Materials and methods

PATIENTS AND TISSUE SPECIMENS

Samples of formalin fixed paraffin wax embedded tumour tissue were examined from 233 oesophageal squamous cell carcinomas curatively removed without additional treatment between 1987 and 1996 at the National Cancer Center Hospital, Tokyo. All patients underwent cervical, mediastinal, and abdominal lymphadenectomy (three field dissection)23 for oesophagectomy, and were classified according to the tumour, node, metastases (TNM) classification24 based on the pathological reports. The pathological stage (pTNM) of each patient was defined at the time of operation. The patients included 57 of pTNM stage I, 108 of stage II, and 68 of stage III. The 211 men and 22 women had a median age of 61 (range 40–86) years. The median follow up period was 33.7 (range 2–117) months.

ANTIBODIES AND IMMUNOHISTOCHEMICAL ANALYSIS

Sections (3 μm) were cut on silane coated glass slides (Muto Pure Chemicals, Tokyo, Japan), deparaffinised in xylene, dehydrated through a graded series of ethanols, and washed in running water. They were then treated with Serotec target unmasking fluid (STUF; Serotec, Oxford, UK) for 10 minutes in a 600 W microwave oven and left to cool for 15 minutes. They were rinsed in two changes of deionised water and phosphate buffered saline, and preincubated with 20% normal rabbit serum in Tris buffered saline at 37°C for 60 minutes. After being washed with Tris buffered saline, separate sections were incubated with the primary monoclonal antibodies 2C5 (3.3 μg/ml; R&D Systems, Abingdon, Oxon, UK), 2F10 (10 μg/ml, R&D Systems), or M23.6.120 (20 μg/ml), which recognise epitopes of the CD44s, v6, and v2 portions respectively, in 1% normal rabbit serum diluted with Tris buffered saline at 4°C overnight in a wet box and then at 37°C for 10 minutes the next morning. After the primary antibody treatment, endogenous peroxidase activity was blocked with 0.3% H₂O₂ in methanol at 4°C for 10 minutes, and the sections were incubated with a 1:400 dilution of biotinylated anti-mouse IgG (Dako, Santa Barbara, California, USA). This was followed by incubation with horseradish peroxidase conjugated avidin-biotin complex (ABComplex; Dako) at room temperature for 60 minutes. Immunostaining was visualised with 3,3'-diaminobenzidine (Sigma, St Louis, Missouri, USA) for 20 minutes, and the reaction was stopped by washing with water. Finally, the sections were counterstained with Mayer's haematoxylin. Between all antibody incubations, they were washed three times for five minutes with Tris buffered saline on a shaking platform. Negative controls included sections treated with 1% normal rabbit serum alone in place of the primary antibody.

The status of CD44s and CD44v was assessed in a coded manner without knowledge of the clinical and pathological features of the case or the clinical outcome. Levels of CD44 protein expression were graded into four subgroups: (a) no detectable expression; (b) faint expression detected in <5% of tumour cells; (c) strong heterogeneous or localised expression in >5% of tumour cells, indicating immunopositive subpopulation; (d) strong homogeneous expression in >50% of tumour cells. For practical and statistical purposes, subgroups (a) and (b) were combined into the group designated “negative”, and subgroups (c) and (d) were combined into the group designated “positive”.

STATISTICAL ANALYSIS

Staining results were compared in all cases with clinicopathological features, including sex, age, pTNM classification, histological grading, vessel invasion, and survival. Each lesion was graded histologically according to the World Health Organisation classification.25 The correlation between CD44 peptide expression and clinicopathological indices was analysed by Fisher's exact test. We defined overall survival as survival from the date of surgery until 26 December 1996. The primary determinant in this study was survival time, as measured from the date of surgery until the time of the last follow up visit or death. Survival curves for 233 patients were constructed according to the Kaplan-Meier method,26and the differences between the curves were calculated. The p value was calculated by the log rank analysis. The prognostic significance of individual parameters in multiparametric analyses was determined by using Cox's proportional-hazards model.27 A p value of less than 0.05 was considered significant.

Results

IMMUNOHISTOCHEMICAL STAINING

All 233 oesophageal squamous cell carcinomas were stained with three monoclonal antibodies directed against the total CD44 (CD44v as well as CD44s), v6 (exon 11), and v2 (exon 7). Moreover, normal oesophageal mucosa adjacent to cancer tissue was examined for reactivity with the same antibodies. All normal squamous epithelium displayed uniform staining for total CD44, v6, and v2 in the prickle cell and basal cell layers (fig 1A–C). Normal mucosa adjacent to the cancer tissue provided positive internal controls for reliably assessing the presence or absence of CD44 immunoreactivity in the squamous cell carcinomas. Total CD44 and v6 were observed in 99% (231 of 233) and 97% (226 of 233) of the cancer specimens respectively (fig2A,B). The high immunoreactivity to total CD44 and v6 was not only shown in normal squamous epithelium but also in cancer tissue. In contrast, the expression of v2 in the cancer tissue was only 30% (70 of 233). The intensity of v2 staining was generally weaker than that of total CD44 and v6 staining. Reactivity with all the CD44 specific antibodies was localised mainly on the tumour cell membranes, and some cells showed positive staining in the cytoplasm of the cells, especially in the case of CD44 v2. Table 1 summarises the v2 staining status of the 233 patients.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Immunohistochemistry of the adjacent normal squamous epithelium. (A) All normal squamous epithelium displayed uniform staining for total CD44 in the prickle cell and basal cell layers. v6 (B) and v2 (C) were also observed in the prickle cell and the basal cell layers. Normal mucosa adjacent to the cancer tissue provided positive internal controls for reliably assessing the presence or absence of CD44 immunoreactivity in the squamous cell carcinomas. Original magnification × 200.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Immunohistochemistry of squamous cell carcinoma. Total CD44 (A) and v6 (B) were observed in 99% (231 of 233) and 97% (226 of 233) of the cancer specimens respectively. In contrast, the expression of v2 (C) in the cancer tissue was only 30% (70 of 233). The intensity of v2 staining was generally weaker than that of total CD44 and v6 staining. Original magnification × 400.

CORRELATION BETWEEN THE EXPRESSION OF V2 AND CLINICOPATHOLOGICAL INDICES

Table 1 gives the relevant clinical characteristics of the 233 patients whose tumours were analysed immunohistochemically. Neither sex nor age correlated with positivity for v2 (p = 0.96 and 0.71 respectively). Also, there was no significant correlation between the expression of v2 and other clinicopathological indices such as histological grading, vessel invasion, pT classification, pN classification, or pTNM stage. The median follow up period was 34.0 months for patients with v2 positive tumours and 34.9 months for those with v2 negative tumours (p = 0.93).

EXPRESSION OF V2 AND PROGNOSIS

The relation between the expression of v2 and survival rate was examined in 233 curatively resected oesophageal squamous cell carcinomas during a follow up period of 2–117 months (median 33.7 months). Figure 3 shows survival curves of patients, stratified according to v2 status. Patients whose tumours were v2 positive had a significantly better prognosis than those whose tumours were v2 negative (p = 0.031). The five year survival rate for patients with v2 positive tumours was 72.4%, whereas the rate was 43.8% for patients with v2 negative tumours. At the conclusion of the study, 51 of 70 patients with v2 positive tumours (73%) were alive, compared with 84 of 163 patients (52%) with v2 negative tumours (p = 0.004). In multivariate analysis with Cox's proportional hazards model, v2 status was an independent prognostic factor (relative risk of death associated with v2 negativity, 2.0; p = 0.008) in all patients studied. However, the hazard ratio was lower than that of pT classification (relative risk of death associated with pT2 and 3, 2.3; p = 0.003) or pN classification (relative risk of death associated with pN1, 2.6; p<0.001). Other indices such as age, sex, histological grading and vessel invasion were not significant independent indicators of prognosis.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Survival curves of all patients studied according to v2 status. Patients whose tumours were v2 positive had a significantly better prognosis than those whose tumours were v2 negative (log rank test; p = 0.031). The five year survival rate for patients with v2 positive tumours (median follow up period 34.0 months) was 72.4%, whereas it was 43.8% for patients with v2 negative tumours (median follow up period 34.9 months).

Figure 4 shows survival curves for patients without lymph node metastasis according to v2 status. As in the data from node negative patients studied, patients whose tumours were v2 positive had a significantly better prognosis than those whose tumours were v2 negative (p = 0.022). The five year survival rate for patients with v2 positive tumours (median follow up period 45.9 months) was 83.8%, whereas the rate was 65.1% for patients with v2 negative tumours (median follow up period 43.0 months). In contrast with the data from all patients including node positivity and negativity, the multivariate analysis with Cox's proportional hazards model showed that only v2 status was an independent prognostic factor (relative risk of death associated with v2 negativity, 4.7; p = 0.037) in the patients without lymph node metastasis. However, pT classification, histological grading, and vessel invasion were not significant prognostic indicators. Tables 2 and 3 give the maximum likelihood estimated from the multivariate analysis.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Survival curves of patients without lymph node metastasis according to v2 status. Patients whose tumours were v2 positive had a significantly better prognosis than those whose tumors were v2 negative (log rank test; p = 0.022). The five year survival rate for patients with v2 positive tumours (median follow up period 45.9 months) was 83.8%, whereas it was 65.1% for patients with v2 negative tumours (median follow up period 43.0 months).  

Discussion

Recent clinicopathological studies have disclosed that the expression of individual variant exons is associated with poorer prognosis in several adenocarcinomas—for example, that of exon 14 (v9) is increased in gastric adenocarcinoma,28 that of exon 11 (v6) in the colon, breast,29 30 and pancreas,22 31 32 and that of exon 7 (v2) in the pancreas21 and breast.22 On the other hand, several reports have provided data indicating that CD44 expression is associated with longer survival in neuroblastoma33 and that decreased expression of v6 correlates with decreased survival in laryngeal squamous carcinoma34 and tumours of squamocellular origin.35 It is difficult to interpret this large body of information and to compare the various studies. However, we can infer from many reports published previously that the promoter and splice control is strictly regulated in normal cells in an organ specific manner, whereas the strict promoter and splice control is lost or changed in many human cancers. To date, there have been no reports on correlations between clinicopathological indices and CD44v in oesophageal cancer. On the basis of this background, we studied the correlation between the expression of CD44v, especially v2, and survival of patients with squamous cell carcinoma.

In our study, all normal squamous epithelium displayed uniform staining of total CD44, v6, and v2 in the prickle cell and the basal cell layers. This result was consistent with other data for tumours of squamocellular origin.33-35 Whereas total CD44 and v6 were observed in 99% and 97% of the cancer specimens respectively, the expression of v2 in the cancer tissue was only 30%. Furthermore, the intensity of v2 immunoreactivity was generally weaker than that of total CD44 and v6. Also, no significant association was found between the expression of v2 and clinicopathological indices of patients, including all stages. Concerning prognosis, patients with v2 positive tumours had a longer survival time than those with v2 negative tumours (p = 0.031). This result suggested that downregulation of v2 was associated with a poor prognosis in patients with oesophageal squamous cell carcinoma. In multivariate results of all patients studied, whereas v2 status was an independent prognostic factor (relative risk of death associated with v2 negativity, 2.0; p = 0.008), the hazard ratio was lower than that of pT classification or pN classification. As expected, the pN classification was an important prognostic factor36-38 (relative risk of death associated with pN1, 2.6; p<0.001).

We then evaluated the association between v2 status and prognosis in the patients without lymph node metastasis, which was the most useful prognostic factor. The result showed not only that v2 positivity gave a significantly better prognosis (p = 0.022) but also that v2 negativity was predictive of mortality (relative risk of death associated with v2 negativity, 4.7; p = 0.037) in the group of node negative patients. Other indices such as pT classification, pTNM stage, histological grading, and vessel invasion were not significant prognostic indicators.

To predict future development of metastasis is of great importance, as patients in such categories may benefit from adjuvant chemotherapy and/or radiotherapy. However, there has been no clear prognostic marker for patients without lymph node metastasis. Our findings here therefore seem to be quite important, because v2 status may assist in determining whether adjuvant therapy should be given to patients with oesophageal squamous cell carcinoma at an early stage.