ATP-binding cassette transporters and sterol O-acyltransferases interact at membrane microdomains to modulate sterol uptake and esterification

Abstract

A key component of eukaryotic lipid homeostasis is the esterification of sterols with fatty acids by sterol O-acyltransferases (SOATs). The esterification reactions are allosterically activated by their sterol substrates, the majority of which accumulate at the plasma membrane. We demonstrate that in yeast, sterol transport from the plasma membrane to the site of esterification is associated with the physical interaction of the major SOAT, acyl-coenzyme A:cholesterol acyltransferase (ACAT)-related enzyme (Are)2p, with 2 plasma membrane ATP-binding cassette (ABC) transporters: Aus1p and Pdr11p. Are2p, Aus1p, and Pdr11p, unlike the minor acyltransferase, Are1p, colocalize to sterol and sphingolipid-enriched, detergent-resistant microdomains (DRMs). Deletion of either ABC transporter results in Are2p relocalization to detergent-soluble membrane domains and a significant decrease (53–36%) in esterification of exogenous sterol. Similarly, in murine tissues, the SOAT1/Acat1 enzyme and activity localize to DRMs. This subcellular localization is diminished upon deletion of murine ABC transporters, such as Abcg1, which itself is DRM associated. We propose that the close proximity of sterol esterification and transport proteins to each other combined with their residence in lipid-enriched membrane microdomains facilitates rapid, high-capacity sterol transport and esterification, obviating any requirement for soluble intermediary proteins.—Gulati, S., Balderes, D., Kim, C., Guo, Z. A., Wilcox, L., Area-Gomez, E., Snider, J., Wolinski, H., Stagljar, I., Granato, J. T., Ruggles, K. V., DeGiorgis, J. A., Kohlwein, S. D., Schon, E. A., Sturley, S. L. ATP-binding cassette transporters and sterol O-acyltransferases interact at membrane microdomains to modulate sterol uptake and esterification.

Sterols are essential components of all eukaryotic membranes where they regulate the fluidity of the bilayer. However, the accumulation of intracellular sterols is cytotoxic and, if not remedied, can lead to cell death. The esterification of sterols with fatty acids is a first-line defense employed by all cells against the toxic effect of increasing sterol levels. Sterol esterification and the subsequent packing of these neutral lipids into cytoplasmic lipid droplets (CLDs) provide a subcellular reservoir that detoxifies and sequesters sterols until they are required for membrane assembly. Sterol esterification is catalyzed by the sterol O-acyltransferases (SOATs), a family of enzymes that is conserved throughout eukaryotic evolution (1). In mammalian cells, the acyl-coenzyme A:cholesterol acyltransferase (ACAT) gene family is tripartite, comprising ACAT1, ACAT2, and diacylglycerol acyltransferase 1. The latter enzyme esterifies nonsterol substrates such as diacylglycerols and retinols. In Saccharomyces cerevisiae, 2 ACAT-related enzymes (Ares), Are1p and Are2p, independently mediate the sterol esterification reaction with varying substrate preferences (2). The deletion of ARE2, which encodes the predominant isoform, results in a 75% decrease in steryl ester formation, whereas a deletion of both members of this gene family results in a viable yeast cell lacking steryl ester. The yeast are1 are2 double mutant is complemented by the expression of cDNAs to the human ACAT1 or ACAT2 orthologs, indicating both structural and functional conservation of these pathways from yeast to metazoans (3, 4).

Sterol esterification is allosterically regulated by the sterol substrates (5). This provides a rapid and effective mode of lipid detoxification and membrane homeostasis that minimizes the requirement for new protein synthesis. Consequently, a key regulatory aspect of the esterification reaction is the provision of the sterol substrate to the acyltransferases. Approximately 50% of the cholesterol substrate for ACAT1 originates from a plasma membrane–associated, endogenous cellular pool. The esterification of this sterol pool is unaffected by energy depletion, inhibition of membrane vesicle trafficking, or defects in the endosomal–lysosomal pathway (6). These observations suggest that transport of the sterol substrate from the plasma membrane to the endoplasmic reticulum (ER)-localized ACAT reaction is facilitated by a nonconventional, nonvesicular mechanism. Numerous lipid-binding proteins have been postulated to act as intracellular sterol transporters (7); however, to date, a role for these proteins in modulating sterol esterification has not been established (8, 9). The predicted size of the monomeric ACAT1 protein is ∼64.8 kDa; however, radiation inactivation experiments indicate that the functional size of native ACAT in rat liver microsomes ranges from 170 to 224 kDa (10). Chemical cross-linking studies indicated that ACAT1 purified from rat adrenal gland microsomes forms oligomeric complexes in vitro. Human ACAT1 clearly interacts with itself, likely as a tetramer (11); however, it has not been determined whether the esterification complex includes accessory proteins such as a sterol transporter.

In order to determine if protein–protein interactions mediate sterol transport to sites of esterification from the plasma membrane, we performed an integrated membrane yeast 2-hybrid (iMYTH) screen utilizing the major yeast ACAT ortholog, Are2p, as an interaction bait. We identified a yeast ATP-binding cassette (ABC) transporter, Pdr11p, as an Are2p-interacting protein, which like its paralog, Aus1p, mediates sterol import during anaerobiosis. Furthermore, we found that the ABC transporters and the SOAT enzyme colocalize to detergent-resistant microdomains (DRMs). Deletion of either AUS1 or PDR11 redistributed the Are2 acyltransferase from a detergent-resistant fraction to a detergent-soluble fraction, which was accompanied by a significant decrease in sterol esterification. We show that sequestration of ABC transporters and SOATs to DRMs also arises in murine tissues. These findings suggest that some plasma membrane ABC lipid transporters act in protein complexes that directly transfer sterols to the ACAT enzymes in order to regulate the esterification pathway based on substrate supply.

MATERIALS AND METHODS

General

Yeast strains used in this study are isogenic with strain W303-1A (12) or derived from the Applied Biosystems (Foster City, CA, USA) deletion collection. The upc2-1 high allele was monitored by PCR analysis of the UPC2 and HAP1 loci (13). Yeast transformations and deletion mutant strains were generated and confirmed by established methods (12). Human hemagglutinin epitope tag (HA) and C-terminal ubiquitin (Cub) fusions of the coding sequence of ARE1 and ARE2 were constructed as previously described (14, 15). The PDR11 and AUS1 open reading frames were tagged with sequences encoding yellow fluorescent protein (YFP) by PCR-based allele replacement (16).

Interaction screens

There were 2 versions of yeast 2-hybrid screens used: the conventional interaction trap (17), and the split-ubiquitin 2 hybrid (iMYTH) (15, 18). In brief, the AUS1 bait strain and an unrelated artificial control bait strain were transformed with N-terminal ubiquitin (Nub)G/NubI control prey plasmids or with prey plasmid expressing a NubG-tagged PDR11 fragment (residues 326–518) previously identified in a large-scale ABC transporter iMYTH screen (19). Cells were spotted onto transformation selection medium (synthetic medium lacking leucine) or interaction selection medium (synthetic medium lacking tryptophan-leucine-adenine-histidine). For the iMYTH genome-wide screen, the C terminus of Are2p was fused in frame with the Cub construct via homologous recombination in the yeast reporter strain THY AP4. Quantitative β-galactosidase activity assays of cell lysates were performed as described (17). Common motifs in the interacting proteins were predicted using T-Coffee (http://www.ebi.ac.uk/Tools/msa/tcoffee/) or prosite (http://prosite.expasy.org/prosite.html).

Analysis of exogenous sterol accumulation and esterification

Upc2-1 yeast strains were grown aerobically for ∼20 h in the indicated medium containing 1% tyloxapol:ethanol (1:1) and 0.01 μCi/ml [4-¹⁴C]cholesterol (49.8 mCi/mM) as described by us and others (13, 20, 21). As indicated, certain strains were grown anaerobically in 0.5% Tween 80, cholesterol (20 μg/ml), and 0.01 μCi/ml [4-¹⁴C]cholesterol. Approximately 10% (representing ∼0.5 nM cholesterol) of this initial dose becomes cell associated, of which the indicated fraction (Fig. 1) becomes esterified. Anaerobic growth was achieved using CO₂-generating gas packs (BD PharMingen, San Diego, CA, USA) and BBL gas jars (Becton Dickinson, San Diego, CA, USA) as described previously (13). Lipids were extracted, analyzed by thin-layer chromatography, and quantified via scintillation counting (13).

Figure 1.

Exogenous cholesterol esterification is independent of the endosomal and vacuolar pathways. A) Exogenous cholesterol redistribution. Cells of the indicated genotype were grown anaerobically in the presence of 4 µg/ml NBD cholesterol for 3 d or stained with FM4-64 (100 µg/ml at 30°C for 90 min), and assessed by fluorescence or light (DIC, differential interfering contrast) microscopy. NBD cholesterol-derived fluorescence is restricted to plasma membranes and subcellular lipid droplets (arrow) in control cells and to the plasma membrane of acyltransferase-deficient (are1Δ are2Δ) cells. B) Esterification of exogenous cholesterol. Cells of the indicated genotypes were grown anaerobically in the presence of 0.01 μCi/ml [4-¹⁴C]cholesterol. Esterification is presented as the mean esterified [4-¹⁴C]cholesterol (percentage of control ± se). In control cells, 78% of exogenous sterol was converted to steryl ester. Statistically significant difference from control is indicated. **P < 0.01 by unpaired Student’s t test.

Fluorescence microscopy

Cells were grown anaerobically, stained with 20 μg/ml 25-(N-[(7-nitro-2-1,3-benzoxadiazol-4-yl)methyl]amino)-27-norcholesterol (NBD cholesterol; 1:1 mixture with Tween 80) or 32 μM FM4-64, and examined by fluorescence microscopy on a Zeiss Axiovert 200M using a ×63 oil-immersion objective (Carl Zeiss GmbH, Jena, Germany). Aus1-YFP and Pdr11-YFP strains were visualized using green fluorescent protein (GFP) filters. Images were taken using a Hamamatsu Orca-ER camera (Boston, MA, USA).

Subcellular fractionation

Plasma membranes were purified from murine brain tissue as described previously (22). DRMs isolated from ∼300 mg mouse liver were homogenized in 1 ml lysis buffer [1% Triton X-100, 25 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (pH 6.5), 150 mM NaCl, 1 mm EDTA, 1 mM PMSF, and protease cocktail (Roche Molecular Biochemicals, Indianapolis, IN, USA)] and held at 4°C for 30 min. The extract was mixed with 1 ml of 2.5 M sucrose and overlaid with 6 ml of 30% sucrose, 25 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 150 mM NaCl, and 4 ml of 5% sucrose and centrifuged for 18 h at 4°C at 39,000 rpm. The gradient was separated into 12 fractions, collected from the top. An aliquot of each fraction was resuspended in sodium dodecyl sulfate (SDS) sample buffer containing 2-ME, and resolved by SDS-PAGE (gradient, 4–15%) and detected after transfer to nylon membrane using antisera against Flotilin-1 (Abcam Inc., Cambridge, MA, USA), ABCG1 (Novus Biologicals, Littleton, CO, USA), and ACAT1 [DM102 provided by T.-Y. Chang (Dartmouth School of Medicine, Dartmouth, NH, USA) (23) and ab128014 (Abcam Inc.)]. ACAT activity was assessed in DRMs isolated from the plasma membrane fraction as previously described (24). The purity of the plasma membrane preparation was assessed by the immunologically detected presence of phosphatidylethanolamine N-methyltransferase (antisera from Jean Vance, University of Alberta, Edmonton, AB, Canada), Na⁺/K⁺-ATPase (ab7671; Abcam), and the SRC tyrosine kinase (ab32102; Abcam). Sterol esterification activity was then assessed on fractions that were positive for both plasma membrane markers, Na⁺/K⁺-ATPase and the -SRC tyrosine kinase, but negative for the ER marker, phosphatidylethanolamine N-methyltransferase. DRMs from yeast were isolated from 10–20 optical density at 600 nm (OD₆₀₀) units of yeast grown at 30°C to log phase and lysed in 750 μl of 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 5 mM EDTA, 1 mM PMSF, and 2.5 μg/ml chymostatin, leupeptin, antipain, and pepstatin by glass bead disruption at 4°C. Cleared lysates (400 µl) were preincubated with Triton X-100 (1% final) for 30 min on ice and mixed with 2 volumes of 60% OptiPrep (Sigma-Aldrich, St. Louis, MO, USA) and overlaid with 2.7 ml of 30% OptiPrep in 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 5 mM EDTA, 1 mM PMSF, and 2.5 μg/ml chymostatin, leupeptin, antipain, and pepstatin. Samples were centrifuged at 45,000 rpm for 2 h. There were 9 fractions of equal volume collected from the top. An aliquot of each fraction was resuspended in SDS sample buffer containing 2-ME, and resolved on a gradient (4–15%) SDS-PAGE and immunoblotted with anti-HA (to detect Are2p and Are1p), anti-YFP (to detect Aus1p and Pdr11p), anti-Pma1p, and anti-Sec22p to detect marker proteins for raft and nonraft fraction. Immunoblots were quantified using ImageJ (National Institutes of Health, Bethesda, MD, USA).

Cross-linking and immunoprecipitation

Yeast microsomal protein preparations (14) were cross-linked with 2 mM dithiobis(succinimidyl propionate) (DSP) for 0.5 h at room temperature (25). YFP-tagged proteins from native or cross-linked preparations were immunoprecipitated with anti-GFP–conjugated agarose beads at 4°C, washed, and incubated at 59°C for 15 min in SDS sample buffer containing 50 mM DTT. Immunoprecipitation of the HA epitope-tagged enzyme from microsomes was performed in the presence of RIPA buffer (9.1 mM Na₂HPO₄, 1.7 mM NaH₂PO₄, 150 mM NaCl, 1% Triton X-100, 0.5% deoxycholate, and 0.1% SDS). Immunoprecipitates were collected on protein G agarose beads, washed in RIPA buffer, resuspended in SDS sample buffer containing 2-ME, and resolved by 6.5% SDS-PAGE and immunoblotting with chicken anti-Are2p (14) or anti-HA antisera.

RESULTS

Sterol esterification in yeast acts independently of the endosomal–vacuolar system

The transport of sterols from the periphery of the cell to the ER is an integral component of cellular sterol homeostasis. As sterols accumulate, they are rapidly esterified, which significantly alters both their chemical and biologic properties (polar to nonpolar, membrane associating to nonassociating). This results in the termination of transport and cytosolic storage of excess sterol in an innocuous lipid droplet before it reaches toxic levels. The transport and esterification of endogenously synthesized sterol are unaffected by the inhibition of vesicular trafficking and independent of the coat-forming protein complexes COPI and COPII in yeast (26) and mammalian cells (27). To assess the subcellular transit of exogenously derived cholesterol to the ER, we used fluorescence microscopy to monitor the movement of NBD cholesterol under anaerobic growth conditions to facilitate sterol uptake and esterification (13). Fluorescence was evident at the plasma membrane and in lipid droplets in control cells, indicating that NBD cholesterol rapidly traverses cellular membranes and becomes a substrate for the SOATs (Fig. 1A). NBD cholesterol was conspicuously absent from any other organelles including the vacuolar–endosomal compartment (coidentified by FM4-64 staining), even in an are1Δ are2Δ mutant, where free sterol could potentially accumulate due to the absence of the esterification reaction (Fig. 1A). To further assess the role of the endocytic pathway in transporting sterol to the esterification reaction, we measured the fraction of exogenously derived [¹⁴C]cholesterol that was esterified in yeast mutants with defects in the endosomal–vacuolar system. We found that abrogation of the endocytic pathway by deletion of END3 (a member of the evolutionarily conserved Eps15 homology domain gene family), or 2 other key components of endocytosis (END1 and END5), had no detectable quantitative impact on esterification of exogenous sterol (Fig. 1B). Similarly, we assessed the VPS (vacuolar protein sorting) pathway that mediates protein transport to the vacuole and thus the integrity of this organelle. In yeast, >50 vps mutants disrupt protein transport to the vacuole, often resulting in aberrant or undetectable vacuoles [as imaged with the fluorescent probe FM4-64 (28); Supplemental Fig. 1]. In order to ascertain the role of this pathway in transporting sterols to the esterification reaction, we assessed the fate of exogenous [¹⁴C]cholesterol during anaerobic growth of 6 vps deletion mutants (vps1, 5, 8, 18, 21, and 28) representing defects at 6 different stages of vacuolar function. We found that disrupting the activity of the vacuole in this manner did not alter exogenous sterol esterification or uptake (Fig. 1B and Supplemental Fig. 2). These data suggest that sterol delivery to the site of esterification is independent of the retrograde vesicular transport system and the endocytic pathway, prompting us to pursue alternate mechanisms such as a direct role of protein-protein interactions in sterol transport.

Multicomponent protein complexes participate in sterol transport to the esterification reaction

Biochemical and cross-linking studies have suggested that mammalian ACAT1 functions as a component of an undefined protein complex (29). We first sought to assess whether yeast Are2p interacts either with itself or with the closely related minor SOAT isoform encoded by the ARE1 gene. Are1p and Are2p were tagged with the 12CA5 hemagglutinin (HA) epitope and expressed from their endogenous promoters. These variants imparted equivalent levels of microsomal protein and enzymatic activity to the untagged enzymes (Fig. 2A and Supplemental Fig. 3). Microsomal preparations from are1Δ are2Δ strains transformed with vector control, or the HA epitope-tagged ARE1 and ARE2 expression plasmids, were immunoprecipitated with anti-HA antibody. Eluted proteins were resolved by SDS-PAGE and immunoblotted using a chicken anti-Are2p polyclonal antibody or an anti-HA antibody (Fig. 2B). Consistent with the differential substrate specificities of the 2 yeast enzymes (3, 30), we found that the Are2 protein self-associates but could not be coimmunoprecipitated with HA Are1p (Fig. 2B). We further confirmed these observations using a yeast 2-hybrid interaction assay (17), whereby we detected homomeric Are1p/Are1p and Are2p/Are2p interactions but saw no evidence for a heteromeric interaction between Are1p and Are2p (Supplemental Table 1).

Figure 2.

Multimerization of sterol esterification enzymes. A) Expression of yeast sterol esterification enzymes. Microsomal proteins from SCY059 (are1Δ are2Δ) strains expressing the indicated genes were separated by 6.5% SDS-PAGE and blotted with αAre2p or αHA antisera. B) Physical interactions of yeast sterol esterification enzymes. Microsomal proteins were immunoprecipitated with anti-HA antibody, eluted, separated by 6.5% SDS-PAGE, and blotted with αAre2p or αHA antisera. Co-IP, coimmunoprecipitation.

Having established the multimeric nature of the yeast SOAT complex, we reasoned that substrate transport to the esterification reaction could be readily achieved by the transient association of a transport protein with the ARE-encoded enzymes. Accordingly, we performed an iMYTH screen designed for the in vivo detection of membrane-associated protein–protein interactions (15). We focused on Are2p-specific interactions due to its marked substrate preference for ergosterol and cholesterol (30, 31). To accomplish this screen, the C terminus of Are2p was fused in frame with the Cub via homologous recombination in a yeast reporter strain. The protein fusion was validated in terms of esterification of exogenous-radiolabeled cholesterol (Supplemental Fig. 4). The orientation of the Are2-Cub bait construct was assessed based on its ability to interact with an ER integral membrane protein Alg5p, fused to either NubI (a positive control) or NubG (a negative control in which an I13G mutation in NubG results in a failure to associate with the Cub bait). Transformation of the Are2-Cub–expressing strain with the positive control Alg5-NubI resulted in the reconstitution of the ubiquitin molecule and subsequent activation of the reporter system, as indicated by growth of colonies on selective medium lacking histidine and adenine, or blue colonies in the presence of 5-bromo-4-chloro-3-indolyl-β-d-galacto-pyranoside (X-gal) (Supplemental Fig. 5). Conversely, transformation with the negative control, Alg5p-NubG, did not yield any colonies on synthetic defined medium lacking tryptophan, adenine, histidine. These results demonstrate that the Cub-transcription factor reporter moiety fused to the C terminus of Are2p is biologically active and will act as an authentic probe in an iMYTH screen. The Are2-Cub bait strain was thus transformed with a yeast genomic library in which genomic fragments were fused to sequences encoding the Nub (15). All positive clones were validated with a bait dependency test in order to ensure that they only reactivated the reporter system in the presence of the Are2p bait. We excluded clones that possessed noncoding open reading frames or autonomously replicating sequences as well as those that conferred interactions in unrelated split-ubiquitin screens (Supplemental Table 2) (32). A total of 11 putative Are2p-interacting proteins remained, many of which have been found to interact with key components of lipid homeostasis (Table 1). To experimentally assess the quantitative role of these proteins in sterol trafficking to the ER, we deleted each candidate gene in a upc2-1 strain and measured esterification of exogenous [¹⁴C]cholesterol (Fig. 3). The upc2-1 strain contains a gain-of-function mutation in UPC2, a transcription factor that facilitates aerobic sterol influx (13). In addition, deletion mutants of MTF1, COS12, and SLU7 were assessed anaerobically for their impact on the esterification of exogenous [¹⁴C]cholesterol. Deletion of COS8, SIP18, MFBI, YJR096W, and PDR11 resulted in a statistically significant decrease in the fraction of exogenous sterol that becomes esterified (Fig. 3). The most striking impact on exogenous sterol transport to the esterification reaction resulted from deletion of the ABC transporter, Pdr11p. Pdr11p and its paralog, Aus1p, mediate the influx of sterols during anaerobiosis (13), and as expected, we found that deletion of AUS1 also significantly decreased the fraction of exogenous sterol that was esterified (Fig. 3). We conclude that the iMYTH interaction screen identified several interrelated components of sterol homeostasis, with the ABC transporters, Pdr11p and Aus1p, having the most profound effect on sterol esterification.

TABLE 1.

Putative Are2p-interacting proteins

Are2p-interacting protein	Protein family or complex	Localization	Function	Human ortholog (% identity)	Known interactions with hits	Other relevant interactions	Interaction domains/motifs
Cos8p	DUP380	ER, NE	Sphingolipid metabolism			FEN1, SUR4, IRE1 (genetic interaction)	3 TMD
							Leu zipper
Cos12p	DUP380	ER	Unknown		Nup84 (physical)		2 TMD
					Pdr11p (genetic correlation)		Leu zipper
Glt1p	NA	Mito	Glutamate synthase activity	DPYD (22)	Mfb1 (negative genetic)	ARE1 (genetic correlation 0.147)	GltS, FMN,
Mfb1p	F-box	MOM	Required for mitochondrial tubulation		Glt1 (negative genetic)	ERG5 (negative genetic 5.4e-07)	F-box
							Leu zipper
Nup84p	NA	NE	Nuclear pore organization and biogenesis	NUP107 (20)	Cos12p (physical)
Pdr11p	ABC transporter (G subfamily, paralogous to yeast Aus1p)	Plasma membrane, microsomes	Sterol uptake	ABCG1 (27)	Cos12p (genetic correlation)	EHT1 (genetic correlation 0.128), Aus1p (this study)	12 TMD,
							ABC2-membrane,
							Leu zipper
Sip18p	NA	Cyto	Phospholipid binding				Leu zipper
YJR096Wp	Aldo-keto reductase (AKR)	Cyto, nucleus	General pentose sugar reductase	AKR1B1 (37)		Aus1p (protein–peptide interaction), ERG6 (negative genetic 2.7 e-02), DAN1 (genetic correlation 0.152)	Leu zipper
Slu7p (essential)		Cyto, nucleosome	Required for the second catalytic step of splicing	SLU7 (24)		ERG2 (negative genetic 3.7 e-16), ERG3 (positive genetic 1.2 e-03), ERG6 (positive genetic 3.1 e-02), INO4 (positive genetic 2.7e-16)	Prp18-interacting factor, Leu zipper
Mtf1p	NA	Mito	DNA-binding, mitochondrial genome maintenance			Erg6p (physical interaction)	Leu zipper
Atp8p	NA	Mito	Aerobic respiration			NA	1 TMD
							Leu zipper

The coding sequence of ARE2 was incorporated into an iMYTH screen of an S. cerevisiae genomic library (15). Reporter gene activation resulted in the identification of 11 fusion proteins that physically interact with Are2p. Complex, functional annotations, localization, and sequence orthologs were ascribed based on the BKLproteome (BIOBASE Biological Databases; https://portal.biobase-international.com). Interactions with other proteins were derived from databases at DRYGIN [http://drygin.ccbr.utoronto.ca (33)] and BKLproteome. Interaction domain/motif assignments were ascribed by SGD (34) or T-Coffee (http://www.ebi.ac.uk/Tools/msa/tcoffee/). C, cytoplasm; DUP380, duplicated 380; M, mitochondria; MOM, mitochondrial outer membrane; N, nucleus; NA, not applicable; NE, nuclear envelope; NS, nucleosome; TMD, transmembrane domain.

Figure 3.

Sterol homeostasis in mutants lacking putative Are2p-interacting partner proteins. The indicated genes were deleted in upc2-1 strains or grown anaerobically as mutations in control backgrounds to facilitate uptake of 0.01 μCi/ml [4-¹⁴C]cholesterol. The data (2 independent sets of triplicates) reflect the mean esterified [4-¹⁴C]cholesterol (percentage of control ± se). In control cells, 80% of exogenous sterol was converted to steryl ester. Statistically significant differences from control are indicated. *P < 0.05 and **P < 0.01 by unpaired Student’s t test.

Aus1p, Pdr11p, and Are2p physically interact

To validate the physical interactions between Are2p and the ABC transporters, Aus1p and Pdr11p, we performed coimmunoprecipitation and cross-linking studies. The AUS1 and PDR11 genes were fused with the YFP-coding sequence by homologous recombination and were shown to mediate anaerobic growth and sterol uptake to levels equivalent to the native proteins (data not shown). Microsomal fractions from these integration strains transformed with an ARE2-HA plasmid or vector control were isolated and observed to contain both the ABC transporters and the esterification enzyme (Supplemental Fig. 6). These microsomal preparations were immunoprecipitated with anti-GFP–conjugated agarose beads. Western blot analysis of the eluted proteins demonstrated that Are2-HAp was detected in all samples except those prepared from vector control strains (Fig. 4A, B). The physical interaction of the individual ABC transporters with Are2p was also detected in the presence of a cleavable covalent cross-linker: DSP (Fig. 4C, D). Interestingly, the 2 ABC transporters likely form a complex with each other in that an iMYTH assay in which we coexpressed AUS1-Cub and PDR11-Nub in a yeast reporter strain activated the appropriate interaction reporters (Fig. 5A). These studies validate the findings of the iMYTH interaction screen, supporting the hypothesis that sterols may be channeled from the ABC transporters to SOATs via protein–protein interactions.

Figure 4.

Are2p, Aus1p, and Pdr11p form a complex in vivo. Aus1-YFP (A and C) or Pdr11-YFP (B and D) strains were transformed with a vector control or pRS424/Are2-HA. Microsomes were solubilized and immunoprecipitated with GFP-conjugated agarose beads followed by denaturing gel electrophoresis and immunoblotting with an α-HA antisera. ns, a nonspecific cross-reacting species. The extracts prior to immunoprecipitation are shown in the input lanes. A and B) ABC transporter and acyltransferases have a shared microenvironment. C and D) ABC transporter and acyltransferases form a complex separated by <15 Å. Cell extracts were incubated with the membrane-permeable cleavable cross-linker DSP, prior to immunoprecipitation, SDS-PAGE resolution, and immunoblotting.

Figure 5.

Membrane properties of yeast ABC-sterol transporters. A) Aus1p and Pdr11p form a complex. The membrane yeast 2-hybrid (iMYTH) testing was carried out as described in Materials and Methods. The AUS1 bait strain and an unrelated control bait strain were transformed with NubG/NubI control prey plasmids or with prey plasmid expressing a NubG-tagged PDR11 fragment (corresponding to aa 326–518). Cells were spotted onto transformation selection medium (T, synthetic medium lacking leucine) or interaction selection medium (I, synthetic medium lacking tryptophan, leucine, adenine, and histidine). B) Fluorescent localization of Aus1-YFP and Pdr11-YFP in upc2-1 strains. Fluorescence (YFP) and differential interfering contrast (DIC) images are shown. Punctate fluorescence at the plasma membrane/cell periphery is indicated (arrows). C) Aus1-YFP and Pdr11-YFP localize to cold DRMs. Cold detergent-resistant proteins from strains expressing the indicated proteins were prepared as in Materials and Methods, resolved by 4–15% gradient SDS-PAGEs and immunoblotted with the indicated antisera (αPma1; DRM marker, αGFP for the indicated ABC transporter-YFP fusions).

Aus1p, Pdr11p, and Are2p localize to plasma membrane cold DRMs

In order to establish a cellular context for the in vivo interaction of these proteins, we assessed the subcellular localization of the ABC transporter–YFP fusion proteins. We confirmed that Aus1p and Pdr11p consistently localized to the plasma membrane by membrane fractionation (Supplemental Fig. 7) but in a punctate pattern based on fluorescence microscopy of the YFP fusion proteins (21, 26) (Fig. 5B). Several mammalian ABC transporters involved in the transport of lipids localize to plasma membranes, often in DRMs. DRMs are sterol- and sphingolipid-rich membrane foci that integrate many eukaryotic metabolic pathways, including lipid influx and efflux (1). Consequently, we investigated whether Aus1p and Pdr11p localized to DRMs. Total membrane proteins from the ABC transporter-YFP–expressing cells were solubilized with cold 1% Triton X-100 detergent and fractionated by OptiPrep density gradient centrifugation at 4°C (35). The isolated DRMs were probed with anti-GFP antibodies to look for expression of Aus1-YFP and Pdr11-YFP (Fig. 5C). In upc2-1 cells, the majority of Aus1p and Pdr11p colocalize with the established DRM marker Pma1p (Fig. 5C; 70 and 82%, respectively, as determined by ImageJ analysis of representative immunoblots).

Are2p, in contrast to the ARE1-encoded acyltransferase, localized to both the ER and plasma membrane fractions, coincident with markers of the ER (Sec61p) and plasma membrane (Gas1p) (Supplemental Figs. 9 and 10). We rationalized that a DRM would be an ideal context in which to esterify excess sterols. We therefore assessed the DRM localization of the yeast acyltransferases in the upc2-1 yeast strain. Although Are1p localized exclusively to a detergent-soluble compartment, the majority of Are2p was found in DRMs coincident with Pma1p and absent from Sec22p-containing, detergent-soluble fractions (Fig. 6A). Strikingly, deletion of either ABC transporter relocalized Are2-HAp from a detergent-resistant to a detergent-soluble fraction (Fig. 6B) with no detectable generalized impact on DRMs as judged by Pma1p distribution. These data suggest that Are2p’s physical interaction with the ABC transporters is necessary for its DRM localization. This shift in localization likely contributes to the decreased exogenous sterol esterification observed in an AUS1 or PDR11 deletion (Fig. 2). This dual dependency on either ABC transporter for Are2p localization to a DRM suggests that it exists in a protein complex with Aus1p and Pdr11p. This is consistent with the heteromultimerization of Pdr11p/Aus1 previously observed (Fig. 5A). We propose that the physical interaction of Aus1p, Pdr11p, and Are2p, in the context of a DRM fraction, provides a platform that efficiently integrates sterol influx with esterification.

Figure 6.

Yeast ABC transporters modulate membrane association of sterol esterification enzymes. In control cells (A), Are1p and Are2p localize to cold detergent-soluble (fractions 7 and 8) or -resistant (peak fractions 2 and 3) microdomains, respectively, coincident with the indicated marker proteins (αPma1; DRM, αSec22; non-DRM). In the absence of either ABC transporter (B), the Are2 acyltransferase becomes detergent soluble (peak fractions 7 and 8). Cold detergent-resistant proteins from indicated strains were prepared as in Materials and Methods, resolved by 4–15% gradient SDS-PAGEs, and immunoblotted with indicated antisera to Pma1p, Sec22, or HA (for Are1p and Are2p).

mABCG1 and mACAT1 colocalize with DRMs in murine lung and brain

Aus1p and Pdr11p are members of the ABC-G subfamily of ABC transporters (36) and share ∼26% sequence identity with human ABCG1. In order to investigate whether mammalian ABC transporters influence the localization and activity of ACAT1, we first assessed the colocalization of these proteins to DRMs. DRMs were isolated from murine lung homogenate or brain plasma membrane preparations by conventional protocols [1% Triton X-100 at 4°C (37)] and assessed with antisera against ABCG1, ACAT1, and flotillin-1 (a marker of mammalian DRMs) (Fig. 7A). We found that both ABCG1 (identified as 2 isoforms by comparison with protein extracts from Abcg1^−/− mice, Supplemental Fig. S8) (38) and an ∼50 kDa ACAT1-specific polypeptide (23) colocalize with flotillin-1. Furthermore, the majority of ACAT activity (Fig. 7B) and protein (Supplemental Fig. S8) fractionated with DRMs in plasma membranes from murine brain. Similar to our prior studies with yeast, the localization of ACAT1 to DRMs partially requires the presence of an ABC-G transporter (Fig. 8). DRMs from ABCG1^−/− murine brain were assessed for the localization of ACAT1 and flotillin-1. We found that in the absence of ABCG1, the subcellular localization of ACAT1 distributed to both detergent-resistant and detergent-soluble fractions, although flotillin-1 remained detergent resistant. ABCG4, a member of the ABC-G subfamily, is functionally redundant with ABCG1 (sharing 69% identity and 82% similarity at the amino acid level) and can physically interact to form heterodimers. To determine whether ABCG4, like ABCG1, plays a role in localizing ACAT1 to a DRM, we isolated DRMs from ABCG4^−/− and ABCG1^−/−ABCG4^−/− tissues. Similar to the DRMs from ABCG1^−/− tissue, we found that ACAT1 from both genotypes (ABCG4^−/− and the double knockout) distributed to both detergent-resistant and -soluble fractions (Fig. 8). Collectively, these data demonstrate that colocalization of members of the ABC-G subfamily and the ACAT family to DRMs is conserved across tissues and organisms. Furthermore, similar to the yeast ABC transporters, mammalian ABCG1 and ABCG4 participate in sequestration of ACAT1 to DRMs. This suggests an evolutionarily conserved mechanism likely acting within DRMs, by which ABC transporters, such as ABCG1 in mammals and Aus1p and Pdr11p in yeast, transport sterol substrates directly to the O-acyltransferases for esterification and thus membrane detoxification.

Figure 7.

Mammalian ABCG1 and ACAT1 localize to DRMs in murine tissues. DRMs were isolated and analyzed as described in Materials and Methods; tissue homogenates were solubilized with cold 1% Triton X-100 and fractionated by sucrose gradient centrifugation. Isolated fractions were run on 4–15% gradient SDS-PAGEs and immunoblotted with the indicated antisera. A) Acat1 and ABCG1 colocalize with flotillin-1 (a DRM marker) in control murine lung. B) ACAT activity is predominantly associated with plasma membrane (PM) DRMs in fractions isolated from murine brain. Plasma membrane DRMs were isolated and analyzed as described in Materials and Methods. Statistically significant differences between detergent-soluble and -insoluble assays are indicated. **P < 0.001.

Figure 8.

Detergent-resistant properties of ACAT1 are altered in the absence of ABCG1 and ABCG4 in murine brain tissues. A) DRMs were isolated from brain homogenates from animals of the indicated genotypes with cold 1% Triton X-100, fractionated by sucrose gradient centrifugation and resolved by SDS-PAGEs, and probed with the indicated antisera. B) ImageJ analysis of immunoblots. Flotillin-1 (a DRM marker) expression peaks in fractions 4–7 for all genotypes. In control tissues, Acat1 expression peaks in DRM fractions (4–7). However, Acat1 expression is present in detergent-resistant and detergent-soluble fractions in mG1^−/−, mG4^−/−, and mG1^−/− mG4^−/− tissues.

DISCUSSION

The esterification of free sterol is an evolutionarily conserved component of eukaryotic membrane homeostasis that limits sterol cytotoxicity. Prior to esterification, the sterol amphipath requires solubilization by proteins and/or membranes, both to function physiologically and to be efficiently delivered to the membrane-associated SOAT reaction. The resulting neutral lipids spontaneously form CLDs at several subcellular locations including the ER and plasma membrane (Supplemental Fig. 11) and become inert and nontoxic. Of the ACAT substrate pool, 50% is derived from the plasma membrane by a nonconventional pathway (39). To delineate the role of protein–protein interactions in subcellular sterol transport to the yeast SOAT, Are2p, we used iMYTH technology to identify 11 putative Are2p-interacting proteins (Table 1). There were 5 gene products found to positively impact sterol esterification in that deletion mutants displayed significant defects in exogenous sterol esterification (Fig. 3). Moreover, we also found that a subset of these gene products conferred resistance to antifungal agents targeting sterol or sphingolipid homeostasis (data not shown) (40), further reinforcing their roles in lipid homeostasis. Many of these interactions identify pathways that are conserved in humans (Table 1). For example, Pdr11p and its paralog, Aus1p, are plasma membrane ABC transporters with orthology to the G subfamily of mammalian ABC half-transporters. Pdr11p and Aus1p have overlapping roles in exogenous sterol uptake (13) and metabolically interact with the esterification reaction in that deletion mutants in either gene significantly reduced the fraction of exogenous sterol that became esterified. The interaction is reciprocal; loss of sterol esterification significantly reduces sterol uptake by these transporters (Supplemental Fig. 2) (26). In confirmation of the iMYTH screening, we found that Aus1p and Pdr11p cross-link and immunoprecipitate with Are2p. Taken together, these studies compellingly suggest that a pool of exogenous sterol is directly channeled via plasma membrane ABC transporters (such as Aus1p and Pdr11p), to the acyltransferase-mediated esterification reaction. This provides a mechanism for rapid and direct transport of excess sterols from the plasma membrane to the ER. This process is optimized by the coresidence of these proteins with their substrate in sterol-rich membrane microdomains (DRMs). Interestingly, deletion of either yeast ABC transporter results in Are2p relocalization from a DRM to a detergent-soluble fraction as well as a significant decrease in the fraction of exogenous cholesterol esterified. This is unlikely to represent a general defect in DRM biogenesis or stability, in that the distribution of the plasma membrane DRM marker, Pma1p, was unchanged in these mutants. This suggests that DRMs not only provide a platform to compartmentalize the sterol flux and esterification reactions but also play a role in regulating the rate of sterol esterification.

The studies described here suggest a model (Fig. 9) in which sterol transport and esterification happen in close proximity, to the extent that, in yeast, the participants of these processes are cross-linked by a cell-permeable reagent (DSP) that theoretically spans 12 Å. Conventionally, the SOATs are integral proteins of the ER membrane in yeast and mammals. Are2p may indeed reside in the ER but bridge to the plasma membrane at membrane contact sites (MCSs) via an interaction with the ABC transporter (Fig. 9A). This would then promote the rapid esterification of newly translocated exogenous sterol. Loss of either ABC transporter resulted in redistribution of the acyltransferase to detergent-soluble microdomains and reduced esterification of exogenous sterol. MCSs among the ER and the plasma membrane (plasma membrane–associated membranes), mitochondria (mitochondria-associated membranes), and lipid droplets are in some instances detergent resistant (37) and have been proposed to facilitate lipid transport in many eukaryotic cells (41). In at least 1 instance, the junctions are bridged by the interaction of an acyltransferase (diacylglycerol acyltransferase 2, on a lipid droplet) with an ER-associated acyl-coenzyme A synthetase (FATP1) (42). Alternatively, a subfraction of the O-acyltransferase pool may mobilize to the plasma membrane in response to exogenous cholesterol loading (Fig. 9B). In this scenario, the absence of either ABC transporter prompts the relocation of the acyltransferase to a detergent-soluble region of the plasma membrane. This also has precedent; treatment of macrophages with acetylated LDLs relocalizes a fraction (10–20%) of the ACATs to the cell surface (43), endocytic recycling compartment (44), and endosome (45).

Figure 9.

Models for coupling of exogenous sterol transport and esterification in S. cerevisiae. Exogenous sterol is imported into the cell via the DRM-residing ABC transporters, Aus1p and Pdr11p. Upon influx, sterols are rapidly and efficiently esterified by Are2p, which also resides in a DRM. The newly synthesized steryl ester (SE) then accumulates in CLDs. A) Are2p serves as a bridge between the ER and plasma membrane (PM) at ER–plasma MCSs. In the absence of either ABC transporter, Are2p relocalizes to a detergent-soluble domain, and the fraction of exogenous sterol that is esterified decreases. By contrast, the minor O-acyltransferase, Are1p, resides solely in the ER in detergent-soluble membranes where it predominantly esterifies sterol biosynthetic intermediates as they are synthesized. B) The transporters and acyltransferases exist as multimeric complexes at the plasma membrane in DRMs. Loss of Aus1p or Pdr11p results in the altered residence of Are2p in a plasma membrane detergent-soluble fraction and a concomitant decrease in the percentage of cellular steryl esters.

Overexpression of human ABCG1 in various cell types redistributes membrane cholesterol to cell surface domains and increases cholesterol esterification, suggesting a physiologic role of the interaction between these proteins in mammals (46). Consistent with this hypothesis, we observed colocalization of the mammalian esterification reaction (ACAT1 protein and activity) and Abcg1 to DRMs in murine lung and brain [also recently observed in cultured human HeLa cells (38)]. Additionally, we found that in the absence of ABCG1 and its homolog ABCG4, ACAT1 was redistributed to a detergent-soluble fraction. ABCG1 and ABCG4, like Aus1p and Pdr11p, likely play a role in anchoring the esterifying enzyme to DRMs. In yeast, the ABC transporters Aus1p and Pdr11p are the only ABC transporters (out of 29) (47) with demonstrated roles in sterol transport (unpublished results) (48). By contrast, in mammalian cells, at least 18 ABC transporters modulate sterol and lipid metabolism. Moreover, at least 5 members of the ABC-G subfamily (ABCG1, G2, G4, G5, and G8) transiently associate with several organelle membranes often in a DRM and impact sterol flux throughout the cell (49–54). This marked redundancy and heterogeneous distribution within the ABC-G subfamily may explain why unlike Are2p, ACAT1 does not homogenously redistribute to a detergent-soluble fraction.

Subcellular sterol distribution between different organelles is rapid and of high capacity. It has been estimated that as much as 1 million sterol molecules transit between the endocytic recycling compartment and the plasma membrane in 1 s (55). The esterification of excess plasma membrane cholesterol by the ACATs is an effective and critical response by the cell to combat toxicity of excess sterol. For example, a 10% increase in cell surface cholesterol results in a 3-fold increase in esterification; the half-time for this response is 10–20 min (38). We describe here a molecular interaction and cellular context that could fulfill these requirements of rapidity and capacity. Are2p, murine ACAT1, Aus1p, Pdr11p, and murine ABCG1 localize to DRMs. However, in what manner does this evolutionarily conserved compartmentalization advantage the cell? It is intriguing that the interaction we describe between ABC transporters and esterification enzymes arises during the challenge of sterol influx. In all eukaryotic cells, at least 2 sets of membranes, the ER and the plasma membrane, require buffering against excess free sterol, in the former case to prevent lipotoxicity and in the latter case to increase membrane capacity to meet cellular needs. Thus, rapid esterification of exogenous sterol as soon as it is absorbed and prior to its arrival at the ER may represent a key regulatory step in the metabolism of sterol in resisting lipotoxicity and in mammals prevent or delay a proatherogenic, obese, or neurodegenerative state.

Glossary

ABC

ATP-binding cassette

ACAT

acyl-coenzyme A:cholesterol acyltransferase

ARE

acyl-coenzyme A:cholesterol acyltransferase-related enzyme

CLD

cytoplasmic lipid droplet

Cub

C-terminal ubiquitin

DRM

detergent-resistant microdomain

DSP

dithiobis(succinimidyl propionate)

ER

endoplasmic reticulum

GFP

green fluorescent protein

HA

human hemagglutinin epitope tag

iMYTH

integrated membrane yeast 2 hybrid

MCS

membrane contact site

NBD cholesterol

25-(N-[(7-nitro-2-1,3-benzoxadiazol-4-yl)methyl]amino)-27-norcholesterol

Nub

N-terminal ubiquitin

SDS

sodium dodecyl sulfate

SOAT

sterol O-acyltransferase

VPS

vacuolar protein sorting

X-gal

5-bromo-4-chloro-3-indolyl-β-d-galacto-pyranoside

YFP

yellow fluorescent protein