Overexpression of the Cytokine Osteopontin Identifies Aggressive Laryngeal Squamous Cell Carcinomas and Enhances Carcinoma Cell Proliferation and Invasiveness

Head and neck squamous cell carcinoma (HNSCC) is the sixth most frequent cancer. Laryngeal squamous cell carcinoma (LSCC) is the most common HNSCC (1–3). Notwithstanding primary prevention, screening, surgical treatment, and radiotherapy, the long-term survival rate of LSCC patients has remained substantially unchanged in the last two decades (4, 5). Stage and histologic grade are prognostic factors in LSCC but do not always distinguish between high-risk and low-risk patients (1–5). Various biological prognostic markers have been identified for LSCC and other HNSCC types: mutation in the p53 tumor suppressor gene (6), amplification of cyclin D1 (7, 8), overexpression of the epidermal growth factor receptor (9) and vascular endothelial growth factor (10), and reduced expression of the CIP/KIP cell cycle inhibitory proteins (11). However, little is known about the molecular mechanisms that govern the establishment and maintenance of the LSCC neoplastic phenotype.

Osteopontin, also known as secreted phosphoprotein 1, is a highly acidic calcium-binding glycosylated phosphoprotein. It is a cytokine (early T lymphocyte antigen-1 or interleukin-28) that regulates T helper cell-1 function (12, 13). In addition, osteopontin binds to the cell surface receptors α_v- or β₁-containing integrins and CD44 (14, 15), thereby supporting chemotaxis, attachment, and migration of many epithelial cell types (16, 17). CD44 is expressed as a standard receptor (CD44s) and in multiple splice isoforms (CD44v), whose expression is altered during tumor growth and progression. Expression of the v6 variant exon of CD44 is necessary for osteopontin binding (18). CD44 splice variants are thought to be correlated with invasive growth and metastasis in many tumor types (18). Osteopontin is overexpressed in many human tumors (e.g., colon, breast, liver, prostate, gastric, ovarian, and lung carcinomas; refs. 19–22).

We have explored osteopontin expression in LSCC. Our results indicate that osteopontin is a promising molecular marker for LSCC risk assessment. At cellular level, osteopontin binding to CD44v6 promoted SCC cell growth and invasion.

Tumors. Archival tumor samples from 58 patients (57 males and 1 female) with laryngeal cancer were retrieved from the files of the Department of Biomorphological and Functional Sciences, Pathology Section, University of Naples “Federico II,” after having obtained informed consent. Clinicopathologic data were recorded (Table 1). Patients underwent surgery at the Institute of Otolaryngology, University of Naples “Federico II,” between 1997 and 2001. The study was approved by the institutional review board committee. The patients' age ranged between 43 and 83 years, with a mean of 64.7 years. Patients underwent otolaryngologic examination, fiberoptic study, and radiological evaluation and were treated by surgery alone, except for patients with locally advanced tumors (T₄), who received postsurgical radiation therapy. The patients underwent a median of 60 months' follow-up, which consisted of clinical and radiological evaluation at 3-month intervals for the first year and at 6-month intervals thereafter. After surgical resection, tissues were fixed in 10% neutral buffered formalin and embedded in paraffin blocks. Sections (4 μm thick) were stained with H&E. Histologic grading and tumor-node-metastasis classification were done according to the recommendations of the International Union Against Cancer (23, 24). The pathologic analysis was done in a blinded manner with respect to the patients' clinical data. For five patients, a 10-μm-thick section was processed for dissection. Paraffin was removed by treatment with xylene for 3 hours at room temperature followed by tissue rehydration through multiple graded ethanol solutions and distilled water. The cancerous region was identified microscopically; normal and tumor tissues were dissected with a sterile 30-gauge hypodermic needle. The collected samples (∼120,000 cells) were placed into 1.5-mL microcentrifuge tubes and processed for RNA extraction.

Immunohistochemistry. Serial tumor sections (4 μm thick) were mounted on poly-l-lysine-coated glass slides. After antigen retrieval, the slides were incubated with anti-osteopontin (5 μg/mL; 10A16, Assay Designs, Ann Arbor, MI) or anti-CD44 (1:100; NCL-CD44v6, clone VFF-7, Novocastra Laboratories Ltd., Newcastle upon Tyne, United Kingdom) monoclonal antibodies (mAb). The sections were incubated (overnight at 4°C) with the primary antibody, biotinylated anti-IgG, and the premixed avidin-biotin complex (Vectastain ABC kits, Vector Laboratories, Burlingame, CA). The immune reaction was revealed with 0.06 mmol/L 3,3′-diaminobenzidine (DAKO, Carpinteria, CA) and 2 mmol/L H₂O₂. The slides were counterstained with hematoxylin. Anti-osteopontin antibody was preincubated with a 5-fold molar excess of osteopontin peptide to ascertain the specificity of the reaction. Control slides, stained with preimmune serum, were included as an additional negative control. The results of the immunohistochemical staining were evaluated separately by two investigators in a blinded manner. The percentage of positive tumor cells was determined by examining at least five representative microscope areas at a ×400 magnification. For osteopontin, samples were assigned to one of the four following categories: 0, absence of positive cells; +, <10% of positive cells; ++, 10% to 50% of positive cells; and +++, >50% of positive cells. Immunohistochemical staining for CD44v6 was expressed as “basal” or “full thickness” based on the epithelium layers stained.

Cell lines. Human normal epidermal keratinocytes (HNEK; neonatal) were cultured in keratinocyte growth medium according to the recommendation of the manufacturer (Cambrex, East Rutherford, NJ). HN and BHY cell lines were derived from a human oral cavity SCC (25), CAL27 and CAL33 cell lines were from human tongue SCC (26), and Hep2 cells were from a LSCC (27); KB are epidermoid cancer cells (28). Tumor cells were maintained in DMEM supplemented with 10% fetal bovine serum, 2 mmol/L l-glutamine, and 100 units/mL penicillin-streptomycin (Life Technologies, Paisley, PA). For cell treatments, recombinant mouse osteopontin protein was from R&D Systems (Minneapolis, MN). U0126 was from Calbiochem (San Diego, CA). Blocking CD44 mAbs were purified from the KM81 hybridoma cell line (TIB-241, American Type Culture Collection, Manassas, VA; ref. 29). For blocking experiments, cells were preincubated for 30 minutes with the KM81 mAb (10 μg/mL) at 37°C, 5% CO₂.

RNA extraction and reverse transcription-PCR. Tissue samples were snap frozen in liquid nitrogen and stored at −80°C before RNA extraction. Tissues were homogenized using a Mixer Mill Homogenizer (Qiagen, Crawley, West Sussex, United Kingdom). Total RNA from the indicated cell cultures and from frozen tissue samples was prepared using the RNeasy Mini kit (Qiagen) according to the manufacturer's instructions. Only tissue samples containing >70% neoplastic cells were used. Total RNA (2.5 μg) was retrotranscribed into cDNA by using the GeneAmp RNA PCR Core kit (Applied Biosystems, Foster City, CA). PCR amplification was done using 2.5 μL of the reverse transcription product in a reaction volume of 25 μL. To exclude DNA contamination, each PCR reaction was also done on untranscribed RNA. The levels of the housekeeping β-actin transcript served as a control of equal RNA loading. Primers were designed with the Primer 3 program (http://www-genome.wi.mit.edu/cgi-bin/primer/primer3_www.cgi) and synthesized by MWG Biotech (Ebersberg, Germany). Primer sequences were osteopontin forward 5′-AGGAGGAGGCAGAGCACA-3′, osteopontin reverse 5′-CTGGTATGGCACAGGTGATG-3′, CD44 (exon 2) forward 5′-GCTTTCAATAGCACCTTGCC-3′, CD44 (exon3) reverse 5′-GTTGTTTGCTGCACAGATGG-3′, CD44 (exon v6) reverse 5′-GTTGCCAAACCACTGTTCCT-3′, β-actin forward 5′-TGCGTGACATTAAGGAGAAG-3′, and β-actin reverse 5′-GCTCGTAGCTCTTCTCCA-3′. Reverse transcription-PCR (RT-PCR) products were loaded on a 2% agarose gel and stained with 0.5 μg/mL ethidium bromide, and the corresponding image was saved with the Typhoon 8600 laser scanning system (Amersham Pharmacia Biotech, Buckinghamshire, United Kingdom). The density and width of each band were quantified using the ImageQuant 5.0 (Amersham Pharmacia Biotech).

Protein studies. Immunoblotting experiments were done according to standard procedures. For tissue protein extraction, samples were snap frozen and homogenized in lysis buffer by using the Mixer Mill apparatus (Qiagen). Protein concentration was estimated with a modified Bradford assay (Bio-Rad, Hercules, CA). Antigens were revealed by an enhanced chemiluminescence detection kit (Amersham Pharmacia Biotech). Anti-osteopontin goat polyclonal antibody (K20) and rabbit polyclonal anti-CD44 (H300) were from Santa Cruz Biotechnology (Santa Cruz, CA). Monoclonal anti-α-tubulin was from Sigma-Aldrich (St. Louis, MO). Anti-phosphorylated mitogen-activated protein kinase/extracellular signal-regulated kinase (ERK) kinase (MEK) 1/2 (Ser²¹⁷/Ser²²¹), anti-phosphorylated p44/42 mitogen-activated protein kinase (Thr²⁰²/Tyr²⁰⁴; ERK), and anti-p44/42 mitogen-activated protein kinase were from Cell Signaling Technology, Inc. (Beverly, MA).

Flow cytometric analysis. Subconfluent cells (1 × 10⁶) were detached from culture dishes. After saturation with 1 μg human IgG per 10⁵ cells, cells were incubated for 20 minutes on ice with antibodies specific for human CD44v6 (R&D Systems) or isotype control antibody. After incubation, unreacted antibody was removed. Cells were then incubated (30 minutes, 4°C) with 100 μL fluorescein-conjugated goat anti-mouse IgG/M (Jackson ImmunoResearch, West Grove, PA) and analyzed on a FACSCalibur cytofluorimeter using the CellQuest software (Becton Dickinson, San Jose, CA). Analyses were done in triplicate. In each analysis, a total of 10⁴ events were calculated.

Chemoinvasion. The cell suspension (1 × 10⁵ cells per well) was added to the upper chamber of Transwell cell culture chambers on a prehydrated polycarbonate membrane filter of 8-μm pore size (Costar, Cambridge, MA) coated with 35 μg Matrigel (Collaborative Research, Inc., Bedford, MA). The lower chamber was filled with complete medium, and when required, purified recombinant osteopontin was added at the concentration of 100 ng/mL. To inhibit Matrigel invasion, cells were preincubated with 10 μg/mL CD44-blocking antibodies (KM81) or with 10 μmol/L U0126. After 24-hour incubation at 37°C, nonmigrating cells on the upper side of the filter were wiped off and migrating cells on the reverse side of the filter were stained with 0.1% crystal violet in 20% methanol for 15 minutes and photographed. The stained cells were lysed in 10% acetic acid. Triplicate samples were analyzed at 570 nm with an ELISA reader (model 550 microplate reader, Bio-Rad). The results were expressed as percentage of migrating cells.

Bromodeoxyuridine incorporation. Cells were seeded on glass coverslips and bromodeoxyuridine (BrdUrd) was added to the cell culture medium at a final concentration of 100 μg/mL (BrdUrd Labeling and Detection kit, Boehringer Mannheim, Mannheim, Germany). After 1-hour incubation, cells were fixed with 70% ethanol/50 mmol/L glycine (pH 2.0). Coverslips were incubated with anti-BrdUrd mouse mAb and with a FITC-conjugated anti-mouse antibody. All coverslips were counterstained in PBS containing Hoechst 33258 (final concentration, 1 μg/mL; Sigma-Aldrich), rinsed in water and mounted in Moviol on glass slides. The fluorescent signal was visualized with an epifluorescent microscope (Axioskop 2, Zeiss) interfaced with the image analyzer software Axiovision (Zeiss, Gottingen, Germany).

RNA silencing. Small inhibitor duplex RNAs targeting human osteopontin were designed with a small interfering RNA (siRNA) selection program available online at http://jura.wi.mit.edu/siRNAext/ and were chemically synthesized by PROLIGO (Boulder, CO). Sense strand for siRNA targeting was (osteopontin siRNA) 5′-AAGCAGCUUUACAACAAAUACCC-3′. As a control, a nonspecific siRNA duplex containing the same nucleotides but in irregular sequence (scrambled) was used. The day before transfection, 1 × 10⁵ cells were plated in 35-mm dishes in DMEM supplemented with 10% fetal bovine serum and without antibiotics. Transfection was done using 360 pmol siRNA and 18 μL Oligofectamine reagent (Invitrogen, Groningen, the Netherlands) following the manufacturer's instruction. Cells were kept in 2.5% serum and BrdUrd incorporation was measured 48 hours after transfection.

Statistical analysis. Statistical evaluation of the data was done with a two-tailed Student's t test when simple comparison between two groups was required; χ² test was used to establish the statistical significance of distributions. Nonparametric Spearman's correlation coefficient method was used to assess the statistical significance of the correlation between clinicopathologic characteristics of tumor and osteopontin expression. Survival curves of the patients were calculated using the Kaplan-Meier method and analysis was done by the log-rank test. Differences were significant at P < 0.05. Statistical analysis was done using the JMP software program version 5.1.1 (SAS Institute, Inc., Austin, TX).

Immunohistochemical detection of osteopontin up-regulation in laryngeal squamous cell carcinomas. Fifty-eight larynx carcinomas at different grades of malignancy and the corresponding normal tissues (Table 1) were tested for osteopontin expression by immunohistochemistry with an anti-osteopontin-specific mAb. Representative immunohistochemical stainings are shown in Fig. 1 and the entire data set is reported in Table 2. Osteopontin was virtually undetectable (<2.0% of cells) in normal tissues (n = 58; data not shown). Hyperplastic epithelia (n = 20) were also constantly negative for osteopontin staining (Fig. 1E; data not shown). In contrast, 93% (54 of 58) of the tumor samples were osteopontin positive (Fig. 1A and C). The signal was confined to tumor cells. The specificity of signal was shown by competition with a molar excess of osteopontin-blocking peptide (Fig. 1B, D, and F). Overall, 37% of T₂ tumors and 73% of T₃/T₄ tumors showed intense (+++) osteopontin immunostaining. Moreover, osteopontin immunostaining was intense (+++) in 21%, 52%, and 78% of G₁, G₂, and G₃ tumors, respectively. Finally, 60% of tumors with lymph node metastases and 92% of tumors with distant metastases had intense osteopontin staining. Accordingly, metastatic tissues were intensely osteopontin positive (data not shown). Thus, osteopontin reactivity was correlated with tumor stage (P < 0.0001) and grade (P < 0.0001) and with the presence of lymph node (P < 0.0046) and distant (P < 0.0001) metastases (Table 3). Importantly, the 5-year survival rate for LSCC patients was negatively correlated with intense osteopontin staining. Five-year survival was 71.40% for osteopontin-positive cases (++) and 51.70% for osteopontin-positive cases (+++) as shown by the Kaplan-Meier survival curves reported in Fig. 2 (P = 0.03, two-sided log-rank test).

The interaction of osteopontin with the CD44 cell surface receptor has been implicated in many signal transduction pathways. CD44 pre-mRNA is encoded by 20 exons. The constant 5′-terminal five exons encode the NH₂-terminal extracellular portion of the CD44 protein, whereas the constant 3′-terminal five exons encode the transmembrane and the short cytosolic tail of the protein. An additional 10 exons (variants v1-v10) are alternatively spliced and encode the extracellular membrane-proximal stem structure (18). Cancer cells often overexpress CD44 variants that include a variable number of “v” exons. The v6 exon has been reported to be important for efficient osteopontin binding (18). Thus, we sought to verify whether CD44v6 molecules were expressed in LSCC. LSCC samples (n = 58) were constantly CD44v6 positive at immunohistochemistry (representative samples are shown in Fig. 3A-C). In tumors, CD44v6-positive cells showed full thickness staining, whereas only basal cells were CD44v6 positive in normal stratified epithelium (data not shown).

Osteopontin and CD44 up-regulation in laryngeal squamous cell carcinomas at protein and mRNA level. Protein lysates were harvested from selected high-stage/grade snap-frozen LSCC samples (T₃, G₃; n = 10) and from the corresponding adjacent normal mucosa from the same patients, and osteopontin protein levels were examined by immunoblotting. As shown in Fig. 4A, the osteopontin protein (M_r ∼60,000) was abundantly expressed in all carcinomas but was barely detectable in matched normal tissues (Fig. 4A). To determine whether up-regulation occurred at transcriptional level, we subjected the same LSCC samples (n = 10) to RT-PCR. Osteopontin mRNA was abundantly overexpressed at mRNA level in tumors with respect to adjacent normal mucosa from the same patients (Fig. 4B; data not shown). Phosphorimaging analysis of band intensity indicated that osteopontin mRNA was 15 ± 3–fold higher in LSCC samples than in the normal tissue counterpart. To validate the results, we analyzed osteopontin mRNA levels in purified dissected tumor cells from five representative high-stage/grade LSCC samples and corresponding normal cells by RT-PCR; the representative samples shown in Fig. 4C showed that osteopontin mRNA accumulation was restricted to tumor cells.

We next examined CD44 expression in LSCC (T₃, G₃) samples (n = 10) using immunoblotting. An intense broad band of a relative molecular mass of 90 kDa was detected in tumors (Fig. 5A; data not shown). In contrast, CD44 was weakly expressed (<10 ± 3–fold compared with LSCC samples) in normal tissue counterparts (Fig. 5A). Then, we used different combinations of amplimers in RT-PCR experiments with RNA extracted from LSCC (T₃, G₃) samples (n = 10). LSCC overexpressed CD44 mRNA species containing the v6 or both v3 and v6 variant exons (Fig. 5B; data not shown).

Osteopontin and CD44v6 overexpression in squamous cell carcinoma cell lines. We evaluated osteopontin expression in six cultured human SCC lines. A primary culture of HNEK was used as a control. Osteopontin protein and mRNA expression was >7-fold higher in SCC than in normal cells (Fig. 6A and B). Moreover, all the SCC cell lines featured high levels (>10-fold with respect to HNEK) of standard and v6-containing CD44 species (Fig. 6C and D). All the cell lines featured abundant cell surface CD44v6 expression by flow cytometry (Fig. 6E; data not shown).

Osteopontin activates intracellular signaling, growth, and invasiveness of squamous cell carcinoma cells. Various SCC cell lines, washed in serum-free medium, were stimulated with exogenous recombinant osteopontin and harvested at different time points. Protein lysates were probed with phosphorylated MEK and phosphorylated ERK (ERK1/2) antibodies. MEK and ERK were readily activated in osteopontin-stimulated cells, peaking at 5 to 15 minutes (Fig. 7A), but not in normal HNEK cells (data not shown). We then examined the ability of SCC cells to synthesize DNA in basal conditions and in the presence of exogenous recombinant osteopontin. Osteopontin stimulated DNA synthesis in SCC cells, washed in serum-free medium, but not in normal HNEK cells (P = 0.01, two tailed Student's t test). Thus, we asked whether osteopontin expression was required for the growth of SCC cells. BrdUrd incorporation was obtained in triplicate after BHY and CAL27 cell transfection with osteopontin or scrambled siRNA. The transient silencing of osteopontin (Fig. 7C, inset) significantly inhibited the growth of BHY cells, whereas the negative control siRNA had virtually no effect (Fig. 7C; data not shown). To determine whether the CD44 receptor mediated these events, we treated BHY, after washing in serum-free medium, with exogenous osteopontin after CD44 blockade with specific antibodies or chemical ERK blockade by the U0126 MEK inhibitor. Activation of MEK and ERK was virtually abrogated by pretreatment with U0126 (Fig. 7A). Stimulation of cell proliferation by osteopontin was obstructed by both anti-CD44 and U0126 (Fig. 7C). We next examined the ability of SCC cells to invade Matrigel in basal conditions and in the presence of exogenous osteopontin. Treatment with osteopontin induced a strong migratory response of tumor cells but not of normal HNEK cells (P = 0.04, χ² test; Fig. 7D). Treatment with CD44-blocking antibodies or U0126 sharply inhibited these effects (P = 0.03, two tailed Student's t test; Fig. 7E).

Here, we show that osteopontin expression was closely correlated with advanced stage, high grade, metastatic disease, and poor survival of LSCC. This is in accordance with our observation that osteopontin affected the signaling and the mitogenic and motile phenotypes of carcinoma cells. Thus, although larger, prospective studies are needed to elucidate the relevance of osteopontin status versus other prognostic factors, our data suggest that osteopontin expression could be exploited as a predictor of outcome in LSCC patients. Interestingly, osteopontin plasma levels have been associated with treatment outcome and survival of HNSCC patients (30).

After its identification as a protein secreted by neoplastic cells (31), osteopontin has been detected in several human tumor types (i.e., gliomas and lung, prostate, gastric, esophageal, and ovarian carcinomas; refs. 16, 17, 19, 20). Osteopontin is also a major determinant of breast (21) and liver (22) cancer metastatization. Numerous growth factors (32), the RAS (33) and SRC (34) oncogenes, and the tumor suppressor p53 (35) regulate the expression of osteopontin. Many of these oncogenic proteins are involved in the pathogenesis of HNSCC (1) and thus might be responsible for osteopontin deregulation in such a tumor type. Osteopontin is able to engage several receptors, including integrins and CD44 variants, and thus may stimulate diverse signaling pathways and influence cellular events that, in turn, favor tumorigenesis and metastasis (16, 17). In particular, osteopontin binds CD44 proteins that contain v6-encoded sequences (14, 15), and osteopontin-CD44v6 binding has been implicated in carcinogenesis (16–18). In agreement with a report of CD44 (the standard CD44 variant in that case) expression in HNSCC (36), we show that osteopontin up-regulation is paralleled by intense expression of CD44 (in particular, CD44v6) in LSCC tissue samples as well as in a panel of SCC cell lines. Thus, CD44v6 is a candidate receptor for osteopontin in LSCC cells. Accordingly, CD44 blockade obstructed osteopontin-mediated cellular effects. CD44 activates a wealth of signaling proteins, among which ERK (37), RAC (38), and RHO (39), as well as secretion of cytokines (40), angiogenic factors, and metalloproteinases (41). This could explain the effects exerted by the osteopontin-CD44v6 axis. On the other hand, osteopontin induces CD44v6 overexpression (42). Of note, our experiments show that osteopontin-CD44v6 binding mediates the effects occurring in SCC cells, but they do not exclude that interactions between osteopontin and integrins (16–18) and between CD44, hyaluronan (16–18), and other membrane receptors, such as members of the MET and ERBB family (43–45), are involved in the osteopontin-CD44 axis as well.

In conclusion, the results of this study suggest that patients affected by LSCC that express high levels of osteopontin protein may be more prone to a poor outcome than patients with low osteopontin-expressing LSCC. Thus, therapies targeted at the molecular mechanism (e.g., ligands that antagonize the interaction of osteopontin with CD44 or antibodies directed against osteopontin and CD44) may prove useful in LSCC patients.

We thank L. Vitiello and L. Racioppi for fluorescence-activated cell sorting analysis and Jean Gilder for text editing.