Amino-terminal Topology of Thromboxane Synthase in the Endoplasmic Reticulum*

The membrane topology of the NH2-terminal portion of human thromboxane synthase (TXS), a member of the cytochrome P450 superfamily, has been investigated. By sequence alignment, the first 6 residues of the mature TXS polypeptide are likely to form a distinctive “tail” structure not found in many other mammalian cyto- chromes P450 in the endoplasmic reticulum membrane. Peptides with either the ultimate 10 or 15 residues of the NHz terminus of TXS were synthesized and used to pro- duce site-directed antibodies. The resulting peptide antibodies were highly specific and recognized human T X S , as shown by binding assays and Western blot anal- ysis. Binding of the peptide antibodies to recombinant TXS in transfected COS-1 and to endogenous TXS in THP-1 cells was analyzed by immunocytochemistry. Se-lective permeabilization of the plasma membrane to im-munoglobulin was achieved with streptolysin 0 , general permeabilization, including the endoplasmic reticulum membrane, was accomplished with Triton X-100. Perme- abilization of the plasma membrane was sufficient to produce binding of both peptide antibodies to their epitopes, indicating that the epitopes for both of the peptide antibodies were exposed on the cytoplasmic side of the endoplasmic reticulum membrane. The results with the peptide antibodies provide direct experi- mental evidence supporting the topological model for membrane-bound cytochrome P450


T X S , as shown by binding assays and Western blot anal-
ysis. Binding of the peptide antibodies to recombinant TXS in transfected COS-1 and to endogenous TXS in THP-1 cells was analyzed by immunocytochemistry. Selective permeabilization of the plasma membrane to immunoglobulin was achieved with streptolysin 0 , general permeabilization, including the endoplasmic reticulum membrane, was accomplished with Triton X-100. Permeabilization of the plasma membrane was sufficient to produce binding of both peptide antibodies to their epitopes, indicating that the epitopes for both of the peptide antibodies were exposed on the cytoplasmic side of the endoplasmic reticulum membrane. The results with the peptide antibodies provide direct experimental evidence supporting the topological model for membrane-bound cytochrome P450 proposed by Nelson and Strobel (Nelson, D. R., and Strobel, H. W. (1988) J.
BioL Chem. 263,6038-6050), in which the NHz terminus is oriented toward the cytoplasmic side of the endoplasmic reticulum membrane.
The cDNA for human lung TXS has been cloned and sequenced, and the amino acid sequence was deduced (Wang et al., 1991;Ohashi et al., 1992). Two cDNA clones of TXS were found in the lung library. The longer predicted protein, desig-* This work was supported in part by National Institute of Health Grant NS-23327. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "uduertisemnt" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The abbreviations used are: TXS, thromboxane synthase; ER, endoplasmic reticulum; PMA, phorbol 12-myristate 13-acetate; PBS, phosphate-buffered saline; FITC, fluorescein isothiocyanate; BSA, bovine serum albumin; STO, streptolysin 0. nated TXS-1, contains 534 amino acids and has an M, of 60,684, whereas the shorter protein, designated TXS-2, contains 460 amino acids and has an M, of 52,408. TXS-2 lacks the conserved cysteine that is believed to serve as the proximal heme ligand in other cytochromes P450 (Ohashi et al., 1992). The primary structure of the enzyme from human platelets has also been deduced recently (Yokoyama et al., 1991). The predicted human lung TXS-1 and human platelet TXS amino acid sequences are essentially identical. The amino acid sequence of TXS has considerable similarity to those of other cytochromes P450, particularly those in family 3 (Ohashi et al., 1992).
The transmembrane topology of mammalian membranebound cytochromes P450, in particular the orientation of the NH2 terminus, remains rather controversial. Current models for the membrane topology of microsomal cytochromes P450 propose a large cytoplasmic domain anchored to the membrane by either one or two NH2-terminal transmembrane segments (Black, 1992). Support for these models is derived from studies using site-specific antibodies and chemical modification to probe the exposure of particular segments of the enzyme in the endoplasmic reticulum (ER) membrane. The results indicated that the bulk of the protein is exposed on the cytoplasmic surface of the ER (Thomas et al., 1977;De Lemos-Chiarandini et al., 1987;Edwards et al., 1991). Nelson and Strobel (1988) conducted a n extensive analysis of hydropathy profiles and of previously reported results and proposed an NH2-terminal transmembrane hairpin loop as the sole membrane anchor, with the NH2 terminus of the polypeptide exposed to the cytoplasmic face of the ER. This proposed loop, involving parts of the first 66 amino acids of microsomal P450, has two transmembrane segments. The more widely accepted structural model has only one transmembrane segment in the NH2-terminal region and the N H 2 terminus oriented toward the lumen (Edwards et al., 1991).
As a member of the cytochrome P450 superfamily, thromboxane synthase shares considerable primary and secondary structure with other cytochromes P450 (Yokoyama et al., 1991;Ohashi et al., 1992). However, the initial 6 residues of the NH2 terminus of mature TXS appear to make up an additional "tail" segment that extends beyond the NH2 terminus of other microsomal cytochromes P450 (Ohashi et al., 1992). Assuming that all microsomal cytochromes P450 adopt a similar membrane topology, the additional tail segment of TXS in the ER membrane provides an important opportunity to determine the orientation of the NH2 terminus in microsomal cytochromes P450.
We have used two peptide antibodies directed against the NH2-terminal segment of human TXS to investigate the transmembrane topology of the NH2-terminal domain of this cytochrome P450 in the ER membrane. The results suggest that some of the first 10-15 residues of the NH2-terminal segment of TXS are exposed on the surface of the cytoplasmic side in the ER membranes, supporting the model proposed by Nelson and Strobel (1988).
EXPERIMENTAL PROCEDURES Structure Prediction-Hydropathy calculations were performed by the method of Kyte and Doolittle (1982). Analysis of sequence similarity and alignment and prediction of secondary structures were performed with the EuGene software package developed by C. B. Lawrence, T. Y.
Shalom, and S. Honda at the Molecular Biology Information Resource at Baylor College of Medicine.
Preparation of Human Platelet Microsomes-Human platelet microsomes were prepared essentially as described by Shen and Tai (1986a). Platelet-rich plasma (Gulf Coast Regional Blood Center, Houston, TX) was centrifuged at 200 x g for 30 min to remove leukocytes and red blood cells. Platelets were collected by centrifugation at 2000 x g for 15 min and washed once with one-third of the original volume of 25 m M Tris-HC1, pH 7.5, in 0.9% NaC1. The platelets were suspended in 3 volumes of same buffer at 4 "C, sonicated (5 x 10 s) with a Sonifier Cell Disruptor (Model W185, Heat Systems-Ultrasonics, Inc.), and then centrifuged at 8000 x g for 30 min. The supernatant liquid was filtered through a layer of glass wool and centrifuged at 100,000 x g for 60 min at 4 "C. The microsomal pellet was suspended in 25 m M Tris-HC1, pH 7.5, in 0.9% NaCl.
Construction of TXS Expression Vectors-'Ib construct a prokaryotic expression vector, two oligonucleotides corresponding to the sequences at the translation start and termination sites were used as primers to amplify the TXS-2 cDNA, which is identical to that of TXS-1 except at the COOH terminus. The primer sequences were 5"CGAGATCTG-GAATGATGGAAGCCTT-3' and 5'-GCAGATCTTGATATAGACACC-3'.
Both primers had the BglII sequence at the 5'-end. The amplified DNA was digested with BglII and subcloned in-frame at the BamHI site adjacent to the coding region for glutathione S-transferase in the pGEX-2T expression vector (Pharmacia LKB Biotechnology Inc.). The desired orientation was confirmed by restriction enzyme mapping. Induction of the fusion protein was performed in Escherichia coli cell as described previously (Smith and Johnson, 1988). For eukaryotic expression, a 2.1-kilobase TXS-1 cDNA containing the entire coding region, including the putative heme-binding cysteine, was first subcloned into the EcoRI site of pGEM7 (Promega Biotec). An XbaYXmaI fragment derived from this plasmid was then subcloned into the eukaryotic expression vector pSVL (Pharmacia) and used for transfection of COS-1 cells.
Cell Cultures-COS-1 cells (ATCC CRL-1650) were grown at 37 "C in a humidified 7% CO, atmosphere on coverslips in a 100-mm dish with Dulbecco's modified Eagle's medium containing 8% calf serum and 2% fetal calf serum to near confluency (-2.5 x lo6 cellddish) and transfected with 3 ml of Dulbecco's modified Eagle's medium containing 15 pg of DNA and 0.75 mg of DEAE-dextran at 37 "C. After 1 h of incubation, 7 ml of Dulbecco's modified Eagle's medium containing 10% fetal calf serum and 0.052 mg/ml chloroquine were added. Five hours later, the medium was changed to Dulbecco's modified Eagle's medium containing 2% fetal calf serum and 8% calf serum, and the cells were cultured for an additional 40 h (DeWitt et al., 1990).
An acute monocytic leukemia cell line, THP-1, was grown on polylysine-coated glass coverslips in a 100-mm dish in RPMI 1640 medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum, 2 m M glutamine, 50 m~ 2-mercaptoethanol, and 50 pg/ml gentamycin at 37 "C in a humidified 5% COz atmosphere. For induction of cell differentiation, the cells were treated with 20 n M PMA in the same medium for 4 days (Sanduja et al., 1991).
Assay for TXS Activity-Synthase activity was assayed by incubating 0.1 m M [14C]arachidonic acid (Du Pont-New England Nuclear) and 20 pg/ml purified sheep prostaglandin H synthase (Kulmacz and Lands, 1987) in 0.5 ml of 0.01 M sodium phosphate buffer, pH 7.2, containing 0.15 M NaCl (PBS) for 1 min and then adding 50 pl of microsomes from human platelets or from transfected COS-1 cells in PBS containing 40 p~ heme. Control incubations included a 700 p~ concentration of the TXS inhibitor U63557A (Biomol Research Laboratories Inc., Plymouth Meeting, PA). M e r 3 min of reaction, the lipid products were extracted with 3 volumes of ethyl ether:methanol:citric acid (30:41) and analyzed for thromboxane Bz by enzyme immunoassay .
Peptide Synthesis-'ho peptides corresponding to either the first 10 residues (NT-10) or the first 15 residues (NT-15) of the NHz-terminal segment of mature human TXS were synthesized by the solid-phase method of Merrifield (1963) using an automatic synthesizer (Model 430A, Applied Biosystems, Inc.) with phenylacetamidomethyl resin and t-butoxycarbonyl amino acids (Moore and Caprioli, 1991). After cleavage from the resin with hydrogen fluoride (Immuno-Dynamics, Inc., La Jolla, CA), the peptide was purified by high performance liquid chromatography using a Vydac C, column (1.0 x 25 cm) with a 30-min gradient of 0-70% acetonitrile in 0.1% trifluoroacetic acid at a flow rate of 2.0 mumin. Molecular masses of the purified peptides were verified by fast atom bombardment mass spectroscopy (Moore and Caprioli, 1991). Each peptide was coupled via carboxyl groups to amino groups of either bovine serum albumin (for NT-10) or ovalbumin (for NT-15) using the isochloroformate method (Erlanger et al., 1959) with a 50:l molar ratio of peptide to carrier. After coupling, unbound peptide was removed by dialysis against water. About 20 mol of each peptide were coupled to 1 mol of carrier protein.
Immunization and Characterization of Peptide Antibodies-Female New Zealand White rabbits (2 k g ) were immunized with 200 pg of peptide-carrier conjugate suspended in Freund's complete adjuvant. Boost injections of 200 pg of conjugate in Freund's incomplete adjuvant were given on days 14, 28, and 42. Blood was collected from the marginal ear vein starting 7 days after the final injection. Antibodies against the synthetic peptides were affinity-purified by the appropriate peptide immobilized on a Sepharose 4B column as described by the manufacturer (Pharmacia). Briefly, serum (20 ml) was loaded onto the peptide-Sepharose 4B column (0.9 x 3.5 cm), and the column was washed with PBS before specific antibody was eluted with 2.2 m M HCl (Ruan et al., 1985). The antibodies were purified further by passage through an ovalbumin-or bovine serum albumin-Sepharose 4B column (0.9 x 3.5 cm) to remove contaminating antibodies against the carrier protein.
Immunoblotting-Human platelet microsomal TXS or crude recombinant glutathione S-transferase-TXS-2 expressed in E. coli cells was solubilized with 1% SDS, mixed with sample buffer (60 m M Tris-HC1, pH 6.8, containing 10% glycerol, 2% SDS, 5% 2-mercaptoethanol, and 0.0013% bromphenol blue), and heated at 100 "C for 3 min. The proteins were separated by electrophoresis on a 10 or 13% polyacrylamide gel (Laemmli, 1970) and then transferred electrophoretically to nitrocellulose membranes (Towbin et al., 1979). After treatment with 1% nonfat powdered milk in PBS at room temperature for 1 h, the membranes were incubated with either antiserum or the affinity-purified anti-peptide antibody at room temperature overnight. After washing three times with PBS, the membranes were incubated with goat anti-rabbit IgG coupled to horseradish peroxidase (Bio-Rad) at room temperature for 3 h. The membranes were washed three times with PBS, and the bands were visualized with a mixture of 10 ml of 0.5% 4-chloro-1-naphthol in methanol and 50 ml of 0.03% HzOz in PBS.
Assay ofAntibody Binding-Binding of antibody to the corresponding synthetic peptides was assessed using microtitration plates (Corning, Corning, NY). The wells were coated with 50 pl of the appropriate peptide solution (30 pg/ml) in PBS at 4 "C overnight. After washing twice with PBS, nonspecific binding sites were blocked by incubation with 1% lysozyme in PBS at 37 "C for 1 h. Incubations with 50 p1 of antibody in PBS containing 1% lysozyme proceeded at room temperature for 2 h, and the plate was washed three times with PBS, followed by a 2-h incubation with 50 pl of goat anti-rabbit IgG-horseradish peroxidase conjugate in 1% 1ysozymelPBS. After washing four times with PBS, 50 p1 of a fresh mixture of equal volumes (1:l) of 3,3',5,5'tetramethylbenzidine solution and HZ02 solution (Kirkegaard & Perry Laboratories, Inc., Gaithersburg, MD) were added. The peroxidase reaction was stopped by addition of 200 pl of 2 N HZS04, and the reaction product was measured by its absorbance at 450 nm using an enzymelinked immunosorbent assay microplate reader (Dynatech Model MR 5000).
Immunocytochemical Studies-THP-1 cells or transfected COS-1 cells grown on coverslips were washed five times with 25 m M HEPES, pH 7.4, containing 2.5 m M magnesium acetate, 25 m M KC1, and 250 m M sucrose (HEPES buffer) and incubated at 0 "C for 10 min with 1 pg/ml streptolysin 0 (Sigma) preactivated with 4 m M dithiothreitol in HEPES buffer at 37 "C for 10 min. After excess toxin was removed by washing twice with HEPES buffer at 0 "C, the cells were incubated at 37 "C for 15 min, fued with 1% formaldehyde for 10 min, and then blocked with 1% myoglobin containing 5 mg/ml glycine in HEPES buffer for 20 min. Fixed cells were incubated with 75 pg/ml purified IgG or with antiserum (1:200 dilution) in HEPES buffer containing 1% myoglobin at room temperature for 1 h. After washing three times with HEPES buffer, the cells were incubated with goat anti-rabbit IgG-FITC conjugate (Sigma; 1:40 dilution) in HEPES buffer for 1 h, washed with the same buffer, and then mounted on glass slides with Mowiol (Calbiochem) containing 2.5% triethylenediamine (Sigma). For permeabilization of cells with Triton X-100, the cells were washed three times with PBS, fixed with 5% formaldehyde in PBS, blocked with 1% myoglobin, and then incubated with the primary and secondary antibodies in the presence of 0.5% Triton X-100 in PBS as described above. Cells stained with the FITC conjugate were examined by fluorescence microscopy (Campbell et al., 1992) using an Olympus microscope equipped for epifluorescence or by confocal microscopy on a Nikon Diaphot inverted microscope fitted with a Molecular Dynamics Multiprobe Model 2001 confocal laser scanning microscopy system. Excitation was with a laser at 488 nm.

RESULTS
Design of NH2-terminal Peptides-The NHz-terminal region of the other mammalian cytochromes P450 is predicted to be a membrane-spanning domain based on the hydropathy profiles and predicted secondary structures, and indeed all models of P450 topology have the NH2 terminus as some sort of transmembrane segment (Heinemann and Ozols, 1982;Heinemann and Johnson, 1985;Black and Coon, 1987;De Lemos-Chiarandini et al., 1987;Sakaguchi et al., 1987). Comparison of the deduced TXS sequence with data bank protein sequences indicates that the NHz-terminal region of mature TXS, beginning at about residue 7, has considerable similarity to other cytochromes P450, particularly those in family 3 (Fig. 1) (Ohashi et al., 1992). A dot matrix homology domain comparison of TXS-1 with a family 3 member, P450 IIIA4, indicated that -60% similarity (i.e. identical residues or conservative mutations) extended over 90% of the length of the TXS polypeptide. This strong overall similarity supports the validity of using TXS as ( A ) . Alignment Of Amino A c i d SEWenCES: A29815; IIIa5 (human hepatic P450 IIIA5), A34101; IIIa6 (rabbit P450 IIIAG), A34236; HFLA (human P450 HFLa), JX0062; GLUC (human hepatic glucocorticoid-inducible P450), A29410; and CN2 (rat pregnenolone 16a-carbonitrile-inducible P450), A25222. In A, amino acids that are identical (-) or similar (*) to those of TXS are indicated for the other cytochromes P450. The numbers to the lee are the position indexes of the first residues shown. The sequences of the NT-10 and NT-15 synthetic peptides used to raise the site-specific antibodies are underlined and numbered I and 2, respectively. In B , the letters H , S, C, and T indicate predicted a-helix, p-sheet, coil, and p-turn structures, respec-P450 IIIA4 were analyzed by the method of Kyte and Doolittle (1982) tively. In C , the hydropathies of the NHz-terminal regions of TXS and using a scan length of 7 amino acids. a model to investigate the NHz-terminal orientation of cytochromes P450. From a n alignment analysis, the first 6 residues or so at the NH2 terminus of mature TXS make up a segment that extends beyond those of the other membrane-bound mammalian cytochromes P450 (Fig. lA) (Nelson and Strobel, 1988).
To characterize the topology of this additional segment in the NHz terminus of mature TXS, peptides containing either the first 10 or 15 residues were synthesized, purified, characterized, and used to raise site-specific antibodies as described under "Experimental Procedures." Characterization of Site-specific Antibodies-Binding of the affinity-purified antibodies to the two synthetic peptides was evaluated by enzyme immunoassay as described under "Experimental Procedures." Each peptide antibody, even at low levels, bound to the corresponding NHz-terminal peptide and did not react with the carrier protein (BSA or ovalbumin) (Fig. 2 A ) . In contrast, preimmune IgG did not bind significantly to either peptide (Fig. 2B). The affinity-purified site-specific antibodies thus exhibited a high titer and specificity for recognition of their target NHz-terminal peptides.
The interactions of the NT-10 antibody with the NT-15 peptide-ovalbumin conjugate and of the NT-15 antibody with the NT-10 peptide-BSA conjugate were also examined. As expected, each affinity-purified antibody cross-reacted with the other NHz-terminal peptide-albumin conjugate (Fig. 3). This further confirmed the ability of the peptide antibodies to recognize the NHz-terminal epitope of TXS.
Binding of the site-specific peptide antibodies to TXS itself was examined with immunoblotting techniques. The affinitypurified antibody against the NT-10 peptide recognized a human platelet microsomal protein with a molecular mass of -61 kDa (Fig. 4 A , lane 11, which is very close to the reported molecular mass of human TXS-1 (Yokoyama et al., 1991;Ohashi et al., 1992). However, the purified NT-15 peptide antibody did not exhibit reactivity with the 61-kDa protein band (data not shown) even though it did bind the TXS NHz-terminal peptide coated on a microtiter plate (Fig. 2). This may indicate that recognition of TXS by the NT-15 antibody is dependent on a structural conformation that is lost upon denaturation of the protein during the immunoblot procedure. As would be expected, unpurified NT-10 antiserum recognized both TXS and albumin ( Fig. 4   significant binding to any platelet microsomal proteins ( Fig.  4 A , lane 3).
To further establish the specificity of the peptide antibodies for TXS, a recombinant fusion protein made up of TXS-2 and glutathione S-transferase was expressed in E. coli cells (Guan and Dixon, 1991) and was then analyzed by electrophoresis and immunoblotting. The fusion protein appeared on a Coomassie Blue-stained electrophoretic gel as a band of -76 kDa (Fig. 4B,  lane 1 ), which is consistent with its expected molecular mass of 77 kDa. The NT-10 antibody recognized the 76-kDa fusion protein on the nitrocellulose blot, and it also recognized a minor 64-kDa protein that may be a proteolytic product of the fusion protein (Fig. 4 B , lane 2). There was no significant binding of the NT-10 antibody to the proteins from control E. coli cells that were not expressing the TXS-2 fusion protein (data not shown). This confirmed that the 61-kDa platelet microsomal protein recognized by the NT-10 antibody was TXS.
Recombinant TXS-The TXS activity of microsomes from COS-1 cells transfected with a vector containing the cDNA of TXS-1 or with the vector alone was assayed as described under "Experimental Procedures." Microsomes from COS-1 cells transfected with TXS-1 cDNA had about seven times the TXS-specific activity compared with human platelet microsomes, and the activity was blocked by a TXS inhibitor (Table I). In contrast, microsomes from COS-1 cells transfected with a control plasmid had no detectable TXS activity (Table I). These results show that the recombinant TXS in the COS-1 cells was highly active and therefore that at least the catalytic region of the protein was likely to be in the native conformation.

Membrane lbpology of NHz-terminal Region of Recombinant
WS-The exposure of the N H 2 terminus of TXS to the cytoplasmic compartment was examined in intact transfected COS-1 cells. As schematized in Fig. 5, STO, a bacterial cytolytic protein (Campbell et al., 19921, was used to selectively permeabilize the plasma membrane to antibodies without perturbing the ER membrane, whereas Triton X-100 was used to permeabilize all cell membranes and to render antibodies accessible to the luminal side of the ER membrane. To confirm the selective permeabilization by STO, transfected COS-1 cells permeabilized with either ST0 or Triton X-100 were incubated with an antibody against protein disulfide-isomerase, a major ER luminal protein. The anti-protein disulfide-isomerase antibody was able to bind its target protein within the ER lumen only in Triton X-100-permeabilized cells and not in STO-permeabilized cells (Fig. 6). These results confirm the validity of using ST0 and Triton X-100 in differential permeabilization of the ER membrane to antibodies, as demonstrated for COS-1 cells by Campbell et al. (1992). Both the NT-10 and NT-15 peptide antibodies bound to Triton X-100-permeabilized COS-1 cells expressing recombinant TXS (Fig. 7, C and D ) but not to the control cells transfected with the vector alone (Fig. 7, E and F). Both antibodies also bound to STO-permeabilized transfected COS-1 cells (Fig. 7, A and B ) with about the same intensity as in cells permeabilized with Triton X-100. Transfected cells were not stained with preimmune IgG (data not shown). It should be noted that only -10% of the TXS cDNA-transfected cells were stained with the TXS peptide antibodies under either permeabilization condition, whereas all of the Triton X-100-permeabilized cells were stained with the protein disulfide-isomerase antibody. Thus, only a small fraction of the cells actively synthesized the recombinant protein in this transient expression system, as has been found previously (Campbell et al., 1992). These results provided clear evidence that the peptide antibodies were specific for native cellular TXS.
To confirm that the immunocytochemical results obtained with recombinant TXS were not artifacts associated with protein overexpression, similar experiments were performed on human THP-1 cells, in which the level of endogenous TXS is markedly increased after induction with PMA (Sanduja et al., 1991). THP-1 cells or PMA-induced THP-1 cells were permeabilized with ST0 or Triton X-100, incubated with the NT-10 or NT-15 antibody, and then stained with the FITC-labeled second antibody as described above for the COS-1 cells. Both peptide

TABLE I lXS activity in microsomes fiom platelets and COS-1 cells
COS-1 cells were transfected with plasmid DNA either with (pSvL TXS-1) or without (pSVL) the TXS-1 coding sequences as described under "Experimental Procedures." Each assay contained 100-120 pg of microsomal protein.

Cells
TXS activity"

+0.7 mM U63557A
Human platelets diluted rabbit anti-protein disulfide-isomerase serum and then stained with the FITC-labeled second antibody as described under "Experimental Procedures." All photomicrographs were taken with the same exposure times (20 s) and are printed at the same magnification (x -500).
antibodies gave the same staining in the PMA-induced THP-1 cells permeabilized with either ST0 (Fig. 8, A and B ) or Triton X-100 (C and D). The significant amount of antibody binding in STO-permeabilized cells indicates that NH2-terminal TXS epitopes recognized by the two antibodies were exposed to the cytoplasmic compartment. Much less intense fluorescence was observed with the STO-permeabilized quiescent THP-1 cells (Fig. 8, E and F ) , as expected from the lower levels of TXS activity (Sanduja et al., 1991). Thus, the endogenous TXS in the THP-1 cells had the same NH2-terminal topology as the recom-  15 ( B , D, and F ) antibody, the cells were stained with the FITC-labeled second antibody (see "Experimental Procedures" for details). Photomicrographs were obtained as described for Fig. 6. binant TXS transiently expressed in COS-1 cells. The diffuse staining of TXS in the THP-1 cells in Fig. 8 was thought to reflect the limited resolution of conventional light microscopy. Therefore, confocal microscopy was used to examine the sub-

Thromboxane Synthase
Amino-terminal Topology cellular localization of TXS in the THP-1 cells and revealed a distinctly reticular pattern of TXS staining (Fig. 9), consistent with an ER location, as found with other cytochromes P450.

DISCUSSION
The similarities in amino acid sequence between TXS and cytochromes P450 establish that TXS belongs to the cytochrome P450 superfamily (Yokoyama et al., 1991;Ohashi et al., 1992). The notable conservation of sequence motifs among the various cytochromes P450 (Nelson and Strobel, 1988) argues for a conservation of general elements of secondary and tertiary structure in the cytochromes P450. A number of alternative models for the membrane topology of mammalian cytochromes P450 have been proposed. The currently favored concepts have the protein anchored by its NHz terminus to the ER membrane (Nelson and Strobel, 1988;Edwards et al., 1991). This arrangement is consistent with the observation by Sakaguchi et al. (1987) that the NHz terminus acts as an uncleaved signal sequence, interacting with the signal recognition particle during insertion of cytochrome P450 in the membrane. Other evidence supporting the NHz-terminal region as the membrane anchor of cytochromes P450 has come from studies using site-specific antibodies (De Lemos-Chiarandini et al., 1987;Edwards et al., 1991), alignment of hydropathy profiles (Nelson and Strobel, 1988), and site-directed mutagenesis (Monier et al., 1988;Kemper and Szczesna-Skorupa, 1989).
De Lemos-Chiarandini et al. (1987) found that an antibody directed to the first 31 residues of native P450 IIBl bound strongly to the purified solubilized enzyme, but not to the microsomal enzyme, and that another antibody directed to residues 24-38 of the same P450 bound equally well to both soluble FIG. 9. Subcellular localization of TXS in PMA-induced THP-1 cells. PMA-induced THP-1 cells permeabilized with ST0 were reacted with the affinity-purified NT-15 antibody, stained with the FITC-labeled second antibody, and then analyzed by confocal microscopy using a 100 x oil objective. Photographs were taken directly from the Silicon Graphics workstation using Imagespace software and a Screenstar 35-mm camera system. and microsomal forms of the protein. Edwards et al. (1991) interpreted these results to imply the presence of a single membrane-spanning segment in the first 20 residues. The ability of FITC to react with the NHz-terminal residue of P450 in a mixture of the protein and liposomes (Bernhardt et al., 1983) was cited by Nelson and Strobel (1988) as evidence that the NHz-terminal residue was exposed to the aqueous environment, a key prediction of the hairpin model they proposed. However, incorporation of the P450 into the liposomal membranes was not demonstrated by Bernhardt et al. (1983), and thorough examinations by Vergeres et al. (1989Vergeres et al. ( ,1991 convincingly showed that the NHz-terminal residue could be labeled by FITC in soluble P450, but not in the microsomal enzyme or in the enzyme incorporated into liposomes. Still, the lack of labeling in the membrane-bound P450 does not rule out the presence of the NHz-terminal residue on the cytoplasmic side of the membrane. Recently, Larson et al. (1991aLarson et al. ( , 1991b expressed rabbit liver P450 IIEl in E. coli cells and found that the active enzyme was localized to the inner membrane even when residues 3-29 were deleted. Yabusaki et al. (1988) obtained similar results when residues 1-30 were removed from a P450 expressed in yeast. Although the nature of the interactions of the mutant cytochromes P450 with the membranes were not established, these results suggest that a single membrane-spanning a-helix at the NHz terminus might not be the only component of the membrane anchor in the protein. The first hydrophobic segment at the P450 NHz terminus has, however, been found to be sufficient to direct correct membrane insertion of test proteins in vitro (Sakaguchi et al., 1987;Vergeres et al., 1989). During revision of this manuscript, Szczesna-Skorupa and Kemper (1993) reported that a chimeric P450 with a 29-amino acid insert containing a potential N-glycosylation site appended to the NH2 terminus of P450 2C1 was in fact glycosylated. This result indicates that the NHz-terminal region of the chimeric P450 was oriented toward the ER lumen. However, the rather long peptide addition in the chimera may have disrupted the normal membrane insertion pattern or prevented reorientation of a part of the protein that is normally only transiently exposed to the ER lumen. The topological arrangement of the P450 NHz terminus in the ER membrane thus has remained controversial, although the single transmembrane segment anchor model is more generally accepted. Resolution of the question centers on obtaining unambiguous information about the disposition of the extreme NHzterminal segment in native cytochromes P450. One of the diffkulties with most mammalian cytochromes P450 is that most of the extreme NHz-terminal segment is hydrophobic, presumably leaving very few amino acid residues to protrude away from the membrane and thus limiting the usefulness of the immunological techniques that have proven so effective in analyzing the topology of the rest of the polypeptide. The NHz-terminal tail present in mature TXS gives this particular P450 additional antigenic epitopes at the NHz terminus, allowing direct immunological study. The tail is not especially long, extending only 6 residues beyond the end of family 3 cytochromes P450 (Fig. lA), and it is not particularly hydrophobic (Fig. IC). The hydropathy profile of the very hydrophobic segment following the extension at the NHz terminus of TXS coincides very well with the putative membrane anchor segment at the NHz terminus of P450 IIIA4 (Fig. 1C). In addition, a dot matrix homology domain comparison of TXS with P450 IIIa4 showed -60% similarity over 90% of the length of the protein sequences. All of this suggests that the presence of the NHz-terminal extension in TXS is not likely to give it an overall topology different from that in the other cytochromes P450, although such a perturbation cannot be ruled out.
' h o antibodies directed to the NHz-terminal segment of TXS were used in this study. One (NT-10) was directed to the ultimate 10 residues of mature TXS, and the other (NT-15) to the ultimate 15 residues. Both antibodies specifically recognized their target peptide sequences , and one (NT-10) recognized authentic denatured TXS on nitrocellulose membranes (Fig. 4). The specificity of the interaction of the two peptide antibodies with native TXS was demonstrated by the observation of their binding only to COS-1 cells expressing enzymatically active TXS, and not to COS-1 cells transfected with a vector lacking TXS cDNA (Fig. 7). These results confirm that the antibodies were indeed specific for the intended targets in the NHz-terminal segment of TXS.
The ability of the NT-15 antibody to bind the cellular TXS in spite of its inability to bind the denatured TXS on nitrocellulose membranes suggests that the epitope recognized by NT-15 requires the native secondary structure in the NHz-terminal segment. Residues 11-15 may contribute part of the NT-15 epitope or be required to preserve some element of secondary structure in residues 1-10, In any case, the binding of both NT-10 and NT-15 antibodies to TXS in STO-permeabilized cells indicates that at least some of the first 10 residues at the NHz terminus of TXS protrude from the membrane into the aqueous surroundings. The somewhat more intense staining with both antibodies seen after treatment of THP-1 cells with Triton X-100 ( Fig. 8 ) is consistent with increased access to the NHz terminus near the membrane surface upon lipid extraction by the detergents. Cytosolic exposure of the NHz-terminal epitopes was observed both with recombinant TXS expressed in COS-1 cells ( Fig. 7) and with endogenous TXS in THP-1 cells (Fig. 8). This offers some reassurance that the observed TXS membrane orientation was not peculiar to the recombinant protein. It remains possible that the mitogen-induced TXS in THP-1 cells has an NHz-terminal topology different from that in quiescent THP-1 cells or in other cells. Larson et al. (1991aLarson et al. ( , 1991b found that the NHz-terminal segment of P450 IIEl was not necessary for the oxygenase catalytic function of the cytochrome or for its interaction with the reductase or with cytochrome b5. Similarly, the extreme NH2 terminus of TXS does not appear necessary for catalytic activity because neither the NT-10 nor the NT-15 antibody inhibited formation of thromboxane B2 from the enzyme in platelet microsomes (data not shown).
Considerable experimental support has accumulated for the concept that mammalian ER cytochromes P450 are essentially globular proteins anchored to the ER membrane by a segment near their NHz termini (Black, 1992). The results of the present study with site-directed antibodies indicate that in one of these cytochromes P450, TXS, the NHz terminus itself is on the cytoplasmic side of the ER membrane. Given these topological constraints, the membrane anchor arrangements of the two models lead to opposite predictions about the orientation of the globular domain of TXS with respect to the ER membrane. The hairpin loop anchor model predicts that the globular domain of TXS is on the same side as the NHz terminus, in the cytoplasm, whereas the single transmembrane anchor model predicts that this domain is on the side opposite the NH2 terminus, in the ER lumen. Because a cytoplasmic orientation of the globular domain has been found for several other cytochromes P450 (Thomas et al., 1977;De Lemos-Chiarandini et al., 1987;Edwards et al., 1991), the simplest interpretation of the present data favors the hairpin anchor model. Some details of the hairpin anchor model are difficult to reconcile with current concepts of the structure of a-helical transmembrane polypeptide segments. Current sequence analysis algorithms predict only one full-length transmembrane a-helical segment at the TXS NH2 terminus (roughly residues 15-32) (Fig. 1). The portion ofTXS that would correspond to the second leg of the hairpin proposed by Nelson and Strobel (1988) is strongly conserved with other cytochromes P450, particularly residues 42-57 (Fig. 11, consistent with some functional importance. However, there are 3 conserved proline residues in this segment of TXS, and such an accumulation of proline residues is unprecedented in the limited number of verified transmembrane polypeptides examined so far. This suggests that the hairpin does not completely traverse the membrane or that the second leg of the hairpin has an unconventional structure. Despite the extensive similarities between TXS and other cytochromes P450 (Fig. 11, it remains possible that TXS has a membrane anchor arrangement different from those in other cytochromes P450. The NH2 terminus of TXS is different from that of other cytochromes P450 and could contain unsuspected signals for a distinct membrane insertion process. In this regard, alteration of just 2 amino acid residues in P450 IIC2 was reported to alter the subcellular location of that protein (Kemper and Szczesna-Skorupa, 1989). Further studies will be needed to determine just how homogeneous the membrane topologies of the ER cytochromes P450 are. In any case, this study defines the membrane topology of the NHz terminus of TXS and sets the stage for detailed studies of the arrangement of the rest of the protein and of the influence of TXS membrane topology on the physiological function of this important protein.