Expression of plasminogen activator inhibitor-1, urokinase receptor and laminin γ-2 chain is an early coordinated event in incipient oral squamous cell carcinoma

Authors


Abstract

Cancer cell invasion is facilitated by extracellular matrix degrading proteases such as plasmin. We have studied the expression of plasminogen activator inhibitor-1 (PAI-1) and urokinase receptor (uPAR) together with the γ2-chain of laminin-5 (lam-γ2) by immunohistochemistry in 20 cases with incipient oral squamous cell carcinoma (SCC). PAI-1-positive neoplastic cells located at the tip of the putative invasive front of grade 1 (incipient) carcinoma were seen in 16 of the 20 cases (75%), whereas adjacent normal and dysplastic epithelium was PAI-1-negative. Clusters of putative invasive neoplastic cells located in the lamina propria were PAI-1-positive in areas with grade 2 incipient carcinoma as were invasive cancer cells in areas of grade 3–4 invasive carcinoma. uPAR immunoreactivity was strongly expressed in numerous stromal cells in the carcinoma area in all 20 lesions, while a few uPAR-positive stromal cells were found in areas with normal and dysplastic epithelium. uPAR-positive neoplastic cell islands located at the front of the lesions were seen in 15 of the 20 cases. The expression pattern of lam-γ2 was very similar to that of PAI-1; however, lam-γ2-positive neoplastic cells were only detected in 11 of the 20 cases (55%) in areas of grade 1 incipient carcinoma. Direct comparison of the 3 components revealed colocalization in neoplastic cell islands in both incipient and invasive SCC. Our results suggest that PAI-1 is a novel potential marker of initial invasion in oral SCC, and that the coordinated expression of PAI-1 with uPAR and lam-γ2 sustain the features of the early invasive cancer cells. © 2006 Wiley-Liss, Inc.

Invasive squamous cell carcinomas (SCC) develop from noninvasive dysplastic precursor lesions or carcinoma in situ. The transition from noninvasive to invasive carcinoma is linked to the proteolytic destruction of the basement membrane, allowing the carcinoma cells to invade the adjacent connective tissue.1, 2, 3 Plasmin(ogen) is believed to play an important role in this process because of its capacity to degrade extracellular matrix proteins, such as fibrin and basement membrane laminin,1, 4, 5 and to activate other matrix degrading proenzymes, such as metalloproteases.6, 7 Plasminogen is an abundant enzyme in the peripheral blood and other body fluids that is converted to active plasmin by either of the 2 well-known plasminogen activators: urokinase-type plasminogen activator (uPA) and tissue-type plasminogen activator (tPA). uPA-directed plasminogen activation mainly occurs on the cell surface after the binding of uPA to its specific cell-surface receptor, uPAR. This leads to focal cell-surface associated generation of plasmin, which mediates proteolysis at the cell surface required for cell migration and invasion.8 Two plasminogen activator inhibitors, PAI-1 and PAI-2, have been characterized of which PAI-1 is the primary physiologic PAI and regulates plasminogen activation in the extracellular matrix.9

The expression of uPA, PAI-1 and uPAR is upregulated in several types of cancer, including oral and pulmonary SCC and breast and colorectal adenocarcinomas.10, 11, 12, 13 The increased levels of one or more of these 3 components measured in tissue extracts or blood are associated with poor prognosis. For example, uPAR is a marker of poor prognosis in pulmonary carcinomas,14 and uPAR, uPA and PAI-1 are all indicators of poor prognosis in breast cancer.15, 16, 17 To understand the role of PAI-1 in plasminogen activation and its link to poor prognosis, nonproteolytic functions of PAI-1 should be considered. For example, the specific binding of PAI-1 to vitronectin suggests that PAI-1 also contributes to cell adhesion and migration.9, 18 Cancer cell invasion involves, in addition to the epithelial cancer cells, a variety of stromal cells (vascular cells, fibroblasts and inflammatory cells), which may contribute to the expression of the different components of the plasminogen activation system. Histological expression studies of SCC of the skin and esophagus have shown that uPA, PAI-1 and/or uPAR mRNA is predominantly expressed by the cancer cells.19, 20, 21 Immunohistochemical analyses of esophagus and head and neck SCC show a similar expression pattern,21, 22, 23 whereas the expression of the 3 components in adenocarcinomas of breast, colon and prostate is predominantly seen in the stromal compartment.24, 25, 26, 27, 28, 29, 30 Thus, the origin of the neoplasm strongly affects cell populations that contribute to the increased plasminogen activation.10, 31 Most, if not all, of the aforementioned expression studies have been performed on manifest invasive cancers, whereas little attention has been given to early carcinomas, including carcinoma in situ lesions and lesions suspected to be invasive (incipient carcinoma).

Lacking specific diagnostic markers to clearly document early invasive growth, the criteria of incipient carcinoma relies solely on the morphologic patterns of epithelial changes. To discriminate carcinoma in situ from incipient oral SCC, the integrity of the basement membrane has been considered as a valuable tool.32 However, since focal absence of basement membrane appears in (noninvasive) dysplastic lesions, laminin or collagen IV immunohistochemical staining cannot be used as reliable markers.32 For SCC of the cervix and the lower anogenital tract, the expression of the γ-2 chain of laminin-5 has been reported to be an indicator of initial invasion.33, 34 The aim of the current study is to describe the expression of PAI-1 and uPAR by immunohistochemistry in incipient oral SCC and to evaluate the significance of PAI-1 and uPAR as indicators of early invasion in oral SCC in comparison with the expression of laminin γ-2 chain.

Material and methods

Tissue samples

All biopsies signed out as squamous cell carcinoma (SCC) consecutively appearing from 1999 to 2002 at the Department of Oral Pathology, Malmö University, were reviewed to select samples containing carcinoma lesions suspected to be invasive (incipient carcinoma). A total of 20 formalin-fixed and paraffin-embedded samples were selected from various locations, including the tongue (n = 8), lip (n = 2), bucca (n = 2), vestibulum (n = 3) and floor of the mouth (n = 5). Histologic grading of the incipient oral SCC, or “questionable invasive SCC” according to Anneroth et al.,35 was done by focusing on the stage of invasion, which expresses the infiltrative characteristics of the front (depth) of the lesion.35 The stage of invasion in the selected samples corresponds at least to grade 1–2 according to Anneroth et al.35 Grades 1 and 2 represent changes beyond carcinoma in situ, including the so-called “pushing” well-delineated infiltrative borders (Fig. 1a) and the presence of solid neoplastic cords and strands in the lamina propria (Fig. 1b). According to the TNM classification, the carcinomas analyzed are T0, N0 and M0 or T1, N0 and M0. The specimens contained adjacent regions of normal epithelium (n = 17), dysplasia (n = 19) and invasive SCC (n = 12) corresponding to grade 3 (n = 12) and grade 4 (n = 2) in the classification (Fig. 1c). The study was performed in accordance with the World Medical Association Declaration of Helsinki, 1996.

Figure 1.

Haematoxylin and eosin-stained sections from incipient and invasive oral squamous cell carcinoma. The carcinomas were graded according to the morphologic changes described in the classification by Anneroth et al.35 (see Material and methods). Incipient carcinoma grade 1 shows pushing and well-delineated epithelial borders (a), whereas the grade 2 carcinoma shows putative invasion of solid cords, bands or strands of neoplastic cells, involving only lamina propria (b). Grade 3 carcinoma is characterized by small groups or thin infiltrating cords of cancer cells reaching muscles and salivary tissues in the submucosa (c), while Grade 4 has a marked, diffuse, widespread, cellular invasion of neoplastic cells in the submucosa (not shown here).

Antibodies

Affinity purified rabbit pAb against PAI-1 (preparation AB-2A) and a preparation of rabbit anti-PAI-1 depleted for anti-PAI-1 IgG (preparation AB-2D) were described earlier.29 Mouse monoclonal antibody (mAb) against PAI-1 (Clone No. 380, IgG1) was purchased from American Diagnostica (Greenwich, CT). Affinity purified rabbit polyclonal antibodies (pAb) and a mAb (clone R2, IgG1) against uPAR have been described previously.36, 37 Clone R2 is directed against domain 3, and thus recognizes full length GPI-anchored uPAR, soluble full-length uPAR as well as cleaved uPAR.38 The uPAR pAbs also recognize these forms of uPAR. A mAb against laminin-5 (clone 4G1, IgG1) directed against laminin γ2-chain precursor39 recognizes the cytoplasmic γ2-chain without reacting with the basement membrane-associated laminin-5 (Henrik Winther, DakoCytomation, personnel communication). A mAb recognizing cytokeratins 5, 6, 8 and 17 (clone MNF116, IgG1), FITC-conjugated goat anti-rabbit IgG, negative control mAb against Aspergillus Niger glucose oxidase (Mab-WF-AF-1, IgG1) and nonimmune rabbit Ig were purchased from DakoCytomation, Glostrup, Denmark. Cy3-conjugated goat anti-mouse IgG was obtained from Jackson Immunoresearch, West Grove, PA. mAb against trinitrophenyl hapten (TNP, IgG1) was described previously.28

Immunoperoxidase staining

Four micrometer paraffin sections were deparaffinized with xylene and hydrated through ethanol/water dilutions. Sections were pretreated with heat-induced antigen retrieval, using a T/T micromed microwave processor (Milestone, Sorisol, Italy) at 99°C in TEG-buffer (pH 9.0) for 10 min for pAb against PAI-1 and mAb against laminin γ2 chain (lam-γ2), or in 10 mM citric acid (pH 6.0) for 10 min for mAb against PAI-1. Pretreatment with proteinase K (DakoCytomation) for 20 min was used for pAb against uPAR and mAbs against uPAR and TNP. Endogenous peroxidase was blocked by incubation in 1% H2O2 for 15 min. The primary antibodies were diluted in TBS containing 0.25% BSA and incubated on the sections overnight in Shandon racks (Thermo Shandon, Pittsburgh, PA) at the following concentrations: uPAR clone R2 (0.2 μg/ml), uPAR pAb (0.2 μg/ml), PAI-1 pAb (0.5 μg/ml), lam-γ2 (10 μg/ml), PAI-1 clone No. 380 (5 μg/ml), and TNP (0.2 μg/ml). After 30 min at room temperature, sections were washed in 50 mM Tris 150 mM NaCl, pH 7.6, containing 0.5% Triton X-100 (TBS-T). The primary antibodies were detected with envision reagents, either anti-rabbit IgG or anti-mouse IgG horseradish peroxidase-conjugated polymers (DakoCytomation). Each incubation step was followed by washes with 6 ml of TBS-T. Sections were developed with 3,3′-diaminobenzidine chromogenic substrate (DAB, DakoCytomation) for 15 min and finally counterstained with hematoxylin.

Immunofluorescence staining

Sections were initially processed as mentioned earlier for the immunoperoxidase staining of uPAR, using proteinase K pretreatment. uPAR pAbs were incubated overnight in TBS-BSA at 3 μg/ml together with mAb against cytokeratin at 1 μg/ml. The following day, the rabbit pAbs were detected with FITC-conjugated goat anti-rabbit (1:200) and the mouse mAb with Cy3-conjugated goat anti-mouse (1:200). After brief washes with TBS, sections were mounted with Prolong Gold antifade (Molecular Probes, Leiden, The Netherlands). The double stained sections were evaluated in a conventional fluorescence microscope and images were obtained using a confocal laser-scanning microscope, LSM 510 META (Carl Zeiss, Jena, Germany) equipped with a 488 nm Argon laser and a 543 nm HeNe laser. The microscope detector setting was the lambda mode with a 135 μm pinhole diameter. Images were collected from 509 to 595 nm. For separation of the specific fluorescence signals, we first obtained FITC, Cy3 and autofluorescence emission specters from single stained sections (for FITC or Cy3 fluorescence) or unstained sections (for autofluorescence). From double labeled sections, the collected fluorescence signal was separated by emission fingerprinting, using the recorded emission specters as described.40 Nomarski differential interference contrast, which is a phase imaging technique, was used to show the tissue structures revealed by refractive index inhomogeneities.

Results

PAI-1 immunoperoxidase staining

The 20 paraffin-embedded samples were immunohistochemically stained with rabbit pAb against PAI-1. PAI-1 immunoreactivity was seen in all cases and was with a few exceptions confined to incipient (grade 1–2) or invasive SCC (grade 3–4). Virtually, all single neoplastic cells and cords of neoplastic cells in these regions were PAI-1-positive (Figs. 2a and 2b), whereas the normal epithelium was negative in all of the 17 samples containing normal mucosa (Fig. 2c). PAI-1 immunoreactivity was detected in 16 of the 20 cases in areas of grade 1 incipient carcinoma and was seen in small foci or solitary basal cells at the tip of the pushing borders (Fig. 2d), whereas areas of grade 2 carcinoma also had strongly stained PAI-1-positive neoplastic cell clusters in lamina propria in 19 of the 20 cases (Fig. 2e and Table I). In one case, we did not detect PAI-1 immunoreactivity in areas of grade 1 and 2 carcinoma, but in areas of grade 3 carcinoma only. In areas with invasive SSC identified in 12 of the cases, larger cancer cell clusters showed PAI-1-immunoreactivity only in the proliferating basal cell layer, whereas suprabasal cells were either negative for PAI-1 or less intensely stained than the basal cells (Fig. 2f). A few PAI-1-positive fibroblast-like cells were seen within the stroma of the cancer area.

Figure 2.

PAI-1 immunoperoxidase staining of oral squamous cell carcinoma. Immunohistochemical staining for PAI-1, using affinity purified rabbit pAb in a lesion (a–c), showing the gradual progression from normal mucosa (N in (a)) to grade 3 invasive carcinoma (IC in (a)). Strong PAI-1 immunoreactivity is seen in cords of invasive cancer cells (b), whereas the adjacent normal mucosa is PAI-1-negative (c). In areas of grade 1 carcinoma, PAI-1 immunoreactivity is seen in foci of neoplastic cells at the tip of the pushing border (d), whereas areas of grade 2 carcinoma show more widespread PAI-1 expression in the neoplastic cells of incipient carcinoma (e). Large clusters of cancer cells in grade 3 carcinoma show PAI-1 immunoreactivity, primarily in the basal cell layers (f, red arrows), whereas all cancer cells in invasive cords are PAI-1 positive (f, black arrows). Two adjacent sections were incubated with anti-PAI-1 pAb (g) and anti-PAI-1 pAb depleted for PAI-1 IgG (h). Strong staining is seen with the 2 anti-PAI-1 antibodies (area of grade 3 carcinoma), whereas the anti-PAI-1 depleted IgG shows no staining. Haematoxylin counterstaining.

Table I. Expression of PAI-1, uPAR and lam-γ2 in Normal and Neoplastic Epithelial Cells in Oral Lesions with Incipient Carcinoma
GradeNormal epithelial cellsIncipient carcinoma neoplastic cellsInvasive SCC neoplastic cells
IIIIII/IV
  • Sections with incipient oral carcinoma were immunohistochemically stained for PAI-1, uPAR and lam-γ2 (see Material and Methods) and the presence of immunoreactive normal or neoplastic epithelial cells in the various progression areas was evaluated (see text).

  • 1

    Number of lesions with immunoreactive cells/number of lesions.

PAI-10/171 (0%)16/20 (75%)19/20 (95%)12/12 (100%)
uPAR0/17 (0%)1/20 (5%)5/20 (25%)12/12 (100%)
lam-γ20/17 (0%)11/20 (55%)19/20 (95%)12/12 (100%)

In the transition zone from dysplastic/hyperplastic epithelium to the carcinoma area, the basal cells were PAI-1-positive in a few of the cases only. Relatively weak PAI-1 immunoreactivity was seen in some of the vessels in the submucosa (data not shown). As a positive control for the PAI-1 immunoreactivity obtained with the pAb, we directly compared this with that obtained with a mAb (clone No. 380) against PAI-1 by immunohistochemically staining adjacent sections with the two antibodies. An identical staining pattern was obtained with these two anti PAI-1 antibodies. The staining intensity obtained with the mAb was however considerably lower than that obtained with the pAb. Negative controls included anti-PAI-1 pAb depleted for anti-PAI-1 IgG and a mAb against Aspergillus Niger. No specific staining was obtained with these two antibody preparations when incubated at concentrations similar to the respective anti-PAI-1 antibody preparations (Fig. 2h).

uPAR immunoperoxidase staining

uPAR immunoreactivity obtained with clone R2 mAb, was seen in all 20 lesions, and uPAR was strongly expressed in areas with incipient and invasive SCC compared to areas with dysplastic and normal epithelium (Figs. 3b and 3c). The predominant uPAR immunoreactivity was seen in stromal cells particularly associated with incipient and invasive SCC. In these areas, we observed numerous uPAR-positive macrophage-like cells (spherical spindle-shaped cells) both in the lamina propria area and in the profound invasive front (Figs. 3c and 3d). uPAR-positive fibroblast-like cells (elongated spindle-shaped cells) were especially evident within the lamina propria area (Fig. 3e). uPAR-positive neutrophils were seen in all cases, most frequently in areas with invasive SCC. In 15 of the 20 cases, we also identified small clusters of uPAR-positive neoplastic cells. These clusters of neoplastic cells (often located among uPAR-positive stromal cells) were observed in areas with incipient (5/20) and invasive SCC (12/12), particularly in the front (Fig. 3f and Table I). There was no correlation between samples showing lack of uPAR positive cancer cells and the location of the individual lesions in the oral cavity.

Figure 3.

uPAR immunoperoxidase staining of oral squamous cell carcinoma. Sections were incubated with mAb R2 against uPAR (a–g), pAb against uPAR (h) or mAb against TNP (i). uPAR immunoreactivity is strongly increased in areas of incipient grade 1 carcinoma (c, framed area in a) compared to the normal mucosa (b, framed area in a). In the normal mucosa, a few uPAR-positive cells are seen interspersed throughout the epithelium, here considered as macrophages or Langerhans cells (b, arrow), whereas a high number of macrophage-like cells are present in areas of incipient carcinoma both within the epithelium (black arrows in c) and within the surrounding stroma (c, red arrows). In incipient grade 2 carcinoma, uPAR immunoreactivity is seen predominantly in the stroma (St) as macrophage-like and fibroblast-like cells at the putative invasive front (d), whereas neoplastic cells (Ca) are negative. uPAR immunoreactivity is seen in fibroblast-like cells located in the lamina propria area (St) with grade 3 carcinoma (Ca in (e)). At the invasive front of grade 3 carcinoma (f), uPAR immunoreactivity is seen in both stromal cells (St) and cancer cells (Ca). Note the strong staining of the cell surface of the cancer cells indicated by arrows (f). The mAb (g) and the pAb (h) against uPAR show the same staining pattern, here of the stromal inflammatory cells (St) in the interface with carcinoma grade 2 (Ca), and no specific staining is seen with the mAb against TNP (i). (g–i) are 3 adjacent sections. Haematoxylin counterstaining.

In areas with dysplastic and normal epithelium, only a few scattered uPAR-positive macrophage-like cells were identified both in the connective tissue area and within the epithelium (Fig. 3b). These areas were devoid of uPAR-positive fibroblast-like cells and only a few neutrophils were observed. As a negative control for the mAb against uPAR, we applied a mAb against TNP. No specific staining was obtained with this mAb when incubated at the same concentration (Fig. 3i). The specificity of the immune reaction obtained with the mAb against uPAR was further evaluated by direct comparison of the staining pattern obtained with rabbit pAb against uPAR. An identical staining pattern was obtained with these 2 uPAR antibodies (g,h).

uPAR double immunofluorescence analyses

In 15 of the 20 investigated cases, we observed small clusters of cancer cells that were positively stained for uPAR. To determine whether these cells were indeed malignant (epithelial) cells and not infiltrating macrophages, we immunohistochemically visualized uPAR with FITC fluorophore and cytokeratin with Cy3 fluorophore in 5 of the cases in which we had identified uPAR-positive cells assumed to be cancer cells. In all these cases, we could clearly identify uPAR and cytokeratin immunoreactivity in the same cells both as cell clusters (Fig. 4) and as single malignant cells. Many of the uPAR-positive single cells located within clusters of cytokeratin-positive cancer cells were found to be cytokeratin-negative and therefore considered to be infiltrating macrophages.

Figure 4.

uPAR and cytokeratin double immunofluorescence analyses of oral squamous cell carcinoma. Sections were incubated with rabbit pAb against uPAR together with a mouse mAb against cytokeratin. The rabbit anti-uPAR is detected with FITC-conjugated goat-anti-rabbit IgG (a) and the mouse anti-cytokeratin with Cy3-conjugated goat anti-mouse IgG (b). In the overlay (c), uPAR immunoreactivity is seen on the surface of some of the cytokeratin-positive cancer cells (white arrows) as well as in some macrophage-like cells located in the stroma (St) that are cytokeratin-negative (Ca, yellow arrows in (a) and (c)). The overlay (c) is added to Nomarski differential interference contrast to delineate the tissue texture.

Laminin γ-2 chain immunoperoxidase staining and comparison with uPAR and PAI-1

The above results suggest that PAI-1 and uPAR are upregulated in oral epithelial cells, during malignant transition. The laminin-γ2-chain (lam-γ2) and uPAR are coexpressed in budding cancer cells at the invasive front of human colon cancers,39 and because laminin 5 has been reported to be a marker for early invasive squamous cell carcinoma of the cervix and vulva,33, 34 we immunohistochemically stained the 20 oral lesions with an antibody against lam-γ2. As expected, we obtained cytoplasmic lam-γ2 immunoreactivity in neoplastic cells and no extracellular or stromal immunoreactivity in all 20 cases. Intense lam-γ2 immunoreactivity was seen in single neoplastic cells and cords of neoplastic cells located at the front of both the incipient and invasive carcinomas, whereas the normal epithelium in the 17 lesions containing normal mucosa was lam-γ2 negative (Fig. 5a). In incipient grade 1 carcinoma, lam-γ2 staining was seen in 11 of the 20 cases in small foci or solitary neoplastic cells located in the tip of the pushing borders (see Fig. 5b and Table I), whereas in areas of grade 2 carcinoma, 19 lesions contained lam-γ2 positive cells located at the tip of the pushing borders as well as in the early invasive single cancer cells or cords of cancer cells (Fig. 5c). The single lesion without lam-γ2 expression in areas of grade 1–2 carcinoma was the very same that lacked PAI-1 immunoreactivity in these areas. The clusters of cancer cells in regions of invasive carcinoma were generally lam-γ2-positive, but some (10–20%) were lam-γ2-negative. Most cancer cell clusters showed lam-γ2 immunoreactivity confined to the proliferating basal cell layer (Fig. 5d). In dysplastic epithelium and in the transition zone from dysplastic/hyperplastic epithelium to the carcinoma area, sporadic neoplastic cells located in the basal cell layer were lam-γ2-positive.

Figure 5.

Laminin γ2 chain immunoperoxidase staining of oral squamous cell carcinoma. A section with both normal mucosa (N in (a)) and areas of invasive grade 3 carcinoma (IC in (a)) immunohistochemically stained with a mAb against lam-γ2 (Laminin-5). Lam-γ2 immunoreactivity is seen in the area of grade 1 carcinoma (b, arrow), whereas no immunoreactivity is seen in the adjacent normal mucosa (N). In an area with incipient grade 2 carcinoma, the lam-γ2 immunoreactivity is confined to the tip of the pushing epithelial border (c, black arrows). A single lam-γ2-positive cancer cell is identified in the stroma (c, red arrow), suggesting initial grade 3 carcinoma. Some larger cancer cell islands in an area of grade 3 carcinoma show lam-γ2 immunoreactivity confined to the basal cell layer (d, red arrows) and strong staining of the invasive cords of cancer cells (d, black arrows). Haematoxylin counterstaining.

These observations urged a direct comparison of the expression of lam-γ2 with that of PAI-1 and uPAR. We therefore analyzed the expression of PAI-1, uPAR and lam-γ2 on adjacent sections from all cases. A direct correlation between PAI-1 and lam-γ2 immunoreactivities was observed in all cases, both in regions of incipient and invasive SCC, and was easily identified because of their distinct expression in epithelial cells. The main difference was that PAI-1 was more broadly expressed than lam-γ2 in areas with incipient and invasive SCC. We found coexpression of PAI-1, uPAR and lam-γ2 in a few foci of putative invasive cancer cells in areas with incipient grade 1 carcinoma (Figs. 6a6c). Generally, in both regions of incipient SCC and invasive SCC, the PAI-1 and lam-γ2-positive neoplastic cells were accompanied by expression of intense uPAR immunoreactivity in stromal cells (Figs. 6d6f). Coexpression of PAI-1, uPAR and lam-γ2 was seen in small clusters of cancer cells in invasive SCC (Figs. 6g6i).

Figure 6.

Colocalization of PAI-1, uPAR and laminin γ2-chain in oral squamous cell carcinoma. Three serial sections obtained from areas of grade 1 carcinoma (a–c), grade 2 carcinoma (d–f) and grade 3 carcinoma (g–i) were immunohistochemically stained for PAI-1 (a), (d), (g), lam-γ2 (b), (e), (h) and uPAR (c), (f), (i), respectively. In the area with grade 1 carcinoma, focal expression of PAI-1, lam-γ2 and uPAR is seen in the same basal layer zone at the tip of the pushing epithelial border (a–c, arrows), probably in the same cells. In the area with grade 2 carcinoma, uPAR is predominant in the stroma (indicated by St in (f)) in contrast to PAI-1 and lam-γ2, which are seen in the same neoplastic cells (arrows in d and e) at the edge of the carcinoma. In areas of grade 3 carcinoma, the same clusters of invasive cancer cells show PAI-1, lam-γ2 and uPAR immunoreactivity (g–i, arrows). Haematoxylin counterstaining.

Discussion

In this study, we have shown that the expression of 2 key regulators of plasminogen activation, PAI-1 and uPAR, is strongly upregulated in incipient and early invasive SCC of the oral cavity. The presence of normal and dysplastic epithelium as well as incipient and invasive carcinoma in most of the samples investigated allowed a semiquantitative comparison of the expression levels between these tissues, and from the analyses we concluded that the expression of PAI-1 and uPAR in epithelial cells is strongly associated with the incipient and invasive carcinoma. Our findings are based on immunohistochemistry using positive and negative control antibodies. Thus, both PAI-1 and uPAR immunoreactivity was obtained with polyclonal and monoclonal antibodies,29, 36, 37 which showed identical staining patterns and negative control antibodies that showed no specific staining. The use of paraffin embedded tissue specimens that had been fixed extensively in formalin for the current immunohistochemical studies prohibited concomitant analysis of uPA, since uPA immunoreactivity cannot be retrieved in such samples.28 However, recent immunohistochemical analyses of uPA in frozen sections from squamous cell carcinomas of the oral cavity showed the presence of uPA, primarily in cancer cells.41 The latter study also showed that the level of uPA activity is strongly increased in the malignant tissue compared to that of the adjacent normal tissue, a finding that substantiated earlier observations of increased uPA mRNA levels in oral SCC.42, 43 Taken together, these results indicate that uPA-directed plasminogen activation is a strongly active process during cancer progression in the oral cavity, the strength of which may be directly related to early recurrence for some of these patients.13

We have subdivided the epithelial changes into areas with apparently nonmalignant mucosa (normal and hyperplastic/dysplastic epithelium), areas suspected to be invasive (incipient carcinoma) and areas with unambiguous invasion, according to Anneroth et al.35 Using this classification, we found that PAI-1 was upregulated in areas corresponding to grade 1 incipient carcinoma in 16 of the 20 cases (75%) investigated as well as in all higher grades, whereas PAI-1 was virtually absent in normal, hyperplastic and dysplatic epithelium. Similarly, lam-γ2 was detected in 11 of the 20 cases (55%) in areas with grade 1 carcinoma and all higher grades. The expression pattern of uPAR was found to be more complex by being expressed in various cell types and not confined to the carcinoma areas. These findings suggest that PAI-1 expression in incipient oral carcinoma is the most effective indicator of initial invasion in oral SCC compared to uPAR and lam-γ2. A larger study of patients with suspected invasive oral SCC analyzed in a follow-up study would be required to determine whether the immunohistochemical detection of PAI-1 can be used as a clinical parameter to discriminate noninvasive from early invasive oral SCC.

We found strong PAI-1 immunoreactivity in the neoplastic cells of incipient carcinoma and invasive carcinoma, but virtually no PAI-1 in the adjacent stromal cells, which is in contrast to the expression pattern found in breast, colon and prostate cancer where PAI-1 is expressed mainly in myofibroblasts.26, 29, 30 Immunohistochemical staining of PAI-1 in oral squamous cell carcinomas has also been reported by Nozaki et al.23 and Yasuda et al,22 who also observed PAI-1 immunoreactivity in cancer cells, but only in ˜30 and 80% of their cases, respectively. Yasuda et al.22 analyzed cryoembedded samples of mainly grade 3 and 4 oral SCC for PAI-1 and, in agreement with our findings, found PAI-1 immunoreactivity at the surface of cancer cell islands (the basal cancer cells). However, these authors also found PAI-1 immunoreactivity in the extracellular matrix in close proximity to the cancer cell islands that interestingly colocalized with vitronectin immunoreactivity, a major binding partner for PAI-1. In our study, PAI-1 immunoreactivity was only observed in cytoplasm of neoplastic cells and a few stromal fibroblast-like cells, while we found only discrete extracellular PAI-1 positive debris in parts of the stroma close to the incipient and invasive carcinoma. This subtle disagreement is likely to be explained by the fact that Yasuda et al.22 analyzed frozen sections. Alternatively, this difference may be explained by the specificity of the pAb used. The pAb employed by Yasuda et al.22 were raised against PAI-1 purified from bovine endothelial cells and affinity purified on a similar PAI-1 preparation, allowing the possibility of co-purifying crossreacting antibodies. In contrast, we used pAb obtained by immunization with PAI-1 purified from HT1080 cells and subsequently affinity purified against recombinant human PAI-1 expressed yeast.29

The expression pattern of PAI-1 was found to be considerably different from that of uPAR. The highest level of uPAR immunoreactivity was observed in all cases in stromal cells associated with incipient and invasive SCC. In addition to the intensely stained stromal cells, we observed clusters of uPAR-positive neoplastic cells in 15 of the 20 cases generally found as invasive cancer cells located in the most profound areas of the invasive lesions. Thus, uPAR was detected in subpopulations of cancer cells and was not expressed by all cancer cells in any of the cases. These findings disagree with those reported by Nozaki et al.,23 who observed uPAR immunoreactivity only in cancer cells. Notably these authors employed the avidin-biotin detection system and detected uPAR only in approximately one third of their cases, whereas we used the more sensitive envision detection system44 and detected uPAR in all cases.

Our observations indicate a complex expression pattern for uPAR in both incipient and definite invasive oral SCC as it is expressed in various cell populations, including macrophages, neutrophils, (myo)fibroblasts and cancer cells, the predominant expression being observed in cells located in the stromal compartment. Predominant expression of uPAR immunoreactivity in stromal cells has been described in various adenocarcinomas, including colon, gastric and hepatocellular carcinoma,24, 45, 46 whereas uPAR immunoreactivity in SCC of the esophagus21 and uPAR mRNA in SCC of the skin20 was predominantly observed in the cancer cells.

Despite the dissimilarity in the expression patterns of uPAR and PAI-1, we observed a noteworthy correlation of their expression in neoplastic cells in areas of incipient as well as invasive SCC. Interestingly, this coexpression also coincided with the expression of the γ2 chain of laminin-5. Laminin-5 is an extracellular protein that links the basement membrane via integrins to hemidesmosomes. However, this interaction is sensitive to proteolytic cleavage mediated by MT1-MMP, which releases domain III of the lam-γ2 chain and subsequently leads to cell motility.47 Interestingly, the keratinocytes upregulate lam-γ2, uPAR and PAI-1 during healing of skin wounds,48, 49, 50 indicating that these molecules are involved in the migratory and invasive phenotype of the leading-edge keratinocytes. Therefore, the concomitant expression of PAI-1 and frequently also of uPAR and lam-γ2 in the same focal microenvironment and even in the same oral neoplastic cells strongly suggests a sustained invasive capacity of these cells. Coexpression of uPAR and lam-γ2 has been reported in invasive cancer cells in human colon adenocarcinoma,39 but apparently these cancer cells do not express PAI-1.26 This divergence may either be explained by the different types of epithelia in the two locations (simple epithelium in the intestine vs. stratified squamous epithelium in the oral cavity) or the unique character of the local mesenchymal environment.

The high expression of lam-γ2 in neoplastic cells of incipient and invasive carcinoma along with its general absence in normal oral mucosa, suggests that lam-γ2, like PAI-1, is a potential marker of initial invasion in oral squamous cell carcinoma. Indeed, lam-γ2 is an established marker of invasion in squamous cell carcinomas of the cervix and the vulva.33, 34 The transition from noninvasive to invasive carcinoma is likely to be mediated by the activity of one or several extracellular matrix-degrading proteases expressed in the local microenvironment,2 such as it has been described in carcinoma in situ lesions of the human breast, where MMP-13 is expressed in myofibroblasts in connection with early invasion.51 In conclusion, our results together with those of Curino et al.41 indicate that the early steps in invasive growth of oral squamous cell carcinomas are associated with the expression of the plasminogen activation cascade. Our findings suggest that PAI-1 is a marker of malignancy in incipient carcinomas of the oral cavity, and that the coordinated expression of PAI-1, uPAR and lam-γ2 could be a molecular profile acquired for early invasion in oral SCC.

Acknowledgements

We thank U. Samuelsson for technical assistance, J. Post for photographic assistance, and O. D. Laerum, J. Rømer and K. Green for critically reading the manuscript. This work was supported by a grant from the European Commission QLK3-CT-2002-02136 (BSN) and the Meyer Foundation.

Ancillary