A heterodimeric nuclear protein complex binds two palindromic sequences in the proximal enhancer of the human erbB-2 gene.

Increased expression of the protein-tyrosine kinase receptor ErbB-2 occurs frequently in human breast and ovarian cancer and causes transformation in experimental systems. Control of transcription of the erbB-2 gene is an important determinant of receptor expression. Within the human erbB-2 promoter, a 100-base pair (bp) region 5' to the TATA box enhances transcription 200-fold. Two palindromes present in this 100-bp region are important for both positive and negative transcriptional control. A nuclear palindrome binding protein (PBP) has been purified to near homogeneity using ion-exchange, DNA-affinity, and gel filtration chromatography. PBP is a heterodimer consisting of a 69-kDa alpha subunit that binds DNA and a 60-kDa beta subunit that appears to enhance subunit binding. DNase I footprinting and electrophoretic mobility shift assays indicate that PBP binds to the half-site of each palindrome with the core recognition sequence TGGGAG. By DNA binding specificity and lack of immunological cross-reactivity, PBP is distinct from NF-kappaB and Ikaros, two proteins with related DNA binding specificities. PBP is proposed to be an important regulator of transcription of the erbB-2 gene.

Increased expression of the protein-tyrosine kinase receptor ErbB-2 occurs frequently in human breast and ovarian cancer and causes transformation in experimental systems. Control of transcription of the erbB-2 gene is an important determinant of receptor expression. Within the human erbB-2 promoter, a 100-base pair (bp) region 5 to the TATA box enhances transcription 200-fold. Two palindromes present in this 100-bp region are important for both positive and negative transcriptional control. A nuclear palindrome binding protein (PBP) has been purified to near homogeneity using ionexchange, DNA-affinity, and gel filtration chromatography. PBP is a heterodimer consisting of a 69-kDa ␣ subunit that binds DNA and a 60-kDa ␤ subunit that appears to enhance subunit binding. DNase I footprinting and electrophoretic mobility shift assays indicate that PBP binds to the half-site of each palindrome with the core recognition sequence TGGGAG. By DNA binding specificity and lack of immunological cross-reactivity, PBP is distinct from NF-B and Ikaros, two proteins with related DNA binding specificities. PBP is proposed to be an important regulator of transcription of the erbB-2 gene.
The erbB-2 gene is the second identified member of the EGF 1 receptor subfamily of receptor tyrosine kinases that also includes ErbB-3 and ErbB-4 (1)(2)(3)(4)(5). erbB-2 was initially identified as the neu oncogene in rat neuro/glioblastomas induced by transplacental mutagenesis with ethylnitrosourea (6 -10). It is widely expressed during fetal development but is low to absent in normal adult tissues (11)(12)(13)(14). Although the activating point mutation in rat neu (Val 664 3 Glu) (15) has not been seen in human tumors (16,17), amplification and overexpression of the human erbB-2 gene is frequent in adenocarcinomas, especially those arising in breast and ovary, where overexpression directly correlates with poorer patient outcomes (18,19). In general, protein expression is the result of gene amplification; however, many tumors overexpress erbB-2 mRNA and protein from single copy genes (19,20). Even with gene amplification, mRNA expression per gene is increased, indicating that transcriptional control mechanisms are likely important (20 -22).
We previously reported the sequence of 3.65 kilobases of the human erbB-2 gene promoter (23,24). This promoter has typical TATA and CAAT elements in the region proximal to the translation start site (ϩ1) and 4 Alu sequences located in the upstream region (24). A 100-bp region upstream of the TATA box increased basal promoter activity 200-fold (25). This 100-bp region contains a strong Sp1 site near the 5Ј end and a functional CAAT box near the 3Ј end. Two palindromic sequences are a prominent feature of this 100-bp control region (23,26,27); the distal palindrome (Pal I) overlaps the major Sp1 site and the proximal palindrome (Pal II) overlaps the CAAT box (25). An AP2 site which is located 5Ј to this 100-bp region increased expression in several breast cancer cell lines that exhibited elevated AP2 activity (21,28).
Similar palindromes are present in rat, mouse, and human erbB-2 promoters (29). Using reporter constructions, these palindromes were shown to exert both positive and negative regulation of the human erbB-2 promoter (25). To investigate control of the erbB-2 promoter via these palindromic sites, we have purified and characterized a protein complex that specifically interacts with active but not with mutant palindromic sequences. This palindromic binding protein (PBP) is a heterodimer consisting of 69-and 60-kDa subunits. PBP interacts with the half sites of both palindromes I and II primarily via the 69-kDa subunit. Although the palindrome DNA half-sites resemble NF-B and Ikaros binding sites, nucleotide competition and immunological studies indicate that PBP is distinct from both.

MATERIALS AND METHODS
Oligonucleotides-Oligonucleotides were synthesized using an Applied Biosystems 380 DNA synthesizer and were labeled with [␥-32 P]ATP and T4 polynucleotide kinase. The DNA probe corresponding to Ϫ329 to Ϫ230 bp of the erbB-2 promoter was generated by polymerase chain reaction and end-labeled with [␣-32 P]dATP and the Klenow fragment of DNA polymerase.
DNA-affinity columns were prepared using established methods (30) in which multimerized double-stranded oligonucleotides were coupled to CNBr-activated Sepharose 4B. WT is the Pal II sequence flanked by TCGA (see Fig. 1). MT is Mb flanked by TCGA. The WT and MT oligonucleotides used for multimerization have the following sequences: oxyuridine oligonucleotide probe and varying amounts of protein without or with competitors in a final volume of 40 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 30 min at room temperature and, where indicated, irradiated with a UV transilluminator for 10 min. Aliquots (20 l) were loaded on a 6% acrylamide gel which 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 150 V for 2-3 h. The remaining 20 l was resolved by SDS-PAGE. Gels were dried and autoradiographed overnight.
DNase I Footprinting-The DNA probe was the Ϫ329 to Ϫ230-bp fragment of the erbB-2 promoter labeled at either end. The binding reactions contained 4 ϫ 10 4 cpm of probe and 10 l of purified PBP. After DNase I cleavage and phenol/chloroform extraction, the products were resolved in a 13% denaturing gel (32). The dried gel was autoradiographed overnight.
Southwestern Blotting-SDS-PAGE was performed as described by Laemmli (33). In native gels, protein samples were loaded without denaturing buffer or heating prior to loading. Gels were run, and all subsequent steps were carried out at 4°C. After electrophoresis, protein was transferred to nitrocellulose and filters were incubated with binding buffer (100 mM KCl, 20 mM HEPES (pH 7.5), 1 mM EDTA, 2 mM dithiothreitol, 5 mM MgCl 2 , and 20% glycerol) with 5% nonfat milk for 30 min. Filters were then incubated in binding buffer with 1% milk, 1 g of poly(dI-dC), 1 g of Mb oligonucleotide, and 10 6 cpm/ml 32 Plabeled Pal II oligonucleotide without or with 20-fold excess unlabeled Pal II oligonucleotide overnight. Filters were then washed 3 times for 10 min each in binding buffer containing 1% nonfat milk. Filters were exposed to Kodak XAR film at Ϫ80°C with an intensifying screen.
Purification of PBP-All steps were carried out at 4°C, and PBP activity was monitored by EMSA at each step. Nuclear extracts from 500 ϫ 15-cm dishes of F9 cells were prepared according to Dignam et al. (34) without ammonium sulfate precipitation. The nuclear extract (50 ml) was diluted with an equal volume of buffer without KCl (10 mM HEPES, pH 7.9, 1 mM EDTA, 5 mM dithiothreitol, 5 mM MgCl 2 , 0.1% Nonidet P-40, and 20% glycerol) and applied to a DE52 column (50 ml bed volume) equilibrated with buffer containing 200 mM KCl. The unabsorbed flow-through from DE52 (100 ml) was diluted with an equal volume of buffer without KCl, and 5 mg of poly(dI-dC) was added. The mixture was rotated for 20 min and centrifuged at top speed in a clinical centrifuge for 20 min, and the supernatant was loaded onto a 2-ml DNA-affinity column containing the mutant oligonucleotide Mb. The unabsorbed flow-through from the Mb DNA-affinity column was loaded onto a 1.5-ml sequence-specific DNA-affinity column containing concatamerized Pal II. The column was washed with 6 column volumes of buffer containing 100 mM KCl, 6 column volumes of buffer containing 200 mM KCl, and bound protein was eluted with 2 ml of buffer containing 600 mM KCl. The 600 mM KCl fraction was diluted to 100 mM KCl, 40 g of poly(dI-dC) was added, and material was chromatographed on a second WT DNA-affinity column of 1-ml bed volume.
The active fraction eluted from the second specific DNA-affinity column was diluted with an equal volume of buffer to give a final concentration of glycerol of 10% and KCl of 300 mM. A 500-l aliquot was loaded onto a Superdex 200 FPLC column. The column was equilibrated and chromatographed in buffer containing 10% glycerol and 300 mM KCl. Individual fractions of 0.5 ml were collected and stored at Ϫ80°C.
Protein Detection-Protein was detected by the Bradford procedure. For detection of small amounts of protein, samples were biotinylated and precipitated with 10% trichloroacetic acid. After electrophoresis in SDS-PAGE gels, protein was transferred to nitrocellulose filters and incubated for 1 h at room temperature with streptavidin (2 g/ml).

erbB-2 Promoter Palindrome-binding Protein Complex
After washing, filters were incubated with a mouse anti-streptavidin antibody and visualized with an alkaline phosphatase-conjugated goat anti-mouse IgG F(ab) 2 . Standard curves were constructed using bovine serum albumin.

RESULTS
Purification of the erbB-2 Palindrome-binding Protein-The erbB-2 promoter contains 2 palindromes, each half consisting of 8 bp (Fig. 1). Pal I has a 5-bp and Pal II a 6-bp spacer region; the 2 palindromes differ by 1 bp in the 5Ј-half and by 2 bp in the 3Ј-half. Because Pal I contains an overlapping binding site for Sp1, Pal II was used in DNA-affinity chromatography to avoid retention of Sp1 or related proteins; Pal II does not bind Sp1 (25). F9 teratocarcinoma cells were used as an enriched source of starting material, and activity was monitored using EMSA.
Nuclear extracts were negatively absorbed to DE52 and to a DNA-affinity column prepared using an oligonucleotide that retained the central 8 bp but was mutant in both halves of Pal II (Fig. 1). Two sequential site-specific DNA-affinity chromatography steps resulted in approximately a 17,000-fold enrichment of PBP (Fig. 2, A and B, and Table I). When this material was chromatographed on Superdex 200 using FPLC, the peak activity migrated anomalously at a position corresponding to ϳ14 kDa (Fig. 2C). As shown in Fig. 2D the peak DNA binding activity consisted of 2 proteins of ϳ69 and 60 kDa designated ␣ and ␤, respectively. A larger M r protein of ϳ150,000, which exhibited weak DNA binding activity, was removed by the Superdex 200 chromatography step. A faster migrating band was weakly detected. Purified PBP is thus a tightly associated complex consisting of approximately equimolar amounts of two proteins.
Interaction of PBP with the erbB-2 Promoter-To identify the sites of interaction with the 100-bp erbB-2 core promoter/enhancer region, purified PBP was used for DNase I footprinting (Fig. 3). Four binding sites were identified corresponding to half of each palindrome. The core binding sequence is TGGGAG.
Previous studies indicated that mutations in either the 5Ј or 3Ј half of the palindromes decreased only modestly the DNA binding and functional activity of the palindromes suggesting that the half-site was the functionally active response element (25). At lower concentrations of PBP, a single complex (Ca) was observed in EMSA on both Pal I and Pal II (Fig. 4). With increasing amounts of PBP, a second slower migrating complex, Cb, was formed. These results support the concept that dimeric PBP binds principally to a half-site core sequence in either palindrome; with higher concentrations of PBP, both half-sites can be occupied simultaneously.
PBP bound to both Pal I and Pal II and DNA binding specificity were confirmed using oligonucleotide competitors in electrophoretic mobility shift assays. As shown in Fig. 5A, binding of PBP to 32 P-labeled Pal I was competed by unlabeled Pal I or Pal II oligonucleotides but not by an oligonucleotide containing the 8-bp central region of Pal II, but with alterations in both of the half-sites were identified by DNase I footprinting. Binding was not competed by a consensus Sp1 binding site oligonucleotide or by an NF-B site oligonucleotide. These results indicate that, although an Sp1 binding site partially overlaps Pal I, PBP does not recognize the Sp1 binding site. While the PBP binding site closely resembles an NF-B site, PBP did not recognize an NF-B site derived from the HIV long terminal repeat (36) either by competition (Fig. 5A) or by direct binding assays (data not shown). PBP binding exhibited similar binding specificity for Pal II (Fig. 5B). Binding was competed by mutations in either palindrome half-site (M3, M5, LSM) but not by mutations in both (Mb). These results confirm PBP binding to either half-site of the palindromes.
Binding to DNA via the ␣ Subunit of PBP-Because the dimeric PBP complex bound to a 6-bp core representing half of The weak Sp1 binding sites II and III have not been shown to respond to Sp1 (25); Sp1 site II corresponds to a region necessary for function of an alternate initiation site (35).

TABLE I Purification of palindromic binding protein (PBP)
Nuclear extracts were prepared from 500 ϫ 15-cm dishes of F9 cells. PBP activity was assayed by EMSA. One unit of activity corresponds to 1 cpm of shifted probe. Protein concentration could not be accurately measured in the fractions derived from step 6. ND, not determined.
Step erbB-2 Promoter Palindrome-binding Protein Complex each palindrome, it was important to determine how the complex recognized DNA. PBP was UV-cross-linked to 5-bromo-dUTP-substituted Pal I or Pal II, and the products were analyzed by SDS-PAGE and autoradiography. As shown in Fig. 5, C and D, the 69-kDa ␣ subunit is the major DNA binding unit. A small amount of the 60-kDa ␤ subunit (Ͻ10%) was detected bound to Pal I but none was bound to Pal II. DNA binding specificity was confirmed using unlabeled oligonucleotides as competitors in cross-linking assays. Binding of the 69-kDa ␣ subunit was competed by Pal I and Pal II oligonucleotides but not by Mb, NF-B, or Sp1 oligonucleotides. The small amount of 60-kDa ␤ subunit binding to Pal I was incompletely competed by palindrome oligonucleotides but was competed by the Sp1 oligonucleotide indicating lack of specificity of the ␤ subunit binding to DNA. The proteins present in the single site EMSA complex formed on Pal I and on Pal II (Fig. 5, A and B) were eluted and separated by SDS-PAGE. Approximately equal amounts of the 69-and 60-kDa species were detected in each complex (Fig. 6A). Thus, although the 69-kDa ␣ subunit was the species crosslinked to the DNA binding site, equal amounts of ␣ and ␤ subunits were present in the DNA-bound complex providing evidence that PBP is a heterodimer.
Southwestern blotting confirmed specific binding of 32 P-labeled Pal II to the 69-kDa ␣ subunit of PBP (Fig. 6B). The higher molecular mass band was not competed, consistent with it corresponding to the smaller amount of higher molecular mass activity shown in Fig. 2C, and the 87-kDa band was inconsistent. When Southwestern blotting was carried out using nondenaturing polyacrylamide gel electrophoretic separation of proteins, strong specific DNA binding was observed (Fig.  6B, right panel). This DNA binding activity was at least 100fold higher using native compared to denaturing gel electrophoresis, a result that could reflect incomplete protein renaturation or higher affinity of the heterodimer ␣␤ PBP compared to the ␣ subunit alone.
Distinguishing PBP from NF-B and Ikaros-Because the half-sites of the erbB-2 palindromes resemble the DNA recognition elements for NF-B (Fig. 1) and because NF-B, like PBP, binds as a heterodimeric protein complex (p65/p50 and rel/p50 (31)), it was important to distinguish PBP from NF-B. DNA recognition appeared distinct because an NF-B binding site did not compete with erbB-2 palindromes for PBP binding (Fig. 5) nor did PBP bind directly to the NF-B site (data not shown). erbB-2 Pal I competed weakly for NF-B binding to its cognate recognition site (Fig. 7A, lanes 2 versus 4). Antibodies specific for p65, rel, and p50 supershifted the NF-B complexes formed on the cognate NF-B binding site oligonucleotide (Fig.  7A, lanes 5, 6, and 7). These antibodies failed to affect PBP complexes formed on either Pal I or Pal II of erbB-2 (Fig. 7B). By both DNA binding specificity and lack of immunological reactivity, PBP thus appears distinct from NF-B. As noted previously, the palindrome response elements also resemble the Ikaros binding site within the TdT promoter (37,38). Ikaros did not, however, bind to the erbB-2 palindromes (data not shown) nor did an antibody specific for Ikaros affect the mobility of the PBP-palindrome complex in EMSA indicating PBP is also distinct from Ikaros (Fig. 7B). DISCUSSION Excessive mitogenic signaling via the ErbB-2 receptor tyrosine kinase may result from mutational activation as occurs with rat neu (15) or from gene amplification and enhanced

erbB-2 Promoter Palindrome-binding Protein Complex
transcription of wild-type erbB-2 (18 -20). Regulation of transcription of the erbB-2 gene is one important determinant of the extent of ErbB-2 expression. It is thus important to identify elements that control transcription of the erbB-2 gene. As assayed using reporter gene constructions in HeLa and CV1 cells, full promoter activity of the proximal ϳ1500 bp of human erbB-2 was retained in the Ϫ330 bp proximal to the translation start site (23). Several breast cancer cell lines are reported to have strong AP2 activity which increased expression via a response element located at Ϫ397 bp (21,28). A 100-bp strong enhancer region is located proximally at Ϫ329 to Ϫ230 bp (25). This region contains a 5Ј Sp1 site and a 3Ј CAAT box. There are 2 dyad symmetries within this 100-bp region that are highly conserved among human, rat, and mouse promoters (29). When placed in front of a minimal TATA box promoter, palindromes enhanced activity. Whereas deletion of Pal I in the context of the erbB-2 promoter reduced activity consistent with an enhancing effect, deletion of the 3Ј-half of Pal II that overlaps the CAAT box increased promoter activity. Function of the palindromes was thus complex and dependent on their context within the promoter.
A dimeric protein complex that specifically binds these erbB-2 palindromes has been purified more than 17,000-fold. The first indication that PBP exists as a complex was its aberrant migration on gel filtration chromatography. The excessive retention of the 69-and 60-kDa proteins on the Superdex 200 column is, perhaps, due to hydrophobic interaction with the column material, and the co-migration of both proteins with the peak of EMSA activity provided evidence that both participated in the active PBP complex. Approximately equal amounts of the 2 subunits were isolated from the DNA-bound complex separated on EMSA. The 69-kDa ␣ subunit contacted the palindrome half-sites. The amount of the 60-kDa ␤ subunit cross-linked to Pal I was small, less specific, and binding to Pal II was not detected. Only binding to the ␣ subunit was detected by Southwestern analysis. Although we cannot exclude that the ␤ subunit is a proteolytic product of ␣, this appears unlikely. Approximately equal amounts of ␣ and ␤ were present in the protein complexes bound to both Pal I and Pal II. Moreover, material purified through 2 site-specific DNA-affinity columns and resolved on Superdex 200 contained approximately equal amounts of ␣ and ␤. Although ␤ did not bind DNA, Southwestern analysis revealed 100-fold higher DNA binding to the native complex than to the ␣ subunit resolved on denaturing gels. While this could be due to incomplete renaturation of the ␣ subunit, the results taken together suggest that PBP is a heterodimeric complex in which binding of the ␣ subunit to DNA is enhanced by the ␤ subunit.
DNase I footprinting confirmed binding to 4 sites within the 100-bp region that correspond to the halves in each palindrome. Each half of the two palindromes appears to be an independent binding site with the sequence TGGGAG. Increasing amounts of PBP resulted in formation of a slower migrating complex on EMSA consistent with occupancy of both halves of the two palindromes.
Cloning will be required to further characterize PBP and its relation to other transcription factors. DNA binding specificity and lack of immunological cross-reactivity indicate PBP is distinct from NF-B and Ikaros, two proteins with related DNA binding specificity. We suggest that PBP will prove an important regulator of erbB-2 transcription and thus biological responses to this tyrosine kinase growth factor receptor.
FIG. 6. Analysis of PBP⅐DNA complexes. A, proteins in the PBPpalindrome complex. DNA⅐PBP complexes in the gel-shifted complexes (Fig. 5, A and B) were identified by autoradiography of the wet gels and excised. Gel slices were fragmented, suspended in 1% SDS heated at 68°C, and the aqueous phase was recovered after centrifugation. After trichloroacetic acid precipitation, the samples were suspended in water and subjected to electrophoresis on a 10% SDS gel; proteins were detected as described under "Materials and Methods." B, detection of the DNA binding subunit of PBP. Southwestern analysis was performed as described under "Materials and Methods." After protein transfer from SDS (left) and native (right) polyacrylamide gels, nitrocellulose was incubated with the 32 P-end-labeled Pal II oligonucleotide without (Ϫ) or with (ϩ) 20-fold excess of unlabeled probe. Filters were autoradiographed at Ϫ80°C with an intensifying screen for 5 days (left) or 4 h (right).