Regulation of the ERBB-2 Promoter by RBPJκ and NOTCH*

Within the human ERBB-2 gene promoter, a 100-base pair region 5′ to the TATA box enhances basal transcription 200-fold. Two palindromes present within this 100-base pair region are important for transcription. The palindrome binding protein was purified to homogeneity and found to be identical to RBPJκ, the mammalian homolog of Drosophila Suppressor of Hairless (Su(H)). Recombinant RBPJκ bound the ERBB-2 promoter with affinity comparable with that seen with well characterized RBPJκ binding sites. RBPJκ activated an ERBB-2palindrome-containing promoter in 293 cells. Because inDrosophila Su(H) acts downstream of NOTCH and because NOTCH·Su(H)/RBPJκ stimulates transcription from target promoters, NOTCH-IC, a constitutively active form of NOTCH, was tested for effects on the ERBB-2 palindrome. NOTCH-IC further increased RBPJκ-mediated transcription on wild type but not mutantERBB-2 palindrome. Thus, RBPJκ can activateERBB-2 transcription and serve as an anchor to mediate NOTCH function on the ERBB-2 gene.

The human ERBB-2 gene, located on chromosome 17q21, encodes a 1255-amino acid protein-tyrosine kinase receptor (1,2). This receptor is widely expressed at higher levels during fetal development than in the adult, where it is detected primarily in epithelial cells (3,4). Amplification and overexpression is frequent in adenocarcinomas, especially in those arising in the breast and ovary where overexpression directly correlates with poorer patient outcomes (5,6). Protein overexpression is frequently the result of gene amplification; however, tumors are reported to overexpress ERBB-2 mRNA and protein from single copy genes. Even with gene amplification, mRNA expression/gene is increased, indicating that transcriptional control mechanisms are important (6,7).
To deduce transcriptional control mechanisms, the human ERBB-2 gene promoter has been sequenced and found to contain typical CAAT and TATA elements and four Alu sequences within 3.65 kilobases of the proximal promoter (8 -10). A 100-bp region upstream of the TATA box increases promoter activity 200-fold. This 100-bp enhancer contains an Sp1 site near its 5Ј end and a CAAT box near its 3Ј end (11). It also contains two palindromic sequences that are conserved in the rat and mouse ERBB-2 promoters (12). Using reporter constructs, these palindromes were shown to exert both positive and negative regulation of the human ERBB-2 promoter (11).
We previously described a palindrome binding protein (PBP) 1 that bound to the half-site of each ERBB-2 palindrome with the core recognition sequence TGGGAG (13). We now report that protein sequence analysis of purified PBP identifies it as RBPJ (recombination signal binding protein of immunoglobulin J gene). RBPJ, which was initially cloned by Matsunami et al. (14) based on recognition of the J recombination signal sequence, was subsequently isolated based on recognition of the Epstein-Barr virus (EBV) C promoter (designated as CBF1) (15)(16)(17) and of the adenovirus pIX gene promoter (18). RBPJ is the mammalian homolog of Drosophila Suppressor of Hairless (Su(H)) (19,20). Further analysis indicated that although RBPJ contained a 40-amino acid region of homology to integrase, it lacked such activity (21) and recognized a composite sequence consisting of the heptamer recombination recognition site and a BamHI linker (15,22). RBPJ is widely expressed with two mouse and three human splicing variants identified (23).
Genetic analyses indicate that Su(H) acts downstream of NOTCH in the signaling pathway of sensory peripheral nervous system development in Drosophila (19,20). A similar pathway for lateral inhibition in neuronal development exists in mammalian species (24). The Drosophila Enhancer of split (E(spl)) and the mouse Hairy enhancer of split (Hes-1) complexes are downstream of NOTCH and Su(H)/RBPJ in this pathway (25,26). In the best studied mechanism, the intracellular domain (IC) of the transmembrane protein NOTCH is translocated to the nucleus with RBPJ that serves as the site-specific DNA binding partner (26,27). The NOTCH-IC⅐RBPJ complex activates transcription of basic helix loop proteins encoded by Hes-1. Events downstream of E(spl)/Hes-1 may be inhibitory (lateral inhibition in neuronal development (28) and inhibition of myogenesis (29)) or stimulatory. RBPJ also acts in a protein complex with the non-DNA-binding EBNA2 to regulate genes involved in EBV latency (15). The alteration of NOTCH that is necessary for nuclear translocation is reported to occur upon ligand binding (27) via proteolysis (30), via translocation that removes much of the ectodomain (31), and with retroviral insertion (32).
Because RBPJ could bind to the two palindromes in the ERBB-2 promoter, we examined its activity using reporter gene constructs. We present evidence that RBPJ stimulates promoter activity via wild type but not mutant ERBB-2 palindromic sequences; activity was markedly enhanced upon co-* This work was supported by National Institutes of Health Grant DK13149 and by a grant from the Markey Charitable Trust. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ expression of the intracellular domain of NOTCH but was not affected by co-expression of EBNA2. These results suggest that ERBB-2 promoter activity is regulated by a protein complex that contains RBPJ.

MATERIALS AND METHODS
Cell Cultures, Transfections, and Reporter Assays-293 cells were grown in Dulbecco's modified Eagle's medium/F-12 medium supplemented with 10% calf serum. F9 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum.
Plasmid DNA was transfected as a calcium phosphate precipitate (33). A ␤-galactosidase expression vector under control of a cytomegalovirus gene promoter was co-transfected with the luciferase reporter gene constructs.
Luciferase activity was measured as described by de Wet et al. (34). Aliquots of cell extract were added to an assay reaction containing 100 mM potassium phosphate (pH 7.8), 5 mM ATP, and 15 mM MgSO 4 in a volume of 0.35 ml. Reactions were initiated by the addition of 0.1 ml of 1 mM luciferin, and light readings were integrated over 10 s with a monolight 2110 luminometer. ␤-Galactosidase activity was measured as described by Norton and Coffin (35). Promoter activity was expressed as light units of luciferase activity/A 420 unit of ␤-galactosidase activity.
Plasmid Constructs-The intracellular region of human NOTCH (NOTCH-IC) was a generous gift from Dr. David Baltimore, MIT. Viral EBNA2 and human RBPJ cDNAs were generous gifts from Dr. Diane Hayward, The Johns Hopkins University. The NOTCH-IC and RBPJ cDNAs were regenerated by polymerase chain reaction and subcloned into a modified pCDNA3 vector containing the luciferase gene translation initiation codon and the hemagglutinin (HA) epitope tag.
Eight copies of the wild type palindrome II (Pal II), four copies of palindrome II mutated in both sites (Mb), or four copies of the Hes-1 RBPJ binding site A were cloned in front of a herpes simplex virus thymidine kinase promoter-driven luciferase reporter construct to generate wtPal II-TK, mutPal II-TK, or Hes-TK, respectively.
Electrophoretic Mobility Shift Assay (EMSA)-Binding reactions contained 2 ϫ 10 4 cpm of DNA probe and varied amounts of protein with or without competitors or antibody in a final volume of 20 l containing 100 mM KCl, 20 mM Tris (pH 7.5), 1 mM EDTA, 5 mM dithiothreitol, 5 mM MgCl 2 , 1 g of poly(dI-dC), and 4% glycerol. The mixtures were incubated for 20 min at room temperature, loaded on a 4 or 6% acrylamide gel that had been run for 30 min at 4°C in 0.5 ϫ TBE (25 mM Tris borate, 0.5 mM EDTA), and electrophoresed at 10 mA for 2 h. Gels were dried and autoradiographed overnight.
Anti-HA antiserum was purchased from BABCO (Berkeley Antibody Co., Berkeley, CA). Oligonucleotides were synthesized using an Applied Biosystems 380 DNA synthesizer. The oligonucleotides used in Figs . Briefly, nuclear extract made from F9 cells was negatively absorbed to DE52 and to a DNA affinity column prepared using a mutant oligonucleotide (Mb) followed by two rounds of specific DNA affinity chromatography using palindrome II (CTGGGAGCGCGCTTGCTCCCAA). The final step was gel filtration chromatography. All active fractions eluting from the Superdex 200 fast protein liquid chromatography column were pooled and separated by SDS-polyacrylamide gel electrophoresis and then transferred onto a nitrocellulose membrane. After visualization by staining with Amido Black 10B, the protein band was excised and subjected to in situ digestion with trypsin. The excised band was treated with polyvinylpyrrolidone to prevent binding of the enzyme to the membrane. Digestion with trypsin (1 g in 30 l of 0.1 M TES, pH 8) was allowed to proceed overnight. The supernatants were then fractionated by reversed-phase HPLC. Fractions were collected manually, based on absorbance at 210 nm. Fractions corresponding to symmetrical peaks in the chromatogram were then subjected to chemical sequence analysis in an ABI470 (Applied Biosystems Inc.) protein sequencer (37) and analyzed by matrix-assisted laser desorption mass spectroscopy on a Bruker Reflex Time-of-Flight instrument. Two HPLC fractions of the protein digest were sequenced. One fraction yielded DGYIHYGQTVK (Fraction 29).
The mass spectrum contained one strong signal at m/z ϭ 1280. The other fraction was Q(TP)(AV)(QL)(VL)(DP)(AV)(DT)(DL)(PV)XSQL (Fraction 46) where residues given in parentheses were observed in the same sequencing cycle, indicating the presence of two peptides in that fraction. The mass spectrum contained a major signal at m/z ϭ 1235 and a weak signal at m/z ϭ 1750. A BLAST computer homology search (38) revealed that all these sequences were identical to tryptic fragments of mouse/Xenopus RBPJ. All molecular masses determined were within 0.05% of the masses calculated for the corresponding RBPJ tryptic fragments.

Palindromes of the ERBB-2 Gene Interact with a Specific
Nuclear Protein, PBP-Previous studies showed that a 100-bp region upstream of the TATA box of the ERBB-2 gene promoter increased basal promoter activity 200-fold (11). Two palindromic sequences are a prominent feature of this 100-bp enhancer element in addition to a strong Sp1 site near the 5Ј end and a CAAT box near the 3Ј end. The distal palindrome (Pal I) overlaps the Sp1 site, and the proximal palindrome, Pal II, overlaps the CAAT box (8,9). When this 100-bp region was used as a probe in EMSA, proteins in F9 nuclear extract formed several complexes (Fig. 1). The complexes that could be effectively competed by an Sp1 consensus sequence are represented by Sp1 and proteins with a recognition motif similar to Sp1. A complex designated as PBP could not be competed by the Sp1 consensus oligonucleotide but was effectively competed by the palindrome sequence. F9 cells thus contain a specific palindrome binding activity.
PBP Is RBPJ-To determine the identity of PBP, it was purified to homogeneity from F9 cells using ion exchange, DNA affinity, and gel filtration chromatographies and SDS-polyacrylamide gel electrophoresis. Fig. 2A summarizes the purification steps. Fig. 2B shows purified PBP before trypsin digestion, and the peptides identified by protein sequence analysis and mass spectroscopy are shown in Fig. 2C. A computer homology search revealed that all three sequences were identical to tryptic fragments of RBPJ, a protein originally identified through binding to the J immunoglobulin recombination signal sequence (14).
To verify that RBPJ had palindrome binding activity, the protein was expressed as a HA-tagged fusion protein and transfected into 293 cells. Transient expression of HA-RBPJ resulted in a large increase of PBP activity that could be competed by wild type Pal II but not by Mb oligonucleotides. An antibody against the HA tag supershifted the complex, supporting the sequence identification of PBP as RBPJ (Fig. 3).
Binding of PBP Is Competed Effectively by Other RBPJ Binding Sites-RBPJ was originally isolated as a protein that bound to the J recombination signal sequence consisting of a heptamer and a nonamer with a 23-bp spacer (14). RBPJ protein is highly conserved from Caenorhabditis elegans to Homo sapiens (14,15,19,20,39). The consensus recognition sequence for the RBPJ family has subsequently been determined as GTGGGAA, which is present in all RBPJ binding sites characterized so far (22). The PBP core site CTGGGAGC is close but not identical to the consensus sequence (13). To compare the affinity of ERBB-2 palindromes with other RBPJ binding sites, palindromes as well as other characterized binding sites were used as unlabeled competitors in EMSA using partially purified F9 nuclear extract and Pal II as a probe. As shown in Fig. 4, the EBV Cp sequence is the optimal binding site. Pal I, Pal II, and half-site mutation M3 have affinity similar to that seen with the Hes-1 and pIX gene promoters. A perfect matching consensus site is located at Ϫ596 of the ERBB-2 gene promoter (8). It is, however, a poor competitor, indicating that flanking sequences significantly affect RBPJ binding affinity for the core consensus sequence.

RBPJ-stimulated Transcription via the ERBB-2 Palindrome Is Augmented by NOTCH-IC but Not by EBNA2-To
study the function of RBPJ on the ERBB-2 promoter palindrome, RBPJ was transfected into 293 cells along with reporter constructs containing wild type or mutant palindrome sequences (Fig. 5). The control reporter TK was induced 3-fold by RBPJ. RBPJ activated wtPal II-TK 16-fold, whereas mut-Pal II-TK was induced no more than the background activation seen with the vector TK. Constitutively active forms of mammalian NOTCH stimulated transcription of the Hes-1 gene presumably through RBPJ and its binding site on the Hes-1 promoter (26). The intracellular domain of NOTCH was transfected to determine whether an active form of human NOTCH (NOTCH-IC) had any effect on the ERBB-2 gene (Fig. 5). NOTCH-IC alone did not increase reporter activity above background. However, co-transfection of NOTCH-IC with RBPJ activated wtPal II-TK activity 82-fold. NOTCH-IC failed to affect the basal stimulation of RBPJ seen with TK and mutPal II-TK. The effect of NOTCH-IC was thus specific to the wild type ERBB-2 palindrome-containing gene and was dependent on RBPJ.
The effects of RBPJ and NOTCH-IC on ERBB-2 palindromes were compared with effects on the Hes-1 promoter. As shown in Fig. 6, NOTCH-IC alone was sufficient to induce activation of the Hes-1 promoter, whereas RBPJ alone did not stimulate the Hes-1 reporter. Co-transfection of RBPJ and NOTCH-IC caused no change in activity beyond that observed with NOTCH-IC alone. EBNA2 is a well described activator of some RBPJ response genes (15,36). It up-regulates Cp and CD23 genes utilizing RBPJ and its binding sites on extended promoter elements. When EBNA2 was transfected into 293 cells, it had no effect on ERBB-2 or Hes-1 promoter activity. EBNA2 also had no effect on activation mediated by RBPJ and/or NOTCH-IC on either the ERBB-2 palindromes or the Hes-1 site. Thus EBNA2 requires interactions additional to those mediated via RBPJ (40), whereas effects of NOTCH⅐RBPJ occur on promoters containing only RBPJ response elements. DISCUSSION The present studies identify the protein (PBP) that binds to the two palindromes of the proximal enhancer of ERBB-2 as RBPJ. Peptide sequences of tryptic fragments of PBP were identical to those of RBPJ (14 -16). Although the ERBB-2 palindrome half-sites do not perfectly match the reported RBPJ consensus binding sequence (22), PBP/RBPJ bound to the palindromic sites with an affinity similar to that observed for the consensus sites. NFB/Rel binding sites, which also resemble the palindrome half-sites, do not compete for PBP binding (13). Additionally, nuclear extracts containing epitope-tagged RBPJ formed a specific complex with ERBB-2 palindromes. Previous studies identified two components of PBP (13). We suggest that these likely represent splicing variants of RBPJ with varying N termini (23).
RBPJ is reported to function as a site-specific DNA binding protein that recruits additional transcription factors to target genes (17,36,41). RBPJ represses transcription from its cognate site in the adenovirus pIX gene promoter (18) and from a Gal4 binding site when expressed as a Gal4 fusion (42). A repression domain in RBPJ was identified that coincided with that required for activation by EBNA2 (42). EBNA2 was thus deduced to activate transcription by both masking the repression domain of RBPJ and by bringing a strong transactivation domain to the RBPJ binding site.
RBPJ is the mammalian homolog of Drosophila Su(H), which functions in the NOTCH pathway and regulates development of the peripheral nervous system (19,20). Su(H) interacts positively with NOTCH (43,44) and negatively with Hairless (45). Fortini and Artavanis-Tsakonas (27) observed that Su(H) was translocated to the nucleus when the transmembrane receptor NOTCH was activated by binding the ligand Delta. Interestingly, an oncogene from a human T lymphoblastic leukemia (TAN-1) was identified as a truncated NOTCH lacking much of the extracellular domain (31). Kopan et al. (30) presented evidence that proteolytic processing of NOTCH could generate NOTCH-IC sufficient to act via a complex with RBPJ in the nucleus. RBPJ binds to the intracellular domain of NOTCH, and expression of RBPJ with ectodomain-deleted forms of NOTCH (␦EC-NOTCH), both with and without retention of the transmembrane domain, activates transcription from Hes-1 and EBV C promoters (26,46). ␦EC-NOTCH⅐RBPJ and EBNA2⅐RBPJ complexes exhibited similar transcriptional enhancement on EBV C promoter sites (46). RBPJ stimulated transcription from a reporter gene containing wild type but not mutant ERBB-2 palindrome sequences. This activity was greatly enhanced by NOTCH-IC. Using immunofluorescence, we found that NOTCH-IC and RBPJ localize in the nuclei of 293 cells (data not shown). These observations support models of RBPJ⅐NOTCH-IC heterodimers that act to enhance transcription from RBPJ response elements. Comparison of ERBB-2 palindrome and Hes-1 elements in reporter gene assays indicates a stronger effect of transfected RBPJ alone on ERBB-2 and a stronger effect of transfected NOTCH-IC alone on Hes-1. This may reflect a higher affinity in vivo of endogenous RBPJ for the Hes-1 site. RBPJ was not inhibitory on either site. On these sites, EBNA2 had no effect. This suggests that stimulatory effects of EBNA2  require interaction with additional proteins such as Spi-1/ Spi-B, which bind to a more extended EBV C element (40).
Several recognition elements have been identified within the extended ERBB-2 promoter. Two AP2 and two Sp1 consensus binding sequences are found at Ϫ397, Ϫ359, Ϫ369, and Ϫ314, respectively, relative to the translation start site (8). Several breast cancer cell lines are reported to have a strong AP2 activity that increases expression via the response element located at Ϫ397 bp (47). A stimulatory factor that binds to the promoter region 3Ј of the TATA box has also been described (48). Analyses of the rat ERBB-2 promoter have identified a transacting factor that binds at Ϫ466 to Ϫ456 bp (49) and inhibition by co-expression of Rb (50), adenovirus E1A (51), and c-Myc (52). However, deletional analysis of the human ERBB-2 gene identified the region between Ϫ329 and Ϫ230 bp as the major strong enhancer region; it contains two functionally important palindromes (11). The activity of RBPJ, which binds to each palindrome half-site (13), is strongly augmented by NOTCH-IC. Interestingly, mice that contain homozygous deletions of ERBB-2 exhibit defects in development of cardiac trabeculae (53), but development of cranial neural crest-derived sensory ganglia and motor neurons was also impaired, implying important roles for ERBB-2 in both cardiac and neural development. Control of ERBB-2 expression by NOTCH signaling pathways that are known to be important in neuronal development in transgenic mice thus appears plausible.