Level of expression of E-cadherin mRNA in colorectal cancer correlates with clinical outcome.

A series of colorectal carcinomas (n = 49) resected from patients with known clinical outcomes were analysed for E-cadherin expression using in situ hybridisation to measure mRNA. Patients surviving 5 years or longer (n = 31) exhibited significantly higher levels of E-cadherin mRNA than those surviving less than 5 years (n = 18, P = 0.003). These preliminary results from this small sample suggest that E-cadherin expression may be a useful prognostic marker in colorectal cancer patients. ImagesFigure 1

Colorectal cancer is one of the most common malignances in the developed world and remains a major publc health problem (King's Fund Forum, 1990). Pathological staging, using the Dukes claation, offers an accurate guide to the outcome of patients who have undergone surgicl resection (Deans et al., 1992(Deans et al., , 1993. However, in Dukes stage B category, many patients who putatively have had a curative resection will succumb to metastasis or tumour recurr. Identification of the lily prognostic outcome of any one individual within this broad division would be of great use but unfortunately remains impossible at this time. E-cadherin is a member of the large cadherin family of homophilic cell-cell adhesion molecules and is expressed in all epithelial tissues (Takeichi, 1990). E-cadherin moleules maintain intercellular connections and, importantly, participate in signalling and communication between neighbouring cells (Takeichi, 1990). Tbere is increasing evience that E-cadherin plays a significant role in neoplastic behaviour. For example, experimental studies have revealed that loss of this mokcule from epithelial cells is associated with the acquisition of the invasive phenotype (Vlminckx et al., 1991). There appears, moreover, to be a quantitative correlation between the level of E-cadherin expression and invasive ability in a range of cell lines derived from human carcinomas (Frixen et al., 1991). A large number of fresh human cancers also have been analysed for E-cadherin expression by immunohistochemistry (Takeichi, 1993), including clinical material denved from patients with colorectal cancer (Dorudi et al., 1993;Kinsella et al., 1993;Nigam et al., 1993).
Generally, these investigations have revealed an inverse relationship between E-cadherin expression and tumour grade, with poorly differentiated tumours exhibiting reduced or absent immunoreactivity (Takeichi, 1993). Although downregulation of E-cadherin has been observed in undifferentiated (Dorudi et al., 1993;Kinsella et al., 1993;Nigam et al., 1993) and advanced colorectal carcinomas (Dorudi et al., 1993), no study has yet examined the possible relationship between expression of this cell adhesion molecule and prognosis in colorectal cancer. Such a relationship, however, has been exaied in three other tumour types: bladder, head and neck and gastric cancer (Brnguier et al., 1993;Mattijsen et al., 1993;Mayer et al., 1993). Uniformly, these authors all reported that a reduced level of E-cadherin was correlated significntly with poor prognosis, but the studies all used frozen material for E-cadherin immunohistochemistry, tended not to have stage-matched material and had lmited follow-up data (of less than 5 years) (Bringuier et al., 1993;Mattijsen et al., 1993;Mayer et al., 1993).
Previously we showed that in colorectal cancer there is a good correlation between the presence of E-cadherin mRNA and protein (Dorudi et al., 1993). Therefore, by using in situ hybridisation of paraffin-embedded, archival material solely from Dukes stage B cancers, we have now been able to examine these relationships in tumours from patients with extended follow-up periods. Although the 5 year survival rate in this group is approximately 70%, identification of those patients with a poor prognosis is of significant clinical and epidemiological importance since 35% of all colorectal cancers are Dukes stage B (Morson, 1990). We show here that retrospective analysis of the relationship between Ecadherin expression and survival in a series of Dukes B colorectal carcnomas, resected from patients who had either survived in excess of 5 years or succumbed to recurrent or metastatic disease within 5 years of surgery reveals an association between poor prognosis and reduced levels of this adhesion molecule.

Patient selection
Formalin-fixed, paraffin-embedded tumours were obtained from the archival store in the Department of Pathology at St.
Mark's HospitaL London, UK. The selection of these tumours was made consecutively on a chronological basis as long as the patients from whom these tumours were rescted fufilled our criteria for this study. Thus, survivors (n = 31) were defined as patients who had been followed up regularly in out-patient cinic and were assessed clnially as diseasefree for a minimum of 5 years, while non-survivors (n = 18) had died as a result of their malignancy within this time. These  Hybndisation to the sections was performed essentially as described by Senior et al. (1988). The tumours were graded and evaluated for E-cadherin positivity using the following scoring system: 3 + denoted a strong homogenous signal, 2 + indicated a clear signal that was heterogenous, 1 + was a recognisable but weak signal, while 0 referred to negative tumours. Examples of tumours scoring 1 + and 3 + are shown in Figure 1. Normal epithelium was, wherever possible, used as an internal positive control (35 out of the 49 tumours examined). However, in every tumour the presence of intact mRNA in all tissue compartments was assessed by using a riboprobe, h"A-10, to detect A-actin mRNA (Ponte et al., 1983) and tumours which did not exhibit a strong signal for P-actin were excluded from the study. Tumours were assessed by one of us (AMH) on two different occasions separated by an interval of 2 weeks with a concordance of over 95%. This observer remained unaware of the clinical outcome of the patients from whom these tumours were derived.
E-cadherin score in relation to tumour grade. Interestingly, four of five (80%) poorly differentiated tumours scoring 3 + were resected from patients surviving more than 5 years. Kaplan-Meier survival curves were also constructed for each E-cadherin score, and these are shown in Figure 2. These data were analysed using the log-rank test and found to be significant (P = 0.001) with extended survival in patients whose tumour E-cadherin scores were high (2+ or 3+).
DisŨ nlike previous reports on the prognostic value of Ecadherin expression (Bringuier et al., 1993;Mattijssen et al., 1993;Mayer et al., 1993), this study has employed in situ hybridisation to measure E-cadherin mRNA rather than the assessment of protein by immunohistochemistry on frozen material. Numerous studies have examined E-cadherin

Results
The E-cadherin scores and grades of tumours in the group of 5 year survivors as compared with those in the non-survivors group are presented in Table I. There was a significant difference in the levels of E-cadherin mRNA between survivors and non-survivors (Fisher's exact test, P= 0.003).    immunoreactivity in human carcinomas, but only three were performed using paraffin-embedded material (Dorudi el al., 1993;Moll et al., 1993;Rasbridge et al., 1993); the remainder all used frozen tumours (see Takeichi 1993 for review). This has made it difficult to examine the majority of archival material. As stringent fixation requirements are necessary for consistent and reliable E-cadherin immunohistochemistry in paraffin-embedded tumours (Rasbridge et al., 1993), in situ hybnidisation constitutes a valuable technique for examining such material. A fixation method comprising 2.5% phenol formol saline was introduced at St. Mark's Hospital in 1987, and this allowed us to analyse tissue processed subsequent to this date using immunohistochemical staining. However, in order to obtain extended survival data we had to use archival material obtained before 1987 and were unable to achieve reactivity with antibodies. Since the presence of E-cadherin mRNA has been shown to be a reliable indicator of the presence of protein (Schipper et al., 1991;Dorudi et al., 1993) we used in situ hybridisation.
It may well be of interest that four of the five poorly differentiated tumours examined in this study which displayed E-cadherin scores of 3+ were derived from patients surviving in excess of 5 years. This is particularly true since high tumour grade is known to be associated with reduced survival in colorectal cancer (Deans et al., 1993), while poorly differentiated bowel carcinomas generally exhibit reduced or absent E-cadherin expression (Dorudi et al., 1993;Kinsella et al., 1993;Nigam et al., 1993).
It is apparent from our results that low or absent Ecadherin expression is associated with reduced survival in patients undergoing curative resections for Dukes stage B colorectal carcinomas. However, the numbers in this study are still small and the analysis retrospective. A large prospective study is necessary before it can be ascertained whether levels of E-cadherin expression in colorectal tumours can truly yield independent prognostic information. If these preliminary findings are confirmed in a prospective study, they would be of considerable clinical relevance as, at present, accurate prediction of the risk of recurrence or metastasis in patients with Dukes stage B cancers is not possible (Morson, 1990). Clearly, such information would be extremely valuable in designating these patients to protocols of adjuvant therapy and might aid in an understanding of the molecular nature of tumour spread.