SEARCH

SEARCH BY CITATION

Keywords:

  • DAG;
  • membrane fusion;
  • membrane traffic;
  • PA;
  • phosphatidylinositides;
  • PLD

Abstract

  1. Top of page
  2. Abstract
  3. Mammalian PC-Selective PLDs
  4. Lipid Products of PLD Activity
  5. PA Enhances Membrane Fusion
  6. PA as a Precursor for DAG
  7. PLDs Act as GAPs for Dynamin
  8. Concluding Remarks
  9. References

The two mammalian phosphatidylcholine (PC)-selective phospholipase D (PLD) enzymes remove the choline head group from PC to produce phosphatidic acid (PA). PA stimulates phosphatidylinositol(4)phosphate 5-kinases, can function as a binding site for membrane proteins, is required for certain membrane fusion or fission events and is an important precursor for the production of diacylglycerol (DAG). Both PA and DAG are lipids that favor negatively curved membranes rather than planar bilayers and can reduce the energetic barrier to membrane fission and fusion. Recent data provide a mechanistic explanation for the role PLDs play in some aspects of membrane traffic and provide an explanation for why some membrane fusion reactions require PA and some do not. PLDs also act as guanosine triphosphatase-activating proteins for dynamin and may participate with dynamin in the process of vesicle fission.

Awareness that mammalian phospholipase D (PLD) enzymes might play some role in membrane traffic arose 15 years ago when the important regulator of membrane traffic, Arf1, was discovered to be a potent activator of PLD (1,2). At the same time, phosphatidylinositides were being recognized as required for many aspects of membrane traffic (reviewed in 3), and a complex positive feedback relationship between PLDs and phosphatidylinositol kinases was discovered (4–7). Phosphatidylinositol(4,5)bisphosphate (PIP2) was found to activate PLDs and the product of PLDs, phosphatidic acid (PA), and to activate type I phosphatidylinositol 5-kinases, which produce PIP2. Over the next decade, evidence that PLDs could affect various membrane processes was obtained (Table 1), but the mechanisms by which this occurred were obscure. In particular, it was not clear if PLD activity was required for membrane traffic or simply altered rates of membrane traffic in response to signal transduction events. Early on, it was hypothesized that PA functions to induce or facilitate negative curvature of membranes, acts as an allosteric regulator of proteins important for membrane traffic, serves as a binding site for proteins and recruits them to membranes, and can be a precursor for production of diacylglycerol (DAG) for any of these processes. Evidence supporting all four of these roles has been obtained, and in addition, PLDs have been found to bind to dynamin and accelerate its guanosine triphosphatase (GTPase) activity (8). This article will focus on recent studies that provide particularly interesting insight into the molecular mechanisms by which PLDs function in membrane traffic and how these specific instances might be generalized to explain the different requirements for lipid-modifying enzymes in different cell types. For a review of earlier research into the role of PLDs in membrane traffic, see Freyberg et al. (9).

Table 1.  PLD location, function and interactionsa
PLDCellular locationTraffic events affectedMolecular interactions
  • ERGIC, the ER–Golgi intermediate compartment; Glut4, insulin responsive glucose transporter 4; PKC, protein kinase C; TGN, trans Golgi Network; TfR, transferrin receptor; GPCR, G protein coupled receptor.

  • a

    PLD1/2 indicates no distinction between PLDs. Molecular interactions are supported by biochemical and cell biological data except where noted as genetic interactions.

PLD1TGN (57)Vesicle budding (57)Actin (62,63)
Golgi and endosomes (20,21)TfR recycling (32)Gβγ(64)
Exocytic vesicles (58)EGFR endocytosis (8)Dynamin (8)
PM (26,58–60)Phagocytosis (62)RalA (60)
ERGIC (61)Exocytosis (26,58–60)SCAMP2 (58)
 Glut4 vesicle fusion (27)Amphiphysins (65)
 TGN-to-PM transport (45)PKC, Arf1 and Rho (17)
 AP180 (66)
PLD2PM (18,23,30,67)GPCR endocytosis (29,30)Gβγ(64)
Golgi (45,68)Phagocytosis (62,69)Dynamin (8)
 MOR1 (70)
 PKC, Arf, Rho and Rac (18)
 Grb2 (67)
PLD1/2 Macropinocytosis (71) 
Dicty PLD Actin-based motility (72) 
SPO14 Prospore membrane fusion (38,40,73)GCS1 (genetic) (74)
 Spo20p (genetic) (75)
 Sma1p (76)
MitoPLDMitochondria outer membrane (41)Mitochondria outer membrane fusion (41) 

Mammalian PC-Selective PLDs

  1. Top of page
  2. Abstract
  3. Mammalian PC-Selective PLDs
  4. Lipid Products of PLD Activity
  5. PA Enhances Membrane Fusion
  6. PA as a Precursor for DAG
  7. PLDs Act as GAPs for Dynamin
  8. Concluding Remarks
  9. References

Two mammalian PLD enzymes have been identified that convert the abundant membrane lipid phosphatidylcholine (PC) to PA (10–12). Both PLD genes are differentially spliced to produce several PLD isoforms. Both enzymes have a modular structure, containing N-terminal lipid and protein-binding PH and PX domains followed by a domain containing two catalytic motif sequences, each forming half of the catalytic site (Figure 1). The PH domain is important for membrane localization and binds PIP2 but is not the binding site for PIP2 that activates the enzyme (13–15). A basic sequence in the catalytic domain binds PIP2 tightly, and this enhances catalytic activity (15). The PX domain accelerates GTP hydrolysis on dynamin (8). Each PLD is activated and inhibited through binding multiple other proteins, many of which are important for membrane traffic and/or organization of the cytoskeleton (Table 1). PLDs are also phosphorylated and palmitoylated (10,16). PLD1 has very low constitutive activity and appears to adopt an inactive conformation until bound by its activators (17). PLD2 has measurable basal activity that can be augmented by activators (18). Both enzymes have been reported on the plasma membrane (PM) and on internal vesicles (18–23), although PLD2 appears less abundant at internal sites than PLD1. Establishing the location of the endogenous enzymes has proven difficult as they are not abundant and many experiments have employed overexpressed tagged proteins that might not reflect the location of endogenous proteins. It is very likely that the location of PLDs is dynamic and sensitive to the growth and signal transduction state of cells.

image

Figure 1. Domain organization of PLDs. PLD1 and 2 contain N-terminal PX domains followed closely by PH domains. The PX domain binds to phosphatidylinositol 3,4,5 trisphosphate and serves as a GAP for dynamin. The PH domain binds to PIP2 for the purpose of helping localize the protein to membranes. A basic sequence in the second PLDc catalytic domain binds PIP2 that modulates enzyme activity. PLDs are activated by protein kinase C (PKC), Arfs and Rho family small GTPases and the regions where they bind are indicated.

Download figure to PowerPoint

Lipid Products of PLD Activity

  1. Top of page
  2. Abstract
  3. Mammalian PC-Selective PLDs
  4. Lipid Products of PLD Activity
  5. PA Enhances Membrane Fusion
  6. PA as a Precursor for DAG
  7. PLDs Act as GAPs for Dynamin
  8. Concluding Remarks
  9. References

The physical characteristics of the lipids produced by PLDs are important for understanding their possible functions in membrane traffic. The most abundant forms of the PC substrate for PLD in mammalian cells have one saturated and one monounsaturated acyl chain. The kink produced by the trans unsaturated bond expands the hydrophobic cross section of this form of PC considerably compared with the same lipid with two saturated acyl chains. When PLD removes the bulky choline head group, the resulting PA has a conical shape that packs poorly in a planar bilayer and well in membranes with negative curvature, as would be found at the neck of a vesicle that is in the process of budding from a donor or fusing to an acceptor membrane (Figure 2). In addition, a patch of PA on a planar membrane would produce packing defects that could facilitate the insertion of hydrophobic amino acids into the membrane as well as negative charges that could interact with basic residues, forming a distinct lipid-binding site for certain proteins. Conversion of PA to DAG by PA hydrolases that remove the negatively charged phosphate would accentuate the hydrophobicity of the membrane surface and its stability in areas of negative membrane curvature. In contrast to the DAG derived from PA, DAG produced by phosphoinositol-specific phospholipase C has unsaturated acyl chains that pack better in planar bilayers. Thus, theoretically, proteins that bind membranes in response to DAG could prefer DAG produced by one pathway over another, and there is evidence that this does occur for the binding of ArfGAP1 to membranes (24,25).

image

Figure 2. Preference of products of PLD for curved membranes. A) Lipids in which the hydrophobic cross section of acyl chains (pink cylinders) matches the hydrophilic cross section of the polar head group (blue cylinders) are stable in planar membranes but do not pack well when membranes are curved. PC with saturated acyl chains is an example. The preferred substrate for PLD, PC with one monounsaturated acyl chain, has a larger hydrophobic than polar cross section but packs well in membranes containing cholesterol. The product of PLD action, PA, has a small polar head group and is more stable in a membrane leaflet with negative curvature. PA hydrolases remove the phosphate head group from PA to produce DAG, which has a smaller polar head group. Phospholipase A enzymes can convert PA to monoacylated lysophosphatidic acid, which prefers membranes with positive curvature. By interconverting lipids, lipid-modifying enzymes can reduce the energy barrier for curving membranes. B) The membrane stalk that forms when vesicles either bud from a donor membrane or fuse to an acceptor membrane is diagramed. PA is shown in blue facilitating negative curvature in one membrane leaflet in the stalk, and lysophosphatidic acid in red facilitates positive curvature in the opposite leaflet.

Download figure to PowerPoint

PA Enhances Membrane Fusion

  1. Top of page
  2. Abstract
  3. Mammalian PC-Selective PLDs
  4. Lipid Products of PLD Activity
  5. PA Enhances Membrane Fusion
  6. PA as a Precursor for DAG
  7. PLDs Act as GAPs for Dynamin
  8. Concluding Remarks
  9. References

PLD at exocytic sites

A number of studies indicate that PLD1 and, to a lesser extent, PLD2 play a role in regulated exocytosis (26,27) and in neurotransmitter release (28). Both PLD1 and PLD2 also are involved in the endocytosis of certain signaling receptors (29–31). In contrast, PLD2, but not PLD1, modulates the recycling of transferrin receptors after endocytosis in Hela cells (32). Recent work by Vitale and colleagues provides insight into the mechanism by which these enzymes might contribute to various pathways of membrane traffic (33). Using a combination of techniques, they showed that PA is present on the PM and on secretory granule membranes located very near the PM and that depleting PLD1 by RNA interference inhibited stimulated exocytosis of chromaffin granules by reducing the number of fusion-competent granules without reducing the initial release kinetics. Adding a lipid that favors positive membrane curvature to the outer leaflet of the PM rescued the defect caused by reducing PLD1 expression. One interpretation of these observations is that production of PA by PLD1 at exocytic sites enhances membrane fusion by reducing the barrier for producing the negative membrane curvature required for a vesicle to pass into a hemifused state that has been postulated to be an intermediate on the pathway to exocytosis (34,35).

Yeast PLD

Additional evidence for a role of PLD in membrane fusion has been found using Saccharomyces cerevisiae. The yeast PLD, Spo14p, is not required for vegetative growth or membrane traffic, except in mutants deficient in the cytosine diphosphate-choline (CDP-choline) pathway for lipid biosynthesis, which normally supplies DAG when nutrients are abundant (36). However, under starvation conditions, Spo14p is required to form the prospore membranes to which all post-Golgi traffic is diverted (37). S. cerevisiae expresses two SNAP25 members, Sec9p and Spo20p, that interact with Sso1/2p and Snc1/2p to mediate fusion of post-Golgi transport intermediates with target membranes. Sec9p functions for fusion at the PM, and Spo20p replaces Sec9p for the formation of prospore membranes and requires PA to locate properly (38). Using liposomes of defined lipid and protein composition, McNew and colleagues have shown that Spo20p forms a less efficient fusion complex than Sec9p and that adding a minor fraction of PA to liposomes increases the fusion activity of Spo20p to equal that of Sec9p (39). Additionally, mutation of juxtamembrane basic amino acids that prevent membrane fusion in vivo(40) has no effect on Spo20p-mediated fusion of liposomes lacking PA but prevents the improvement of fusion when PA is added. These observations suggest that PA helps Spo20p form the acute negative membrane curvature that is a necessary intermediate for membrane fusion. The juxtamembrane basic residues of Spo20p bind PA to increase its local concentration where it is needed (39). As these authors point out, it is possible that in other membrane traffic events known to be influenced by PLD, the requirement for PA is to lower the energy barrier to fusion. Just as in the case of yeast, where Sec9p forms a more energetically stable SNARE bundle that does not need PA to induce membrane fusion, the requirement for PA or other lipids like DAG might depend upon the energetics of a particular membrane fusion machine. This could differ at different membrane surfaces or even in the same membrane traffic step in different cell types, depending upon the isotype of a fusion protein that is expressed.

MitoPLD

A recent report investigating a different member of the PLD superfamily reinforces this theme. Using bioinformatics approaches, Frohman and colleagues identified a protein with a predicted mitochondrial import sequence and a single PLD catalytic motif (41). This MitoPLD protein was found on the outer mitochondrial membrane as a homodimer, as one would expect for a monomer carrying only half the active site. MitoPLD retained its mitochondrial import sequence, probably for use as a membrane anchor. A model of the structure of MitoPLD based on a related bacterial PLD family member predicted that the catalytic site of MitoPLD would face away from the membrane surface to which it was anchored. Overexpression of the wild-type protein resulted in aggregation of all mitochondria in the cell. Eliminating MitoPLD by small interfering RNA (siRNA), or the overexpression of a catalytically dead enzyme, resulted in fragmented mitochondria. Epistasis experiments indicated that MitoPLD acts downstream of the tethering of mitochondria by the fusion protein Mitofusin. The substrate for MitoPLD was shown to be cardiolipin and its products PA. Taken together, the data suggest that MitoPLD produces PA on the opposed mitochondrial membrane surface brought close by Mitfusin. Whether the product PA functions in the process of membrane curvature required for membrane fusion or acts allosterically to enhance the action of proteins involved in fusion remains to be discovered.

PA as a Precursor for DAG

  1. Top of page
  2. Abstract
  3. Mammalian PC-Selective PLDs
  4. Lipid Products of PLD Activity
  5. PA Enhances Membrane Fusion
  6. PA as a Precursor for DAG
  7. PLDs Act as GAPs for Dynamin
  8. Concluding Remarks
  9. References

Early observations that mammalian PLDs can stimulate coat protein I (COPI) coat formation in vitro and that PLDs are required for membrane traffic events in the secretory pathway (42–45) were not confirmed by siRNA experiments knocking down PLDs. However, the amount of time (usually several days) required for reduction of protein activity by siRNA allows cells to compensate for the loss of one protein by increasing the expression of another, and this possibility severely limits the interpretation of negative results obtained with siRNA. Much of the PA produced in cells is rapidly converted to DAG, and as demonstrated by Bankaitis and colleagues in yeast and Malhotra and colleagues in mammalian cells (46–48), DAG is required for membrane traffic in the secretory pathway and can be provided by several distinctly regulated pathways. Insight into a mechanism by which DAG might function in membrane traffic was provided recently by Egea and colleagues (25), who showed that DAG is required for the formation of COPI vesicles for retrograde transport between the Golgi and the endoplasmic reticulum (ER). Treating cells with propranolol to inhibit PA hydrolase or U73122 to inhibit phosphoinositol-specific phospholipase C decreased Golgi DAG by 65% and caused COPI-coated vesicles to be frozen in a late stage of budding before they formed the constricted membrane necks prior to budding from the Golgi complex. This suggests that Golgi DAG levels are normally maintained by at least two pathways using two precursor lipids, PA and PI. ArfGAP1 is a GTPase-accelerating protein for Arf1 that functions in the secretory pathway between the ER and the Golgi complex (49–51). It is also an Arf effector required for the budding of COPI vesicles (52,53), as is its ortholog in yeast (54). ArfGAP1 binds to membranes enriched in DAG with monounsaturated acyl chains (55), the type of DAG that results from PA hydrolases acting on PA produced by PLD. ArfGAP1 binds to membranes by inserting large hydrophobic residues into the membrane that either stabilize or generate negative curvature (24,56). Propranolol decreased the amount of ArfGAP1 that bound to the Golgi (25), suggesting that a critical role for the DAG derived from PA is recruit ArfGAP1 for COPI vesicle budding.

PLDs Act as GAPs for Dynamin

  1. Top of page
  2. Abstract
  3. Mammalian PC-Selective PLDs
  4. Lipid Products of PLD Activity
  5. PA Enhances Membrane Fusion
  6. PA as a Precursor for DAG
  7. PLDs Act as GAPs for Dynamin
  8. Concluding Remarks
  9. References

Recent, intriguing observations by Ryu and colleagues show that PLD proteins have functions in addition to their lipase activities. Both PLD1 and PLD2 are modular proteins and bind to both membranes and other proteins. The PX domains of PLD1 and 2 were found to stimulate dynamin GTPase activity in a classic, arginine finger-dependent manner, and expressing the PX domain alone accelerated internalization of the epidermal growth factor receptor (EGFR) (8). Previously, PLD enzymatic activities had been shown to affect the rate of EGFR internalization (29). Ryu and colleagues showed that at low EGF concentrations, catalytically dead PLDs accelerated EGFR internalization but at high EGFR concentration, the enzyme activity was also involved. At low EGF concentration, siRNA of PLDs decreased EGFR internalization, and this was rescued by transfecting the cells with plasmids expressing siRNA-resistant lipase-deficient PLDs but not PLDs lacking the arginine finger (8). Although in these studies both PLD1 and 2 were equally active, the two enzymes are not usually redundant in membrane traffic events. It is possible that overexpressing the enzymes or their PX domains masks differences in membrane or protein binding that in vivo might cause the two enzymes to interact with dynamin in different places or under different conditions.

Concluding Remarks

  1. Top of page
  2. Abstract
  3. Mammalian PC-Selective PLDs
  4. Lipid Products of PLD Activity
  5. PA Enhances Membrane Fusion
  6. PA as a Precursor for DAG
  7. PLDs Act as GAPs for Dynamin
  8. Concluding Remarks
  9. References

One of the more interesting questions remaining to be answered is why lipid-modifying enzymes, such as PLD1 and PLD2, appear to be required for certain membrane traffic events but not others that involve an essentially identical physical process. One possibility often mentioned is that they play a regulatory role rather than being part of the core machinery. Another possibility, supported by the recent report by McNew and colleagues (39), suggests that protein machines involved in membrane traffic may differ importantly in the efficiency with which they function such that in some cases, modification of membrane lipids is required to attain biologically significant rates of activity, whereas other analogous protein assemblies do not require the help. A third possibility is that the modification of lipids for processes of membrane coat formation, fission and fusion is part of the core process, but the nature of the modification and thus, the enzymes required, differs according to the specific proteins and lipids involved. For example, lipids that favor negative curvature may always be required to form the stalk intermediate required for membrane fission and fusion, but whether the lipid is PA or DAG or some more exotic lipid, and whether this lipid is generated acutely at the site where it is needed or diffuses into the site and is concentrated, will depend upon the nature of the membrane where the event occurs.

References

  1. Top of page
  2. Abstract
  3. Mammalian PC-Selective PLDs
  4. Lipid Products of PLD Activity
  5. PA Enhances Membrane Fusion
  6. PA as a Precursor for DAG
  7. PLDs Act as GAPs for Dynamin
  8. Concluding Remarks
  9. References