SEARCH

SEARCH BY CITATION

Keywords:

  • integrin α3;
  • integrin β4;
  • integrin β5;
  • lymph node metastasis;
  • distant metastasis

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

BACKGROUND.

The objective of the current study was to identify biomarkers that reflect the clinical course of squamous cell carcinoma of the tongue (TSCC).

METHODS.

TSCC tissue samples from 66 patients were subjected to gene expression analysis by real-time polymerase chain reaction. Eleven integrin family genes and 14 genes used for normalization, including housekeeping genes and genes that encode desmosomal, cytoskeletal, and extracellular matrix molecules, were considered. Multivariate statistical analysis was performed on 154 expression ratios of integrin genes with clinical parameters.

RESULTS.

In principal-component analysis, the first principal component was related to the outcome of death, and the second principal component mainly reflected the tendency for cervical lymph node (LN) metastasis. The former axis consisted of the variance of the integrin β4 gene (ITGB4) and ITGB5 expression levels, and the latter axis agreed with the expression level of the integrin α3 gene (ITGA3). Multivariate logistic regression analysis with cervical LN metastasis as the response variable concordantly identified ITGA3/junction plakoglobin gene (JUP) expression (P = .02) and ITGB5/paxillin gene (PXN) expression (P = .04) as significant factors. Only ITGB4/JUP expression was identified as a significant factor in terms of the outcome of death (P < .00049) by a Cox proportional hazards model. The group with high ITGB4/JUP levels exhibited a significantly high death rate on a Kaplan-Meier curve (P < .0001; Wilcoxon and log-rank tests).

CONCLUSIONS.

The expression levels of ITGA3, ITGB4, and ITGB5 with functional normalization by desmosomal or cytoskeletal molecule genes were selected as candidate biomarkers for cervical LN metastasis or for the outcome of death in TSCC. Cancer 2008. © 2008 American Cancer Society.

Tongue cancer is the most frequent malignant neoplasm in the oral region. Previous studies reported a poor prognosis and overall 5-year disease-specific survival rates of approximately 30% among patients who presented with lymph node (LN) metastasis.1, 2 Tongue cancer tends to metastasize earlier than squamous cell carcinomas (SCC) of other oral sites. Therefore, optimization of the individual treatment plan based on biologic malignancy is essential for minimizing functional loss after treatment and for controlling the cancer. Currently, histopathologic grade, conditions of LN metastasis, tumor size, and serum tumor markers are used as prognostic factors for patients with cancer of the tongue.3, 4 Recently, attempts to develop an improved diagnostic system have been made on the basis of gene expression levels.5, 6 In oral cancer, the matrix metalloproteinases, cadherins, integrins, et cetera, have been considered as candidate biomarkers, but none have successfully attained the level of practical application.7, 8 The status of LN metastasis is related closely to outcome in patients with tongue cancer and currently is used as the effective clinical parameter. Nevertheless, it cannot be applied to early stages before metastasis. In addition, any judgment reached after the manifestation of metastasis is too late to obtain the best therapeutic effect. The establishment of biomarkers that reflect biologic functions, such as cell cycle, proteolysis, adhesion, motility, immune response, and vascularization, would enable clinicians to make an accurate diagnosis of malignancy at an early stage and allow the optimization of treatment.

The integrin (ITG) molecule is composed of a transmembrane-type heterodimer that consists of an α chain and a β chain. It functions as a cell-surface receptor that connects the cytoplasm and the extracellular matrix (ECM). Currently, human ITG forms a superfamily that is made up of 18 α chains and 8 β chains, which form 24 pairs of ITG receptors.9, 10 These ITG receptors are related to specific binding to ECM ligands. Generally, 10% of the α chains and β chains compose intracellular domains that are bound to the cytoskeleton through cytoplasmic anchor proteins.11–13 Specific cell adhesion by ITG receptors mediates extracellular mechanical and chemical signals to the interior of cells. This ITG signaling, which modulates many different signal-transduction cascades, supports cell survival, apoptosis, proliferation, and motility and also influences cell differentiation.9, 12–14

We previously performed gene expression analysis of oral SCC (OSCC) by using a combinational DNA (cDNA) microarray. Consequently, it was determined that the ITG α3 (ITGA3) and ITG β4 (ITGB4) genes exhibited significantly high expression levels in OSCCs with LN metastasis.7 The results from that study specifically indicated the potential of ITG family molecules to serve as biomarkers of malignancy in SCC of the tongue. Quantitative real-time polymerase chain reaction (qPCR) gene expression analysis was used to determine the expression of 6 ITG α chains and 5 ITG β chains that reportedly are active in malignant tumor behavior.7, 15–20 In addition to housekeeping genes, various other genes related to ITGs in terms of function or tissue localization were evaluated to determine the optimum normalization molecules. In data analysis, an inclusive examination of the biologic significance indicated from gene expression data was performed by using multivariate analysis. In the current exploratory study, certain ITG gene expression ratios were identified that are related specifically to the outcome of death as well as distant organ metastasis, both of which have been difficult to predict with existing diagnostic systems.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

The tumor samples that were used for the current gene expression analyses were collected at the time of biopsy or surgical resection from 66 patients with tongue cancer who were treated at the Dental Department of Niigata University Medical and Dental Hospital, the Special Dental Care and Oral Surgery of Shinsyu University Hospital, and the Division of Oral Surgery of Nagaoka Red Cross Hospital from 1999 to 2005 (Table 1). The study protocol for this project was approved by the ethics committees of each institution. A written letter of consent was processed after obtaining the patients' informed consent to participate in this study.

Table 1. Clinicopathologic Data From 66 Patients With Squamous Cell Carcinoma of the Tongue
Clinicopathologic factorNo. of patients (%)
  • Y-K grade indicates Yamamoto-Kohama grade.

  • *

    Major width of the tumor.

  • Tumor (T) category according to the International Union Against Cancer (UICC) TNM classification of malignant tumors of the lip and oral cavity.

  • Lymph node (N) category according to the UICC TNM classification of malignant tumors of the lip and oral cavity.

  • §

    Histopathologic classification of oral squamous cell carcinoma according to Yamamoto et al., 1983.21

Average age [range], y62.38 [21–91]
Sex
 Men41 (62.1)
 Women25 (37.9)
Average observation period [range], d795 [115–1821]
Tumor size, mm*
 >2020 (30.3)
 20–4032 (48.5)
 ≥4014 (21.2)
Tumor status
 T123 (34.8)
 T222 (33.3)
 T35 (7.6)
 T416 (24.2)
Lymph node metastasis
 N030 (45.4)
 N113 (19.7)
 N2b12 (18.2)
 N2c11 (16.7)
Histologic Y-K grade§
 2–323 (34.8)
 4c33 (50)
 4d10 (15.2)
Distant metastasis
 Negative60 (90.9)
 Positive6 (9.1)
Local recurrence
 Negative64 (97)
 Positive2 (3)
Chemotherapy
 No43 (65.2)
 Yes23 (34.8)
Radiotherapy
 No39 (59.1)
 Yes27 (40.9)
Outcome
 Alive54 (81.8)
 Dead12 (18.2)

Extraction of Total RNA From Carcinoma Tissues

Cancer tissue specimens were preserved by immersion in RNAlater (Ambion Inc., Austin, Tex) immediately after sampling. The extraction of total RNA was performed after homogenization by using an Ultra-TurraxT8 (IKA Labortechinik, Staufen, Germany) in TRIzol reagent (Invitrogen Corp., Calif) according to the standard protocol for the reagent. Synthesis of first-strand cDNA was performed by reverse transcription using total RNA (2 μg) as a template (Super Script II; Invitrogen Corporation, Carlsbad, Calif).

Gene Expression Analysis by qPCR

Gene expression analysis was performed by qPCR (Smart Cycler; Cepheid, Sunnyvale, Calif) using cDNA synthesized from the cancer specimens. qPCR by real-time monitoring with a TaqMan probe (TaqMan Gene expression assays; Applied Biosystems, Foster City, Calif) was performed according to the following protocol: 600 seconds at 95 °C, followed by thermal cycles of 15 seconds at 95 °C, and 60 seconds at 60 °C for extension. Relative standard curves representing several 10-fold dilutions (1:10:100:1000:10,000:100,000) of cDNA from an OSCC tissue sample were used for linear regression analysis of other samples. In this study, 11 ITG family genes (ITGA1, ITGA2, ITGA3, ITGA5, ITGA6, ITGAv, ITGB1, ITGB3, ITGB4, ITGB5, and ITGB6), 3 housekeeping genes (actin β [ACTB], glyceraldehyde-3-phosphate dehydrogenase [GAPDH], and 18-second ribosomal RNA [18sRNA]), 7 ECM genes (tenascin C [TNC], fibronectin 1 [FN1], laminin α3 [LAMA3], LAMA4, LAMA5, collagen I α1, and vitronectin), and 4 of the desmosomal anchor protein genes (junction plakoglobin [JUP], plectin 1, paxillin [PXN]. and the cytoskeletal molecule keratin 5 [KRT5]) were considered as subject genes.

Statistical Analysis

Expression levels of the 11 ITG family genes mentioned above were normalized with the expression levels of the 14 functionally related genes so that, in total, 154 gene expression ratios were calculated for each of the 66 patients with tongue SCC. The correlations between cervical LN metastasis or death and these ITG gene expression ratios were subjected to univariate analysis with the Mann-Whitney U test. Taking into consideration the possibility of multiple testing errors, P values were used simply to assess involvement in the clinical outcomes without assigning any level of significance. Because many factors were analyzed in this study, combinations of univariate and multiple multivariate statistical methods were applied with a stepwise procedure in which multiple testing with the univariate method was used only to optimize the number of variables. Consequently, 12 different ITG gene expression ratios that exhibited P values ≤.01 were used expediently for the subsequent multivariate statistical analyses. For clinical parameters, age, sex, tumor size, cervical LN metastasis, multiple cervical LN metastasis (≥4), distant metastasis, experience of chemotherapy, radiation therapy, and death were included as variables in the multivariate analyses. The correlations and loading between the variables were reviewed by principal-component analysis (PCA). Subsequently, a multivariate logistic regression analyses using cervical LN metastasis as a response variable and the outcome of death as an endpoint were performed by using a Cox proportional hazards model and a Kaplan-Meier curve (SPSS version 12.0; SPSS Inc., Chicago, Ill). A P value ≤.05 was assigned for the level of significance.

Histologic and Immunohistochemical Observations

Histologic examination of tumor specimens was performed on 10% formalin-fixed, paraffin-embedded sections stained with hematoxylin and eosin. The same sections were subjected to immunohistochemical staining for ITG species, D2-40, and CD31 by using the ChemMate Envision System (DakoCytomation, Copenhagen, Denmark). After incubation with the primary antibodies overnight at 4 °C, the sections were treated with polymer-immune complexes (EnVision+ peroxidase, rabbit/mouse; DakoCytomation; 1:1 dilution) for 1 hour at room temperature. Peroxidase reaction products were observed by incubation with 0.02% 3′3-diaminobenzimine (Dohjin Laboratories, Kumamoto, Japan) in 0.005% hydrogen peroxide. Finally, the sections were counterstained with hematoxylin. For control studies, the primary antibodies were replaced with preimmune rabbit immunoglobulin G (IgG) or mouse IgG1.

Antibodies

Mouse monoclonal antibody against human ITG α3 (clone P1B5, IgG1; 1:75 dilution) and rabbit polyclonal antibody against human ITG β5 (1:200 dilution) were obtained from Chemicon International (Calif). Rabbit polyclonal antibody against human ITG β4 (1:100 dilution) was obtained from Santa Cruz Biotechnology (Santa Cruz, Calif). For the demonstration of lymphangiogenesis and angiogenesis, a mouse monoclonal antibody against human podoplanin and against human CD31 (Dako Cytomation) were used, respectively.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Univariate Analysis

According to results from the Mann-Whitney U tests with 154 gene expression ratios, many gene expression ratios for ITGA3 and ITGB5 exhibited low P values for the status of cervical LN metastasis (data not shown). Similarly, an accumulation of low P values among the gene expression ratios for ITGB4 and ITGB5 were observed for the outcome of death (data not shown). Because the objective of the univariate step was to optimize the number of factors that were adequate for subsequent multivariate analyses, we considered minimizing the type I error rather than increasing the type II error. ITGA3/GAPDH, ITGA3/KRT5, ITGA3/JUP, ITGB4/KRT5, ITGB4/JUP, ITGB5/ACTB, ITGB5/FN1, ITGB5/TNC, ITGB5/LAMA3, ITGB5/LAMA4, ITGB5/LAMA5, and ITGB5/PXN ratios that exhibited P values <.01 were used for the subsequent multivariate analyses.

Multivariate Analysis

Based on the results of PCA using clinical parameters and the ITG gene expression ratios as variables, the first through seventh principal components displayed eigenvalues >1, and their cumulative proportion of variance reached 0.729. The first, second, and third principal components each included >2 variables with factor loading >0.5 and eigenvalues >2 (Table 2). For the first principal component, distant metastasis, multiple LN metastasis, and death displayed a factor loading of 0.455, 0.418, and 0.555, respectively (Table 2). Therefore, the first principal component (Z-1) reflected death caused by uncontrollable tumor expansion from the clinical viewpoint. This first principal component was defined by the distribution of the 10 gene expression ratios of ITGA3, ITGB4, and ITGB5 that displayed factor loading from 0.5 to 0.7. In contrast, for the second principal component (Z-2), only 2 gene expression ratios, ITGA3/JUP and ITGA3/KRT5, displayed factor loading >0.6, and positive cervical LN metastasis status coincidentally had 1 of the highest factor loadings (0.67) among the clinical parameters. Conversely, because factor loading for the outcome of death and for distant metastasis was 3.47 and 0.32, respectively, the second principal component reflected cervical LN metastasis but was not related directly to uncontrollable tumor advancement. The finding that factor loading for tumor size was 0.43 in the second principal component suggests that tumor size was involved as a factor in the formation of cervical LN metastasis to some extent. Nevertheless, because the factor loading for tumor size was −0.159 in the first principal component, its correlation with the outcome of death was weak. A scatter plot of the factor loading of the first and second principal components that visually presents the relations between molecular and clinical parameters is provided in Figure 1.

thumbnail image

Figure 1. Two-dimensional scatter plot of the first (Z-1) and second (Z-2) principal component axes. The Z-1 axis had a tendency to be related mainly to an outcome of death, and the Z-2 axis had a tendency to be related mainly to cervical lymph node (LN) metastasis. ITGA3 indicates integrin α3 gene; GAPDH, glyceraldehyde-3-phosphate dehydrogenase gene; KRT5, keratin 5 gene; JUP, junction plakoglobin gene; ITGB5, integrin β5 gene; TNC, tenascin C gene; LAMA3, laminin α3 gene; PXN, paxillin gene; ACTB, actin β gene; FN1, fibronectin 1 gene.

Download figure to PowerPoint

Table 2. Principal Component Analysis With Clinical Parameters and the 12 Gene Expression Ratios
ParameterZ-1Z-2Z-3
  • Z-1 indicates the first principal component; Z-2, second principal component; Z-3, third principal component, LN, cervical lymph node; ITGA3, integrin α3 gene; GAPDH, glyceraldehyde-3-phosphate dehydrogenase gene; KRT5, keratin 5 gene; JUP, junction plakoglobin gene; ITGB5, integrin β5 gene; TNC, tenascin C gene; LAMA3, laminin α3 gene; PXN, paxillin gene; ACTB, actin β gene; FN1, fibronectin 1 gene.

  • *

    Principal components with eigenvalues >1 were considered informative.

  • The ratio of the variance explained by the individual component to the total variance.

  • Represents cumulation with former proportion of variance.

  • §

    Major width of the tumor.

  • ||

    The most significant clinical factor for each principle component.

Eigenvalue*5.1643.1472.173
Proportion of variance, %23.514.39.9
Cumulative proportion of variance, %23.537.847.7
Factor loading
 Age0.087−0.2750.430
 Sex−0.237−0.0440.338
 Observation period0.287−0.585−0.377
 Tumor size§−0.1590.430−0.192
 LN metastasis0.3980.674||−0.032
 Distant metastasis0.4550.3230.544
 Multiple LN metastasis ≥40.4180.3140.214
 Death outcome0.555||0.3470.566||
 Chemotherapy−0.1200.554−0.481
 Radiotherapy0.0570.519−0.174
 ITGA3/GAPDH0.6440.079−0.089
 ITGA3/KRT5−0.1570.691−0.210
 ITGA3/JUP0.0920.623−0.166
 ITGB5/TNC−0.6910.0440.299
 ITGB5/LAMA3−0.5370.3380.265
 ITGB5/PXN−0.529−0.2360.321
 ITGB4/KRT50.7070.0730.355
 ITGB4/JUP0.7780.0070.338
 ITGB5/ACTB−0.7450.2280.295
 ITGB5/FN1−0.5140.0750.141
 ITGB5/LAMA4−0.5380.2660.109
 ITGB5/LAMA5−0.6490.1070.262

The ITGA3/JUP and ITGB5/PXN ratios were identified as variables that affected cervical LN metastasis in multivariate logistic regression analysis (Table 3). The ITGA3/JUP ratio had a positive correlation, and the ITGB5/PXN ratio had a negative correlation.

Table 3. Multifactorial Logistic Regression Analysis With Cervical Lymph Node Metastasis
VariableβSE of βPOR
  1. β indicates logistic regression coefficient; SE, standard error, OR, odds ratio; ITGA3, integrin α3 gene; JUP, junction plakoglobin gene; ITGB5, integrin β5 gene; PXN, paxillin gene.

ITGA3/JUP4.0131.733.02155.287
ITGB5/PXN−2.0951.023.0410.123
Invariable0.2270.582.6971.255

Only the ITGB4/JUP ratio was identified as a significant variable in regression analysis using a Cox proportional hazards model for the outcome of death (Table 4). The cutoff ratio of 0.15 was set with reference to the results of a receiver operating characteristic curve; then, a survival curve was calculated according to the Kaplan-Meier method for the groups with ITGB4/JUP ratios >0.15 and <0.15 (Fig. 2). The group with ITGB4/JUP ratios <0.15 had higher survival rates (P < .0001; Wilcoxon test and log-rank test).

thumbnail image

Figure 2. Kaplan-Meier survival curves for patients with squamous cell carcinoma of the tongue who had integrin β4 gene/junction plakoglobin gene (ITGB4/JUP) ratios of 0.15, >0.15, and <0.15 (P < .0001; generalized Wilcoxon test and log-rank test).

Download figure to PowerPoint

Table 4. Cox Proportional Hazard Model for Uncontrollable Death
VariableβSE of βPOR95% CI for OR
Lower limitUpper limit
  1. 95% CI indicates 95% confidence interval; OR, odds ratio; β, logistic regression coefficient; SE, standard error; ITGB5, integrin β5 gene; JUP, junction plakoglobin gene.

ITGB4/JUP13.873.98.000491,055,671432.172,578,695,704

Histologic and Immunohistochemical Observations in Relation to ITG Gene Expression Levels

Expansive growth of alveolus tumor nests was observed among tumors with low ratios of both ITGB4/JUP and ITGA3/JUP (Fig. 3A and 3B). An invasion of small nests of cells from the margin of the differentiated carcinoma cell mass was observed in a patient who presented solely with a high ITGA3/JUP ratio (Fig. 3C and 3,D). In patients who had high ratios of both ITGA3/JUP and ITGB4/JUP, invasive growth of small cell nests or of single cells that consisted of undifferentiated carcinoma cells was observed along the entire tumor cell mass (Fig. 3E and 3F). Nevertheless, histologic characteristics that represented a difference between tumors with high ITGA3 expression and tumors with high ITGB4 expression could not be extracted. The precise prediction of a high-malignancy trait that was related specifically to distant metastasis and an outcome of death was practically impossible based on histology.

thumbnail image

Figure 3. Histology of hematoxylin-and-eosin staining in paraffin-embedded sections. (A,B) Tongue cancers with low levels of both the integrin α3 gene (ITGA3)/junction plakoglobin gene (JUP) ratio and the integrin β4 gene (ITGB4)/JUP ratio were correlated with low invasive morphology without metastasis. (C,D) Tongue cancers with high levels of only the ITGA3/JUP ratio exhibited invasive morphology with lymph node metastasis at early stages, although regional control could be achieved. (E,F) Tongue cancers with high levels of both the ITGA3/JUP ratio and the ITGB4/JUP ratio exhibited invasive morphology; and, because of lung metastasis, an outcome of death followed.

Download figure to PowerPoint

The α3, β4, and β5 integrins displayed specific distribution in OSCC tissues on immunohistochemical analysis. The α3 integrin was localized atypically to carcinoma cells in tumor tissues (Fig. 4A). Immunoreactivity for the β4 integrin was observed in the basal area of normal epithelium (data not shown) and either was localized at the margin of carcinoma cell nests or small cell groups or was observed on the cytoplasm of migrating single cells at the invasive front of the tumor (Fig. 4B and 4D). The β5 integrin was immunoreactive from the parabasal to the acanaceous cell layers in the epithelium (data not shown). In invasive tumor tissues, β5 integrin was localized to differentiated carcinoma cells in the central part of cancer cell nests (Fig. 4C). The β4 integrin-immunopositive cancer cells that invaded normal blood and lymphatic vessels, along with the involvement of muscle tissues, were stained at the invasive front of highly metastatic tumors (Fig. 4D–4F).

thumbnail image

Figure 4. Immunohistochemistry of tongue squamous cell carcinoma tissues demonstrating the expression of α3 integrin (A), β4 integrin (B), and β5 integrin (C). Serial sections adjacent to the section shown in Figure 3F were immunostained with the β4 integrin antibody (D), the CD31 antibody for the demonstration of angiogenesis (E), and the D2-40 antibody for lymphangiogenesis (F).

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

The results from the current study indicate that there are 2 types of cervical LN metastasis in cancers of the tongue. One is simple LN metastasis, which can be controlled regionally, and the other is characterized by uncontrollable expansion, leading to death. Although it is difficult to distinguish between these 2 types of cervical LN metastasis by using the currently available diagnostic systems, the gene expression levels of ITGA3, ITGB4, and ITGB5 may serve as accurate biomarkers based on the biologic functions of cancer cells that are essential for individualized cancer therapy.

PCA was effective for understanding the correlation between clinical parameters and ITG gene expression ratios. Factors that were predictive of cervical LN metastasis, distant metastasis, and death were distributed between the first (Z-1) and second (Z-2) principal component axes on the scatter plot for factor loading. The findings suggested that these clinical status factors were influenced by 2 different biologic characteristics. The probability that death would be the outcome in OSCC could be presented numerically by establishing the appropriate biomarker of the ITGB4 or ITGB5 gene expression ratio distributed along the first principal component axis (Z-1). Conversely, because ITGA3/JUP or ITGA3/KRT5 was distributed along the second principal component axis (Z-2), these ratios had the potential to serve as excellent biomarkers for evaluating the probability of regionally controllable cervical LN metastasis.

Multivariate logistic regression analysis of cervical LN metastasis status and response variables indicated that the ITGA3/JUP and ITGB5/PXN ratios were significant explanatory variables. The ITGA3/JUP ratio was correlated with simple cervical LN metastasis in the PCA results and had the highest regression coefficient. This indicated that the ITGA3/JUP ratio was the most potent biomarker for predicting latent cervical LN metastasis. The ITGB5/PXN ratio was 1 of the factors related to an outcome of death in the PCA results. These results suggested that the possibility of cervical LN metastasis could be judged by the summation of 2 independent biologic traits (ie, factors that represent regionally controllable cervical LN metastasis and locally uncontrollable tumor expansion, respectively).

Only the ITGB4/JUP ratio was related to an outcome of death by the Cox proportional hazards regression model (hazard ratio, 1,055,671; P < .00049). This result was supported by the finding that the majority of the ITG gene expression ratios were distributed around the first principal component Z-1 axis, which reflected an outcome of death according to the PCA results. The Kaplan-Meier survival curve also clearly indicated the diagnostic value of ITGB4/JUP. In addition, according to the results from the generalized Wilcoxon test (P < .0001) and the log-rank test (P < .0001), the explanatory power of this factor was evident in both the short term and the long term.

The α3 integrin, as a subunit of the α3β1 integrin complex, is comprised of receptors for ECM molecules in tissues such as fibronectin, laminin 5, laminin 10, and laminin 11.10 It has been established that α3β1 integrin is involved in the maintenance of basement membrane integrity, cell proliferation and motility, and the survival of migrating keratinocytes through adhesion to the basement membrane component laminin 5.22–25 A recent report indicated the early arrest of tumor cells through the interaction of α3β1 integrin with laminin 5 in exposed basement membranes in the pulmonary vasculature.26 However, the clinical significance of α3β1 integrin remains controversial. Some studies reported that low expression of α3 integrin was related to poorly differentiated histology, LN metastasis, and a resultant low disease-free survival rate.24, 27 In contrast, another study presented results indicating a correlation between positive expression of α3β1 integrin and invasiveness of carcinoma.7, 28 Finding a proper explanation for these discrepant clinical results is a difficult task. Nevertheless, because most of the previous reports obtained results from nonquantitative immunohistochemical analyses, we expect that further investigations using quantitative procedures will help to resolve the issue.

The β4 integrin originally was identified as a tumor-associated antigen and is composed of α6β4 integrin with an α6 subunit. In epithelial structures, the function of the α6β4 integrin is to bind intracellular intermediate filaments and ECM components as a hemidesmosome-associated, laminin-binding integrin. The adhesion between α6β4 integrin and basement membrane component laminin 5 triggers signal transduction toward the inside of the cell in addition to maintaining epithelial structure.29 A noteworthy point is that α6β4 integrin is involved not only in the stabilization of the epithelial structure but also in cell motility, functions that appear to be the opposite of each other. It has been suggested that an association between the cytoplasmic domain of α6β4 integrin and F-actin controls the formation of filopodia and lamellae in carcinoma cells and that, through this process, cell motility is effected.30 Clinical evidence that the expression of α6β4 integrin triggers the invasive process and the associated poor prognosis has been provided in malignant tumors.31, 32 Their functional diversity implies that the cytoskeletal association of the α6β4 integrin is different from the anchoring of normal epithelial cells. The results of this study demonstrated that tongue SCCs with low tissue expression levels of ITGB4 exhibited β4 integrin immunoreactivity only in the basal cell layers of the tumor, as observed in normal epithelium. In contrast, in tongue SCCs with high tissue expression levels of ITGB4, cytoplasmic localization of β4 integrin was observed in highly invasive carcinoma cells. A similar finding was reported in carcinomas of other organs.33 These findings support the functional specificity of β4 integrins in carcinoma.

Moreover, because the reported function of integrin signaling closely relates to the results of the current study, it is believed that the role of such signaling in the survival of carcinoma cells is significant. It is essential for the formation of metastasis to avoid apoptosis, to maintain cell motility, and for growth of the nascent colony at the invasive front of the tumor or in metastatic tissue. It was reported that tumor cells obtain these abilities through autocrine laminin 5-α6β4 integrin ligation and that this process is regulated through vascular endothelial growth factor or protein kinase B-nuclear factor κB signaling.34, 35 Similarly, it was suggested that β5 integrin signaling is involved in the survival of cells, although clinical evidence has not been reported. The β5 integrin is a subunit of the αvβ5 integrin and functions as a receptor for fibronectin and vitronectin.10 A previous report demonstrated that anoikis, anchorage-related apoptosis in epithelial and endothelial cells that are detached from matrix or that are attached through the wrong molecules, requires the cytoplasmic domain of β5 integrin. This result suggests that down-regulation of the αvβ5 integrin through simultaneous up-regulation of the αvβ6 integrin protects SCC cells from anoikis by activation of the Akt survival signal.36 These findings were believed to indicate a mechanism for isolated tumor cells engaged in the process of metastasis to promote the anchorage-independent survival, migration, and growth of nascent tumors. The results of the current study indicated that an increase in the expression level of ITGB4 and a decrease in ITGB5 were associated with distant metastasis and an outcome of death. These findings are consistent with the aforementioned molecular function of the β4 and β5 integrins. Our results support the notion that a key factor in regrowth from isolated metastatic or migrating tumor cells is the acquisition of resistance to anoikis by OSCC cells.

Gene expression data are influenced by various conditions, such as bias from cell populations in tumor tissue, sampling sites, and degradation of molecules. Under such conditions, devices for obtaining consistent gene expression data were considered to be an inevitable subject of study in terms of the clinical application phase. In general, normalization of gene expression data has been performed by obtaining expression ratios to housekeeping genes. Nevertheless, as a result of the improvement in quantitative determination by qPCR, it was revealed that this technique is not always appropriate for normalization, because the expression levels of housekeeping genes change greatly, depending on the tissue type or sampling conditions.37 To optimize the normalization of a pair of genes, we tested the ratios of 154 ITG family genes to ligand ECM molecules, cytoskeletal components, and cytoplasmic anchor molecules with a consideration of localization and/or functional relevance. Consequently, the messenger RNA expression levels of KRT5, which compose an intermediate filament by binding to α6β4 integrins on the cytoplasmic side of the hemidesmosome, and the expression levels of JUP which is a component of the attachment plaque that lines the cytoplasmic side of the desmosome,38 were revealed as effective for normalization. JUP functions in the binding of adhesion molecules, such as cadherin, to cytoskeletal filament, such as keratin, at the attachment plaque of the desmosome.39 All of the molecules α3 integrin, β4 integrin, β5 integrin, keratin 5, and JUP localize and function in epithelial cells.15, 20 Therefore, increases in the ITGA3/KRT5, ITGB4/KRT5, ITGB4/JUP, and ITGA3/JUP ratios suggest a decrease of the desmosome, the activation of integrin signaling, and permanent rearrangement of the hemidesmosomal structure in OSCC tissue. In consequence, these changes in the gene expression ratios can be understood as carcinoma cells invading and metastasizing aggressively as a result of a decrease in binding among OSCC cells along with the acquisition of resistance to apoptosis. Biologic rationales for these ITG gene ratios and their clinical effectiveness as biomarkers remain to be investigated.

Reliable criteria for precise estimation of the probability of distant metastasis have yet to be established. ITG gene expression data are useful as biomarkers for bringing objective diagnostic criteria into the most challenging diagnostic issue related to effective therapy for patients with oral cancer. Tongue cancer tends to metastasize from quite early stages; conversely, tumor size is not a critical factor in the outcome of death. Therefore, the establishment of a precise diagnosis that estimates the probability of latent distant metastasis, even at early stages of the disease, will enable clinicians to make appropriate decisions regarding the suitability of chemotherapy and incision areas. Such early-stage decisions are considered essential for improving the ability to control disease progression in patients with OSCC.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES