Journal list menu

Volume 593, Issue 17 p. 2428-2451
Review Article
Free Access

Phosphatidic acid in membrane rearrangements

Mikhail A. Zhukovsky

Corresponding Author

Mikhail A. Zhukovsky

Institute of Protein Biochemistry and Institute of Biochemistry and Cell Biology, National Research Council, Naples, Italy

Correspondence

C. Valente, D. Corda, or M. A. Zhukovsky, Institute of Protein Biochemistry and Institute of Biochemistry and Cell Biology, National Research Council, Via Pietro Castellino 111, Naples 80131, Italy

Tel: +39 081 6132225 (CV); +39 081 6132543 (MAZ); +39 081 6132536 (DC)

E-mails: [email protected] (CV), [email protected] (MAZ), [email protected] (DC)

Search for more papers by this author
Angela Filograna

Angela Filograna

Institute of Protein Biochemistry and Institute of Biochemistry and Cell Biology, National Research Council, Naples, Italy

Search for more papers by this author
Alberto Luini

Alberto Luini

Institute of Protein Biochemistry and Institute of Biochemistry and Cell Biology, National Research Council, Naples, Italy

Search for more papers by this author
Daniela Corda

Corresponding Author

Daniela Corda

Institute of Protein Biochemistry and Institute of Biochemistry and Cell Biology, National Research Council, Naples, Italy

Correspondence

C. Valente, D. Corda, or M. A. Zhukovsky, Institute of Protein Biochemistry and Institute of Biochemistry and Cell Biology, National Research Council, Via Pietro Castellino 111, Naples 80131, Italy

Tel: +39 081 6132225 (CV); +39 081 6132543 (MAZ); +39 081 6132536 (DC)

E-mails: [email protected] (CV), [email protected] (MAZ), [email protected] (DC)

Search for more papers by this author
Carmen Valente

Corresponding Author

Carmen Valente

Institute of Protein Biochemistry and Institute of Biochemistry and Cell Biology, National Research Council, Naples, Italy

Correspondence

C. Valente, D. Corda, or M. A. Zhukovsky, Institute of Protein Biochemistry and Institute of Biochemistry and Cell Biology, National Research Council, Via Pietro Castellino 111, Naples 80131, Italy

Tel: +39 081 6132225 (CV); +39 081 6132543 (MAZ); +39 081 6132536 (DC)

E-mails: [email protected] (CV), [email protected] (MAZ), [email protected] (DC)

Search for more papers by this author
First published: 31 July 2019
Citations: 100
Edited by Felix Wieland

Abstract

Phosphatidic acid (PA) is the simplest cellular glycerophospholipid characterized by unique biophysical properties: a small headgroup; negative charge; and a phosphomonoester group. Upon interaction with lysine or arginine, PA charge increases from −1 to −2 and this change stabilizes protein–lipid interactions. The biochemical properties of PA also allow interactions with lipids in several subcellular compartments. Based on this feature, PA is involved in the regulation and amplification of many cellular signalling pathways and functions, as well as in membrane rearrangements. Thereby, PA can influence membrane fusion and fission through four main mechanisms: it is a substrate for enzymes producing lipids (lysophosphatidic acid and diacylglycerol) that are involved in fission or fusion; it contributes to membrane rearrangements by generating negative membrane curvature; it interacts with proteins required for membrane fusion and fission; and it activates enzymes whose products are involved in membrane rearrangements. Here, we discuss the biophysical properties of PA in the context of the above four roles of PA in membrane fusion and fission.

Abbreviations

AGPAT4, acylglycerophosphate acyltransferase type 4

C2 domain, second conserved domain of protein kinase C

Cer-1-P, ceramide-1-phosphate

CtBP1-S/BARS, C-terminal-binding protein 1-short form/brefeldin A ADP-ribosylation substrate

DAG, diacylglycerol

DAGPP, diacylglycerolpyrophosphate

DGKs, diacylglycerol kinases

DHAP, dihydroxyacetone phosphate

DOPC, dioleoyl phosphatidylcholine

Drp1, dynamin-related protein 1

EEA1, Early Endosome Antigen 1

F-BAR domain, FCH-Bin/amphiphysin/RVS domain

FYVE domain, Fab 1-YOTB-Vac 1-EEA1 domain

G3P, glycerol-3-phosphate

LPA, lysophosphatidic acid

LPAAT, lysophosphatidic acid acyltransferase

LPC, lysophosphatidylcholine

MICAL-L1, molecule interacting with CasL-Like1

Munc13, mammalian uncoordinated 13

NSF, N-ethylmaleimide-sensitive factor

PA, phosphatidic acid

Pah1, phosphatidic acid phosphatase 1

PAK, p21-activated kinase

PAP, phosphatidic acid phosphatase

PA-PLA1, phosphatidic acid-selective phospholipase A1

PC, phosphatidylcholine

PE, phosphatidylethanolamine

PG, phosphatidylglycerol

PH domain, Pleckstrin Homology domain

PI, phosphatidylinositol

PIPs, phosphatidylinositol phosphates

PKCs, protein kinases C

PKDs, protein kinases D

PLA, phospholipase A

PLD, phospholipase D

PS, phosphatidylserine

PtdIns(4,5)P2, phosphatidylinositol-4,5-bisphosphate

PtdIns4P, phosphatidylinositol-4-phosphate

PtdIns4P5K, phosphatidylinositol-4-phosphate-5-kinase

PX domain, phox (phagocyte oxidase) homology domain

RasGRPs, Ras guanyl nucleotide-releasing proteins

SNARE, soluble N-ethylmaleimide-sensitive factor attachment protein receptor

TGN, trans-Golgi network

Unc-13, uncoordinated protein 13

Vps1p, vacuolar protein sorting-associated protein 1

Phosphatidic acid (PA), or 1,2-diacyl-sn-glycero-3-phosphate, is one of the simplest glycerophospholipids and, in general, one of the simplest membrane phospholipids. The PA molecule consists of a three-carbon glycerol backbone, to which two fatty acids and a phosphate are ester-linked at positions 1, 2 and 3 (denoted as sn-1, sn-2 and sn-3) respectively. These fatty acids can vary in length (usually 14–22 carbons long) and saturation (the number of double bonds) [1]. Often, a saturated fatty acid bonded to carbon-1, whereas an unsaturated fatty acid bonded to carbon-2. In lysophosphatidic acid (LPA), or 1-acyl-sn-glycero-3-phosphate, only one fatty acid is bonded to carbon-1. PA is a minor membrane phospholipid, it usually constitutes 1-4% of membrane lipids [2-6]. However, this substance is present in virtually all cellular membranes ([7] and references therein). PA is a signalling molecule that regulates several processes in the cell, including cell proliferation, survival, cytoskeletal organization, vesicular trafficking, secretion, reproduction, response to hormones and abiotic stresses [5, 7-19]. Turnover rate of PA is high [20, 21], and the cell could regulate the level of PA by the balance of PA synthesis and degradation [7, 15]. Such regulation has important implication for the cell function and is achieved through the PA-mediated binding and regulation of a wide range of effector proteins [22-24] at a certain time and in a particular subcellular membrane region [25]. In cell, PA can be produced by three major routes [24-26] (Fig. 1):

Details are in the caption following the image
Schematic representation of the biosynthetic pathways for PA production. Three major enzymatic pathways can generate PA. The de novo synthesis routes to the formation of LPA and then PA are shown by dashed arrows. Solid arrows represent the synthesis routes most often involved in signalling processes. The dashed green arrows show the production of LPA from DHAP via the 1-acyl-DHAP (Ac-DHAP) pathway: acylation followed by reducing. The dashed purple arrows show the synthesis of LPA from DHAP via G3P pathway: reducing followed by acylation. LPAAT catalyses the acylation of LPA to PA (dashed blue arrow). Phospholipase A2 (PLA2) catalyses the inverse reaction converting PA to LPA (solid blue arrow). The solid light blue arrow shows the hydrolysis of PC mediated by PLD as an alternative pathway to produce PA. The enzymes and reactions that are involved in the DAG pathway are shown in orange. DAG can be phosphorylated by DGK to form PA. The opposite reaction is catalysed by PAP. PA and DAG are in dynamic equilibrium and this balance affects the composition and curvature of the biological membranes. In addition, DAG is an important precursor for triacylglycerol (TAG) synthesis. Acyl-CoA:diacylglycerol acyltransferase (DGAT) catalyses the formation of an ester linkage between the free hydroxyl group of DAG and a fatty acid.
  1. de novo synthesis pathway in which the final step is acylation of LPA by lysophosphatidic acid acyltransferases (LPAATs) [27-29]. Five members of LPAAT family are present in mammals [28]. LPA can be produced from dihydroxyacetone phosphate (DHAP) via two different pathways. One of these pathways is the reduction of DHAP to glycerol-3-phosphate (G3P) via glycerol-3-phosphate dehydrogenase (GPDH), followed by the acylation of G3P to LPA via glycerol-3-phosphate acyltransferase (GPAT) [24, 26]. Another pathway is the acylation of DHAP to 1-acyl-DHAP (Ac-DHAP) via DHAP acyltransferase (DHAP AT), followed by the reduction of Ac-DHAP to LPA via Ac-DHAP reductase [5, 30, 31].
  2. the phosphorylation of diacylglycerol (DAG) by diacylglycerol kinases (DGKs) [32-34]. Ten members of DGK family are known ([13] and references therein);
  3. hydrolysis of phospholipids by phospholipase D (PLD) [1, 35-39]. Among the six known PLD isoforms, the mammalian PLD1 and PLD2 are the best studied ([13, 39] and references therein). PLD6 (MitoPLD) generates PA via hydrolysis of cardiolipin on the outer surface of the mitochondria [36, 40, 41]. At present, no enzymatic activity has been reported for PLD5 and for the two Endoplasmic Reticulum-localized PLD isoforms, namely PLD3 and PLD4 [42].

In turn, PA can be metabolized to DAG by phosphatidic acid phosphatases (PAPs) [43, 44] or to LPA by phospholipase A (PLA1 and PLA2) [45-47]. In plant and yeast cells, PA can be phosphorylated by PA kinase (PAK) into diacylglycerolpyrophosphate (DAGPP), and PA can be produced from DAGPP by DPP (DAGPP phosphatase) [5, 48]. Hence, directly or indirectly, PA is involved in biosynthesis and homeostasis of most phospholipids [25].

In this review, rather than discussing the role of PA as lipid mediator involved in cell growth, proliferation and cytoskeletal organization (reviewed elsewhere, see [3, 14-16]), we will focus on the structural and chemical properties of PA. In particular, we will highlight how such properties influence the processes of fusion and fission of PA-containing membranes.

Structural properties of phosphatidic acid

Phosphoric acid H3PO4 contains three hydroxyl groups. Phosphates are esters of phosphoric acid. In phosphoglycerides, also known as glycerophospholipids, glycerol (with one or two fatty acid ‘tails’) is ester-linked to one of three hydroxyl groups of phosphoric acid. In most phosphoglycerides such as phosphatidylcholine (PC), phosphatidylethanolamine (PE), phosphatidylglycerol (PG), phosphatidylserine (PS), phosphatidylinositol (PtdIns), a polar headgroup (such as choline in case of PC) is ester-linked to the other hydroxyl group (Fig. 2). Hence, in each of these phosphoglycerides, the phosphate contains only one hydroxyl group and two ester bonds. At physiological pH, this phosphate group is usually deprotonated. Hence, if a polar headgroup is electrically neutral (as in case of PS and PG), the lipid possesses a charge (−1) at neutral pH (Fig. 2). However, in PA, no headgroup (other than hydrogen) is present, and, hence, in the PA molecule, phosphate contains two hydroxyl groups and one ester bond (Fig. 2). Thus, PA contains a unique phosphomonoester headgroup rather than phosphodiester headgroup that is present in other phosphoglycerides [7]. At physiological pH, at least one of two PA hydroxyls is usually deprotonated and, hence, PA is negatively charged (Fig. 2).

Details are in the caption following the image
Schematic representation of the common headgroups of glycerophospholipids and their charges and shapes at physiological conditions. Glycerophospholipids with fatty acyl chains in positions sn-1 (R1) and sn-2 (R2) and with the common headgroups in position sn-3 (X) are shown. Some of the common glycerophospholipid headgroups with their charges and shapes are indicated.

Depending on their molecular shapes, various lipids are able to form structures of different curvatures. According to their shapes, lipids can be classified under three categories ([49-53] and references therein). If a tail of particular lipid is wider than its headgroup, this lipid can be modelled as a cone. The monolayer consisting of such lipid molecules tends to bulge spontaneously towards the hydrocarbon tails and to form inverted hexagonal phase (HII), with an aqueous core surrounded by the hydrophilic headgroups, and hydrophobic tails aggregating around the perimeter. This monolayer has negative curvature, and cone-shaped lipids are said to have negative spontaneous curvature (Fig. 3). If, however, a polar headgroup of a lipid is wider than its tail, this lipid can be modelled as an inverted cone. The leaflet consisting of such molecules has tendency to bulge spontaneously towards the layer of polar heads and to form spherical micelles, with headgroups contacting surrounding aqueous solvent, and lipid tails are directed inward. Such monolayer has positive curvature, and inverted cone-shaped lipids are said to have positive spontaneous curvature (Fig. 3). If the width of polar head is similar to the width of hydrophobic tail of a lipid, this lipid has a cylindrical shape. Such lipids form monolayers that are flat or almost flat, and spontaneous curvature of cylindrical lipids is equal or close to zero (Fig. 3). Cone-shaped and inverted cone-shaped lipids can be called nonbilayer lipids.

Details are in the caption following the image
Membrane lipid structures. Illustration of lipid shapes and their influence on membrane structure. Lipids with similar widths of polar head and hydrophobic acyl chains have a cylindrical shape (i.e. PC, PS, PG and PI shown in red). They do not induce spontaneous membrane curvature and assemble as flat membrane such as bilayers. Lipids with a wide polar headgroup and a single acyl chain have an inverted cone shape (i.e. LPC, LPA, shown in green). These lipids induce spontaneous positive membrane curvature favouring the formation of spherical micelles. Lipids with a small headgroup and a bulky acyl tail have a cone shape (i.e. PA, DAG and PE shown in yellow). They induce a negative membrane curvature promoting the formation of the inverted hexagonal phase or inverted micelles.

Under physiological conditions, PA [54] as well as PE [55], DAG [49] and cholesterol [56] are lipids that have cone shape and, hence, possess a negative spontaneous curvature. Such lipids as PtdIns(4,5)P2 [57] and one-tailed lipids lysophosphatidylcholine (LPC) [58] and LPA [54] possess inverted cone shape and, hence, have a positive spontaneous curvature. Two-tailed PC, having headgroup larger than the headgroup of PE, is a lipid forming monolayers that are almost flat. Shape of this lipid, in first approximation, is cylindrical, and spontaneous curvature of this lipid is close to zero [59, 60] (Fig. 3). Similarly, PG is a cylindrical lipid under physiological conditions [61]. PS, the most abundant acidic phospholipid, also has a cylindrical shape at neutral pH. However, when the pH is decreased through 3.5, that is below the pKa value, the shape of PS becomes conical, and this lipid exhibits a transition from bilayer to inverted hexagonal HII phase [62-64]. At neutral pH, the shape of PS is cylindrical because of repulsion of charged headgroups, whereas at low pH, these headgroups become protonated, repulsion between the headgroups is reduced, and PS molecules become cone-shaped [62]. The inverted hexagonal HII phase of PS is also observed under some other conditions, for example in the presence of lithium ions [65].

The shape of lipid is strongly dependent upon the degree of unsaturation of acyl tails. Cis double bond induces a ‘kink’ in acyl chain and thus causes it to occupy larger volume. Hence, lipid containing such unsaturated chain has more negative spontaneous curvature [66-68]. Acyl chains containing trans double bonds are more straight than chains that contain ‘kinked’ cis double bonds, and thus the effect of trans double bonds on lipid shape is much lower [66, 69]. The influence of increased saturation of acyl chains on lipid shape is similar to the influence of the shortening of these chains [70] or increase in size of polar headgroups [68]. Due to the more pronounced cone shape, unsaturated lipids containing cis double bonds are prone to create lipid packing defects, that is locations on the membrane surface where the hydrophobic interior of the membrane is exposed to the solvent [68]. Such defects have a tendency to recruit proteins ([68, 71, 72], and references therein).

Furthermore, except PA, few other lipids, such as PS, PG, PtdIns and various phosphatidylinositol phosphates (PtdInsPs), contain a negatively charged headgroup (Fig. 2). However, PA is a unique molecule that has both properties: anionic and very small headgroup. Ceramide-1-phosphate (Cer-1-P) also possesses this combination of properties: cone shape and negatively charged headgroup. However, the content of Cer-1-P in cells is even lower that the content of PA [11].

Chemical properties of PA

Based on the unique properties of PA: headgroup that is small, negatively charged and, moreover, phosphomonoester, this lipid is able to establish functionally relevant interactions with proteins, lipids and also with itself.

PA–protein interactions

In the studies of PA–protein interactions, various assays were used. In some of these assays, lipids were immobilized on nitrocellulose filter [73-76], coupled to the beads [22, 77], or organized into liposomes [78-81]. These assays are well discussed in Refs [3, 82, 83]. The liposome-based assay is the most used one because of the advantage that liposomes form a membrane bilayer mimicking the natural arrangement of lipids [3, 82].

The peculiar structural property of PA (described above) allows this lipid to be a very effective binding site for various proteins. In general, cone-shaped lipids induce packing defects [84-86]. It has been argued that PA is able to create such defects in the packing of lipids in membranes [7, 87, 88]. Hydrophobic residues of some proteins are involved in the interaction with PA or PA-containing membranes. Mutations of hydrophobic residues to alanine [76, 89-93], to hydrophilic residues [76, 89, 94] or to glycine [95, 96] (Table 1) contribute to the disruption of protein–PA interactions, whereas mutation of glycine (G) to hydrophobic valine (V) in position 12 (G12V) in oncogenic K-Ras4B protein results in preferential interactions with PA [97]. Most probably, hydrophobic residues interact with the PA acyl chains within the hydrophobic interior of the membrane [80, 89, 90, 94, 98-100] that is exposed to the solvent due to the presence of the membrane packing defects. Moreover, the negative charge of the PA headgroup facilitates the interaction of the PA-induced membrane packing defects with positively charged protein segments.

Table 1. Amino acid mutations that reduce the binding of proteins to PA.
Protein Mutation References
Mammalian
Raf-1 R398A [266]
R398A/R401A [267]
K399A, R398A/R401A, R398A/K399A, K399A/R401A, R398A/K399A/R401A, H402A, N404A, I405A, M409A, G410A [90]
mTOR R2109A, R2042A/K2095A/R2109A [268]
p47phox R70Q [111]
PKCε W23A/R26A/R32A, I89N, Y91A [89]
Dvl-2 H490A, K494A [269]
RPK118 K58Q/K59Q [270]
PtdIns4P5KIγ K97A/R100A, H126A/H127A, K97A/R100A/H126A/H127A [236]
Epac1 R82A [271]
Bazooka K458A/R504A/K506A/R546A/R635A/R637A/R682A [272]
PLC-β1 Y952G/I955G [96]
KSR1 R612A/R615A [267]
Nir2 D1128A [273]
PDZGEF K428A, R429A, K428A/R429A [274]
Sos-1 H475E/R479E [81]
R97A/K98E/R99A [275]
PKCα K209A/K211A, K197A/K209A/K211A [112]
Syntaxin1A R262A/R263A, K260A/R262A/R263A/K264A/K265A [79]
TAPAS-1 W19A/W20A, W19D/W20D [76]
CIDEA K167E/R171E/R175E [22]
CIN85 K645A/K646A/R648A/R650A [276]
PP1 S292A [277]
DOCK1 K1699A/K1701A/R1702A, V1614A/R1615A/T1616A/M1617A/R1628A/R1630A/K1699A/K1701A/R1702A [92]
DOCK2 K1697A/K1698A/R1699A, V1618A/R1619A/E1620A/M1621A/R1631A/R1633A/K1697A/K1698A/R1699A [91]
Lipin1β K153A/K154A/R155A/R156A, K153A/K154A/R155A/R156A/K157A/R158A/R159A/R160A/K161A [278]
Arf1 G2A [77]
Arf6 G2A [77]
Rac1 G12V/R185K [279]
Drp1 stalk domain A352G/Y354G/I355G/L360G [95]
IQGAP1 K1455A/K1456A/K1458A/K1464A/K1465A [280]
LATS1 H659A/K661A/K662A [74]
RA-RhoGAP K95A/R96A/R100A/H104A [281]
Insect
LKB1 K539A/K540A/K541A/K546A/R547A/R548A/K550A/K551A [282]
Mollusk
PKC Apl II R256G/R257S [283]
Fungal
Spo20p K66E/K68E/R71E/K73E/L67P [284]
Opi1p K112R/K119R/K121R/K125R/K128R [285]
K136A/K137A/R138A [286]
Sso1p R260G/K261G/K263G [198]
Ups1 R25E, K61E/K155E [80]
I78D, V106E [94]
Plant
ABI1 R73A [287]
TGD2 K221A [288]
G192S, G234R [289]
TGD4 S128A, R129A, K135A, K138A [290]
PABD domain of SnRK2.4 R266A/K278A/K279A/K294A/K300A [291]
MAP65-1 S428G/K429G [292]
RbohD160 R149A/R150A/R156A/R157A, R149G/R150G [293]
14-3-3 K56E [294]
MtDef4 R39A/R40A [75]
WEREWOLF K51A, R52A, R58A, K51A/R52A, R58A/R60A [295]
NsD7 K36E, R39E [78]
RGS1 K259E [296]
GEF8 K13A, K18A, K13A/K18A [73]
AtSPHK1 deletion of V182SGDGI187 [297]
PID K119G/K120G/K121G [298]
Protozoan
PfAPH K138A, K140A, K138A/K140A, K163A, K165A, K163A/K165A, I143A/F144A, H145A [93]
TgAPH K130A/K132A, K155A/K157A, L135A/F136A [93]
Bacterial
MscL F80W/R98Q/K99Q/K100Q [299]
PA39111 L58W, V69W, L114W, L56R, V69R, L58R, L114R [300]

Phosphatidic acid-binding sites often consist of stretches of positively charged amino acid residues [3, 83]. Consistent with this observation, a vast majority of PA-binding sites shown in Table 1 contain at least one (R/K/H)(R/K/H) motif or (R/K/H)x1–3(R/K/H) motif, where x is any residue. In most of these sites, arginines (R) and lysines (K) are present, and only few of these sites contain histidines (H).

Phosphomonoester headgroup of PA has a remarkable property to form hydrogen bonds [101, 102]. According to the electrostatic/hydrogen bond switch mechanism, the formation of the hydrogen bond of positively charged residue (lysine or arginine) with the PA headgroup causes deprotonation of this headgroup. Thus, the PA charge increases from −1 to −2, and this change stabilizes protein–lipid interaction [99]. This process is possible only in phosphomonoester headgroup containing two, rather than one, hydroxyl groups. Lysine is more effective in docking to PA headgroups than arginine [99]. Possible reason is reduced hydrogen bond donating capacity of arginine due to the substantial delocalization of charge in its guanidinium group [99]. The theory of electrostatic/hydrogen bond switch was further developed in [103, 104]. The results reported in Refs [4, 105, 106] support the electrostatic/hydrogen bond switch mechanism. Mobeen Raja stressed that PA-induced stabilization of potassium channel KcsA is more pronounced than cardiolipin-induced stabilization of this protein, due to the electrostatic and hydrogen bond interactions between PA and positively charged residues [106]. Consistent with the electrostatic/hydrogen bond switch mechanism, pH-dependent binding of transcriptional repressor Opi1 to PA has higher affinity compared with pH-independent binding of Opi1 to methyl-PA [4]. Observation that membrane binding and PAP (PA phosphatase) activity of Lipin 2 are stimulated by di-anionic PA can also be explained by electrostatic/hydrogen bond switch mechanism [105].

Thus, the specific interactions of PA with proteins take place due to the unique properties of PA: headgroup that is small, negatively charged and, moreover, phosphomonoester.

Moreover, it is not surprising that the binding of proteins with PA is specific. Although PS in more abundant negatively charged membrane lipid than PA, the proteins that bind PA rarely bind also PS, due to the stabilization of protein–PA interaction according to the electrostatic/hydrogen bond switch mechanism described above [7]. Many proteins exhibit high-affinity interaction with PA ([12, 83, 107-109] and references therein) (Table 1). At least 50 proteins that bind to PA or are regulated by PA were found in yeast, plants and animals [107]. However, no PA-recognition consensus sequence motif was identified so far [5, 14, 25, 108] in a contrast to the known lipid-binding domains, such as Pleckstrin Homology (PH), Phox (phagocyte oxidase) Homology (PX), Fab 1-YOTB-Vac 1-EEA1 (FYVE), or second conserved domain of protein kinase C (C2) domains. According to the current view, such motif does not exist; PA effectors use cationic and/or surface exposed hydrophobic residues for binding to PA [3, 19, 25]. Indeed, in the structures of PA-binding proteins and peptides obtained using X-ray crystallography [78, 80, 94, 110-114] and NMR [75, 99, 100], the binding of positively charged and hydrophobic residues to PA was often observed.

Phosphatidic acid is known to stabilize the conformation of various proteins and peptides [22, 106, 115-119]. The circular dichroism (CD) spectroscopy [22, 117-119] and Fourier-transform infrared (FTIR) spectroscopy [116] results show that upon interaction of α-synuclein [117, 118], potassium channel KcsA [119], nicotinic acetylcholine receptor nAChR [116] and mutant peptide of brown adipocyte protein CIDEA [22] with PA-containing lipid bilayers, the α-helical [22, 117-119] or β-sheet [116] content increases. The stabilizing effect of other negatively charged lipids, such as PS [117] or PG [119], was less pronounced. It has been suggested that electrostatic interactions between the positively charged residues and negatively charged PA headgroups [106, 116, 119] or the influence of PA on membrane fluidity [115] might cause the PA-induced protein stabilization.

PA–lipid interactions

Phosphatidylcholine headgroup contains a quaternary amine, whereas PE headgroup contains a primary amine, in which only one hydrogen atom is replaced, see Fig. 2. Unlike PC quaternary amine, PE primary amine is able to form a hydrogen bond with PA and thus facilitates PA deprotonation and increases negative charge of PA [11, 99]. In this way, PE facilitates the binding of proteins to PA-containing membranes [108]. Moreover, PE also promotes the interaction of proteins with PA, because cone-shaped PE molecules increase the exposure of PA headgroups which are usually buried deep, close to the glycerol backbone [108].

When present in PC bilayer, LPA carries more negative charge than PA, despite identical phosphomonoester headgroups [120]. This phenomenon can be explained by intramolecular hydrogen bond in LPA, between the sn-2 hydroxyl (absent in PA) and phosphomonoester. Such bonding facilitates the dissociation of proton [7]. However, in bilayers rich in PE, the charge of LPA is similar to the charge of PA. In such bilayers, primary amines of PE headgroup provide a large amount of hydrogen bond donors that overcome the hydrogen bonding ability of the sn-2 hydroxyl of LPA [11].

By definition, the pKa value of an acid is equal to the value of pH at which the half of the acid molecules is deprotonated. PA contains a phosphomonoester headgroup, and the pKa value of such group is within physiological range. Hence, in response to relatively slight changes in pH in the physiological range, the PA molecule changes its protonation state noticeably [7]. Due to this property, PA acts as a pH biosensor [4, 7]. The role of PA as pH biosensor allows this lipid to regulate various biological processes, such as nutrient sensing and cell-growth signalling in yeast [7]. The authors of Ref. [121] suggested that the presence of multivalent ions is crucial for the pH-sensing ability of PA. Furthermore, it was suggested that electrostatic-hydrogen bond switch mechanism allows PA-containing membranes to use physiological pH changes to trigger adsorption and desorption of proteins [103].

PA–PA interactions

As we pointed out, phosphomonoester headgroup of PA is able to form hydrogen bonds [99, 101, 102]. Not only PA–PE, PA–lysine and PA–arginine bonds (described above), but also PA–PA bonds can form. At low pH or in the presence of divalent cations, attraction of PA molecules overcomes the electrostatic repulsion of anionic headgroups and, thus, PA microdomains (clusters) are formed [20, 102, 122-124]. PA clusters can also form as a result of ‘explosive’ generation of PA by PA-producing enzymes [76, 125]. PA microdomains were experimentally observed, using molecular sensor for PA, at the plasma membrane, near sites of active exocytosis [125]. As already mentioned, PA molecules recruit proteins to the membranes, and PA microdomains can be patches of increased surface density of proteins that bind to PA [20, 126, 127]. Highly charged PA microdomains could cause conformational changes in PA-associated proteins and be patches of negative membrane curvature [20].

The authors of Ref. [115] suggested that the fluidity of a membrane decreases with an increase in PA content, due to the very small headgroup of PA that leads to the lateral condensation of the lipid tails. Some lipids, such as cholesterol, are known to reduce membrane fluidity [56], and due to such influence, cholesterol indirectly modulates the function of some membrane proteins [128], such as cholecystokinin receptor [129, 130] or GABA receptor [131]. But, due to the low content of PA in membranes, the putative ability of PA to affect membrane fluidity should not influence the properties of membrane proteins noticeably.

However, interestingly, Miner et al. hypothesized that an increase in DAG levels, due to deleting the DAG kinase Dkg1 converting DAG into PA, augments yeast vacuole fusion through altered membrane fluidity. It has been suggested that such change in the physical properties of the membrane modulates the activity of the Ypt7 protein that is required for fusion [132].

DAG: a PA lipid metabolite

As we already pointed out, PA can be metabolized to DAG by PAPs [43] (Fig. 1). The molecule of DAG consists of glycerol backbone, to which two fatty acids are linked through ester bonds in positions 1 and 2. DAG is a lipid second messenger that regulates a wide variety of cellular processes, proliferation, differentiation, adhesion, apoptosis, actin cytoskeleton reorganization and neurotransmitter release [10, 49]. DAG molecule has a cone shape [58] and, furthermore, the most pronounced cone shape of all lipids [49]. Moreover, due to its extreme cone shape and lack of charge, DAG induces unstable, asymmetric regions in membranes, and existence of such DAG-induced regions are suggested to be essential for membrane fusion and membrane fission ([49] and references therein). DAG binds to and activates various proteins containing cysteine-rich C1 domains, such as protein kinases C (PKCs), protein kinases D (PKDs), DAG kinases (DGKs), chimaerins, Ras guanyl nucleotide-releasing proteins (RasGRPs), uncoordinated protein 13 (Unc-13), mammalian uncoordinated 13 (Munc13) proteins ([49, 133-135] and references therein). Interestingly, DAG activates DGKs that produce PA from DAG, thus providing negative-feedback mechanism for the termination of its own signalling [134].

Due to its small and electrically neutral polar head, DAG possesses an unique capacity to undergo rapid, within seconds, transbilayer translocations known as flip-flop ([136-139] and references therein). Other phospholipids, including PA, undergo transbilayer translocations at a much slower rate, within minutes to hours ([137, 140] and references therein). DAG does not facilitate the flip-flop of other lipids [137]. However, in the presence of DAG-binding proteins, the transbilayer translocation of DAG might be inhibited [141].

The role of PA in membrane fusion and membrane fission

Both proteins and lipids play important roles in membrane rearrangements. Although proteins mediate membrane fusion (Box 1) and fission (Box 2), lipids are involved in the regulation of these processes [142-147]. Such lipid-mediated regulations of membrane rearrangements are only beginning to be explored. Sterols ([56, 148-151] and references therein), phosphoinositides ([152-159] and references therein), PA and DAG ([20, 111, 147, 160-164] and references therein) are among few lipids that have regulatory role in membrane fusion and fission.

Box 1. Membrane rearrangements: fusion

Membrane fusion and fission are two opposite processes. Membrane fusion is a process by which two biological membranes unite into one membrane. If we consider fusion of two membrane-enclosed compartments, contacting (also known as proximal, juxtaposed, apposing) monolayers are leaflets that face each other through a water gap before the fusion process starts and then make the initial contact, whereas distal monolayers are opposite leaflets [188, 244, 245]. Inner or outer monolayers of membrane-enclosed compartments play the role of contacting leaflets during membrane fusion, depending on relative position of these compartments. For example during exocytosis, synaptic vesicles fuse with cell plasma membrane from inside, and, hence, inner and outer monolayers of plasma membrane are contacting and distal monolayers, respectively, whereas for synaptic vesicle, inner and outer monolayers are distal and contacting monolayers respectively.

The fusion process starts by the formation of stalk, a local hourglass-shaped lipidic connection between the contacting membrane monolayers [59, 245-247]. The formation of the stalk intermediate [248] is a common step in all membrane fusion processes [249]. Radial expansion of a stalk leads to the formation of hemifusion diaphragm, a bilayer formed by two distal monolayers of the apposed membranes [247, 250, 251]. Evolution of hemifusion intermediate leads eventually to the formation of a fusion pore, a structure involving connection of contacting monolayers as well as connection of distal monolayers of fusing bilayers. Formation of a fusion pore establishes continuity between aqueous spaces initially separated by opposing membranes. The edge of fusion pore is covered with the hydrophilic headgroups of lipids. Expansion of fusion pore completes fusion reaction.

Box 2. Membrane rearrangements: fission

Membrane fission is a process that involves a splitting of one membrane-enclosed compartment into two. This process usually evolves through three steps: membrane neck formation, hemifission and formation of two separate membranes [252-259]. During membrane fission, initially saddle-shaped membrane neck forms: both monolayers remain continuous, but aqueous volume consists of two parts connected by a narrow ‘capillary’. Then, coalescence of one of the leaflets takes place [257]. We will use the term ‘contacting monolayer’ to name this leaflet, and the term ‘distal monolayer’ to name the opposite leaflet. Thus, hemifission intermediate forms [253], and the aqueous volume breaks into two parts, although distal monolayer remains continuous. At a later stage, distal monolayer divides, and as a result, two separate membrane-enclosed compartments form.

Neck intermediate of membrane fission is similar to the pore of membrane fusion process, whereas hemifission resembles hemifusion intermediate. During membrane fission, membrane neck is followed by hemifission and then by two separate membranes, whereas during membrane fusion, stalk is followed by hemifusion, then fusion pore forms and finally a united membrane-enclosed compartment appears. The difference between fusion pore and membrane neck (sometimes referred to as ‘fission pore’) is that fusion pore expands, whereas neck narrows.

If we consider kiss-and-run fusion (an unconventional fusion between secretory vesicles and a target membrane where the vesicle opens and closes transiently) [260], the monolayers that were contacting during membrane fusion become distal during fission, and vice versa, monolayers that were distal during fusion become contacting during fission.

As a result of membrane fission, two disjoint compartments (as in case of cell division) or, alternatively, one compartment inside another compartment (as in case of endocytosis) can form.

Although different fusion and fission reactions are mediated by many dissimilar proteins, often processes driven by dissimilar proteins are regulated by a common lipid, such as PA [40].

Phosphatidic acid, LPA and DAG are involved in many membrane fusion and membrane fission events, such as fission of Golgi membranes [140, 162, 163, 165-169], fission during endocytic transport [170, 171], vacuole fission [172] and fusion [132, 147, 173-175], mitochondrial dynamics [17, 40, 95, 176-185], exocytosis [30, 40, 76, 79, 125, 186-194], vesicle-vesicle fusion [164], myoblast fusion [195, 196], osteoclast fusion [197], membrane fusion during sporulation [198-201], membrane fusion during fertilization [202, 203]. Fusion of PA-containing vesicles was studied in Refs [87, 204].

Special attention should be payed to studies where the role of PLD-mediated PA production in a variety of cellular processes relied on the use of primary alcohols (e.g. n-butanol). Alcohols are not PLD inhibitors, instead they compete with water to hydrolyse PC leading to the production of phosphatidylalcohol (instead of PA) in the transphosphatidylation reaction. Here the inaccuracy concerns the assumption that this reaction will be complete and the metabolically stable phosphatidylalcohol product will not exert inhibitory effects on the PLD/PA-driven cellular processes and will not affect other cellular pathways. Unfortunately, several membrane transport routes blocked by 1-butanol [205-208] are not affected by PLD specific inhibitors [208] and have been analysed in cells where the Golgi apparatus is fragmented under primary alcohol treatments [209, 210]. Thus, the generated data need further verification with PLD-isoform selective knockdowns, genetic knockouts and/or inhibitors.

According to Refs [13, 14, 40, 144, 187, 211], PA is able to influence membrane fusion and fission through the following four main mechanisms, see Fig. 4:

Details are in the caption following the image
Schematic representation of membrane rearrangements influenced by PA. (A) The increment in the local concentration of PA (in yellow) leads to the accumulation of negative charges (as indicated) and membrane packing defects allowing the recruitment of membrane fission/fusion-inducing proteins containing positively charged residue(s) (indicated as green star). (B) Local accumulation of PA (in yellow) generates membrane rearrangements by promoting negative membrane curvature. (C) PA (in yellow) stimulates the activity of enzymes whose products are involved in membrane rearrangements [e.g. PtdIns4P5K that produces PtdIns(4,5)P2]. (D) PA, generated by LPAAT or DAG kinase activities, acts as a substrate for enzymes producing other lipids: PA is converted into DAG by PAP enzyme or into LPA by PLA2 respectively (as indicated).
  1. PA is a substrate for enzymes producing other lipids;
  2. PA plays a role in membrane rearrangements by generating negative membrane curvature (see also Box 3);
  3. PA interacts with proteins required for membrane fusion and fission;
  4. PA activates enzymes whose products are involved in membrane rearrangements.

Box 3. The dependence of fusion and fission reactions on monolayer spontaneous curvature

The dependence of the fusion reaction on monolayer spontaneous curvature is not obvious. For a fusion stalk, there are two curvatures in different projections: positive, if you look from above (around the stalk), and negative in profile. The energy of fusion stalk should be calculated with respect to the energy of the preceding intermediate that is two flat membranes. Calculations show that lipids of positive and negative spontaneous curvature, respectively, inhibit and promote stalk formation (and, hence, membrane fusion), if present in the contacting monolayer. Furthermore, lipids of positive and negative spontaneous curvature, respectively, promote and inhibit fusion pore formation, if present in the distal monolayer [50, 261].

Thus, fusion can be restricted by the presence of lipids of positive spontaneous curvature or the lack of lipids of negative spontaneous curvature in the contacting monolayer. However, fusion block can be rescued, at least partially, by the addition of a lipid of positive spontaneous curvature to the distal monolayer [262, 263].

The dependence of the fission reaction on monolayer spontaneous curvature is (as in the case of fusion) not obvious. The neck is characterized both by negative curvature (along the neck profile) and positive curvature (around the neck) [264]. The energy of hemifission intermediate should be calculated with respect to the energy of the preceding intermediate that is constricted neck (but not two flat membranes as in the case of fusion stalk).

According to the current view, inverted cone-shaped lipids (i.e. lipids of positive spontaneous curvature) present in the distal monolayer promote membrane fission [256, 265].

It is quite possible that in any given event of membrane rearrangements, PA could influence via a combination of these mechanisms.

PA is a substrate for enzymes producing other lipids

Results of few studies indicate that conversion of PA to DAG [140, 147, 181, 212] or to LPA [177] might influence membrane rearrangements. The authors of Refs [140, 212] suggested that DAG undergoes flip-flop, influences membrane curvature and thus plays a role in membrane fission during peroxisome division [212] or Golgi vesicle and tubule formation [140, 213]. According to Ref. [212], DAG also recruits dynamin-like GTPase Vps1p that promotes membrane fission. Interestingly, in Ref. [140], the translocation from the cytosolic (outer) to the lumenal (inner) leaflet of Golgi is suggested, whereas in Ref. [212], quite opposite, the translocation of DAG from the lumenal (inner) to the cytosolic (outer) leaflet of peroxisome is hypothesized. However, the presence of cone-shaped lipids in the distal monolayer should rather inhibit membrane fission (Box 3).

The authors of Refs [181] and [177] suggest that conversion of PA to other lipids, DAG and LPA, plays a role in mitochondrial dynamics: fission and fusion. In Ref. [181], it is suggested that PA recruits the phosphatase Lipin 1 that converts PA to DAG and promotes mitochondrial fission. According to Ref. [177], PA-PLA1-catalysed consumption of cone-shaped PA and/or production of inverted cone-shaped LPA at mitochondrial constriction sites might play a role in mitochondrial fission. However, the relationship between DAG-producing Lipin 1 and LPA-producing PA-PLA1 in the regulation of mitochondrial fusion and fission should be clarified in the future.

In summary, the role, in membrane rearrangements, of the shape of LPA and DAG, produced from PA, is often discussed, and the importance of DAG flip-flop is sometimes suggested. Nonetheless, in order to elucidate the role that inverted cone-shaped LPA and cone-shaped DAG play in membrane fusion and fission, one needs to clarify, in which monolayer these lipids are present, and whether such an influence can be explained by existing models of fusion and fission (Box 3). Furthermore, one should take into account that, due to flip-flop, cone-shaped DAG is equilibrated fast between two monolayers. Hence, membrane curvature generated by DAG is able to influence membrane rearrangements only if these processes are so fast that they occur before DAG redistribution.

In Refs [147, 214], quite different model is discussed. During the priming step of soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE)-mediated yeast vacuole fusion, the normal function of hexameric Sec18 (yeast orthologue of mammalian N-ethylmaleimide-sensitive factor, NSF) is the reactivation of vacuolar cis-SNARE complexes for additional rounds of membrane fusion. However, if concentration of PA in vacuolar membrane is high, Sec18 is inactive because Sec18 protomers remain associated with vacuoles through interactions with negatively charged PA. Upon hydrolysis of PA to neutral DAG by PA phosphatase Pah1, Sec18 is released from the membrane, Sec18 hexamers form, and these hexamers are able to be recruited to cis-SNARE complex to perform its priming function. Thus, PA-to-DAG conversion promotes membrane fusion, because a protein that is needed for fusion is sequestered due to its interaction with PA, but freed due to the conversion of PA to DAG.

Moreover, DAG-activated C1 domain-containing Munc13 proteins mediate neurotransmitter release by exocytosis (i.e. membrane fusion) [215, 216].

PA plays a role in membrane rearrangements by generating negative membrane curvature

Surface density of the cone-shaped lipid PA in the inner leaflet of the plasma membrane (contacting leaflet in exocytosis) was decreased due to PLD1 down-regulation mediated by RNA interference [125, 188, 190] or due to the mutation of five cationic residues within PA-binding site of syntaxin1A that sequesters PA to sites of fusion [79]. As a consequence, membrane fusion was blocked. This block was rescued by addition of inverted cone-shaped LPC to the outer leaflet of plasma membrane (distal leaflet in exocytosis) [79, 125, 188, 190]. These observations confirm that, at least in this system, PA promotes fusion due to its cone shape (Box 3).

The fission of post-Golgi carriers directed towards the basolateral plasma membrane has been analysed at the molecular level [159, 162, 217, 218]. This process is mediated by brefeldin A ADP-ribosylation substrate (BARS), the shorter splice isoform of the CtBP1 protein, known as CtBP1-S/BARS ([219-222] and references therein). This protein is a member of C-terminal binding protein (CtBP) family. CtBP1-S/BARS is a fission-inducing protein involved in several intracellular fission events along the intracellular transport pathways [217, 218, 223, 224] as well as during the Golgi ribbon unlinking in mitosis [225-227]. During the formation of basolaterally directed post-Golgi carriers, CtBP1-S/BARS localizes on the outer surface of trans-Golgi network (TGN) where it is a key component of a well-defined protein complex that controls and mediates the fission of these carriers [159]. Once incorporated into this protein complex, CtBP1-S/BARS binds to and activates LPAATδ (also known as acylglycerophosphate acyltransferase type 4, AGPAT4), an enzyme that converts LPA into PA [162].

In this system, the presence of cone-shaped PA in the outer (distal) leaflet of TGN should, according to the current view, inhibit the process of membrane fission rather than promote it (Box 3). It is reasonable to conclude that the presence of PA-recruited fission-inducing protein CtBP1-S/BARS overcomes the membrane fission-inhibiting effect of cone-shaped lipid PA.

PA interacts with proteins required for membrane fusion and fission

Interaction of PA with proteins inducing membrane rearrangements promotes these rearrangements

Phosphatidic acid is known to play an important role in membrane fusion mediated by SNARE proteins ([79, 147, 198-200, 214] and references therein). Interactions of PA with various SNARE proteins, such as sporulation-specific yeast homolog of SNAP-25 (Spo20p) [200], Sso1p and Sso2p (yeast homologs of syntaxin1) [198, 199], syntaxin1A [79], stimulates membrane fusion. Multiple mutations of cationic residues in juxtamembrane PA-binding sites in Sso1p [198] and syntaxin1A [79] inhibit fusion. The authors of Ref. [79] suggested that syntaxin1A sequesters PA to sites of fusion, where PA promotes the fusion process due to its cone shape (Box 3).

Phosphatidic acid interacts with the mitochondrial outer membrane protein Ugo1, a component of the mitochondrial fusion machinery, and thus promotes mitochondrial fusion [184]. PA stimulates membrane insertion of Ugo1 and assembly of Ugo1 homodimeric complexes [184]. Furthermore, mitochondrial outer membrane protein mitofusin 1 drives mitochondrial fusion, and PA stimulates this process [40].

Moreover, in few systems PA promotes membrane fission via interaction with fission-inducing proteins. The authors of Ref. [171] studied generation of recycling endosome (RE) membrane tubules and their subsequent fission. Two proteins, molecule interacting with CasL-Like1 (MICAL-L1, via its coiled coil domain) and syndapin2 [via its FCH-Bin/amphiphysin/RVS (F-BAR) domain] are recruited to PA-enriched membranes, and direct interactions between these two proteins stabilize their association with the membrane, allowing enhanced tubulation mediated by syndapin2.

Experiments based on the use of specific PLD2 inhibitors indicate that PLD2-mediated PA production recruits ArfGAP1 to the Golgi membranes. ArfGAP1 is involved in the molecular mechanisms that drive the formation of Golgi tubules [168].

The CtBP1-S/BARS-mediated activation of LPAATδ enzyme and the local production of PA support the fission of basolaterally directed post-Golgi carriers [162]. CtBP1-S/BARS binds PA [163], and we further hypothesized that a runaway process takes place: CtBP1-S/BARS-bound LPAATδ converts LPA into PA, more CtBP1-S/BARS is recruited to the PA-containing Golgi membrane, and so on, until the surface densities of CtBP1-S/BARS and PA become high enough to induce membrane fission [162].

Interaction of PA with proteins inducing membrane rearrangements inhibits these rearrangements

SNARE-mediated yeast vacuole fusion begins with the priming stage in which Sec18 (yeast homolog of NSF) and its co-chaperone Sec17 (yeast homolog of α-soluble NSF attachment protein, α-SNAP) disassemble inactive cis-SNARE complexes. The authors of Ref. [147] found that PA sequesters Sec18 from cis-SNARE complexes. Thus, PA inhibits membrane fusion through restraining of a protein that is required for fusion (see section ‘PA is a substrate for enzymes producing other lipids’). We can see that PA is able to promote [79, 198-200] and inhibit [147] SNARE-mediated membrane fusion.

Mechanoenzyme dynamin-related protein 1 (Drp1) plays an important role in mitochondrial fission [228-230]. Drp1 binds to PA present in mitochondrial membrane. This interaction does not stimulate enzymatic activity of Drp1, but instead keeps this protein inert [95, 176]. Moreover, PA-producing enzyme MitoPLD interacts with Drp1, and thus PA is produced near Drp1 and is able to restrain it [176].

As we pointed out above, PA stimulates mitochondrial fusion mediated by mitofusin 1 [40]. Interestingly, MitoPLD also binds directly to mitofusin 1 [176]. Thus, PA regulates both mitochondrial fission and fusion processes.

Moreover, these phenomena are very similar to the processes involved in CtBP1-S/BARS-mediated fission of post-Golgi carriers [162] described above. We can see that the same theme is common for disparate fusion and fission processes. A protein that mediates membrane rearrangements (CtBP1-S/BARS, Drp1 or mitofusin1) binds to an enzyme that produces PA (LPAATδ for CtBP1-S/BARS, MitoPLD for Drp1 and mitofusin1). PA recruits more copies of these membrane rearrangement-mediating proteins, and thus a runaway process takes place. This way, PA promotes CtBP1-S/BARS-mediated fission of post-Golgi carriers, inhibits Drp1-mediated mitochondrial division and promotes mitofusin1-mediated mitochondrial fusion.

In summary, PA promotes membrane fusion or inhibits fission, if it activates fusion proteins [79, 184, 198-200] or inhibits fission proteins [176]. Furthermore, PA promotes membrane fission or inhibits fusion, if it activates fission proteins [162, 168, 171] or inhibits fusion proteins [147].

PA activates enzymes whose products are involved in membrane rearrangements

Phosphatidic acid binds to and activates the three isoforms (α, β, γ) of phosphatidylinositol-4-phosphate-5-kinase (PtdIns4P5K) [231-235]. Several studies have shown that PA generated by PLD, as well as DGKα and DGKζ, activates PtdIns4P5K in vivo, in contrast to PA produced by DGKε ([235] and references therein). According to Refs [232, 235], species of PA with two unsaturated acyl chains are much better activators of PtdIns4P5K than those containing one saturated and one unsaturated acyl chain or two saturated chains. PA species containing acyl tails composed of 14 carbons or longer are sufficient to bind PtdIns4P5K [232]. PtdIns4P5K is an enzyme that generates phosphatidylinositol-4,5-bisphosphate [PtdIns(4,5)P2] ([235, 236] and references therein). PA-binding site on γ isoform of PtdIns4P5K was identified [236]. PA binding is not only involved in recruiting PtdIns4P5KIγ to membranes but also may induce a conformational change [236]. PtdIns(4,5)P2 plays an important role in endocytosis and exocytosis [158, 192, 237-239], myoblast fusion [195, 240] and phagocytosis [241]. Interestingly, PtdIns(4,5)P2 stimulates PLD, an enzyme that produces PA. Thus, reciprocal stimulation of PLD and PtdIns4P5K enzymes enables a rapid feed-forward stimulation loop for a fast generation of PA and PtdIns(4,5)P2 ([242, 243] and references therein).

In summary, PA activates an enzyme, PtdIns4P5K, whose product PtdIns(4,5)P2 is involved in membrane fission and fusion.

Concluding remarks

Although several studies support the role of PA in the regulation of intracellular membrane transport and a number of PA-binding effectors have been described, the mechanisms by which PA controls membrane transport remain poorly characterized. Based on the unique biophysical properties of PA, in this review we have reported the main mechanisms through which this lipid induces membrane rearrangements, namely membrane fission and fusion. However, PA can be also metabolized to DAG and LPA, two lipids also able to strongly affect membrane curvature. Future biochemical and biophysical studies using in vitro reconstitution systems will help defining the exact role of the PA molecule and of PA-binding effectors in the PA-driven membrane rearrangements. Finally, novel imaging tools to monitor the spatiotemporal membrane localization of PA will improve our knowledge on the subcellular distribution, metabolism and role in membrane transport of this unique lipid in vivo.

Acknowledgements

The authors’ research work was supported by the Italian Association for Cancer Research (AIRC) (to DC IG10341 and IG18776; to AL IG20786 and IG15767), the AIRC-Fondazione Cariplo TRansforming IDEas in Oncological research project (TRIDEO) (to CV IG17524), the PRONAT project, the PRIN project No. 20177XJCHX, the SATIN POR Project 2014–2020 and the Italian-MIUR Cluster project Medintech (CNT01_00177_962865).