Int. J. Cancer: 50,208-21 4 (1 992) 0 1992 Wiley-Liss, Inc.
WC
PJblication of tne lnternattonai Un on Aga nst Cancer Puol cat on oe Lnion Internationale Conrre e Cancer
TYPE-1 PLASMINOGEN ACTIVATOR INHIBITOR IN HUMAN BREAST CARCINOMAS David REILLY’’, Lise CHRISTENSEN’, Mogens DUCH’,Naimh NOLAN4, Michael J. DUFFY’and Peter A. ANDREASEN” ‘Institute of Molecular Biology, University of Jrhus, Denmark; ‘Education and Research Center, St. Vincent’s Hospital, Dublin, Ireland; ’Department of Pathology, Rigshospitalet, Copenhagen, Denmark; ‘Pathology Department and ’Nuclear Medicine Department, St. VitzcentS Hospital, Dublin, Ireland. We have investigated the occurrence of type-I inhibitor of plasminogen activators (PAL I) in human breast tumors. PAL I levels, measured by enzyme-linked immunosorbent assay, were significantly higher in malignant breast carcinomas (n = 178) than in benign breast tumors (n = 25). The levels of PAL I were found to be correlated with those of urokinase-type plasminogen activator @-PA). The presence of PAL I in tumor extracts was also demonstratedby immunoblotting analysis. Immunohistochemical investigations by the use of monoclonal and polyclonal antibodies showed that PAL Iwas mostly localized in the tumor islands, associated with the tumor cells; in addition, it was present in vessel walls and in normal duct epithelia, but absent from the stroma. Analysis of RNA extracted from tumors by polymerase chain reaction revealed the presence of PAI-I mRNA. We conclude that PAI-I is present in human breast carcinoma cells, and that it is-at least partiallyproduced locally, either by the cancer cells or by other cells in the tumors. We have previously demonstratedthat a high level of u-PA in human breast carcinomas is associated with poor prognosis. These results, combined with our present findings, present 2 possibilities: either the cancer cells need PAI-I in order to utilize the u-PA-mediated pathway of plasminogen activation for invasion and metastasis; or PAI-I represents a defense mechanism against tumor invasion.
Proteolytic degradation of the extracellular matrix is considered an early and crucial feature of the metastatic process (Poste and Fidler, 1978; Liotta et al., 1980; Liotta, 1986). Increasing evidence of proteolytic enzymes playing a role in cancer invasion and metastasis has been obtained both from in vitro invasion and metastasis assays and from in vivo studies of levels and localizations of these proteases (Liotta et al., 1991). An important and highly regulated cascade of such proteolytic events involves the plasminogen activation system. Plasminogen activators are serine proteases which catalyze the conversion of the abundant zymogen plasminogen into plasmin, another serine protease of broad-spectrum trypsin-like specificity. The effect of plasmin appears to be either direct degradation of extracellular matrix components such as fibronectin, laminin and proteoglycans, or an indirect effect via the activation of latent collagenase and subsequent degradation of cotlagen. Of the 2 distinct types of plasminogen activators, the urokinase-type (u-PA) has been implicated in both normal and pathological processes of tissue remodelling, organ involution and invasion; the tissue-type (t-PA) plays an important role in fibrinolysis and thrombolysis (Dan0 et al., 1985; Skriver et al., 1987; Saksela and Rifkin, 1988; DUAL, 1991). Regulation of these proteases is achieved at many levels, including inhibition by at least 2 types of fast-acting specific inhibitors, PAI-1 and PAI-2. These inhibitors are products of different genes, have different biochemical properties and probably also different biological functions (Kruithof, 1988; Andreasen et al., 1990). It was previously demonstrated that human breast carcinomas contain u-PA (Tissot et al., 1984; Clavel et a/., 1986) and its mRNA (Sappino et al., 1987). In a recent study, we measured u-PA levels in tumor extracts by ELISA; the level of u-PA protein varied strongly between different tumors, but showed a strong inverse correlation with disease-free interval and survival (DufTy et al., 1990). Janicke et al. (1990) reported similar
findings. We now report the presence of PAI-1 protein and mRNA in human breast carcinomas. MATERIAL AND METHODS
Ma teria 1 The following materials were obtained from the indicated sources. Alkaline phosphatase-conjugated rabbit IgG against mouse IgG (Dakopatts, Copenhagen, Denmark); nitroblue tetrazolium (Grade 111) and 5-bromo-4-chloro-3-indolyl phosphate (Sigma, St. Louis, MO); nitrocellulose filters for protein blotting (type GSWP00010, Les Ulis, France); DNase I (Grade 1, Boehringer, Mannheim, Germany); avian myeloblastosis virus reverse transcriptase and Thermus aquaticus DNA polymerase I (Stratagene, La Jolla, CA). All other materials were those described previously (Andreasen et al., 1987; Christensen and Strange, 1987; Duffy et al., 1990), or of the best grade commercially available. Tissues Tumors were collected at the time of surgery and freed of visible necrosis, fat and connective tissue. They were frozen immediately in liquid nitrogen and stored at -70°C until further use. Antibodies Murine monoclonal IgGs against human PAI-1 were from hybridoma clones 1, 2 and 3, as described previously (Nielsen et aL, 1986), or from clone 6, resulting from another independent immunization. Of these, IgGs from clones 2 and 6 may have identical epitopes, while IgGs from the other clones recognize different epitopes, all of which are different from that recognized by clones 2 and 6 (data not shown). Rabbit polyclonal IgG against human PAI-1 (Andreasen et al., 1986) were adsorbed against human fibronectin in order to remove contamination with anti-fibronectin IgG. A portion of the fibronectin-adsorbed IgG was also adsorbed against PAI-1, which had been purified by immunoaffinity chromatography with monoclonal IgG from hybridoma clone 2 (Nielsen et al., 1986). The adsorbed antibodies were controlled for lack of anti-fibronectin and anti-PAI-1 reactivity, respectively, by ELISA. Monoclonal IgG against the trinitrophenyl group was that described by Shulman et al. (1978). A monoclonal IgG, L26, a B-lymphocyte marker, was purchased from Dakopatts (code M755). ELISA Tumor extracts were prepared by one of 2 methods: as previously described (Durn et al., 1990); by homogenization of tumor pieces with a glassiglass homogenizer in a buffer of 0.1 M Tris, pH 8.1,0.5% Triton X-100, 10 mM EDTA and 10 kg/ml of Trasylol, followed by centrifugation at 10,000g for 15 min.
‘To whom correspondence and reprint requests s ould be sent, at the Institute of Molecular Biology, University of rhus, 130 C.F. Mdler’s AM, 8000 k h u s C, Denmark.
A
Received: July 1, 1991 and in revised form August 20,1991.
209
PAI-I IN BREAST CARCINOMAS
The protein concentrations of the extracts were determined by the Bradford method. There was no interference from the Triton X-100 in the extraction buffer in the amounts in which it was added to the assay. The extracts were analyzed for PAI-1 by ELISA, using a mixture of IgGs from hybridoma clones 1, 2, 3 and 6 on the solid phase. This ELISA measures free PAI-1 and PAI-1 in complex with u-PA and t-PA (Lund et al., 1988). The u-PA ELISA employed measures free u-PA and u-PA/PAI-1 complexes (Lund et al., 1988).
Immunoblottinganalysis For immunoblotting analysis, PAI-1 was partly purified from tumor extracts by immunoaffinity chromatography on columns of monoclonal anti-PAI-1 IgG from hybridoma clone 1 immobilized on Sepharose (Nielsen et al., 1986). The columns were equilibrated in 0.1 M Tris, PH 8.1. The samples, usually containing a total of 4 to 5 mg protein, were then applied, and the columns washed with equilibration buffer. The columns were further washed in the same buffer with 1 M NaCI. Bound protein was then eluted with 0.1 M CH,COOH, pH 2.9, 1 M NaCI, dialysed against 0.03% SDS and lyophilized. Subsequently, the proteins were subjected to SDS-PAGE and transferred electrophoretically to nitrocellulose filters, which were stained immunochemically with mouse monoclonal antiPAI-1 IgG from hybridoma clone 1 or mouse monoclonal IgG of irrelevant specificity. For staining, the nitrocellulose filters were first washed with 0.05 M Tris, PH 7.4, 0.15 M NaCl (TBS), containing 0.3% Tween 20 (TBS-Tween). In order to block remaining binding sites on the filters, they were incubated with TBS with 30% rabbit serum. The filters were then washed with TBS-Tween, and incubated with the monoclonal antibodies at a concentration of 5 pg/ml overnight at 4°C. After 3 washes with TBS-Tween, the filters were incubated with alkalinephosphatase-conjugated rabbit anti-mouse IgG, diluted 1:100 from the commercial stock solution. Filters were washed 3 times with TBS-Tween and once in 0.05 M Tris, pH 7.4. Finally, the filters were washed with 0.1 M Tris, pH 9.5, 0.1 M NaCI, 5 mM MgCI, and developed using nitroblue tetrazolium (0.33 mg/ml) and 5-bromo-4-chloro-3-indolyl phosphate (0.17 mg/ ml) in the same buffer. Immunohistochemical stainings Frozen sections were cut from tumor samples at 5 Km, air-dried at room temperature for 24 hr and fixed in acetone for 10 min. Two sections were placed on the same slide representing the tumor at 2 different levels. These were stained with monoclonal IgG against PAI-1, polyclonal antibodies against PAI-1 o r control antibodies, including monoclonal IgG of irrelevant specificity (L26) and polyclonal antibodies that had been absorbed against PAI-1 in order to remove anti-PAI-1 reactivity (see above). All primary antibodies were incubated at a concentration of 10 pg/ml for 20 hr at 4°C and then for 2 hr at room temperature. The sites of reaction of the primary antibodies were visualized by conventional immunoperoxidase techniques (Christensen and Strange, 1987). The chromogen was 3-amino-9-ethylcarbazole,and Mayer’s hemalun was used as nuclear counterstaining. In some of the tumors, 2 sections for immunohistochemical staining were cut from each end of the tissue block, and sections from the intervening 100 to 200 mg of tissue were extracted for ELISA. Other tumors were cut in serial sections, and different sections were stained with different antibodies. RhLA extraction and analysis
Total RNA was extracted from solid tumors as described by Krieg et al. (1983), with minor modifications. Briefly, 0.2 to 0.5 g of tumor tissue, frozen in liquid nitrogen, was homogenized by mechanical disruption, using a Braun Mikro-dismembrator. Nucleic acids were extracted with 0.3 M Na acetate, pH 7.5,
0.5% SDS, 5 mM EDTA, followed by phenol extraction and ethanol precipitation. The precipitated nucleic acids were re-dissolved in sterile H,O. R N A was then selectively precipitated with an equal volume of 4 M L E I . All samples were treated with 20 pg/ml DNase I and finally re-dissolved in a buffer of 10 mM Tris, PH 7.4, 1 mM EDTA. Concentration and quality of RNA was checked spectrophotometrically and by agarose gel electrophoresis. For analysis of R N A samples for the presence of PAI-I mRNA by polymerase chain reaction (PCR), single-stranded cDNA was first synthesized by reverse transcription, as described by Krug and Berger (1988), with minor modifications. Briefly, 3 pg of total R N A were suspended in 9 pl sterile H,O and heated to 65°C for 5 min, placed on ice, and mixed with 1.5 p1 of a buffer of 500 mM Tris-HC1, pH 8.3,500 mM KCl, 100 mM MgCI,, 10 mM dithiothreitol, 10 mM EDTA, 100 pgirnl BSA. One microliter 10 mM dNTP mixture, 1 pl 150 pg/ml oligo dT, 0.75 pl 10 mM spermidine, and 0.75 pl 80 mM NaPP, were added. Finally, 10 U of avian myeloblastosis virus reverse transcriptase, diluted in a buffer of 10 mM K, HPO,, pH 7.4, 10% glycerol, 0.2% Triton X-100 and 2 mM dithiothreitol were added and the reaction was carried out at 37°C for 1 hr. The 2 oligonucleotides used for amplification of PAI-1 cDNA in the reverse transcription reaction mixture were 5’-GAGGTGCCTCTCTCTGCCCTCACCAACATT-3‘ and 5’-AGCCTGAAACTGTCTGAACATGTCG-3’, corresponding to sequences in the PAI-1 cDNA at nucleotides 925-955 and 1082-1108 (Pannekoek et al., 1986). The oligonucleotides were synthesized on an Applied Biosystems (Foster City, CA) Oligonucleotide Synthesizer. PCR amplification was done according to the Stratagene protocol, except that the annealing temperature was set at 62°C. Denaturation and annealing times were 1.2 min, and extension was 1 min. After 44 rounds of amplification, the PCR products were analyzed by electrophoresis in 3% agarose gels. Statistics All statistical analyses were carried out with IBM Statistics and Graphics System Package, Version 1.2. The data were analyzed by non-parametric methods, using Spearman rank correlation coefficient.
RESULTS
Demonstration of PAI-1 in tumor extracts by ELISA and immunoblottinganalysis The levels of PAI-1 and u-PA were measured by ELISAs in extracts of 178 breast carcinomas and 25 benign breast tumors. The PAI-1 ELISA measured free PAI-1 as well as complexes between PAI-1 and u-PA or t-PA. The u-PA ELISA measured free u-PA as well as u-PAiinhibitor complexes. These facts indicated that the measured levels of PAI-1 and u-PA were mutually independent. It was found that the levels of both PAI-1 (Fig. 1) and u-PA (not shown) were significantly higher in the carcinomas than in the benign tumors. In addition, there was a significant correlation between the levels of the 2 proteins (Fig. 2). In contrast, no significant correlation was found between.PA1-1 and t-PA levels (data not shown). Among the carcinomas, PAI-1 levels did not correlate with a number of histological parameters, including tumor volume, grade and cellularity; lymphocyte, mast cell or eosinophil infiltration; axillary-node status; and lymphatic invasion or skin invasion. The PAI-1 level was in the range 1 to 100 ngimg total protein. For comparison, the plasma level of PAI-1 is in the range 0.02 to 0.7 ng/mg protein, and the serum level around 2.5 ngimg protein (Kruithof, 1988).
210
REILLY ET AL. 100
I
MALIGNANT
BENIGN
FIGURE1-PAI-1 antigen levels in breast carcinomas and benign tumors. Extracts from 178 carcinomas and 25 benign tumors were prepared by method 1 (see “Material and Methods”) and analyzed for PAI-1 by ELISA. PAI-1 levels greater than 10 ng/mg total protein were found in 78% of the malignant breast carcinomas. In contrast, levels higher than 10 ngiml were found in only 20% of the benign samples. There was a significant difference between the 2 groups ( p < 0.0005). inn
.--
I
r
I
.
.
0.1
. . . 1.0
I
.. . . . .
I I
10
100
U-PA (ngirng protein)
FIGURE2 - Correlation between PAI-1 and u-PA antigen levels in breast carcinomas. Extracts from 178 tumors were prepared by the first method (see “Material and Methods”) and analyzed for PAI-1 and u-PA by ELISA. Slope 0 . 5 3 , ~< 0.0005. The presence of PAI-1 in tumor extracts was also demonstrated by immunoblotting analysis. PAI-1 was first partly purified from the extracts by immunoaffinity chromatography on a column of monoclonal anti-PAI-1 IgG immobilized on Sepharose. The partly purified preparations were subjected to immunoblotting analysis, incubating the filters with either monoclonal anti-PAI-1 IgG or monoclonal IgG of irrelevant specificity. Incubation with anti-PAI-1 IgG, but not IgG of irrelevant specificity, gave staining of a band co-migrating with purified PAI-1; no other bands were stained (Fig. 3). Fibrin-agarose zymography of 30 tumor extracts (not shown) revealed the presence of u-PA exclusively as an M, 54,000 band. The M, 36,000 degradation product was not found. I n addition, no PAI-1 /u-PA inhibitor complexes were detected, in agreement with the results of the immunoblotting analysis.
FIGURE 3 - Immunoblotting analysis of PAI-1 in extracts of breast carcinomas. PAI-1 was partly purified from tumor extracts by immunoaffinity chromatography. The partly purified preparations were subjected to SDS-PAGE, and then transferred electrophoretically to nitrocellulose filters. The filters were stained immunochemically. Lanes a and a’ contain 35 ng of purified PAI-1, and lanes b, c, b’, and c’ representative tumor samples. Lanes a-c were stained with monoclonal anti-PAI-1 IgG from hybridoma clone 1, and lanes a’-c’ with monoclonal antitrinitrophenyl IgG. The position of PAI-1 is indicated to the left.
Immunohistochemical localization of PAI-1 in breast tumors Sections from 43 breast carcinomas were stained immunohistochemically for PAI-1 with monoclonal IgG from different hybridoma clones, or with polyclonal IgG. The sections were cut from frozen tumor samples. Formalin-futed tissue was found unsuitable due to loss of antigenicity during processing. All anti-PAI-1 antibodies consistently showed PAI-1 immunoreactivity associated with the tumor cells, although in variable amounts in different areas. Also epithelial cells from benign-looking ducts and glands were stained. PAI-1 immunoreactivity appeared to be associated with the basement membrane of the ducts. PAI-1 was also present in blood vessels. Stromal cells were not stained (Fig. 4). The staining was heterogeneous, and the staining intensity varied from one tumor to the other. However, no clear correlation could be demonstrated between staining intensity and the histological carcinoma sub-type. Among the monoclonal IgGs tested, that from hybridoma clone 6 gave the strongest staining; IgG from hybridoma clones 1 and 2 also stained, but more weakly. The following controls assured the specificity of the staining: Monoclonal IgG of irrelevant specificity and polyclonal IgG, that had been adsorbed against PAI-1 in order to remove anti-PAI-1 reactivity, did not give any staining (Fig. 4b and d ) . This indicates that the staining is due to specific antigenantibody reaction and not a non-immunological binding of antibodies to tissue components. When adjacent sections of a tumor were stained with monoclonal anti-PAI-1 IgGs with different epitopes and with polyclonal IgG, the same areas were found to contain PAI-1 immunoreactivity (not shown), indicating that cross-reacting antigens in the tissue did not account for the staining. The immunoblotting analysis showed that monoclonal antiPAI-1 IgG reacted specifically with only 1 band in extracts of the tumors (Fig. 4), again indicating the absence of crossreacting antigens in the tissue. Parts of some of the tumors used for staining were extracted, and the amounts of PAI-1 in the extracts were determined by ELISA. The ELISA values were compared blindly with a semi-quantitative grading of the intensity of the staining with
PAI-1 IN BREAST CARCINOMAS
211
FIGURE 4 - Immunohistochemical localization of PAI-1 in human breast carcinomas. Sections were cut from frozen tumor samples, and incubated with various anti-PAI-1 and control antibodies. ( a ) and (b) show step sections of an invasive mammary-duct carcinoma stained with a polyclonal antibody against PAL1 ( a ) and with the polyclonal antibody adsorbed against purified PAI-1 antigen (b).The tumor cells show a heterogeneous PAL1 immunoreactivity, whereas the normal duct (bottom left quarter of the micrograph) is more uniformly stained, with a tendency to concentrate towards the basement-membrane area. Blood vessels are stained as well (bottom center) ( a ) . The absorbed antibody gives no staining reaction (b). Magnification 87-fold. ( c ) and (d) show step sections of a well differentiated duct carcinoma stained with a monoclonal antibody against PAI-1 (from hybridoma clone 2) ( c ) and an irrelevant MAb L26 against B lymphocytes (d). The tumor cells of the irregular gland-like structures (center) show a distinct immunoreactivity for PAI-1. The same is true for the epithelial cells and the basement-membrane area of a normal-looking duct (top right) and endothelial cells (arrows) ( c ) . No cellular staining is seen with the irrelevant antibody (d). Magnification 87-fold. (e) and (f) show close-up photographs of step sections from an invasive tumor island of a large-cell duct carcinoma. Tumor cell nuclei appear empty, due to freezing artifacts. In (e) the tumor cells are intensely stained with a MAb to PAI-1 (from hybridoma clone 6). In (f) the tumor cells fail to stain by the irrelevant antibody L26 against B-lymphocytes. Magnification 250-fold.
212
REILLY ET AL.
monoclonal IgG on a scale from 0 to 4. T h e ELISA values vary in parallel with the grading of the staining (Fig. 5). PAI- 1 mRNA RNA was extracted from 24 breast carcinomas and analyzed for the presence of PAI-1 mRNA by PCR analysis. First, single-strand cDNA was synthesized by the use of reverse transcriptase with tumor R N A as a template. Synthesized PAI-1 cDNA was then amplified by the use of 2 primers, expected to result in a product of 183 b p (see “Material and Methods”). A number of measures were taken to exclude signals created by contaminants. Firstly, primers were chosen belonging to 2 different exons (Loskutoff et al., 1987), thereby excluding the possibility of amplification due to trace amounts of genomic DNA surviving the DNase treatment during the R N A preparation; such contaminants would result in a PCR product of about 2 Kb. Secondly, before amplification, all samples and controls were treated with RNase 1, in order to remove RNA, and with Alu I, which will cleave any contaminating doublestrand PAI-1 cDNA between the 2 primers, but not cleave the single-strand cDNA resulting from the reverse transcriptase reaction. Thirdly, controls were run in which either reverse transcriptase or RNA was omitted from the reverse transcriptase reaction. As shown in Figure 6, no PCR product was obtained from the controls without R N A or reverse transcriptase. In contrast, all tumor R N A samples resulted in clear bands, which by comparison to molecular size markers were found to have the expected size (Fig. 6), and to give the expected restriction fragments (not shown). We therefore concluded that the breast tumors contain PAI-1 mRNA.
0 0
2
4
6
8
10
1
P A I - 1 (nghng protein)
FIGURE5 - Comparison of intensity of immunohistochemical staining of PAI-1 and PAI-1 ELISA values in breast carcinomas. Sections of 15 breast carcinomas were stained immunohistochemically by the use of monoclonal anti-PAI-1 IgG from hybridoma clone 1. The intensity of the staining was graded on a scale from 0 to 4 (ordinate). Other pieces of the same tumors were extracted by the second method (see “Material and Methods”), and the PAI-1 level determined by ELISA (abcissa). The linear correlation coefficient for the line drawn, calculated by non-weighted linear regression analysis, was 0.63. With 15 samples, this corresponds to a p value of 0.019.
FIGURE6 - Demonstration of PAI-1 mRNA in human breast carcinomas by PCR. RNA from tumors was used as a template for reverse transcription, and the resulting single-strand cDNA amplified by PCR. Lanes 1 to 24 represent the PCR products obtained from tumor samples. Lanes 25 and 27 represent controls, in which RNA was omitted during reverse transcription. Lanes 26 and 28 represent controls, in which RNA from the tumor samples shown in lanes 11 and 12 was incubated for reverse transcription in the absence of reverse transcriptase. The PCR products were electrophoresed in 3% agarose gels. Size markers (unnumbered lane between lanes 14 and 15): 220,234,398,453,517, 653, 1033, 1766 and 2176 bp. DISCUSSION
The amount of u-PA in extracts of primary human breast carcinomas is inversely correlated with the disease-free interval of the patient (DuEy et al., 1990; Janicke et al., 1990). This observation is in agreement with the hypothesis that u-PA is implicated in tumor invasion and metastasis (Dan0 etal., 1985; Skriver et al., 1987; Saksela and Rifkin, 1988; Duffy, 1991). We show here that the level of PAI-1 is higher in breast carcinomas than in benign tumors, and that the level of PAI-1 in the tumor extracts is correlated with the level of u-PA. Although the correlation between PAI-1 and u-PA levels is highly significant, we have not found any significant relationship between the PAI-1 level and the risk of disease recurrence at the present stage of patient follow-up. Yet the presence of PAI-1 in these tumors at a level correlating with that of u-PA indicates that PAI-1 may play a role in regulation of plasminogen activation, and thereby of invasion and metastasis, in breast cancer. Our findings are in agreement with the previous report of PAI-1 being present within tumor cells in the transplantable murine Lewis lung carcinoma (Kristensen et al., 1990), which also contains high levels of u-PA (Skriver et al., 1984). The co-existence of a protease and its inhibitor within a tissue is not unique to the plasminogen activation system. Laskowski and Kato (1980) pointed out that the presence of secretory trypsin inhibitors in the pancreas of vertebrates, together with high levels of trypsinogen, may serve to prevent premature activation of trypsinogen and other pancreatic zymogens. PAI-1 has also been reported to be a marker of invading trophoblastic cells, representative of a controlled invasion process which is believed to involve the plasminogen activation system (Feinberget al., 1989; Yeh and Kurman, 1989). The present findings pose important questions as to which cells are responsible for PAI-1 biosynthesis in breast tumors. PAI-1 occurs in high levels in blood platelets, and is also present in blood plasma (Kruithof, 1988; Andreasen et al., 1990). However, the high level of PAI-1 in the tumor extracts, as compared with that in plasma and serum (see above), excludes plasma, serum and platelet contamination as the sole sources of PAI-1 in tumor extracts. Also, the demonstration of PAI-1 mRNA in the tumors excludes the possibility that blood platelets are the only source of PAI-1 in the tumors, and shows that at least part of the PAI-1 must b e produced locally. The association of PAI-1 antigen with the tumor cells is compatible with the idea that at least part of
213
PAI-1 I N BREAST CARCINOMAS
the PAI-1 is produced by these cells. A similar staining of vessel walls and normal ducts suggests PAI-1 production there as well. Irrespective of the cell type producing PAI-1 in and around these tumors, the localization of PAI-1 antigen may also in part reflect a redistribution following release from producer cells. PAI-1 binds strongly to vitronectin, and the localization of PAI-1 may be related to vitronectin localization. It has also been demonstrated that u-PA bound to its cellular receptor, u-PAR, is able to bind PAI-1 (Cubellis et al., 1989), and that the receptor-bound u-PAIPAI-1 complexes may subsequently become internalized and degraded (Cubellis et al., 1990; Estreicher et al., 1990; Jensen et al., 1990). PAI-1 production by one cell type may be related to a binding to that cell type of PAI-1 produced by other cells. Alpha,-proteinase inhibitor production has been reported to be enhanced by its target protease elastase in macrophages, apparently through the binding of alpha,-proteinase inhibitorielastase complexes to a serpin-protease complex receptor (Perlmutter et al., 1988, 1990). The correlation between the levels of PAI-1 and u-PA in the tumors is compatible with the existence of similar mechanisms in the present context.
Production of PAI-1 may act as a mechanism, by which the tumor protects itself against the plasminogen-activationdependent extracellular proteolysis, accomplished by the u-PA present in the tumor. Alternatively, PAI-1 may be produced by the cells surrounding the tumor cell clusters as a defense mechanism against the u-PA-dependent tissue destruction being imposed upon it by the tumor cells. Mapping of the PAI-1- and u-PA-producing cells in breast carcinomas by in situ hybridization will, in combination with immunohistochemical stainings, shed more !ight on these questions. ACKNOWLEDGEMENTS
The technical assistance of Ms. H. Baasch and Ms. T. Olesen is gratefully acknowledged. This work was supported by grants from Astrid Thaysens Legat (grant 02/90), the Danish Cancer Society (grants 90-069 and 90-066), the Danish Medical Research Council (grant 12-85419, the Danish Biotechnology Programme, the Health Research Board of Ireland, the Irish Cancer Society, and the EEC Concerted Action Scheme (4th Medical Health Programme).
REFERENCES
ANDREASEN, P.A., GEORG,B., LUND, L.R., Rrccro, A. and STACEY, S., Plasminogen-activator inhibitors: hormonally regulated serpins. Mol. cell. Endocrinol., 68,l-19 (1990). ANDREASEN, P.A., NIELSEN,L.S., KRISTENSEN, P., GR0NDAHLHANSEN, J., SKRIVER, L. and DAN0, K., Plasminogen-activator inhibitor from human fibrosarcoma cells binds urokinase-type plasminogen activator, but not its pro-enzyme.J. bid. Chem., 261,7644-7651 (1986). ANDREASEN, P.A., PYKE,C., Rrccro, A,, KRISTENSEN, P., NIELSEN, L.S., LUND,L.R., BLASI,F. and DAN^, K., Plasminogen-activator inhibitor type-1 biosynthesis and mRNA are increased by dexamethasone in human fibrosarcoma cells. Mol. cell. Eiol., 7,3021-3025 (1987). CHRISTENSEN, L. and STRANGE,L., Universal immunoperoxidase staining protocol to optimize the use of polyclonal and monoclonal antibodies. J. Histotechnol., 10, 11-15 (1987). CLAVEL,C., CHAVANEL, G. and BIREMBAUT, P., Detection of the plasmin system in human mammary pathology using immunofluorescence. CuncerRes., 46,57434747 (1986). CUBELLIS, M.V., ANDREASEN, P.A., RAGNO,P., MAYER,M., DAN0, K. and BLASI, F., Accessibility of receptor-bound urokinase to type-1 plasminogen-activator inhibitor. Proc. nut. Acad. Sci. (Wush.), 86, 4828-4832 (1989). CUBELLIS, M.V., WUN,T.-C. and BLASI,F., Receptor-mediated internalization and degradation of urokinase is caused by its specific inhibitor PAI-1. EMEOJ., 9,1079-1085 (1990). P., DAN@, K., ANDREASEN, P.A., GR0NDAHL-HANSEN, J., KRISTENSEN, NIELSEN, L.S. and SKRIVER, L., Plasminogen activators, tissue degradation and cancer.Advunc. Cancer Res., 44,139-266 (1985). DUFFY, M.J., Plasminogen activators and cancer. Blood Cougul. Fibrinol., 1,681-687 (1991). DUFFY,M.J., REILLY,D.C., O'SULLIVAN, C., O'HIGGINS, N., FENNELLY, J.J. and ANDREASEN, P.A., Urokinase-plasminogen activator, a new and independent prognostic marker in breast cancer. Cancer Res., 50,6827-6829 (1990). ESTREICHER, A., MUHLHAUSER, J., CARPENT~ER, J.-L., ORCI,L. and VASSALLI, 3.-D., The receptor for urokinase-type plasminogen activator polarizes expression of the protease to the leading edge of migrating monocytes and promotes degradation of enzyme inhibitor complexes. J. Cell Biol., 111,783-792 (1990). FEINBERG, R.F., KAO, L.-C., HAIMOWITZ, J.E., QUEENAN, J.T., WUN, T.-C., STRAUSS, J.F. and KLIMAN, H.J., Plasminogen-activator inhibitor types 1and 2 in human trophoblasts. PAI-1 is an immunocytochemical marker of invading trophoblasts. Lab. Invest., 61,20-26 (1989). JANICKE, F., SCHMITT,M., HAFTER,R., HOLRIEDER, A., BABIC,R., ULM,K., GOSSNER, W. and GRAEFF,H., Urokinase-type plasminogenactivator (u-PA) antigen is a predictor of early relapse in breast cancer. Fibrinolysis, 4,69-78 (1990). JENSEN,P.H., CHRISTENSEN, E.I., EBBESEN,P., GLIEMANN, J. and ANDREASEN, P.A., Lysosomal degradation of receptor-bound urokinase-
type plasminogen activator is enhanced by its inhibitors in human trophoblastic choriocarcinoma cells. Cell Reg., 1,1043-1056 (1990). KRIEG,P., AMTMANN, E. and SAUER,G., The simultaneous extraction of high-molecular-weight DNA and of RNA from solid tumors. Anal. Eiochem., 134,288-294 (1983). KRISTENSEN, P., PYKE,C., LUND,L.R., ANDREASEN, P.A. and DAN@, K., Plasminogen activator inhibitor type 1 in Lewis lung carcinoma. Histochemistly, 93,559-566 (1990). KRUITHOF, E.K.O., Inhibitors ofplusminogen activators. Dissertation, Faculty of Science, University of Lausanne, Switzerland (1988). KRUG,M.S. and BERGER,M.S., First-strand cDNA synthesis primed with oligo(dT). Meth. Enzymol., 152,316-375 (1988). LASKOWSKI, M. and KATO,I., Protein inhibitors of proteinases. Ann. Rev. Eiochem., 49,593-626 (1980). LIOTTA, L.A., Tumor invasion and metastasis. Role of the extracellular matrix. Rhoads Memorial Lecture. Cuncer Res., 46, 1-7 (1986). LIOTTA,L.A., STEEG,P.S. and STETLER-STEVENSEN, W.G., Cancer metastasis and angiogenesis: an imbalance of positive and negative regulation. Cell, 64,327-336 (1991). LIOTA, L.A., TRYGGVASON, K., GARBISA, s., HART,I., FOLTZ,C.M. and SHAFIE, S . , Metastatic potential correlates with enzymatic degradation of basement membrane collagen. Nature (Lond.), 284, 67-68 (1980). LOSKUTOFF, D.J., LINDERS, M., KEIJER,J., VEERMAN, H., VANHEERH., Structure of the human plasminogenIKHUIZEN, H. and PANNEKOEK, activator inhibitor 1gene: non-random distribution of introns. Eiochemistry, 26,3763-3768 (1987). LUND,L.R., GEORG,B., NIELSEN,L.S., MAYER,M., DAN@,K. and ANDREASEN, P.A., Plasminogen activator inhibitor type-1: cell-specific and differentiation-induced expression and regulation in human cell lines, as evaluated by enzyme-linked immunosorbent assay. Mol. cell. Endocrinol., 60,43-53 (1988). NIELSEN,L.S., ANDREASEN, P.A., GR0NDAHL-HANSEN, J., HUANG, J.-Y., KRISTENSEN,P. and DAN0, K., Monoclonal antibodies to human 54,000-molecular-weightplasminogen-activator inhibitor from fibrosarcoma cells. Inhibitor neutralization and one-step affinity purification. Thromb. Huemostus., 55,206-212 (1986). PANNEKOEK, H., VEERMAN, H., LAMBERS, H., DIERGARDE, P., VERWEIJ, C.L., VANZONNEFELD, A.-J. and VANMOURIK, J.A., Endothelial plasminogen activator inhibitor (PAI): a new member of the Serpin gene family. EMBOJ., 5,2539-2544 (1986). G.I., RIVETNA, M., SCHASTEEN, C.S. and PERLMUTTER, D.H., GLOVER, FALLON, R.J., Identification of a serpin-enzyme complex receptor on human hepatoma cells and human monocytes. Proc. nut Acud. Sci. (Wash.), 87,3753-3757 (1990). PERLMUITER, D.H., TRAVIS,J. and PUNSAL, P.I., Elastase regulates the synthesis of its inhibitor, alpha,-proteinase inhibitor, and exagger-
214
REILLY ET AL.
ates the dcfect in homozygous PiZZ alpha,-proteinase inhibitor deficiency. J. din. Invest., 81, 1774-1780 (1 988). POSTE,G. and FIDLER,I.J., The pathogenesis of cancer metastasis. Nature (Lond.), 283,139-146 (1978). SAKSELA, 0.and RIFKIN, D.B., Cell-associated plasminogen activation: regulation and physiological functions. Ann. Rev. Cell Biol., 4,93-126 ( 1988). SAPPINO, A.-P., Busso, N., BELIN,D. and VASSALLI, J.-D., Increase of urokinase-type plasminogen activator gene expression in human lung and breast carcinomas. Cancer Rex, 47,4043-4046 (1987). SHULMAN, M., WILDE,C.D. and KOHLER,G., A better cell line for making hybridomas secreting specific antibodies. Nature (Lond.), 276, 269-270 (1978).
SKRIVER, L., LARSSON, L.-I., KIELBERC, V., NIELSEN, L.S., ANDRESEN, P.B., KRISTENSEN, P. and DAN@, K., Immunocytochemical localization of urokinase-type plasminogen activator in Lewis lung carcinoma. J . CellBiol., 99,753-758 (1984). SKRIVER, L., LUND,L.R., ANDREASEN, P.A., KRISTENSEN, P. and DANB, K., Plasminogen activators and tumor invasion. In: H. Tschesche (ed.), Proteinuses in injammation and tumor invasion, pp. 281-303, de Gruyter, Berlin (1987). TISSOT, J.-D., HAUERT,J . and BACHMAN, F., Characterization of plasminogen activators from normal breast and colon and from breast and colon carcinomas. Int. J. Cancer, 34,295-302 (1984). YEH,I.-T. and KURMAN, R.J., Functional and morphological expressions of trophoblast. Lab. Invest., 61,l-3 (1989).
WC
PJblication of tne lnternattonai Un on Aga nst Cancer Puol cat on oe Lnion Internationale Conrre e Cancer
TYPE-1 PLASMINOGEN ACTIVATOR INHIBITOR IN HUMAN BREAST CARCINOMAS David REILLY’’, Lise CHRISTENSEN’, Mogens DUCH’,Naimh NOLAN4, Michael J. DUFFY’and Peter A. ANDREASEN” ‘Institute of Molecular Biology, University of Jrhus, Denmark; ‘Education and Research Center, St. Vincent’s Hospital, Dublin, Ireland; ’Department of Pathology, Rigshospitalet, Copenhagen, Denmark; ‘Pathology Department and ’Nuclear Medicine Department, St. VitzcentS Hospital, Dublin, Ireland. We have investigated the occurrence of type-I inhibitor of plasminogen activators (PAL I) in human breast tumors. PAL I levels, measured by enzyme-linked immunosorbent assay, were significantly higher in malignant breast carcinomas (n = 178) than in benign breast tumors (n = 25). The levels of PAL I were found to be correlated with those of urokinase-type plasminogen activator @-PA). The presence of PAL I in tumor extracts was also demonstratedby immunoblotting analysis. Immunohistochemical investigations by the use of monoclonal and polyclonal antibodies showed that PAL Iwas mostly localized in the tumor islands, associated with the tumor cells; in addition, it was present in vessel walls and in normal duct epithelia, but absent from the stroma. Analysis of RNA extracted from tumors by polymerase chain reaction revealed the presence of PAI-I mRNA. We conclude that PAI-I is present in human breast carcinoma cells, and that it is-at least partiallyproduced locally, either by the cancer cells or by other cells in the tumors. We have previously demonstratedthat a high level of u-PA in human breast carcinomas is associated with poor prognosis. These results, combined with our present findings, present 2 possibilities: either the cancer cells need PAI-I in order to utilize the u-PA-mediated pathway of plasminogen activation for invasion and metastasis; or PAI-I represents a defense mechanism against tumor invasion.
Proteolytic degradation of the extracellular matrix is considered an early and crucial feature of the metastatic process (Poste and Fidler, 1978; Liotta et al., 1980; Liotta, 1986). Increasing evidence of proteolytic enzymes playing a role in cancer invasion and metastasis has been obtained both from in vitro invasion and metastasis assays and from in vivo studies of levels and localizations of these proteases (Liotta et al., 1991). An important and highly regulated cascade of such proteolytic events involves the plasminogen activation system. Plasminogen activators are serine proteases which catalyze the conversion of the abundant zymogen plasminogen into plasmin, another serine protease of broad-spectrum trypsin-like specificity. The effect of plasmin appears to be either direct degradation of extracellular matrix components such as fibronectin, laminin and proteoglycans, or an indirect effect via the activation of latent collagenase and subsequent degradation of cotlagen. Of the 2 distinct types of plasminogen activators, the urokinase-type (u-PA) has been implicated in both normal and pathological processes of tissue remodelling, organ involution and invasion; the tissue-type (t-PA) plays an important role in fibrinolysis and thrombolysis (Dan0 et al., 1985; Skriver et al., 1987; Saksela and Rifkin, 1988; DUAL, 1991). Regulation of these proteases is achieved at many levels, including inhibition by at least 2 types of fast-acting specific inhibitors, PAI-1 and PAI-2. These inhibitors are products of different genes, have different biochemical properties and probably also different biological functions (Kruithof, 1988; Andreasen et al., 1990). It was previously demonstrated that human breast carcinomas contain u-PA (Tissot et al., 1984; Clavel et a/., 1986) and its mRNA (Sappino et al., 1987). In a recent study, we measured u-PA levels in tumor extracts by ELISA; the level of u-PA protein varied strongly between different tumors, but showed a strong inverse correlation with disease-free interval and survival (DufTy et al., 1990). Janicke et al. (1990) reported similar
findings. We now report the presence of PAI-1 protein and mRNA in human breast carcinomas. MATERIAL AND METHODS
Ma teria 1 The following materials were obtained from the indicated sources. Alkaline phosphatase-conjugated rabbit IgG against mouse IgG (Dakopatts, Copenhagen, Denmark); nitroblue tetrazolium (Grade 111) and 5-bromo-4-chloro-3-indolyl phosphate (Sigma, St. Louis, MO); nitrocellulose filters for protein blotting (type GSWP00010, Les Ulis, France); DNase I (Grade 1, Boehringer, Mannheim, Germany); avian myeloblastosis virus reverse transcriptase and Thermus aquaticus DNA polymerase I (Stratagene, La Jolla, CA). All other materials were those described previously (Andreasen et al., 1987; Christensen and Strange, 1987; Duffy et al., 1990), or of the best grade commercially available. Tissues Tumors were collected at the time of surgery and freed of visible necrosis, fat and connective tissue. They were frozen immediately in liquid nitrogen and stored at -70°C until further use. Antibodies Murine monoclonal IgGs against human PAI-1 were from hybridoma clones 1, 2 and 3, as described previously (Nielsen et aL, 1986), or from clone 6, resulting from another independent immunization. Of these, IgGs from clones 2 and 6 may have identical epitopes, while IgGs from the other clones recognize different epitopes, all of which are different from that recognized by clones 2 and 6 (data not shown). Rabbit polyclonal IgG against human PAI-1 (Andreasen et al., 1986) were adsorbed against human fibronectin in order to remove contamination with anti-fibronectin IgG. A portion of the fibronectin-adsorbed IgG was also adsorbed against PAI-1, which had been purified by immunoaffinity chromatography with monoclonal IgG from hybridoma clone 2 (Nielsen et al., 1986). The adsorbed antibodies were controlled for lack of anti-fibronectin and anti-PAI-1 reactivity, respectively, by ELISA. Monoclonal IgG against the trinitrophenyl group was that described by Shulman et al. (1978). A monoclonal IgG, L26, a B-lymphocyte marker, was purchased from Dakopatts (code M755). ELISA Tumor extracts were prepared by one of 2 methods: as previously described (Durn et al., 1990); by homogenization of tumor pieces with a glassiglass homogenizer in a buffer of 0.1 M Tris, pH 8.1,0.5% Triton X-100, 10 mM EDTA and 10 kg/ml of Trasylol, followed by centrifugation at 10,000g for 15 min.
‘To whom correspondence and reprint requests s ould be sent, at the Institute of Molecular Biology, University of rhus, 130 C.F. Mdler’s AM, 8000 k h u s C, Denmark.
A
Received: July 1, 1991 and in revised form August 20,1991.
209
PAI-I IN BREAST CARCINOMAS
The protein concentrations of the extracts were determined by the Bradford method. There was no interference from the Triton X-100 in the extraction buffer in the amounts in which it was added to the assay. The extracts were analyzed for PAI-1 by ELISA, using a mixture of IgGs from hybridoma clones 1, 2, 3 and 6 on the solid phase. This ELISA measures free PAI-1 and PAI-1 in complex with u-PA and t-PA (Lund et al., 1988). The u-PA ELISA employed measures free u-PA and u-PA/PAI-1 complexes (Lund et al., 1988).
Immunoblottinganalysis For immunoblotting analysis, PAI-1 was partly purified from tumor extracts by immunoaffinity chromatography on columns of monoclonal anti-PAI-1 IgG from hybridoma clone 1 immobilized on Sepharose (Nielsen et al., 1986). The columns were equilibrated in 0.1 M Tris, PH 8.1. The samples, usually containing a total of 4 to 5 mg protein, were then applied, and the columns washed with equilibration buffer. The columns were further washed in the same buffer with 1 M NaCI. Bound protein was then eluted with 0.1 M CH,COOH, pH 2.9, 1 M NaCI, dialysed against 0.03% SDS and lyophilized. Subsequently, the proteins were subjected to SDS-PAGE and transferred electrophoretically to nitrocellulose filters, which were stained immunochemically with mouse monoclonal antiPAI-1 IgG from hybridoma clone 1 or mouse monoclonal IgG of irrelevant specificity. For staining, the nitrocellulose filters were first washed with 0.05 M Tris, PH 7.4, 0.15 M NaCl (TBS), containing 0.3% Tween 20 (TBS-Tween). In order to block remaining binding sites on the filters, they were incubated with TBS with 30% rabbit serum. The filters were then washed with TBS-Tween, and incubated with the monoclonal antibodies at a concentration of 5 pg/ml overnight at 4°C. After 3 washes with TBS-Tween, the filters were incubated with alkalinephosphatase-conjugated rabbit anti-mouse IgG, diluted 1:100 from the commercial stock solution. Filters were washed 3 times with TBS-Tween and once in 0.05 M Tris, pH 7.4. Finally, the filters were washed with 0.1 M Tris, pH 9.5, 0.1 M NaCI, 5 mM MgCI, and developed using nitroblue tetrazolium (0.33 mg/ml) and 5-bromo-4-chloro-3-indolyl phosphate (0.17 mg/ ml) in the same buffer. Immunohistochemical stainings Frozen sections were cut from tumor samples at 5 Km, air-dried at room temperature for 24 hr and fixed in acetone for 10 min. Two sections were placed on the same slide representing the tumor at 2 different levels. These were stained with monoclonal IgG against PAI-1, polyclonal antibodies against PAI-1 o r control antibodies, including monoclonal IgG of irrelevant specificity (L26) and polyclonal antibodies that had been absorbed against PAI-1 in order to remove anti-PAI-1 reactivity (see above). All primary antibodies were incubated at a concentration of 10 pg/ml for 20 hr at 4°C and then for 2 hr at room temperature. The sites of reaction of the primary antibodies were visualized by conventional immunoperoxidase techniques (Christensen and Strange, 1987). The chromogen was 3-amino-9-ethylcarbazole,and Mayer’s hemalun was used as nuclear counterstaining. In some of the tumors, 2 sections for immunohistochemical staining were cut from each end of the tissue block, and sections from the intervening 100 to 200 mg of tissue were extracted for ELISA. Other tumors were cut in serial sections, and different sections were stained with different antibodies. RhLA extraction and analysis
Total RNA was extracted from solid tumors as described by Krieg et al. (1983), with minor modifications. Briefly, 0.2 to 0.5 g of tumor tissue, frozen in liquid nitrogen, was homogenized by mechanical disruption, using a Braun Mikro-dismembrator. Nucleic acids were extracted with 0.3 M Na acetate, pH 7.5,
0.5% SDS, 5 mM EDTA, followed by phenol extraction and ethanol precipitation. The precipitated nucleic acids were re-dissolved in sterile H,O. R N A was then selectively precipitated with an equal volume of 4 M L E I . All samples were treated with 20 pg/ml DNase I and finally re-dissolved in a buffer of 10 mM Tris, PH 7.4, 1 mM EDTA. Concentration and quality of RNA was checked spectrophotometrically and by agarose gel electrophoresis. For analysis of R N A samples for the presence of PAI-I mRNA by polymerase chain reaction (PCR), single-stranded cDNA was first synthesized by reverse transcription, as described by Krug and Berger (1988), with minor modifications. Briefly, 3 pg of total R N A were suspended in 9 pl sterile H,O and heated to 65°C for 5 min, placed on ice, and mixed with 1.5 p1 of a buffer of 500 mM Tris-HC1, pH 8.3,500 mM KCl, 100 mM MgCI,, 10 mM dithiothreitol, 10 mM EDTA, 100 pgirnl BSA. One microliter 10 mM dNTP mixture, 1 pl 150 pg/ml oligo dT, 0.75 pl 10 mM spermidine, and 0.75 pl 80 mM NaPP, were added. Finally, 10 U of avian myeloblastosis virus reverse transcriptase, diluted in a buffer of 10 mM K, HPO,, pH 7.4, 10% glycerol, 0.2% Triton X-100 and 2 mM dithiothreitol were added and the reaction was carried out at 37°C for 1 hr. The 2 oligonucleotides used for amplification of PAI-1 cDNA in the reverse transcription reaction mixture were 5’-GAGGTGCCTCTCTCTGCCCTCACCAACATT-3‘ and 5’-AGCCTGAAACTGTCTGAACATGTCG-3’, corresponding to sequences in the PAI-1 cDNA at nucleotides 925-955 and 1082-1108 (Pannekoek et al., 1986). The oligonucleotides were synthesized on an Applied Biosystems (Foster City, CA) Oligonucleotide Synthesizer. PCR amplification was done according to the Stratagene protocol, except that the annealing temperature was set at 62°C. Denaturation and annealing times were 1.2 min, and extension was 1 min. After 44 rounds of amplification, the PCR products were analyzed by electrophoresis in 3% agarose gels. Statistics All statistical analyses were carried out with IBM Statistics and Graphics System Package, Version 1.2. The data were analyzed by non-parametric methods, using Spearman rank correlation coefficient.
RESULTS
Demonstration of PAI-1 in tumor extracts by ELISA and immunoblottinganalysis The levels of PAI-1 and u-PA were measured by ELISAs in extracts of 178 breast carcinomas and 25 benign breast tumors. The PAI-1 ELISA measured free PAI-1 as well as complexes between PAI-1 and u-PA or t-PA. The u-PA ELISA measured free u-PA as well as u-PAiinhibitor complexes. These facts indicated that the measured levels of PAI-1 and u-PA were mutually independent. It was found that the levels of both PAI-1 (Fig. 1) and u-PA (not shown) were significantly higher in the carcinomas than in the benign tumors. In addition, there was a significant correlation between the levels of the 2 proteins (Fig. 2). In contrast, no significant correlation was found between.PA1-1 and t-PA levels (data not shown). Among the carcinomas, PAI-1 levels did not correlate with a number of histological parameters, including tumor volume, grade and cellularity; lymphocyte, mast cell or eosinophil infiltration; axillary-node status; and lymphatic invasion or skin invasion. The PAI-1 level was in the range 1 to 100 ngimg total protein. For comparison, the plasma level of PAI-1 is in the range 0.02 to 0.7 ng/mg protein, and the serum level around 2.5 ngimg protein (Kruithof, 1988).
210
REILLY ET AL. 100
I
MALIGNANT
BENIGN
FIGURE1-PAI-1 antigen levels in breast carcinomas and benign tumors. Extracts from 178 carcinomas and 25 benign tumors were prepared by method 1 (see “Material and Methods”) and analyzed for PAI-1 by ELISA. PAI-1 levels greater than 10 ng/mg total protein were found in 78% of the malignant breast carcinomas. In contrast, levels higher than 10 ngiml were found in only 20% of the benign samples. There was a significant difference between the 2 groups ( p < 0.0005). inn
.--
I
r
I
.
.
0.1
. . . 1.0
I
.. . . . .
I I
10
100
U-PA (ngirng protein)
FIGURE2 - Correlation between PAI-1 and u-PA antigen levels in breast carcinomas. Extracts from 178 tumors were prepared by the first method (see “Material and Methods”) and analyzed for PAI-1 and u-PA by ELISA. Slope 0 . 5 3 , ~< 0.0005. The presence of PAI-1 in tumor extracts was also demonstrated by immunoblotting analysis. PAI-1 was first partly purified from the extracts by immunoaffinity chromatography on a column of monoclonal anti-PAI-1 IgG immobilized on Sepharose. The partly purified preparations were subjected to immunoblotting analysis, incubating the filters with either monoclonal anti-PAI-1 IgG or monoclonal IgG of irrelevant specificity. Incubation with anti-PAI-1 IgG, but not IgG of irrelevant specificity, gave staining of a band co-migrating with purified PAI-1; no other bands were stained (Fig. 3). Fibrin-agarose zymography of 30 tumor extracts (not shown) revealed the presence of u-PA exclusively as an M, 54,000 band. The M, 36,000 degradation product was not found. I n addition, no PAI-1 /u-PA inhibitor complexes were detected, in agreement with the results of the immunoblotting analysis.
FIGURE 3 - Immunoblotting analysis of PAI-1 in extracts of breast carcinomas. PAI-1 was partly purified from tumor extracts by immunoaffinity chromatography. The partly purified preparations were subjected to SDS-PAGE, and then transferred electrophoretically to nitrocellulose filters. The filters were stained immunochemically. Lanes a and a’ contain 35 ng of purified PAI-1, and lanes b, c, b’, and c’ representative tumor samples. Lanes a-c were stained with monoclonal anti-PAI-1 IgG from hybridoma clone 1, and lanes a’-c’ with monoclonal antitrinitrophenyl IgG. The position of PAI-1 is indicated to the left.
Immunohistochemical localization of PAI-1 in breast tumors Sections from 43 breast carcinomas were stained immunohistochemically for PAI-1 with monoclonal IgG from different hybridoma clones, or with polyclonal IgG. The sections were cut from frozen tumor samples. Formalin-futed tissue was found unsuitable due to loss of antigenicity during processing. All anti-PAI-1 antibodies consistently showed PAI-1 immunoreactivity associated with the tumor cells, although in variable amounts in different areas. Also epithelial cells from benign-looking ducts and glands were stained. PAI-1 immunoreactivity appeared to be associated with the basement membrane of the ducts. PAI-1 was also present in blood vessels. Stromal cells were not stained (Fig. 4). The staining was heterogeneous, and the staining intensity varied from one tumor to the other. However, no clear correlation could be demonstrated between staining intensity and the histological carcinoma sub-type. Among the monoclonal IgGs tested, that from hybridoma clone 6 gave the strongest staining; IgG from hybridoma clones 1 and 2 also stained, but more weakly. The following controls assured the specificity of the staining: Monoclonal IgG of irrelevant specificity and polyclonal IgG, that had been adsorbed against PAI-1 in order to remove anti-PAI-1 reactivity, did not give any staining (Fig. 4b and d ) . This indicates that the staining is due to specific antigenantibody reaction and not a non-immunological binding of antibodies to tissue components. When adjacent sections of a tumor were stained with monoclonal anti-PAI-1 IgGs with different epitopes and with polyclonal IgG, the same areas were found to contain PAI-1 immunoreactivity (not shown), indicating that cross-reacting antigens in the tissue did not account for the staining. The immunoblotting analysis showed that monoclonal antiPAI-1 IgG reacted specifically with only 1 band in extracts of the tumors (Fig. 4), again indicating the absence of crossreacting antigens in the tissue. Parts of some of the tumors used for staining were extracted, and the amounts of PAI-1 in the extracts were determined by ELISA. The ELISA values were compared blindly with a semi-quantitative grading of the intensity of the staining with
PAI-1 IN BREAST CARCINOMAS
211
FIGURE 4 - Immunohistochemical localization of PAI-1 in human breast carcinomas. Sections were cut from frozen tumor samples, and incubated with various anti-PAI-1 and control antibodies. ( a ) and (b) show step sections of an invasive mammary-duct carcinoma stained with a polyclonal antibody against PAL1 ( a ) and with the polyclonal antibody adsorbed against purified PAI-1 antigen (b).The tumor cells show a heterogeneous PAL1 immunoreactivity, whereas the normal duct (bottom left quarter of the micrograph) is more uniformly stained, with a tendency to concentrate towards the basement-membrane area. Blood vessels are stained as well (bottom center) ( a ) . The absorbed antibody gives no staining reaction (b). Magnification 87-fold. ( c ) and (d) show step sections of a well differentiated duct carcinoma stained with a monoclonal antibody against PAI-1 (from hybridoma clone 2) ( c ) and an irrelevant MAb L26 against B lymphocytes (d). The tumor cells of the irregular gland-like structures (center) show a distinct immunoreactivity for PAI-1. The same is true for the epithelial cells and the basement-membrane area of a normal-looking duct (top right) and endothelial cells (arrows) ( c ) . No cellular staining is seen with the irrelevant antibody (d). Magnification 87-fold. (e) and (f) show close-up photographs of step sections from an invasive tumor island of a large-cell duct carcinoma. Tumor cell nuclei appear empty, due to freezing artifacts. In (e) the tumor cells are intensely stained with a MAb to PAI-1 (from hybridoma clone 6). In (f) the tumor cells fail to stain by the irrelevant antibody L26 against B-lymphocytes. Magnification 250-fold.
212
REILLY ET AL.
monoclonal IgG on a scale from 0 to 4. T h e ELISA values vary in parallel with the grading of the staining (Fig. 5). PAI- 1 mRNA RNA was extracted from 24 breast carcinomas and analyzed for the presence of PAI-1 mRNA by PCR analysis. First, single-strand cDNA was synthesized by the use of reverse transcriptase with tumor R N A as a template. Synthesized PAI-1 cDNA was then amplified by the use of 2 primers, expected to result in a product of 183 b p (see “Material and Methods”). A number of measures were taken to exclude signals created by contaminants. Firstly, primers were chosen belonging to 2 different exons (Loskutoff et al., 1987), thereby excluding the possibility of amplification due to trace amounts of genomic DNA surviving the DNase treatment during the R N A preparation; such contaminants would result in a PCR product of about 2 Kb. Secondly, before amplification, all samples and controls were treated with RNase 1, in order to remove RNA, and with Alu I, which will cleave any contaminating doublestrand PAI-1 cDNA between the 2 primers, but not cleave the single-strand cDNA resulting from the reverse transcriptase reaction. Thirdly, controls were run in which either reverse transcriptase or RNA was omitted from the reverse transcriptase reaction. As shown in Figure 6, no PCR product was obtained from the controls without R N A or reverse transcriptase. In contrast, all tumor R N A samples resulted in clear bands, which by comparison to molecular size markers were found to have the expected size (Fig. 6), and to give the expected restriction fragments (not shown). We therefore concluded that the breast tumors contain PAI-1 mRNA.
0 0
2
4
6
8
10
1
P A I - 1 (nghng protein)
FIGURE5 - Comparison of intensity of immunohistochemical staining of PAI-1 and PAI-1 ELISA values in breast carcinomas. Sections of 15 breast carcinomas were stained immunohistochemically by the use of monoclonal anti-PAI-1 IgG from hybridoma clone 1. The intensity of the staining was graded on a scale from 0 to 4 (ordinate). Other pieces of the same tumors were extracted by the second method (see “Material and Methods”), and the PAI-1 level determined by ELISA (abcissa). The linear correlation coefficient for the line drawn, calculated by non-weighted linear regression analysis, was 0.63. With 15 samples, this corresponds to a p value of 0.019.
FIGURE6 - Demonstration of PAI-1 mRNA in human breast carcinomas by PCR. RNA from tumors was used as a template for reverse transcription, and the resulting single-strand cDNA amplified by PCR. Lanes 1 to 24 represent the PCR products obtained from tumor samples. Lanes 25 and 27 represent controls, in which RNA was omitted during reverse transcription. Lanes 26 and 28 represent controls, in which RNA from the tumor samples shown in lanes 11 and 12 was incubated for reverse transcription in the absence of reverse transcriptase. The PCR products were electrophoresed in 3% agarose gels. Size markers (unnumbered lane between lanes 14 and 15): 220,234,398,453,517, 653, 1033, 1766 and 2176 bp. DISCUSSION
The amount of u-PA in extracts of primary human breast carcinomas is inversely correlated with the disease-free interval of the patient (DuEy et al., 1990; Janicke et al., 1990). This observation is in agreement with the hypothesis that u-PA is implicated in tumor invasion and metastasis (Dan0 etal., 1985; Skriver et al., 1987; Saksela and Rifkin, 1988; Duffy, 1991). We show here that the level of PAI-1 is higher in breast carcinomas than in benign tumors, and that the level of PAI-1 in the tumor extracts is correlated with the level of u-PA. Although the correlation between PAI-1 and u-PA levels is highly significant, we have not found any significant relationship between the PAI-1 level and the risk of disease recurrence at the present stage of patient follow-up. Yet the presence of PAI-1 in these tumors at a level correlating with that of u-PA indicates that PAI-1 may play a role in regulation of plasminogen activation, and thereby of invasion and metastasis, in breast cancer. Our findings are in agreement with the previous report of PAI-1 being present within tumor cells in the transplantable murine Lewis lung carcinoma (Kristensen et al., 1990), which also contains high levels of u-PA (Skriver et al., 1984). The co-existence of a protease and its inhibitor within a tissue is not unique to the plasminogen activation system. Laskowski and Kato (1980) pointed out that the presence of secretory trypsin inhibitors in the pancreas of vertebrates, together with high levels of trypsinogen, may serve to prevent premature activation of trypsinogen and other pancreatic zymogens. PAI-1 has also been reported to be a marker of invading trophoblastic cells, representative of a controlled invasion process which is believed to involve the plasminogen activation system (Feinberget al., 1989; Yeh and Kurman, 1989). The present findings pose important questions as to which cells are responsible for PAI-1 biosynthesis in breast tumors. PAI-1 occurs in high levels in blood platelets, and is also present in blood plasma (Kruithof, 1988; Andreasen et al., 1990). However, the high level of PAI-1 in the tumor extracts, as compared with that in plasma and serum (see above), excludes plasma, serum and platelet contamination as the sole sources of PAI-1 in tumor extracts. Also, the demonstration of PAI-1 mRNA in the tumors excludes the possibility that blood platelets are the only source of PAI-1 in the tumors, and shows that at least part of the PAI-1 must b e produced locally. The association of PAI-1 antigen with the tumor cells is compatible with the idea that at least part of
213
PAI-1 I N BREAST CARCINOMAS
the PAI-1 is produced by these cells. A similar staining of vessel walls and normal ducts suggests PAI-1 production there as well. Irrespective of the cell type producing PAI-1 in and around these tumors, the localization of PAI-1 antigen may also in part reflect a redistribution following release from producer cells. PAI-1 binds strongly to vitronectin, and the localization of PAI-1 may be related to vitronectin localization. It has also been demonstrated that u-PA bound to its cellular receptor, u-PAR, is able to bind PAI-1 (Cubellis et al., 1989), and that the receptor-bound u-PAIPAI-1 complexes may subsequently become internalized and degraded (Cubellis et al., 1990; Estreicher et al., 1990; Jensen et al., 1990). PAI-1 production by one cell type may be related to a binding to that cell type of PAI-1 produced by other cells. Alpha,-proteinase inhibitor production has been reported to be enhanced by its target protease elastase in macrophages, apparently through the binding of alpha,-proteinase inhibitorielastase complexes to a serpin-protease complex receptor (Perlmutter et al., 1988, 1990). The correlation between the levels of PAI-1 and u-PA in the tumors is compatible with the existence of similar mechanisms in the present context.
Production of PAI-1 may act as a mechanism, by which the tumor protects itself against the plasminogen-activationdependent extracellular proteolysis, accomplished by the u-PA present in the tumor. Alternatively, PAI-1 may be produced by the cells surrounding the tumor cell clusters as a defense mechanism against the u-PA-dependent tissue destruction being imposed upon it by the tumor cells. Mapping of the PAI-1- and u-PA-producing cells in breast carcinomas by in situ hybridization will, in combination with immunohistochemical stainings, shed more !ight on these questions. ACKNOWLEDGEMENTS
The technical assistance of Ms. H. Baasch and Ms. T. Olesen is gratefully acknowledged. This work was supported by grants from Astrid Thaysens Legat (grant 02/90), the Danish Cancer Society (grants 90-069 and 90-066), the Danish Medical Research Council (grant 12-85419, the Danish Biotechnology Programme, the Health Research Board of Ireland, the Irish Cancer Society, and the EEC Concerted Action Scheme (4th Medical Health Programme).
REFERENCES
ANDREASEN, P.A., GEORG,B., LUND, L.R., Rrccro, A. and STACEY, S., Plasminogen-activator inhibitors: hormonally regulated serpins. Mol. cell. Endocrinol., 68,l-19 (1990). ANDREASEN, P.A., NIELSEN,L.S., KRISTENSEN, P., GR0NDAHLHANSEN, J., SKRIVER, L. and DAN0, K., Plasminogen-activator inhibitor from human fibrosarcoma cells binds urokinase-type plasminogen activator, but not its pro-enzyme.J. bid. Chem., 261,7644-7651 (1986). ANDREASEN, P.A., PYKE,C., Rrccro, A,, KRISTENSEN, P., NIELSEN, L.S., LUND,L.R., BLASI,F. and DAN^, K., Plasminogen-activator inhibitor type-1 biosynthesis and mRNA are increased by dexamethasone in human fibrosarcoma cells. Mol. cell. Eiol., 7,3021-3025 (1987). CHRISTENSEN, L. and STRANGE,L., Universal immunoperoxidase staining protocol to optimize the use of polyclonal and monoclonal antibodies. J. Histotechnol., 10, 11-15 (1987). CLAVEL,C., CHAVANEL, G. and BIREMBAUT, P., Detection of the plasmin system in human mammary pathology using immunofluorescence. CuncerRes., 46,57434747 (1986). CUBELLIS, M.V., ANDREASEN, P.A., RAGNO,P., MAYER,M., DAN0, K. and BLASI, F., Accessibility of receptor-bound urokinase to type-1 plasminogen-activator inhibitor. Proc. nut. Acad. Sci. (Wush.), 86, 4828-4832 (1989). CUBELLIS, M.V., WUN,T.-C. and BLASI,F., Receptor-mediated internalization and degradation of urokinase is caused by its specific inhibitor PAI-1. EMEOJ., 9,1079-1085 (1990). P., DAN@, K., ANDREASEN, P.A., GR0NDAHL-HANSEN, J., KRISTENSEN, NIELSEN, L.S. and SKRIVER, L., Plasminogen activators, tissue degradation and cancer.Advunc. Cancer Res., 44,139-266 (1985). DUFFY, M.J., Plasminogen activators and cancer. Blood Cougul. Fibrinol., 1,681-687 (1991). DUFFY,M.J., REILLY,D.C., O'SULLIVAN, C., O'HIGGINS, N., FENNELLY, J.J. and ANDREASEN, P.A., Urokinase-plasminogen activator, a new and independent prognostic marker in breast cancer. Cancer Res., 50,6827-6829 (1990). ESTREICHER, A., MUHLHAUSER, J., CARPENT~ER, J.-L., ORCI,L. and VASSALLI, 3.-D., The receptor for urokinase-type plasminogen activator polarizes expression of the protease to the leading edge of migrating monocytes and promotes degradation of enzyme inhibitor complexes. J. Cell Biol., 111,783-792 (1990). FEINBERG, R.F., KAO, L.-C., HAIMOWITZ, J.E., QUEENAN, J.T., WUN, T.-C., STRAUSS, J.F. and KLIMAN, H.J., Plasminogen-activator inhibitor types 1and 2 in human trophoblasts. PAI-1 is an immunocytochemical marker of invading trophoblasts. Lab. Invest., 61,20-26 (1989). JANICKE, F., SCHMITT,M., HAFTER,R., HOLRIEDER, A., BABIC,R., ULM,K., GOSSNER, W. and GRAEFF,H., Urokinase-type plasminogenactivator (u-PA) antigen is a predictor of early relapse in breast cancer. Fibrinolysis, 4,69-78 (1990). JENSEN,P.H., CHRISTENSEN, E.I., EBBESEN,P., GLIEMANN, J. and ANDREASEN, P.A., Lysosomal degradation of receptor-bound urokinase-
type plasminogen activator is enhanced by its inhibitors in human trophoblastic choriocarcinoma cells. Cell Reg., 1,1043-1056 (1990). KRIEG,P., AMTMANN, E. and SAUER,G., The simultaneous extraction of high-molecular-weight DNA and of RNA from solid tumors. Anal. Eiochem., 134,288-294 (1983). KRISTENSEN, P., PYKE,C., LUND,L.R., ANDREASEN, P.A. and DAN@, K., Plasminogen activator inhibitor type 1 in Lewis lung carcinoma. Histochemistly, 93,559-566 (1990). KRUITHOF, E.K.O., Inhibitors ofplusminogen activators. Dissertation, Faculty of Science, University of Lausanne, Switzerland (1988). KRUG,M.S. and BERGER,M.S., First-strand cDNA synthesis primed with oligo(dT). Meth. Enzymol., 152,316-375 (1988). LASKOWSKI, M. and KATO,I., Protein inhibitors of proteinases. Ann. Rev. Eiochem., 49,593-626 (1980). LIOTTA, L.A., Tumor invasion and metastasis. Role of the extracellular matrix. Rhoads Memorial Lecture. Cuncer Res., 46, 1-7 (1986). LIOTTA,L.A., STEEG,P.S. and STETLER-STEVENSEN, W.G., Cancer metastasis and angiogenesis: an imbalance of positive and negative regulation. Cell, 64,327-336 (1991). LIOTA, L.A., TRYGGVASON, K., GARBISA, s., HART,I., FOLTZ,C.M. and SHAFIE, S . , Metastatic potential correlates with enzymatic degradation of basement membrane collagen. Nature (Lond.), 284, 67-68 (1980). LOSKUTOFF, D.J., LINDERS, M., KEIJER,J., VEERMAN, H., VANHEERH., Structure of the human plasminogenIKHUIZEN, H. and PANNEKOEK, activator inhibitor 1gene: non-random distribution of introns. Eiochemistry, 26,3763-3768 (1987). LUND,L.R., GEORG,B., NIELSEN,L.S., MAYER,M., DAN@,K. and ANDREASEN, P.A., Plasminogen activator inhibitor type-1: cell-specific and differentiation-induced expression and regulation in human cell lines, as evaluated by enzyme-linked immunosorbent assay. Mol. cell. Endocrinol., 60,43-53 (1988). NIELSEN,L.S., ANDREASEN, P.A., GR0NDAHL-HANSEN, J., HUANG, J.-Y., KRISTENSEN,P. and DAN0, K., Monoclonal antibodies to human 54,000-molecular-weightplasminogen-activator inhibitor from fibrosarcoma cells. Inhibitor neutralization and one-step affinity purification. Thromb. Huemostus., 55,206-212 (1986). PANNEKOEK, H., VEERMAN, H., LAMBERS, H., DIERGARDE, P., VERWEIJ, C.L., VANZONNEFELD, A.-J. and VANMOURIK, J.A., Endothelial plasminogen activator inhibitor (PAI): a new member of the Serpin gene family. EMBOJ., 5,2539-2544 (1986). G.I., RIVETNA, M., SCHASTEEN, C.S. and PERLMUTTER, D.H., GLOVER, FALLON, R.J., Identification of a serpin-enzyme complex receptor on human hepatoma cells and human monocytes. Proc. nut Acud. Sci. (Wash.), 87,3753-3757 (1990). PERLMUITER, D.H., TRAVIS,J. and PUNSAL, P.I., Elastase regulates the synthesis of its inhibitor, alpha,-proteinase inhibitor, and exagger-
214
REILLY ET AL.
ates the dcfect in homozygous PiZZ alpha,-proteinase inhibitor deficiency. J. din. Invest., 81, 1774-1780 (1 988). POSTE,G. and FIDLER,I.J., The pathogenesis of cancer metastasis. Nature (Lond.), 283,139-146 (1978). SAKSELA, 0.and RIFKIN, D.B., Cell-associated plasminogen activation: regulation and physiological functions. Ann. Rev. Cell Biol., 4,93-126 ( 1988). SAPPINO, A.-P., Busso, N., BELIN,D. and VASSALLI, J.-D., Increase of urokinase-type plasminogen activator gene expression in human lung and breast carcinomas. Cancer Rex, 47,4043-4046 (1987). SHULMAN, M., WILDE,C.D. and KOHLER,G., A better cell line for making hybridomas secreting specific antibodies. Nature (Lond.), 276, 269-270 (1978).
SKRIVER, L., LARSSON, L.-I., KIELBERC, V., NIELSEN, L.S., ANDRESEN, P.B., KRISTENSEN, P. and DAN@, K., Immunocytochemical localization of urokinase-type plasminogen activator in Lewis lung carcinoma. J . CellBiol., 99,753-758 (1984). SKRIVER, L., LUND,L.R., ANDREASEN, P.A., KRISTENSEN, P. and DANB, K., Plasminogen activators and tumor invasion. In: H. Tschesche (ed.), Proteinuses in injammation and tumor invasion, pp. 281-303, de Gruyter, Berlin (1987). TISSOT, J.-D., HAUERT,J . and BACHMAN, F., Characterization of plasminogen activators from normal breast and colon and from breast and colon carcinomas. Int. J. Cancer, 34,295-302 (1984). YEH,I.-T. and KURMAN, R.J., Functional and morphological expressions of trophoblast. Lab. Invest., 61,l-3 (1989).
