Photolabeling of prostaglandin endoperoxide H synthase-1 with 3-trifluoro-3-(m-[125I]iodophenyl)diazirine as a probe of membrane association and the cyclooxygenase active site.

Previous studies of the crystal structure of the ovine prostaglandin endoperoxide H synthase-1 (PGHS-1)/S-flurbiprofen complex (Picot, D., Loll, P. J., and Garavito, R. M. (1994) Nature 367, 243-249) suggest that the enzyme is associated with membranes through a series of four amphipathic helices located between residues 70 and 117. We have used the photoactivatable, hydrophobic reagent 3-trifluoro-3-(m-[125I]iodophenyl)diazirine ([125I]TID) which partitions into membranes and other hydrophobic domains to determine which domains of microsomal PGHS-1 are subject to photolabeling. After incubation of ovine vesicular gland microsomes with [125I]TID, ovine PGHS-1 was one of the major photolabeled proteins. Proteolytic cleavage of labeled PGHS-1 at Arg277 with trypsin established that [125I]TID was incorporated into both the 33-kDa tryptic peptide containing the amino terminus and the 38-kDa tryptic peptide containing the carboxyl terminus. This pattern of photolabeling was not affected by the presence of 20 mM glutathione, indicating that the photolabeling observed for PGHS-1 was not due to the presence of [125I]TID in the aqueous phase. However, nonradioactive TID as well as two inhibitors, ibuprofen and sulindac sulfide, which bind the cyclooxygenase active site of PGHS-1, prevented the labeling of the 38-kDa carboxyl-terminal tryptic peptide. These results suggest that [125I]TID can label both the cyclooxygenase active site in the tryptic 38-kDa fragment and a membrane binding domain located in the 33-kDa fragment. Cleavage of photolabeled PGHS-1 with endoproteinase Lys-C yielded a peptide containing residues 25-166 which was labeled with [125I]TID. This peptide contains the putative membrane binding domain of ovine PGHS-1. Our results provide biochemical support for the concept developed from the crystal structure that PGHS-1 binds to membranes via four amphipathic helices located near the NH2 terminus of the protein.

The prostaglandin endoperoxide H synthase (PGHS) 1 isozymes catalyze the conversion of arachidonic acid to prostaglandin endoperoxide H 2 , the committed step in the biosynthetic pathway of prostaglandins and thromboxane (1). PGHS isozymes have two distinct activities: a cyclooxygenase activity, which catalyzes the oxygenation of arachidonic acid to yield PGG 2 , and a peroxidase activity, which reduces the 15-hydroperoxyl group of PGG 2 to form prostaglandin endoperoxide H 2 (2,3). Two isoforms of PGHS are known and are designated PGHS-1 and PGHS-2. PGHS-1 is the original "cyclooxygenase" characterized from ovine vesicular glands. PGHS-1 is expressed constitutively in most tissues (4). PGHS-2 is the more recently discovered isozyme whose expression is inducible and has been implicated in inflammation (4). PGHS isozymes have considerable homology in amino acid sequence and are thought to have similar structures (5)(6)(7)(8)(9). Nonsteroidal anti-inflammatory drugs (NSAIDs) such as aspirin, ibuprofen, and sulindac sulfide inhibit both isoforms of PGHS by binding the cyclooxygenase active site and blocking cyclooxygenase activity (10,11). The peroxidase activity of the isozymes is not appreciably affected by most NSAIDs consistent with the idea that PGHS isozymes have distinct cyclooxygenase and peroxidase active sites (12,13).
Both PGHS isozymes are glycoproteins (14) which are associated with the membranes of the endoplasmic reticulum (ER) and nuclear envelope (15,16). The isozymes are both inserted through the ER membrane via NH 2 -terminal signal peptides, which are cleaved in the mature enzymes (5)(6)(7)17). Both isozymes exhibit the characteristics of integral membrane proteins in that they are associated with microsomal membranes and require detergent for solubilization (2,3,17). Additionally, following purification, PGHS-1 can be reincorporated into liposomes, indicating that the protein can bind to lipid bilayers directly (18). However, unlike other integral membrane proteins, the PGHS isozymes do not appear to contain transmembrane domains. This prediction is based on analysis of the crystal structure of solubilized ovine PGHS-1 (19), and is supported by biochemical studies showing that six different evenly spaced regions of the isozymes are located in the lumen of the ER (20).
An interesting mechanism for the association of the PGHS isozymes has been proposed by Garavito and co-workers based on examination of the crystal structure of ovine PGHS-1 (19,21). Residues 70 -117 2 of ovine PGHS-1 contain four consecutive, short amphipathic ␣-helices which form a hydrophobic face along one side of the enzyme with the hydrophobic residues pointing away from the body of the enzyme. This region of the enzyme is proposed to be a membrane binding domain, associating with a single leaflet of the ER membrane but with-out fully traversing it. Should this prediction hold true, PGHS-1 will be the first monotopic membrane protein for which the structure has been determined. However, there is currently no biochemical evidence supporting this prediction.
3-Trifluoro-3-(m-[ 125 I]iodophenyl)diazirine ([ 125 I]TID) is a hydrophobic, photoactivatable reagent which partitions into lipid bilayers making this reagent a useful tool for determining regions of proteins which are membrane-associated (22)(23)(24). In order to identify regions of PGHS-1 which are associated with the lipid bilayer, we examined the photolabeling by [ 125 I]TID of PGHS-1 present in microsomal membranes prepared from ovine seminal vesicles. Our results indicate that [ 125 I]TID photolabels a region of PGHS-1 which contains the putative membrane binding domain. We also found that [ 125 I]TID can photolabel the cyclooxygenase active site of PGHS-1.

EXPERIMENTAL PROCEDURES
Materials-[ 125 I]TID and ECL Western blotting reagents were from Amersham Corp. Protein A-Sephacryl 4B, ibuprofen, Tween 20, trypsin, chicken egg white trypsin inhibitor, and reduced glutathione were from Sigma. Endoproteinase Lys-C was from Boehringer Mannheim. Goat anti-rabbit IgG-horseradish peroxidase was from Bio-Rad. Frozen ovine vesicular glands were obtained from Oxford Biomedical Research. Nonradioactive TID was the generous gift of Dr. Jonathan Cohen (Harvard Medical School). Sulindac sulfide was a gift from Merck.
Preparation of Microsomes from Ovine Vesicular Glands-Frozen ovine vesicular glands were sliced into thin strips with a razor blade and homogenized with a Polytron homogenizer in a HEPES buffer consisting of 20 mM HEPES, 20 mM glutamic acid, 2 mM magnesium acetate, and 200 mM sucrose, pH 7.5. The homogenate was then centrifuged at 10,000 ϫ g for 10 min to remove cell debris and mitochondria. The resulting supernatant was centrifuged at 200,000 ϫ g for 1 h (3), and the microsomal pellet was resuspended in HEPES buffer by homogenization with a Dounce homogenizer; this preparation was then cleared by centrifugation for 5 min at 10,000 ϫ g. The protein concentration of the cleared microsomal suspension was determined by the method of Bradford (25).
Labeling of Microsomal Membranes with [ 125 I]TID-Ovine vesicular gland microsomes were diluted to a final protein concentration of 3 mg/ml of HEPES buffer, and [ 125 I]TID was added at a final concentration of approximately 5 M. Following a 10-min incubation at 0°C, the microsomes were transferred to a 1-cm diameter plastic culture dish and were irradiated at a distance of 5-10 cm with a hand held UV illuminator (366 nm) for 10 min at 0°C (24).
In some experiments, microsomes were incubated with glutathione, ibuprofen, sulindac sulfide, or nonradioactive TID for 10 min at 0°C prior to the addition of [ 125 I]TID. Incubations with [ 125 I]TID and photolabeling were then performed using the procedure described above.
Trypsin Treatment and Immunoprecipitation of [ 125 I]TID-labeled PGHS-1-Following photolabeling of ovine vesicular gland microsomes with [ 125 I]TID, the membranes were collected by centrifugation at 200,000 ϫ g for 1 h and resuspended at a final protein concentration of 3 mg/ml. The resulting microsomal suspension was incubated with trypsin (10:1 microsomal protein:trypsin) for 15 min at 25°C. A 40-fold excess of trypsin inhibitor was then added, and the microsomal proteins were solubilized by addition of Tween 20 to a final concentration of 1%. The solution was sonicated for 30 s and centrifuged for 10 min at 10,000 ϫ g. The solubilized protein was incubated with a mixture of antipeptide antibodies against the amino and carboxyl termini of ovine PGHS-1 (20), and protein A-Sephacryl 4B was added. The immobilized protein A-antibody-PGHS-1 complex was collected by centrifugation at 1000 ϫ g for 1 min and was washed extensively with 0.1 M Tris-HCl, pH 7.4, containing 1% Tween 20 essentially as described previously (26). PGHS-1 was eluted from the complex by boiling the sample in SDS-PAGE-loading buffer. Samples were subjected to electrophoresis on 15% SDS-PAGE gels, and the proteins visualized by silver staining (27). 125 I-Labeled peptides were identified by autoradiography. Proteolytic Digestion of [ 125 I]TID-photolabeled PGHS-1 with Endoproteinase Lys-C-Following photolabeling, microsomes (600 g of microsomal protein) were solubilized by the addition of Tween 20 to a final concentration of 1%, followed by sonication and centrifugation at 10,000 ϫ g for 10 min. Solubilized PGHS-1 was immunoprecipitated using the mouse monoclonal antibody cyo-7 (26) coupled to Protein A-Sephacryl 4B. PGHS-1 was eluted from the Sephacryl beads by boiling in 0.5% SDS. The sample was diluted to 0.05% SDS in 10 mM Tris-HCl, pH 8.0, and 0.3 unit of endoproteinase Lys-C was added. Following incubation overnight at 37°C, the sample was lyophilized and resuspended in SDS-PAGE loading buffer. Proteolytic products were resolved on a 15% SDS-PAGE gel and transferred electrophoretically to nitrocellulose. Western blotting was performed on the filters, and the immunoreactive bands were visualized using enhanced chemiluminescence (ECL) detection as detailed previously (20). After waiting for several hours to allow the chemiluminescence to fade, [ 125 I]TID-labeled peptides were identified by autoradiography.
Following photolabeling of ovine vesicular gland microsomes with [ 125 I]TID, PGHS-1 was one of the most prominently radiolabeled proteins (Fig. 1). After tryptic digestion of photolabeled PGHS-1, both the 33-kDa amino-terminal and the 38-kDa carboxyl-terminal peptides were found to be labeled by [ 125 I]TID. This result was unexpected because the putative membrane binding domain of PGHS-1 is located near the NH 2 terminus, and it was anticipated that only the 33-kDa aminoterminal tryptic peptide would be labeled by [ 125 I]TID. To determine if the labeling of either of the tryptic peptides was due to the presence of [ 125 I]TID in the aqueous phase, photolabeling was performed in the presence of 20 mM reduced glutathione, which scavenges any [ 125 I]TID present in the aqueous phase (22). Reduced glutathione did not affect the photolabeling of either tryptic peptide of PGHS-1 (Fig. 1), indicating that the photolabeling was not due to the presence of  A, B, and C) or presence of 20 mM reduced glutathione (D, E, and F) and then incubated for 10 min at 0°C with [ 125 I]TID in low light. Finally, the microsomal samples were photolabeled by irradiation at 366 nm for 10 min at 0°C. Microsomes were then collected by centrifugation, resuspended, and solubilized, incubated in the absence or presence of trypsin, and immunoprecipitated with anti-PGHS-1 antibodies. Samples were resolved on 15% SDS-PAGE gels. The gels were subjected to silver staining (left panel) and exposed to film for autoradiography (right panel). Lanes A and D, microsomal membranes; lanes B and E, immunoprecipitated PGHS-1; lanes C and F, trypsin treated, immunoprecipitated PGHS-1.
[ 125 I]TID in the aqueous phase.

Competition for [ 125 I]TID Photolabeling of PGHS-1 by Nonradioactive TID and NSAIDs-[ 125 I]TID
can nonspecifically photolabel regions of proteins present in the hydrophobic environment of the membrane; in addition, [ 125 I]TID bound in a saturable manner to specific hydrophobic binding sites can photolabel regions of proteins which are not membrane-associated (24,31,32). To distinguish between these two possibilities, we examined the ability of nonradioactive TID to block photolabeling of PGHS-1 by [ 125 I]TID. When [ 125 I]TID labeling was performed in the presence of a 20-fold molar excess of nonradioactive TID, photolabeling of the 38-kDa carboxyl-terminal tryptic peptide of PGHS-1 was greatly diminished with little effect on the photolabeling of the 33-kDa amino-terminal tryptic peptide (Fig. 2). Silver staining established that the pretreatment of microsomes with nonradioactive TID did not affect the generation of 38-and 33-kDa tryptic fragments from PGHS-1 (data not shown). The photolabeling of the 33-kDa tryptic peptide in the presence of nonradioactive TID is consistent with the association of a region or regions of this peptide with membranes. The competition of the [ 125 I]TID photolabeling of the 38-kDa COOH-terminal tryptic peptide by nonradioactive TID suggests that the PGHS-1 protein contains a saturable, hydrophobic binding site(s) within the 38-kDa peptide which binds [ 125 I]TID in the absence of competitor. Thus, photolabeling of the larger 38-kDa carboxyl-terminal peptide is not consistent with the association of this peptide with the ER membrane.
The cyclooxygenase active site of PGHS-1 is a hydrophobic channel which contains segments of the peptide chains found in both the 33-kDa amino-terminal tryptic peptide and in the 38-kDa carboxyl-terminal tryptic peptide (19). We reasoned that the labeling of one or both of the tryptic peptides might result from the binding of [ 125 I]TID within the cyclooxygenase active site. A comparison of the structures of [ 125 I]TID and several NSAIDs suggest that a hydrophobic compound such as [ 125 I]TID could occupy the cyclooxygenase active site (Fig. 3). Accordingly, the ability of two NSAIDs, ibuprofen and sulindac sulfide, to compete for labeling of PGHS-1 by [ 125 I]TID was After photolabeling, the microsomal samples were treated with trypsin and immunoprecipitated as described in the text. Immunoprecipitated PGHS-1 and PGHS-1 tryptic fragments were resolved on 15% SDS-PAGE gels and the gels exposed to film for autoradiography. Lane A, no pretreatment; lane B, 100 M nonradioactive TID pretreatment.

FIG. 3. Structures of [ 125 I]TID and several nonsteroidal anti-inflammatory drugs.
PGHS-1 Membrane Binding examined (Fig. 4). Incubation of microsomes with ibuprofen or sulindac sulfide prior to photolabeling with [ 125 I]TID did affect the labeling pattern seen with ovine PGHS-1, decreasing somewhat the intensities of labeling of the 33-kDa tryptic peptide and essentially eliminating the labeling of the 38-kDa tryptic peptide. Pretreatment of PGHS-1 with the NSAIDs did not affect the generation of the 38-and 33-kDa peptides as determined by silver staining (data not shown). These results suggest that much of the photolabeling of PGHS-1 by [ 125 I]TID, particularly in the 38-kDa carboxyl-terminal tryptic peptide, is caused by the binding of [ 125 I]TID within the cyclooxygenase active site.
Photolabeling of Ovine PGHS-1 Occurs in a Region Containing the Proposed Membrane Binding Domain-Exhaustive digestion of [ 125 I]TID photolabeled ovine PGHS-1 with endoproteinase Lys-C generated a 20.5-kDa peptide which was reactive with an anti-peptide antibody raised against residues 25-35 of ovine PGHS-1 (Fig. 5); in contrast, this 20.5-kDa fragment was unreactive with an anti-peptide antibody raised against residues 203-217 of ovine PGHS-1 3 (data not shown). Importantly, this 20.5-kDa peptide was radiolabeled (Fig. 5). The molecular mass and immunoreactivity are completely consistent with this peptide being an endoproteinase Lys-C fragment, 25 Ala-Lys 166 , derived from the NH 2 terminus of ovine PGHS-1.

DISCUSSION
[ 125 I]TID has been a useful tool for identifying transmembrane domains of proteins (22)(23)(24). [ 125 I]TID partitions efficiently into membrane lipid bilayers, and photolabeling of proteins with [ 125 I]TID occurs predominately in domains which are in direct contact with membranes (22). There is, however, a precedent for the incorporation of [ 125 I]TID into regions of a protein which are not in contact with the lipid bilayer. Photolabeling of the nicotinic acetylcholine receptor with [ 125 I]TID could be partially inhibited by receptor agonists and antagonists apparently because these agents exclude [ 125 I]TID from the receptor binding site (31). In order to distinguish between the photolabeling of membrane-associated regions of the nicotinic acetylcholine receptor and photolabeling of the receptor binding site, [ 125 I]TID photolabeling was performed in the presence of excess non-radioactive TID (24,31,32). The reasoning behind these experiments was that nonradioactive TID would compete with [ 125 I]TID for saturable binding sites, whereas nonradioactive TID could not compete for photolabeling of membrane-associated domains because the membranes could not be saturated with TID at submillimolar levels (24,31,32).
We have found that [ 125 I]TID is a useful probe for analyzing the structure of ovine PGHS-1. The photolabeling of ovine PGHS-1 is similar to that of the nicotinic acetylcholine receptor in that PGHS-1 was photolabeled by [ 125 I]TID in both a nonspecific manner consistent with the association of regions of the enzyme with the ER membrane and in a specific manner consistent with the occupation of a hydrophobic pocket in the enzyme by [ 125 I]TID. Our data suggest that a region of ovine PGHS-1 contained in the NH 2 -terminal 277 amino acids is associated with the ER membrane and that the COOH-terminal half of the enzyme does not contain a membrane-associated region. Additionally, a peptide composed of residues 25-166 of ovine PGHS-1 is photolabeled by [ 125 I]TID, although we have not proven unequivocally that the labeling occurs specifically in the putative membrane binding domain encompassing residues 70 -117. Our data are consistent with the prediction that residues 70 -117 form a novel membrane binding domain (19,21). Examination of peptides photolabeled by [ 125 I]TID in the presence of nonradioactive TID and NSAIDs by amino acid sequencing will provide a valuable test of this prediction and should allow for the determination of which amino acids of PGHS are important for membrane association.
The ability of both nonradioactive TID, ibuprofen, and sulindac sulfide to block photolabeling of the COOH-terminal half of PGHS-1 by [ 125 I]TID suggests that [ 125 I]TID can photolabel the cyclooxygenase active site of PGHS-1. Thus, [ 125 I]TID may also be useful for identifying amino acids which are exposed in the hydrophobic cyclooxygenase pocket. Furthermore, because [ 125 I]TID labeling can be performed with membrane-associated enzyme, this labeling procedure may be used with the membrane-bound enzyme to test predictions made from the crystal PGHS-1 was immunoprecipitated from solubilized microsomes, denatured, and subjected to exhaustive digestion with endoproteinase Lys-C as described in the text. Samples were resolved on a 15% SDS-PAGE gel and transferred to a nitrocellulose filter. Western blotting (left panel) was performed on the filter using a rabbit anti-PGHS-1 antibody raised against a peptide corresponding to residues 25-35 of ovine PGHS-1 (20). After waiting several hours for the chemiluminescence to fade, the filter was exposed to film for autoradiography (right panel). Lane A, photolabeled microsomes; lane B, endoproteinase Lys-C digest of immunoprecipitated, photolabeled PGHS-1. structure of detergent-solubilized PGHS-1. Finally, [ 125 I]TID may prove to be a valuable tool for comparing structural differences between the cyclooxygenase active sites of PGHS-1 and -2. Recent studies have indicated that isozyme specific NSAIDs can be developed (10,11,33,34), and comparison of the labeling of the active sites may reveal some of the differences in the structures of the active sites of the isozymes which lead to NSAID selectivities (33)(34)(35).