Sixty patients with previously untreated squamous cell carcinoma of the tongue and who were seen at Korea University Medical Center were included in this study from 1997 to 2004. All the patients underwent primary surgery with the intent to cure. The patients received postoperative adjuvant radiation therapy in a multidisciplinary setting. Forty-one patients received radiation therapy after surgery (68.3%) and 19 patients received surgery alone (31.7%). This study was approved by the Institutional Review Board of Korea University Medical Center.

Immunohistochemistry for Glut-1 was performed using consecutive 4µm serial sections of formalin-fixed, paraffin-embedded tissues that were placed onto positively charged glass slides with using a single-staining procedure. Two observers reviewed the immunostained slides. All of the interpretations of the immunohistochemistry were done without the pathologist having any knowledge of the patients' clinical outcomes. The polyclonal rabbit anti-human Glut-1 antibody (A3536, DakoCytomation, Copenhagen, Denmark) was used to determine the expression of Glut-1. For antigen retrieval, the sections were microwaved at 600 W in 0.01 M citrate buffer (pH 6.0, Labvision, UK) twice for 5 minutes each time. After blocking with 10% human serum, the slides were incubated with the primary antibody (1 : 200) for 1 hour at room temperature. Biotinylated linked HRP and streptabidin-HRP (K0675, DakoCytomation, Copenhagen, Denmark) were each applied for 10 min, respectively. After the color was developed with using DAB solution and the samples were counterstained with hematoxylin, the tumor cells were considered positive for the Glut-1 antigen if they displayed cytoplasmic staining. After the slides were scanned at low magnification (×40), four areas (per case) of maximum Glut-1 expression were selected. The degree of positive Glut-1 staining on the section was assessed at high magnification (×200) based on the semi-quantitative scale of 0 to 3. A positive Glut-1 expression was designated as those samples with membranous and/or cytoplasmic staining with an intensity of grade 3, and the percentage of tumor cells staining for Glut-1 was measured at low magnification (×40). The mean value of the examined fields was the final value.

For the detection of Ki-67, we performed immunostaining with the anti-Ki-67 mouse monoclonal antibody (clone MIB-1, DakoCytomation, CA, USA) based on the above-mentioned procedure. After the slides were scanned at low magnification (×40), the cells with positively stained nuclei were then counted at high magnification (×400). The percentage of the Ki-67 expression was quantified by determining the number of positive cells expressing nuclear Ki-67 among the total number of tumor cells per high power field. Two to three hundred tumor cells were counted and the Ki-67 index was calculated as the percentage of Ki-67 positive cells per positive and negative cells.

To assess the microvessel density, the anti-CD31 mouse monoclonal antibody clone JC70A (Dako, Copenhagen, Denmark) was used to stain the microvessels. Primary antibody (1 : 20 dilution) was applied at room temperature for 60 min, and the slides were stained by the avidin-biotin method. After the slides were scanned at low magnification (×40), four areas (per case) of high vascularization were chosen for counting the microvessels. The microvessel density (MVD) was expressed as the number of vessels per field. The mean value from four fields was recorded as the MVD for each tumor.

Fisher's exact test was applied to assess the association between the categorical variables. The coefficient of correlation( r) between the expressions of the markers was calculated by using Spearman's rank test. The disease free survival (DFS) and overall survival (OS) curves were calculated using the Kaplan-Meier method and they were compared by the log-rank test. The OS was determined from the date of surgery to the date of death or the last follow-up visit, and DFS was determined to the date of disease relapse or the date of the last follow-up visit. The Cox proportional hazards regression model was used for multivariate analyses. All the statistical analyses were performed using a statistical software package (SPSS, Version 10.0, Inc., Chicago, IL.). Statistical significance was defined as p values less than 0.05. All p values were two-sided.

Analysis of the Glut-1 expression included the following parameters: the staining intensity of tumor cells on a grading scale of 0 to 3 and the percentage of positively stained tumor cells. A positive Glut-1 expression was designated as those samples with membranous and/or cytoplasmic staining with an intensity of grade 3 (Fig. 1) so as to exclude false positive cases such as nonspecific or vaguely immunostained cases. The distribution of the Glut-1 expression was diffuse with a membranous pattern (Fig. 1). The median percentage of Glut-1 positive tumor cells was 60.0% (range: 0.0~90.0%). High and low Glut-1 expressions were defined according to the median value (high ≥60.0% versus low <60.0%).

The clinical and pathological characteristics of the patients enrolled in this study are summarized in Table 1. The follow-up duration was from 4.10 months to 117.13 months. The median age of the patients was 54.9 years (range: 20.0~79.4 years). 30 patients relapsed during follow-up and 22 of these patients died. There were 19 cancer-related deaths, but there were three deaths not related to cancer: 1 case of acute myocardial infarction, 1 case of subarachnoid hemorrhage and 1 case of pneumonia. 30 patients were alive without evidence of disease relapse and eight patients were alive with disease. The correlation of the Glut-1 expression with the clinical and pathological characteristics is summarized in Table 2. A high Glut-1 expression (≥60.0%, median value) was more frequently observed in tongue base squamous cell carcinoma, which is known to have a poorer prognosis than the other sites of the tongue such as the anterior or lateral tongue. Lymph node metastasis status was also associated with a high Glut-1 expression (p=0.010) although the T stage was not related with the Glut-1 expression. The number of patients with a high Glut-1 expression was significantly greater for the patients with stage III/IV disease than for the patients with stage I/II disease. However, other characteristics, including gender, age, the smoking status and the degree of differentiation, were not associated with the Glut-1 expression (Table 2). The patients treated with adjuvant radiotherapy had high Glut-1 and Ki-67 expressions (p=0.006). Since radiotherapy was administered to those patients with a high probability of relapse or a high tumor burden, these findings suggest that the Glut-1 expression might be related with an aggressive tumor phenotype.

To assess the relationships of the Glut-1 expression with cellular proliferation, the percentage of Glut-1-positive tumor cells were compared with the Ki-67 expression by using Spearman's rank test. The Glut-1 expression showed a correlation with the Ki-67 expression (r=0.406, p=0.001, Fig. 2A). The Glut-1 expression was also correlated with microvessel density, although its significance was weak (r=0.267, p=0.039). To examine the spatial relationship of Glut-1 with proliferation and angiogenesis, we compared the expressions of both proteins in the same field. We found the positive Ki-67 expression in the same region as the Glut-1 expression (Fig. 3). Microvessels represented by CD31 staining were observed in the interstitial area (Fig. 3). While the Glut-1 expression was diffuse and wide spread over the tumor areas, the Ki-67 expression was dominantly observed at the periphery of tumor nests that were close to microvessels. Therefore, these findings suggest the presence of hypoxia might stimulate angiogenesis and maintain proliferation of tumor cells in the hypoxic compartment of the tumors.

Among the 30 cases with relapse, 27 were loco-regional and three were at distant sites: lung and bone metastasis. Tumor relapse was not significantly related with the Glut-1 expression (Table 2). On the univariate analysis, the patients with a high Glut-1 expression were significantly associated with poorer overall survival as illustrated by the Kaplan-Meier curves (Fig. 4A). However, the disease-free survival (p>0.05, Fig. 4B) was not significantly different among the compared. Multivariate analysis was performed to examine the independent prognostic significance of several variables, as compared with the established clinical staging; these variables included the Glut-1 expression and the other clinical and pathological characteristics such as age, stage, location of the primary site, the presence of tumor cells in the resection margin and the treatment modality. The expression of Glut-1 failed to display independent prognostic significance for survival (p>0.05).