Cell adhesion molecules in bladder cancer: soluble serum E-cadherin correlates with predictors of recurrence.

Sera from 40 patients with newly diagnosed bladder cancer (28 superficial tumours (pTa and pT1) and 12 muscle-invasive tumours) were assessed by enzyme-linked immunosorbent assay (ELISA) to determine the concentrations of soluble E-cadherin (sE-cadherin), soluble E-selectin (sE-selectin), soluble vascular cell adhesion molecule-1 (sVCAM-1) and soluble intercellular adhesion molecule-1 (sICAM-1). Corresponding frozen sections of primary tumour were analysed for E-cadherin expression using the monoclonal antibody, HECD-1 and standard immunohistochemistry. Patients with bladder cancer had significantly higher concentrations of sE-cadherin compared with a control group (P = 0.017). No difference was found between the two groups with regard to sE-selection (P = 0.403), sVCAM-1 (P = 0.942) and sICAM-1 (P = 0.092). High levels of sE-cadherin were related to poor histological grade (P = 0.009), number of superficial tumours at presentation (P = 0.008) and a positive 3 month check cytoscopy in superficial disease (P = 0.036). Abnormal E-cadherin expression was associated with increasing tumour stage (P = 0.009) and grade (P = 0.03). There was no correlation between high levels of soluble E-cadherin in sera and abnormal E-cadherin expression by the tumour (P = 0.077). Elevated levels of sE-cadherin are found in sera of patients with bladder cancer and correlate with known prognostic factors.

Of newly diagnosed superficial bladder tumours, approximately 30% are multifocal at presentation (Lutzeyer et al., 1982), 60-70% will recur (Greene et al., 1973) and 10-20% will undergo stage progression (Lutzeyer et al., 1982). Intense debate surrounds the exact aetiology of synchronous and metachronous bladder cancer. Mucosal abnormalities, namely dysplasia and carcinoma in situ, adjacent and distal to the primary tumour, are associated with a poor prognosis, in terms of tumour recurrence and progression in superficial disease (Althausen et al., 1976;Smith et al., 1986). In contrast, recent reports, based on molecular biological studies (Sidransky et al., 1992;Lunec et al., 1992) have added a new dimension suggesting a common clonal origin for concomitant urothelial tumours, at least in some cases. Lateral intraepithelial spread of transformed cells from the origin of a bladder carcinoma or dispersal of tumour cells are possible mechanisms underlying multifocal disease. In support of these theories, a single instillation of intravesical mitomycin C (Tolley et al., 1988) or epirubicin (Oosterlinck et al., 1993) immediately following transurethral resection has been reported to increase time to first recurrence.
Decreased intercellular adhesiveness favours detachment of tumour cells (Takeichi, 1991) and this may play a role in multifocality, the development of recurrent superficial tumours and progression to metastatic disease. At least four families of cell adhesion molecules are thought to be involved in cell-cell adhesion (cadherins, selectins, immunoglobulins and integrins). The most widely studied have been Ecadherin, a cell-surface glycoprotein restricted to epithelial tissue (Behrens et al., 1985;Shimoyama et al., 1989), involved in calcium-dependent homotypic cellcell adhesion, and vascular cell adhesion molecules [E-selectin, vascular cell adhesion molecule 1; (VCAM-1) and intercellular adhesion molecule 1 (ICAM-1)].
In normal urothelial cells, E-cadherin is homogenously expressed at all cell-cell borders (Bringuier et al., 1993) but not by the cell border in contact with the basement membrane. Loss of expression of E-cadherin on cell membranes has been reported in patients with high-grade, muscle-invasive bladder tumours (Bringuier et al., 1993;Otto et al., 1994;Lipponen and Eskelinen, 1995) and high-grade prostate cancers (Umbas et al., 1992), in which abnormal immunohistochemical patterns including cytoplasmic staining, heterogenous cell membrane expression and negative staining have been reported. Abnormal expression of Ecadherin is associated with reduced recurrence-free survival (Lipponen et al., 1995), along with higher progression rates in patients with superficial bladder tumours (Otto et al., 1994) and also reduced survival in patients with muscle-invasive tumours (Bringuier et al., 1993). It has been suggested that reduced E-cadherin expression, associated with an increase in autocrine motility factor receptor (gp78) expression identifies a group of patients with superficial bladder cancer at high risk of stage progression and earlier cancer-related death (Otto et al., 1994). Using bladder carcinoma cell lines and primary murine urothelium as a model for tumour recurrence, E-cadherin appears to be an important determinant of the pattern of intraepithelial expansion and the ability of tumour cells to colonise primary urothelium (Rebel et al., 1994).
Soluble forms of E-cadherin (sE-cadherin) have been detected in the urine of healthy individuals and patients with bladder cancer (Banks et al., 1995). Soluble E-cadherin fragments have also been found in the sera of patients with gastric and hepatocellular carcinomas (Katayama et al., 1994). The molecular weight of the main form of sEcadherin detected in urine, 80 kDa, was consistent with that found in serum in the latter study. Furthermore, an 80 kDa fragment is unlikely to represent a product filtered from the  (Lauri et al., 1991) and VCAM-1 in malignant melanoma (Rice and Bevilacqua, 1989). ICAM-1 is associated with disease progression in a group of patients with malignant melanoma (Natali et al., 1990). Elevated circulating soluble variants of E-selectin, VCAM-1 and ICAM-1 have been found in a variety of cancer patients (Banks et al., 1993). Elevated soluble ICAM-1 is also associated with reduced survival in patients with malignant melanoma (Harning et al., 1991), and in patients with liver metastases in gastric, colonic, gallbladder and pancreatic cancers (Tsujisaki et al., 1991). Finally, a recent study has linked soluble VCAM-1 and soluble E-selectin with the formation of new blood vessels (angiogenesis), which in turn may promote new tumour growth (Koch et al., 1995).
The primary objective of this study was to determine the levels of soluble forms of E-cadherin (sE-cadherin), E-selectin (sE-selectin), VCAM-1 (sVCAM-1) and ICAM-1 (sICAM-1) in sera of patients with newly diagnosed transitional cell carcinoma of the bladder. Our first aim was to assess if any of these factors correlated with multifocal synchronous or metachronous tumours in patients with superficial disease or the presence of metastases at presentation in those patients with muscle-invasive disease. A further objective was to compare levels of sE-cadherin in sera with expression of Ecadherin by the primary bladder tumours assessed by immunohistochemistry.

Materials and methods Patients
Samples of venous blood were collected aseptically from 40 patients (36 males and four females) with newly diagnosed transitional cell carcinoma of the bladder before cystoscopy and transurethral resection of tumour (TURT). Patient controls consisted of 13 age-[median age=65 years (95% confidence interval) (CI)= 58 years to 74 years] and sex-(eight males, five females) matched patients undergoing elective surgery for non-malignant conditions including peptic ulceration, diverticular disease and inguinal herniae. Bladder tumours were staged using the TNM system (Union Internationale Contre le Cancer, 1992). Seventeen patients had Ta tumours, 11 had Tl tumours and 12 patients had muscle-invasive (T2 -T4) tumours. Four patients with superficial tumours and three patients with muscle-invasive tumours had concomitant carcinoma in situ. Tumours were graded using the WHO system (Mostofi, 1974). There were five GI, 16 G2 and 19 G3 tumours. The median age of the patients was 71 years (CI = 67 years to 74 years). Three of 28 (11%) patients with superficial disease were given intravesical chemotherapy following transurethral resection: one patient was given a single dose of intravesical mitomycin C immediately after TURT; the other two patients underwent a course of six instillations of intravesical epirubicin at weekly intervals. In patients with superficial disease, the number of tumours at presentation and the outcome of the 3 month check cystoscopy were noted. In patients with muscle-invasive disease, the presence of metastases were determined by computerised tomography (CT) scanning, chest radiography and the operative findings at cystectomy.

Preparation of venous blood
Venous blood samples were immediately centrifuged at 400 g for 10 min, allowing separation of serum, which was stored at -20°C. Before analysis, samples were slowly thawed and gently mixed. ELISA Levels of circulating sE-cadherin (Takara Schuzo Co., Kyoto, Japan), sE-selectin, sVCAM-1 and sICAM-1 (R&D Systems, Abingdon, Oxon, UK) were measured by commerically available ELISA kits. The immunoassay is based on simultaneous reaction of the soluble adhesion molecule in the sample to two anti-human adhesion molecule antibodies, one antibody coated to the walls of the microtitre wells, the other conjugated to the enzyme horseradish peroxidase. The reaction between peroxidase and substrate results in colour development, intensities of which are proportional to soluble adhesion molecule levels. Monoclonal antibodies, HECD-1 and SHE13-6 were used in the sE-cadherin immunoassay. Their affinities for both the intact 120 kDa form of E-cadherin on epithelial surfaces and the soluble 80 kDa fragment of E-cadherin in serum have been confirmed by Western blot analysis (Katayama et al., 1994).
Following addition of stop solution, the optical density of each well was determined immediately by ELISA plate reader (MR 5000, Dynatech, West Sussex) using dual wavelength correction (A = 450 nm, A = 630 nm for correction). Corrected concentrations of unknown sera were calculated automatically from standard curves. E-cadherin immunohistochemistry Samples of bladder tumours were collected at the time of TURT, and immediately stored at -80°C. Frozen sections of tumour corresponding to serum analysis were available for 31 patients (22 superficial and nine muscle-invasive tumours). Positive controls consisted of biopsies of normal urothelium from three patients with no evidence of urothelial tumour and one normal colonic biopsy. For each run, negative controls were prepared by staining duplicate sections of tumour using the methods described below, but omitting the primary antibody.
Frozen sections were cut at 5 ,um, air dried and fixed in acetone for 10 min. Sections were rinsed in 5 mM Tris buffered saline (TBS) pH7.6 for 5 min and covered with normal rabbit serum (NRS) for 10 min. Excess serum was removed and primary anti-E-cadherin monoclonal antibody, HECD-l (R&D Systems) was applied at a dilution of 1:400 for 30 min at room temperature. After further washing in TBS, sections were covered with biotinylated rabbit antimouse immunoglobulin diluted 1:500 in NRS for 30 min. Following incubation in secondary antiserum, sections were again rinsed in TBS and covered with tertiary antiserum consisting of a streptavidinbiotinperoxidase complex (1 MI of streptavidin, 1 4ul of biotinylated horseradish peroxidase and 100 jul of TBS) for 30 min. The reaction was developed in diaminobenzidine (BDH, Poole, Dorset, UK) and sections were then counterstained with haematoxylin (BDH).
A tumour was defined as having normal E-cadherin expression if tumour cells displayed positive staining along cell-cell borders. All other staining patterns were classified as abnormal.
Statistical analysis Statistical analysis was performed using Minitab for Windows (Release 9) software. Data were presented as medians with 95% confidence intervals (Wilcoxon confidence interval of the median). Distributions were compared using the Mann -Whitney U-test. In certain cases, the Mann -Whitney confidence interval for the difference between two medians was quoted to illustrate the practical significance. Proportions were compared using Fisher's exact test. In the above analyses, a P-value of < 0.05 was considered statistically significant.

Bladder cancer vs control group patients
The median concentration of sE-cadherin in serum of patients with bladder cancer was 3955 ng ml-' compared with a median concentration of sE-cadherin in control patients of 1013 ng ml-' (P = 0.017; CI = 161 -3976 ng ml'-( Figure 1). No significant difference was found between patients with bladder cancer and controls with regard to sEselectin (P=0.403), sVCAM-1 (P=0.942) or sICAM-1 (P = 0.092).
Tumour bulk Of 13 patients with solitary superficial tumours, four patients had tumours greater than 5 cm in diameter. There was no significant difference in sE-cadherin concentrations in patients with tumours greater or less than 5 cm in diameter (P= 0.9385; CI = 2099 -1369 ng ml-'). Similarly, the number of multiple superficial tumours was not related to sE-cadherin concentration (>three tumours vs two or three tumours, P=0.871; CI= -6454 to 6607 ng ml-').
Outcome of 3 month check cystoscopy Three patients with superficial bladder cancer were not assessed cystoscopically at 3 months following diagnosis owing to poor general health. One patient with T1G3 disease had a cystectomy. Seven of 24 patients (29%) with superficial bladder cancer had a positive 3 month check cystoscopy. Elevated levels of sE-cadherin at presentation were positively correlated with a positive 3 month cystoscopy (P= 0.036; CI=65-7374 ng ml-'). The median concentration of sEcadherin in patients with a recurrence at 3 months was 7805 ng ml-' (CI=3973-9198 ng ml-l), compared with a median concentration of 1828 ng ml-' (CI = 1082-4564 ng ml-') for those patients disease free at 3 months.
sE-cadherin and metastatic disease Eight of twelve patients with muscle-invasive tumours had no evidence of metastatic disease on CT scanning. Three of the eight patients were later treated by cystectomy, three underwent radiotherapy and two were treated by transurethral resection alone. One of the three patients treated by cystectomy was found to have metastatic lymph node involvment at the time of surgery. A systematic assessment for metastases was not performed in the remaining four patients as they were considered too frail for radical treatment. Preliminary data suggest that patients with muscle-invasive disease but no evidence of metastases at presentation have a wide range of concentrations of sEcadherin in sera (median=2588 ng ml-'; CI= 1010-6737 ng ml-').

E-cadherin immunohistochemistry
In normal urothelium and normal colon, E-cadherin was expressed at cell-cell borders as expected. Of 31 tumours, 17 (55%) showed a similar staining pattern to normal urothelium. Abnormal E-cadherin expression was detected in the remaining 14 tumours (heterogenous membrane staining in 13 and perinuclear staining in one). Correlation of staining patterns with tumour stage and grade is shown in Table II. Abnormal E-cadherin staining pattern correlated with tumour stage (Ta vs TI and muscle invasion; Fisher's exact test, P=0.009). The three well-differentiated tumours were found to have normal E-cadherin expression. In contrast, 14 of 28 (50%) moderately and poorly differentiated tumours displayed abnormal E-cadherin expression (urothelial controls/GI vs G2/G3, Fisher's exact test, P= 0.03). Abnormal E-cadherin staining did not correlate with elevated sE-cadherin in serum (P=0.077). Abnormal E-cadherin expression was not related to the number of superficial tumours at presentation (Fisher's exact test, P = 0.350), or the presence of tumour recurrence at the 3 month check cystoscopy (Fisher's exact test, P = 1.000). Of seven patients with non-metastatic muscle invasive disease, three had heterogeneous E-cadherin expression. Conversely, one patient with evidence of perivesical lymphatic metastatic disease, detected at cystectomy, had normal E-cadherin staining.

Discussion
This is the first report of elevated sE-cadherin in the sera of patients with bladder cancer. In this study, all patients were newly diagnosed in comparison with a previous study reporting significantly elevated levels of sVCAM-l and sICAM-l in six patients with bladder cancer (Banks et al., 1993), who were at various stages of therapy, making interpretation difficult. Also, it is known that interleukin 2 leads to induction of sICAM-1 in sera of patients with malignant melanoma (Becker et al., 1992).
Multiple tumours at presentation and the presence of tumours at 3 months are associated with shorter recurrencefree intervals and higher recurrence rates respectively in multivariate analysis (Parmar et al., 1989;Reading et al., 1995). Parmar has suggested that patients with the longest tumour-free interval and the lowest recurrence rates were those who have solitary superficial tumours at presentation that do not recur at 3 months. Eleven of 24 (46%) patients in this study were in this category and had lower concentrations of sEcadherin at presentation (P = 0.011; CI = 690-7460 ng ml-') compared with the other 13 patients with superficial bladder cancer. Patients with the lowest risk of recurrence, as defined by Parmar, had a median sE-cadherin concentration of 1462 ng ml-I at presentation compared with a median sEcadherin concentration of 5648 ng ml-' for the remainder (Figure 4). There was no significant difference between these two groups with regard to the concentrations of sE-selectin (P = 0.118), sVCAM-1 (P = 0.543) or sICAM-1 (P= 0.954). Several lines of evidence indicate that E-cadherin plays a role in regulating the cohesiveness of epithelial tissues (Takeichi, 1991). Regulation of the expression of E-cadherin on tumour cells is unclear. In prostate cancer, a correlation has been shown between chromosome 16q deletions and abnormal expression of E-cadherin (Umbas et al., 1992). E-cadherin may, therefore, behave as a classical tumour-suppressor gene. Deletion of one allele may be accompanied by inactivating point mutation in the corresponding allele. Post-translational modification of the protein product may also affect function. It is known that three molecules (a, ,B and y catenins) form bridges between the cytoplasmic tail of E-cadherin and the cytoskeleton that may be necessary for E-cadherin to function normally (Nagafuchi and Takeichi, 1988).
Our findings are consistent with a previous study showing that abnormal E-cadherin expression correlates with high stage, although, in our series, abnormal E-cadherin expression was commonly detected not only in muscle-invasive disease (Bringuier et al., 1993;Otto et al., 1994) but also in TI disease.
In this study, abnormal E-cadherin staining was not accompanied by high serum levels of sE-cadherin. Conversely, in some patients, homogeneous expression of Ecadherin was found despite elevated levels of sE-cadherin in serum. It is possible, therefore, that levels of sE-cadherin in serum reflect the rate of turnover of E-cadherin whereas Ecadherin expression also reflects genetic abnormalities. ) 4000 ' 2000 : E cn 0 Group 1 Group 2 Group 3 (n= 11) (n= 7) (n= 6) Superficial bladder tumours Figure 4 Dot plot comparing concentrations of soluble Ecadherin in sera of patients with superficial bladder cancer with low, intermediate and high risk for tumour recurrence. Group 1 patients (low risk), solitary tumour at presentation, no recurrence at 3 months; group 2 patients (intermediate risk), either (a) solitary tumour at presentation, tumour recurrence at 3 months or (b) multiple tumours at presentation, no recurrence at 3 months. Group 3 patients (high risk), multiple tumours at presentation, tumour recurrence at 3 months.
Our results show that elevated sE-cadherin is found in patients who are at risk of developing early tumour recurrence following TURT, and may prove to be a useful serum marker. Levels of sE-selectin, sVCAM-1 and sICAM-1 in sera appear to have less significance in bladder cancer. It is possible that sE-cadherin reflects some aspect of the pathogenesis of multifocal tumours and development of tumour recurrence. Interestingly, it has been demonstrated that sE-cadherin itself interferes with cell -cell adhesion (Wheelock et al., 1987) in cultured epithelial cells. This implies that sE-cadherin may be directly involved in the adhesive process, rather than merely being a degradation product of intact E-cadherin. It is also conceivable that sEcadherin detected in serum is derived from cellular Ecadherin given that the transmembrane form of E-cadherin (a 120 kDa glycoprotein) can be degraded to a 80 kDa soluble product in vitro (Takeichi, 1991;Wheelock et al., 1987) and that sE-cadherin has also been detected in urine (Banks et al., 1995). Further studies are needed to determine if elevated sE-cadherin in sera is associated with high recurrence rates in superficial bladder cancer and to determine its relationship with progression to muscle invasion, the development of metastatic disease and reduced survival.
The precise relationship between abnormal expression of E-cadherin and genetic aberrations such as chromosome 16q deletion needs to be evaluated in bladder tumours as well as elucidation of factors regulating the degradation of the intact trans-membrane form of E-cadherin to the soluble form.