Prostaglandin F2 alpha in benign and malignant breast tumours.

Prostaglandin F2 alpha (PGF2 alpha) was determined by radioimmunoassay in 57 breast carcinomata, 16 fibroadenomata, and 33 sclero-cystic-disease (SCD) specimens. In 41 cases of carcinoma and 10 cases of fibroadenoma, histologically non-malignant tissue was also obtained from the same breast. PGF2 alpha levels were significantly elevated in breast cancer when compared with the normal tissues and benign diseases (P less than 0.005 for each group). High PGF2 alpha levels were positively correlated with differentiation, positive oestrogen and progestagen receptor status, and low mitotic index. Tumours with good prognosis (less than 20 mm, negative lymph nodes, some degree of differentiation) showed significantly higher PGF2 alpha levels than tumours with a bad prognosis (greater than 20 mm, positive nodes and undifferentiated). A tendency for elevated PGF2 alpha levels was observed with negative lymphatic permeation, postmenopausal status, low grade of nuclear and cellular polymorphism and high degree of elastosis and fibrosis. No correlation was observed between PGF2 alpha levels and host-cell reaction. Plasma levels of 15-keto-13, 14-dihydro-PGF2 alpha were not elevated in cancer patients when compared with the SCD-group. The present study demonstrates that PGF2 alpha levels are high in tumours with good prognosis. However, since other authors have suggested that a high PGE2 production is a bad prognostic index, it is possible that conversion of PGE2 to PGF2 alpha by 9-keto-reductase explains this relationship. Nevertheless, the presented results question the unrestricted use of prostaglandin-synthesis-inhibitors in the treatment of breast cancer.

Prostaglandins (PGs), especially of the E-series, have been shown to be elevated in a large number of human and experimental tumours. Special efforts have been made to investigate the features and the role of PG-synthesis in human breast cancer (Bennett et al., 1975;Powles et al., 1976;Kibbey et al., 1980;Greaves et al., 1980;Malachi et al., 1981;Campbell et al., 1983).
A considerable volume of research on the mechanisms of action of PGs in a wide variety of cells and tissues, indicates that PGs are possibly involved in tumour initiation, tumour promotion, cell proliferation and differentiation, the immune response, tumour metastasis, osteolysis and hypercalcaemia (Karmali, 1980;Honn et al., 1981a;Droller, 1981;Goodwin, 1981).
The present study was designed to describe the basic features of PGF2X, production in benign and malignant breast tumours. Normal glandular breast tissues were used as controls. We determined thoroughly investigated in human mammary cancer.
In order to find out whether the high levels of prostaglandin F2a found in cancerous tissues could be correlated with either metastatic potential or with other prognostic unfavourable variables, the PGF2a, levels were examined in relation to the size of the tumour, axillary lymph node status, lymphatic vessel permeation, histological type and differentiation of the tumour, mitotix index, oestrogen and progestagen receptor status, and age and menopausal status of the patient.
Tumourassociated host-cells can produce considerable amounts of prostaglandins (Humes et al., 1977;Brune et al., 1978). Therefore the numbers of host-derived cells and the amount of necrosis were evaluated by means of quantitative microscopy. The epithelial cellularity was also evaluated by morphometric determination of the mean nuclear density and mean nuclear area.

Materials and methods
There were 165 specimens from 106 patients who underwent surgery for a breast lump. Each specimen was divided into two representative parts and immediately immersed, either in acetone cooled by solid CO2 (-70°C) for PG-investigations, or in Bouin's liquid for histopathological examination.
The tissue samples for PGF2a investigation were then stored at -300C until radioimmunoassay was performed. Sections 8pm thick were cut from the stored routine embedded blocks, processed by standard methods, and stained by haematoxylin and eosin.
Histologically proven normal breast tissue from 41 breasts with a carcinoma, and 10 from breasts with an adenofibroma, were investigated. In addition 6 lymph nodes, two of which showed metastasis, were examined.
In blood removed at the time of surgery the serum level of 15-keto-13, 14-dihydro-PGF2a, the main metabolite of PGF2,, was measured in 11 cancer patients and in 16 patients with sclerocysticdisease. The estimation of the PGF metabolite rather than PGF2a as such was preferred, since the former more closely reflects the true production in vivo and is less liable to erroneous changes during blood sampling.
The age and menopausal status of the patients and the tumour size at anatomopathological examination were recorded. Menopause was defined as at least one year after the last menstrual period.
Twenty-one cancerous lesions were examined for oestrogen and progestagen receptor status, determined according to Noel et al. (1982). In our laboratory, the oestrogen receptor status and progestagen receptor status are considered positive for levels higher than 10 and 100 fmole mg'protein respectively.
The PGF2. amounts were expressed as ng PGF2. mg-1 protein and as ng PGF2a per cellularity index (g wet w tissue divided by the mean nuclear density). The number of evaluated sam'ples is higher when the cellularity index is used, as there was not always enough tissue to measure the protein content. Also, the histological variables could not always be measured.

Histopathology
The slides were independently reviewed by three of the authors and were re-evaluated by a senior pathologist in cases of discordance. The results of the PG investigation were not known at that time.
Tumours were classified according to the WHO classification (1981). Whenever there was a combination of more than one histological type, the tumour was assigned to the group to which its most extensive component belonged. Tumours with a high amount of intraductal necrosis were classified as comedocarcinomata (Haagensen, 1971). The malignant tumours were graded according to Bloom & Richardson (1957).
Inflitration, fibrosis, elastosis and host-cell reaction were determined as negative, mild, moderate or strong. Nuclear and cellular polymorphism were calssified as mild, moderate or strong. The nuclear: cytoplasmic ratio was recorded as small, moderate or high.
Axillary lymph node status was recorded positive or negative depending on whether cancerous cells were present or not. Lymphatic vessel permeation was recorded as positive when cancerous cells were present in the lymphatic vessels in the tumour, or when the lymph node status was positive.
Quantitative microscopy Nuclear density of the carcinomatous cells, nuclear size, mitotic index and necrosis, were evaluated morphometrically, according to Weibel (1979), using a planimeter (Kontron -MOP AM 02) and a microprocessor (HP 98 15 A).
The necrosis and total area of all the carcinomatas were determined with semi-automatic histomorphometry. Presence of necrosis was evalulated as percentage of the total area of representative tumour tissue.
Mean nuclear density and mean nuclear area of the malignant epithelial cells were counted on a projection microscope at x 400 magnification using the planimeter and microprocessor. Twenty areas of 6680 pm2 were counted according to Weibel (1979).
The numbers of mitoses were counted in 20 random fields at x 400 magnification. Once focussed, no further adjustment was allowed and the structures that could be differently interpreted were not counted. The mitotic index was expressed as the mean of the number of mitoses counted in the 20 evaluated fields.

Radioimmunoassay of PGF2a
Before radioimmunoassay of PGF2a the acetone was evaporated under nitrogen and the weight of the tissue determined. Tris buffer (50 M, pH=8.00 at 25°C) was added (3mlg-' tissue) and sonicated for 90min (Bransonic). Ice was regularly added to the bath fluid of the sonication apparatus to keep temperature below 10°C. The supernatant was separated from the tissue after centrifugation at 10,OOOg (Eppendorf centrifuge). The extraction yield for PG was checked by adding trace amounts of radiolabelled standards.
A precipitate was formed with bovine-y-globulin after adding polyethyleneglycol (PEG) 4000. Radioactivity was counted in a Packard 460 scintillation counter with quench correction. The protein content of the breast tissue extracts was determined (Bradford, 1976) and the PGF2a content expressed as ng PG mg-' protein.
The extraction yield of PGF2a from the specimens was 80.8 + 4.1% (n = 6), calculated by adding trace amounts of radioactive PGF2 to the tissues.

Statistical analysis
The used statistical test were: Wilcoxon (u value), one way analysis of variance (ANOVA; F value) and significance of the correlation coefficient (linear regression; t value) (Goldstein, 1964). When no statistical test is mentioned, the Wilcoxon test was used to compare two groups of samples.

PGF2a levels in relation to the histopathological groups
The results according to the different histopathological groups are shown in Figure 1 and are analysed statistically in Table I. The PGF2a yield from malignant tissue (CA) was higher than from non-malignant tissues from breast with a carcinoma  (N-CA), fibroadenomata (FA), sclero-cystic-disease tissues (SCD) and normal glandular breast tissues (N) (all P<0.0003).
Amounts from fibroadenomata (FA) were higher than from normal glandular breast tissues (N) (P=0.002). We found no difference in the PGF2a yields from the N and N-CA groups (P = 0.43).
PGF2a yield tended to be higher from SCD specimens than from N specimens (P=0.069).
In 12 out of 26 patients a measurable 15-keto-13,14-dihydro-PGF2. level was found, i.e. >25pgml-P. Only 3 patients showed plasma levels >100pgml-P (475, 456 and 1065pgml-'). We found no correlation between plasma levels of 15keto-13, 14-dihydro-PGF2a and PGF2e breast tissue levels. There was no apparent difference between the plasma levels found in the CA group and the SCD group.
Relationship between the breast cancer PGF2a levels and the different prognostical variables Tumour oestrogen and progestagen receptor contents Oestrogen-positive tumours yielded more PGF2 mg-1 protein, than did oestrogen-negative tumours (P=0.004); progesterone-positive tumours yielded more PGF2a mg-1 protein (P = 0.05) and per cellularity index (P=0.037) (Table II and Figure   2).
Age and menopausal status Tumour PGF2a levels mg'-protein tended to be higher with older patients (>60 years) (ANOVA; F = 2.70; P = 0.06), and with the cellularity index the P value was 0.03 (ANOVA; F = 3.33).
There was no significant difference between tumour PGF2. levels of pre-and postmenopausal patients (P=0.29) (Table III).
Histological type and differentiation Statistical analysis of all the histological types was difficult, because the infiltrating ductal carcinomatas composed the only substantial group. They yielded more PGF2a than did the other histological types (Table IV). The two cellular intracanalicular fibroadenomata yielded 0.2 and 4.0 ng PGF2a mgprotein respectively. Histological differentiation was recorded as small, moderate or high. For statistical analysis we divided  (24) 14.80+2.14 The PG results are given as ngmg-1 protein, mean+s.e. The number of evalulated samples is given in brackets.
Oestrogen receptor pos = oestrogen receptor contents > 10 fmol mg -1 protein Oestrogen receptor neg = oestrogen receptor content < 10 fmol mg -I protein Proesterone receptor pos = progesterone receptor content .100 fmol mg-' protein Progesterone receptor neg = progesterone receptor content < 100 fmol mg -I protein Differentiation undiff= undifferentiated diff= some degree of differentiation (small, moderate or high). *the tumours into two groups: undifferentiated and some degree of differentiation (small, moderate or high). The tumours with some degree of differentiation yielded more PGF2a than did the undifferentiated tumours (P = 0.004, ng mg 1 protein; P = 0.013, ng cellularity index-1) (Table II and Figure 3). ANOVA between the various groups with some degree of differentiation showed no significant differences (F=0.16, P=0.85).
Lymph node metastasis and lymphatic vessel permeation We found, at most, a weak tendency for higher PGF2a levels in the lymph node negative group compared with the node positive group (P = 0.189) (Table III). PGF2a was determined in 6 lymph nodes, the values for the negative lymph nodes being 0.1, 7.7, 1.9 and 9.2 ng mg-1 protein; those for the positive lymph nodes were 3.3 and 9.7 ng mg'-protein.
Lymphatic vessel permeation was present in 82% of the cases and there was little or no difference (P = 0.209) between this group and the small negative group (Table III).
Size and density of nuclei of carcinoma cells We observed no correlation between the PGF2a levels and the mean nuclear density or size (P=0.32; r= -0.064 and P = 0.49; r = -0.002 respectively). We observed also no correlation with the mean nuclear area (P = 0.30; r = 0.076).

Mitotic index The mitotic index correlated
inversely with the ng PGF2amg -protein (P = 0.05; r = -0.227), but when the cellularity index was used, the significance was P = 0.13 (r = -0.159).  (25) 20-<39mm (21) .40 mm (7) Menopausal status pre (21) post (32) Lymph nodes negative (25) positive (28) Lymphatic permeation negative (7) positive ( The PG results are given as ngmg-' protein, mean+ s.e.m. The number of evaluated samples is given in brackets. aCases difficult to classify were not evalulated. Nuclear and cellular polymorphism and the nuclear: cytoplasmic ratio Tumours with a low degree of nuclear and cellular polymorphism and a low nuclear: cytoplasmic ratio showed at most a weak tendency to yield more PGF2. (ANOVA, F=0.60; P = 0.55; and F = 1.64, P = 0.20 respectively) (Table  III).
Host cell reaction and necrosis PG levels tended to correlate inversely with the host-cell reaction (ANOVA, F=2.01, P=0.15), but there was little or no relationship to the presence of mast cells (P=0.37), or to the amount of necrosis (ANOVA, F=0.13; P=0.88) (Table III).

Cumulation of variables A gradual increase in
statistical significance occurred when several characteristics were combined. Mean PGF2a levels ng mg 1 protein tended to be higher when the lymph nodes were negative (P = 0.189), and the relationship marginally strengthens when tumours <20 mm with negative lymp nodes are compared with tumours >20mm with positive lymph nodes (P=0.104). Tumours with some degree of differentiation < 20 mm and with negative lymph nodes yielded more PGF2a than did undifferentiated tumours >20mm with positive lymph nodes (P = 0.045) ( Table V).

Discussion
The present study has examined PGF2a tissue levels in benign and malignant breast tumours. This is to our knowledge the first study which has determined by radioimmunoassay the PGF2a tissue levels in breast cancers and compared them with histologically proven non-malignant tissues from the same breast with correlations in terms of PGF2 mg-1 protein and cellularity. Bennett et al. (1975Bennett et al. ( , 1977 using bioassay first described elevated levels of "prostaglandin-like' material in extracts of human mammary cancer. Greaves et al. (1980) performed radioiommunoassay of PGE and PGF and correlated the results with tissue weight in a small group of breast cancers (n= 16). Rolland et al. (1980) determined the PGE2 production from added arachidonic acid in 91 breast cancers. Bishop et al. (1980) and Malachi et al. (1981) determined PGE2 and Fulton et al. (1982 PGE2and PGF2a by radioimmunoassay in breast cancer and correlated the results with tissue wet w. Watson et al. (1984) measured PGE2 and PGF2a, by Gas Liquid Chromatography -Mass Spectrometry in 100 mammary carcinomatas.
Diff =differentiated: some degree of differentiation (small, moderate or high).
non-malignant tissue of the same breast confirms the findings in the studies cited above. We also found statistically significant different PGF2a yields from cancers, compared with fibroadenomata, sclero-cystic-disease and normal glandular breast tissue (Table I). Stamford et al. (1980) showed an increase of prostaglandin-like material in extracts of blood draining breast carcinomatas. In the study of Powles et al. (1977) 15-keto-13, 14-dihydro-PGE2 was elevated, especially in the group of breast cancer patients who had metastasis. In the study of Malachi et al. (1981) the plasma PGE2-metabolite concentration did not reflect the PGE2 tissue levels, and no difference was found between the benign and malignant cases, but both concentrations were higher than those of the healthy controls. In this study none of the patients had overt metastasis at the time of biopsy. We could not demonstrate a higher level of 15-keto-13, 14-dihydro-PGF2a in the plasma of the cancer patients than in the SCD group, or a correlation between the plasma levels of the PGF2a-metabolite and the PGF2a tissue levels. Bennett et al. (1977Bennett et al. ( , 1983 showed that the ability of malignant breast tumours to produce prostaglandin-like material correlates inversely with patient survival. Rolland et al. (1980) concluded that a high PGE2 production occurs very early in the development of a malignant tumour, and that an elevated prostaglandin production seems to be associated with metastasis. This relationship does not seem to exist for PGF2a tissue levels.
In the present study high PGF2a levels correlate with good prognostic variables (differentiation, positive oestrogen and progestagen receptor status, low mitotic index). Tumours with a good prognosis (some degree of differentiation, <20mm, no lymph node metastasis) yielded more PGF2a than did tumours with a bad prognosis (undifferentiated, > 20 mm, lymph node metastasis) Table V), although the relationships with these uncombined variables were weak.
We also observed tendencies for higher PGF2a levels with negative lymphatic permeation, postmenopausal patients, fibrosis, elastosis and a low nuclear: cytoplasmic ratio (Table III). Malachi et al. (1981) found no correlation between PGE2 tissue levels and survival, histological type and stage. The data on tumour recurrence of Bennett et al. (1983) argued against an important role for PGE2 in bone metastasis, and Blamey (personal communication) did not find a correlation between PGE2 and bone metastasis. In a series of rat mammary carcinomatas PGE2 levels correlated inversely with metastatic potential (Kibbey et al., 1979). Fulton et al. (1982) and Campbell et al. (1983) concluded that oestrogen receptorpositive tumours synthesized more PGE2, whereas Rolland et al. (1980) found a tendency for the opposite. Wilson et al. (1980) and Watson et al. (1984) found little or no correlation between prostaglandins and oestrogen receptor status. In the present study PGF2, correlated with receptors for steroid hormones. Host-derived inflammatory cells are thought to contribute to the PG-production by cancers (Greaves et al., 1980;Honn et al., 1981a).  correlated total "PGE2-equivalents" with the extent of inflammation and necrosis in Xirradiated squamous carcinomata of head and neck. We found no significant relationship between the PGF2a levels and the intensity of the host cell reaction, presence of mast cells or necrosis. This corresponds with the finding that most of the inflamatory cells were lymphocytes, and that coincubation of carcinogen-induced rat bladder tumour cells with lymphocytes did not significantly change the PGE2 yield (Owen et al., 1980). PGs might be involved in tumour initiation, tumour promotion, cell proliferation, cell differentiation, the immune response, tumour metastasis and hypercalcaemia. Initiation of carcinogenesis most commonly requires oxidation, which can occur during PG synthesis (Honn, 1981a;Marnett, 1982;Zenser et al., 1982).
Views on the role of PGs in tumour growth are controversial. Some authors claim that increased PG synthesis represents a part of the homeostatic response to limit tumour growth. Others claim that PGs are involved in the initiation and the enhancement of tumour growth. Furthermore the effect of prostaglandin synthesis inhibitors on tumour cell growth, -enhancement or inhibitioncan vary according to the tumour cell type and to the concentration of the anti-inflammatory drug used (Karmali, 1980). Impressive effects of PGs on cells in vitro are the induction of maturation and differentiation. PGE regulates proliferation and differentiation of stem cells in the bone marrow in vitro (Bockman, 1982). Certain PGs notably PGA, PGD and PGI2, can induce differentiation in vitro of mouse mammary carcinoma, neuroblastoma, B16 melanoma and human malignant melanoma (Jubiz et al., 1979;Rudland & Warburton, 1982;Prasad, 1982;Bregman & Meyskens, 1983;Simmet & Jaffe, 1983).
It is also possible that PGs influence the host/tumour interplay. PGE2 is thought to be a factor involved in the failure of the immune system to eliminate tumours (Goodwin, 1981), but it is not yet known whether inhibition of prostaglandin synthesis will enhance a putative immune reaction of the host to all types of tumours (Kelly & Parker, 1979, Stringfellow & Fitzpatrick, 1979Favalli et al., 1980). Investigation into the control of metastasis to bone and of hypercalcaemia associated with malignancy was prompted by the finding that some PGs cause osteolysis. Unfortunately, in man there are no consistent data proving the effectiveness of prostaglandin synthesis inhibitors on bone metastasis and hypercalcaemia (Powles et al., 1982). Numerous platelet anti-aggregating substances e.g. apsirin, indomethacin, dipyridamole, flurbiprofen and benorylate have been investigated as possible antimetastatic agents with both positive and negative results (Bennett, 1982;Honn 1981a). Of considerable interest is the hypothesis proposed by Honn et al. (1981b) that tumour cells can alter the TXA2/PGI2 balance in favour of platelet aggregation. But the extents to which thromboxane inhibitors or prostacyclines, which reduce platelet aggregation, are of value in human cancers are not known.
The present study demonstrates that PGF2a yield from tumours is high in patients with good prognosis. However, since other authors (Bennett et al., 1977(Bennett et al., , 1983Rolland, 1980) have suggested that a high PGE2 production is a bad prognostic index, it is possible that conversion of PGE2 to PGF2a by 9-keto-reductase explains this relationship. Nevertheless, the presented results question the unrestricted use of prostaglandin-synthesisinhibitors in the treatment of breast cancer.