becomes covalently attached to the COOH terminus of a protein. As a result, GPtdIns-linked proteins are found exclusively on the luminal leaflet of secretory vesicles and on the topologically equivalent extracellular surface of the plasmalemma. At least in vitro, the protein moiety can be detached by PLD- or PLC-mediated cleavage (227).

PtdIns itself can also be the source of DAG. PLC preferentially cleaves PtdIns(4,5)P₂, but also has measurable activity towards PtdIns (314, 392), liberating DAG and phosphoinositol. The functional impact of this source of DAG is not clear.

While the preceding reactions contribute to its metabolism, the most important path for PtdIns conversion is its phosphorylation. Despite being minor constituents of membranes, the resulting polyphosphoinositides are key regulators of membrane traffic and protein activity. As such, their function is discussed in detail in the sections devoted to endocytosis, phagocytosis, and macropinocytosis. However, a brief primer on their biosynthesis and catabolism is presented below. Of note, in the following sections the kinases that phosphorylate only PtdIns (the PtdInsKs) are distinguished from other phosphoinositide kinases (the PtdInsPKs) that phosphorylate a single polyphosphoinositide or multiple phosphoinositides.

F. Polyphosphoinositides

1. PtdIns(4)P

PtdIns(4)P is highly enriched in the Golgi apparatus and has on occasion been used as a marker of this organelle (85a). It is also found at lower concentrations on endosomes, the plasma membrane, and the ER. PtdIns(4)P is created from PtdIns by PtdIns 4-kinase (PtdIns4K; not to be confused with PtdInsP4K). Two types of PtdIns4K (type II and III PtdIns4K) have been described in mammalian cells, which differ in their molecular weight and sensitivity to PtdInsP3K inhibitors and adenosine (26). PtdIns(4)P can be consumed by degradation or by conversion to more complex inositide species. It is dephosphorylated by PtdIns(4)P phosphatases such as the yeast Sac1 and its mammalian orthologs (167, 275) and can serve as a comparatively poor substrate of PLC (314, 392). PtdIns(4)P can also be consumed by conversion into PtdIns(4,5)P₂, a reaction catalyzed by the PtdOH-sensitive type I PtdInsP5K (176).

PtdIns(4)P is key to sphingolipid synthesis in the Golgi complex, where it interacts with various proteins that include CERT, FAPP2, and OSBP1 (138). In addition, the adaptor protein AP-1 is recruited to the trans-Golgi network (TGN) upon binding both PtdIns(4)P and Arf1 (155, 345, 388). In this manner PtdIns(4)P contributes to the regulation of anterograde traffic from the TGN.

2. PtdIns(5)P

PtdIns(5)P is by far the least characterized phosphoinositide (78). It is not very abundant, and suitable probes for its detection are lacking. As a result, its distribution, metabolism, and functions have not been fully characterized. PtdIns(5)P can be made when PtdIns(3,5)P₂ is dephosphorylated by myotubularins, which are 3-phosphatases (207, 365). It is used as a substrate by type II PtdInsP5K (more accurately called type II PtdInsP4K, as it will be referred to hereafter) to generate PtdIns(4,5)P₂ (302), but this pathway is likely a minor contributor to the overall biosynthesis of PtdIns(4,5)P₂ (307). PtdIns(5)P may also play a role in traffic between the late endosome and the plasma membrane, where PtdIns(3,5)P₂ and PtdIns(4,5)P₂, respectively, are enriched. To our knowledge, only one specific binding partner of PtdIns(5)P has been identified in animal cells: the inhibitor of growth 2, which is a tumor suppressor (137).

3. PtdIns(4,5,)P₂

PtdIns(4,5)P₂ is most enriched at the inner leaflet of the plasma membrane, where it contributes 1–2 mol% of the phospholipids (249). It is produced through phosphorylation of PtdIns(4)P by type I PtdInsP5K and, to a lesser extent, from PtdIns(5)P by type II PtdInsP4K (58). Three isoforms of type I PtdInsP5K (termed α, β, and γ with some discrepancy between different species) have been described, and three splice variants of γ (87, 90, and 93 kDa) are expressed (175, 176, 225). Dephosphorylation of PtdIns(3,4,5)P₃ by PTEN is another source of PtdIns(4,5)P₂. PtdIns(4,5)P₂ itself can be dephosphorylated by a remarkable number of 5′ phosphatases that are highly regulated and include synaptojanin, OCRL, and Inpp5B (283).

Furthermore, PtdIns(4,5)P₂ is metabolized by lipases and kinases that, in the process, generate several key signaling molecules. It is phosphorylated by any one of several PtdInsP3K isoforms to produce PtdIns(3,4,5)P₃, a critical mediator of cellular activation that is discussed in more detail below. PtdIns(4,5)P₂ can also be hydrolyzed into IP₃, the quintessential Ca²⁺-mobilizing agent, and DAG, which recruits C1 domain-containing proteins, most prominently the classical isoforms of PKC (140). The PLC family that mediates the hydrolysis of PtdIns(4,5)P₂ currently consists of 13 members, divided among 6 isotypes (β, γ, δ, ε, ζ, and η). The PLC-γ isoforms are regulated by receptor and nonreceptor tyrosine kinases, while the β, ε, and η isoforms are sensitive to G protein subunits (51, 186). Interestingly, in addition to mobilizing Ca²⁺ by breaking down PtdIns(4,5)P₂ into IP₃, all of the isoforms require Ca²⁺ for optimal activity.

The functional contribution of PtdIns(4,5)P₂ is not limited to its role as a substrate for the production of second messengers, however. Its unique electronegativity (3.5–4 negative charges at physiological pH) and ability to segregate in specialized membrane domains make it a preferred ligand of a variety of polycationic proteins, such as MARCKS (250). The latter and other similarly abundant polycationic proteins are thought to sequester a significant fraction of plasmalemmal PtdIns(4,5)P₂ and can make it available when they detach upon phosphorylation. PtdIns(4,5)P₂ is ligated not only by polycationic proteins, but also by a variety of proteins with domains that recognize PtdIns(4,5)P₂ stereospecifically via PH (pleckstrin homology), ENTH (epsin NH₂-terminal homology), FERM (band 4.1/ezrin/radixin/moesin), and other domains (151, 209). As a result, PtdIns(4,5)P₂ has been reported to interact with myriad proteins; some of these influence cytoskeletal structure by capping or severing actin filaments (315), while others like PLD can generate second messengers or modulate the activity of GTPases (284). The specific roles of PtdIns(4,5)P₂ in endocytic processes are detailed in subsequent sections.

4. PtdIns(3,4,5)P₃

PtdIns(3,4,5)P₃ is a major signaling motif and survival signal. Its concentration in resting cells is minute and, even after stimulation by growth promoters, PtdIns(3,4,5)P₃ makes up <1% of the total lipid (349). The enzymes that generate PtdIns(3,4,5)P₃ from PtdIns(4,5)P₂ are the family of class I PtdInsP3-kinases, which are largely plasmalemmal enzymes. The PtdInsP3K isoforms are divided into subgroups and vary in their subunit composition and in the means whereby they become activated. In light of their functional importance in health and disease, the PtdInsP3Ks have been studied extensively (102, 224). While a comprehensive overview is outside the scope of this review, the reader is referred to recent reviews for more details (124, 377). Briefly, the type I PtdInsP3Ks consist of two subunits: a catalytic or p110 subunit and a regulatory subunit. Three isoforms of the catalytic subunit (p110α, β, and δ) bind a regulatory subunit (p85, p55, or p50) that is controlled by tyrosine phosphorylated proteins. The p110γ subunit, on the other hand, binds either p84 or p101 as its regulatory subunit and is also regulated by G protein-coupled receptors. PtdIns(3,4,5)P₃ can be dephosphorylated by both 3′ phosphatases, such as PTEN (54), and type II inositol 5′ phosphatases like SHIP. The resulting phosphoinositide products, PtdIns(4,5)P₂ and PtdIns(3,4)P₂, respectively, have very distinct functional connotations, described in their respective sections. Activating mutations in the catalytic subunit of type I PtdInsP3K and inhibitory mutations in PTEN are associated with a variety of human cancers (61). In fact, coupling these mutations together is sufficient to induce tumorigenesis in mice (192).

By virtue of the additional phosphate group, the net charge of PtdIns(3,4,5)P₃ is even greater than that of PtdIns(4,5)P₂. However, the contribution of PtdIns(3,4,5)P₃ to the surface charge of the membrane is probably minor, in view of its low abundance. Instead, the physiological responses elicited by PtdIns(3,4,5)P₃ are attributable to stereospecific recognition of its head group by proteins bearing a specific type of PH or PX (phagocyte oxidase homology) domain (217). A number of important signaling proteins, such as PLC-γ, Akt, Vav, and PDK isoforms, bear PtdIns(3,4,5)P₃-binding domains that recruit and/or activate them at the membrane.

5. PtdIns(3,4)P₂

Like PtdIns(3,4,5)P₃, PtdIns(3,4)P₂ is a potent survival signal. Different pathways can lead to the biosynthesis of PtdIns(3,4)P₂. It can be formed from PtdIns(4)P by phosphorylation at the 3′ position of the inositol ring by the class II PtdInsP3K, from phosphorylation of PtdIns(3)P by type II PtdInsP4K (73), or from dephosphorylation of PtdIns(3,4,5)P₃ by 5′ phosphatases. Elimination of PtdIns(3,4)P₂ can occur by removal of the 4′ phosphate by the phosphatases, Inpp4A and Inpp4B, or of the 3′ phosphate by PTEN, yielding PtdIns(3)P and PtdIns(4)P, respectively. The biological effects of PtdIns(3,4)P₂ have been attributed to the recruitment of proteins bearing PH or PX domains. These include the kinase Akt and the adaptor proteins TAPP1 and TAPP2.

6. PtdIns(3)P

PtdIns(3)P is particularly abundant in the cytoplasmic leaflet of the membrane of early endosomes. It has also been detected on internal vesicles of multivesicular bodies. PtdIns(3)P is made primarily by phosphorylation of PtdIns by class III PtdIns3K, which is also known as Vps34 (22). PtdIns can also be phosphorylated by class II PtdInsP3K, but this reaction is felt to be quantitatively less important. Furthermore, PtdIns(3)P can be made by dephosphorylation of PtdIns(3,4)P₂ by Inpp4. PtdIns(3)P is degraded by 3′-phosphatases of the myotubularin family (308). Mutations in members of the myotubularin family cause myotubular myopathy and Charcot-Marie-Tooth syndrome type 4B (32). Although present at comparatively low concentrations in the cell, PtdIns(3)P plays very important roles in membrane traffic, directing endosome progression and phagosome maturation and regulating autophagy, and the production of reactive oxygen species (ROS). As described for the other phosphoinositides, PtdIns(3)P functions by recruiting a defined population of proteins that stereospecifically recognize its head-group. A sizable number of proteins bearing FYVE (an acronym for Fab 1/YOTB/Vac 1/EEA1) or certain types of PX domains have been described to bind PtdIns(3)P selectively (209). The role of these proteins in endocytic processes is discussed in detail in later sections.

7. PtdIns(3,5)P₂

PtdIns(3,5)P₂ is found primarily on late endosomal compartments (94, 162). It is produced by phosphorylation of PtdIns(3)P on position 5′ by PtdInsK-FYVE (FYVE domain-containing PtdIns kinase), also known as Fab1 in yeast (281). Like PtdIns(3)P, PtdIns(3,5)P₂ is dephosphorylated on the 3′ position by myotubularins, generating the rare phosphoinositide PtdIns(5)P. More commonly, it is dephosphorylated on the 5′ position by Sac3, also known as Fig4 in yeast. Notably, Sac3/Fig4 knockdown or mutation often has baffling effects on PtdIns(3,5)P₂ levels and sometimes causes them to decrease inexplicably as though the phosphatase produces PtdIns(3,5)P₂ rather than eliminating it (70, 96). This discrepancy suggests a reciprocal relationship between PtdInsK-FYVE/Fab1 and Sac3/Fig4, where both proteins depend on each other for optimal performance, even though they are functionally at odds with one another. Indeed, this antagonistic kinase-phosphatase pair associates physically forming a heteromeric complex where they are held together by ArPtdInsKfyve (Vac14 in yeast) (40, 323). Such an interaction explains why mutations in the gene encoding Sac3/Fig4 are associated with increased PtdIns(3,5)P₂ levels and with some forms of amyotrophic lateral sclerosis (69) and Charcot-Marie-Tooth peripheral neuropathy (70).

Although its physiological role and mode of action are unclear, PtdIns(3,5)P₂ has been implicated in the osmotic stress response, in autophagy, and in the control of ionic channel activity. Available evidence most strongly suggests that PtdIns(3,5)P₂ regulates sorting of endocytic membranes for delivery to the TGN (313) and to intraluminal vesicles (281). Formation of PtdIns(3,5)P₂ may act as a checkpoint during the early-to-late endosomal transition. Like other phosphoinositides, PtdIns(3,5)P₂ can recruit a growing list of proteins through stereospecific interactions, and the PROPPIN motif has been described to bind PtdIns(3,5)P₂, although it also binds PtdIns(3)P in some cases (94).

G. Bis(monoacylglycero)phosphate/lysobisphosphatidic acid

Bis(monoacylglycero)phosphate or lysobisphosphatidic acid (BMP/LBPA) is an unconventional phospholipid. It is a polyglycerophospholipid, consisting of two monoacylated glycerol molecules, combined together through a single phosphate group. It is anionic and makes up <1% of the total phospholipids in mammalian cells (369).

1. Biosynthesis and distribution of BMP/LBPA

BMP/LBPA is most abundant in lysosomes and late endosomes, specifically in multivesicular bodies (MVBs), where it accounts for ∼15% of the total phospholipid content (194). It is further enriched in certain subpopulations of intraluminal vesicles (ILVs), where it is estimated to constitute ∼70% of the phospholipid content (195). This fivefold difference between ILVs and the ILV-containing vacuole is not surprising because BMP/LBPA is virtually absent from the limiting membrane of MVBs (261a).

Like the polyphosphoinositides, BMP/LBPA seems to be synthesized at the location where it is found, namely, the MVB. In this respect, it differs from the major, conventional phospholipids that are made in the ER or Golgi complex. The in vivo precursors of BMP/LBPA are unknown, but it is clear that its formation does not require PtdIns(3)P, the other uniquely endosomal phospholipid (294). Instead, BMP/LBPA is most likely derived from phosphatidylglycerol, but quantitative considerations suggest that phosphatidylglycerol condensation may not account for all of the BMP/LBPA present in cells (170). In view of its restricted localization to mitochondria, cardiolipin is unlikely to account for the remainder of the BMP/LBPA, but another pathway involving conversion of PtdCho has been implicated (305).

2. Catabolism and functional role of BMP/LBPA

The degradation of BMP/LBPA is even less clear than its synthesis. Even though it is resistant to phospholipase degradation (242), it is possibly degraded by lysosome-specific lipases at a very slow rate or recycled through back-fusion of the ILVs with the limiting membrane and retrograde traffic.

MVBs form as ILVs bud into the lumen of late endosomes. At least some of these vesicles deliver ubiquitinated transmembrane proteins into the hydrolytic environment of the acidic interior by a process requiring the ESCRT complex. To the extent that it is found in ILVs, it is reasonable to speculate that BMP/LBPA may be necessary for this inward budding. Indeed, even in pure lipid model membranes, BMP/LBPA is capable of inducing spontaneous inward vesiculation (241). Because this inward budding is pH dependent, it is conceivably caused by transmembrane flipping of the protonated, uncharged species, with a resulting alteration in membrane curvature. However, the situation in vivo is likely more complex. This is suggested by the observation that Alix, a BMP/LBPA-interacting protein, is also required for ILV formation. Lastly, it must be pointed out that more than one type of ILV exists and that it is not clear whether the vesicles generated by the ESCRT system are the same as those containing BMP/LBPA.

IV. SUBORGANELLAR DISTRIBUTION OF PHOSPHOLIPIDS

Not only does the phospholipid composition vary between compartments, but also between subdomains within compartments. These differences can be in the transversal or in the lateral direction: the composition of the two monolayers of a single membrane is often very different, and lipids are nonuniformly distributed in the plane of the membrane (or even the individual leaflet). These inhomogeneities have profound functional consequences. The mechanisms underlying their formation and their functional consequences are discussed next.

A. Transmembrane Topology: Flippases, Floppases, and Scramblases

Although DAG can flip spontaneously across the bilayer, most phospholipids translocate across membranes very slowly. Those with acidic head-groups, like PtdOH and BMP/LPBA, can be induced to translocate in vitro under low pH (367b), an event that may be of significance in the late compartments of the endocytic or secretory pathways, which are richly endowed with V-ATPases. For other phospholipids, such as PtdCho, transbilayer movement occurs at a glacial pace, on a time scale that is not physiologically relevant (23, 251).

Despite this low intrinsic rate of translocation, phospholipids are known to traverse biological membranes frequently, sometimes at astonishingly high rates. Such translocation is mediated by specialized lipid-transport proteins: the flippases, floppases, and scramblases (91). Flippases are energy-dependent proteins that transfer lipids from the extracellular leaflet, or in the case of intracellular organelles, the luminal leaflet, to the cytoplasmic leaflet of the respective membrane. Aminophospholipid flippases transfer and maintain virtually all the PtdSer and a large fraction of the PtdEtn on the cytoplasmic aspect of the plasmalemma. Generation of this asymmetric distribution requires the investment of energy and, while their identity has been long debated, it is generally agreed that these enzymes require ATP to catalyze vectorial translocation. Recent evidence strongly suggests that flippases are in fact members of the P4 family of ATPases, which are related to cation-translocating pumps (289, 359, 378). Yeast cells express five P4 ATPases. The P4 ATPase Drs2p has been studied most (409). It localizes to late Golgi membranes, where it has been shown to flip fluorescently-tagged versions of PtdSer and PtdEtn. Mammalian cells express 14 different P4 ATPases. Remarkably, the ability of the mammalian P4 ATPases to flip aminophospholipids has not been unequivocally documented to date (120). Thus, although mutation or deletion of genes encoding some of the ATPases cause severe phenotypic defects, it is unclear whether these result from defective lipid flipping or from some other derangement.

By actively pumping PtdSer inward, flippases prevent the exposure of this aminophospholipid on the outer surface of normal cells. This in turn precludes recognition by PtdSer receptors, enabling phagocytes to distinguish between healthy and apoptotic cells (104). A second consequence of flippase activity is the accumulation of PtdSer, an anionic species, on the inner aspect of the membrane, thereby contributing to the generation of a considerable negative surface charge. Other contributors to the surface charge and its functional implications are discussed in more detail below.

Floppases are similarly energy-dependent lipid transporters. However, unlike flippases, they actively move phospholipids from the inner to the outer leaflet of the plasma membrane or, in the case of endomembrane organelles, from the cytoplasmic to the luminal leaflet. ATP-binding cassette (ABC) transporters are transmembrane proteins that translocate a large variety of substances, including phospholipids, at the expense of ATP hydrolysis. The ABC transporter ABCB4 is necessary for bile formation (141); knocking out this gene in mice results in impaired PtdCho secretion into bile (340). Thus this PtdCho exporter is often considered a floppase for PtdCho (184a).

In addition to being actively created through investment of metabolic energy, phospholipid asymmetry could in principle also be generated by asymmetrically trapping randomly moving lipids on one side of the bilayer. For example, certain lipids may interact with, and be retained by, cytoskeletal proteins, which are available only to lipids exposed cytosolically. While this source of lipid asymmetry has not been formally demonstrated (52, 143, 296), protein-mediated mooring cannot be discounted.

Scramblases are energy-independent transporters that catalyze the spontaneous translocation of phospholipids across the bilayer. These proteins are Ca²⁺-dependent enzymes that facilitate the entropic, bidirectional movement of lipids. Unlike flippases and floppases that generate the uneven distribution of phospholipids, scramblases collapse the asymmetry. They are essential for the externalization of PtdSer that occurs during apoptosis and also during platelet activation. Indeed, defects in scrambling cause the bleeding disorder Scott syndrome (412). As was the case for flippases, the molecular identity of the scramblase(s) was the subject of controversy. A compelling candidate is transmembrane protein 16F (TMEM16F), which was recently demonstrated to promote scrambling when transfected heterologously, and to be mutated in a Scott syndrome patient (352).

B. Lateral Segregation of Subdomains

In addition to transmembrane asymmetry, membranes exhibit lateral (cis) asymmetry. Within a given membrane, phospholipids are frequently enriched in hot spots that may be transient or long-lived. An obvious example is that of PtdIns(3,4,5)P₃, which is often enriched in one area of the cytoplasmic leaflet of the plasmalemma. During phagocytosis, for instance, PtdIns(3,4,5)P₃ accumulates selectively at the phagosomal cup (239), with little enrichment apparent in other regions of the plasma membrane, despite the ostensible continuity of the bilayer. Likewise, during chemotaxis, PtdIns(3,4,5)P₃ is selectively present at the leading edge of the cells (163) and, in some polarized epithelia, PtdIns(3,4,5)P₃ is abundant in the basolateral membrane yet seemingly absent from the apical membrane (48). These gradients of PtdIns(3,4,5)P₃ may be established by various mechanisms that segregate biosynthetic or catabolic enzymes; these could include the local accumulation and/or activation of lipid kinases or phosphatases at sites of receptor stimulation or at specialized structures like the epithelial tight junction. However, diffusional barriers may be necessary to prevent homogenization by lateral diffusion. Proteins such as septins or components of the ESCRT complex have been invoked as components of such diffusion barriers (103).

In addition to the comparatively large domains mentioned above, lipids are also segregated into much smaller and dynamic subdomains. Different types of microdomains have been described, but the so-called “lipid rafts” are the most thoroughly investigated (334). Lipid rafts are nanoscale assemblies where lipids are stabilized within a liquid-ordered phase. The size and lifespan of such rafts have been hotly debated, but the current thinking is that they are metastable clusters 20–70 nm in diameter. Sphingolipids and cholesterol provide the core for the formation of rafts, at least on the outer leaflet of the plasma membrane, but phospholipids also participate. Lipids with longer and saturated acyl chains are thought to reside preferentially in rafts. In this regard, it is noteworthy that saturated PtdSer species are more abundant in the plasmalemma than in other cellular membranes. Some PtdIns(4,5)P₂ species have also been postulated to associate with rafts (56). Conversely, lipids with unsaturated chains tend to be excluded from rafts.

Segregation of the membrane into liquid-ordered and liquid-disordered microdomains can be driven solely by lipids, but association with proteins also influences lipid partition in the lateral plane. This confers an additional, dynamic dimension to lipid architecture. One example is provided by the local enrichment of PtdIns(4,5)P₂ at nascent endosomal pits (125). The underlying mechanism has not been fully defined, but tight association with proteins containing BAR or other PtdIns(4,5)P₂-binding domains is likely to confine the phosphoinositide to the region of the pit. Moreover, protein rearrangements that occur in the course of receptor signaling can also cause phospholipids to redistribute, based on hydrophobic and head-group matching (97).

Microdomains that form in the membrane are not always free to diffuse laterally. Indeed, transmembrane proteins anchored to the cytoskeleton can act as pickets of cytoskeletal fences that corral membrane components. These corrals, however, are not permanent structures and membrane components can escape during the course of transient openings. Moreover, phospholipids can “hop” across these barriers (126).

V. BIOPHYSICAL PROPERTIES OF PHOSPHOLIPID MEMBRANES

The chemical nature of phospholipids can affect the properties of their constituent membranes. Certain phospholipids distort bilayers, conferring positive or negative curvature. Others contribute negative charge to a membrane, affecting the interactions of the bilayer with proteins and with other membranes.

A. Curvature

During vesicle or tubule formation, membranes undergo acute changes in curvature. A monolayer is said to exhibit negative curvature (concavity) when the head groups point towards the center of the arc. Conversely, positive curvature (convexity) occurs when the head groups point away from the center of the curvature (268). Any bend within a bilayer creates positive curvature in one monolayer and negative curvature in the contralateral one. During endocytosis, for example, the nascent invagination imposes positive curvature on the cytoplasmic leaflet and negative curvature on the extracellular leaflet, which will become luminal. At the neck of the forming endosome, on the other hand, the external leaflet experiences positive curvature while the cytoplasmic monolayer experiences negative curvature.

Phospholipids with conical morphology (i.e., those with a head that is smaller in diameter than are the acyl tails) are thought to induce, or at least prefer and thereby stabilize, negative curvature. The opposite is true for any lipid with a head that is larger than the tails. PtdOH and lysophospholipids, respectively, exemplify negative and positive curvature-inducing lipids.

Although these simplistic predictions may be true for synthetic membranes, it is likely that lipid-protein interactions are predominately responsible for dictating membrane curvature within cells. Indeed, certain proteins are associated with specific types and degrees of curvature. The most compelling feature of proteins that bind curved membranes is the Bin-Amphiphysin-Rvs (BAR) domain (12, 86). Classical BAR domains can interlock to produce a banana-shaped dimer that preferentially binds to positively curved membrane buds or tubes, whereas I-BAR domains produce a cigar-shaped dimer that prefers negatively curved membrane invaginations (406). It is not always clear whether these proteins are sensing and stabilizing “intrinsic,” spontaneously generated curvature, or generating the curvature themselves. Lastly, membrane curvature can be generated by the application of mechanical tension. The acto-myosin cytoskeleton is chiefly responsible for the exertion of lateral tension on biological membranes and is presumably an important contributor to curvature. In support of this, plasma membrane blebbing, an example of a large-scale change in membrane curvature, is promoted by myosin II (257).

B. Surface Charge

Proteins associate with membranes through stereospecific, hydrophobic, and/or electrostatic interactions. Biological membranes are thought to be, to varying degrees, electronegative (248). The more abundant nonionic or zwitterionic lipids make no contribution to the surface charge, but a considerable fraction of the lipids are anionic, and very few like sphinganine and sphingosine, can be cationic. PtdOH, PtdSer, and the phosphoinositides are all negatively charged at physiological (cytosolic) pH. Polyphosphoinositides like PtdIns(4,5)P₂ and PtdIns(3,4,5)P₃ are polyanionic, making a disproportionate contribution to the surface charge. Through Coulombic interactions, electronegative membranes attract cationic solutes, notably polycationic proteins. Salient examples include MARCKS, c-Src, Rac1, and K-Ras, all expressing polycationic motifs (397). In many instances, these cationic motifs are flanked by a hydrophobic determinant, such as prenylation, acylation, or a cluster of hydrophobic amino acids. The combined effect of electrostatic attraction and hydrophobic partition provides a powerful means for these proteins to associate with anionic membranes (399). Conversely, proteins that are anionic and hydrophilic will be repelled from the membrane.

The anionic surface charge of biological membranes, notably the inner aspect of the plasma membrane, affects not only the recruitment and retention of soluble proteins, but also the structure and function of transmembrane proteins. A variety of ion channels and exchangers have clusters of basic residues in their cytoplasmic tails or in loops linking transmembrane domains (8). These are exposed and sensitive to the membrane charge, which can result in distortion of the protein structure and, consequently, alteration of its function. Thus electrostatic association of ion channels with phosphoinositides on the inner leaflet modifies their open probability and voltage dependence and, in the case of ion exchangers can explain their ATP dependence and osmotic sensitivity (8).

It is most likely that the surface charge of the inner leaflet changes during the course of endocytic events, with accompanying functional changes. This speculation is based on the observation that the phosphoinositide composition of the membrane changes drastically upon membrane scission, and the recent report that, at least in yeast, PtdSer is selectively excluded from endosomes (107). Localized generation of PtdOH during phagocytosis or macropinocytosis could also contribute to altering the surface charge. The predicted net loss of negative charge is anticipated to cause considerable remodeling of the proteins that associate electrostatically with the membranes, and to alter the responsiveness of some transmembrane proteins.

The surface charge of the membrane and its dynamic changes during endocytic processes may have protein-independent functions. The mutual Coulombic repulsion of anionic phospholipids potentially contributes to their redistribution and lattice formation (342). Furthermore, the lateral pressure profile of a membrane fluctuates as its surface charge is changed (211).

VI. ENDOCYTOSIS AND PROGRESSION

A. Types of Endocytosis

Endocytosis is a generic term that describes all forms of plasma membrane internalization. Some types of endocytosis result from small invaginations of 100–200 nm in diameter, whether vesicular, flask-like, or tubular in shape. Others involve the formation of large membranous vacuoles that can reach several microns in diameter; these include macropinosomes and phagosomes. Following scission, the small endocytic vesicles generally fuse with a common compartment, the early endosome. Vesicles generated by clathrin-mediated endocytosis (CME), caveolar endocytosis, flotillin-dependent endocytosis, and Arf6-dependent endocytosis all converge at early (sorting) endosomes. In contrast, the tubules generated by the CLIC/GEEC seem to take a different route, at least initially (34). The large vacuoles formed by macropinocytosis and phagocytosis also merge with early endosomes, but being considerably larger, generate hybrid organelles distinct from the normal sorting endosomes.

CME is by far the most thoroughly studied of endocytic pathways. It involves the coat protein clathrin and the large GTPase dynamin. While dynamin is also used by other endocytic pathways, only CME utilizes clathrin. Caveolar endocytosis relies on caveolin and cavins as “coat-forming” proteins (150). It is characterized by flask-like invaginations (caveolae) that are rather stable and very slowly internalized, if at all (288). Whether their scission requires dynamin or a dynamin-like protein, such as EHD2, is still debated. In polarized epithelia, caveolae are restricted to the basolateral membrane, and they are absent from leukocytes, which do not normally express any of the caveolin isoforms. Flotillin-dependent endocytosis is similar to caveolar endocytosis, but depends instead on flotillin-1 and -2. Surprisingly, these molecules are also found on late endosomes (90, 341), where their function is unclear. Both dynamin-dependent (7a) and dynamin-independent (136) uptake of flotillin-containing vesicles has been described. Like caveolae endocytosis, flotillin-induced invaginations bud infrequently (92). Finally, Arf6-dependent endocytosis is related to membrane recycling (272, 300), especially of MHC-I. In most cases, it is dynamin independent.

These clathrin-independent pathways often involve cholesterol and sphingolipids (166, 245). There is evidence connecting the Arf6-dependent pathway (46) and also CLIC/GEECs with PtdIns(4,5)P₂ (229), and caveolae with PtdSer (108), but the available data are largely preliminary and indirect. Because the information relating caveolae, CLIC/GEECs, and flotillin- or Arf6-dependent endocytosis with phospholipids is scant, these pathways were excluded from this review. Instead, we directed our focus to the three forms of endocytosis where a strong dependence on phospholipids has been well documented: CME, macropinocytosis, and phagocytosis. The role of phospholipids is analyzed not only in the context of membrane invagination and scission, but also with regards to the progression/maturation of the resulting endomembrane structures. To the extent that caveolae and vesicles formed via flotillin- or Arf6-dependent endocytosis also merge with sorting endosomes, the analysis of endosome progression can apply to their cargo as well.

B. Clathrin-Mediated Endocytosis

CME facilitates the internalization of a variety of membrane constituents, notably receptors. Some, typified by the transferrin receptors (TfR), are recycled constitutively even in the absence of ligand, while others like epithelial growth factor receptors (EGFR), are only taken up in response to ligand binding. By internalizing activated receptors, CME modulates signaling; in addition, it contributes to nutrient uptake and helps the cell sample the extracellular milieu.

CME is mechanistically distinct from all other endocytic processes. It is characterized by the recruitment and assembly of cytosolic clathrin triskelia into a plasmalemmal lattice that gradually curves into a soccer ball-like structure that undergoes scission, yielding clathrin-coated vesicles (FIGURE 2). Clathrin, however, does not bind to the membrane directly; it requires bridging by a veritable army of adaptor and accessory proteins. In addition, the large GTPase dynamin is necessary for vesicle scission (238).

Unlike phagocytosis and macropinocytosis, CME is largely an actin-independent process, at least in mammalian cells. While yeast cells require actin for CME (182, 183), in higher organisms actin is necessary for CME formation only on the apical surface of epithelial cells (171). This may be attributable to the unique mechanical properties of the apical brush border. Indeed, CME becomes actin dependent in other membranes as well when they are subjected to (osmotically induced) tension (42). CME is also distinct from phagocytosis and macropinocytosis in that it does not require PtdIns(3,4,5)P₃. Instead, the major phospholipid implicated in CME formation is PtdIns(4,5)P₂, by the mechanisms detailed below.

1. Endosome formation via clathrin

According to the model of McMahon and Boucrot, CME formation can be divided conceptually into five steps: nucleation, cargo selection, coat assembly, scission, and uncoating (253). During the first step, nucleation factors bind to endocytic hot spots, identifying these locations as suitable for clathrin coat assembly and pit formation. It is unclear why certain spots are favored over others on the plasma membrane. A local enrichment in anionic phospholipids could be one of several signals that identify a favorable location (280). Once the nucleation factors are recruited, they are thought to induce the initial budding of the plasmalemma. Accordingly, the nucleating proteins FCHO 1 and 2 contain F-BAR domains, which are known to associate with comparatively flat areas of positive curvature (157). The FCHO proteins also recruit Eps15 and intersectin, which subsequently recruit AP2 and possibly other adaptor proteins that are cargo-specific. These adaptor proteins bind both the cargo and the neighboring plasma membrane and contribute to the second step in CME formation, i.e., cargo selection (259, 285).

Clathrin is the most visually obvious feature of CME. Although it was initially thought to induce membrane curvature, clathrin does not bind to the plasma membrane directly nor does it interact with any phospholipids (99). More likely, its role is to stabilize positively curved membranes by binding BAR domain-containing proteins like endophilin and amphiphysin, which can sense and/or generate positive curvature in phospholipid bilayers. At this point in the formation process, when endophilin and amphiphysin are recruited, the clathrin-coated pit has reached its apex of positive curvature at the bud and negative curvature at the neck. Endophilin and amphiphysin additionally recruit dynamin, a large GTPase that is necessary for vesicle scission (112, 159, 358). After the vesicle separates from the plasma membrane, it undergoes uncoating, which allows it to fuse with vesicles of the endolysosomal system. At this point, CME formation is complete and the fully formed endocytic vesicle can undergo progression so that cargo is properly processed.

PtdIns(4,5)P₂ is necessary for the early steps in CME formation, namely, nucleation, cargo selection, coat assembly while scission and uncoating depend on PtdIns(4,5)P₂ elimination (FIGURE 2). Remarkably, with the exception of clathrin itself, almost every protein associated with CME formation binds directly to PtdIns(4,5)P₂!

CME is initiated in PtdIns(4,5)P₂-rich membranes that have, at the outset, low curvature. PtdIns(4,5)P₂, formed predominantly by type I PtdInsP5K (286), serves to recruit several proteins associated with CME to the membrane and dictates the size of vesicles (11). Even proteins like Arf6 that are upstream of type I PtdInsP5K stimulate CME formation by amplifying PtdIns(4,5)P₂ production (201). The domains that mediate PtdIns(4,5)P₂ binding during CME make up a diverse list. The F-BAR domain found in FCHO proteins associates with anionic phospholipids, like PtdIns(4,5)P₂ (157). The clathrin-binding molecules epsin and AP180 have ENTH and ANTH domains, respectively, that similarly interact with PtdIns(4,5)P₂ (67). Classical BAR-domain containing proteins, such as endophilin and amphiphysin, bind electronegative phospholipids, including PtdIns(4,5)P₂ and PtdSer (293). Sorting nexin 9 (Snx9) has an unconventional PX domain that, unlike many other PX domains, prefers PtdIns(4,5)P₂ (228, 299). Finally dynamin, through its PH domain, also binds PtdIns(4,5)P₂ (3). In addition to these proteins with stereotypical PtdIns(4,5)P₂-binding domains, the cargo-recognizing adaptor AP2 binds simultaneously to PtdIns(4,5)P₂, cargo and clathrin (76). Accordingly, depleting cells of PtdIns(4,5)P₂ prevents AP2 recruitment, and CME formation is reduced because forming endosomes lack the quintessential bridge between cargo and clathrin (410). Multiple methods have been employed to reduce PtdIns(4,5)P₂ levels, including pharmacological inhibition of type I PtdInsP5K (41), RNA interference to knock down type I PtdInsP5K (63), and overexpression of 5′ phosphatases (11). These techniques suffer from limitations because they are either imperfectly selective, or are chronic, generalized treatments that affect the entire cell. An attractive method involves breaking down PtdIns(4,5)P₂ acutely and selectively at the plasma membrane by acutely recruiting a phosphatase to a specific location via a chemically induced heterodimerization system (1, 410). Nonetheless, all of these diverse approaches have converged to support the notion that PtdIns(4,5)P₂ is required for efficient CME; indeed, addition of exogenous PtdIns(4,5)P₂ can rescue CME when PtdIns(4,5)P₂ synthesis is inhibited (41). Unexpectedly, despite the widespread requirement for the phosphoinositide, some clathrin lattices still assemble at the membrane when PtdIns(4,5)P₂ is depleted (1). However, these form only pits that are static, nonbudding, and infrequent.

In contrast to the requirement for PtdIns(4,5)P₂ in nucleation and coat assembly, scission and clathrin uncoating are associated with PtdIns(4,5)P₂ depletion (82). It is easy to envisage how PtdIns(4,5)P₂ hydrolysis induces uncoating, as it promotes dissociation of the multiplicity of adaptors that affix clathrin to the membrane. PtdIns(4,5)P₂ is depleted from clathrin-coated pits primarily by phosphatases, notably the synaptojanins, but possibly also PLC and type I or II PtdInsP3Ks. The synaptojanin family consists of two proteins, synaptojanin 1 and 2. Both splice variants of synaptojanin 1 are recruited to clathrin-coated pits and bind numerous effectors of the endocytic machinery (292). Endophilin is considered the major factor in recruiting synaptojanin 1 to forming pits (327). Synaptojanin 2B, a Rac1 effector (235), also localizes to the clathrin-coated pits (11, 276, 361) and its knockdown affects CME formation (312).

In addition to the synaptojanins, other 5′ phosphatases have been linked to CME formation. The lipid phosphatase OCRL is recruited to forming vesicles by AP2 and clathrin (237, 367) and another 5′ phosphatase, SHIP2, is recruited by intersectin (271). All of these phosphatases ensure that PtdIns(4,5)P₂ is depleted in a stringently controlled manner.

Unlike phagocytosis and macropinocytosis (described in detail below), where PLC activity and 3′ phosphorylation have been convincingly shown to contribute to the localized depletion of PtdIns(4,5)P₂, the contribution of these pathways is less clear for CME. Several receptors that are taken up via CME activate type I PtdInsP3K. Furthermore, type II PtdInsP3K is activated by clathrin and localizes to clathrin-coated pits (128, 407). Yet, it is generally agreed that inhibition of PtdInsP3K does not impair CME formation (344), although the isoform of type II PtdInsP3K that associates with clathrin, type II PtdInsP3Kα, is rather refractory to the conventional PtdInsP3K inhibitors (93). Evidence for CME-associated PLC activity is scant and CME is not associated with Ca²⁺ signaling. Nonetheless, certain isoforms of PLC are attracted and retained at the membrane by PtdIns(4,5)P₂ (72). Interestingly, they preferentially hydrolyze phosphoinositides in the context of small liposomes which have very positively curved membranes, like the membranes of late-stage clathrin-coated vesicles, where the phospholipid head-groups are easily accessible for modification (6). By acting selectively on curved membranes, the phospholipases could, in principle, promote scission and uncoating without interfering with the earlier PtdIns(4,5)P₂-dependent steps of nucleation and cargo selection that occur in the flatter clathrin-coated lattices and early pits.

Theoretical models provide evidence that the elimination of PtdIns(4,5)P₂ in positively curved regions of the membrane may in fact be sufficient to induce scission (222, 223). When PtdIns(4,5)P₂ is preferentially depleted from the invaginated region yet enriched in the neck region, a lipid phase boundary force is created at the neck-bud interface, which is sufficient to pinch off the vesicle. This holds true in an artificial system, where fission can be induced by recruitment of PtdIns(4,5)P₂-specific phosphatases to endophilin-associated tubules (62). Thus it is possible that the role of dynamin in fission is to deform the membrane sufficiently to produce differential phosphatase recruitment and formation of the lipid phase boundary. Indeed, the precise mechanism whereby dynamin mediates fission remains controversial (329, 343). In cell-free systems, dynamin induces membrane curvature (tubulation) (356) but does not cause fission per se until tension is applied to the membrane (311). Within cells, the forming endosome may reach a state where scission is ultimately caused by interfacial tension, attributable in turn to curvature-induced lipid modification. Circumstantial evidence to support this hypothesis includes the observation that dynamin dissociation and GTPase activity precede fission (30, 298). Furthermore, in yeast, CME is dynamin-independent (130), although a yet unidentified functional homolog could be fulfilling a similar role.

The generation of a lipid boundary force may be a conserved characteristic of plasmalemmal fission events. Indeed, several PtdIns(4,5)P₂-consuming enzymes are recruited to the cleavage furrow during cell division (33, 114, 178) and, as discussed below, lipases and phosphatases collaborate to eliminate PtdIns(4,5)P₂ from forming phagosomes and macropinosomes as well.

In the final step of CME, uncoating, the loss of PtdIns(4,5)P₂ causes a variety of proteins to dissociate from the newly formed vesicle. Predictably, those proteins that were recruited and retained by the phosphoinositide are released upon PtdIns(4,5)P₂ hydrolysis. In accordance with this notion, impairing PtdIns(4,5)P₂ depletion by silencing the responsible phosphatases prevents/delays uncoating of the clathrin-coated vesicle (82).

In summary, CME critically depends on the formation and depletion of PtdIns(4,5)P₂ for virtually all its stages: nucleation, cargo selection, coat assembly, scission, and uncoating. In contrast, much less is known about the role of other phospholipids in CME. PtdSer can interact with BAR domain-containing proteins and, being highly enriched in the inner leaflet of the plasmalemma, is likely to contribute to their recruitment during endocytosis. This, however, has not been demonstrated directly, since methods to selectively manipulate PtdSer content are lacking. Lastly, PtdOH generation by DGK was reported to be required for optimal EGFR internalization (10). This effect is thought to reflect a regulatory, rather than an essential role of PtdOH in CME.

2. Endosomal progression

Cargo sorting occurs at the time of CME and also during endosomal progression (FIGURE 2). Some of the internalized molecules are intended for degradation and must therefore be targeted to lysosomal compartments, while others need to be recycled. The latter include bystander, bulk membrane components that were inadvertently internalized along with the selected cargo. Careful routing of the individual components requires a precisely orchestrated series of events to direct their traffic from the first endocytic station, the sorting endosome, to their final destination. This involves a highly complex series of membrane fusion and fission reactions. Briefly, endocytic vesicles quickly merge with early endosomes, where active cargo sorting ensues. Whether sorting endosomes and other compartments of the endocytic pathway are permanent organelles that are maintained in a steady state by material in- and out-flux, or undergo maturation in a manner analogous to phagosomes (see below) is still a matter of debate (306). Some cargo is rapidly recycled from sorting endosomes to the plasma membrane by a vesicular pathway regulated by Rab4 (343a). Other recycling cargo takes a slower, more circuitous route, dwelling temporarily in a juxtanuclear population of Rab11-enriched vesicles, the recycling endosomes. Another small GTPase, Rab5, is required for progression of early endosomal material to degradative compartments (49). It enables the delivery of tubulovesicular structures to MVB/late endosomes by a process that involves also Rab7. Some of the endosomal material is delivered retrogradely to the TGN by a process requiring sorting nexins (55), while a Rab9-dependent pathway also fosters retrograde delivery from late endosomes (226). The final station in the endocytic pathway is the lysosome, a highly acidic, degradative organelle.

Because of its complexity, the endosomal progression sequence is, at present, only partially understood. It is clear, nevertheless that phospholipids play key functions at various stages of the process.

A) PtdIns(3)P IN ENDOSOMAL PROGRESSION.

PtdIns(3)P is the phospholipid most intimately associated with endosomal maturation (FIGURE 2). As previously mentioned, PtdIns(3)P can be synthesized by the class III PtdIns3K Vps34, which is inhibited by wortmannin, or by the wortmannin-insensitive class II PtdIns3K, and can also be formed through sequential hydrolysis of PtdIns(3,4,5)P₃ and PtdIns(3,4)P₂ by the appropriate phosphatases. The available evidence points to Vps34 as the predominant source of PtdIns(3)P in the endocytic pathway. It localizes to early endosomes together with its critical binding partner Vps15 (in yeast) or p150 (in humans) (395). Interestingly, two small GTPases known to direct endosomal traffic, Rab5 and Rab7, associate with Vps34 and promote its activity (269, 348).

Two well-defined protein domains have been identified that preferentially bind PtdIns(3)P, the FYVE and PX domains, and proteins involved in endosome maturation often contain one of these. FYVE domains are quite specific for PtdIns(3)P, while PX domains can be more promiscuous, occasionally preferring other phosphoinositides to PtdIns(3)P. The tethering molecule early endosome antigen 1 (EEA1) contains a FYVE domain, as well as NH₂- and COOH-terminal Rab5-binding domains, and mediates homotypic fusion of endosomes through its interactions with the endosomal SNARE proteins, syntaxin 6 and 13 (71, 246, 335, 336). Rabenosyn-5, a Rab5 effector, also has a FYVE domain and interacts with the Sec1p/Munc18-like protein Vps45, presumably to control SNARE-mediated fusion. Another FYVE domain-containing protein is Hrs or ESCRT-0. As its alternative name suggests, Hrs is crucial for initiating assembly of the ESCRT complex, a sophisticated multiprotein molecular machine that produces intraluminal vesicles (156), thereby internalizing ubiquitinated membrane-associated proteins for degradation.

PX domain-containing proteins include members of the sorting nexin (Snx) family, the yeast SNARE protein Vam7, and the kinesin KIF16B. The Snx family consists currently of 36 proteins reported to have a diversity of functions (85). Most relevant to the present discussion are Snx1, 2, 5, and 6, which contribute to form the retromer complex. The retromer consists of five proteins: Vps35, Vps26, Vps29, and two Snx subunits. These can be one copy of either Snx1 or -2 and one of either Snx5 or -6. The retromer complex plays an important role in retrieving acid hydrolases from endosomes back to the TGN. Silencing either Vps26 (18) or Snx1 and 2 (309) together causes the cation-independent mannose-6 phosphate receptor to remain within the endolysosomal system, where it is degraded instead of recycled. By binding to PtdIns(3)P, while at the same time inducing or stabilizing membrane curvature through their BAR domain, the Snx subunits of the retromer are thought to favor tubulation or scission of endosome-to-TGN carriers.

The degradation of PtdIns(3)P during endosome maturation occurs via three different pathways: dephosphorylation, phosphorylation, and hydrolysis. PtdIns(3)P is dephosphorylated by the myotubularins, which are active 3-phosphatases. Interestingly, the myotubularin MTM1 binds directly to the Vps34-Vps15 complex and competitively prevents its interaction with Rab5 and Rab7, suggesting an additional, phosphatase-independent mechanism of antagonizing PtdIns(3)P accumulation (22). PtdIns(3)P can also be eliminated upon phosphorylation by the 5-kinase Fab1/PtdInsK-FYVE, producing PtdIns(3,5)P₂. The latter is a marker of late endosomes, suggesting a sequential progression of phosphoinositides during endocytosis, with PtdIns(3)P in early compartments and PtdIns(3,5)P₂ in later ones. Finally, PtdIns(3)P can be degraded by lysosomal phospholipases. This is made possible by the fact that PtdIns(3)P is forced into the ILVs of MVB by the inward budding process mediated by the ESCRT complex.

PtdIns(3)P can be visualized with a genetically encoded probe that consists of a fluorescent protein tagged with one or more tandem FYVE domains (e.g., 2FYVE^EEA1-GFP) (132, 135) or a PX domain (PX^p40phox-GFP). Unfortunately, no selective inhibitor of class III PtdIns3K is commercially available (256), making it difficult to ascertain whether a regulatory mechanism is specific to PtdIns(3)P or affects multiple 3′ polyphosphoinositides. Alternative approaches to antagonizing PtdIns(3)P include silencing Vps34, introducing cytosolic antibodies to Vps34 (37), overexpressing a construct containing multiple FYVE domains (294), using inositol phosphates as competitive inhibitors (184), or inducibly recruiting myotubularins to PtdIns(3)P-containing compartments (115). In summary, PtdIns(3)P plays a critical role in the sorting and progression of cargo at the early stages of endocytosis by recruiting and retaining a variety of proteins bearing FYVE and PX domains. PtdIns(3)P subsequently disappears by a combination of dephosphorylation, phosphorylation to more complex species, and inward budding, signaling the transition to late endosomes.

B) BMP/LBPA IN ENDOSOMAL PROGRESSION.

Like PtdIns(3)P, BMP/LBPA is produced within maturing endosomes, although its cellular precursors are less well known. The functional relevance of BMP/LBPA was demonstrated using antibodies to this unique lipid; in cells allowed to internalize anti-BMP/LBPA antibodies by fluid-phase uptake late endosome traffic and morphology are altered (194). Cells that take up such antibodies accumulate cholesterol in their late endosomes (193), recapitulating the defects found in Niemann-Pick type C disease, a cholesterol-storage disorder. Conversely, adding exogenous BMP/LBPA fosters cholesterol removal from Niemann-Pick C fibroblasts (65). Sphingolipid hydrolysis during endosomal progression has also been shown to depend on BMP/LBPA (326).

BMP/LBPA mediates the formation of ILVs in a PtdIns(3)P-independent (and presumably ESCRT-independent) manner (FIGURE 2). In this regard, it is interesting to note that yeast cells do not possess BMP/LBPA, which may explain why the ESCRT complex is essential for ILV formation in yeast, while higher organism can form ILVs when the ESCRT machinery is depleted (304). In fact, BMP/LBPA can spontaneously form ILVs in pure lipid liposomes (241). No protein is necessary for this phenomenon, but it requires imposition of a pH difference between the lumen and the exterior of the liposome, in a direction that is analogous to the gradient that exists across the membrane of endosomes/lysosomes in living cells. The most prevalent form of BMP/LBPA has its two fatty acids in a thermodynamically unstable position. Steric effects cause the free ends of the fatty acids to move progressively apart, imparting a conical shape to the molecule. Although the bulk of BMP/LBPA is negatively charged at endosomal-like pH, the protonation of even a tiny subset of BMP/LBPA allows this wedge-shaped lipid to flip to the outer leaflet of the limiting membrane, where it deprotonates and therefore irreversibly accumulates, inducing invagination.

The invagination process in cells, however, may be drastically different, inasmuch as BMP/LBPA is not found on the limiting membrane of MVBs. Instead, the invagination process is likely regulated by proteins, such as Alix, a distant homolog of the yeast protein Vps31p (241). Alix preferentially binds liposomes containing BMP/LBPA, and knock-down of Alix reduces BMP/LBPA levels and the number of ILVs. Interestingly, and somewhat paradoxically, Alix is also a binding partner of ESCRT-I and a negative regulator of ILV formation in synthetic (pure lipid) membranes (110). The significance of these findings is not yet obvious.

C) PtdIns(3,5)P₂ IN ENDOSOMAL PROGRESSION.

PtdIns(3,5)P₂ is found on late endosomes and has been implicated in lysosomal traffic. As mentioned previously, it is synthesized from the early endosome lipid, PtdIns(3)P, by the 5′ kinase PtdInsK-FYVE/Fab1 (FIGURE 2). Its first identified role was in the regulation of MVB sorting (281). PtdIns(3,5)P₂ binds Vps24 of ESCRT III, and in yeast it also interacts with two related proteins, Ent3p and Ent5p, which are implicated in MVB sorting and Golgi-to-vacuole traffic (94). Additionally, PtdIns(3,5)P₂ has been suggested to play a role in the opposite pathway, recycling from the endolysosomal system to the TGN. In accordance with these proposed roles, knocking PtdInsK/Fab1 out (in yeast) or down (in mammalian cells) causes late endosomes and lysosomes to become enlarged. This is expected because PtdInsK/Fab1 serves in two mechanisms that remove membrane from these compartments, MVB formation and recycling to the TGN, while leaving the anterograde traffic that delivers membrane to these compartments unscathed. Furthermore, because PtdIns(3,5)P₂ affects the activity of ion channels, the enlarged compartments may represent an osmotically driven swelling response. A role for PtdIns(3,5)P₂ in osmoregulation is also suggested by the observation that hyperosmotic stress induces a massive increase in the content of this lipid.

C. Phagocytosis

Like CME, phagocytosis is a receptor-mediated internalization process. However, in contrast to CME, phagosome formation requires PtdInsP3K activity, actin polymerization, and Rho-family kinases. Cargo ubiquitylation and clathrin are not required for phagocytosis, and the dependence on dynamin is unclear (119).

Phagocytosis occurs when particles greater than 0.5 μm in diameter engage receptors on the surface of professional (monocytes/macrophages, neutrophils, dendritic cells) or nonprofessional phagocytes (e.g., fibroblasts, epithelial cells) (FIGURE 3). These phagocytic receptors recognize the particle either directly or indirectly through serum proteins that coat the surface of the particle. Such coat proteins, called opsonins, include the antibody immunoglobulin G (IgG) or the complement end-product C3bi. Under natural conditions, particles associate with more than one type of opsonin and engage a variety of both opsonic and nonopsonic phagocytic receptors. For simplicity, however, under laboratory conditions individual phagocytic receptors are often studied in isolation, using experimental paradigms where only a single receptor type is engaged. This facilitates dissection of the downstream mediators and effectors. Among these models, phagocytosis of IgG-coated particles by Fcγ receptors (FcγRs) is by far the most thoroughly analyzed. The information below is based largely on studies of this prototypical system.

During phagocytosis of opsonized particles, FcγRs cluster as they bind a particle coated with multiple IgG molecules. Lateral clustering by multivalent target particles is a sine qua non requirement for activation of FcγR and several other phagocytic receptors. Members of the FcγR family contain (or associate with proteins containing) the immunoreceptor tyrosine-based activation motif (ITAM) within their cytoplasmic tails. Upon clustering, ITAMs are phosphorylated on their dual Y-X-X-I/L motifs by Src-family tyrosine kinases, such as Hck, Lyn, and Fgr, and possibly other kinases (116, 134, 147, 385). Phosphorylation on both Y-X-X-I/L motifs that constitute the ITAM is necessary for optimal phagocytosis, likely to recruit the tyrosine kinase Syk, which contains two suitably spaced SH2 domains that selectively bind phosphotyrosine moieties. Syk is indispensable for FcγR-mediated phagocytosis; this was demonstrated using macrophages from syk^−/− mice, which are unable to ingest IgG-coated particles (83, 190). Once Syk is phosphorylated by Src-family kinases or by other Syk molecules, it recruits and/or phosphorylates a host of adaptor and effector proteins, such as the Grb2-associated-binding protein 2 (Gab2), the linker of activated T cells (LAT), class I PtdInsP3K, PLCγ, and Vav (363, 401).

Ultimately, this FcγR-initiated signaling cascade induces a biphasic actin remodeling response. Like construction scaffolding, actin filaments at the phagosomal cup are assembled and then disassembled as the pseudopod progresses around the particle. The spike in actin polymerization is tied to the recruitment of numerous guanine nucleotide-exchange factors (GEFs) such as Vav and DOCK180, which enlist the actin nucleator Arp2/3 via the Rho-family GTPases Rac and Cdc42 and the Wiskott-Aldrich syndrome family proteins WAVE and WASP (244). When actin polymerization is inhibited, the membrane cannot protrude to envelop the particle (17). Once the particle is surrounded by the phagocytic membrane, the opposite reaction is needed: disassembly of actin filament must occur, allowing the nascent phagosome to enter the cell, without being obstructed by the shell of cortical actin that normally lines the plasmalemma. The actors involved in actin depolymerization have not been studied directly. Likely candidates include actin-severing proteins and Rho family GTPase-activating proteins (GAPs). The sequence of actin polymerization and depolymerization must be timed accurately to ensure optimal engulfment, yet the timing mechanism(s) remain poorly understood. How phagosomes then separate from the plasmalemma is even less clear; virtually nothing is known about the phagosome scission process. Despite these gaping holes in our knowledge, it is becoming apparent that phospholipid metabolism is important at all phases of phagosome formation.

1. Phagosome formation

As described for CME, PtdIns(4,5)P₂ is an important determinant of phagocytosis (FIGURE 3). Mirroring actin dynamics, the abundance of PtdIns(4,5)P₂ changes as the pseudopod extends to surround the particle (39). Initially PtdIns(4,5)P₂ levels increase modestly as the membrane makes contact with the target and the nascent pseudopods contour the base of the target. Several activators of type I PtdInsP5K, the enzyme that makes the predominant pool of plasmalemmal PtdIns(4,5)P₂, are enriched or activated at the forming phagosome, notably Arf6, PtdOH, and Rac (31). PtdIns(4,5)P₂ appears to support its own buildup by stabilizing type I PtdInsP5K on the inner aspect of the plasmalemma (106). The kinase exposes a positively charged region on its membrane-interacting face, which associates electrostatically with the negative surface charge of the inner leaflet; together with PtdSer, PtdIns(4,5)P₂ is the main contributor to this electronegativity. Interestingly, different isoforms of type I PtdInsP5K have unique functions during phagocytosis. Data from knockout mice and transiently knocked-down cell lines suggest that PtdInsP5Kγ regulates actin remodeling during receptor clustering, whereas PtdInsP5Kα promotes actin polymerization during pseudopod extension (236). Regardless of the underlying mechanism, the local accumulation of PtdIns(4,5)P₂ is associated with, and most likely contributes to, the localized increase in actin polymerization at the base of the phagocytic cup (328). PtdIns(4,5)P₂ is known to activate actin nucleators and cross-linkers, while simultaneously inhibiting actin-severing and barbed end-capping proteins. Not surprisingly, depletion of PtdIns(4,5)P₂ reduces the phagocytic efficiency (77). Remarkably, retention of PtdIns(4,5)P₂ also reduces phagocytosis (106, 328). This seemingly paradoxical observation can be readily explained by the need to disassemble actin to complete particle engulfment. PtdIns(4,5)P₂ catabolism during phagosome formation and closure occurs by one of three pathways: hydrolysis by PLC, phosphorylation by PtdIns3K, or dephosphorylation by 4′ and 5′ phosphatases.

Hydrolysis of PtdIns(4,5)P₂ at the phagocytic cup is largely attributable to PLC (FIGURE 3). Both isoforms of PLCγ have been found to localize to the phagosomal cup (39). Their dual SH2 domains serve to recruit them to the phosphotyrosine residues of the activated adaptors, such as LAT. Once at the cup, PLCγ is activated via Syk-dependent phosphorylation and converts PtdIns(4,5)P₂ into DAG and IP₃. DAG causes recruitment of C1 domain-containing proteins, such as conventional and novel PKC isoforms, and IP₃ induces Ca²⁺ mobilization from the ER (206). Neither PKC activation nor Ca²⁺ mobilization is thought to be essential for particle ingestion, yet pharmacological inhibition of PLC activity ablates phagocytosis. The disappearance of the substrate PtdIns(4,5)P₂ may therefore be more important than the generation of the products.

PtdInsP3K also contributes to the elimination of PtdIns(4,5)P₂ from sites of phagocytosis, by phosphorylating its 3′ position to produce PtdIns(3,4,5)P₃ (FIGURE 3). Interestingly, inhibiting PtdInsP3K activity with wortmannin abolishes pseudopod extension, but not the formation of the phagocytic cup or the accumulation of actin at its base (17). In addition, as particle size decreases, phagocytosis becomes progressively less dependent on PtdInsP3K activity (379). Thus uptake of targets of diameter ≤1 μm is virtually insensitive to wortmannin! It is possible that large particles need both PLC and PtdInsP3K activity to eliminate PtdIns(4,5)P₂, while smaller particles can rely on PLC alone. Alternatively, and in our view more likely, products of PtdInsP3K activity may be required to recycle components of the actin machinery from the base of the cup to the tips of pseudopods and/or to facilitate the focal exocytosis of endomembranes that may be required to fully surround larger targets (24, 80). Neither of these reactions may be essential for entrapment of smaller particles.

PtdIns(3,4,5)P₃ accumulates markedly in nascent phagosomes. The accumulation seems to be confined to the cup, without spreading to the bulk (extraphagosomal) plasma membrane (239). How this restricted localization is attained is not clear; diffusional barriers or a source-sink mechanism may be at work. PtdIns(3,4,5)P₃ seems to play multiple roles at sites of phagocytosis: it recruits the motor protein myosin X (81), which has been speculated to promote phagosomal closure by a purse-string mechanism. In addition, PtdIns(3,4,5)P₃ helps stabilize several PLC isoforms at the membrane, by interacting with their PH domains (109). Finally, it generates a positive-feedback loop by interacting with Gab2 (142). This adaptor protein helps recruit the p85 regulatory subunit of PtdIns3K to the phagosomal cup; the PtdIns(3,4,5)P₃ synthesized by this kinase in turn helps stabilize Gab2, which contains a PH domain, at the cup.

The remnant PtdIns(4,5)P₂ is eliminated by 4′- and 5′-phosphatases. Several 5′-phosphatases are recruited to the phagosome. These include Inpp5B, Inpp5E, OCRL, and SHIP (7, 38, 165). PtdIns(4,5)P₂ is considered the preferred substrate of the Inpp5 isoforms and of OCRL, even though they dephosphorylate PtdIns(3,4,5)P₃ as well (283). In the case of SHIP, PtdIns(3,4,5)P₃ is the canonical substrate, but this phosphatase also displays activity against PtdIns(4,5)P₂, at least in vitro. Recent evidence indicates that knockdown of OCRL and Inpp5B impairs the disappearance PtdIns(4,5)P₂ at the phagosomal cup (38). Moreover, when recruitment of OCRL and Inpp5B to the membrane is prevented, invasion of host cells by the bacterium Yersinia pseudotuberculosis is inhibited. The (pre)vacuoles formed by Yersinia fail to seal, implying that hydrolysis of PtdIns(4,5)P₂ by the phosphatases is required for scission (320). It remains unclear whether the PtdIns(4)P generated by the 5′-phosphatases plays an active role in the sealing process.

The three pathways of PtdIns(4,5)P₂ catabolism combine to eliminate a major contributor to the negative surface charge of the inner leaflet of the membrane lining the engulfed particle. Among other effects, the reduced electronegativity weakens the association of PtdInsP5K with the membrane (106). Thus focal elimination of PtdIns(4,5)P₂ also indirectly reduces PtdIns(4,5)P₂ production.

Another important negatively charged phospholipid produced at sites of phagocytosis is PtdOH (322). Accumulation of PtdOH on nascent phagosomes was documented using a fluorescent probe, although caution should be exercised in interpreting these experiments (79), which used a domain derived from Raf1 that has proven unreliable. The major source of PtdOH is PLD activity, which cleaves the choline moiety off PtdCho. In phagocytes, PLD1 localizes to late endosomes/lysosomes. It accumulates on the forming phagosome, possibly via exocytosis directed to sites of ingestion, and persists during phagosome maturation (albeit these dynamics are more obvious for transfected GFP-PLD1 than for endogenous PLD1). PLD2, on the other hand, is constitutively at the plasmalemma and disappears once the phagosome seals. The motifs involved in targeting PLD1 and PLD2 to different compartments are unclear. PLD1 is palmitoylated and requires cofactors like Rho, Arf, and PKC for optimal activity; PtdIns(3)P has additionally been shown to affect both the localization and the activity of PLD1. In contrast, PLD2 has basal activity independently of any known cofactors. It can bind Grb2 and acts as a GEF for Rac2, an important regulator of the NADPH oxidase in phagocytes.

Early studies showed that pharmacological inhibition of both PLD isoforms blocks phagocytosis (203). This prompted the suggestion that PtdOH may be a key participant in phagocytosis by modulating Rac and PtdInsP5K activity. An effect of PtdOH on membrane curvature can also be envisaged. Subsequent experiments using dominant-negative PLD constructs (177) and siRNA-mediated silencing (79) also support a role for PLD isoforms in particle ingestion. On the other hand, neutrophils from knockout mice lacking PLD1 or PLD2 perform phagocytosis normally (279). Thus the requirement for PLD remains uncertain.

It is noteworthy that PtdOH could be conceivably generated also by DAG kinases. Considering the active generation of DAG at sites of phagocytosis, this alternative mechanism could contribute importantly to the accumulation of PtdOH, which may have complicated the assessment of the role of PLD and PtdOH in phagosome formation.

2. Phagosome maturation

In many ways, phagosome maturation parallels endosome progression. After sealing, the nascent phagosome undergoes a gradual metamorphosis, becoming first an early phagosome and then a late phagosome that subsequently merges with lysosomes, giving rise to the phagolysosome. The intermediate stages include the formation of ILV, rendering the late phagosome a type of MVB. Ubiquitinated cargo, including the activating FcγRs themselves, is thereby destined for degradation (208).

As during endosome progression, PtdIns(3)P and likely also BMP/LBPA and PtdIns(3,5)P₂ are involved in phagosome maturation (FIGURE 3). These shared features are not surprising, considering that the phagosome fuses with components of the endosomal compartment as it progresses to become a phagolysosome. However, phagocytosis has unique features. Because professional phagocytes such as macrophages and dendritic cells are also professional antigen-presenting cells, phagosome maturation in these cells has to contend with the added complexity of MHC II loading and delivery to the surface for presentation. Furthermore, these cells often take up toxins and pathogens, which need to be neutralized. As such, professional phagocytes contain specialized machinery to oxidize, permeabilize, and starve bacteria and other microbes of essential nutrients (117). These weapons are exceptionally effective but also dangerous, especially if deployed haphazardly against the host cells. To avoid unwanted damage, phagosomes have unique permeability and self-repair properties. In addition, professional phagocytes minimize the generation of potentially toxic microbicidal agents by differentiating dangerous intruders from innocuous cargo, such as apoptotic cells that are also internalized by phagocytosis (36).

PtdIns(3)P is a key marker and maturation determinant of the early phagosome. As in endosomes, it is important for fusion with other early compartments, for ESCRT assembly, and for retromer recruitment. Vps34, the class III PtdIns3K, is responsible for most, if not all, the PtdIns(3)P generated by phagosomes (379). Shortly after scission, and often even before sealing, phagosomes undergo fusion with early endosomes. This step appears to proceed normally in the absence of PtdIns(3)P, to the extent that Rab5 is still acquired by phagosomes when PtdIns3K inhibitors are added to the cells immediately after phagosome formation (380). In fact, addition of wortmannin or LY294002 causes phagosomes to retain Rab5 for inordinately long periods of time. Inhibitor-treated phagosomes also acquire (some) Rab7, yet are unable to fully mature to the phagolysosomal stage. Inhibiting PtdIns(3)P formation also prevents the recruitment of Hrs, the ESCRT-0 component, to the phagosome and presumably impairs the generation of (at least one type of) ILVs (381). PtdIns(3)P is also required to recruit the retromer complex to maturing phagosomes. Although its function has not been explored in mammalian phagosomes, studies in Caenorhabditis elegans implicate the retromer in retrieval of the phagocytic receptor, CED-1, which is associated with clearance of apoptotic cells (64).

Most of the above functions of PtdIns(3)P in phagosome maturation are equivalent to those it fulfills during endocytosis. In addition, PtdIns(3)P has other, specialized responsibilities during phagocytosis. First, it is essential for the generation of ROS within maturing phagosomes. Recruitment of p40^phox, one of the soluble subunits of the NADPH oxidase complex, is required for phagosomes to generate superoxide; this is accomplished by association of the PX domain of p40^phox with phagosomal PtdIns(3)P (185, 404). Moreover, some types of phagocytosis, especially of bacteria, are accompanied by recruitment of components of the autophagy system to the phagosome (35). The precise role of this recruitment is not entirely clear, but it may serve to repair damage induced by the pathogens or incurred in the process of generating microbicidal products. While its function remains to be defined, we do know that delivery of autophagic components to phagosomes depends on PtdIns(3)P.

Less is known about the contribution of PtdIns(3,5)P₂ and BMP/LBPA to phagosome maturation. BMP/LBPA has been detected on the late phagosome (381), but its exact role is undefined. Although extrapolation is risky, it can be tentatively postulated, by analogy with endocytosis, that BMP/LBPA is involved in ILV formation. PtdIns(3,5)P₂ has not been reported on phagosomes. Nevertheless, because PtdIns(3)P is the main determinant of PtdInsK-FYVE recruitment and this enzyme is ubiquitously expressed, it is safe to assume that PtdInsK-FYVE and PtdIns(3,5)P₂ will be proven to exist on maturing phagosomes as well.

Finally, PtdSer, which is abundant in the plasma membrane, is also present on maturing phagosomes. While its content has not been analyzed rigorously by chemical means at the individual stages of the process, recent experiments using a PtdSer-specific fluorescent probe demonstrated that PtdSer is present on the cytoplasmic leaflet throughout the maturation sequence, and that its concentration does not seem to vary greatly, at least within the first 30–45 min (396). The persistence of PtdSer on the phagosomal membrane is anticipated to allow docking of proteins bearing C2 domains and to confer some electronegativity to its cytosolic aspect, which should in turn affect its ability to recruit and retain polycationic proteins. Notably, phagosomes containing the pathogens Legionella pneumophila and Chlamydia trachomatis, which divert traffic from the endocytic to the ER and Golgi pathways, respectively, are devoid of PtdSer, reflecting the composition of the destination compartments (398).

Often microbes actively manipulate phospholipid signaling as a means to establish a chronic infection. Microbes that infect phagocytic cells employ one of three strategies to avoid the degradative environment of the phagolysosome: arresting phagosome maturation, escaping the phagosome, or remodeling the phagosome into a different, novel compartment (FIGURE 3). Mycobacterium tuberculosis uses the former strategy and inhibits phagosome maturation. A glycosylated PtdIns, mannose-capped lipoarabinomannan, found in Mycobacterium tuberculosis has been shown to block PtdIns(3)P-dependent trafficking (123), ostensibly by inhibiting class III PtdIns3K. Listeria monocytogenes, on the other hand, relies on the second strategy and escapes the phagosome. The bacterial PtdIns-specific PLC abets this escape by producing DAG, which recruits factors that promote permeabilization of the phagosome (295). Finally, Legionella pneumophila exploits the third strategy by remodeling the phagosome into a new compartment. It uses PtdIns(4)P to anchor its protein SidC, which aids in recruiting vesicles derived from the rough endoplasmic reticulum (301).

D. Macropinocytosis

Macropinocytosis shares some of the features of both phagocytosis and CME. Like CME, macropinocytosis generates fluid-filled vacuoles but, similar to phagocytosis, it requires large-scale actin remodeling and PtdInsP3K activity (17). It is clathrin-independent, produces vesicles of heterogeneous size (from 0.2 to 5 μm), and is preceded by membrane ruffling (FIGURE 4). Simplistically, macropinosomes form as membrane extensions flop into each other or fold onto the plasma membrane, undergoing fusion and thereby creating large fluid-filled vacuoles that pinch off the surface membrane. Some cells, like immature dendritic cells and macrophages, undergo constitutive ruffling as a means of sampling the environment for antigenic determinants (318). Once internalized, these soluble factors are processed and loaded onto the class II major histocompatibility complex (FIGURE 4), which is then diverted to the cell surface where antigens are presented to cells of the adaptive immune system. Other cells require specific initiators to induce macropinocytosis. Examples of macropinocytic inducers include growth factors, phorbol esters, and chemotactic molecules (354). Genetically, introduction of dominant-active alleles of Ras, Src, or Rac isoforms can force cells to undergo constitutive macropinocytosis in the absence of any external stimuli. Due to the ubiquitous ability of cells to form macropinosomes, several intracellular pathogens, such as Salmonella and some viruses, exploit this pathway to gain entry into host cells.

A general pathway leading to macropinocytosis (355) can be inferred from the mode of action of the known initiators: activation of receptor tyrosine kinases fosters the recruitment of SH2 domain-containing adaptors like Grb2, which in turn activate Ras through the appropriate exchange factors. Ras stimulation leads to activation of class I PtdIns3K via its p85 subunit and consequent activation of Rac and Cdc42. The ensuing cycles of actin polymerization and depolymerization underpin the observed ruffling. Different inducers of macropinocytosis seem to enter this pathway at different stages. Growth factors stimulate the pathway from the top, while phorbol esters and constitutively active Rac affect the pathway further downstream. Care must be taken to distinguish classical macropinocytosis, that occurs at the cell periphery, from circular dorsal ruffles or waves that occur on the dorsal surface of the cell (92). These morphologically similar but spatially distinct pathways tend to rely on different signaling pathways, even though they are both stimulated by growth factors.

Macropinocytosis is often diagnosed based on its sensitivity to PtdInsP3K inhibitors or to actin perturbants like the cytochalasins and latrunculins. These properties, however, are shared with phagocytosis and with some forms of endocytosis. An additional criterion has been widely applied to identify macropinocytosis, namely, its susceptibility to inhibition by amiloride. Amiloride is a potent inhibitor of Na⁺/H⁺ exchange and in all likelihood inhibits macropinocytosis by compromising the regulation of the intracellular pH (196). As such, this drug can inhibit not only macropinocytosis, but a plethora of other cellular responses as well, and must therefore be used with extreme caution.

1. Macropinosome formation

Our understanding of the role of phospholipids in macropinocytosis is incomplete and confounded by the use of different induction conditions. Nevertheless, several generalizations can be made, particularly with regard to growth factor-induced macropinocytosis, which is more commonly studied than constitutive macropinocytosis.

In a manner analogous to that described for phagocytosis, PtdIns(4,5)P₂ undergoes localized biphasic changes, initially increasing during the ruffling stage and disappearing at the time of macropinosome closure (FIGURE 4). The exact factors underlying the initial rise in PtdIns(4,5)P₂ are unknown, but the acknowledged activators of PtdInsP5K (Rac, PtdOH, and Arf6) are detected at sites of macropinocytosis and are therefore likely candidates (139, 144). Disappearance of PtdIns(4,5)P₂ depends on PtdInsP3K and PLC activity. Phosphoinositide phosphatases have been detected in sealed macropinosomes (353), but their participation in PtdIns(4,5)P₂ hydrolysis has not been formally established.

Inhibition of PLC abolishes both ruffling and macropinocytosis (9). Calcium released from the ER by IP₃ does not appear to play an essential role; DAG, in contrast, seems to be critical. In fact, addition of exogenous DAG or of a DAG mimetic, such as phorbol 12-myristate 13-acetate, is sufficient to elicit macropinocytosis (354). Notably, the most common targets of DAG, the conventional and novel PKC isoforms, have an ambiguous role in macropinocytosis, since only some of their inhibitors have an effect (187, 321).

Type I PtdInsP3Ks are present and active at sites of macropinocytosis. PtdInsP3K inhibition prevents both “spontaneous” macropinocytosis in macrophages and oncogene-transformed fibroblasts, and also growth factor-induced macropinocytosis. Interestingly, the ruffling that precedes and is required for macropinosome formation is inhibited by PtdInsP3K antagonists only in some cases (118), but not in others (390). This suggests that random contacts between ruffles are insufficient for membranes to merge into macropinosomes, and that another, PtdInsP3K-dependent process is involved.

Unlike phagocytosis (380), macropinocytosis is profoundly depressed when the small GTPase Rab5 is inhibited (215). Expression of a dominant-negative allele of Rab5 consistently inhibits macropinosome formation, although the underlying mechanism is not known. It is tempting to speculate that Rab5 functions by recruiting 5′-phosphatases to complete the hydrolysis of PtdIns(4,5)P₂, thereby enabling scission.

2. Macropinosome progression

Less is known about the maturation of macropinosomes than about their formation (FIGURE 4). They are initially very similar to the plasmalemma, inasmuch as they do not concentrate receptors when they form, unlike phagocytosis and CME. The fate of the nascent macropinosome varies, depending on the cell type. In A431 and 3T3 cells, they tend to recycle back to the surface (188). In other cells, macropinosomes undergo traditional maturation and fuse with the lysosome. As they mature, they progressively shrink; this is unlikely to reflect osmotic changes, but rather the occurrence of net membrane fission. However, the size of macropinosomes occasionally increases as they fuse with one another. Thus macropinosomes possess the machinery for both homotypic fusion and sorting/recycling. By homology with other endocytic compartments, these processes likely require PtdIns(3)P. Accordingly, several PtdIns(3)P-associated Snx proteins, including those associated with the retromer complex (Snx1 and Snx5), are recruited to the maturing macropinosome (386). PtdInsK-FYVE is also recruited to maturing macropinosomes and likely generates PtdIns(3,5)P₂. Overexpression of a catalytically-inactive mutant of PtdInsK-FYVE or incubation with a PtdInsK-FYVE inhibitor prevents fusion between macropinosomes and lysosomes or late endosomes (189). Beyond this, little else is known about the fate of macropinosomes. Clearly, much remains to be learned.

VII. CONCLUDING REMARKS

For too long, phospholipids were regarded largely as an inert solvent where membrane proteins were embedded and traveled laterally. It is now glaringly apparent that lipids in general, and phospholipids in particular, are active and very dynamic partners in virtually all aspects of membrane function, from signaling to cytoskeletal assembly. In particular, phospholipids are key to membrane traffic. The relationship is bidirectional: phospholipids control membrane fusion and fission, while membrane trafficking affects the distribution of phospholipids–the conductor travels with the train. Despite fluctuations due to membrane turnover and the concomitant metabolism of a fraction of lipids during signaling, the phospholipid composition of nonreplicating membranes remains surprisingly stable within narrow limits. It is this illusion of stasis that obscured the functionality of lipids for decades.

Technical limitations have also contributed to our ignorance of phospholipid biology and, to a large extent, these limitations persist. While fluorescent probes have been designed and implemented for the study of some lipids in their native environment, this applies to only a handful of lipid classes. For the remaining lipids, detection is limited to mass methods that require extraction followed by analysis, precluding spatial (subcellular) analysis and severely limiting the study of minor, evanescent species. These limitations were inevitably reflected in the scope of this review.

Much remains to be learned about the dynamic functional roles of lipids in biological membranes, and the rate of progress will be dictated by the evolution of techniques to visualize, quantify, and manipulate individual lipid species. We foresee further development of fluorescently tagged lipid-specific probes that can be encoded genetically or otherwise introduced into live cells, and of new and better head-group-specific antibodies. In combination with emerging superresolution microscopy techniques, these beacons will reveal much about lipid distribution and dynamics. The increasing sensitivity and flexibility of mass spectrometric methods will facilitate analysis of minor species and will further inform of the length and saturation of the acyl chains, variables that profoundly impact the biological function of phospholipids. In addition, (over)expression of lipid-modifying enzymes, or their selective inhibition by knockout or gene silencing strategies, together with the use of new pharmacological agents will provide key information of lipid function.

The current intractability of lipids, however, should neither distract us from their importance nor deter us from studying them in much greater depth.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.