Molecular cloning, expression, and partial characterization of two novel members of the ovalbumin family of serine proteinase inhibitors.

A human placental lambda gt11 cDNA library was screened for sequences encoding proteins related to human proteinase inhibitor 6 (PI6), and two plaques were identified that displayed weak hybridization at high stringency. Isolation and characterization of the DNA inserts revealed two novel sequences encoding proteins composed of 376 and 374 amino acids with predicted molecular masses of approximately 42 kDa. The novel proteins displayed all of the structural features unique to the ovalbumin family of intracellular serpins including the apparent absence of a cleavable N-terminal signal sequence. The degree of amino acid sequence identity between the novel serpins and PI6 (63-68%) significantly exceeds that of any other combination of known intracellular serpins. The two novel serpins encoded by the two novel cDNA sequences have been designated as proteinase inhibitor 8 (PI8) and proteinase inhibitor 9 (PI9). The putative reactive center P1-P1' residues for PI8 and PI9 were identified as Arg339-Cys340 and Glu340-Cys341, respectively. PI9 appears to be unique in that it is the first human serpin identified with an acidic residue in the reactive center P1 position. In addition, the reactive center loop of PI9 exhibits 54% identity with residues found in the reactive center loop of the cowpox virus CrmA serpin. Two PI8 transcripts of 1.4 kilobases (kb) and 3.8 kb were detected by Northern analysis in equal and greatest abundance in liver and lung, while the 1.4-kb mRNA was in excess over the 3.8-kb mRNA in skeletal muscle and heart. Two PI9 transcripts of 3.4 and 4.4 kb were detected in equal and greatest abundance in lung and placenta and were weakly detected in all other tissues. PI8 and PI9 were expressed in baby hamster kidney and yeast cells, respectively. Immunoblot analyses using rabbit anti-PI6 IgG indicated the presence of PI8 in the cytosolic fraction of stably transfected cells that formed an SDS-stable 67-kDa complex with human thrombin. PI9 was purified to homogeneity from the yeast cell lysate by a combination of heparin-agarose chromatography and Mono Q fast protein liquid chromatography and migrated as a single band in SDS-polyacrylamide gel electrophoresis with an apparent molecular mass of 42 kDa. Purified recombinant PI9 failed to inhibit the amidolytic activities of trypsin, papain, thrombin, or Staphylococcus aureus endoproteinase Glu-C and did not form an SDS-stable complex when incubated with thrombin. The cognate intracellular proteinases that interact with PI8 and PI9 are unknown.

The mammalian serine proteinase inhibitors, or serpins, are a superfamily of single chain proteins that typically range between 40 and 60 kDa in molecular mass, resemble ␣ 1 -proteinase inhibitor in overall structure, and include antithrombin III, plasminogen activator inhibitors 1 and 2, ␣ 1 -antichymotrypsin, and ␣ 2 -antiplasmin (1). The majority of serpins participate in the regulation of several proteinase-activated physiological processes including blood coagulation, fibrinolysis, complement activation, extracellular matrix turnover, cell migration, and prohormone activation, to name a few (2). Serpins inhibit proteolytic events by forming a 1:1 stoichiometric complex with the active site of their cognate proteinases, which is resistant to denaturants (3).
In addition to the mammalian serpins, several viral serpins have been identified and implicated as virulent factors. These serpins include the SERP-1 gene product produced by tumorigenic myxoma virus (4) and the CrmA serpin produced by the cowpox virus (5). Insight into the mode of action of these viral serpins has been derived from the findings that the SERP-1 gene product inhibits the serine proteinases of the fibrinolytic system and also inhibits C1 esterase, the first enzyme in the complement cascade (4). Recently, CrmA was found to attenuate the host inflammatory response by acting as a specific inhibitor of the interleukin-1␤ converting enzyme (ICE), 1 a novel cytosolic cysteine proteinase (5). These findings are significant since they suggest that the proteinase specificity of mammalian serpins may extend beyond serine proteinases and include some cysteine proteinases. In this regard, an ␣ 1 -antichymotrypsin-like serpin was recently purified from bovine chromaffin granules of adrenal medulla and found to inhibit a novel cysteine proteinase involved in enkephalin precursor processing (6).
Aside from the serpins that regulate proteinase activity, several members of this superfamily lack a proteinase inhibitory capability and have other physiological roles. These latter serpins were originally identified by data base searching and include thyroxine-binding globulin (7), angiotensinogen (8), and ovalbumin (9). Ovalbumin represents the parent prototype of a unique family of serpins, within the serpin superfamily, that lack a typical amino-terminal cleavable signal sequence but have been found to reside intracellularly, extracellularly, or both (10). Therefore, members of the ovalbumin serpin family may function as dualistic molecules with an intracellular and/or extracellular function. The serpins previously classified as members of the ovalbumin family are plasminogen activator inhibitor-2 (PAI-2) (11), an elastase inhibitor (EI) isolated from monocyte-like cells (12,13), a squamous cell carcinoma antigen (SCCA) (14), maspin (15), and a novel cytoplasmic serpin isolated in this laboratory and designated as cytoplasmic antiproteinase (16,17). A functionally inactive form of cytoplasmic antiproteinase was purified earlier from human placenta and designated as placental thrombin inhibitor or PTI (18). Recently, the Genome Database organization has recommended that this serpin be designated as proteinase inhibitor 6, or PI6 (19). Purified PI6 inhibits the amidolytic activities of a broad spectrum of trypsin-like serine proteinases including thrombin, trypsin, urokinase, and factor Xa (16). In addition, trypsin inhibition by PI6 appears to involve a two-step mechanism that results in the formation of a tight inhibitory complex that is pseudoreversible and behaves similar to the proteinase inhibitory complex formed with ␣ 2 -antiplasmin (17). In the present study, we report the molecular cloning, expression, and partial characterization of two novel human PI6 homologs that are divergent within their reactive centers and are predicted to inhibit distinct proteinases. In addition, the reactive center loop of one novel PI6 homolog exhibits a high degree of structural similarity to the serpin encoded by the crmA gene carried by the cowpox virus. These two novel PI6 homologs have been designated as proteinase inhibitor 8 (PI8) and proteinase inhibitor 9 (PI9) as recommended by the Genome Database nomenclature committee.

EXPERIMENTAL PROCEDURES
Materials-Penicillin-streptomycin-neomycin, methotrexate, ampicillin, dithiothreitol, and N ␣ -benzoyl-DL-arginine p-nitroanilide were obtained from Sigma. D-Val-Leu-Lys-p-nitroanilide and Phe-Pip-Arg-pnitroanilide were obtained from Helena Laboratories. Carbobenzoxy-L-Phe-Leu-Glu-p-nitroanilide was a product of Boehringer Mannheim. Prestained SDS-PAGE standards (low range) were from Bio-Rad. Nitrocellulose membranes were purchased from Schleicher and Schuell. Tissue culture T-75 flasks were obtained from Corning. Dulbecco's modified Eagle's medium was a product of Mediatech. Fetal bovine serum was obtained from Hyclone Laboratories. 125 I-labeled protein A was purchased from DuPont NEN. Epicurian Coli TM XL-1 Blue supercompetent cells were from Stratagene. Amylose resin, pMAL-c2 vector, anti-MBP antiserum, XmnI, EcoRI, and NcoI were products of New England Biolabs. Papain and cystatin were purchased from Calbiochem. Isopropyl-1-thio-␤-D-galactopyranoside was obtained from U. S. Biochemical Corp. Protein A-Sepharose and the SDS low M r standard kit were purchased from Pharmacia Biotech Inc. L-1-Tosylamido-2phenylethyl chloromethyl ketone-treated trypsin and Staphylococcus aureus endoproteinase Glu-C were obtained from Worthington Biochemical Corp. All other reagents were the highest grade commercially available.
Construction, Expression, and Purification of a Maltose-binding Protein/PI6 Fusion Protein-PI6 cDNA was cleaved with NcoI to generate a 1.8-kilobase DNA fragment that was purified by gel electrophoresis (20). This fragment, encoding PI6 residues 18 -376, was ligated into vector pMAL-c2 that had previously been cleaved with XmnI and EcoRI, and subsequently treated with calf intestinal phosphatase prior to ligation (21). Epicurian Coli TM XL-1 Blue supercompetent cells (New England Biolabs) were transformed with ϳ50 ng of pMAL-c2/PI6 DNA essentially as outlined by the manufacturer, and 5 l of the transformation mixture was plated on Luria-Bertani (LB)/ampicillin plates. Individual colonies were grown in LB/ampicillin medium and examined for fusion protein synthesis by SDS-PAGE (22) following induction with 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside. Cells containing plasmids that coded for a MBP/PI-6 fusion protein were grown to 5 ϫ 10 8 cells/ml at 37°C with shaking in LB/glucose/ampicillin for 5 h. Isopropyl-1-thio-␤-D-galactopyranoside was then added to a final concentration of 0.1 mM, and the culture was incubated with vigorous shaking at 22°C for an additional 24 h. The cells were harvested by low speed centrifugation, resuspended in 20 mM Hepes (pH 7.5), containing 200 mM NaCl,1 mM EDTA, 6 mM mercaptoethanol, 5 mM phenylmethylsulfonyl fluoride and 0.02% NaN 3 (column buffer), and lysed by sonication. Cell debris was removed by high speed centrifugation, and the supernatant was applied directly to a column of amylose resin (1.6 ϫ 8 cm) previously equilibrated at 4°C with column buffer. Following sample application and wash, the MBP/PI6 fusion protein was eluted from the affinity resin in column buffer containing 10 mM maltose. Examination of the maltose eluent by SDS-PAGE indicated a prominent 78-kDa protein that was immunoreactive to commercially available rabbit anti-MBP IgG. Antibodies against the MBP/PI6 fusion protein were generated in rabbits (23), and the IgG fraction was purified by protein A-Sepharose column chromatography. This IgG fraction was further purified by passage through a MBP-Affi-Gel 15 column to remove anti-MBP IgG.
Cell Expression-The full-length cDNA for PI8 was directionally cloned into expression vector Zem229R and expressed in baby hamster kidney (BHK) cells (24). BHK cells transfected with "empty" Zem229R expression plasmid (PI8-free) served as control cells in this study. BHK/PI8 and BHK/Zem 229R cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, penicillinstreptomycin-neomycin, and 1 M methotrexate. PI9 was expressed in yeast as follows. An expression plasmid, PI9/pDPOT, was created by cloning the cDNA for PI9, along with DNA fragments for the AdH24c promoter and the TP11 terminator, into the yeast shuttle plasmid pDPOT. Spheroplasts from the S. cerevisiae yeast strain ZM118 were transformed separately with either PI9/pDPOT or pDPOT (PI9-free) and selected for growth on glucose medium. Transformants were grown in 1% peptone, 1% yeast extract, 1.5% glucose, and 1.5% ethanol at 30°C for 3 days (25).
Preparation of BHK Cell Cytosol-The cytosolic fractions of BHK/PI8 and BHK/Zem 229R cells were prepared essentially as described (16). Briefly, each cell line in a T-75 flask was dislodged from the flask by brief trypsinization followed by centrifugation (1500 rpm; 7 min; 25°C). The cells (ϳ10 6 cells) were resuspended in 5 ml of 20 mM Hepes (pH 7.5) containing 150 mM NaCl, recentrifuged, and resuspended in 1 ml of 20 mM Hepes (pH 7.5) containing 1 mM DTT. Cells were then ruptured by six sequential passages through a 25-gauge 1.58-cm needle, and the needle effluent was collected directly into thick walled polycarbonate ultracentrifuge tubes placed on ice. The cytosolic fractions were harvested following centrifugation of the cell lysate (100,000 ϫ g; 30 min; 4°C) in a Beckman ultracentrifuge (model L5-75) and subsequently reconstituted with 0.15 M NaCl.
Purification of Recombinant PI9 -Yeast cells transformed with the PI9/pDPOT were suspended in lysis buffer (10 mM Tris-HCl (pH 7.4) containing 2 mM EDTA, 50 mM NaCl, 43 g/ml phenylmethylsulfonyl fluoride, 2 g/ml pepstatin, and 1 g/ml leupeptin) and lysed using glass beads and vortexing. The lysate was centrifuged, and the supernatant was dialyzed at 4°C against 50 mM Hepes (pH 7) containing 1 mM DTT. The retentate (ϳ30 ml) was applied to a column (1.6 ϫ 20 cm) of heparin-agarose equilibrated at 4°C with 50 mM Hepes (pH 7) containing 1 mM DTT at a flow rate of 0.5 ml/min. The column was washed with equilibrating buffer, and PI9 eluted from the column in a linear NaCl gradient (0 -0.5 M) dissolved in equilibrating buffer. PI9containing fractions from the heparin-agarose column were pooled and dialyzed at 4°C against 25 mM Tris-HCl (pH 8.0) containing 1 mM DTT.
The dialyzed sample was concentrated by ultrafiltration to ϳ5 ml and applied to a Mono Q HR5/5 column equilibrated at room temperature with 25 mM Tris-HCl (pH 8), 1 mM DTT. PI9 was eluted from the column in a linear gradient of NaCl (0 -1 M)in equilibrating buffer. Fractions eluted from the Mono Q column were subjected to SDS-PAGE, and pure fractions were pooled and stored at Ϫ80°C.

Molecular Cloning and Nucleotide Sequencing of Two Novel
Intracellular Serpins-In previous studies, PI6 was identified and subsequently isolated in a functionally active form from the cytosolic fraction of a monkey kidney epithelial cell line, BSC-1 (16), as well as human placenta (17). Cloning and sequencing of the cDNA encoding human placental PI6 revealed that it was a new member of the serpin superfamily and showed greatest sequence identity with other members of the ovalbumin branch of intracellular serpins (17). In an effort to identify additional relatives of PI6, a human gt11 placental cDNA library was screened with a 209-base pair polymerase chain reaction-generated 32 P-labeled probe corresponding to the codons of amino acid residues 67-149 of PI6. Several plaques were identified that showed weak hybridization with the probe at high stringency, suggesting the existence of cDNAs encoding proteins related to PI6. Several of these plaques were isolated and their inserts subjected to DNA sequence analysis. Two of these clones, designated as H2-2-11 and H3-1-11, contained inserts with open reading frames encoding proteins of 376 and 374 amino acids, respectively ( Fig. 1A and B). Both proteins have predicted molecular masses of approximately 42 kDa. The 5Ј-regions of both cDNAs contain a Kozak consensus sequence (29) between nucleotide bases 109 -115 of H2-2-11 and 89 -95 of H3-1-11 that include an in frame initiation codon. A second Kozak sequence also exists 117 nucleotide bases downstream of the first initiation codon and includes the codon for Met 41 of both proteins. The importance of these alternative translational start sites and the extent of their utilization in vivo is unknown. The 3Ј-untranslated region of the H3-1-11 cDNA contains an AATAAA consensus sequence located 99 nucleotide bases downstream of the termination codon for nascent mRNA cleavage and polyadenylation (Fig. 1A) (30). However, a polyadenylation consensus sequence was not found in the 3Ј-untranslated region of the H2-2-11 cDNA after sequencing 151 nucleotides downstream from the translational termination codon (Fig. 1B), suggesting that the nucleotide sequence of the H2-2-11 3Ј-untranslated region is incomplete.
A computer search of the NBRF protein data base revealed that the proteins encoded by the H2-2-11 and H3-1-11 cDNAs were novel but showed considerable amino acid sequence identity with members of the ovalbumin branch of the serpin superfamily of proteinase inhibitors. The H3-1-11 and H2-2-11derived amino acid sequences showed 68 and 63% identity with PI-6, respectively. We have provisionally designated the proteins encoded by the H3-1-11 and H2-2-11 cDNAs as PI8 and PI9, respectively, according to recommendations provided by the Genome Database nomenclature committee. Similar to PI6, PI8 and PI9 exhibit a high degree of amino acid sequence identity to other human members of the ovalbumin family of cytoplasmic serpins including EI (PI8, 51%; PI9, 49%), PAI-2 (PI8, 46%; PI9, 45%), and SCCA (PI8, 46%; PI9, 45%). In addition, PI8 showed 63% amino acid sequence identity to PI9. The two novel cytoplasmic antiproteinases exhibit all the structural features previously demonstrated to be unique to the ovalbumin family of serpins that can be summarized as follows (10): (a) PI8 and PI9 lack an N-terminal extension, and the open reading frame begins at amino acid residue 23 of ␣ 1proteinase inhibitor; (b) PI8 and PI9 lack a C-terminal extension terminating at Pro 376 , the equivalent of Pro 391 in ␣ 1proteinase inhibitor; (c) PI8 and PI9 both have a Ser at position 375 of ␣ 1 -proteinase inhibitor in place of a highly conserved Asn found among serpins distantly related to the ovalbumin family; and (d) PI8 and PI9 appear to lack a typical N-terminal cleavable signal sequence. The new cytoplasmic antiproteinases also have a potential N-glycosylation consensus sequence (NX(T/S)) at Asn 8 and Asn 78 of PI8 and Asn 6 and Asn 23 of PI9 (Fig. 1, A and B).
Alignment of the deduced primary structure of PI8 and PI9 with the amino acid sequences of PI6 (Fig. 2) and other human FIG. 2. Comparison of the amino acid sequences for PI6, PI8, and PI9. Amino acid residues common to the three homologs are boxed. Hyphens are introduced for optimal alignment. The complete amino acid sequence for human PI6 has been reported previously (17,48).

Identification of Two Novel Ovalbumin-type Serpins
members of the ovalbumin serpin family (data not shown) identified the putative reactive center P 1 -P 1 Ј residues of PI8 as Arg 339 -Cys 340 , respectively, which are identical to PI6. However, the regions flanking the P 1 -P 1 Ј residues in PI6 and PI8 are highly divergent. The P 2 -P 6 residues of PI6 and PI8 show no identity, while Arg 342 in the P 3 Ј position was conserved in both serpins. Since residues in the vicinity of P 1 have been previously shown to influence both proteinase target specificity and the inhibitory potency of several serpins (31), these findings suggest that PI6 and PI8 interact with the active sites of distinct cognate proteinases that have trypsin-like substrate specificity. In contrast, alignment of the PI9 amino acid sequence identified the putative P 1 -P 1 Ј residues as Glu 340 -Cys 341 , respectively. The identification of an acidic P 1 residue in the PI9 reactive center is unique to the human serpin superfamily. The only other serpins identified with an acidic P 1 residue in their reactive centers are CrmA (Asp) (32) and a recently described rabbit ␣-1-antiproteinase E (Glu) (33). The latter serpin failed to inhibit the amidolytic activities of trypsin, thrombin, and S. aureus endoproteinase Glu-C (33).
These observations prompted us to determine the overall structural relatedness of the cytoplasmic antiproteinases and the crmA protein relative to other intracellular serpins. Previous studies have reported ϳ30% amino acid sequence identity between CrmA and several extracellular serpins including antithrombin III, human and murine ␣ 1 -antichymotrypsin, human and murine ␣ 1 -proteinase inhibitor, and human heparin cofactor II (32). However, to our knowledge, no study has compared the primary structures of CrmA and recently discovered members of the mammalian intracellular serpins. By employing the NBRF program ALIGN, CrmA was found to share ϳ39% amino acid sequence identity with PI6, EI, and PAI-2, ϳ37% identity with PI8 and PI9, and ϳ35% identity with SCCA (data not shown). As a result, these intracellular serpins appear to represent the closest mammalian relatives of the viral serpin reported to date. These findings are consistent with the previous observation that CrmA lacks a cleavable N-terminal signal sequence and an N-terminal extension common to the ovalbumin family of intracellular serpins (10). In addition, a comparison of the CrmA reactive center loop with the reactive center loops of the mammalian intracellular serpin family revealed a remarkable degree of structural conservation with PI9 (Fig. 3). The amino acid sequence of the PI9 reactive center loop shares ϳ54% of the structurally conserved residues found in the reactive center loop of CrmA. An insignificant degree of structural conservation is found between the amino acid residues in the reactive centers of CrmA, EI, SCCA, and PAI-2, ranging from 0 to 7%. A noteworthy feature characteristic of only the cytoplasmic antiproteinase homologs and CrmA is the presence of a conserved Cys residue in the reactive center P 1 Ј positions. Since the reactive center has been previously demonstrated to be the most divergent domain of the serpin superfamily (34), the cowpox virus may have acquired a mammalian intracellular serpin gene and the reactive center of the viral serpin either converged or remained relatively conserved with the reactive center of PI9.
Expression, Purification, and Partial Characterization of PI8 and PI9 -The full-length cDNAs for PI8 and PI9 were ligated into mammalian cell and yeast expression vectors and expressed in BHK cells and yeast cells, respectively, as described above. Using a rabbit anti-PI6 IgG preparation, an immunoreactive doublet that migrated with an apparent M r of 42-44 kDa was detected in the BHK/PI8 cell cytosolic fraction by Western blotting (Fig. 4). The immunoreactive doublet was consistently observed in several BHK/PI8 cytosolic fractions with varying intensity of the lower M r band ranging from ϳ20 to 50% of the doublet. Although not investigated further, the duplicity of the putative PI8 may be due to proteolytic cleavage during the preparation of the cytosol. A faint immunoreactive band at ϳ42 kDa was also detected in the cytosolic fraction derived from BHK/Zem 229R cells that served as a control (data not shown). The cytosolic fraction of the BHK/PI8 cells inhibited the amidolytic activity of trypsin toward S-2251 in a dose-dependent manner, indicating that PI8 recognizes trypsin-like serine proteinases. The cytosolic fraction derived from the BHK/Zem 229R cells also inhibited trypsin amidolytic activity in a dosedependent manner, but this inhibition was ϳ10% of that observed for an equivalent concentration of BHK/PI8 cytosol. Furthermore, overnight incubation of the BHK/PI8 and BHK/ Zem 229R cytosolic fractions with anti-PI6 IgG essentially blocked the ability of these cytosolic fractions to inhibit trypsin, while incubation with preimmune rabbit IgG was without effect (data not shown). Incubation of the BHK/PI8 cytosolic fraction with human thrombin resulted in the migration of the putative PI8 as a single band with an apparent M r of 43 kDa and the formation of a PI8/thrombin SDS-stable complex that migrated with an apparent M r of ϳ67 kDa (Fig. 4). Consistent The boxed residues designate amino acid residue identities between the reactive center sequences of the mammalian serpins and the CrmA reactive center sequence. The solid circles represent accepted mutations in the PI9 reactive center that are structurally conserved to the correponding amino acid residues in CrmA. The mammalian reactive center sequences were searched for structurally conserved amino acid alignments with CrmA based on the grouping system previously reported by Dayhoff (49), including D, E; R, K; Y, F; I, V; and S, T. The arrow designates the cleavable P 1 -P 1 Ј peptide bond in the reactive center sequences.
FIG. 4. Immunoblot analysis of PI8 and PI9 following incubation with human thrombin. Samples of PI8 (cytosolic fraction) and purified recombinant PI9 were incubated with either buffer control (TBS) or 50 nM human thrombin dissolved in TBS for 30 min at 37°C. Each incubation mixture was subjected to 12% SDS-PAGE following reduction and electrophoretically transferred to a nitrocellulose membrane. PI8-, PI9-, and PI8-thrombin complexes were visualized by incubating with rabbit anti-PI6 IgG followed by incubation with 125 Ilabeled protein A and autoradiography. with the trypsin inhibition data, incubation of the BHK/Zem 229R cytosolic fraction with 125 I-labeled thrombin also resulted in a 125 I-labeled 67-kDa SDS-stable complex that appeared to be ϳ10% of that observed in the BHK/PI8 cytosol (data not shown). In addition, incubation of the BHK/PI8 and BHK/Zem 229R cytosolic fractions with anti-PI6 IgG completely blocked formation of the 125 I-labeled 67-kDa SDS-stable complex (data not shown). Collectively, the above observations provide evidence for the synthesis of PI8 following transfection of BHK cells with PI8 cDNA based on (a) its cross-reaction with anti-PI6 IgG, (b) its enhanced ability relative to mock-transfected cells to inhibit trypsin amidolytic activity, and (c) its enhanced ability to form an SDS-stable, 67-kDa complex with thrombin. Our data also suggest the low level, constitutive synthesis of hamster PI6 and/or PI8 by the BHK cells that apparently cross-react with rabbit anti-human PI6 IgG.
Recombinant PI9 was purified to homogeneity from transformed yeast cell lysates in a two-step procedure using heparinagarose column chromatography and Mono Q fast protein liquid chromatography. Initial experiments indicated that ϳ20% of the total protein in the transformed yeast cell lysate migrated in SDS-PAGE with a molecular mass of ϳ42 kDa (Fig. 5). In contrast, this protein was not observed in the mock-transfected yeast cell lysate by this technique (data not shown), suggesting that the 42-kDa protein was PI9. In the absence of a functional assay for PI9, the presence of high expression levels of the putative PI9 in the yeast cell lysates facilitated identification of this protein in column effluents by SDS-PAGE. As seen in Fig.  5, the PI9 pool from heparin-agarose was ϳ90% pure, while the PI9 derived from Mono Q fast protein liquid chromatography migrated as a single band in SDS-PAGE with an apparent molecular mass of 42 kDa (Fig. 5). Amino-terminal amino acid sequence analysis of the purified, putative PI9 indicated that the protein, like PI6 (16), was derivitized at the amino terminus. Treatment of the protein with methanolic HCl failed to release a potential N-formyl group, suggesting that the protein was acetylated at the N terminus. In order to demonstrate that the isolated protein was indeed PI9, the purified protein was cleaved with trypsin and the tryptic digest fractionated on a C 4 reverse phase HPLC column. One peptide (T28) was isolated and yielded an amino-terminal sequence of LAHVGEV, which is identical to that observed in the PI9 deduced sequence at residues 206 -212, thus confirming the identity of the isolated protein as PI9.
The purified recombinant PI9 was then tested for its ability to inhibit the amidolytic activities of trypsin and papain toward S-2251 and N ␣ -benzoyl-DL-arginine p-nitroanilide, respectively. At a 500:1 inhibitor:enzyme molar ratio, PI9 failed to inhibit the amidolytic activities of these proteases under conditions where trypsin and papain amidolytic activities were completely inhibited by a 10-fold molar excess of soybean trypsin inhibitor and egg white cystatin, respectively (data not shown). In addition, purified PI9, at a 500:1 inhibitor:enzyme molar ratio, failed to inhibit the amidolytic activities of human thrombin and S. aureus endoproteinase Glu-C and did not form a high M r complex with human thrombin as judged by Western blotting using the rabbit anti-PI6 IgG to detect PI9 (Fig. 4).
Northern Analysis-To identify transcripts that encode PI8 and PI9 and determine their human tissue distribution, radiolabeled probes were used to hybridize immobilized poly(A) ϩ mRNA by Northern analysis. Northern analysis with a PI8 32 P-labeled 1.1-kb probe revealed two transcripts of 1.4 and 3.8 kb (Fig. 6A). Both transcripts were equally abundant in all tissues except in skeletal muscle and heart, where the 1.4-kb transcript exceeded the levels of the 3.8-kb transcript. The transcripts encoding PI8 were detected at the greatest levels in skeletal muscle, liver, lung, and placenta. Two mRNA species of 3.4 and 4.4 kb were detected with a PI9 reactive center oligonucleotide probe (Fig. 6B). Both mRNA species encoding the PI9 reactive center were detected at the highest levels in placenta and lung and were weakly detected in all tissues examined. In addition, two minor PI9 mRNA species of ϳ7.5-8.0 kb were also detected in placenta. The hybridization of the Northern blots was of sufficient stringency to preclude hybridization of inexact nucleotide matches, eliminating the possibility of nonspecific hybridization. Our previous studies identified a single PI6 transcript of 1.4 kb that was expressed in all tissues except brain and detected in greatest abundance in skeletal muscle and placenta. Therefore, the transcripts encoding the cytoplasmic antiproteinases appear to have an overlap- ping but differential tissue distribution with PI8-and PI9related proteins encoded by multiple transcripts. DISCUSSION In the present study, we have cloned and sequenced two human cDNAs encoding novel proteins that exhibit all the structural features characteristic of the ovalbumin branch of the serpin superfamily. In particular, the new proteins showed greatest amino acid sequence identity with a recently discovered member of the ovalbumin serpins that has been designated by the Genome Database organization as proteinase inhibitor 6, or PI6. The extent of amino acid sequence identity between the new serpins and PI6 (63 or 68%) significantly exceeds that reported for any other combination of the ovalbumin serpin family members, which typically range between 45 and 50% (10). The new serpins have been designated at PI8 and PI9 according to recommendations made by the Genome Database nomenclature committee. Based on primary structure identity and the presence of a unique Cys residue conserved in the reactive center P 1 Јposition of the cytoplasmic antiproteinases, these serpins appear to represent a distinct subfamily within the ovalbumin branch of the serpin superfamily. Since serpins are generally classified by the reactive center P 1 specificity residue (1, 2), PI8 appears to be an Arg-serpin and, like PI6, inhibits trypsin amidolytic activity and forms an SDSstable complex with human thrombin. On the other hand, PI9 was found to be unique in that it is the first human serpin identified with an acidic residue in the reactive center P 1 position and has been classified as a Glu-serpin. PI6 has been previously shown to function as a proteinase inhibitor of several prototypical serine proteinases (16,17), and PI8 and PI9 both show complete conservation of the PI6 reactive center hinge region that conforms to the structural motif P 17 EEGTEAAAATP 8 recently identified in the majority of inhibitory serpins (35). Serpins that carry unconserved mutations in the reactive center hinge region typically lack serine proteinase inhibitory activity since steric hindrance impedes partial insertion of the hinge peptide into the antiparallel A-␤ sheet (36,37), which normally provides a source of favorable interactions that contribute to the overall stability of the proteinase-serpin inhibitory complex. Therefore, since the reactive center hinge sequence of PI9 conforms precisely with the hinge sequences of other inhibitory serpins, this novel cytoplasmic antiproteinase is likely to function as active site-directed proteinase inhibitor.
The novel cytoplasmic antiproteinases displayed all of the structural characteristics common to the mammalian serpins of the ovalbumin family, including the apparent absence of a typical N-terminal cleavable signal sequence (10). These findings suggest that, like PI6 (16,17,38), EI (12,13), PAI-2 (39 -41), and SCCA (14), the new serpins also reside in the cytoplasm of cells. Nonetheless, similar to other members of the ovalbumin serpin family, PI8 and PI9 have consensus sites for the potential attachment of N-linked carbohydrate. Therefore, the possibility that PI8 and PI9 are secreted and function in the extracellular milieu under certain conditions cannot be ruled out at present. Previous studies have demonstrated that PAI-2 (40,41) and SCCA (14) can exist intracellularly and/or extracellularly. For example, PAI-2 has been detected predominantly in the cytosolic fraction of resting monocytes as an unglycosylated functionally active inhibitor of urokinase (40,41). Upon activation of the monocytes with phorbol esters, the majority of intracellular PAI-2 is efficiently glycosylated and secreted (40,41). Consequently, the degree of glycosylation has no effect on the inhibition kinetics of urokinase by PAI-2 (40). Likewise, SCCA has been reported to be localized in the cytoplasm of normal squamous epithelial cells while the corre-sponding carcinoma cells secrete a glycosylated form of SCCA (14), apparently in a manner similar to that described for PAI-2. To date, PI6 has been detected only in the cytosolic fraction of all cultured cells examined, and treatment of cells with phorbol esters had no effect on the intracellular localization of functionally active PI6 (16,18,38). Given the high degree of amino acid sequence identity in regions outside the reactive centers of the cytoplasmic antiproteinases, PI8 and PI9 may also behave like PI6 and be confined to the cytoplasm of cells.
To date, the only intracellular serpin with a defined intracellular proteinase target is the viral serpin encoded by the crmA gene of the cowpox virus (5). CrmA functions as a specific inhibitor of the ICE (5), which has recently been shown to represent a prototype of a larger family of ICE-like homologs (42)(43)(44)(45)(46). The ICE family of cysteine proteinases have been intimately linked to both the negative and positive regulation of apoptosis (44). Mammalian intracellular inhibitors of these cysteine proteinases have not been identified. In the present study, we have made the previously unreported observation that CrmA shows the greatest degree of amino acid sequence identity to the mammalian intracellular serpins of the ovalbumin family, suggesting a possible origin of the crmA gene. In addition, the reactive center of CrmA shows considerable structural similarity to the reactive center of PI9, including a conserved Asp to Glu switch in the P 1 specificity site. Moreover, all of the cytoplasmic antiproteinases have a unique Cys residue conserved in the P 1 Ј position and found only in the corrresponding position of CrmA. CrmA has been shown recently to form a noncovalent tight inhibitory complex with ICE (47), suggesting that the geometric compatibility between serpins and some proteinases is important in determining inhibitory potency. Therefore, since the cytoplasmic antiproteinases are the only mammalian serpins with a Cys residue in the P 1 Ј position, they may regulate proteinases by a mechanism analogous to CrmA.