Twenty-two samples of normal colorectal mucosa (NM) were collected from 10 patients during colonoscopy, at least 2 samples from each patient. One sample from each patient was fixed in formalin and embedded in paraffin for immunohistochemistry; the others were frozen at −80°C. All patients had normal colonoscopy. Twenty-two MAs were also identified in 6 patients, and removed after operation for colorectal cancer on surgical specimens; the precise localization of the lesion was made staining the resected mucosa with a 0.1% methylene-blue solution in saline, and observing it under a dissecting microscope in order to take and process only the fields of the lesion, as previously described (16,19). All MAs, defined as dysplastic aberrant crypt foci at histology were referred to as MAs. Ten MAs were fixed in formalin and embedded in paraffin, the others were frozen at −80°C. Finally, 22 samples of colorectal cancer (K) were collected from 11 patients operated on for cancer, fixed in formalin or frozen at −80°C. All cases of cancer were Dukes’ C adenocarcinomas.

The patients enrolled in the present study, who underwent colonoscopy or surgical resection for colorectal cancer at the University Hospital of Modena, provided written informed consent to the study protocol, which was approved by the local Ethics Committee.

Ten samples of NM, 10 of MA, and 10 of K were fixed in a 10% formalin solution for 1 h. They were then dehydrated, embedded in paraffin wax, and sectioned (3-μm thick); the sections were placed on Super Frost Plus microscope slides (Menzel). Prior to immunohistochemistry, routine histology of all tissue samples was carried out after hematoxylin and eosin (H&E) staining of the sections. Slides were dried overnight at 37°C, dewaxed in 2 changes of fresh xylene, and rehydrated in a descending alcohol series. Antigen retrieval involved treatment with a protease (Pronase 1:20; DakoCytomation) for 7 min at 37°C. Prior to immunohistochemical staining, the sections were washed with PBS and blocked with 20% Swine Serum (Normal; DakoCytomation) in PBS for 30 min to reduce nonspecific antibody binding. Primary antibody was then applied at appropriate dilutions for 30 min at room temperature. Polyclonal rabbit anti-mouse BCL6 antibody (DakoCytomation) was used as primary antibody at a dilution of 1:200 in PBS. Endogenous peroxidase activity was blocked by treating the slides with 0.3% hydrogen peroxide in methanol for 12 min. After washing with PBS, biotinylated anti-mouse secondary antibody (1:20; DakoCytomation) was incubated for 30 min, and the slides were incubated with pre-diluted streptavidin-HRP conjugates for 30 min according to the manufacturer’s instructions. In order to reveal the color of the antibody staining, DAB substrate solution (freshly made immediately prior to use: 0.05% DAB-0.015% H₂O₂ in PBS) was used. Slices were counterstained with Mayer’s hematoxylin, dehydrated in ascending alcohol series and covered with a coverslip (Histovitrex mounting medium; Carlo Erba). For all these procedures, the negative control was obtained by staining samples without the primary antibody. Human B cell lymphomas were used as positive control.

One or 2 slices for each sample were scored as described below for immunofluorescence analysis. For each group of colorectal tissue and lesions, about 25 ×100 microscopic fields were scored. Moreover slices for each group of normal mucosa and lesions were scored by 2 independent observers and there was <5% variance between the results of the 2 counts. To quantify the staining area, the slices were processed with an ImageJ software. The images were converted into 16-bit images and then thresholded (the same value was used in all analyzed images). The resulting thresholded images were binary and they showed only the staining area. The total staining area was calculated by the software.

Immunofluorescence analysis was carried out to evaluate the expression of BCL6 protein. Six samples of NM, 6 of MA and 6 of K, were fixed in 4% paraformaldehyde in PBS, cryoprotected in 15% sucrose in PBS, and frozen in isopentane cooled in liquid nitrogen. Horizontal cryosections of the samples were cut (10-μm thick), and H&E staining was performed on sections to control tissue integrity. After a treatment with 3% BSA in PBS for 30 min at room temperature, the cryostatic sections were incubated with the primary antibodies (mouse anti-BCL6; DakoCytomation) diluted 1:25 in PBS containing 3% BSA for 1 h at room temperature. After washing in PBS, the samples were incubated for 1 h at room temperature with the secondary antibodies diluted 1:20 in PBS containing 3% BSA (sheep anti-mouse FITC-conjugated; Sigma). After washing in PBS and in H₂O, the samples were counterstained with 1 μg/ml DAPI in H₂O and then mounted with anti-fading medium (0.21 M DABCO and 90% glycerol in 0.02 M Tris, pH 8.0). Negative control samples were not incubated with the primary antibody. The confocal imaging was performed on a Leica TCS SP2 AOBS confocal laser scanning microscope. For DAPI and FITC double detection, samples were sequentially excited with the 405-nm/25-mW lines of a blue diode laser and the 488-nm/20-mW lines of the Argon laser. The emission signals from DAPI and FITC were detected by 2 photomultiplier tubes. Excitation and detection of the samples were carried out in sequential mode to avoid overlapping of signals. Sections were scanned with laser intensity, confocal aperture, gain and black-level setting kept constant for all samples. Optical sections were obtained at increments of 0.3-μm in the z-axis and digitized with a scanning mode format of 512×512 or 1,024×1,024 pixels and 256 grey levels. The confocal serial sections were processed with Leica LCS software to obtain 3-dimensional projections. Image rendering was performed by adobe Photoshop software.

The original green fluorescent confocal images were converted to grey-scale and median filtering was performed. An intensity value ranging from 0 (black) to 255 (white) was assigned to each pixel. Background fluorescence was subtracted and the immunofluorescence intensity was calculated as the average for each selected area. To quantify BCL6 expression, all blocks were completely sectioned and 3–4 slices were examined at a magnification of ×40 for each patient. Starting randomly, 3 microscopic fields for each section were used for sampling all sections within the unbiased sampling frame. The fluorescence intensity at the selected areas, linearly correlated with the number of pixels, was quantitatively analyzed using the standard imaging analysis software of the NIS-Elements system. To each sample was assigned a code number and the score, referred to as immunofluorescence intensity score, determined by an observer who was blind to tissue groups during analysis (20).

Whole cell lysates were obtained from 6 samples of NM, 6 of MA and 6 of K, extracted with hypotonic buffer (50 mM Tris-Cl, pH 7.8, containing 1% Nonidet P-40, 140 mM NaCl, 0.1% SDS, 0.1% Na deoxycholate, 1 mM Na₃VO₄ and freshly added protease inhibitor cocktail). Lysates were then cleared by centrifugation for 15 min in a refrigerated centrifuge (1,500 rpm) and immediately boiled in SDS sample buffer. Protein extract (40 μg) from each sample (NM, MA and K) was electrophoresed on SDS-PAGE and transferred to nitrocellulose membranes. The protocols of the western blotting were performed as previously described (20). The membranes were blocked with 3% dry milk and 2% BSA in PBS-T and were then incubated with mouse anti-BCL6 (DakoCytomation) antibody, diluted 1:1,000 overnight at 4°C under agitation. After washing, the membranes were incubated with secondary HPR-conjugated goat anti-mouse IgG antibody (1:10,000) for 30 min at room temperature. Immunoreactive proteins were detected with ECL (Amersham). Furthermore, the membranes were stripped and incubated with anti-mouse β-tubulin (Sigma) to control and correct for loading error. Densitometry analysis was performed using a Kodak Image Station 440CF system (Rochester, NY, USA).

All quantitative data for NM, MA and K are reported as means ± SD. The difference in BCL6 average expression in the different groups of colorectal lesions was tested for statistical significance using ANOVA one-way analysis, followed by Student-Newman-Keuls tests. P<0.05 was considered to indicate a statistically significant difference between groups.

The quality of staining obtained by the immunohistochemical method was generally very good; BCL6 staining was evident as brown spots both in the cytoplasm and in the nucleus of stromal cells, easily allowing their identification even at low magnification. Examples of BCL6 staining are shown in Fig. 1, in which panels A and B show low levels of BCL6-positive cells (fibroblast-like cells and leukocytes) in NM, especially in the stroma adjacent to the crypts. In all samples of normal mucosa, appreciable BCL6 staining of epithelial cells was not found. In MAs, BCL6 expression was higher compared to normal mucosa and it was very evident in the stromal compartment (Fig. 1C and D). Moreover, a consistent epithelial positivity was observed in MA and K: the protein aggregated in small clumps appears distributed in the cytoplasm at the lateral and basal portions of the epithelial cells and also partially in their nucleus (Fig. 1G). Concerning the stromal compartment, the samples of K were characterized by a higher number of BCL6-positive cells with respect to MA (Fig. 1E and F); in carcinoma, positive cells were not evenly distributed in the tissue, but tended to gather in some areas, usually at the edge of the neoplastic tissue (Fig. 1E), not only in the stroma inside the tumor (Fig. 1F), but also among the neoplastic cells at the margins of the lesions (Fig. 1H). Inside the epithelial-positive cells, BCL6 protein accumulated predominantly at the apical region of cells towards the lumen of colonic crypts, as well as at the nuclear level (Fig. 1H). The semi-quantitative evaluation of immunostaining intensity reported as immunohistochemistry intensity score in Fig 3A, showed that the level of BCL6 protein in both the epithelial and the stromal compartment had an increasing trend from NM to K, with statistical significance of the different expression among the groups. In order to obtain a better resolution of subcellular structures in very thick samples, the confocal analysis of BCL6 was performed in a subset of NM, MA and K samples (Fig. 2). Moreover, this technique allowed to define the precise distribution pattern and the quantification of BCL6 protein. The staining patterns of BCL6 varied from a diffuse mode to a granular one, both in the cytoplasm and in the nucleus of stromal cells, appearing consistent and unchanged in the same samples, without appreciable variations in intensity and localization among the cells. In all NM samples, a few scattered areas in the stroma compartment exhibited a moderate BCL6-reactivity, although the epithelial cells were not marked (Fig. 2A–C). The results of immunofluorescence analysis were in line with those of immunohistochemistry: intense staining was evident in MA, in stromal cells surrounding colonic crypts (Fig. 2D–F); moreover, a moderate expression appeared in epithelial cells, but only at the cytoplasmic level (Fig. 2G–I). BCL6-positive cells were observed throughout the resected tissue in all colon cancer lesions and they were particularly numerous within the central tumor stroma (Fig. 2J–L). Some cells (both the epithelial and stromal ones) showed a strong BCL6 immunoreactivity in the cytoplasm as well as in the nucleus inside some tumor areas (Fig. 2M–O); in particular, they appeared located deeply, inside the tumor lesion. These features were confirmed by the measurement of the BCL6 expression levels, both in the epithelial and in the stromal compartment, by means of immunohistochemistry and immunofluorescence, that showed statistically significant differences in the 3 different types of samples (Fig. 3A and B). These results were in agreement with those obtained by immunohistochemical analysis.

To confirm the findings of the morphological evaluation previously described, cell lysates of NM, MA and K were analyzed by western blotting. The protein profile of BCL6 clearly showed 1 single weak band at the expected molecular weight (98-kDa) in NM, an intense band in MA and a very intense band in K (Fig. 4A). Densitometric analysis and normalization (with equal amounts of protein loading) of the immunoreactivity signal from protein extracts of NM, MA and K showed that the level of BCL6 protein has an increasing trend from NM to K, with statistical significance of the different expressions (Fig. 4B).