Defective Transforming Growth Factor β Signaling Pathway in Head and Neck Squamous Cell Carcinoma as Evidenced by the Lack of Expression of Activated Smad2¹

Members of the TGF-β³ superfamily of peptides regulate cellular growth and differentiation through interaction with specific cell surface receptors that phosphorylate and activate intracellular signaling molecules related to the Drosophila gene Mothers against decapentaplegic (Mad; Ref. 1). Different members of the Smad protein family, the vertebrate homologues of Mad proteins, have different roles in signaling and appear to be pathway-restricted. The pathway-restricted Smad2 and Smad3 are phosphorylated after stimulation by TGF-β and activin, whereas Smad1 and Smad5 are phosphorylated and translocate to the nucleus after stimulation by bone morphogenetic protein (1, 2). After phosphorylation, the pathway-restricted Smads form hetero-oligomers with the common mediator Smad, Smad4. The Smad complex then translocates into the nucleus. In the nucleus, Smad proteins regulate transcriptional responses through interactions with DNA and other DNA-binding proteins (2, 3, 4, 5). Recently, the inhibitory Smads, Smad6 and Smad7, have been shown to bind to the TβR-I, precluding the phosphorylation of the pathway-restricted Smads and, consequently, interfering with TGF-β and bone morphogenetic protein signaling (6, 7, 8).

The demonstration that TGF-β has antiproliferative effects in a variety of cell types has led to the hypothesis that inactivation of the TGF-β-signaling pathway contributes to tumor development or progression (9). Thus, the TGF-β-signaling pathway has been postulated to be a tumor suppressor pathway, which is operational in normal cells and functions by delaying progression through the cell cycle. Consistent with this hypothesis, alterations have been identified in many components of this pathway. Mutations in Smad4 (DPC4) and Smad2 (MADR2) have been reported in pancreatic, colon, and lung cancer (10, 11, 12, 13, 14). Both of these Smads have been mapped to 18q21, a region often deleted in several human cancer types.

It has been reported previously (15, 16) that, in pancreatic cancer, Smad6 and Smad7 are overexpressed, suggesting that this is a mechanism of repression of the TGF-β-signaling pathway. In an analysis of human epithelial skin tumors, however, no differences were found in Smad2, Smad6, and Smad7 expression (17). None of these studies examined the expression of activated Smad2 (the Smad2 phosphorylated, Smad2-P). In this study, we have observed variable expression of the Smads that are not pathway-restricted. However, expression of the TGF-β/activin pathway-restricted Smad2 was greatly diminished in poorly differentiated HNSCC. This decrease in the expression of Smad2 was independent of the expression of other components of the TGF-β pathway.

HNSCCs were selected from the archives of the Pathology Department at the H. Lee Moffitt Cancer Center and Research Institute (Tampa, FL) to evaluate protein expression in the context of different histopathological stages in the progression from normal epithelium to invasive cancer. The selected cases had to fulfill all of the following criteria: (a) presence of areas of “non-neoplastic” squamous epithelium “adjacent” to the tumor and lacking histopathological features of dysplasia or malignancy; (b) presence of areas of dysplasia and/or in situ carcinoma; (c) presence of superficial and infiltrating component; (d) presence of at least two different degrees of differentiation; (e) rapid specimen processing with no delayed fixation; (f) free of necrosis; (g) patients who did not receive any previous treatment (with the exception of a diagnostic biopsy); and (h) patients who were initially diagnosed, treated, and followed at our institution, ensuring collection of clinical data, adherence to Institutional Research Board requirements, and patient confidentiality.

The diagnosis of squamous cell carcinoma was made by a head and neck pathologist using standard criteria that included the presence of intracellular bridges, keratinization, keratin pearl formation, various degrees of cytological atypia, mitotic rate, and architectural abnormalities. The degree of differentiation was assigned as follows: (a) well differentiated, large cells with obvious keratinization, intercellular bridges, or keratin pearl formation; (b) moderately differentiated, heterogeneous population of cells of moderate size with only focal keratinization, poorly preserved intercellular bridges, and higher degree of cytological atypia; and (c) poorly differentiated, anaplastic population lacking keratinization and intercellular bridges and with minimal resemblance to squamous epithelium. Definitive stromal infiltration required desmoplastic reaction associated with ragged tumor borders. The superficial component of the invasive carcinoma was defined as that region of the tumor in continuity with the overlying squamous epithelium that invaded the underlying stroma up to a distance of approximately 300 μm. Within this distance, there is a high probability of angiolymphatic invasion. Deep invasion was defined as those areas of the tumor invading to a distance more than 300 μm and/or with tumor tongues separated from the main mass. Clinicopathological staging was assigned following the recommendations of the American Joint Committee on Cancer (18).

Five-μm sections from formalin-fixed, paraffin-embedded tissues were cut and placed on poly-l-lysine-coated slides. Slides were subjected to deparaffination in xylene and hydration through a series of decreasing alcohol concentrations, following standard procedures. Endogenous peroxidase was quenched with a 3% solution of H₂O₂ for 20 min at 37°C. Slides were then washed in deionized water for 5 min. Antigen retrieval was performed by placing the slides in a clear plastic container with a vented top containing citrate buffer [10 mm citric acid, 10 mm sodium citrate (pH 6.0)] in a microwave oven set on high 2 times for 5 min each. The slides were then allowed to cool for 10–20 min, rinsed in deionized water, placed in PBS for 5 min, and drained. Blocking serum was applied, and the slides were incubated in a humid chamber for 20 min at room temperature. After blotting, the following primary antibodies were applied at room temperature at a 1:100 dilution: anti-TβR-II (COOH-terminal domain; Ref. 19); anti-Smad2 (Upstate Biotechnology, Lake Placid, NY); anti-Smad2-P (Upstate Biotechnology); anti-Smad3 (Zymed Laboratories, San Francisco, CA); anti-Smad4 (Upstate Biotechnology); anti-Smad6 (Zymed Laboratories), anti-Smad7 (Santa Cruz Biotechnology); and Ki-67 (Dako Corporation, Carpinteria, CA). After 1 h, the slides were rinsed with PBS and placed in PBS for 5 min. For detection, the Vectastain ABC Kit, Rabbit IgG, Elite series (Vector Laboratories, Inc., Burlingame, CA) was used following the manufacturer’s specifications. The biotinylated secondary antibody was applied for 20 min at room temperature in a humid chamber. At the end of this incubation, the slides were rinsed and placed in PBS for 5 min, followed by the addition of the avidin-biotin complex. The slides were incubated in a humid chamber for 30 min at room temperature and then rinsed and placed in PBS for 5 min. 3,3′-diaminobenzidine, prepared following the manufacturer’s specifications, was applied to the slides, and color development was monitored for 5 min. Slides were rinsed in water, counter-stained with modified Mayer’s hematoxylin for 30 s, washed in running water for 10 min, dehydrated, cleared, and mounted with resinous mounting medium.

The Optimas 6.5 software (Media Cybernetics, Inc., Seattle, WA) was used to quantitate protein expression. The ROIs (normal squamous epithelium, dysplasia, carcinoma in situ, and superficial and deep components of the tumor) were stored as TIFF images using identical magnifications (×40) and camera settings. The total number of nuclei in a given ROI was determined, and constant thresholds were set to discriminate between the brown color of positive nuclei and the negative (blue) nuclei. A protein index was determined as the percentage of positive versus total nuclei.

Protein quantitation, expressed as an index, was compared between the different groups using the Wilcoxon rank-sum (Mann-Whitney U) test.

Genomic DNA and total RNA were obtained from fresh-frozen tissues stored at the H. Lee Moffitt Cancer Center Tissue Procurement Laboratory. cDNA was synthesized from 10 μg of total cellular RNA by reverse transcription with Moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.) using random hexamers as primers. The γ-actin sense (5′-CCCGGTGCTTCTGACCGAGGCC-3′) and antisense (5′-CAGACTGAGTACTTGCGCTC-3′) primers were used as positive controls for the genomic DNA and the reverse transcription reaction. These primer combinations allowed us to distinguish by size the origin (DNA or spliced mRNA) of the PCR products (20). The sequences of the DPC4 and Smad2 sense and antisense primers have been described before (11). The PCR mixture contained 50 mm Tris (pH 9.0), 50 mm KCl, 1.5 mm MgCl₂, 0.01% gelatin, 100 ng of each primer, and 1 unit Taq polymerase (Life Technologies, Inc.). After an initial denaturation at 94°C for 5 min, amplification was carried out for 30 cycles at 94°C for 30 s, 55°C for 1 min, and 72°C for 1 min, followed by a final elongation step at 94°C for 2 min, 55°C for 2 min, and 72°C for 10 min. The PCR products were resolved on a 2% agarose gel and stained with ethidium bromide.

A decrease in the expression of the TGFβ type II receptors is believed to be responsible, at least in part, for the resistance of squamous cell carcinoma to the antiproliferative effects of TGF-β. We have reported previously (21) that TβR-II expression decreases as tumors become less differentiated and more biologically aggressive, suggesting that aberrant TβR-II expression is a contributing factor to the pathogenesis of squamous cell carcinoma. However, well-differentiated squamous cell carcinomas and some poorly differentiated squamous cell carcinomas retain expression of TβR-II. In an effort to understand the mechanisms whereby TβR-II-positive tumors are resistant to the antiproliferative effects of TGF-β, both TβR-II-positive and -negative cases were examined by immunohistochemistry for the expression of other signaling molecules of the TGF-β pathway (Smads). In normal squamous epithelium (Fig. 1), the distribution pattern of protein expression across epithelial layers suggests a favored location for the various TGF-β-signaling molecules. This pattern seems to correlate with the accepted interpretation of the role that Smad molecules play in cellular proliferation, as revealed by Ki-67 immunostaining. Thus, proliferating basal and parabasal layers tend to express with higher intensity the inhibitory Smads (specifically Smad7), whereas upper, more quiescent, and mature layers tend to predominantly express the TGF-β pathway-specific Smad2. The TGF-β pathway-specific Smad3 and the common pathway Smad4 are expressed in all of the layers, and the TβR-I and TβR-II are distributed across layers with a variable pattern (data not shown). Thirteen squamous cell carcinomas were included in this study (see Table 1). The age of the patients ranged from 54 to 82 years with an average of 66. Twelve patients (92%) were male and one (8%) was female. The anatomical location of the tumors included lip, tongue, floor of the mouth, larynx, and the tonsillar region. As can be seen in Table 1, the common Smads (Smad4, Smad6, and Smad7) showed variable expression. The TGF-β pathway-specific Smad3 was consistently expressed in a nuclear location in both the tumor and the non-neoplastic adjacent epithelium. Expression of the other TGF-β pathway-specific Smad, Smad2, was more difficult to ascertain because differences in expression and intensity were observed with different antibodies. The cases listed in Table 1 were evaluated using three different antibodies on consecutive sections of the same tissue block as described in “Materials and Methods.” For two of the antibodies (see “Materials and Methods”), no competing peptide was available; however, for one of the antibodies, competing peptide was used to ascertain binding specificity. The data presented in Table 1 describe Smad2 expression consistently obtained with two of these antibodies. As can be seen, Smad2 is lost in 38% of cases. It is preserved in 75% of the well-differentiated tumors and 50% of the poorly differentiated tumors. However, all of the poorly differentiated tumors lacked expression of the activated pathway-specific Smad2-P. Both molecules, Smad2 and Smad2-P, were expressed simultaneously in only three tumors (23%). Fig. 2 depicts the expression of TGF-β-signaling molecules in a representative TβR-II-positive, well-differentiated tumor. As can be seen, the non-neoplastic squamous epithelium expresses Smad2 (Fig. 2,A), nuclear Smad2-P (Fig. 2,B), Smad3 (Fig. 2,C), Smad4 (Fig. 2,D), Smad6 (Fig. 2,E), and Smad7 (Fig. 2,F). In the adjacent tumor, however, the expression of Smad2-P is lost (Fig. 2,H). In this particular tumor, other Smads (Smad4, Smad6, and Smad7) are expressed (Fig. 2, J–L), and both TβR-I and TβR-II expression was maintained (data not shown). Fig. 3 shows the quantitation steps followed to measure Smad2-P in a representative case. Smad2-P-positive nuclei (brown) and -negative nuclei (blue) are counted by the software to determine the protein index. Image analysis quantitation shows (Table 2) that the non-neoplastic epithelium adjacent to the tumor has a Smad2-P index significantly higher (44.3 ± 14.7) than either moderately differentiated (16.19 ± 3.0) or poorly differentiated (6.52 ± 5.7) tumors. This is independent of levels of TβR-II expression (data not shown).

Impaired activation of the TGF-β-signaling pathway, via loss of Smad2-P, could be responsible for the lack of responsiveness of TβR-II-positive tumor cells to the antiproliferative effects of TGF-β. It has been reported previously (11) that lack of Smad2 expression could be attributable to either deletions or mutations in the MADR2/Smad2 gene. To examine alterations in the MADR2/Smad2 gene in HNSCC tumors, we analyzed genomic DNA and RNA as described by Eppert et al. (11). Smad2- and Smad4-specific primers were used to amplify genomic DNA obtained from the same tissue areas where immunohistochemistry was performed. In all of the cases examined, both MADR2/Smad2 and DPC4/Smad4 were amplified (data not shown) implying that, in the cases examined, there were no MADR2/Smad2 or DPC4/Smad4 gene deletions. Because no gene deletions were observed, RNA from these samples were examined by reverse transcription-PCR analysis for gene expression using primers specific for the MADR2/Smad2 and the DPC4/Smad4 genes. All of the tumors examined expressed RNA for both Smad2 and Smad4 (Fig. 4). Smad2 cDNA PCR amplification was done using overlapping primer combinations that have been used previously (11) for mutational analysis by single-strand conformational polymorphism. Single-strand conformational polymorphism analysis using the conditions described before did not detect any mutations in the tumors examined (data not shown). It could be argued that in some tumors a deletion of the gene could be masked by contamination with normal tissue. However, our data are in agreement with what has been reported in other studies of head and neck tumors (22), where mutations of the TβR-II, MADR2/Smad2, and DPC4/Smad4 genes are found only rarely, suggesting that, in these tumors, lack of Smad2 protein expression is attributable to mechanisms other than genetic alterations.

HNSCC affects an estimated 500,000 patients annually worldwide. In the United States, the annual incidence is estimated at 41,000 cases with 12,300 associated deaths (23). Although significant progress has been made in treatment, advanced HNSCC remains a disease with overall poor prognosis (24). Therefore, it is important to better understand the carcinogenic mechanisms involved, thereby allowing identification of high-risk individuals for prevention, early diagnosis, better therapeutic options, and improved prognosis (25).

In analogy to colon cancer, several groups have proposed that tumorigenesis in HNSCCs is a multistep process driven by the accumulation of genotypic and phenotypic alterations (26, 27, 28). Several groups have examined the expression of oncogenes and tumor suppressor gene products in squamous cell carcinomas. Accumulation of the p53 protein (suggesting possible mutations) and cyclin D1 overexpression have been documented (29, 30). However, no significant association between immunohistochemical detection of p53 protein and survival has been found in patients with advanced head and neck cancer (31, 32, 33). Therefore, it has been concluded that although p53 protein accumulation occurs early in the pathogenesis of oral cavity cancers, it does not appear to have prognostic significance (33). Recently, 100 archived squamous cell carcinomas of the oral cavity were examined to identify tumor-associated alterations in normal adjacent mucosa (34). No correlation was found in expression of p53, Mdm2, Bcl-2, and p21^WAF1/CIP1 between tumors and adjacent mucosa.

TGF-β signaling requires the formation of hetero-oligomeric receptor complexes, most likely a heterotetramer, containing two molecules each of TβR-I and TβR-II (35, 36, 37). A model for the mechanism of action of this heterotetramer complex has been proposed by Wrana et al. (36). In this model, TβR-II is constitutively autophosphorylated. Ligand binding causes recruitment and phosphorylation of TβR-I, the kinase activity of which is essential for the phosphorylation of downstream substrates and biological responses. After complex formation and phosphorylation of TβR-I, a phosphorylation cascade of downstream cytoplasmic substrates, including the Smad proteins, is activated (1, 38). Downstream events in the TGF-β-signaling pathway include complex formation of Smad2 or Smad3 with Smad4, translocation of the Smad/Smad4 complex to the nucleus, and eventual activation of target genes (1). In the absence of ligand, the inhibitory Smads, Smad6 and Smad7, are localized predominantly in the nucleus. Upon TGF-β receptor activation, they accumulate in the cytoplasm (39) and associate with the ligand-activated TGF-β receptor complex in the cell membrane, antagonizing TGF-β family signaling by preventing the activation of signal-transducing Smad complexes.

Studies (9) on the TGF-β antiproliferative effects in a variety of cell types strongly suggest that inactivation of the TGF-β-signaling pathway contributes to tumor development or progression. Thus, we hypothesized that a defect in any of the components of the TGF-β-signaling pathway might render the pathway inactive, contributing to the malignant phenotype of HNSCC. A decrease in the number of TβR-II-binding sites has been associated with loss of responsiveness to TGF-β (40, 41, 42, 43, 44, 45, 46, 47, 48, 49). This decrease in TβR-II expression correlates with tumorigenicity in many cell lines (48, 50). We have reported previously (21) that TβR-II expression decreases as HNSCCs become less differentiated and more biologically aggressive. In a small subset of tumors, however, TβR-II expression was maintained.

Analysis of the expression of TGF-β intracellular signaling molecules (Smads) in HNSCC reveals that expression of the activated pathway-specific Smad2-P is lost in a significant percentage of these tumors. Lack of expression of Smad2 and/or Smad2-P could further contribute to signaling abnormalities at a postreceptor level in tumors with abnormal receptor expression and be responsible for defective signaling in tumors with normal receptor expression (TβR-II; Ref. 21). Interestingly, the expression of the TGF-β pathway-specific Smad3 was not lost in any of the tumors examined. It is possible, however, that examination of an activated form of Smad3 would reveal changes in patterns of expression. In our study, four TβR-II-positive tumors express Smad2-P, suggesting that the inability to respond to TGF-β antiproliferative response might be attributable to other defects, perhaps in genes that are downstream targets of the Smads.

Expression of the Smads that are not TGF-β-specific (4, 6, and 7) in tumors was similar to that of adjacent non-neoplastic squamous epithelium. This is not surprising, because many different pathways use these molecules, and their absence may be incompatible with cell survival. This has also been reported (17) in skin tumors where no significant differences were observed in the expression of Smad4, Smad6, and Smad7 between tumor and nontumoral cells. This is in contrast to reports (15, 16) in pancreatic cancer, however, where overexpression of the inhibitory Smads, Smad6 and Smad7, has been proposed as another mechanism for inactivation of the TGF-β-signaling pathway. Our data supports the hypothesis that aberrant expression of critical molecules may play a role in the inactivation of the TGF-β pathway at different levels contributing to HNSCC tumorigenesis.

Loss of heterozygosity in chromosome 18q has been reported frequently in HNSCCs (51, 52, 53). Both the DPC4/Smad4 and MADR2/Smad2 genes have been mapped to 18q (10, 11), and genetic alterations in the DPC4/Smad4 and MADR2/Smad2 genes have been reported in human colon and pancreatic cancers, in which 18q is frequently lost (54). In head and neck cancers, however, our amplification of both DPC4/Smad4 and MADR2/Smad2 genes has failed to reveal loss of this chromosomal region. This is in agreement with what has been recently reported by Takebayashi et al. (55) who analyzed the minimally lost areas of 18q in 57 HNSCCs. In this analysis, both the MADR2/Smad2 and the DPC4/Smad4 genes were always present, suggesting that other tumor-suppressor gene(s) may be involved in tumors with observed loss of heterozygosity in 18q.

Mutations of DPC4/Smad4, Smad3, and MADR2/Smad2 genes have been reported in pancreatic (10), colon (11, 12), lung (13, 14), and ovarian cancers (56). Analysis of other Smad genes (i.e., Smad 1, 5, and 6) in a variety of tumors, including those from colon, breast, lung, and pancreas, revealed no mutations (22, 54). The genetic changes found in Smad3 in over 40% of ovarian cancers are not accompanied by amino acid changes, probably representing a polymorphism not involved in tumorigenesis (56). We found that the lack of expression of Smad2 protein in head and neck tumors is not attributable to genetic alterations, as reported for colon and pancreatic cancers. This is in agreement with what has been reported in head and neck, hepatocellular, breast, lung, ovary, and prostate carcinomas (14, 22, 45, 56, 57, 58, 59, 60, 61, 62, 63), where mutations of the TβR-II, MADR2/Smad2, and DPC4/Smad4 genes are rare. This suggests that genetic instability in other Smad genes does not account for the widespread resistance to TGF-β.