Argyrophilic nucleolar organizer regions and bromodeoxyuridine and 3[H]‐thymidine labelling indices in colorectal cancer

The count of argyrophilic nucleolar organizer regions (AgNORs) has been proposed as a useful method for evaluating cell replication in human tumours. The current study was undertaken to compare AgNOR values in colorectal cancers with two better established methods for investigating cell proliferation such as bromodeoxyuridine (BrdUrd) and 3[H]‐thymidine (3[H]dT) labelling indices (LIs).

AgNOR proteins and the proliferative activity of cells has been suggested (Trere' et al. 1989, Vagner-Capodano el a/. 1987, particularly in myogenic stromal tumours of the stomach and in cervical intraepithelial neoplasias (Sinn et a. 1989, Thickett er al. 1989). Accordingly, in several tissues, the increase in the area occupied by AgNORs has been shown to correlate positively with elevated Bromodeoxyuridine (BrdUrd) labelling index (Leek et al. 199 1,Trere' et al. 199 1 ). Recently Ruschoff et al. (1990) proposed that the mean area per silver-stained NOR particle is a good way to provide reproducibility of AgNOR value. The mean area of the NOR particles is inversely related with AgNOR number and higher NOR counts with smaller mean NOR particle areas characterize tumours with unfavourable outcome (Chomette & Auriol 199 1 ).
In the present study we compared mean AgNOR values (number, area and arca per silverstained NOR particle) per nucleus of 40 colorectal canccrs with two established parameters of cell proliferation, namely >[HJdT and BrdUrd labelling indices (Lls). To ensure the reliability and reproducibility of AgNOR evaluation we carried o u t a control study by indcpcndcntly repeating the same measurements (number, area and area per silver-stained NOR particle) in two centres with different operators and different computer-assisted image analysers. Then we investigated the relationship between AgNOR values, BrdUrd L1, '[H]dT LI and clinicopathological variables of the tumours. Our results indicate that AgNORs do not correlate with either 3[H]dT or BrdUrd Lls, suggesting that the evaluation of AgNORs cannot be a reliable estimatc of proliferative activity in colorectal cancers.

M A T E R I A L S A N D M E T H O D S Patients
We examined 40 volunteers (20 females and 20 males) with colorectal cancer. The mean age was 66.6 years, (range 41-83). 14 tumours were from the right colon (i.e. proximal to the splenic flexure: 11 from the caecum and ascending colon, one from hepatic flexure and two from the splenic flexure), 13 from the left colon (two from the descending and 1 1 from the sigmoid) and 13 from the rectum or rectosigmoid junction. Tumours were staged according to the Dukes stage (Dukes 1932) in stage A ( n = 2), stage B ( n = 20), stage C ( t i = 14), stage D ( I I = 4). Two patterns of tumour growth were defined: expanding (12 = 29), when the tumour showed a well delineated and circumscribed border of growth, and infiltrating ( n = 1 l), when single tumour cells or clusters of cells spread through and beyond the bowel wall (Jass et , Ponz de Leon rt al. 1992. Tumours were graded according to Dukes & Bussey ( 1958) as well ( I ? = 9). modcratcly ( I I = 28) and poorly ( I I = 3) differentiated.
Bromodeoxyuridine labelling 40 patients were intravenously given 500 mg of 5'-bromo-2'-deoxyuridine (BrdUrd) (Sigma, St Louis MO, USA) sterilized by filtration, dissolved in 100 ml isotonic solution, 2-6 h before surgery. Immediately after surgery, wedge-shaped samples of colonic neoplasia were taken, fixed in 70% ethanol overnight and embedded in paraffin. Four pm sections were prepared. The immunohistochemical detection of nuclei that had incorporated BrdUrd was carried out using the avidin-biotin peroxidase complex method (Vectastain ABC Kit PK-4002, Vector, Burlingame, CA, USA) and anti-BrdUrd monoclonal antibody (Dakopatts, Glostrup, Denmark) diluted I :SO. One thousand or more cells were scored for each carcinoma. BrdUrd-positive cells were easily identified at histology because the nuclei stained dark-brown against a clear background. BrdUrd LI was defined as the fraction of BrdUrd labelled nuclei in the total scored tumour nuclei.

3[H]dT-thymidine labelling
In 20 of the 40 cases, small fragments of neoplastic tissue were incubated in a shakerbath for 60 min at 37°C immediately after surgery. The incubation medium was composed of 2 ml Eagle's minimal essential medium supplemented with 10% FCS and L-glutamine [GIBCO, Grand Island, NY, USA], and 222 kBq or 6 pCi/mmol of '[H]dT (specific activity of 925 GBq or 25 Ci/mmol, Amersham, Aylesbury, UK). After fixation in Bouin's solution for 1 h and in 80% ethanol overnight, the fragments were embedded in paraffin. Four pm sections were dewaxed, rehydrated and coated with Ilford F-5 photographic emulsion. The slides were exposed for 3-6 days at 4"C, developed in D19b Kodak solution and stained with haematoxylin. The cells accepted as positively labelled were those with at least 6 black grains per nucleus against a clear background. "[H]dT L1 was defined as the fraction of labelled nuclei in the total neoplastic nuclei in each microscopic field at x 1000 magnification with an oil immersion lens.

AgNOR staining
Tumour specimens parallel to the wedge-shaped samples used for BrdUrd labelling were routinely fixed in 10% buffered formalin and embedded in paraffin. Sections, 2 p m thick, were dewaxed and post-fixed for 30 min in a 3: 1 ethanol acetic acid solution. Then, the sections were rehydrated and incubated in the dark for 20 min at 37°C in a medium made up of one volume of 2%) gelatin in 1% aqueous formic acid and two volumes of 50O10 silver nitrate. Quantitative evaluation of intherphase AgNORs was performed using an automated image analyser. The slides were scanned by the operator under the microscope at x 400 magnification. The selected image was captured into digital memory and visualized on the monitor. Here, the silver-stained NORs appeared as well-defined, dark grey structures on a lighter background. The operator defined the grey threshold which permitted the selective identification of dots corresponding to AgNORs, by interactively comparing the video-image with the original microscopic image. When the structures defined by the chosen grey threshold were superimposable on the silver-stained dots in the original field the measurement was performed. By means of this morphometric technique, the interphase AgNOR number and area (in square microns) were evaluated.
For each case a total of 100 randomly selected nuclei of neoplastic cells were evaluated and thc mean AgNOR number and area per nucleus were calculated from the data stored on a file. The mean area per silver-stained NOR particle was also calculated by dividing AgNOR area per AgNOR number in each nucleus (Ruschoff ef d. 1990).
To evaluate the reliability and the reproducibility of AgNOR measurements, number and area were independently assessed in two centres with two different computer-assisted image analysers-system 1: ASEM Pc IBM compatible with SISTEMA MONO (Immagini e Computer, Milano); system 2: COMMODORE-AMIGA 3000 with BIOIMAGE ANALYSIS (Eurosoft, International, Italy).
Student's t-test and linear regression were peformed for statistical analysis.

RESULTS
The mean number of AgNOR per nucleus in the 40 cases of colorectal carcinomas ranged from a minimum of 1.35 (SD k 0.33) to a maximum of 6.48 (SD k 1.08) with a medium value of 2.3 1 in system 1 and from a minimum of 1.70 (SD + 0.42) to a maximum of 5.60 (SD + 0.99) with a median value of 2.7 in system 2. Analogously, the mean area of AgNOR proteins per nucleus varied from a minimum of 2.33 ,urn2 (SD f 0.78) to a maximum of 7.32 ,urn2 (SD k 1.79) with a median of 4.65 ,urn2 in system 1 and from a minimum of 4.02 pm2 (SD f 1. 2 1)  of 5.15 in system 2. As shown in Figure 1 the results of AgNOR evluation obtained with two analyzing systems were reproducible and comparable ( r = 0.75 P < 0.001 for number; r = 0.62, P < 0.00 1 for area). Also, the range of mean area per NOR particle did not differ in the two systems (0.62 to 4.26 pm? with a median of 1.79 pm2 for system 1 and 0.99 to 3.56 pm2 with a median of 2.05 ,urn2 in system 2) and the values were significantly correlated ( r = 0.63, I ' < 0.00 1 ).

Figure 2. Correlation between AgNOR values (a number/nucleus and h area/nucleus) and BrdUrd Labelling Index (system 1).
In the 40 specimens analysed, the BrdUrd LI ranged from 6.0 to 35.4% No significant differences were found in the values of BrdUrd LI in relation to the different time intervals which elapsed between BrdUrd administration and surgery. No significant rclationships were detected between AgNOR values (number and area/nucleus) and BrdUrd LI (Figure 2). Analogously, no significant correlation could be found when AgNOR values were compared with '[H]dT LI (Figure 3), which ranged from 7.0 to 23.4 in the 20 cases studied. In addition, no relationships were observed when BrdUrd Ll or '[H]dT LI were correlatcd with mean area per silver-stained NOR particle.
As previously reported (Roncucci et 01. 1992), the two Lls did show a significant correlation with clinical and pathological parameters, like pattern of growth, grade and stage of the tumour. Analogously, no correlations were found between AgNOR values and the same parameters.

DISCUSSION
The aim of this study was to establish whether AgNOR values could relate with parameters of cell proliferation like BrdUrd and " [H]dT LIs in colorectal carcinoma. The results failed to demonstrate significant correlations of AgNOR with the rate of cell proliferation. A significant correlation instead was seen between BrdUrd and "H]dT LIs. Thymidine and its analogue BrdUrd are incorporated into the nuclei during the S phase, thus providing reliable tools for evaluating tissue proliferative activity. However, both molecules require living cells to be incorporated and cannot be applied to tissues already processed for routine histology. On the contrary, AgNORs have the advantage of being visualized in paraffinembedded material, but counting methods are tedious and time consuming. Furthermore, variations in section thickness and interobserver variability are known to influence the accuracy and reproducibility of counting.
For these reasons, the reliability of AgNOR values was assured in our study by repeating twice the measurements with two independent image-analyser systems. The results obtained were rather comparable, with coefficients of variation in the order of 0.6-0.7 ( P <0.001), and emphasized the lack of correlation with BrdUrd and '[HldT Lls.
The quantity of AgNORs has been repeatedly shown to directly relate with cell replication rate (Trere et al. 1989, Hittmair et ul. 1994  Moreover, the AgNOR number correlated with DNA index and S phase determined by flow cytometry in breast lesions (Gin er al. 1989) and non-Hodgkin lymphoma (Crocker e t a / . 1988). Recently, AgNOR size has been shown to correlate closely with proliferative indices investigated by in vivo injection of BrdUrd in various normal and neoplastic rodent tissues (Leek et al. 199 1 ) and by in vitro BrdUrd incubation in several human malignant tumours of different origin (Trere el al. 199 1 ). In a study o n 12 established cell lines derived from different tumour types including five colorectal carcinomas, a direct relationship has resulted between AgNOR protein quantity and growth rate of neoplastic cells evaluated by ' [H]dT labelling independently of the cell line origin (Derenzini er al. 1990a).
However, other studies have failed to find positive correlations between the quantity of AgNORs proteins and the cell proliferation rate investigated by Ki-67 or flow cytometry in lung and breast carcinomas (Soomro & Whimster 1990, Hahn et al. 1990) and also in rectal carcinomas (Griffiths et al. 1989, Kram et al. 1989. Analogously, our results excluded significant correlation between values of AgNOR and parameters of cell proliferation in colorectal cancer. Moreover, in contrast to previous studies, the search for correlation was not only restricted to the study of AgNOR number, but also involved AgNOR area and area of single silver-stained NOR particle. In fact, as recently demonstrated by Hittmair et al. (1994), the distribution and quantity of AgNORs are strongly influenced by the phase of the cell cycle, the AgNOR number being strictly related to mitotic index and AgNOR area to S phase fraction.
However, in our study, none of the AgNOR parameters investigated showed correlation with proliferation indices. Therefore, AgNORs may be either related to other parameters of cell replication such as PCNA, Ki-67, Alpha Polymerase or, (probably better), to phases of increased ribosomal production and peptide synthesis not necessarily linked to cell proliferation as suggested by Griffiths el al. (1993). Indeed, Mourad et al. ( 1 991, 1994) showed that the mean number of AgNORs is a reflection of the total chromosome count or ploidy rather than or proliferative activity. AgNOR counts are the result of the spatial arrangement of AgNOR-carrying chromosomes in different phases of the cell cycle.
In conclusion, as AgNOR values do not appear to relate to the more direct paramcters of cell kinetics, it follows that their significance as a biomarker of cell proliferation remains questionable, at least in colorectal neoplasms.

A C K N O W L E D G E M E N T S
The authors are grateful to Mrs Romana lotti and Elisabetta Mattioli for technical assistance, and t o Miss Joanne Stenton for the English revision of the manuscript. The work was supported in part by funds of the Ministero dell 'Universita' e della Ricerca Scientifica (quota 40%).