In addition to their role in the development and function of the reproductive system, estrogens have significant anti-inflammatory properties. Although both estrogen receptors (ERs) can mediate anti-inflammatory actions, ERβ is a more desirable therapeutic target because ERα mediates the proliferative effects of estrogens on the mammary gland and uterus. In fact, selective ERβ agonists have beneficial effects in preclinical models involving inflammation without causing growth-promoting effects on the uterus or mammary gland. However, their mechanism of action is unclear. The purpose of this study was to use microarray analysis to determine whether ERβ-selective compounds produce their anti-inflammatory effects by repressing transcription of proinflammatory genes. We identified 49 genes that were activated by TNF-α in human osteosarcoma U2OS cells expressing ERβ. Estradiol treatment significantly reduced the activation by TNF-α on 18 genes via ERβ or ERα. Most repressed genes were inflammatory genes, such as TNF-α, IL-6, and CSF2. Three ERβ-selective compounds, ERB-041, WAY-202196, and WAY-214156, repressed the expression of these and other inflammatory genes. ERB-041 was the most ERβ-selective compound, whereas WAY-202196 and WAY-214156 were the most potent. The ERβ-selective compounds repressed inflammatory genes by recruiting the coactivator, SRC-2. ERB-041 also repressed cytokine genes in PBMCs, demonstrating that ERβ-selective estrogens have anti-inflammatory properties in immune cells. Our study suggests that the anti-inflammatory effects of ERB-041 and other ERβ-selective estrogens in animal models are due to transcriptional repression of proinflammatory genes. These compounds might represent a new class of drugs to treat inflammatory disorders.
Menopause is associated with increased production of inflammatory cytokines, such as ILs and TNF-α (1, 2). The elevated cytokine levels likely contribute to the increased incidence of inflammatory diseases after menopause, such as osteoporosis, and neurodegenerative and cardiovascular diseases (3, 4, 5, 6). Because estrogens are known to have anti-inflammatory properties (2), and estrogens in the form of hormone therapy (HT)3 lower cytokine levels in postmenopausal women (2, 7), it seems practical to attempt to prevent inflammatory diseases associated with menopause by replacing estrogens after the onset of menopause. Observational studies supported this rationale for using HT as a chemopreventative intervention by showing that HT decreased osteoporosis, Alzheimer’s disease, and cardiovascular disease (8, 9, 10). It was therefore quite surprising when the Women’s Health Initiative (WHI) trial found that estrogen plus progestin increased the risk of heart disease and dementia (11). However, consistent with observational studies, the WHI did find that HT prevented osteoporosis and fractures (11). Furthermore, recent secondary analysis of data from the WHI found that the increase risk of heart disease occurred mainly in women who started HT long after menopause (12). A reduction in coronary heart disease occurred in women who started HT close to the onset of menopause, which is the typical time HT is prescribed (12). The protective effect of estrogens on osteoporosis and heart disease in younger women suggest that inflammatory pathways should remain an important therapeutic target of estrogens for treating women close to the onset of menopause.
A key to generate selective estrogens to prevent inflammatory diseases is to understand the mechanisms whereby estrogens exert their anti-inflammatory actions. Estrogens inhibit the release cytokines from multiple cell types (13, 14), suggesting that proinflammatory genes are major targets for the two forms of estrogen receptors (ER), ERα and ERβ (15, 16). Although ERs can activate or repress gene transcription, most studies indicate that ERs exert their anti-inflammatory actions by repressing genes that promote inflammation, such as IL-6 and TNF-α (14, 17). Many inflammatory genes are activated by NF-κB transcriptional factors. Estrogens have been shown to inhibit NF-κB binding to the IL-6 promoter (17, 18), suggesting that ERs work by blocking NF-κB-mediated activation of inflammatory genes. In contrast, we found that estrogens did not repress the TNF-α gene by inhibiting the binding of NF-κB, but instead repressed the TNF-α gene by recruiting steroid receptor coactivator 2 (SRC-2), which acts as a transcriptional repressor (19). These results indicate that there are likely multiple mechanisms whereby estrogens exert anti-inflammatory effects, suggesting that estrogenic drugs can be designed to alter various targets involved in the regulation of inflammatory genes.
Although the anti-inflammatory actions of estrogens likely play an important role in the prevention of inflammatory diseases in women, the major property of estrogen that needs to be eliminated is its proliferative effects on the uterus and mammary gland. Clearly, it is important to develop estrogens that lack proliferative effects, but retain their anti-inflammatory actions. One approach to achieve this goal is exemplified by the compound, WAY-169916, which inhibits NF-κB transcriptional activity but is devoid of estrogenic activity on breast cells (20). Another approach is to design estrogens that selectively regulate ERβ transcriptional pathways. Estrogens exert their proliferative effects through ERα as demonstrated by the observation that the ERα knockout mice show little mammary gland development, whereas ERβ knockout mice develop normal mammary glands (21). In contrast, ERβ acts as a tumor suppressor and inhibits the proliferation of breast cancer cells (22, 23, 24). We previously showed that estradiol (E2), acting through ERβ, is also a potent inhibitor of the transcription of the TNF-α gene (25). Based on these findings, we hypothesize that ERβ-selective ligands should retain the anti-inflammatory properties of E2, but lack growth-promoting activity. The development of highly selective ERβ agonists (26, 27) provides a unique opportunity to investigate the anti-inflammatory action of ERβ, which is expressed in immune cells (16, 28). Two such compounds, ERB-041 (2-(3-fluoro-4-hydroxyphenyl)-7-vinyl-1,3 benzoxazol-5-ol) and WAY-202196 (3-(3-fluoro-4-hydroxyphenyl)-7-hydroxynaphthonitrile), have showed potent anti-inflammatory effects in animal models for diseases such as rheumatoid arthritis, inflammatory bowel disease, endometriosis, and/or sepsis, but did not have proliferative effects on the mammary gland or uterus (27, 29, 30, 31, 32, 33, 34). These findings suggest that ERβ-selective compounds might be useful to treat a wide range of inflammatory diseases, including some that are not associated with menopause. However, the molecular mechanisms whereby the ERB-041 and other synthetic ERβ-selective agonists produce their anti-inflammatory effects are not known. The purpose of this study was to investigate whether the synthetic ERβ-selective drugs produce anti-inflammatory effects by causing transcriptional repression of proinflammatory genes.
Materials and Methods
ERB-041 (35), WAY-202196 (34), and WAY-214156 (2,8-dihydroxy-6H-dibenzo[c,h]chromene-4,12-dicarbonitrile) and propyl-pyrazole-triol (36) were obtained from the Wyeth Research compound library. In a competitive radioligand-binding assay, the IC50 of ERB-041 was 5.4 nM and it was >220-fold selective over ERα. WAY-202196 had an IC50 of 2.7 nM and was >75-fold selective in this assay. Finally, WAY-214156 was 100-fold selective for ERβ with an IC50 of 4.2 nM. For comparison, estradiol binds equally well to ERα and ERβ with an IC50 of ∼3–4 nM. All other compounds were obtained as previously described (25, 37).
Cell culture and virus production
Human U2OS osteosarcoma cell lines stably expressing a tetracycline-regulated ERα (U2OS-ERα) or ERβ (U2OS-ERβ) cDNA were prepared and maintained as previously described (38). Cells were maintained in phenol-free DMEM/F-12 containing 5% FBS, 2 mM glutamine, 50 U/ml penicillin, and 50 μg/ml streptomycin. U2OS-ERα and U2OS-ERβ cells were grown in 5% stripped FBS with 50 μg/ml hygromycin B and 500 μg/ml zeocin. The lentivirus producing a short hairpin to SRC-2 or a scrambled hairpin sequence was produced as described previously (19). For virus infection, U2OS-ERβ cells were incubated with culture medium containing the lentivirus supplemented with polybrene (8 μg/ml). Following transduction, cells were selected with 4 μg/ml puromycin (19).
38). The data were analyzed using the Microarray Suite Version 5.0 with the default parameters. The comparative data generated for each treated group were analyzed further in Microsoft Excel. TNF-α-induced genes were selected for further analysis only if they had a 1.2 signal log ratio mean value (2.3-fold change) and were statistically significant (p < 0.05) in at least three separate experiments.2 for 18 h. Total RNA extraction, generation of labeled cRNA, and hybridization of fragmented samples to the U95Av2 arrays (n = 3 for untreated, n = 4 for TNF-α-treated, n = 4 for TNF-α- and E2-treated samples) were performed as previously described (
Quantitative real-time PCR
Total RNA was prepared using the Aurum Total RNA kit (Bio-Rad) from cells treated as indicated in the figure legends. Reverse transcription reactions were performed using 1 μg of total RNA with an iScript cDNA Synthesis kit (Bio-Rad). Quantitative real-time PCR detection of gene expression was performed with the Bio-Rad iCycler Thermal Cycler System using iQ SYBR Green Supermix (Bio-Rad). The sequences of the primers used for PCR are available per request. Relative expression was determined from cycle threshold (Ct) values, which were normalized to β-actin as the endogenous control. Experiments were performed at least three times and the mean ± SE was calculated and statistical analysis was performed using the Prism curve-fitting program (Graph Pad Software, version 3.03).
Chromation immunoprecipitation (ChIP)
Following treatments as indicated in the figure legends, cells were cross-linked, washed, collected, and lysed as previously described (38, 39). A total of 50 μl of each sample was saved as total input. Immunoprecipitations were performed overnight at 4°C with anti-SRC-2 Ab (ab9261) (Abcam) or IgG to detect nonspecific background. DNA fragments were purified with a QIAquick PCR Purification kit (Qiagen). The −167 to +42 region of the TNF-α gene was amplified with the primers, 5′-CCCCGCGATGGAGAAGAAACCGAGA-3′ and 5′-GCTGGTCCTCTGCTGTCCTTGCTGA-3′. The primers used to amplify the keratin 19 estrogen response element (40) were: forward, 5′-TCCAGCCTGGGTGACAGAGC-3′ and reverse, 5′-TCCAAGTTCACCCCAACCTGA-3′. Real-time PCR data were analyzed as previously described (41). Experiments were done in triplicate and the mean ± SE was calculated and statistical analysis was performed using the Prism curve-fitting program (Graph Pad Software, version 3.03).
Western blot analysis
Buffy coats were obtained from healthy blood donors obtained from Stanford Blood Center (Palo Alto, CA). PBMCs were isolated by the Ficoll-Paque (Amersham Pharmacia) gradient procedure and cultivated in phenol red-free RPMI 1640 medium supplemented with 2 mM glutamine, 50 U/ml penicillin, 50 μg/ml streptomycin, 0.1 mM nonessential amino acids, and 10% charcoal dextran-stripped heat-inactivated FBS. After 3 days, cells were plated in 6-well cell culture dishes and treated for experiments. Cells were treated with 10 ng/ml LPS (Sigma-Aldrich) for 90 min in the absence or presence of 10 nM E2 or 100 nM ERB-041 for 6 h. Total RNA was isolated and the levels of mRNA were measured as described in the “Quantitative real-time PCR” section.
E2 represses TNF-α-inducible proinflammatory genes
To identify target genes activated by TNF-α that are repressed by 17β-estradiol (E2), we used microarrays. Previously characterized U2OS osteosarcoma cells stably transfected with a tetracycline-inducible human ERβ were treated with doxycycline to induce the expression of ERβ (38). The cells were treated with TNF-α for 1 h in the absence or presence of E2 and gene expression profiles were determined by using GeneChips that contain 12,555 probe sets encoding ∼10,000 independent human genes. Overall, 49 genes were significantly induced with TNF-α compared with untreated cells. Eighteen genes with known functions were activated by TNF-α and repressed by coadministration of E2 (Table I⇓). Among these genes, 10 are classified as cytokines and/or chemokines, which are crucial for inflammatory responses. Fifteen of the 18 genes repressed by E2 are involved in the immune response and/or TNF-α signal transduction. Most of these genes were not known to be repressed by estrogens. These data support a role for ERβ in the regulation of inflammatory genes.
ERβ is more potent than ERα at repressing cytokine genes
Quantitative real-time PCR was used to verify E2-mediated repression of 18 genes induced by TNF-α. Although identified in U2OS cells expressing ERβ, all of these genes were also repressed by E2 in U2OS cells expressing ERα (Table II⇓). Significant repression of TNF-α-inducible genes by E2 was rapid, occurring with all genes after a 30- or 60-min treatment. To determine the relative potency of repression with ERα and ERβ, we treated U2OS-ERα and U2OS-ERβ cells with increasing concentrations of E2. E2 produced a dose-dependent repression of the TNF-α (Fig. 1⇓A), small-inducible cytokine A4 (SCYA4, Fig. 1⇓B), IL-6 (Fig. 1⇓C), MCP1 (Fig. 1⇓D), and CSF2 (Fig. 1⇓E) genes in both U2OS-ERβ and U2OS-ERα cells. The IC50 values showed that ERβ is more potent than ERα at repression of the TNF-α, SCYA4, IL-6, and MCP1 genes, whereas no difference in repression of the CSF2 gene was observed with ERα and ERβ (Table III⇓).
ERB-041 is the most selective, but least potent, of the ERβ-selective compounds tested
It was previously shown that the synthetic ERβ-selective compound, ERB-041, promotes anti-inflammatory effects in rats (27). However, the mechanism whereby ERB-041 causes its anti-inflammatory actions is not known. We investigated the possibility that ERB-041 and related compounds exert their anti-inflammatory effects by causing transcriptional repression of inflammatory genes. In addition to ERB-041, we selected two other ERβ-selective compounds, WAY-202196 and WAY-214156, because these compounds were also ERβ selective in a competitive radioligand-binding assay (data not shown) and represent additional chemical scaffolds. U2OS-ERα and U2OS-ERβ cells were treated with increasing doses of ERB-041, WAY-202196, WAY-214156, or the ERα-selective agonist, propyl-pyrazole-triol (PPT) (36) for 2 h and then exposed to TNF-α for 1 h. In both cell lines, TNF-α produced a profound induction of TNF-α and CSF2 mRNA as determined by real-time PCR (data not shown). ERB-041 repressed the activation of the TNF-α (Fig. 2⇓A) and CSF2 (Fig. 2⇓B) genes only in the U2OS-ERβ cells. WAY-214156 and WAY-202196 were selective for ERβ, but repression of the TNF-α (Fig. 2⇓, C and E) and CSF2 (Fig. 2⇓, D and F) genes was also observed in the U2OS-ERα cells. As expected, PPT showed a much greater repression activity of the TNF-α and CSF2 genes in U2OS-ERα cells (Fig. 2⇓, G and H). Comparison of the IC50 values (Table IV⇓) demonstrates that ERB-041 is the most ERβ-selective compound, whereas WAY-214156 and WAY-202196 were more potent than ERB-041. The inhibitory effects of the drugs on the inflammatory genes were not due to a toxic effect, because the repression that occurred at 1 μM was abolished by the ER antagonist, ICI 182780 (data not shown).
SRC-2 is involved in ERB-041-mediated transcriptional repression
We previously showed that the coactivator SRC-2 is required for repression of the TNF-α gene by E2 via ERα (19). To determine whether SRC-2 is also involved in ERB-041-mediated repression of the TNF-α gene, we used a lentivirus to express a short hairpin RNA (shRNA) directed to SRC-2 in U2OS-ERβ cells. Western blots show that the cells expressing a SRC-2 shRNA have about an 80% knockdown of SRC-2 protein compared with U2OS-ERβ cells that express a nontargeted, scrambled shRNA (Fig. 3⇓A). E2 inhibited TNF-α activation by 82%, and ERB-041 by 80% in the U2OS-ERβ cells expressing scrambled shRNA, whereas only a 50% repression by E2, and 44% repression by ERB-041 was observed in the SRC-2 knockdown cells (Fig. 3⇓B).
To further investigate the role of SRC-2 in the transcriptional repression of the TNF-α gene by the ERβ-selective compounds, we performed ChIP. U2OS-ERβ cells were treated for 3 h with TNF-α before the addition of ERB-041, WAY-202196, WAY-214156, or E2 for 2 h. As a positive control, we show that all ligands recruited SRC-2 to the keratin 19 gene (Fig. 3⇑C) which is a known target gene activated by E2 (19, 38, 40). No recruitment of SRC-2 was observed in cells treated with only TNF-α (Fig. 3⇑D). ERB-041, WAY-202196, WAY-214156, as well as E2, induced the recruitment of SRC-2 to the TNF-α gene (Fig. 3⇑D). The RNA interference and ChIP studies demonstrate that the ERβ-selective compounds require SRC-2 for repression of the TNF-α gene.
ERB-041 represses cytokine genes in PBMCs
Our findings demonstrate that ERβ-selective estrogens repress inflammatory genes in the U2OS osteosarcoma cell line. To determine whether ERβ-selective estrogens exert similar effects in immune cells, we examined the effects of ERB-041 in PBMCs. We selected these cells because PBMCs have been shown to express ERβ and are responsive to estrogens (42, 43). Human PBMCs were maintained in charcoal dextran-stripped FBS for 3 days to deplete cells of intracellular estrogen before exposure of the drugs. The PBMCs were treated with E2 or ERB-041 for 6 h in the presence of LPS, which was used to activate cytokine genes because LPS caused a stronger effect than TNF-α in PBMCs (data not shown). LPS produced a large activation of the TNF-α, MCP-1, and CSF2 genes (Fig. 4⇓, A–C, respectively). Both E2 and ERB-041 inhibited the activation of these genes. Our results demonstrate that E2 and ERB-041 represses inflammatory genes in immune and U2OS cells. However, other cell types may not respond to E2 or the ERβ-selective compounds in the same way.
To begin to investigate how estrogens cause anti-inflammatory effects, we initially studied the effects of E2 on the expression of the TNF-α gene in a well-defined culture system previously shown to be responsive to estrogens (19, 25). TNF-α is one of the most potent cytokines and also stimulates the expression of other cytokine genes that trigger the inflammatory response. We found that in U2OS cells expressing either ER, TNF-α activates its own gene and that E2 represses the TNF-α-mediated autoinduction of its gene (19, 25). Our first goal in this study was to identify additional inflammatory genes induced by TNF-α, which could be used as markers to study the anti-inflammatory effects of estrogens in other in vitro and in vivo systems. Our microarray studies showed that TNF-α activated 49 genes in U2OS cells. E2 repressed 18 of these genes. Most of these genes were known proinflammatory genes, such as TNF-α, IL-6, and IL-8 (4, 6, 44, 45). The repression by E2 was rapid, occurring within 30–60 min after the treatment. These results suggest that ER acts directly at the promoter of these genes to repress transcription. It is unclear why E2 represses only 18 of the TNF-α-inducible genes. E2 can inhibit the action of NF-κB (17, 18), which is known to act as a transcriptional activator of inflammatory genes. Most of the genes induced by TNF-α did contain an NF-κB regulatory element, which indicates that these genes were not repressed because of the absence of an NF-κB site. We showed previously that the repression of the TNF-α gene requires a composite element that consists of c-jun and NF-κB, which are both required to form a platform for ER binding to the TNF-α promoter (37). These findings suggest that NF-κB probably requires additional elements for repression that are absent from the genes not repressed by E2.
The repression of inflammatory genes by E2 occurred with both ERα and ERβ. However, ERβ was modestly more potent than ERα in transcriptional repression for most of the TNF-α-induced genes tested. This is in contrast to the effects of ER on transcriptional activation, where ERβ generally has weaker activity than ERα in response to estrogens (46). Based on their preclinical in vivo activity (27), selective ERβ agonists would be expected to have an improved side-effect profile compared with currently available estrogen-based therapies that equally activate both ERs.
We examined three ERβ-selective ligands that were developed based on their selective binding to ERβ. Two of them, ERB-041 and WAY-202196, have already been shown to have potent anti-inflammatory effects in animal models (27, 34). Here, we show that ERB-041 and WAY-202196, as well as WAY-214156, were very effective at transcriptional repression of TNF-α-activated proinflammatory genes. However, the compounds did exhibit differences in ER subtype selectivity and potency. ERB-041 was the most selective of the compounds tested. It caused transcriptional repression in the presence of ERβ, but not ERα. It was also the least potent of the three compounds tested. WAY-202196 was the most potent compound with ERβ, but the least selective at repressing the TNF-α and CSF2 genes.
We previously showed that the coactivator SRC-2 is required for E2-mediated transcriptional repression of the TNF-α gene. In this study, we showed by RNA interference and ChIP that the ERβ-selective ligands also require SRC-2 for repression. Silencing of SRC-2 in U2OS-ERβ cells caused a partial reversal of ERB-041- and E2-mediated repression of the TNF-α gene. The incomplete reversal of repression by ERB-041 and E2 is probably a consequence of the presence of residual SRC-2 protein as we were unable to obtain a cell line with a complete knockdown. These results demonstrate that silencing of SRC-2 results in the same effect on TNF-α expression in the presence of both ERB-041 and E2, and suggest that the mechanism of repression by ERB-041 is similar to E2-mediated repression. WAY-202196 and WAY-214156 also recruited SRC-2 to the TNF-α promoter. The greater potency of WAY-202196 and WAY-214156 with ERβ compared with ERB-041 is likely due to their capacity to create a more favorable conformation for the recruitment of cofactors at the promoters of the inflammatory genes.
In this study, we showed that ERβ-selective compounds repressed genes that produce cytokines, which are involved in inflammatory diseases such as rheumatoid arthritis, endometriosis, Alzheimer’s disease, osteoporosis, and cardiovascular disease (3, 4, 5, 6, 47, 48). ERB-041 repression of inflammatory cytokine genes could be a mechanism whereby ERB-041 produces beneficial effects in animal preclinical models for rheumatoid arthritis and endometriosis (27, 29). In contrast, ERB-041 did not show bone protection in rats (27). The lack of effect in the bone compared with the other inflammatory models might be due to poor penetration of the drug into bone tissue or tissue-specific regulation of inflammatory genes. Although estrogens have anti-inflammatory actions in the brain (49) and cardiovascular system (2, 50), and ERβ is expressed in these tissues (51, 52), it is unclear whether ERβ-selective compounds will protect against neuroinflammation or atherosclerosis. However, estrogen protects against vascular injury in animal models and inhibits the expression of inflammatory mediators in smooth muscle cells through an ERβ-dependent mechanism (53, 54). These findings suggest that the vasculature might be an important therapeutic target for ERβ-selective agonists. It has been demonstrated that PBMCs express ERβ (42, 43), which is present in monocytes, B and T lymphocytes (55). In this study, we found that ERB-041 represses several cytokine genes in PBMCs. Whereas our study did not define the precise cell type in PBMCs that is inhibited by ERB-041, our findings suggest that immune cells are a potential therapeutic target for ERβ-selective estrogens to modulate inflammatory responses. Although the precise clinical indications are still being defined, the mechanistic studies presented here and animal studies (30), suggest that these ERβ-selective compounds might represent a new class of drugs to prevent and treat inflammatory disorders. Ultimately, clinical trials with synthetic, as well as natural, ERβ-selective compounds derived from plants (56) are needed to assess the potential translation of drugs that are targeted to ERβ for the prevention and treatment of human inflammatory disorders.
We thank Pierre Chambon, Jan-Åke Gustafsson, and Didier Trono for providing plasmids, David Rosen for assistance with PBMC isolation, Mike Malamas for synthesizing ERB-041, and Rick Mewshaw for synthesizing WAY-202196 and WAY-214156.
H. A. Harris is a full-time employee of Wyeth Research.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
↵1 This work was supported by National Institutes of Health Grant DK061966 (to D.C.L.).
↵2 Address correspondence and reprint requests to Dr. Dale C. Leitman, University of California, MS 1258, P.O. Box 0556, San Francisco, CA 94143-0556. E-mail address:
↵3 Abbreviations used in this paper: HT, hormone therapy; E2, estradiol; ER, estrogen receptor; SRC-2, steroid receptor coactivator 2; ChIP, chromatin immunoprecipitation; PPT, propyl-pyrazole-triol; shRNA, short hairpin RNA.
- Received June 15, 2007.
- Accepted October 23, 2007.
- Copyright © 2008 by The American Association of Immunologists