Expression of cyclooxygenase-2 and peroxisome proliferator-activated receptor-γ and levels of prostaglandin E2 and 15-deoxy-Δ12,14-prostaglandin J2 in human breast cancer and metastasis


  • The contents of this article are solely the responsibility of the authors and do not necessarily represent the official view of the National Cancer Institute.


Cyclooxygenase-2 (COX-2) expression and peroxisome proliferator-activated receptor-γ (PPARγ) inactivation are linked to increased risk of human breast cancer. The purpose of our study was to examine the relationship between COX-2 (with the resulting prostaglandins E2, PGE2) and PPARγ (and its natural endogenous ligand 15-Deoxy-Δ12,14-prostaglandin J2, 15d-PGJ2) at various stages during the development of human breast cancer and its progression to metastasis. Human breast tissue specimens were collected from normal breasts or from individuals with fibrocystic disease and served as controls (n = 22). Tissues were also collected from uninvolved (n = 25), tumor (n = 25) and lymph node metastasis (n = 15) regions from breast cancer patients. COX-2 and PPARγ mRNA expression were increased and downregulated, respectively, in tissues from cancer patients compared to controls. Metastatic tissues tended to have higher alterations compared to non-metastatic tissues (p < 0.05). These altered expressions in COX-2 and PPARγ were paralleled by increases in the tissue levels of PGE2 and decreases in 15d-PGJ2. A significant inverse correlation was found between PGE2 and 15-d-PGJ2 (r = −0.51, p < 0.05). Significant correlations (p < 0.05) were also obtained between COX-2 and PPARγ mRNA (inverse, r = −0.72) and between COX-2 and PGE2 (direct, r = 0.68). Increases in COX-2 mRNA expression and levels of PGE2 and down-regulation of PPARγ mRNA expression and 15d-PGJ2 levels were characterized as predictors of breast cancer risk (p < 0.05). Our results suggest that the altered expression of COX-2 and PPARγ and the subsequent modulation in the tissue levels of PGE2 and 15-d-PGJ2 may influence the development of human breast cancer and its progression to metastasis. © 2002 Wiley-Liss, Inc.

The high prevalence of breast cancer1 and the limited options for treatment provide a strong rationale for identifying new, selective molecular targets that can be nutritionally or pharmacologically modulated and, thereby, offer a potential for chemoprevention. Among the regulatory molecules that have been characterized as holding great promise for breast cancer prevention are cyclooxygenase-2 (COX-2) and peroxisome proliferator-activated receptor-γ (PPARγ).2 Cyclooxygenase is the rate-limiting enzyme in prostaglandin (PG) synthesis, of which 2 isoforms were identified: the constitutive COX-1 and the inducible COX-2. PGs produced by COX-1 mediate various physiological responses, whereas PGs produced by COX-2, predominantly PGE2, induce inflammation and are potent mediators of a number of signal transduction pathways that modulate cell adhesion and growth and are implicated in cancer development.3 COX-1 and COX-2 are the primary targets of non-steroidal anti-inflammatory drugs (NSAIDs), which inhibit the activity of these enzymes as a major mode of their anti-tumorigenic action.4 Recent findings, however, suggest that some of the cancer preventive potential of NSAIDs is mediated via their role as ligands of PPARs.5 Activation of PPARs, members of the nuclear hormone receptor family, plays a key role in controlling cell differentiation and apoptosis.6, 7 The gene encoding PPARγ is transcribed into 3 mRNA species, PPARγ1, γ2 and γ3, derived from alternative splicing and promoter usage.7, 8 Little is known about the expression pattern of PPARγ3, except it is the predominant PPARγ type in macrophages;9 whereas PPARγ2 is confined to adipose tissue.10 PPARγ1, however, is the ubiquitously expressed PPARγ subtype.7, 10 The biologically active cyclopentenone PG, 15-deoxy-Δ12,14-prostaglandin J2 (15d-PGJ2) is an endogenous ligand of PPARγ.11 15d-PGJ2 can inhibit COX-2 activity12 and was found to induce apoptosis, inhibit proliferation and prevent the growth of human breast cancer cells in the nude mouse model.11, 13

Independent studies have shown that COX-2 and PPARγ are induced14, 15, 16 and inactivated,17 respectively, in human breast cancer. Inhibition of COX-2, e.g., by celecoxib, a COX-2 specific inhibitor18, 19 and activation of PPARγ, e.g., by GW7845, a PPARγ-specific ligand,20 inhibited the development of rat mammary gland carcinogenesis. Additionally, various in vitro studies have demonstrated that COX-2 selective inhibitors21 and PPARγ-ligands13, 22 can significantly attenuate the growth of human breast cancer cells. Taken together, these observations suggest that targeting COX-2 or PPARγ can provide an effective approach for breast cancer prevention in humans. Attempts to use agents that specifically target either or both molecules for cancer chemoprevention, however, should be preceded by elucidating the profile of COX-2 and PPARγ involvement in breast cancer and determining the relationship between the 2 target molecules at various stages of this malignancy.

To provide information on the association between COX-2 and PPARγ and human breast cancer development, the present study examines the expression of both genes and the corresponding levels of PGs (PGE2, product of COX-2 and 15d-PGJ2, endogenous PPARγ-ligand) in human breast tissues at various cancer stages. Elucidating the interrelationship between these factors in breast carcinogenesis could facilitate the evaluation and feasibility of targeting either (or both) molecules as an effective strategy to prevent breast cancer in asymptomatic individuals.


COX, cyclooxygenase; DMBA, 7,12-dimethylbenz[a] anthracene; 15d-PGJ2, 15-deoxy-Δ12,14-PGJ2; MNU, N-methyl-N-nitrosourea; NFκB, nuclear factor-κ B; NSAIDs, non-steroidal anti-inflammatory drugs; PGs, prostaglandins; PPARγ, peroxisome proliferator-activated receptor γ.


Study subjects

A total of 87 human breast tissue specimens (from 47 patients) were obtained from the Human Tissue Bank of Fox Chase Cancer Center and were approved for processing by the Center's Institutional Review Board. The tissue specimens were procured by the Tissue Bank staff from 1990–2000 and were stored at −80°C after collection. Table I presents the characteristics of study subjects and the types of examined tissue specimens. Information such estrogen and progesterone receptors status and menopausal status were not available. Samples collected from subjects with no cancer but diagnosed with fibrocystic disease, fibroadenomas or having reactions to minor surgical procedures (n = 22) were designated as controls. Cases (n = 25) were divided into 2 groups: 1) breast cancer patients with invasive and in situ ductal carcinoma (Grades II/III) but no sign of metastasis (n = 10), from which pairs of uninvolved (to serve as preneoplastic tissue) and tumor specimens were collected; and 2) patients with invasive and in situ ductal and lobular carcinoma and documented metastasis in at least 2–3 lymph nodes (n = 15), from which both uninvolved, tumor and metastatic tissues were obtained. The total number of tissue specimens processed was, therefore, n = 87 samples divided as controls (n = 22), uninvolved (n = 25), tumor (n = 25) and metastasis (n = 15). Histopathological analyses were also obtained for the study subjects from the Human Tissue Bank database and were linked to the studied markers.

Table I. Characteristics of Study Subjects and Types of the Examined Human Breast Tissue Specimens
GroupAge (mean ± SD)PathogenesisnCollected tissue specimens1
  • 1

    Number in parentheses represents the number of specimens collected from each tissue type.

  • 2

    Normal tissues collected from subjects with fibrocystic diseases, fibroadenomas and normal tissues from reaction to minor surgical procedures.

  • 3

    Tumor tissue obtained from cancer patients with invasive and in situ ductal carcinoma (Grades II and III).

  • 4

    Metastatic tissues were collected from breast cancer patients with invasive and in situ ductal and lobular carcinoma and metastasis in at least 2–3 lymph nodes.

Controls50 ± 11Normal222Normal
Cancer cases58 ± 13No metastasis310Uninvolved (10) and tumor (10)
  With lymph node metastasis415Uninvolved (15), tumor (15) and matastasis (15)
All subjects55 ± 13 47Normal (22), uninvolved (25), tumor (25) and metastasis (15)

COX-2 and PPARγ expression

To examine mRNA expression, we used PCR primers that have already been described for COX-2,23 PPARγ124 and GAPDH23 (COX-2 5′ primer: TTC AAA TGA GAT TGT GGG AAA ATT GCT, 3′ primer: AGA TCA TCT CTG CCT GAG TAT CTT; PPARγ1 5′ primer: CCG CTC GAG CGG GCC GCC GTG GCC GCA GAA, 3′ primer: AGG AAT TCA TGT CAT AGA TAA CG; GAPDH 5′ primer: CCA CCC ATG GCA AAT TCC ATG GCA, 3′ primer: TCT AGA CGG CAG GTC AGG TCC ACC). These primer pairs yield amplified products of 305 bp for COX-2, 790 bp for PPARγ1 and 593 bp for GAPDH. Total RNA from the human breast tissue specimens, isolated using TRI reagent (Sigma, St. Louis, MO), was subjected to RT-PCR analysis using Titanium™ One-step RT-PCR protocol (Clontech, Palo Alto, CA) according to the manufacturer's instructions and following the cycling conditions described previously for COX-2,23 PPARγ24 and GAPDH.23 Briefly, RT-PCR reactions were carried out in a buffer containing 1 μg RNA, Moloney murine leukemia virus-reverse transcriptase (MMLV-RT), 20 μM oligo(dT) primers, 40 mM Tricine, 20 mM KCl, 3 mM MgCl2, 0.2 mM each dNTP, 1× Taq enzyme mix provided by the manufacturer, 20 U recombinant RNase inhibitor and 45 μM PCR primer mix in a total volume of 50 μl. Using a hot-lid Gene-Amp 9700 thermocycler (Perkin Elmer, Norwalk, CT), RNA was reverse transcribed at 50°C for 60 min followed by 5 min at 95°C. PCR cycling programs consisted of 25 cycles of 94°C for 1 min, 55°C for 1 min and 72°C for 1 min for COX-2 and GAPDH and 94°C for 30 sec, 65°C for 30 sec and 68°C for 1 min for PPARγ and were followed by 68°C for 2 min. Under these conditions, the yield of the amplified products was linear with respect to the input RNA and cycle number. The assay resolved a 2-fold difference in the amount of input RNA. Furthermore, the yields were linear when the PCR reactions were carried out for 20, 25 or 30 cycles. PCR products were separated in 2% agarose gels and visualized by ethidium bromide staining. We ran negative controls to rule out contamination of RNA with genomic DNA. In this control, we used PCR enzyme mix without RT (Titanium™ Taq DNA polymerase, Clontech) instead of the RT-Titanium Taq enzyme mix included in the one-step kit. Additionally, the standard one-step reaction was preceded by heating at 94°C for 5 min to inactivate reverse transcriptase before using the thermal cycling procedures. This reaction ensured that no cDNA synthesis would occur before inactivation. To rule out other sources of contamination, RT-PCR reactions were carried out in mixtures containing no RNA. Digitized images of the stained cDNA products were captured as 8-bit digital TIFF files by using a DC290 Digital Camera (Eastman Kodak Co., Rochester, NY). The intensity of each band was measured using Kodak Digital Science 1D-Image Analysis 3.6 (Eastman Kodak Co.). Transcripts of COX-2 or PPARγ present were integrated and normalized to the corresponding level of GAPDH and were expressed as arbitrary density units (ADU). Using anti-human COX-2 and PPARγ antibodies (Santa Cruz Biotech, Santa Cruz, CA), protein expression was carried out in a representative set of specimens (n = 6–8) from control, tumor and metastatic tissues (uninvolved was not analyzed), using Western blot analysis, essentially as described previously for COX-215 and PPARγ.17 Equal protein loading (30 μg/lane) was checked against the “house-keeping” COX-1 protein.

Levels of PGE2 and 15d-PGJ2

Analysis of the human breast tissue levels of PGE2 and 15d-PGJ2 were carried out as described earlier by enzyme immunoassay (EIA).23 Briefly, tissue specimens (100 mg) were minced, sonicated in 0.1 M Tris-HCl buffer (pH 7.4) and centrifuged at 800g for 15 min to separate cell debris and the fat layer. The resulting supernatant was acidified to pH 3.5 with glacial acetic acid and extracted 3 times with 3 vol of ethyl acetate. Extracts were combined, evaporated to dryness and resuspended in 200 ml of the EIA buffer and stored at −80°C until analyzed. EIA for PGE2 and 15d-PGJ2 was carried out using a Correlate-EIA kit (Assay Designs Inc., Ann Arbor, MI) according to the manufacturer's instructions. The cross- reactivities of these EIA kits for a number of eicosanoids were determined by the manufacturer. The competitive PGE2 EIA analysis exhibits cross-reactivity with PGE1 (70%), PGE3 (16.3%) and PGF (1.4%) whereas 15d-PGJ2 analysis cross-reacts with PGJ2 (49%), Δ12-PGJ2 (5.9%) and PGD2 (4.9%). Color intensity was measured at 405 nm using SpectraMax 250 microplate reader (Molecular Devices, Sunnyvale, CA) and the levels of PGE2 and 15d-PG2 were expressed as ng/g tissue. Standard curves and positive and negative controls were generated for PGE2 and 15d-PGJ2 and assayed simultaneously with the samples. We did not observe any inter- or intra-day differences in the levels of PGE2 or 15d-PGJ2, based on the data from the standard curves or the positive controls.

Statistical analysis

Given the 4 groups of tissue examined in our study, i.e., 1) control, 2) uninvolved, 3) tumor and 4) metastasis; the expression of COX-2 and PPARγ mRNA and the levels of PGE2 and 15d-PGJ2 were analyzed by Student's t-test to evaluate the differences between Groups 2–4 compared to controls (Group 1). Comparisons were also made between all non-cancerous (1+2) and cancerous (3+4) tissues. ANOVA test was also carried out among the groups for COX-2 and PPARγ expression and for the levels of PGE2 and 15d-PGJ2. Correlation analyses between: 1) COX-2 and PPARγ mRNA expression; 2) COX-2 mRNA and levels of PGE2; 3) PPARγ mRNA and levels of 15d-PGJ2; and 4) PGE2 and 15d-PGJ2 were carried out across tissue specimens and within each tissue type using Spearman's Test. Correlations were also evaluated between cancer status (tissue types were ranked 1–4, as above) and the examined markers. The expression of COX-2 and PPARγ mRNA and the levels of PGE2 and 15d-PGJ2 were evaluated as biomarkers for breast cancer risk using ordinal logistic regression analysis. The association between these biomarkers was examined with stages (i.e., control, preneoplastic, tumor and metastasis). The variables were considered one at a time as a predictor-of-risk. The association of stage with base-line covariates was limited to age, a confounding factor that had a statistically significant difference between cases and controls (p < 0.05). Each biomarker was, therefore, adjusted for age when examined as a predictor-of-risk. All tests were carried out using the SAS System release 8.0 (SAS Institute, Cary, NC).


COX-2 and PPARγ mRNA expression

The expression of COX-2 and PPARγ was evaluated in human breast tissue specimens (Fig. 1a). Levels of COX-2 mRNA and its detection frequency were higher in tissues obtained from cancer cases compared to controls. Over 60% of the tumor tissue (with or without metastasis) showed induction in COX-2 transcripts compared to only 25% in the preneoplastic uninvolved tissue and <15% in the control tissue (Fig. 1b). Expression of COX-1 mRNA was also measured in these samples, but no significant variation in the transcript levels was noted among different tissue types (data not shown). Control and uninvolved tissues showed similar levels of COX-2 mRNA, but they were lower than tumors or metastatic tissues (p < 0.05). In tissue samples positive for COX-2 mRNA, levels of the transcripts correlated positively with the progression from normal to tumor and metastatic breast cancer (r = 0.8, p < 0.05).

Figure 1.

Expression of COX-2 and PPARγ in human breast tissue. mRNA expression was measured by RT-PCR. Single stranded cDNA was reverse transcribed from total RNA and used in one-step analysis for PCR amplification with specific primers. Protein expression was measured by Western blot analysis. Representative RT-PCR and Western blots (a) are shown for COX-2, PPARγ and housekeeping GAPDH (mRNA) or COX-1 (protein) in control (C), uninvolved (U), tumor (T) and metastatic (M) breast tissue. The average levels of COX-2 (b) and PPARγ (c) mRNA in human breast tissue specimens was quantified as arbitrary density units (A.D.U.) by measuring the band intensity relative to GAPDH. Numbers in parentheses represent the detection frequency of COX-2 mRNA in each tissue type. Data are means ± SD for n number of tissue specimens in each group (see Table I). *Statistically significant difference (p < 0.05) from the control and uninvolved tissue types and from each other using multiple Student's t-tests. Significant differences among groups (p < 0.05) was also observed using ANOVA test.

In contrast to the tissue-type dependent detectability of COX-2, PPARγ1 transcripts were observed in all examined samples but with varying intensities. Levels of PPARγ1 mRNA were significantly lower in tumors and metastatic breast tissues compared to control or uninvolved tissues from which they arise (Fig. 1c, p<0.05). This downregulation of PPARγ1 in cancerous tissue vs. controls was reflected by a significant inverse correlation between the transcript levels and the progression from normal to tumor and metastatic states (r = −0.67, p < 0.05). A significant inverse correlation (r = −0.72, p < 0.05) was also obtained between COX-2 and PPARγ mRNA measured in the same tissue specimens (Fig. 2a, Table II).

Figure 2.

Correlations between COX-2 and PPARγ expression (a) and levels of PGE2 and 15d-PGJ2 (b). Correlation analysis was carried out using Spearman's test. (a) Total of 33 samples from the different cancer stages had detectable COX-2 mRNA.

Table II. Correlational Analyses Between COX-2, PPARγ, PGE2 and 15D-PGJ2 in Human Breast Cancer
  • 1

    Values represent correlation coefficient (r) using Spearman's test. Value for association between COX-2 and PPARγ mRNA was estimated only from samples positive for COX-2. All values were statistically significant (p < 0.05) except as noted.

  • 2

    p = 0.081.


Levels of PGE2 and 15d-PGJ2

Tissue levels of PGE2 and 15d-PGJ2 were analyzed by EIA. Concentrations of PGE2 in the control tissue were approximately 1.5-, 5.5- and 11-fold lower than the levels detected in uninvolved, tumor and metastatic tissues, respectively (Fig. 3a). In contrast, control tissues contained 2-, 15- and 70-fold higher levels of 15d-PGJ2 compared to uninvolved, tumor and metastasis tissues, respectively (Fig 3b). In the control tissue, PGE2 was 6-fold higher than 15d-PGJ2. This difference was wider in the cancerous tissues as the levels of PGE2 were increasing whereas those of 15d-PGJ2 were decreasing. This relationship between PGE2 and 15d-PGJ2 was reflected by a significant inverse correlation between the two markers (r = −0.51, p < 0.05) when measured in the same tissue specimens (Fig. 2b, Table II). Compared to normal individuals or breast cancer cases, patients with metastasis contained significantly higher levels of PGE2 and lower levels of 15d-PGJ2 (Fig. 4). In patients with metastasis, tissue levels of PGE2, but not 15d-PGJ2, showed significant gradual increases from uninvolved to tumor to metastatic tissue (Fig. 4). Increased PGE2 and decreased 15d-PGJ2 were correlated with the progression from normal to tumor and metastatic states (r = 0.8 for PGE2 and −0.6 for 15d-PGJ2, p < 0.05).

Figure 3.

Levels of PGE2 (a) and 15d-PGJ2 (b) in human breast tissue specimens. Differences in the levels of PGE2 and 15d-PGJ2 between the control (C), uninvolved (U), tumor (T) and metastatic (M) groups were described using box-and-whisker plot analysis. The box-and-whisker plot represents a characterization of the data showing the 10th and 90th percentile of data (end of the whisker), the 25th and the 75th percentile (ends of the box), the 50th percentile (solid line within the box), the average of the data (dashed line) and the outliers, i.e., data values beyond the 10th and 90th percentile (•). The differences between U and T vs. C or M were statistically significant (p < 0.05) using multiple Student's t-tests. Significant differences was noted among groups (p < 0.05) using ANOVA test.

Figure 4.

Distribution and levels of PGE2 (a) and 15d-PGJ2 (b) in different tissue types from breast cancer patients with and without metastasis. Uninvolved (U), tumor (T) or metastasis (M) tissues were collected from the same patient and compared to each other and to controls (C). Values represent means ± SEM for n number of samples from the different types of tissue specimens collected in each group (see Table I). In each panel, all comparisons (t-test) were significantly different from each other (p < 0.05), except: (a vs. b) and (d vs. b or c) in (a) and (a vs. b), (b vs. c), (d vs. e or f) and (e vs. f) in (b).

Tissue level of PGE2 was significantly associated with COX-2 mRNA expression (r = 0.68, p < 0.05, Table II). In contrast, levels of 15d-PGJ2 and PPARγ1 were marginally correlative (r = 0.41, p = 0.081), perhaps due to the fact that subtypes other than PPARγ1 exist, i.e., PPARγ2 that is confined to adipose tissue. COX-2 and PPARγ1 mRNA expression as well as levels of PGE2 and 15d-PGJ2 were evaluated as biomarkers for breast cancer risk using ordinal logistic regression analysis adjusted for age. Each of the 4 markers was a predictor-of-risk, where increased COX-2 expression, elevated PGE2, downregulation of PPARγ1 mRNA or decreased 15d-PJG2 was linked to increased risk of human breast cancer (p < 0.05, data not presented).


Separate studies have shown that induction of COX-2 and inactivation of PPARγ may occur during the development and progression of human breast cancer. Inhibition of COX-2 or activation of PPARγ prevents mammary carcinomas in experimental animals and may exhibit a similar effect in humans. Therefore, elucidating the relationship between the two molecules may be relevant for breast cancer preventive regimens that employ combinational targeting of multiple, complementary cancer-related pathways. In the present study, we have investigated the expression of COX-2 and PPARγ in human breast tissue specimens at different stages of cancer. These analyses were accompanied by and inter-correlated with measures of the endogenous levels of PGE2, a major COX-2 product and 15d-PGJ2, an endogenous PPARγ-ligand. Over-expression and increased detection frequency of COX-2 mRNA were observed in breast cancer tissues compared to controls (Fig. 1a,b). These results are supported by previous reports showing that COX-2 mRNA14 and protein15, 16 expressions are increased in human breast cancer and linked to poor prognosis.16 In the tumor tissue, COX-2 detection frequency varies between 5%,15 38%,16 60% (our study) and 100%.14 This variation was suggested to be due to the possibility that COX-2 induction may be confined to cancer cases over-expressing HER-2.22

Increases in COX-2 expression are usually accompanied by increases in COX enzymatic activity and enhanced capacity of the tissues to synthesize PGs.25 In the present study, levels of PGE2 were significantly higher in breast cancer cases compared to controls (Fig. 3a) and were positively associated with COX-2 mRNA expression (r = 0.68, p < 0.05, Table II). Elevated levels of PGE2 were previously reported in human breast cancer26, 27, 28 with a tendency of tumors with worse prognostic grades to have more PGs.29 Direct evidence for the role of COX-2 was demonstrated when COX-2 transgenic animals had higher rates of mammary carcinomas compared to wild-type carriers.30 COX-2 and PGE2 may foster cancer development by a variety of mechanisms including: 1) controlling cell growth and proliferation; 2) rendering cells resistant to apoptosis; 3) mediating tumor-associated immune suppression; 4) promoting cell adhesion to the extra-cellular matrix which augments metastasis; and 5) inducing angiogenesis.21, 31, 32 Therefore, COX-2 and PGE2 may play a multifunctional role in cancer development as well as in its progression to metastasis. Indeed, our findings that cancer patients with metastasis exhibited higher expression and detection frequencies of COX-2 mRNA (Fig. 1) and elevated PGE2 compared to non-metastatic cases (Figs. 3a,4a) implicate a possible role of a COX-2-mediated pathway in metastasis. In support, increased COX-2 expression and PGE2 levels were associated with metastatic cancer in murine mammary glands33 and in human breast cancer cells.34 Moreover, PGE2 levels were higher in metastasized cases compared to non-metastatic and associated with a shorter post-surgical survival time.35 In fact, data showing that breast tumors with high PGE2 content tend to have histological evidence of lymphatic, nodal and vascular invasion led Rolland et al.36 to propose that high PGE2 can be a marker for high metastatic potential. Our results support this notion by demonstrating that COX-2 expression and PGE2 levels are positively associated with tumorigenesis (r = 0.8, p < 0.05) and can provide valuable markers of breast cancer risk (p < 0.05).

Inhibition of COX-2 and PGE2 synthesis by regular NSAIDs, such as indomethacin37 and flurbiprofen38 or COX-2 specific inhibitors, such as celecoxib,18, 19 inhibited both the initiation and promotion stages of carcinogen-induced rat mammary carcinogenesis. Several studies, however, have demonstrated unequivocally that, in addition to their COX-dependent mechanisms, the effects of NSAIDs may be mediated via a variety of COX-independent pathways. Particular attention has been paid recently to the role of NSAIDs as ligands to PPARγ5 because of the involvement of these receptors in the genesis of human breast cancer. PPARγ expression is downregulated in rodent mammary gland carcinomas compared to normal tissues.39 Moreover, semi-quantitative analyses indicated that normal breast tissues contain more than 2-fold higher levels of PPARγ protein compared to tumors.17 In the present study PPARγ1 mRNA, the ubiquitously expressed subtype, decreased as breast cancer developed and progressed to metastasis (Fig. 1a,c, r = −0.67 between PPARγ mRNA and cancerous status, p < 0.05). Inactivation of PPARγ in breast cancer can be, partly, due to the lower tissue levels of 15d-PGJ2 as detected in cancer cases (Fig. 3b). Inhibition of 15d-PGJ2 synthesis in patients with metastasis compared to non-metastatic cases (Fig. 4b) may implicate its importance in cancer chemoprevention. Downregulation of PPARγ expression and lower tissue levels of 15d-PGJ2 were identified as markers of breast cancer risk (p < 0.05). In support of this observation, levels of COX-2 (and PGE2) and PPARγ (and 15d-PGJ2) in controls were, respectively, lower and higher than those in uninvolved tissues from breast cancer patients. These results may be relevant to the role of COX-2 and PPARγ in the risk of breast cancer. Furthermore, the implication of COX-2 and PPARγ in the risk to metastasis can be substantiated from comparing the patten of their expression in the uninvolved tissues from cancer patients (lower COX-2 and higher PPARγ) to that in the uninvolved tissues from cancer patients with metastasis.

A wide variety of PPARγ-agonists, including 15d-PGJ2, were found to inhibit cell growth and colonogenic capacity and induce terminal differentiation and apoptotic cell death in mammary cancer cells, thereby promoting the emergence of phenotypic changes associated with a more differentiated and less malignant status.13, 22, 40 15d-PGJ2 is synthesized when PGD synthase (PGDS) catalyzes the isomerization of PGH2 (formed through the action of COX on arachidonic acid) to PGD2, which undergoes a spontaneous dehydration to form 15d-PGJ2.41 It is well established that during inflammation, PGs are formed to counterbalance each other where PGE2 mediates the inflammatory reactions42 whereas 15d-PGJ2 contributes to its eventual resolution.43 These observations may be relevant to the inverse correlations obtained between PGE2 and 15d-PGJ2 in human breast tissue specimens (Fig. 2b, r = −0.51, p < 0.05, Table II). In support, among various PGs, only 15d-PGJ2 exerted a marked inhibition on COX-2 expression and PGE2 synthesis.12, 44 In cancer cells, the counterbalancing effect of PGE2 and 15d-PGJ2 is unlikely to take place because of the significant inhibition of PGDS45 and the subsequent lower synthesis of 15d-PGJ2. This relationship may explain the significant down-regulation of 15d-PGJ2 synthesis and the subsequent decrease in PPARγ expression in breast cancer. It may be relevant here to mention that the ratio of PGE2:15d-PGJ2 was 6 in the control tissue (1,000 in tumors), which is comparable to a urinary ratio of 10 between derivatives of PGD246 and PGE247 in healthy women.

It is known that normal mammary tissues contain a high proportion of adipose component (that expresses PPARγ) whereas tumor tissues are not. This observation is highly relevant to the interpretation of the present results because we used preparations from blocks of tissues without separating the epithelial from the adipose tissue. Various studies, however, have demonstrated that mammary epithelium and adipose tissues predominantly express PPARs,48, 49 where PPARγ1 is the ubiquitously expressed subtype whereas PPARγ2 is only confined to adipose tissue.7 Similarly, COX-2 expression and increased levels of PGE2 occur in both stromal50 and epithelial51 cells of mammary tumor and normal tissues.52 Indeed, levels of COX-2 and PPARγ1 mRNA and protein were comparable in both mammary stromal and epithelial cells50 (A.F. Badawi, unpublished data). A paracrine loop involving PGs has been demonstrated between mammary epithelial and stromal compartments,53, 54 so that COX-2 induction in stromal cells with concomitant increased PGE2 synthesis or reduced 15d-PGJ2 could have additional effects on PPARγ and on the tumorigenic potential of mammary epithelial cells.

In breast cancer, COX-2 and PPARγ modulate different signaling pathways. The correlation between the 2 molecules (and between PGE2 and15d-PGJ2) observed in the present study (Fig. 2, Table II) implicates that COX-2 and PPARγ may act coordinately to influence breast tumorigenesis. Direct evidence for the coordinated actions between PPARγ and COX-2 has been substantiated by studies showing that activation of the PPAR pathway induces apoptosis and inhibits inflammation by inhibiting COX-2 expression.44, 55 Additionally, a variety of studies have demonstrated that COX-2 inhibitors and PPARγ-ligands influence tumorigenesis by concomitantly modulating the expression of several target genes and transcription factors. For example, activation of PPARγ17, 44, 56 or inhibition of COX-25, 57 antagonizes the activities of transcription factors (NFκB, AP-1, STAT) and modulates the effect of genes controlling cell migration (E-cadherin, β-catenin), apoptosis (bcl-2, bax) and angiogenesis (VEGF).

In conclusion, it appears that COX-2 and PPARγ may contribute to breast cancer induction either directly or via their coordinate effects on various cancer-related genes and transcriptional factors. We noted recently58 that both molecules can be potential targets for combinational chemoprevention of human breast cancer.


The authors thank Dr. C. Spittle for careful reading and Ms. M. Climaldi for her assistance with the manuscript preparation.