| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Diabetes Center (A.P., L.L., E.M.K., F.S.) and Department of Medicine (F.S), University of California, San Francisco, California 94143
Address all correspondence and requests for reprints to: Fred Schaufele, S-1230, 513 Parnassus, University of California San Francisco, San Francisco, California 94143-0540. E-mail: freds{at}diabetes.ucsf.edu.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
dimer complex by comparing, in living human breast cancer cells, the amounts of energy transfer between fluorophores attached to different domains of ER. Estradiol, 4-hydroxytamoxifen, and fulvestrant all promoted the rapid formation of ER dimers with equivalent interaction kinetics. The amino- and carboxyl-terminal ER domains both contain activation functions differentially affected by these ligands, but the positions of only the carboxyl termini differed upon binding with estradiol, 4-hydroxytamoxifen, or fulvestrant. The association of a specific ER dimer conformation with the binding of ligands of different clinical effect will assist the identification of a SERM with optimal tissue-selective estrogenic and antiestrogenic activities. These studies also provide a roadmap for dissecting important structural and kinetic details for any protein complex from the quantitative analysis of energy transfer. |
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Clinically, estrogens also are involved in the initiation of many breast tumors and are required for the continued growth of a subclass of breast tumors. For those tumors, aromatase inhibitors that deplete estrogen synthesis from nonovarian tissues in postmenopausal women are an effective treatment for postmenopausal breast cancer patients (4, 5, 6). Alternatively, drugs such as tamoxifen and fulvestrant (also referred to as faslodex or ICI 182780) that bind the estrogen receptor (ER) without activating the ER in breast tissue provide treatment for estrogen-dependent breast tumors in women of all ages (7, 8, 9, 10, 11). For breast cancer patients, the positive effects of these treatments in blocking tumor proliferation clearly outweigh their adverse effects on nonbreast tissues (12). However, that is not true for healthy individuals. A drug with an ideal spectrum of tissue-selective actions not only would be preferred for women choosing hormone replacement therapies, but also would minimize the side effects of breast cancer treatment and could be used to reduce the risk of breast cancer in at-risk individuals (13, 14).
Our current understanding is that the preferred ER ligand would be antiestrogenic or neutral in the breast, endometrium, and vasculature, but estrogenic in bone and estrogen-responsive regions of the brain pertinent to cognitive performance and hot flashes. A ligand with this highly selective spectrum of tissue-selective activities currently does not exist. Identifying the ideal selective estrogen receptor modulators (SERMs) or SERM combination will be challenging but the existence of ER-binding drugs, such as tamoxifen and fulvestrant, with differential, tissue-selective agonist and antagonist activities offers hope that an ideal SERM may be attainable (14, 15, 16). A key is to identify the ER activities associated with the specific effects of estrogen and each SERM in specific tissues, and then screen for novel drugs with that preferred subset of activities (14, 15, 16, 17, 18, 19).
Some ligand-regulated ER activities for estradiol and the known SERMs are well defined, whereas other activities are not. Estradiol binding to its receptor results in ER dimer formation. The ER dimer then associates with genes through direct DNA binding or interaction with other promoter-binding factors (20, 21, 22, 23, 24). Transcriptional activation by estradiol binding to ER is accompanied by the release of activation function-1 (AF-1) in the ER amino-terminal domain (NTD) and by the modification of a coactivator binding pocket (AF-2) in the carboxyl-terminal ligand binding domain (LBD) (21). The structure of AF-2 in the estradiol-bound LBD favors coactivator binding (25, 26), whereas elements of the same pocket are used to bind corepressors in many unliganded nuclear receptors (27, 28) or in some SERM-bound ERs (29). Unlike estradiol, tamoxifen and fulvestrant do not promote the formation of the AF-2 coactivator binding pocket (30, 31) and therefore block ER action in tissues that rely on AF-2, including the breast. By contrast, tamoxifen activation of ER in other tissues must depend on less characterized molecular pathways. Our goal is to measure those molecular events in living cells to improve SERM characterization that eventually will aid the discovery of SERMs with unique properties, some of which may help improve therapy.
We characterized the ligand-regulated interdomain structure of the ER dimer in living human breast cancer MCF7 and T47D cells by quantifying the amounts of fluorescence resonance energy transfer (FRET) between cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP) attached to the NTD and LBD of ER. FRET provides information about the interaction kinetics between molecules in living cells as well as about the structure of the interacting complex (32). All ligands examined (estradiol, 4-hydroxytamoxifen, and fulvestrant) promoted dimer association with similar on-rates and with similar equilibrium interaction kinetics. Structurally, the relative positions of the fluorophores attached to the LBD were different for the estradiol, 4-hydroxytamoxifen, and fulvestrant-bound ER dimers, whereas the positions for the NTDs were similar. This showed how the interdomain ER dimer conformations differed upon binding to tissue-selective ligands that activate different subfeatures of ER activity.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Dimer Structures can be detected in living cells (33, 34). This potentially powerful technique has suffered from an inability to understand in biologically meaningful terms what a FRET signal means. We previously outlined a strategy for extracting from FRET data the effect of a stimulus of the formation and conformation of a protein complex in live cells (35, 36). That strategy is demonstrated here for ligand regulation of ER.
The cDNAs for CFP or YFP were subcloned as in frame fusions to the amino terminus of ER (CFP-ER or YFP-ER) (see Fig. 1A
) and coexpressed in human breast cancer MCF7 and T47D cells. Similar or different amounts of energy transfer from CFP to YFP in cells incubated with different ligands indicate similar or different ER interaction kinetics and/or positions of the fluorophore-tagged amino termini (AF-1s) in the ER dimer. Similarly, CFP or YFP are fused to the carboxyl terminus of ER (ER-CFP or ER-YFP) (Fig. 1C) and coexpressed to quantify the effect of ligand on the interactions or positions of the LBDs in ER complexes. Finally, coexpression of CFP-ER with ER-YFP (Fig. 1B) provides a measurement of the interactions or positions of AF-1 and the LBD in the ER complex. Methods for distinguishing effects of ligand on dimer formation vs. dimer structure (32, 37) were applied as discussed below.
|
Fusions with CFP Are Functional remains functional after fusion of the fluorescent proteins to either part of the protein. Prior studies indicated that ER remained transcriptionally competent after fusion of fluorescent proteins to its amino or carboxyl termini (33, 34, 38, 39, 40, 41). We similarly established that the slightly different fusion proteins used in the current studies remained transcriptionally active (Fig. 2A). Transient expression of either ER-CFP or CFP-ER in ER-null HeLa cervical carcinoma cells was accompanied by an estradiol-regulated increase in the amount of chloramphenicol acetyltransferase expressed from a promoter containing a single copy of the estrogen response element of the Xenopus vitellogenin gene upstream of the TATA box of the herpes simplex thymidine kinase. Incubation with 4-hydroxytamoxifen (the active metabolite of tamoxifen) did not enhance expression of the reporter, and there was no ligand activation of the reporter when no ER was expressed. This demonstrated that ER fused at its amino or carboxyl terminus with a fluorescent protein was capable of binding estradiol and of forming competent transcriptional activation complexes.
|
(42). We therefore examined whether ligand addition affected the steady-state amounts of the ER fusion proteins. Nuclear extracts were prepared 20 or 40 min after the addition of vehicle, estradiol, 4-hydroxytamoxifen or fulvestrant to MCF7 breast cancer cells transiently expressing CFP-ER or ER-YFP. Western blots were performed with an anti-ER antibody that detects both the endogenous ER and the slower migrating ER fusion proteins (Fig. 2B
). These studies showed that both the CFP-ER and ER-YFP fusion proteins, like the endogenous ER, were mildly down-regulated by the addition of fulvestrant over the ligand incubation periods used in most of the FRET studies (discussed below). Longer ligand exposures also were accompanied by 4-hydroxytamoxifen stabilization of the endogenous ER and its transiently expressed fusion proteins (data not shown). These studies indicated that turnover of the ER fusion proteins was correctly regulated by ligand.
Fluorescence microscopy showed that the unliganded CFP-ER, ER-CFP, YFP-ER, and ER-YFP also, as appropriate, targeted predominantly to the cell nucleus when expressed in MCF7 breast cancer cells (Fig. 2C), T47D breast cancer cells, or HeLa cervical carcinoma cells (data not shown). The addition of estradiol, 4-hydroxytamoxifen, or fulvestrant all resulted in the previously described (38) rapid redistribution of ER to discrete sites in the cell nuclei (Fig. 2C). Note that, under conditions under which 30–40% of the transfected cells displayed ER fusion protein fluorescence poorly visible through the microscope eyepieces, the amounts of ER fusion protein detected by Western blot (Fig. 2B) were marginally more than that of the endogenous ER. Because cells highly overexpressing the fusion proteins were avoided during image collection and analysis, the FRET studies discussed in the following sections were conducted under conditions in which functional ER fusion proteins were expressed at levels similar to, or only moderately above, those of the endogenous ER. This enabled FRET detection of the ligand-regulated interactions at near-normal amounts of ER.
Finally, the FRET studies of ER interaction kinetics (discussed below) showed that those interactions were identical when measured for any combination of fluorophore-tagged ER. This further demonstrated that interaction was being driven by ER regardless of fluorophore position. Together with the studies on ligand regulation of ER-mediated promoter activity, ER amount, and ER intranuclear distribution, these studies showed that fusion of the fluorescent protein to either end of ER did not result in any overt defects in ER response.
Increased Energy Transfer between Tagged ER in Estradiol-Treated Cells
MCF7 and T47D cells grow in response to estradiol but not in response to tamoxifen or fulvestrant. Both cell lines therefore represent good models for an initial investigation of the effects of ligand on the relative positions of AF-1 and the LBD in ER. We initially compared the amounts of energy transfer in MCF7 and T47D cells grown for 25 min (MCF7) or 4 h (T47D) in the presence of 10–8 M estradiol or an equivalent amount of the drug vehicle, ethanol. The parameters of image collection for energy transfer analysis by fluorescence microscopy are provided in Materials and Methods.
The method of determining energy transfer from CFP (the donor fluorophore) to YFP (the acceptor fluorophore) for a representative MCF7 cell coexpressing CFP-ER and ER-YFP is shown in Fig. 3, A–D. Briefly, images collected from 133 control cells expressing only ER-YFP (Fig. 3A) were used to quantify the amounts of background-subtracted fluorescence (measured in the cell nucleus where ER predominantly is localized) that ER-YFP individually emits into the acceptor channel (excite YFP and collect YFP), into the donor channel (excite CFP and collect CFP), and into the FRET channel (excite CFP and collect YFP). The contributions of YFP to the donor and FRET channels are constants (1 and 15% of the amount of fluorescence in the acceptor channel, respectively) that depend upon the physical properties of the fluorophore and the collection parameters (32). Similarly, images of 205 control cells expressing only CFP-ER were used to quantify fluorescence amounts that CFP emits into the FRET channel (62% of the amount of fluorescence in the donor channel), and into the acceptor channel (0%) (Fig. 3B). Therefore, in cells coexpressing CFP-ER and ER-YFP, the fluorescence in the acceptor channel originates completely from ER-YFP, whereas both CFP-ER and ER-YFP contribute to fluorescence in the donor and FRET channels.
|
and ER-YFP (Fig. 3C, no ligand), the amount of ER-YFP fluorescence was quantified and the constants derived from the ER-YFP control cells were used to subtract the fluorescence that YFP contributes to the donor and FRET channels (Fig. 3D). The remaining fluorescence in the donor and FRET channels originates from CFP-ER. With energy transfer, the energy emitted from CFP as cyan-colored light is decreased and is instead transferred to YFP to be emitted as yellow-colored light; the result is that the FRET/donor ratio is higher than if there were no energy transfer. When the cells were not incubated with ligand, the amount of CFP-ER fluorescence in the FRET channel relative to the amount of CFP fluorescence in the donor channel was 0.75 ± 0.11 (mean ± SD from 266 cells), which was marginally above that measured in the CFP-ER control cells (0.62 ± 0.03). This slightly higher FRET/donor ratio results from a weak interaction of CFP-ER with ER-YFP, which brings the CFP and YFP into close enough proximity (<80 Å) to result in energy transfer. By contrast, in the estradiol-treated cells, energy transfer was more readily detected as a statistically significant (P < 0.001) higher FRET/donor ratio than in the unliganded cells for all three combinations of CFP- and YFP-tagged ER in both MCF7 and T47D cells (Fig. 3E). Thus, estradiol treatment caused an increase in energy transfer between fluorophores attached to interacting ER.
Estradiol Promotes Self-Association of ER
The amount of energy transferred from a donor-labeled protein to an acceptor-labeled protein is related to the interaction kinetics of the interacting proteins and to the positions of the fluorophores in the interacting complexes (i.e. the structure of the complex) (32). Thus, the higher value of energy transfer in estradiol-treated cells indicated that, for all three different combinations of labeled proteins in both cells types, estradiol either 1) improved interaction of ER with itself (bringing the fluorophores into contact) or 2) caused a shift in the position of the fluorophores of an already existing complex that led to a higher level of energy transfer. The challenge is to distinguish changes in energy transfer arising from changes in interaction and/or changes in structure. We previously have discussed how to extract such biochemically relevant information from the energy transfer readout (32, 35, 37).
In cells that express relatively more YFP-labeled ER, the YFP-labeled ER will be better capable of finding and interacting with CFP-labeled ER. If the energy transfer observed is due to an interaction between two ER, then the energy transfer from the CFP-labeled ER should increase in relationship to increasing amounts of YFP-labeled ER present in a cell. This increase in the FRET/donor ratio above the baseline 0.62 value was observed with increasing acceptor amounts when 276 different estradiol-treated MCF7 cells expressing CFP-ER and ER-YFP were measured (Fig. 4A, closed squares). In cells treated for 25 min with 10–8 M estradiol, the measurements from multiple cells fit well (R2 = 0.82) to a curve that describes an interaction between two proteins. However, the weaker FRET detected in the 266 vehicle-treated cells (Fig. 4A, open circles) did not fit well to this bimolecular interaction curve (R2 = 0.27).
|
). FRET was observed in a small number of MCF7 cells not treated with ligand indicating that some MCF7 cells not treated with ligand supported ER association similar to that observed in estradiol-treated cells (see Fig. 3A). It is difficult to conclude whether the unliganded ER can associate in a small subpopulation of MCF7 cells or whether this reflects an experimental inability to completely remove ligand. Nevertheless, the data indicate that estradiol treatment leads to enhanced, more uniform self-association of ER in MCF7 and T47D cells.
|
), the amounts of YFP-labeled ER required to reach half-maximal interaction with CFP-labeled ER (measured at half the maximal amount of energy transfer) will be a representation of the kinetics of interaction at equilibrium (Kd). Note that the term is not meant to be used in the classical sense (the equilibrium dissociation constant or Kd), because we cannot yet ascribe the amounts of fluorescence in terms of absolute units of concentration, such as molarity. Instead, the Kd defined here relies on relative fluorescence units that depend upon the collection conditions. On the equipment used in the current studies, 1.18 U of fluorescence from YFP are collected for every 1.00 U collected from an equimolar amount of CFP (determined by expressing both fluorescent proteins on a common translation unit, but spaced far enough apart to permit no energy transfer of CFP to YFP). The loss of donor fluorescence to FRET, which increases nonlinearly with acceptor level, also will affect the acceptor/donor measurement and is more difficult to correct for (32, 43). In the current study, the loss of donor to FRET is low, eventually reaching 15% with maximal energy transfer at saturating acceptor levels. Thus, in the current studies, the Kd values measured provide a good comparison of the relative association of the tagged ER. However, we caution that, in other studies, a high level of energy transfer will more substantively introduce errors into the acceptor/donor ratio and into comparisons of Kd values for interactions with different energy transfer efficiencies.
The Kd measured for all three combinations of interaction (CFP-ER with YFP-ER, CFP-ER with ER-YFP, and ER-CFP with ER-YFP) should be similar, because all three measurements are of the same interaction between two or more ER. As expected, there was no statistically significant difference in the Kd values measured for all three combinations of ER interaction the MCF7 cells (2.65 ± 0.36, 2.50 ± 0.64, and 3.98 ± 1.13; or 3.04 ± 0.47 on average). The Kd values measured in T47D cells also were not statistically different. However, the Kd values were more variable in T47D cells (4.44 ± 2.19) in which a higher and variable amount of autofluorescence (the background fluorescent from cells not expressing any fluorescent protein) made more difficult the accurate quantification of fluorescence required for measuring FRET. Therefore, the remaining studies were conducted primarily in MCF7 cells.
DNA Binding and AF-2 Are Not Required for ER Dimerization
The estradiol-induced association of ER in living cells may be caused by the interactions of ER with common third factors. We therefore introduced into CFP-ER and ER-YFP a mutation (K362A) that disrupted the surface of the ER LBD required for interaction with p160 coactivators and with N-CoR/SMRT corepressors (25, 44). A different mutation (K210E, R211D) was constructed in amino acids of the ER DNA binding domain that make critical contacts with specific bases in the estrogen response element (45). As shown in Fig. 4 and Table 2, the Kd values for interaction of both mutants were the same as for the unmutated ER. Thus, the ability of ER to associate in the presence of estradiol was not altered by the inhibition of cofactor or DNA binding. These mutants also did not affect ER association in the absence of ligand. By contrast, association of CFP-ER and ER-YFP induced by 10–7 M estradiol (Fig. 5A) as well as by 10–7 M 4-hydroxytamoxifen and 10–7 M fulvestrant (Fig. 5, B and C) was eliminated by a L504E/L508E/L511E triple mutation in the ER LBD that disrupts dimerization in vitro (46). Overall, the results are consistent with a ligand induction of dimerization by the bulk of estradiol-liganded ER that does not require DNA binding or AF-2-interacting cofactors.
|
|
to punctuate foci within the cell nucleus (Fig. 5, inset images). By contrast, mutation of the cofactor interaction surface or of estrogen response element-specific DNA binding had no effect on ligand-regulated relocation (Fig. 4). This suggested that dimerization, or some other aspect of ER function dependent upon leucines 504, 508, and 511, is required for ligand-induced relocation of ER.
Similar Dimerization of ER Bound with Estradiol, 4-Hydroxytamoxifen, or Fulvestrant
Our primary goal was to define the similar and different effects of estradiol and SERMs on the ER dimer. We therefore compared ER dimer formation under saturating amounts of estradiol (10–8 M), 4-hydroxytamoxifen (10–7 M), and fulvestrant (10–7 M) for four different combinations of coexpressed CFP-tagged ER and YFP-tagged ER (ER-CFP with YFP-ER, CFP-ER with YFP-ER, CFP-ER with ER-YFP, and ER-CFP with ER-YFP). Estradiol, 4-hydroxytamoxifen, and fulvestrant all induced FRET that fit the bimolecular interaction curve for all four combinations of fluorophore-tagged ER (average R2, 0.81 ± 0.7). The Kd for ER self-association was averaged for each ligand from the four different measurements. The Kd values measured for estradiol (2.9 ± 0.4), 4-hydroxytamoxifen (4.4 ± 1.5), and fulvestrant (2.2 ± 0.8) were not statistically different (P > 0.05), although this conclusion may be affected moderately by the differential changes in endogenous ER amount induced by each ligand (see FRET Analysis in Materials and Methods).
The similar Kd values of ER dimerization at equilibrium may mask differences in the on- and off-rates of association promoted by each ligand. To measure the on-rate of association, we made repetitive FRET measurements from individual MCF7 cells at 1 min time intervals before and after the addition of each ligand (Fig. 6A). Measurements were averaged from multiple cells that were first selected for low levels of FRET in the absence of ligand. To further ensure that the measurements were comparable, all were done under conditions under which the FRET amount was close to saturated by an excess of the acceptor (YFP)-labeled ER over the donor (CFP)-labeled ER. The Kon was established as the amount of time required to reach one half of the maximal level of association. Note that because these experiments are conducted in living cells, the on-rate defined here also is a reflection of how rapidly the ligand gains access to the cell. No statistically significant differences in the biological Kon were detected for the association of ER promoted by estradiol, 4-hydroxytamoxifen, or fulvestrant in MCF7 cells (Fig. 6B). Thus, all three ligands equivalently promote the dimerization of ER measured directly in the cellular environment.
|
Dimer Conformations, the maximal amount of energy transfer will occur when the maximal amount of CFP-labeled ER is interacting with YFP-labeled ER. For classical binding curves between two proteins under different conditions, different binding maxima (Bmax) would imply different proportions of the tracer protein (CFP-labeled ER) interacting with the driver protein (YFP-labeled ER) at equilibrium binding. However, the Bmax derived for FRET will additionally be affected by the position and orientation of the fluorophores in the interacting complexes.
The Bmax for FRET between ER was determined for each of the four combinations of CFP- and YFP-labeled ER expressed in MCF7 cells (Fig. 7). When querying interaction between fluorophores attached to the carboxyl termini of ER (ER-CFP coexpressed with ER-YFP; right-most bars), the Bmax of FRET was statistically higher upon incubation with fulvestrant than with estradiol (*, P < 0.01); the Bmax for 4-hydroxytamoxifen was intermediate between that measured in the presence of estradiol or fulvestrant. If the different Bmax values detected were a function of different ligand-specific, maximal proportions of CFP-tagged ER bound with YFP-tagged ER, the Bmax of interaction would be expected to be higher for fulvestrant than for estradiol for all combinations of coexpressed CFP- and YFP-labeled ER. Because the Bmax for FRET between fluorophores attached to the two amino termini (CFP-ER coexpressed with YFP-ER) or attached to one amino terminus and one carboxyl terminus (CFP-ER coexpressed with ER-YFP or YFP-ER coexpressed with ER-CFP) were identical for each ligand, the very different Bmax values for FRET measured between the carboxyl termini indicate a ligand-specific difference in the positions or orientations of the fluorescent proteins attached to the carboxyl termini of ER. This interpretation is bolstered by the very similar interaction kinetics measured for each ligand.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Dimers-helix (helix 12) near the end of the ER LBD is essential to the proper formation of AF-2 (25). However, the lack of structural information on the full-length ER dimer has impeded our understanding about the effects of ligand binding on the positions of AF-1 in the NTD. The ability of estradiol and tamoxifen to release AF-1 previously was shown to be cell-selective in cultured cell models (47, 48, 49). In addition, tamoxifen and fulvestrant showed a differential ability to release AF-1 (29, 47, 48, 49), although both ligands block the formation of AF-2 (30, 31, 50). The differential, cell-specific effects of SERMs on AF-1 activity has led to the attractive speculation that SERM binding to LBD is uniquely communicated to AF-1 in the NTD (21), and that this SERM-selective communication contributes to part of the tissue-selective nature of SERM, and estrogen, function.
The nature of the communication between the LBD and AF-1, and even whether it is SERM-selective, remains unknown owing to the difficulties in obtaining structural information for the full-length, active ER (51). Our FRET analyses showed, in living cells, that there was no gross difference in the overall positions and/or orientations of fluorophores attached to AF-1 in ER dimers bound with estradiol, 4-hydroxytamoxifen, or fulvestrant. We also observed that the ER dimers activated by these three different ligands had similar affinities, which indicated that their differential effects on transcription and proliferation in MCF7 cells were not related to some differences in the kinetic nature of the ER dimer. Finally, all three ligands entered the cell and activated dimer formation with the same temporal kinetics. Because there were no interdomain differences in the positions of AF-1 in the estradiol, 4-hydroxytamoxifen or fulvestrant-bound ER dimers, and no difference in the time required to acquire those kinetically similar structures, we suggest that the SERM-specific differences in AF-1 activity do not arise from gross structural differences in the interdomain positions of AF-1 after the binding of estradiol, 4-hydroxytamoxifen, or fulvestrant to the ER LBD.
Ligand-Selective Differences in the Positions of the ER LBDs
Our studies showed that, in living MCF7 cells, the conformations of the full-length ER dimers bound with estradiol, 4-hydroxytamoxifen, or fulvestrant differ primarily in the positions of the ER carboxyl termini. It is difficult to ascribe differences in FRET amounts to actual distances between the fluorophores, because the amount of FRET is a function both of fluorophore distance and of fluorophore orientation (32). As an indicator of scale only, calculations from the FRET data suggested that the positions of the carboxyl-terminal fluorophores were approximately 7 Å closer in the fulvestrant-bound ER dimer than in the estradiol-bound ER dimer. We emphasize the inherent inaccuracy in these estimates, which depend upon the assumption that the orientations of the fluorophores were the same in the estradiol-, 4-hydroxytamoxifen-, and fulvestrant-bound dimers. Nevertheless, our data indicated a structural difference in the positions and orientations of the fluorophores attached to the ER dimers bound by estradiol, 4-hydroxytamoxifen, and fulvestrant that was on a 5–10 Å scale.
The differences in the positions of the carboxyl termini described here were detected in the physiological environment of living human breast cancer cells. Interestingly, these differences correlated well with crystallographic studies of the isolated ER LBD. Those x-ray structures indicated that helix 12 is oriented at an angle that is more than 100° different in the estradiol-bound and 4-hydroxytamoxifen-bound human ER LBDs (30, 31). Because helix 12 is located only 50 amino acids from the carboxyl terminus of human ER to which we tagged our fluorophores, our studies likely detected that ligand-regulated movement. Structures of the fulvestrant-bound human ER LBD have not been published but structures of the rat ERß LBD bound with ICI 164384, a chemical cousin of fulvestrant, showed helix 12 to be unstructured (50). Our data suggested that fulvestrant also caused some unique change in the carboxyl termini of the human ER LBD, possibly associated with the position or structure of helix 12 and/or the remaining 50 amino acids in ER.
To date, all structures of ER, or of all nuclear receptors, have been done with the isolated LBDs (51, 52, 53). The FRET measurements provided the first direct confirmation that the ligand-selective differences in the positions of helix 12 occur within the full-length ER. Moreover, it showed that this shift occurred in the dynamic, physiological environment of the living cell and under conditions in which the ER was not constrained by crystal packing considerations. Our data therefore further validated the assumption that those crystal structures faithfully represent the in vivo conformations of ER. In addition, our FRET data showed that this functionally important shift readily can be detected in assays that are highly amenable to rapid screening (37). This may have important consequences to our abilities to identify new compounds with novel specificities, particularly when one considers that the FRET assay has been shown to be readily adaptable to directly investigate a wide variety of interactions, including those with fragments of nuclear receptor-interacting cofactors (33, 54) or with phage-display-selected peptide probes of ER structure (40).
Dimerization and ER Relocation in the Cell Nucleus
The rapid redistribution of ER to discrete sites in the cell nuclei (Fig. 2C) remains of unknown functional importance (32, 38, 41, 55, 56). The current studies showed that dimerization precedes, and is necessary for, ligand-induced relocation. Dimerization detected by FRET was observed rapidly after ligand addition (Kon of 7–8 min), before the visible concentration of the bulk of the ER at those sites (usually >10 min). Even after ER is strongly relocalized (20 min after ligand addition), higher magnification studies (data not shown) indicated that the levels of ER dimer energy transfer detected at the subnuclear foci were not statistically different from those detected away from the subnuclear foci. We emphasize that slight movements in those very small subnuclear foci during image collection limit confidence in that specific conclusion. But both observations together indicate that dimerization likely occurs within the general nucleoplasm and does not require subnuclear foci formation to occur. Moreover, our studies with the dimerization-defective mutant (Fig. 5) demonstrated that some activity affected by the L504E/L508E/L511E triple mutant, mostly likely dimerization, was required for formation of the intranuclear foci. These studies still do not shed any light on the functional nature of the subnuclear foci but do strongly suggest that it is an active process dependent on ER dimerization.
FRET for Measuring SERM Action
Prior studies showed some FRET between CFP-ER and YFP-ER coexpressed in human HEK293 kidney cells and MCF7 cells (33, 34). Like most FRET studies to date, those studies established that an interaction could be observed in the physiological environment but that a large amount of variability in the FRET signal precluded firm conclusions about the FRET signal. Here, we applied a more advanced analysis (32, 35, 37) in a large number of cells to extract the biochemical and structural contributors to FRET. Those analyses take into account that the high variations in FRET signal originate from the extent by which interaction kinetics will affect the amount of acceptor-labeled protein required for a productive interaction with the donor-labeled protein. We applied those advanced techniques with the fluorophores in different positions to establish the similarities and differences in interdomain structures affected by different ligands. Such extensive studies were possible only after collecting large numbers of images and applying semiautomated image analysis. Full automation of image collection and analysis will be required for more extensive studies. We envisage full automation to be a solvable technical hurdle that, once overcome, will allow similar analyses on many types of interactions.
The strength of the FRET approach is that it permits the analyses of interaction and structure between any two proteins in living cells with time after the addition of a ligand, drug, or physiological event. However, there are limitations. Some molecular studies indicate that cell-specific differences in the ability of the SERM-bound ER dimers to bind cofactors contributes to the organ-selective agonistic and antagonistic actions of those ligands (57). Such interactions could be investigated by FRET as long as the positions of the fluorophores in the interacting complex are less than 80 Å apart. That requirement has been difficult to achieve for interactions of a nuclear receptor with any of the larger full-length nuclear receptor cofactors end-labeled with a fluorescent protein (data not shown). However, FRET has been detected with smaller cofactor fragments or peptides (33, 40, 54) that may act as surrogate probes of interaction.
A large advantage of the FRET technique is the ability to compare, with time, alterations in conformation and interaction. We found that ER dimers bound by estradiol, 4-hydroxytamoxifen, or fulvestrant had similar equilibrium interaction kinetics in living cells, which compares favorably with the similar in vitro dissociation rates observed for isolated LBDs bound to these ligands (58). Our temporal studies in living cells also indicated that the ER conformations induced by estradiol, 4-hydroxytamoxifen, or fulvestrant all were attained within an equivalent short time period after the addition of each ligand. The similar speeds by which FRET was acquired with the fluorophores in different positions showed that the initial interaction was not followed minutes later by the movement of other domains into their final positions. This suggests that ligand addition may cause a rapid folding of ER, which is then followed by dimer formation. Indeed, we and others have previously found that FRET within a dual-labeled CFP-ER-YFP (37, 59) increases more rapidly after ligand addition than does FRET between CFP-ER and ER-YFP (37). For the nuclear receptors in which ligand addition leads to an ordered, stepwise cycling of cofactor interactions (60, 61, 62, 63), it will be interesting to place the timing of conformational changes in relationship to changes in the association and dissociation of different complexes along the ligand-initiated interaction pathway. This would allow the visualization and investigation of the dynamic molecular machines that all life depends upon.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
was inserted into the pECFP-C1, pEYFP-C1, pECFP-N1, and pEYFP-N1 (Clontech, Palo Alto, CA). These expression vectors were introduced into MCF7 (tet-off; Clontech), T47D (American Type Culture Collection, Manassas, VA) or HeLa cells (tet-off; Clontech) using the Effectene transfection reagent (Qiagen, Valencia, CA). Twenty-four hours before transfection, cells were washed thoroughly with PBS and grown in estrogen-free medium throughout transfection and image collection. Estrogen-free media consisted of phenol red-free, modified DMEM/F-12 Ham’s medium containing 10% charcoal-stripped newborn calf serum. A total of 100 ng of CFP-labeled ER expression vector was transfected per well of a six-well dish together with variable amounts (30–300 ng) of YFP-labeled ER expression vector to achieve the distribution of YFP-labeled ER that enabled the calculations of the bimolecular interaction curves (Figs. 4 and 5).
For studies of the transcriptional activation of an estrogen-responsive promoter, HeLa tet-off cells, grown in media containing serum charcoal-stripped of estrogens, were transiently transfected with the 100 ng of the estrogen response element-
TCO estrogen-responsive chloramphenicol acetyltransferase reporter and 20 ng of either the ER-CFP expression vector, the CFP-ER expression vector, or the same expression vector with no cDNA inserted (No ER). Twenty-four hours after transfection, 10–6 M estradiol, 10–6 M 4-hydroxytamoxifen, or the same amount of drug vehicle (No Ligand) were added. Extracts were collected 24 h later, and the amounts of reporter activity normalized to protein in the extract were compared.
Western blots were conducted on 1.75 µg of nuclear extracts prepared as previously described (64) from transfected MCF7 cells. After transfer to nitrocellulose and blocking, the blots were probed with a 1:1000 dilution of anti-ER rabbit antibody (05-820; Upstate Biotechnology, Lake Placid, NY) and a 1:2000 dilution of horseradish peroxidase-linked antirabbit IgG secondary antibody (Santa Cruz Biotechnology, Santa Cruz, CA). The blots were then developed with enhanced chemiluminescence detection solutions (Amersham Biosciences, Piscataway, NJ).
Image Collection
Cells destined for image collection were grown on No 1 borosilicate glass coverslips. One day after transfection, the cells were treated with estradiol (Sigma, St. Louis, MO), 4-hydroxytamoxifen (Sigma), or fulvestrant (Tocris Bioscience, Ballwin, MO) for 20 min (MCF7 cells) or 4 h (T47D cells). Acceptor, donor, and FRET images were acquired for each cell with a x40, 0.95 numerical aperture Plan Apochromat air objective (Olympus, Tokyo, Japan) and the CFP/YFP filter set (Chroma, Brattleboro, VT) previously described (37). To visualize the subnuclear domains where ER concentrates after ligand addition (Fig. 2C), a x100, 1.40 numerical aperture Plan Apochromat oil objective (Olympus) was used.
Within an experiment, three images (acceptor, donor, and FRET) were acquired with identical integration times, typically 300–700 msec, depending on transfection efficiency. This provided for capture of ER-tagged CFP expressed at low to mid levels, near to that of endogenous ER (Fig. 2B). By design, the acceptor amounts were expressed at variable levels surrounding CFP by varying the amounts of vector transfected. Quantifying FRET at acceptor/donor levels greater than 6–7 required quantification at acceptor amounts that would saturate acceptor image intensity. Therefore, a fourth image was acquired of the acceptor at 0.25x integration times in addition to the acceptor, donor, and FRET images captured at 1.00x integration times. Images from 522 control cells expressing varying amounts of acceptor (no donor) showed that the amounts of background-subtracted fluorescence detected in the 0.25x acceptor channel were accurately collected as 25.10 ± 0.23% that in the 1.00x acceptor channel. This enabled collecting higher levels of acceptor with shorter integration times, and then multiplying those acceptor levels by 4 to obtain acceptor/donor levels out to greater than 10. This facilitated the calculations of the binding curves (Figs. 4 and 5).
For some experiments, particularly the initial experiments in T47D cells that were conducted 4 h after ligand addition, ligand-regulated changes in the amounts of ER fusion protein present necessitated corresponding changes in image integration time. This was problematic for the fulvestrant-incubated cells in which the number of fluorescent cells was dramatically decreased 4 h after ligand addition. In the second-generation, MCF7 cell experiments that form the bulk of the studies presented, ligand incubation times were dramatically shortened to a time period in which the changes in turnover were less pronounced (Fig. 2B). By maintaining collection parameters constant and by designing post-image collection analysis to eliminate cells not containing a minimum CFP fluorescence intensity of 200 U above background (on a 12-bit scale), cells of comparable CFP fluorescence intensity were collected on average for each ligand (for example, in one experiment, average CFP intensities of 629, 712, and 692 for estradiol, 4-hydroxytamoxifen, and fulvestrant compared with 2310, 2105, and 2237 for YFP).
FRET Analysis
Fluorescence intensities within the acceptor, donor, and FRET images were processed using Metamorph Image Analysis software (Molecular Devices, Downingtown, PA), and the amounts of energy transfer were determined from background-subtracted acceptor, donor, and FRET fluorescence amounts measured in the nucleus of a cell according to previously published procedures (35, 37, 40). An example of the FRET calculation is also provided in Results. FRET amounts from multiple cells were fitted to a bimolecular association curve using Prism software (GraphPad, San Diego, CA). Statistical significance of values were determined from the 95 or 99% confidence intervals calculated in Prism and, where appropriate, by ANOVA using InStat software (GraphPad).
The FRET calculations use fluorescence bleedthrough corrections that are physical properties of the fluorophores not affected by the level of fluorescence intensity. In addition, the FRET/donor ratio is a marker of the efficiency of energy transfer from the donor to the acceptor. As a result, the FRET/donor measurements are the same for the same interaction under differing protein amounts (except, of course, where those protein amounts affect the kinetics of interaction). Thus, the FRET/donor ratio provides information about the quality of the interaction regardless of whether protein amounts are altered by, for instance, changing ER amounts upon fulvestrant incubation. However, the FRET binding curves can be affected by the relative amounts of fluorophore-tagged and endogenous protein present because the endogenous protein will compete for interactions between the fluorescent fusion proteins. This competition becomes minimized as more ectopic ER is expressed, such as at the higher acceptor/donor ratios. Thus, the Bmax levels determined will not be affected by endogenous ER or the variations in endogenous ER observed upon ligand addition.
In contrast, Kd values can be affected by competition by endogenous protein, particularly if only tracer amounts of donor-labeled protein are present. In the current experiments, this was not an issue for comparison of the estradiol and 4-hydroxytamoxifen-treated MCF7 cells, in which 20–40 min of ligand did not appreciably affect endogenous ER amount. By contrast, a decrease in the level of endogenous ER in the fulvestrant-treated cells would be expected to artificially right-shift the Kd because the same amount of fluorescence represents lower amounts of ER (endogenous plus expressed). The current studies showed a slight, non-statistically significant left shift in the Kd of fulvestrant-mediated dimerization (2.2 acceptor/donor units vs. 2.9 for estradiol). We therefore caution that a stronger left-shift in the fulvestrant binding curve may be masked by the overall effect of altering endogenous ER in these studies. It is important for users of the FRET techniques to be cognizant of these limitations in their interpretations.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
A.P., L.L., and E.M.K. have nothing to declare. F.S. has received lecture fees from Wyeth.
First Published Online September 28, 2006
Abbreviations: AF-1, Activation function-1; CFP, cyan fluorescent protein; ER, estrogen receptor; FRET, fluorescence resonance energy transfer; LBD, ligand binding domain; NTD, amino-terminal domain; SERM, selective estrogen receptor modulator; YFP, yellow fluorescent protein.
Received for publication February 13, 2006. Accepted for publication September 19, 2006.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
activity in vivo. J Biol Chem 278:6912–6920
