Primary Structure of Human Thromboxane Synthase Determined from the cDNA Sequence*

Polymerase chain reaction techniques have been used to isolate a cDNA clone containing the entire protein coding region of thromboxane A2 synthase (EC 5.3.99.5) from a human lung cDNA library. The cDNA clone hybridizes with a single 2.1-kilobase mRNA species in phorbol ester-induced human erythroleukemia and monocytic leukemia cell lines. A second cDNA, differing only by an insert of 163 base pairs near the 3’-end of the translated region, was also found to be present in the same library. The proteins predicted from both nucleic acid sequences include the three polypeptide sequences determined from amino acid se- quencing of the purified human platelet enzyme, five potential sites for N-glycosylation, and a hydrophobic region that may serve to anchor the synthase in the endoplasmic reticulum membrane. The longer pre- dicted protein, designated thromboxane synthase-I, contains 534 amino acids, with a M, of 60,684, whereas the shorter protein, designated thromboxane synthase-11, contains 460 amino acids and has a M, of 52,408. Although thromboxane synthase-I1 lacks the conserved cysteine that serves as the proximal heme ligand in the other cytochromes, significant sequence similarities exist among thromboxane synthase-I and -11 and several P450s, particularly those in family 3. The overall amino acid identity is considerably less than 40%, making it likely that thromboxane synthase

$ On leave from the First Department of Internal Medicine, Tokyo Medical and Dental University, Tokyo 113, Japan.
The abbreviations used are: TXA, thromboxane A2; PCR, polymerase chain reaction; SDS, sodium dodecyl sulfate; SSC, standard saline citrate; PMA, phorbol 12-myristate 13-acetate; bp, base pair(s); kb, kilobase(s). monocyte-macrophages and is also produced in lung, kidney, spleen, brain, and gastrointestinal tracts (8)(9)(10)(11). Its biosynthesis involves arachidonate release, prostaglandin endoperoxide formation, and isomerization to TXA. The last step is catalyzed by TXA synthase. Besides forming thromboxane, purified TXA synthase also fragments prostaglandin H P into a hydroxyheptatrienoic acid and malondialdehyde; the ratio of products from the two pathways is about 1:l but is somewhat variable (11,12). The enzyme exhibits absorbance spectrum features characteristic of a cytochrome P450 but has no monooxygenase activity (11,13). During catalysis, TXA synthase undergoes a "suicide" inactivation (14) that may be a physiological mechanism for limiting TXA formation. The level of TXA in cells can be increased by erythroid differentiation factor, diacylglycerols and 1,25-dihydroxy vitamin DJ (15-17), Although the biosynthesis of TXA has been extensively studied, little is known regarding the chemistry of the enzyme due to the difficulty in obtaining a sufficient amount of active protein (11-13). We have recently reported the sequence of a cDNA encoding a part of TXA synthase (18). As a part of our investigation into the mechanisms of enzyme catalysis and into the control of expression of the gene, we report here the isolation of a full-length cDNA clone which contains the entire protein coding region of TXA synthase.

Cloning and Sequencing of T X A Synthase cDNA-We recently
obtained a -700-bp TXA synthase cDNA using the "nested primer" PCR technique (18). This cDNA was used as a probe to screen a Xgtll human lung cDNA library (Clontech, Palo Alto, CA). The probe was 32P-labeled by random primer and used to hybridize replicate filters containing about 2.5 X lo5 plaques under the following conditions: 6 X SSC (1 X SSC is 15 mM sodium citrate, pH 7.0, and 150 mM NaCl), 0.5% SDS, 2 X Denhardt's solution (50 X Denhardt's solution contains 1% each of Ficoll, bovine serum albumin, and polyvinylpyrrolidone), and 10 mM EDTA at 54 "C. The filters were washed with 2 X SSC and 0.5% SDS at 68 "C for 30 min, then with 0.5 X SSC and 0.5% SDS at 68 "C for 30 min. The EcoRI fragments of the inserts of the positive plaques were subcloned into the pGEM7-Zf (Promega, Madison, WI) or M13 mp18 and sequenced using Sequenase (U. S. Biochemical Corp.) and the dideoxy chain termination method (19). Compression regions were sequenced with inosine mix substituted for guanosine mix.

RNA Blot Anulysis-A human erythroleukemia cell line (HEL)
and a monocytic leukemia cell line (THP-1) were obtained from American Type Culture Collection, Rockville, MD. Both cell lines were induced to differentiate by incubating with culture medium containing 50 nM PMA at 37 "C for 48 h. Total RNA was prepared as described (20). Poly(A+) RNA was prepared by using oligo-dT column chromatography. The RNA blot analysis was performed as follows. RNA preparations were size-fractionated by electrophoresis in a 1.0% agarose gel containing formaldehyde, transferred to Zetaprobe membrane (Bio-Rad), and then hybridized with the 1.9-kb cDNA which had been labeled with 32P by random primers. Hybridization was performed at 54 "C in 1 X Church buffer (0.5 mM NaH2P04 789 Thromboxane Synthase cDNA pH 7.2, 1 mM EDTA, and 7% SDS; Ref. 21); the filter was washed at 68 "C in 1 X Church buffer for 1 h and then exposed for 16 h to Kodak (Rochester, NY) XAR-5 x-ray film. Reverse Transcription and Polymerase Chain Reaction-Four oli-g6nucleotides were synthesized for reverse transcription and PCR. To amplify the TXA synthase cDNA from the lung library, PCR was carried out as described above for 35 cycles using P3 and P4 as primers and 0.5 pg of cDNA as the template; each cycle consisted of 1 min at 94 "C, 1 min at 52 "C, and 1 min at 72 "C. The PCR product was digested with HindIII and BglII and then ligated into pBluescript I1 SK (Stratagene, La Jolla, CA) for sequencing.

RESULTS AND DISCUSSION
Our strategy for the isolation of the full-length cDNA clone involved screening of the human lung Xgtll cDNA library with the 700-bp probe isolated previously (18). The location of this fragment relative to the full-length cDNA is indicated as PCR in Fig. 1 (panel A ) . Screening of approximately 5 X lo5 plaques from this library led to the identification of two positive clones, designated TXS 2 and TXS 4 ( Fig. 1). TXS 4 contained a 1.0-kb insert, whereas TXS 2 yielded EcoRI fragments of 0.4 and 0.5 kb. The nucleotide sequences of these inserts were determined. The nucleotide sequence of the 3' portion of TXS 4 was identical with that of the overlapped 5' region of TXS 2. However, the coding region was incomplete a t the 3'-end. To obtain the full-length sequence, the library was rescreened by hybridization with the 3'-end EcoRI-fragment of TXS 2 (0.5 kb). Screening of 2.5 x lo5 plaques from the library yielded one additional positive clone (TXS 3), which had a 1.0-kb insert and included an additional 400 bp at the 3'-end of the previous sequence. The entire cDNA sequences of the three clones were determined by dideoxy nucleotide sequencing using "13 universal primer, T7 primer, SP6 primer, and unique synthetic oligonucleotides within the cDNA. The sequencing strategy is outlined by arrows ( Fig. 1, panel B ) . Of particular importance is the observation that no sequence differences were found in the extensive overlapping segments of TXS 2 and TXS 4, or in those of TXS 2 and TXS 3, indicating that all three cDNA originated from the same mRNA.
To demonstrate unequivocably that TXS 2, TXS 3, and TXS 4 were indeed from the same cDNA rather than from cross-hybridization with other members of the P450 superfamily (see below), we carried out reverse transcription-PCR using poly(A+) RNA from PMA-treated HEL cells as template and primers corresponding to the nucleotide sequences near the translational start and stop sites (see "Materials and Methods" section for details). A restriction enzyme digestion analysis of the resulting PCR product indicated that it was the same as the composite full-length lung cDNA except for the presence of an additional DNA fragment of about 160 bp between the HindIII and KpnI sites. A re-examination of the lung cDNA library by PCR, using primers corresponding to the sequences upstream of the HindIII site and downstream of the KpnI site (P3 and P4) gave roughly equal amounts of two PCR products, indicating that equal populations of two mRNA were present in the source of the lung library. One of these PCR products was about 370 bp long (as was expected from TXS 3), whereas the other was about 160 bp longer. The sequence of the longer fragment (530 bp) was determined, revealing a 163-bp segment (positions 1369-1531 in Fig. 2) flanked by sequences identical to those in TXS 3. While this manuscript was under review Yokoyama et al.
reported the isolation and sequencing of a TXA synthase clone from a human platelet library. Their sequence is almost exactly the same as that of the longer message we isolated from the human lung library. This confirms that the PCR product with the 163-bp insert is derived from the TXA synthase message itself rather than from a related cytochrome P450 message. It is not known whether the platelet library contains cDNA corresponding to that of the shorter message found in the lung library.
The total length of TXA synthase cDNA including the 163bp insert but not the poly(A) tail is 2067 bp. The cDNA contained a 1602-bp open reading frame encoding 534 amino acids, flanked by 355 bp of 5'-and 169 bp of 3"untranslated regions (Fig. 1). The cDNA lacking the 163-bp insert contained a 1380-bp open reading frame encoding 460 amino acids. The nucleotide sequences and deduced amino acid sequences are presented in Fig. 2; the longer protein is designated TXA synthase-I, the shorter one TXA synthase-11. More than 90% of the sequence was determined on both DNA strands. The published partial amino acid sequences obtained from the NHz terminus and two tryptic peptides of the purified human platelet enzyme, including a total of 83 amino acids (12), are consistent with the deduced amino acid sequences of both TXA synthase-I and -11 except for three residues: Asn + Gln at position 269, Asn -P Gln at position 270, and Glu -+ Asp at position 278. These discrepancies may be due to genetic polymorphism or to difficulties in interpretation of amino acid sequencer data. The agreement between the peptide sequences and amino acid sequences deduced from the cDNA confirms the correct reading frame and is strong evidence for the authenticity of the cDNA clone. The amino acid compositions of TXA synthase-I and -11 deduced from the cDNA sequences both agree reasonably well with that of the purified protein (13). The molecular weight of TXA synthase-I is calculated to be 60,684, whereas that of TXA synthase-I1 is 52,408.
The number of forms of TXA synthase, and their exact size, remains controversial. Two distinct species of TXA synthase, with molecular masses of 53 and 56 kDa, were detected in WI-38 human lung fibroblasts (ll), and molecular mass values of 50 and 59 kDa have been reported for the human platelet protein (11,13). Alternate splicing of the TXA synthase message might account for the presence of multiple forms of the synthase protein. A variable extent of glycosylation of the synthase might contribute to a variation in molecular weight. Five potential N-glycosylation sites (Asn-Xaa-Thr/Ser) are present at positions 56, 104, 107, 185, and 313.
The ATG site shown in Fig. 2 is almost certainly the start of translation because the following 25 amino acids are identical to the NH2-terminal sequence of the purified platelet enzyme. Furthermore, a termination codon sequence is located in frame at -63-position from the first methionine site. Also, the nucleotide sequence surrounding the first methionine corresponds well to the consensus sequence for a eukaryotic initiation site (23). Two methionines are observed at the NH, terminus, but the NHz-terminal sequence of the purified enzyme starts with the second methionine. It is likely that the first methionine residue is removed proteolytically at the translational stage, a process known to occur in both prokaryotes and eukaryotes (24); removal of the NH2-terminal methionine has been found for another eukaryotic P450 (25).
Although the 3"untranslated region of this clone did not contain the typical polyadenylation signal, AATAAA (26), another consensus sequence, CACTG, which also may be involved in the polyadenylation and/or cleavage of the mRNA at 3'-end (27), is observed at nucleotides 1691-1695.
Recently, we have obtained a genomic clone of the human TXA synthase which does have an AATAAA sequence 69 bp downstream from the CACTG sequence.' We have constructed a 1.9-kb cDNA containing the entire region covered by TXS 2, TXS 3, and TXS 4. When this cDNA was employed as a hybridization probe in RNA blot analysis, a single 2.1kb mRNA was identified in the PMA-treated HEL cells and THP-1 cells (Fig. 3). These results are consistent with the observation that monocytes/macrophages and activated erythroid cells produce TXA (15, 28). The size of our cDNA clone (1.9 kb without the 163-bp insert) is very close to that of the full-length cDNA (2.1 kb), if the poly(A) tail is included. Any additional sequence is most likely located at the 3'-end of the mRNA, because the AATAAA sequence is present 69 bp downstream from the CACTG sequence.
Comparison of the deduced sequence of TXA synthase with data bank sequences indicates that, as predicted from spectroscopic studies (12), the synthase has considerable sequence similarity with other P450s, particularly those in family 3. Fig. 4 presents detailed comparison of four regions of the synthase sequence with selected members of family 3, and with P450cam, the bacterial protein whose three dimensional structure has been determined (29). The sequence identity with the family 3 members is about 30% for Region 1, 40% for Region 2,50% for Regions 3 and 4, and about 35% overall for synthase-I and 30% overall for synthase-11. This level of sequence similarity is considerably less than that expected for an authentic member of family 3 (30), thus indicating that TXA synthase represents a new family in the P450 superfamily. The common NH2-terminal portion of synthase-I and synthase-I1 contains a strongly hydrophobic segment whose secondary structure is predicted to be helical (indicated by M in Fig. l(panel C)). As shown in Fig. 4, the sequence of this segment is similar to those of the NHz-terminal segments proposed to act as membrane anchor in eukaryotic P450s (31), and it may serve the same function for TXA synthase.
The heme prosthetic group and the residues that form the heme-binding pocket are important structural features in other P450s. Detailed structural studies in P450cam have revealed that helix I forms the hydrophobic backbone through the center of the molecule and contributes part of the hemebinding pocket (29). This helix is also believed to provide interactions with oxygen that are important to the catalytic function of P450cam (see Ref. 32 for a recent review). Two residues in helix I have been found to be conserved in all P450s besides TXA synthase: the glycine and threonine residues found at positions 250 and 253 in P450cam. Significantly, some of the greatest sequence similarity between TXA synthase and other P450s is found in the helix I area (Region 3 in Fig. 4); the conserved glycine is present in the synthase, but the threonine, which is believed to be involved in the oxygen-heme binding (29), is replaced by isoleucine.
The differences in the helix I region of the synthase thus need to be considered as possible contributors to the marked catalytic differences between the synthase and other P450 enzymes. Three other helical segments found in P450cam also are conserved (althollgh to a lesser extent) in TXA synthase-I and -11 (,Helices E, F, and J in Fig. 4).
The carboxyl-terminal region of other P450s contains a conserved cysteine residue thought to be the proximal ligand for the heme iron, and also helix L, considered to form part of the helix-heme-helix sandwich structure in other P450 enzymes (31). The carboxyl-terminal region is where the sequence of TXA synthase-I diverges from that of TXA synthase-11, due to the 163-bp insert at position 1369 in the cDNA sequence. Both the cysteine pocket (with the crucial heme ligand at position 480) and the helix L regions are present in synthase-I; both are absent in synthase-11. TXA synthase-I1 thus might be expected to be unable to form heme ligands in the usual P450 fashion. However, one of the 11 cysteines in the deduced synthase sequences is in a short segment with considerable similarity to thyroid peroxidase. The segments corresponding to helices E, F, I, J, and L, and the cysteine pocket, in P450cam, and the postulated transmembrane segment in the eukaryotic cytochromes P450, are indicated by horizontal arrows. Data on sequences other than TXA synthase are from GenBank. The accession numbers are as follows: IIIA4, human P450IIIA4 (A29815; also known as 3A4 or nifedipine oxidase); HFLA, human P450 HFLa (JX0062; also known as 3A7 or HFL33); GLUC, human hepatic glucocorticoid-inducible P450 (A29410; also known as HLp or 3A3); and CAM, Pseudomonas putida camphor hydroxylase (A25660).
The sequence in the synthase (beginning at position 413) is REAAQDCE, whereas that in the peroxidase (starting at position 908) is RAAAQDSE (33). The presence in the synthase of a cysteine in place of a serine makes this particular cysteine an interesting candidate for an alternative proximal heme ligand.
The evidence for the presence of two distinct TXA synthase messages in the lung tissue raises the possibility of alternate splicing of the mRNA, the observation of only one species of synthase message in the HEL cells by both reverse transcription-PCR and RNA blotting (Fig. 3), indicates that any alternate splicing occurs in a tissue-specific manner. The differences between the two predicted sequences in strongly conserved segments near the carboxyl terminus suggests they might differ in catalytic activity. It is noteworthy that the ratio of the alternate products of catalysis by TXA synthase,