I. INTRODUCTION

Top Previous Next References

Cross-talk between members of the nuclear receptor superfamily theoretically can multiply the possible modes of gene regulation, leading to a greater and more flexible array of transcriptional responses to environmental changes. In the central nervous system (CNS), such gene regulation conceivably can help to coordinate behavioral responses of the organism to climatic and social stimuli (217). Such cross-talk can also underlie metastatic processes. The activation of the estrogen-dependent growth responses by a nonestrogen such as the growth factor, insulin growth factor (IGF), may promote the growth of various cell types. Cross-talk between IGF and estrogens, for example, can lead to cell proliferation in breast carcinoma (41 and references therein) and hence is of considerable interest as a target for adjunct therapy.

Estrogen (ER) and thyroid hormone receptors (TR) are members of the nuclear receptor superfamily that bind the low-molecular-weight ligands, estrogens and thyroid hormones, respectively. They transduce these signals into gene regulation events. These receptors have a modular protein structure with high homology in the central DNA binding domain. They are ligand-activated transcription factors that influence transcription from target genes (43, 105). Nuclear receptors bind enhancer elements on DNA called hormone response elements to regulate transcription from genes (43, 105).

How do nuclear receptors regulate gene transcription? Gene activation events require the recruitment of specific coactivators by ligand-bound nuclear receptors (33, 116, 122). This leads to the remodeling of chromatin, since many coactivators possess histone acetyltransferase (HAT) activity (27, 178). The remodeling and "opening" up of chromatin leads to gene activation (27). Unlike the ER, the TR can regulate transcription even in the absence of ligand (213). Two well-characterized corepressors termed nuclear corepressor (NCoR) and SMRT exist for the TRs and the retinoic acid receptors (RAR) (71, 93). In the absence of ligands, TR and RARs recruit corepressors, which have histone deacetylase activity and antagonize HAT coactivators. This, in turn, represses gene transcription (68).

Estrogens are critical in the control of reproduction in both male and female mammals (88). The deletion of the ER isoform causes infertility in both male and female mice (104, 128). The nonreproductive functions of estrogens include maintenance of bone mass (64 and references therein) and cardioprotective effects (170). The central role of estrogens in mammalian reproduction is reflected in the neuroendocrine control of gonadotrophin production from the anterior pituitary and feedback regulation of gonadotropin releasing hormone (GnRH) in the hypothalamus (88, 89). Recently, neuroprotective effects as well as effects on mood and cognition have also been described (99, 147, 204).

Unlike estrogens, which have a more focused role, the thyroid hormones triiodothyronine (T₃) and thyroxine (T₄) exhibit a large range of actions (135, 208). In the adult homeothermic animal, they exert control over lipogenesis, lipolysis, and thermogenesis. They are critical for growth, development, and differentiation (134, 172). In humans, neonatal hypothyroidism results in cretinism, a disorder characterized by mental retardation and skeletal defects (35). The effects of thyroid hormone on the brain are especially well illustrated by thyroid hormone control of myelin basic protein and the consequent myelination of axons in the brain (45). This article attempts to review the relevant information on estrogen and thyroid hormone interactions with a perspective on neuroendocrine functions.

	II. HORMONAL INDUCTION OF GENES IN THE CENTRAL NERVOUS SYSTEM AND IN CELL LINES

Top Previous Next References

Estrogens are necessary for the induction of the primary female-typical sexual behavior lordosis (91) in several species. Estrogens regulate the expression of several genes in the brain, some of which are responsible for the facilitation of lordosis (reviewed in Ref. 145). For example, the nonapeptide oxytocin (OT) and its receptor, the oxytocin receptor (OTR), are expressed in the uterus during pregnancy (96) and are thought to be important for gestation, parturition, and lactation. Mice that do not have OT (OT knock-out mice) cannot lactate and therefore cannot nurse pups (123). The expression of OT and OTR in the CNS is believed to be important for the facilitation of affiliative (24, 44) and sexual behaviors (7, 8, 22) that are, in turn, required for optimal reproduction.

Estrogen administration to ovariectomized female rats increases OTR mRNA in the ventromedial hypothalamus (VMH), a brain region critical for lordosis (153). Estrogens also upregulate OTR mRNA in the medial amygdala, hippocampus, and anterior pituitary but do not change the concentration of OTR mRNA in the caudate putamen or arcuate nucleus of the rat (153). Concomitantly, estrogen treatment increases oxytocin binding in the bed nucleus of the stria terminalis, lateral ventral septum, amygdala, and VMH (20, 59, 188). In situ hybridization studies show that OTR mRNA is highest in the rat VMH at proestrous, showing a correlation with the estrogen surge that occurs during this part of the estrous cycle (12). When antisense oligonucleotides are infused into the VMH of estrogen-primed female rats, the rats show significantly higher rejection behavior and lower lordosis, thus demonstrating the need for an intact OT-OTR system for reproductive success (107).

In a similar manner, estrogen treatment also upregulates preproenkephalin (PPE) mRNA in the rodent VMH (97, 168, 169). The importance of the expression of this gene is underscored by the ability of antisense oligonucleotides against PPE in the hypothalamus to reduce lordosis behavior (121).

B.  Isoforms From Genes for ER and TR: Distinct and Overlapping Functions

1.  Two ER genes:  and

Among vertebrates, the ER exists in two isoforms,  and , which are products of different genes (186). The newly discovered ER isoform was cloned from rat prostate and the ovary. The isoforms exhibit considerable homology in the DNA binding domain and COOH-terminal AF2 domain but high divergence in the NH₂-terminal transactivation AF-1 domain (65). Both isoforms can bind several ligands with similar affinities (92); they also bind the consensus the estrogen response element (ERE) with similar affinities (31). The dissociation of ER and ER from such a consensus ERE is similarly affected in the presence and absence of ligand at elevated temperature (136). They also bind many EREs (72) derived from estrogen-regulated genes that have deviations from the consensus ERE sequence. Both receptors contain a functionally conserved AF-2 domain, which can be stimulated by binding the steroid receptor coactivator (SRC1) (31, 186). Despite limited homology in the NH₂-terminal AF-1 domain, both ER and ER contain a mitogen-activated protein kinase (MAPK) phosphorylation site that results in enhanced transcription (186). Hall and McDonnell (66) show that despite similar binding affinities for several ligands, activation of transcription from simple target promoters containing EREs by ER is dependent on pure agonists. On the other hand, ER can activate transcription when bound to agonists and partial agonists. If a chimeric ER receptor containing the A/B domain of the ER is tested for transcriptional activation, antiestrogens such as tamoxifen, which showed no transcriptional activation with ER, could now show some degree of transactivation (110). Jones et al. (76) investigated the ability of ER and ER to activate transcription from a number of different promoters that are estrogen responsive but lack classical EREs in human breast, bone, and uterine cell lines. These included a collagenase promoter containing an AP-1 element important in estrogen induction, a nonconsensus ERE containing complement C3 promoter, and a transforming growth factor (TGF)- promoter containing both ERE and Sp1 elements. All antiestrogens studied were agonistic on the collagenase reporter in the uterine cell lines when ER was transfected, but tamoxifen alone was agonistic when ER was transfected (76). Also, the ability of ER to repress transactivation of NFB in osteoblasts occurs only in the presence of 17-estradiol, whereas ER can repress NFB transactivation in the absence or presence of ligand (152). This suggests important mechanistic differences, possibly arising from differences in amino acid sequence in the AF-1 domain (66, 195).

At AP-1 sites, classical estrogens such as diethylstilbestrol and 17-estradiol activated transcription when bound to ER but were antiestrogens when bound to ER (137). On the human RAR-1 promoter, ER activates transcription in response to estrogens through nonclassical ERE and not by direct DNA-receptor binding. However, in response to estrogens, ER does not activate this promoter; it activates it in response to tamoxifen, raloxifene, and ICI-164,384 (220). Therefore, the ER isoforms show considerable promoter site specificity. In vivo and in vitro, heterodimerization between ER and ER has been shown (129, 143), and tissues that coexpress both isoforms are thought, therefore, to respond differently to various ER ligands compared with tissues that express predominantly one isoform (65, 66). Therefore, there is likely a considerable contribution of ER to the pharmacology of estrogens and antiestrogens.

The four TR isoforms are protooncogene products derived by differential splicing of two different genes: TR1 and TR2 are from the TR gene, while TR1 and TR2 are from a separate TR gene. In chicken, a shorter TR1 transcript lacking the NH₂-terminal A/B domain is also present. Also, two TR1 transcripts, which possess very short A/B domains, are also present in the chicken. Xenopus laevis has several transcripts with homology to TR1 but none similar to TR2 (98). Again, although there is considerable homology in the central DNA binding domain among the isoforms, significant dissimilarity exists in the NH₂-terminal A/B domain. Not all TR isoforms can bind ligand; the TR2 isoforms lacks the ability to bind ligand due to a loss of 40 amino acids in the COOH-terminal hormone-binding domain. The role of TR2 in physiology is unclear; ex vivo studies in cell culture implicate it as a dominant negative inhibitor of TR action. However, the potency of dominant negative action is lower than the unliganded TR isoforms, possibly due to deficient interactions with corepressors (182).

The transcriptional properties of different TR isoforms have been poorly studied, but there do exist some differences. The differences in the TR isoforms could allow for differential interactions with other proteins, thereby regulating transcription. For example, unliganded TR2 can bind SRC-1, unlike TR1 and TR1 (125). The TR2 is a more potent mediator of ligand-independent activation than TR1 or TR1 of T₃ target genes such as the thyroid stimulating hormone (TSH) subunit gene and the TRH gene (70, 94, 171). This ability is independent of NCoR and may be due to differential binding of coactivators. The TR2 isoform also is unable to mediate ligand-independent repression on the growth hormone promoter, unlike the TR1 and TR1 isoforms, due to lack of NCoR binding ability (70). Zhu et al. (215) have noted an increased ability of TR1 to transactivate from a F2 thyroid hormone response element (TRE) compared with TR1 (215). Differential interaction with other proteins, including other nuclear receptors, may therefore play a role in thyroid hormone physiology.

Since the consensus DNA sequences bound by ER and TR share a common half site, it is possible that competition between the two receptors may lead to antagonism of the other's effect. Indeed, this was first demonstrated for the vitellogenin ERE by Glass et al. (60); the thyroid hormone receptor could decrease ER-mediated transactivation.

Steroid hormone receptors have been shown to decrease ligand-dependent TR transactivation from a TRE (210). Estrogens were also shown to suppress the T₃ effect on the -glycoprotein hormone subunit promoter (207). In both pituitary-derived GH₃ cells and JEG-3 choriocarcinoma cells, T₃ mediates suppression of the -glycoprotein hormone subunit promoter. 17-Estradiol suppressed this inhibition. In vitro synthesized ER could bind to the TRE present in this promoter, thereby suggesting competition between these two nuclear receptor systems as a distinct possibility (207). However, in both these studies, pure TR and ER isoforms were not used. Hence, the investigation of possible differential interactions between distinct ER and TR isoforms is of interest.

Two different promoters were initially used to examine the interactions between the ER and the TR. The consensus ERE derived from the vitellogenin gene promoters has long been used as a model system to explain molecular mechanisms by which estrogen regulates genes. In CV-1 kidney fibroblast cells, the consensus ERE linked to a CAT reporter has been shown previously to be transcriptionally upregulated by estrogen-liganded ER (218). Transiently transfected TR1 was able to inhibit this ER-mediated induction, but TR1 and TR2 had no effect (218). Contrasting to another promoter, when three tandem copies of the estrogen response EREs from the progesterone receptor promoter are used in the CV-1 cell line, no TR isoform could inhibit the ER-mediated induction (173). This suggested that the interactions between TR and ER isoform were different on different promoters.

The interaction of ER, the classical ER isoform, and the ligand-binding TR isoforms has been observed on a consensus vitellogenin ERE linked to a minimal thymidine kinase promoter in kidney fibroblast, CV-1, cells (190, 218). CV-1 cells were chosen since they have low endogenous ER and TR isoforms (215). On cotransfection of ER and TR1 expression vectors, the T₃-liganded TR1 isoform could interfere with the ER induction of this simple promoter (Fig. 1A) (190, 218). However, the TR1 or TR2 isoforms did not have any effect on the ER induction of this promoter (190, 218) (Fig. 1, B and C). This demonstrates that on a simple consensus ERE, there is considerable difference in the modulation of transcriptional activity of ER by the ligand-binding TR isoforms.

View larger version (31K): [in this window] [in a new window]   Fig. 1. Effect of the ligand binding thyroid hormone receptor (TR) isoforms on estrogen receptor (ER)-mediated induction of the consensus estrogen response element (ERE) in CV-1 cells. CV-1, a kidney fibroblast cell line, was grown in DMEM supplemented with 10% fetal bovine serum. The reporter plasmid is the ERE-tk-CAT and has a single consensus vitellogenin ERE upstream of a minimal thymidine kinase promoter linked to chloramphenicol acetyltransferase (CAT) (218). The cells seeded in 6-well plates (Falcon) were transfected at 60% confluency. The expression plasmids used were the pSG-hER coding for the ER protein and the pCDNAI-rTR1, -1, and -2 coding for the TR isoforms (218). The pSV--galactosidase (pSVgal) plasmid coding for the enzyme -galactosidase was used as a normalization control for transfection efficiency and lysate preparation. The reporter plasmid (200 ng), the expression plasmids for the nuclear receptors (80 ng), and the pSVgal (80 ng) plasmids were cotransfected into CV-1 cells using the Effectene reagent (Qiagen) according to the manufacturer's instructions. Twenty-four hours after transfections, phenol red-free media supplemented with hormone-free sera were added to the cells. Either 17 -estradiol (E) (107 M) or triiodothyronine (T) (106 M) or both (E+T) were added to wells. A set of wells received the vehicle, ethanol, alone. Forty-eight hours after hormone treatment, cells were lysed using Reporter lysis buffer (Promega) according to the manufacturer's instructions, and CAT and -gal assays were performed on every sample. The CAT activity was normalized to the -gal activity for every sample. Results (fold over vehicle control) represent means ± SE (n = 5/treatment group). Statistical comparisons between treatment groups were done using ANOVA followed by Student-Newman-Keuls post hoc tests. A (TR1): *P < 0.001 compared with the vehicle-treated group. #P < 0.001 compared with the estrogen-treated group. B (TR1): *P < 0.05 compared with the vehicle-treated group. C (TR2): *P < 0.001 compared with the vehicle-treated group. [Modified from Vasudevan et al. (190).]

ER could also induce this promoter in CV-1 cells, albeit at a lower level than the ER isoform (Fig. 2). When the ligand binding TR isoforms were cotransfected with ER in CV-1 cells, the TR isoforms showed differential effects on ER-mediated induction from the consensus ERE. In contrast to the inhibitory effect of the TR1 isoform on the consensus ERE, the TR1 isoform stimulated ER-mediated transcription (Fig. 2A). The TR1 isoform also stimulated ER-mediated transcription (Fig. 2B), while the TR2 isoform inhibited ER-mediated transcription (Fig. 2C). However, neither TR1 nor TR2 had any effect on ER-mediated transcription. This shows that a single TR isoform can lead to differential transcriptional outcomes depending on the ER isoform present in the cell.

View larger version (31K): [in this window] [in a new window]   Fig. 2. Effect of the ligand binding TR isoforms on the ER-mediated induction from the consensus ERE in CV-1 cells. The protocol for transfection as well as the reporter and expression plasmids for the TR isoforms has been detailed in the legend to Figure 1. The ER expression plasmid used was the pCMV-ER (137). The CAT activity was normalized to the -gal activity for every sample. Results (fold over vehicle control) represent means ± SE (n = 8 or 9/treatment group). Statistical comparisons between treatment groups were done using ANOVA followed by Student-Newman-Keuls post hoc tests. A (TR1): *P < 0.01 compared with the vehicle-treated group. #P < 0.001 compared with the vehicle-treated group. P < 0.001 compared with the estrogen-treated group. B (TR1): *P < 0.001 compared with the vehicle treated group. #P < 0.001 compared with the estrogen-treated group. C (TR2): *P < 0.01 compared with the vehicle-treated group. #P < 0.001 compared with the vehicle-treated group. P < 0.01 compared with the estrogen-treated group. [Modified from Vasudevan et al. (190).]

Are the interactions between the various TR and ER isoforms different on a physiologically relevant promoter? To address this question, the estrogen-responsive, behaviorally relevant PPE and OTR promoters cloned upstream of reporter genes were transfected into CV-1 cell lines and neuronal (SK-N-BE2C) cell lines. The PPE promoter has two EREs located within 450 bp of the transcription start site (77), whereas the OTR promoter has a distal ERE located ~4 kb from the transcription start site (11).

Similar to the inhibitory effect by the TR1 on ER-mediated transcription from the consensus ERE, the TR1 inhibited ER induction from the PPE promoter (Fig. 3A). In contrast, the TR1 isoform stimulated ER-mediated transcription (Fig. 3B), demonstrating differences in the interaction of a given TR isoform with an ER isoform. In the CV-1 cell line, the TR isoforms had no effect on ER or ER-mediated induction of the PPE promoter (192). Again, similar to the consensus ERE, the ability of the TR1 isoform to inhibit or stimulate ER-mediated transcription of the PPE depends on the ER isoform present in the cell (192).

View larger version (19K): [in this window] [in a new window]   Fig. 3. TR1 has opposing effects on ER-mediated (A) and ER-mediated (B) induction from the rat preproenkephalin (PPE) promoter in CV-1 cells. The protocol for transfection and the expression and normalization plasmids used have been described in the legends to Figures 1 and 2 and in previous studies (137, 218). The rat PPE gene promoter (437 to +53 bp) containing two putative EREs was inserted into pUTKAT1 (77, 218). The CAT activity was normalized to the -gal activity for every sample. Results (fold over vehicle control) represent means ± SE. Statistical comparisons between treatment groups were done using ANOVA followed by Student-Newman-Keuls post hoc tests. A (n = 8/treatment group): *P < 0.001 compared with the vehicle-treated group. #P < 0.01 compared with the estrogen-treated group but >0.05 compared with vehicle-treated group. B (n = 5/treatment group): *P < 0.05 compared with the vehicle-treated group. #P < 0.01 compared with estrogen-treated group. [Modified from Vasudevan et al. (192).]

To test the interactions between the TR and ER isoforms in a cell line with neuronal properties and to compare these interactions with those occurring in a nonneuronal cell line, the OTR promoter containing the distal ERE was investigated in both CV-1 and SK-N-BE2C cell lines (189). Again, the TR1 isoform was capable of inhibiting the ER-mediated induction from the OTR promoter in both CV-1 and SK-N-BE2C cell lines. However, a complex pattern of interactions emerged when the TR isoforms were expressed in conjunction with the ER isoform. Although the TR2 isoforms were inhibitory to ER-mediated induction of this OTR promoter in either cell line, the TR1 effect on ER induction depends on the cell line. In the CV-1 cell line, this isoform stimulated the ER induction while inhibiting it in the neuronal cell line (189). A neuronal specific cofactor, NIX1, which can bind liganded TR1 and downregulate transcription, could be responsible for this phenomenon (63). The expression of the cofactor appears to be confined to dentate gyrus, the amygdala, as well as thalamic and hypothalamic regions and may contribute to the differences in transcriptional activation observed with the TR1 isoform in these cell lines. Therefore, cell-specific effects are also important in the interactions between nuclear receptor isoforms. These data underscore the need to test physiological promoters in different cell lines.

Table 1 summarizes the transcriptional pattern obtained using different promoters and various combinations of ER and TR isoforms. The different outcomes show that the interaction between the ER and TR isoforms demonstrates considerable promoter specificity. It is easy to visualize that different levels of TR and ER isoforms in cells may, therefore, allow for flexible regulation of EREs depending on stimuli.

View this table: [in this window] [in a new window]   Table 1. ER compared with ER in the specificities of interactions with different TR isoforms: promoter and cell type dependence

A) COMPETITIVE DNA BINDING. What are the possible mechanisms by which the TR isoforms interact with the ER isoforms? Since the consensus DNA sequences, the hormone response element (HRE), bound by both ER and TR, are very similar, TR can interfere with ER-mediated transcription by competing with the ER for binding to the ERE on DNA (60). A TR1 mutant called the TR P-box mutant has a mutation in the DNA binding region; this disallows binding by this isoform to DNA (209). If DNA binding were important in the inhibitory interaction between TR1 and the ER isoform, then such a mutant should not be able to inhibit the ER induction. With the consensus ERE in CV-1 cells, inhibition by the TR1 isoform was lost when the TR P-box mutant is used (190, 218), implicating DNA binding and hence competition for the ERE as important in this inhibition. In addition, the TR1 and TR1 isoforms can bind to the consensus ERE, thus making inhibition by competitive DNA binding possible (218). However, differences arise when "physiological" promoters such as the OTR and PPE promoters are used. Despite lack of DNA binding ability, the TR P-box mutant could nonetheless inhibit the ER-mediated induction of both PPE (Fig. 4A) and OTR promoters, suggesting that DNA competition may not be a universal mechanism in inhibition. Also, to check if the levels of ER isoforms expressed in the CV-1 cell lines significantly differ, binding of [³H]estradiol to extracts of cells transfected with either ER or ER was done. There was no difference in the levels of ER or ER; this also correlates well with the similar level of transcriptional activation promoted by either isoform in response to 17-estradiol. The nuclear corepressor NCoR mediates basal repression by unliganded TR isoforms (30). However, because the inhibition detailed in Table 1 is ligand dependent, this must represent a novel, NCoR-independent mechanism.

View larger version (20K): [in this window] [in a new window]   Fig. 4. Mechanisms of TR-mediated interference of ER-driven transcription. A: a TR1 mutant unable to bind DNA, the TR P-box mutant, can still cause inhibition of the ER-mediated induction from the rat PPE promoter in CV-1 cells. B: SRC-1 overexpression rescues TR1 inhibition of ER-mediated induction from the rat OTR in CV-1 cells. A: CV-1 cells were cotransfected with PPE reporter plasmids and the expression plasmids for ER and the TR P-box mutant (209) as detailed in the legend to Figure 1. [Modified from Vasudevan et al. (192).] B: SRC-1, a general steroid receptor coactivator, was overexpressed in CV-1 along with the TR1 and ER isoforms as detailed in the legend to Figure 1. The samples that were transfected with SRC-1 expression vector (4th, 5th, and 6th bar from left) were treated with 17-estradiol (E) (107 M) or both 17-estradiol and triiodothyronine (E + T). Corresponding hormone treatment was given to a set of samples that received the empty control SRC-1 expression plasmid (first 3 bars from left). [Modified from Vasudevan et al. (189).] Forty-eight hours after hormone treatment, cells were lysed, and each sample was assayed for both -gal and CAT activity. The reporter gene activity was normalized to the -gal activity for every sample. Results (fold over vehicle control) represent means ± SE. Statistical comparisons between treatment groups were done using ANOVA followed by Student-Newman-Keuls post hoc tests. A (n = 8/treatment group): *P < 0.01 compared with vehicle-treated group. #P < 0.01 compared with estrogen treatment. B (n = 5/treatment group): *P < 0.05 compared with the vehicle-treated group. #P < 0.05 compared with the vehicle-treated SRC-1 group. P < 0.001 compared with the estrogen-treated SRC-1 group.

B) RESCUE OF TR1 INHIBITION BY THE EXPRESSION OF A COACTIVATOR. The p160 group of coactivators, of which SRC-1 is a member, has been shown to bind both the ER ligand binding domain and the TR (103). Therefore, SRC-1 appeared to be an attractive candidate that could be tested for its ability to restore transcriptional activation by ER on the rat OTR and PPE promoters in the presence of the inhibitory TR1 isoform. To explore the idea that coactivators are sequestered by the TR isoforms, we overexpressed a general steroid coactivator, the SRC-1, along with TR1 and ER. On both PPE and OTR (Fig. 4B), promoters as well as the minimal promoter containing the consensus ERE, SRC-1 overexpression could rescue TR1 inhibition, suggesting that squelching of common coactivators is an important mechanism of inhibition by the TR isoforms. Recently, Auger et al. (10) have shown that reduction in brain SRC-1 levels by antisense oligonucleotide injection reduces the ability of the ER to defeminize the brain during postnatal sexual differentiation in Sprague-Dawley rats. Inhibition of T₃-dependent transcriptional activation by other nuclear receptors such as the glucocorticoid and estrogen receptor has been reported to be due to titration of essential coactivators (209). The ligand-bound TR and TR proteins could interfere with progesterone receptor (PR)-mediated transactivation from a progesterone responsive reporter in the CV-1 cell line. Deletion of the DNA binding region did not affect the inhibitory properties; however, deletion of the six amino acids in the ligand binding domain needed for binding coactivators abolished the interference (214). Therefore, squelching of proteins has important consequences for gene regulation. For example, the coactivator proteins RIP140 and TIF2 compete for a common binding site on the glucocorticoid receptor (GR), allowing TIF2 to relieve the inhibitory effect of RIP140 on GR action (180). Although it is not clear if ER-containing neurons also coexpress SRC-1, the widespread distribution of SRC-1 in the brain makes this likely (10). Also, many ER containing VMH neurons coexpress SRC-1, thus making such alleviation of inhibition possible in vivo (10). Therefore, the ability of the TR1 isoform to bind DNA as well to squelch coactivators such as SRC-1 provides a rationale for the inhibition observed with this isoform.

The PPE gene plays a role in analgesic responses which can help the female to put up with somatosensory stimuli during mating which otherwise would be treated as noxious (17). Therefore, PPE gene induction represents a causal route which allows us to link a hormone's genomic effects with a specific behavior, lordosis. In the rat VMH, PPE mRNA in the afternoon of proestrous was significantly higher than diestrous (56). In the female ewe, PPE mRNA increased in the VMH both during lactation and with estrogen treatment (21). On a single dose of 17-estradiol-3-benzoate given to female ovariectomized mice, PPE mRNA was upregulated in the VMH, medial amygdala (MeAmyg), and arcuate nucleus (ARC) at 24 and 48 h (154). However, PPE is not regulated in the caudate putamen or in the cortical amygdala by estrogens (154). A single dose of estradiol benzoate to ovariectomized female rats resulted in a biphasic increase in PPE mRNA producing cells both in the ARC and VMH with a peak at 48 h (151). This biphasic response of PPE consists of a primary peak at 1 h and a second peak between 24 and 48 h postinjection. The rapid first peak was stress induced and could be blocked by adrenalectomy or constant low levels of corticosterone. A peak of plasma corticosterone also coincided with this peak. In the medial amygdala, the antiestrogen tamoxifen blocked the second peak of PPE mRNA expression. These data indicate that both steroids and noxious mild somatosensory stimuli interact to give increases in PPE expression (177). This is consistent with a role for PPE in female reproductive behavior. Acute, mild stress in the form of male approach behavior may activate limbic and hypothalamic circuits known to be important for the full display of reproductive behavior (177).

The OTR gene is also critical for reproductive success. Infusion of oxytocin into the VMH increases sexual behavior and maternal behavior. How does it do so? It is thought that most laboratory tests for sexual behavior and maternal behavior involve unfamiliar, novel and potentially threatening surroundings for rodents. Exposure to such apparatuses or to an unfamiliar animal could trigger inhibitory stress responses. In Swiss Webster mice pretreated with estrogens, peripherally administered OT increased entries into open arms in the elevated plus maze (108). In mice given intracerebroventricular injections of OT, entries into open arms were increased compared with mice given arginine vasopressin (108). Therefore, estrogen upregulation of OT could be a vital component in anxiolytic actions, decreasing stress and facilitating social interactions (109).

Therefore, both PPE and OTR are downstream genes with proven behavioral roles and are upregulated by a primary reproductive effector, estrogen. These downstream gene products are expressed in behaviorally relevant neurons that possess ERs. Indeed, they can be visualized as systems that link a small hormonal signal with an identifiable behavior.

	III. PHYSIOLOGICAL DATA AND THEIR IMPLICATIONS

Top Previous Next References

Neuronal and genetic mechanisms for lordosis behavior have been worked out in such detail (reviewed in Ref. 145) that the behavior virtually stands as an expression system for ER transcriptional activation. In turn, interactions between liganded TRs and ER function can be charted thereby.

The neuroanatomy of estrogens liganded to ER or ER has been charted in considerable detail (144, 146, 175). VMH neurons binding estrogens sit on top the lordosis behavior neuronal circuit (148). Hormone implant experiments establish that it is the estrogenic sensitivity of these neurons that accounts for hormonal facilitation of lordosis. New RNA and protein synthesis are required for the behavioral facilitation. Specific, hypothalamically expressed genes have the following properties: they are turned on by estrogens, and their products facilitate lordosis (150). Therefore, in the manner of a logical syllogism, their induction comprises part of the mechanisms by which estrogenic hormones turn on the behavior.

These particular genes in no way exhaust the possibilities of hormone-stimulated messages, which are behaviorally relevant. New DNA-microarray experiments have revealed hitherto unimagined genes that are hormone sensitive (113) which indicate linked glial/neuronal and possible leptomeningeal/neuronal cooperation in neuroendocrine function.

Normal gene expression for ER is required for normal lordosis behavior (128, 132). In dramatic contrast, active gene expression for ER actually suppresses lordosis; ER knock-out female mice show the behavior during a larger portion of their estrous cycles than wild-type female littermate controls (126).

Therefore, estrogen-dependent lordosis behavior is well suited to look for thyroid/estrogenic interactions. Thyroid hormone administration, in fact, reduces lordosis behavior performance both in female rats (38) and in female mice (114). The molecular mechanisms involved are not necessarily simple and require further investigation; that is, against all predictions, the contribution of the TR gene to the regulation of female reproductive behavior is diametrically opposite to that of the TR gene (36).

B.  Differences in Isoforms From Nuclear Receptor Genes: Use of Knock-out Models

1.  TR gene knock-out mice

Differences in the physiological roles of the different isoforms have been explored using knock-out mice models. Despite similar ligand binding characteristics, the differential distribution of the TR isoforms as well as data obtained from knockout mice suggest a unique role for the various TR isoforms. The TR1 and the TR isoforms have both common and specific roles in vivo. Predicated on the high concentration of TR2 in the anterior pituitary (69), the TR2 isoform plays a major role in the negative-feedback regulation of TSH by thyroid hormone. Lack of TR2, therefore, causes hyperthyroidism in mice (2). On the other hand, lack of the TR1 isoform results in a mild hypothyroidism in mice (75). The distribution of the TR2 isoform in the developing retina of the mouse is also indicative of an important developmental role for this isoform. In rodents, cones contain different opsins sensitive to different wavelengths. The TR2 isoform is responsible for a commitment to M-cone (M, middle or green wavelengths) identity. Deletion of this isoform results in a lack of M cones and a concomitant increase in S-opsin (S, short wavelength) immunoreactive cones (119). The inability for one isoform to substitute for another is also exemplified by T₃-controlled type 1 deiodinase expression (6). Although both TR1 and TR are present in both liver and kidney, expression of the deiodinase was highly dependent on TR in the liver and completely dependent on TR in the kidney (6). Another example, the TR1 isoform, is implicated in hearing loss, whereas the TR knock-out mice remain unaffected (2).

The TR and TR isoforms also play distinct roles in the facilitation of lordosis in female mice. Deletion of the TR1 isoform resulted in decreased lordosis behavior in female mice, whereas loss of the TR isoforms resulted in increased lordosis (36). OT immunoreactivity in the paraventricular nucleus (PVN) was elevated in TR knock-out female mice treated with estradiol compared with wild-type mice given the same treatment, implicating OT increase in the PVN as important in increased lordosis (36). Both behavioral and molecular data on the cross-talk between the ER and TR isoforms on the PPE and OTR promoters point to opposing effects of the TR and TR isoforms.

Reproductive and affiliative behaviors are differentially affected according to which of the ERs is deleted in mice. ER knock-out (ERKO) female mice show virtually no lordosis (128, 132), whereas ER knock-out mice (BERKO) not only show normal lordosis behavior but express this behavior during a larger portion of the estrous cycle than wild-type littermate controls (126). ERKO females have striking deficits in maternal behavior. Dramatically, aggressive behaviors in young adult BERKO males are heightened (124), whereas they are markedly suppressed in ERKO males (131). Note that there appears to be genotype/age interactions in aggressive behavior by BERKO mice, in that the relatively inexperienced young BERKO mice are more aggressive in resident-intruder tests. Finally, the increase in locomotor activity in both genetic females and genetic males, following estrogen administration, depends absolutely on the patency of the ER gene but not the ER gene (127).

It has been suggested that the inability of ERKO mice to induce OTR in response to estradiol benzoate treatment (212) is a factor in their failure to promote social interactions. Antagonistic effects of two ER isoforms expressed in the same cell have also been reported. On a consensus ERE promoter in HeLa cells, Hall and McDonnell (66) have noted that coexpression of ER along with ER reduces the transactivation seen with ER. We were interested in investigating if there is a similar effect on a physiological promoter in both the cell lines. However, although ER did not affect the transactivation observed with the ER isoform on the OTR promoter in response to 17-estradiol in the CV-1 cell line (Fig. 5), it decreased the transcriptional activation observed on this promoter in the SK-N-BE2C cell line (Fig. 5). Again, cell-specific effects such as the expression of tissue-specific cofactors may play a role in this phenomenon. The inability of estrogen-liganded ER to activate transcription from a physiological OTR promoter may also help explain a result obtained from ERKO mice. The OTR gene promoter is upregulated by estrogens in several brain regions. Unlike in the rat, estrogens do not induce OTR expression in the mouse hippocampus but induce it in the cortex (212). However, the ERKO mouse, which does not have ER but has ER, lacks estrogen induction of OTR in many brain regions, such as the cortex, as monitored by OT binding (212). This may be explained by the differential distribution of ER isoforms as well as the inability of the ER to induce the OTR. In the mouse hippocampus, ER expression is sparse, although ER expression is intense (174). It has been suggested that ER distribution in the rat brain results in overall "global" functions for ER in estrogen action, augmenting cognitive processes and neuronal regeneration (176).

View larger version (18K): [in this window] [in a new window]   Fig. 5. Effect of the coexpression of ER and ER on oxytocin receptor (OTR) gene transcription in CV-1 and SK-N-BE2C cells. Both ER and ER expression plasmids were cotransfected into CV-1 and SK-N-BE2C cells at equal concentrations along with the OTR-luciferase construct. A corresponding set of samples received the ER or the ER expression construct alone with the OTR-luciferase reporter construct. After treatment of each set with 107 M 17-estradiol or ethanol for 48 h (n = 6/treatment group), cells are lysed and assayed for -gal and luciferase activity. The results are analyzed using ANOVA followed by Student-Newman-Keuls post hoc test to compare between treatment groups. CV-1 cells: *P < 0.001 compared with vehicle treatment of the ER group. #P < 0.01 compared with vehicle treatment of the ER and ER group. SK-N-BE2C cells: P < 0.01 when compared with vehicle treatment of the same group. P < 0.01 when compared with estrogen treatment of the ER group. [Modified from Vasudevan et al. (189).]

However, the cortex of the mouse expresses both ER and ER isoforms. Although the ER isoform consistently supported 17-estradiol induction of the rat OTR promoter in both cell lines tested, the ER isoform failed to induce the OTR gene promoter under identical conditions. This would predict that brain regions rich in ER but poor in ER such as the hippocampus would fail to support a significant induction of OTR gene expression in wild-type mice. When the predominant ER isoform present in brain areas is the noninducing ER isoform, this is precisely the case. The inability of ER to induce transcription in response to 17-estradiol has also been reported in transiently transfected NG108-15 neuroblastoma glioma cells using a neuropeptide Y-Y1 receptor gene promoter (117). This receptor promoter has two half-site ERE, and estrogen-mediated transcription is strictly dependent on the presence of transfected ER. Coexpression of both ER isoforms abolished the ER-mediated transactivation, suggesting an antagonistic effect of ER on this physiological promoter (117).

Do relative amounts of ER and ER contribute to this result? As monitored by [³H]estradiol binding, ER and ER appeared to be expressed equivalently to each other in both CV-1 cells (190) and SK-N-BE2C (189) cell lines. Although not statistically significant, ER binding to [³H]estradiol appeared to be slightly lower and may have contributed to the noninducibility of the rat OTR promoter by ER. In contrast, in CV-1 kidney fibroblast cells (Fig. 3B), ER was capable of mediating the 17-estradiol induction of the natural PPE gene promoter fragment. Also, ER is capable of promoting neurite elongation in SK-N-BE2C human neuroblastoma cells in response to added 17-estradiol, thus proving that both cell lines are responsive to estrogens via both ER and ER (140). Therefore, the nonresponsiveness of the OTR gene promoter on transfection of the ER isoform in both the cell lines tested is promoter and ER isoform specific. The lower transcriptional efficiency of ER has been noted using the consensus ERE construct, ERE-tk-luc, in COS-1 and HepG2 cells. With the use of Gal4 DNA binding fusion proteins fused to the AF-1 domains of either ER or ER, it was determined that the AF-1 activity of ER was negligible compared with ER (32). On promoters and in cell lines which require both AF-1 and AF-2 activity, ER appears to be a poorer transcriptional activator than ER (32).

The above data suggest that isoforms deriving from closely related genes for nuclear receptors play unique roles that are clearly not equivalent in whole animal studies. The ERKO mice do not exhibit the same behavioral phenotype as the BERKO mice in several behavior tests, especially those designed to elicit socially motivated responses. Similarly, the TRKO females treated with estrogens do not display the same levels of sexual receptivity as the TRKO females (36). Hence, the KO are not equivalent to the KO mice in either ER or TR knock-out models. Table 2 highlights theoretical scenarios for functional relations between ERKO and BERKO gene products. On the right side of Table 2 are summarized some of the data gathered from genetic females and genetic male mice on a number of behavioral and histochemical assays. The variety of relations between ER and ER are reminiscent of the differences seen between ER and ER in their molecular interactions, reviewed above, with a given TR isoform in the cell culture model systems. For both the molecular and the behavioral studies, the specificities of interactions among ER gene products and TR gene product isoforms additionally provide us with internal controls.

View this table: [in this window] [in a new window]   Table 2. Behavioral phenotypes shown by ERKO and ERKO mice

In the early 1940s, Beadle and Tatum (13), in their famous genetic and biochemical studies on the fungus Neurospora crassa, found that the loss of an enzyme led to a specific biochemical defect. The underlying cause of lack of enzyme activity was discovered to be a loss of a gene, thus leading to the well-known "one gene-one enzyme" hypothesis (13, 158, 183). The data presented above show us the combinatorial possibilities of nuclear receptor gene product interactions. These force us to redirect our thinking into a new, more organismal framework whereby patterns of gene expressions and interactions in the central nervous system underlie patterns of behavior.

D.  Physiological Implications of Thyroid Hormone Modulation of Estrogen Action

1.  Neuroendocrine data

The involvement of thyroid hormone in the neuroendocrine control of reproduction has been documented especially well in starlings and sheep. In these species inhabiting temperate latitudes, seasonal reproduction ensures the birth of young in conditions that maximize survival. Therefore, the termination of the breeding season in sheep and the initiation of anestrous occur during the long-day period (spring and summer). An endogenous rhythm, which is entrained by such changes in day length, controls the timing of seasonal reproduction (78, 166, 167). In starlings, thyroidectomy of starlings prevented the start of photorefractoriness and allowed for continuation of the breeding season (201). Thyroxine rise during long days in European starlings is permissive for the neuroendocrine shift to the nonbreeding season, which is primarily dictated by day length (15). In the thyroidectomized male American tree sparrow, administration of T₄ given intracerebroventricularly could restore all components of seasonality. Therefore, T₄ was capable of acting centrally to program already photostimulated male American male sparrows (203).

However, in mammals, both retinal photoreceptors and the pineal gland are required for reproductive responses to photoperiod (120). Long days induced a drop in lutenizing hormone (LH) in thyroid-intact ewes, though thyroidectomy blocked this effect (112). However, thyroidectomy had no effect on the circadian pattern of circulating melatonin or prolactin and the change that occurs with the photoperiod (34). Thyroid hormones may play a permissive role to photoperiod; they need to be present at the end of the breeding season for anestrous to commence. This critical period of neuroendocrine responses to thyroid hormone is late in the breeding season. The minimal effective duration of exposure to circulating levels of thyroid hormone was 60-90 days beginning in late December such that anestrous could develop in spring (184). Do thyroid hormones also play a role in the maintenance of anestrous once it develops? In ewes thyroidectomized (THX) just as they entered anestrous, the timing of the LH rise late in anestrous, indicative of the next breeding season, was the same as non-THX controls. Therefore, although thyroid hormones play a role in initiating anestrous, they do not have any role in the maintenance of the anestrous and the timing of the subsequent breeding season (185) in sheep. However, in the male American tree sparrow, T₄, T₃, and reverse T₃ given intracerebroventricularly could allow for thyroid hormone-dependent photoperiodic testicular growth. The order of potency was T₄ > T₃ > reverse T₃ (203). These data demonstrate that thyroid hormone may play slightly different roles in the maintenance of nonreproductive conditions in mammals and birds.

The mechanisms of thyroid hormone-induced anestrous in sheep have centered mainly on the gonadal hypothalamic-pituitary axis. There is no effect of THX in female ewes on the ability of a rise in estrogens to elicit the LH surge or in the ability of progesterone to suppress LH secretion (199). However, there is intensified estrogen-mediated negative feedback in control ewes compared with THX ewes (199). High-frequency pulses of both GnRH and LH are observed in THX ewes that did not make the transition to anestrous (198). Central infusion of T₄ to THX and ovarectomized ewes given Silastic implants of estradiol benzoate restored anestrous to these ewes. This demonstrated that thyroxine acts centrally in the brain in ewes to promote changes in GnRH and LH that signal anestrous (194). Also, TR has been colocalized in 46% of GnRH neurons in sheep (74). In rats, hyperthyroid rats had 25% less LH on proestrous, showing depression of the LH surges. Also, the amount of estrogens required to initiate a LH surge was greater in hyperthyroid animals (53). Hyperthyroid animals could, however, respond to GnRH, suggesting that the pituitary was not a site of action for thyroid hormone. The hypothalamus is more plausible, since stimulation of the arcuate nuclei-median eminence area (ARC-ME) resulted in hyperthyroid rats secreting less LH than control rats (53). In rats devoid of the thyroid gland, the synthesis and metabolism of LH was not affected, but the secretion of LH was higher (52). Propylthiouracil (PTU), a goitrogen, given transiently to neonatal rats dramatically increases sperm production and testis size in the adult rat. However, it leads to a significant drop in GnRH-stimulated LH production. Gonadal feedback is enhanced in PTU-treated males resulting in chronically reduced circulating levels of LH and follicle stimulating hormone (86).

Thyroid hormone elevation has also been shown to have an adverse effect of reproduction in rodents (38, 114). Concomitant administration of T₄ to ovariectomized rats (38) and mice (114) treated with estrogens has been shown to reduce lordosis, compared with ovariectomized rodents that received estrogens alone. TR knock-out female ovariectomized and estrogen-treated mice deleted for TR isoforms showed higher lordosis than the TRWT, suggesting that TR may exert an inhibitory influence on ER-controlled reproduction (36).

Because reproductive behavior is controlled by estrogens via the ER, it is possible that a reduction in ER target genes such as OT, OTR, and PPE could be responsible for TR-mediated inhibition (149, 217). Indeed, injections of thyroid hormone to estrogen-treated female rodents lead to a decrease of OT mRNA in the PVN (39). Ex vivo studies indicate that thyroid hormone upregulates the human OT promoter fivefold through the composite element containing an imperfect ERE located at 148/172 bp upstream of the transcriptional start site (4). TR1 protein can bind to this composite element and interfere with the transcriptional induction by estrogens (4). Thyroid hormone elevation also reduces the expression of another estrogen-induced gene in the VMH, the PPE gene, which facilitates lordosis behavior (37).

In other species, thyroid hormone has also been shown to modulate estrogen action. In the fish tilapia, Oreochromis neoloticus, three distinct populations of GnRH exist: the terminal nerve neurons in the forebrain, the preoptic neurons, and the midbrain neurons. In castrated male tilapia, terminal nerve neurons express GnRH, which is lowered by exogenous T₄ treatment (138). Interestingly, in the tilapia, the ontogeny of terminal nerve neuron GnRH is concomitant with a decrease in T₄ levels. Sexually mature tilapia have low levels of thyroid hormones but high levels of terminal nerve GnRH (197). In oviparous species such as the clawed toad, Xenopus laevis, metamorphosis is dependent on thyroid hormone while vitellogenesis is strongly dependent on estrogens (155). T₃ could enhance ER production and autoinduction and thereby enhance the estrogenic activation of vitellogenin genes (156).

One of the most prominent effects of estrogens is to promote mitosis in the uterine luminal epithelium, stroma, and myometrium. The ability of hypothyroid rats to increase the mitotic index in these uterine regions is reduced compared with the euthyroid controls (87). This diminished uterine response is not due to a shift in the dose-response curve of the estrogen; rather, it is possible that thyroid hormones have a direct effect on the uterus such that it lowers its responsiveness to estrogens (57). In pregnancy, despite lower levels of free T₄ and free T₃, there is no rise in serum TSH (50). A similar absence of TSH rise is seen in postmenopausal women receiving estrogen replacement therapy (1). Although estrogen treatment did not augment the serum concentrations of TSH in euthyroid or untreated hypothyroid rats, it increased the suppressive T₃ effect on serum TSH in hypothyroid animals. An increase in the number of pituitary nuclear receptors for T₃ was seen after estrogen treatment in rats, suggesting that the augmentation of the T₃ effect may be due to increase in thyroid hormone receptors (51). In bone tissue, where estrogens promote bone mineral density, hyperthyroidism has been shown to increase bone turnover and decrease bone density (49). In the pituitary, there was a significant increase in weight and total cellular RNA when ovariectomized rats were given estrogens. This increase was inhibited by concomitant administration of T₃ (216). Estrogen produces an anorectic effect in rats, presumably by acting on the hypothalamus. T₄ given daily subcutaneously could antagonize the estrogen-mediated anorectic effect. It has been proposed that this antagonism could be related to thyroxine's reductive effect on blood glucose level and subsequent decrease in satiety (211).

In the gonadectomized rat, there is a subset of estrogen-driven physiological responses that require thyroid hormonal interplay. These include the estrogen suppression of somatic growth and the effects of estrogens on serum triglycerides. It also includes estrogenic suppression of LH secretion (negative feedback control) (48). Surprisingly, this subset of T₃-dependent estrogen responses also can be promoted by tamoxifen acting as an estrogen agonist. Because T₃ regulates growth and energy metabolism, this may provide an interactive mechanism for relating metabolic state to reproductive biology (48).

Thyroid hormone may also affect the clearance rate of estrogen, although there are conflicting studies on this theme. Hypothyroidism lowers the clearance rate of estrogen in women (102). In another study in women, hyperthyroidism increased the clearance rate of estrogen (164). This is also predicted by data in the male Japanese quail (165). Also, in hyperthyroid rabbits, the clearance rate of estrogen is increased (187). In female Japanese quail, thyroidectomy actually increased the clearance rate of estrogen, contrary to the data in males (141). However, in male cytomegalous monkeys, there was no change in the clearance rate of estrogen on thyroid hormone treatment (18).

	IV. ROLE OF PROMOTER AND CELL SPECIFICITY IN DISTINCT TRANSCRIPTIONAL RESULTS

Top Previous Next References

How do differences in sequences bound by the ER and TR isoforms explain their transcriptional differences in the context of cell lines and promoters? Different hormone response elements are allosteric mediators of receptor conformation. Therefore, hormone response elements not only position receptors close to basal transcription complexes but also serve to direct the mode of regulation of the target gene (100). Studies with the consensus vitellogenin A2 ERE, or the imperfect pS2, vitellogenin B1 or oxytocin (OT) ERE show that the A2 was the most potent activator of transcription followed by the OT ERE (205). DNase footprinting revealed that MCF-7 proteins protected the OT and A2 EREs to a greater extent than the pS2 or B1 EREs. Although the receptor-interacting domains of the glucocorticoid receptor interacting protein 1 (GRIP) and SRC-1 bound effectively to ER, TIF2 was bound less by B1-bound ER than A2-bound ER, suggesting that allosteric modulation of ER conformation by different EREs influences coactivator recruitment (205).

Different isoform conformations within the cell could also have an effect in the recruitment of coactivators. The TR2 isoform, for example, can bind p160 class of coactivators in the absence of the hormone and, therefore, mediate ligand-independent activation of target genes. This is mediated by contacts in the unique NH₂ terminus of TR2 and an internal interaction domain of SRC-1 and GRIP-1 coactivators. These contacts are different from the LXXLL motifs that mediate hormone-dependent coactivator contacts and hence hormone-dependent transcriptional activation (206). The NH₂-terminal region in the human PR (hPR) also modulates differential coactivator and repressor binding. The hPR exists as two different isoforms: hPRA, which is a strong ligand-dependent repressor of transcription, and the hPRB, which is a transcriptional activator in most cell and promoter contexts. An inhibitory domain (ID) present in the hPRA and B is active only in the hPRA isoform, facilitating the interaction between this isoform and the transrepres