Expression of the BCL-2 protein in normal and dysplastic bronchial epithelium and in lung carcinomas.

Although expression of the bcl-2 protein has been investigated in a number of non-haematological malignancies, little is known of its distribution in premalignant lesions. Expression of bcl-2 was investigated immunohistochemically in archival biopsies of normal (n = 8) and dysplastic bronchial epithelium (n = 56) and in 31 bronchial resection margins and their corresponding carcinomas. All dysplasias had lost the prominent basal staining pattern seen in histologically normal epithelium. Two were negative and six had occasional basal positive cells. In 37 cases up to 66% of the epithelial cells throughout the full epithelial thickness were bcl-2 positive with weak to moderate staining intensity. In 11 cases, all severe dysplasias, strong expression was observed in > 90% of the epithelial cells. Four patterns of bcl-2 expression in dysplasias were identified and an increasingly aberrant pattern of bcl-2 expression correlated with an increasing grade of dysplasia (Spearman's rank correlation, P < or = 0.0001). Sixty-five per cent of the carcinomas contained bcl-2-positive cells. Patients with non-small-cell lung carcinomas (n = 27) in which > 50% of the tumour cells were bcl-2 positive showed a survival advantage compared with those with 0-25% bcl-2-positive cells (P = 0.02). No correlation was found between p53 expression (Walker et al., 1994) and bcl-2 expression in dysplasias or carcinomas.

Lung cancer is the commonest cancer in the UK, accounting for one in six of all new cancer cases (Cancer Research Campaign, 1992). Most patients present with already advanced diseas, and the prognosis remains poor despite improvements in clinical treatment (Roth, 1992;Souhani, 1992;Gazdar, 1994). Conventional screening studies for early detection have had little effect on overall survival (Frost, 1986;Tockman et al., 1992), possibly because lesions that are clinically informative may have already progressed to disseminated disease (Gazdar, 1994).
Lung carcinomas arise after a series of morphological and genetic changes within the bronchial epithelium, and it may take years to progress from normal epithelium to invasive cancer. The morphological changes are thought to progress from hyperplasia, to metaplasia/dysplasia, to carcinoma in situ and finally to invasive and metastatic cancer (Gazdar, 1994;Lee, 1992). Improvements in the treatment of this disease which may prolong survival rely on early recognition of the molcular changes at a time of absent or minimal histopathological change before the acquisition of invasiveness (Gazdar, 1994).
Moleular and cytogenetic changes have been described in preinvasive bronchial ksions (Sundaresan et al., 1992) as well as in normal epithelium adjacent to lung cancers (Lee et al., 1987(Lee et al., , 1992Sozzi et al., 1991). Abormal expression of the p53 protein has been reported in dysplastic bronchial epithelium and in normal bronchial epithelium of cancer patients (Bennett et al., 1993;Nuorva et al., 1993;Hirano et al., 1994;Walker et al., 1994), suggesting that aberration in p53 function may be a very early process in the development of lung cancers. Such studies prompt further investigations to establish the nature of the changes in dysplastic epithelial cells which may be involved in the development of malignant potential.
The bcl-2 gene codes for a 26 kDa protein with lipophilic character and no substantial homology with any other protooncogene products (Cleary et al., 1986;Tsujimoto and Croce, 1986;Hockenbery et al., 1990) and may contribute to malig-nancy by preventing programmed cell death or apoptosis (Hockenbery et al., 1990;Jacobson et al., 1993;Kerr et al., 1993). This proto-oncogene was first described as a result of the chromosomal translocation t(14; 18) seen in a laige number of follicular B-cell lnes and the majority of malignant human follicular B-ell lymphomas (Tsujimoto et al., 1985;Korsmeyer, 1992). In this translocation the bcl-2 gene on chromosome 18 becomes juxtaposed with the IgH gene on chromosome 14, resulting in overexpression of the bcl-2 protein, and conferring affected lymphocytes with reistance to apoptosis (Ckary et al., 1986;Hockenbery et al., 1990). High levels of bcl-2 expression prevent cell death from a wide variety of cell stresses and cytotoxic cheicals, including growth factor depletion, heat shock, ionising radiation, excess calcium influx and a range of chemotherapeutic drugs (Tsujimoto, 1989;Sentmen et al., 1991;Miyashita and Reid, 1992;Lotem and Sachs, 1993).
In oncogenesis, deregulation of bcl-2 expression may contribute to the accumulation of oncogenic mutations by suppressing the apoptotic deletion of cells that normally follows the induction of DNA damag (Kerr et al., 1993). Pathological expression of bcl-2 has so far been investigated mainly in haematologicl malignancies (Pezzella et al., 1990;1991;Korsmeyer, 1992;Piris et al., 1994), but only in a few epithelial or neural tumours (Castle et al., 1993;Leek et al., 1994;Pilotti et al., 1994;Ramini and Lu, 1994;Segal et al., 1994;Silvestrini et al., 1994). The bcl-2 protein is expressed in some small-cell lung cancer (SCLC) cell lines (Ikegaki et al., 1994) and in 28% of non-small-cell lung cancers (NSCLC) (Pezzella et al., 1993). In NSCLC, bc)-2 positivity is associated with better prognosis (Pella et al., 1993). Although it is present in a number of lung carcinomas, bcl-2 is absent in differentiated cells of normal bronchial epithelium (Pezzella et al., 1993). Pezzella et al. (1993) suggest that the presence of bcl-2 in differentiated cells may be an indicator of malignancy, but before drawing such conclusions some knowledge of the expression of this molecule in premalignant lesions would be useful.
In this study, we have investigated immunohistochemically the expression of the bcl-2 protein in normal and dysplastic bronchial epithelium and in lung carcinomas. Comparison has been made with the expression of the p53 protein in these tissues, detrmined previously (Walker et al., 1994). paraffin-embedded bronchial biopsies and four resction margins, taken from patients who did not have lung cancer at the time of removal and which contained epithelium reported as histologically normal, were also obtained from the files.
Thirty-one formalin-fixed, paraffin-embedded specimens of lung carcinoma and their corresponding bronchial resection margins were colleted prospectively by Dr N Pendleton from lobectomies or pneumonectomies performed at the Cardiothoracic Centre, Liverpool, UK. Patients received no other form of therapy either before or after surgery and were staged using UICC guidelines. Full clinical data were available for these cases.

Imunnohistochemistry
Bcl-2 immunoreactivity was determined using methods similar to those described in Walker (1994), except that microwave antigen retrieval was essential. Sections were microwaved in 10 mM citrate buffer pH 6.0 for 20 min using a 650 W microwave oven at full power before staining. A monoclonal antibody to bcl-2 (clone 124, Dako) was used at 1:40. Negative controls using normal rabbit serum at 1:400 or Tris-buffered saline (TBS) in place of the primary antibody and lymphoid tissue as positive control were included in each staining run. Lymphocytes in each section acted as additional internal positive controls. Sections were reviewed for bcl-2 positivity and the intensity of stained cells scored as negative, weak, moderate or strong. Weak staining was defined as that which was only apparent at high magnifiction ( x 400), while moderate and strong staining was visible at all magniiations. In normal and dysplastic epithelium, the proportions of bcl-2-positive epithelial cells and their distribution according to the thickness of the epithelium containing these cells was recorded. Dysplasias were classified into four categories (A-D) ( Table  1) according to the number and distribution of positive cells and their staining intensity. Inter-observer variability (x) for this classification was 0.87 (95% confidence interval 0.77-0.97). In tumours, the distribution of stained tumour cells across the sections was noted and the percentage of positive cells assessed independently by two pathologists.
Immunoreactivity to p53 in cases that had not previously been investigated was determined as described in Walker et al. (1994) using the CMI antibody (Novaastra).

Statistical analysis
The signifa of associations were determined using the Fisher-Irwin exact probability test or the chi-squared test. Spearman's rank correlation was used to compare the severity of dysplasia with bcl-2 staining patterns. Survival &il2. in .M --PepNbum *u C Wakr et a 165 analysis was by the log-rank test. Two-tailed probabilities are quoted for all statistical tests.

Relts
Normal epithelium AU bronchial resection margins from lung cancer patients (30/30), bronchial biopsies (8/8) and resected bronchial tissues from non-cancer patients (4/4) showed a similar pattern of immunohistochemical staining in histologically normal epithelium when stained with a monoclonal antibody (clone-124) to the bcl-2 protein AU cases examined showed long stretches of normal epithelium positive for bcl-2, although the intensity of stain varied between cases and within sections; in a few cases areas of epithelium negative for bcl-2 were present In bcl-2-positive regions of normal epithelium basal cells were stained usually with a moderate to strong intensity, while the more differentiated cells were negative, resulting in a prominent basal staining pattern (Figure 1). The intracellular distribution of this stain was cytoplasmic, with many cells showing perinuclear membranous sining. In some cases occasional brush border cells stained intensely with similar intracellular distribution; sometimes these cells were clearly ciliated.
Bronchial glands stained cytoplamically with variable intensity, some being strongly stained. Perinuclear membranous stain was occasionally noted in glandular epithelial cells. In all tissues scored, lymphocytes were strongly stained.
Dysplastic epithelium Fifty-six bronchial biopsies with dysplastic epithelium were investigated for expression of the bcl-2 protein. None showed the pattern of staining typical of histologically normal epitheliuim AU had lost the prominent basal layer of stained cells seen in normal epithelium. In two cases dysplastic epithelium was completely negative for bcl-2, although lymphocytes in the sections were positive. In 28 cases the intensity of the stain in bcl-2-positive cells was weak. In the lemaining 26 cases the intensity was similar to or increased compared with bcl-2-positive normal epithelium. In all positive dysplasias the proportion and distribution of positive cells was asd. In six cases only a few positive very weakly stained cells were evident in basal locations. In 37 cases bcl-2-positive cells of weak or moderate intensity were found in suprabasal locations and in many cases extended throughout the fuIl thickness of the epithelium. In these cases positive cells were present in varying proportions, from focal positivity to up to approximately two-thirds of the epithelial cells. In 11 cases strong expression was observed in the majority of the epithelial cells and throughout the full thickness. In dyspWlaias all bcl-2-positive cells showed cytoplasmic stain; some showed perinuclear membranous stain. Occasional positive nuclei were observed both in cells in mitosis and in cells not under going mitosis.
Based on the number and distribution of positive cells in the epithelium and their staining intety, four patterns of bcl-2 expression were identified (Table I and Figure 2). Bcl-2 and severity of dysplasia Dysplasias of all histological grades showed bcl-2 staining patterns A, B and C, while pattern D was only found in severe dysplasia (Table II). Severe dysplasias with the most pleomorphic cells showed the strongest stain. Comparing mild and moderate dysplasias with severe dysplasias, pattern A was more often found in the mild/moderate group (P = 0.042), while pattern D was only found in the severe  (Table II). An increasingly aberrant pattern (from A to D) correlated with an increasing grade of dysplasia (Spearman's rank correlation coefficient of 0.49, P < 0.0001). Thus, staining tended to be more prominent, with a greater proportion of bcl-2-positive cells in the upper layers of the epithelium with increasing severity of dysplasia.
In the system of categorisation given in Table I, groups B and C differed only in the ining intensity of bcl-2-positive cells. Even if groups B and C were combined in the Spearman rankl correlation analysis a trend towards the more aberrant patterns of bcl-2 expression was still obtained as severity of dysplasia increased (correlation coefficient of 0.46, P = 0.0002).
Bcl-2 and p53 expression in dysplastic epitheliwn For many of the samples of dysplastic epithelium examined in this study, expression of p53 had already been investigated (Walker et at., 1994). Where sufficient tissue was available, new cases were also eamined for p53 expression using the CMI antibody. There was no correlation betweii p53 expression and bcl-2 expression.

Bronchial carcinomas
As described above, the expression of the bcl-2 protein was examined in the histologically normal epithelium in bronchial margins from lung carcinoma resections. In the corresponding carcinomas, 20/31 tumours and 16/27 NSCLCs contained bcl-2-positive cells (Table Ill). Eklven out of 31 tumours were completey negative for bcl-2, even though lymphoid tiss and, in many sections, adjacent normal epithelium stained strongly. A further 14 tumours had bcl-2negative areas within the tumour. In some bcl-2-positive tumours, staining was focal and patchy, although in positive areas stain was evident in all tumour cells. In others bcl-2 was expressed more intensely at the periphery of tumour islands. Six tumours showed intense stain throughout the entire tumour. In the remaining positive tumours intensity varied between the tumours and within the tumour sections from weak to strong. In tumours stain was cytoplasmic and in some cases perinuckar membranous. In some tumours occasional clls, either in mitosis or with pleomorphic nucki, showed intense nuclar staining for bcl-2.
Although the number of tumours investigated was small, full clinical information was available for these cases. Bcl-2 expresson was found in all histological types (Table EII). No correlation was found between the degree of differentiation, the UICC stage or TMN score and bcl-2 expression.
A scatter diagam for the percentage bcl-2-positive tumour cells in the 27 NSCLCs eamined in this study is shown in Figure 3. In survival analysis for the NSCLC tumours in this series, the tumour that had 1% tumour cells positive for bd-2 was included in either the negative group or the positive group with no significnt difference to the results obtained. No sigificnt survival advantage was shown for patients with tumours which were negative for bcl-2 compared with those with bcl-2-postie tumours (ogrank test 0-1% vs 20-100%, chi-squared 3.43, P = 0.06). In contrast, patients in the group whose tumours had 50-100% bcl-2-positive tumour cells had a significntly longr survival than those with 0-25% bcl-2-positive tumour cells (Figure 4)  Mild + moderate vs severe: A vs B + C + D, P = 0.042* (Fisher's exact test); A + B wsC + D, P = 0.0040 (chi-squared test); A + B + C vs D, P = 0.004* (FLsber's exact test). *Two-tailed probabilities. Bcl-2 and p53 expression in bronchial carcinomas Immunohistochemial expression of the p53 protein in these tumours had been determined previously (Walker et al., 1994). No correlation was obtained between bcl-2 expression and p53 expression.

Dbcw_o
Although expresson of the bcl-2 proten has so far been investigated in many haematological malignancies and in a number of solid tumours, few studies have eamined the distribution of this protein in premalignant lesions.
The bcl-2 protein has been detected by immunohis tochemical procedures in a limited number of non-lymphoid tisus under different physiological conditions: in long-lived, post-mitotic cells (neurones), complex organised epithelia (skin and gastric intestinal mucosa) and in glandular epithelium under hormonal control and growth factor control (Hockenbery et al., 1991;McDonnel et al., 1992). In these tisus in which apoptosis accounts for cell turnover, bcl-2 is topographially restricted to the long-lived progenitor cells that renew lineages and selected post-mitotic cells that require an extended lifespan (Hockenbery et al., 1991). In preous studies bronchial epithelial mucosa has been reported either not to express bcl-2 (Hockenbery et al., 1991) or to show positivity in basal cells, with the more differentiated cells being negative (Lu et al., 1993;Pezzella et al., 1993). We have found that ls with their nucei in the basal layer of histologically normal bronchial epithelium express bcl-2, resulting in a predominantly basal saining pattern. However, although the majority of differentiated cells were negative, in some cases occasional well-diferentiated columnar brush border cells also expressed bl-2. lbe intacellular distribution of the stain in these occasional differentiated bronchial epithelial cells and the complance with the predkied pattern of bcl-2 expression in all other cell types in these sections and control lymphoid tissue suggests that these cls do ideed express bcl-2. Furthermore, the monoclonal antibody clone 124 used in this study is reported to have satisfatory specificity and has been used in a number of other studies of both fresh and archival material (Pezzella et al., 1990, Lauwers et al., 1994Leek et al., 1994;Pilotti et al., 1994;Silvestrini et al., 1994;Ramini and Lu, 1994). These bcl-2-positive differentiated cells were seen in the histologically normal epithelium from cancer and non-cancer patients and are therefore unlikely to reflect an early event in the transformation to malignancy. Glandular epithelia such as breast and thyroid (Hockenbery et al., 1991;Pilotti et a!., 1994), and in this study bronchial mucosal glands, express bc!-2, thus bcl-2 positivity m these differentiated bronchial epithehal cells may be related to secretory function.
lThe bd-2 protein is an integral membrane protein and is usually a d with a cytoplamic location, being present in mitochondrial membranes, nuclear outer membranes and endoplasmic rtulum (Korsmeyer, 1992;Akao et al., 1994;Lithgow et al., 1994). In this study, occasional cells with bdl-2-postive nucli were noted in bronchial dysplasias and carcinomas. In support of this, in some epithelial cell lines the bcl-2 protein has been shown to have occurred transiently in mitotic nuclei (Wilingham and Bhalla 1994;Lu et al., 1994), tng that the anti-apoptotic function of bcl-2 extends into mitosis.
Expression of the bcl-2 protein was altered in all bronchial epithelial dysplasias examined, suggesting that deregulation of bdl-2 expression occurs concomitantly with the histological disorganisation that accompanies dysplasia. It is possible that the loss of the prominently stained basal layer seen in histologically normal epithelium and the appearance of bcl-2positive cells in the upper layers of the epithelium in dysplastic lesions results from the change from a pseudostratifled columnar to a stratified epithelium in which growth control has become aberrant. It is likely that initial changes in bcl-2 expression seen in all dysplasias is secondary to other genetic events which deregulate growth. As bcl-2 is associated in normal cells with protection from apoptosis (Korsmeyer, 1992;Kerr et al., 1994), bcl-2-positive cells in the upper layers of the epithelium may evade stringently controlled normal differentiation and apoptosis and have a growth advantage compared with bcl-2-negative cells. While staining patterns A, B and C were found in all histological grades of dysplasia, pattern D was only observed in some of the severe dysplasias. The change or changes resulting in this pattern of bcl-2 expression may therefore be a relatively late, but not obligatory event in the progression to invasive neoplasia.
This expression pattern may result from a specific genetic lesion in either the bcl-2 gene or in a gene which controls bcl-2 expression, occurring relatively late in the development of malignancy. t(14; 18) translocations which cause overexpression of bcl-2 in some lymphomas (Korsmeyer, 1992) have not been reported in lung cancers, but immunohistochemical overexpresson of bcl-2 has been reported in folicular lymphomas in the absence of this translocation (Pezzella et al., 1990).
Aberrant expression of the bcl-2 protein has been reported in other premalignant lesions. In gastric epithelial dysplasias, alterations in the spatial distribution of bcl-2-positive cells were observed, with bcl-2-positive cells being present in extended regions of the epithelium (Lauwers et al., 1994). In premalignant keratinocytic tumours, bcl-2 was found to be expressed in 73% of tumours due to Bowen's disease and in 25% of cases of actinic keratosis, in contrast to surrounding bcl-2-negative keratinocytes (Nakagawa et al., 1994).
In the initial design of this study, a limited number of bronchial tumours were examined to determine whether we obtained a similar distribution of bcl-2-positive cells in tumour sections to that reported in the literature for other solid tumours. The bcl-2 protein was found to be distributed similarly but present in a higher percentage of the bronchial carcinomas examined in this study than reported by Pezzella et al. (1993); however, our tumour numbers were much lower. Because clnical information was available for these cases, analysis of the chnicopathological data was carried out, although the conclusions drawn are limited by the small number of cases examined. Bcl-2 was expressed in tumours of all grades of histological differentiation. The only positive correlation obtained was for survival, and this only when survival data for patients with tumours with 0-25% bcl-2positive tumour cells were compared with those for patients with tumours with >50% bcl-2-positive cells. Few studies on the prognostic sign of bcl-2 in NSCLC are as yet published. In the study by Pezzella et al. (1993) a survival advantage for patients with bcl-2-positive tumours was found, particularly for patients with squamous cell carcinomas and in those who were over 60 years of age. In this study the bcl-2-positive group was compared with the bcl-2negative group (Pezzella et al., 1993). It is notable that Silvestrini et al. (1994), in their study of the prognostic signii of bcl-2 expression in breast carcinomas, compared survival data for patients with more than 30% bcl-2positive cells with those for patients with lower positivity.
The relationship of bcl-2 expression to prognosis and survival in non-haematological malignancies is currently poorly defined. In some studies, bcl-2 expression is associated with a survival advantage (NSCLC, Pezzella et al., 1993;breast, Silvestrini et al., 1994) and markers of good prognosis (breast, Leek et al., 1994), while in others it has no effect (neuroblastoma, Ramini and Lu, 1994) or is related to poor prognosis (neuroblastoma, Castle et al., 1993). While it can be argued that retention of bcl-2 expression in tumours may protect against apoptosis and lead to abnormal accumulation of malignant cells (Korsmeyer, 1992;Kerr et at., 1994), overexpression of bcl-2 has been found to lead to growth inhibition in some cultured cancer cell lines (Pietenpol et al., 1994). These observations suggest that further investigation of bcl-2 expression in a larger series of NSCLCs than hitherto reported would be justified to assess the clinical utility of this marker.
Both the bcl-2 and p53 genes are involved in the genetic control of apoptosis (Haldar et al., 1994;Miyashita et al., 1994a). Recent research has shown that there is a negative response element in the bcl-2 gene through which p53 may directly or indirectly transcriptionally down-regulate the bcl-2 gene (Miyashita et al., 1994b). In breast and some thyroid carcinomas and non-Hodgkin's lymphoma, an inverse correlation between the immunohistochemical expression of p53 and bcl-2 has been demonstrated (Pezzella et al., 1992;Silvestrim et al., 1994;Leek et al., 1994;Pilotti et al., 1994). In contrast, no correlation between p53 and bcl-2 expression was observed in this study in either bronchial dysplasias or carcinomas.
In this study, we have shown that abnormal expression of the bcl-2 protein is present in bronchial dysplasias of all histological grades, suggesting that changes in the expression of this molecule arise early in the transformation of normal to dysplastic epithelium. Further alterations in its expression may occur late in the progression to malignancy. Before expression of this protein could be used as a biomarker for malignancy, further investigation of the expression of this gene and its biological and prognostic significance in lung cancer is essential.