Alternative splicing of human prostaglandin G / H synthase mRNA and evidence of differential regulation of the resulting transcripts by transforming growth factor beta 1 , interleukin 1 beta , and tumor necrosis factor alpha

Prostaglandin G/H synthase (PGG/HS) is the rate-limiting enzyme in the conversion of arachidonic acid to prostaglandins and thromboxanes. We screened a human lung fibroblast cDNA library with an ovine PGG/HS cDNA and isolated a 2.3-kilobase clone (HCO-T9). Sequence analysis of this clone showed that (a) it contained the entire translated region of PGG/HS and (b) it displayed an in-frame splicing of the last 111 base pairs encoded by exon 9, which resulted in the elimination of the N-glycosylation site at residue 409. Polymerase chain reaction amplification with specific oligonucleotides of reverse-transcribed mRNA from diverse human tissues and cultured cells yielded 400- and 300-base pair fragments that corresponded, respectively, to the intact and spliced transcripts. The expression of these two transcripts in cultured human lung fibroblasts was differentially regulated by serum, transforming growth factor beta 1, interleukin 1 beta, tumor necrosis factor alpha, and phorbol 12-myristate 13-acetate, as each of these conditions stimulated preferentially the expression of the unspliced transcripts. The elimination of one of the four N-glycosylation sites by the alternative splicing of exon 9 and the differential regulation of this process by relevant cytokines and growth factors may represent a mechanism for the regulation of PGG/HS enzymatic activity under physiological or pathological conditions.

PG G/H synthase (PGG/HS) (EC 1.14.99.1), a membranebound homodimer of two 70-kDa polypeptides. PGG/HS has a bisdioxygenase (cyclo-oxygenase) activity mediating the bisoxygenation of arachidonic acid to the hydroperoxide PGGZ, and a hydroperoxidase activity mediating the reduction of PGGz to the endoperoxide PGHz, which is the precursor of PG and thromboxanes (4,5). PGG/HS is a hemoprotein with a protoporphyrin IX prosthetic group which is required for both cyclo-oxygenase and hydroperoxidase activities. The enzyme contains four oligosaccharides bound to asparagine residues (6). At least one of these residues appears to play a crucial role in the activity of the enzyme.
The primary structure of PGG/HS from ovine vesicular glands (7)(8)(9) and murine 3T3 cells (10) has been deduced from full-length cDNA sequences. Similarly, the primary structure of the human enzyme has been deduced from sequences of a genomic clone (1 1). Northern hybridizations employing ovine or murine cDNAs have identified either 2.7-kb transcripts that correspond to the size of the cDNA (12,13) or transcripts of 2.7 and 5.5 kb (14)(15)(16) in a variety of cells and tissues from different species, including humans. In addition, hybridizations under low stringency conditions showed the presence of 4.0-kb transcripts in murine and ovine tissues, suggesting the existence of a related gene (17,18). More recently, the avian equivalent of this gene has been cloned (19).
PG have been shown to down-regulate the production of extracellular matrix proteins in fibroblasts by diverse mechanisms (3). PG production is stimulated by various cytokines, such as interleukin 18 (IL-l@), tumor necrosis factor a (TNFa), transforming growth factor B1 (TGF-P1), epidermal growth factor, and platelet-derived growth factor (20)(21)(22)(23)(24). Consequently, PG may modulate the effects that these cytokines have on extracellular matrix protein production (22). The availability of a cDNA for the human PGG/HS would allow a better understanding of the complex regulatory processes involved in PG production in various tissues under diverse conditions.
Here, we report the cloning of a cDNA for the human PGG/ HS and present evidence of an alternative splicing of the corresponding mRNA that eliminates 111 bp encoded by exon 9 of the gene. The alternative splicing would result in the elimination of one of the four functionally required N-glycosylation sites in the enzyme. Furthermore, we demonstrate the differential regulation of PGG/HS mRNA splicing by serum, TGF-P', IL-1P, TNF-a, and phorbol esters. These observations suggest that the alternative splicing we identified may play an important role in the regulation of the enzymatic activity of PGG/HS under normal or pathologic conditions.

MATERIALS AND METHODS
Cell Culture-Primary fibroblast cultures were established from human lung, dermis, and synovium and were cultured as described (25). Confluent fibroblasts were incubated for 24 h in Eagle's minimal essential medium (MEM) supplemented with 1, 5, or 10% FCS, 1% (v/v) vitamins, and 2 mM L-glutamine. After this period, the cells were incubated for different intervals in media containing either no serum or serum at the same concentrations as those received during the previous 24 h. The cultures were incubated under control conditions or with one of the following additives: 0.1 p~ phorbol 12-myristate 13-acetate (PMA), 200 units/ml TNF-a (Boehringer Mannheim), 50 units/ml IL-10 (Boehringer Mannheim) or various concentrations of TGF& (10-500 PM) (Collaborative Research Inc., Bedford, MA).
Isolation and Characterization of the cDNA-Single-stranded cDNA was prepared by reverse transcription of 5 pg of human lung fibroblast poly(A+) mRNA with Moloney leukemia virus reverse transcriptase (GIBCO-BRL) and oligo(dT)lo primers (30), and the second strand was synthesized using RNase H and DNA polymerase I (Boehringer Mannheim). EcoRI linkers (Promega Biotec, Madison, WI) were attached to the double-stranded cDNA, and it was ligated to an EcoRI-predigested X vector (X-ZAP II; Stratagene, La Jolla, CA), and packaged in uitro using a commercial packaging extract (Giga Pack I1 Plus; Stratagene, La Jolla, CA). A bacteriophage cDNA library was obtained containing approximately 4.5 X IO5 colonies and was screened by plaque hybridization with the gel-purified 1.6-kb ovine PGG/HS cDNA previously radiolabeled with [ ( U -~~P I~C T P by nick translation to a specific activity of 1.5 X 10* cmp/pg DNA. Hybridizations were carried out for 48 h at 42 "C in a buffer containing 6 X SSC, 5 X Denhardt's solution, 1% sodium dodecyl sulfate, and 100 pg/ml salmon sperm DNA. A positive clone was identified that was plaque-purified, and the Bluescript phagemide containing the cDNA was excised from the X vector with the helper virus R408. Progressive bidirectional deletions of the double-stranded cDNA were prepared with exonuclease 111 (Erase-a-Base System, Promega Biotec) and sequenced by the dideoxy chain termination method (31) using modified T7 polymerase (Sequenase, U. S. Biochemical Corp.).
Polymerase Chain Reaction (PCR) Amplification of PGG/HS Transcripts-Total RNA (5-10 pg) or poly(A+) mRNA (2 pg) from various human tissues and cultured cells was employed to synthesize singlestranded cDNA with Moloney leukemia virus reverse transcriptase and oligo(dT)lo primers. Amplification of the cDNA was performed by 40 cycles of PCR (32) in 50 pl of Taq DNA polymerase mixture (Perkin-Elmer Cetus Instruments) with 25 PM of two primers flanking a region from nucleotides 1011 to 1415. The reaction cycles were 94 "C for 30 s, 52 "C for 30 s, and 72 "C for 1 min. The 5"primer (5'-GGAGACCATCAAGATTGT) is complementary to the antisense strand, and the 3"primer (5"GTTCTCCAAACCGTACTT) is complementary to the sense strand. The PCR-amplified cDNAs were directly cloned into the pCR-1000 vector (Invitrogen, San Diego, CA) and sequenced by the dideoxy chain termination method.
In other experiments, 4 pg of total RNA from lung fibroblasts incubated under various conditions were reverse-transcribed using random hexamers and amplified by PCR in the presence of 25 pCi of [a-"'PIdCTP for 20 cycles. The reactions were chloroform-extracted and precipitated, resuspended in H20, and divided into two aliquots and analyzed by electrophoresis in 3% agarose and 5% acrylamide, respectively. In other reactions, 3 or 8 pg of total RNA were reversetranscribed and amplified by PCR in the presence of 50 pCi of [a-'"PIdCTP in a reaction volume of 200 pl, and 38-pl aliquots were taken at cycles 15, 18, 21, 24, 27. Each aliquot was divided and analyzed by electrophoresis in agarose and acrylamide gels as described above. Acrylamide gels were fixed in 7% acetic acid, dried, and subjected to autoradiography. The segments of the gels corresponding to the bands detected by autoradiography were excised, and the radioactivity was determined by scintillation counting. The counts/min obtained were divided by the cytidine content of both strands of the cDNA corresponding to the intact and spliced transcripts flanked by the oligonucleotide primers used for amplifications; that is cpm in 400 bp/210 and cpm in 300 bp/144. The resulting values were used for further calculations.

RESULTS
Cloning and Sequencing of a Full-length Human PGGIHS cDNA and Evidence of Alternative Splicing-Northern hybridizations of poly(A+) mRNA from normal human lung fibroblasts with the 1.6-kb ovine PGG/HS cDNA showed hybridization to 2.7-kb and 5.5-kb transcripts (Fig. 1, lane 1). A bacteriophage cDNA library was prepared from this mRNA and a positive clone containing a 2.3-kb cDNA, referred to as HCO-T9, was obtained. Northern hybridization analysis of normal human lung fibroblast poly(A+) mRNA with the HCO-T9 clone showed transcripts of 2.7 and 5.5 kb (Fig. 1, lane 2). A partial restriction map and the sequencing strategy for the HCO-T9 clone are shown in Fig. 2.
The nucleotide sequence of the HCO-T9 clone was determined and compared with the sequence of the human genomic clone for PGG/HS (ll), which encodes for the 2.7-kb transcript (Fig. 3). This analysis showed that the HCO-T9 clone has the following features.

T H G R W F W E F V B A T F I R E M L M R L V 1 1 5
E " 345 Expression of Intact and Spliced PGGIHS Transcripts in Various Tksues and Cultured Cells-To investigate whether transcripts corresponding to the intact and alternatively spliced forms of the PGG/HS gene were expressed in different human cells, PCR amplification of cDNA prepared from fetal and adult cartilage and from dermal, lung, and synovial fibroblast mRNA was performed. The ovine 1.6-kb (intact) and the human HCO-T9 (spliced) cDNAs were also amplified as controls. The 18-mer primers used were designed to bind regions of perfect homology between the human and ovine cDNAs. PCR amplification of the reverse transcribed mRNA from the five different human sources resulted in two fragments of approximately 400 and 300 bp in every case (Fig. 4). PCR amplification of mRNAs extracted from synovial tissue produced fragments of the same sizes (data not shown). PCR amplification of the ovine 1.6-kb cDNA yielded a single fragment of the expected 400 bp, whereas amplification of the HCO-T9 yielded a single fragment of 300 bp (Fig. 4). The 300-bp and 400-bp fragments amplified from the human mRNA and the 400-and 300-bp fragments amplified from the 1.6-kb ovine and the HCO-T9 cDNAs, respectively, were cloned in the pCR-1000 vector and sequenced. Analysis of the sequences showed that ( a ) the 400-bp fragments from both species were derived from transcripts corresponding to an intact exon 9 and (b) the human 300-bp fragment lacked the last 111 bp encoded by exon 9. We conclude that there was indeed an alternative splicing of the human PGG/HS mRNA a t a position corresponding to nucleotide 1185 of exon 9 of the gene.

Regulation of the Expression of Intact and Spliced PGGIHS Transcripts by Growth Factors and
Cytokines-The possibility that there may be differences in the enzymatic activity of the proteins translated from the intact and alternatively spliced transcripts led us to investigate the relative expression of these transcripts in normal human lung fibroblasts, under in vitro conditions known to up-regulate the expression of the gene (12,14,17,(34)(35)(36)(37)(38)(39). Fibroblasts incubated with increasing concentrations of TGF-PI showed a dose-dependent increase (up to 17-fold at 500 PM TGF-PI) in the steady-state levels of the 2.7-kb PGG/HS transcript (Fig. 5A). It was also noted that the 5.5-kb transcript described above was coordinately expressed with the 2.7-kb PGG/HS transcript. To determine the effects of the higher concentrations of TGF-B1 on the relative proportions of intact and spliced PGG/HS transcripts, total RNA from control and TGF-&-treated cells was reverse-transcribed and amplified by PCR. Aliquots taken at different cycles were electrophoresed in agarose and acryl-

FIG. 5. Northern hybridizations and PCR analysis of mRNA from normal human lung fibroblasts stimulated with TGF-8,.
A, confluent normal human lung fibroblasts were incubated in MEM (1% FCS) with various concentrations of TGF-6, for 24 h. Total RNA was isolated as described under "Materials and Methods" and 10-pg aliquots were analyzed by Northern hybridizations with the HCO-T9 clone (specific activity, 1 x 10" cpm/pg) and a murine glyceraldehyde-3-phosphate dehydrogenase cDNA (specific activity, 1 X 10" cpmlpg). Lane 1, control; lanes 2-6, TGF-Dl, 10,50,100,250, and 500 p~, respectively. B, 8 pg of total RNA from control and 500 PM TGF-&-treated cells was reverse-transcribed and amplified by PCR. Equal aliquots taken at different cycles were electrophoresed in 5% acrylamide and processed as described under "Materials and Methods." An autoradiograph of a 15-min exposure of the gel is shown. C, the radioactivity from gel slices corresponding to the autoradiographic bands was corrected for the dCTP content of the templates, as described under "Materials and Methods," and their logarithms were plotted against the number of cycles. The r values were calculated from cycles 15-21.0, 400-bp control ( r = 0.9984); W, 300-bp control ( r = 0.9990); 0, 400-bp TGF& ( r = 0.9965); 0, 300hp TGF-6, ( r = 0.9533). I), 4 pg of total RNA were reverse-transcribed and PCR-amplified for 20 cycles, and the samples were processed as described under "Materials and Methods." An autoradiograph of a 15-min exposure of the gel is shown. Lane 1, control; lanes 2-6, TGF-PI, 10, 50, 100, 250, and 500 pM, respectively. amide as described under "Materials and Methods." The autoradiographs of the acrylamide gel showed a progressive amplification of the intact (400 bp) and spliced (300 bp) PGG/ HS transcripts, with higher levels of both transcripts in TGF-PI-treated cells at all cycles of amplification examined (Fig.  5B). The relative extent of amplification of the intact and spliced transcripts showed a linear increase from cycles 15 to 21 (Fig. 5C). It was, therefore, possible to obtain an accurate estimate of the relative ratios of the 400-bp/300-bp PCRamplified products at cycle 18, the midpoint of the linear phase of the reaction. This ratio increased from 1.67 in control cells to 2.58 in TGF-&-treated cells. A similar analysis of cells treated with increasing concentrations of TGF-0, showed a dose-dependent increase in the levels of the intact and spliced transcripts (Fig. 50), with a more pronounced increase in the levels of the intact transcripts relative to those of the spliced transcripts (Table I).
TNF-a and IL-lP caused a 1.6-and 2.3-fold increase in the steady-state levels of the 2.7-kb PGG/HS transcript, respectively, and a coordinated increase on the 5.  Table I).
Treatment of fibroblasts with serum or PMA showed a time-dependent increase in the steady-state levels of the 2.7kb PGG/HS transcript that reached 18-and 10-fold higher levels over control cells, respectively, (Fig. 8A). In addition, there was a coordinate increase in the expression of the 5.5kb transcript and of a minor transcript of 4.4 kb. PCR analysis various treatments was reverse-transcribed and amplified by PCR in the presence of [n-'*P]dCTP, and the products were resolved in 5% acrylamide gels as described under "Materials and Methods." The radioactivity of the gel slices corresponding to the autoradiographic bands was divided by the cytidine content of the templates, and the ratios of the amplified cDNAs were calculated with the following equation. Ratio = (cpm in 400 bp/210)/ (cpm in 300 bp/144).

Experiment
Condition 400-bp 300-bp Ratio 400/300  A, confluent normal human lung fibroblasts were incubated in MEM (5% FCS) with or without 200 units/ml TNF-a for 24 h. Total RNA was isolated as described under "Materials and Methods", and 5-pg aliquots were analyzed by Northern hybridizations with the HCO-T9 clone (specific activity, 4.5 X 10' cpm/pg) and a murine glyceraldehyde-3-phosphate dehydrogenase cDNA (specific activity, 2.6 X lo8 cpm/pg). Lane I , control; lane 2, TNF-a. B, three pg of total RNA were reverse-transcribed and PCR-amplified. Equal aliquot taken a t a different cycle was electrophoresed on a 5% acrylamide gel and processed as described under "Materials and Methods." An autoradiograph of a 30-min exposure of the gel is shown. C, the radioactivity of gel slices corresponding to the autoradiographic bands was corrected for the dCTP content of the templates as described under "Materials and Methods", and their logarithms were plotted against the number of cycles. The r values were calculated from cycles 15- at 20 cycles showed a parallel increase in the intact and spliced transcripts (Fig. 8B). Estimation of the 400-bp/300bp ratios in serum-deprived cells showed an increase from 1.17 to 2.07 and 2.1 following 6 h of serum stimulation and 4 h of PMA stimulation, respectively (Table I).

DISCUSSION
We have been studying the role of PG as regulators of various human fibroblast functions and their participation in the modulation of the effects of several cytokines on collagen production by these cells (2,22). Cytokines such as IL-1P and TNF-a cause inhibition of collagen production, whereas TGF-PI is a potent stimulant. All three cytokines also stimulate under the treatments explored. Whether the 4.4-and 5.5-kb transcripts are from related genes or represent immature forms of PGG/HS mRNA remains to be determined. The presence of two pools of PGG/HS in human endothelial cells (40) and monocytes (41) and the identification of two antigenically distinct variants of PGG/HS localized to specific cell types in the rat ovary (42) suggest that the 5.5and 4.4kb transcripts may correspond to genes coding for different proteins with cyclooxygenase activity. The cloning of a murine PMA-induced 4-kb cDNA corresponding to a protein with a predicted amino acid sequence that is 80% homologous to PGG/HS supports this possibility (43).
The HCO-T9 clone we isolated is a 2.3-kb cDNA containing the entire translated coding region of the human PGG/HS, as compared with an ovine cDNA (7) and a human genomic clone (ll), except that it lacked a fragment corresponding to the last 111 bp encoded by exon 9. This observation suggested an alternative splicing of PGG/HS transcripts. PCR amplification of reverse-transcribed mRNA and the presence of a consensus 5"splice donor site at the area around position 1185 (33) confirmed this notion. Because the alternative splicing occurs in-frame, it is expected that it will be translated into an identical protein, except for the lack of residues The intact and spliced PGG/HS transcripts were found in mRNA extracted directly from tissues (synovium and cartilage) or from cultured cells (dermal, synovial, and lung fibroblasts), indicating that alternative splicing of PGG/HS mRNA is a naturally occurring phenomenon. Treatment of normal human lung fibroblasts with the inducers of PG production, TGF-PI, TNF-a, IL-lP, serum, and PMA, increased the steady-state levels of PGG/HS mRNA. PCR amplification of RNA from stimulated cells showed that each of these 396-432.
conditions increased the expression of the intact and spliced transcripts. However, all conditions caused a preferential stimulation of the intact transcripts (Table I).
PGG/HS is a membrane-bound glycosylated protein with bisdioxygenase and hydroperoxide activities that requires a heme group for each 70-kDa monomer for the expression of both catalytic activities. The bisdioxygenase and hydroperoxide activities reside on different domains of the protein (44,45). It has been proposed that the region responsible for the bisdioxygenase activity is located in the stretch encompassing residues 476-546. This region is highly conserved among different species and contains the active site SerS3' (10). It has also been shown that the bisdioxygenase activity of the enzyme requires a tyrosine residue that has been identified by site-directed mutagenesis as Tyr3'r' (46). The hydroperoxide active site appears to reside around His2"-Arg2"R residues (47). It has been postulated that the His3m and Hiszz6 residues are the axial and distal heme ligand sites, respectively (10). Substitution of Hisqo9 by site-directed mutagenesis completely abrogated both enzymatic activities of PGG/HS (48). The alternative splicing of the PGG/HS transcripts from exon 9 demonstrated here spares all these critical regions and residues.
The human PGG/HS has four potential glycosylation sites located at residues AsnIn3, and Am4@'. These are considered to be involved in maintaining the conformation of the tertiary structure, as well as in anchoring the enzyme to the cell membrane. Site-directed mutagenesis has shown that substitution of the Am4'' of the ovine enzyme completely abrogates its cyclooxygenase activity (48). The corresponding residue in the human enzyme is eliminated by the alternative splicing of PGG/HS transcripts from exon 9. The elimination of Am4@' is the most important alteration of the primary structure of the human enzyme that would result from the alternative splicing of the PGG/HS transcripts described here. The consequences of such alteration on the function of the translated protein are, at the moment, only speculative. However, the findings that the expression of nonspliced transcripts is preferentially stimulated by cytokines and growth factors known to increase PG production suggests that the alternative splicing of transcripts encoded by exon 9 may be a mechanism to limit the conversion of arachidonic acid to PG and thromboxanes at an early step of the biosynthetic pathway. Critical functional studies of the protein coded by the spliced transcripts are in progress.