The Amyloid &Protein Precursor Promoter A REGION ESSENTIAL FOR TRANSCRIPTIONAL ACTIVITY CONTAINS A NUCLEAR FACTOR BINDING DOMAIN*

A manifestation of Alzheimer's disease is the presence of amyloid depositions in brains of afflicted individuals. A major component of these depositions is the amyloid beta-protein, which is a truncated form of the larger amyloid beta-protein precursor (APP). To investigate the regulation of APP gene expression, the APP promoter and selected deletions were placed 5' to the reporter gene chloramphenicol acetyltransferase. The promoter deletions were transfected into different cell lines that showed variant levels of endogenous APP transcripts. Transient transfection assays showed that 96 base pairs 5' to the transcriptional start site are sufficient for cell type-specific promoter activity. A nuclear factor that binds to this region in a sequence-specific manner was identified by mobility shift electrophoresis, DNase footprinting, and methylation interference. The DNase-protected region covers about 25 base pairs on both strands (position -31 to -55). Mutations within this domain revealed a sequence of 12 base pairs that is crucial for factor binding. This sequence overlaps with the consensus sequences for transcription factors AP-1 and AP-4. However, competition experiments suggest that the nuclear factor that binds to the APP promoter is distinct from both AP-1 and AP-4. Factor binding to the characterized recognition sequence is observed in nuclear extracts originating from human, mouse, and rat cells, suggesting a high degree of conservation.

. The APP gene is differentially expressed in all major tissues. In brain it is expressed primarily, but not exclusively, in neurons (Schmechel et al., 1988). The apparent overexpression of the APP gene in Down's syndrome  and in certain areas of the brain in Alzheimer's disease (Cohen et al., 1988;Higgins et al., 1988;Johnson et al., 1990) indicates that this overexpression might be a critical requirement for Alzheimer's disease neuropathology (Goldgaber et al., 1987). Recent experiments with transgenic mice showed that APP overexpression indeed leads to amyloid deposition (Quon et al., 1991;Wirac et al., 1991a). These observations illustrate the importance of elucidating the mechanism of APP gene expression.
The proximal APP promoter region is devoid of CCAAT and TATA elements but contains numerous consensus sequences for known regulatory transcription factors (La Fauci et al., 1989; al., 1988). Of these, only two HOX-1.3 elements have been shown to actually be recognized by the corresponding transcription factor (Goldgaber et al., 1991). The APP promoter mediates neuron-specific gene expression of a reporter gene in transgenic mice (Wirac et al., 1991b). Phorbol esters, interleukin-1, and other cytokines increase levels of APP transcripts (Donnelly et al., 1990;Goldgaber et al., 1989;Mobley et al., 1988). However, no promoter elements that are involved in such increases have been identified. These results, together with the complex pattern of APP gene expression, suggest that the APP promoter contains sequence elements that are targets for regulatory transcription factors. The purpose of this study was to identify promoter regions that are essential for the expression of the APP gene. This was accomplished by analyzing promoter deletions in transient transfection assays. A proximal promoter that extends to position -96 upstream of the transcriptional start site was found to be essential for high levels of expression. This domain contains a recognition sequence for a nuclear binding factor.
The @-galactosidase gene was excised from pCHllO as a HindIII-XmnI restriction fragment and subcloned into pUC18. The @-actin promoter was cloned 5' to the @-galactosidase gene into the SalI restriction site of the polycloning site of pUC18 as an XhoI-HinfI (-277 to +1) restriction fragment, in which the HinfI site had been converted to SalI with linkers (Quitschke et al., 1989a). The entire @galactosidase gene, including the p-actin promoter, was excised as a BamHI-BarnHI restriction fragment in which the BamHI sites were converted to NotI restriction sites with linkers. This fragment was cloned into the NdeI restriction site of pUC18, which had been converted with linkers to a NotI restriction site. The chloramphenicol acetyltransferase (CAT) gene was excised from pSV2-CAT as a HindIII-BarnHI restriction fragment. The BamHI site was bluntended, and the CAT gene was then cloned between the HindIII restriction site and the blunt-ended NarI restriction site of pUC18 containing the @-galactosidase gene.
The structure of the APP promoter has been described in detail elsewhere (La Fauci et al., 1989;Salbaum et al., 1988). The APP promoter fragment used in this study extends from the BarnHI restriction site at position +lo0 downstream of the transcriptional start site to the EcoRI restriction site a t position -2832 upstream of the transcriptional start site (Fig. L4). Deletions at the 5' end of this fragment were produced using selected restriction sites. Deletion 1 was introducedat position -1359 (SspI), deletion 2 at -488 (HindIII), deletion 3 at -303 (XbaI), deletion 4 at -204 (EagI), deletion 5 at -96 (NarI), and deletion 6 at -49 (PuuII). Resulting restriction fragments were cloned into the polycloning site of pCAT2bGAL. In addition, the chicken @-actin promoter was cloned into the polycloning site of pCAT2bGAL as a reference promoter. In this control plasmid, both the @-galactosidase and the CAT genes are transcribed from identical @-actin promoter fragments.
Expression Assays-Transfected cells were harvested, disrupted in three freeze-thaw cycles, and centrifuged at 12,000 X g for 5 min. The supernatant was used for subsequent enzyme assays. @-Galactosidase activity was determined by using chlorophenol red-@-D-galactopyranoside (Boehringer Mannheim) as a substrate. Five percent of cell extract from 25-cmZ flasks was incubated a t 37 "C in 5 mM chlorophenol red-@-D-galactopyranoside, 1 mM MgClz, 100 mM Hepes pH 7.6 for 5-30 min, depending on transfection efficiency. Substrate conversion was determined spectrophotometrically at a wavelength of 570 nm. The @-galactosidase activity was determined to normalize all extracts with regard to transfection efficiency and other variables prior to performing CAT assays.
CAT assays (Gorman, 1985) were performed with cell extracts adjusted to identical @-galactosidase activity and quantitated by liquid scintillation counting of the acetylated and nonacetylated forms of ~-threo-[dichloroacetyl-l-~~C]chloramphenicol (Amersham Corp.), which were excised from thin layer chromatography plates. CAT assays were adjusted so that total conversion into the monoacetylated forms was within the linear range of less than 40%. The CAT activity resulting from each APP promoter construct was normalized to the CAT activity from the @-actin promoter, which was assigned the value 100% in each cell line. Each construct was tested in a t least four separate assays.
RNA Isolation and Northern Anulysis-Cytoplasmic RNA was isolated from cultured cells as described (Maniatis et al., 1989). Approximately 15 pg of each RNA sample was subjected to Northern blot analysis using nick-translated radiolabeled probes. The APP probe was an approximately 1-kb EcoRI fragment (Goldgaber et al., 1987) from a human APP,,, cDNA (base pairs 2020-3076). The @actin probe was an approximately 1.6-kb fragment representing a full-length chicken @-actin cDNA (Eldridge et al., 1985).
DNase Footprinting and Methylation Interference-DNase footprinting (Galas and Schmitz, 1978) was carried out on a 5' endlabeled APP promoter fragment extending from position -96 (NarI) t o +53 (SrnaI) (Fig. 1B). This fragment was incubated with nuclear extract from rat brain as described above, and 0.1 unit of DNase I was added for 2 min followed by electrophoresis in 1% agarose in 0.5 X Tris-borate-EDTA. The DNA was then transferred to diethylaminoethylcellulose paper (Whatman) and autoradiographed. The bound and the free DNA fragments were eluted in 1.5% NACl overnight, extracted with phenol-chloroform, and precipitated with ethanol. The DNase I-digested bound and free fragments were electrophoresed in an 8% sequencing gel.
Methylation interference was assayed by partially methylating the same APP promoter fragment with dimethyl sulfate (Maxam and Gilbert, 1981) before mobility shift electrophoresis. After isolating the bound and the free fragments as described above for DNase footprinting, the partially methylated fragments were cleaved with piperidine, separated in an 8% sequencing gel, and autoradiographed.
Nuclear Extracts-Nuclear extracts from cultured cells were prepared as described elsewhere (Heberlein and Tjian, 1988). Nuclear extracts from rat brain were obtained by a modification of the same protocol. Rat brains (5-15 g) were homogenized in 5 volumes of buffer A (10 mM Hepes pH 7.6, 15 mM KCl, 2 mM MgC12, 0.1 mM EDTA, 1 mM dithiothreitol, 1 mM NaS205, 0.2 mM phenylmethylsulfonyl fluoride). The homogenate was filtered multiple times through lens paper to remove connective tissue and large debris. Subsequently, the homogenate was centrifuged at 15,000 X g for 10 min, and the supernatant was discarded. The pellet was rehomogenized in 5 volumes of buffer A containing 1.2 M sucrose. The homogenate was filtered through lens paper once more and centrifuged in a swinging bucket rotor at 100,000 X g for 30 min. The pellet was rehomogenized in buffer A and centrifuged at 15,000 X g for 10 min. The supernatant was discarded, and the pellet was homogenized in a total volume of 18 ml of buffer A. To this, 2 ml of buffer B (50 mM Hepes pH 7.6, 1 M KCl, 30 mM MgCI,, 0.1 mM EDTA, 1 mM dithiothreitol, 1 mM NaS206, 0.2 mM phenylmethylsulfonyl fluoride) and 2 ml of 4 M (NH&S04, pH 7.6 were added. The viscous solution was mixed in a rotator for 20 min and centrifuged at 100,000 X g for 30 min. The pellet was discarded, and 0.3 g of solid (NH4)&04 was added per ml of supernatant and stirred until dissolved. The precipitated protein was pelleted and redissolved in buffer C (25 mM Hepes pH 7.6, 40 mM KCl, 12.5 mM MgClZ, 0.1 mM EDTA, 1 mM dithiothreitol, 0.5 mM NaSZO3, 0.1 mM phenylmethylsulfonyl fluoride, 10% glycerol) to a final concentration of 2-5 mg of protein/ml. The protein solution was dialyzed against 500 ml of buffer C for 2-4 h. Insoluble material was pelleted, and the supernatant was aliquoted and stored at -80 "C for use in binding studies. All operations were carried out at 4 "C.
Oligonucleotides and Fragments-Complementary double-stranded oligonucleotides were synthesized on an Applied Biosystems DNA synthesizer. They were deblocked and gel-purified prior to labeling and hybridization. Most double-stranded oligonucleotides that were used as competitors and probes are indicated in Figs. 6 and 8. An additional double-stranded oligonucleotide containing a CCAAT binding domain from the chicken @-actin promoter (position -106 to -71) was used as a control for the quality of the nuclear extracts ( Fig. 4), 5'-AGGTGCGGCGGCAGCCAATCAGAGCGGCGCGCT-CCG. The recognition sequence for the CCAAT binding factor is underlined (Quitschke et al., 1989a).
Restriction fragments derived from the APP promoter are described in Fig. l. Additional restriction fragments used as mobility shift competitors were obtained from the chkken @-actin promoter and the chicken cardiac a-actin promoter. The chicken @-actin promoter fragment extended from position -106 to +1 and included both TATA and CCAAT domains as well as an spl and a CArG consenses sequence (Quitschke et al., 1989a). The cardiac a-actin promoter extended from position -100 to +16 and included a TATA and a CArG element (Quitschke et al., 198913).  (Wu et al., 19871, AP-2 (Williams et al., 1988), and AP-4 (Mermod et al., 1988). The positions of relevant restriction sites (vertical lines) are indicated with respect to their distance from the main transcriptional start site (+l). The position numbers of restriction sites mark the first 5'nucleotide that is part of that restriction site. B, the sequence of the proximal promoter segment from position -96 to +105. The positions of relevant restriction sites and reference nucleotides are indicated. This portion of the promoter also contains consensus sequences AP-4 (TCAGCTGACT) and AP-1 (TGACTCG) (underlined). C, schematic structure of the expression vector pCAT2bGAL. All APP promoter constructs and the chicken @-actin promoter (Quitschke et al., 1989a) were cloned into the polycloning site of this vector. pCAT2bGAL, which was used for transfection studies (Fig.  IC). This plasmid contains the transcriptional unit of the bacterial @-galactosidase gene, which is transcribed from the constitutive chicken @-actin promoter (Quitschke et al., 1989a) and serves as an internal control for the expression of the reporter gene CAT (Gorman, 1985). The colinear arrangement of the CAT and @-galactosidase transcriptional units in pCAT2bGAL minimized experimental variations resulting from differences in transfection efficiencies and handling procedures. The reporter gene CAT is transcribed from the APP promoter or its 5' deletion fragments. In addition, a control plasmid was generated in which the @-actin promoter was introduced into the polycloning site of pCAT2bGAL. The control plasmid was used to compare the level of expression from the @-actin promoter with that from the APP promoter and its deletion constructs (Fig. IC). Plasmids containing the promoter constructs were transfected into PC-12, H4, C6, C2C12, and 10T1/2 cells by the calcium phosphate-DNA coprecipitation method (Gorman, 1985).

Deletion Analysis of the Human APP Promoter in Trans-
In addition, the steady state level of endogenous APP transcript was determined by Northern blot analysis in these cell lines and the human retinoblastoma line Y79 (Fig. 2). The primary APP transcripts in all cell lines migrated to a position corresponding to a size of about 3.3 kb. This represents the approximate size of the three major APP transcripts APP770, APPT5,, and APPss5. The highest level, by far, of APP transcript was observed in the human retinoblastoma cell line Y79. In this cell line, the level of APP transcript exceeds that of @-actin transcript, which was determined for reference. PC-12 and H4 cells contained lower levels of APP transcript than @-actin transcript. The level of APP transcript in 10T1/2, C6, and C2C12 cells was exceedingly low and barely detectable. plasmic RNA was isolated from the cell lines C2C12, C6, H4, PC-12, Y79, and 10T1/2 and analyzed by Northern blotting. The blot was sequentially probed with a 1-kb APP cDNA probe and the full-length 1.6-kb chicken$ actin cDNA (Eldridge et al., 1985) as indicated. 10T1/2, C6, and C2C12 are shown. The CAT activities obtained from the full-length APP promoter ( A ) and its deletions (1-6) were normalized to those from the &actin promoter, which were assigned the value 100% in each cell line. Each bar represents the mean value of at least four independent assays.
The cell lines PC-12, H4, C6, C2C12, and 10T1/2 grow as fibroblast-like monolayers. In contrast, the cell line Y79 grows in suspension, which complicates transfection procedures. Therefore, it was excluded from transient transfection experiments even though it showed the highest level of endogenous APP transcript (Fig. 2). The CAT expression from each APP promoter construct was compared with the CAT expression from the @-actin promoter in the control plasmid, which was assigned the value 100% in each cell line. The highest level of APP promoter expression relative to the expression from the @-actin promoter occurs in PC-12 cells followed by H4, 10T1/2, and C6 cells (Fig. 3). The lowest level of CAT activity from the APP promoter is observed in C2C12 cells. These relative levels of expression are in rough agreement with the levels of APP transcript observed in the different cell lines (Fig. 2).
Furthermore, deleting the APP promoter from position -2832 (APP) to position -96 (deletion 5) has no significant effect on promoter activity in any of the cell lines, regardless of level of expression relative to the @-actin promoter. However, further deleting the promoter to position -49 (deletion 6) causes a 7-10-fold decline in promoter activity in all cell lines. These results suggest that a crucial promoter element is present within the 96 base pairs upstream of the transcriptional start site that is either disrupted or removed in deletion 6.
Nuclear Factor Binding to the Proximal APP Promoter Element-To determine if this sequence contains a nuclear factor binding domain, the promoter region extending from position -96 to +53 (NarI-SmaI; Fig. 1B) was 5' end-labeled, and factor binding was examined by mobility shift electrophoresis. Nuclear extracts were obtained from rat brain (Manning et al., 1988) and the human retinoblastoma line Y79 (Fig. 2) since these contain high levels of APP transcript. The cell line Y79 was also selected because it is of human origin since it is possible that a nuclear factor might recognize a sequence that is specific for the human promoter studied here. A 36base-pair oligonucleotide containing a ubiquitous CCAAT binding recognition sequence from the chicken P-actin promoter (Quitschke et al., 1989a) was included in the mobility shift analysis to control for the quality of the nuclear extract.
Both rat brain and Y79 nuclear extracts show the characteristic mobility shift observed with the CCAAT binding domain of the @-actin promoter (Quitschke et al., 1989a(Quitschke et al., , 1989b (Fig. 4A). In addition, both extracts generate a distinct mobility shift with the APP promoter fragment from position -96 to +53 (Fig. 4B). The bound APP promoter fragment migrates to the same position with both rat brain and Y79 nuclear extracts. Identical mobility shift patterns have also been observed with nuclear extracts from PC-12 (rat), C6 (rat), C2C12 (mouse), 10T1/2 (mouse), and H4 (human) cells, indicating that the binding factor is structurally conserved in these species (not shown). To further define the binding region, a different fragment of the APP promoter was used as a probe. This AuaI-AuaI restriction fragment extends from position -77 to -13 ( Fig. 1B) and includes the PuuII restriction site at position -49. This fragment also produces a mobility shift with rat brain and Y79 nuclear extracts (Fig.   4C).
Mobility shift competition was performed to determine the specificity of binding and whether the factor binding to the A I~-A u u I restriction fragment is identical to the factor bind-;' to the entire NarI-SmaI restriction fragment. The 5' endlabeled NarI-SmaI fragment was incubated with nuclear extract from rat brain with increasing concentrations of unlabeled fragment as competitor. Significant Competition for binding occurs when a 50-fold molar excess of competitor is added (Fig. 40, lanes 1-4). The level of competition was the  (Quitschke et al., 1989a). The assay was performed with nuclear extracts from rat brain (Br) and the human retinoblastoma line Y79. The bound ( b ) and free ( f) fragments are indicated. B, mobility shift assay using an APP promoter fragment from position -96 to +53 (NarI-SrnaI; Fig. 1B) as a probe. C, mobility shift assay using an APP promoter fragment from -77 to -13 (AvaI-AoaI) as a probe. D, mobility shift competition. The 5' end-labeled APP promoter fragment from position -96 to +53 (NarI-SrnaI) was incubated with nuclear extract from rat brain (lane I ). The mobility shift was competed with a lo-, 30-, and 50-fold molar excess (IOx, 30x, and 50x) of unlabeled fragment (lanes 2-4). In lanes 5-7, the APP promoter fragment from position -77 to -13 ( A d -A u a I ) was used as an unlabeled competitor. Lunes 8 and 9 show mobility shifts with a 100-fold molar excess ( 1 0 0~) of the chicken cardiac-a actin (-100 to +16) (Quitschke et al., 1989b) and the chicken &actin (-107 t o +1) (Quitschke et al., 1989a) promoters. same when the AuaI-AuaI (position -77 to -13) fragment was used as competitor (Fig. 40, lanes 5-7). No significant competition for binding was observed with a 100-fold molar excess of either cardiaca (lane 8 ) or p-actin (lane 9) promoter fragments. This suggests that factor binding is sequencespecific and that the recognition sequence is located within the promoter fragment extending from position -77 to -13.
Mapping of Factor Binding Site by DNase Footprinting and Methylation Interference-To more precisely define the binding domain, the fragment from position -96 to +53 was 5' end-labeled on either strand. The labeled fragment was incubated with rat brain nuclear extract and partially digested with DNase I. A protected region is observed on both strands, which covers approximately 25 base pairs and extends from position -55 to -31 (Fig. 5). These results suggest that the nuclear factor binding domain extends from position -55 to -31 upstream of the transcriptional start site. Incidentally, deletion 6 at position -49, which displays diminished promoter activity (Fig. 3), interrupts this binding domain (Fig.  1B).
The same fragment was also analyzed by methylation interference assay to determine structurally relevant DNA contact points. Partial inhibition of binding was observed with methylation of two G residues on each strand. The G residues that cause binding interference are located within the DNaseprotected domain determined by the footprinting assay (Fig. 5).
Effect of Sequence Mutations on Binding Activity-To identify the recognition sequence for factor binding, complementary double-stranded oligonucleotides that included the pro- For both assays, the APP promoter fragment from position -96 to +53 was 5' end-labeled and incubated with rat brain nuclear extract. A, DNase footprinting of coding (C) and noncoding ( N ) strands. Free moter region from position -58 to -30 were synthesized. Within this domain, blocks of transverse mutations were introduced, and binding activity was analyzed by mobility shift and competition assays (Fig. 6 ) . Mutating the peripheral nucleotides from -58 to -54 and -30 to -38 (Fig. 6, lane M I ) has no appreciable effect on binding activity. Oligonucleotides containing these transverse mutations display nuclear factor binding that is indistinguishable from the binding to the wild type oligonucleotides and show the same degree of binding competition. However, mutating three additional nucleotides on the 5' side of the binding domain (lane M 2 ) causes a significant reduction in factor binding. Mutating blocks of 3 base pairs at a time beyond that point (lanes M3-M 5 ) causes almost complete inhibition of binding for the next 9 base pairs. Mutating an additional three base pairs (lane M6) further downstream has only a marginal effect on factor binding. These results suggest that the core binding recognition seauence (underlined) is comDosed of the nine base Dairs (GGA)TCAGCTGAC(TCG). The' core is flanked by 3-base Dairs on either side. which disDlav a limited effect on factor binding when mutated (Fig. 6): " Nuclear extracts from cell lines used for APP expression studies were also analyzed to determine if factors present in these cells bind to the same recognition sequence. Electrophoretic mobility shift assays were performed with nuclear extracts from six cell lines (Fig. 7). The extracts were incubated with labeled oligonucleotides containing the unmodified native sequence (Fig. 7, lane W T ) and mutations Ml"6 ( Fig. 6 A ) . The nuclear extracts from all sources display the same binding pattern as is observed with nuclear extract from rat brain (Figs. 6B and 7). This indicates that nuclear factors present in cells of mouse, human, and rat origin bind to the same recognition sequence.
The 9-base pair core domain and its flanking nucleotides b FIG. 6. Effects of mutations on factor binding. A, block mutations in oligonucleotides containing the nuclear factor binding domain were tested for their ability t o bind and compete the nuclear factor from rat brain. WT indicates the unmodified APP sequence domain of an oligonucleotide from position -58 to -30. Oligonucleotides MI-M6 contain various block transverse mutations. The mutated sequence elements are underlined. B, mobility shift assay with the 5' end-labeled nucleotides described in A using rat brain nuclear extract. C, mobility shift competition assay with a 75-fold molar excess of the unlabeled oligonucleotides described in A as competitors.
T h e left lane shows a mobility shift assay without competitor.  contain the recognition sequences for transcription factors AP-4 (CAGCTG) and AP-1 (TGACTCG) (Mermod et al., 1988;Hu et aZ., 1990). It was investigated whether either of these factors contribute to the binding activity described here.

5'-G G A A A G A~A A T G L G I U
In one case, a complementary pair of oligonucleotides was synthesized in which parts of the APP sequence were replaced to generate a perfect AP-1 site (Bohmann et aZ., 1987), while eliminating the upstream AP-4 site (Fig. SA, lane M7). These base replacements resulted in a profound reduction in binding and a different mobility shift pattern (Fig. SB, lane M7).
The original studies on AP-4 and AP-1 were performed on the SV40 enhancer, which contains both sequence elements in the vicinity of each other (Mermod et al., 1988). However, the AP-1 element is in opposite orientation to the AP-4 element. A 29-base pair oligonucleotide identical to that part of the SV40 enhancer sequence was analyzed (Fig. 8A, lane  SV40).
A mobility shift assay with this oligonucleotide results in strong binding activity; however, the bound fragment migrates faster than the native APP promoter fragment (Fig. 8A, lane  SV40). More importantly, neither the M7 nor the SV40 fragments are able to compete for the binding factor to the wild type APP fragment, whereas they are able to compete for their own binding factor (Fig. 8, B and C). These experiments indicate that this APP promoter binding factor is distinct from AP-1 and AP-4.

DISCUSSION
The APP promoter does not contain a CCAAT or TATA element in the vicinity of the main transcriptional start site. However, the GC content of the proximal promoter region (-96 to +1) exceeds 75% and contains multiple CpG elements. It is thus similar to promoters of housekeeping genes, which may be transcriptionally regulated through DNA methylation (Bird, 1986;Gardiner-Garden and Frommer, 1987). In addition, the APP promoter studied here contains several consensus sequences of known potential binding sites for nuclear regulatory factors (Fig. 1). However, no function has been assigned to any of these binding sites, and only two HOX-1.3 recognition sequences have been shown to be recognized by the corresponding transcription factor (Goldgaber et al., 1991). The APP promoter was dissected by introducing 5' deletions at selected restriction sites. These deletion constructs were analyzed by transient transfection in cell lines derived from different species and tissues. In each cell line, the activity of the APP promoter constructs was compared with the activity of the chicken 8-actin promoter. The activity of this promoter is dependent on a ubiquitous CCAAT binding factor and is constitutively expressed in all cell lines examined to date (Quitschke et al., 1989a;Seiler-Tuyns et al., 1984;Vourio et al., 1990). Consequently, the ,&actin promoter serves as a suitable control for variations in promoter expression due to differences in experimental procedures as well as differences in metabolic conditions of cells that may affect general transcriptional activity.
After normalizing the activities from the APP promoter constructs with those from the P-actin promoter, it became apparent that APP promoter expression varied in the different cell lines. This suggested that the human APP promoter contained sufficient sequence information to mediate cell type-specific expression. In addition, in each cell line, promoter expression remained largely constant despite a reduction in promoter size from -2832 to -96 base pairs upstream from the transcriptional start site and the removal of two HOX-1.3 sites, four NF-KB sites, one AP-1 site, one heat shock element, and one AP-2 site (Fig. lA). The expression from the APP promoter differs in each cell line with respect t o expression from the P-actin promoter as determined by transient transfection. Similarly, there are variations in the levels of APP transcript between the cell lines with respect to the level of P-actin transcript (Figs. 2 and 3). In general, the cell lines with the highest level of expression from the APP promoter also contained the highest level of endogenous APP transcript (Figs. 2 and 3). This suggests that an endogenous regulatory mechanism exists, which confers levels of APP expression that are specific for each cell line. This cell type-specific level of expression from the APP promoter is maintained with only 96 base pairs upstream from the transcriptional start site (Fig. 3). Sequence elements further upstream appear to have a negligible effect, if any, on base-line promoter expression. The reduction in APP promoter expres-sion in deletion 6 a t position -49 further substantiates the notion that the observed cell type-specific difference in APP promoter expression is mediated by sequence elements in the proximal promoter domain that are eliminated in deletion 6.
In a different study, APP promoter deletions were analyzed by transient transfection of HeLa and PC-12 cells as reported by Lahiri and Robakis (1991). Those authors report two blocks of regulatory elements within the analyzed promoter region. One block, from position -600 to -460, acts as a positive regulator and another, from position -450 to -150, acts as a negative regulator. The reason for the discrepancies with the results reported here is currently unclear, but it may be related to the use of different methods for normalizing the CAT activity between transfection experiments.
DNase footprinting, methylation interference, and mobility shift assays indicate a nuclear factor that binds in a sequencespecific manner to a region that extends from position -55 to -31. Using synthetic oligonucleotides with transverse mutations, the recognition sequence was determined to contain a core domain of 9 base pairs. The core recognition sequence (underlined) with flanking nucleotides GGAT[CAGC(TG]--ACTCG) contains the overlapping recognition sequences AP-4 [CAGCTG] and AP-1 (TGACTCG). A close association between AP-4 and AP-1 binding sites is not unique and has been reported in other eukaryotic promoters (Comb et al., 1988;Gabudza et al., 1989;Mermod et al., 1988).
In estimating the contribution of these two sequence elements to the factor binding described here, the following points should first be considered. Mutating the first 3 base pairs of the 9-base pair core domain from TCA to GAC completely eliminates the AP-4 consensus sequence but leaves the AP-1 consensus sequence intact. Nevertheless, factor binding is completely abolished, suggesting that the binding factor is not AP-1. Furthermore, competition of factor binding to the wild type sequence is not abolished with a 100-fold molar excess of an oligonucleotide containing the AP-1 consensus sequence (Fig. 8), and mutating the 5"flanking sequence GGA to TTC has a profoundly reducing effect on binding activity to this promoter fragment. This sequence element is not part of the recognition sequence for AP-4, and on that basis alone it is questionable whether the binding factor described here is AP-4. Further evidence that AP-4 is not involved in the binding to this APP promoter region is obtained from mobility shift experiments using an oligonucleotide that contains an AP-4 binding site as part of the native SV40 enhancer (Mermod et al., 1988). This fragment generates a mobility shift pattern that is different from the pattern obtained with the APP promoter fragment. In addition, the SV40 enhancer fragment does not compete with the binding to the APP promoter fragment (Fig. 8 B ) .
These experiments indicate that the APP promoter binding factor is distinct from both AP-1 and AP-4. However, other DNA binding factors that share the helix-loop-helix motif of AP-4 (Hu et al., 1990) include Pan (Nelson et al., 1990), the immunoglobulin gene enhancer binding proteins (Murre et al., 1989), muscle-specific differentiation factors (Lassar et al., 1989;Wright et al., 1989;Braun et al., 1989), and DFOsophila regulatory factors (Caudy et al., 1988;Murre et al., 1989). These factors have very similar core recognition sequences, and it is conceivable that the binding factor observed here belongs to the same class of binding factors. Alternatively, the APP promoter binding factor may be unrelated to AP-1 and AP-4. For example, two factor binding sites, which are near each other within the enhancer of the proenkephalin gene, are required for the transcriptional response to CAMP and phorbol ester. These binding elements contain overlap-ping recognition sequences with AP-1 and AP-4; however, they are also recognized by binding factors related to NF-1 (Chu et al., 1991;Comb et al., 1988).
The interpretation of binding and transient transfection studies seems to suggest that the nuclear factor described here is a transcriptional activator, since deletion 6 at position -49 both interrupts the binding domain and displays diminished promoter activity. However, the experiments do not preclude the possibility that the binding factor acts as a negative regulator that modifies transcription by a positive factor. If this is the case, the sequence elements responsible for transcriptional activation must nevertheless be located near the recognition sequence of the binding factor described here, since deleting the promoter from position -96 to -59 profoundly reduces its expression. Further studies are required to characterize this APP promoter binding factor and determine its precise functional significance.