Synergistic enhancement of nuclear receptor function by p160 coactivators and two coactivators with protein methyltransferase activities.

Nuclear receptors (NRs) activate gene transcription by binding to specific enhancer elements and recruiting coactivators of the p160 family to promoters of target genes. The p160 coactivators in turn enhance transcription by recruiting secondary coactivators, including histone acetyltransferases such as CREB-binding protein (CBP) and p300/CBP-associated factor (p/CAF), as well as the recently identified protein methyltransferase, coactivator-associated arginine methyltransferase 1 (CARM1). In the current study, protein arginine methyltransferase 1 (PRMT1), another arginine-specific protein methyltransferase that shares a region of high homology with CARM1, was also found to act as a coactivator for NRs. PRMT1, like CARM1, bound to the C-terminal AD2 activation domain of p160 coactivators and thereby enhanced the activity of NRs in transient transfection assays. The shape of the graphs of reporter gene activity versus the amounts of CARM1 or PRMT1 expression vector indicated a cooperative relationship between coactivator concentration and activity. Moreover, CARM1 and PRMT1 acted in a synergistic manner to enhance reporter gene activation by both hormone-dependent and orphan NRs. The synergy was most evident at low levels of transfected NR expression vectors, where activation of reporter genes was almost completely dependent on the presence of NR and all three exogenously supplied coactivators, i.e. GRIP1, CARM1, and PRMT1. In contrast, with the higher levels of NR expression vectors typically used in transient transfection assays, NR activity was much less dependent on the combination of coactivators, suggesting that target gene activation occurs by different mechanisms at high versus low cellular concentrations of NR. Because multiple coactivators are presumably required to mediate transcriptional activation of native genes in vivo, the low-NR conditions may provide a more physiologically relevant assay for coactivator function.

Nuclear receptors (NRs) 1 belong to a large superfamily of transcriptional activator proteins that include the receptors for steroid and thyroid hormones, retinoids, and vitamin D (1)(2)(3)(4); the family also includes so-called orphan receptors for which an activating ligand is unknown or not required (5,6). NRs contain a highly conserved DNA binding domain (DBD) in the central region of the polypeptide chain that binds specific enhancer elements in the promoters of target genes. Activation of transcription is directed by two transcriptional activation domains: AF-1, located in the N-terminal end, and AF-2, located within the C-terminal hormone-binding domain (7)(8)(9)(10). Transcriptional activation involves two general mechanisms. Alteration of the chromatin structure around the promoter by ATPdependent nucleosome remodeling and histone modification facilitates the recruitment of the RNA polymerase II transcription preinitiation complex (11)(12)(13). The NRs also make direct or indirect physical contact with components of the transcription preinitiation complex (14,15). Activation is accomplished with the help of complexes of coactivator proteins that bind to the activation domains of the NRs (13,16).
A growing list of potential coactivators has been defined primarily by virtue of their ability to enhance the function of NRs in transient transfection experiments (13, 16 -19). Among the most well characterized of NR coactivators are a family of three related 160-kDa proteins called the p160 coactivators (SRC-1, GRIP1/TIF2, and pCIP/RAC3/ACTR/AIB1/TRAM1). The p160 coactivators contain a cluster of three conserved NR box motifs (LXXLL, where L is leucine and X is any amino acid), located approximately in the middle of the ϳ1400-amino acid polypeptide chain. The NR boxes bind to NR AF-2 regions by fitting into a conserved hydrophobic cleft on the surface of essentially all NRs that act as transcriptional activators (20 -22). In addition, the C terminus of the p160 coactivators binds to the AF-1 regions of some but not all NRs (23)(24)(25).
After binding to the activated NR through at least one of these two NR interaction domains, the p160 coactivators propagate the activating signal through at least two activation domains, AD1 and AD2, which act by recruiting a number of secondary coactivator proteins. AD1, located near amino acid 1000, binds p300 and CREB-binding protein (CBP), two related proteins that bind to and serve as coactivators for a large number of DNA-binding transcriptional activator proteins, in-cluding NRs (26 -28). CBP and p300 can acetylate histones and have been associated with chromatin remodeling (29). In addition, their ability to acetylate nonhistone proteins (30,31) and the ability of even acetyltransferase-negative mutants to act as coactivators in some settings (28) suggest that they contribute to transcriptional activation through multiple molecular mechanisms. For example, CBP and p300 associate with p300/CBPassociated factor (p/CAF), another coactivator with protein acetyltransferase activity (26). They also interact with components of the basal transcription machinery and thus may help to recruit the transcription preinitiation complex to the promoter (32,33).
We recently demonstrated that AD2, located at the C terminus of p160 proteins, can also transmit the activating signal received from NRs (25). The action of AD2 is independent of AD1 and requires neither CBP nor p300. We also identified a novel protein, coactivator-associated arginine methyltransferase 1 (CARM1), which binds to AD2 and acts as a secondary coactivator for NRs, i.e. its effect on NR function is totally dependent upon the presence of a primary coactivator in the form of a p160 coactivator (34). By amino acid sequence homology, CARM1 belongs to a family of previously identified arginine-specific protein methyltransferases, which includes the mammalian proteins, protein arginine methyltransferase 1 (PRMT1), PRMT2, and PRMT3, and the yeast protein arginine methyltransferase 1 (RMT1) (34). PRMT1 and CARM1 are very different in size, but they share a high degree of sequence homology in the central region, which contains the arginine-specific protein methyltransferase activity (34) (Fig. 1). The coactivator function of CARM1 and the homology between CARM1 and PRMT1 led us to test whether PRMT1, like CARM1, can bind to p160 coactivators and function as a coactivator for NRs. We also investigated the functional relationship between PRMT1 and p160 coactivators and the possibility of synergistic enhancement of NR function by p160 coactivators, CARM1, and PRMT1.
Protein-Protein Interaction in Vitro-GST pull-down assays were conducted essentially as described previously (42,43). GST fusion proteins were isolated from Escherichia coli BL21 (CARM1) or DH5␣ (other proteins) after induction with 0.1 mM isopropyl thio-␤-D-galactoside for 5 h. Bacterial cells were harvested, resuspended in NETN buffer (100 mM NaCl, 1 mM EDTA, 20 mM Tris-HCl, pH8.0, and containing 0.5% detergent, either Nonidet P-40 or IGEPAL CA-630). Cell debris was removed by centrifugation (13,000 ϫ g for 30 min at 4°C), and GST fusion proteins were isolated by incubation of the supernatant overnight at 4°C with rotation with glutathione-agarose beads (Sigma). Beads were washed by centrifugation (500 ϫ g for 5 min at 4°C) once with NETN and then twice with NETN containing 0.01% detergent. Proteins to be tested for interaction with GST fusion proteins on the beads were synthesized by transcription and translation in vitro in the presence of [ 35 S]methionine using the TNT-T7-coupled reticulocyte lysate system (Promega). The binding assay was conducted by incubating beads containing 1-2 g of GST or GST fusion protein with slow rotation for 60 min at 4°C with 10 l of the in vitro synthesis reaction in a 100-l total volume of NETN containing 0.01% detergent. Beads were washed four times by centrifugation with NETN containing 0.01% detergent. GST fusion protein and interacting 35 S-labeled protein were eluted from beads with Laemmli sample buffer (44) and analyzed by SDS-polyacrylamide gel electrophoresis and autoradiography.
Cell Culture and Transfection-CV-1 cells (45) were transfected in 6-well culture dishes, and cell extracts were assayed for luciferase and ␤-galactosidase activity as described previously (25). A Rous sarcoma virus (RSV)-␤-galactosidase reporter plasmid was used as the internal control to monitor for transfection efficiency, and ␤-galactosidase activity was assayed in the cell lysates as described previously (46). Data shown are the means and standard deviations of results from three transfected cultures. When hormone was required, transfected cells were grown in medium containing charcoal-treated serum (Gemini BioProducts) and 20 nM concentration of the appropriate hormone for each NR: estradiol for ER, triiodothyronine for TR, or dihydrotestosterone for AR.

RESULTS
Binding of PRMT1 to p160 Coactivators and NRs-Because CARM1 binds specifically to the C-terminal region of p160 coactivator GRIP1 (34), we tested for a similar binding activity of the homologous protein, PRMT1. When a GST-PRMT1 fusion protein was affixed to glutathione-agarose beads, it specifically bound all three members of the p160 coactivator family (GRIP1, SRC-1a, and ACTR), which were synthesized in vitro in the presence of [ 35 S]methionine ( Fig. 2A). In contrast, the p160 coactivators did not bind the control protein, GST. In similar GST pull-down assays, PRMT1 specifically bound the C-terminal AD2 region of GRIP1 (amino acids 1121-1462) (Fig.  2B), which was previously shown to bind CARM1 (34). Also, like CARM1, PRMT1 failed to bind an N-terminal fragment of GRIP1 (amino acids 5-765) containing the NR box motifs or a central fragment (amino acids 730 -1121) containing NR box III and AD1 (Fig. 2B). The N-terminal and central fragments thus served as negative controls.
GST pull-down assays also revealed specific protein-protein interaction of GST-PRMT1 with three different NRs, i.e. ER␣, TR␤1, and AR (Fig. 2C). The binding was observed in the presence or absence of a 20 nM concentration of the appropriate hormone for each NR.
Synergistic, p160-dependent Enhancement of NR Function by PRMT1 and CARM1-CARM1 was previously shown to function as a secondary coactivator for NRs in transient transfection assays, i.e. it enhanced NR function through its association with the C-terminal AD2 region of p160 coactivators (34). Because PRMT1 shares extensive sequence homology with CARM1 and can bind to the AD2 region of p160 coactivators and to NRs as well, we tested PRMT1 for coactivator function with ER, TR, and AR. CV-1 cells were transiently transfected with an NR expression vector, a corresponding luciferase reporter gene (driven by a mouse mammary tumor virus (MMTV) promoter containing the natural hormone response elements for AR or engineered enhancer elements for ER or TR), and expression vectors for various coactivators. Transfected cells were grown with the appropriate hormone for the NR used, and luciferase activity was determined from the transfected cell extracts. In the presence of a hormone-activated NR and GRIP1, PRMT1 enhanced reporter gene activity (Fig. 3). For example, transient expression of AR caused hormone-dependent reporter gene activity (Fig. 3C, histograms 1 and 2), and GRIP1 enhanced that activity 10-fold (histogram 3). Cotransfection of PRMT1 plasmid DNA caused a further 3-fold enhancement (histogram 4). Similar results were seen with ER␣ and TR␤1 (Fig. 3, A and B). Because PRMT1 bound directly to NRs as well as to p160 coactivators, we tested whether PRMT1 was able to serve as a direct or primary coactivator for NRs, i.e. without exogenously expressed p160 coactivators. In the absence of exogenous GRIP1, the activity observed with hormoneactivated NR (Fig. 3, A-C, histogram 2) was enhanced only slightly or not at all by the addition of PRMT1 (histogram 5). In the presence of GRIP1, the enhancement of NR function by PRMT1 was also dependent upon the presence of hormone (compare histograms 6 and 8), indicating that PRMT1 was acting as a coactivator for NRs rather than an independent transcriptional activator. The fact that the activity of PRMT1 depended on the presence of a p160 coactivator indicated that PRMT1, like CARM1, was acting as a secondary coactivator.
When fixed amounts of reporter gene, NR vector, and p160 vector were cotransfected with varying amounts of PRMT1 or CARM1 expression vector, both CARM1 and PRMT1 enhanced reporter gene activity in a dose-dependent manner (Fig. 4, A-C, graphs a and b) and had approximately equivalent coactivator effects on NR function. The range of CARM1 and PRMT1 vectors tested produced upward curving graphs, i.e. the slopes of the dose-response curves increased along with the amounts of CARM1 and PRMT1 vector used, suggesting a cooperative relationship between coactivator concentration and activity. When the reporter gene activity was replotted on a logarithmic scale, the data fit straight lines, consistent with a cooperative dose-response curve ( Fig. 4E shows data for AR replotted from Fig. 4C). Similar coactivator effects of PRMT1 were also observed when other members of the p160 coactivator family, SRC-1a or ACTR, were substituted for GRIP1 (Fig. 5, graphs a-c).
Because both PRMT1 and CARM1 acted as secondary coactivators for NRs, we tested for additive or synergistic coactivator effects on NR function when both proteins were expressed together with a p160 coactivator in the transient transfection assays. In the presence of the reporter gene, NR, hormone, and GRIP1, varying doses of PRMT1 plasmid DNA were cotransfected with a fixed amount of CARM1 plasmid DNA (0.5 g)

FIG. 2. Binding of PRMT1 to GRIP1 and NRs.
A, in GST-pull down assays, 35 S-labeled p160 coactivator, synthesized in vitro, was incubated with GST or GST-PRMT1 recombinant protein bound to agarose beads. Bound proteins were analyzed by SDS-polyacrylamide gel electrophoresis and autoradiography. For comparison, 10% of the labeled protein initially incubated with the agarose beads (10% input) is shown. B, in GST pull-down assays performed as described in A, the indicated 35 S-labeled GRIP1 fragments were tested for binding to GST and GST-PRMT1. C, in GST-pull down assays performed as described in A, 35 S-labeled ER␣, TR␤1, and AR were tested for binding to GST or GST-PRMT1 in the presence or absence of the appropriate hormone for each NR. The results presented are from a single experiment representative of three separate experiments. M.M., molecular mass. and vice versa (Fig. 4, graphs c and d). The effects of combining CARM1 and PRMT1 were synergistic. For example, in transfections with AR, adding 0.5 g of CARM1 DNA and 0.5 g of PRMT1 DNA together (Fig. 4D, histogram 5) caused a synergistic enhancement of activity when compared with the sum of their separate effects at the same doses (histograms 3 and 4). In other words, compared with the activity observed with NR plus GRIP1 and no methyltransferase expression vectors (histogram 2), the increase in activity achieved by the simultaneous addition of CARM1 and PRMT1 (histogram 5 minus histogram 2) was considerably greater than the sum of the enhancements caused by adding CARM1 and PRMT1 in separate assays (histogram 3 minus histogram 2 plus histogram 4 minus histogram 2).
Similar synergistic effects were observed for PRMT1 and CARM1 when ER␣ or TR␤1 was used instead of AR (Fig. 4, A  and B, graphs c and d), or when SRC-1a or ACTR was used instead of GRIP1 as the p160 coactivator for AR (Fig. 5, graphs  d and e). The synergistic effects of CARM1 and PRMT1 were dependent upon the presence of p160 coactivator and hormone (Fig. 3, compare histograms 7 and 8 with 6). Thus, an activated NR and a p160 coactivator were both required before CARM1 and PRMT1 could enhance reporter gene activity.
Synergy between CARM1 and PRMT1 Depends upon NR Levels-The coactivator effects of CARM1 and PRMT1, expressed separately or together, were examined at different levels of AR by varying the amount of transfected AR expression vector and fixing the amounts of all coactivator expression vectors ( Fig. 6). At lower levels of AR expression vector (0.1 and 0.3 g), the addition of either CARM1 or PRMT1 without the other (graphs c and d) produced a relatively small enhancement over the activity observed with AR plus GRIP1 (graph b); however, CARM1 and PRMT1 added together (graph e) had a strong synergistic coactivator effect. In contrast, at higher levels of AR expression vector (0.6 and 1.0 g), CARM1 or PRMT1 each enhanced reporter gene activity (graphs c and d) above the level observed with AR plus GRIP1 (graph b), but adding CARM1 and PRMT1 together produced no further increase (graph e).
Because the degree of synergy increased at the lower levels of AR, we extended the analysis to even lower levels (0.01 g) of AR expression vector. Under these conditions, the activity of hormone-activated AR was extremely low (Fig. 7, histogram 1) compared with the activity at higher levels of AR expression vector (Fig. 6, graph a). Coexpression of GRIP1 did not enhance the reporter gene activation observed with 0.01 g of AR vector (Fig. 7, compare histograms 1 and 5). In the presence of GRIP1, coexpression of either CARM1 or PRMT1 produced a modest 2-5-fold enhancement (compare histograms 6 and 7 with 5) which was GRIP1-dependent (compare histograms 6 and 7 with 2 and 3, respectively). Simultaneous coexpression of all three coactivators (GRIP1, CARM1, and PRMT1) produced a dramatic, synergistic, 35-fold enhancement of reporter gene activity (histogram 8). This activity was almost completely abolished by omission of any one of the three coactivators (compare histogram 8 with histograms 4, 6, and 7) or AR (data not FIG. 4. Synergistic enhancement of NR function by CARM1 and PRMT1. A-C, transfections and reporter gene assays were conducted as described in the legend for Fig. 3, with 0.25 g of pSG5.HA-GRIP1 and varied or fixed (0.5 g) amounts of pSG5.HA-PRMT1 and/or pSG5.HA-CARM1 as indicated (graphs a-d). Transfected cells were grown with the appropriate hormone for each NR. Insets show the same graphs on a smaller scale. All data for a given NR are from the same experiment, and data for Figs. 3 and 4 are from the same experiment. D, selected data from C are shown in histogram form. E, data from C were replotted on a semilog scale, and straight lines were fitted to graphs a-d.
shown). Similar synergy among these three coactivators and a similar dependence of synergy on low levels of NR expression vector were observed with ER␣ and TR␤1 (data not shown). Thus, coactivator combinations had their largest effects when NR was expressed at low levels, and under these conditions, there was a very high degree of coactivator synergy.
CARM1 and PRMT1 Coactivator Function Requires GRIP1 with a Functional AD2 Domain-The coactivator effects of CARM1 and PRMT1 on NRs required the presence of a p160 coactivator. The fact that CARM1 and PRMT1 both bind to the C-terminal AD2 region of GRIP1 suggests that this physical interaction is required for enhancement of NR function by CARM1 and PRMT1. To test this hypothesis, GRIP1 deletion mutants lacking AD1, AD2, or both regions were tested with AR, CARM1, and PRMT1 in the transient transfection system.
As described above, wild type GRIP1 functioned in a highly synergistic manner with CARM1 and PRMT1 when 0.01 g of AR expression vector was used (Fig. 7, histograms 1-8). The results observed with GRIP1⌬AD1 and the two methyltransferases (histograms 13-16) closely resembled the activation patterns observed with wild type GRIP1 (histograms 5-8). However, the activity contributed by the combined methyltransferases was about 90% less when coexpressed with the GRIP1⌬AD2 mutant (histograms 9 -12) or the GRIP1⌬AD1/ ⌬AD2 double mutant (histograms 17-20) compared with their activity in the presence of wild type GRIP1. Results with GRIP1 mutants lacking AD2 closely resembled the activity levels seen when no GRIP1 vector was transfected (Fig. 7,  histograms 1-4). By immunoblot analysis, all three mutant GRIP1 species were expressed at similar levels in COS cells (Ref. 25 and data not shown). Furthermore, at higher levels of AR and in the absence of coexpressed coactivators, the GRIP1⌬AD2 mutant still retained about half of the coactivator function of wild type GRIP1 (25), demonstrating that its loss of synergistic activity with CARM1 and PRMT1 was not due to a complete loss of function. However, even in the absence of a GRIP1 AD2 domain, the combination of CARM1 and PRMT1 caused a small but reproducible enhancement of NR function (Fig. 7, histograms 4, 12, and 20). Therefore, although a small amount of coactivator activity from CARM1 and PRMT1 was independent of exogenously expressed p160 coactivator, the great majority of the CARM1 and PRMT1 coactivator activity required the presence of a p160 protein with an intact AD2 domain. These results support the conclusion that the binding of CARM1 and PRMT1 to the AD2 region is necessary and therefore presumably responsible for their functional dependence on GRIP1.
Synergistic Action of CARM1 and PRMT1 on Orphan NRs-AR, ER, and TR represent Class I (AR and ER) and Class II (TR) NRs for which known ligands are required to activate the receptor by inducing an NR conformation that can interact with coactivators (13). A third class of NRs are the orphan NRs, for which ligands are unknown or not required for activation (5, 6). Orphan NRs that function as transcriptional activators generally have AF-2 domains that share homology with those of the hormone activated NRs and bind p160 coactivators, and the p160 coactivators can thus enhance transcriptional activation by the orphan NRs (35,39,47). We therefore tested the ability of CARM1 and PRMT1 to serve as synergistic secondary coactivators for one subfamily of orphan NRs, which consists of ERR1, ERR2, and ERR3. As indicated by their name, the ERRs are related to the ERs both in sequence and in function (35,48,49). They do not bind estrogen, but they can bind to and activate transcription from estrogen response elements and SF-1 (steroidogenic factor-1) response elements, and they also bind p160 coactivators in vitro and in vivo. All of these activities occur in the absence of any added ligand, indicating that the ERRs are constitutively active (i.e. ligand-independent) transcriptional activators.
Transient transfection experiments were performed with low amounts (0.01 g) of ERR expression vectors, since these conditions with the hormone binding NRs were shown to optimize coactivator synergy (Figs. 6 and 7) and may give rise to NR concentrations closer to physiological levels than those resulting from the use of higher levels (0.1-1 g) of NR expression vectors. In tests with GRIP1, CARM1, and PRMT1, these conditions produced patterns of coactivator synergy for the ERRs (Fig. 8A) very similar to the pattern observed for AR (Fig. 7). The reporter gene activity stimulated by each ERR, in the absence of coexpressed coactivators, was very low (Fig. 8A,  histograms 1) and was indistinguishable from the activity of the reporter gene in the absence of ERRs (compare histograms 1 and 9). Coexpression of any single coactivator with the ERR caused a negligible (less than 2-fold) increase of reporter gene activity (histograms 2, 6, and 7). The addition of GRIP1 expression vector with either CARM1 or PRMT1 expression vector caused a moderate 2-6-fold enhancement of the low basal activity (histograms 3 and 4). As with the hormone-binding NRs, a combination of GRIP1, CARM1, and PRMT1 vectors caused a dramatic enhancement of activity (histogram 5). The activity in all of these cases was entirely dependent on the presence of GRIP1 and ERR vectors (histograms 6 -11). Thus, at low ERR levels, GRIP1, CARM1, and PRMT1 together caused a dramatic and synergistic enhancement of reporter gene activity that was almost entirely dependent on the presence of all three coactivators.
In the experiments performed with ERR3 (Fig. 8A), an internal control for transfection efficiency was performed by including in the transfections a ␤-galactosidase reporter gene driven by the RSV promoter. ␤-Galactosidase activity was reproducibly enhanced 3-fold by the combination of CARM1 and PRMT1, whereas GRIP1 had little if any effect on ␤-galactosidase activity (Fig. 8B). Because the activity of the RSV and various other so-called constitutive promoters is often enhanced somewhat by transcriptional coactivators, they are not perfectly appropriate controls for transfection efficiency in these experiments. Therefore, in this entire study the luciferase activity of the NR-regulated reporter genes was not normalized with the ␤-galactosidase activity. However, because the effects of the coactivators on ␤-galactosidase activity was much less than the effects on the NR-controlled luciferase activity (3-fold versus 33-fold in the ERR3 experiment), ␤-galactosidase activity does serve as a valid control to show that differences in transfection efficiency cannot explain the observed coactivator effects on NR-dependent luciferase activity.
Synergistic Enhancement of GRIP1 Activity by PRMT1 and CARM1-Although PRMT1 bound NRs as well as GRIP1, the ability of PRMT1 and CARM1 to enhance NR function depended on the presence of a p160 coactivator. These results suggested that CARM1 and PRMT1 enhanced NR function by enhancing or mediating the coactivator activity of GRIP1. We therefore tested whether PRMT1 and/or CARM1 could directly enhance GRIP1 activity in an assay with no NR involvement. When CV-1 cells were transiently transfected with GAL4 DBD fused to GRIP1, there was weak activation of the GK1 luciferase reporter gene, which contained Gal4 response elements; this activity was severalfold higher than the activity observed with Gal4 DBD alone (Fig. 9B, histograms 1 and 2). Coexpression of PRMT1 with Gal4DBD-GRIP1 enhanced reporter gene activity, although relatively high quantities of PRMT1 expression vector were required for a moderate effect (Fig. 9A, lower  graph). CARM1 and PRMT1 together caused a synergistic enhancement of GRIP1 activity, which depended on the dose of PRMT1 vector used (Fig. 9A, upper graph). For example, when used separately, 0.5 g of CARM1 vector and 0.5 g of PRMT1 vector each caused a 50% increase in reporter gene activity (Fig. 9B, compare histograms 3 and 4 with 2). When expressed together, the same amounts of CARM1 and PRMT1 vectors increased reporter gene activity more than 5-fold (histogram 5). The synergistic enhancement of Gal4DBD-GRIP1 activity by PRMT1 and CARM1 was observed at all concentrations of Gal4DBD-GRIP1 expression vector tested (Fig. 9C). When Gal4DBD-GRIP1 was replaced by Gal4 DBD alone, the combination of CARM1 and PRMT1 caused a very small enhancement of activity, and the activity thus achieved was negligible compared with that observed with Gal4DBD-GRIP1 and the two methyltransferases (data not shown).
PRMT1 Had No Autonomous Activation Domain-To begin examining the mechanism of PRMT1 coactivator function, PRMT1 was fused to Gal4-DBD. When this fusion protein was expressed in CV-1 cells, the resulting activity of the GK1 reporter gene was no greater than that observed with Gal4 DBD alone (Fig. 10, histograms 1 and 3). In contrast, Gal4DBD-CARM1 had an activity almost 500 times higher than Gal4 DBD alone (histogram 5). To test the functional integrity of the Gal4DBD-PRMT1 expression vector, it was coexpressed with a VP16-PRMT1 fusion protein. This provided a two-hybrid test of the ability of PRMT1 to form homodimers or homo-oligomers, an ability that has been demonstrated previously (50). Although VP16-PRMT1 and Gal4 DBD produced no activity (histogram 2), the two PRMT1 fusion proteins together produced a high reporter gene activity (histogram 4), indicating that both fusion proteins produced PRMT1 proteins that can form homo-oligomers.

DISCUSSION
Mechanism of PRMT1 Coactivator Function-CARM1, a protein methyltransferase, was recently identified as a new coactivator for NRs (34). CARM1 bound to the C-terminal AD2 region of p160 coactivators and enhanced NR function only in the presence of a p160 coactivator. These findings suggested a model whereby p160 coactivators bind directly to NRs and act as primary coactivators, whereas CARM1 is recruited indirectly to the NR by binding to the C-terminal region of p160 coactivators and thus acts as a secondary coactivator. The evidence presented here indicated that PRMT1, which shares sequence homology with CARM1 and is also an arginine-specific protein methyltransferase, is also a coactivator for NRs and functions as a secondary NR coactivator in a manner similar to CARM1. Both PRMT1 and CARM1 enhanced the function of ER, TR, AR, and ERRs, indicating their ability to act as coactivators for a wide range of NRs, including hormoneactivated as well as hormone-independent NRs. Like CARM1, PRMT1 bound to the C-terminal region of p160 coactivators, and its enhancement of NR function depended on the presence of a p160 coactivator with an intact AD2 region, which is the binding site for CARM1 and PRMT1. Thus, although PRMT1 can bind directly to NRs in GST pull-down assays, this binding interaction is obviously not sufficient for the functional enhancement of NRs by PRMT1 and thus at most might play a subsidiary role to the required binding between PRMT1 and p160 coactivator. The dependence on NR and any required activating hormone demonstrated that CARM1 and PRMT1 were not directly activating the reporter gene but were acting as coactivators; and the dependence on a p160 coactivator indicated roles for CARM1 and PRMT1 as secondary rather than primary coactivators. The coactivator effects of CARM1 and PRMT1 were lost when the C-terminal AD2 region of GRIP1 was deleted but not when the AD1 region (the binding site for CBP and p300) was deleted, indicating that the Cterminal region of GRIP1 is essential for CARM1 and PRMT1 coactivator function. These data support the conclusion that the interaction of PRMT1 with the C-terminal region of p160 is the key binding interaction that recruits PRMT1 to the promoter. Further support for these conclusions was provided by the ability of PRMT1, as well as CARM1, to enhance the function of GRIP1 fused to Gal4DBD in the absence of NRs.
Activated NRs bound to enhancer elements of target genes stimulate initiation of transcription with the help of coactivators. How PRMT1 and CARM1 pass along the activating signal from NRs and p160 coactivators to the transcription machinery still remains to be determined. Both CARM1 and PRMT1 can methylate histones in vitro (34), suggesting nucleosome remodeling as one possible mechanism for their action. Histone methylation could theoretically cooperate with other types of histone modification by coactivators, including acetylation and phosphorylation, to remodel nucleosomes and thus help recruit a transcription preinitiation complex (51). Of course, just as acetylation of nonhistone proteins by coactivators may also play a role in transcriptional activation (29 -31), so methylation of nonhistone proteins by PRMT1 and CARM1 must also be considered. In addition to histones, PRMT1 can methylate a number of proteins, including heterogeneous nuclear ribonucleoprotein A1, nucleolin, fibrillarin, and poly(A)-binding protein II (52)(53)(54), which play diverse roles in RNA metabolism, including post-transcriptional steps of gene regulation. An increase in luciferase activity in transient transfection assays could result from coactivator effects on the efficiency of transcriptional activation or any subsequent step in gene expression required to produce a functional protein. Recent studies provide some examples of possible post-transcriptional mechanisms by which protein methyltransferase-type coactivators could act. In yeast arginine methylation of some heterogeneous nuclear ribonucleoproteins has been linked to their nuclear export efficiencies, and a genetic relationship was found between the function of mRNA cap-binding protein 80 and an arginine-specific protein methyltransferase (55).
The protein acetylation activities of CBP, p300, and p/CAF are required for their coactivator function with some transcriptional activators but not with others (28). Thus CBP, p300, and p/CAF use multiple mechanisms and multiple domains to transmit activating signals to the transcription machinery. By analogy, PRMT1 and CARM1 may use mechanisms in addition to or even instead of protein methylation to enhance reporter gene expression. CARM1 exhibited an autonomous transcriptional activation activity when fused to Gal4 DBD (Fig. 10); this activity is located in a domain separate from the methyltransferase region, suggesting the possibility that regions other than the methyltransferase domain of CARM1 may contribute to the transmission of the activating signal. 2 In contrast, fulllength PRMT1 fused to Gal4 DBD failed to exhibit any transcriptional activation activity (Fig. 10). Thus, PRMT1 is apparently not capable of activating transcription by itself when recruited or tethered to the promoter. The transmission of an activating signal by PRMT1, whether by the methyltransferase domain or some other domain, apparently requires action by another protein, either to activate PRMT1 or to act in cooperation with PRMT1. To understand the specific mechanisms of each methyltransferase-type coactivator, it will be important to map their functional domains and investigate possible downstream targets that bind to or are methylated by these coactivators.
Mechanism of Coactivator Synergy-Although CARM1 and PRMT1 both act as arginine-specific protein methyltransferases and as secondary coactivators for NRs, they are unlikely to be redundant in their functions because of their ex-tensive structural and functional differences. CARM1 and PRMT1 share extensive regions of amino acid homology, primarily in a 125-amino acid sequence in the central region of their polypeptide chains (34) (Fig. 1). A smaller subregion believed to be the S-adenosylmethionine binding region is also homologous to other S-adenosylmethionine dependent methyltransferases (e.g. DNA-methyltransferases) (52). Aside from the 125-amino acid homology region and one much smaller region of homology (Fig. 1), there is no convincing sequence homology between extensive portions of the N-and C-terminal regions of these two proteins. In addition, CARM1 (608 amino acids) is much longer than PRMT1 (353 amino acids) and thus has extensive regions with no counterpart in PRMT1. CARM1 has an autonomous activation domain, whereas PRMT1 does not (Fig. 10). Although CARM1 and PRMT1, like other protein arginine methyltransferases, both transfer methyl groups from S-adenosylmethionine to specific arginine residues of specific target proteins, their protein substrate specificities (i.e. proteins that they methylate efficiently) are quite different. The only known common substrate is histone H2A (34). In contrast, CARM1 methylates histone H3, but PRMT1 does not; PRMT1 but not CARM1 methylates heterogeneous nuclear ribonucleoprotein A1, histone H4, and an arginine residue in a glycinerich peptide representing a region of fibrillarin (34,52,53). Thus, substantial indirect evidence suggests that although CARM1 and PRMT1 share regions of extensive homology, they also have extensive regions with no homology and have distinct functions and modes of action. The fact that their coactivator activities for NRs are synergistic also indicates that they rely on different mechanisms for their coactivator activities.
One attractive hypothesis to explain the synergistic coactivator functions between CARM1 and PRMT1 is their abilities to methylate different proteins in the coactivator complex, chromatin, or other components of the transcription machinery or RNA processing machinery. Their complementary abilities to methylate histones H3 and H4 suggest one possible mechanism of synergy, but methylation of other proteins could also play a role. Methylation of histones may cooperate with acetylation and other types of covalent histone modifications to remodel chromatin structure and/or provide a signal to facilitate assembly of the active transcription initiation complex (51). Of course, the synergistic coactivator function of CARM1 and PRMT1 could involve other types of complementary mechanisms in addition to or instead of protein methylation, such as interactions with other components of the transcription or RNA processing machinery. Further work will be required to characterize the domains of CARM1 and PRMT1 required for binding to p160 coactivators, for protein methylation, and for transmitting the activating signal to the transcription machinery.
A cooperative relationship was observed between reporter gene activity and the amount of CARM1 or PRMT1 expression vector used in transfections (Fig. 4). Such a cooperative doseresponse curve may suggest that multimers of these methyltransferases function more efficiently as coactivators than monomers or that the binding or action of one CARM1 or PRMT1 molecule facilitates the binding or action of a second CARM1 or PRMT1 molecule. Such a mechanism could account for the cooperative curves observed as well as the synergy between CARM1 and PRMT1. The precise mechanisms remain to be elucidated, but previously published results and our preliminary data indicate that CARM1 and PRMT1 can form homo-oligomers (50) (Fig. 10 and footnote 3). It should be stated that although we have demonstrated a requirement for p160 and methyltransferase-type coactivators in these assays, other coactivators endogenous to CV-1 cells (such as CBP, p300, and p/CAF) may also be necessary for the observed NR function.
Significance of Coactivator Synergy-When low levels of NR expression vectors were used in the transient transfection assays, a remarkable degree of synergy was observed among the NRs and three different coactivators (GRIP1, CARM1, and PRMT1) such that if any one of these four components was not expressed, there was an almost complete loss of reporter gene activity (Figs. 7-8). In other words, the activity of low levels of NRs was almost undetectable unless GRIP1, CARM1, and PRMT1 were all coexpressed with the NR. This extremely high level of synergy was observed only when very low levels of NR expression vectors were used. In contrast, at the higher levels of transfected NR expression vectors more typically used in transient transfection assays, the effects of the NRs were readily observable in the absence of any coexpressed coactivators; p160 coactivators alone or p160 coactivator combined with either CARM1 or PRMT1 caused moderate enhancement of reporter gene activity (Fig. 6). However, the addition of a third coactivator failed to cause further stimulation. How should one interpret these dramatically different levels of requirement for coactivators? Because transient transfections generally result in higher than physiological levels of expression, the lower NR concentrations achieved by transfecting very low amounts of NR expression vector may be closer to the physiological levels of NR. Thus it is possible that some cells expressing low endogenous levels of NRs may require the presence of all three of these coactivators (GRIP1, CARM1, and PRMT1), as well as other coactivators not tested for in these studies, for NRs to activate their target genes efficiently. In contrast, high levels of NRs appear capable of activating transcription with less help from coactivators (i.e. with help from fewer coactivators and/or lower levels of coactivators).
In this report, we have identified a new coactivator for NRs (PRMT1), demonstrated synergy between p160 coactivators and two different methyltransferase-type coactivators, determined specific physical interactions between p160 and methyltransferase coactivators that underlie this synergy, and defined conditions that make NR function almost entirely dependent upon the presence of three different coactivators (GRIP1, CARM1, and PRMT1). Although many putative coactivators for NRs have been reported, most have been studied separately from other coactivators. However, the current state of knowledge suggests that many coactivators may need to cooperate to mediate the process of transcriptional activation. Our finding of conditions where multiple coactivators are required for NR function is consistent with this prediction and furthermore provides a new set of experimental conditions that may be important and valuable for investigating coactivator function. Thus, for future coactivator studies it may prove advantageous to use in parallel two sets of experimental conditions: one with higher levels of NR expression vector, where effects of individual coactivators and pairs of coactivators can be observed; and a second set of conditions with lower levels of NR expression vectors, where there is a high degree of requirement for multiple coactivators. We suggest that the conditions producing a requirement for multiple coactivators may more closely resemble the physiological state. It will also be important to search for other coactivator combinations that exhibit synergy and for other experimental conditions that may allow a dependence on higher numbers of coactivators.