Viral transformation increases vitamin K-dependent γ-carboxylation of glutamate

EXPERIMENTAL

CELL

RESEARCH

32-40

192,

(19%)

Viral Transformation Increases Vitamin K-Dependent r-Carboxylation of Glutamate DAVID

T.BERG,DONB.MCCLURE,JENNA Departments

of Molecular Lilly

D. WALLS,S.BETTYYAN,*

Genetics and *Department of Biochemistry, Corporate Center, Indianapolis, Indiana

turelfunction use. cc’ 1991

Mammalian cells contain a microsomal vitamin K-dependent carboxylase activity which catalyzes the y-carboxylation of glutamate. While most cells have a limited ability to fully y-carboxylate proteins, it has been suggested that the ability of transformed cells to perform this complex post-translational modification may play a role in tumor biology. In this study, we examined the effect of transformation by adenovirus oncogenes on the ability of cells to efficiently y-carboxylate a vitamin K-dependent protein. Several morphologically transformed BHK-21 cell lines (BHK-Ad) were isolated following the chromosomal integration of the viral oncogenes ElA/ElB from human adenovirus type 12 (Ad12). The lines were capable of growing in soft agar and low serum and produced functional ElA as determined by promoter activation studies. Using a vector for the expression of the vitamin K-dependent recombinant human protein C (HPC), a regulator of the clotting cascade, Ad-transformed and nontransformed lines secreting rHPC were generated. The rHPC from the transformed and nontransformed cell lines displayed identical serine protease activities, and there were no apparent differences in the proteolytic processing of the proteins, although a minor difference in the proportion of each HPC glycoform was observed. However, the functional anticoagulant activity, which depends on the y-carboxyglutamic acid (Gla) content, was -70% higher in the Ad-transformed lines. Approximately 90% of the rHPC from the Ad-transformed lines exhibited a calcium-dependent (high Gla) elution profile on anion-exchange resin, compared to only 15 to 26% from the nontransformed cell clones. By analyzing endogenous microsomal carboxylase, we determined that en- 50% following transformazyme activity increased tion. Overall, our data demonstrate that transformation can increase the potential of a cell to efficiently y-carboxylate a protein and lend support to the suggested involvement of this post-translational modification in tumor cell function. Further, our results demonstrate a potential means of altering cells to enable full modification of vitamin K-dependent factors for struc-

’ To whom

reprint

requests

should

0014.4827/91$3.00 Copyright 8 1991 by Academic Press, All rights of reproduction in any form

Lilly Research 4628.5

studies Academic

Press,

W. GRINNELL~

Laboratories,

and potentially

for therapeutic

Inc.

INTRODUCTION The vitamin K-dependent y-carboxylation of glutamate is a post-translational modification that is known to occur on a number of proteins, including coagulation factors, bone osteocalcin, urinary Gla (y-carboxyglutamate) protein, pulmonary surfactant-associated proteins, and atherosclerotic plaque Gla protein (reviewed in [l]). The biological importance of this modification has been extensively studied with several of the proteins involved in the coagulation pathway (i.e., prothrombin, factors VII, IX, and X, and proteins C and S [2-41). Although y-carboxylation can be required for the efficient secretion of vitamin K-dependent proteins [5], it is primarily important in calcium-dependent protein/ phospholipid interactions. Recently, Hauschka et al. [6] reviewed evidence that would suggest a role of y-carboxylation in tumor biology. Further, studies with a variety of tumor cells have demonstrated that they do in fact contain vitamin Kdependent carboxylase activity, albeit at low activity compared to normal liver [7-121. In addition, evidence that tumor cells and malignant tissue contain low levels (0.01 to 0.1%) of Gla protein has been presented [6,13]. While nonmalignant tissues and cultured cells also contain carboxylase activity [ 14-191, it is clear, from studies attempting to express the cDNAs for several vitamin K-dependent proteins, that y-carboxylation is not performed efficiently by most continuous cell lines. For example, attempts to produce functional vitamin K-dependent proteins such as recombinant human protein C (HPC) and factor IX in a number of different cell lines [20-261 have resulted in the secretion of partially carboxylated or partially active protein or a mixture of fully active and partially active molecules, even at comparatively low levels of expression. In contrast, we have found that the expression of human protein C in the adenovirus-transformed cell line 293 results in secreted protein that is completely carboxylated and has full

be addressed. 32

Inc. reserved.

ANDBRIAN

CELLULAR

TRANSFORMATION

functional anticoagulant activity, even at high levels of secretion [al, 271. The observation that a fully processed vitamin K-dependent protein could be produced in this virally transformed cell line suggested the possibility that transformation might affect the ability of the cell to correctly y-carboxylate this protein. In the present study, we examined the effect of transformation on the ability of the baby hamster kidney BHK-21 cell line to correctly perform the vitamin K-dependent modification of HPC. Using the adenovirus oncogenes ElA and ElB and expression vectors for recombinant human protein C, we have isolated transformed and nontransformed cell lines secreting rHPC. Through studies on functional properties and calcium dependency, we have found that adenovirus transformation of the BHK-21 cell line dramatically increases the amount of correctly modified HPC secreted from the cell. In addition, we demonstrate that transformation results in a 50% increase in the microsomal carboxylase activity. Overall, our results demonstrate the profound effect cell transformation can have on the vitamin K-dependent processing of a protein, as well as providing a potential means for improving the carboxylation of recombinant vitamin K-dependent proteins for use as therapeutic agents. MATERIALS

AND

METHODS

Reagents and materials. Restriction endonucleases were purchased from Bethesda Research Laboratories and from New England RioLabs, factor X (KabiVitrum) was from Helena Laboratories, and rabbit brain cephalin was from Sigma. Ham’s F12 and Dulhecco’s modified Eagle’s medium were purchased from GIBCO, fetal bovine serum was from HyClone, and bovine serum albumin and human transferrin were purchased from Miles Laboratories. Lilly supplied bovine insulin and hygromycin B (Hy). Methotrexate (MTX) was purchased from Sigma. Vitamin Kl (Aquamephytome) was purchased from Merck Sharp and Dohme. Sheep polyclonal antibody to HPC was from American Diagnostica. Peroxidase-conjugated goat anti-mouse IgG was purchased from Tago and ABTS (2,2’-azinodi(3. ethylbenzothiazoline-sulfonate)) was purchased from Bionetics. Biotinylated antibodies and the ABC detection kit were from Vector. Nitrocellulose membranes were obtained from Schleicher & Schuell. The synthetic peptide FLEEL was from Sigma (P-5523, lot 12535900) All reagents used were of the highest quality and purity. Cell lines. The following cell lines were obtained from the American Type Culture Collection: adenovirus-transformed human kidney 293 (ATCC CRL 1573), monkey kidney MK2 (ATCC CCL 7); and baby hamster kidney BHK-21 (ATCC CCL 10). The adenovirus-induced Syrian hamster tumor cell lines AV12-664 (ATCC CRL 9595), SA7, and SV20 were obtained from Dr. .J. P. Burnett, Lilly Research Laboratories. These three tumor lines had been transformed by human adenovirus type 12, simian adenovirus 7, and simian virus 20, respectively. Each of the cell lines was grown in Dulhecco’s modified Eagles medium (DMEM) supplemented with 10% fetal bovine serum and 10 *g/ml vitamin KI. Construction of expression uectors. The multicistronic expression plasmids were constructed using the cDNA coding sequence isolated previously by Beckmann et al. [28]. The cistrons encoding the hygromycin phosphotransferase (HyPR) and murine dihydrofolate reductase (dhfr) cDNAs were obtained from plasmid pSV2-hyg (211 and

AND -/-CARBOXYLATION pSV2-dhfr (American Type Culture Collection ATCC 37146), respectively. The BL transcriptional unit, composed of the enhancer sequence of the P2 strain of BK virus [29] and the adenovirus type 2 (Ad2) major late promoter, was obtained from pBLP-CAT [30]. The HPC expression vector pLPC-hd, which also contains cistrons for the expression of HyPR and dhfr, was constructed as described previously [21]. DNA transfection and drug selection. One day prior to transfection, cells were plated at a density of lo6 cells/55 cm*. Calcium phosphate-DNA precipitates were prepared [31] with 10 to 50 pg of expression plasmid DNA and no carrier DNA. Four hours after transfection, the culture medium was replaced, and 2 to 3 days after transfection the culture medium was replaced with selection medium containing 200 fig Hy/ml. Clonal cell lines resistant to Hy were isolated 4 to 5 weeks after drug selection was applied. These recomhinant cell lines were expanded, and the presence of HPC in the conditioned culture medium was determined by the ELISA described helow. In the transient expression studies using chloramphenicol acetyltransferase (CAT) expression plasmids, the cells were analyzed for CAT activity as described previously [32]. Adenovirus transformation of BHK-21 cells. Human adenovirus 12 (Ad12, ATCC VR 1089) DNA was isolated from virus grown in HeLa cells as described previously [33]. BHK-21 cells were plated at a density 0.5 X lo6 cells/55 cm’. For viral transformation, BamHI-digested human adenovirus type 12 DNA was used. To obtain Adtransformed rHPC-producing lines, cells were also transfected with the HPC expression vector pLPC-hd [27] including, in addition to the HPC cistron, a cistron for the expression of the HyPR gene. Two days later, the cells were trypsinized and replated in 0.45% agarose (Sigma type VII) containing DMEM, 1% fetal bovine serum, and 200 pg/ml hygromycin B. The resulting hygromycin-resistant clones, capable of growing in the low-serum-containing soft agar, were isolated, expanded, and analyzed for HPC secretion by ELISA. The presence of the adenovirus-transforming region, including the early region 1A (ElA), was determined by Southern blot analysis using the “P-labeled Sal1 C fragment of adenovirus 12 DNA [34]. The procedures for the isolation of total cellular DNA and Southern blot analysis were as previously described [32]. Detection of HPC antigen. A solid-phase sandwich enzyme-linked immunosorhent assay (ELISA), with a sensitivity of 5 rig/ml, was used to measure the level of recombinant HPC (rHPC) antigen in culture media. Ninety-six-well plates were coated with the goat antiHPC polyclonal antibody. The HPC antigen was sandwiched between the polyclonal antibodies and a hiotinylated goat anti-HPC polyclonal antibody. The ELISA was developed with avidin-conjugated horseradish peroxidase and o-phenylenediamine color indicator. Alternatively, the HPC antigen was sandwiched between the polyclonal antibodies and a murine monoclonal antibody designated anti-HPCB. In this case, the ELISA was developed with horseradish peroxidaseconjugated goat anti-murine IgG. Purified plasma-derived HPC was used as standard. There was no difference in the immunoreactivity of purified plasma-derived HPC and the rHPCs. SDS-PAGE and Western blot analysis. Recombinant human protein C was subjected to SDS-polyacrylamide gel electrophoresis [35] on 10% SDSpolyacrylamide gels (acrylamide/bis 30/0.8) under either reducing or nonreducing conditions. Protein was detected by silver staining or by electrophoretic transfer of protein from the gel to a nitrocellulose membrane for Western blot analysis. After transfer to nitrocellulose the membrane was blocked for 1 h in 5% nonfat dry milk in phosphate-buffered saline, pH 7.4, and incubated with a sheep anti-HPC polyclonal antibody followed by biotinylated rabbit antisheep IgG. The blot was developed using the Vector ABC kit and 4-chloro-l-naphthol. Calcium dependency ofrHPC. nation of calcium-dependent recombinant cell lines were

To obtain material for the determielution from anion exchange [36], the grown in a modified mixture of Dul-

34

BERG

becco’s modified Eagle and Ham’s F-12 media containing 1 &ml human insulin, 1 pg/ml human transferrin, and 10 Kg/ml vitamin Kl. Conditioned media were collected, adjusted to a final concentration of 5 mM benzamidine and 4 mM EDTA, pH 7.4, and then adsorbed to an anion-exchange column (Pharmacia Fast Flow Q resin) as described by Yan el al. 1361. After washing with three column volumes of 20 mM Tris, 0.15 M NaCl, 5 mM benzamidine, 2 mM EDTA, pH 7.4, and three column volumes of 20 mM Tris, 0.15 A4 NaCl, 5 mM benzamidine, the bound rHPC was eluted with 20 mA4 Tris, 0.15 A4 NaCI, 10 mM CaCl,, 5 mM benzamidine, pH 7.4. Residual material was eluted in a high salt buffer of 20 mM Tris, 0.5 M NaCl, and 5 mM benzamidine. Functional actiuity of rHPC. The zymogen HPC was converted to the activated protein C (APC) by treatment with immobilized rabbit thrombomodulinbovine thrombin as described previously [37]. The anticoagulant activity of the recombinant APC (rAPC) was measured with an activated partial thromboplastin time clotting assay. The amidolytic activity of the rAPC was determined by the hydrolysis of a tripeptide substrate H-D-Phe-Pip-Arg-p-nitroanilide (Helena Laboratories S-2238) at pH 7.4,25”C. One unit of APC was defined as the amount of APC required to release 1 pmol of p-nit,roanilide in 1 min at 25”C, pH 7.4, using an extinction coefficient for p-nitroanilide at 405 nm of 9620 M-i cm-’ [38]. Plasma-derived APC was purified as previously described [21] and used as a standard for both functional assays. Carboxylasc activity assay. A soluble microsomal fraction from both Adl2-transformed and nontransformed BHK-21 cells was prepared from a crude microsomal suspension as described by Houser et al. (391. Cells were grown in roller culture, removed by trypsin treatment, and washed in Hanks’ balanced salt solution (GIBCO). The washed cell pellets were resuspended in 2 vol of SIK buffer (250 mM sucrose, 25 mi’vf imidazole, 80 mM KCI, pH 7.4). The cells were lysed by freeze/thaw and clarified by centrifugation at 10,OOOg for 10 min at 4°C. The microsomes were isolated from the postmitochondrial supernatant by centrifugation at 105,OOOg for 1 h, then washed in SIK buffer, and resuspended in 1 vol of SIK buffer containing 1.5% Triton X-100. The microsomes were Dounce-homogenized and centrifuged for 1 h at 105,OOOg to remove the insoluble material. The protein content in the soluble microsome fraction was determined using the Bio-Rad protein assay. The carboxylase activity was determined by the incorporation of ‘“CO, into the synthetic FLEEL (Phe-Leu-Glu-Glu-Leu) peptide, essentially as described previously [40]. The reaction contained 2.5 mM FLEEL dissolved in 0.5 ml soluble microsome fraction, 625 pM NADH, 50 PCi [“‘Clbicarbonate, and 10 @g/ml vitamin Kl. After 30 min at 25OC, an equal volume of trichloroacetic acid was added and incubated at 4°C for 15 min, and the trichloroacetic acid-precipitable material was removed by centrifugation at 5OOOg for 10 min. Samples of each supernatant, containing the trichloroacetic acid-soluble FLEEL peptide, were gassed with 100% CO, for 1 to 2 h prior to scintillation counting. RESULTS

Function of rHPC from Several Cell Lines As indicated above, the functional anticoagulant acof HPC depends on the degree of y-carboxylation and we previously demonstrated that rHPC secreted from the Ad-transformed 293 cell line was fully y-carboxylated and had full anticoagulant activity. We have compared the anticoagulant activities of rHPC secreted from several additional adenovirus-transformed or induced tumor lines (AV12-664, SA7, SV20) with those produced from two non-adenovirus-transformed lines tivity

ET

AL.

TABLE Functional

Source

Activities Transformed

of HPC

Plasma derived6 AV12-664 cell line SA7 cell line SV20 cell line MK2 cell line BHK-21 cell line

1

of rHPC Produced and Nontransformed

Viral transformation

+ + + -

from Cell

rHPC expression level (~g/lO” cells)

0.23 0.19 0.15 0.16 0.05

AdenovirusLines Functional anticoagulant activity” NJ/w) 250 280 + 33 (6) 250 (2) 240 (2) 138 & 21 (4) 120 t 23 (11)

a The zymogen HPC was converted to the activated protein C (APC) by treatment with immobilized rabbit thrombomodulin-bovine thrombin as described previously [37]. The APC activity was measured with an activated partial thromboplastin time (APTT) clotting assay 1701. One unit of anticoagulant activity was defined as a equivalent of 4 pg plasma HPC which is present normally in 1 ml of plasma. The number in parentheses is the number of samples analyzed. h Plasma-derived APC was purified as previously described [21].

(MK2, BHK-21). Expression vectors for rHPC were introduced into each of the cell lines as indicated under Material and Methods and stable cell lines secreting rHPC were isolated. As shown in Table 1, the rHPC secreted from the three transformed lines had anticoagulant activity comparable to that of plasma-derived HPC and approximately twice that of the rHPC secreted by the nontransformed cells. While the above data are suggestive, a direct comparison of the same cell before and after transformation would be required to determine if carboxylation potential increases following transformation of the cell. Because BHK-21 cells can be readily transformed by human adenovirus type 12 [34], we chose to use this line for further studies. Adenovirus Transformation

and Expression of HPC

As described under Materials and Methods, viral transformation of the BHK-21 cell line was accomplished using the BamHI restriction fragment of human AdI2, which contains the coding regions for the ElA (and linked early region lB, ElB) viral oncogene [41]. Viral transformants were identified by their ability to grow in low-serum-containing soft agar. Following transfer to culture dishes for attachment-dependent growth, the clones displayed an epitheloid morphology suggestive of adenovirus transformation. Five Adtransformed clones also producing rHPC were isolated as described under Materials and Methods. Two clones, designated BHK-Ad-21 and BHK-Ad-37, were chosen for further study. DNA from both lines was examined by Southern blot analysis using a 32P-labeled probe to the EIA/ElB cod-

CELLULAR

A

BHKsAd-21

W

Ad.12

TRANSFORMATION

B

q BHK

-21

EIBHK

-Ad

23.19.46.6-

2.32.0-

pBL.CAT

1.4-

‘pTPL.CAT’

12345

FIG. 1. Analysis of human adenovirus type 12.transformed BHK-21 cells. (A) Southern blot analysis of chromosomal DNA isolated from a BHK-21 subclone (BHK-Ad-21) transfected with BamHI-digested adenovirus 12 DNA. Total DNA was isolated, digested with the indicated restriction endonucleases, and analyzed using a %-labeled probe encompassing the transforming region of the virus (Ad12 Sal1 C fragment) (lanes 1 and 2). Lane 3, undigested chromosomal DNA. As a hybridization control, Ad12 linear DNA, representing 10 genome equivalents, was analyzed (lanes 4 and 5). (B) BHK-21 cells and an Adl2-transformed derivative (BHK-Ad) were transfected with CAT expression plasmids pBL-CAT and pTPLCAT and, 72 h later, cell extracts were prepared and analyzed for CAT activity as described under Materials and Methods. The results are the average of three determinations. Purified CAT (PL Biochemicals) was used as a control.

ing region to obtain direct proof that the two clones chosen for study were adenovirus transformed. The results with clone BHK-Ad-21 are shown in Fig. 1A. The ElA/ElB probe hybridized to a high-molecular-weight species in undigested chomosomal DNA (lane 3). Several bands were detected with the probe following digestion of the DNA with BumHI and EcoRI (lanes 1 and 2), suggesting several integration sites. Compared to the intensity of the control bands of digested Ad12 DNA (lanes 4 and 5), the minor band at approximately 18 and 6.6 kbp in the BumHI and EcoRI digested samples, respectively (lanes 1 and 2), represented approximately 1 copy of the integrated Ad12 DNA. The two bands of greater intensity in each lane represented approximately 5 to 6 copies, each from a single-site multiple-integration event. A similar analysis confirmed the presence of the Ad12 EIA/ElB coding regions in the BHKAd-37 cell line (data not shown). An experiment to test for functional ElA protein was performed in the two cell lines by analysis of promoter strength from two ElA-inducible promoters [27, 321. For these experiments, plasmids pBL-CAT and pTPL-CAT, in which the ElA-inducible BL and TPL promoters drive expression of the CAT gene, were separately introduced into the nontransformed and Ad-transformed BHK lines. Three davs nest-transfection. -I the level ~- ~~ of -- CAT ----

AND

35

-y-CARBOXYLATION

enzyme activity in each line was determined. As shown in Fig. lB, the level of CAT expression from both of the plasmids was 12- to 15-fold higher in the Adl2-transformed cell line than in the nontransformed BHK line, indicating the presence of functional ElA tumor protein. While it was not feasible to test for the presence of functional ElB, the Southern analyses demonstrated that the ElA and ElB genes had remained linked in the two transformed lines. Overall, the data above demonstrate that the Ad12 DNA-transfected cells used in the study were functionally transformed. Functional Activities and Calcium Dependency of rHPCs from Transformed and Nontransformed BHK Cells The functional activities of the rHPC from the two AdlB-transformed and from two nontransformed cell lines (designated BHK-21-1 and -7) were determined following activation to APC by thrombin/thrombomodulin. The amidolytic activity of APC depends only on the presence of an intact serine protease domain, whereas the anticoagulant activity (activated partial thromboplastin time) depends not only on the presence of the intact serine protease domain but also on the degree of y-carboxylation of glutamate in the N-terminal light chain domain [36, 421. The amidolytic activities of the rAPC from bhe Ad-transformed and nontransformed lines were identical (Table 2). However, the anticoagulant activities of the rAPC from the Adtransformed lines were both significantly higher than those of the rAPC from the nontransformed lines, sug-

TABLE

2

Functional Activity of rHPC Secreted from AdenovirusTransformed and Nontransformed BHK-21 Cells Functional activity (unitslmg)b Source

of HPC”

BHK-21-l BHK-21-7 BHK-Ad-21 BHK-Ad-37

Ad-transformed + +

Anticoagulant

Amidolytic

91 * 17 (2)' 111 * 21 (4) 213 f 18 (10)

22 22 24 22

200

*

36(9)

-t 4 (2) Ik 4 (2) f 3 (5) rk 3 (3)

“Expression levels in serum-free medium were from 100 to 200 rig/ml for BHK-21-l and -7, and 3 to 6 pg/ml for BHK-Ad-21 and -37. b The zymogen HPC was converted to the APC as described in the footnote to Table 1. The amidolytic activity of the rAPC was determined by the hydrolysis of a tripeptide substrate H-D-Phe-Pip-Arg NH-@-NO2 (S-2238) at pH 7.4,25”C. One unit of APC was defined as the amount of APC required to release 1 rmol of p-nitroanilide in 1 min at 25”C, pH 7.4, using an extinction coefficient forp-nitroanilide at 405 nm of 9620 M-i cm-’ [38]. ’ The number in parentheses is the number of independently produced samples assayed in duplicate or triplicate

36

BERG

gesting that the rHPC from the Adl2-transformed lines might have a higher content of Gla. Using the procedure of Yan et al. [36], the calcium-dependency profile of rHPC from the transformed and nontransformed cell lines was examined to determine if the observed differences in functional anticoagulant activity were due to differences in the Gla content. In this procedure, fully y-carboxylated HPC (containing 9 residues of Gla) elutes from an anion-exchange resin in 10 mA4 CaCl, (CaCl, fraction) and incompletely carboxylated HPC (containing 6 to 7 residues of Gla) remains bound but can be subsequently eluted with 0.5 M NaCl (NaCl fraction) [5, 361. As shown below, noncarboxylated or poorly carboxylated protein does not appear to bind to the column under the conditions used. Serumfree conditioned culture medium from the Adl2-transformed and nontransformed BHK cells was adsorbed to the anion-exchange resin, washed, and sequentially eluted as described under Materials and Methods. As shown in Fig. 2A, 50 to 60% of the rHPC from the two

W Flow Through 0 Calcium Fraction El NaCl Fraction

A

BHK.21.1

B

f EiHK-Ad-i1 t BHK.21.7 BHK.Ad.37

100-l

t .E 20 n 0

+K

-K

-K/+K

BHK-Ad.37 FIG. 2. Calcium pseudoaffinity separation of rHPC from Adtransformed and nontransformed BHK-21 cells on anion-exchange chromatography. rHPC in conditioned cell culture medium was adsorbed to Pharmacia Fast Flow Q resin and sequentially eluted with CaCl, and NaCl as indicated under Materials and Methods. The amount of rHPC in each fraction was determined by ELISA. (A) Comparison of the rHPC fractionation profile for the two nontransformed (BHK-21-l and BHK-21-7) and Ad-transformed (BHK-Ad-21 and BHK-Ad-37) cell lines. Each cell line was grown in the presence of 10 kg/ml vitamin Kl. (B) Comparison of the rHPC fractionation profile for BHK-Ad-37 continually grown in the presence (+K) and absence (-K) of vitamin Kl and from the vitamin K-depleted cells supplemented for 24 h with vitamin Kl (-K/+K).

ET

AL.

TABLE Functional

Protease

and

3

Anticoagulant

Activities

of the

Calcium-Dependent rHPCs from Adenovirus-Transformed and Nontransformed BHK-21 Cells Functional activity (units/mg)” Source

of HPC

BHK-21-l BHK-27 -7 BHK-Ad-21 BHK-Ad-37 a See footnote determinations

Ad-transformed

Anticoagulant

-. + t to Table 2. Results of the same sample.

230 280 195 200 are the

averages

Amidolytic 21 24 22 22 of duplicate

nontransformed cell lines examined did not bind to the resin and only a portion of the bound material eluted in the calcium (9 Gla) fraction. In contrast, 94% of the rHPC from the two AdlB-transformed lines bound to the column and virtually all of the bound material eluted in the calcium fraction. Thus, as was suggested by the functional data in Table 2, the Adl2-transformed lines secreted predominantly fully carboxylated rHPC compared to 15 to 26% fully carboxylated material secreted by the nontransformed lines. To ensure that we had fractionated the high Gla material, the functional anticoagulant activity of the calcium fraction was determined. As shown in Table 3, the calcium-eluted rHPC from both Ad-transformed lines gave essentially the same specific activity as that present in the crude conditioned medium (Table 2). This was not unexpected because the majority of the rHPC in the conditioned culture medium eluted in this fraction. The functional activity of the calcium-eluted rHPC from the nontransformed BHK cells was very high, in contrast to the low functionality in the conditioned medium, confirming the separation of high-Glalhigh-specific-activity material. The specific activity of the calcium-eluted rHPC from the nontransformed lines was actually higher than that of the calcium-eluted rHPC from the two Adl2-transformed lines. A possible reason for this is discussed below. For comparative purposes, rHPC from BHK-Ad-37, grown for 30 days in vitamin K-depleted medium to reduce the level of y-carboxylation, also was analyzed. This rHPC had a functional activity of 10 to 20 U/mg, which is 5 to 10% of that obtained in the presence of vitamin K. As shown in Fig. 2B, the rHPC produced from the Adl2-transformed line BHK-Ad-37 grown in the absence of vitamin K exhibited an elution profile similar to that of the material produced from the nontransformed cells grown in the presence of vitamin K; i.e., a high percentage of the total material was present in the flow-through fraction. Addition of vitamin K to

CELLULAR

BHK-Ad- BHK-21. m --37 21 7 1 %

TRANSFORMATION

5n 8N

kDa -97 -68

sci HC ;;I YL

- 43 IX' -

LC-

29

& - 18 123456

FIG. 3. Western blot analysis of rHPC from transformed and nontransformed BHK-21 cells. Samples (100 ng) of the calcium-dependent rHPCs isolated by anion-exchange chromatography from two Ad-transformed and two nontransformed HPC-producing lines (Table 2) were subjected to electrophoresis on reducing 8.75% polyacrylamide-SDS gel and were transferred to nitrocellulose, and the rHPC was detected using a sheep polyclonal antibody against HPC and the Vectastain ABC detection method (Vector Laboratories). Lanes 1 and 2, rHPC from Ad-transformed lines BHK-Ad-37 and -21, respectively. Lanes 3 and 4, rHPC from nontransformed clones BHK-21-7 and -1, respectively. Lane 5, rHPC secreted from the human 293 cell line; lane 6, rHPC secreted from the 293 cell line following treatment with the n-mannosidase inhibitor swainsonine (SW). The locations of the single chain (SC), the (Y, 8, y forms of the heavy chain (HC), and the light chain (LC) are indicated.

the depleted culture resulted in the secretion ofpredominantly calcium-dependent rHPC. These data clearly demonstrate that the observed difference in HPC elution profiles between the Ad-transformed and nontransformed lines is due to a vitamin K-dependent process. Along with the very low functional activity, the data also suggest that the rHPC that does not bind to the resin is very poorly carboxylated. Processing of the rHPC from Transformed and Nontransformed Cells To determine if there were any major differences in the processing of the protein from the Ad-transformed and nontransformed BHK-21 cell lines, samples were analyzed by SDS-PAGE Western blot analysis (Fig. 3). HPC from human plasma is composed primarily of a two-chain disulfide-linked heterodimer consisting of three heterogeneous glycoforms (01,/3, and y) [4,5,43] of the serine protease-containing heavy chain (HC) and the y-carboxylated light chain (LC). Approximately 10% of the HPC remains as single chain. For comparative purposes, we analyzed purified rHPC from the human 293 cell line (lane 5), which has a processing pattern very similar to that of human plasma-derived HPC [21]. rHPC produced in the presence of the glycosylation inhibitor swainsonine was included as a control for

AND

37

+ARBOXYLATION

reduced carbohydrate heterogeneity (lane 6). The migration patterns of the different HPC peptides were identical between samples from the Adl2-transformed and nontransformed BHK-21 cell lines. In addition, there were no observable differences between the migration pattern of rHPC from the crude culture medium and the ion-exchange fractions (data not shown). By densitometric scanning, the relative amounts of the a and 0 glycoforms were the same before and after purification; however, the amount of the a-form HPC from the nontransformed lines was slightly higher, suggesting that transformation affects the glycosylation of the protein (Table 4). In the crude culture medium, the amount of the single-chain protein was the same between cell lines, and although the amount of the singlechain protein was reduced following purification, the amount present was identical between the cell lines. Thus, adenovirus transformation does not appear to significantly modify the overall proteolytic processing of the protein. The migration of each of the HPC chains from each of the BHK cell lines was slightly slower than that from the 293 cell, consistent with the observation that the overall carbohydrate content of HPC from BHK cells (16% carbohydrate by weight) is higher than that of HPC from 293 cells (13% carbohydrate by weight) [36]. Effect of Viral Transformation Carboxylase Activity

on Endogenous

One possible explanation for the increase in efficiency of y-carboxylation by the virally transformed cells TABLE

4

Processing Patterns of Purified and Nonpurified rHPC Secreted from Transformed and Nontransformed BHK-21 Cells Percentage of total rHPC heavy chain” Sample BHK-21-l Crude Purified’ BHK-12-7 Crude Purified BHK-Ad-21 Crude Purified BHK-Ad-37 Crude Purified

Single chain (% of total)

a form

p form

30 17 * 2

72 71 f 3

28 29 f 3

28 k 3 15

72 -+ 2 71

29 -+ 2 28

30 k 4 17 t 3

61 f 5 61 I 3

39 f 5 40 f 3

33 17

59 60

41 40

0 Determined from duplicate densitometric scans of Western blots. The percentages k the SEM are given for independent samples analyzed three or more times. b Samples were purified by anion-exchange chromatography as described under Materials and Methods.

38

BERG ET AL.

would be an increase in the cell’s carboxylase activity. To test this possibility, we compared the ability of microsomal fractions from Ad-transformed and nontransformed BHK-21 cells to incorporate 14C0, into the synthetic peptide FLEEL, a substrate for y-carboxylation. The level of 14C0, incorporation into the synthetic substrate was 790 ? 30 dpm/mgprotein (n = 4) using microsomal extracts from the nontransformed BHK-21 cell line. However, the levels of 14C0, incorporation using microsomal protein from two independent viral transformants were 1230 t 30 and 1100 + 40, representing a 40 to 55% increase in the level of carboxylase activity. Background levels without FLEEL were less than 50 dpm/mg. Although we routinely used the quinone form of vitamin K along with NADH to supply reducing equivalents for microsomal reductase activities, with this assay there was no significant difference in the measured levels of carboxylation compared to that obtained with the hydroquinone. Thus, the observed difference in 14C0, incorporation between transformed and nontransformed cells was not due to differences in microsomal reductase activity. In short, Ad transformation appears to increase the level of carboxylase activity in BHK-21 cells. The level of 14C0, incorporation into endogenous substrate (TCA precipitable material) was too low and variable to accurately determine if propeptide-containing intracellular material increased in the nontransformed cells. DISCUSSION

The relationships among the coagulation cascade, ycarboxylation by tumor cells, and tumorigenesis have been recently reviewed [6]. Anticoagulants, such as warfarin, have been demonstrated to inhibit both the growth and metastatic activities of a variety of tumor cells (see [6, 44, 451 for references). These effects are thought to be due primarily to the indirect effect of limiting fibrin generation that is induced by the procoagulant activities elicited by tumor cells, e.g., tissue factor [46,47], cathepsin B-like soluble thiol protease [48-531, and factor XIII-like fibrin stabilizing activity [54]. However, direct effects on tumor cells have been reported following inhibition of vitamin K-dependent carboxylation, including inhibition of growth, mitosis, and cell motility [55-571. Interestingly, the activity of the procoagulant-soluble thiol protease activity is reduced by warfarin [50], as is the secretion of tumor procathepsin B [58], suggesting some involvement of the vitamin Kdependent pathways in its production. Hilgard et al. [59] suggested that warfarin’s antimetastatic activity was a direct effect on the tumor cells and demonstrated a vitamin K-dependence for metastasis [60, 611. In addition, Hauschka et al. [6] have suggested that the reduced calcium dependence for the growth of tumor cells may be related to a vitamin K-dependent protein involved in

calcium transport. Taken together, these data suggest that the carboxylation potential of tumor cells may be important in tumor growth and metastasis. In this paper we have directly examined the effect of cellular transformation on the ability of cells to perform the complex post-translational modification of y-carboxylation of glutamate. Using the vitamin K-dependent factor HPC, our data demonstrate that transformation of cells does in fact increase their ability to y-carboxylate and lend credence to the suggested involvement of this modification in tumor biology. We show that following viral transformation, the amount of fully y-carboxylated rHPC secreted from the cell increases from -20 to -90% of the total. Although the majority of the rHPC secreted from the nontransformed cells was not the 9 Gla, calcium-dependent protein, the small amount of material that was calcium-dependent actually had a higher functional anticoagulant activity than the calcium-dependent rHPC from the transformed cell line (Table 3). Studies by Hau and Salem [62] and Yan et al. [36] have suggested that an increase in the degree of sialylation of HPC would decrease the specific activity of the protein. As shown in Table 4, the carbohydrate processing of the rHPC was slightly different following transformation. Possibly, the rHPC from the Adl2-transformed cells, in addition to being fully carboxylated, is also more heavily sialylated, slightly reducing its functional activity from the expected 250 U/mg for plasma HPC. In support of this, the G glycoprotein of vesicular stomatitis virus has been shown to have increased amounts of sialylated hybrid oligosaccharides when produced from BHK-21 cells transformed with Rous sarcoma virus [63]. Because of the low yields in the expression and purification of the calcium-dependent material from the nontransformed cell lines, we are unable at this time to purify sufficient amounts for comparative carbohydrate analyses. However, using previously reported methods [36], we have found that the sialic acid content of the rHPC from the 293 cell line, which has a very high functional activity (-325 U/mg), is approximately half that in the purified rHPC from the BHK-Ad-37 cell line. The level of rHPC secreted by the Ad-transformed cultures was from 10 to 30 times higher than that from the nontransformed cell (Table 2). Although we have not performed a detailed analysis of transcription, translation, and secretion rates, the reason for the difference in expression level is likely due to a combination of (1) the promoter activation by the ElA tumor protein in the transformed cell and (2) a lower secretion rate for the rHPC in the nontransformed line. Previously, we reported that the rate of rHPC secretion from the human 293 cell line dropped 5 to lo-fold if y-carboxylation of the protein was reduced, either by growth in vitamin K-depleted medium or following treatment of cells with chloro-K, a competitive vitamin K analog [5]. Similarly,

CELLULAR

TRANSFORMATION

we found that the level of rHPC secreted from BHKwhen the cells were grown in Ad-37 dropped -lo-fold vitamin K-depleted medium (data not shown), concomitant with the drop in the degree of carboxylation (Fig. 2). Thus, both the level of secretion and the degree of y-carboxylation of rHPC from the nontransformed cells in the presence of vitamin K were essentially the same as those observed in rHPC from the transformed cells grown under vitamin K-depleted conditions. We do not believe that the higher expression level from the transformed cells in any way could have resulted in the observed increase in the cell’s carboxylation potential. Overexpression of the protein would more likely have resulted in saturation of the carboxylase machinery and a reduction, not an increase, in the amount of fully carboxylated protein. This in fact has been observed with prothrombin, where increasing levels of secretion above 100 ng/106 cells/day from recombinant CHO cells resulted in a dramatic reduction in the y-carboxylation of the protein [64]. It has been reported that conditions reducing the cell’s ability to y-carboxylate (such as warfarin treatment) can result in a slight increase in carboxylase activity presumably due to substrate induction [65], and increased accumulation of propeptide-containing intracellular substrate could stimulate carboxylation as propeptide has been shown to stimulate carboxylation of FLEEL [66, 671. If the reduced ability of the nontransformed lines to y-carboxylate rHPC resulted in intracellular substrate accumulation, we may actually have underestimated the potential increase in microsomal carboxylase activity as a result of transformation. Although we have not observed any increase in total intracellular HPC antigen by ELISA, e.g., following vitamin K depletion, we would not have detected extensively degraded material that might serve as substrate. It is unclear whether the observed -50% increase in carboxylase activity following transformation is sufficient to account for the elevated amounts of fully y-carboxylated rHPC secreted from the transformed lines. In fact, a high level of carboxylase activity in a cell line does not necessarily indicate an ability to efficiently carboxylate and secrete a Gla-containing protein. For example, under our assay conditions human liver cell line HepG2 has a relatively high level of carboxylase activity (4320 dpm/mg). However, factor IX [24] and HPC (S. P. Little, M. T. Lai, N. U. Bang, and S. B. Yan, unpublished observation) produced in this line do not appear to be fully y-carboxylated based on lower than expected functional activities. Thus, other factors in the secretion pathway of these Gla-containing proteins may affect the degree of fully modified protein secreted from the cell. Cellular transformation may be affecting processes such as ER transport and trafficking. However, for the BHK-Ad cell lines the increase in carboxylase activity is consistent with the increase in the carboxyl-

AND

39

+ARBOXYLATION

ation of secreted HPC and offers a plausible explanation for our results. The mechanism for the increased carboxylase activity following viral transformation is not clear. The adenovirus ElA tumor protein is known to both activate and repress the transcription of several genes (see [68, 691) and we speculated that the increase in carboxylase activity could be due to ElA-dependent gene activation. However, we have found that the transient expression of the ElA gene product in the nontransformed HPC-producing cells does not result in an increase in the amount of fully carboxylated protein (data not shown). Thus, the observed effect on HPC modification appears to require the transformation of the cell and cannot be explained by ElA-induced gene act.ivation. The ability to modulate a cell capacity to perform the complex post-translational modification of y-carboxylation will make it possible to improve the functionality of recombinant proteins, such as human protein C, protein S, factor IX, and factor VII. Producing a fully modified and fully functional protein would certainly be desirable not only for studies of the processing, secretion, and modification of these complex proteins, but also for their potential use as therapeutic agents. We gratefully for HPC assay

thank JoAnn support.

Henry,

John

Ivancic,

and Yolanda

Roe

REFERENCES 1.

Suttie, search,

2.

Stenflo, J., and S&tie, J. W. (1977) Annu. Reu. Biochem. 157-172. Suttie, J. W. (1985) Annu. Reu. Biochem. 54,459-477.

13 . 4. 5.

6. 7. 8. 9. 10. 11. 12. 13. 14.

Esmon,

J. W. (1988) in Current Advances p. 529, Elsevier, New York.

C. T. (1987)

Science

235,

in Vitamin

K Re46,

1348-1352.

Grinnell, B. W., Walls, J. D., Berg, B. T., Boston, J., McClure, D. B., and Yan, S. B. (1989) in Genetics and Molecular Biology of Industrial Microorganisms (Hershberger, C. L., Queener, S. W., and Hegeman, G., Eds.), pp. 226-237, Amer. Sot. Microbiol., Washington, DC. Hauschka, P. V., Haroon, Y., Buchthal, S. D., and Bell, R. G. (1986) Huemostasis 16, 273-287. Bell, R. G., Buchthal, S. D., McAllister, C. G., and Laux, D. C. (1983) ASBC Abstract, l/18/83. Buchthal, S. D., McAllister, C. G., Laux, (1982) Rio&em. Biophys. Res. Commun.

D. C., and Bell,

R. G.

109, 52-62.

Taverso, H. P., Hauschka, P. V., and Gallop, P. M. (1980) C&if. Tissue Int. 30, 73-76. Nishimoto, S. K., and Price, P. A. (1980) J. Biol. Chem. 255, 6579-6583. Shoyab, M., and Todary, G. J. (1980) J. Biol. Chem. 25, 87358739. Gray, P. N.. Delaney, R., and Johnson, B. C. (1979) XIth Znt. Con@. Biochem., 666. Hauschka, P. V. (1977) Anal. Biochem. 80, 212-223. Buchthal, 1082.

S. D., and Bell,

R. G. (1983)

Biochemistry

22,

1077-

BERG 15.

Friedman, P. A., Hauschka, P. V., Shia, M. A., and Wallace, (1979) Biochem. Biophys. Acta 583,261-265.

16. Hauschka, P. M.

(1976)

P. V., Friedman, P. A., Traverso, Biochem. Biophys. Res. Commun.

17. Hauschka, 18.

P. V., Frenkel, J., DeMuth, (1983) J. Biol. Chem. 258, 176-182. Lian, J. B., andFriedman, P. A. (1978)

J. P., and Gallop, 71, 1207-1213.

R., and Gundberg, J. Biol.

J. K.

Chem.

253,6623-

6626.

Oppenheimer, Vitamin-K New York.

C., and Wydro, Research (Suttie,

Walls,

J., and Yan,

S. B. (1987)

R. (1988) in Current Advances in J. W., Ed.) pp. 165-171, Elsevier,

41.

A., Cazenave, J., Tolstoshev, (London) 316,268-270.

P., and Lecocq,

J. (1985)

Nature

Anderson, W. (1983)

S. B., and Grinnell,

B. W. (1989)

81, 139-149.

28. Beckmann,

R. J., Schmidt, G. R., and Long,

Crabtree,

R. J., Santerre, R. F., Plutzsky, G. L. (1985) Nucleic Acids Res.

J.,

13,

5233-5247. 29. Berg, D. T., Walls, J., and Grinnell, B. W. (1988) Nucleic Acids Res. 16, 9057. 30. Grinnell, B. W., Berg, D. T., and Walls, J. D. (1988) Mol. Cell. Biol.

8,3448-3457.

31. Wigler, Chasin,

M., Pellicer, A., Silverstein, L. (1979) Proc. N&Z. Acad.

S., Axel, R., Urlaub, G., and Sci. USA 76, 1373-1376.

32. Grinnell, B. W., Berg, D. T., and Walls, J. (1986) Mol. Cell. Biol. 6,3596-3605. 33. Kuchler, R. J. (1977) Biochemical Methods in Cell Culture and Virology, Stroudsberg.

pp.

117-240,

Dowden,

Hutchingson

and

N. and Donati, M. B. (1981) in Malignancy and the Hemostatic System (Donati, M. B., and Garattini, D., Eds.) pp. 65-81, Raven Press, New York.

48. Curatolo,

J. S. (1984) in The Adenoviruses (Ginsberg, Ed.), pp. 35-125, Plenum, New York. Laemmli, U. K. (1970) Nature (London) 227,680-685.

35. 36. Yan, S. C. B., Razzano, McClure, 655-661.

37. Vagano A. (1986)

P., Chao, Y. B., Walls, J. D., Berg, D. B., and Grinnell, B. W. (1990) BiolTechnology

D’Angelo, J. Clin.

S., Comp, P. D., Esmon, Inuest. 77, 416-425.

38. Pfleiderer,

39.

50.

Ross,

53.

40. Suttie,

H. S.,

D. T., 8,

(Perlmann, Academic G. R., and

J. W., Hoskins, J. A., Engleke, J., Hopfgartner, A., Ehrlich, H., Bang, N. U., Belagaje, R. M., Schoner, B., and Long, G. L. (1987) Proc. Natl. Acad. Sci. USA 84,634-638. May 16, 1990 version received

August

20, 1990

Mort, J. S., Recklies, A. D., and Poole, A. R. (1980) Biochim. Biophys.Acta 416,134-143. Poggi, A., Colucci, M., Delaini, F., Semerao, N., and Donati, M. B. (1980) Eur. J. Cancer 16,1641-1642. Recklies, A. D., and Mort, J. S. (1985) Cancer Res. 45, 2302% Sloane, B. F., Dunne, 1151-1153. Sloane,

J. R., and Honn,

B. F., and Honn,

K. V. (1984)

K. V. (1981) Cancer

212,

Science

Metastasis

3,249-

(1966)

Biochem.

263. 54.

Laki, K., Tyler, H. M., and Yancey, Biophys. Res. Commun. 24, 776-781.

55.

Kirsch, W. M., Schultz, D., Van Buskirk, (1974) J. Med. 5, 69-82. Lisnell, A., and Mellgren, J. (1963) Stand. 57,145.

56.

57. Thornes,

S. T.

J. J., and Young, Acta

R. D., Edlow, D. W., and Wood, J. 123,305-316.

kinsMed.

Puthol.

S. (1968)

H. E.

Micorbiol. Johns

Hop-

58. Hauschka, P. V. (1984) J. Cell Biol. 99, 87a. 59. Hilgard, P. (1981) in Malignancy and the Hemostatic 60. 61.

(Donati, M. Press, New Hilgard, P. Hilgard, P., C. G. (1977) Hau, L., and

System B., and Garattini, D., Eds.) pp. 103-111 Raven York. (1977) Brit. J. Cancer 35,891-892. Schulte, H., Wetzig, G., Schmitt, G., and Schmidt, Brit. J. Cancer 35, 78-85. Salem, H. H. (1988) Thromb. Haemostasis 60,267-

270. 63. Hunt, L. A. (1987) Biochim. Biophys. Acta 924,175-184. M., Cantor, A., Furie, B., and Furie, B. (1987) 64. Jorgensen, Chem.

65.

C. T. and D’Angelo,

G. (1970) in Methods in Enzymology G. E., and Lorand, L., Eds.), Vol. 19, pp. 514-521, Press, San Diego. Houser, R. M., Carey, D. J., Dus, K. M., Marshall, Olson, R. E. (1977) FEBS Lett. 75, 226-230.

Received Revised

L., Colucci, M., Cambini, A. L., Poggi, A., Morasca, L., M. B., and Semerao, N. (1979) Brit. J. Cancer 40, 228-

233. 49.

62.

34. Sussenbach,

G. L., and DeWolf,

2307.

A., Joseph, M., Halfpap, L., Insely, M., K., and Woodbury, P. (1985) Nature (LonD. T., Yan,

D., Harvey, V. S., Bach, R., Davis, Cancer Res. 43, 4334-4342.

47. Semeraro,

26. Busby,

J. D., Berg,

H. S., Ed.)

43. Kisiel, W. (1979) J. Clin. Inuest. 64, 761-769. 44. Rickles, F. R., and Hancock, W. W. (1988) Semin. Thromb. Hemostusis 14, 126-132. 45. Rickles, F. R., Wayne, W., Hancock, W. W., Edwards, R. L., and Zacharski, L. R. (1988) Semin. Thromb. Hemostatis 14, 88-94. 46. Dvorak, H. F., Van DeWater, L., Bitzer, A. M., Dvorak, A. M.,

52.

S., Kumar, Berkner, K., Kurachi, don) 316,271-273.

(Ginsberg,

J. E., and Griffin, J. H. (1983) in Progress in Hematology (Brown, E. D., Ed.) pp. 265-278 Grune & Stratton, New York.

51.

R. J., Wasley, L. C., Furie, B. C., and Shoemaker, C. B. (1986) J. Biol. Chem. 261,9622-9628.

in The Adenoviruses New York.

42. Gardner,

25. Kaufman,

Gene

Graham, F. L. (1984) pp. 339-399, Plenum,

Donati,

23. Anson, D. S., Austen, D. E. G., and Brownlee, G. G. (1985) Nature (London) 315,683-685. 24. de la Salle, H., Altenburger, W., Elkaim, R., Dott, K., Dieterle’,

27. Walls,

AL.

C. M.

Traverso, R. D., Hauschka, P. V., and Gallop, P. V. (1979) in Vitamin K-Metabolism and Vitamin K Dependent Proteins. (Suttie, J. W., Ed.) pp. 311-314, University Park Press, Baltimore. Suttie, J. W. (1986) Thromb. Res. 44, 1299134. Grinnell, B. W., Berg, D. T., BiolTechnology 5,1189-1192.

ET

J. Biol.

262,6729-6734.

Vermeer, C., and Soute, B. A. M. (1988) in Current Advances in Vitamin-K Research. (Suttie, J. W., Ed.) pp. 25-39, Elsevier, New York.

J. E., and Suttie, J. W. (1987) J. Biol. Chem. 262, 66. Knobloch, 15,334-15,337. C., and 67. Ulrich, M. M. W., Furie, B., Jacobs, M. R., Vermeer, Furie,

B. C. (1988)

68. Friedman, 305. 69.

J. Biol.

Chem.

D. J., and Ricciardi,

263,9697-9702. R. P. (1988)

Hurst,

H. C., and Jones,

N. C. (1987)

70. Comp,

P. C., and Esmon,

C. T. (1979)

Virology

165,303-

Genes Dev.

1, 1132-1146.

Blood

1272-1278.

54,