Expression of biologically active human antithrombin III by recombinant baculovirus in Spodoptera frugiperda cells.

Antithrombin III (ATIII) is a plasma-borne serine protease inhibitor that plays a pivotal role in the regulation of hemostasis. The cDNA for ATIII has been available, but genetic studies on the functional domains of ATIII have not progressed because of the absence of an expression system that will yield sufficient quantities of biologically active protein for biochemical analyses. In the present studies the cDNA of the human antithrombin III gene was inserted into the vector pVL 1393, which is suitable for cotransfection of Spodoptera frugiperda (Sf9) insect cells with Baculovirus wild-type DNA. Recombinant virus particles were selected by the presence of occlusion-negative plaques. Upon infection with purified recombinant virus, Sf9 cells secreted 10-35 micrograms of ATIII/1 x 10(6) cells. Southern analysis of DNA from infected cells demonstrated incorporation of the full-length cDNA into the Baculovirus recombinant, and RNase protection experiments verified the presence of full-length transcript. This recombinant ATIII protein was immunologically reactive with antisera raised against native human ATIII, formed stable complexes with thrombin, and was heparin-accelerated at the same concentration as native human ATIII. In addition, the recombinant ATIII retained specificity for the same molecular species of heparin that activates authentic human ATIII. This is the first successful production of active, recombinant ATIII in quantities that will allow purification on the milligram scale and permit a biochemical analysis of genetically engineered variants.


Expression of Biologically Active Human Antithrombin I11 by Recombinant Baculovirus in Spodoptera frugiperda
Antithrombin I11 (ATIII) is a plasma-borne serine protease inhibitor that plays a pivotal role in the regulation of hemostasis. The cDNA for ATIII has been available, but genetic studies on the functional domains of ATIII have not progressed because of the absence of an expression system that will yield sufficient quantities of biologically active protein for biochemical analyses. In the present studies the cDNA of the human antithrombin I11 gene was inserted into the vector pVL 1393, which is suitable for cotransfection of Spodopteru frugiperda (Sf9) insect cells with Baculovirus wild-type DNA. Recombinant virus particles were selected by the presence of occlusion-negative plaques.
Upon infection with purified recombinant virus, Sf9 cells secreted 10-35 pg of ATIII/l X lo6 cells. Southern analysis of DNA from infected cells demonstrated incorporation of the full-length cDNA into the Baculovirus recombinant, and RNase protection experiments verified the presence of full-length transcript. This recombinant ATIII protein was immunologically reactive with antisera raised against native human ATIII, formed stable complexes with thrombin, and was heparin-accelerated at the same concentration as native human ATIII. In addition, the recombinant ATIII retained specificity for the same molecular species of heparin that activates authentic human ATIII. This is the first successful production of active, recombinant ATIII in quantities that will allow purification on the milligram scale and permit a biochemical analysis of genetically engineered variants.
Antithrombin I11 (ATIII)' is a plasma-borne serine protease inhibitor that acts as a key regulatory element in the blood clotting cascade (1)(2)(3). This regulation is principally due to the inhibition of thrombin (4), although ATIII inhibits several other proteases in the cascade less efficiently, including factors IXa and Xa (5), XIa (6), and XIIa (7). Although the mechanism underlying the inhibition of thrombin by ATIII is not completely understood, it is agreed that a small fragment of ATIII is cleaved and released during the process (8)(9)(10). Thrombin and ATIII remain in tight association in a 1:l * This work was supported by Grant R01-FM34001 from the National Institutes of Health and Grant 86-SlO7 from the American Heart Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
' $ To whom correspondence should be addressed.
' The abbreviations used are: ATIII, antithrombin 111; SDS, sodium dodecyl sulfate; Sf9, Spodoptera frugiperda 9; PME, P-mercaptoethanol; PAGE, polyacrylamide gel electrophoresis; HEPES, 4-(2-hydrox-yethy1)-1-piperazineethanesulfonic acid; PIPES, 1,4-piperazinediethanesulfonic acid. stoichiometry in complexes that are stable to boiling in SDS and electrophoresis in the presence of PME (1). Although the precise nature of the ATIII-thrombin association has not been clearly demonstrated, it is assumed that it represents the covalent acylation intermediate that arises during serine protease hydrolysis. The anti-protease activity of ATIII, as well as several other plasma-borne serine protease inhibitors, is dramatically activated by the glycosaminoglycan heparin (1). This activation results in a several thousandfold acceleration in complex formation between ATIII and thrombin and is initiated by the binding of heparin to ATIII (1). Since the binding of heparin to ATIII appears to be a prerequisite for acceleration, elucidation of the structure and location of the heparinbinding site in ATIII is fundamental to the mechanistic understanding of its activation. To date, there have been several studies indicating that the first third of the molecule contains the heparin-binding domain. 'H nuclear magnetic resonance studies demonstrate that a conformational change occurs in the first third of the molecule upon heparin addition (ll), and analysis of CNBr fragments of ATIII show heparin binding to this region (12). Studies employing chemical modification of lysine residues 107, 125, 133, and 136 (13-E), which alter the heparin binding capacity, also support the premise that the first third of the molecule contains the heparin-binding site. More recently, the heparin-binding site has been narrowed to the region of the molecule containing residues 124-145 (16, 17). Because the heparin-binding domain has been narrowed to such a short region of the molecule, the next logical step in finding the exact amino acids involved in heparin binding is to employ genetic manipulation of the protein sequence.
In the present studies we have taken the first step toward this goal by generating a recombinant Baculovirus containing the human ATIII cDNA in place of the major coding region of the polyhedrin gene (18). Sf9 insect cells infected with the recombinant virus produce and secrete ATIII which is biologically active, heparin-activated, and strongly reactive with antisera raised against authentic human ATIII purified from human plasma. These studies represent the first successful expression of biologically active recombinant human ATIII in quantities sufficient for biochemical analysis.
EXPERIMENTAL PROCEDURES Materials-The ATIII cDNA (phATII1-113) cloned into pBR322 was a generous gift from Dr. Savio Woo (Howard Hughes Medical Institute, Houston, TX). ATIII was purified from human plasma as previously described (16). This preparation is referred to as authentic ATIII throughout the paper. Thrombin was a generous gift from Dr. John Fenton. Anti-human ATIII antisera and radioimmune assay grade bovine serum albumin were from Sigma. Na"'1odide and Hybond N were purchased from Amersham Corp. Electrophoresis reagents were purchased from Bio-Rad. Protein G-Sepharose was from

R.
Scientific. All other common reagents were from Sigma, Cal-Biochem, and Bethesda Research Laboratories. Construction and Transfection-The ATIII cDNA was cut out of pBR322 with PstI and MstI and subcloned into Bluescript (Stratagene) at the PstI and SmaI sites. The gene was then digested out of Bluescript with PstI and Xba, and inserted into vector pVL 1393 at those sites in the polylinker. The resulting pVL 1393-AT111 vector was purified by cesium chloride centrifugation (18). Two X lo6 Sf9 cells were cotransfected with 2 pg of the purified vector and 1 pg of wild-type viral DNA as described (18). Recombinant virus particles were purified by plaque assay (18), employing visual and microscopic inspection for the presence of occlusion-negative plaques. T-25 flasks seeded with 3 X lo6 cells were infected with second-round-purified recombinant plaques, wild-type virus, and control media from uninfected cultures. The media from cells infected with recombinant isolates 1, 2, wild-type virus, and uninfected control medium, as well as the cells from the flasks, were used in many of the following experiments. The recombinant Baculovirus isolates 1 and 2 are referred to as BrATIII-1 and BrATIII-2 throughout the rest of the paper.
Southern Analysis-Cells infected with BrATIII-1 and BrATIII-2, wild-type virus, and uninfected cells were harvested for DNA extraction (18). Briefly, the cells were removed from the flasks and lysed (0.03 M Tris, pH 7.5, 0.01 M Mg(OAc)a, 1% Nonidet P-40), and the nuclei were pelleted by centrifugation at 1200 X g. The pellets were resuspended in an extraction buffer (0.1 M Tris, pH 7.5, 0.1 M Na2EDTA(2H20), 0.2 M KCl), subjected to proteinase K digestion, and solubilized in the detergent Sarkosyl. After extraction and precipitation the total DNA from each pellet was resuspended in 500 pl of 0.1 X TE (10 mM Tris-C1, 1 mM EDTA, pH 8.0). Ten p1 of each suspension was double-digested with PstI and BamHI to cut out any recombinant gene. Both uncut and cut samples of the total DNA preparations along with the vector containing the gene were electrophoresed into 1% agarose gels and transferred to Hybond N by capillary transfer (19). The transfers were prehybridized at 65 'C and then hybridized at 55 "C overnight with the 3ZP-phATIII-113 gene (19).
RNA Nuclease Protection Assay"Sf9 cells seeded at 1 X lo' cells/ plate were either innoculated with recombinant virally infectious media at a multiplicity of infection of 301, or left uninfected. The cells were allowed to incubate for 48 h and then were harvested for total RNA. The method of RNA purification was adapted from the Chomczynski and Sacchi (20) single-step method using an acid guanidinium thiocyanate/phenol/chloroform extraction. Briefly, the cells were removed from the dishes and pelleted by centrifugation at 600 X g. After removal of the media, the cells were resuspended in 2 ml of solution D (4 M guanidinium thiocyanate, 25 mM sodium citrate, pH 7, 0.5% Sarkosyl, 0.1 M PME) and mixed by pipetting up and down. Next, 0.2 ml of 2 M sodium acetate, pH 4,2 ml of phenol (water saturated), and 0.4 ml of chloroform/isoamyl alcohol (24:l) were added sequentially and mixed thoroughly after each addition. After addition of the chloroform solution, the suspension was mixed for 10 s and then left on ice for 15 min. The samples were centrifuged for 20 min at 10,000 X g at 4 "C. The RNA contained in the aqueous phase was precipitated with 1 ml of isopropyl alcohol at -20 "C for 1 h. After pelleting the RNA by centrifugation at 10,000 X g for 20 min at 4 "C, the RNA was resuspended with 0.3 ml of solution D, transferred to microcentrifuge tubes, and again precipitated with 0.3 ml of isopropyl alcohol. The pelleted and rinsed (with 75% EtOH) RNA was resuspended in TE and stored at -70 "C.
A full-length antisense riboprobe was made by linearizing the Bluescript containing ATIII with BamHI and using the T7 promoter for in uitro transcription. Total RNA from infected and uninfected of "P-antisense ATIII riboprobe using NaOAc and EtOH. The pre-Sf9 cells, 10 and 30 pg of each, were coprecipitated with 4 X lo6 cpm cipitates were resuspended with 20 pl of buffer containing 80% formamide, 40 mM PIPES, pH 6.4, 400 mM NaCl, and 1 mM EDTA (21) and split into two tubes (10 pl each), thus resulting in 5 and 15 pg of total RNA. After denaturation at 75 "C for 5 min, the RNAs were allowed to anneal at 42 "C for 4 h. The unprotected RNA was digested simultaneously with 100 ng of RNase A and 100 units of RNase T1 in RNase buffer (10 mM Tris-HC1, 5 mM EDTA, 300 mM NaCl, pH 7.7) for 1 h at 30 "C. The reaction was stopped by the addition of 30 p1 of 10% SDS and 50 pg of proteinase K. The reactions were incubated for an additional 30 min at 37 "C. The protected RNA was extracted with phenol/chloroform and precipitated with isopropyl alcohol. Protected RNA was separated on 5% denaturing polyacrylamide gels, dried under vacuum, and visualized by autoradiography.
ATZZZ Linkage Assay-Twenty-five p1 of media harvested from cells infected with recombinant viruses (BrATIII-1 and BrATIII-2), wild-type virus, uninfected control cultures, and 300-1,000 ng of authentic ATIII were incubated with 16 ng of '251-thrombin at 37 "C for 30 min. The reactions were terminated by the addition of 25 p1 of SDS sample buffer containing 2 mM @ME and boiled for 5 min. Thirty pI of each sample was separated by SDS-PAGE on 10% polyacrylamide gels (23). The gels were dried and exposed to x-ray film (XAR-5) overnight at -70 "C in the presence of a Cronex Lightning plus enhancement screen.
Immunoprecipitation-ATIII-thrombin complexes were formed as described above, with the exception that the reactions were not terminated with SDS sample buffer or boiled. The complexes were immunoprecipitated at 37 "C for 60 min using anti-AT111 antisera at a final dilution of 1:lOO. The antibody complexes were precipitated by a 30-min incubation with 100 p1 of Protein G-Sepharose (a 1:5 slurry in phosphate-buffered saline), and pelleted by centrifugation in a microcentrifuge for 30 s at 10,000 X g. The pellets were washed with a buffer consisting of 20 mM HEPES, pH 7.5, 1% Nonidet P-40, 0.5% SDS, 1 mM phenylmethylsulfonyl fluoride, 0.1 mg/ml ovalbumin, and centrifuged again. The samples were resuspended in 50 pl of SDS sample buffer containing 2 mM PME and analyzed by SDS-PAGE on 10% polyacrylamide gels followed by autoradiography as described above.
Zmmunoblotting-Twenty-five p1 of each medium sample and 50 ng of ATIII were separated on 10% polyacrylamide gels and transferred to nitrocellulose for 1 h at 70 V using a semidry blotting apparatus. The blots were blocked overnight at room temperature with 3% radioimmune assay grade bovine serum albumin in Trissaline, pH 7.5 (10 mM Tris, 0.15 M NaCl). Anti-AT111 antiserum was added at a 1:lOO dilution, and the incubation was continued for another 4 h at room temperature. Following removal of the antiserum and successive 5-min washes, once in Tris-saline, twice in Tris-saline + 0.05% Nonidet P-40, and once more in Tris-saline, the blots were transferred to blocking solution containing 10' cpm/ml of '251-Protein G. Following a 1-h incubation at room temperature, the "'I-Protein G was decanted, and the wash protocol described above was repeated. An autoradiogram of the dried nitrocellulose was prepared by exposure to x-ray film overnight in the presence of an enhancement screen at -70 "C.
Heparin Actiuation-Twenty-pl samples of medium harvested from BrATIII-1-infected cells were preincubated with 0, 1, 10, 25, 50, and 100 ng of heparin at 37 "C for 30 min. One pl of lZ5I-thrombin (16.8 ng/pl; 25,787 cpm/ng) was added, and the reactions were incubated for an additional 30 s at 37 "C. The reactions were terminated by the addition of 25 pl of SDS sample buffer containing 2 mM PME, separated by SDS-PAGE on 10% polyacrylamide gels, dried, and exposed to x-ray film as described above.
Endoglycosidase F Digestion-One-hundred and fifty pl of medium harvested from BrATIII-2-infected Sf9 cells and 1.8 pg of ATIII (in 25 p1 of Excel1 medium) were digested for approximately 20 h with 3 and 1 unit(s) of endoglycosidase F (Behring Diagnostics), respectively. The digested proteins were precipitated with ice-cold 10% trichloroacetric acid, pelleted by centrifugation for 10 min at 10,000 dried. The protein pellets were resuspended in 25 pl of reduced SDS-X g in a microcentrifuge, rinsed with ice-cold acetone, repelleted, and PAGE sample buffer and separated by SDS-PAGE on a 10% polyacrylamide gel along with undigested ATIII (450 ng) and 25 p1 of medium harvested from the BrATIII-2-infected cells. The gel was transferred to nitrocellulose, incubated with anti-AT111 antiserum, and processed for autoradiography as described above.
Chromogenic Assay for Thrombin Actiuity-Reactions were carried out in duplicate wells of a 96-well cell culture plate. Thirty ng of human thrombin was incubated with 25 ng of heparin in the presence of known amounts of purified human ATIII (18-180 ng) or conditioned medium (1-50 pl) harvested from uninfected Sf9 cells or cells infected with BrATIII-1 or BrATIII-2. The final reaction volumes were adjusted to 100 p1 with phosphate-buffered saline containing 2 mg/ml ovalbumin. At the end of a 40-min incubation at 37 "C, the reactions were placed on ice, and 50 pl of a 1 mM solution of Chromozyme-TH substrate was added to each well. The reactions were hrought to room temperature and incubated for an additional 10 min. The reactions were terminated by the addition of 100 pl of 5 M acetic acid and quantitated hy ahsorhance at 405 nm. infected, wild-t.ype-infected, and uninfected Sf9 cells was isolated according to the Raculovirus manual (see "Experimental Procedures"). Two pg of each DNA preparation were digested with I'stI and RarnHI, separated on 1% agarose gels, and capillary transferred to Hyhond N paper. The blots were prohed with "l'-phATII1-113 a t 55 "C and visualized by autoradiography after a 36-h exposure. 1Mnc.s I and 2, uninfected; lanes 3 and 4. wild-t-we; lnnes 6 and 7, HrATIII-1; lanes H and 9, HrATIII-2; lnnm 10 and 11. pVI, 139.7 containing ATIII. Digested (+); undigested (-). 1mne.s 5 and 12 contain standards.

Generation and Characterization of Recombinant
RrATIII-1 and RrATIII-2 were prepared by the infection of 3 x 10" Sf9 cells. Seventy-two-h post infection supernatants from RrATIII-1-and RrATIII-2-infected cells were harvested and assayed and yielded titers of 8.5 and 9.0 X 10" pfu/ml, respectively. Several of the experiments described below were done with the harvested medium or the DNA extracted from the infected cells.
To determine if the full-length ATIII cDNA insert was successfully transferred into the Raculovirus genome, samples of total cellular DNA from wild-t.ype Raculovirus-infected cells, and from cells infected with RrATIII-1 and RrATIII-2 were digested and probed with a ~".'P-phATIII-l1.7 (Fig. I R ) . The autoradiogram shows that labeled phATIII probe specifically hybridized t o high molecular weight DNA from cells infected with both RrATIII-1 and HrATIII-2. In contrast, the DNA from wild-type virus-infected cells and uninfected cells did not show specific hybridization. When the DNA samples were double-digested with I?stI and HarnHI, the specific hybridization was localized to a 1.4-kilobase band that migrated identically to the authentic phATIII-113 cIINA that was oriRinally cloned into the pVI, 1.79.7 vector using P s t I and XhaI sites. These data demonstrate that both RrA7'11I recombinants which were plaque purified from the primary transfection contain the full-length AT111 cDNA insert.
Production of Riolo,qically Actiw ATIII by Sf9 C d s Infwtpd with HrATIII Rrcornhinants-Cultures of St3 cells were infected with RrATIII-I and RrATIII-2 or with wild-type virus as described above. Seventy-two h after infection, the culture supernatants were harvested and assaved for the presence of biologicallv active ATIII using the thrombin linkage assav (see "Experimental Procedures"). Rrieflv, 20-pl aliquots of the media were incubated with ""I-thrombin and analyzed for the presence of high molecular weight complexes by SDS-PAGE and autoradiographv. When ""I-thrombin was incubated with medium from cells infected with HrATIII-I or RrATIII-2, complexes of approximately 80 kDa were formed (Fig. 2, lanes I and 2 ) . In contrast. complexes were absent when '"'I-thrombin was incuhated with medium harvested from uninfected cultures or cultures infected with wild-type virus (Fig. 2,  lanes 3 and 4 ) . These data show that the production of a protein capable of forming complexes with thrombin is dependent on infection with the recombinant viruses. BrATIII Expressed in Sf9 Cells Is Recognized by Antihuman ATIII Antiserum-To confirm the identity of the protein involved in the formation of complexes as ATIII, we examined the ability of anti-AT111 antisera to recognize the complexes in an immunoprecipitation experiment. "'I-Thrombin-inhibitor complexes were allowed to form in samples of conditioned medium from BrATIII-1-and BrATIII-2infected cells as described above. Anti-AT111 antiserum was added, and immunoreactive material was recovered by incubation with Protein A-Sepharose, followed by centrifugation. The samples were then solubilized in SDS sample buffer and resolved by SDS-PAGE on 10% polyacrylamide gels (Fig. 3). Complexes formed by the addition of "'I-thrombin to medium harvested from BrATIII-1-and BrATIII-2-infected Sf9 cells were specifically recognized by the anti-AT111 antiserum (Fig.  3, lanes 1 and 4 ) , confirming the identity of the inhibitor in these complexes as human ATIII. These complexes were not immunoprecipitated by control antisera raised against a related serine protease inhibitor, protease nexin-I (Fig. 3, lanes  2 and 5 ) , or in the absence of antisera (Fig. 3, lanes 3 and 6). Immunoprecipitable complexes were absent when 12'I-thrombin was incubated with medium harvested from wild-type virus-infected cells (Fig. 3, lanes 7-9), further demonstrating that the inhibitor was not produced in the absence of recombinant virus. It should be noted, however, that the apparent mass of the complexes is smaller than those formed with authentic human ATIII (Fig. 3, lane 10).
Immunoblot Analysis and Endoglycosidase Digestion-Since the immunoprecipitation data of "'I-thrombin-inhibitor complexes suggested a difference in the molecular masses of authentic and recombinant ATIII, we examined this difference more carefully. Although this difference appeared to be small, even a relatively large difference could be obscured in the high molecular mass range on 10% polyacrylamide gels. T o more accurately quantitate this difference, samples of medium harvested from BrATIII-1-and BrATIII-2-infected cells, and authentic ATIII, were subjected to electrophoresis and immunoblot analysis. In this gel system, the authentic ATIII migrated with an apparent molecular mass of 58 kDa (Fig. 4, lane I), in precise agreement with previously published values (24). The recombinant ATIII expressed by both Rr-ATIII-1-and BrATIII-2-infected cells (Fig. 4, lanes 3 and 7, 2 and 6 respectively), had an apparent molecular mass of 50 kDa, 8 kDa less than authentic ATIII purified from plasma. The Sf9 insect cells have been shown to glycosylate at the universal consensus sites, but the precise nature of the carbohydrate groups added can differ substantially from those added to a protein in its native expression system (24). T o determine if the apparent difference in molecular mass might be due to glycosylation differences, the immunoblot experiment shown in Fig. 4 was repeated with and without endoglycosidase F digestion, to remove N-linked carbohydrate. Both the recombinant and authentic AT111 showed an approximate 4 kDa loss in mass after treatment with endoglycosidase F (data not shown). This apparent difference in molecular mass is addressed further in experiments discussed below.
Recombinant ATIII Is Heparin Actioated-Since the original goal of these studies was to develop an ATIII expression system to examine the mechanism of heparin activation by genetic manipulation, we next examined the ability of the recombinant ATIII to be activated by heparin and to recognize the same molecular species of heparin that selectively activates authentic human ATIII. T o determine if the recombinant ATIII retains specificity for the same species of heparin that activates authentic ATIII, two different types of heparin were compared in an activation assay. These two different preparations of heparin, termed "high" and "low" affinity. were prepared by the repeated passage of commercial porcine mucosal heparin over an affinity column of authentic human ATIII covalently coupled to Sepharose. On each cycle, the retained heparin was eluted with 1.0 M NaCl and is referred to as "high affinity" heparin. The flow-through heparin is referred to as "low affinity" heparin. Both preparations were quantitatively standardized according to uronic acid content c " -.)  (lanes 1, 4, and 7), anti-PN-I antisera at the same dilution (lanes 2,5, and R ) , or no antisera (lanes  3,6, and 9 ) . After the incubation, Protein G-Sepharose was added to precipitate the IgG fraction. The precipitated antibody complexes were harvested by centrifugation, separated by SDS-PACE on 10% polyacrylamide gels, and visualized by autoradiography. Lunes 1-3,  HrATIII-I; lanes 4-6, RrATIII-2; lanes 7-9, wild-type; lane IO, aut hentic ATIII-""I-thromhin complexes. TH, thromhin.  thrombin ('251-Th) were added, and the reactions were allowed to proceed for an additional 30 s. The reactions were rapidly terminated by the addition of SDS sample buffer. Free thrombin and complexes were separated by SDS-PAGE on 10% polyacrylamide gels and visualized by autoradiography. Bands corresponding to the position of high molecular weight '251-thrombin-ATIII complexes were excised and quantitated by gamma counting. ( O " --O ) , high affinity heparin; (U), low affinity heparin. (10). Samples of medium harvested from BrATIII-2-infected Sf9 cells were preincubated with various concentrations of high and low affinity heparin for 30 min at 37 "C. lZ5I-Thrombin was added, and the incubation was continued for 1 min. The reactions were terminated by the addition of SDS sample buffer, followed by the separation of free and complexed thrombin by SDS-PAGE on 10% polyacrylamide gels. Following autoradiography, the bands corresponding to the positions of ATIII-thrombin complexes were excised and quantitated by gamma counting. The high affinity heparin displayed a sharp dose-response curve and was maximally active in accelerating complex formation at a concentration of 100 ng/ml (Fig. 5 ) . In contrast, the low affinity preparation was at least 100-fold less active and never approached the same efficacy over the concentration range examined. These data clearly demonstrate that the recombinant ATIII produced by the insect cells retains a high degree of specificity for the same molecular species of heparin that activates authentic human ATIII.
Finally, it was of interest to examine the kinetics of high affinity heparin activation, comparing recombinant and authentic plasma ATIII in parallel. While the experiment shown in Fig. 5 demonstrates that the specificity for high affinity heparin is retained by recombinant ATIII, a different type of experiment was required to gain the resolution needed to examine the kinetics. To accomplish this, the linkage assay shown in Fig. 5 was repeated with authentic ATIII and recombinant ATIII in parallel, under conditions where the concentration of ATIII was approximately equal to or less than the concentration of '251-thrombin. Using this approach we were able to more accurately define the relationship between heparin concentration and activation (Fig. 6). The activation of both authentic and recombinant ATIII was sharply dependent on the concentration of heparin, and nearly identical activation curves were seen for each. The difference in the amount of high affinity heparin required to achieve half-maximal activation of each was less than 0.3 ng. These data are of particular importance, since they demonstrate that the kinetic interactions of high affinity heparin with authentic ATIII and recombinant ATIII are functionally identical.
BrATIII-1 and BrATIII-2 Produce Full-1engthATZII Tran-High Affinity Heparin, ng FIG. 6. Kinetic interaction of authentic and recombinant ATIII with high affinity heparin. Thirty-six ng of authentic ATIII purified from human plasma (W) and 5 ~1 of medium from BrATIII-1-infected Sf9 cells (M) were incubated with the indicated concentrations of high affinity heparin for 30 min at 37 "C in a reaction volume of 40 pl. At the end of the incubation, 18 ng of Iz51-thrombin was added (10 pl), and the incubation was continued for an additional 30 s. The reactions were terminated by the addition of 50 p1 of boiling SDS-PAGE sample buffer. Twenty-pl aliquots of the reaction mixtures were subjected to SDS-PAGE to separate lZ5I-ATIII complexes from free '251-thrombin. Regions of the gels corresponding to the position of lZ5I-ATIII complexes were excised and quantitated as described in the Fig. 5 legend. The amount of activation observed at each heparin concentration has been normalized to the maximum amount of activation obtained.
scripts-To determine if the apparent difference in mass between the recombinant ATIII and authentic ATIII was due to post-translational modification, or an RNA splice or truncation, we examined the ability of RNA from cells infected with BrATIII-1 to protect full-length in vitro transcripts of the authentic ATIII cDNA. Full-length 32P-phATIII antisense RNA was transcribed from the Bluescript construct using the T7 promotor (see "Experimental Procedures"). Five and 15 pg of total RNA purified from BrATIII-1-infected Sf9 cells was hybridized to the probe, followed by RNase treatment (Fig. 7). Parallel control samples were hybridized to RNA from uninfected cells. RNA purified from cells infected with BrATIII-1 (Fig. 7, lanes 1 and 2 ) completely protected the full-length probe (shown in lane 5 for comparison). In contrast, RNA from uninfected cells afforded no protection (Fig.  7, lanes 3 and 4 ) . These data demonstrate that the BrATIII-1 transcript is indeed full-length and thus does not account for the apparent difference in molecular mass observed between authentic and recombinant ATIII. It should also be noted that only a single size class of protected RNA was seen, demonstrating the existence of a single RNA species.
Partial Purification and NH2-terminal Sequence of Recombinant ATIZI-The immunoprecipitation data shown in Fig.  4 demonstrate a difference in the molecular masses of the complexes formed between the recombinant and native ATIIIs and thrombin. This difference was estimated to be comparable to the difference observed between uncomplexed authentic and recombinant ATIII on immunoblots. Since the active site of ATIII, i.e. the thrombin clip site, is 38 residues in from the carboxyl-terminus, and the molecular mass difference is also observed in complexes, then the carboxylterminal end of the recombinant must be intact. Even though the RNase protection experiments verify that a full-length transcript is produced by the recombinant virus, we tested the possibility that alternative processing at the amino-terminal end of the protein might account for the apparent mass difference.
T o accomplish this, 100 ml of medium harvested from BrATIII-1-infected cells was concentrated and subjected to ion-exchange chromatography over DEAE-Sephacel. The col-  1 (lanes 1-4). The reaction was stopped by the addition of SDS, followed by proteinase K digestion (see "Experimental Procedures"). Protected RNA was separated by 5% denaturing PAGE and visualized by autoradiography. Lanes I and 2, RNA from BrATIII-1 infected cells, 15 and 5 pg, respectively; fanes 3 and 4, RNA from uninfected cells, 15 and 5 pg, respectively; fane 5, "'P-antisense ATlII RNA; lane 6, standards, given in number of bases.
umn was eluted with a 0-1.0 M NaCl gradient. Fractions were assayed for antithrombin activity using a chromogenic thrombin substrate assay (see "Experimental Procedures"). The majority of the ATIII activity eluted in a sharp peak at approximately 0.38 M NaCl (data not shown). A 1.5-ml aliquot of the peak activity fraction was precipitated with 10% trichloroacetic acid, pelleted and subjected to SDS-PAGE on a 10% polyacryamide gel. Parallel lanes were transferred to polyvinylidene difluoride; one lane was stained for protein, the other was incubated with anti-AT111 antibody to immunolocalize the band corresponding to recombinant ATIII (Fig.  8). The two lanes were aligned, and the protein band corresponding to the immunoreactive band was excised and subjected to protein sequence analysis. One major sequence was obtained (Table I). This sequence, Gly-Gly-Ser-Pro-Val-Asp-Ile, matches the NHp-terminal residues of authentic ATIII, with the exception of the first residue which is a His in authentic ATIII. However, this is most likely an artifact due to contaminating glycine in the sample, since the band was recovered from a Tris-glycine-buffered SDS gel. Since the next 6 residues matched exactly, and this sequence does not occur anywhere else in ATIII, we conclude that the recombinant ATIII is processed properly in the insect cell expression system and that the difference in molecular mass is not due t o alternative processing at the amino terminus.
Quantitation of ATIII Production by RrATIII Recombinants-Finally, it was of interest to determine the relative levels of ATIII expression by Sf9 cells infected with BrATII-1 and BrATIII-2 since one of the goals of this study was to produce recombinant ATIII in quantities sufficient for biochemical analyses. A chromogenic assay that specifically measures thrombin activity was employed. Separate flasks of Sf9 cells (3 X 10") were infected with BrATIII-1 and BrATIII-2 obtained from second-round plaque purification, a t a mul-  Table I).   (Fig. 9 R ) . A standard curve of thrombin inhibition was generated in parallel using authentic human ATIII (Fig.  9A). The data are presented as the percent remaining thrombin activity uersu.9 pl of conditioned medium added (see "Experimental Procedures" for details). In the standard curve, 65 ng of authentic ATIII were required to achieve 50% thrombin inhibition under these assay conditions (Fig. 9 A ) . While medium harvested from uninfected Sf9 cells showed no thrombin inhibitory activity, medium harvested from RrATIII-1-and BrATIII-2-infected cells were potently inhibitory. Ba.sed on the volumes of medium required for 50% inhibition, and the total volume of medium harvested from the cells, we estimate a recombinant ATIII production of 10 and 35 pg from 3 x 10' Sf9 cells infected with BrATIII-1 and BrATIII-2, respectively.
These experiments have been repeated at a 50-fold higher multiplicity of infection, and production was found not to he significantly increased (data not shown).

DISCUSSION
In the present studies we used the Baculovirus-insect cell expression system to produce recombinant human ATIII. The  cotransfection of Sf9 cells with wild-type Baculovirus and pVL 1393 vector carrying a full-length human ATIII cDNA with polyhedrin gene-flanking sequences resulted in the formation of two independent recombinant Baculovirus carrying the ATIII cDNA. Both of the recombinants, designated BrATIII-1 and BrATIII-2, were purified by two rounds of plaque purification, are stable, and give rise to the production of recombinant ATIII in Sf9 insect cells.
The recombinant ATIII produced by BrATIII-1 and Br-ATIII-2 appear to be identical t o each other and retain all of the biological properties of authentic ATIII that were tested. Both are immunologically reactive with antisera raised against ATIII purified from human plasma, form complexes with thrombin, and are maximally activated by heparin at the same molar concentration required to maximally activate authentic ATIII. The heparin activation data are of particular importance, since many studies are now directed a t determining the mechanism of heparin activation.
The only difference observed between authentic ATIII and the recombinant forms generated in insect cells infected with BrATIII-1 and BrATIII-2 is the apparent molecular mass determined by SDS-PAGE. We initially thought this difference was probably due to differences in glycosylation, but endoglycosidase treatment suggested that both the authentic and recombinant ATIII were glycosylated to the same degree. While we still have not determined the reason for the apparent mass difference, three lines of evidence suggest that it is not due to a difference in the polypeptide chain per se. First, RNase protection experiments showed that the transcript produced in the insect cells completely protected full-length antisense transcript generated in uitro, demonstrating that the full-length transcript is produced in infected cells. Second, the complexes formed between thrombin and recombinant ATIII showed the same apparent mass difference. Since thrombin cleaves ATIII a t a site 38 residues from the carboxyl terminus (Arg393-Ser394), the modification(s) causing this ap-parent difference must be to the amino side of Arg393. Finally, the recombinant ATIII was partially purified and the amino terminus was sequenced. The amino terminus of the recombinant and authentic ATIII were identical for the first 7 residues. Therefore, the recombinant signal sequence is processed properly, and the difference in mass is not due to a proteolytic cleavage at the amino terminus. Thus, we conclude that the apparent mass difference is due either to posttranslational modifications of the complete polypeptide chain, which has been shown to occur with other proteins expressed in Baculovirus (24), or is simply an SDS-PAGE artifact. Importantly, this apparent mass difference does not affect any of the biological or biochemical properties of activities of the recombinant ATIII that we studied.
These studies were initiated so that an expression system could be developed for human ATIII for use in site-directed mutagenesis experiments, primarily in the region of heparin binding. Previous attempts to express active ATIII have been largely unsuccessful, with the exception of expression in COS cells (25). Unfortunately, the level of transient expression in that system was so low (1 ng/106 cells) that its utility value in site-directed mutagenesis studies is doubtful. Using the Baculovirus expression system, we have obtained expression at a 30,000-60,000-fold higher level than in COS cells. Most importantly, this recombinant ATIII appears to retain functional identity to authentic ATIII. The recombinant ATIII presented in these studies is activated by heparin and retains specificity and the same apparent affinity for the same molecular species of heparin that activates authentic ATIII. ATIII expressed in the Baculovirus expression system should now prove to be a valuable tool in further dissecting the heparin-binding site(s) of ATIII through genetic manipulation.