The lipid flippase ATP8A1 regulates the recruitment of ARF effectors to the Trans-Golgi Network

Abstract

Formation of transport vesicles requires the coordinate activity of the coating machinery that selects cargo into the nascent vesicle and the membrane bending machinery that imparts curvature to the forming bud. Vesicle coating at the trans-Golgi Network (TGN) involves AP1, GGA2 and clathrin, which are recruited to membranes by activated ARF GTPases. The ARF activation at the TGN is mediated by the BIG1 and BIG2 guanine nucleotide exchange factors (GEFs). Membrane deformation at the TGN has been shown to be mediated by lipid flippases, including ATP8A1, that moves phospholipids from the inner to the outer leaflet of the TGN membrane. We probed a possible coupling between the coating and deformation machineries by testing for an interaction between BIG1, BIG2 and ATP8A1, and by assessing whether such an interaction may influence coating efficiency. Herein, we document that BIG1 and BIG2 co-localize with ATP8A1 in both, static and highly mobile TGN elements, and that BIG1 and BIG2 bind ATP8A1. We show that the interaction involves the catalytic Sec7 domain of the GEFs and the cytosolic C-terminal tail of ATP8A1. Moreover, we report that the expression of ATP8A1, but not ATP8A1 lacking the GEF-binding cytosolic tail, increases the generation of activated ARFs at the TGN and increases the selective recruitment of AP1, GGA2 and clathrin to TGN membranes. This occurs without increasing BIG1 or BIG2 levels at the TGN, suggesting that the binding of the ATP8A1 flippase tail to the Sec7 domain of BIG1/BIG2 increases their catalytic activity. Our results support a model in which a flippase component of the deformation machinery impacts the activity of the GEF component of the coating machinery.

INTRODUCTION

Vesicular traffic facilitates the transport of soluble and integral membrane proteins between distinct compartments of the secretory and endocytic pathways [1]. The formation of transport vesicles from distinct secretory and endosomal compartments involves the participation of multiple distinct cellular machineries. At the trans-Golgi network (TGN), the coating machinery selects the proteins to be transported and consists of soluble coat proteins that are recruited to membranes by members of the ARF superfamily after the ARFs are activated by the Sec7 domain-containing GEFs. On the other hand, the membrane deformation machinery refers to the various proteins and lipids involved in reshaping cellular membranes during vesicle formation. At the TGN, one of the mechanisms mediating membrane deformation involves the transmembrane lipid flippases that translocate phospholipids to the cytoplasmic leaflet of membrane bilayers to induce curvature necessary for vesicle formation [2-4]. A connection between the coating and the membrane deformation machineries has been described in the yeast S. cerevisiae through analyzing the role of the Drs2p flippase in the bi-directional transport between TGN and early endosomes and in the release of exocytic vesicles containing different cargoes [5-7]. This work documented an interaction between Drs2p and the yeast GEFs Gea1p and Gea2p, by showing that the C-terminal tail of the Drs2p flippase interacts directly with the Sec7 domain of the GEFs. Moreover, the flippase-GEF interaction was shown to have functional importance in trafficking by documenting that a mutation within the Sec7 domain of Gea2p necessary for binding the flippase, generates secretory granules/vesicles with abnormal morphology [6, 8]. Importantly, it was shown that the GEF-flippase interaction stimulates the catalytic activity of the flippase [9-11], and that the GEF stimulates the flippase activity synergistically with PI4P [6, 11, 12]. Thus, the coating machinery appears to regulate the membrane deformation machinery. However, whether functional flippase-GEF interactions are also found in mammalian cells has not been previously reported.

Vesicle coating at the Golgi and the TGN of mammalian cells requires ARF activation by GEFs belonging to a three-member subfamily that includes GBF1, BIG1 and BIG2 (the so called “large” GEFs) in a 15 member family characterized by a highly conserved Sec7 domain [13-17]. The Sec7 domain, initially identified in the yeast Sec7p, represents the catalytic domain of GEFs and it alone can facilitate ARFs activation [18, 19]. All the “large” GEFs are >200kDa cytoplasmic proteins that transiently associate with membranes of specific compartments. When membrane-bound, GEFs mediate GDP/GTP exchange on ARFs to allow their conformational shift from an inactive to an activated form that can interact with downstream effectors such as coats. The mammalian GEFs GBF1, BIG1 and BIG2 are inhibited by the fungal metabolite Brefeldin A (BFA), and vesicular traffic in the secretory and endocytic pathways is inhibited when cells are treated with BFA [20, 21].

GBF1 (the mammalian ortholog of Gea1/2p) mainly localizes to the membranes of the ER-Golgi-Intermediate Compartment (ERGIC) and the Golgi, where it facilitates ARF activation required for the formation of COPI-coated vesicles [22-25]. Additionally, GBF1 is involved in membrane recruitment of several other effectors such as Mint3, GCC8, GMAP210, golgin-160, golgin-97 and FAPP2 [26]. BIG1 and BIG2 (the mammalian orthologs of yeast Sec7p) predominantly localize to the TGN and recycling endosomes (RE), and participate in the formation of clathrin coated vesicles containing AP1 and GGA adaptors [14, 25-27]. BIG1-mediated events are required for glycosylation and function of β1 integrin [28], while BIG2 catalyzes coating events involved in the trafficking of E-cadherin, β-catenin, GABA A receptor, transferrin and Mannose 6 Phosphate Receptor [29-32]. Moreover, both GEFs appear involved in vesicular trafficking of furin and both regulate traffic of proteins involved in cell migration [15, 32, 33].

The mammalian flippase ATP8A1 is an enzyme with 10 transmembrane spanning regions that belongs to the P4-ATPase (class 1a) family and localizes in the TGN, endosomes and the plasma membrane [34-36]. Structurally, P4-ATPase flippases are heterodimers composed of a catalytic α subunit and a β subunit belonging to the CDC50 protein family [37]. The association between an α subunit and a CDC50 is required for stability, ER export and catalytic activity of the P4-ATPases [35]. P4-ATPase flippases are ATP-powered pumps that actively translocate phospholipids (mainly phosphatidylserine) towards the cytoplasmic leaflet of the cell membrane leading to an asymmetric distribution of phospholipids across the two leaflets [38-40]. This asymmetry leads to the curving of the membrane bilayer and is believed to contribute to the formation of nascent buds from a usually more planar membrane of the donor compartment [41].

The documented in the yeast S. cerevisiae functional relationships between the coating (Gea2p) and the curvature-inducing (Drs2p) events prompted us to explore whether an analogous interaction between GEFs and flippases may occur in mammalian cells, and whether such an interaction might be functionally relevant. Herein, we demonstrate that the mammalian flippase ATP8A1 and two GEFs, BIG1 and BIG2 co-localize and interact at the TGN, and that the interaction occurs through the C-terminal tail of the flippase and the catalytic Sec7d of the GEFs, the same domains shown to mediate the interaction of the yeast proteins. We also show that the flippase-GEF interaction has functional consequences in mammalian cells, and that the overexpression of the ATP8A1 flippase, but not of tail-less ATP8A1 unable to bind GEFs, results in increased generation of activated ARFs at the TGN and increased ARF-dependent membrane recruitment of only BIGs-dependent coating machinery (AP1, GGA2 and clathrin). Our results indicate that interactions between GEFs and flippases are evolutionarily conserved in organisms as diverse as yeast and humans, and suggest that the flippase components of the deformation machinery functionally intersects with the GEF components of the coating machinery.

RESULTS

1. ATP8A1 co-localizes with BIG1 and BIG2 at the TGN and recycling endosomes (REs).

The yeast Drs2p flippase shows the highest similarity to the mammalian ATP8A1, with analogous domain organization and extensive overall sequence homology (~46% identity and ~68% conserved amino acid sequence substitutions) (Figure 1A-B). The yeast protein has a significantly longer N-terminal domain, which accounts for its larger size. The C-terminal tail or the yeast Drs2p is also longer, and shows limited overall sequence conservation with the C-terminal tail of the mammalian ATP8A1 (~26% identity and no conserved substitutions). The region of Drs2p (residues 1250-1270) shown to mediate GEF binding (green box in Figure 1B) [6] contains only a single identical amino acid and no conserved substitutions. However, there are regions of high sequence identity between the yeast and the mammalian flippases, which might play a role in their functions (discussed later).

Figure 1. Domain structure and amino acid sequence conservation between yeast and mammalian flippases and GEFs.

(A) Schematic representation of Drs2p and ATP8A1 showing overall domain organization including the N-terminus (yellow), the transmembrane motifs (grey), the Actuator (A) domains (green) that have phosphatase activity and dephosphorylate the Phosphorylation (P) domains (blue), the Nucleotide binding (N) domains (orange) that phosphorylate the P domain, and the C-terminus tails (purple). Amino acid numbers are shown to show differences in flippase size between the species. (B) Alignment of Saccharomyces cerevisiae Drs2p (P39524·ATC3_YEAST) and Homo sapiens ATP8A1 (Q9Y2Q0·AT8A1_HUMAN) C-tail sequences. Identical residues are in dark blue. The C-terminus tail sequences are highlighted in pink. The amino acids within Drs2p tail that bind Sec7d (amino acids 1250-1270) are enclosed in a green box. (C) Schematic representation of Sec7p, BIG1 and BIG2 showing overall conservation of structure and domain organization including: DCB (dimerization and cyclophilin-binding domain) (yellow), HUS (homology upstream of Sec7d) (green), the Sec7 domain (purple), HDS1–4 (homology downstream of Sec7 1–4) (blue). The position of the catalytic “glutamic finger” residues within the Sec7d is indicated (red). (D) Alignment of Saccharomyces cerevisiae Sec7p (P11075 · SEC7_YEAST), Homo sapiens BIG1 (Q9Y6D6· BIG1_HUMAN) and Homo sapiens BIG2 (Q9Y6D5 · BIG2_HUMAN) Sec7 domain sequences. Identical residues are in dark blue and conserved residues are in light blue. The Sec7 domains sequences are highlighted in pink. The amino acids within the Sec7 domain that are important for interaction with Drs2p flippase (amino acids 912-1002) are enclosed in a green box. The catalytic “glutamic finger” residues within the Sec7d is indicated (red box).

In yeast, the Drs2p flippase localizes mainly to the TGN and has been shown to interact with the Golgi localized GEF Gea2p [6, 42]. The exact cellular site of this interaction is unclear since Gea2p has been previously localized to the Golgi and only recently reported to be present at the interface between the Golgi and the TGN compartments [6, 43]. Drs2p is necessary for recruiting the coat clathrin and adaptors such as AP-1 and GGAs to allow the formation of vesicles that mediate cargo exit from the TGN [44]. This coating mechanism is well conserved from yeast to mammalian cells.

The mammalian ATP8A1 has been localized to the general Golgi-TGN region [45, 46], and to provide insight into a possible flippase-GEF interaction, we re-examined the localization of ATP8A1 in HeLa cells. We utilized a full-length human ATP8A1 tagged at the C-terminus with HA due to the lack of workable antibodies that detect the endogenous protein. The construct was co-expressed with CDC50A, as previous studies demonstrated that ATP8A1 requires the CDC50A chaperone as its β-subunit to exit the ER and reach its subcellular localization [35, 45, 47]. ATP8A1-HA localization was assessed 24 hrs post transfection by immunofluorescence relative to organellar markers (Figure 2A). As shown in Figure 2B, ATP8A1 extensively co-localizes with the TGN marker Golgin-245 (0.67 ± 0.08 mean Pearson value ± SE) and with transferrin receptor (TfR) (0.65 ± 0.10 mean Pearson value ± SE), a marker of recycling endosomes (RE), but shows significantly lower colocalization with the cis-Golgi marker GM130 (0.54 ± 0.14 mean Pearson value ± SE). This suggests that ATP8A1 is concentrated at the TGN and RE.

Figure 2. ATP8A1 localizes to the TGN and RE.

(A) HeLa cells were co-transfected with Fl-ATP8A1 tagged with HA and CDC50-myc, and grown overnight. Cells were then fixed and processed by immunofluorescence with anti-HA (to detect the flippase) and either anti-GM130 (cis Golgi marker), anti-Golgin 245 (TGN marker) or anti-TfR (RE marker). A higher magnification of sections from the merged images are shown in right panels. Bars = 10 μm. (B) The amount of FL-ATP8A1 that co-localized with GM130, Golgin 245 or TfR was assessed and is presented as a Pearson coefficient. The means were compared by unpaired t test (***, P<0.0001; NS, not significantly different). n >20 transfected cells.

To assess the likelihood of ATP8A1 interacting with a specific “large” GEF, we defined the distribution of full-length ATP8A1 relative to GBF1, BIG1 and BIG2, the three GEFs required for coating at the TGN and RE [14, 15, 25-27]. GBF1 predominantly localizes to the ERGIC and the Golgi [22-25], but although at a lower concentration, GBF1 is also detected at the TGN [25]. BIG1 localizes predominantly to the TGN, while BIG2 has a wider distribution that includes the TGN and the RE [48, 49]. As shown in representative images in Figure 3A and quantified in Figure 3B, ATP8A1 exhibits relatively low co-localization with GBF1 (0.36 ± 0.13 mean Pearson value ± SE), but shows higher co-localization with BIG1 (0.58 ± 0.14 mean Pearson value ± SE) and BIG2 (0.55 ± 0.10 mean Pearson value ± SE), suggesting that ATP8A1 may preferentially interact with BIG1 and/or BIG2 to facilitate vesicle coating at the TGN and the RE.

Figure 3. ATP8A1 colocalizes with BIG1 and BIG2.

(A) HeLa cells were co-transfected with FL-ATP8A1 tagged with HA and CDC50-myc, and grown overnight. Cells were then fixed and processed by immunofluorescence with anti-HA (to detect the flippase) and either anti-GBF1, anti-BIG1 or anti-BIG2. A higher magnification of sections from the merged images are shown in right panels. Bars = 10 μm. (B) The amount of FL-ATP8A1 that co-localized with GBF1, BIG1 or BIG2 was assessed and is presented as a Pearson coefficient. The means were compared by unpaired t test (***, P<0.0001; NS, not significantly different). n >20 transfected cells.

2. Dynamic behavior of BIG1 and ATP8A1 containing structures.

ATP8A1 and BIG1 show the highest level of co-localization at the TGN in static images, and to gain insight into the extent of their co-localization during membrane trafficking, we assessed their dynamics in live cells. First, HeLa cells were singly transfected with BIG1-GFP or doubly transfected with FL-ATP8A1-mCherry and CDC50-myc, and 24 hrs after transfection, live cells were imaged every 8 seconds for 5 minutes. BIG1-GFP structures or ATP8A1-mCherry structures we tracked and their trajectories were color coded with rainbow LUT in which blue color indicates the starting position of the particle, and red color indicates the latest position in all the imaged frames.

Monitoring the dynamics of BIG1-GFP (Figure 4A), we observed relatively immobile and long-lived TGN structures and two types of forming tubule like structures. The first type was fast and thin, usually budding from the TGN to travel to PM, and those accounted for ~30% of the recorded events (Figure 4A, filled arrowheads). Interestingly, some of the tubule like structures (5% of the recorded events) were traveling in the opposite direction, from the cell periphery to the TGN (not shown), possibly reflecting the role of BIG1 in recycling events [50]. The second type were more long-lived tubule like structures, formed from the TGN that appeared relatively stable and immobile, and remained largely stationary throughout the imaging (Figure 4A, empty arrowhead). Full time course is presented in Supplementary video 1.

Figure 4. Dynamics of BIG1 and ATP8A1 in live cells.

HeLa cells were transfected with BIG1-GFP (A) or co-transfected with FL-ATP8A1-mCherry and CDC50-myc (B), grown overnight and BIG1-GFP or FL-ATP8A1-mCherry containing particles were imaged every 8 seconds for 5 minutes. Temporal projections were created for the entire time of data acquisition and each particle was color-coded to reflect its motility over the 5-minute period. Particle color-coding spans the entire color bar (least motile particles are in white and blue, most motile particles are in red). A. The movement of BIG1-GFP tubule like structures that originate from the TGN and appear to travel towards the plasma membrane are shown (filled arrowheads). Long-lived tubule like structures containing BIG1-GFP were also observed (empty arrowheads). B. ATP8A1-mCherry tubule like structure (filled arrowheads) originates from the TGN and moves towards the plasma membrane. Long-lived tubule like structure containing ATP8A1-mCherry is also observed (empty arrowheads). A higher magnification of sections from the images are shown in right panels. Bars= 10 μm.

The dynamics of FL-ATP8A1-mCherry (Figure 4B) were similar to the events observed with BIG1-GFP. Most of the observed tubule like structures were polarized towards PM (30% of the events recorded, filled arrowhead), with occasional (5% of the recorded) tubule like structures traveling in the opposite direction (not shown). The latter might suggest the dual role of ATP8A1 in cargo transportation and recycling of endosomal proteins. It was previously reported that flippases are mostly found to participate in cargo transport to the PM [51]. As for BIG1-GFP, we also observed long-lived tubule like structures containing FL-ATP8A1-mCherry (Figure 4B, empty arrowhead). Full time course is presented in Supplementary video 2.

To examine the dynamics of BIG1 and ATP8A1 simultaneously, HeLa cells were triple transfected with BIG1-GFP, FL-ATP8A1-mCherry and CD50-Myc, and 24 hrs after transfection, live cells were imaged every 8 seconds for 5 minutes (Figure 5). We observed extensive colocalization of BIG1 with ATP8A1 (Figure 5A-C, filled arrowheads) in highly dynamic tubule like structures (~30% of the events recorded). Moreover, we detected tubule like structures positives for only BIG1 or ATP8A1 (Figure 5A-B, empty arrowheads). We also represented cropped higher magnification images showing the tracks that BIG1-GFP and FL-ATP8A1-mCherry comprised individually during the tubule formation dynamic (Figure 5C). However, it is important to point out that despite the fact that BIG1 and ATP8A1 colocalized, they seem to follow independent kinetic patterns during trafficking, suggesting a high heterogeneity of processes that both proteins are involved in. Full time course is presented in Supplementary video 3.

Figure 5. BIG1 and ATP8A1 traffic together in dynamic transport tubules.

HeLa cells were triple transfected with BIG1-GFP, FL-ATP8A1-mCherry and CDC50-myc, and grown overnight. Cells were then imaged and BIG1-GFP (A) or FL-ATP8A1-mCherry (B) containing structures were tracked simultaneously every 8 seconds for 5 minutes. The tubule like structures where BIG1 and ATP8A1 act together are indicated with filled arrowheads. The tubule like structures where BIG1 and ATP8A1 act independently are indicated with empty arrowheads. Temporal projections were created for the entire time of data acquisition and each particle is color-coded to reflect its motility over the 5-minute period. Particle color-coding spans the entire color bar (least motile particles are in white and blue, most motile particles are in red). C. A higher magnification of the sections shown in A-B is represented. A time course presenting the dynamic of the structures where BIG1-GFP (cyan) and FL-ATP8A1-mCherry (magenta) traffic together are represented (filled arrows). Individual sections for BIG1-GFP or FL-ATP8A1-mCherry are also represented (right panels). Bars =10 μm.

3. The GEF-flippase interaction is mediated by the Sec7 domain of the GEF and the C-terminal tail of the flippase

It has been shown that the C-terminal cytoplasmic tail of the yeast flippase Drs2p interacts with the catalytic Sec7 domain of the yeast GEF Gea2p [6]. To determine whether mammalian flippases and GEFs also interact and whether the same domains are mediating binding, we expressed in HeLa cells either a full length ATP8A1 (FL-ATP8A1) or a mutant lacking the C-terminal tail (ΔT-ATP8A1), both tagged at their C-termini with HA. The cells were lysed 24 hrs after transfection, and the lysates were incubated with glutathione beads coated with bacterially expressed Sec7 domain of BIG2 (amino acids 637-842) tagged at its N-terminus with GST or with beads coated with only GST. The amount of FL-ATP8A1 or ΔT-ATP8A1 bound to the beads was assessed after SDS-PAGE of the bound material and immunoblotting with anti-HA. As shown in Figure 6A, equivalent amounts of FL-ATP8A1 and ΔT-ATP8A1 are detected in the input fractions (lane 1 and 6; Western anti-HA tag panel). FL-ATP8A1 is recovered on BIG2 Sec7d beads (lane 3), but not on control GST beads (lane 2), despite equivalent amounts of GST-Sec7 and GST being present on the beads (Ponceau-stained panel). In contrast, ΔT-ATP8A1 does not bind to the BIG2 Sec7d-containing beads (lane 8) or the control beads (lane 7). Thus, mammalian ATP8A1 flippase interacts with the Sec7 domain of BIG2, and that interaction requires the C-terminal tail of the flippase.

Figure 6. The C-terminal tail of ATP8A1 interacts with the Sec7 domain of BIG2.

(A) HeLa cells were transfected with HA-tagged FL-ATP8A1 or ΔT-ATP8A1 and lysed 24 hrs later. The lysates were incubated with beads coated with GST or a chimera of GST fused to the Sec7 domain of BIG2. The starting lysate (input), the material bound to beads coated with GST alone (GST lanes) or GST-Sec7/BIG2 chimera were analyzed by SDS-PAGE and immunoblotted with anti-HA antibodies (Western anti-HA panel). FL-ATP8A1 but not ΔT-ATP8A1 bound to beads coated with GST-Sec7/BIG2, but not to beads coated with GST. Equivalent amount of purified GST or GST-Sec7/BIG2 were loaded on the beads (Ponceau stain panel). (B) HeLa cells were lysed and the lysates were incubated with beads coated with GST (GST) or a chimera of GST fused to the tail of ATP8A1 (GST-A1-tail). The starting lysate (input) and the bound and unbound material were analyzed by SDS-PAGE and immunoblotted with the indicated antibodies. BIG1 and BIG2 bound to GST-A1-tail beads, but not to GST alone beads. GBF1 and GAPDH didn’t bind to GST or GST-A1-tail beads. Similar amount of purified GST or GST-A1 tail were loaded on the beads, with GST-A1-tail construct showing partial degradation (Ponceau stain panel). (C) Gels analogous to that in (B) were quantified by densitometry, and the data are presented as % of total antigen in the input that bound to beads loaded with FL-ATP8A1 tail. The data represent mean ± SD of 3 independent experiments. The separated blots represent different parts of the same gel. Full-length blots are included in the Supplementary Information file.

To further confirm the flippase-GEF interaction and to probe whether ATP8A1 also interacts with BIG1, a reverse pull-down assay was used, in which glutathione beads were coated either with GST or GST fused to the ATP8A1 tail (amino acids 1049-1163), and then incubated with HeLa lysates. The bound material was analyzed by SDS-PAGE and immunoblotting with different antibodies. As shown in Figure 6B and quantified in Figure 6C, the beads coated with GST-flippase tail, but not beads coated with GST alone, bound significant amounts of endogenous BIG1 (24 ± 7.9 % of total input) and BIG2 (18.5 ± 1.5 % of total input), indicating that both GEFs interact with the flippase. Importantly, minimal amounts of the Golgi-localized GBF1 bound to the beads (only 5± 0.9% of total input recovered), indicating strong selectivity in the GEF-flippase interactions. Additionally, GAPDH was used as a negative control and was not recovered on the beads, further confirming the selectivity of the binding assay.

4. Expression of the ATP8A1 flippase increases levels of activated ARFs on TGN membranes.

Previous reports showed that the binding of the yeast Gea2p GEF to the Drs2p flippase stimulates the catalytic activity of the flippase [11]. We aimed to assess whether the reciprocal also might be true and whether a flippase (ATP8A1) could impact the catalytic activity of GEFs (BIG1 and/or BIG2) to increase ARF activation. Thus, we measured the levels of activated ARF at the TGN under conditions of increased levels of ATP8A1 at the TGN. As a negative control, we used the tail-less ATP8A1 unable to bind BIG1 and BIG2. Our analysis was based on the recruitment of a fluorescent GAT domain reporter (the GAT domain of GGAs (GGA-GAT), which is recruited to membranes via interaction with the GTP-bound activated form of ARFs [52-54]. In this assay, the amount of GGA-GAT recruited to membranes directly correlates with the levels of activated ARFs on the membrane. A GFP-tagged GGA-GAT construct was co-transfected with either FL-ATP8A1 or ΔT-ATP8A1 into HeLa cells. After 24 hrs, the total and TGN localized fluorescence intensity of GFP-tagged GGA-GAT was measured, and the percent of total intensity that was TGN associated (GGA-GAT/TGN %) was calculated. As shown in Figure 7A and quantified in Figure 7B, cells expressing FL-ATP8A1 that can interact with BIG1 and BIG2 showed a significantly higher level of GGA-GAT in the TGN region than cells expressing the tail-less ATP8A1 unable to interact with BIG1 and BIG2. These results suggest that increasing the levels of TGN-localized ATP8A1 flippase increases the levels of activated ARFs at the TGN membrane.

Figure 7. Expression of FL-ATP8A1, but not ΔT-ATP8A1 increases ARF activation at the TGN.

(A) HeLa cells were co-transfected with CDC50-myc, GGA-GAT tagged with GFP and either FL-ATP8A1 or ΔT-ATP8A1 tagged with HA, and grown for 24 hours. Cells were fixed and processed for immunofluorescence with anti-HA (to detect the flippase) and anti-GFP (to detect GGA-GAT). A higher magnification of boxed regions from the merged images are shown in right panels. Bars = 10 μm. (B) Images analogous to those in A were used to manually trace cell outlines and ROIs containing flippase signal (outlines in A), and the intensity of GGA-GAT within both was measured and used to quantify the % of total cellular GGA-GAT recruited to the TGN region. At least 20 cells expressing GGA-GAT and either FL-ATP8A1 or ΔT-ATP8A1 constructs were assessed. ROI i: Region of Interest of the entire cell fluorescence. ROI ii: Region of Interest of the TGN region fluorescence (See methods). The means were compared by unpaired t test (***, P<0.0001) and significantly more GGA-GAT was recruited to the TGN in cells expressing FL-ATP8A1.

The increased levels of TGN-localized activated ARF in cells expressing ATP8A1 can result from increased levels of BIG1s at the TGN or increased catalytic activity of the same number of TGN-localized GEFs. To gain insight into the possible mechanism responsible for the increase in ARF activation, we assessed whether increasing the levels of ATP8A1 or tail-less ATP8A1 unable to bind BIGs, at the TGN affects the recruitment of endogenous GEFs to the membranes. We assessed the recruitment of the Golgi-localized GBF1 (as a negative control) and of the TGN-localized BIG1 and BIG2 in cells expressing FL-ATP8A1 or ΔT-ATP8A1. As shown in Figure 8A and quantified in Figure 8B, the percentage of cellular GBF1 present at the TGN was not significantly different in cells expressing FL-ATP8A1 or ΔT-ATP8A1. Importantly, the levels of BIG1 and BIG2 at the TGN also were not different in cells expressing FL-ATP8A1 or ΔT-ATP8A1. Thus, the observed increase in ARF activation in cells expressing ATP8A1 at the TGN is not due to increased levels of TGN-associated BIG1 or BIG2. Instead, our results suggest that the increase in ARF activation might be due to increased catalytic activity of the BIGs.

Figure 8. Expression of FL-ATP8A1 or ΔT-ATP8A1 doesn’t cause an increase in GEF recruitment to the TGN.

(A) HeLa cells were co-transfected with CDC50-myc and either FL-ATP8A1 or ΔT-ATP8A1 tagged with HA, and grown for 24 hours. Cells were fixed and processed for immunofluorescence with anti-HA (to detect the flippase) and the indicated antibodies. Bars = 10 μm. (B) Images analogous to those in A were used to quantify the % of total cellular GBF1, BIG1 or BIG2 recruited to the TGN region in cells expressing FL-ATP8A1 or tail-less ATP8A1. The means were compared by unpaired t test and no significant difference was found in GEF recruitment in cells expressing FL-ATP8A1 or ΔT-ATP8A1 flippase. n>20 cells.

5. Expression of the ATP8A1 flippase enhances GGA2, AP-1, clathrin and Golgin-245 recruitment to the TGN.

Activated ARFs mediate the recruitment of coats and other effectors to the membrane. GBF1-mediated ARF activation facilitates the recruitment of the COPI coat at the ER-Golgi interface [22-25], as well as the recruitment of several additional ARF effectors such as Mint3, golgin-97 and GCC8 to the TGN [26]. BIG1 and BIG2 have been shown to be involved in clathrin vesicle formation at the TGN [14, 27], and are necessary for generating activated ARFs that facilitate the recruitment of AP-1 and GGA2/3 adaptors [14, 25-27]. In addition, BIG1 and/or BIG2 facilitate the recruitment of the ARF effector golgin-245 (although golgin-245 seems to be also regulated at least partially by GBF1 [26]). We reasoned that if ATP8A1 increases ARF activation by BIG1 and BIG2, but doesn’t affect GBF1 activity, then expression of ATP8A1 might lead to increased membrane association of effectors selectively recruited through BIGs-mediated ARF activation, but not those recruited through GBF1-mediated ARF activation. Thus, we first examined the membrane association of BIGs-dependent and GBF1-dependent effectors in cells expressing FL-ATP8A1 or ΔT-ATP8A1 unable to bind BIGs. As shown in Figure 9A and quantified in Figure 9B, cells expressing FL-ATP8A1 showed a significantly higher recruitment of BIGs regulated coats and effectors: clathrin, AP-1, GGA2 and golgin-245 than cells expressing ΔT-ATP8A1.

Figure 9. Expression of FL-ATP8A1, but not ΔT-ATP8A1 causes increased recruitment of clathrin, GGA2, AP1 and Golgin-245 to the TGN.

(A) HeLa cells were co-transfected with CDC50-myc and either FL-ATP8A1 or ΔT-ATP8A1 tagged with HA, and grown for 24 hours. Cells were fixed and processed for immunofluorescence with anti-HA (to detect the flippase) and the indicated antibodies. Bars = 10 μm. (B) Images analogous to those in A were used to quantify the % of total cellular clathrin, GGA2, AP1 or Golgin-245 recruited to the TGN region in cells expressing FL or tail-less ATP8A1. The means were compared by unpaired t test (***, P<0.0001). Expression of FL-ATP8A1, but not ΔT-ATP8A1 induced a significant increase in effector recruitment to TGN. n>20 cells.

We then assessed the recruitment of GBF1-dependent effectors. As shown in Figure 10A and quantified in figure 10B, cells expressing FL-ATP8A1 or ΔT-ATP8A1 showed no significant difference in the recruitment of GBF1 regulated proteins: golgin-97, Mint3 and GCC8. Thus, the expression of ATP8A1 flippase capable of binding BIG1 and BIG2 increases levels of activated ARF at the TGN, and results in increased recruitment of only BIGs-specific TGN effectors to membranes.

Figure 10. ATP8A1 expression doesn’t recruit GBF1-dependent ARF effectors to the TGN.

(A) HeLa cells were co-transfected with CDC50-myc and either FL-ATP8A1 or ΔT-ATP8A1 tagged with HA, and grown for 24 hours. Cells were fixed and processed for immunofluorescence with anti-HA (to detect the flippase) and the indicated antibodies. Bars = 10 μm. (B) Images analogous to those in A were used to quantify the % of total cellular Golgin-97, Mint3 or GCC88 recruited to the TGN region in cells expressing FL or tail-less ATP8A1. The means were compared by unpaired t test (ns, not significant). Expression of FL-ATP8A1 or ΔT-ATP8A1 has no significant effect on the recruitment of these effector proteins to TGN. n>20 cells.

DISCUSSION

Membrane traffic is a highly regulated process that delivers proteins and lipids to their final destination throughout the cell. Membrane transport along the secretory and endocytic pathways requires the coordinated function of two machineries: the protein sorting machinery that selects a subset of proteins for transport from the donor compartment, and the membrane remodeling machinery that generates membrane curvature necessary to form a transport intermediate from the donor membrane. Protein sorting is facilitated by cytoplasmic coat proteins, such as the COPI heptameric coatomer that selects cargo for transit at the ER-Golgi interface and the AP1 and GGA1-3 adaptors and clathrin that mediate transport at the TGN-endosomal interface. In both transport pathways, the coats/adaptors are recruited to the cytosolic face of the membrane bilayer by binding to activated (GTP-loaded) forms of ARF GTPases. Coat recruitment is absolutely dependent on ARF activation by the large GEFs (GBF1, BIG1 and BIG2), since treating cells with the drug Brefeldin A (BFA) that inhibits only the large GEFs causes the dissociation of coats from membranes and inhibits cargo traffic [3, 29, 30, 33, 50]. GEFs are soluble proteins that rapidly cycle between a cytosolic and membrane associated pools, and their membrane attachment defines the site and time of ARF activation and the subsequent coat recruitment.

In addition to the sorting machinery, membrane traffic requires membrane deformation that induces curvature to a normally more planar donor compartment. Membrane bending is mediated by the activity of lipid flippases that transport lipids (predominantly PS, but also PC and PE) from the inner to the outer membrane leaflets leading to bilayer asymmetry and bending of the membrane [55]. Previous evidence suggests that flippases play a role in shaping biological membranes and in regulating traffic processes in mammals: for example, when overexpressed in HeLa cells, the flippase ATP10A causes plasma membrane tubulation and deformation by increased flipping of phosphatidylcholine (PC) [56]. Moreover, mice carrying mutations in the ATP8A2 flippase develop neurodegenerative disease due to an impairment in axonal transport processes [57]. Similarly, in COS-1 cells, ATP8A1 plays a crucial role in regulating membrane traffic through recycling endosomes by facilitating the recruitment of the membrane fission protein EHD1 to these organelles [58]. Furthermore, a recent study conducted in primary mouse alveolar type 2 (AT2) cells demonstrated that the interaction between ATP8A1 and the Adaptor Protein 3 complex (AP-3) at early endosomes is essential for the appropriate transport of soluble proteins to the Lamellar Bodies [59]. ATP8A1, in conjunction with CDC50A, also mediates the trans-bilayer translocation of phosphatidylethanolamine (PE) required for the reorganization of the plasma membrane to generate ruffles and facilitate CHO cells migration [60].

The sorting and the deformation machineries appear to interact, as shown in the yeast S. cerevisiae where the Sec7 domain of the GEF Gea2p (ortholog of GBF1) binds the C-terminal tail of the flippase Drs2p (homolog of ATP8A1) [6, 8]. Importantly, the interaction has functional consequences, as mutations in the Sec7 domain of Gea2p that is required for Drs2 binding impair secretion [6]. Moreover, it has been shown that Gea2p binding to Drs2p acts synergistically with PI4P to activate the flippase activity of Drs2p [12]. Thus, a component of the sorting machinery (Sec7p GEF) directly engages a component of the deformation machinery (Drs2 flippase) to stimulate its flippase activity.

Whether a relationship between a GEF and a flippase is evolutionarily conserved in mammalian cells has not been previously examined, and we undertook a series of experiments to assess their possible interaction in human HeLa cells. We focused on the sorting and coating machineries at the TGN, and decided to explore the relationship between the BIG and BIG2 GEFs known to localize to the TGN and the ATP8A1 flippase. We chose ATP8A1 because it is the most homologous to the yeast Drs2p flippase which localized to the TGN and endosomal membranes, where it facilitates clathrin-coated vesicle formation [4, 34]. We first defined the localization of ATP8A1 by immunofluorescence after transfecting tagged ATP8A1, and showed that the mammalian flippase also targets predominantly to the TGN and RE, with lower amounts detected on Golgi membranes (Figure 2). ATP8A1 localization suggested that it likely interacts with TGN and RE-localized GEFs, and indeed, ATP8A1 showed extensive co-localization with BIG1 and BIG2, but not with the Golgi-localized GBF1 (Figure 3). Moreover, when probed in live cells, ATP8A1 and BIG1 showed extensive co-localization in relatively static large TGN elements, as well as in dynamic tubules extending from the TGN, suggestive of a cooperative function in membrane traffic from the TGN (Figure 4 and 5).

The extensive co-trafficking of BIG1 and ATP8A1 in punctate and tubular structures in live cells suggested possible interaction, and this was confirmed by affinity pull-down assays (Figure 6). The mammalian BIG1and BIG2 appear to interact (either directly or within larger complexes) with ATP8A1, and the interaction seems to be mediated by the catalytic Sec7 domain of BIG2 and the C-terminal tail of ATP8A1, the same domains shown to mediate the interaction between the yeast Gea2p and Drs2p. In the yeast Gea2p, single or double mutations in the C-terminal third of the Sec-7 domain (amino acids ~643-733), abolished the interaction with Drs2p [6], and it is likely that the same domains mediate the intersections of the mammalian BIG1 and BIG2 with ATP8A1 since these regions of the Sec7 domain are highly conserved. The Sec7 domain of the yeast Sec7p shows extensive identity with the Sec7 domains of human BIG1 (54.6%) and BIG2 (53.8%) over ~130 amino acids (Figure 1C-D).

It has been shown that the ~20 amino acids region (amino acids 1250-1270) of Drs2p C-terminal tail is important for Gea2p recognition [6]. Interestingly, the C-terminal tail of Drs2p shows limited identity (25.6%) to the ~100 amino acids tail of ATP8A1 (Figure 1A-B).

The cytoplasmatic tail of yeast Drs2p has been shown to act as an autoinhibitory domain regulating its own lipid translocation activity [42]. The binding of the GEF to the tail has been proposed to relieve inhibition and thereby stimulate the catalytic activity of the flippase [11]. The Drs2p C-terminal tail (amino acids 1267-1273) also interacts with phosphatidylinositos-4-phosphate (PI4P), and Drs2p (in complex with Cdc50p) can be activated in vitro by PI4P incubation [6, 11]. Moreover, the flippase activation appears to be synergistically stimulated by simultaneously binding both the PI4P and the Gea2p GEF [11, 61]. It has been shown that the C-terminal tail of the ATP8A2 flippase is important for its proper folding and also has a regulatory function that can autoinhibit its PS flipping activity [62]. This view is further supported by cryo-EM structures of intermediates during phospholipid transport by the human ATP8A1 [63]. Thus, it is possible that analogous mechanisms are utilized to regulate flippase activity through GEF binding in yeast and in mammals.

Because the binding of the yeast Gea1p or Gea2p GEFs to the Drs2p flippase stimulates the catalytic activity of the flippase [6], we wanted to assess whether the reverse also could be true, i.e., whether a flippase could stimulate the catalytic activity of a GEF. We explored this using two cellular read-outs that measure ARF activation at the TGN in response to increased levels of the ATP8A1 flippase. First, we assessed the membrane recruitment of the GGA-GAT domain construct, which acts as a “sensor” to detect ARF activation by binding specifically to the GTP-bound form of ARF1 [53]. We observed significantly increased GGA-GAT recruitment to the TGN in cells expressing FL-ATP8A1, but not in cells expressing the tail-less ATP8A1 unable to bind GEFs (Figure 7), suggesting that promoting flippase-GEF interactions results in increased ARF activation. Such an increase could result from an increased recruitment of GEFs to the TGN membrane or be due to increased catalytic activity of the same number of membrane-associated GEFs. Importantly, we didn’t observe an increase in the levels of recruitment of GBF1, BIG1 and BIG2 to the TGN when ATP8A1 levels were increased (Figure 8), suggesting that the increased ARF activation likely results from increased catalytic activity of the TGN-localized BIGs. We stress that because this effect was only observed with full length ATP8A1 but not ΔT-ATP8A1 (i.e., requires the tail domain of ATP81), it is unlikely to be caused by flippase-induced changes in cytoplasmic leaflet lipid composition, because both, full-length and tail-less flippases are enzymatically active [42, 61-64].

Second, we assessed ARF activation by measuring the recruitment of ARF effectors, and more specifically, the recruitment of BIG1- and BIG2-dependent effectors versus GBF1-dependent effectors. We show that increasing the levels of ATP8A1 at the TGN stimulates the recruitment of clathrin, GGA2, AP-1 and Golgin-245 to TGN membranes (Figure 9-10). These effectors have been shown to be coupled to BIG1 and BIG2, with AP-1 distribution affected when BIG1 and BIG2 are depleted [15], and AP1, GGA2 and clathrin not being recruited to membrane when BIG1 and BIG2 (but not GBF1) are inhibited with BFA [25, 26]. Similarly, Golgin-245 recruitment to the TGN requires at least partial GEF activity of BIGs [25]. In contrast, multiple effectors dependent on GBF1-mediated ARF activation (Golgin-97, Mint3 and GCC88) did not show increased recruitment in cells expressing increased levels of ATP8A1. Thus, only ARF activation mediated by BIGs and the selective subsequent recruitment of the cognate effectors were stimulated by ATP8A1.

Together, our findings support a model (Figure 11/Graphical abstract) in which at resting state, TGN-localized ATP8A1 is inactive through autoinhibition (purple C-terminus autoinhibits the flippase), and the coating machinery components (BIG1, BIG2, ARF-GDP, adaptors and clathrin) are freely diffusing within the cytosol. To initiate vesicle formation, the ATP8A1 autoinhibitory tail binds PI4P and is now able to interact with the Sec-7 domain of BIG1 and BIG2 in a process that releases the autoinhibition within the flippase and recruits/positions the GEFs on the TGN membrane in close proximity to the flippase. Our results suggest that the flippase tail binds the catalytic Sec7d of the BIGs, and that such interactions may stimulate the catalytic activity of the GEF to generate activated ARFs at the TGN membrane. This then causes increased recruitment of BIG-specific ARF effectors such as the AP1/GGA clathrin adaptors. The membrane-binding of the adaptors then facilitates the sorting of cargo proteins into newly forming vesicles. Together, our results imply a close communication between the coating (GEF) and membrane deformation (flippase) machineries in mammalian cells. Effective coordination between the GEF and flippase activities could act as a way to synchronize the coating and membrane deformation machinery, leading to optimal vesicle formation.

Figure 11 (GRAPHICAL ABSTRACT). Working model for ATP8A1 and BIG1/2 function in effector recruitment at the TGN.

(A) ATP8A1 (C-terminal tail is in purple) resides in the TGN in a complex with its chaperone CDC50, and its flippase activity is auto-inhibited by its C-terminal tail. BIG1, BIG2 (the flippase-binding Sec7d is in pink), inactive (GDP-bound) ARFs, the adaptors (AP-1 or GGA2) and clathrin are cytosolic. Soluble and transmembrane cargo proteins are not sorted into a transport intermediate. (B) ATP8A1 binds PI4P and adopts a conformation that allow its C-terminal tail to interact with the Sec7 domain of BIG1 or BIG2. This leads to membrane recruitment of the BIGs, which then generate active (GTP-bound) membrane associated ARFs that recruit adaptors, which subsequently recruit clathrin triskelions. The adaptors also bind and sort cargo proteins by directly binding transmembrane protein or binding transmembrane receptors occupied by soluble proteins into the deformed patch of the TGN membrane.

EXPERIMENTAL PROCEDURES

Antibodies:

The following commercially available antibodies were used: monoclonals anti-GBF1, anti-GM130, anti-Golgin 245, anti-GGA2, and anti-Mint3 (catalog no. 612116, 610823, 611280, 612613, 611380, BD Transduction Laboratories); monoclonal anti-HA (catalog no. MMS-101R BioLegend); polyclonal anti-HA, monoclonal anti-BIG1, monoclonal anti-BIG2, monoclonal anti-AP1 (catalog no. H6908, MABS1247, MABS1246, A5968 Millipore Sigma); monoclonal anti-TfR (catalog no. 13-6800, Thermo Fisher Scientific); polyclonal anti-Clathrin, (catalog no. 1085-1-AP, Proteintech); monoclonals anti-GAPDH, anti-Golgin 97 (catalog no. ab8245, ab169287, Abcam); polyclonal anti-GCC88 (catalog no. GTX120148, GeneTex). Secondary anti-mouse antibody conjugated to HRP (catalog no. 1030-05, Southern Biotech). Secondary antibodies conjugated to Alexa-488 and Alexa-594 (catalog no. A11034, A11029, A11012, A11032, Invitrogen).

Reagents:

ECL Western blotting reagent was from Thermo Fisher Scientific (Waltham, MA). SuperSignal^® West Femto Maximum Sensitivity Substrate was from Thermo Scientific, (Chicago, IL). Complete Protease Inhibitors Cocktail, EDTA-free was from Thermo Scientific (IL, USA). 3-12% BN-PAGE gels and molecular weight standards for native gels (cat#LC0725) were purchased from Invitrogen (Madison, WI).

Plasmids:

the GGA-GAT-GFP tagged construct was kindly provided for Dr James Casanova (University of Virginia Health System, Charlottesville, VA, USA); the CDC50-myc tagged constructs were kindly provide for Dr Robert Molday (University of British Columbia, Vancouver, Canada); the carboxyl terminal HA tagged (FL-ATP8A1-HA) construct was constructed by cloning the full length of ATP8A1 into the pcDNA4 C-terminus HA tagged vector with HindIII and XhoI restriction enzymes. N-terminal GST tagged ATP8A1-tail (GST-ATP8A1-tail) was constructed by cloning the ATP8A1(1049-end) into the pGEX5X-3 with SmalI and XhoI. BIG2-Sec7 (637-842) construct was cloned into the N-terminal pGEX5X-3 with SalI and NotI. ATP8A1-HA Forward Primer is: GATCAAGCTTGATGCCCACCATGCGG; ATP8A1-HA Reverse Primer is: GATACTCGAGACCATTCGTCGGGCCT. ATP8A1 tail -GST-Forward Primer is: GATCCCCGGGAACTCCTTGATGTGGTGTAC; ATP8A1 tail -GST-Reverse Primer is: GATCAACTCGAGACCATTCGTCGGGCCTCTG. BIG2-Sec7-GST Forward Primer is: GATCGTCGACTTTGAGGTCATCAAGC. BIG2-Sec7-GST Reverse Primer is: GATGCGGCCGCTAGTAGATTTGGTT.

Sequence alignment:

for Saccharomyces cerevisiae Drs2p (P39524·ATC3_YEAST), Homo sapiens ATP8A1 (Q9Y2Q0·AT8A1_HUMAN), Saccharomyces cerevisiae Gea2p (P39993 · GEA2_YEAST), Homo sapiens BIG1 (Q9Y6D6· BIG1_HUMAN) and Homo sapiens BIG2 (Q9Y6D5 · BIG2_HUMAN) were retrieved from www.uniprot.org and performed using the Clustal Omega Program (1.2.4).

Cell culture and transfection:

Human HeLa (CCL-2) cell line was obtained from ATCC, The Global Bioresource Center, USA. Cells were cultured in vitro in MEM Eagle medium (Cellgro, Manassas, VA) supplemented with L- glutamine, 10% fetal bovine serum, 100 units/ml penicillin, 100 mg/ml streptomycin, and 1 mM sodium pyruvate (Cellgro, Manassas, VA) at 37°C in humidified atmosphere. Cells were grown at 37°C in 5% CO₂ in 6-well dishes until ~70% confluent and were transfected using Mirus TransIT-LT1 transfection reagent from Mirus Bio Corporation (Madison, WI), according to the manufacturer's instructions.

GST pull-Down, SDS-PAGE electrophoresis and Western blotting:

HeLa cells were transfected or not with HA-tagged FL-ATP8A1 or ΔT-ATP8A1 constructs, 24 h later cells were lysed in RIPA buffer (150 mM sodium chloride, 1.0%NP-40, 0.5% sodium deoxycholate, 50 mM Tris, pH 7.4) supplemented with protease inhibitor cocktail by repeated passage through a 26 and 27-gauge needle. The homogenate was centrifuged at 1000 g for 15 min at 4 °C in a microcentrifuge to remove unbroken cells and nuclei. The supernatant was incubated (as specified in each experiment) 1 h at R.T with BIG2 Sec7-GST beads, ATP8A1-tail-GST and GST beads as a control. Samples were washed 3 times in PBS. GST beads were then resuspended in SDS sample-buffer, boiled for 3 min and resolved by 8% SDS-PAGE prior to transfer to NitroPure nitrocellulose membrane (Micron Separations, Westborough, MA) by wet transfer overnight at 30 mV. Membranes were blocked with 5% fat free milk and probed with antibodies as indicated in each figure. The Western blot band intensity measurements were performed with ImageJ.

Imaging:

HeLa cells were seeded overnight on glass coverslips, transfected and ~18hours later processed for immunofluorescence. Cells were washed in PBS, fixed in 3% paraformaldehyde for 10 min, and quenched with 10 mm ammonium chloride. Cells were permeabilized with 0.1% Triton X-100 in PBS. The coverslips were then washed with PBS and blocked in PBS containing 2.5% goat serum, 0.2% Tween 20 for 5 min followed by blocking in PBS containing 0.4% fish skin gelatin, and 0.2% Tween 20. Cells were incubated with primary antibodies for 1 h at room temperature. Coverslips were washed with PBS containing 0.2% Tween 20 and incubated with secondary antibodies for 45 min. Nuclei were stained with DAPI dye. Coverslips were washed as described above and mounted on slides in ProLong Gold antifade reagent (Invitrogen, Madison WI). Fluorescence patterns were visualized using a Leitz Wetzlar microscope with epifluorescence and Hoffman Modulation Contrast optics by Chroma Technology, Inc (Bellows Falls, VT, USA). Images were captured with a 12-bit CCD camera from QImaging (Surrey, BC, Canada) and processed with iVision-Mac software. Confocal imaging was on a Nikon TE2000 inverted microscope. The system was equipped with laser and filter sets to visualize and image FITC, TRITC and DAPI fluorescence. Images were captured using a Hamamatsu C9100-50 EMCCD camera (Hamamatsu Photonics K.K., Hamamatsu city, Japan) and 60X Plan APO oil-immersion objective. The imaging system was operated by Nis Elements 5.0 Imaging Software (Melville, NY, USA).

Co-localization quotients:

At least 20 transfected cells perimeters were selected for each condition and the Pearson coefficient between FL-ATP8A1 (green) and GM130, Golgin 245, TfR, GBF1, BIG1 or BIG2 (red) localization were calculated by ImageJ. The dispersion graph data was represented as absolute values, where each dot corresponds to one transfected cell.

Percent of GGA-GAT/effectors on TGN quotients:

At least 20 transfected cells entire perimeters (ROI i) and TGN perimeters (ROI ii) were selected and the fluorescence intensity measured by ImageJ. Afterward the % of GGA-GAT/effectors in the TGN was calculated by using a spreadsheet. The dispersion graph data was represented as absolute values, where each dot corresponds to one quantified transfected cell.

Live-cell microscopy:

HeLa cells were plated and grown for 48h before transfection on 25 mm glass coverslips (Electron Microscopy Sciences, Hatfiled, PA, USA). Cells were transfected either separately or co-transfected with BIG1-GFP tagged and Fl-ATP8A1-mCherry tagged (together with CDC50-myc tagged) and imaged 24h post-transfection. For live-cell imaging coverslips were placed in titan chambers and located in thermostage with CO₂ supply (temperature was rising between 36 and 37 °C, 5% CO₂) followed by a change of cell medium for phenol red-free FluoroBrite^™ DMEM medium supplemented with 10% fetal bovine serum. The 5-min long movies of representative cells were performed with Nikon inverted microscope Eclipse Ti2 (Nikon, Melville, NY, USA). Temporal projections were generated in ImageJ using the Temporal-color code plugin.

Supplementary Material

1

Supplementary Video 1.

Live-cell imaging of HeLa cell expressing the BIG1-GFP plasmid from Figure 4 A and 4 A’. Images composing the movie were acquired every 8 s for 5 min.

2

Supplementary Video 2.

Live-cell imaging of HeLa cell expressing the FL-ATP8A1-mCherry and CDC50-myc plasmids from Figure 4 B and 4 B’. Images composing the movie were acquired every 8 s for 5 min.

3

Supplementary Video 3.

Live-cell imaging of HeLa cell expressing the BIG1-GFP (Cyan), FL-ATP8A1-mCherry (Magenta), and CDC50-myc plasmids from Figure 5. Images composing the movie were acquired every 8 s for 5 min.

HIGHLIGHTS.

• The lipid flippase ATP8A1 interacts with the Guanine Nucleotide Exchange Factors BIG1/BIG2 at the trans-Golgi network (TGN) in mammalian cells.

• ATP8A1 and BIG1/BIG2 interaction stimulates activated ARFs at TGN, recruiting AP1, GGA2, and clathrin.

• ATP8A1 co-localizes with BIG1/BIG2 in static and mobile TGN elements, linking membrane deformation with coating processes.

DATA AVAILABILITY

All data obtained in this study will be available from the corresponding author by written request.