Comparative evaluation of urokinase-type plasminogen activator receptor expression in primary breast carcinomas and on metastatic tumor cells



The urokinase-type plasminogen activator receptor (uPAR, CD87) plays a central role in the plasminogen activation cascade, which participates in extracellular matrix degradation, cell migration and invasion. Here we performed a comprehensive immmunohistochemical evaluation of uPAR expression in primary tumor cells, tumor-surrounding fibroblasts, lymph node metastases and micrometastatic cells in bone marrow of patients with breast carcinomas at the time of primary diagnosis. Variable degrees of uPAR staining of tumor cells were observed in 84 of 93 (90%) carcinomas, whereas intratumoral fibroblasts were uPAR-positive in 70 (75%) carcinomas. The fraction of uPAR-positive primary tumor cells but not fibroblasts was positively correlated with the presence of tumor cells in bone marrow (p = 0.037), whereas no correlation with lymph node metastasis was found. Immunophenotyping of bone marrow and lymph node specimens revealed expression of uPAR on metastatic tumor cells in 10 of 13 and 22 of 23 cases, respectively. Direct comparison to the autologous primary tumor cells showed different uPAR staining scores in most patients with evidence for both up- and downregulation of uPAR on metastatic cells. Our results indicate that uPAR plays an active role in breast cancer metastasis and may therefore be a promising target for new biologic therapies. © 2003 Wiley-Liss, Inc.

The plasminogen activation system plays an essential role in fibrinolysis and extracellular matrix degradation, including cancer invasion, metastasis and angiogenesis, and its components are expressed in tumor cells and normal cells, including tumor-surrounding fibroblasts.1, 2, 3 The urokinase-type plasminogen activator (uPA), a serin-proteinase, catalyzes the conversion of the proenzyme plasminogen to the active broad-spectrum serin-proteinase plasmin, which is a central regulator of the activation of other proteinases, including matrix metalloproteinases (MMPs).1, 2, 4 In general, plasmin is generated pericellular at the cell surface by complex formation of uPA and its receptor uPAR (CD87), and its generation is regulated by plasminogen activator inhibitors PAI-1 and PAI-2.1, 2 uPAR is an Mr 55,000–60,000 glycoprotein, attached to the cell membrane by a glycosyl-phosphatidyl-inositol (GPI) anchor, which has no intracellular domain and no transmembrane peptide part.1, 5, 6 Therefore, signal transduction processes, which have been related to uPAR, require contact formation with direct signal transducing molecules.1, 2, 5, 6 The signal transduction cascades induced lead to cell adhesion, migration, differentiation, chemotaxis and proliferation.1, 2, 5, 6 In breast cancer, elevated uPAR protein in the primary tumor is associated with poor prognosis due to an increased rate of metastatic relapse.1, 2 Nevertheless, uPAR is expressed by both tumor cells and surrounding fibroblasts and it is still under debate which type of uPAR expression contributes most to tumor cell spreading to regional lymph nodes or distant organs. It is also unclear to which degree the expression of uPAR on tumor cells is being modulated during their passage from the primary lesion to the metastatic site.

Bone marrow represents a relevant site of distant metastasis in breast cancer. Recent studies on more than 2,500 patients with primary disease have indicated that the presence of immunostained tumor cells in bone marrow is associated with an unfavorable prognosis.7, 8, 9, 10, 11 There is, however, an ongoing discussion whether this parameter is an independent prognostic factor relevant in all subgroups of patients. Staining of large cohorts of bone marrow from control patients has demonstrated the high specificity of single monoclonal antibodies against cytokeratins (CKs).9 The detection of disseminated tumor cells in bone marrow has introduced a new opportunity to evaluate which of the diverse biologic characteristics of the primary tumor might favor the early blood-borne dissemination of its cells.

The aim of this study was to perform a comprehensive immunohistochemical analysis of the uPAR expression pattern on primary tumor cells, surrounding fibroblasts, disseminated tumor cells present in bone marrow, and lymph node metastases in patients with primary diagnosis of breast cancer. Results of these immunostainings were compared and related to clinical and pathologic data. Our findings yield important information on the role of uPAR in the onset of breast cancer metastasis with important implications for adjuvant biologic therapies targeting uPAR.


APAAP, alkaline phosphatase antialkaline phosphatase; CK, cytokeratin; mAb, monoclonal antibody; MMPs, matrix metalloproteinases; MNCs, mononucleated interphase cells; PBS, phosphate-buffered NaCl solution, pH 7.4; uPA, urokinase-type plasminogen activator; uPAR, urokinase-type plasminogen activator receptor.


Patients and specimens

Between April 1999 and May 2001, 93 breast cancer patients had undergone surgical treatment at the Department of Gynecology, University Hospital Hamburg-Eppendorf, in Hamburg and the Department of Gynecology and Obstetrics, University Hospital Kiel, in Kiel, Germany. From all patients, bone marrow aspirates were obtained before surgery from the upper iliac crests after providing written informed consent. The primary tumors were completely removed and classified by a certified pathologist. The following clinicopathologic data were collected from patient files and pathology reports: Tumor-Node-Metastasis (TNM) classification, histologic type and grading, estrogen and progesterone receptor status detected by immunohistochemistry (Table I). This study has been approved by the local ethics committee.

Table I. Correlation Between the Fraction of UPAR-Positive Tumor Cells on the Primary Tumor and Primary Tumor Characteristics
CharacteristicAll patientsFraction of uPAR-positive tumor cells in %
  • a

    Values in parentheses are the percentages of all patients with the characteristic. Deviations from 100% are due to rounding errors.

  • b

    Other histological types were tubular (n = 3), medullary (n = 3), ductal-lobular (n = 3), papillary (n = 1) and mixed-type carcinoma (n = 1).

  • c

    Statistically significant (p = 0.006, generalized exact Fisher test; p = 0.047, linear association test).

Total939 (10)a8 (9)21 (23)37 (40)18 (19)
Histologic type      
 Ductal carcinoma525 (10)3 (6)16 (31)20 (39)8 (15)
 Lobular carcinoma302 (7)3 (10)3 (10)13 (43)9 (30)
 Otherb112 (18)2 (18)2 (18)4 (36)1 (9)
Tumor size      
 pT1473 (6)3 (6)10 (21)24 (51)7 (15)
 pT2445 (11)5 (11)11 (25)12 (27)11 (25)
 pT3/421 (50)0 (0)0 (0)1 (50)0 (0)
Tumor grade      
 G1101 (10)1 (10)1 (10)3 (30)4 (40)
 G2513 (6)6 (12)9 (18)23 (45)10 (20)
 G3325 (16)1 (3)11 (34)11 (34)4 (13)
Lymph node status      
 pN0667 (11)6 (9)14 (21)25 (38)14 (21)
 pN1/2272 (7)2 (7)7 (26)12 (44)4 (15)
Estrogen receptor status      
 Negative212 (10)2 (10)10 (48)7 (33)0 (0)
 Positive727 (10)6 (8)11 (15)30 (42)18 (25)c
Progesterone receptor status      
 Negative273 (11)3 (11)8 (30)10 (37)3 (11)
 Positive666 (9)5 (8)13 (20)27 (41)15 (23)

uPAR staining of tissue sections

Tissue samples from all primary tumors (n = 93) and positive lymph nodes from 23 patients classified as pN1/2 (total, n = 27) were analyzed for uPAR expression. All samples were routinely formalin-fixed and paraffin-embedded. Antibodies were diluted in 10% AB serum (Biotest, Dreieich, Germany) in PBS. After deparaffinization, 2–4 μm thick sections were blocked for 20 min with 10% AB serum (Biotest) in PBS. After the blocking step, the slides were incubated for 45 min with monoclonal antibody (mAb) R2 (IgG1), directed against uPAR, at a concentration of 5 μg/ml.12 The specificity of R2 has previously been tested and confirmed with Western blotting.12 MOPC-21 (Sigma, Deisenhofen, Germany), an unrelated mouse myeloma immunoglobulin, served as the isotype control at a concentration of 5 μg/ml. MOPC-21 represents an intact monoclonal antibody of the IgG1 subtype with κ light chains, as is mAb R2 (IgG1, κ) specific for uPAR.12 The secondary layer was EnVision (Dako, Hamburg, Germany), consisting of an alkaline phosphatase-labeled dextran polymer conjugated with goat antimouse and goat antirabbit immunoglobulins, which was applied undiluted for 30 min. EnVision conjugated with alkaline phosphatase is only available with goat antimouse and goat antirabbit immunoglobulins. The goat antimouse immunoglobulins bind to mAb R2. In a third step, the slides were incubated with an alkaline phosphatase antialkaline phosphatase (APAAP) complex (Dako) at a dilution of 1:100 for 30 min in order to intensify the staining. The mouse antibody of the APAAP complex is bound by the goat antimouse antibodies of EnVision. The staining was developed with a substrate solution containing 0.2 mg/ml naphtol AS-MX phosphate (Sigma) diluted in dimethyl-formamide (Merck, Darmstadt, Germany) and 1 mg/ml fast blue BB salt (Sigma) diluted in 0.1 M Tris buffer (pH 8.2). Levamisole (0.25 mg/ml; Sigma) was used to block endogenous phosphatase activity. The colon cancer cell line HT-29 (ATCC HTB 38, American Type Culture Collection, Rockville, MD) and the breast cancer cell line MCF7 (ATCC HTB 22) served as positive controls. Without prior dehydration, the tissues were mounted with Aquatex (Merck) and covered with coverslips. The sections were evaluated independently by 2 investigators (A.H., L.R.) unaware of the clinical data. The degree of concordance was 85% and a consensus decision was made for the remaining 15% of the samples.

Semiquantitative evaluation of immunostaining

Immunostaining of tumor cells was evaluated by examination of staining intensity (negative, weak, moderate, strong) and of percentage of positive tumor cells with the following classes: 0%, 1–10%, 11–50%, 51–80%, > 80%. This is the usual and widespread base to describe immunohistochemical staining in routine histology with therapeutic relevance. These classes were proposed on the Second Consensus Meeting of the 17th Conference of the German Society of Pathology in Mainz in 1986 to describe the expression of estrogen and progesterone receptor in breast carcinoma because this classification led to the best results in interobserver reproducibility.13 Immunostaining of stromal fibroblasts was scored by examination of staining intensity only (negative, weak, strong), because the percentage of stained stromal fibroblasts did not vary considerably.

Bone marrow preparation

The procedure for bone marrow preparation has been described in detail previously.9 Briefly, during primary surgery, bone marrow samples were obtained by needle aspiration (mean volume, 5 ml/sample) and stored in heparinized tubes. After centrifugation using a Ficoll-Hypaque density gradient (density, 1.077 g/mol; Pharmacia, Freiburg, Germany) at 540g (30 min), mononucleated interphase cells (MNCs) were washed and at least 7 × 105 cells were centrifuged onto glass slides at 150g (5 min). The cytospin preparations were overnight dried and either stained immediately (n = 90) or stored at −80°C (n = 3) for a maximum time of 5 weeks.

Immunocytochemical detection of tumor cells in bone marrow

Ninety-three patients were screened for the presence of CK-positive cells in bone marrow. In total, at least 2 × 106 MNCs of each bone marrow specimen were analyzed.9 mAb A45-B/B3 (IgG1; Micromet, Munich, Germany), directed against a common epitope of CK polypeptides, including the CK heterodimers 8-18 and 8-19,9 was used at 2 μg/ml to detect tumor cells in the cytospin preparation. A negative staining control was accomplished by using an unrelated mouse myeloma IgG1 antibody (2 μg/ml; Sigma) on the patients' bone marrow specimens. The breast carcinoma cell line BT-20 (ATCC HTB 19) served as a positive control for CK immunostaining in each staining batch. The specific reaction of the primary antibody was developed with the APAAP technique (Dako), combined with the new fuchsin stain, to indicate antibody binding, as previously described in detail.9 Cytospins were analyzed with the automated cellular imaging system (ACIS; ChromaVision Medical Systems, San Juan Capistrano, CA).

Immunocytochemical staining for uPAR of tumor cells in bone marrow

Bone marrow from a subset of 37 patients was selected for phenotyping of CK-positive cells in bone marrow. For each patient, 2 to 12 × 106 MNCs (median, 5.5 × 106) were analyzed. Slides were fixed according to the manufacturer's instructions with Solution B of the Epimet Kit (Micromet) and then washed in PBS. All antibodies were diluted in antibody diluent with background reducing elements (Dako). After blocking with 10% AB serum (Biotest) in PBS for 20 min, the slides were incubated with mAb R2 at a concentration of 5 μg/ml for 45 min. MOPC-21 (Sigma) served as the IgG1 isotype control (5 μg/ml). Slides were then incubated with EnVision (Dako), which was applied undiluted for 30 min, followed by incubation with the APAAP complex (Dako) at a dilution of 1:100 for 30 min. The substrate was again a solution of fast blue BB salt and naphtol AS-MX phosphate. After blocking with 5% mouse serum (Dako) in PBS for 10 min, the tumor cells were detected with the mAb A45-B/B3 directly labeled with the fluorochrome Cy3 (Micromet), which was incubated for 45 min at a concentration of 2 μg/ml. The specificity of the antibody reaction was confirmed with a monoclonal isotype control antibody directed against fluorescein isothiocyanate (FITC) and directly labeled with Cy3 (2 μg/ml; Micromet). Finally, the slides were incubated for 3 min with DAPI (Sigma). After washing with PBS, the slides were mounted with mounting medium (Sigma) and covered with coverslips. The colon cancer cell line HT-29 (ATCC HTB 38) and the breast cancer cell line MCF7 (ATCC HTB 22) served as positive controls. The MNCs were screened by fluorescence microscopy for the presence of CK-positive cells and the expression of uPAR was then determined in the bright field. The cytospins were evaluated by 2 independent observers (A.H., K.P.). The degree of concordance was 89% and a consensus decision was made for the remaining 11%. The number of CK-positive cells (both uPAR-positive and uPAR-negative cells) in bone marrow samples of each patient ranged from 1 to 15 (median, 3) cells.

Statistical analysis

The potential association between the expression of uPAR on tumor cells and fibroblasts and the detection of occult metastatic cells in bone marrow and other tumor characteristics was statistically evaluated with the generalized exact Fisher test and the linear association test using the SPSS software package 10.0. Statistical 2-sided p-values below 0.05 were considered significant. Due to multiple testing, the associated p-values are to be interpreted exploratorily.


Expression of uPAR on primary tumor cells

We have stained tissue sections of 93 primary breast carcinomas using mAb R2 directed against domain 3 of uPAR. The observed staining patterns were complex with regard to the type of cells stained, the staining intensity and the fraction of positive cells per specimen. In general, tumor cells showed a stronger staining than the surrounding stromal fibroblasts. However, the fraction of uPAR-positive tumor cells varied considerably. For further analyses, we subdivided the specimens into 5 categories (Table I). In 9 specimens (10%), uPAR staining on tumor cells was not detected, whereas uPAR was expressed in variable degrees in the remaining 84 tumor specimens; the fraction of uPAR-positive tumor cells ranged from 1% to more than 80% (Table I, Fig. 1).

Figure 1.

uPAR expression on sections of a primary tumor, a lymph node metastasis and micrometastatic tumor cells in bone marrow from individual breast cancer patients. Primary tumor: hematoxylin-eosin stained section of a lobular breast carcinoma (a) and strong uPAR staining intensity of the tumor cells (b), phase contrast image of negative staining with the IgG1 isotype control antibody (c), 200× magnification. Lymph node metastasis: hematoxylin-eosin stained section (d) and moderate uPAR staining of metastatic tumor cells (e), phase contrast image of negative staining with the IgG1-isotype control antibody (f), 200× magnification. Bone marrow micrometastases: uPAR staining of a 3 tumor cell cluster in bone marrow preparation (g), CK and DAPI staining of the same 3-cell cluster (h), 400× magnification. Fast blue stain visualizes uPAR antibody binding (b, e, g), and the fluorochromes Cy3 (orange) and DAPI (blue) visualize CK antibody binding and nuclear DNA, respectively (h).

There was no significant correlation between the fraction of uPAR-positive tumor cells and the histologic type, the size (pT stage) or the grade (i.e., degree of differentiation) of the primary tumor, the metastatic lymph node involvement (pN stage) and the progesterone receptor status (Table I). In contrast, we observed a significant positive correlation between the fraction of uPAR-positive cells and the expression of the estrogen receptor (p = 0.006, generalized exact Fisher test; p = 0.047, linear association test). Forty-eight (67%) of the 72 estrogen receptor-positive tumors coexpressed uPAR on more than 50% of their tumor cells, whereas such a high percentage of uPAR expression was only found in 7 (33%) of the 21 estrogen receptor-negative tumors (Table I).

The additional analysis of staining intensity of tumor cells (i.e., division into 4 subjective categories: negative, weak, moderate, strong) did not add any further information. In total, 53 (57%) specimens were moderately or strongly stained, whereas 40 (43%) samples showed only weak or no staining of tumor cells.

Expression of uPAR on tumor-surrounding stromal cells

We evaluated the uPAR staining pattern of the stromal fibroblasts and performed a separate analysis of either the intratumoral fibroblasts or the peritumoral fibroblasts. In 23 (25%) of 93 tumors, intratumoral fibroblasts showed no uPAR staining, whereas positive staining of variable intensity was found in the remaining 75% of the specimens (Table II). There was no significant correlation with the histologic type and the size (pT stage) of the primary tumor, the metastatic lymph node involvement (pN stage) and the progesterone receptor status. Nevertheless, we observed a significant negative correlation with the primary tumor grade (p = 0.011, generalized exact Fisher test), although the linear association to uPAR expression within the subgroups of tumors with different grades was only of borderline significance (p = 0.092, linear association test). Another significant correlation was found between the staining intensity of the intratumoral fibroblasts and the estrogen receptor status (p = 0.029, generalized exact Fisher test; p = 0.027, linear association test; data not shown).

Table II. Correlation Between the Bone Marrow Status and the Expression of uPAR on Primary Tumors
CharacteristicAll patientsPatients with CK-positive cells in bone marrow
  • a

    Values in parentheses are the percentages of all patients with the characteristic.

  • b

    Statistically significant (p = 0.037, linear association test).

  • c

    No peritumoral stroma in the pathological specimen of one patient.

Fraction of uPAR-positive primary  tumor cells in %   
 > 80188(44)b
Staining intensity of uPAR-positive  intratumoral fibroblasts   
Staining intensity of uPAR-positive  peritumoral fibroblastsc   

The additional analysis of uPAR staining on the peritumoral fibroblasts revealed positive staining in 34 (37%) of the samples, but this type of staining showed no significant correlation with the other tumor characteristics analyzed in this study (data not shown).

Correlation between uPAR expression in primary tumor and detection of occult metastatic cancer cells in bone marrow

To investigate whether uPAR expression on primary tumor cells or the surrounding stromal fibroblasts was associated with hematogenous tumor cell dissemination, we obtained bone marrow aspirates from 93 patients and analyzed these samples for the presence of CK-positive tumor cells. Despite the fact that all of the 93 patients analyzed in the present study were free from any clinical signs of overt distant metastasis (TNM stage M0), 29 (31%) patients had CK-positive cells in their bone marrow (Table II). No significant correlations were found between the presence of CK-positive cells in bone marrow and the conventional histopathologic characteristics of the primary tumor (i.e., histologic type, tumor size, tumor grade, lymph node status, hormone receptor status; data not shown). Braun et al.9 reported that the detection of CK-positive cells in bone marrow was unrelated to the presence or absence of lymph node metastasis (p = 0.13). However, they observed a significant correlation between the number of lymph nodes with metastases and the presence of CK-positive cells in bone marrow (< 0.001) as well as tumor size and grade.9 We observed similar trends in our study (especially if we pooled different tumor stages or grades), but the total number of cases might be too small to obtain significant correlations. Thus, our present findings are not necessarily in contrast to the findings of the large study on more than 500 patients by Braun et al.9

In contrast, there was a linear association between the incidence of CK-positive cells in bone marrow and the fraction of uPAR-positive primary tumor cells (p = 0.037, linear association test; Table II). The rate of a CK-positive bone marrow finding increased from 11% in patients with uPAR-negative tumors to 44% in patients with tumors that had more than 80% uPAR-positive tumor cells. There was no such correlation with regard to the uPAR expression on fibroblasts (Table II).

uPAR expression on occult metastatic cells in bone marrow

Coexpression of uPAR on CK-positive cells in bone marrow was further investigated in a subgroup of 37 patients using an immunocytochemical double-labeling method (Fig. 1). This subgroup was representative of the entire group of 93 patients in terms of their distribution to the different clinicopathologic and biologic factors analyzed in this study. Individual CK-positive cells were discovered in 13 (35%) of 37 patients. In 10 of 13 bone marrow-positive patients, CK-positive cells coexpressed uPAR. uPAR expression was not restricted to tumor cells, but also observed on CK-negative bone marrow cells. Comparison between uPAR expression on primary and disseminated tumor cells revealed a considerable degree of heterogeneity of both types of cells.

uPAR expression on lymph node metastases

Subsequent staining of lymph nodes from 23 patients with histopathologic lymph node involvement (stages pN1/pN2) revealed uPAR expression on metastatic tumor cells in 22 cases. In 11 cases, more than 50% of the tumor cells were stained (Table III, Fig. 1). Comparing the fraction of uPAR-positive tumor cells in the primary lesion with that in the autologous metastatic cells showed different scores in 19 patients. In 11 patients, the fraction of uPAR-positive cells was higher in the primary tumor, whereas the opposite was observed in 8 patients (Table III).

Table III. uPAR Expression on Tumor Cells in Primary Tumors and Lymph Node Metastases
Patient numberHistologyPrimary tumorLymph node metastases
Fraction of uPAR-positive tumor cells in %uPAR staining intensityFraction of uPAR-positive tumor cells in %uPAR staining intensity
7Ductal11–50Weak> 80Moderate
12Ductal51–80Moderate> 80Weak
13Ductal-lobular51–80Weak> 80Moderate
21Ductal> 80Weak11–50Weak
22Ductal> 80Moderate51–80Moderate
23Lobular> 80Weak11–50Strong

In 8 of the 23 lymph node-positive patients, we detected CK-positive cells in bone marrow (data not shown). Five of these 8 patients were in the group of 11 patients with more than 50% uPAR-positive tumor cells in their nodes. uPAR coexpression on CK-positive cells in bone marrow was observed in 2 of the 8 patients. Both patients had lymph nodes with a high fraction (51–80%) of uPAR-positive tumor cells.


Our objective was to investigate the potential involvement of uPAR in the metastatic spread of primary human breast cancer. Our results show that uPAR was frequently expressed on the primary tumor cells and this expression was associated with blood-borne spread of cancer cells. In contrast, uPAR expression of the stromal fibroblasts present in the primary tumor was not related to metastatic spread. uPAR was also frequently expressed by metastatic tumor cells in bone marrow and lymph nodes. Direct comparison to the expression pattern in the autologous primary tumor cells indicated a considerable modulation of uPAR expression on metastatic tumor cells. We detected uPAR expression immunohistochemically, using mAb R2, which exhibits high affinity and specificity for domain 3 of human uPAR and labels both uPA binding and unoccupied uPAR.12, 14 This seems to be important in the context of the findings by Del Vecchio et al.,15 who showed that 70% of the uPAR molecules in breast carcinomas are occupied with uPA. Immunohistochemical evaluation of breast cancer tissue sections makes it possible to distinguish between the localization of uPAR on tumor cells or stromal fibroblasts, which may provide further information about invasive and metastatic behavior of the tumor. This differentiation might be especially important for uPAR expression since uPAR localizes the active components of the urokinase-type plasminogen activator system at the cell surface.

Our present evaluation showed that 90% of the primary breast cancer specimens harbored tumor cells that expressed uPAR. This observation is in accordance with those reports, demonstrating that uPAR-positive tumor cells are present in the majority of the primary breast cancer specimens.14, 16–19 The group of Dublin et al.,19 who also applied monoclonal antibody R2, observed uPAR-positive staining of tumor cells in 97% of the breast cancer specimens. However, in other published studies, uPAR expression on tumor cells was observed in less than 50% of the cases, which in part might be due to the different antibodies and secondary antibody systems used in these studies.20, 21, 22, 23, 24 Pyke et al.20 found uPAR-positive tumor cells in only 13% of their paraffin-embedded specimens despite using the R2 antibody. In breast cancer specimens, uPAR may be present in a cleaved form, it may be complexed with uPA, uPA + PAI-1 or PAI-2, or it may be bound to vitronectin.1, 5 Carriero et al.14 observed that acid pretreatment of the sections increased uPAR staining of breast cancer cells by dissociation of uPAR bound uPA. Luther et al.17 suggest that tumor cells may express uPAR with a varying glycosylation pattern. The different methods of pretreatment procedures may also have an influence on the structure of uPAR. Pathologists usually prefer to analyze formalin-fixed tumor sections because the morphology of the tissue is better preserved than on fresh-frozen tissue sections. Moreover, fresh-frozen sections are only available in specialized centers, whereas formalin-fixed sections are available in each hospital. Thus, working in the field of translational research, results obtained on formalin-fixed tissue can be translated into clinical practice.

The most important finding was the significant linear correlation between the fraction of uPAR-positive tumor cells and the presence of micrometastatic cells in bone marrow. The incidence of CK-positive cells in bone marrow increased with a higher number of uPAR-positive primary tumor cells. uPAR expression in the primary tumor was not associated with the lymph node status, which is consistent with the findings of several other groups.14, 15, 19, 22, 24 These observations differ from experimental studies in rats, which developed metastatic lesions in liver, spleen and lymph nodes due to an increased uPAR expression of breast cancer cells.25 uPAR facilitates cell motility by regulating cell adhesion and detachment during migration and invasion of tumor cells.1, 5 These processes play an important role in the early events of the metastatic cascade and may therefore explain how uPAR expression on primary tumor cells may facilitate their release into the blood stream. The plasminogen activation system promotes neovascularization, which may specifically increase the hematogenous metastatic spread. An earlier study demonstrated that tumor angiogenesis seemed to be associated with the detection of micrometastatic tumor cells in bone marrow without affecting lymphatic spread, which might explain the observed difference between the influence of uPAR on hematogenous versus lymphatic spread.26

We did not find a correlation between uPAR expression on stromal fibroblasts present in the primary tumor specimens and the detection of micrometastatic tumor cells in bone marrow. This finding is consistent with the observation by Costantini et al.,16 who found a significant association between the invasiveness of primary breast carcinomas and uPAR expression on the primary tumor cells, but not on the stromal cells. Our findings do not support the assumption that the fibroblastic expression of uPAR has a stronger impact on metastasis than the uPAR expression of the tumor cells.19, 24

The investigation of uPAR expression on disseminated tumor cells in bone marrow and lymph node metastases revealed that uPAR was expressed in the majority of the metastatic cells. Our finding is consistent with the recent report of Fisher et al.,27 who investigated uPAR mRNA expression of 25 osseous metastases by in situ hybridization. They observed moderate to high positive signals in 18 cases and low positive signals in 5 cases and uPAR expression was mainly localized to the tumor cells. However, in only 3 patients were the autologous primary tumors investigated in parallel with the matching osseous metastases. To our best knowledge, our present report is the first comprehensive comparative analysis of uPAR expression on autologous tumor cells from various tissue sources of the same patient. We observed that the expression pattern of uPAR on metastatic tumor cells was quite heterogeneous. For bone marrow, this result confirms the preliminary findings of Tögel et al.28 However, their double-labeling technique (immunogold staining combined with neufuchsin stain), originally introduced for uPAR in the report of Heiss et al.,29 is difficult to evaluate and therefore less reproducible than our present technique. Thus, staining protocols based on fluorescence-conjugated antibodies have become the better choice.30

Comparison between uPAR expression on primary and metastatic tumor cells revealed a considerable modulation of uPAR expression in individual patients with evidence for both up- and downregulation of uPAR during the dissemination process or at the metastatic site. uPAR enhances the activation of pro-uPA, plasminogen and MMPs and thereby promotes the release and activation of growth factors from the surrounding extracellular matrix, which may explain the high incidence of uPAR expression on metastatic tumor cells in lymph nodes and bone marrow.1, 4, 6 uPA activates hepatocyte growth factor/scattor factor (HGF/SF), and plasmin activates latent transforming growth factor-β (TGF-β) and releases active basic fibroblast growth factor (bFGF) from its ECM-binding sites.1 MMPs are involved in tumor cell growth by cleaving insulin-like growth factor-binding protein (IGF-BP), thereby releasing IGF and removing transmembrane precursors of growth factors, including transforming growth factor-α (TGF-α). MMPs play an important role in angiogenesis by providing the proangiogenic growth factors vascular endothelial growth factor (VEGF), fibroblast growth factor-2 (FGF-2) and TGF-β.4 Intracellular signal transduction cascades, which are activated through uPAR, lead to proliferation.1, 2, 5, 6 Tumor cells enter a state of dormancy with diminished proliferation due to reduced uPAR expression.31 Thus, the lack of uPAR expression might be a marker of the dormant state of occult metastatic cells. Clinical follow-up analyses are required to prove this hypothesis. In gastric cancer, patients with CK-positive cells coexpressing uPAR in bone marrow had indeed an increased risk of metastatic relapse,29 supporting the assumption that uPAR expression might help occult metastatic cells to escape from dormancy. As a result, uPAR expression might be a promising therapeutic target for future clinical trials with agents directed to uPAR.2 Agents that specifically target uPAR have become available now, and they might become promising tools for a dormancy-inducing therapy in breast cancer. Because this therapy is directed against metastatic cells, it will be more informative to type these cell directly for uPAR expression rather than stratify patients based on immunostaining of the primary tumor.


The authors gratefully acknowledge the excellent technical assistance of Antje Andreas, Katrin Baack, Cornelia Coith and Sonja Santjer and the provision of monoclonal antibody A45-B/B3 and the Epimet-Kit from Micromet (Munich, Germany). They thank Dr. U. Schüppler and Dr. S. Ostertag (Department of Gynecology and Obstetrics, University Hospital Kiel) for providing bone marrow samples and patient data, Carina Bergner for helping to prepare the article and Dr. S. Riethdorf for many helpful discussions.