Cytoplasmic induction and over-expression of cyclooxygenase-2 in human prostate cancer: implications for prevention and treatment


Dr El-Nasir Lalani, Department of Histopathology, Division of Investigative Sciences, Imperial College School of Medicine, Hammersmith Hospital Campus, Du Cane Road, London W12 0NN, UK.


Objective To assess the level and morphological distribution of cyclooxygenase (COX)-1 and -2 in human prostates and to determine any association with the Gleason grade of prostate cancer.

Materials and methods The study comprised 30 samples from patients with benign prostatic hyperplasia (BPH) and 82 with prostate cancer. Immunohistochemistry was used to assess the expression of COX-1 and -2, and 13 samples were also assessed using immunoblotting (six BPH and seven cancers).

Results For both BPH and prostate cancer, COX-1 expression was primarily in the fibromuscular stroma, with variable weak cytoplasmic expression in glandular/neoplastic epithelial cells. In contrast, COX-2 expression differed markedly between BPH and cancer. In BPH there was membranous expression of COX-2 in luminal glandular cells and no stromal expression. In cancer the stromal expression of COX-2 was unaltered, but expression by tumour cells was significantly greater (P = 0.008), with a change in the staining pattern from membranous to cytoplasmic (P < 0.001). COX-2 expression was significantly higher in poorly differentiated than in well differentiated tumours (P < 0.001). These results were supported by immunoblotting, which showed similar levels of COX-1 in both BPH and cancer, but four times greater expression of COX-2 in cancer than in BPH.

Conclusion This is the first study to assess the co-expression of COX-1 and COX-2 proteins in benign and malignant human prostates, and showed the induction and significantly greater expression of COX-2 in cancer, which was also associated with tumour grade. The regular use of nonsteroidal anti-inflammatory drugs is associated with a reduced incidence of cancers. The present results provide the basis for a potential role for COX-2 inhibitors in the prevention and treatment of prostate cancer.


Prostate cancer is the commonest adenocarcinoma and second commonest cause of cancer death in men in the Western hemisphere [1]. The progression of prostate cancer is a multistep process and involves dysregulation of cell proliferation and death, development of invasive and metastatic phenotype, and the acquisition of hormone insensitivity [2].

Epidemiological studies of prostate cancer show an association between increased incidence of the disease and a high dietary intake of fat, especially animal fat [3]. This is thought to be related to the growth stimulating effects of the n-6 polyunsaturated fatty acids (PUFAs), e.g. arachidonic acid, present in animal fats, mediated by PGs and leukotrienes [3]. Prostaglandins are derived from n-6 PUFAs by a series of enzymic reactions. The first rate-limiting step in PG synthesis from arachidonic acid involves the cyclooxygenases (COXs), also known as PG G/H synthases. Consistent with the model that PGs derived from dietary fats may predispose to various cancers, several population-based studies have shown that regular use of aspirin and other NSAIDs results in a 40–50% decrease in the relative risk of colorectal cancer [4,5]. This is relevant because NSAIDs act through inhibiting COXs. In a recent population-based case-control study, there was a trend towards reduced risk of advanced prostate cancer with regular use of NSAIDs [6].

There are two COX isoforms, COX-1 and COX-2, that differ both in their regulation and tissue distribution [7]. COX-1 is a constitutively expressed ‘housekeeping’ gene involved in processes like gastric acid secretion, vascular homeostasis and water reabsorption by the renal collecting tubules. In contrast, COX-2 is inducible and thought to be involved in differentiative processes such as inflammation and ovulation [7]. Of the two isoforms, COX-2 is the most consistently up-regulated in many cancers, including oesophagus [8], stomach [9], colon [10], lung [11], pancreas [12] and head and neck [13].

Recent studies on rat mammary glands suggest that hormonal influences on cancer development may also be mediated by COX-2 gene expression and PG synthesis [14,15]. COX-2 inhibitors are chemopreventive against colon [16] and lung cancers [17] in mouse models. Furthermore, COX-2/ApcΔ716 double-gene ‘knockout’ mice have fewer and smaller intestinal polyps than ApcΔ716 knockout mice [18]. Evidence that increased levels of COX-2 may be important in the development of prostate cancer comes from preliminary results in human [19,20] and canine prostates [21]. Thus the aim of the present study was to assess the co-expression of COX-1 and -2 proteins in adult human benign and malignant prostates, and to analyse their expression pattern.

Materials and methods

The study included 112 specimens (formalin-fixed and fresh-frozen) of prostates obtained during TURP (30 of BPH and 82 of adenocarcinoma), from the Department of Histopathology and Human Biomaterials Resource Centre, Imperial College School of Medicine, Hammersmith Hospital Campus, London, and the Department of Histopathology, West Middlesex University Hospital, London. The median (range) age of the patients was 70 (51–92) years. The tumours were graded according to the Gleason system by one author(E-N.L.). Of the adenocarcinomas, 10 were well differentiated (Gleason grade 2–4), 32 moderately differentiated (Gleason grade 5–7) and 40 poorly differentiated (Gleason grade 8–10).

For immunohistochemistry (IHC) all specimens were fixed in 10% neutral buffered formalin, paraffin embedded and processed routinely; 4 µm thick serial sections were taken onto poly- l-lysine-coated slides. A three-step immunoperoxidase method (described previously [22]) was used to detect the expression of COX. Briefly, sections were de-waxed in xylene, hydrated through graded alcohols and water, and immersed in 0.3% v/v H2O2 in distilled water for 30 min to block endogenous peroxidases. Antigens were retrieved by microwaving at 750 W for 15 min in 0.01 mol/L trisodium citrate buffer (pH 6.0). Sections were rinsed well in standard PBS (pH 7.2–7.4) and nonspecific binding sites blocked with 10% normal rabbit serum (Dako, Denmark) for 30 min. Sections were incubated with COX-1 (160110) or COX-2 (160112) mouse mAb (Cayman Chemicals, Ann Arbor, MI) at a dilution of 1 : 300 and 1 : 200 in PBS, respectively, for 16 h at 4 °C. After rinsing with PBS, sections were incubated with biotinylated rabbit antimouse immunoglobulins (Dako) at a dilution of 1 : 200 in PBS for 45 min. Sections were rinsed with PBS and incubated with avidin-biotin horseradish peroxidase complex solution (Dako) for 30 min, rinsed with PBS and immersed for 5–10 min in a peroxidase substrate solution containing 0.05% w/v 3,3′-diaminobenzidine (Sigma Chemical Co., Poole, UK) and 0.02% v/v H2O2 in PBS. Sections were counterstained with Cole's haematoxylin (Pioneer Research Chemicals, UK), dehydrated, cleared, and mounted in mounting medium. Normal colon sections known to express COX-1 and COX-2 were used as positive controls, while for negative controls the primary antibody was replaced with PBS. The immunostaining was evaluated independently by two authors (S.M. and E-N.L.) analysing the intensity, distribution and pattern of immunostaining. If there was a discrepancy, a consensus was reached after further evaluation. The intensity of immunostaining was graded as 0 (negative), 1 (weak), 2 (moderate) and 3 (strong).

Western blotting was used on 13 fresh-frozen samples (six BPH and seven prostate cancers); 30 sections (15 µm thick) were cut from each sample using a cryostat (at −25 °C) and placed in pre-chilled Eppendorf microfuge tubes. The tubes were transferred to ice and 300 µL of lysis buffer (715 mol/L 2-mercaptoethanol, 10% glycerol, 2% SDS, 40 mmol/L Tris pH 6.8, 1 mmol/L EDTA) containing a cocktail of protease inhibitors (Boehringer Mannheim, UK) was added to each tube. After 20 min the lysates were centrifuged at 13 000 rpm at 4 °C for 5 min. The supernatants were transferred to clean microfuge tubes and the protein concentration determined using a protein assay reagent (Bio-Rad Laboratories, Hercules, CA) and Ultraspec III spectrophotometer (Pharmacia Biotech, UK). About 30 µg of total protein from each sample was subjected to SDS-PAGE, the proteins transferred to nitrocellulose membranes (Millipore, UK), and the membranes blocked with 5% non-fat milk (Marvel, Cadbury Schweppes, UK) in Tris-buffered saline solution containing 0.5% Tween 20 (TBST) for 1 h at room temperature. The blots were then probed overnight with COX-1 or COX-2 mouse mAb (Cayman Chemicals) at a dilution of 1 : 1000, washed in TBST for 1 h with a change of buffer every 10 min, and probed with horseradish peroxidase-conjugated rabbit antimouse immunoglobulins (Dako) at a dilution of 1 : 1000. After another wash in TBST for 2 h with a change of buffer every 15 min, the blots were placed in enhanced chemiluminescence solution (ECL, Amersham, UK) and exposed to X-ray film (Hyperfilm, Amersham). The relative expressions of the different proteins were measured using a densitometer (Molecular Dynamics, UK).

Fisher's exact test was used to compare the intensity and pattern of expression of COX-1 and COX-2 between benign and malignant samples, and among the various grades of malignancy, with P < 0.05 was considered to indicate significant differences.


On IHC, in all BPH samples COX-1 was constitutively expressed in the fibromuscular stroma ( Fig. 1A); the expression was heterogeneous and cytoplasmic, with perinuclear accentuation. There was occasional faint expression in the endothelial cells and lymphocytes. Weak COX-1 expression was also detected in the luminal glandular cells, but in only seven samples (23%), in which the expression was heterogeneous and cytoplasmic granular/diffuse.

Figure 1.

Immunohistochemical detection of COX-1 and COX-2 expression in BPH and prostate cancer. In BPH, COX-1 was expressed (A) in the stroma but not in the glandular epithelium and COX-2 was expressed (B) in the glandular epithelial cells but not in the stroma. The encircled area at increased magnification in (C) shows basolateral COX-2 expression. In prostate cancer COX-1 was expressed (D) both in the stroma and the neoplastic epithelium (arrow). In contrast, COX-2 was expressed in the neoplastic epithelial cells (arrow) but not in the stroma (E). Membranous expression is seen in the adjacent non-neoplastic glandular epithelium. × 200 (A, B, D and E); × 500 (C). S, stroma; g, gland.

In contrast, there was constitutive membranous expression of COX-2 in the luminal glandular cells of all the BPH samples, which was more intense basally and basolaterally ( Fig. 1B,C). The intensity of expression decreased toward the apices. In three samples (10%) there was also weak heterogeneous focal cytoplasmic expression. There was no COX-2 expression in the stroma ( Fig. 1B).

In prostate cancers, COX-1 expression was marginally greater in tumour cells than in benign glands, within the same section. There was weak to moderate cytoplasmic expression in 61 samples of prostate cancer (74%) which was heterogeneous ( Fig. 1D). COX-1 expression in the fibromuscular stroma was similar to that in the BPH samples ( Fig. 1A,D). There was no difference in the levels or patterns of expression in the stroma surrounding malignant cells compared with that juxtaposed with benign glands in the same section. COX-1 expression in the malignant cells did not correlate with tumour grade (P = 0.72).

In contrast, COX-2 expression in prostate cancer was heterogeneous, cytoplasmic, moderate to strong (and increased with increasing grade) and granular/diffuse ( Fig. 1E). COX-2 intensities in the malignant and benign glands were significantly different (P = 0.008; Table 1). The distribution of glandular staining also changed, from membranous in BPH ( Fig. 1C) to cytoplasmic in prostate cancers ( Fig. 1E) (P < 0.001; Table 1). The fibromuscular stroma of BPH and prostate cancer was consistently COX-2-negative. The intensity of COX-2 expression increased with increasing tumour grade (P < 0.001; Table 1). The shift from membranous to cytoplasmic within the cell also correlated with increasing tumour grade (P = 0.01; Table 1).

Table 1.  The intensity of COX-2 cytoplasmic expression in benign and malignant prostates, with grade and localization
CharacteristicBPHProstate cancer; differentiation
No. of samples30103240
Pattern of cytolocalisation
Cytoplasmic expression score

On immunoblotting, there was up to four times more COX-2 protein in prostate cancer than in BPH samples, but no detectable difference in COX-1 levels between BPH and prostate cancer. The results from six samples (three BPH and three prostate cancers) are shown in Fig. 2.

Figure 2.

An immunoblot showing COX-1 and COX-2 proteins in three prostate cancers (lanes 1–3) and three BPH samples (lanes 4–6). Equal amounts of protein (30 µg/lane) were loaded onto an 8% SDS-polyacrylamide gel, electrophoresed and subsequently transferred onto nitrocellulose membrane. The blots were probed overnight with COX-1 or COX-2 mAbs. Purified ovine COX-1 and COX-2 were used as electrophoresis standards (lane 7). There are single bands for both COX-1 and COX-2 at 70 and 72 kDa, respectively. Densitometric analysis of the blots showed four times as much COX-2 protein in prostate cancer than in BPH samples, while COX-1 levels were similar in both.


This study showed that COX-1 and COX-2 are differentially expressed in benign prostates, and that the expression of COX-2 differed significantly between BPH and prostate cancer samples. COX-2 was expressed in luminal glandular epithelial cells of BPH and in the neoplastic epithelial cells of prostate cancers. However, the level of expression in the epithelial cells of prostate cancers was significantly higher than that in BPH. The results of IHC were supported by immunoblotting. The intracellular localization of COX-2 in the epithelial cells of BPH and prostate cancer also differed. While in BPH it was mainly localized at the basal and basolateral cell membrane, in prostate cancers it was predominantly cytoplasmic. In contrast, COX-1 protein expression was not significantly different between BPH and prostate cancer; in both tissues staining was primarily stromal, with variable weak cytoplasmic expression in the epithelial component. The expression of COX-1 on IHC did not vary (P = 0.72), while COX-2 expression increased with grade (P < 0.001).

The present findings closely parallel those of Tremblay et al.[21], who recently reported a study of COX-1 and -2 protein expression using IHC and immunoblotting in a canine model of prostate cancer. They reported constitutive COX-1 expression in stromal fibroblasts and vascular endothelium in both benign and malignant prostates, and faint COX-1 expression in the neoplastic cells of some tumours. In contrast, COX-2 was only detected in the neoplastic cells in 75% of their cases. The authors did not assess the relationship between COX-2 expression and tumour grade. Tremblay et al. found no COX-2 expression in benign epithelial cells, in contrast to the present findings. This difference might be attributable to the following factors: (i) this study utilized mAbs to both COX-1 and COX-2 were, whereas Tremblay et al. used polyclonal antibodies; (ii) no antigen-retrieval method was applied to the canine samples; and (iii) COX immuno-expression in human and dog prostates may be species-specific.

COX-2 was expressed in cell membranes in the luminal epithelial cells of BPH, which is consistent with the co-localization of the enzyme and its substrate. Phospholipids such as linoleic acid are major constituents of the lipid bilayer of the cell membrane and linoleic acid is a precursor of arachidonic acid, from which PGs are synthesized [23]. It is unclear why in benign prostate COX-2 localizes to the membrane and COX-1 to the cytoplasm. This feature of COX-2 expression was lost during cellular transformation, as the expression became cytoplasmic. The observed patterns of expression cannot be ascribed to differences in either fixation or methods of detection, as both patterns were maintained in sections containing benign glandular components (membranous) and prostatic adenocarcinoma (cytoplasmic) together.

The present finding of significantly increased levels of COX-2 in prostate cancer is consistent with data showing that in human prostate cancers, radiolabelled arachidonic acid is converted to PGE-2 at a rate almost 10 times higher than that observed in benign prostates [24]. We therefore conclude that COX-2 rather than COX-1 is likely to be responsible for this increased conversion rate.

There are many mechanisms by which COX-2 may play a role in carcinogenesis, and some or all of these may be involved in prostate cancer development and progression. Many of these are likely to result fromCOX-2-induced increases in PG synthesis. Evidence that increased synthesis of PGs has both growth-promoting and positive-feedback effects in prostate cancer comes from a study by Tjandrawinata et al.[25]. They showed that treatment of prostate cancer cells with exogenous PGE-2 resulted not only in increased mitogenesis that was inhibited by the NSAID flubiprofen, but also inCOX-2 up-regulation. Over-expression of COX-2 has been shown to up-regulate Bcl-2 expression, with an associated decrease in apoptosis [26]. Consistent with this, the human prostate cancer cell line LNCaP, which over-expresses COX-2, shows induction of apoptosis and down-regulation of Bcl-2 expression when treated with NS-398, a selective inhibitor of COX-2 enzyme function [27]. Bcl-2 expression has been closely associated with the androgen-independent phenotype of prostate cancer [2] and it therefore represents a potential pathway through which COX-2 may induce progression of prostate cancer to an androgen-independent state.

Other effects of COX-2 over-expression that may contribute to the malignant phenotype include decreased expression of E-cadherin, with consequent loss of cell-to-cell adhesion [26], over-expression of matrix-metalloproteinase 2 with an associated increase in invasiveness [28], and modulated production of angiogenic factors by cancer cells [29]. COX-2 over-expression in cancer cells has also been shown to inhibit immune surveillance [30] and increase metastatic potential [28]. Furthermore, COXs may play a role in the bioactivation of several polycyclic aromatic hydrocarbons and aromatic amines, two classes of carcinogens which induce extrahepatic neoplasia [31].

To our knowledge the present is the first study analysing the expression of COX-1 and COX-2 proteins in benign and malignant human prostates. Given that the actions of COX-2 induce features of the malignant phenotype, there is a powerful argument that COX-2 should be evaluated further as a promising therapeutic target, both in prostate cancer and in other cancers.


This work was partly funded by a grant from The Friends of Hammersmith Hospital. We thank Dr R.W. Stirling, Consultant Histopathologist, West Middlesex University Hospital, London for providing us with some of the prostate cancer samples. We also thank Prof. A. Wanji, University of Hong Kong for reviewing the manuscript and constructive suggestions.


S. Madaan, MS, FRCS, Clinical Research Fellow.

P.D. Abel, ChM, FRCS, Reader in Urology and Honorary Consultant.

K.S. Chaudhary, PhD, Senior House Officer.

R. Hewitt, PhD, Clinical Scientist.

M.A. Stott, FRCS, Consultant Urologist.

G.W.H. Stamp, FRCPath, Chairman and Professor of Histopathology.

E.N. Lalani, PhD, MRCPath, Reader in Molecular & Cellular Pathology and Honorary Consultant Histopathologist.