Transcriptional regulation of the mouse presenilin-1 gene.

The presenilin-1 (PS-1) gene encodes at least three separate mRNA transcripts from its 12 exons, which are spread over 50 kilobase pairs of mouse DNA. The first transcript begins with exon 1A, whereas the other transcripts begin with exon 1B. Different portions of exon 1B are spliced to give long and short mRNAs. The expression of all of these transcripts depends on a single promoter located just upstream of exon 1A. Although this region lacks a TATA box and a number of common initiator sequences, it does contain a CAAT box, a heat-shock responsive element, a polyomavirus enhancer activator-3 site, an Ets 1-3 site, and multiple-Sp1 and multiple-Ap2 binding sites, which are typically found in eukaryotic promoters. We have combined a reporter gene with various portions of this putative PS-1 promoter and measured firefly luciferase activity relative to an internal renilla luciferase standard. We identified a 25-base pair fragment spanning the 5'-transcription start site of exon 1A as containing the core of the promoter activity. The sequences downstream of this region had undetectable promoter activity, suggesting that this core element is the gene's only promoter, and it controls expression of all three transcripts. Although human PS-1 mRNA expression is clearly different from the mouse PS-1 mRNA pattern, the human and mouse core promoters do share limited homology.

The presenilin-1 (PS-1) gene encodes at least three separate mRNA transcripts from its 12 exons, which are spread over 50 kilobase pairs of mouse DNA. The first transcript begins with exon 1A, whereas the other transcripts begin with exon 1B. Different portions of exon 1B are spliced to give long and short mRNAs. The expression of all of these transcripts depends on a single promoter located just upstream of exon 1A. Although this region lacks a TATA box and a number of common initiator sequences, it does contain a CAAT box, a heatshock responsive element, a polyomavirus enhancer activator-3 site, an Ets 1-3 site, and multiple-Sp1 and multiple-Ap2 binding sites, which are typically found in eukaryotic promoters. We have combined a reporter gene with various portions of this putative PS-1 promoter and measured firefly luciferase activity relative to an internal renilla luciferase standard. We identified a 25-base pair fragment spanning the 5-transcription start site of exon 1A as containing the core of the promoter activity. The sequences downstream of this region had undetectable promoter activity, suggesting that this core element is the gene's only promoter, and it controls expression of all three transcripts. Although human PS-1 mRNA expression is clearly different from the mouse PS-1 mRNA pattern, the human and mouse core promoters do share limited homology.
Alzheimer's disease is a devastating neurological disorder and the most common cause of dementia. The genetics of this disorder suggests that multiple genes are involved. To date, mutations in four genes have been found to be associated with Alzheimer's disease phenotypes including the amyloid precursor protein gene on chromosome 21 (1,2), the apolipoprotein E gene on chromosome 19 (3)(4)(5), the presenilin-1 (PS-1) 1 gene on chromosome 14 (6), and the presenilin-2 (PS-2) gene on chromosome 1 (7). Although an unknown gene on chromosome 12 appears to associate with a large percentage of late-onset Alzheimer's patients (8), the majority of familial Alzheimer's disease cases are associated with mutations in the PS-1 gene. To date, over 30 independent mutations in PS-1 have been described in unrelated Alzheimer's families displaying an early age-of-onset phenotype. Most of these mutations are missense mutations that result in single amino acid changes (6, 9 -15). Deletions found in exon 4 and exon 9 cause additional mutations as do several truncations of the RNA transcripts arising through differential splicing (16). Although clustering of these mutations within the protein suggests the location of functionally important domains, the exact function of presenilin proteins is a matter of active investigation.
One approach to find gene function is to study the regulation of PS-1 gene expression. Using in situ hybridization, we and others demonstrate that PS-1 mRNA is most highly expressed in neurons of the brain (17). 2 Immunohistochemistry revealed that the PS-1 protein was abundant in neurons, but was also associated with amyloid plaques and some glial cell types (18,19). In contrast, Sherrington et al. (6) reported that PS-1 mRNA is widely expressed in a variety of organs throughout the body. This raises the question why mutations in the PS-1 gene product appear to confer a disease state in familial Alzheimer's patients without apparent effect on their peripheral organs. The situation is further compounded because PS-1 mRNA and protein levels have not been reported thus leaving open the possibility that pathological regulation of PS-1 gene expression in familial Alzheimer's disease patients, compared with age-matched healthy controls, contributes to the disease state.
Mutations in the PS-1 gene's promoter and non-protein encoding regions are not known and reports on the gene's wildtype sequence are lacking. Similarly, no functional analysis of the gene's ability to promote transcription has been reported. Combined with recent reports that PS-1 knockout mice are embryonic lethal (20), knowledge of the PS-1 gene sequence and its transcriptional regulation may be important clues that help to identify PS-1 function in both normal and diseased states. To begin answering some of these questions, we now report a detailed sequence of the mouse PS-1 gene including its promoter, the entire protein encoding region, and ending in exon 12. In addition to the usefulness of this sequence in creating additional PS-1 knockout mice, we have defined the PS-1 promoter's transcriptional regulatory elements. Our data on promoter activity suggests that the presenilin-1 gene is preferentially expressed and transcribed in neurons.

EXPERIMENTAL PROCEDURES
Isolation and Characterization of Genomic Clones-Labeled oligonucleotides and polymerase chain reaction (PCR) products of the mouse PS-1 cDNA were used as probes to screen mouse libraries for genomic PS-1 clones. Based on the mouse PS-1 cDNA sequence (GenBank TM accession no. L42177), an upstream primer of sequence 5Ј-CG-GAGAGAGAAGGAACCAAC-3Ј and a downstream primer of sequence 5Ј-TCAGCTCTTCGTCTTCCTCCTCATC-3Ј were used, with Quick Clone Mouse Brain cDNA (CLONTECH) as template, to amplify a portion of the mouse PS-1 cDNA by PCR. Amplification reactions were performed in a 100-l volume containing 1 ϫ PCR buffer II (Perkin-Elmer), MgCl 2 (1.5 mM), dATP, dGTP, dCTP, and dTTP (0.2 mM each, Perkin-Elmer), DNA primers at 0.5 M, 1 l of cDNA template (0.1 ng), and Ampli-Taq DNA polymerase (5 units, Perkin-Elmer). The reaction cycle was 95°C for 1 min, 50°C for 1 min and 72°C for 2 min for a total of 30 cycles. This PCR product was gel-purified and labeled with [␣-32 P]dCTP and a Random Primers DNA labeling system (Life Technologies, Inc.). Labeled probe was used to conventionally screen a mouse strain 129/SVJ genomic library in Lambda Fix-II vector (Stratagene) as described previously (21). Screening identified four independent phage clones designated Ph-1, Ph-2, Ph-3, and Ph-4 (see Fig. 2). Following digestion with NotI and/or EcoRI, restriction enzyme fragments were subcloned into pBluescript-II-KS(ϩ) phagemid vector (Stratagene) using a DNA ligation kit (Stratagene). DNA sequence was determined using an Applied Biosystems model 373A automated DNA sequencer with dye terminator chemistry and protocols recommended by the manufacturer. Additional oligonucleotide probes from the 5Јuntranslated region of the mouse PS-1 cDNA were labeled with [ 32 P]ATP and T4-polynucleotide kinase and used to identify plasmid subclones by hybridization. Based on the partial sequence of phage clone Ph-2, the PCR primers 1C-US-for (GATCACAGTCTAGGTTGCT-GGTGTG) and 1C-US-rev (TGGGGCAAGGGACACAAATAAG) were used to further screen a mouse ES-129/SVJ genomic library in a P1 vector (Genome Systems Inc.) by PCR. Of the three P1 clones identified, P1-10809 was digested with EcoRI or HindIII, and these restriction enzyme fragments were subcloned and sequenced as described above.
Computation of Sequence Similarities-Comparison of the mouse PS-1 promoter with other eukaryotic promoter sequences was performed using the BLAST network service and the eukaryotic promoter data base Release 45 available from the National Center for Biotechnology Information.
Construction of PS-1 Promoter Firefly Luciferase Reporters-Mouse genomic DNA fragments containing portions of the putative PS-1 promoter were subcloned upstream of the firefly luciferase gene into the promoterless pGL3-basic vector (Promega). Based on the sequence of genomic DNA, PCR primers were designed to incorporate XhoI sites into the forward primers and HindIII sites into the reverse primers. Corresponding to different locations in the genomic DNA (see Fig. 2), these primers were used to PCR amplify fragments which were purified with a Wizard purification system (Promega), digested with the appropriate restriction enzymes and repurified with the Wizard kit. Cleaved PCR products were ligated into pGL3 plasmid cleaved with the same restriction enzymes and transformed into competent bacteria, and clones containing plasmids with inserts were verified by DNA sequencing.
For transient transfection, Neuro2a, P19, retinoic acid-treated-P19, dimethyl sulfoxide-treated P19, and NIH/3T3 cells were plated in sixwell tissue culture dishes at 9 ϫ 10 4 cells/well and allowed to recover for 1 day. Cells containing PS-1-promoter/reporter constructs were then co-transfected with 0.3 pmol of one of the promoter firefly luciferase plasmid constructs, pGL3 basic vector or pGL3 promoter plasmid (which contains an SV40 promoter upstream of the firefly luciferase gene, Promega), and 0.3 pmol of pRL-TK plasmid (which contains a herpes simplex virus thymidine kinase promoter upstream of the renilla luciferase gene, Promega), using the Lipofectin procedure (Life Technologies, Inc.) as described in the manufacture's protocol.
Relative Luciferase Activity Measures-Transfected cells were cultured for 24 h, washed twice with 2 ml of Ca 2ϩ -and Mg 2ϩ -free phosphate-buffered saline, and lysed with Passive lysis buffer (Promega). Firefly luciferase and renilla (sea pansy) luciferase activities were measured sequentially using a Dual-Luciferase Reporter assay system (Promega) and a model TD-20E Luminometer (Turner Design). After measuring the firefly luciferase signal (LA F ) and the renilla (sea pansy) luciferase signal (LA R ), the relative luciferase activity (RLA) was calculated as: RLA ϭ LA F /LA R , where relative RLA was calculated as a percentage, i.e. %RLA ϭ RLA/(RLA) max . To compare the relative luciferase activity in one cell line with another, an index of relative luciferase activity was calculated as: IRLA ϭ RLA/RLA SV40 , where RLA SV40 is the ratio of firefly luciferase signal with an SV40 promoter in pGL3 divided by the renilla luciferase signal in pRL-TK.

RACE Detects
Multiple Transcripts-As a prelude to cloning the PS-1 promoter, the exact 5Ј-end of PS-1 mRNA from mouse brain was identified by the RACE technique. Using the antisense oligonucleotide "101-80-reverse" found in exon 2 of mouse PS-1, 5Ј-RACE gives a major broad band of 210 bp and a minor band of 430 bp from single-stranded cDNA templates complementary to mouse brain mRNA (Marathon Ready cDNA, CLONTECH, data not shown). Each of these bands was isolated from agarose gels, subcloned into the pGEM-T vector (Promega) and sequenced. Sequencing revealed the presence of three different PS-1 transcripts which appear to derive from two unique transcriptional start sites (Fig. 1). This information suggests that the PS-1 gene may contain two promoters and that differential splicing of exons, should they exist in genomic DNA, might generate multiple transcripts.
Isolation of the Mouse PS-1 Gene-A mouse genomic DNA library in Lambda FIX-II was screened with the 5Ј-portion of the mouse PS-1 cDNA ( Fig. 2A, probe A). Of the positively FIG. 1. Structure of three different presenilin-1 transcripts from mouse brain. DNA sequencing of the cloned products of 5Ј-RACE of mouse brain cDNA revealed the presence of three independent transcripts (A, B, and C), which appear to derive from two unique transcription start sites marked by the vertical arrows. The distance between the two transcription start sites is 410 bp. The sizes of exon 1A, exon 1B, and exon 1C are 141, 371, and 139 bp, respectively. hybridizing phage clones, four were selected for restriction mapping with EcoRI and NotI as shown in Fig. 2B. Only one phage clone, Ph-2 hybridized to oligonucleotides from the 5Јuntranslated region of the mouse PS-1 cDNA. Primers from the phage arms allowed sequencing into the genomic DNA insert. The insert's sequence allowed the PCR primer pair "1C-USforward and 1C-US-reverse" to be chosen and used to identify a P1 clone of the mouse PS-1 genomic DNA as shown in Fig. 2A. Clone P1-10809 was identified through a positive PCR reaction product with these primers and hybridization to the PS-1 cDNA fragment probe A (Fig. 2). P1-10809 was then restricted, mapped and its entire sequence subcloned into multiple pBluescript-II-KS-(ϩ) plasmid vectors as shown by the thick lines in Fig. 2B. Each subclone was sequenced on an Applied Biosystems 373A automated DNA sequencing machine using protocols supplied by the manufacturer.
Characterization of the PS-1 Gene's Exon-Intron Structure-The sequence of almost 50 kbp of the P1-10809 clone was compared with the mouse PS-1 cDNA sequence and regions of homology aligned with the MacVector DNA analysis program (IBI, New Haven, CT). The first nucleotide on the 5Ј end of the RNA transcript is usually designated as nucleotide ϩ1 of exon 1. In our case, PS-1 appears to have two different 5Ј-endsequences which are associated with three different length RNA transcripts (Fig. 1, transcript-A, transcript-B, and transcript-C). The alignment of transcript-A with genomic sequence shows that a "G," designated conventionally as position ϩ1 in exon 1, corresponds to the transcription start site. The presenilin gene's exon 1A extends from position ϩ1 to ϩ141, which is spliced to exon 2, whose 5Ј-end begins at gene position number ϩ11,210, to give transcript A. The alignment of transcript B with genomic DNA shows that a "C" at position ϩ411 corresponds to the alternative transcription start site. We define this second start site as beginning in exon 1B, which extends from position ϩ411 to ϩ781 and is spliced to exon 2 (gene position number ϩ11,210) to give transcript B. The alignment of transcript C with genomic DNA shows the same "C" at position ϩ411 as the alternative transcription start site followed by a shorter sequence alignment. We define exon 1C as extending from position ϩ411 to ϩ549 which is spliced to exon 2 (gene position number ϩ11,210) to give transcript C. Thus, the 5Ј-untranslated regions of PS-1 RNA transcripts include exon 1A, exon 1B, exon 1C, and exon 2 together with a portion of exon 3 ( Fig. 2C and Table I). The protein-encoding portions of the gene begin with the ATG codon at gene position number ϩ11,420 where translation initiates in exon 3 followed by exons encoding the remainder of the protein until stopping at a TAG codon in exon 12 (position ϩ45,627). 983 bp downstream from this TAG stop codon lies the putative polyadenylation signal AATTAA at position ϩ46,612. Interestingly, the Intron between exon 1 and exon 2 is about 10 kbp, between exon 3 and exon 4 is about 12 kbp and between exon 5 and exon 6 is about 8 kbp, whose sequences, together with the rest of the PS-1 exons and introns, can be found in GenBank TM (accession no. AF007560). Characterization of the Mouse Presenilin-1 Promoter-The DNA located upstream and surrounding the transcription initiation sites typically confers the gene's promoter activity. When the DNA surrounding the transcript A initiation of transcription site was compared with its human PS-1 genomic DNA counterpart (Fig. 3), 3 the region of maximal similarity extends from positions Ϫ39 to ϩ117 of the mouse PS-1 sequence. This region is rich in guanosine (G) and cytosine (C) residues containing the sequence motifs: GCCGGAAGT resembling an Ets 1-3 element (30) and GGGCGGG motif resembling an Sp-1 hexanucleotide element, which is commonly found in the promoters of other genes. The mouse sequences upstream of this region do not share similarity with the human sequence nor do they contain the most common eukaryotic promoter element, a TATA box (Fig. 4). Instead, this unique mouse sequence contains two CAAATA motifs, at positions Ϫ365 and Ϫ281, which resemble CAAT boxes found in other eukaryotic promoters. This region unique to mouse also contains an Ap-2 binding element at position Ϫ80 (CCCAGCCC) and a sequence similar to a heat shock inducible element at position Ϫ220 (CTC-GAATCGCAG). Putative Sp-1 hexanucleotide binding sites with the sequence GGGCGG or CCGCCC are found downstream from the ϩ1 site of transcription initiation with exon 1A containing two Sp1 motifs and the exon 1A/exon 1B intron containing five Sp1 motifs. Also downstream of the Cap (transcription initiation) site are two additional Ap-2 sites and another Ets 1-3 motif.
We employed a Dual-Luciferase reporter assay system (Promega) to test whether these elements function to promote transcription. In general, we assayed the promoter activity of DNAs flanking the transcription initiation site of PS-1 by inserting these DNA fragments in front of a basic, promoterless firefly luciferase reporter gene in plasmid pGL3. Constant amounts of pGL3 containing PS-1 promoter fragments and of pRL-TK plasmid containing a herpes simplex virus thymidine kinase promoter driving expression of sea pansy luciferase, were co-transfected into a constant number of cells. After 24 h, lysates of transfected cells were sequentially assayed for firefly luciferase (LA F ) and sea pansy luciferase activity (LA R ) so that a ratio of firefly to sea pansy activities could be calculated for each PS-1 promoter fragment as its relative luciferase activity or RLA. Of all the fragments tested, plasmid LUC 29 with the fragment Ϫ327 to ϩ206 showed the greatest ratio of firefly to sea pansy activity ( Fig. 5 and Table II), which we defined as 100% activity. Larger fragments in LUC 1 (Ϫ2232 to ϩ1436), LUC 3 (Ϫ499 to ϩ1171), and LUC 16 (Ϫ276 to ϩ519) display only a small percentage of the LUC 29 activity, suggesting the presence of negative elements that apparently reduce their activities. Interestingly, the high activity of LUC 29 is not found in its flanking fragments such as LUC 2 (Ϫ2232 to Ϫ496) and LUC 23 (ϩ188 to ϩ519), which both lack significant promoter activity. The LUC 23 result is particularly interesting because the alternative transcription start site begins at position ϩ411 of exon 1B/exon 1C and apparently lacks meaningful promoter activity.
To more accurately define the minimal or core regions conferring promoter activity, we studied the Ϫ327 to ϩ206 region of the PS-1 gene in greater detail. Sequence comparison showed this region to contain a CAAT box (Ϫ281), a heat shock element (Ϫ220), an AP2 site (Ϫ80), a PEA-3 site (Ϫ53), an Ets 1-3 site (Ϫ7), and Sp1 sites (ϩ25, ϩ119, and ϩ161). To find which of these elements and/or new elements were functionally active, we resected this region and tested smaller fragments for promoter activity. Since LUC 24 and LUC 23 lacked significant activity, we initially focused on the fragments from Ϫ440 to ϩ91 as shown in Fig. 5 3. Comparison of the mouse and human presenilin-1 promoters. Mouse PS-1 transcription begins with "G" at position ϩ1 of exon 1A and human PS-1 transcription begins with "A" (data not shown.). By DNA sequence similarity searching with BLAST network service available from National Center Biotechnology Information, regions of mouse/human homology are found around the transcription initiation sites for both genes. Consensus binding sites for the transcription factors ETS1 and SP1 are underlined and are conserved in both mouse and human genes. less activity than LUC 6 (Ϫ327 to ϩ91) which contains this CAAT box. A negative element must reside upstream of this CAAT box because the activity of LUC 4 (Ϫ440 to ϩ91) is about half that of LUC 6 (Ϫ327 to ϩ91). The heat shock element at Ϫ220 may not play a role in PS-1 promoter activity as fragments containing (LUC 8, Ϫ261 to ϩ91) and lacking (LUC 10, Ϫ192 to ϩ91) this element have similar activities. The AP2 site at Ϫ80 and/or the PEA-3 site at Ϫ53 appear to play positive roles in PS-1 promoter function as LUC 12 (Ϫ87 to ϩ91) has about 4-fold more activity than LUC 13 (Ϫ32 to ϩ91) which lacks these sites. Similarly, the Ets 1-3 site at position Ϫ7 plays a positive role as judged by the RLA activity of LUC 14 (Ϫ9 to ϩ91) at 7.9% and LUC 15 (ϩ42 to ϩ91) at 0.7%. While the Sp1 site at position ϩ161 does not appear to contribute when LUC 17 (Ϫ276 to ϩ206) is compared with LUC 18 (Ϫ276 to ϩ148), the Sp1 sites at ϩ25 and ϩ119 appear to be very active in the PS-1 promoter as negative and positive elements, respectively, when LUC 26 (Ϫ9 to ϩ41) is compared with LUC 27 (Ϫ9 to ϩ16) and LUC 7 (Ϫ276 to ϩ91) is compared with LUC 18 (Ϫ276 to ϩ148).
Based on these experiments, we tested whether the region from Ϫ87 to ϩ41 could contain the core promoter activity in two ways. First, LUC 25 (Ϫ87 to ϩ41) had an RLA promoter activity of 28%. Second, the deletion of this region to give LUC 30 (delete Ϫ87 to ϩ41 from Ϫ327 to ϩ206) decreased activity from 100% (LUC 29) to 0.2% (LUC 30). Taken together, these results strongly suggest that the Ap2, PEA-3, Ets 1-3, and Sp1 elements comprise the major functional elements of the PS-1 promoter in the region Ϫ87 to ϩ41.
Cell-specific Transcription-Using in situ hybridization to human brain slices, we found that PS-1 RNA was most abundant in neurons and below the limits of detection in other brain cells. 2 This result suggested that the PS-1 promoter may preferentially function in neurons. To test this idea further, we compared the activity of the promoter-fragment/reporter plasmids LUC 1, LUC 3, LUC 4, LUC 27, and LUC 29 in different cell types. As reported above, the mouse Neuro-2A cell line of neuroectodermal lineage supports more RLA promoter activity from LUC 29 and LUC 4 than from LUC 27, LUC 3, and LUC 1 ( Fig. 6 and Table II). In contrast, the mouse NIH/3T3 fibroblast cell line supports only minimal promoter activity with each of these promoter/reporter constructs (LUC 29, LUC 27, LUC 4, LUC 3, or LUC 1). To further test the idea that the PS-1 promoter activity is great in neurons, we transfected the mouse embryonal carcinoma cell line P19 with these reporter constructs. P19 cells are uniquely differentiated by all-trans-retinoic acid treatment into a neuron-like phenotype (22) or by dimethyl sulfoxide treatment into a muscle-like phenotype (23). Retinoic acid-treated P19 cells support as much as 2.5-fold more relative luciferase activity from plasmid LUC 29 compared with untreated P19 cells. Untreated P19 cells support as much as 1.3-fold more relative luciferase activity compared  5. Mouse presenilin-1 promoter-reporter constructs and their relative luciferase activity (%RLA). A, structural organization of PS-1 promoter. Top line represents the region of the PS-1 gene which was analyzed for promoter activity where boxes for exon 1A and exon 1B are Mouse Presenilin-1 Promoter Activity 23494 with dimethyl sulfoxide-treated P19 cells. These results are consistent with the hypothesis in which PS-1 transcription is preferred in neuron-like cells.

DISCUSSION
From promoter to polyadenylation signal, the full sequence of the mouse presenilin-1 gene and its exon-intron structure set the stage to describe some of its unique functions. In contrast to the reported PS-1 cDNA sequence, 5Ј-RACE surprised us by amplifying three different mRNA transcripts which share two unique transcription start sites. Sequence analysis showed that transcript A begins with exon 1A while transcript B and transcript C begin with exon 1B. Exon 1C is a fragment of exon 1B sharing its 5Ј-end at position ϩ411, but only extending to position ϩ549. This example of alternative splicing in exon 1B versus exon 1C yields multiple RNA transcripts and has also been described for exon 9 in the human PS-1 gene (16). Two distinct transcription start sites, however, have been reported for only a few genes including human catechol-O-methyl transferase (24), mouse neurotrophin-3 (25), and rat aromatic Lamino acid decarboxylase (26). In these cases, each transcriptional start site was associated with a distinct promoter so that a stoichiometry of one promoter per transcription start site was observed.
Our characterization of promoter activities for the PS-1 gene, however, revealed a much different picture. Using a promoterfragment coupled to the firefly luciferase reporter with sea pansy Renilla luciferase as an internal standard, we found that the Ϫ327 to ϩ206 fragment (LUC 29) contains most of the PS-1 promoter activity. The known sequence motifs which apparently contain this activity are a CAAT box (Ϫ281), an Ap2 site (Ϫ80), a PEA-3 site (Ϫ53), an Ets 1-3 site (Ϫ7), and an Sp1 site (ϩ25). While this region overlaps some of exon 1A, deletion of the Ϫ87 to ϩ41 region in LUC 30 reduces promoter activity by 50-fold. To measure promoter activity around the alternative transcription start site, we tested LUC 23 (ϩ118 to ϩ519) containing Sp1, Ap2, and Ets 1-3 sites and found these exon 1B/exon 1C sequences to confer about 1% of the activity surrounding the exon 1A promoter. These results suggest to us that the region surrounding the ϩ1 position of exon 1A may promote the expression of transcript A, transcript B, and transcript C. Alternatively, a weak promoter controlling transcription initiation at position ϩ410 in exon 1B/exon 1C may amount to only 1% of the transcription initiation at position ϩ1. By cloning all of the products of the 5Ј-RACE into plasmid vectors and counting each clone carrying exon 1C, we estimate the abundance of transcript C to approach 30% of all of the PS-1 transcripts (data not shown), further supporting the idea that the major promoter at ϩ1 functions to control transcription initiation from both the ϩ1 and the ϩ410 sites. Quantitative measurement of transcript A, transcript B, and transcript C levels will help to further resolve this issue. The high homology between human and mouse promoters combined with our description of multiple start sites and alternative splicing for the mouse PS-1 gene reasonably suggests how the human PS-1 promoter may function.
Recently, PS-1 was reported to be expressed predominantly in neurons of the central nervous system (17). This result matches our own data that PS-1 RNA, by in situ hybridization, is strongly expressed in neurons and at undetectable levels in other cell types. 2 Similarly, several immunohistochemical studies report primarily neuronal localization of PS-1 protein with weak staining of amyloid plaques and some glia surrounding those plaques. On the other hand, Sherrington et al. (6) showed that Northern blots of RNA from different organs all hybridized to a PS-1 cDNA probe, suggesting that PS-1 RNA is ubiquitously expressed. At present, these results can not be easily reconciled.
While not rigorous proof, our data clearly shows preferential promoter activity in neuron-like cells supporting a cell-typespecific pattern of PS-1 expression. We find the greatest amount of PS-1 promoter activity in the mouse Neuro2a neuroblastoma cell line, followed by the P19 embryonal carcinoma cell line and almost no activity in the mouse NIH/3T3 fibroblast cell line (Table II). To further confirm this finding, we employed the P19 mouse embryonal carcinoma cell line because of its unique ability to be differentiated into a muscle-like phenotype following dimethyl sulfoxide treatment or into a neuron-like phenotype (P19-RA-neuron) following all-trans-retinoic acid treatment (22,23). If our hypothesis that PS-1 promoter activity is preferred in neuron-like cells, then we would predict that P19 cells differentiated with retinoic acid into neuron-like cells would display more PS-1 promoter activity than P19 cells differentiated with dimethyl sulfoxide into muscle-like cells. As clearly shown in Fig. 6 and Table II, P19-RA-neuron cells display the most PS-1 promoter activity followed by untreated P19 cells and the least activity in P19 dimethyl sulfoxidetreated muscle cells. These results, combined with the Neuro2a and NIH/3T3 results, indicate a clear pattern of PS-1 promoter activity which is preferred in neurons.
The mechanisms controlling neuron-specific promoter activity are poorly understood. The most direct mechanism would be for a positive regulator, that is only present in neuronal cells, to singularly activate the neuron-specific promoter. Alternatively, a negative regulator, that is only present in non-neuronal cells, could globally repress the neuron-specific promoter in all but labeled. Open boxes represent genomic DNA fragments corresponding to the mouse PS-1 gene (top line) which were cloned upstream of the firefly luciferase reporter gene in the plasmid pGL3-Basic (Promega). Open boxes are labeled on the left with the name of the promoter-reporter plasmid as LUC no. and with a nucleotide number of the 5Ј-end of the fragment based on ϩ1 being the "G" at the beginning of exon 1A. Letters above the open box refer to a restriction enzyme cleavage site. Numbers to the immediate right of the open box denote the 3Ј-end of the promoter fragment. The numbers on the left-hand side are the percentage of relative luciferase activity calculated as described under "Experimental Procedures" followed by the number of times that construct has been transfected into cells and its activity measured, which is in parentheses. B, Fine structure map of the PS-1 promoter and promoter-reporter construct activity strategy. Top line represents the region of the PS-1 gene with putative promoter elements, exon 1A, exon 1B, and restriction enzyme site positions labeled. Open boxes of promoter-reporter constructs are as in A. Letters above the open box refer to the end of the promoter fragment shown in Fig. 3.

TABLE II
Neuron-preferred activity of total and core promoter regions of the mouse PS-1 gene LUC29 and LUC27 were transiently transfected into the cell lines to measure the activities of total-and core-promoter, respectively. An SV40-promoter driving firefly luciferase in pGL3-basic plasmid (Promega) was also transfected as a control. An IRLA was calculated for the total-and the core-promoter as IRLA ϭ RLA/RLA SV40 . neuronal cells. Depending upon the exact DNA elements within the promoter, some combination of positive and negative controls of transcriptional activity might also yield neuron-preferred promoter function. Going beyond our characterization of the regions conferring PS-1 promoter activity in Neuro2a cells, we may now look at the data to suggest which of the DNA elements might confer neuron-preferred promoter function. The region showing the highest activity in Neuro2a neuron-like cells extends from Ϫ329 to ϩ206 (LUC 29) and contains a CAAT box at Ϫ281, a heat-shock inducible element at Ϫ218, an Ap2 site at Ϫ80, a PEA-3 site at Ϫ53, an Ets 1-3 site at Ϫ7, and an Sp1 site at ϩ25. The CAAT box could be the source of about a third of the positive control of neuron-specific activity as its deletion reduces promoter activity by about a third when comparing LUC 6 (Ϫ327 to ϩ91) with LUC 8 (Ϫ261 to ϩ91, Fig. 4). The heat shock element probably does not contribute to neuron-specific activity as its deletion does not affect promoter activity when comparing LUC 8 (Ϫ261 to ϩ91) to LUC 10 (Ϫ192 to ϩ91, Fig. 4). Based on the 4-fold greater activity of LUC 12 (Ϫ87 to ϩ91) compared with LUC 13 (Ϫ32 to ϩ91), it appears that both the Ap2 site and the PEA-3 site are good candidates for the positive control of neuron-specific promoter function. Ap2 sites are reported to be most frequently found in promoters active in cells of neural crest lineage, and several examples exist of their involvement with neuron-specific activity (27)(28)(29). In contrast, the 5-fold less activity of LUC 26 (Ϫ9 to ϩ41) compared with LUC 27 (Ϫ9 to ϩ16) implicates the Sp1 site at ϩ25 as a negative regulator of neuron-specific promoter function. These same data could also be interpreted as the Ets 1-3 site having a positive function, possibly as part of a core promoter element from Ϫ9 to ϩ16. Direct measurement of LUC 27 (Ϫ9 to ϩ16) shows that Neuro2a and P19-RA-neuron cells have more activity than do P19 dimethyl sulfoxide-treated muscle or NIH/3T3 non-neuronal cells supporting the idea that this 25-bp region contributes to neuron-preferred promoter activity. The major transcription start site at position ϩ1 is FIG. 6. Cell type-specific PS-1 promoter activity. PS-1 promoter-reporter constructs LUC 29, LUC 27, LUC 4, LUC 3, and LUC 1 were transiently transfected into Neuro2a neuroblastoma (N2a), undifferentiated P19 (P19), all-trans-retinoic acid differentiated neuron-like P19 (P19N), dimethyl sulfoxide-differentiated muscle-like P19 (P19M), and NIH/3T3 fibroblast cells. An SV40 promoter driving firefly luciferase in pGL3-basic plasmid (Promega) and pRL-TK (thymidine kinase promoter driving renilla luciferase gene) was also transfected into each cell line as external and internal controls, respectively. After measuring luciferase activity from all combinations of plasmids, the IRLA was calculated as RLA/RLA SV40 , where RLA SV40 is the ratio of firefly luciferase signal in the external control divided by the renilla luciferase signal in the internal control, to compare the activity of different promoter fragments in different cell lines. Plasmid LUC 29 transfected into N2a cells showed the greatest IRLA value which we defined as 100% activity. located in this proposed core promoter element. The ETS-1 transcription factor prefers binding to the Ets 1-3 binding site found in this core by a ratio of five to one over the PEA-3 binding site (30). This finding is particularly interesting as the ETS-1 transcription factor is thought to be specific for B cells and resting T cells of the immune system and not been previously described for neuronal cells. Sp1 binding sites appear to be ubiquitously distributed in all promoters of all cell types and their ability to function as negative elements appears to be novel.
In summary, we described a complete sequence of the mouse presenilin-1 gene from its tip to its tail. This sequence has shown us that there are two independent transcription start sites. Functional testing of the DNA regions surrounding these start sites showed that they both were apparently controlled by a single, major promoter that includes the ϩ1 position of exon 1A. This promoter was also quite interesting because it is mostly active in neuron-like cells. Further characterization can now progress to a complete description of those positive and negative DNA elements and transcription factors which function to control presenilin-1 gene expression.