| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Phosphorylation
Department of Molecular Endocrinology and Bone Biology, Merck Research Laboratories (L.L., J.E.F., N.W., A.S., Q.S., J.Y., F.C., S.W., E.T.B., A.S., S.P.R., L.P.F., A.A.R.), West Point, Pennsylvania 19486; Department of Medicinal Chemistry, Merck Research Laboratories (S.K., H.Y.C., Q.T., F.D., M.L.H.), Rahway, New Jersey 07065; and Department of Biochemistry and Biophysics, University of Pennsylvania School of Medicine (G.A.R.), Philadelphia, Pennsylvania 19104
Address all correspondence and requests for reprints to: Dr. Alfred Reszka, Molecular Endocrinology and Bone Biology, Merck Research Laboratories, WP26A-1000, West Point, Pennsylvania 19486.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
(ER) serine 118 (Ser118) phosphorylation modulates activation function-1 (AF1) function. Correct positioning of helix 12 promotes agonist-dependent recruitment of cyclin-dependent kinase-7 to catalyze this event. In this study we show robust cyclin-dependent kinase-7-independent, AF2 antagonist-induced Ser118 phosphorylation. Estradiol (E2) and ICI-182,780 (ICI-780) induce Ser118 phosphorylation of wild-type ER and either of two helix 12 mutants, suggesting AF2-independent action, probably via shedding of 90-kDa heat shock protein. With E2 treatment, the predominantly nuclear, phosphorylated ER in COS-1 cells is detergent soluble. Although levels of ICI-780-induced phosphorylation are profound, Ser118-phosphorylated ER is aggregated over the nucleus or in the cytoplasm, fractionating with the cell debris and making detection in cleared lysates improbable. Selective ER modulators (SERMs) elicit a mixed response with phosphorylated ER in both detergent-soluble and -insoluble compartments. Apparent ligand-induced loss of ER protein from cleared lysates is thus due to ligand-induced redistribution into the pellet, not degradation. The COS-1 response to ICI-780 can be mimicked in MCF-7 cells treated with a proteasome inhibitor to block authentic ligand-induced degradation. With SERMs and antagonists, the magnitude of Ser118-phosphorylated receptor redistribution into the insoluble fraction of COS-1 cells correlates with the magnitude of authentic ER degradation in MCF-7 cells. A strong inverse correlation with ligand-induced uterotropism in vivo (P < 0.0001) and direct correlation with AF2-independent transrepression of the matrix metalloprotease-1 promoter in endometrial cells in vitro are seen. These data suggest that ligand-induced Ser118 phosphorylation of ER can be AF2 independent. Furthermore, they identify translocation of Ser118-phosphorylated ER out of the nucleus, leading to cytoplasmic aggregation, as an antagonist pathway that may precede receptor degradation. |
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
(ER) transcriptional activity and phosphorylation of serine 118 (Ser118) in the activation function-1 (AF1) domain can be induced by ligands such as estradiol (E2) and hydroxytamoxifen (TOT) or by signaling through the MAPK pathway (1, 2). The importance of Ser118 phosphorylation is suggested by the impact of this residue on transcriptional regulation. Mutation of Ser118 to alanine or glutamate can, respectively, suppress or enhance agonist-induced transcription using various promoters (1, 3). In mutants lacking the AF2 domain, these Ser118 replacements can have a similar impact on ligand-independent transcription in a promoter- and cell context-based fashion. Regarding mitogen-induced transcription, the alanine substitution also renders the full-length receptor refractive to Ras-induced transcription from an estrogen response element (2). Although there is no link between Ser118 phosphorylation and DNA binding, it does have an impact on transcription, even in cell-free systems (4, 5). This might be related to the requirement of phosphorylated Ser118 for effective recruitment of coactivators, such as steroid receptor RNA activator (6), or for blocking interaction with corepressors, such as nuclear receptor corepressor (7).
Despite the fact that both ligand and mitogens can stimulate phosphorylation of the same residue on ER (Ser118), the ligand-dependent pathway is independent of MAPK signaling (8, 9). In this regard, MAPK inhibition can completely abrogate mitogen- or phorbol ester-induced Ser118 phosphorylation, but it has no effect on E2-stimulated phosphorylation at the same site. In a search for the relevant kinase involved in ligand-dependent Ser118 phosphorylation of ER, cyclin-dependent kinase-7 (cdk7) was identified as the mediator of this effect (10). E2-induced phosphorylation was linked to recruitment of cdk7 via its inclusion, along with Leu-x-x-Leu-Leu-bearing p62, as a component of transcription factor II H (TFIIH). Agonist-induced recruitment of the complex leads to phosphorylation in a helix-12-dependent fashion in vitro. This, in consideration of the more modest stimulation of Ser118 phosphorylation by TOT vs. E2 and the near absence of response to ICI-182,780 (ICI-780) in cleared lysates from transfected cells, suggests an AF2-dependent phosphorylation event (9, 10).
In contrast to this very compelling model, however, certain other studies have suggested that AF2 antagonists can also stimulate ER phosphorylation of Ser118. In the earliest work describing ligand-dependent phosphorylation of ER, both TOT and ICI-164,384 (ICI-384) were identified as ligands eliciting Ser118 phosphorylation (1). ICI-384 was about one third as effective as E2 in inducing the response. Others have shown equivalent levels of Ser118 phosphorylation elicited by E2, TOT and ICI-384 (3, 8). The reasons for these apparent differences in antagonist responses have not been resolved to date. In any case, putative phosphorylation by these means would not be predicted to be mediated through cdk7/TFIIH, based on the lack of responsiveness of this complex to treatment with antagonist, as discussed above. This suggests the possibility of an alternate mechanism.
Generally accepted mechanisms for transcriptional antagonism by ligands such as TOT, ICI-384, and ICI-780 include impaired receptor dimerization (11), the formation of unique conformations capable of recruiting corepressors (12), and proteasome-mediated receptor degradation (13). With partial antagonists such as TOT, the recruitment of corepressors vs. coactivators determines whether the ligand serves in the agonist or antagonist mode. The ratio of corepressor to coactivator does not seem to have an impact on the transcriptional activities of the pure antagonists, which may act primarily through their induction of the degradation pathway. To this extent, ligands such as ICI-780 and GW7604 can be distinguished from TOT and raloxifene (RAL) based on their ability to trigger ER degradation (14, 15). In this regard, the ligands have a clear effect of reducing the levels of ER in nuclear extracts or cleared lysates of cells treated with these antiestrogens. Some studies have examined the effects of ICI-384, ICI-780, and RU 58668 in comparison with those of TOT and concluded that these pure antagonists may act by disrupting nucleoplasmic shuttling and excluding the receptor from the nucleus (16, 17, 18). Interestingly, RU 58668 was shown to diminish the levels of ER in the cytosol and nuclear extracts, but this did not represent actual receptor degradation. Instead, the receptor was found in the cytosolic pellet.
In the present study, we examined ligand-dependent phosphorylation of ER in COS-1 and HEC-1A cells in response to treatment with both agonist and antagonist. The data suggest that in addition to the agonist-induced, AF2-dependent mechanism, there is a robust antagonist-induced, AF2-independent, cdk7-independent phosphorylation mechanism. Depending on the ligand, the phosphorylated receptor under these circumstances may not necessarily be detected in cleared lysates. However, it is most often abundantly present in the extranuclear Triton-insoluble compartment. The proportion of the receptor that remains Triton soluble after ligand treatment correlates with AF2-independent transcriptional regulation at AP1 and with uterotropism in vivo.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Ser118 Phosphorylation phosphorylation at Ser118 using both endogenous receptor in MCF-7 cells and transfected human ER in COS-1 cells. To account for all receptor in these cell types, we examined both Triton-soluble receptor found in whole-cell cleared lysates and insoluble receptor, which pellets with the cell debris (18). In MCF-7 cells treated for 30 min or 4 h, ICI-780, but not E2, induced a decline in ER protein expression, which occurred in the Triton-soluble compartment (Fig. 1
, A and B). Degradation was more extensive by 4 h, and this could be blocked by treatment with the proteasome inhibitor, MG132 (Fig. 1C). Quantitative analyses showed that in the absence of the proteasome inhibitor, the relative proportion of ER found in the lysate pellet (vs. the Triton-soluble lysate) rose with ICI-780, but not E2, treatment (Fig. 1B, bottom panel). Interestingly, inhibition of proteosomal degradation by treatment with MG132 caused a substantial rise in the relative proportion of ER in the Triton-insoluble pellet (Fig. 1C). Indeed, most of the receptor saved from proteosomal degradation by MG132 (
70%) was found in the Triton-insoluble compartment.
|
-transfected COS-1 cells. In the dimethylsulfoxide (DMSO)-treated control, the distribution of ER in cleared lysates and in Triton-insoluble pellets was fairly even (Fig. 1
, A and E). Treatment with E2 did not substantially disturb the distribution between these compartments, but ICI-780 shifted most of the receptor into the Triton-insoluble fraction. This response to ICI-780 was similar to that seen in MG132-treated MCF-7 cells. If analyses were limited to the cleared lysate fraction, this could convey the impression that the receptor is degraded in these cells, as has been previously reported for antagonist effects in this same cell background (1). However, in considering ER accumulation in the Triton-insoluble compartment, there was no apparent ligand-induced degradation in COS-1 cells even after 24-h exposure to either ICI-780 or E2 (Fig. 1E). Although high ER expression in COS-1 cells could impede effective degradation of ER in response to ligand, Coomassie staining revealed no identifiable bands corresponding to ER in the Triton-insoluble fraction after treatment with ICI-780 in comparison with DMSO or E2 controls (data not shown), suggesting that expression levels did not approach those of other recognizable markers, such as actin. Because the receptor accumulates in the lysate pellet in response to ICI-780, the data suggest no net loss of expression. A similar lack of antagonist-induced degradation was previously reported when anionic detergent was used to prepare cleared lysates (3).
Analyses of Ser118 phosphorylation showed a generally low basal phosphorylation for the expressed ER in COS-1 cells (DMSO control; Fig. 1A), although a weak signal could be detected in the Triton-insoluble compartment in several experiments (Figs. 1E, 2B, 5, and 7). There was also measurable basal phosphorylation in the MCF-7 background in both fractions (Fig. 1A, DMSO control). As expected, treatment of MCF-7 cells with E2 induced a rise in Ser118 phosphorylation, although the magnitude of the effect was only about 2-fold. In ER-transfected COS-1 cells treated with E2, there was a robust, time-dependent rise in predominantly Triton-soluble Ser118-phosphorylated ER (85.1 ± 8.0% of total phospho-Ser118 signal; n = 48), with higher specific phosphorylation at 4 h vs. 30 min of treatment. A more detailed time-course analysis in COS-1 cells showed a steady time-dependent rise in phosphorylation, with peak levels occurring at 2–4 h, the latter time point giving more consistent results (Fig. 7A). A slightly higher phospho-Ser118 signal at 24 h (Fig. 1E) was attributable to somewhat higher expression at this time point. Dose-dependent effects of E2 were generally maximal at 100 nM with some exceptions for which a higher dose was required; thus, a ligand concentration of 1 µM was chosen for most experiments to ensure a consistent and robust response. This concentration was also used for most selective ER modulators (SERMs) and antagonists, although dose responses were also established for certain ligands (see below).
|
|
|
, A and E). Interestingly, a striking rise in Ser118-phosphorylated receptor was detected in the Triton-insoluble fraction (95.9 ± 3.9% of total phospho-Ser118 signal; n = 53) of COS-1 cells. As with E2, the rise in phospho-Ser118 stimulated by ICI-780 was maximal after a 4-h treatment period, and there was no diminution by 24 h (Fig. 1E). Low concentrations (3 and 10 nM) of ICI-780 induced no detectable change in the overall distribution of ER between the Triton-soluble and -insoluble pools (data not shown). At these low doses, the Ser118-phosphorylated receptor reached, respectively, 1% and 10% of the maximum seen at 1 µM, and the signal was exclusively limited to the Triton-insoluble compartment. The absence of the phosphoreceptor in cleared lysates at all tested concentrations of ICI-780 explains the inability to detect this ligand’s phosphorylation-inducing effect in previous studies (9, 10).
In the MCF-7 background, we (Fig. 1A) and others (14) have observed that ligands such as ICI-780 induce ER degradation. Additional examination in MCF-7 cells showed that there was no net decrease in Ser118 phosphorylation in response to ICI-780 treatment, although overall protein expression had declined (Fig. 1, A and B). Thus, the specific phosphorylation of ER rose in response to this antagonist. When these analyses were extended to include MCF-7 cells treated with MG132 (Fig. 1C), there was a more than 10-fold induction in total Ser118 phosphorylation after ICI-780 treatment. As with COS-1 cells, the majority (88% average) of Ser118-phosphorylated ER was recovered in the Triton-insoluble compartment. Because there was a similarity in the responses seen in COS-1 vs. MG132-treated MCF-7 cells, we hypothesized that the Triton-insoluble compartment could somehow be related in both cell types. To test this, we looked for a correlation between relative levels of ER remaining after ligand treatment (percent expression vs. DMSO control) in MCF-7 cells (Fig. 1D, y-axis) and the percentage of Ser118-phosphorylated receptor in COS-1 cells that remained Triton soluble after ligand treatment (x-axis). The analysis included ICI-780, Ral, TOT, and 18 different ER-selective SERMs (SERMs) from the dihydrobenzoxathiin and chromane classes (19, 20). Linear regression analyses showed a strong positive correlation (r = 0.68; P = 0.0007). The lack of substantial degradation in COS-1 (vs. MCF-7) cells made it an ideal system for investigation of antagonist-induced Ser118 phosphorylation, because the liganded receptor remains intact for subsequent analyses.
AF2-Independent Ser118 Phosphorylation
In an effort to define more clearly the role of the AF2 domain in eliciting the ligand-dependent phosphorylation response, we examined the effects of E2 and ICI-780 on ER AF2 mutants, L540Q and D538A/E542A/D545A (triple; Fig. 2A). The Ser118Ala (S118A) mutant was also included to confirm the specificity of the antiphospho-Ser118 ER antibody. In comparison with the wt receptor, which showed a robust induction of ER Ser118 phosphorylation in cleared lysates (Fig. 2A, top panels), the S118A mutant showed no detectable phosphorylation response to E2 treatment despite comparable levels of receptor in Triton-soluble and -insoluble (lysate pellet) compartments. Similarly, there was no detectable Ser118 phosphorylation signal from the S118A mutant in the Triton-insoluble fraction after treatment with ICI-780 (Fig. 2A, bottom panels). With this mutant, the ER protein still accumulated in the Triton-insoluble compartment with normal proportions, suggesting that Ser118 phosphorylation was not required for redistribution of ER after antagonist treatment.
In a previous study, the triple mutant showed only AF1-mediated (minimal) transcriptional induction in response to E2 in CV-1 cells, the parental cell line for COS-1 (21). The L540Q mutant is transcriptionally dead in response to E2 in a promoter- and cell context-dependent fashion, but it can confer transcriptional activity to the pure antagonist, ICI-384 in several cell types (22, 23). In response to E2, the triple mutant showed no loss of ability to induce Ser118 phosphorylation, and the phosphorylated receptor remained predominantly in the Triton-soluble compartment (Fig. 2A, top). A slight band shift was attributed to loss of the three acidic residues. In the L540Q mutant, a moderate loss of overall Ser118 phosphorylation in response to E2 treatment was observed, and again, the phosphorylated receptor remained mostly Triton soluble.
With ICI-780, the triple mutant showed a partial decline in Ser118 phosphorylation, and the phosphorylated receptor remained mainly within the Triton-insoluble compartment. The L540Q substitution did not block ICI-780-induced Ser118 phosphorylation. However, the response differed from that of the wt receptor in that phosphoreceptor was recovered almost exclusively in the Triton-soluble fraction (Fig. 2A, compare top and bottom panels), and the magnitude of phosphorylation was more comparable with that seen with the wt receptor when treated with E2. This reflects the effect of the L540Q substitution to allow some agonist activity in response to an otherwise pure antagonist, as noted above. Together, the Ser118 phosphorylation responses to E2 and ICI-780 with both wt and AF2 mutant receptor support an AF2-independent mechanism for ligand-induced phosphorylation of ER at Ser118 by antagonist.
It is now established that AF2-dependent ER Ser118 phosphorylation is catalyzed by cdk7, which acts as a component of TFIIH, whereas mitogens act on ER via MAPK (8, 9, 10). In light of evidence that antagonist-dependent ER Ser118 phosphorylation may be AF2 independent, we also investigated whether cdk7 or MAPK activities were required for the response to ICI-780 (Fig. 2B). ER-transfected COS-1 cells were pretreated for 1 h with U-0126 or the cdk inhibitor, R-roscovitine, as described in Materials and Methods. Cells were then treated with E2 or ICI-780 for 4 h, and the phosphorylation response was followed by immunoblot analyses. For quantification of the response, relative ER protein expression and Ser118 phosphorylation were normalized vs. the E2 effect (Fig. 2B, lanes 7 and 8), as indicated below each treatment group. In comparison with cells treated with DMSO alone, neither of the two kinase inhibitors altered the distribution of ER or Ser118-phosphorylated ER to Triton-soluble or -insoluble fractions. For simplicity, therefore, phosphorylation levels are reported as the sum of the amounts detected in the Triton-soluble and -insoluble fractions. R-Roscovitine, but not U-0126, lowered total ER expression levels. The effects of the inhibitors on ER Ser118 phosphorylation were therefore examined in the context of ER expression levels, and specific phosphorylation of the receptor (total relative phosphorylation/total relative ER protein level) was calculated for each treatment, as indicated below each test group. For these analyses, the total specific phosphorylation level was 1.0 with E2 treatment (reference group) and 0.77 with ICI-780 treatment.
Consistent with a previous report (8), MAPK inhibition with the MAPK kinase inhibitor, U-0126, had little impact on ligand-induced phosphorylation, and total specific phosphorylation was 0.75 with E2 (compare to 1.0) and 0.69 with ICI-780 (compare to 0.77). Although R-roscovitine caused a 50% decline in total ER expression in the E2-treated group, there was a disproportionately larger decline in total relative Ser118 phosphorylation (76%). Thus, the total specific phosphorylation level was reduced to 0.48 vs. that with E2 alone. This is consistent with previous reports suggesting a cdk7-mediated effect (9, 10). Interestingly, and consistent with an AF2-independent mechanism, R-roscovitine did not block ICI-780-induced Ser118 phosphorylation. Total specific phosphorylation rose to 1.44, which was 1.8-fold greater than that with ICI-780 in the absence of any kinase inhibitor. These data are consistent with cdk7 playing a large role in agonist-induced, but not antagonist-induced, Ser118 phosphorylation.
Cytoplasmic Localization of Phospho-ER in Response to ICI-780
To examine the subcellular localization of the Ser118-phosphorylated ER in response to agonist and antagonist, we compared total ER staining to that of phospho-Ser118 in COS-1 cells. Previous reports have suggested that ER nucleocytoplasmic shuttling in COS-1 cells is disrupted in response to ICI-780 treatment, with more than 90% of the cells showing predominant cytoplasmic localization by 3 h, most of which is aggregated (16). A similar response in green fluorescence protein-ER-transfected MCF-7 cells was reported after 12–16 h of treatment with ICI-780 (17). We found similar translocation of ER from the nucleus to the cytoplasm after 4-h treatment with ICI-780 (Fig. 3A). For these analyses, we examined cells with high (left panels) and low (right panels) expression. In control (DMSO-treated) cells, the receptor was limited to the nucleus with lower expression levels, although there was detectable ER in the cytoplasm with high expression. At higher magnification (inset box), the cytoplasmic staining in these cells was found to be largely diffuse, with essentially no punctate staining. Treatment of ER-expressing cells with ICI-780 for 4 h led to the translocation of ER into discrete punctate structures in the cytoplasm, as expected. High expression seemed to increase the intensity of staining at each discrete site. However, the aggregation effect was also seen in cells exhibiting low expression. This suggests that high overexpression is not required for this cytoplasmic aggregation response to the antagonist.
|
from ICI-780-treated cells was phosphorylated at Ser118 (Fig. 3B). In vehicle-treated controls (bottom panels), ER (rhodamine staining) was predominantly limited to the nucleus, and there was no substantial phosphorylation [fluorescein isothiocyanate (FITC)] signal. Consistent with a previous report (24), E2 treatment resulted in the receptor adopting a somewhat punctate appearance within the nucleoplasm, but not in the nucleoli (Fig. 3B, top left panel). There was a coincident phospho-Ser118 signal in the E2-treated cells, which yielded a yellow composite image limited to the nucleus (top right panel). Treatment of the L540Q mutant with ICI-780 resulted in nuclear staining comparable to that seen with the wild-type (wt) receptor in response to E2 (data not shown). Treatment of the wt receptor with ICI-780 (Fig. 3B, middle panels) caused a redistribution of the ER with three general types of localization: 1) cytoplasmic/aggregated, as described above; 2) perinuclear/aggregated; and 3) highly punctate staining over the nucleus. In most cases the ER signal overlapped with the phospho-Ser118 fluorescence, although the more distal cytoplasmic phosphorylation signal could at times be less intense than that seen closer to the nucleus. Rarely, small spots of non-phosphorylated ER could also be found in the cytoplasm (data not shown). A comparison between biochemical fractionation (Figs. 1 and 2) and immunofluorescence staining suggests that the aggregated phosphorylated ER is the main constituent of the Triton-insoluble fraction in ICI-780-treated cells. Meanwhile, the ER in the nucleus after E2 treatment is mostly Triton soluble.
Heat Shock Protein (90 kDa; HSP90) Inhibition Induces Ser118 Phosphorylation, Aggregation, and Degradation of ER
The strong phospho-Ser118 responses of wt and AF2 mutant forms of ER to agonist and antagonist suggest the existence of alternate, AF2-independent mechanisms that may also be active within these cells. HSP90 interacts with ER and is reported to contribute to the DNA binding affinity of the receptor (25, 26). Interestingly, both agonist and antagonist ligands can trigger dissociation of ER from HSP90 (18). Previous studies have also shown that the HSP90 inhibitor, geldanamycin (GDA), triggers ER degradation in MCF-7 cells (27, 28). Thus, GDA can mimic some effects of an ER antagonist. We therefore examined the possibility that HSP90 inhibition could mimic antagonist effects on ER phosphorylation, subcellular localization, and Triton solubility in our system.
Initial analyses in transfected COS-1 cells focused on subcellular localization (Fig. 4). As with MCF-7 cells, treatment of ER-transfected COS-1 cells with GDA triggered ER degradation within 4 h (Fig. 4, bottom left panel), an effect that was confirmed by immunoblot analyses (Fig. 5). In this regard, responses to GDA differed from those to ICI-780, which did not trigger degradation in COS-1 cells. To examine GDA’s effects on phosphorylation and localization before degradation, we limited exposure to 15 min (Fig. 4, middle panels) and included ICI-780 for comparison (Fig. 4, top panels). At this early time point, there was no detectable GDA-induced change in ER expression (Fig. 5, top blot). In control cells treated with ICI-780, ER (rhodamine red X staining) was ejected from the nucleus by 15 min and was often diffuse within the cytoplasm (Fig. 4, top), although careful examination revealed that aggregates were present within the larger pattern of most cells. Under these circumstances, the Ser118-phosphorylated receptor (inset image in Fig. 4, top left panel) was found in both the diffuse and aggregated compartments. Cells with mostly aggregated receptor could also be found (Fig. 4, top center), although this pattern did not predominate until treatments lasted 1 h (data not shown) or 4 h (Fig. 3). The presence of diffuse ER overlaying the punctate pattern in the cytoplasm at 15 min, therefore, seemed to represent a transient effect that lasted only until the receptor became fully aggregated.
|
to the cytoplasm at 15 min, and the receptor appeared diffuse within this compartment (Fig. 4, middle left). A distinct aggregated pattern was also observed within the pattern in each cell, suggesting an early stage of aggregate formation. In other cells, aggregation was more complete (Fig. 4, middle center). In this regard, aggregate formation in response to GDA treatment preceded receptor degradation, suggesting that neither receptor overexpression nor treatment-induced aggregation is capable of inhibiting degradation in COS-1 cells. Most striking was the observation that the HSP 90 inhibitor, like ICI-780, stimulated robust Ser118 phosphorylation of ER (Fig. 4, middle left and center panels), which was absent in the S118A mutant (middle right panel). As with ICI-780, both the diffuse and aggregated ER was Ser118 phosphorylated in response to GDA treatment (inset image, Fig. 4, middle left panel). Immunoblotting of Triton-soluble and -insoluble fractions revealed that less than half of the Ser118-phosphorylated ER was soluble after treatment with GDA at 10, 15, and 60 min (Fig. 5). Consistent with the immunofluorescence staining (Fig. 4, bottom left panel), there was a substantial reduction in overall ER expression at 4 h, at which time no detectable Ser118-phosphorylated receptor could be found in Triton-soluble fractions. Together, the immunofluorescence and immunoblotting data suggest that HSP90 inhibition can, like ICI-780, trigger ER phosphorylation. The response differs, however, in that GDA triggers receptor degradation in COS-1 cells, whereas ICI-780 does not.
SERMs Exhibit Partial Agonist Effects on ER Ser118 Phosphorylation and Partitioning
SERMs can modulate the activity of ER, with agonist and antagonist responses seen in different cell types. It is generally believed that levels of coactivator can drive the agonistic SERM responses at the ER AF2 domain, as was described for TOT (12). Within the same cell types, different SERMs can also display unique degrees of agonism. For instance, although TOT has been shown to display agonist activities in Ishikawa cells, it is a full antagonist in MCF-7 breast cancer cells (29), whereas RAL is an antagonist in both cell types. TOT was previously found to display only minimum transcription from an estrogen response element in COS-1 cells, although this was greatly enhanced by overexpression of Ras, which leads to MAPK-induced Ser118 phosphorylation (2). This suggests that AF1 activity plays a key role in TOT-induced transcription in this system, although some AF2 activity could also exist. To test for possible AF2 activity in COS-1 cells, we compared TOT and RAL to E2, ICI-780, and four SERMs (Fig. 6A) from the dihydrobenzoxathiin class (19). Using an ER ligand-binding domain (LBD)-galactosidase-4 (Gal4) DBD construct, AF2-specific transcriptional effects of these ligands were tested in both antagonist (Fig. 6B) and agonist modes. None of the ligands displayed any AF2-dependent transcriptional activity when tested in the agonist mode (data not shown). Like ICI-780, all tested SERMs were E2 antagonists with roughly equivalent maximal transcriptional inhibitory activity. Consistent with a model for AF2-independent phosphorylation, all ligands induced robust ER Ser118 phosphorylation (Figs. 6C and 7). Consistent with the greater (uterine) agonism of TOT, the phosphorylated ER with this ligand was predominantly found in the Triton-soluble compartment (78.4 ± 14.1%; n = 4), much like with E2 treatment, as noted above. In contrast, the Ser118-phosphorylated ER with RAL treatment tended to favor the Triton-insoluble compartment with only 28.4 ± 8.4% (n = 19) soluble phosphoreceptor. By this measure, RAL trended to mimic the effects of ICI-780, although the levels of Triton-soluble phosphoreceptor were consistently and substantially higher than with ICI-780 itself. The four different ER-selective SERMs showed activities that ranged between those of RAL and TOT (Figs. 6C and 7B; 38–51% soluble; n = 4 each).
|
). Between the 30 min and 4 h points, the RAL-induced Ser118-phosphorylated ER showed consistent partitioning into both compartments. Similarly, the Ser118-phosphorylated ER partitioned mostly to the Triton-soluble compartment at all time points. In separate analyses, dose effects of RAL were compared with those of SERM-A, which showed similar partitioning characteristics (Fig. 7B; 37.5 ± 12.3% soluble; n = 8). Both ligands induced ER phosphorylation at concentrations as low as 1–10 nM, although maximal effects were not seen below 100-1000 nM. The dose-dependent effects of the ligands led to consistent partitioning of phosphorylated receptor mostly into the Triton-insoluble compartment, as seen in adjacently loaded replicate samples. By this measure, there was no recognizable relationship between ligand dose and the partitioning effect. The consistent relative partitioning of ER in response to each ligand across dose and time suggests that the response is specific and not simply triggered by overexpression of the receptor and/or mass action in these cells. These data are also consistent with an AF2-independent mechanism for Ser118 phosphorylation of ER.
Partitioning of Ser118-Phosphorylated ER Correlates with Matrix Metalloprotease-1 (MMP1) Promoter Transrepression in Endometrial Cells in Vitro and Uterotropism in Vivo
Although TOT, RAL, and ICI-780 have antiproliferative and ER antagonist effects in breast cancer cells, it is well established that TOT is more like E2 in its effect on the uterus than is RAL or ICI-780. Likewise, the partitioning of Ser118-phosphorylated ER with TOT was more comparable to that with E2, whereas RAL had effects closer to those of ICI-780 (Fig. 6C). Based on these initial observations, we hypothesized that partitioning of ER may correlate with biological responses to the ligands. As discussed above (Fig. 1D), a correlation between partitioning of Ser118-phosphorylated ER in COS-1 cells and receptor degradation in MCF-7 cells was observed. Parallel comparisons with MCF-7 proliferation, however, showed no similar relationship (data not shown). This suggests that TOT and other SERMs may have antiproliferative properties in MCF7 cells that are unrelated to ER degradation and/or partitioning. Despite this, we did find other biological responses that seemed to correlate with the partitioning effects of these ligands.
One consistent observation arising from these analyses was that the more uterotropic SERMs and SERMs tended to display greater Triton solubility of the Ser118-phosphorylated receptor in the COS-1 background, and they tended to trigger little or no receptor degradation in MCF-7 cells. To test for a possible correlation with uterotropism, we examined the effects of approximately 70–100 SERMs/SERMs on uterine wet weight in the immature rat in comparison to their effects on ER Ser118 phosphorylation and Triton solubility in COS-1 cells in vitro (Fig. 8A). In addition to RAL and TOT, most of the SERMs tested were from the dihydrobenzoxathiin or chromane class of ER-selective ligands (19, 20). Seventy-two ligands were tested for uterine effects in 20-d-old Sprague Dawley rats after 3 d of oral dosing, as described in Materials and Methods. At necropsy on d 4, uterine wet weights were measured (plotted on the y-axis of Fig. 8A). Effects on uterine weights were compared with percent Triton solubility of the Ser118-phosphorylated ER (x-axis). Linear regression showed that there was a strong correlation between the in vitro phosphorylation and partitioning effect of each ligand in COS-1 cells and the respective uterotropism in vivo (r = 0.67; P < 0.0001). We also examined the antagonistic effects of more than 100 SERMs on E2-induced uterotropism in the immature rat, which showed an inverse correlation with the percent Triton solubility of the phosphorylated receptor (r = –0.48; P < 0.0001; data not shown). This raised the possibility that ligand-dependent effects on partitioning of Ser118-phosphorylated ER might correlate with some form of transcriptional regulation in uterine cells, perhaps different from the traditional AF2-dependent model.
|
), Ral, TOT, and SERMs A–D, like ICI-780, function as antagonists in AF2-dependent transcription in COS-1 cells. A similar analysis of AF2-dependent transcription in HEC-1A human endometrial cancer cells also showed a pure antagonist profile for these same ligands with even greater suppressive effects (100%) on E2-dependent transcription (Fig. 8B). Thus, there was no correlation to standard AF2-dependent transcription in either of the two cell types. E2 was previously shown to repress MMP-1 expression in endometrium-derived fibroblasts in coculture with epithelial stromal cells (30), an effect that is dependent on AP1 (31). In this regard, transcriptional repression represents the agonist response to ligand. We hypothesized that SERMs might display partial agonist behavior in such a transcription system, which might be influenced by receptor localization within or outside the nucleus. To test this, we transfected HEC-1A cells with an MMP1 promoter-luciferase reporter along with ER, as described in Materials and Methods. We stimulated the MMP1 promoter at AP-1 by treatment with phorbol ester [10 nM 12-O-tetradecanoyl phorbol 13-acetate (TPA)] in the presence or absence of ligand, and the degree of promoter inhibition was assessed.
Consistent with repression of the MMP1 promoter as an agonist response, E2 treatment dose-dependently suppressed luciferase activity, with an average maximal response of 82% vs. vehicle control (Fig. 8C). Interestingly, TOT displayed transcriptional repression activity equivalent to that seen with E2. RAL and SERMs A–D each displayed some E2-like activity, albeit with somewhat milder average suppression levels. Furthermore, ICI-780, despite exhibiting a small initial rise, showed a most modest suppression of MMP1 expression (20% at its top doses). Thus, one could elicit varied agonist transcriptional responses from each of six ER AF2 antagonists in this system. For comparison purposes, we examined the relative levels of Ser118-phosphorylated ER in the Triton-soluble and -insoluble compartments of transfected HEC-1A cells in response to each of these ligands (Fig. 8D). Despite the different cell background and the 4-fold lower receptor expression levels (vs. COS-1), the Ser118-phosphorylated ER fractionation pattern in HEC-1A cells was remarkably similar to that seen in COS-1 cells, as shown by regression analysis (Fig. 8E; r = 0.95; P = 0.0002). E2 and TOT maintained the phosphorylated receptor in a mostly Triton-soluble state, ICI-780 treatment yielded mostly insoluble receptor, and RAL and SERMs A–D elicited mixed responses, favoring the Triton-insoluble compartment. Overall, the fractionation pattern for Ser118-phosphorylated ER seemed to closely match the relative degree of MMP1 promoter suppression by each ligand, as indicated by regression analyses (r = 0.83; P = 0.0099).
A similar correlation was observed for the effects of these eight ligands on partitioning of the phospho-ER in COS-1 vs. MMP1 repression in HEC-1A (r = 0.91; P = 0.0018). This was expected, because ligand-induced changes in receptor fractionation are similar in HEC-1A and COS-1 cells. We thus extended the correlation analyses to examine the effects of 77 SERMs, including Ral, TOT, and SERMs, from the dihydrobenzoxathiin and chromane classes (Fig. 8F). Ligand effects on Triton solubility of Ser118-phosphorylated ER in COS-1 cells (Fig. 8F, x-axis) were plotted against repression of the MMP1 promoter in HEC-1A cells (y-axis). The degree of transrepression in HEC-1A correlated well with the percentage of phosphorylated ER found in the Triton-soluble compartment in COS-1 (r = 0.64; P < 0.0001). Thus, although there was no correlation of Triton solubility of the liganded receptor in either COS-1 or HEC1A cells to AF2-dependent transcription in either background, there was a strong correlation with transrepression of the MMP1 promoter in the endometrial cells. Together, the data suggest that a greater Triton solubility of the receptor tracks with greater agonism both in the uterus in vivo and in transcriptional repression of the MMP1 promoter in endometrial cells in vitro.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
are myriad and can include induction/repression of transcription, receptor accumulation/degradation, and induction of phosphorylation at each of several sites, including Ser104/106, Ser118, and Ser167 (1, 2, 3, 32). Ligand-mediated induction of Ser118 phosphorylation was considered to be an agonist response because of the more modest effects of TOT vs. E2 and the lack of effect of ICI-780. Moreover, the known ligand-induced kinase complex, TFIIH, is recruited to ER in an AF2-dependent fashion in vitro (9, 10). In this study we show that ligand-induced phosphorylation in COS-1, HEC-1A, and MCF-7 cells can be either agonist or an antagonist induced. However, the partitioning of the phosphorylated receptor, which correlates with certain types of agonist potential inherent in each respective ligand, can influence the detection of ER phosphorylation in cleared lysates of COS-1 and HEC-1A cells prepared using typical methods. For the antagonist response, we propose an alternate phosphorylation mechanism that is ckd7 independent, probably mediated through shedding of HSP90. For the AF2 agonist response, our data are consistent with the cdk7-dependent model, although shedding of HSP90 may also be involved.
Regarding the basic discovery of antagonist-induced ER phosphorylation, the data presented here are actually consistent with previous publications using the same cell system (9, 10). In considering the varied responses to ICI-384 in inducing ER phosphorylation in different studies using COS-1 cells (1, 3), it is notable that greater detection of antagonist-induced phosphorylation occurred when an anionic detergent was used, and less was observed with nonionic detergent. Furthermore, lysis with sodium dodecyl sulfate enabled detection of a robust ICI-780-induced, Ser118 phosphorylation-associated, band shift (8). Rather than use more stringent lysis conditions, we modified our approach to include separate analyses of the levels of total and Ser118-phosphorylated ER in both the cleared lysate and the pelleted Triton-insoluble material. When our analyses were limited to the cleared lysate fractions, we observed little or no substantial increase in Ser118 phosphorylation in response to ICI-780, as was previously reported. Indeed, we saw an antagonist-dependent loss of total ER in this fraction, which was consistent with the first suggestion that antagonists induce ER turnover in these cells (13). However, when the analyses included measurement of the ER found in the Triton-insoluble material, there was no ligand-dependent degradation of ER protein levels with any of more than 100 tested AF2 agonists, antagonists, and SERMs. Instead, the less uterotropic ligands shifted the distribution of ER from the Triton-soluble to -insoluble fraction. This could indeed give a false impression that ER levels were in decline should the analyses be limited to the cleared lysate fraction (or nuclear extracts).
In contrast to the responses seen in COS-1 cells, ER was authentically degraded in response to treatment of MCF-7 cells with ICI-780 or many SERMs. With ICI-780 treatment, the decline was limited to the Triton-soluble compartment, without substantial changes in levels in the insoluble fraction. The increase in the relative proportion of receptor found in the Triton-insoluble fraction of MCF-7 cells in response to ICI-780 suggested that this compartment might be involved in the degradation process itself. This was supported by the observation that levels of Triton-insoluble receptor increased several fold upon inhibition of proteosomal degradation with MG132, whereas levels of Triton-soluble receptor rose only slightly. A similar response was seen in the COS-1 and HEC-1A cells in the absence of any proteosomal inhibition. Therefore, although antagonist-induced degradation in MCF-7 cells is authentic, the parallel response seen in COS-1 and HEC-1A cells involves compartmentalization of the phosphorylated receptor in a Triton-insoluble state without degradation.
Analyses of subcellular localization of the Ser118-phosphorylated ER in COS-1 cells showed that the E2-treated receptor was distributed throughout the nucleus with only infrequent cytoplasmic staining. In contrast, ICI-780 caused both rapid phosphorylation and translocation of the ER outside the nuclear envelope. Initially, this receptor was more diffuse within the cytoplasm and less frequently found in large aggregates, although careful examination of these cells showed that the aggregation process had already begun. With additional time, the proportion of cells displaying the exclusively aggregated pattern increased, and by 4 h, the receptor was most often aggregated. Time- and dose-dependent cytoplasmic accumulation of mouse ER, without assessment of the phosphorylation state, has been quantified in COS-1 cells in response to ICI-384 and ICI-780 (16). At the lowest doses of ICI-780, the receptor remained mostly nuclear, although the little ER that was found within the cytoplasm was entirely punctate. Increasing doses caused nuclear levels to decline and cytoplasmic punctate staining to increase. Others have described antagonist-induced cytoplasmic accumulation and aggregation in CHO and COS-7 cells (18), suggesting that the phenomenon is not unique to one cell type. We add to these observations the novel discovery that the cytoplasmic aggregated receptor is actually Ser118 phosphorylated in a ligand-dependent fashion. Because the initial extranuclear ER was a mixture of both diffuse/phosphorylated and punctate/phosphorylated receptor, we propose that aggregation of the receptor is not the trigger for the phosphorylation event. Furthermore, because ICI-780 triggers aggregation of the S118A mutant, Ser118 phosphorylation is not needed for aggregation. Importantly, although Ser118 phosphorylation is suggested to activate AF1-dependent transcription (1, 3), antagonist-induced exclusion from the nucleus should prevent this from having a positive impact.
One important hypothesis arising from the observation that an antagonist could induce cytoplasmic accumulation of ER is that nuclear exclusion should somehow contribute to regulation of ER-controlled transcription. We were unable to link expulsion of the receptor from the nucleus to ligand effects in inhibiting AF2-dependent E2-stimulated transcription. This is illustrated in our transcription assays in both COS-1 and HEC-1A cells, in which all SERMs exhibited a pure antagonist response. The AF2 antagonists used in these analyses included ICI-780, which induces most of the ER to become Triton insoluble; TOT, which allows most of the receptor to remain Triton soluble; and RAL and SERMs A–D, which showed a mixed effect. Previous work has already shown that TOT does not trigger expulsion of ER from the nucleus, and indeed, a larger proportion of the receptor resides in the nucleus after treatment (16). The equivalence of TOT and all other AF2 antagonists to suppress E2-induced AF2-mediated transcription points to the already established importance of coactivators and corepressors in the response.
However, there may be a role for partitioning the receptor in the regulation of other types of transcription and in certain biological responses to ligand. We found a surprisingly strong correlation between Triton solubility of the Ser118-phosphorylated receptor in COS-1 cells and ligand effects on uterine wet weight in the rat after several days of in vivo treatment. Likewise, a correlation was found between the ability of ligands to antagonize the effects of E2 on the uterus and the level of Triton-insoluble receptor. Although the exact mechanism for ligand-dependent uterotropism is outside the scope of these analyses, the data suggest a possible mechanistic link between ligand attributes that dictate receptor distribution within the cell and those that control uterine growth. To extend the in vivo observations, we examined ER-mediated regulation of the MMP1 promoter in HEC1A endometrial cancer cells. The AP1-dependent (31) MMP1 promoter is repressed by E2 in fibroblasts derived from the endometrium (30) as well as in both primary osteoblasts and osteoblast-like cells (33). We found similar repressive effects of E2 in HEC-1A cells using a luciferase reporter assay stimulated by phorbol ester. Interestingly, unlike with the purely antagonistic effects on AF2-dependent transcription, there was a partial agonist response at the MMP1 promoter to each of the tested SERMs. This included an almost pure agonist response to TOT and strong partial agonist responses to RAL and SERMs A–D. ICI-780 elicited the weakest response on this promoter. Other SERMs showed effects that covered the spectrum. Consistent with a possible role for receptor partitioning in controlling the repression response, the magnitude of MMP1 promoter repression by these SERMs was paralleled by the differential partitioning of Ser118-phosphorylated ER in response to these same ligands. This was true when partitioning was measured in either HEC-1A or COS-1 cells. Whether MMP1 repression or that of any other gene in the uterus plays a key role in uterine growth in vivo remains to be established. In any case, the responsible genes would probably be regulated in an AF2-independent fashion, because numerous AF2 antagonists are quite uterotropic.
One question of particular interest regards the mechanism by which AF2 antagonists could induce Ser118 phosphorylation. ICI-780 does not position helix 12 in a fashion compatible with TFIIH-mediated Ser118 phosphorylation as seen in purely in vitro assays (10). Nonetheless, we observed strong phosphorylation in COS-1 cells after treatment with ICI-780 and each of more than 100 AF2 antagonists. Furthermore, we observed both agonist- and antagonist-dependent phosphorylation at this site using either of two helix 12 mutants. The positive effects of AF2 antagonists and the relatively normal phosphorylation of AF2 mutants induced by agonist and antagonist alike suggested the need to identify common responses shared by all ligands regardless of the transcriptional effects of each. Other than the fact that agonists and antagonists both bind into the same pocket of the LBD, there is little similarity in responses at the molecular level. It was reported, however, that both E2 and the antagonist RU 58668 share the ability to trigger HSP90 dissociation from ER (18). We, therefore, hypothesized that HSP90 could play a role in suppressing Ser118 phosphorylation through its binding to the unliganded ER. By extension, ligand-dependent dissociation of HSP90 from the receptor might be sufficient to trigger Ser118 phosphorylation.
To test this, we inhibited HSP90 function using GDA, which was previously shown to trigger ER degradation in MCF-7 cells (27, 28). Given that both ICI-780 and GDA induce receptor degradation in MCF-7 cells and that ICI-780 does not induce ER degradation in COS-1 cells, we originally predicted that GDA would have no negative impact on ER expression in the COS-1 background. However, GDA was very effective in causing receptor degradation, which was readily evident by 4 h. The magnitude of the effect diminished with brief exposure, and little degradation was seen at 15 min. Interestingly, there was a strong similarity between the effects of GDA and ICI-780 at this early time point. Both treatments triggered movement of the ER from the nucleus to the cytoplasm, both induced phosphorylation at Ser118, and both induced ER aggregation. Due to the rapid phosphorylation response, we suggest that the GDA effect on phosphorylation was due to inhibition of HSP90 associated with ER rather than through activation of an alternate pathway. Indeed, previous studies of MAPK have shown no positive effect of GDA on activation in either the basal state or in response to mitogen treatment (34, 35, 36). There are no reported effects of GDA on TFIIH or cdk7. In any case, the initial similarities in the ER responses to GDA and ICI-780 support the hypothesis that ligand-dependent translocation of ER from the nucleus to the cytoplasm as well as cytoplasmic aggregation could be results of HSP90 dissociation.
In the search for an enzyme responsible for AF2-independent phosphorylation, we tested the possibility that one of the known ER kinases could be involved. In the case of cdk7 and MAPK, we saw no evidence that either could account for ICI-780-induced phosphorylation. Although R-roscovitine is known to inhibit cdk2, -7, and -9 in the submicromolar range (37), only cdk7 is implicated in AF2-dependent phosphorylation of Ser118. Consistent with previous reports implicating cdk7 in agonist-induced phosphorylation (9, 10), R-roscovitine suppressed E2-mediated phosphorylation at Ser118. Interestingly, neither this inhibitor nor one targeting MAPK kinase-1 was able to suppress ICI-780-induced Ser118 phosphorylation at this site. This clearly suggests that AF2-independent induction of Ser118 phosphorylation by ICI-780 involves a unique mechanism that requires additional investigation.
In summary, we suggest a model for ligand-induced ER phosphorylation in which both agonist and antagonist are active. The agonist effect is cdk7 dependent, whereas the antagonist response involves an unidentified kinase. The antagonist response may be triggered by HSP90 dissociation, which should also contribute to agonist-induced phosphorylation. Although Ser118 phosphorylation could have a positive impact on transcription, the partitioning of ER in the Triton-soluble and -insoluble compartments may override this. Although receptor partitioning does not seem to correlate with AF2-dependent transcription, it may play a role in modulating the MMP1 promoter or others. The biological responses that these genes control may play a role in controlling uterine growth, lipid lowering, and bone protection, all of which can be included as AF2-independent agonist responses to the SERMs.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
-pcDNA3.1+ expression construct was a gift from Dr. Hilary Wilkinson (Merck & Co., West Point, PA). L540Q, D538A/E542A/D545A, and S118A ER mutants were generated as previously described (2, 38, 39) and expressed using pcDNA3.1+ (Invitrogen Life Technologies, Inc., Carlsbad, CA). For ER-Gal4 transcription assays, the Gal4 DNA-binding domain (DBD) was transferred from the pM vector (BD Clontech, Mountain View, CA) and inserted downstream of the cytomegalovirus promoter of pcDNA3.1+. The LBD of human ER was then fused in-frame with the Gal4 DBD. The Gal4 luciferase reporter was constructed by inserting five copies of the consensus Gal4-binding site upstream of the luciferase reporter gene in the promoter-less plasmid pGL2-Basic (Promega Corp., Madison, WI). The –179 to 63 MMP1-luciferase pGL2-Basic (MMP1-luc) reporter construct (31) was a gift from Dr. Dwight Towler (Washington University School of Medicine, St. L
0.03; **, P 
