Estradiol Represses Human T-cell Leukemia Virus Type 1 Tax Activation of Tumor Necrosis Factor- (cid:1) Gene Transcription*

Adult T-cell leukemia is caused by human T-cell leukemia virus type I (HTLV-I). The HTLV-I Tax protein is essential for clinical manifestations because it activates viral and cellular gene transcription. Tax enhances production of tumor necrosis factor- (cid:1) (TNF- (cid:1) ), which may lead to bone and joint destruction. Because estrogens might prevent osteoporosis by repressing TNF - (cid:1) gene transcription, we investigated whether estrogens in-hibit the transcriptional effects of Tax on the TNF - (cid:1) promoter. Tax activated the (cid:2) 1044, (cid:2) 163, and (cid:2) 125 TNF - (cid:1) promoters by 9–25-fold but not the (cid:2) 82 promoter, demonstrating that Tax activation requires the (cid:2) 125 to (cid:2) 82 region, known as the TNF response element (TNF-RE). Three copies of the TNF-RE upstream of the minimal thymidine kinase promoter conferred a similar magnitude of activation by Tax. We demonstrated that c-Jun, NF (cid:3) B, p50, and p65 interact with and activate the TNF-RE by using mutational analysis of the TNF-RE, Tax mutants that selectively activate NF (cid:3) B or the cAMP-response element binding protein/activating transcription factor pathway, and gel shift assays with nuclear extracts. Estradiol markedly repressed Tax-ac-tivated transcription of the TNF - (cid:1) gene with estrogen receptor (ER) (cid:1) or (cid:4) . Nuclear extracts from U2OS cells stably transfected with ER (cid:1) demonstrated that ERs interact with the TNF-RE. Our studies provide evidence that ERs repress Tax-activated TNF - (cid:1) transcription by interacting with a c-Jun and NF (cid:3) B platform on the TNF-RE. Estrogens may ameliorate bone and inflammatory joint diseases in patients infected with HTLV-I by repressing transcription of the TNF - (cid:1) gene. (cid:2) repression of Tax-induced TNF - (cid:1) gene expression

The human T-cell leukemia virus type 1 (HTLV-1) 1 is the causative agent of adult T-cell leukemia (ATL), which is a fatal T-lymphoproliferative disorder (1,2), and a chronic progressive disease of the central nervous system termed HTLV-1-associated myelopathy/tropical spastic paraparesis (3). HTLV-1 infection is also associated with several autoimmune disorders such as Sjogren's syndrome and arthropathy, which is a chronic inflammatory disorder of joints similar to idiopathic rheumatoid arthritis (4,5).
In addition to genes such as gag, pol, and env encoding retroviral structural proteins, HTLV-1 also encodes for regulatory proteins such as Tax (6,7). Tax is a 40-kDa zinc finger protein that is involved in the etiology of ATL and its associated diseases by stimulating viral and cellular gene expression (8). Tax regulates gene expression mainly by activating a variety of transcription factors that interact with promoters of target genes (9 -16). Tax likely participates in the pathogenesis of diseases associated with HTLV-I infection by inducing multiple cytokine genes including interleukin-2, interleukin-2R␣, granulocyte-macrophage colony stimulating factor, interleukin-6, and tumor necrosis factor-␣ (TNF-␣) (17)(18)(19)(20)(21)(22)(23)(24).
Transgenic mice expressing Tax display thymic aplasia, neurofibromas, and skeletal alterations such as an increased number of osteoclasts (25)(26)(27)(28). Tax-expressing mice also exhibit a bone phenotype similar to that observed in HTLV-1-infected humans. Tax promotes bone diseases and hypercalcemia by increasing the expression of several cytokines. For example, T-cells infected with HTLV-1 express constitutively high levels of TNF-␣ (29), leading to increased serum levels of TNF-␣ (30). Because TNF-␣ acts as an inhibitory factor for the proliferation of osteoblasts and promotes the differentiation of precursor cells to mature osteoclasts (31), excessive TNF-␣ production leads to bone resorption and hypercalcemia (32), two characteristic features of ATL.
Albrecht et al. (33) reported that Tax could activate the TNF-␣ promoter in a region that binds NFB. We previously identified a region in the TNF-␣ promoter, the TNF-response element (TNF-RE), that is activated by TNF-␣ (34). We also found that estradiol (E 2 ) represses TNF-␣ activation of the TNF-␣ promoter in the presence of estrogen receptor ␣ (ER␣) or ␤ (ER␤) (35). Intriguingly, repression by ERs does not require direct DNA binding, because deletion of the DNA binding domain of ER does not prevent E 2 from inhibiting TNF-␣ activation of the TNF-␣ gene (35). This observation suggests that repression by ERs is not mediated through DNA binding but most likely through protein-protein interactions with other transcription factors at the TNF-␣ promoter.
Several studies (36,37) indicate that HTLV-1 infection exhibits gender-specific differences in clinical outcomes, which are thought to result from different levels of sex hormones. Women infected with HTLV-1 have a lower level of Tax expression and viral load as well as a decreased incidence of ATL compared with men (36). Furthermore, Hisada et al. (37) reported that the mortality from ATL is 4-fold higher in males relative to females. These observations suggest that higher levels of estrogens in women may attenuate the clinical effects of HTLV-1 by inhibiting the action of Tax. To test this hypothesis, we investigated whether estrogens repress Tax-mediated stimulation of TNF-␣ expression. The results shown here demonstrate that ERs repress Tax activation of TNF-␣ gene transcription in the presence of E 2 by interacting with c-Jun and NFB. E 2 repression of Tax-induced TNF-␣ gene expression suggests that estrogens may provide a potential therapeutic approach to ameliorate HTLV-1-associated bone and inflammatory joint diseases.

MATERIALS AND METHODS
Cell Culture, Transient Transfection, and Luciferase Assays-U937 (human monocytic leukemia cells) and U2OS (human osteosarcoma cells) were maintained as described previously (38). For transient transfection assays, cells were collected, transferred to a cuvette, and electroporated using a Bio-Rad Gene Pulser as described previously (38,39). Following electroporation, cells were resuspended in phenol redfree Dulbecco's modified Eagle's medium/F-12 media containing 5% fetal bovine serum, 2 mM glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin and plated in 12-well tissue culture dishes. Cells were treated with 17␤-estradiol for 24 h following transfection and then collected and lysed in 200 l of 1ϫ lysis buffer (Promega, Madison, WI). Luciferase assays were performed according to the manufacturer's instructions (Promega) using a Monolight luminometer. All experiments presented in the Figs. 1-7 legends were performed at least three times, and the data were similar between experiments. U20S cells stably transfected with a plasmid that expresses a tetracycline repressor were purchased from Invitrogen. These cells were then stably transfected with full-length ER␣ cloned downstream of a cytomegalovirus promoter that contains two tetracycline-responsive elements (pcDNA TO vector). The U20S-ER␣ cells were selected with hygromycin and zeocin. Individual clones were screened for the presence of ER␣ by reverse transcription PCR and Western blotting.
Electrophoretic Mobility Shift Assays-Binding reactions were performed using purified NFB p50 or c-Jun (Promega). 1 l of undiluted c-Jun or 1 l of diluted NFB p50 (diluted 1:25 in buffer containing 20 mM HEPES (pH 7.9), 400 mM KCl, 1 mM EDTA (pH 8.0), 1 mM EGTA (pH 8.0), 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 10% glycerol) was added to 15 l of total binding buffer (10 mM HEPES, (pH 7.9), 50 mM KCl, 0.2 mM EDTA (pH 8.0), 0.6 mM EGTA (pH 8.0), 10% glycerol, 2.5 mM dithiothreitol, 200 of g bovine serum albumin, and 2 g of poly(dI-dC)). Binding reactions were incubated for 15 min at 4°C. Following binding, anti-c-Jun (Cell Signaling Technologies) or anti-NFB p50 (gift from Dr. Warner Greene, Gladstone Institute of Virology and Immunology, San Francisco, CA) were added to the reactions, which were incubated for an additional 15 min at 4°C. Radiolabeled wild-type TNF-RE (containing the Ϫ125 to Ϫ82 region of the TNF-␣ promoter) probe was then added (40,000 cpm per reaction), and binding reactions were allowed to incubate for an additional 15 min at room temperature. The resulting complexes were electrophoresed through a 5% nondenaturing polyacrylamide gel with 1ϫ TBE running buffer (200 volts, 4°C). Following electrophoresis, gels were dried and exposed to film or examined using a Storm PhosphorImager and analysis software (Amersham Biosciences).
Quantitative Real Time PCR-U2OS-vector control or U2OS-ER␣ stable cells were transiently transfected with 0.5 g of Tax expression plasmid (a gift from Dr. Warner Greene) and treated with E 2 or ethanol for 24 h. Following treatment, total RNA was isolated using Trizol (Invitrogen). Reverse transcription reactions were performed using 500 ng of total RNA, 250 ng of random primers (Invitrogen), 200 units of MuLV reverse transcriptase (Invitrogen), 1ϫ reverse transcriptase buffer, 1 mM dNTP mix, 7.5 mM MgCl 2 , and 40 units of RNase inhibitor (Roche Molecular Biochemicals). Reactions were incubated at 25°C for 10 min, 48°C for 40 min, and 95°C for 5 min. Real-time PCR detection of TNF-␣ expression was performed using the pre-developed TaqMan assay reagents target kit for TNF-␣ (Applied Biosystems, Foster City, CA) and the ABI PRISM ® 7700 (Applied Biosystems). Control reactions were performed using primers and a probe to detect ␤-glucoronidase (GUS). GUS primers were GUS forward, 5Ј-CTCATTTGGAATTTTGC-CGATT-3Ј; GUS reverse, Ј-CCGAGTGAAGATCCCCTTTTTA-3Ј; and probe 6FAM-TGAACAGTCACCGACGAGAGTGCTGG-TAMRA (Operon, Alameda, CA). Average threshold cycle was calculated using the sequence detection software supplied with the ABI PRISM ® 7700.

Tax Activation and ER␣ and ER␤ Repression of the TNF-␣
Promoter in U937 Cells-We used the human monocytic leukemia cell line U937 to identify regions within the TNF-␣ promoter that may be responsive to HTLV-I Tax. This cell line is known to express the TNF-␣ gene in response to cytokines (34). U937 cells were co-transfected with the luciferase reporter containing several deletions of the TNF-␣ promoter (Ϫ1044 to Ϫ82) and a Tax expression vector. Tax activated the Ϫ1044, Ϫ163, and Ϫ125 TNF-␣ promoter constructs by about 9 -25fold ( Fig. 1). Deleting the TNF-␣ promoter from Ϫ125 to Ϫ82 abolished Tax activation. The Ϫ125 to Ϫ82 region of the TNF-␣ promoter was previously termed the TNF-response element, because it is activated by TNF-␣ (34). Tax activated three copies of the TNF-RE upstream of the minimal thymidine kinase promoter by a similar magnitude to the Ϫ1044 TNF-␣ promoter ( Fig. 2A). This finding demonstrates that the TNF-RE contains elements that confer responsiveness to Tax. Expression of ER␣ or ER␤ resulted in a marked repression (88 and 73%, respectively) of Tax-stimulated TNF-RE activity in response to E 2 . The repression by E 2 was dose-dependent with maximal effect observed at 1 nM (Fig. 2B). E 2 also repressed Tax activation of the TNF-RE in an adult human osteoblastic (AHTO) cell line (40) that is immortalized by the SV-40 large T oncogene (Fig. 2C), demonstrating that the effect of E 2 also occurs in bone cells.
ER␣ and ER␤ Repress Endogenous TNF-␣ Expression in U2OS Cells-To investigate whether the repression by E 2 is physiologically relevant, we performed quantitative real time PCR analysis to determine whether E 2 also represses Taxmediated activation of the endogenous TNF-␣ gene. For these studies, we used U2OS-ER␣ cells, which are human osteosarcoma cells stably transfected with a tetracycline-inducible cytomegalovirus promoter that drives the expression of the ER␣ cDNA. The U2OS-ER␣ cells or a vector control cell line (U2OS-vector) were induced with doxycycline, transfected with the Tax expression plasmid, and treated with E 2 for 12 h. Quantitative PCR analysis demonstrates that Tax activates the endogenous TNF-␣ gene by 3-fold compared with the FIG. 1. A Tax-responsive element is localized to the ؊125 to ؊82 region in the TNF-␣ promoter. U937 cells were transfected with plasmids containing a luciferase reporter fused to various regions of the TNF-␣ promoter or three copies of the TNF-RE (Ϫ125 to Ϫ82) upstream of the minimal thymidine kinase promoter in the presence or absence of a Tax expression plasmid (1 g). After 24 h cells were harvested, and luciferase activity was assayed. Each data point is the average of triplicate determinations. The S.E. was Ͻ10%. (Fig. 3). E 2 produces a 50% reduction in Tax-induced TNF-␣ mRNA levels in the U2OS-ER␣ cells. These results demonstrate that, similarly to transient transfection assays, ER␣ represses Tax activation of the endogenous TNF-␣ gene expression in an E 2 -dependent manner.

U2OS-vector control cells without Tax
c-Jun/CRE and NFATp/NFB Elements Are Required for Tax Activation of the TNF-RE-We next sought to explore the mechanism whereby Tax activates the TNF-␣ promoter. Several studies showed that ETS-, ATF-2-, NFATp-, NFB-, and c-Jun/CREB-related transcription factors bind to the TNF-RE in the TNF-␣ promoter (34,(41)(42)(43). NFATp/NFB and c-Jun/ CRE are apparently the most critical elements that regulate the TNF-␣ promoter function, because previous studies eliminated the ETS binding site as being central to TNF-␣ promoter activity (44). To investigate the role of NFATp/NFB (5Ј-GGGTTTCTCC-3Ј) and c-Jun/CRE (5Ј-TGAGCTCA-3Ј) elements in Tax activation of the TNF-␣ promoter, transfection assays were performed using luciferase reporters containing the wild type TNF-RE or mutations of the c-Jun/CRE or NFATp/NFB binding sites upstream of the thymidine kinase promoter. As observed in previous experiments, Tax activates the wild type TNF-RE ϳ10-fold (Fig. 4), whereas mutations in the c-Jun/CRE or NFATp/NFB binding site severely diminished Tax activation. As expected, mutations in both c-Jun/ CRE and NFATp/NFB binding sites also dramatically impaired Tax activation. No differences were observed between the levels of Tax activation with the single mutation reporters compared with the double mutation reporter constructs. Therefore, maximal Tax activation of the TNF-␣ promoter requires both c-Jun/CRE and NFATp/NFB elements.
NFB Activity Is Necessary for Tax-mediated Activation of the TNF-RE-To further dissect the pathways involved in Tax activation of the TNF-␣ promoter, we utilized two previously characterized Tax mutants, Tax M22 and Tax M47. The Tax M22 mutant is unable to activate NFB while maintaining its ability to activate transcription factors of the CREB/ATF family (45)(46)(47), whereas the Tax M47 mutant activates the NFB pathway but not the CREB/ATF pathway (48). As shown in Fig.  5, wild type and Tax M47 activated the TNF-RE, which was inhibited by E 2 . In contrast, Tax M22 was ineffective at activating the TNF-RE, and no repression by E 2 was observed. These results demonstrate that Tax has to activate NFB but not CREB/ATF to stimulate the TNF-␣ promoter.
Purified NFB p50 and c-Jun Bind to the TNF-RE-We investigated the ability of c-Jun and NFB to bind the TNF-RE by electrophoretic mobility shift assays. Binding reactions were done with purified p50 and c-Jun, the DNA binding components of NFB and AP-1, respectively. As shown in Fig. 6, both purified NFB p50 and c-Jun bind to the TNF-RE probe ( lanes  1 and 2, respectively). When reactions containing both NFB p50 and c-Jun were incubated for 15 or 60 min prior to binding to the TNF-RE probe, there were shifted complexes that correspond to the individual proteins (compare lanes 3 and 4 with  lanes 1 and 2). By using selective antibodies, it is clear that the slower migrating complex (marked by the arrow) contains both NFB p50 and c-Jun. (lanes 5 and 6). These results indicate that NFB p50 and c-Jun bind simultaneously to the TNF-RE.
ER␣ Binds to a c-Jun/NFB Complex in the TNF-RE-To begin to probe the mechanism whereby E 2 represses Taxmediated activation of the TNF-␣ promoter, we investigated our hypothesis that ERs bind to the TNF-␣ promoter indirectly via interactions with transcription factors bound to the TNF-RE, because we reported previously that ERs do not bind directly to the TNF-RE (35). For these studies, we used U20S FIG. 2. A, ER␣ and ER␤ repress Tax-activated transcription of the TNF-RE in U937 cells. U937 cells were transfected with the TNF-REtk-luc plasmid (3 g), a Tax expression vector (500 ng), and an expression vector for ER␣ (1 g) or ER␤ (1 g). Cells were treated with 10 nM E 2 for 24 h, and luciferase activity was measured. B, dose-dependent repression of ER mediated repression of Tax activation of the TNF-RE. U937 cells were transfected as described above and then treated with increasing concentrations (10 Ϫ13 -10 Ϫ6 M) of E 2 . After 24 h cells were harvested, and luciferase activity was assayed. C, ER␣ and ER␤ repress Tax-activated transcription of the TNF-RE in human (AHTO) osteoblasts. AHTO cells were transfected with the TNF-RE-tk-luc plasmid (3 g), a Tax expression vector (500 ng), and an expression vector for ER␣ (1 g) or ER␤ (1 g). Cells were treated with 10 nM E 2 for 24 h, and then luciferase activity was measured. Each data point is the average of triplicate determinations. The S.E. was Ͻ 10%.

FIG. 3. Tax stimulates endogenous TNF-␣ gene expression.
Tetracycline-inducible U2OS-ER␣ cells were transfected with the Tax expression plasmid. The cells were treated with doxycycline to induce ER␣ expression and maintained in the absence or presence of 10 nM E 2 for 24 h. Quantitative real time PCR was performed for TNF-␣ mRNA, which was normalized to ␤Ϫglucuronidase mRNA. cells stably transfected with ER␣. Following treatment with doxycycline to induce ER␣ expression, the cells were transfected with the Tax expression plasmid and treated with E 2 for 12 h. Binding reactions were performed with U20S-ER␣ nuclear extracts and specific antibodies to c-Jun, NFB p50, NFB p65, ATF-2, and ER␣ to identify candidate proteins that bind the TNF-RE (Fig. 6, lanes 2-6 respectively). A predominant single shifted band is observed with the nuclear extract (Fig. 7). This band is supershifted with antibodies to c-Jun, NFB p50, and NFB p65, but not with an antibody to ATF-2. This pattern demonstrates that a p50 and p65 heterodimer binds to the NFB site in the TNF-RE. Furthermore, c-Jun binds to the TNF-RE but not ATF-2, suggesting that the complex that binds to the c-Jun/CRE element is a homodimer of c-Jun or, more likely, a heterodimer with another transcription factor. There is also a strong supershift with the ER␣ antibody, indicating that it is also present in the complex bound to the TNF-RE. Taken together, these data suggest that the complex that binds to the TNF-RE contains c-Jun and NFB and that ER␣ represses the TNF-␣ promoter by binding to these factors via protein-protein interactions. DISCUSSION Our studies demonstrate that ERs repress Tax activation of the TNF-␣ promoter in transient transfection assays. The results from quantitative PCR analysis showed that the repression by E 2 is not an artifact of reporter plasmids, because E 2 produced a similar magnitude of repression of the endogenous TNF-␣ gene. Deletion studies showed that the Ϫ125 to Ϫ82 region of the TNF-␣ promoter is necessary for Tax activation. This region contains binding sites for ETS, c-Jun, ATF-2, NFATp, and NFB transcription factors that might mediate  6. Purified NFB p50 and c-Jun bind to the TNF-RE. Electrophoretic mobility shift assays were performed using purified NFB p50 or c-Jun. Binding reactions containing either purified protein were allowed to incubate for either 15 or 60 min at 4°C prior to the addition of NFB p50-or p65-specific antibodies. The presence of c-Jun and NFB p50 is confirmed by supershifting with specific antibodies (marked by the asterisk). The arrow indicates supershifted complexes.

FIG. 7. ER␣, NFB, and c-Jun interact on the TNF-RE.
Tetracycline-inducible U2OS-ER␣ cells were transfected with the Tax expression plasmid and then treated with 10 nM E 2 for 24 h and doxycycline to induce ER␣ expression. Nuclear extracts were prepared and then incubated with c-Jun, NFB p50, NFB p65, ATF-2, or ER␣ antibodies. Proteins or protein complexes that bound to a radiolabeled TNF-RE probe were detected by autoradiography. The arrow indicates supershifted complexes.
the Tax activation of the TNF-␣ promoter. Tsai et al. (42) reported that a c-Jun/ATF-2 heterodimer binds to the c-Jun/ CRE in the TNF-RE. The CREB/ATF-defective M47 Tax mutant activated the TNF-RE more than wild-type Tax, suggesting that CREB and ATF-2 are not the factors activated by Tax in these cells. Carter et al. (49) showed that M47 Tax is more stable then wild type Tax, which can explain the finding that M47 Tax produced a greater activation of the TNF-RE compared with wild type Tax. Activation of the NFB element by Tax is essential for maximum activation of the TNF-␣ promoter, because the M22 Tax mutant deficient in activating the NFB pathway is unable to stimulate the TNF-RE reporter. Our in vitro DNA binding studies using nuclear extracts from Tax-stimulated human osteosarcoma cells showed that NFB p50, NFB p65, c-Jun, and ER␣ but not ATF-2 interact with the TNF-RE element. These results indicate that the activation of the TNF-␣ promoter by Tax requires c-Jun and NFB but not ATF-2. Furthermore, c-Jun and NFB must bind simultaneously to the TNF-RE to activate the TNF-␣ promoter, because a mutation in either site abrogates activation by Tax. Other studies reported that related factors such as c-Fos do not interact with the TNF-RE (34).
Our results suggest the following model whereby ER represses the activation by Tax at the TNF-RE. After the activation of NFB by Tax, the NFB p50/p65 complex binds next to c-Jun on the TNF-RE, leading to the activation of the TNF-␣ promoter. The ER is then recruited to the promoter by binding to a platform comprised of c-Jun and NFB. Support for this model of ER action is provided by evidence that ER␣ or ER␤ do not directly bind to the TNF-␣ promoter (35) and that NFB p65 and ER can interact (50 -53). Furthermore, ER␣ and c-Jun proteins interact directly (54) in glutathione S-transferase pulldown and mammalian two-hybrid assays (55). Alternatively, ER could be recruited to the TNF-␣ promoter through a bridging protein in direct contact with the c-Jun/NFB complex.
Examination of the sequence within the TNF-RE region of the TNF-␣ promoter reveals that there is only one nucleotide separating the c-Jun/CRE and NFATp/NFB elements. Previous experiments have found that nucleotide insertions resulting in either 0.5, 1, or 1.5 extra turns of the DNA helix abolishes TNF-␣ activation (44). Therefore, it is likely that c-Jun and NFB on the TNF-␣ promoter are critical for providing the initial platform for ER binding to the TNF-␣ promoter, and disruption of the platform prevents the assembly of other factors necessary for activation of the TNF-␣ gene by Tax.
One key question that remains is what are the additional proteins in the complex that mediate repression. There is evidence suggesting that coactivators such as glucocorticoid receptor-interacting protein-1 (GRIP1) are involved in steroid receptor repression of gene transcription. GRIP1 potentiates ERmediated repression of the TNF-␣ gene (35) and glucocorticoid receptor-mediated repression of the collagenase gene (56). Another protein that may be part of the repression complex is p300/CBP, because it is involved in Tax-mediated activation of HTLV-1 transcription in vitro (57). Whether these proteins are an integral part of the repression complex at the TNF-␣ promoter or whether other proteins participate in repression remains to be determined.
Individuals infected with HTLV-1 suffer from various bone and joint diseases most likely due to an increased expression of various cytokines, including TNF-␣. Hypercalcemia, which results from enhanced bone resorption, is a prevalent complication observed in patients with ATL and has been linked to early death (58). The cause of hypercalcemia in ATL patients is unknown, but it has been suggested that TNF-␣ plays a role in this disease because it is associated with increased serum levels of TNF-␣ (30). Tax is also strongly expressed in human synovial cells, which leads to joint destruction. In a transgenic mouse model of HTLV-I infection, mice expressing Tax exhibit skeletal alterations resembling Paget's disease, a chronic disorder that results in enlarged and deformed bones and have chronic inflammatory polyarthropathy (8,59). Analysis of the joints in these transgenic mice demonstrated enhanced expression of several cytokines, including TNF-␣ (59).
Intriguingly, similarly to HTLV-I infected patients, many postmenopausal women develop bone and joint diseases that might occur from excessive TNF-␣ production. A prominent role for TNF-␣ in the pathogenesis of osteoporosis is supported by animal studies. For example, overexpression of TNF-␣ in mice produces profound hypercalcemia from enhanced bone resorption (60). Furthermore, the loss of bone mineral density observed in mice after an oophorectomy can be prevented with TNF-binding proteins (61) or a soluble TNF receptor that prevents the action of TNF-␣ (62). TNF-␣ might cause osteoporosis by inducing several proteins responsible for differentiating precursor monocytic cells into bone resorbing osteoclasts (63,64).
Estrogens are used extensively in postmenopausal women to prevent osteoporosis (65,66). Several studies indicate that the bone-sparing effect of estrogens is at least in part due to its ability to repress TNF-␣ gene transcription and down-regulate TNF-␣ levels (67,68). Other studies indicate that TNF-␣ is involved in the destruction of the articular cartilage that is observed in osteoarthritis (69). Osteoarthritis is a prevalent condition that is exacerbated by estrogen deficiency in postmenopausal women and shows some improvement with estrogen replacement (70), possibly by decreasing TNF-␣ levels. Whereas estrogens are useful in postmenopausal bone and joint diseases, our study suggests that they also might be a potential therapeutic approach for HTLV-1-associated bone and inflammatory joint diseases by directly targeting TNF-␣ gene expression. Estrogens are known to exhibit anti-inflammatory properties (71), but it is not known if this effect occurs through ER␣, ER␤, or both of these ERs. Our results demonstrating that ER␤ is more effective than ER␣ at inhibiting TNF-␣ and Tax-activation of TNF-␣ gene transcription suggest that ER␤ may be the predominate receptor that mediates the anti-inflammatory effects of estrogens. The finding that E 2 is a potent repressor of Tax activation of the TNF-␣ gene may also account for the observation that females infected with HTLV-I exhibit less severe clinical manifestations and lower mortality compared with males (37).