Expression of Human Recombinant Plasminogen Activators Enhances ...

MOLECULAR AND CELLULAR BIOLOGY, May 1989, p. 2133-2141 0270-7306/89/052133-09$02.00/0 Copyright © 1989, American Society for Microbiology

Vol. 9, No. 5

Expression of Human Recombinant Plasminogen Activators Enhances Invasion and Experimental Metastasis of H-ras-Transformed NIH 3T3 Cells JONATHAN H. AXELROD,1 REUVEN REICH,2 AND RUTH MISKIN1*

Department

Weizmann Institute of Science, Rehovot 76100, Israel,' and Laboratory of Developmental Biology and Anomalies, National Institute of Dental Research, Bethesda, Maryland 208922

of Biochemistry,

Received 9 November 1988/Accepted 8 February 1989

The gene transfer technique was used to examine the role of plasminogen activator (PA) in the invasive and metastatic behavior of tumorigenic cells. H-ras-transformed NIH 3T3 clonal cells producing a very low level of PA were generated and further transfected with an expression plasmid containing a cDNA sequence encoding either the urokinase-type or the tissue-type human PA. Compared with the parental transformed cells, clonal cells expressing high levels of both types of recombinant PA invaded more rapidly through a basement membrane reconstituted in vitro. Furthermore, cells expressing high levels of recombinant urokinase-type PA also caused a higher incidence of pulmonary metastatic lesions after intravenous injection into nude mice. Both activities were reduced by the serine proteinase inhibitor EACA; invasion was also suppressed by antibodies blocking the activity of human PAs and by the synthetic collagenase inhibitor SC-44463. These findings provide direct genetic evidence for a causal role of PA in invasive and metastatic activities.

Tumor cell invasion and metastasis are complex processes affected by a multiplicity of factors whose molecular nature is scarcely defined (13, 24, 44). Hydrolytic enzymes, including proteinases, were long ago implicated in tumor metastasis to account in part for the ability of tumor cells to detach from the primary lesion, to penetrate through basement membrane surrounding blood vessels, and finally to implant themselves into a remote organ (for reviews, see references 8 and 33). One of the enzymes whose role in metastasis is most controversial is plasminogen activator (PA), a serine proteinase that converts the ubiquitous extracellular zymogen plasminogen into the trypsinlike proteinase plasmin. Plasmin, in turn, dissolves the fibrin network of the blood clot, degrades interstitial glycoproteins such as fibronectin and laminin, and converts procollagenases into collagenases necessary for degradation of basement membrane collagen (8). Such a spectrum of activities renders the plasminogen activation system an ideal candidate to participate in tumor invasion and metastasis, since high levels of PA are closely associated with neoplasia-related phenomena (8, 41). The hypothesis concerning the role of the plasminogen activation system in tumor invasion has recently been supported by studies measuring cellular invasion in vitro in assay systems designed to simplify study of tumor cell invasion and metastasis (30, 42). These studies have demonstrated that the passage of tumor cells through a barrier of basement membrane in vitro requires a proteolytic cascade of which the final proteinase collagenase is generated by PA-activated plasmin. That PA may also be involved in tumor invasion under physiological conditions is suggested by the immunocytochemical localization of urokinase-type PA (uPA) in areas of invasive growth of the Lewis lung carcinoma (47). Furthermore, the metastatic spread of the Hep-3 human carcinoma in an avian system was significantly inhibited by antibodies specifically blocking the activity of human uPA without affecting either the local growth of the tumor or the avian uPA activity (37, 38). Similarly, inhibition by antibod*

ies of surface-localized uPA on B16 melanoma cells reduced the capacity of the cells to generate experimental metastasis in mice (20). In contrast to the latter cases, in numerous other studies the role of PA in tumor metastasis was inferred from circumstantial rather than direct evidence (for reviews and detailed lists of references, see references 6, 8, 14, 22, 26, 34, 40, and 43). For example, in sublines derived from transplanted tumors, such as the B16 melanoma and the Lewis lung carcinoma, comparison was made between PA production and metastatic capacity (6, 9, 14, 40, 54). Comparison was also made between PA levels of spontaneously arising human tumors of the colon, breast, and prostate and PA produced by their invasive and metastatic derivatives (5, 22, 26, 27, 36). Taken together, these studies have yielded inconsistent results, and their validity suffers from their correlative nature. To examine directly whether PA plays a causal role in tumor cell invasion and metastasis, we used in this study the gene transfer technique. We generated H-ras-transformed NIH 3T3 cells expressing a very low level of PA and further transfected into them a recombinant gene encoding one of the two known molecular types of human PA, uPA or tissue-type PA (tPA). We report here on (i) the effect of high expression of recombinant PA on the invasion of cells through a basement membrane reconstituted in vitro and (ii) experimental metastasis after intravenous injection of cells into nude mice. MATERIALS AND METHODS

Cloning of full-length cDNA encoding human uPA. Five oligonucleotides encoding selected peptides along the uPA protein (17, 51) were chemically synthesized (7) to serve as specific probes to monitor the enrichment of uPA mRNA and to select for clones containing uPA-coding sequences. The oligomers were 14 to 17 nucleotides in length and were synthesized as unique sequences or in pools containing several oligomers. The oligomers were labeled with [y-32P] ATP (5,000 Ci/mmol; Amersham Corp., Arlington Heights, Ill.) and T4 polynucleotide kinase (Amersham) (28) to a

Corresponding author. 2133

2134

AXELROD ET AL.

specific activity of approximately 1012 cpm/,mol and purified by DE-52 (Whatman, Inc., Clifton, N.J.) ion-exchange chromatography. Poly(A)-containing RNA was prepared from the urokinase-rich HEp-3 human epidermoid carcinoma propagated in nude mice and was enriched for uPA about threefold on two successive sucrose gradients before use for cDNA synthesis. Synthesis of double-stranded cDNA and construction of a cDNA library in bacteriophage XgtlO were performed as described by Huynh et al. (21) except that the cDNA was synthesized according to Maniatis et al. (25). Recombinant phage (about 2.5 x i05 PFU) were grown in Escherichia coli POP13 (hifA) and screened on replicate nitrocellulose filters with 32P-labeled probes. A total of 36 positive clones were obtained. Recombinant phage that hybridized with more than one probe were plaque purified and further characterized by restriction mapping and DNA sequence analysis (28). A full-length cDNA clone was constructed by ligations of two partial clones through a common BamHI site. The clone consisted of 2,296 nucleotides: 69 noncoding nucleotides at the 5' end, 1,296 coding nucleotides, and 931 noncoding nucleotides at the 3' end, followed by a poly(A) tail of more than 80 residues. The coding sequence started with 60 base pairs encoding the 20 amino acids making up the signal peptide of prepro-urokinase (51) and was followed by the sequence encoding the entire uPA protein, which is in complete concordance with the published amino acid sequence (17, 51). This cDNA was ligated into the SmaI-EcoRI site of plasmid pUN121 (35), yielding a plasmid designated pcUK176. Construction of the uPA cDNA expression plasmid pcUKLTR6. To drive transcription of the recombinant human uPA gene, a 1,300-base pair SmaI-SmaI fragment containing the enhancer-promoter element and the start of transcription from the Harvey sarcoma virus 5' long terminal repeat (LTR) (10) was purified by agarose gel electrophoresis and inserted, by using T4 DNA ligase (25), at a unique SmaI site located at the immediate 5' end of the uPA cDNA insert of plasmid pcUK176. The resulting construct, designated pcUKLTR6, is shown in Fig. 3. Construction of the tPA cDNA expression plasmid pTcHSV. Plasmid pTcHSV contained the entire tPA cDNA sequence as previously published by Pennica et al. (39) and was constructed from four fragments (see Fig. 4). The bulk of the tPA-coding sequences located in a 1,573-base pair BstXIBglII fragment were taken from phage clone ATcll, and the 5' noncoding sequence and 400 base pairs of the coding region up to the BstXI site were taken from phage clone XTc37. To drive transcription of the tPA gene, a HindIIIBamHI fragment containing the Harvey sarcoma 5' LTR promoter-enhancer sequence was inserted upstream of the cDNA into the HindIII-SmaI-restricted plasmid pLSV (23). All fragments were purified by agarose gel electrophoresis before ligation. DNA transfection. For transfection, 106 cells were plated in 9-cm-diameter dishes with Dulbecco modified Eagle medium (DMEM; GIBCO Laboratories, Grand Island, N.Y.) supplemented with 10% heat-inactivated fetal calf serum (Kibbutz Beth Aemek, Maale Hagalil, Israel) and incubated for about 16 h. DNA transfection was performed by the calcium phosphate technique essentially as described previously (15). Briefly, 10 ,ug of plasmid DNA and 1 p,g of plasmid DNA containing the selectable marker pSV2neo (50) or pSVgpt (32) were placed in 62 ,ul of 2 M CaCl2, and 1 ml of transfection buffer was added. A fine precipitate was formed at room temperature within 3 to 6 min, and the mixture was placed on the cells for 6 h at 37°C and 5% CO2.

MOL. CELL. BIOL.

The medium was then removed, and the cells were subjected to glycerol shock for 1 min with 5 ml of phosphate-buffered saline (PBS) containing 10% glycerol. The cells were then washed twice with 5 ml of PBS and returned to standard culture conditions. After 2 days, the cells were split 1:5 and the medium was replaced by selective medium supplemented with G418 sulfate (600 ,g/ml; GIBCO) for cells transfected with pSV2neo or with mycophenolic acid (2 jig/ml), xanthine (150 ,ug/ml), and hypoxanthine (15 ,ug/ml) for cells transfected with pSV2gpt. Growing clones were detected approximately 21 days after transfection. Cell cultures. All eucaryotic cells were grown and maintained in DMEM supplemented with glutamine (4 mM), sodium pyruvate (1 mM), penicillin (2,000 U/ml), streptomycin (100 ,ug/ml), and 10% heat-inactivated fetal calf serum at 37°C in a humidified atmosphere of 5% CO2. Cells cloned by using the selectable marker pSV2neo were maintained with G418 sulfate (600 ,ug/ml). Cells were routinely grown to subconfluency and passaged twice a week. Establishment of H-ras-transformed fibroblasts. NIH 3T3 fibroblasts were transfected, along with the selectable marker pSV2gpt (32), with plasmid pEJ6.6, containing the activated form of the H-ras oncogene as cloned from the bladder carcinoma cell line EJ (45). Transformed clones resistant to mycophenolic acid were isolated, and one clone, ras 6, was selected for further experiments. Assays of PA. To determine PA production, 2 x 105 cells were seeded in 3-cm-diameter dishes and incubated for 24 h. Medium was then changed to DMEM containing 0.5 mg of bovine serum albumin (Sigma Chemical Co., St. Louis, Mo.) previously acid treated (52) to inactivate inhibitors of proteinases. After 24 h, conditioned medium and cells were collected and tested for PA. Quantitative determination of PA activity was performed in a fibrinolytic assay as previously described (49). Briefly, the assay was performed in multiwell plates coated with insoluble "25I-labeled fibrinogen and containing purified human plasminogen (20 ,ug/ml). The ability of plasmin, generated from plasminogen by PA, to release soluble '25I-labeled fibrin degradation products was measured. Results are expressed as international units (International Laboratory for Biological Standards, World Health Organization, London, England) of uPA accumulated in the growth medium during 24 h and calculated per milligram of cellular protein in the culture. Protein was determined by the method of Bradford (3). Control assays in which plasminogen was omitted were performed for all cellular clones and yielded negative results, which indicated an absolute dependence of fibrinolysis on plasminogen. Zymographic analysis of PA was performed in sodium dodecyl sulfate-polyacrylamide minigels as previously described (31). Briefly, samples of conditioned medium were electrophoresed through sodium dodecyl sulfate (Bio-Rad Laboratories, Richmond, Calif.)-polyacrylamide gels containing plasminogen and casein as substrates for PA and plasmin, respectively. After electrophoretic separation, the gel was washed with 2.5% Triton X-100 (Fisher Scientific Co., Pittsburgh, Pa.) to remove the sodium dodecyl sulfate, incubated for 3 h at 37°C to allow proteolysis, and stained with Coomassie brilliant blue (G-250; Bio-Rad). On the darkly stained casein background, PA activity was visualized as clear bands in which the casein was degraded. No bands could be detected in control gels run in the absence of plasminogen. Invasion assay. The invasion assay was performed as previously described (1, 42). Briefly, polyvinylpyrrolidonefree polycarbonate filters (Nuclepore Corp., Pleasanton,

PLASMINOGEN ACTIVATORS ENHANCE INVASION AND METASTASIS

VOL. 9, 1989

Calif.) of 8-p.m-diameter pore size (smaller than the cells) were coated with an extract of basement membrane components (Matrigel; 12.5 or 25 ,ug per filter, as indicated) and placed in modified Boyden chambers. The cells to be studied were collected by short exposure to EDTA (1 mM) and suspended in DMEM containing 0.1% bovine serum albumin; then 2 x 105 cells were placed in the upper compartment of each Boyden chamber. Fibronectin (10 ,ug/ml of DMEM), purified as described previously (11), was placed in the lower compartment to serve as a chemoattractant. After incubation for the indicated times, the cells on the lower surface of the filter were stained and quantitated with an image analyzer (Optomax IV) attached to an Olympus CK2 microscope. Data are expressed as the area (square micrometers per high-power field) of the bottom surface of the filter occupied by the cells, which is proportional to the number of cells on this surface. When indicated, c-amino caproic acid (EACA; 50 p.g/ml; Aldrich Chemical Co., Inc., Milwaukee, Wis.), the collagenase inhibitor SC-44463 (50 ,ug/ml; G. D. Searle and Co., Chicago, Ill.), or the indicated sera were added to the upper compartments of the Boyden chambers 30 min before the addition of cells. Collagenase assay. Collagenase type IV activity was measured by using a modified solid-phase radioassay (29) as previously described (42). Briefly, collagen type IV extracted from Engelbeth-Holm-Swarm tumors was iodinated by the Bolton-Hunter method, and a solution of the labeled collagen was applied to microdilution plates (Removawell; Dynatech Laboratories, Inc., Alexandria, Va.) and allowed to bind overnight. Medium samples (100 R1) were taken at 6 h from the upper parts of the Boyden chambers used for invasion experiments and added to the microdilution plates together with inhibitors of serine proteinases. The samples were incubated at 37°C, and the amount of labeled collagen released from the solid phase was measured after 24 h. Experimental metastasis. Cells of subconfluent cultures were lightly trypsinized (0.125% trypsin; GIBCO), suspended in DMEM with 10% fetal calf serum, spun down, suspended in PBS, counted for viable cells by trypan blue exclusion, and adjusted to the necessary concentration. When indicated, cells were harvested by EDTA treatment as follows. Cells were left in PBS containing 1 mM EDTA at 37°C for 2 to 3 min and then supplemented with an equal volume of 4 mM CaCl2 in DMEM containing 0.1% bovine serum albumin (freshly prepared). The cells were spun down and resuspended in PBS. Cells (5 x 105 to 7.5 105, as indicated) in 0.5 ml of PBS were injected into the lateral tail vein of 5- to 6-week-old CD-1 nude mice (Weizmann Institute Breeding Center). No significant reduction in cell viability was detected at the end of the injection time. The mice were sacrificed 17 to 21 days postinjection; the lungs were removed, fixed for 24 h in Bouin solution, and stored in 70% alcohol. Before fixation, when indicated, the lungs were blotted dry on Kimwipe tissues and weighed. In some experiments, lungs were stained with India ink to simplify counting of metastatic nodules (55). Macroscopic lesions on the lung surface were counted with the aid of a dissecting microscope. When indicated, EACA was mixed with the cells at 4 mg/ml just before the cells were injected into mice. x

2135

NIH/3T3 CELLS COTRANSFECTION WITH PLASMIDS pEJ6.6 AND pSV2gpt

STEP 1

AND GROWTH IN gpt-SELECTION MEDIUM

ras 6

CELLULAR CLONE COTRANSFECTION pTcHSV

/ \

pTcHSV

ISTEP

/

2

I-I

WITH pSV2neo AND THE INDICATED

EXPRESSION PLASMIDS, AND SELECTION

pcUKLTR6

WITH G418

NR CLONES

CLONES

UNR

TNR

(control)

WuPA)

(tPA)

FIG. 1. Scheme for generation of NIH 3T3 cells transformed with the H-ras gene and expressing high levels of recombinant human PAs. See text for explanation.

tumorigenic cellular clone, ras 6, producing a low level of murine PA, was isolated after transfection of NIH 3T3 cells with the H-ras oncogene as described in Materials and Methods. When grown in culture, ras 6 cells exhibited a transformed morphology and, after subcutaneous injection into nude mice, elicited visible tumors within 1 week. However, no increase in PA level was detected in ras 6 cells compared with the low PA activity measured in the parental NIH 3T3 cells (Fig. 2). In a second step, tumorigenic cells producing high levels of recombinant human PA were generated by transfection of ras 6 cells with the gene construct pcUKLTR6 or pTcHSV, encoding human uPA or tPA, respectively (Fig. 3 and 4). In these constructs, the LTR enhancer-promoter sequence of Harvey sarcoma virus was fused upstream of the PA-coding sequences to drive efficient transcription. Independent experiments indicated that the same constructs, when lacking the PA-coding sequences and transfected into NIH 3T3 cells, did not affect expression of

'0 0)

100

4) 0

-

0 = 0a) OL..( o C

-0

U = a a) c 0 EY

501

-

E

n)

RESULTS Establishment of H-ras-transformed NIH 3T3 cells producing high levels of human PAs. The cells used for testing the effect of recombinant human PA on tumor cell invasion and metastasis were generated in two

steps

(Fig. 1). First,

a

FIG. 2. PA activities secreted by different clonal cells. Cells of the indicated clones were grown, and samples of conditioned media were collected and tested for PA activity in the fibrin plate assay as described in Materials and Methods.

2136

AXELROD ET AL.

MOL. CELL. BIOL.

none grew in the absence of serum (data not shown). The levels of PA activity secreted by the control cell lines were low and similar to that of the parental NIH 3T3 cells, whereas the activities secreted by UNR1O and TNR9 cells were more than 30-fold higher (Fig. 2). PA activities were also determined by a zymographic analysis performed in sodium dodecyl sulfate-polyacrylamide gels. The activity secreted by UNR1O cells comigrated with that of standard human urokinase, whereas the TNR9 activity migrated somewhat more slowly, similarly to the tPA produced by human Bowes melanoma cells (Fig. 5). PA activities produced by UNR1O and TNR9 cells were also tested with rabbit antiserum that neutralizes human uPA activity without blocking murine PAs and with goat antiserum elicited against human tPA that also inhibits murine tPA. The anti-uPA serum quenched the UNR1O but not the TNR9 activity, whereas the anti-PA serum caused an opposite effect; the control sera had no effect on either PA type (Fig. 5). These results indicate that UNR1O cells produced a high level of recombinant human uPA; also, since TNR9 cells were generated by exactly the same procedure as was used for UNR1O cells, the results also strongly suggest that tPA produced by TNR9 cells was of human origin. This latter interpretation is more likely than the alternative one, namely, that endogenous mouse tPA was highly expressed exclusively in TNR9 cells. No changes in the pattern of recombinant PA expression by any of the cellular clones have been observed since production of the clones about 1.5 years ago. Effect of recombinant PAs on basement membrane invasion in vitro. To assess the effect of recombinant PA on the invasive capacity of H-ras-transformed NIH 3T3 cells, a recently described (1) in vitro assay was used. The assay was performed in a modified Boyden chamber separated into two compartments by a filter. The upper surface of the filter was coated with a reconstituted basement membrane composed

pUNI121/

-p5&UK- LTR 6 human u-PA cDNA

EcoRI FIG. 3. Expression plasmid pcUKLTR6 containing the cDNA encoding human uPA. Symbols: E , Harvey sarcoma virus (HaSV) LTR; O, human cDNA sequences; -*, direction of transcription. The bacterial vector was pUN121.

endogenous murine PA (unpublished results). Along with the recombinant genes, we transfected the pSV2neo plasmid, which confers resistance to the drug G418 sulfate, to serve as a selectable marker. Colonies surviving in the presence of G418 sulfate were selected and assayed for PA expression. Virtually all expressed the recombinant PA, although at somewhat different levels. Cellular clones expressing high levels of human uPA and tPA were designated UNR and TNR, respectively. As control cells for this study, we transfected ras 6 cells with the pSV2neo plasmid alone and selected for clones (designated NR) that were resistant to the drug. Three control clones lacking a PA construct (NR1, NR2, and NR3) and four clones expressing high levels of PA (UNR9, UNR10, TNR6, and TNR9) were chosen for further studies. The different cellular clones exhibited similar growth rates when grown in culture in the presence of different concentrations of fetal calf or bovine serum, and S

XTC11

k) Tc 37

f-LTR

E .OE 11

SgOnE

ESgkE E

L

II. sqhl

1. Eco RI

2 Klenow, 4 dNTPs a BstX I

2. Klenow,4dNTP 3 3BomH I Linker T4 DNA Ligose 4 Bom HI, Bst XI 5A/hter ''

Sx

/

4

'

cDN4A-5'

'

cDNA-D

-

T4 DNA Ligose

pR 322

J)

hu t-PAeONA

\

-oyUS

yI@ti@

son'

FIG. 4. Construction of expression plasmid pTcHSV, containing the cDNA encoding human tPA. Symbols: _, tPA cDNA combined from XTc37 and ATc11; Harvey sarcoma virus (HaSV) LTR; c, simian virus 40 (SV40) early gene polyadenylation signal. The direction of transcription is indicated by the arrow inside the circle, and the bacterial vector pBR322 is indicated. Abbreviations for restriction nucleases: B, BamHI; Bg, BglII; Bx, Bxtxl; E, EcoRI; H, HindIII; S, SmaI. c,

VOL. 9, 1989


r') rr~)

aU)

4)

0

m

TNR9

UNRIO

cL

'

O

L

CL

o

c

c

-.

:

I

-0

c

.^

7.

z

z

2137

o

0

c

a

c

E

0

.t 10

C>

.2I FIG. 5. Zymographic analysis of PA secreted by different clonal cells, performed in the absence and presence of specific PA antisera. Samples (5 .1l) of conditioned media collected from the indicated cells were applied to the gel and analyzed as described in Materials and Methods. Markers were standard uPA (1 mU) and tPA secreted by Bowes melanoma cells. Samples of medium from UNR1O and TNR9 cells were incubated before application to the gel with anti-uPA, anti-tPA, or control serum or in the absence of serum (-). Incubation was at 37°C for 30 min and then at 4°C for 1 h; sera were diluted 1:100.

mainly of collagen IV, laminin, and heparan sulfate proteoglycans. Cells were seeded into the upper chamber, and a chemoattractant was added to the lower chamber. Cells derived from malignant tumors were shown to cross the membrane barrier more readily than did cells derived from nonmalignant tumors (1, 42). We compared cells expressing high levels of human recombinant PA with counterpart control cells expressing low levels of murine PA with respect to ability to cross the barrier of reconstituted basement membrane (Fig. 6 and 7). The highest invasive activity was manifested by UNR1O cells, followed by TNR6 and TNR9 cells. The invasiveness of these cells was 2.5- to 7-fold greater than that of control NR3 and NR2 cells and about 7-fold greater than that of the parental NIH 3T3 cells. Also, invasion by NR3 cells occa-

+ W E=L1I lOOr0:c

20

UNRIO

10

5 tTNR6

NR2

/~~~~~~~~,.,R

Hours FIG. 6. Invasive activities of different clonal cells. Invasive activities of cells of the indicated clones were measured as described in Materials and Methods on filters coated with 25 pLg of Matrigel. Inset: Chemotactic migration measured for 6 h.

E 5j 1 l In~~~~~~~~~L

JNRlO NR3 3h

)JNR1O NR3 6h

UNR10 TNR9

NR3 3T3, 6h FIG. 7. Effects of specific PA antisera on invasive activities of different clonal cells. Experiment 1: Anti-human uPA or control rabbit serum was added at a final dilution of 1:50 to the upper part of a Boyden chamber. UNR10 or NR3 cells were added 30 min later and tested for invasion on filters coated with 12.5 ,g of Matrigel. Experiment 2: As described for experiment 1 except that anti-human tPA and control goat serum were added and the filters were coated with 25 ,g of Matrigel.

sionally was more efficient than that by NIH 3T3 cells (Fig. 6 and 7). Because chemotaxis is a prerequisite for invasion in the assay, we compared the cellular clones for chemotactic response to the chemoattractant present in the lower compartments of the Boyden chambers. The chemotaxis assay was performed as previously described (42), in the same manner as the invasion assay but with filters lacking the reconstituted basement membrane. However, no significant differences were detected in the chemotactic migration of the different cellular clones (Fig. 6, inset). Inhibitors of PA were added to the invasion assay to examine whether the high invasive capacity of UNR1O and TNR9 cells could be attributed to the high expression of recombinant PA. In the experiments shown in Fig. 7, antiuPA, anti-tPA, or control sera were added to the invasion chambers. Compared with control rabbit serum, anti-uPA serum inhibited the invasion of UNR1O cells by 55 and 45% after 3 and 6 h, respectively. Compared with control goat serum, anti-tPA serum inhibited the invasion of TNR9 cells by 54% after 6 h and showed no inhibitory effect on invasion of UNR1O and NR3 cells. The serine proteinase inhibitor EACA also inhibited the invasion of UNR1O and TNR9 cells, by 67 and 51%, respectively (Fig. 8), but did not affect the chemotactic migration of these cells (not shown). It was recently reported that invasion in vitro could be inhibited by collagenase inhibitors and that PA affected the process through conversion of plasminogen into plasmin which, in turn, activated latent collagenases (30, 42). In the invasion assay used here, plasminogen was provided through the Matrigel forming the reconstituted basement membranes, which was previously reported to contain enough plasminogen to allow slow but full activation of collagenase necessary for invasion (42). To test whether recombinant human PAs affect invasion through the same proteolytic pathway, we measured collagenase IV activity in samples taken from the upper compartments of the Boyden chambers. In addition, we examined the effect of the synthetic collagenase inhibitor SC-44463, which was previously shown to inhibit efficiently cellular invasion in the same

2138

-~

MOL. CELL. BIOL.

AXELROD ET AL.

10

N

E I

0

C)

5k c

TABLE 1. Effect of recombinant uPA on experimental metastases caused by H-ras-transformed NIH 3T3 fibroblasts *

control

I] EACA{50pg/ml) l SC-44463(5Opg/mo M EACA+SC-44463

TNR6

UNRIO

FIG. 8. Effects of proteinase inhibitors on invasive activities of UNR1O and TNR9 cells. Invasive activities of UNR1O and TNR9 cells were tested for 6 h as described in Materials and Methods on filters coated with 25 ,ug of Matrigel. When indicated, the inhibitors EACA and SC-44463, either alone or in combination, were added to the upper part of a Boyden chamber 30 min before the addition of cells. assay (42). The levels of active collagenase found in UNR10, TNR6, and TNR9 cells were similar and about 2- to 2.5-fold higher than the levels found in NR2 and NR3 control cells; activities of the latter cells were close to the activities of the parental ras 6 and NIH 3T3 cells (Fig. 9). Furthermore, the drug SC-44463 inhibited the invasion of UNR1O and TNR6 cells about 75%, and EACA increased this effect only slightly (Fig. 8). Taken together, these results indicate that expression of high levels of recombinant human PAs increases the invasive capacity of cells in an in vitro assay by, to a large extent, elevating the collagenase activity in the cellular environ-

ment.

Effect of recombinant PA on experimental metastases. To test the effect of recombinant human PA on the metastatic

E 00

rn)

I0

0

0

5

-~~~~~~~~~111 0'

0: 0

,PO

a

%o

:p

Aso

,P

FIG. 9. Collagenase IV activities in conditioned media of different cellular clones. Collagenase IV activity was measured as described in Materials and Methods. Data represent means + standard errors of three different experiments.

Expta

Cellular clone

No. of cells

injected

Mean±lung lesions SEM

mouseb

Mean wet lung M wt SEM (g)b wt±SM()c

18 ± 4 A 0.14 ± 0.1 B 3±2 B

0.24 ± 0.01 A 0.17 ± 0.01 B 0.17 ± 0.00 B

(105)

1

UNR1O NR3 NR1

5 5 5

2

UNR1O NR3

7.5

72 ± 18 A 12 ± 3 B

ND ND

3

UNR9 NR3

7.5

145 ± 23 A 23 ± 8 B

0.76 ± 0.06 A 0.19 ± 0.01 B

4c

UNR9 NR3

7.5 7.5

74 ± 14 A 7±3 B

0.57 ± 0.09 A 0.17±0.02 B

a Cells were harvested by EDTA treatment in experiment 4 and with trypsin in all other experiments. b Means for a given experiment followed by different letters are significantly different (P < 0.003) by the Duncan multiple-range test. ND, Not done. c Correlation coefficient of lung weight to number of metastatic lesions is 0.96 (n = 50).

growth of tumorigenic cells in vivo, we compared UNR cells and NR control cells for the ability to produce experimental lung lesions after intravenous injection into CD-1 nude mice (12). In these experiments, clonal cells propagated in parallel cultures were injected at the indicated numbers into the tail vein (Materials and Methods). The mice were sacrificed 17 to 21 days postinjection, and their lungs were weighed and scored for macroscopic surface lesions. Metastatic lesions were not detected in mice injected with 106 NIH 3T3 cells (results not shown) but were found in mice injected with control NR1 and NR3 cells (Table 1), which confirmed previous reports obtained with different clones of ras-transformed NIH 3T3 cells (2, 4, 16, 53). However, inoculation of mice with TNR9 or UNR1O clones expressing recombinant PA resulted in about sixfold more lung lesions than in mice inoculated with the control cells. This difference was found regardless of whether the cells used for injection were harvested by trypsin or by EDTA and irrespective of the number of challenging cells in the inoculum, although for each cellular clone, increasing the number of injected cells caused a higher incidence of lung lesions (Table 1). During the course of this study, we noticed some variability between the number of metastatic lesions caused by the same cellular clone in different experiments (compare experiments 3 and 4 in Table 1). Similar variability was previously reported and attributed to several factors, such as the state of cellular growth in culture and the harvesting conditions (20, 34). It is possible that the nude mice used at the age of 5 to 6 weeks also contributed to this diversity, perhaps as a result of changes between experiments in their natural killing capacity, which was reported to be maximal at 6 weeks (19). Zymographic analysis of random samples from lungs of mice injected with UNR1O cells revealed in all cases a band of human uPA in addition to bands of murine uPA and tPA, both highly expressed in the lung tissue (not shown). The pulmonary lesions were identified as fibrosarcoma by histological examination. Counting of metastatic areas in histological sections of several experiments yielded results similar to those obtained by macroscopic counting of surface lung nodules (not shown). We also noted that the increase in the number of lung lesions was directly related to wet lung weight measured before lung fixation. The difference in lung


VOL. 9, 1989

TABLE 2. Effect of EACA on experimental metastases caused by injection of 5 x 10i UNR10 cells

None EACAC

Mean lung lesions + SEM/mousea

Mean wet lung wt + SEM (g)

19 + 5 2 ± 0.4

0.26 ± 0.03 0.17 ± 0.01

a Means were significantly different (P < 0.003) according to the Duncan multiple-range test. b Correlation coefficient of lung weight to number of metastatic lesions is 0.96 (n = 50). c EACA (2 mg per mouse) was mixed with cells just before injection.

weight between UNR and NR cells was statistically significant (Tables 1 and 2), and histological examination confirmed that fibrosarcomas were the cause of the increase in lung weight; this increase fulfilled a criterion for metastasis as previously suggested (18). To test whether uPA activity was involved in the formation of lung nodules after injection of UNR cells, we tried to block the effect of uPA by injecting mice with EACA simultaneously with the cells. This treatment caused a significant reduction in the number of lung lesions and in the net weight of lungs (Table 2). The inhibitory effect of EACA on experimental metastasis could not be attributed to a general toxic effect, as the drug (4 mg/ml) did not affect cellular growth measured for 3 days. These experiments show that cells expressing high levels of recombinant PA exhibit an increased capacity to form experimental metastases and that a serine proteinase(s) is to a large extent responsible for the increase. DISCUSSION The choice of cells used in this study and their further genetic modification were based on the following considerations. In separate experiments, we found that highly expressed cDNA constructs encoding either human uPA or tPA did not function as oncogenes by themselves when transfected into NIH 3T3 cells, as determined by the morphology of the transfected cells and by the inability of these cells to generate tumors in nude mice (unpublished results). Therefore, to test the effect of recombinant PAs on the formation of metastasis, the PA-coding DNA sequences had to be introduced into cells that were already tumorigenic. We estimated that the effect of the recombinant PA gene on the metastatic potential could be best tested by transfecting the gene into tumorigenic cells that exhibit some capacity to metastasize yet express endogenous PA at a level low enough to impose a rate limitation on the entire metastatic process. We chose as recipient cells NIH 3T3 mouse fibroblasts because we found that they produce an extremely low level of murine PA activity and, on transformation by the H-ras oncogene, have been reported to form metastatic lesions in nude mice (2, 4, 16, 53). We were able to transform NIH 3T3 cells with the H-ras gene without increasing the PA level, probably by introducing a copy number too low to increase the endogenous PA activity (46). Cells derived from a single ras-transfected clone were further transfected with a DNA construct containing the human PA-coding sequence attached to a strong promoter-enhancer element. Introduction into cells of this single gene generated a phenotype with a higher capacity to invade in vitro and to elicit experimental metastases in vivo in mice, which provides genetic evidence for the direct involvement of the PA gene product in these activities.

2139

The conclusion on the involvement of PA in cellular invasion was reinforced by experiments performed with a clone designated UK6.6 (not shown). This clone was generated by a one-step transfection of NIH 3T3 cells with plasmid pcUKLTR6 along with the selectable plasmid pSV2gpt. Missing the ras gene, UK6.6 cells were not tumorigenic and therefore were not tested for metastasis formation. With respect to PA, collagenase IV, and invasiveness, however, they exhibited a phenotype very similar to that of UNR1O cells, which contained both the ras and human uPA genes. These results thus relate the three activities to the uPA recombinant gene and not to the ras gene. Direct evidence linking PAs to cellylar invasion is provided by the specific inhibitory effect of anti-uPA and anti-tPA sera on the invasion of cells producing human uPA and tPA, respectively. We noted, however, that unlike the complete inhibition caused by these antisera when tested for inhibition of PA in a fibrinolytic assay, invasion was not entirely inhibited, as has been previously reported (42). In the invasion assay, this difference could be attributed, at least in part, to better accessibility of PA to its substrate, plasminogen, found in the reconstituted membrane matrix (42) than to the soluble antibody when PA is secreted by cells attached to the membrane. However, enzymes other than PAs, such as cathepsin B, may affect cellular invasion and account for the incomplete inhibition of the PA-neutralizing antibodies (48). The inhibitory effect displayed by the specific antisera strongly suggests that the inhibition of invasion and probably also of experimental metastasis by EACA, a general inhibitor of serine proteinases, is also largely due to an effect on serine proteinases of the plasminogen activation pathway, i.e., PA and plasmin. Control experiments indicate that this inhibition cannot be attributed to a toxic effect of the drug on the cells. However, to completely establish the inhibitory effect of EACA on metastasis formation, the clearance time of the drug from the circulation and its relationship to the kinetics of metastasis formation must be studied. It was previously shown that PA promotes cellular invasion in vitro through activation of plasminogen into plasmin, which generates active collagenase, whose activity is the permissive event in the invasion (30, 42). This was demonstrated for cellular invasion through the amniotic membrane (30) and through the Matrigel barrier containing plasminogen. Two of our findings indicate that recombinant uPA and tPA initiate the same proteolytic cascade. First, the collagenase inhibitor SC 44463 strongly inhibited in vitro invasion of cells producing either of the human PA types. Second, the levels of active collagenase correlated with the levels of PA activity. These results also indicate that the proteolytic cascade involved in cellular invasion through the Matrigel could be initiated by either of the two PA types. This finding is consistent with the involvement of tPA in the in vitro invasion of the murine melanoma-derived cell line K-1735M2 and with the reports on the involvement of uPA in the invasion of HIT-1080 cells, derived from a human fibrosarcoma (42), and of B16/BLC cells, derived from a murine melanoma (30). Concerning experimental metastasis, future experiments will indicate whether expression of recombinant tPA exerts an effect similar to that of recombinant uPA demonstrated here. During this study, we found that in tumors elicited by subcutaneous injection of UNR1O cells into mice, expression of the recombinant uPA gene was highly variable and in several cases lost entirely. Analysis of tumor-derived DNAs indicated that the foreign DNA was retained in all cases

2140

AXELROD ET AL.

without any detectable change in the restriction and methylation pattern. The variability in recombinant uPA expression is currently under investigation; however, it motivated us to study the effect of recombinant PA on experimental metastasis rather than on spontaneous metastasis arising from a primary tumor site. Our present finding that uPA promotes experimental metastasis indicates that the protease is active in a late metastatic step of extravasation or the subsequent colonization of the UNR tumorigenic cells, as previously suggested for B16 melanoma cells (20). By contrast, in the avian system human uPA was reported to interfere with an early step of intravasation of HEp-3 tumor cells transplanted onto the chorioallantoic membrane and metastasizing into chicken embryo organs (37). Our findings demonstrate that introduction into cells of a gene expressing high levels of either recombinant uPA or tPA yields a phenotype with a higher capacity to invade and, for uPA, to produce experimental metastases. The results thus indicate a causal role for PA-mediated proteolysis in these activities. This conclusion does not rule out the involvement of alternative proteolytic pathways in the degradation of biological barriers. In fact, such a complementary mechanism is suggested to account for the metastatic capacity of the PA-poor control NR cells, which, although inferior to that of UNR cells, was reproducibly evident. ACKNOWLEDGMENTS We thank Oded Lachman for excellent technical assistance, Mordechai Bodner for synthesis of oligonucleotides, L. Ossowski for critical reading of the manuscript and for a gift of uPA antiserum, E. L. Wilson for a gift of tPA antiserum, G. R. Martin for a gift of the SC-44463 collagenase inhibitor and for generous support in the performance of the invasion experiments, and A. Meshorer and I. Rubinstein for assistance in histological examination. This research was supported by grants from the Leo and Julia Forscheimer Center for Molecular Genetics at the Weizmann Institute of Science and from the Basic Research Foundation, administered by the Israel Academy of Science and Humanities (to R.M.). LITERATURE CITED 1. Albini, A., Y. Iwamoto, H. K. Kleinman, G. H. Martin, S. A. Aaronson, J. M. Kozlowski, and R. W. McEwan. 1987. A rapid in vitro assay for quantitating the invasive potential of tumor cells. Cancer Res. 47:3239-3245. 2. Bernstein, S. C., and R. A. Weinberg. 1985. Expression of the metastatic phenotype in cells transfected with human metastatic tumor DNA. Proc. Natl. Acad. Sci. USA 82:1726-1730. 3. Bradford, M. M. 1976. A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Anal. Biochem. 72:248-254. 4. Bradley, M. O., A. R. Kraynak, R. D. Storer, and J. B. Gibbs. 1986. Experimental metastasis in nude mice of NIH/3T3 cells containing various ras genes. Proc. Natl. Acad. Sci. USA 83:5277-5281. 5. Camiolo, S. M., G. Markus, L. S. Englander, M. R. Suita, G. H. Hobika, and S. Kohgen. Plasminogen activator content and secretion in explants of neoplastic and benign tumor prostate tissue. Cancer Res. 44:311-318. 6. Carlsen, S. A., I. A. Ramshaw, and R. C. Warrington. 1984. Involvement of plasminogen activator production with tumor metastasis in rat model. Cancer Res. 44:3012-3016. 7. Caruthers, M. H. 1982. p. 71. In H. G. Gassen and J. A. Lang (ed.), Chemical and enzymatic synthesis of gene fragments: laboratory manual, p. 71-80. Verlag-Chimie, Weinheim, Federal Republic of Germany. 8. Dano, K., P. A. Andreasen, J. Grondahl-Hansen, B. P. Kristensen, L. S. Nielsen, and L. Skriver. 1985. Plasminogen activators, tissue degradation and cancer. Adv. Cancer Res. 44: 139-266.

MOL. CELL. BIOL. 9. Eisenbach, L., S. Segal, and M. Feldman. 1985. Proteolytic enzymes in tumor metastasis. I. Plasminogen activator in clones of Lewis lung carcinoma and T10 sarcoma. J. Natl. Cancer Inst. 74:77-85. 10. Ellis, R. W., D. Defeo, J. M. Maryak, H. A. Young, T. Y. Shih, E. H. Chang, D. R. Lowy, and E. M. Scolnick. 1980. Dual evolutionary origin for the rat genetic sequences of Harvey murine sarcoma virus. J. Virol. 36:408-420. 11. Engvall, E., and E. Rouslahti. 1977. Binding of soluble form of fibroblast surface protein, fibronectin, to collagen. Int. J. Cancer 20:1-5. 12. Fidler, I. J. 1978. General considerations for studies of experimental cancer metastasis. Methods Cancer Res. 15:399-439. 13. Fidler, I. J., D. M. Gersten, and I. R. Hart. 1978. The biology of cancer invasion and metastasis. Adv. Cancer Res. 28:149-250. 14. Giraldi, T., G. Sava, L. Perissin, and S. Zorzet. 1985. Proteinases and proteinase inhibition by cytotoxic and antimetastatic drugs in transplantable solid metastasizing tumors in mice. Anticancer Res. 5:355-360. 15. Graham, F. L., and A. J. van der Eb. 1973. A new technique for the assay of infectivity of human adenovirus 5 DNA. Virology 52:456-467. 16. Greig, R. G., T. P. Koestler, D. L. Trainer, S. P. Corwin, L. Miles, T. Kline, R. Sweet, S. Yokoyama, and G. Poste. 1985. Tumorigenic and metastatic properties of "normal" and rastransfected NIH/3T3 cells. Proc. Natl. Acad. Sci. USA 82: 3698-3701. 17. Gunzler, W. A., G. J. Steffens, F. Otting, S.-M.A. Kim, E. Frankus, and L. Flohe. 1982. The primary structure of high molecular mass urokinase from human urine. Hoppe-Zeyler's Z. Physiol. Chem. 363:1155-1165. 18. Hagmar, B., W. Ryd, and L. J. Erkell. 1983. Methods in metastatic research. Invasion Metastasis 3:1-21. 19. Hanna, N. 1980. Expression of metastatic potential of tumor cells in young nude mice is correlated with low levels of NK cell-mediated cytotoxicity. Int. J. Cancer 26:675-680. 20. Hearing, V. J., L. W. Law, A. Corti, E. Appelia, and F. Blasi. 1988. Modulation of metastatic potential by cell surface urokinase of murine melanoma cells. Cancer Res. 48:1270-1278. 21. Huynh, T. Y., R. A. Young, and R. W. Davis. 1984. Constructing and screening cDNA libraries in AgtlO and Agtll, p. 499. In D. Glover (ed.), DNA cloning: a practical approach. IRL Press, New York. 22. Kirchheimer, J. C., H. Pfluger, P. Ritschl, G. Hienert, and B. R. Binder. 1985. Plasminogen activator activity in bone metastasis of prostatic carcinomas as compared to primary tumors. Invasion Metastasis 5:344-355. 23. Laub, O., L. B. Rall, M. Truett, Y. Shaul, D. N. Standring, P. Valenzuela, and W. J. Rutter. 1983. Synthesis of hepatitis B surface antigen in mammalian cells: expression of the entire gene and the coding region. J. Virol. 48:271-280. 24. Liotta, L. A. 1986. Tumor invasion and metastasis-role of the extracellular matrix: Rhoads Memorial Award Lecture. Cancer Res. 46:1-7. 25. Maniatis, T., E. F. Fritsch, and J. Sambrook. 1982. Molecular cloning: a laboratory manual. Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. 26. Markus, G. 1984. The role of hemostasis and fibrinolysis in the metastatic spread of cancer. Semin. Thromb. Hemostasis 10: 61-70. 27. Markus, G., S. M. Camiolio, S. Kohga, J. M. Madeja, and A. Mittelman. 1983. Plasminogen activator secretion of human tumors in short-term organ culture, including a comparison of primary and metastatic colon tumors. Cancer Res. 43:55175525. 28. Maxam, A., and W. Gilbert. 1980. Sequencing with basespecific chemical cleavage. Methods Enzymol. 65:499-560. 29. Menzel, J., and W. Borth. 1983. Influence of plasma fibronectin on collagen cleavage by collagenase. Collagen Related Res. 3:217-230. 30. Mignatti, P., E. Robbins, and D. B. Rifkin. 1986. Tumor invasion through the human amniotic membrane: requirement for a proteinase cascade. Cell 47:487-498.

VOL. 9, 1989


31. Miskin, R., and H. Soreq. 1981. Sensitive autoradiographic quantification of electrophoretically separated proteases. Anal. Biochem. 118:252-258. 32. Mulligan, R. C., and P. Berg. 1981. Selection for animal cells that express the Escherichia coli gene coding for xanthineguanine phosphoribosyl-transferase. Proc. Natl. Acad. Sci. USA 78:2072-2076. 33. Mullins, D. E., and S. T. Rohrlich. 1983. The role of proteinases in cellular invasiveness. Biochim. Biophys. Acta 695:177-214. 34. Nicolson, G. L. 1982. Cancer metastasis. Organ colonization and the cell-surface properties of malignant cells. Biochim. Biophys. Acta 695:113-176. 35. Nilsson, B., M. Uhlen, S. Josephson, S. Gatenbeck, and L. Philipson. 1983. An improved positive selection plasmid vector constructed by oligonucleotide mediated mutagenesis. Nucleic Acids Res. 11:8019-8030. 36. O'Grady, P., H. R. Lijnen, and M. J. Duffy. 1985. Multiple forms of plasminogen activator in human breast tumors. Cancer Res. 45:6216-6218. 37. Ossowski, L. 1988. Plasminogen activator dependent pathways in the dissemination of human tumor cells in the chick embryo. Cell 52:321-328. 38. Ossowski, L., and E. Reich. 1983. Antibodies to plasminogen activator inhibit human tumor metastasis. Cell 35:611-619. 39. Pennica, D., W. E. Hohnes, W. J. Kohr, R. N. Harkons, G. A. Vehar, C. A. Ward, W. F. Bennett, E. Yelverton, P. H. Seeburg, H. L. Heyneker, D. V. Goeddel, and D. Collen. 1983. Cloning and expression of human tissue-type plasminogen activator cDNA in E. coli. Nature 310:214-221. 40. Ramshaw, I. A., P. Badenoch-Jones, A. Grant, M. Maxted, and C. Claudianus. 1986. Enhanced plasminogen activator production by high metastatic variant cell line of a rat mammary adenocarcinoma. Invasion Metastasis 6:133-144. 41. Reich, E. 1978. Activation of plasminogen: a general mechanism for producing localized extracellular proteolysis, p. 155-169. In R. D. Berlin, H. Herrmann, I. H. Lepow, and J. M. Tanzer (ed.), Molecular basis of biological degradation processes. Academic Press, New York. 42. Reich, R., E. Thompson, Y. Iwamoto, G. R. Martin, J. R. Deason, G. C. Fuller, and R. Miskin. 1987. Effects of inhibitors of plasminogen activator, serine proteinases and collagenase IV on the invasion of basement membrane by metastatic cells in mice and humans. Cancer Res. 48:3307-3312. 43. Sappino, A.-P., N. Busso, D. Belin, and J.-D. Vassali. 1987.

44. 45.

46.

47.

48.

49. 50.

51.

52. 53.

54.

55.

2141

Increase of urokinase-type plasminogen activator gene expression in human lung and breast carcinomas. Cancer Res. 47: 4043-4046. Schirrmacher, V. 1985. Cancer metastasis: experimental approaches, theoretical concepts, and impacts for treatment strategies. Adv. Cancer Res. 43:1-73. Shih, C., and R. A. Weinberg. 1982. Isolation of a transforming sequence from a human bladder carcinoma cell line. Cell 29:161-169. Sistonen, L., J. Keski-Oja, I. Ulmanen, E. Hoitta, B.-J. Wikgren, and K. Alitalo. 1987. Dose effect of transfected c-Ha-rasvaI 12 oncogene in transformed cell clones. Exp. Cell Res. 168: 518-530. Skriver, L., L.-I. Larsson, V. Keilberg, L. S. Nielsene, P. B. Andresen, P. Kristensen, and K. Dano. 1984. Immunocytochemical localization of urokinase-type plasminogen activator in Lewis lung carcinoma. J. Cell Biol. 9;9:753-757. Sloane, B. F., K. V. Honn, J. G. Sadler, W. A. Turner, J. J. Kimpson, and J. D. Taylor. 1982. Cathepsin B activity in B16 melanoma cells: a possible marker for metastatic potential. Cancer Res. 42:980-986. Soreq, H., and R. Miskin. 1981. Plasminogen activator in the rodent brain. Brain Res. 216:361-374. Southern, P. J., and P. J. Berg. 1982. Transformation of mammalian cells to antibiotic resistance with a bacterial gene under control of the SV40 early region promoter. Mol. Appl. Genet. 1:327-341. Steffens, G. J., W. A. Gunzler, F. Otting, E. Frankus, and L. Flohe. 1982. The complete amino acid sequence of low molecular mass urokinase from human urine. Hoppe-Zeyler's Z. Physiol. Chem. 363:1043-1058. Strickland, S., E. Reich, and M. Sherman. 1976. Plasminogen activator in early embryogenesis: enzyme production by trophoblast and parietal endoderm. Cell 9:231-240. Thorgeirsson, U. P., T. Turpeenniemi-Hujanen, J. E. Williams, E. H. Westin, C. A. Heilman, J. E. Talmadge, and L. A. Liotta. 1985. NIH/3T3 cells transfected with human tumor DNA containing activated ras oncogenes express the metastatic phenotype in nude mice. Mol. Cell. Biol. 5:259-262. Wang, B. S., G. A. McLoughlin, J. P. Richie, and J. A. Mannick. 1979. Correlation of the production of plasminogen activator with tumor metastasis in B16 mouse melanoma cell lines. Cancer Res. 40:288-292. Wexler, H. 1966. Accurate identification of experimental pulmonary metastases. J. Natl. Cancer Inst. 36:641-645.