CARM1 mediates the ligand-independent and tamoxifen-resistant activation of the estrogen receptor α by cAMP

cAMP activates endogenous ERα through PKA

A literature survey and pilot experiments suggested that experimental parameters might strongly influence signaling cross-talk between cAMP and ERα. We therefore wished to ensure that we could recapitulate and study the phenomenon in our own setting. The response of endogenous ERα was assessed by a transactivation assay with human MCF7 breast cancer cells. A typical result obtained with a transfected luciferase reporter gene for ERα is shown in Supplemental Figure S1. As expected from previous studies, ERα is activated by both the cognate ligand estradiol (E2) and a cell-permeable analog of cAMP. Since the response is blocked by the PKA inhibitor H89 and mimicked by overexpression of the catalytic subunit of PKA (Fig. 1; Supplemental Fig. S1), we conclude that cAMP activates ERα through PKA, and that they can be used interchangeably despite quantitative differences in the response. Intriguingly, in early experiments, PKA overexpression occasionally failed to turn on ERα. Prompted by a report that the PKA phosphorylation of S236 in the DNA-binding domain of ERα inhibits dimerization and, as a consequence, DNA binding (Chen et al. 1999b), we analyzed the effects of PKA more carefully by performing a dose response experiment (Supplemental Fig. S1B). The bimodal response of ERα to increasing PKA concentrations suggests that low and intermediate levels of PKA can activate ERα, and that the inhibitory phosphorylation of S236 becomes dominant at high PKA levels.

Domain requirements for signaling cross-talk

To dissect the ERα domain requirements in transfection experiments (Fig. 1), we switched to ER-negative cell lines such as SKBr3 or 293T. The N-terminal A/B and the C-terminal E/F domains are associated with the transcriptional activation functions AF1 and AF2, respectively. Interestingly, deletion of either domain does not abolish the response to cAMP or PKA (Fig. 1A,B). The N-terminally truncated receptor can still be activated in the absence of E2, and the C-terminal truncation mutant, only poorly active in this experimental system, is further stimulated. Whereas the constitutively active A/B domain by itself does not seem to be sufficient to respond to cAMP/PKA as a fusion protein with the Gal4 DNA-binding domain, the HBD responds robustly in this context, suggesting that the HBD may be the primary target domain (Fig. 1C). The very C-terminal and poorly conserved F domain, which was present in Gal4 construct Gal4.ERα but is not part of the HBD itself, is dispensable for this response (Supplemental Fig. S2A). The role of known ERα phosphorylation sites was explored with a panel of point mutants. Apart from S236, S305 was of particular interest because its phosphorylation by p21-activated kinase 1 and PKA had been demonstrated to mediate ligand-independent activation (Wang et al. 2002) and tamoxifen resistance (Michalides et al. 2004), respectively. Phosphorylation of S305 had also been shown to increase the hormone sensitivity of ERα by blocking acetylation of K303 (Cui et al. 2004). Although there are quantitative effects of mutating these serine residues, none of these mutations abolishes the response to cAMP/PKA (Fig. 1D). These findings are consistent with the idea that the direct phosphorylation of ERα by PKA, which may occur, is not a major determinant for this type of signaling cross-talk.

Signaling cross-talk depends on AF2 and on a novel ERα-specific surface of the HBD

An AF2 mutant of ERα or a dominant-negative mutant coactivator block cAMP activation (Supplemental Fig. S2). Although these findings indicated that AF2 is required, this did not exclude the possibility that other surfaces of the HBD might also be involved. In this context, the ERα mutant G400V provided a valuable hint. It has a slightly lower affinity for E2, but still achieves full activation at higher E2 concentrations (Tora et al. 1989). In contrast, it had been reported that its response to the neurotransmitter dopamine, most likely mediated by cAMP/PKA, is completely abolished (Smith et al. 1993). We therefore revisited this result with a Gal4-ERα HBD fusion protein, which provides a larger dynamic range than wild-type ERα. Whereas E2 induces G400V almost 40-fold, PKA is completely ineffective (Supplemental Fig. S2B).

Residue G400 is not part of AF2. In fact, in the dimeric HBD, it is located on the opposite side of the subunit as well as far from AF2 of the other HBD monomer (Supplemental Fig. S3). G400 forms part of a turn between helix 5 (H5) and the only β sheet in the HBD (Fig. 2A; Supplemental Fig. S3). According to the crystal structure, a valine substitution would result in the isopropyl group of the side chain protruding outside of the structure into a solvent-accessible area, which appears to be part of a larger groove-like structure in the HBD dimer. In addition to the loop, several amino acids from H6 and H7 belonging to one subunit, and from H8 to H9 from the other subunit, contribute to the formation of this groove. Among this group, the amino acids R412, S433, S463, and S464, which are most proximal to G400 and carry charges or hydrophilic side chains, were selected for mutagenesis (Fig. 2). In addition, L429 was also selected because its side chain appears to form part of the base of the groove and is most proximal to G400. Interestingly, G400, R412, S463, and S464 are highly conserved in ERα sequences from a large range of vertebrate species (Supplemental Fig. S3A).

Initially, our panel of alanine substitution mutants were tested as Gal4-ERα HBD chimeras (Supplemental Fig. S4). The key results were confirmed in the context of full-length ERα by a transactivation assay in ER-negative 293T cells (Fig. 2). Whereas the full-length version of L429A also yielded a higher maximal response in response to cAMP, it remains to be further examined whether it is activated by lower concentrations of E2 as its Gal4 fusion variant (Supplemental Fig. S4). Since ERβ, which is completely refractory to cAMP (Supplemental Fig. S5), contains an alanine in the position corresponding to S464, we speculated that the double mutant S463/4A might be defective because of the substitution of S464. As can be seen in Figure 2, an alanine substitution abolishes the cAMP response but not the response to E2, whereas S464D behaves like wild type for both responses. While we cannot formally exclude that S464 is phosphorylated by PKA, we consider this very unlikely in view of our other findings (see below), and the fact that two different groups have shown that S305 is by far the major PKA phosphorylation site (Cui et al. 2004; Michalides et al. 2004). The functionality of S464D suggests that it is the specific nature of the substitution in S464A that leads to the cAMP signaling defect. Thus, the specific cAMP response defects of G400V and S464A suggest that the groove that these residues line might represent a novel regulatory or interaction surface of ERα for the recruitment of a coregulator in response to cAMP.

Coactivator-associated arginine methyltransferase 1 (CARM1) is a mediator of signaling cross-talk

Since our attempts to identify such proteins by an unbiased biochemical approach proved unsuccessful, we turned to a candidate protein approach. We were intrigued by a report that apo-ERα cycles on and off a target promoter along with several known coregulators, but in the notable absence of AF2 coactivators of the p160 family (Métivier et al. 2004). It seemed possible that one or several of these cycling factors might engage in transcriptionally productive complexes following cAMP/PKA stimulation. We decided to examine more closely the histone acetyltransferases CBP/p300, and CARM1 (for review, see Bedford and Clarke 2009; Lee and Stallcup 2009).

The overexpression of p300, CBP, or CARM1 in MCF7 cells markedly increased the response of the endogenous ERα to cAMP signaling (Fig. 3A; Supplemental Fig. S6A). It had no effect on the activity of apo-ERα in the absence of cAMP. Whereas CBP/p300 can interact directly with ERα through both AF1 and AF2 even in the absence of ligand (Kobayashi et al. 2000; Dutertre and Smith 2003), CARM1 is known to be recruited indirectly to liganded ERα; for example, through the p160 coactivator GRIP1 (Chen et al. 1999a, 2000). We therefore investigated the recruitment of CBP and CARM1 to ERα in response to cAMP (Supplemental Fig. S6B). As expected, endogenous ERα could be constitutively coprecipitated with HA-tagged CBP from transfected MCF7 cells. In contrast, under our experimental conditions, HA-CARM1 only immunoprecipitated ERα after stimulation with cAMP. The cAMP dependence for this interaction was confirmed by a reciprocal immunoprecipitation (IP) of endogenous CARM1 with endogenous ERα (Fig. 3B), and, using a chromatin IP (ChIP) experiment, we found that CARM1 is recruited to the ERα target gene pS2 (officially known as TFF1) in response to both E2 and cAMP (Fig. 3C). Thus, regarding the response to cAMP, IP and ChIP results correlate, including with earlier reports showing the recruitment of ERα itself to the pS2 gene (Al-Dhaheri and Rowan 2007; Fenne et al. 2008), and support the conclusion that endogenous CARM1 is recruited to ERα upon cAMP signaling.

We next used an RNAi experiment (see Supplemental Fig. S7 for validation) to determine whether CARM1 is also necessary. Knocking down CARM1 expression in MCF7-SH cells with a specific shRNA construct reduced the cAMP-induced activity of endogenous ERα, as assayed with a transfected reporter gene or by analyzing expression of the endogenous pS2 gene (Fig. 3D,E), whereas it did not affect the E2-induced activity. Note that the latter is consistent with previous reports (Chen et al. 1999a, 2000) that also indicated that E2-induced ERα activity may be largely CARM1-independent under these experimental conditions, and even for some endogenous target genes (Frietze et al. 2008). The fact that CARM1 is required for some E2-induced responses (see also Yadav et al. 2003) supports the notion that coregulator requirement can be highly tissue- and context-dependent.

To facilitate the next series of experiments, we verified that the cAMP response of the Gal4-ERα HBD chimera (Gal4.ERα), like that of endogenous ERα in MCF7 cells, can be further augmented by overexpression of CARM1, and that there is no effect of CARM1 overexpression on Gal4.ERα activity in the absence of cAMP (Fig. 4A).

First, we tested whether CARM1 activity is required by comparing the augmentation by wild-type CARM1 and by a methyltransferase mutant CARM1. To our surprise, the enzymatic activity turned out to be dispensable (Supplemental Fig. S8A), as reported recently for the coactivation of NFκB by CARM1 (Jayne et al. 2009). Next, we speculated that the cAMP-induced recruitment of CARM1 to ERα is nevertheless necessary for the transcriptional response, and that it might involve a new regulatory or interaction surface on the ERα HBD. We tested this notion by using our set of HBD point mutants (Fig. 4). They were coexpressed as Gal4-ERα HBD chimeras along with CARM1 in 293T cells. The experiment showed that, with the notable exceptions of the single-point mutant G400V and the double mutant S463/4A, all other mutants can be potentiated (Fig. 4B; Supplemental Fig. S8B). L429A again was the most active one. S463/4A was not only refractory to cAMP and/or CARM1 overexpression, it also failed to interact with it, as judged by an IP experiment (Fig. 4C).

Direct phosphorylation of CARM1 by PKA and interaction with ERα

With CARM1 established as a necessary mediator, it seemed plausible that it might be the target of PKA. We used an antibody directed against the PKA-phosphorylated motif RRX-phospho-S/T to immunoprecipiate PKA substrates from MCF7 extracts, and probed for the presence of CARM1 in the immunoprecipitate. The result shows that endogenous CARM1 becomes phosphorylated in cells treated with cAMP (Fig. 5A). We then looked for potential PKA phosphorylation sites in CARM1 using Scansite 2.0 (http://scansite.mit.edu). At medium stringency, the motif scan yielded only one hit: S448 within the sequence KRQS. The importance of this residue for in vivo phosphorylation of HA-CARM1 by PKA was assessed by expressing both the wild-type and S448A mutant versions in transfected 293T cells. The IP was done as before, except that this time the presence of exogenous HA-CARM1 in the immunoprecipitates was revealed with an anti-HA antibody (Fig. 5B). In contrast to wild-type CARM1, the mutant S448A appears to be completely refractory to PKA-induced phosphorylation. The phosphoserine mimic S448E is constitutively recognized by the antibody, further corroborating the prediction of this residue as a PKA target site. To confirm that CARM1 is phosphorylated directly by PKA, we incubated purified recombinant CARM1 with a commercial preparation of PKA. Again, only wild-type, but not the S448A mutant, could be phosphorylated (Fig. 5C). These results argue very strongly that S448 is phosphorylated by PKA, and that it is the major, if not the only, PKA phosphorylation site of CARM1.

The impact of CARM1 phosphorylation on the interaction with ERα was investigated with a GST pull-down experiment using purified recombinant interaction partners (Fig. 6A). CARM1 was first phosphorylated with PKA before incubation with a GST fusion of the ERα HBD. Unlike the mutant S448A, which could not be phosphorylated, wild-type CARM1 was able to interact with the ERα fusion protein. Note that the GST-ERαHBD fusion protein does not get detectably phosphorylated by PKA that remains present during the GST pull-down (data not shown). Thus, the phosphorylation of CARM1 by PKA promotes the direct interaction with apo-ERα in the absence of other cellular proteins. The mutant S448E, which resembles the phosphorylated form, interacts constitutively with ERα. These results argue that phosphorylation of S448 by PKA is sufficient to promote the interaction with the ERα HBD.

Since the mutant S448A fails to be phosphorylated by PKA and to interact with ERα, it was to be expected that it would also be defective for stimulating the cAMP response of ERα in vivo (Fig. 6B). This clearly demonstrates that S448 is necessary for this response. Interestingly, the phosphoserine mimic S448E behaves like wild type in that it does not stimulate ERα transcriptional activity without stimulation by PKA. As will be discussed further below, this suggests that the phosphorylation of S448 may not be sufficient for stimulation of transcriptional activity.

Elevated CARM1 phosphorylation in tamoxifen-resistant cells promotes ERα binding and tamoxifen-induced activity

One setting where our findings might be physiologically relevant is in the context of breast cancer progression. There is already evidence that tamoxifen resistance correlates with and may be due to increased PKA activity (Michalides et al. 2004; Zwart et al. 2007). Using the tamoxifen-resistant MCF7 subclone LCC2 (Brünner et al. 1993), we explored the possibility that its altered behavior correlates with changes in the CARM1–ERα relationship. While ERα and CARM1 levels are comparable between wild-type and LCC2 cells, a CARM1 IP revealed that endogenous CARM1 is constitutively associated with ERα in LCC2 cells (Fig. 7A). This may be explained by the fact that PKA phosphorylation of CARM1 displays a high basal level in LCC2 cells (Fig. 7B), which in turn may be due to elevated levels of PKA in LCC2 relative to MCF7 cells (data not shown). Indeed, the PKA inhibitor H89 abolishes the constitutive interaction in LCC2 cells (Fig. 7C). Wild-type MCF7 and LCC2 cells also differ in the effect of (hydroxy-)tamoxifen (OHT) on the CARM1–ERα interaction. The interaction of the endogenous proteins is sensitive and resistant to OHT in wild-type MCF7 and LCC2 cells, respectively.

The functional impact on the activity of endogenous ERα became apparent with a luciferase reporter gene assay (Fig. 7D). OHT activates ERα as well as cAMP in LCC2 but not wild-type MCF7 cells, and this activity depends on both PKA activity and CARM1. It is important to note that the endogenous ERα target gene pS2 responds exactly the same way (Fig. 7E). Its activation by OHT in LCC2 cells depends on both CARM1 and PKA activity. This OHT response, and the resistance to OHT, can be recapitulated in ER-negative 293T cells by either overexpressing wild-type CARM1 and stimulating with cAMP, or overexpressing the CARM1 phosphoserine mimic S448E even in the absence of cAMP (Supplemental Fig. S9).

Earlier efforts to unravel the mechanism of ligand-independent activation of ERα by cAMP

Past efforts to unravel the mechanism were hampered by various difficulties inherent to this system. There is no doubt that PKA phosphorylates ERα directly and does modulate its activity in various ways (Aronica and Katzenellenbogen 1993; Chen et al. 1999b; Cui et al. 2004; Michalides et al. 2004; Al-Dhaheri and Rowan 2007). One obvious complication is that the phosphorylation of S236 in the DNA-binding domain inhibits even DNA binding of ERα (Chen et al. 1999b). Hence, the level and, perhaps, the duration, of activation become important parameters that are difficult to control. It had been shown that cAMP/PKA stimulates the recruitment of cyclin D1 to unliganded ERα (Lamb et al. 2000) and of the coactivator SRC-1 to the unliganded chicken PR (Rowan et al. 2000), but whether the latter is true for ERα and whether either cofactor is necessary for the response were not determined. In fact, a mutational analysis indicated that none of the known phosphorylation sites of SRC-1 is necessary (Coleman et al. 2003).

CARM1 is the link between PKA and ERα

Once we established that none of the known PKA phosphorylation sites on ERα are necessary for ligand-independent activation of ERα by cAMP/PKA, it was clear that the target had to be a cofactor. With CARM1, we believe to have closed the gap between PKA and ERα. CARM1 is phosphorylated directly by PKA at S448, and both CARM1 and this phosphorylation are necessary for the response. S448 resides within the PKA consensus phosphorylation site R/KXXS/T, and this whole motif (KRQS) is highly conserved in evolution, but not in other protein arginine methyltransferases such as PRMT1. The known structure of CARM1 (Troffer-Charlier et al. 2007; Yue et al. 2007) does not allow one to predict the molecular effects of the phosphorylation of S448, which is located in the catalytic core domain, for binding to ERα and/or coactivator activity. The discovery of this phosphorylation site further emphasizes that CARM1 is subject to regulatory inputs. As-yet-unknown kinase(s) phosphorylate(s) human CARM1 at two different serine residues—S217 and S229—to inhibit the methyltransferase activity (Higashimoto et al. 2007; Feng et al. 2009). Whereas the phosphorylation of S448, which can be mimicked by a phosphoserine mutant, proved to be sufficient to promote the binding to the ERα HBD, it is insufficient to activate ERα in the absence of ligand. The most likely explanation is that there is yet another factor that PKA signaling must modify to allow the transcriptional activation of unliganded ERα. Future work will address these additional mechanistic questions, as well as the questions of whether CARM1 also mediates the activation of PR by cAMP, and whether the phosphorylation of CARM1 modulates the activity of other CARM1-dependent transcription factors.

A novel interaction groove in the ERα HBD

Our mutational analysis of the ERα HBD revealed that a novel regulatory surface is involved in mediating the response to cAMP. We speculate that a groove that is formed at the interface of the two subunits of the HBD homodimer might serve as the binding surface for CARM1. Determining the structure of the complex will be necessary to prove this hypothesis, and to determine whether this groove accommodates CARM1 through phosphorylated S448. Interestingly, this groove is located on the “back” side of the ERα HBD with respect to AF2, and has not received much attention. Other ERα coregulators might use this surface as well. ERα S464, which is located at the edge of the groove, seems to be a critical residue for this response. While this residue is highly conserved in ERα across vertebrates, it is an alanine in human ERβ. Considering that the ERα S464A mutant cannot be activated by cAMP/PKA, this sequence divergence between the two ER isoforms provides an explanation for why unliganded ERβ cannot be activated by cAMP/PKA. This does not exclude that cAMP/PKA could modulate certain ERβ target genes through alternate mechanisms. It was demonstrated that activation of ERβ by cAMP/PKA could be monitored in transfections with reporter genes containing promoter-proximal cryptic AP1 sites (Coleman et al. 2003), but it should be emphasized that all of our reporter genes were devoid of these. Our results suggest that cAMP signaling would help to discriminate, at least at certain target genes, between ERα and ERβ effects in cells that coexpress the two isoforms, and thereby to fine-tune estrogenic responses.

Building signal-specific coactivator complexes for signal-specific responses

With CARM1 binding the ERα HBD on the “back” side, this potentially leaves the AF2 surface available for binding other coactivators. To allow this, one must assume that CARM1 binding somehow results in the release or reorganization of the Hsp90 complex and in the repositioning of helix 12 of the ERα HBD to form AF2. This is indeed supported by our finding that AF2 is required for cAMP/PKA activation of ERα. Whether CARM1 and AF2 coactivators are bound simultaneously—and, if so, which AF2 coactivator(s)—remains to be established, but it is intriguing to speculate that CARM1 might act as a pioneer factor (Fig. 8). Its main function and mode of action might be to allow and to promote the recruitment of yet other factors. In addition to those that might be recruited to the “CARM1 remodeled” ERα HBD, the C-terminal CARM1 transactivation domain (Teyssier et al. 2002) and interactors such as Flightless I (Lee et al. 2004) might contribute.

Interestingly, although prolonged treatment with cAMP leads to the proteasome-dependent degradation of GRIP1 (Hoang et al. 2004), early during induction, cAMP stimulates the recruitment of GRIP1 to the ERα target gene pS2 (Fenne et al. 2008). Although it was not determined whether this is necessary for cAMP activation of ERα, it is intriguing to speculate that, in this case, the recruitment of GRIP1 to ERα may be CARM1-dependent, rather than the other way around, as in the presence of E2 (Fig. 8).

Conversely, CARM1 may also be recruited in two different ways (Fig. 8). Following phosphorylation by PKA, it can bind unliganded ERα directly, whereas in the presence of estrogen, it can be incorporated into ERα complexes indirectly through a p160 coactivator such as GRIP1. Even if CARM1 is built into the complex in both cases, the functional requirements, topology, stoichiometry, and dynamics of these complexes are likely to be different. For example, the arginine methyltransferase activity of CARM1, and thus its ability to methylate histone H3 and other proteins, is necessary for CARM1 function in the context of E2-induced ERα–coactivator complexes (Lee et al. 2002), but it is dispensable for the cAMP response. Such differences may affect the selection of additional CARM1 and/or ERα partners through other protein surfaces.

Thus, depending on how ERα is activated, it will recruit different coregulator complexes. This prediction is supported by a recent study that examined an analogous situation in which the recruitment of several coregulators to ERα target genes was shown to be different depending on whether ERα was activated by E2 or IGF-1 (Cascio et al. 2007). As a consequence, one might expect that the activating signal of ERα determines target gene selection and/or regulation. We recently carried out a gene expression profiling analysis with MCF7 cells that shows just that: ERα regulates considerably different sets of genes depending on how it is activated (Dudek and Picard 2008).

Pathological and physiological significance

With the CARM1–ERα connection, we add another twist to the already established link between cAMP/PKA signaling and tamoxifen resistance of breast cancer cells (Michalides et al. 2004; Zwart et al. 2007). Our results obtained with the tamoxifen-resistant MCF7 variant LCC2 indicate that increased PKA activity also preserves ERα activity in the presence of OHT through the recruitment of CARM1. The finding that OHT abolishes the cAMP-induced interaction in wild-type MCF7 cells but not the constitutive interaction in tamoxifen-resistant LCC2 cells argues that high and, perhaps, persistent PKA activities may be required for tamoxifen resistance. The fact that the OHT stimulation of ERα activity could be recapitulated in ER-negative cells by overexpression of the CARM1 phosphoserine mimic mutant lends further support to these notions. Unfortunately, we could not directly test the CARM1 requirement for tamoxifen resistance in a proliferation assay, since a prolonged knockdown of CARM1 proved to be lethal (data not shown). Manipulating the CARM1–ERα interaction may yield interesting insights and turn out to be a promising therapeutic avenue in the future.

A large number of extracellular signals lead to the activation of adenylate cyclase, and the ensuing increase in cAMP and activation of PKA contribute to their cellular and physiological effects. Considering that cAMP can activate ERα, it is therefore conceivable that far more of these extracellular factors than previously thought also signal through ERα. We now have the tools to explore the pathological and physiological consequences of ligand-independent activation of ERα by cAMP.

Plasmids

Details about the plasmids used can be found in the Supplemental Material.

Cell culture, transfection, RNAi, and luciferase assays

Where indicated, E2 (to 10–100 nM), the PKA inhibitor H89 (to 5 μM), and 8-Br-cAMP (to 1 mM from a 100 mM stock in H₂O) were added. Firefly luciferase activities were standardized to a Renilla luciferase transfection control. Unless indicated otherwise, the data shown are averages of triplicate samples (error bars indicate standard deviations). Additional details are given in the Supplemental Material.

Antibodies and recombinant proteins

The following primary antibodies were used: the mouse monoclonals 2GV10 against the Gal4 DNA-binding domain (White et al. 1992), HA.11 against the HA tag (Babco), H90-10 against Hsp90 (a gift from David O. Toft), His-1 against the His6 tag (Sigma), anti-phospho-Ser/Thr (Upstate Biotechnologies), the rabbit monoclonal 100G7E against substrates phosphorylated by PKA (Cell Signaling), the rabbit polyclonal sera CARM1-421A against CARM1 (Bethyl Laboratories), and HC-20 against the ERα C terminus (Santa Cruz Biotechnologies). Recombinant proteins were expressed in Escherichia coli and purified on glutathione-Sepharose (Amersham) or TALON Metal Affinity Resin (Clontech) as directed by the manufacturers. After elution from the beads, His6-tagged proteins were dialyzed against 10 mM Tris-HCl (pH 7.02), 6.25 mM MgCl₂, and 10% glycerol, and stored at −80°C.

IP experiments

The standard procedure for all IP experiments was as follows (except for those with the anti-HA monoclonal) (see the Supplemental Material): Cells were lysed by sonication in 20 mM Tris-HCl (pH 7.5), 1 mM EDTA, 10% glycerol, 20 mM Na-molybdate, and protease inhibitors. Five-hundred micrograms of each extract were first incubated with the primary antibody for 2 h at 4°C, and then with 20 μL of protein G-sepharose beads for an additional 16 h at 4°C. Beads were washed four times with a buffer containing 20 mM Tris-HCl (pH 7.5), 1 mM EDTA, 10% glycerol, 20 mM Na-molybdate, 50 mM NaCl, and 0.1% Tween.

ChIP experiments

The recruitment of CARM1 to the pS2 promoter upon stimulation of MCF7-SH cells with 100 nM E2, 1 mM 8-Br-cAMP, or vehicle for 45 min was determined by ChIP with the polyclonal CARM1 antiserum. Further details are given in the Supplemental Material.

Quantitative RT–PCR analysis of pS2 expression

MCF7 and LCC2 cells were induced for 4 h, and gene expression was analyzed by standard procedures as detailed in the Supplemental Material.

In vitro phosphorylation and interaction assays

Five-hundred nanograms of His6-tagged CARM1 were incubated with 70 ng of the catalytic β subunit of PKA (Stressgen) and 20 mM ATP in 20 μL of PKA buffer (10 mM Tris-HCl at pH 7.02, 6.25 mM MgCl₂) for 1 h at 30°C. The reaction was then added to 10 μL of glutathione-sepharose beads loaded with GST-ERαHBD in 200 μL of pull-down buffer (20 mM HEPES-KOH at pH 7.9, 4.3 mM Na-phosphate, 10% glycerol, 100 mM NaCl, 100 mM KCl, 5 mM MgCl₂, 0.2 mM EDTA, 1 mM DTT, 0.2 mM PMSF) and incubated overnight at 4°C. The beads were washed four times with pull-down buffer.