Cell transfection and luciferase assay.

Transfections of 70% confluent C3H10T1/2, HeLa, COS7, and C2C12 cells in 60-mm-diameter dishes were performed according to the 2-[bis(2-hydroxyethyl)amino]ethanesulfonic acid (BES)-buffered saline protocol as previously described (20). Two micrograms of reporter constructs was cotransfected with the indicated amounts of expression vectors for HERP or HES1. The total amount of plasmid DNA was adjusted to 9 μg with the control plasmid DNA lacking the cDNA. Transactivation of reporter genes was evaluated by luciferase assay as described before (44). These assays were done in triplicate and repeated several times.

Immunoprecipitations and Western blots.

For in vivo protein interaction studies, 70% confluent 293T cells in 10-cm-diameter dishes were transfected by the BES-buffered saline method as described above with 15 μg of pCEP4 Flag-N-CoR plus 5 μg of pcDNA3.1(−) HA-HERP1. For the mSin3A interaction studies, the cells were transfected with 5 μg of pcDNA3.1(−)-Flag-HERP1 plus 15 μg of the myc-mSin3A expression vector. For HDAC1 interaction studies, the cells were transfected with 10 μg of pcDNA3 HDAC1-cFlag plus 10 μg of pcDNA3.1(−) HA-HERP1. For endogenous HDAC1 interaction studies, the cells were transfected with 20 μg of either pcDNA3.1(−)-Flag empty, pcDNA3.1(−)-Flag-HERP1, or pcDNA3.1(−)-Flag-HERP2. For HES1 interaction studies, the cells were transfected with 10 μg of pcDNA3.1(−)-Flag-HERP1 plus 10 μg of pc3Gal-HES1. The cells were lysed in a lysis buffer (20 mM Tris-HCl [pH 8.0], 5 mM MgCl₂, 10% glycerol, 0.1% NP-40, 100 mM KCl) supplemented with freshly prepared protease inhibitors. The cell extracts were incubated with either anti-Flag M2 (Sigma) or control normal mouse immunoglobulin G (IgG) coupled to protein G agarose (Sigma) for 1 h at 4°C. After washing three times, bound proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) followed by Western blot analysis with anti-myc antibody (9e10, ATCC CRL1729), anti-HA antibody (Y-11; Santa Cruz Biotech), anti-HDAC1 antibody (06-720; Upstate Biotechnology), or anti-GAL4 DNA-binding domain (GAL4-DBD) antibody (sc-577; Santa Cruz Biotech).

In vitro interaction assay.

GST pull-down experiments were carried out as described previously (20). Various deletion mutants of mouse GST-HERP1 fusion proteins were prepared from Escherichia coli as described previously (20). In vitro-translated ³⁵S-labeled proteins were prepared using the TNT coupled transcription-translation system (Promega). Labeled proteins were incubated with equal amounts of GST fusion protein for 1 h at 4°C. Bound proteins were analyzed by autoradiography after SDS-PAGE.

Electrophoretic mobility shift assay.

In vitro-translated proteins of HERP1 and HES1 were prepared using the TNT coupled transcription-translation system. Nuclear proteins were extracted as described before with minor modifications as follows (2). 293T cells were transfected with plasmids described in the figure legends. Three days after transfection, cells were harvested with ice-cold lysis buffer (20 mM HEPES [pH 7.6], 20% glycerol, 1.5 mM MgCl₂, 0.2 mM EDTA, 0.1% Triton X, 10 mM KCl, 1 mM dithiothreitol, 1 μg of leupeptin per ml, 1 μg of pepstatin per ml, 1 μg of aprotinin per ml, 1 μM phenylmethylsulfonyl fluoride) and homogenized with 10 strokes of a Dounce homogenizer. The cells were centrifuged for 10 min at 900 × g at 4°C. The pellet was resuspended in the same volume of lysis buffer augmented to 500 mM KCl and transferred to a microcentrifuge tube. The lysate was incubated at 4°C for 1 h and then centrifuged for 15 min at 16,000 × g at 4°C. Supernatants were collected and frozen at −80°C.

Gel mobility shift assays were carried out as described previously (24, 41, 45). The annealed oligonucleotides were labeled at both ends by filling in with Klenow enzyme in the presence of [α-³²P]dCTP. Protein-DNA complexes were formed by incubation of proteins described above with 25 fmol of radiolabeled nucleotides in 40 μl of buffer (25 mM HEPES [pH 7.5], 100 mM KCl, 20% glycerol, 0.1% Nonidet P-40, 10 μM ZnSO₄, 1 mM dithiothreitol). Poly(dI-dC) was included as a nonspecific competitor at 5 ng/μl. After incubation with probes at room temperature for 20 min, DNA-protein complexes were resolved by electrophoresis on a 5% acrylamide gel. The dried gel was exposed to X-ray film at −80°C as well as being subjected to PhosphorImager analysis (Storm 840; Molecular Dynamics).

HERP repression activity resides in the bHLH domain.

Having confirmed the intrinsic repressor activity of HERP, we expected that repression activity would reside in the modified, but conserved, tetrapeptide motif in the carboxyl terminus of HERP (YQPW or YRPW), as it does in the WRPW motif of HES. After all, Runt domain proteins carry a variant (WRPY) of the tetrapeptide motif (WRPW) and yet are active in the recruitment of TLE/Groucho and transcriptional repression (16). Therefore, one might anticipate that the YQPW motif of HERP also would have a repression activity. Accordingly, a series of HERP1 deletion mutants were expressed in GAL fusion proteins and tested. Unexpectedly, deletion of YQPW had little effect on HERP's repression activity (Fig. ​(Fig.2C,2C, ΔY, NBO), and the C-terminal region containing YQPW (CY) produced only weak repression. Similarly, the Orange domain, previously described as a putative repression domain (12), also had little effect (O). To our surprise, the bHLH domain alone retained repression activity (B) fully comparable to that of wild-type HERP1. Similar results were obtained using segments of HERP2 (data not shown). Thus, the repression activity of HERP resides primarily in the bHLH domain, while the YQPW motif plays a relatively minor role, if any. This finding is in sharp distinction to the function of the bHLH domain of hairy, a Drosophila homologue of HES that has no intrinsic repression activity (17). These unexpected results prompted us to directly compare repression domains of HERP and HES within a single experiment (Fig. ​(Fig.2D).2D). In contrast to HERP, the WRPW-containing carboxyl-terminal region of HES1 (HES1-CW) has full repression activity, whereas its bHLH domain had no effect (HES1-B). Together, these results indicate that the closely related proteins HES1 and HERP1 repress transcription using very distinct mechanisms.

HERP associates with mSin3 corepressor complex.

The relative dispensability of the tetrapeptide feature of HERP is in sharp contrast to the WRPW motif of HES that is required to recruit the TLE/Groucho corepressor. The distinct repression domains of HERP and HES imply that HERP1 is likely to engage different corepressors, and several were tested. Sin3 is a major corepressor that participates in a number of transcriptional repression activities in mammalian cells, Drosophila, and Saccharomyces cerevisiae (5, 27). When both mSin3A and HERP1 were simultaneously expressed in cells, we found a strong association between them (Fig. ​(Fig.3A,3A, lane 1). The association was specific, as it was not observed when HERP1 was absent (lane 2) or when control mouse IgG was used for mock immunoprecipitation (lane 3). These data suggest that HERP specifically associates with the mSin3A complex in cells. To further support this conclusion, we next studied whether HERP also associates with the corepressor N-CoR, which is known to associate with the mSin3 complex. Consistently, N-CoR also was found specifically associated with HERP1 in the cells (Fig. ​(Fig.3B).3B). Furthermore, we found a strong and specific interaction between HERP1 and HDAC1, a known subunit of the mSin3 complex (Fig. ​(Fig.3C).3C). An additional study further verified the association of an endogenous HDAC1 with both HERP1 and HERP2 (Fig. ​(Fig.3D).3D). These findings demonstrate that HERP recruits the mSin3 complex containing histone deacetylases, as well as another corepressor, N-CoR.

The bHLH domain of HERP directly associates with mSin3A and N-CoR.

Given that HERP associates with the mSin3 complex in cells, we further asked whether HERP directly contacts mSin3A, N-CoR, and HDAC1 in vitro. Incubation of full-length GST-HERP1 or GST alone with in vitro-translated mSin3A protein clearly showed a strong specific interaction between HERP1 and mSin3A (Fig. ​(Fig.3E,3E, lanes 2 and 7). A deletion analysis to determine the interaction domain of HERP1 revealed that the bHLH domain is responsible for this interaction (lanes 3, 4, and 9) and that neither the Orange domain nor the YQPW-containing C terminus is involved (lanes 5, 6, and 10). Interestingly, the same set of HERP1 regions that contains the bHLH domain also mediates the interaction with N-CoR (Fig. ​(Fig.3F,3F, lanes 13, 14, and 19). These results (as summarized in Fig. ​Fig.3G)3G) indicate that the bHLH domain directly engages both mSinA and N-CoR corepressors. HERP1 does not interact with HDAC1 in these in vitro interaction assays (data not shown), indicating that the association of HERP1 with HDAC1 observed in vivo (Fig. ​(Fig.3C3C and D) is indirect and is mediated by either N-CoR, mSin3A, or other subunits of the mSin3 complex. Similar indirect HDAC associations with transcription factors have been reported previously (5, 27). Importantly, these findings are in full agreement with the results of our repression domain mapping studies (Fig. ​(Fig.2C2C and D) and identify the bHLH domain of HERP1 as a docking site for an HDAC-containing mSin3 corepressor complex to mediate transcriptional repression by HERP1.

HES and HERP bind both distinct and common DNA sequences.

In certain cell types, HERP and HES are individually expressed, whereas in other cells they are coexpressed. Having established that HERP is a transcriptional repressor, we questioned why there are two Notch effectors coexpressed within some cell types. Given the difference in amino acid sequences of their DNA-binding basic domains (Fig. ​(Fig.1B),1B), HES and HERP may bind distinct DNA sequences. To address this possibility, we studied DNA binding activities of HES and HERP using various bHLH-binding DNA sequences (class A, B, and C) (15, 24, 41) as probes in gel shift assays. Although HERP1 binds to all the tested probes (Fig. ​(Fig.4A,4A, lanes 7 to 12), and HES1 also binds most probes (lanes 2 to 6) except the class A probe (lane 1), HES1 and HERP1 show distinct preferences for different DNA sequences. For instance, HES1 binds to class B and C-1 probes at equal efficiency (lanes 2 and 3), but only weakly to the other class C probes (lanes 4 to 6). Although HERP1 and HES1 bind similarly to the class B probe (lanes 2 and 8), the binding of HERP1 to C-1 and C-2 probes was much weaker than that of HES1 (compare lanes 3 and 4 and lanes 9 and 10). Unlike HES1, HERP1 bound the class A probe (lane 7), raising the possibility that HERP1 could directly compete with tissue-specific transcriptional activators for class A sequences. HERP2 showed DNA binding preferences nearly identical to that of HERP1 (data not shown). Thus, as summarized in Fig. ​Fig.4A4A (bottom), while HES and HERP can bind to most of the same DNA sequences, they do so with clearly different preferences.

Open in a separate window

Open in a separate window

Open in a separate window

FIG. 4

DNA binding properties of HES and HERP. (A) HES1 and HERP1 show distinct DNA binding preferences. Electrophoretic mobility shift assays were performed using various DNA probes. Autoradiographs after short and long exposures are shown. Top-strand oligonucleotide sequences used in this assay are indicated (bottom panel). Consensus binding elements are underlined. Sequences of classes A and B are derived from those described previously (24). The C-1, C-2, C-3, and C-4 oligonucleotide sequences are from the regulatory region of the known HES/E(spl) target genes; achaete (41), E(spl) genes (45), hASH (9), and HES1 itself (46), respectively. Radioactivities of each band were measured by a PhosphorImager with ImageQuant software to assess the relative binding activities shown in the bottom panel. The radiolabeling efficiency of the probes was comparable among the different probes, and excess amounts of the free probes were confirmed in all lanes in a parallel experiment with the same results (data not shown). (B) HES1 and HERP1 form a heterodimer. HES1 and HERP1 proteins were either individually translated (lanes 2 to 7, 12 to 17, 22 to 27, 32 to 37, and 42 to 47) or cotranslated (lanes 8 to 10, 18 to 20, 28 to 30, 38 to 40, and 48 to 50). In vitro-translated proteins prepared in parallel using [³⁵S]methionine are shown to verify equal protein product (lanes 42 to 50). (C) A gel shift assay was performed using either in vitro-translated proteins (lanes 1 to 5) or nuclear extract from the cells transfected with the expression vectors for HERP1 and HES1 (lanes 6 to 17). These proteins were incubated with the DNA probe C-1 at room temperature for 20 min, and the indicated antibodies (lanes 4, 5, 7, 9, 11 to 13, and 15 to 17) were added to the mixture, followed by additional incubation for 30 min on ice. DNA-protein complexes were resolved by electrophoresis on a 5% acrylamide gel. Note that nuclear extracts from the cells expressing both HES1 and HERP1 showed a dramatic increase of DNA binding activity. Symbols: thin arrow, nonspecific band; thin open arrow, supershifted homodimer band; thick open arrow, heterodimer band; thick solid arrow, supershifted heterodimer band. The amount of HES1 protein in the nuclear extract from the cotransfected cells (lanes 10 and 14) was half of that from singly transfected cells (lane 6), as confirmed by Western blot analysis, and the same was true for HERP1 (data not shown).

Formation of the HES-HERP heterodimer both in vitro and in vivo.

The finding that HES and HERP can also bind certain DNA sequences with the same efficiency (e.g., Fig. ​Fig.4A,4A, lanes 3 and 8), together with their coexpression within single cells, raises the question of whether HES and HERP compete for the common DNA binding site or otherwise interact. When HES1 or HERP1 is incubated independently with the class B probe in an electrophoretic mobility shift assay, each shows a single specific band but with distinct mobility (Fig. ​(Fig.4B,4B, lanes 2 to 4 and 5 to 7). Surprisingly, when the two are simultaneously incubated with the probe, a single new main band with intermediate mobility appears. Two important features are noted regarding this intermediate band. First, the intermediate band is generated at the expense of the respective homodimers of HES1 and HERP1, as the homodimer bands are no longer present (lanes 8 to 10). Second, the intermediate band has at least a two- to threefold-higher DNA binding activity than the sum of those of the two homodimers (compare, e.g. lanes 3, 6, and 9). Essentially identical data were obtained using the class C-1 and C-2 probes (lanes 12 to 20 and 22 to 30). These data suggest that the intermediate band represents a HES-HERP heterodimer and that the heterodimer is strongly preferred to the homodimers across a variety of different DNA docking sequences. As shown in lanes 41 to 50, the protein concentrations of HES1 and HERP1 remained the same when the two were coexpressed (compare lanes 42 to 47 and lanes 48 to 50). Thus, they represent higher binding activities of the heterodimer for the DNA rather than altered protein concentrations. No such heterodimer was observed with the class A probe (lanes 32 to 40) (which only HERP binds), further supporting the notions that the band with intermediate mobility represents a HES-HERP heterodimer and that stable heterodimer formation requires the two basic domains of HES and HERP both to contact the DNA. That the intermediate band indeed contains both HES1 and HERP1 was directly demonstrated using specific antibodies against these proteins (Fig. ​(Fig.4C,4C, lanes 4 and 5; see below).

But is the heterodimer formed in vivo? Nuclear extracts from cells expressing only HES1 or only HERP1 did not have detectable specific DNA binding activities (Fig. ​(Fig.4C,4C, lanes 6 and 8), suggesting low binding affinities of the homodimers. Although addition of HES1-specific antibody caused little change (lane 7), addition of HERP1-specific antibody revealed a supershifted band (lane 9). Most importantly, nuclear extract from cells expressing both HES1 and HERP1 proteins showed a marked increase in DNA binding activity (lane 10). Addition of either HES1- or HERP1-specific antibody (lanes 11 and 12), but not a control IgG (lane 13), supershifted the band, indicating that the band contains both HES1 and HERP1. Similar supershifted bands were also observed when in vitro-translated proteins were incubated with the antibodies (lanes 4 and 5). Furthermore, simultaneous addition of both antibodies caused an additional supershift (lane 17), suggesting that the supershifted bands engendered by the single antibodies (lanes 11 and 12 or 15 and 16) contain both HES1 and HERP1. We observed essentially identical results using HERP2 (data not shown). These data demonstrate that HES and HERP form a heterodimer both in vitro and in vivo and that the heterodimer has a striking DNA binding activity and is the exclusive entity that forms in cells expressing both proteins.

HERP associates with HES in solution.

In order to form a DNA-bound heterodimer, HERP might associate with HES before DNA binding. We studied this possibility first in vitro by a GST pull-down approach. In vitro-translated HES1 was efficiently associated with GST-HERP1 but not with GST alone (Fig. ​(Fig.5A,5A, lanes 1 and 3). A reciprocal study using in vitro-translated HERP1 and GST-HES1 confirmed a strong specific interaction between HES1 and HERP1 proteins. A similar observation has been made by others previously (30). Furthermore, we found that this association was also observed in vivo. Total cell extracts from the cells expressing tagged HERP1 and/or HES1 were subjected to immunoprecipitation followed by Western blot analysis. A specific in vivo interaction was observed between the two proteins (Fig. ​(Fig.5B,5B, lanes 1 to 3). The interaction is very strong, as a majority of HES1 protein was found associated with HERP1 in the immunoprecipitate (compare lanes 1 and 4). These data confirmed the presence of a tight association between HES and HERP in solution. Whether this interaction is more stabilized upon binding appropriate DNA sequences by the heterodimer remains to be determined (Fig. ​(Fig.4B).4B). Furthermore, whether such protein-protein interactions precede or are required for DNA binding of the heterodimer can only be answered by kinetic studies.

Open in a separate window

FIG. 5

HERP1 and HES1 associate in solution. (A) HERP1 directly associates with HES1 in the absence of DNA in vitro. In vitro protein-protein interaction assays were carried out as described for Fig. ​Fig.3E.3E. (B) HERP1 associates with HES1 in cells. 293T cells were transfected with pcDNA3.1(−)-Flag-HERP1 or pcDNA3.1(−) empty vector plus pc3Gal-HES1. The extracts from the transfected cells were incubated with an equal amount of anti-Flag M2 antibody or normal mouse IgG. Bound proteins were separated by SDS-PAGE followed by Western blot analysis with anti-GAL4-DBD antibody. Ip, immunoprecipitation; Ab, antibody. Numbers to the right of the gels are molecular masses in kilodaltons. interaction was studied as described above using anti-HA antibody. (D) Both Flag-HERP1 and -HERP2 associate with endogenous HDAC1. Cells were transfected with either pcDNA3.1(−)-Flag empty, pcDNA3.1(−)-Flag-HERP1, or pcDNA3.1(−)-Flag-HERP2. The interaction was studied as described above using anti-HDAC1 antibody. (E) HERP1 interacts with mSin3A via its bHLH domain. In vitro-translated ³⁵S-myc-mSin3A was incubated with equal amounts of the indicated GST-HERP1 fusion proteins for 1 h at 4°C. Bound proteins were analyzed by autoradiography after SDS-PAGE. (F) HERP1 interacts with N-CoR via its bHLH domain. Assays were performed as described above using ³⁵S-N-CoR. (G) Summary of the domain mapping studies. Ip, immunoprecipitation; Ab, antibody. Numbers to the right of the gels are molecular masses in kilodaltons.

HES and HERP: similar domains with different functions.

The present study has revealed that the repression activity of HERP resides primarily in the bHLH domain rather than the YQPW motif, suggesting that the tetrapeptide motif may be dispensable for the HERP family. This is in sharp contrast to the well-established essential roles of the WRPW motif in the HES/E(spl) family (17, 18, 43, 49). However, two exceptions are reported: the WRPW domain of the HES/E(spl) family does not appear to be required for suppression of neurogenesis in zebrafish or for suppression of SCUTE activity in the sex determination pathway in Drosophila (12, 47). Thus, the requirement for the WRPW motif of HES/E(spl) is not absolute. It is an interesting possibility that the YQPW motif of HERP might have a more significant role in repressing target gene expression in other contexts.

Another distinct feature of HERP is the presence of a glycine in its basic domain at a position that is invariably occupied by proline in the HES/E(spl) family. The strict conservation of glycine among all the HERP family members suggests an important role of the residue. Given that HERP had different DNA binding preferences than those of HES1, the proline-to-glycine substitution might contribute to defining different DNA binding preferences for the two proteins. In an analogous situation, however, a role for the proline in the DNA binding of HES1 is yet to be established, since a proline-to-asparagine mutation in an E(spl) protein largely diminished its DNA binding activity, whereas a proline-to-threonine mutation had little effect (48). Thus, further studies are needed to determine the contribution of the glycine in defining the DNA binding specificity of HERP.

Recruitment of an HDAC complex by HERP.

HERP recruits HDAC1 as a subunit of the mSin3 complex. Interestingly, Chen et al. have shown for Drosophila that Groucho can interact with Rpd3, an orthologue of mammalian HDAC (8). This raises the possibility that mammalian TLE might also recruit HDAC and, therefore, that HES/E(spl) and HERP might share a partly common repression mechanism that includes chromatin remodeling by HDACs. However, HES and HERP engender their own unique protein-protein interactions. For instance, Groucho is involved in additional interactions with histones H3 and H1 (8, 42). In addition, the mSin3 complex (which is recruited to HERP) also has HDAC-independent repression activity (29), likely mediated by multiple subunits of the complex as well as other associating proteins such as N-CoR. It has recently been shown that HES1 retains its repression activity in cells lacking N-CoR (25). Thus, HES and HERP employ different repression mechanisms involving heterologous sets of proteins. The mSin3 complex is recruited to a number of DNA binding transcription factors including the nuclear hormone receptors (1, 21, 36), PLZF (11), MeCP2 (26, 39), Ski (40), p53 (35), and Mad (6). Although Mad is a bHLH-Zip transcriptional repressor, it uses its N-terminal amphipathic α-helical region to recruit the mSin3 corepressor complex rather than its bHLH domain (6). To our knowledge, the present study is the first to show that a bHLH domain of a transcription factor can serve as an interface to recruit the mSin3 complex and its component HDACs.

HERP, a new heterodimer partner for HES/E(spl).

bHLH proteins form a dimer through their HLH domains that properly positions and orients the basic domains for specific DNA sequences, the E box (CANNTG) and its variants. A basic domain of each subunit of a dimer recognizes a half-site of the E box (7, 33). Thus, each of three dimers, HES-HES, HERP-HERP, and HES-HERP, should have its own DNA binding specificity. Consistent with this idea, the HERP homodimer showed a different DNA binding sequence profile than that of the HES homodimer (Fig. ​(Fig.4A).4A). The finding that HES and HERP homodimers bind common DNA sequences but with distinct preferences suggests that they may regulate both common and different target genes.

Regulation by homodimers may occur primarily in cells that express either HES or HERP, since in the presence of both proteins homodimers disappear and a HES1-HERP1 heterodimer becomes the exclusive species for most binding sites tested. The heterodimer showed a moderate increase in DNA binding activity in vitro (Fig. ​(Fig.4B),4B), whereas it showed a drastic increase in vivo (Fig. ​(Fig.4C).4C). This discrepancy might reflect posttranslational modification or additional cofactors present only in vivo. In any case, the higher DNA binding activity in vivo further supports the physiological relevance of the heterodimer. Heterodimer formation may be essential to bring a DNA binding activity above a threshold level for any physiologically meaningful repression, and thus, it may represent an on-and-off switch for Notch signaling rather than simply providing a linear increase of signal. Alternatively, it may represent an efficient mechanism to amplify the signal or to change target genes.

One critical issue regarding the heterodimer formation is whether Notch stimulation can simultaneously up-regulate expression of HES and HERP in a single cell type. We have indeed observed the coexpression of HES and HERP mRNAs after Notch stimulation in certain cell types (23). Interestingly, the degree of HERP mRNA induction is typically severalfold higher than that of HES. Thus, it is possible that HERP protein might be abundant enough to generate HERP homodimers after formation of the HES-HERP heterodimer, which could regulate a wider spectrum of genes including both the homodimer-specific and the heterodimer-specific target genes.

We are grateful to Robert Eisenman, Christopher Glass, Stuart Schreiber, Ryoichiro Kageyama, Ronald Evans, and Tetsuo Sudo for critical reagents. We thank T. Saluna for technical assistance and members of IGM for useful discussions. T.I. thanks Nobuko I. for her understanding and encouragement. Y.H. thanks M. D. Schneider and A. I. Schafer for support.

This work was done during the tenure of a research fellowship from the American Heart Association, Western States Affiliate (to T.I.). This work was supported in part by grants from the National Institutes of Health (to L.K.).