Expression of interleukin 6 (IL-6) correlates with oestrogen receptor in human breast carcinoma

Multifunctional cytokines play important and only partially defined roles in mammary tumour development and progression. Normal human mammary epithelial cells constitutively produce interleukin 6 (IL-6), IL-8 and a non-secreted form of tumour necrosis factor. Transformation of mammary epithelial cells by different oncogenes is frequently associated with alterations of cytokine/growth factor production and responsiveness. In the present study we analysed the expression of IL-6 in 149 cases of invasive breast carcinoma and the data have been correlated with clinico-pathological variables including tumour size, histological grade, nodal status, and oestrogen and progesterone receptors, Ki67 and p53, protein expression. Though the majority of breast carcinomas expressed at least low levels of immunoreactive IL-6, we found that expression of this cytokine was inversely associated with histological tumour grade (P = 0.0017), but not with tumour size and nodal status. Ki67 positivity was inversely correlated with IL-6 expression (P = 0.027). Among biological parameters analysed, a direct association was found between the percentage of IL-6-positive cells and that of oestrogen (P = 0.00005) and progesterone (P = 0.025) receptor-positive cells. No correlation was observed between IL-6 and p53 protein expression. These data indicate that down-regulation of IL-6 is associated with highly malignant mammary carcinomas. It will be of interest to evaluate whether alterations of cytokines that are constitutively produced by mammary cells are also associated with high-grade tumours. © 1999 Cancer Research Campaign


IL-6 and p53 protein expression
After fixation in 10% formalin, processing and paraffin embedding, 3-to 5 µm-thick sections were prepared. Sections were dewaxed and dehydrated, and the endogenous peroxidase activity was blocked by incubation for 15 min with 0.3% hydrogen peroxide in methanol. Sections were stained with two different anti-IL-6 antibodies (rabbit polyclonal antibody and IL-6 mouse IgG1 monoclonal antibody, Genzyme Co., Cambridge, MA, USA; product codes LP-716 and 1618-01), p53 mouse IgG2b monoclonal antibody (DAKO A/S, Denmark; clone DO7). Antibody dilutions (1:100 or 1:500) were incubated overnight; immunoreactivity was revealed with the avidin-biotin immunoperoxidase complex technique (Vector Elite ABC kit; Vector Laboratories, Burlingame, CA, USA). Sections were scored independently by two pathologists (GF and FB) using the following method: (a) the intensity of staining for each antigen was scored from 0 to 3 (0, absent; 1, weak; 2, moderate; and 3, strong); (b) the proportion of malignant cells positively stained was scored from 0 to 4 (0, no positive cells; 1, Ͻ 10% positive cells; 2, 11-50%; 3, 51-75%; and 4, 76-100%). The two scores were then added to yield the total score (0-7). Immunoreactivity was defined as negative when the score was 0, weak when it was 1-3, and strong when it was Ͼ 4. Tumours with scores ranging from 2 to 7 were considered as positive. Appropriate controls were included in every experiment. Previous studies (Basolo et al, 1993a) had shown that the sensitivity of immunostaining for IL-6 with the two antibodies used in this study was high. We have now shown that cytoplasmic positivity is maintained even in cultured cells releasing as little as 100-200 pg of IL-6 per 10 6 cells. As specificity controls, we showed that immunostaining of mammary tissue sections by IL-6 antibodies was blocked by incubation with recombinant IL-6 (Genzyme; 300 ng cytokine ϩ 1 µg antibody in 100 µl for 6 h), but not with equivalent amounts of recombinant IL-2 or IL-8 (Genzyme).

Ki67 labelling index
The nuclear proliferation-associated antigen Ki67 was assayed on frozen sections using the Ki67 mouse monoclonal antibody (Dako A/S) diluted 1:10 in phosphate-buffered saline (PBS) and incubated overnight. Immunoreactivity was revealed as stated above. Negative control sections were incubated with normal mouse serum. As previously reported (Campani et al, 1991), the evaluation of Ki67 was expressed as the percentage of positive cells. For each tumour section, at least 25 microscopic fields (40 ϫ objective) were analysed. The percentage was calculated dividing the number of Ki67-positive neoplastic cells by the total number of cells counted, and the result multiplied by 100. Tumours were classified as highly proliferating if more than 10% (median value of the distribution) of cells were immunoreactive (Campani et al, 1991).

Hormone receptor analysis
The presence of oestrogen receptor (ER) was determined using the Abbott ER-ICA monoclonal kit (Abbott Laboratories, Abbott Park, MI, USA), as described (Campani et al, 1991). Briefly, rat anti-human ER antibody was incubated with frozen sections followed by a goat anti-rat IgG and revealed with a rat peroxidase-anti-peroxidase complex. Peroxidase activity was detected by the incubation of the antibody complex with 3,3′-diaminobenzidine and hydrogen peroxide. Expression of progesterone receptor (PR) was determined by the same method as ER, using the Abbott PR-ICA monoclonal kit (Abbott Laboratories). Negative controls were incubated with normal rat serum. ER and PR expression was quantitated as a percentage value as described for Ki67. Tumours with ≤ 5% positive cells were classified as negative. a Tumours with scores of 0 and 1 were considered as negative; tumours with scores ranging from 2 to 7 were considered as positive. b Statistical analysis: χ 2 with Fisher's correction (the NK group as not included). c NK, not known. a Tumours with scores of 0 and 1 were considered as negative; tumours with scores ranging from 2 to 7 were considered as positive. b Statistical analysis: χ 2 with Fisher's correction (the NK group was not included). c Tumours with ≤5% positive cells were considered as negative; tumours with Ͼ 5% positive cells were considered as positive. d NK, not known. e Low expression means ≤10% of positive cells; high expression, means Ͼ 10% of cells.

Statistical methods
A χ 2 test with Fisher's correction was used for the correlating clinico-pathological parameters with IL-6 immunoreactivity. For continuous variables, Sperman's test was used to analyse the correlation between the percentages of IL-6-positive cells and positivity for p53, ER, PR and K67.

IL-6 protein expression
The histotype of all cases was determined according to the widely accepted criteria of Rosen and Oberman (Rosen and Oberman, 1993). IL-6 expression was studied by immunostaining with polyclonal and monoclonal antibodies. Both reagents produced clear Correlation between IL-6 expression and expression of p53, PR and ER in breast carcinoma. The percentage of IL-6-positive cells in each case was related to the positivity for p53, PR or ER. Statistical analysis was performed by the Spearman test for continuous variables. No correlation was found between IL-6 and p53 expression (r ϭ Ϫ0.042, t ϭ Ϫ0.51; P ϭ 0.6); a direct correlation was observed between IL-6 expression and positivity for both ER (r ϭ 0.381, t ϭ 3.7, P ϭ 0.00037) and PR (r ϭ 0.22, t ϭ 2.27; P ϭ 0.025) cytoplasmic reactivity, but the staining pattern produced by the polyclonal rabbit antibody was more intense and reproducible (Figure 1 A, B). Reactivity of both antibodies was totally abolished by preincubation with recombinant-IL-6 (r-IL-6), but not with either r-IL-2 or r-IL-8. The immunoreactivity of each tumour was scored from 0 to 7 as reported in Materials and Methods. The staining pattern was heterogeneous (i.e. in positive sections the intensity of staining ranged from negative to strong). As reported in Table 1, the majority of cases expressed immunoreactive IL-6 (82%). Among cases of IL-6-negative carcinoma, staining was clearly limited to remaining normal ducts and lobules.

IL-6 expression in relation to clinico-pathological parameters
In Table 1, patients were grouped according to IL-6-negativity and IL-6-positivity. We analysed the data in relation to the following clinico-pathological variables: tumour size, histological grade and nodal status. A significant association was found between IL-6 expression and histological grading of the tumours; i.e. low-grade tumours (grade I) had higher IL-6 expression than high-grade tumours (grade II and III; P ϭ 0.0017). No statistically significant association was, however, found between IL-6 expression and tumour size or nodal status.

DISCUSSION
Over the last few years, the role of cytokines in cancer has been the subject of numerous investigations (Porter and Lippman, 1994). It has been reported that, in the mammary gland, cytokines play a role in growth and differentiation (Massagué et al, 1992;Basolo et al, 1993b;Danforth and Sagias, 1993), extracellular matrix production (Silberstein et al, 1992), angiogenesis (Garver et al, 1994) and as immunomodulating factors (Allione et al, 1994). Recently, we have shown that epithelial cells of the normal mammary gland produce constitutively IL-6, IL-8 and a nonsecreted form of TNF (Basolo et al, 1993a). In addition, we observed that upon transformation in vitro by several different oncogenes (except for int-2), MEC lose the ability of producing and responding to IL-6 (Basolo et al, 1993b).
It has also been demonstrated in vivo that, as compared to normal mammary tissue, the expression of IL-6 in neoplastic tissues is slightly reduced in non-invasive ductal carcinoma, but strongly reduced in the invasive forms of ductal and medullary carcinoma. This finding suggests an inverse relationship between tumour aggressiveness and IL-6 expression, in agreement with observations in tumours of the salivary gland (Gandour-Edwards et al, 1995a) and thyroid (Basolo et al, 1998). These results have been confirmed by experiments in vitro. Primary cultures from ductal carcinomas produce reduced amounts of IL-6 as compared to primary cultures obtained from the normal mammary gland (Basolo et al, 1996b).
IL-6 expression appears to relate to prognosis in tumours other than those of breast. An inverse relationship between IL-6 expression and biological aggressiveness has been reported in salivary gland tumours (Gandour-Edwards et al, 1995a), malignant melanoma cells have been shown to be unresponsive to exogenous IL-6 (Lu and Kerbel, 1993;Silvani et al, 1995), invasive pituitary malignancies appear to overexpress IL-6 (Gandour-Edwards et al, 1995b). In ovarian cancer patients, IL-6 serum levels have been correlated with a poor prognosis, since patients with low IL-6 levels had better survival than patients with high levels of this cytokine (Scambia et al, 1995). The above results prompted us to investigate whether alterations of IL-6 expression were correlated with tumour aggressiveness as evaluated by clinico-pathological and biological parameters. Although no association was observed between IL-6 expression and tumour size or nodal status, a statistically significant association was found with the histological grade of tumours. Approximately one-fifth (25/121) of investigated ductal carcinomas were classified as grade I. None of these 25 well-differentiated tumours was IL-6-negative. On the contrary, 20/96 (21%) grade II or III ductal carcinomas were IL-6-negative. Since high grade indicates less differentiated tumours, this finding suggests that reductions of IL-6 expression are associated with late stages of tumorigenesis. This was confirmed by the observation that highly proliferating tumours (as evaluated by Ki67 immunoreactivity) were characterized by low percentages of IL-6-positive cells (r ϭ Ϫ0.28, t ϭ Ϫ2.28, P ϭ 0.027). Other interesting correlations of IL-6 expression with other biological parameters (ER, PR) were detected.
On the contrary, IL-6 expression was directly correlated with the expression of ERs and PRs that represent important mammary differentiation markers. Eighty-three out of 94 (88%) ER-positive tumours were also reactive with antibody to IL-6 (P ϭ 0.02; Table  2), and the percentage of ER-positive cells was directly related to that of IL-6-positive cells (P ϭ 0.00005; Figure 2). Equivalent results were obtained with regard to the percentages of PR-positive cells (P ϭ 0.025; Figure 2). The potential interaction between cytokines and steroid hormones has already been suggested by studies of breast cyst fluid (Reed et al, 1992). It was observed that IL-1 and IL-6 do increase oestrogen synthesis by stimulating aromatase and oestradiol-dehydrogenase activities in breast cancer cells. In addition, it has been reported that ER-positive human breast cancer cells express the IL-6 receptor and that their proliferation is inhibited by addition of IL-6 (Chen et al, 1988;Chiu et al, 1996). In experiments in vitro we found that insertion of the int-2 gene into MEC results in the up-regulation of IL-6-receptor thus promoting cell proliferation in response to exogenous IL-6 and glucocorticoids (Basolo et al, 1993a). Taken together, these results indicate a possible autocrine and/or paracrine role of IL-6 at least in the subgroup of mammary tumours that express the IL-6receptor. The existence of a complex interplay between IL-6 pathways and steroid hormones is further exemplified by the finding that in human osteoblasts the IL-6 promoter is inhibited by oestrogen in the absence of a functional ER binding site (Stein and Yang, 1995). In conclusion, it appears that IL-6 is capable of modulating steroid hormone responsiveness and vice-versa.
Overexpression of the p53 protein has been correlated with poor patient prognosis in breast cancer (Stenmark-Askmalm et al, 1995;Horne et al, 1996), indicating that p53 represents an important prognostic marker. However, in this study we failed to detect correlations between p53 overexpression and IL-6 expression ( Table 2, P ϭ 0.2; Figure 2, P ϭ 0.6). It is of interest to note that p53 appears to modulate IL-6 expression. Experiments in HeLa cells demonstrated that the IL-6 promoter is repressed by wildtype p53 and up-regulated by mutant forms (Margulies et al, 1993;Stenmark-Askmalm et al, 1995).
Since MEC constitutively produce not only IL-6, but also IL-8 and a non-secreted form of TNF, further studies are needed to ascertain if the expression of the ER and PR differentiation markers is also associated with altered expression of the above cytokines. In this context it should be recalled that Fas (APO-1/CD95) is expressed on the surface of normal and transformed MEC (Keane et al, 1996) and that interaction of Fas with ligands of the TNF family may trigger programmed cell death (Hug, 1997). It will be of interest to evaluate whether Fas expression is downregulated in high grade mammary tumours as a means of escaping the apoptotic response as already reported for colon carcinoma (von Reyhet et al, 1998).