Phospholipid Flippases in Membrane Remodeling and Transport Carrier Biogenesis

Abstract

Molecular mechanisms underlying the formation of multiple classes of transport carriers or vesicles from Golgi and endosomal membranes remain poorly understood. However, one theme that has emerged over three decades is the dramatic influence of membrane lipid remodeling on transport mechanisms. A large cohort of lipid transfer proteins, lipid transporters and lipid modifying enzymes are linked to protein sorting, carrier formation and SNARE-mediated fusion events. Here we focus on one type of lipid transporter, phospholipid flippases in the type IV P-type ATPase (P4-ATPase) family, and discuss recent advances in defining P4-ATPase influences on membrane remodeling and vesicular transport.

Introduction

Transverse movement of phospholipid between leaflets of a protein-free membrane bilayer (flip-flop) is an energetically unfavorable and slow process [1]; however, rates of lipid flip-flop can be greatly accelerated by lipid transporters in biological membranes called flippases, floppases or scramblases [2]. Flippases are ATP-powered pumps from the P4-ATPase family that transport specific lipid substrates from the exofacial leaflet (extracellular leaflet of plasma membrane or luminal leaflet of Golgi/endosome membranes) to the cytosolic leaflet. Floppases are ABC transporters that catalyze lipid transport in the opposite direction and scramblases mediate energy-independent bi-directional lipid transport [2]. Most of the flippases are αβ heterodimers formed from a catalytic α subunit (the P4-ATPase) and a β subunit from the Cdc50 (TMEM30A) protein family [3]. The α subunit nomenclature is inconsistent between species, but P4-ATPases are designated ATP8A1 through ATP11C in mammals; ALA1 through ALA12 in Arabidopsis thaliani; Drs2, Neo1, Dnf1, Dnf2, Dnf3 in Saccharomyces cerevisiae; and TAT-1 through TAT-5 in C. elegans. No high-resolution structure is available for a P4-ATPase, but homology models for the catalytic subunit have been generated to provide a framework for structure/function studies into how these enzymes transport their substrate [4].

Influence of phospholipid flippases on membranes

The unidirectional transport of lipid across a membrane bilayer by a P4-ATPase has a number of consequences. Removal of phospholipid from the exofacial leaflet coupled with its addition to the cytosolic leaflet creates an imbalance in surface area between the leaflets, which imparts molecular crowding stress on the cytosolic leaflet, and a packing defect stress on the exofacial leaflet [5]. A number of P4-ATPases transport phosphatidylserine (PS), thereby displacing negative charge to the cytosolic leaflet, significantly changing the electrostatic properties of the membrane, and exacerbating the crowding stress by concentration of negatively charged lipid. The interleaflet phospholipid gradient stores potential energy that the cell can harness to drive membrane bending towards the cytosol (positive curvature), to displace lipids capable of spontaneous flip-flop (cholesterol, ceramide, diacylglycerol) to the exofacial leaflet, to facilitate extraction of lipid from the cytosolic leaflet by lipid transfer proteins, or to transduce signals by activation of a scramblase and dissipation of the gradient (Figure 1). In addition, phospholipid asymmetry generated by P4-ATPases influences the activity of other transporters and channels in the membrane, the association of soluble proteins with the membrane, and the biophysical properties of the membrane [6–9].

Figure 1. P4-ATPase-mediated lipid transport imparts stress on membranes that can be relieved through membrane bending, induced flop of other lipids, exchange of lipid by transfer proteins, or activation of a scramblase.

P4-ATPases convert the chemical potential energy in ATP to potential energy stored in a lipid gradient within the membrane. In the example shown, the flippase-catalyzed unidirectional transport of PS creates molecular crowding stress in the cytosolic leaflet, packing defect stress in the luminal leaflet, and a charge difference between the leaflets. The potential energy in the membrane can be used by the cell to drive at least four different processes. 1) Exchange of cytosolic leaflet lipids with a bulky headgroup (such as phosphatidylinositol-4-phosphate or PS) for sterol or ceramide by lipid transfer proteins can relieve molecular crowding stress and facilitate membrane remodeling in the Golgi (sphingolipid synthesis and sterol loading). 2) Membrane bending to support budding or scission of protein transport carriers. 3) Displacement of lipids capable of spontaneous flip-flop to the luminal leaflet. Cholesterol is an example of a lipid that flip-flops rapidly between leaflets in an energy-independent manner. Coupling of PS flip to the cytosolic leaflet with cholesterol flop to the luminal leaflet could play an important role in lateral segregation of sterol, sphingolipid and protein in raft-like structures for sorting into exocytic carriers. 4) Rapid exposure of PS in the outer leaflet of cells through activation of a scramblase. This signaling event plays a critical role in apoptosis, blood clotting, immune system suppression, fusion of myoblasts and bone formation.

Several observations support the biological significance of these effects of P4-ATPase catalyzed lipid transport on vesicular transport. Golgi cisternae and endosomes are highly curved tubular-vesicular structures in budding yeast, but drs2 and neo1 mutants display a loss of Golgi curvature and protein transport defects [10,11]. In addition, the PS flippase activity of Drs2 is required to generate the curvature and anionic surface charge on these membranes needed for membrane recruitment of an ArfGAP (Gcs1) via its curvature and charge-sensitive ALPS (ArfGAP Lipid Packing Sensor) motif [9]. The PS flippase activity of ATP8A1 and ATP8A2, mammalian orthologs of Drs2, is similarly required to recruit EHD1 (Eps15 Homology Domain-containing protein 1) to recycling endosomes [12]. TAT-5 deficiency in C. elegans causes substantial shedding of plasma membrane derived extracellular vesicles [13,14]. This observation suggests that TAT-5 flippase activity normally counterbalances a force on the plasma membrane that promotes outward bending (negative curvature). P4-ATPases are required to support COPI-[11,15], clathrin adaptor- [10,16–19], and retromer-dependent trafficking routes [20–22], suggesting the coat proteins are taking advantage of flippase membrane-bending power to facilitate carrier formation.

Many different transport factors are argued to drive the membrane curvature needed for carrier formation (such as coat proteins, small GTPases and BAR-domain proteins [23]) and so it is difficult to tease out the precise contributions of any one protein to membrane deformation. For example, P4-ATPase loss of function trafficking defects could be a direct result of a loss of curvature in the donor membrane, or a secondary effect of perturbing the activity of other membrane-bending, cargo sorting, or membrane scission factors. However, a recent study demonstrates a gain-of-function condition where a P4-ATPase activity induces membrane curvature [24]. Expression the phosphatidylcholine (PC) flippase ATP10A in cells that do not normally express this P4-ATPase induces tubulation of the plasma membrane. PC is normally enriched in the plasma membrane outer leaflet of cells and its redistribution to the cytosolic leaflet should impart stress on the membrane to drive curvature. It remains to be determined how ATP10A influences membrane architecture and endocytosis in cells where it is normally expressed. In contrast, several other P4-ATPases are known to play critical roles in protein trafficking and are functionally intertwined with well-characterized transport machineries.

Trafficking pathways requiring ATP8A1, ATP8A2 and TAT-1

The group of orthologous PS/phosphatidylethanolamine (PE) flippases have been linked to endosomal recycling pathways in mammalian cells [12] and C. elegans [25,26](Figure 2). Depending on the cell type, ATP8A1 and ATP8A2 localize to endosomes, Golgi or the plasma membrane, although the primary site of residence appears to be Rab11-positive recycling endosomes (REs) [3,12]. Depletion of ATP8A1 in COS-1 cells results in a defect in transferrin recycling and its accumulation in REs, and a defect in shiga toxin transport from REs to the Golgi [12]. Morphologically, REs accumulates long thin tubules in ATP8A1 depleted cells and phenocopies cells depleted for EHD1, which is thought to facilitate fission of tubular carriers. In fact, EHD1 is a PS-binding protein with preference for curved membranes and its recruitment to the RE requires ATP8A1 activity [12]. These observations are consistent with prior reports showing a critical role for TAT-1 in RME-1 (EHD1 ortholog)-dependent endosomal protein recycling in C. elegans, although RME-1 recruitment to tat-1 endosomes appeared normal [25–27]. ATP8A2 is expressed in neurons, but it can replace the endocytic recycling function of ATP8A1 in COS-1 cells and primary neurons from Atp8a2^−/− mice accumulate transferrin receptors internally [12]. Thus, ATP8A1, ATP8A2 and TAT-1 likely function similarly to support endocytic recycling by recruiting EHD1/RME-1 to help drive scission of tubular carriers.

Figure 2. Roles for ATP9A and ATP8A1 in endosomal recycling pathways.

ATP8A1 (TAT-1 in C. elegans) and EHD1 (RME-1 in C. elegans) are required for transport of endocytosed proteins from recycling endosomes to the Golgi and for return back to the plasma membrane. Transport of PS to the cytosolic leaflet by ATP8A1 helps recruit EHD1 to recycling endosomes (RE) to potentially drive scission of tubular transport carriers. ATP9A (TAT-5 in C. elegans and Neo1 in S. cerevisiae) and Snx3-retromer is required for protein transport from early endosomes (EE) to the Golgi. MON2 and DOPEY2 are also linked to the function of ATP9A in this pathway.

Deficiency of ATP8A2 causes severe neurological disease that might be attributable to defects in endosomal recycling. An ATP8A2 mutation causes a striking disease in humans called cerebellar ataxia, mental retardation, disequilibrium syndrome (CAMRQ) where affected individuals walk on all fours [28]. In addition, the wabbler-lethal mouse carries a mutation Atp8a2^−/− and this mouse displays progressive motor neuron degeneration and early death [29]. Interestingly, the phenotypically similar wobbler mouse, a model for amyotrophic lateral sclerosis, harbors a mutation in Vps54, a subunit of the Golgi-Associated Retrograde Protein (GARP) complex involved in tethering endosome-derived carriers to the Golgi [30]. Mutations in four different genes (VLDLR, WDR81, CA8 and ATP8A2) cause CAMRQ and WDR81 function is also linked to endosomal trafficking [31,32]. WDR81 is a BEACH (Beige and Chediak Higashi) domain and WD40 repeat protein (SORF-1 in C. elegans) that regulates phosphatidylinositol-3-phosphate (PI3P) levels on endosomes [31]. In brain, the very low density lipoprotein receptor (VLDLR) is a receptor for reelin, a secreted patterning factor involved in neuronal migration and synaptic plasticity [33]. A speculative model is that ATP8A2 and WDR81 mutations perturb recycling of VLDLR (and other neuronal membrane proteins) through the endosomal system to cause CAMRQ. The human, mouse and C. elegans genetics support an endosomal recycling role for ATP8A/TAT-1, consistent with foundational studies implicating their budding yeast ortholog, Drs2, in protein trafficking [9,10,15,19,34,35].

Trafficking pathways requiring Drs2

Drs2 localizes to the TGN/endosomal system and its flippase activity is acutely required for sorting of AP-1/clathrin cargos [10,18]. Drs2 also contributes to trafficking pathways mediated by AP-3 and GGA adaptor proteins, but does so redundantly with Dnf1, Dnf2 and Dnf3 P4-ATPases [19]. The specific requirement for Drs2 in AP-1 function can be attributed to its ability to flip PS because Drs2 mutations that perturb recognition pf PS, but not PE, cause these trafficking defects [9]. Moreover, gain of function mutations in Dnf1 that greatly increase its ability to flip PS allow these Dnf1^PS variants to replace Drs2’s trafficking functions [9]. Surprisingly, AP-1 appears to be recruited normally to the trans-Golgi network (TGN) in drs2 mutants, but is nonfunctional [18]. This result implies that AP-1/clathrin coat assembly does not provide adequate energy to bend the Golgi membranes in the absence of the flippase. Moreover, fungal tetrameric adaptors lack clathrin binding motifs found in metazoan adaptors, and AP-2 and AP-3 can function independently of clathrin in yeast [17,36]. In addition, all of the yeast adaptors, including AP-2, are functionally linked to flippases [17–19,37]. We speculate that cells compensate for weak (or lost) interactions between fungal tetrameric adaptor proteins and clathrin heavy chain by using flippases to drive curvature of the transport carriers. The observed loss of AP-1 function does not explain all of the drs2 trafficking defects because endocytic recycling of an exocytic v-SNARE, Snc1, back to the plasma membrane requires Drs2 but is unaffected in clathrin adaptor or retromer mutants [19,34,38,39].

We recently discovered a surprising role for COPI in the Drs2-dependent recycling of Snc1 (Figure 3) [15]. Snc1 is ubiquitinated on several lysine residues within the SNARE motif and mutation of the preferred lysine, or attachment of a potent deubiquitinase to GFP-Snc1, disrupts its recycling [15,40]. Ubiquitin is well known to form a sorting signal on membrane proteins for endocytosis and sorting of protein into intraluminal vesicles [41], but these studies implicate ubiquitin as a recycling sorting signal and COPI as the coat acting in this pathway [15,40]. The clathrin-like α and β’ subunits of COPI contain WD40 repeats that form twin β-propeller structures, which bind to polyubiquitin with specificity for the chain length (3 or more) and linkage (K63)[15]. Deletion of a single propeller from β’COP perturbs Snc1 recycling, but replacing this propeller with a small domain that binds specifically to K63-linked polyubiquitin (the NZF domain from Tab2) fully restores Snc1 recycling [15]. Rcy1 is an F-box protein (potential substrate adaptor for SCF ubiquitin ligases) needed for Snc1 recycling [42,43], but Rcy1 may function in this pathway through binding and activation of Drs2 rather than as a ubiquitin ligase [15,39]. The role of Drs2 PS flippase activity in this pathway is only partially understood. Drs2 helps establish the negative charge and curvature required to recruit Gcs1 to TGN/EE membranes, and this ArfGAP binds to both COPI and Snc1 [44,45]. However, the gcs1 mutant displays a weaker Snc1 recycling defect than drs2 or COPI mutants [9], and so it is likely that additional Drs2 effectors are acting in this pathway. Whether or not flippases, polyubiquitin and COPI contribute to SNARE (e.g. VAMP2) recycling in animal cells remains to be determined.

Figure 3. Model for how a PS flippase (Drs2-Cdc50) contributes to the polyubiquitin- and COPI-dependent recycling of an exocytic SNARE.

Snc1 is a t-SNARE that functions in the fusion of exocytic carriers with the plasma membrane. After endocytosis, Snc1 recycles through the endosomal system back to the Golgi for re-packaging into exocytic carriers. This recycling process requires Drs2-Cdc50, Rcy1, the ArfGAP Gcs1, COPI, and modification of Snc1 by K63-linked polyubiquitin. Rcy1 binds to a C-terminal regulatory domain on Drs2 and likely stimulates flippase activity. PS flip to the cytosolic leaflet increases membrane curvature and negative charge to recruit Gcs1, which binds to both COPI and Snc1. In addition, the twin β propellers of α and β’ COP bind directly to K63-linked polyubiquitin and this interaction is essential for Snc1 recycling.

Trafficking pathways requiring ATP9A, ATP9B, TAT-5, and Neo1

These orthologous P4-ATPases are also linked to protein trafficking in the Golgi and endosomal system, and are unusual because they do not require a β subunit from the Cdc50 family to function [11,20–22,46–48]. In addition, phospholipid flippase activity has not yet been definitely demonstrated for any of these proteins through purification and reconstitution. However, knockdown or mutations in TAT-5 and Neo1 cause a loss of PE asymmetry in C. elegans [13], and both PE and PS asymmetry in budding yeast [49]. Thus, it is likely that these proteins are PE or PE/PS flippases. Neo1, ATP9A and ATP9B localize to the Golgi and endosomal system, and although TAT-5 localizes to the plasma membrane, it requires endosomal recycling to maintain cell surface localization [14,20,46]. Neo1, ATP9A and TAT-5 engage in a conserved set of interactions with Snx3, DOPEY2 (Dopl in yeast, PAD-1 in C. elegans) and MON2 (an ArfGEF-related protein), and are functionally linked to retromer [21,22,50].

The core retromer heterotrimer (Vps26, Vps29, Vps35) forms two distinct complexes with sorting nexins, either a SNX-BAR heterodimer or a PX-SNX (Snx3), and the retromer-Snx3 complex specifically retrieves Neo1 from endosomes back to the Golgi [22]. In addition, mutation of the sorting signal in Neo1 required for interaction with Snx3 not only disrupts Neo1 trafficking, but another retromer-Snx3 cargo as well. Thus, it appears Neo1 is needed to form the retromer-Snx3 carriers [22]. SNX3 and ATP9A/TAT-5 are also required for retromer-dependent retrieval of Wntless from endosomes to the Golgi so it can be reused for Wnt secretion in animal cells [21]. SNX3 interacts with the DOPEY-MON2 complex, and depletion of ATP9A/TAT-5, DOPEY2 or MON2 disrupts Wntless trafficking and Wnt signaling. This is consistent with prior studies demonstrating functionally important interactions between Dopl, Mon2 and Neo1 in budding yeast [22,50]. In C. elegans, MON-2 plays a role in TAT-5 recycling from the endosomal system redundantly with sorting nexins, although curiously independent of core retromer subunits, and it appears the DOPEY (PAD-1) interaction with TAT-5 regulates its PE flippase activity. Like tat-5, pad-1 and mutants defective in TAT-5 recycling excessively shed extracellular vesicles [14].

Neo1 and Dop1 are essential for viability in budding yeast [11,51], in contrast to the near normal growth of retromer mutants [52]. Therefore, Neo1 and Dop1 must have other, retromer-independent functions necessary for growth. Conditional neo1 mutants display several Golgi-associated defects, including mislocalization of COPI-dependent retrograde cargos targeted to the ER [11]. Thus, the remaining four P4-ATPases do not normally compensate for loss of Neo1. However, neo1Δ, dop1Δ or mon2Δ cells grow well when ANY1, encoding a conserved PQ-loop membrane protein, is knocked out [53]. Growth of neo1Δ any1Δ cells requires Drs2 and it appears that Drs2 can carry out Neo1’s essential function these cells [53]. How Anyl segregates Neo1 and Drs2 functions is still unclear, although it has been suggested that Any1 could be a scramblase that dissipates the phospholipid gradients formed by Drs2 and Neo1 in the Golgi [53].

Concluding remarks

Membrane remodeling by P4-ATPases plays a crucial role in protein trafficking, although there are many gaps in our understanding of the molecular basis for this role. The activity of ATP9A, TAT-5 and Neo1 in facilitating retromer-Snx3 carrier formation (or cargo sorting) from endosome for delivery to the Golgi is a great example of a highly conserved P4-ATPase function in a specific trafficking pathway [21,22]. However, there is no direct evidence that these P4-ATPases are bona fide flippases, comparable to studies with Drs2 or ATP8A1 [54,55]. The biochemical functions of MON2, DOPEY, and Any1 “regulators” of ATP9A/TAT-5/Neo1 remain obscure, as does the essential function of Neo1 and TAT-5 and their relationship to COPI-dependent transport from Golgi cisternae. The most clearly conserved function of ATP8A/TAT-1/Drs2 is in endosomal recycling. Drs2 has been linked to COPI and AP-1 function for trafficking between the Golgi and endosomes, but the nature of the donor compartment (TGN, recycling endosome, early endosome?) and directionality of transport is not clear. The tubular carriers ATP8A1/ATP8A2/TAT-1 help generate from recycling endosomes in animal cells are linked to EHD1 (RME-1) activity, but the precise function of this ATPase and how cargo is sorted into these carriers is unknown. While recent advances in the field have been impressive, we have a long way to go before the final chapter on phospholipid flippases in protein transport can be written.