The Human Phosphatidylinositol Phosphatase SAC1 Interacts with the Coatomer I Complex*

The Saccharomyces cerevisiae SAC1 gene encodes an integral membrane protein of the endoplasmic reticulum (ER) and the Golgi apparatus. Yeast SAC1 mutants display a wide array of phenotypes including inositol auxotrophy, cold sensitivity, secretory defects, disturbed ATP transport into the ER, or suppression of actin gene mutations. At present, it is not clear how these phenotypes relate to the finding that SAC1 displays polyphosphoinositide phosphatase activity. Moreover, it is still an open question whether SAC1 functions similarly in mammalian cells, since some phenotypes are yeast-specific. Potential protein interaction partners and, connected to that, possible regulatory circuits have not been described. Therefore, we have cloned human SAC1 (hSAC1), show that it behaves similar to ySac1p in terms of substrate specificity, demonstrate that the endogenous protein localizes to the ER and Golgi, and identify for the first time members of the coatomer I (COPI) complex as interaction partners of hSAC1. Mutation of a putative COPI interaction motif (KXKXX) at its C terminus abolishes interaction with COPI and causes accumulation of hSAC1 in the Golgi. In addition, we generated a catalytically inactive mutant, demonstrate that its lipid binding capacity is unaltered, and show that it accumulates in the Golgi, incapable of interacting with the COPI complex despite the presence of the KXKXX motif. These results open the possibility that the enzymatic function of hSAC1 provides a switch for accessibility of the COPI interaction motif.

The Saccharomyces cerevisiae SAC1 gene encodes an integral membrane protein of the endoplasmic reticulum (ER) and the Golgi apparatus. Yeast SAC1 mutants display a wide array of phenotypes including inositol auxotrophy, cold sensitivity, secretory defects, disturbed ATP transport into the ER, or suppression of actin gene mutations. At present, it is not clear how these phenotypes relate to the finding that SAC1 displays polyphosphoinositide phosphatase activity. Moreover, it is still an open question whether SAC1 functions similarly in mammalian cells, since some phenotypes are yeast-specific. Potential protein interaction partners and, connected to that, possible regulatory circuits have not been described. Therefore, we have cloned human SAC1 (hSAC1), show that it behaves similar to ySac1p in terms of substrate specificity, demonstrate that the endogenous protein localizes to the ER and Golgi, and identify for the first time members of the coatomer I (COPI) complex as interaction partners of hSAC1. Mutation of a putative COPI interaction motif (KXKXX) at its C terminus abolishes interaction with COPI and causes accumulation of hSAC1 in the Golgi. In addition, we generated a catalytically inactive mutant, demonstrate that its lipid binding capacity is unaltered, and show that it accumulates in the Golgi, incapable of interacting with the COPI complex despite the presence of the KXKXX motif. These results open the possibility that the enzymatic function of hSAC1 provides a switch for accessibility of the COPI interaction motif.
Phosphatidylinositol (PtdIns) 1 phosphates act as signaling components in various intracellular membranes and influence membrane trafficking, cytoskeletal organization, motility, and cellular survival, depending on their subcellular localization and the availability of specialized PtdIns phosphate-binding proteins, lipases, PtdIns kinases, and PtdIns phosphatases (for reviews, see Refs. 1 and 2). Several PtdIns phosphatases have recently gained much attention, because their loss of function is associated with disease. For example, mutations in the OCRL1 gene, which encodes a phosphatidylinositol-(4,5)bisphosphate 5-phosphatase of the trans-Golgi network, are responsible for the oculocerebrorenal syndrome of Lowe (3). The tumor suppressor PTEN (phosphatase and tensin homologue deleted on chromosome ten) (for a review, see Ref. 4) functions as a plasma membrane-associated PtdIns-3-phosphatase that negatively regulates the PtdIns-3-kinase/AKT pathway (5). Furthermore, myotubularin 1 (MTM1) belongs to a family of PtdIns-3-phosphatases and was originally identified by positional cloning of the gene responsible for X-linked myotubular myopathy (6,7). Interestingly, PTEN and the MTM1 family harbor a core CX 5 R(T/S) catalytic motif, which was first identified in protein phosphatases (8,9).
None of the human PtdIns phosphatases containing a SAC domain have been characterized in depth for their biological function. Information about human SAC1 (hSAC1) is hitherto restricted to an EST (KIAA0851), which maps to the C3CER1 segment on chromosome 3p21.3. C3CER1 is commonly eliminated in SCID-derived tumors (17,18). Human SAC2 was enzymatically characterized and found to exert 5-phosphatase activity specific for PtdIns(4,5)P 2 and PtdIns(3,4,5)P 3 (19).
Yeast strains with mutations in SAC1 exhibit an array of phenotypes including inositol auxotrophy (20), cold sensitivity (21), secretory defects in chitin deposition (13), multiple drug sensitivity (22), and ATP transport deficiencies into the ER (23). Mutations in yeast SAC1 are capable of suppressing le-thality of some secretory pathway (SEC) mutants, like SEC14 deficiency (24). SEC14, a phosphatidylinositol/phosphatidylcholine transport protein, is responsible for establishing a critical phospholipid composition in yeast Golgi membranes and is required for secretory competence (25). This is in line with the finding that ySac1p negatively regulates a pool of PtdIns(4)P in the yeast Golgi that is important for forward trafficking of chitin synthases to the cell periphery (13).
In eukaryotic cells, three vesicle-based protein cargo trafficking systems have evolved, which are each defined by a special set of membrane-covering protein coats. Clathrin-coated vesicles allow transport from the plasma membrane to the trans-Golgi network and from the trans-Golgi network to endosomes. Coatomer II (COPII)-coated vesicles have been assigned to ER to Golgi anterograde transport (26,27). In contrast, coatomer (COPI)-coated vesicles are involved in membrane traffic mainly between Golgi and ER or intra-Golgi compartments (28,29). The COPI coatomer consists of seven proteins (␣, ␤, ␤Ј, ␥, ␦, ⑀, and ). Formation of the COPI coat depends on the GTPase ARF1 (ADP-ribosylation factor 1), which in its GTP-bound form inserts into the target membrane and attracts coatomer components prior to vesicle budding. GTP hydrolysis by ARF correlates with uncoating of COPI proteins followed by vesicle fusion with the target membrane (26,30).
ySac1p is described as an integral protein of ER and Golgi membranes with the SAC domain pointing toward the cytoplasm. There is still controversy about the number of functional transmembrane (TM) domains and thus the topology of SAC1, since ySac1p was demonstrated to utilize its two Cterminal TM domains (31), whereas rSAC1 apparently only uses the first one (11). In addition to this controversy, the subcellular localization of endogenous mammalian SAC1 is still open, and it is not clear whether it is involved in the secretory pathway to the same extent as in yeast, which of its functions might depend on its enzymatic activity, or which binding proteins might aid in its biological functions.
To address these questions, we characterized human SAC1 and show that it localizes to ER and Golgi. We generated a PtdIns phosphatase inactive variant by mutation of the core CX 5 R(T/S) motif (hSAC1-C/S) and find it accumulated in the Golgi. We also demonstrate for the first time that wild type hSAC1 interacts with members of the coatomer I complex and with itself. Mutation of a putative COPI-interaction motif, K(X)KXX, in hSAC1 causes a loss in COPI binding and accumulation in the Golgi. Interestingly, although the hSAC1-C/S mutant has an unaltered COPI interaction motif, it fails to efficiently bind to coatomer I. Our findings suggest that the localization of hSAC1 might not only depend on the mere presence of an intact COPI interaction motif and that the PtdIns phosphatase function of hSAC1 is needed to achieve appropriate usage of the K(X)KXX motif.

EXPERIMENTAL PROCEDURES
Molecular Cloning of hSAC1 and Plasmids-The EST KIAA0851 was found to contain an open reading frame that is highly homologous to the yeast enzyme SAC1 (31.8% identity, 46.1% similarity). Human SAC1 cDNA was amplified from brain mRNA (Invitrogen) by reverse transcriptase-PCR using the following primers: primer a, 5Ј-GAG AGA GAA GGA AGG AGG TGG T-3Ј; primer b, 5Ј-TGT GGA AAA GTA TGC CTG CTA ATA GTG-3Ј. PCR products were subcloned into pCRII-TOPO vector (Invitrogen) generating pCRII-hSAC1 and confirmed to be identical to KIAA0851 (nucleotide coordinates 40 -2272) via sequencing. The hSAC1 ORF is covered by KIAA0851 nucleotides 70 -1833.
The hSAC1-C/S mutation was introduced into pCRII-hSAC1 with a Transfomer TM site-directed mutagenesis kit (Clontech) using mutagenic primer 5Ј-GTT CCG AAG CAA TAG CAT GGA TTG TCT AG-3Ј and selection primer 5Ј-GTG ACT GGT GAG GCC TCA ACC AAG TC-3Ј according to the manufacturer's instructions. The hSAC1-KEKID mutation was generated by PCR using the following primers: primer c, 5Ј-GCA CAA TCC ATC TGG TGG CA-3Ј; mutagenic primer (primer d), 5Ј-A TCC TCA GTC TAT CGC TTC TGC CTG GA-3Ј with the changed nucleotides underlined. The respective fragment was cloned into pCRII-TOPO, and an EcoRI fragment thereof was swapped into pGFP-hSAC1wt (see below). All mutations were confirmed by sequencing. An N-terminal FLAG tag (amino acids MDYDDDDKATAA, with hSAC1 amino acids 2-5 underlined) was fused to hSAC1wt and hSAC1-C/S coding sequences using standard PCR techniques. Insertion into the mammalian expression vector pcDNA3.1 (Invitrogen) generated pcDNA3.1-FLAG-hSAC1wt and pcDNA3.1-FLAG-hSAC1-C/S. To generate Schizosaccharomyces pombe expression vectors pESP-hSAC1wt and pESP-hSAC1-C/S, the BamHI site of pESP1 (Stratagene) was exploited to introduce the hSAC1 variants following the PCR-mediated addition of BamHI sites directly to the respective coding sequences. Products were confirmed by sequencing. BamHI-flanked ORFs from pESP-hSAC1wt and pESP-hSAC1-C/S were used to generate Cu 2ϩinducible hSAC1 variants in pYEX-BX (Clontech), resulting in pYEX-BX-hSAC1wt and pYEX-BX-hSAC1-C/S, respectively. To build pGFP-hSAC1wt and pGFP-hSAC1-C/S, the same BamHI-fragments were used to generate intermediates in pENTR3C prior to Gateway TM -mediated shuttling (Invitrogen) into gEGFP-A30, an appropriately modified derivative of pEGFP-C2 (Clontech).
Expression and Purification of Recombinant GST-hSAC1 Proteins in S. pombe-The vectors pESP-hSAC1wt and pESP-hSAC1-C/S, containing GST-hSAC1 fusion sequences, were transformed into the S. pombe strain SP-Q01 (Stratagene) and induced as described by the manufacturer (Stratagene). Cells were disrupted by French press in lysis buffer containing 0.3 M sorbitol, 0.1 M NaCl, 5 mM MgCl 2 , 10 mM Tris/Cl, pH 7.2, and 1% Triton X-100 followed by batch purification of GST fusion proteins with glutathione 4B FF-Sepharose (Amersham Biosciences) according to the manufacturer. Washings were done with lysis puffer, and elution of proteins occurred in 10 mM Tris/Cl, 10 mM glutathione, pH 7.5. For storage at Ϫ80°C, glycerol was added to 25% final concentration.
Phosphatase Activity Assay-To measure phosphatase activity, a modified version of a malachite green assay was used (32). Phospho inositides were purchased from Cell Signals Inc., and phosphatidylserine was obtained from Sigma (P-1060). 1 g of recombinant GST-hSAC1wt and GST-hSAC1-C/S in 25 l of storage buffer were incubated (1 h, 32°C) with 25 l of liposomes, prepared by sonification in reaction buffer (500 M phosphatidylserine, 100 M diC 16 -PtdInsP isoform, 200 mM sodium acetate, 100 mM Tris-base, 100 mM Bis-Tris, 20 g/ml porcine gelatin, pH 6.0). Reactions were stopped by the addition of 20 l of 100 mM N-ethylmaleimide and centrifugation (14,000 ϫ g, 15 min). 25 l of the reaction supernatant was transferred to a multiwell plate, incubated with (20 min, room temperature) 50 l of malachite green solution (33) to quantify released phosphate at 600 nm. Phosphate release was calibrated to a dilution series of KH 2 PO 4 .
Peptides, Immunization, and Antibodies-Peptides were synthesized according to standard procedures followed by purification on a reverse phase high pressure liquid chromatograph. Peptides used for competition experiments were hSAC1/578 -587 (PRLVQKEKID), hSAC1/41-55 (TLAVKKDVPPSAVTR), and crosstide (biotin-GRPRTSSFAEG). hSAC1 antibodies were raised against recombinant GST-hSAC1 wild type protein in rabbits (Eurogentec). GST-reactive immunoglobulins were removed by adsorption to GST-Sepharose beads (serum 252 and serum 253). Additionally, polyclonal peptide antisera 69 and 7889 were raised against hSAC1 amino acids 570 -587 and 41-55, respectively. The following antibodies were commercially available and used accord- Immunoprecipitations and Western Blotting-Transfer of proteins to nitrocellulose was performed according to standard procedures. For hSAC1 immunoblotting, membranes were blocked with 5% skim milk in TBST (50 mM Tris/Cl, pH 7.5, 150 mM NaCl, 0.1% Tween 20) and washed with TBST. Visualization was done using suitable horseradish peroxidase-conjugated secondary antibodies and ECL. For immunoprecipitation, cell lysates were prepared by scraping subconfluent cells on ice into lysis buffer (20 mM Hepes, pH 7.9, 150 mM NaCl, 0.1% Triton X-100, Complete™), followed by passage through a 26-gauge needle (five times), clearance of the lysate by centrifugation (10 min, 6000 ϫ g, 4°C), and preincubation (15 min, 4°C) with either protein A-or protein G-Sepharose 4B FF (Amersham Biosciences). Precleared lysates were incubated (2 h, 4°C) with antibodies, followed by the addition and further incubation (45 min, 4°C) of 50 l of protein A-or protein G-Sepharose and five washes in lysis buffer. For oligomerization experiments, precleared lysates were incubated (32°C, 30 min) prior to antibody reactions. Samples were boiled in 2ϫ sample buffer (Roti®load 1; Roth) and subjected to SDS-PAGE. Lipid Binding Assay-For assessment of phospholipid binding properties, PIP Strips TM (Echelon, Inc.) were blocked (1 h, room temperature) with TBST including 3% bovine serum albumin, incubated either with purified GST, GST-hSAC1wt, or GST-hSAC1-C/S at a concentration of 0.5 g/ml in blocking buffer at 4°C overnight. Following three washes in TBST, PIP Strips TM were incubated (1 h, room temperature) with anti-GST antibody in TBST followed by standard secondary antibody incubation and ECL to detect GST-tagged proteins bound to the phospholipid spots on the membrane.
Cell Lines and Media-Cells were maintained at 37°C in a humidified atmosphere with 5% CO 2 . All media were supplemented with 10% fetal calf serum unless stated otherwise. Human glioblastoma cells U373MG were kept in Eagle's minimum essential medium (BioWhittaker), human prostate carcinoma cells PC-3 were kept in RPMI1640 (Invitrogen), human colon carcinoma cells HCT116, HeLa (cervix carcinoma) cells were kept in Dulbecco's modified Eagle's medium with 1% penicillin/streptomycin (Invitrogen), and monkey kidney COS-7 cells were kept in Dulbecco's modified Eagle's medium (Invitrogen). Mammalian expression constructs (10 g) were transfected using Fugene 6 (Roche Applied Science) according to the manufacturer's instructions. To generate U373MG cell lines stably expressing FLAG-hSAC1wt, FLAG-hSAC1-C/S, GFP-hSAC1wt, and GFP-hSAC1-C/S, the constructs and appropriate vector controls were transfected and selected in the presence of G418 (1 mg/ml; Calbiochem). Resistant clones were routinely kept in medium containing 200 g/ml G418. Prior to experiments, cells were adapted for 24 h to G418-free medium.

Identification of Human SAC1 and Generation of a PtdIns
phosphatase Inactive Mutant-To clone hSAC1 cDNA, we used a set of primers flanking the ORF found in the EST KIAA0851. A cDNA fragment identical to KIAA0851 nucleotides 40 -2272 was amplified via reverse transcriptase-PCR from brain mRNA (see "Experimental Procedures"). Two other human ESTs (KIAA0966 and KIAA0274) also contain a SAC1 homology domain within their sequence. KIAA0966 was recently designated hSAC2 (19), whereas KIAA0274 still awaits characterization.
The hSAC1 ORF (587 amino acids (aa)) exhibits 32% identity to ySac1p and 95% identity to rSAC1 (11) and is predicted to code for a protein of 64 kDa with a pI of 6.69. Inspection of the primary amino acid sequence reveals ( Fig. 1) two potential TM domains at the C terminus (aa 521-543, aa 550 -569) also present in ySAC1. A third TM domain at the N terminus of hSAC1 is not consistently predicted using different software tools and is therefore omitted in Fig. 1. A highly conserved phosphatase signature motif, CX 5 R(T/S), starts at aa 389 ( Fig.  1). Furthermore, a putative leucine zipper (hSAC1, aa 98 -126) is conserved in the mammalian SAC1 family members but absent in Caenorhabditis elegans or ySac1p ( Fig. 1) (17).
To generate a PtdIns phosphatase inactive mutant, cysteine 389 was replaced by serine to generate hSAC1-C/S. Sequences covering the ORFs for wild type and mutant hSAC1 were subcloned into various expression constructs (see "Experimental Procedures") to analyze their functions.
hSAC1, but Not hSAC1-C/S, Complements Inositol Auxotrophy of ⌬SAC1 Yeast Cells-To assess whether wild type hSAC1 is functionally equivalent to ySac1p in yeast cells and to confirm that the hSAC1-C/S mutant is inactive, we inducibly expressed hSAC1wt, hSAC1-C/S, and ySAC1 in a yeast strain deleted for the SAC1 gene (⌬SAC1; see "Experimental Procedures"). Immunoblotting demonstrated that induction in this strain yielded equal protein levels for both hSAC1 variants (Fig. 2).
One important phenotype of the ⌬SAC1 yeast strain is inositol auxotrophy (20), the inability to grow on inositol-free hSAC1 Interacts with COPI medium. As shown in Fig. 2, the ⌬SAC1 inositol auxotrophy phenotype could be rescued by Cu 2ϩ -inducible expression of hSAC1 to an extent comparable with expression of wild type ySac1p. In contrast, the hSAC1-C/S mutant was not able to complement inositol auxotrophy similar to the empty vector control. This result demonstrates (i) functional similarity between the yeast and the human SAC1 proteins and (ii) a successful impairment of hSAC1 enzymatic function through mutation of the core phosphatase motif.
The hSAC1-C/S Mutant Protein Displays Reduced Lipid Phosphatase Activity-To assess whether the failure of the hSAC1-C/S mutant to complement a ySac1p-dependent func-tion reflects its inability to process PtdIns phosphates, we expressed and purified GST-conjugated wild type and C/S mutant hSAC1 variants (Fig. 3A). Recombinant proteins were incubated with each of the seven phosphorylated isoforms of PtdIns, and phosphate release was determined employing a malachite green assay (see "Experimental Procedures"). GST-hSAC1wt enzyme efficiently dephosphorylated PtdIns(4)P and, to a lesser extent, PtdIns(3)P, whereas other PtdInsP species were not processed (Fig. 3B). In contrast, no significant dephosphorylation was observed on any of the substrates when GST-hSAC1-C/S was used. We could not detect significant activity of wild type enzyme toward PtdIns(3,5)P 2 , as reported for ySac1p hSAC1 Interacts with COPI or rSAC1. There are several possibilities to explain this discrepancy (see "Discussion"). In conclusion, mutation of the core phosphatase motif in hSAC1 abolishes its ability to dephosphorylate PtdIns phosphates.
Wild Type hSAC1 Forms Oligomers-Given that the C/S mutation disables the PtdIns phosphatase function of hSAC1, we tested whether this mutant acts as a classical dominant negative in the sense of forming inactive heterodimers with the wild type enzyme. To this end, we generated extracts from U373MG glioblastoma cells stably expressing GFP-tagged variants of hSAC1wt or hSAC1-C/S (see below; see "Experimental Procedures"). Such clones expressed comparable amounts of wild type and mutant GFP-hSAC1 fusion proteins (see Figs. 4A and 6A). To these extracts, purified GST-hSAC1wt was added and subsequently immunoprecipitated using a GST-specific antibody. Equal amounts of GST-hSAC1wt were detected in the immunoprecipitates as revealed by immunoblotting with GST antibody. Analysis of this material with GFP antibody revealed efficient binding of GST-hSAC1wt to GFP-hSAC1wt but strongly reduced binding to GFP-hSAC1-C/S (Fig. 4A). Using GST-hSAC1-C/S in such experiments revealed neither binding to GFP-hSAC1wt nor GFP-hSAC1-C/S (data not shown). To further support these results, we cotransfected expression constructs coding for FLAG-tagged hSAC1wt (pcDNA3.1-FLAG-hSAC1wt; see "Experimental Procedures") with constructs expressing either GFP-hSAC1wt or GFP-hSAC1-C/S into COS-7 cells. Two days after transfection, cells were lysed and subjected to immunoprecipitation employing an anti-FLAG monoclonal antibody. Equal amounts of FLAGtagged hSAC1wt were precipitated in each case. Efficient interaction between the FLAG-tagged and the GFP-tagged hSAC1 variant was only detected in GFP-hSAC1wt and not in GFP-hSAC1-C/S samples (Fig. 4B). These results demonstrate that hSAC1-C/S is not likely to function as a dominant negative mutant, because it does not form a complex with the wild type protein. They also show that the wild type hSAC1 protein can form oligomers, although we cannot formally exclude the possibility that a bridging factor(s), which might not be bound by the C/S mutant, is responsible for the interaction between hSAC1wt proteins.
hSAC1wt and hSAC1-C/S Display Equivalent PtdInsP Binding Properties-As stated above, two biochemical features of the wild type enzyme, PtdIns phosphatase activity and selfassociation, are eliminated in the C/S mutant. To test whether mutation of the core phosphatase motif would also influence the lipid binding capacity of hSAC1, we subjected the respective purified GST fusion proteins to a lipid blot analysis (see "Experimental Procedures"). Both GST-hSAC1wt and GST-hSAC1-C/S protein bound with highly comparable affinity to FIG. 4. Wild type hSAC1 forms oligomers. A, U373MG cells expressing GFP (clone 3), GFP-hSAC1wt (clone 5), or GFP-hSAC1-C/S (clone 6) were seeded. 16 h later, extracts were prepared and normalized, aliquots were removed for input analysis (input, lanes 1-3), and the rest were incubated (30 min, 32°C) with 2 g of recombinant GST-hSAC1wt. Immunoprecipitation using an anti-GST antibody (IP: ␣-GST, lanes 4 -6) was followed by SDS-PAGE and immunoblotting employing an anti-GFP (WB: ␣-GFP) or anti-GST (WB: ␣-GST) antibody. GST-hSAC1wt interacts efficiently with GFP-hSAC1wt. IgG h.c. and IgG l.c., immunoglobulin G heavy and light chain, respectively. B, COS-7 cells were transfected with pcDNA3.1-FLAG-hSAC1wt plus constructs expressing either GFP-hSAC1wt or GFP-hSAC1-C/S and lysed 48 h later. Precleared cell extracts were incubated (30 min, 32°C) and subjected to immunoprecipitation with anti-FLAG antibody followed by immunoblotting with either anti-FLAG (␣-FLAG) or anti-GFP (␣-GFP) antibodies. Molecular weight standards and 10% of starting material (10% input) are depicted.  1 (lane 4), or GFP-hSAC1-C/S clone 6 (lane 5) are shown next to human kidney the same subset of offered phospholipids, whereas the GST control was unable to interact with any lipid (Fig. 5). The GST-hSAC1 variants displayed highest affinity to the monophosphorylated PtdIns phosphates, followed by efficient association with PtdIns(3,5)P 2 and PtdIns(3,4,5)P 3 . Only weak interaction was observed with PtdIns(3,4)P 2 , PtdIns(4,5)P 2 , or unphosphorylated PtdIns, whereas lysophosphatidic acid, lysophosphocholine, and sphingosine 1-phosphate were not bound (Fig. 5). To our knowledge, this is the first time that the phospholipid-binding profile of a SAC1 family member has been reported. Moreover, hSAC1-C/S displays essentially unaltered PtdInsP binding, which argues in favor of undisturbed protein folding and modification prior to purification from S. pombe (also see "Discussion").
hSAC1-C/S Accumulates in the Golgi-There have been conflicting reports on SAC1 subcellular localization. The yeast protein was shown to be part of the Golgi and the ER (for a review, see Ref. 10), whereas rSAC1 was found to be an ERassociated protein when ectopic expression of an epitopetagged version was used (11). We therefore investigated localization of endogenous hSAC1 using new antisera raised against hSAC1 peptides or whole protein (see "Experimental Procedures"). Such rabbit antisera recognized a single protein species of 64 kDa widely expressed in normal human tissues and tumor cell lines (Fig. 6A, data not shown). Using serum 69, we could clearly demonstrate that hSAC1 exhibits both ER and Golgi localization in HeLa cells, as corroborated by co-staining with antibodies raised against Golgi-specific protein p58 (34) or the ER membrane-associated translocation complex protein Sec61␣ (35) (Fig. 6B). GFP-tagged hSAC1 variants now allowed us to address whether the enzymatic activity of hSAC1 has an influence on subcellular localization. To this end, U373MG glioblastoma cells were stably transfected with GFP-conjugated hSAC1 wild type or the C/S variant. Clones exhibiting similar expression levels of ectopically expressed GFP-hSAC1 variants (Fig. 6A) were selected for immunofluorescence analysis. As supported by respective costaining with antibodies against the Golgi marker golgin-97 (36) or Sec61␣, GFP-SAC1wt localizes to ER structures and perinuclear Golgi structures, very reminiscent of the localization of endogenous hSAC1 (Fig. 6, C and D). In contrast, the C/S mutant clearly exhibits accumulation in the Golgi with only small amounts left in the ER (Fig. 6, C and D). This difference in localization (also see Table I) was also observed when other sets of GFP-hSAC1wt-versus GFP-hSAC1-C/S-expressing U373MG clones were analyzed or when life cells were recorded (Fig. 6E).
hSAC1 Wild Type but Not hSAC1-C/S Interacts with COPI-The accumulation of the PtdIns phosphatase inactive mutant in the Golgi as opposed to the wider distribution of the wild type enzyme poses several interesting questions: (i) To what extent is the PtdIns phosphatase function of hSAC1 essential to maintain the wider distribution? (ii) Are differential interactions to ER or Golgi proteins responsible for the divergent subcellular localization of the two hSAC1 variants? (iii) Does hSAC1 influence vesicle transport dynamics between the ER and the Golgi? Some of these questions could be addressed with a better knowledge about hSAC1-binding proteins. To this end, we generated preparative immunoprecipitates from U373MG cells stably expressing GFP-tagged or FLAG-tagged versions of hSAC1wt or hSAC1-C/S, respectively. Silver staining of these immunoprecipitated samples revealed consistently, in addition to the respective hSAC1 fusion proteins, protein species of about 150 and 100 kDa only when using lysates expressing hSAC1wt and not hSAC1-C/S or control proteins (Fig. 7A, data not shown). To identify the proteins interacting with hSAC1wt, respective gel pieces were excised and subjected to MALDI-TOF mass spectrometry analysis (see "Experimental Procedures"). The 100-kDa protein species could unequivocally be identified as ␤-coatomer protein (␤-COP), whereas the 150-kDa protein turned out to be ␣-coatomer protein (␣-COP).
To independently corroborate the interaction, we subjected U373MG cells stably expressing GFP-hSAC1 variants to immunoprecipitation with anti-GFP antibody followed by immunoblotting with antibodies raised against ␣-COP, ␤-COP, ␥-COP, or ⑀-COP. In all of these cases, the respective member of the coatomer I complex could be identified in immunoprecipitates containing wild type hSAC1 but not catalytically inactive hSAC1-C/S (Fig. 7B). A virtually identical result was obtained in experiments with stably expressed FLAG-tagged hSAC1 variants, demonstrating that the interaction is clearly not tag-dependent (Fig. 7B). Furthermore, when ␥-COP protein was immunoprecipitated from the respective stable U373MG cell line lysates, GFP-tagged hSAC1wt, but not hSAC1-C/S, could be detected (data not shown).
To demonstrate that also endogenous hSAC1 interacts with the COPI complex, human PC-3 prostate carcinoma cells were lysed and immunoprecipitated using our new hSAC1 antibodies 252 and 69. Clearly, both hSAC1 antisera immunoprecipitated hSAC1 protein together with COPI (represented by detection of ␥-COP), whereas preimmune serum controls did not (Fig. 7C). Conversely, immunoprecipitation using ␥-COP antibody revealed endogenous hSAC1 as a complex partner (Fig.  7D), clearly demonstrating that hSAC1 binds to the coatomer I complex.
A Putative COPI Interaction Motif, KEKID, Is Essential for the hSAC1/COPI Interaction-Inspection of the hSAC1 primary amino acid sequence revealed a motif, KEKID-COOH, at  Fig. 6, C and D) or transiently transfected COS-7 cells (see Fig. 8C) were analyzed by immunofluorescence microscopy. The percentage of cells showing localization predominantly to the indicated organelle is depicted. At least 120 cells per sample were counted. the extreme C terminus, which resembles the KXKXX motif described to be functional in coatomer I-dependent ER retrieval (37). This motif is absent in ySac1p. Since both hSAC1 variants contain this motif, we first wanted to demonstrate its functionality by using peptide competition experiments. Immunoprecipitation of GFP-hSAC1wt from a U373MG clone followed by ␥-COP immunoblotting revealed the interaction already described. The addition of a peptide encompassing the last 10 amino acids of hSAC1 containing the KEKID motif (hSAC1/ 578 -587) efficiently blocked the interaction between GFP-hSAC1wt and the coatomer I complex, whereas a peptide corresponding to another region of hSAC1 (hSAC1/41-55) or an unrelated peptide (crosstide; see "Experimental Procedures") did not compete (Fig. 8A).
To further validate the finding that the KEKID motif is responsible for the hSAC1/COPI interaction, we generated a GFP-hSAC1 mutant in which the two lysines in the KEKID motif were replaced by alanine (resulting in AEAID) to generate the construct GFP-hSAC1-K2A. This mutant was transiently expressed in HCT116 cells. Lysates were immunoprecipitated using GFPantibody and subjected to ␥-COP immunoblotting. Whereas transiently expressed GFP-hSAC1wt interacts with COPI, no binding could be detected using the GFP-hSAC1-K2A mutant (Fig. 8B). The GFP-hSAC1-C/S mutant displayed also in this cell type strongly reduced COPI binding. These findings demonstrate that hSAC1 harbors a functional COPI interaction motif at its extreme C terminus and that this motif cannot be used efficiently when the PtdIns phosphatase function of hSAC1 is abrogated.
The KEKID Mutant Accumulates in the Golgi-To demonstrate that the KEKID motif of hSAC1 is functional in vivo, we transfected constructs coding for GFP-tagged hSAC1-K2A or for the respective wild type and C/S enzymes into COS-7 cells. Since we reproducibly observed that cells with very high expression of the three GFP-tagged hSAC1 variants tend to (i) form aberrant cytoplasmic aggregates of GFP-positive material and (ii) undergo apoptosis (data not shown), we only evaluated cells that exhibited fluorescence intensity comparable with U373MG clones stably expressing GFP-hSAC1. Each variant displayed a distribution of localization phenotypes (quantified in Table I), which, however, clearly allowed us to conclude that the GFP-hSAC1-K2A mutant accumulates in the Golgi apparatus to an extent very similar to GFP-SAC1-C/S, whereas the wild type protein again mainly displayed ER plus Golgi localization (Fig. 8C). FIG. 7. hSAC1 binds to the coatomer I complex. A, U373MG cells stably expressing GFP-hSAC1wt, GFP-hSAC1-C/S, or GFP proteins were immunoprecipitated with anti-GFP antibody, subjected to SDS-PAGE, and analyzed using silver staining. Two protein species of 150 and 100 kDa predominantly interacting with GFP-hSACwt were excised and subjected to tryptic digests, followed by MALDI-TOF mass spectrometry (not shown). They were identified as ␣-COP and ␤-COP, respectively (long arrows). Positions of GFP, GFP-hSAC1 fusion proteins, and molecular weight standards (MW) are indicated. IgG h.c., immunoglobulin G heavy chain; IgG l.c., immunoglobulin G light chain. B, GFP-or FLAG-tagged hSAC1 variants were immunoprecipitated from respective U373MG clones using anti-GFP (␣-GFP; lanes 1-3), or anti-FLAG (␣-FLAG; lanes 4 -6) antibodies. Immunoprecipitated proteins were separated by SDS-PAGE, transferred to nitrocellulose, and incubated with antibodies against ␣-COP, ␤-COP, ␥-COP, and ⑀-COP.
To demonstrate equal expression of ␥-COP in various clones, lysates were directly analyzed by immunoblotting (lower panel, ␥-COP/lysate). Equal immunoprecipitation of GFP-hSAC1 or FLAG-hSAC1 fusions from respective U373MG clones was demonstrated in similar, independent experiments (not shown). C, PC-3 lysates were subjected to immunoprecipitation using hSAC1 antisera 252 and 69. The corresponding preimmune sera (Pre) served as controls. Immunoprecipitated proteins were analyzed by immunoblotting with ␥-COP (upper panel) and hSAC1 (antiserum 253) antibodies, respectively. D, the same experiment as in C, except that goat anti-␥-COP antibody was used for immunoprecipitation. Positions of endogenous hSAC1, endogenous ␥-COP, and molecular weight standards (MW) are indicated. control, goat antibody raised against phospho-Ser 193 AFX.

DISCUSSION
We show that the endogenous hSAC1 protein can be found localized to the ER and the Golgi apparatus in nonsynchronized human cells. Two mutants of hSAC1 were generated, one eliminating the PtdIns phosphatase function (hSAC1-C/S) and the other destroying a putative COPI interaction signal at the extreme C terminus of hSAC1 (hSAC1-K2A). Both mutants show the same accumulation in the Golgi apparatus when expressed as GFP fusion proteins, whereas the wild type protein, GFP-hSAC1wt, behaves like the endogenous protein and displays ER plus Golgi localization. We could further demonstrate for the first time that the coatomer I complex interacts with hSAC1, that a peptide harboring the cognate C-terminal K(X)KXX motif (KEKID) efficiently competes with hSAC1 for binding to COPI and that mutation of the two key lysine residues eliminates the hSAC1/COPI interaction. Surprisingly, this KEKID motif is present in the PtdIns phosphatase inactive hSAC1-C/S mutant, which we show is unable to efficiently interact with COPI.
Lipid Binding and Substrate Specificity-We consider it unlikely that the hSAC1-C/S protein displays aberrant folding due to the point mutation at amino acid 389, since we could demonstrate essentially unaltered PtdIns binding capability when comparing with the respective wild type protein. We have for the first time characterized the lipid binding potential of hSAC1 and show that it binds to monophosphorylated PtdIns phosphates with highest affinity, to PtdIns(3,5)P 2 and PtdIns(3,4,5)P 3 with medium affinity, and to PtdIns(3,4)P 2 , PtdIns(4,5)P 2 , and unphosphorylated PtdIns with lowest affinity. Analysis of more mutants, however, is necessary to actually map the lipid-binding domain of hSAC1.
We could further demonstrate that hSAC1 dephosphorylates PtdIns(4)P and PtdIns(3)P, whereas PtdIns(3,5)P 2 apparently did not serve as a substrate. This is in contrast to reports using ySac1p or rSAC1 as enzymatic source. This discrepancy could arise from different posttranslational modifications due to the expression systems used (Sf9 insect cells for rSAC1 versus S. pombe for hSAC1). Moreover, we used full-length hSAC1  Table I. protein, whereas other studies analyzed truncated versions of SAC1 (11,31,38). Nevertheless, PtdIns(4)P and PtdIns(3)P always were the two phosphoinositides to be converted most efficiently.
In conclusion, hSAC1 appears to display affinity to more phospholipids than it actually uses as substrates. We cannot exclude, however, the possibility that the expanded PtdIns-binding profile as compared with the catalytic profile reflects that hSAC1wt requires slightly distinct hydrolysis conditions for various PtdIns isoforms that we might have missed in our assay.
COPI: First Interaction Partner for SAC1-Despite a history of more than 14 years in SAC1 research, interaction partners for this protein have not been identified. Recently, Bsp1p/ Ypr171p was reported to bind the proteins Sjl2p and Sjl3p in a region confined to the SAC homology domain. Bsp1p/Ypr171p might act as an adaptor protein for linking its binding partners to the cortical actin cytoskeleton. However, no interaction was observed with Sjl1p or ySac1p (39).
We demonstrate for the first time that hSAC1 binds to members of the COPI complex (␣, ␤, ␥, ⑀). Although inspection of the human hSAC1 sequence might have implied such an interaction due to the presence of a cognate K(X)KXX sequence, this motif is not evolutionarily conserved in the SAC1 family. Several studies have documented the interaction between coatomer I and a C-terminal K(X)KXX motif of transmembrane proteins that have to be retrieved to the ER. It is still controversial which subunit(s) of COPI binds the dilysine signal. Genetic screens revealed that mutations in ␣, ␤Ј, ␥, ␦, and -COP subunits interfere with KKXX retrieval (37,40). Selective cross-linking of dilysine peptides disclosed ␥-COP as the essential subunit (41,42), whereas ␣-COP emerged as the major interaction partner for KKXX motifs in a combinatorial screening approach (43). Interestingly, a KXKXX motif did not function as a strong ER retrieval motif in this setting. The inconsistencies in identifying a single COPI subunit for dilysine motif binding might be explained by the finding that there are probably at least two dilysine binding sites on the assembled coatomer I complex that might not necessarily lie on the same subunit (44).
Genetic screens in yeast revealed that disruption of SAC1 in a mutant background of SEC21 aggravated the SEC21 phenotype, whereas other mutants in the secretory pathway (like SEC14) were suppressed by a mutant SAC1 allele (24). Sec21p is the yeast homologue of ␥-COP. It is, however, currently unknown whether the C-terminal motif of ySac1p (PLKRD), lacking a lysine at the Ϫ4 or Ϫ5 position, is able to interact directly with Sec21p or other members of the yeast coatomer I complex.
Dynamics of ER/Golgi Localization-It is surprising that we could alter the subcellular distribution of hSAC1 by inactivating its PtdIns phosphatase function. Accumulation of the hSAC1-C/S mutant in the Golgi could be explained by a loss in COPI binding and might reflect that wild type hSAC1 usually functions by assuring its retrograde transport from the Golgi to the ER. The question remains why then inactivation of PtdIns phosphatase activity causes a loss in COPI binding, especially since the cognate retrieval motif, KEKID, is present. It has been reported that ␣-COP binds PtdIns(3,4,5)P 3 and to a lesser extent PtdIns(3,4)P 3 and PtdIns(4,5)P 2 (other PtdIns derivatives were not tested (45)). Inactive hSAC1 could therefore cause an alteration of the phospholipid composition that directly or indirectly influences attraction of COPI subunits to their target membrane. In this case, hSAC1 would be an essential player in COPI vesicle generation and might impose an additional layer of regulation on top of GTP-bound ARF or the p23/p24 family of coatomer receptors (for a review, see Ref. 30).
This scenario might only apply to our experimental system if the number of hSAC1-C/S molecules per COPI vesicle is high enough to outnumber endogenous hSAC1wt proteins, since the latter might still generate the necessary PtdIns phosphates in the test cells used. In this regard, it is important to keep in mind that the hSAC1-C/S mutant does not act as a classical dominant negative, since it does not form heterodimers (Fig. 4). Determination of PtdIns levels at hSAC1-C/S-containing COPI vesicles might solve these questions, yet this is beyond the scope of this work.
On the other hand, hSAC1 might act as a classical transmembrane cargo without enzymatic impact on coatomer I formation. The fact that endogenous SAC1 protein interacts with members of the coatomer complex I can be taken as a strong support for the notion that hSAC1 does indeed use both TM domains and does acquire a "J"-like topology with the N terminus and the KXKXX motif exposed to the cytoplasm. As such, COPI subunits might capture it and route it for antero-or retrograde mode of transport. Since mutation of the two key lysine residues in the hSAC1-K2A mutant disabled interaction with COPI and caused accumulation in the Golgi, the similar behavior of hSAC1-C/S might therefore imply that its KEKID motif is not effectively used. Since motifs of the KXKXX type apparently are less efficiently retrieved than polypeptides with a KKXX motif (43), and since coatomer I offers more than two dilysine binding sites (see above), successful retrieval of hSAC1wt might depend on, for example, oligomerization to fully exploit otherwise weak COPI interaction. The failure of the C/S mutant to oligomerize might in this way be at the cost of its efficient retrieval. It is not clear at present whether the C/S mutation directly or indirectly affects the oligomerization competence of hSAC1. In this respect, it might be elucidating to introduce mutations into the leucine zipper of hSAC1 to assess whether this classical dimerization motif is actually used and whether it has an influence on COPI binding and/or subcellular localization.
Alternatively, inefficient binding of the hSAC1-C/S KXKXX motif to COPI might be caused by removing it spatially from COPI access. Such removal might be induced by a mutantspecific conformational change with impact on the second TM domain. This impact could be profound in the sense that the whole second TM domain flips through the membrane, thereby rendering the KEKID motif luminal. Such alternative TM usage is reported for Escherichia coli lactose permease LacY, dependent on the phospholipid composition of the inner membrane (46), or other mammalian and viral proteins (47,48). Alternatively, the impact might be subtle and solely move the second TM domain further into the bilayer. Interestingly, alternative TM predictions for the second TM domain (Fig. 1, data not shown) extend into aa 574, which would leave only 8 amino acids between the bilayer and the KEKID motif, a distance reported to be too short to allow ER retrieval of a KKXXcontaining transmembrane protein (49).
We have started to shed new light on the function of the human homologue of ySac1p. We used ectopic expression of mutant proteins and learned about novel interaction partners and a dynamic distribution of hSAC1 between the ER and Golgi. It might also be very interesting to analyze whether other point mutations of ySac1p, like those contributing to multiple drug sensitivity, perform in a similar way in the human counterpart and how they might influence COPI binding or vesicle trafficking. to Martin Steegmaier and Norbert Kraut for critical reading of the manuscript, and to Wolfgang Rettig for constant support.