A Yeast Golgi E-type ATPase with an Unusual Membrane Topology

E-type ATPases are involved in many biological processes such as modulation of neural cell activity, prevention of intravascular thrombosis, and protein glycosylation. In this study, we show that a gene of Saccharomyces cerevisiae , identified by similarity to that of animal ectoapyrase CD39, codes for a new member of the E-type ATPase family (Apy1p). Overexpression of Apy1p in yeast cells causes an increase in intracellular membrane-bound nucleoside di- and triphosphate hydro-lase activity. The activity is highest with ADP as substrate and is stimulated similarly by Ca 2 1 , Mg 2 1 , and Mn 2 1 . The results also indicate that Apy1p is an integral membrane protein located predominantly in the Golgi compartment. Sequence analysis reveals that Apy1p contains one large NH 2 -terminal hydrophilic apyrase domain, one COOH-ter- minal hydrophilic domain, and two hydrophobic stretches in the central region of the polypeptide. Although no signal sequence is found at the NH 2 -terminal portion of the pro- tein and no NH 2 -terminal cleavage of the protein is ob- served, demonstrated by the detection of NH 2 -terminal tagged Apy1p, the NH 2 -terminal domain of Apylp is on the luminal side of the Golgi apparatus, and the COOH-termi-nal hydrophilic domain binds to the cytoplasmic face of the Golgi membrane. The second hydrophobic stretch of Apy1p is the transmembrane domain. These results indicate that Apylp is a type III transmembrane protein; however, the size of the Apy1p extracytoplasmic NH 2 terminus is much larger than those of other type III transmembrane proteins, suggesting that a novel

E-type ATPases (E-ATPases) 1 are found in most eukaryotic cells and hydrolyze nucleotide tri-and/or diphosphates in the presence of Ca 2ϩ or Mg 2ϩ (1). They play important roles in many biological processes including the modulation of neural cell activities (2), prevention of intravascular thrombosis (3,4), and regulation of immune responses (5). The molecular identity of the E-ATPases was revealed recently by the purification and cloning of a soluble apyrase from potato tubers (Solanum tuberosum) (6). The deduced amino acid sequence for potato apyrase had sequence similarities with various other polypeptides in the data base, including CD39, a human and mouse lymphoid cell antigen (7), garden pea NTPase (8), and Toxo-plasma gondii NTPase (9). All of these proteins contain four highly conserved sequences called apyrase conserved regions (ACR1-4). The sequences of ACR1 and ACR4 are similar to those of the actin-hsp70-hexokinase ␤ and ␥-phosphate binding motifs, respectively, suggesting a possible role in nucleotide binding (6). CD39 was subsequently shown to exhibit E-ATPase activity (5), which confirmed the existence of a novel protein family for E-ATPases. Since then, additional members of the E-ATPase family have been identified (10 -16).
Most E-ATPases are transmembrane proteins (5, 10 -12, 14, 17, 18). Transmembrane proteins of the endoplasmic reticulum (ER) and ER-derived cell organelles are inserted cotranslationally into the ER membrane in a signal recognition particle-dependent manner (19 -21). During the insertion process, membrane topology is determined by interaction between topogenic signals within the nascent protein and the translocation machinery; these signals are only partially understood (22). Four types of single-spanning membrane proteins can be distinguished based on the topogenic sequence involved in the insertion (23,24). Type I membrane proteins have a cleaved NH 2terminal signal sequence followed by a transmembrane anchor segment, and their mature NH 2 terminus is extracytoplasmic (N exo ). Type II proteins contain an uncleaved signal/anchor sequence resulting in the cytoplasmic localization of the NH 2 terminus (N cyto ). In type III proteins, the NH 2 terminus is translocated to the lumen of the ER (N exo ), and the signal sequence is not cleaved (23). Type IV proteins have a N cyto /C exo topology like type II proteins but have the transmembrane segment very close to the COOH terminus and are inserted into the ER membrane by uncharacterized machinery (24,25). All reported membrane-bound E-ATPases have a type II-like orientation.
In the lumen of the Golgi apparatus, proteins and lipids become glycosylated. Nucleotide sugars are synthesized in the cytosol and transported into the Golgi lumen via specific carrier proteins (26,27). After transfer of sugar residues to proteins and lipids by glycosyltransferases, the resulting nucleoside diphosphates are converted to nucleoside monophosphates by nucleotide diphosphatases. In this way, nucleoside diphosphates that are inhibitors of glycosyltransferases do not accumulate in the lumen of the Golgi apparatus. The nucleoside monophosphates exit the lumen of Golgi in exchange with cytosolic sugar nucleotides. It has been shown previously that a Saccharomyces cerevisiae GDPase (GDA1) is required for protein and lipid mannosylation (17,28). Gda1p was recognized as an E-ATPase because of its similarity to potato apyrase (6) and to animal ectoapyrase CD39 (5). Deletion of the gene has a minor effect on the growth of yeast but does result in decreased mannosylation of membrane proteins (28). A sequence homology search in the GenBank data base revealed another gene from S. cerevisiae (GenBank accession number P40009) with high similarity in the ACR1-4 motifs to members of the E-ATPase family. This gene is on chromosome V, encodes a hypothetical 71.9-kDa protein, and was proposed recently to be the second E-ATPase found in yeast (14). In this study, we demonstrate that this 71.9-kDa protein is indeed a novel E-ATPase. Interestingly, it is located in the Golgi and has an unusual membrane topology.
DNA Constructs-DNA manipulations were carried out according to standard protocols (32). To clone a full-length APY1 gene, two primers were designed to amplify the DNA fragment encoding the putative 71.9-kDa protein from yeast chromosomal DNA by PCR. Yeast SK1 chromosomal DNA was isolated by standard protocol (33). The sense primer XZ101 (5Ј-CCCAAGCTTATCAACCAACGGCTTCTTCTCGCA-3Ј), with a HindIII site at the 5Ј-end, contained a sequence of the open reading frame (nucleotides 1-23 in Fig. 1). The antisense primer XZ102 (5Ј -CGCGGATCCTCAGGCGTAGTCCGGGACGTCATATGGGTAATC-ATATAGCCTACTGTCCTTAAATTTGGA-3Ј) contained a BamHI site at the 5Ј-end, an antisense sequence encoding an influenza hemagglutinin epitope (HA) recognized by monoclonal antibody 12CA5, and a sequence complementary to the APY1 gene (nucleotides 2134 -2163 in Fig. 1). A stop codon was introduced between the BamHI site and the HA-tagged sequence in the antisense primer. The resulting PCR fragment was cloned into the 2-m vector pRS316 (34; kindly provided by Dr. Neil Hunter, Harvard University); the resulting construct was named pGZ98. To express APY1-HA behind the glycerol dehydrogenase promoter, a PCR fragment was generated using pGZ98 as a template, primer XZ102 and primer XZ105 (5Ј-CGCGGATCCGTAACCATGCTC-ATAGAAAACACTAAT-3Ј, which contained a Kozak sequence right in front of the start codon, a BamHI site at the 5Ј-end, and nucleotide sequences 274 -294 ( Fig. 1)). The resulting PCR product was subcloned into the expression vector pG1 (35; kindly provided by Sean Burgess, Harvard University); the resulting construct was named as pGZ103.
Two primers were used to amplify a DNA fragment encoding an NH 2 -terminal myc-tagged Apy1p-HA. The sense primer XZ111 (5Ј-CG-CGGATCCGTAACCATGGAGCAAAAGCTGATATCTGAAGAGGACT-TGCTCATGAAAACACTAAT-3Ј) contained a BamHI site, a Kozak sequence, a start codon, a sense sequence encoding the myc epitope (EQKLISEEDL), and nucleotide sequence 277-294 in Fig. 1. The antisense primer was XZ102. The resulting PCR product was subcloned into pG1 (pGZ105).
To delete the segment between amino acid residues 500 and 630 of Apy1p, the PCR product of primer XZ105 and primer XZ123 (5Ј-CGC-GGATCCTCAGGCGTAGTCC-GGGACGTCATATGGGTATTTATTAG-AAAGGCTGGA-3Ј, containing a BamHI site, an antisense HA sequence and an sequence complementary to nucleotide sequence 1756 -1773) was subcloned into pG1 with BamHI (pGZ118). To introduce an addi- tional glycosylation site (D619N) in the N371I mutant, the PCR product of primer XZ105 and primer XZ122 (5Ј-ACGCGTCGACTCACTTAAAT-TTGGAAAAATTAGCCATAGAAAACGC-3Ј, containing an SalI site and a sequence complementary to nucleotides 2113-2145 with the mismatched nucleotide underlined) was digested with SalI and used to replace the corresponding SalI fragment of pGZ114 to form pGZ120. To delete amino acids 552-630 from Apy1p, the PCR product of primer XZ111 and primer XZ124 (5Ј-CGCGGATCCT-CAATCTGATCTCCTTA-GAAATTTCAA-3Ј, containing a BamHI site and a sequence complementary to nucleotides 1906 -1929) was digested with BamHI and subcloned into pG1 to form pGZ121. All the constructs were verified by DNA sequencing.
Membrane Preparation and Cell Fractionation-Cells were grown to A 600 ϳ 4.0 in the proper yeast drop-out medium containing 2% glucose at 30°C. NaN 3 was added to 10 mM, and the culture was harvested by centrifugation and washed once in 10 mM NaN 3 . Cells were resuspended at 0.25 g/ml in homogenization buffer (50 mM Tris-HCl, pH 7.5, 0.3 M sucrose, 5 mM EDTA, 1 mM EGTA, 5 mg/ml bovine serum albumin, 2 mM dithiothreitol, 2.5 mg/ml chymostatin, 1 mM phenylmethylsulfonyl fluoride) and lysed by passage through a French pressure cell (SLM-Amino, Urbana, IL) at 20,000 p.s.i. The lysate was centrifuged at 10,000 ϫ g for 20 min; the supernatant fraction was then centrifuged at 120,000 ϫ g for 1 h. The membrane pellet was resuspended in 10 mM Tris-HCl, pH 7.0, 1 mM EGTA, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride.
Isolation of Golgi membranes by differential centrifugation was done as described (36). Briefly, spheroplasts (300 A 600 units) were lysed by dilution in hypo-osmotic buffer. The lysate was centrifuged at 1,000 ϫ g for 10 min to generate the P1 (pellet) and the S1 (supernatant) fractions, and the latter was centrifuged at 13,000 ϫ g (P13) and 120,000 ϫ g (P120) for 20 and 60 min, respectively. The P13 and P120 pellets were resuspended in 0.8 M sorbitol, 10 mM triethanolamine, pH 7.2, 1 mM EDTA.
Protease Protection Assay-To examine the protease accessibility of Apy1p, 25 l of the P120 pellet containing Golgi-enriched vesicles in 0.8 M sorbitol, 10 mM triethanolamine, pH 7.2, 1 mM EDTA were added to the following solutions: (a) 25 l of 0.8 M sorbitol, 10 mM triethanolamine, pH 7.2, 1 mM EDTA; (b) 25 l of proteinase K (800 g/ml) in 0.8 M sorbitol, 10 mM triethanolamine, pH 7.2, 1 mM EDTA; (c) 25 l of 2% Triton X-100 in 0.8 M sorbitol, 10 mM triethanolamine, pH 7.2, 1 mM EDTA; (d) 25 l of 2% Triton X-100, 800 g/ml proteinase K, 0.8 M sorbitol, 10 mM triethanolamine, pH 7.2, 1 mM EDTA. All samples were incubated on ice for 45 min, and trichloroacetic acid was added to a final concentration of 10%. The samples were pelleted and washed with ice-cold acetone. The dried pellets were resuspended in sample buffer and subjected to 10% SDS-PAGE.
Measurement of Nucleoside Di-and Triphosphatase Activity-To measure apyrase activity, crude membranes of yeast cells (15 g) were suspended in 90 l of buffer A (20 mM HEPES-Tris, pH 7.4, 120 mM NaCl, 5 mM KCl, and 1 mM EGTA) with or without 10 mM CaCl 2 and preincubated for 5 min at 37°C. The nucleotidase reactions were initiated by the addition of 10 l of the same buffer containing 20 mM nucleotide. The divalent cation-stimulated apyrase activity was determined by measuring the inorganic phosphate released as described by Ames (37) and by subtracting values obtained with EGTA alone from those with 10 mM CaCl 2 plus chelator (14).
Immunoblotting and Immunofluorescence Staining-Anti-HA monoclonal antibody (12CA5, 1:2,000 dilution for immunoblotting) was purchased from Berkeley Antibody Corp. (Berkeley, CA). Anti-myc monoclonal antibody (9E10, 1:500 for immunoblotting) was purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Peroxidase-conjugated goat anti-mouse antibody (1:2,000) was purchased from Sigma. Immunoblotting was done as described previously (5), except that the incubation with the first antibody was for 5 h. Indirect immunofluorescence of the Apy1P-HA protein was performed by standard methods (39) using the anti-HA antibody 12CA5 at 1:150 dilution and fluorescein isothiocyanate-conjugated goat anti-mouse antibody (1:128, Sigma). Samples were observed and photographed with a Nikon microphot SA epifluorescence microscope.

Cloning of the APY1 Gene from Yeast Chromosomal DNA-
The full-length DNA sequence encoding the yeast 71.9-kDa protein (Apy1p) was cloned by PCR using yeast DNA as a template. Fig. 1 shows the sequence of the clone which has a few differences compared with the yeast genomic data base sequence. The ATG initiation codon at nucleotides 273-275 is the most likely translational initiation site because there are four stop codons, shown shadowed, preceding the ATG in the same open reading frame. The apyrase conserved regions (ACR1-4) are indicated in boldface. There are two potential N-glycosylation sites, shown in bold and italics, at amino acid residues 371-373 and 532-534.
Expression of Apy1p-To determine whether this gene can be expressed in yeast cells, the sequence for a HA tag was added in-frame to the 3Ј-end of the open reading frame, and the resulting cDNA was expressed behind a constitutive promoter in the yeast vector pG1. As shown in Fig 1 and 2), 1% Triton X-100 (lanes 3 and 4) on ice for 30 min, then centrifuged at 100,000 ϫ g for 20 min in a TLA100.3 rotor. Both pellets and supernatant fractions were analyzed by 10% SDS-PAGE, and Apy1p was detected by immunoblotting with anti-HA. dase F (Fig. 2B), a single band at 72 kDa (lane 1) was observed, showing that the 73-kDa protein band is the glycosylated form of Apy1p.
After the cells expressing APY1-HA were mechanically lysed as described under "Materials and Methods," the lysate was subjected to high speed centrifugation to separate the membrane fraction (pellet) and cytosol fraction (supernatant). Apy1p-HA was detected only in the membrane fractions, not in the cytosol (data not shown). To address further the membrane association of Apy1p (Fig. 2C), the membrane fraction was extracted with 0.1 M sodium carbonate, pH 11.0 (lanes 1 and 2), and 1% Triton X-100 (lanes 3 and 4). The treated samples were centrifuged at 100,000 ϫ g to produce pellet and supernatant. Nearly all Apy1p sedimented in the membrane pellet after the treatment with alkaline carbonate buffer but was extracted from the membrane into supernatant fraction with Triton X-100. These data indicate that Apy1p is an integral membrane protein. A 15-kDa band, present in lanes 3 and 4, is probably a degradation product of Apy1p. Moreover, a considerable amount of Apy1p was Triton X-100-insoluble and remained in the pellet after extraction.
Function of Apy1p-It was proposed that Apy1p might be a new member of the E-ATPase family because Apy1p contains four ACR motifs (14). To study the function of Apy1p, crude membrane fractions isolated from yeast cells overexpressing Apy1p and control yeast cells were assayed for their nucleotidase activities. As shown in Fig. 3A, expression of HA-tagged Apy1p increased the membrane-associated nucleotidase activity on various nucleoside di-and triphosphates. Both ADPase and TDPase activities are more than 20-fold higher than those of membranes from control yeast cells. Neither 1 mM azide (inhibitor of F-type ATPases) nor 0.5 mM vanadate (inhibitor of P-type ATPases) inhibited the activities. It is known that E-ATPases are resistant to these inhibitors (1), so Apy1p appears to be a new E-ATPase. Fig. 3B shows the ADP and TDP concentration dependence of nucleotidase activity. These activities were determined under conditions where the activities were linear with respect to time, and the substrate concentration did not change during the assay. ADPase and TDPase activities reached maximum values at concentrations of 4 mM and 2 mM, respectively. The nucleoside phosphatase activity of Apy1p was highest with ADP as the substrate. The activity of Apy1p ADPase was stimulated similarly by Ca 2ϩ , Mn 2ϩ , and Mg 2ϩ (Fig. 3C). Maximum activation was obtained at 5 mM divalent cation. The effect of cations on Apy1p activity is different from that on yeast GDPase (17) and human UDPase (14), in which Ca 2ϩ is more effective than Mn 2ϩ or Mg 2ϩ .
Golgi Localization of Apy1p-To study the cellular localization of this yeast E-ATPase, yeast cells expressing Apy1P-HA were subjected to subcellular fractionation by differential centrifugation. Spheroplasts were lysed in a hypo-osmotic buffer, and the lysate was subjected to sequential centrifugation at 1,000 ϫ g, 13,000 ϫ g, and 120,000 ϫ g. Most of the ER, vacuolar membrane, and plasma membrane are found in the 13,000 ϫ g pellet (40,41), whereas the 120,000 ϫ g pellet (P120) is a Golgi-enriched fraction, containing the majority of the Golgi enzymes ␣-1,3-mannosyltransferase (Mnn1p) (36,42), guanosine diphosphatase (Gda1p) (17), and endopeptidase Kex2p (43). We routinely found that about 70% of Apy1p-HA was in the P120 fraction, suggesting that Apy1p is localized in the Golgi fraction. To confirm further the Golgi localization of Apy1p-HA, immunofluorescence experiments with monoclonal anti-HA antibody were done. Yeast cells containing APY1-HA on a 2-m plasmid displayed a punctuated staining pattern scattered throughout the cytoplasm but excluded from the vac-uole and nucleus, as judged by Nomarski optics and in double staining experiments with DAPI (Fig. 4, D-F). This signal was absent in control cells (Fig. 4, A-C). The staining pattern is typical of the yeast Golgi apparatus (44); it is also seem for endopeptidase Kex1p, dipeptidyl aminotransferase A, Ca ϩ2 -ATPase Pmr1p, Mnn1p, Gda1p, and Kex2p (36,(45)(46)(47)(48)(49). Because the copy number of the 2-m plasmid can vary between 10 and 40 copies per cell within a population (50), the number of distinct fluorescent spots corresponding to Apy1p-HA also varies from cell to cell, ranging from a weak signal to a very strong signal. All of the cells expressing Apy1p-HA displayed a Golgi staining pattern (Fig. 4E); however, cells exhibiting a large amount of the protein, presumably because of a high dosage of the APY1-HA gene, also had a brightly stained vacuole (5-10% of cells of a pep4-3 strain, data not shown). Thus, cells producing high levels of Apy1p-HA showed mislocalization of the protein to the vacuole, which has also been reported for other yeast Golgi enzymes (36,48).
Membrane Arrangement of Apy1p-Hydrophobicity analysis using the Kyte and Doolittle algorithm (59) predicts two adjacent hydrophobic stretches in the polypeptide (amino acid residues 446 -467 and 501-517), which can potentially serve as transmembrane domains. The NH 2 -terminal 445-amino acid hydrophilic segment of Apy1p contains the apyrase domain. The apyrase domains of other E-ATPases are all located extracellularly, through either a cleaved (6,8,9) or a noncleaved signal sequence (5,14,17). Surprisingly, there is no signal sequence in the NH 2 -terminal portion of Apy1p; in addition, this domain (445 amino acids) is much larger than the extracellular domain of known type III membrane proteins (Ͻ100 amino acids) (51). This finding raises the question as to whether the apyrase domain is located in the cytoplasm or in the lumen of the Golgi. Both possibilities are problematic: the apyrase domain would not be expected to be in the cytoplasm because of its nucleotidase activity; on the other hand, the presence of the large domain in the lumen of the Golgi in the absence of a signal sequence would be unprecedented. Accordingly, we set out to determine the membrane topology of Apy1p.
To determine whether the NH 2 terminus might have an irregular cleavable signal sequence, Apy1p was tagged with a myc epitope at the NH 2 -terminal end and with a COOH-terminal HA tag. As is shown in Fig. 5A, Apy1p could be detected by both anti-HA and anti-myc antibodies, indicating that the NH 2 -terminal domain of Apy1p is intact. The results in Fig. 5B further support the conclusion that no NH 2 -terminal cleavage has occurred because addition of the myc tag composed of 10 amino acids to Apy1p-HA (myc-Apy1p-HA) increased the size of the protein by 1 kDa (Fig. 5B, lane 1).
To determine the location of the NH 2 -and COOH-terminal ends of Apy1p, a protease digestion protection assay was used. Golgi membrane fractions containing myc-Apy1p-HA were iso-lated and subjected to proteinase K digestion as described under "Materials and Methods." As is shown in Fig. 6A, proteinase K digestion reduced the size of Apy1p by 2 kDa, detected by anti-myc antibody (lane 2). However, no band was visible in the protease-treated sample examined with the anti-HA antibody (lane 6). These data indicate that the NH 2 terminus of Apy1p is protected from protease digestion, suggesting that it is in the Golgi lumen and that the COOHterminal end of Apy1p is not protected and is in the cytoplasm. The finding that protease digestion only removes 2 kDa from the COOH-terminal end of the protein suggests that most of the COOH-terminal hydrophilic domain is not available to the protease.
To demonstrate further that the NH 2 -terminal domain of Apy1p is in the lumen and the COOH-terminal end is in the cytosol, two glycosylation sites (Asn-371 and Asn-532) of Apy1p located on NH 2 -and COOH-terminal sides of the putative transmembrane domains were mutated to see if glycosylation of Apy1p would be affected. As is shown in Fig. 6B, the N371I mutation eliminated the formation of the 73-kDa glycosylated Apy1p (lanes 1-4), whereas the N532I mutation did not affect glycosylation (lanes 5-8). These results indicate that the NH 2terminal domain is indeed in the lumen, that glycosylation of Asn-371 contributes to the formation of the 73-kDa form of Apy1p, and that Asn-532 is not utilized for glycosylation, which is consistent with the cytoplasmic location of the COOH-terminal domain. Introduction of a new glycosylation site (D619N) into the COOH-terminal end of the Apy1p N371I mutant did not result in a glycosylated form of Apy1p (data not shown), supporting the cytoplasmic location of this domain.
The localization of the apyrase activity of Apy1p was investigated by measuring the ADPase activity of crude membranes. In the absence of additions, the nucleotidase activity is low; however, in the presence of digitonin, a nonionic detergent, and of alamethicin, a pore-forming antibiotic that allows ADP to traverse the membrane, the ADPase activity increases 2-and 4-fold, respectively (Fig. 7). This result also suggests that the active site of the enzyme is in the lumen of the Golgi. Triton X-100 appears to inhibit the enzymatic activity.
Because the NH 2 -and COOH-terminal domains of Apy1p are not on the same side of the membrane, these results also indicate that Apy1p has an odd number of transmembrane domains, probably one, whereas there are two potential hydrophobic stretches in Apy1p. To determine which hydrophobic FIG. 4. Subcellular localization of Apy1p by indirect immunofluorescence gives a punctate staining pattern typical of the yeast Golgi complex. As described under "Materials and Methods," transformants with pG 1 (panels A-C) or APY1-HA expressing plasmid pGZ103 (panels D-F) were fixed, solubilized, and treated with affinitypurified anti-HA monoclonal antibodies. The cells were then reacted with fluorescein isothiocyanate-coupled anti-mouse antibody; in addition, cells were treated with DAPI. The appearance of cells as detected by differential interference contrast microscopy (DIC, panels A and D), DAPI fluorescence (panels B and E), and indirect immunofluorescence for HA (panels C and F) is shown. stretch is the real transmembrane domain, two COOH-terminal truncated versions of Apy1p were constructed. The ⌬491-630 mutant lacked the COOH-terminal hydrophilic domain and the second hydrophobic stretch, and it had a HA tag at the end of the 33-amino acid hydrophilic loop connecting the two hydrophobic stretches. The ⌬552-630 mutant lacked the COOH-terminal 79 amino acids following the second hydrophobic segment. As is shown in Fig. 8A, ⌬491-630 Apy1p was not glycosylated (lanes 1 and 2), indicating that the NH 2 -terminal domain was not in the lumen. Fig. 8B shows that ⌬552-630 Apy1p was glycosylated (lanes 1 and 2), indicating that the NH 2 -terminal domain was in the lumen. These data support the view that the second hydrophobic stretch of Apy1p is the transmembrane domain.
Proposed Membrane Topology Model of Apy1p-Based on the results obtained above, the membrane topology of Apy1p is shown in Fig. 9. The intact NH 2 -terminal apyrase domain (amino acids 1-500) is located in the Golgi lumen with glycosylation of residue Asn-371. The COOH-terminal 113-amino acid hydrophilic domain is in the cytoplasm; since most of this domain is protected from protease digestion, it may bind to the membrane through its many positive charges. Furthermore, the second hydrophobic stretch (amino acids 501-517) of Apy1p is the likely transmembrane domain. DISCUSSION In this study, we have demonstrated that a putative yeast 71.9-kDa protein (Apy1p) is a new member of the E-ATPase family. Apy1p hydrolyzes various nucleotides, and its activity is highest with ADP. By membrane fractionation and immunofluorescence staining, Apy1p was found to be localized in the Golgi apparatus. Based on the results of protease protection assays and site-directed mutagenesis experiments, we have shown that Apy1p has a large NH 2 -terminal extracellular domain, a transmembrane domain, and a smaller COOH-terminal cytoplasmic domain; the protein does not have an explicit signal sequence.
It is interesting that there are two E-ATPases in yeast, and both are located in the Golgi. Yeast GDPase is involved in protein and lipid mannosylation. Its null mutant (gda1) is viable, but it has a partial block in O-and N-glycosylation of secreted proteins and a decrease in the transport of GDPmannose into the Golgi lumen (17,28). The APY1 null mutant (apy1) is viable and grows at approximately the same rate as the wild type at 30°C; the double deletion mutant (gda apy1) is also viable and grows approximately four times more slowly than the wild type strain. 2 Because the yeast GDP-mannose transporter null mutation is lethal, suggesting that glycosylation of protein and lipid in the Golgi is essential (52), it appears that the activities of Gda1p and Apy1p are not essential for protein and lipid glycosylation, probably because there is yet another way to hydrolyze nucleotide diphosphates. As the enzymatic properties of these two enzymes are different, further experiments are required to determine whether they have different biological functions.
Another possible function of Apy1p is in the conversion of ADP to AMP, which is then used as an antiporter in the translocation of ATP into the Golgi. It has been reported that an ATP transport activity is required for phosphorylation of proteoglycans and secretory proteins in the lumen of the Golgi apparatus and that the translocation of ATP into the Golgi lumen is coupled to the exit of AMP (27,53). This hypothesis is consistent with the view that the antiport molecules for sugar nucleotide transport into the Golgi are nucleoside monophosphates (27). In a recent paper (54) it has been suggested that a plant ectoapyrase may be involved in phosphate transport; it is possible that Gda1p and Apy1p are a link to transport of phosphates in the Golgi.
Apylp has a broad substrate specificity, whereas yeast GDPase (Gdp1p) is quite specific for GDP (17). The divalent cation preference of these enzymes is also different (17). The family of E-type ATPases embraces enzymes that are specific for ATP (55,56), for nucleotide di-and triphosphates (1), and nucleoside diphosphates (17). At present, the structural basis for the substrate specificity is not known and cannot be deduced by examination of the primary structures of the enzymes.
The most surprising finding in this study is that a large NH 2 -terminal apyrase domain is located in the Golgi lumen although it lacks a signal sequence for translocation through the membrane. Only a few membrane proteins with a long extracellular NH 2 terminus lacking a signal sequence are known. Some G protein-coupled receptors belong in this class, but their NH 2 -terminal segment is less than 100 residues; above this size limit, an NH 2 -terminal signal sequence is employed (51). It has been reported that Neu differentiation factor has a 240-amino acid extracellular NH 2 terminus and lacks a typical signal sequence, but the NH 2 -terminal 13 amino acids are cleaved, suggesting the existence of some kind of NH 2terminal signal (57). To our knowledge, Apy1p has the longest translocated NH 2 terminus reported so far. If translocation of the NH 2 terminus of Apy1p takes place by a new mechanism, Apy1p may be classified as a type V transmembrane protein.
We predict that hypothetical protein C33H5.14 of Caenorhabditis elegans, a candidate E-ATPase member (14), also belongs to this new subgroup because the primary sequence suggests that it has a large hydrophilic NH 2 -terminal apyrase domain lacking a signal sequence and a single hydrophobic segment.
The mechanism for the translocation of the NH 2 -terminal domain of Apy1p across the membrane should be different from those of type I and III membrane proteins. It has been shown that the orientation of the insertion of the transmembrane segments of membrane proteins correlates best with ⌬(C-N), the difference in charges within an arbitrary window of 15 residues flanking the transmembrane segment on either side (58). Because the end of the transmembrane segment with the most positive charges is retained in the cytoplasm, the value of ⌬(C-N) can be used to predict the orientation of the transmembrane segment. Similar to type III proteins, the value of ⌬(C-N) for Apy1p is positive in accordance with our localization of the NH 2 terminus and COOH terminus in the lumen of the Golgi and in the cytoplasm, respectively.
Compared with other E-ATPases, Apy1p has a long COOHterminal domain. Most of it may bind to the cytoplasmic face of the Golgi membrane, probably because of its highly positive charge. The functional significance of this binding is unknown; however, deletion of two-thirds of the COOH-terminal domain does not affect the membrane topology of APy1p. It is worthy of notice that a single tyrosine kinase phosphorylation site is present at the end of the COOH-terminal domain (622-629 KFKDSRLY). It will be interesting to find out whether this site can really be phosphorylated and whether there is any activity change when this site is mutated or deleted.
In summary, we have shown that a yeast hypothetical 71.9-kDa protein, Apy1p, possesses a cation-stimulated nucleotidase activity and is located mainly in the Golgi. We also demonstrate that Apy1p has a large translocated NH 2 terminus and an unusual membrane topology. Based on the distinct features of Apy1p, we propose that Apy1p and a hypothetical protein of C. elegans belong to new subgroup of transmembrane proteins.