Functional interaction between the mouse notch1 intracellular region and histone acetyltransferases PCAF and GCN5.

The Notch receptor that plays an important role in cell fate determination is intracellularly cleaved by interaction with the ligand. The cleaved intracellular region (RAMIC) of Notch is translocated into the nucleus and interacts with a DNA-binding protein RBP-J to activate transcription of genes that regulate cell differentiation. Although RAMIC has been shown to facilitate the RBP-J-mediated transactivation by displacing the histone deacetylase corepressor complex from RBP-J, there is no evidence demonstrating the involvement of histone acetyltransferases (HATs) in the transactivation. Here we show that mouse Notch1 RAMIC interacts with two conserved HATs, mouse PCAF and GCN5, and recruits each of the HATs to RBP-J. The ankyrin repeats and the transactivation domain of RAMIC and the N-terminal regions of PCAF and GCN5, respectively, are required for the interaction. We also show that not only mouse Notch1 but also Drosophila Notch RAMIC interacts with mouse PCAF and GCN5 in mammalian cells. Furthermore, the RBP-J-mediated transactivation activity of RAMIC is repressed by two HAT inhibitor proteins, E1A and Twist. These results suggest that HATs including PCAF and GCN5 play an important role in the RBP-J-mediated transactivation by RAMIC.

The Notch receptor that plays an important role in cell fate determination is intracellularly cleaved by interaction with the ligand. The cleaved intracellular region (RAMIC) of Notch is translocated into the nucleus and interacts with a DNA-binding protein RBP-J to activate transcription of genes that regulate cell differentiation. Although RAMIC has been shown to facilitate the RBP-J-mediated transactivation by displacing the histone deacetylase corepressor complex from RBP-J, there is no evidence demonstrating the involvement of histone acetyltransferases (HATs) in the transactivation. Here we show that mouse Notch1 RAMIC interacts with two conserved HATs, mouse PCAF and GCN5, and recruits each of the HATs to RBP-J. The ankyrin repeats and the transactivation domain of RAMIC and the Nterminal regions of PCAF and GCN5, respectively, are required for the interaction. We also show that not only mouse Notch1 but also Drosophila Notch RAMIC interacts with mouse PCAF and GCN5 in mammalian cells. Furthermore, the RBP-J-mediated transactivation activity of RAMIC is repressed by two HAT inhibitor proteins, E1A and Twist. These results suggest that HATs including PCAF and GCN5 play an important role in the RBP-J-mediated transactivation by RAMIC.
The Notch family consists of single transmembrane receptors that are involved in cell fate determination in multiple steps of development (1). Notch has been found in a variety of organisms from nematode and fruit fly to higher vertebrates, and four Notch proteins have been identified in mammals. It was shown that the Notch receptor is proteolytically processed in the trans-Golgi network into the N-terminal fragment containing most of the extracellular region and the remaining C-terminal fragment, and that the two fragments are presented on the cell surface as a functional heterodimer (2,3). The Notch signal triggered by interaction with the ligand like Delta (4,5) or Jagged/Serrate (6, 7) blocks differentiation of stem or progenitor cells and keeps them in a multipotential state. Recent studies indicate that interaction with the ligand induces a second cleavage at a site within or near the transmembrane region of the C-terminal fragment of the heterodimer, resulting in the release of the intracellular region and its nuclear translocation (8,9).
The translocated nuclear Notch intracellular region binds to a ubiquitous DNA-binding protein Drosophila Suppressor of Hairless (Su(H)) 1 or its mammalian homolog RBP-J/CBF1 (10), and thereby transcription of the downstream target genes such as Enhancer of split [E(spl)] complex genes (11,12) or mammalian homologs of Hairy and Enhancer of split, HES-1 and HES-5 genes (4,5,13,14) is up-regulated. Basic helix-loophelix (bHLH) proteins encoded by the E(spl) and HES genes antagonize other bHLH transcription factors that induce cell differentiation (15). Thus the Notch signal can be mimicked by forced expression of the Notch intracellular region (16 -18). The intracellular region of mouse Notch1 (hereafter designated as RAMIC) contains two evolutionarily conserved RBP-J-binding domains (10,19); the transmembrane-proximal RAM domain and the CDC10/ankyrin (ANK) repeats flanked by two nuclear localization signals (NLSs) (see Fig. 1). While the RAM domain binds to RBP-J strongly (19 -21), the ANK repeats bind to it weakly (18,22). Hence, IC lacking the RAM domain can activate transcription of the HES genes but less strongly than RAMIC (13,18). In the C-terminal region of RAMIC, there are the glutamine-rich OPA sequence and the PEST sequence, the functions of which are not clear yet. We recently identified a novel transcriptional activation domain (TAD) between the second NLS and the PEST sequence of mouse Notch1 (23). Although it was reported that Drosophila Notch RAMIC also contains a potent TAD in the region corresponding to the TAD of mouse Notch1 RAMIC (24), the amino acid sequences in the corresponding regions of the two species are not homologous.
RBP-J is known to associate with the viral protein Epstein-Barr virus (EBV) nuclear antigen 2 (EBNA2) which is essential for the immortalization of human primary B lymphocytes (10). EBNA2 has an acidic TAD similar to the TAD of the virus protein VP16 (25) and binds to RBP-J to activate transcription from the EBV promoter through the EBNA2-responsive element (EBNA2RE) carrying canonical RBP-J-binding sites (26,27). Since the EBV promoter and the HES-1 promoter are stimulated by either RAMIC or EBNA2, respectively (18,28), it is expected that the cellular (RAMIC) and viral (EBNA2) proteins activate transcription from a similar set of promoters by a common mechanism. Evidently, provision of the TAD is at least one mechanism of the RBP-J-mediated transactivation. In addition, several lines of evidence have suggested that 1 The abbreviations used are: Su(H), Suppressor of Hairless; bHLH, basic helix-loop-helix; RAM, RBP-J-associating molecule; ANK, ankyrin; NLS, nuclear localization signal; TAD, transcriptional activation domain; EBV, Epstein-Barr virus; EBNA2, EBV nuclear antigen 2; EBNA2RE, EBNA2-responsive element; HDAC, histone deacetylase; HAT, histone acetyltransferase; CBP, CREB-binding protein; PCAF, P300/CBP-associated factor; tk, thymidine kinase; CMV, cytomegalovirus; DBD, DNA-binding domain; IP, immunoprecipitation; PAGE, polyacrylamide gel electrophoresis; TSA, trichostatin A; GTF, general transcription factor; mAb, monoclonal antibody; EMSA, electrophoretic mobility shift assay. RBP-J is associated with a putative corepressor(s) and that RAMIC and EBNA2 displace the corepressor(s) from RBP-J to facilitate the transactivation (18,21,23,27,29).
It has been demonstrated that hypoacetylated histones accumulate within transcriptionally silenced genes, while hyperacetylated histones accumulate within transcriptionally active genes (30,31). In fact, many transcriptional repressors and corepressors have been shown to associate with histone deacetylase (HDAC) activity that generates a repressed chromatin structure with decreased accessibility to the transcription machinery (31,32). Recently, Kao et al. (33) demonstrated that RBP-J interacts with a corepressor complex including the SMRT/NCoR family of nuclear receptor corepressors and histone deacetylase 1 (HDAC1) (34). They showed that SMRT/ NCoR competed with TAN-1 (human Notch1 RAMIC) for RBP-J-binding and antagonized the RBP-J-mediated transactivation activity of TAN-1. It was also shown that the RBP-J/CBF1-interacting corepressor, CIR can interact with the similar HDAC complex including HDAC2 and SAP30 (35). On the other hand, a number of transcriptional activators have been shown to associate with histone acetyltransferases (HATs) such as the global coactivator P300/CBP (36) and the prototype HAT PCAF, which was originally isolated in human as a P300/ CBP-associated factor by virtue of its sequence similarity to yeast GCN5 (37).
Here we describe the functional interaction between mouse Notch1 RAMIC and two conserved HATs, mouse PCAF and GCN5 (38). We show that the interaction is conserved across species and RAMIC can recruit each of the HATs to RBP-J. Furthermore, we show that the RBP-J-mediated transactivation activity of RAMIC is repressed by two HAT inhibitor proteins, E1A and Twist, which are structurally unrelated to each other. Our results suggest that HATs including PCAF and GCN5 may be involved in the RBP-J-mediated transactivation by Notch RAMIC.
pCMX-E1A was constructed by subcloning the HindIII-BglII fragment excised from pBJ9⍀Ad5-E1A12S (41) into the HindIII/BamHI sites of pCMX. The coding region of mouse Twist was obtained by polymerase chain reaction amplification with specific primers, 5Ј-AT-GATGCAGGACGTGTCCAGCTCG-3Ј (forward) and 5Ј-TGCTAGTGG-GACGCGGACATGGAC-3Ј (reverse), using AKR mouse genomic DNA as template. The amplified fragment was cloned into pGEM-T vector-Easy and the sequence of both strands was confirmed. The EcoRI-FspI and FspI-SalI fragments were excised from the plasmid, and the two fragments were simultaneously ligated to the EcoRI/SalI sites of pFlag-CMV2, resulting in pFlag-CMV2-Twist.
Cell Culture-SV40 transformed monkey kidney COS7 cells, human embryonic kidney 293T cells, and murine fibroblast NIH3T3 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Murine myoblast C2C12 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 15% fetal bovine serum. The C2C12-derived cell line L9, which contains the pGa981-6 reporter construct with the blasticidin S resistance gene, was maintained in Dulbecco's modified Eagle's medium supplemented with 15% fetal bovine serum and 10 g/ml blasticidin S (Kaken Seiyaku).
Luciferase Assay-COS7, L9, and NIH3T3 cells were seeded on 6-well plates at 2 ϫ 10 5 cells per well 18 h before transfection and transiently transfected with various combinations of expression plasmid DNA indicated in each figure legend, using LipofectAMINE reagent (Life Technologies, Inc.). The total amount of plasmid DNAs was kept constant (1-2 g) by adding the same expression plasmids without inserts. Transfected cells were harvested 48 h after transfection and luciferase activities were measured according to the manufacturer's instructions (TOYO INK) in a Berthold luminometer, LumatLB9501. Ethanol or 0 -100 ng/ml trichostatin A (Sigma) was added to the culture medium at 24 h before harvesting. In all the experiments, the relative luciferase activity was normalized to the ␤-galactosidase activity obtained by transfection with 50 ng of pCMX-LacZ. Each experiment was repeated more than two times and the representative results are shown as the average of triplicate values with standard deviations.
Western Blot Analysis and Immunoprecipitation (IP)-For Western analysis, whole cell lysates of transfected NIH3T3 cells or nuclear extracts of transfected COS7 cells were mixed with SDS sample buffer, boiled for 5 min, and separated by SDS-PAGE. Separated proteins were transferred onto a nitrocellulose filter (Amersham Pharmacia Biotech) by Semi-Dry (Bio-Rad). After blocking with phosphate-buffered saline containing 2% skim milk, the filter was incubated with the primary antibody for 1 h at room temperature, washed with phosphate-buffered saline containing 0.05% Tween 20, and incubated with the horseradish peroxidase-conjugated secondary antibody for 1 h. Immunoreactive bands were visualized by enhanced chemiluminescence (ECL Plus detection system, Amersham Pharmacia Biotech). For immunoprecipitation, 293T cells seeded on 6-cm dishes at 5 ϫ 10 5 cells per plate 18 h prior to transfection were transiently transfected using the CellPhect transfection kit (Amersham Pharmacia Biotech). After 40 h of incubation, cells were washed, scraped in phosphate-buffered saline, and lysed in IP buffer (25 mM HEPES, pH 7.9, 150 mM KCl, 0.1% Nonidet P-40, 1 mM EDTA, 1 mM dithiothreitol, 10% glycerol, 0.1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml leupeptin, and 1 g/ml pepstatin A). After the cell lysates were rotated for 30 min at 4°C, the cell debris was removed by centrifugation and the supernatant was used as whole cell extract. The whole cell extracts were incubated with anti-Flag M2-affinity gel (Sigma) by rotating for 3 h at 4°C. Immuno-precipitates were washed three times with ice-cold IP buffer. Bound proteins were eluted in SDS sample buffer and subjected to Western blot analysis.
Electrophoretic Mobility Shift Assay (EMSA)-Preparation of nuclear extracts from COS7 cells and EMSA for GAL4 fusion proteins were performed as described previously (23). The binding reaction contained 20 mM HEPES (pH 7.9), 50 mM NaCl, 5 mM MgCl 2 , 10 M ZnSO 4 , 10% glycerol, 0.1 mg/ml bovine serum albumin, 1 g poly(dI-dC), 0.5 ng of 32 P-labeled oligonucleotide probe MH100, and 4 g of nuclear extracts. The reaction mixture in a final volume of 30 l was incubated for 30 min at 20°C and loaded onto a native 4% polyacrylamide gel in a buffer (pH 7.5) containing 6.7 mM Tris, 3.3 mM NaOAc, and 1 mM EDTA. After electrophoresis at 130 V for 2.5 h at 4°C, the gels were dried and analyzed using an Imaging Analyzer BAS1500 (Fuji Film).

TSA Derepresses Transcactivation Activity of the Fusion Protein between RBP-J and the Mouse Notch1 TAD-Previously
we reported that the fusion protein between RBP-J and the C-terminal portion of mouse Notch1 RAMIC (RBP-J-IC⌬ANK, Fig. 1) failed to activate transcription of the pGa981-6 luciferase reporter gene carrying 12 canonical RBP-J-binding sites (23), although the C terminus of mouse Notch1 activates transcription autonomously when fused to the yeast GAL4 DBD (Ref. 23 and see Fig. 5). We also showed that the repressed transactivation activity of the fusion protein was relieved by coexpression of either the RAM domain or the ANK repeats of RAMIC, suggesting that a putative corepressor(s) would associate with the RBP-J-IC⌬ANK fusion protein and might be displaced by the coexpressed RBP-J-binding domains. Since RBP-J interacts with HDAC corepressor complexes (33,35), we asked whether the fusion protein between RBP-J and the Cterminal portion of mouse Notch1 RAMIC is repressed by the HDAC activity.
The reporter plasmid pGa981-6 or pGa50-7 (without RBP-Jbinding sites) was transfected into COS7 cells with the plasmid encoding RBP-J or RBP-J-IC⌬ANK, and the cells were treated with increasing concentrations (0 -100 ng/ml) of a histone deacetylase-specific inhibitor, trichostatin A (TSA; Ref. 42). Transcription of pGa981-6 was activated by TSA in a dose-dependent manner (7-fold with 100 ng/ml TSA) when RBP-J-IC⌬ANK, but not when the non-chimeric RBP-J, was transfected ( Fig. 2A). This activation was dependent on the presence of RBP-J-binding sites as well as on the fused C terminus of RAMIC, arguing against the general effect of TSA on transcription. Next we investigated whether the effect of TSA was restricted to episomal DNA or it was also applicable to chromosomally integrated DNA. As shown in Fig. 2B, transcription was markedly activated by TSA (45-fold with 100 ng/ml TSA) when RBP-J-IC⌬ANK was transiently transfected into L9, a line of mouse myoblast C2C12 cells that was established by stable transfection with the pGa981-6 reporter. When RBP-J alone was transfected into L9, weak transactivation was observed by high concentration of TSA (2.7-fold with 100 ng/ml). Thus, inhibition of HDAC activity results in derepression of the transactivation activity of the fusion protein, either with the transient reporter or with the stable one. These results are consistent with the notion that Notch RAMIC activates RBP-J-mediated transcription by competing with the HDAC corepressor complex for RBP-J-binding (33), and by providing the TAD for RBP-J (23).
Mouse Notch1 RAMIC Interacts with Histone Acetyltransferases PCAF and GCN5 in Vivo-Several studies have shown that HDACs and HATs play opposite roles in regulating transcription of the same genes and it has been postulated that acetylation of the N-terminal histone lysine residues alters nucleosomal structure and increases the accessibility of specific promoters to the transcription machinery (30 -32). Therefore, we explored the possibility that mouse Notch1 RAMIC associates with HATs to facilitate the RBP-J-mediated transactivation. Although IC lacks the RAM domain (Fig. 1), it can bind to RBP-J via the ANK repeats and activate transcription through the RBP-J-binding sites, weakly but significantly (18,23). To test the association of mouse Notch1 IC with HATs, we chose two conserved HATs, mouse PCAF and GCN5 (38). The originally isolated human GCN5 (hGCN5-S) lacks the sequence similar to the N-terminal region of human PCAF (37), whereas mouse PCAF and GCN5 as well as the recently identified other human GCN5 (hGCN5-L) are homologous to human PCAF along their length (Refs. 38 and 43, and see Fig. 6A). Expression plasmids encoding six-Myc-tagged mouse Notch1 IC (amino acids 1810 -2531, Fig. 1) and Flag-tagged mouse PCAF or GCN5 were transiently introduced into 293T cells. The cell extracts were subjected to immunoprecipitation using anti-Flag antibody and probed in Western blot analysis with antibodies against Myc and Flag. Myc-IC was coprecipitated with Flag-PCAF, Flag-GCN5, or the control Flag-RBP-J, but no such coprecipitation was observed for extracts with the empty vector (Fig. 3A, lanes 4 -7). Similar amounts of Flag-tagged proteins were precipitated in the presence or absence of Myc-IC, and almost equivalent amounts of Myc-IC were expressed in cells transfected with each Flag-tagged construct. The results show that mouse Notch1 IC interacts with the two conserved HATs, mouse PCAF and GCN5, in vivo.
To address the functional relevance of the interaction between IC and the HATs, mammalian two-hybrid assays were performed using luciferase reporter constructs carrying the reiterated GAL4-binding sites. A fusion construct between GAL4 DBD-(1-147) and mouse PCAF or GCN5 (GAL4-PCAF or GAL4-GCN5) and a fusion construct between the TAD of VP16 and mouse Notch1 IC (VP16-IC) were transiently co-transfected into NIH3T3 cells. Transcription of the TK-MH:100ϫ4 reporter construct under the control of thymidine kinase promoter of herpes simplex virus was not activated by GAL4-PCAF per se (Fig. 3B, lane 3 versus lane 1), consistent with the previous reports (44 -46). Transcription of the TK-MH:100ϫ4 reporter construct was slightly (2-3-fold) activated by GAL4-GCN5 (lane 5 versus lane 1). However, transfection of VP16-IC with GAL4-PCAF and with GAL4-GCN5 enhanced the transcriptional activity 7-and 12-fold, respectively, in comparison with VP16 alone (lanes 3, 4, 5, and 6). Transcriptional activity by transfection of VP16-IC was enhanced 25-fold with GAL4-PCAF and 40-fold with GAL4-GCN5 using the pGL3-G5B reporter construct that contains the minimal TATA promoter of adenovirus E1b (lanes 9, 10, 11, and 12), and this was possibly due to its low level of basal transcription activity. When VP16-IC was transfected with the GAL4 DBD alone, no enhancement of the TK-MH:100ϫ4 reporter (lanes 1 and 2) and marginal enhancement of the pGL3-G5B reporter (lanes 7 and 8) were observed. We conclude that physical interaction of mouse Notch1 IC with mouse PCAF and GCN5 in vivo causes transcriptional activation of the GAL4-dependent luciferase reporter plasmids.
To examine whether the RAM domain has a role in the interaction of RAMIC with the two HATs, we compared VP16-RAMIC and VP16-IC for their transactivation activities of the pGL3-G5B reporter when co-transfected with GAL4-PCAF (Fig. 4A, lanes 2 and 3) or with GAL4-GCN5 (Fig. 4B). The results indicate that the activities of the two constructs were indistinguishable, suggesting that the RAM domain would not be required for the interaction of RAMIC with PCAF and GCN5. To ensure the integrity of VP16 fusion constructs, we tested for their ability to activate transcription of the pGL3-G5B reporter by co-transfection with GAL4-RBP-J. Only fusion constructs that contain the RAM domain activated the transcription (Fig. 4B), in agreement with the earlier reports showing that the RAM domain is a primary binding domain to RBP-J (19 -21). Although a weak but significant interaction Interaction Domains of RAMIC with PCAF and GCN5-Using the deletion constructs of VP16-RAMIC, we examined which domains of RAMIC are involved in the interaction with PCAF and GCN5. We first analyzed the involvement of the ANK repeats. VP16-IC activated transcription of the pGa981-6 reporter or the ptk-HES1 reporter, which contains two canonical RBP-J-binding sites, and its transactivation activity is abrogated by the M1 mutation (18) or deletion of the ANK repeats (data not shown). When they were transfected with GAL4-PCAF, VP16-IC (M1) exhibited a 4-fold decreased transactivation activity (Fig. 4A, lane 4 versus lane 3), and VP16-IC⌬ANKN and VP16-IC⌬ANK exhibited no transactivation activity (lanes 5 and 6). When transfected with GAL4-GCN5, VP16-IC (M1) exhibited a 10-fold decreased transactivation activity, and VP16-IC⌬ANKN and VP16-IC⌬ANK exhibited no transactivation activity (Fig. 4B). Protein expression of the VP16-RAMIC deletion constructs in transfected NIH3T3 cells was confirmed by Western blot analysis using anti-VP16 antibody (Fig. 4C). These results indicate that the ANK repeats are required, although not sufficient (Fig. 4A, lanes 13 and 14, and We then asked whether the C-terminal portion of IC is involved in its interaction with PCAF and GCN5. VP16-IC⌬PEST exhibited an enhanced transactivation activity in the presence of GAL4-PCAF (Fig. 4A, lane 8 versus lane 3), suggesting that the deleted C-terminal part (2399 -2531) has an inhibitory role in the interaction. Alternatively, the truncated protein may be more stable in vivo because it lacks the PEST sequence involved in high rates of protein turnover (47). A further deletion into the C-terminal half of the TAD (VP16-IC⌬OPA/PEST) dramatically reduced transactivation activity (lane 10 versus lane 8) and a deletion removing the whole TAD (VP16-RAMIC⌬C and VP16-IC⌬C) abrogated it completely (lanes 11 and 12). When internal deletion constructs between the ANK repeats and the TAD were transfected with GAL4-PCAF, lower but significant transactivation activities were exhibited relative to their non-deleted forms (lanes 7, 9 versus lanes 3,8), suggesting that this deleted region (2098 -2193) may function as a linker between the two functional domains, the ANK repeats and the TAD. The transactivation profile with GAL4-GCN5 is similar to that with GAL4-PCAF (Fig. 4B). The TAD is also important, but not sufficient (Fig. 4A, lane 15 and Fig. 4B), for the interaction with PCAF and GCN5.
To confirm the functional interaction between IC and the two HATs, we carried out a complementary experiment using fu- sion constructs of GAL4-ICs (Fig. 5A) and VP16-PCAF. DNA binding activity of the GAL4-ICЈ fusion proteins expressed in COS7 cells were examined by EMSA using the nuclear extracts and 32 P-labeled oligonucleotide probe containing a GAL4-binding site (Fig. 5B). As we reported previously (23), ANK-containing constructs showed smears and aggregates in the wells, probably because of homophilic interaction between the ANK repeats. As shown in Fig. 5C, GAL4-ICЈ and GAL4-TAD comparably activated transcription of the TK-MH:100ϫ4 reporter (lanes 3 and 6), since they contain the TAD. Co-transfection of VP16-PCAF enhanced the transactivation activity of GAL4-ICЈ (lanes 3-5), but not that of GAL4-TAD (lanes 6-8). The transactivation activity of GAL4-ICЈ⌬N (2194 -2531) (23) was not enhanced by VP16 (data not shown). GAL4-ICЈ⌬C per se has no transactivation activity (Fig. 5D), because it lacks the TAD (Fig. 5A). Transcriptional activity was not increased by cotransfection of VP16-PCAF (Fig. 5D, lanes 7 and 8), while it was increased by co-transfection of VP16-RBP-J (lanes 9 and 10). Transcriptional activity of GAL4-ICЈ⌬C (M1) was not increased by either VP16-PCAF or VP16-RBP-J (lanes 13-16). These data are consistent with the above observation that the ANK repeats and the TAD are both required for the interaction of IC with the two HATs.
Involvement of PCAF and GCN5 in RBP-J-mediated Transactivation of RAMIC-The functional interaction of Notch RAMIC with PCAF and GCN5 prompted us to examine whether RAMIC can recruit the HATs to RBP-J. First, we tried to detect a ternary complex composed of RBP-J, RAMIC, and PCAF. Myc-PCAF was coimmunoprecipitated with Flag-RBP-J in the presence, but not in the absence, of co-transfected VP16-mNotch1 RAMIC (Fig. 7A, lanes 4 and 5). Western blot analysis of whole cell extracts showed that comparable levels of the proteins were expressed in each transfected cells. To verify the recruitment of the HATs to RBP-J by RAMIC, we carried out a mammalian tri-hybrid assay. As previously reported (21,39), transcription of the TK-MH:100ϫ4 reporter was activated by co-transfection with GAL4-RBP-J and RAMIC (Fig. 7B, lane 2), and the addition of VP16-PCAF or VP16-GCN5 resulted in augmentation of the transactivation activity (lanes 3-6). Cotransfection with GAL4-RBP-J and RAMIC⌬C lacking the TAD (Fig. 1) failed to activate transcription (Fig. 7B, lane 7), and the addition of VP16-PCAF or VP16-GCN5 had a negligible effect on transcriptional activity (lanes [8][9][10][11]. The TAD of RAMIC seems to be important for the interaction with the two HATs, also when RAMIC is tethered to RBP-J. It has been recently demonstrated that the structurally unrelated viral oncoprotein E1A and the cellular bHLH protein Twist bind to the HAT domain of PCAF (41,49,50). In addition, both E1A and Twist inhibit the HAT activity of PCAF and repress its function as a transcriptional coactivator (41,49). To analyze the involvement of PCAF and GCN5 in the RBP-Jmediated transactivation by RAMIC, we investigated the effect of the two HAT inhibitor proteins. As shown in Fig. 8A, E1A repressed transactivation of the ptk-HES1 reporter by RAMIC in a dose-dependent manner. Expression of another HAT inhibitor protein, Twist also resulted in a dose-dependent repression of the transactivation activity (Fig. 8C), although the repression was weaker than that by E1A. The expression level of RAMIC was not affected by coexpression of E1A (Fig. 8B) or Twist (Fig. 8D). These repressive effects of the two HAT inhibitor proteins suggest that the HAT activity is important for the RBP-J-mediated transactivation by RAMIC. Since E1A and Twist were also shown to inhibit the HAT activity of P300/CBP (49,50), it is possible that several HATs including PCAF and GCN5 may be involved in the RBP-J-mediated transactivation by Notch RAMIC. DISCUSSION We have demonstrated here that the mouse Notch1 intracellular region (RAMIC) interacts with the two HATs, mouse PCAF and GCN5 in vivo. Extensive analyses using the mammalian two-hybrid assays have shown that the ANK repeats and the TAD of RAMIC, both of which play important roles in the RBP-J-mediated transactivation (18,23), are also involved in the interaction with PCAF and GCN5. Furthermore, we have shown that RAMIC can recruit PCAF and GCN5 to RBP-J, and that the transactivation activity of RAMIC through the RBP-J-binding sites was repressed by two structurally unrelated HAT inhibitor proteins, E1A and Twist. Taken together, our data suggest the important role of HATs including PCAF and GCN5 in the RBP-J-mediated transactivation by Notch RAMIC.
Drosophila Notch RAMIC (IC) interacted with mouse PCAF and GCN5 in mammalian cells, suggesting that the interaction is evolutionarily conserved. The ANK repeats are well conserved between mouse Notch1 and Drosophila Notch (the ho-mology between the two species is 66%), but the sequence homologous to the mouse Notch1 TAD is not found in Drosophila Notch although it is reported that Drosophila Notch contains a potent TAD, C-terminal to the ANK repeats (24). The replacement of Notch RAMIC for the HAT interaction between the distantly related species implies the functional importance of the interaction in Notch signaling. We have also shown that the N-terminal regions of mouse PCAF and GCN5 are involved in the interaction with mouse and Drosophila Notch RAMICs. The less conserved N-terminal regions of the two HATs have not been characterized well with the exception of their interaction with P300/CBP (37,38), whereas the C-terminal regions were shown to be involved in the interaction with several transcription factors, namely, ADA2 (51), RXR/RAR (52), NF-Y (53), interferon regulatory factor (IRF; 54), and HTLV-1 Tax (55). To our knowledge, Notch RAMIC is the second protein that has been experimentally shown to interact with the Nterminal regions of PCAF and GCN5. Our finding would contribute to the elucidation of the hitherto unidentified function of the N-terminal regions of the two HATs. Since no remarkable difference was found between PCAF and GCN5 in their interaction with RAMIC, it remains to be investigated whether the roles of the two HATs in the Notch signaling are redundant or distinct.
It has been shown that RAMIC plays two important roles in the RBP-J-mediated transactivation: displacement of the HDAC corepressor complex (33) and recruitment of the TAD (23). Derepressive effect of TSA on the fusion protein of RBP-J with the mouse Notch1 TAD clearly indicates that RBP-J is usually repressed by the HDAC activity. Marked transactivation of the transient and stable reporter genes was dependent on the C-terminal portion of RAMIC fused to RBP-J, reinforcing the important role of the TAD in RBP-J-mediated transactivation. Kao et al. (33), however, showed that expression of the Xenopus Notch/Su(H) target gene ESR1 in animal caps of embryo was weakly derepressed by TSA treatment in the presence and absence of injected Xenopus Delta, an inducer of the Notch signal. Their observation seems contradictory to our results, but the weak derepression of our stably integrated reporter gene was actually observed by treatment with a high dose of TSA, even though the non-chimeric RBP-J was used for transfection. Because it is generally accepted that genes embedded in chromosome are more repressed than genes located in episomal DNA, the chromosomally integrated genes may be more sensitive to TSA. It is reported that derepressive effects of TSA on some genes are restricted to chromosomal DNA (56,57).
In addition to the displacement of the HDAC corepressor complex and the provision of the TAD, we have demonstrated here that RAMIC can recruit PCAF and GCN5 to RBP-J. Different but partially overlapping domains of RAMIC are involved in each of these events. The RAM domain (18,21,23) and the ANK repeats (23) are responsible for displacing the corepressor complex from RBP-J, and the C-terminal TAD should associate directly or indirectly with general transcription factors (GTFs) for initiation of transcription (23). One can envisage that PCAF or GCN5 connects RAMIC with GTFs because PCAF and GCN5 were found in similar, large multisubunit protein complexes including several of the TATA-binding protein-associated factors (58 -60). PCAF also interacts with a phosphorylated, elongation-competent form of RNA polymerase II (61). However, neither GAL4-PCAF nor GAL4-GCN5 activates transcription efficiently by itself in mammalian cells (Refs. 44 -46 and Fig. 3B), and the ANK repeats and the TAD are both required for the interaction of RAMIC with PCAF and GCN5 (Figs. 4 and 5). For these reasons, it seems unlikely that the interaction between the RAMICs TAD and GTFs is mediated by PCAF or GCN5. Although we cannot rule out the possibility that PCAF and GCN5 are partly involved in the interaction between RAMIC and GTFs, they would mainly contribute to the RBP-J-mediated transactivation by acetylating histones to alter the chromatin structure. It is currently believed that the exceedingly small amounts of RAMIC produced by the ligand interaction can transduce the Notch signal (8,9). PCAF or GCN5 recruited to the promoter of the target genes would catalyze rapid acetylation of the histones or probably other proteins, and then RAMIC should associate with GTFs for intiation of transcription. In this way, a weak transactivation activity of the small amounts of RAMIC could be amplified.
The repressive effect of E1A and Twist on the transactivation activity of RAMIC through the RBP-J-binding sites (Fig. 8) suggests the positive role of HATs. PCAF interacts with the nuclear receptor coactivators SRC-1/NCoA (44,62) and ACTR/ PCIP (62,63) in addition to P300/CBP, all of which contain the HAT activity. PCAF or GCN5 and its interacting HATs may form a multi-HAT complex and coordinately regulate activation of the RBP-J-mediated transcription by RAMIC. Therefore, it is possible that E1A and Twist repress the transactivation activity of RAMIC by inhibiting the HAT activity of other HAT proteins like P300/CBP (48,49), or by disrupting the functional multi-HAT complex (37,64). It is important to analyze the involvement of other HATs in the RBP-J-mediated transactivation by RAMIC.
It was recently shown that transcription from the EBV promoter is repressed by EBNA3C, another EBV protein which links RBP-J with HDAC1 (65). On the other hand, the acidic TAD of EBNA2 interacts with P300/CBP and PCAF (66,67). The transactivation activity of EBNA2 through the myc promoter, although it is independent of RBP-J, is enhanced by coexpression of CBP or PCAF (66). P300 coactivates expression of the EBV-encoded latent membrane protein 1 mediated by EBNA2, and PCAF potentiates the coactivation (67), indicating that EBNA2 associates with HATs to facilitate transcriptional activation of the EBV promoter. Thus the recruitment of HATs seems to be a common mechanism for the RBP-J-mediated transactivation by RAMIC and EBNA2. Further studies will clarify the more detailed mechanism for the RBP-J-mediated transactivation by RAMIC and the physiological roles of HATs including PCAF and GCN5 in the Notch signaling.