Identification of cis-Elements Regulating Exon 15 Splicing of the Amyloid Precursor Protein Pre-mRNA*

Alternative splicing of exon 15 of the amyloid precursor protein (APP) pre-mRNA generates two APP isoform groups APPex15 (containing exon 15) and L-APP (without exon 15), which show a cell-specific distribution in non-neuronal cells and neurons of rat. Both APP isoforms differ in regard to functional properties like post-translational modification, APP secretion, and proteolytic production of Aβ peptide from APP molecules. Since Aβ generation is an important factor in the development of Alzheimer's disease, one could anticipate that these major APP isoforms might contribute differentially to the mechanisms underlying neurodegeneration in Alzheimer's disease. In this study, we established an APP minigene system in a murine cell system to identifycis-acting elements controlling exon 15 recognition. A 12.5-kilobase pair genomic fragment of the murine APP gene contained all cis-regulatory elements to reproduce the splicing pattern of the endogenous APP transcripts. By using this approach, two intronic cis-elements flanking exon 15 were identified that block the inclusion of exon 15 in APP transcripts of non-neuronal cells. Point mutation analysis of these intronic regions indicated that pyrimidine-rich sequences are involved in the splice repressor function. Finally, grafting experiments demonstrated that these regulatory regions cell-specifically enhance the blockage of a chimeric exon in the non-neuronal splicing system.

Amyloid plaques represent characteristic neuropathological lesions in brains of patients with Alzheimer's disease. These amyloid deposits are mainly composed of a short, highly hydrophobic peptide (40 -42 amino acid residues) termed A␤ (1,2). A␤ is derived by proteolytic processing of the amyloid precursor protein (APP), 1 which is expressed abundantly in almost every cell type including neurons (3,4). The human APP gene is located on chromosome 21 (5,6); it spans ϳ300 kb, and the coding region is distributed over 18 exons (7). The APP pre-mRNA molecules are targets for post-transcriptional processing by alternative splicing of exons 7 (6,8,9), 8 (9), and 15 (10) generating eight different isoforms of a type I transmembrane glycoprotein ranging in length from 677 to 770 amino acids.
Alternative splicing of APP exon 15 was first discovered in peripheral mononuclear leukocytes and activated microglial cells (10). Therefore, isoforms lacking exon 15 were denoted as leukocyte-derived APP (L-APP). Detailed expression studies revealed a ubiquitous distribution of L-APP transcripts in rat tissues (11). In peripheral tissues L-APP mRNA molecules constitute between 25 and 72% of total APP transcripts. In contrast, the expression levels of L-APP in the central nervous system, including hippocampus, cerebellum, and cerebral cortex, is low (Ͻ10%), and no significant L-APP amounts were detected in cultured primary rat neurons. These findings suggested that exon 15 is recognized by a cell type-specific mechanism in rat.
Exclusion of 18 amino acids encoded by exon 15 has important consequences for the functional properties of the corresponding APP isoforms. Fusion of exon 14 with exon 16 creates a new post-translational modification motif ENEGSG. This motif is recognized by xylosyltransferase that attaches a chondroitin sulfate glycosaminoglycan chain via xylose to the serine residue of the ENEGSG sequence. Chondroitin sulfate glycosaminoglycan chain-modified L-APP is denoted as appican (12,13). The biological function of appican is not known, but it may be similar to the function of other chondroitin sulfate proteoglycans in biological processes as wound healing, cell adhesion, neurite outgrowth, and axon guidance (14 -16). Evidence for appican function in mediating cell adhesion came from expression studies on appican in rat C6 glioma cells. Appican-expressing cells attach more avidly to substrate surface than untransfected cells (17). In former cells appican is enriched in the extracellular matrix and displays adhesion-promoting properties for neuronal cells (17). Furthermore, alternative splicing of exon 15 regulates the basolateral and apical secretion of APP and L-APP isoforms in polarized Madin-Darby canine kidney II cells by either creating a new apical or destroying a basolateral sorting signal in L-APP molecules, respectively (18). In context of Alzheimer's disease pathogenesis it is relevant that exclusion of the region encoded by exon 15 from the mature APP also affects its processing to the A␤ peptide by decreasing the relative amount of generated A␤ (18).
In order to get some insight into the splicing regulation of exon 15, in the present study we performed a molecular anal-ysis of this splicing process. We demonstrate that the cell type-specific recognition of exon 15 in non-neuronal and neuronal cells is conserved between mouse and rat. To characterize the regulatory requirements for the cell type-specific usage of exon 15, an APP minigene expression system was established in murine cell lines NIH3T3 and P19. By using this approach we were able to identify two intronic regions flanking exon 15, which we refer to as "upstream control region" (UCR) and "downstream control region" (DCR). Finally, we present evidence that both intronic regions play a key role for the cell type-specific exclusion of exon 15 in APP mRNA molecules in non-neuronal cells.

MATERIALS AND METHODS
Animals and Tissue Preparation-All tissues were prepared from 2to 3-month-old female NMRI mice, provided by the animal facility of Center for Molecular Biology, Heidelberg, Germany. The animals were killed by cervical dislocation, and the tissues were immediately removed, washed in ice-cold phosphate-buffered saline, and quickly frozen in liquid nitrogen.
Cell Culture Conditions-The following murine cell lines NIH3T3, P19, N2a, and AtT20 were cultured in Dulbecco's modified Eagle's medium plus 10% fetal calf serum (Sigma and Boehringer Ingelheim) as adherent monolayers on 10-cm tissue culture dishes at 37°C under 5% CO 2 atmosphere.
Isolation of Genomic Clones-The genomic APP clone L12 ( Fig. 2A) was isolated from a Lambda Fix II 129 SVJ mouse genomic library (Stratagene) using an ␣-32 P-labeled human APP cDNA probe encoding exons 14 -17. Minigene Constructs-All clones were created by using standard methods (19). For cloning the APP minigene MG1, two DNA fragments of the genomic APP clone L12 were used ( Fig. 2A) as follows: fragment I, SacI site of exon 14 to XbaI site of intron 14 (ϳ5.5 kb), and fragment II, XbaI site of intron 14 to EcoRI site of exon 16 (ϳ6.9 kb) ( Fig. 2A). The recipient vector pSKE1416X for fragment I and II was created by inserting a mouse APP cDNA fragment (exons 14 -16) that was amplified by PCR using the primer pair sE14EcoRV (5Ј-ACGATTGATA-TCATGAGCTGCTTCAGAAGGAGC-3Ј) and aE16XhoI (5Ј-AGCAC-CCTCGAGCAGTTTTTGATGGCGGACTTC-5Ј; boldface nucleotides indicate restriction sites) in the EcoRV and XhoI site of a Bluescript (SKϩ) plasmid. Finally, an XbaI polylinker was introduced between the SacI (exon 14) and EcoRI (exon 16) sites of the APP cDNA fragment. Fragment II was introduced into the EcoRI and XbaI site of pSKE1416X generating the pSK1516X plasmid. Afterward fragment I was inserted into SacI and XbaI site of pSK1516X producing the APP minigene pSKMG1. The EcoRV/XhoI fragment of pSKMG1 was inserted into the ScaI/XhoI site of the mammalian expression vector pZeoSV (Invitrogen) generating the APP minigene MG1. All cloning steps were controlled by DNA sequencing. MG2 was constructed by deleting a 2.8-kb NcoI DNA fragment from pSK1516X, followed by a religation and exchanging the wild type XbaI/XhoI fragment of MG1 against the shortened one of pSK1516X.
Deletions constructs A1 to A6 and B1 to B8 were generated by PCR cloning, amplifying 5Ј-or 3Ј-truncated PacI/AflII fragments (Figs. 4 and 5). Minigene constructs ⌬UCR, ⌬DCR, dUCR, dDCR, UCR1, DCR1, RCU, RCD, M2, and PM34.1 were produced by using a recombinant PCR approach. Mutagenized PacI/AflII fragments of the genomic APP region were inserted in the parental APP minigene MG2. Point mutations PM13, PM14, PM15, PM23, PM25, PM26, and PM35 were introduced in the PacI/AflII fragment of MG2 by a modified protocol of the megaprimer PCR method (20). Instead of Taq DNA polymerase we used Pfu DNA polymerase (Stratagene), and therefore no Klenow treatment of newly synthesized megaprimer was necessary. All introduced mutations were confirmed by sequencing. MG4 was generated by PCR using the primer pair sxE15EcoRV (5Ј-ACGATTGATATCGTGAATGCGTC-TGTTCCA-3Ј), axE17XhoI (5Ј-AGCACCCTCGAGTCAGGGTCGTTC-CTCCAG-3Ј; bold-face nucleotides mark restriction sites), and a human genomic DNA preparation as template. The PCR product was inserted into the ScaI/XhoI sites of the pZeoSV vector. E16⌬90 was produced by deleting a 90-bp DNA fragment in exon 16 of MG4 by recombinant PCR. For creating the chimeric constructs C1 and C2, the genomic APP DNA fragments shown in Fig. 8A were fused by recombinant PCR to construct E16⌬90. Detailed PCR strategies and oligonucleotide sequences are available on request.
DNA Transfection and Total RNA Preparation-The P19 and NIH3T3 cells were transfected by the calcium phosphate coprecipita-tion method. One 10-cm dish of 15% confluent cells was transfected with 10 g of plasmid DNA. Each transfection was performed at duplication. 48 h after transfection cells were washed three times with ice-cold phosphate-buffered saline and harvested. The total RNA was isolated by using RNeasy columns (Qiagen) or High Pure total RNA Isolation columns (Roche Molecular Biochemicals) according to the manufacturer's protocol.
Reverse Transcriptase-Polymerase Chain Reaction (RT-PCR)-In general for first-strand cDNA synthesis 2 g of total RNA preparations were reverse-transcribed with 200 units of SuperScript TM (Life Technologies, Inc.) and 50 pmol of oligo(dT) 17 in a total reaction volume of 20 l under the conditions provided by the manufacturer's protocol.
To verify the endogenous APP mRNA splicing pattern, a PCR reaction with the primer pair sm1805 (5Ј-AGGAGATTCAAGATGAAGTCG-3Ј, last 21 nt of exon 13) and am2225 (5Ј-CCCACATCTTCAGCAAA-GAAC-3Ј, first 21 nt of exon 17) was performed. To measure the level of minigene-derived mRNA isoforms, the primer pair spZEO1 (5Ј-AGC-TATTCCAGAAGTAGTGAG-3Ј) and apZEO2 (5Ј-GACACTATAGAA-CAGCTGATC-3Ј) was used, recognizing unique sequences in the pZeoSV vector. In a standard quantitative PCR reaction one-hundredth of a cDNA preparation was amplified in a Perkin-Elmer thermocycler 9600 using the following reaction conditions: 200 nM sense primer, 50 nM 33 P-end-labeled and 150 nM unlabeled antisense primer, 0.2 mM dNTP, 1.5 mM MgCl 2 , and 2.5 units of Taq DNA polymerase (Life Technologies, Inc.) with the supplied PCR buffer in a final reaction volume of 50 l. Since pilot experiments showed that the PCR amplification rates of the endogenous APP transcripts and of the minigenederived APP transcripts were linear between the 22nd and 28th cycle, we used the following cycle parameters in our quantitative RT-PCR assay: 94°C 10 s, 60°C 15 s, 72°C 40 s for 24 cycles (endogenous APP ex15 /L-APP) and 26 cycles (MG-APP ex15 /MG-L-APP) plus a final extension of 7 min at 72°C. Since the signals for MG-L-APP PCR product were very weak in the P19 cells, one-twenty-fifth of a cDNA preparation was amplified with spZEO1 and apZEO2 (pilot experiments showed that the amount of cDNA template did not change the ratio of both PCR products). Generated PCR products were separated electrophoretically on a 6% denaturing urea-polyacrylamide gel, and the dried gels were exposed to x-ray films (Kodak X-Omat AR) at room temperature between 2 days and 2 weeks. The amount of incorporated radioactivity was determined for each band by using a PhosphorImager (Molecular Dynamics) and the quant modus of the computer program Mac BAS 1.0. For quantification each band was corrected by a corresponding background value, and the relative amount of both isoforms with (APP ex15 ) and without exon 15 (L-APP) was calculated to total APP (L-APP ϩAPP ex15 ).

RESULTS
APP mRNA Isoforms without Exon 15 (L-APP) Are Ubiquitously Expressed in Murine Tissues, Including Brain, but Not in Neurons-A previous report on the APP expression established the ubiquitous distribution of L-APP in rat tissues with the exception of primary cultured neurons (11). Since in the present study we used the mouse system to identify cis-acting elements involved in the splicing regulation of exon 15, we first examined the expression pattern of exon 15-containing (APP ex15 ) and lacking (L-APP) APP mRNA isoforms in a number of organs of three adult female mice. To assay the relative amounts of APP ex15 and L-APP in these murine tissues, cDNA was prepared and served as a template in a radioactive PCR using a primer pair flanking exon 15 (Fig. 1A). After separation of the radioactively labeled PCR products by denaturing polyacrylamide gel electrophoresis, two major bands of 420 and 366 bp were visible in the autoradiogram (Fig. 1B), corresponding to APP ex15 and L-APP, respectively. The identity of these bands was confirmed by comparison with the mobility of DNA fragments amplified from corresponding plasmid standards and sequencing (data not shown). APP ex15 and L-APP were present in all tested tissues (Fig. 1B, lanes 1-14). A quantitative determination of the relative L-APP amounts by phosphorimaging revealed that the relative L-APP portions in the non-neuronal tissues varied between 45% (adrenal gland) and 65% (lung) with the mean value of 55% (Table I). In contrast, the relative intensities of the L-APP bands in cDNA preparations from cerebellum or cortex were markedly weaker (Fig. 1B, lanes 13 and 14; Table I) consisting of 7 or 8%, respectively, and suggesting a predominant production of APP ex15 isoforms in the central nervous system. To clarify which cell types in the central nervous system contribute to L-APP production, primary cultures of embryonic mouse neurons (embryonal day 11) and primary cultures of microglial cells derived from newborn mice were cultured, and the relative APP ex15 and L-APP amounts were measured by the quantitative RT-PCR assay described above. As shown in Fig. 1C, murine primary neurons do not synthesize detectable amounts of L-APP, whereas in microglial cells both APP ex15 and L-APP are expressed, the latter accounting for more than a half of the total APP mRNA (Table I; Fig. 1C, compare lanes 16 and 17). Thus, L-APP is produced in murine brain exclusively by cells of non-neuronal origin, and skipping of exon 15 appears to be a very rare event during post-transcriptional processing of APP pre-mRNA transcripts in neurons. These results are consistent with the data reported by Sandbrink et al. (11) for rat tissues and suggest that exon 15 splicing is regulated in a cell type-specific manner.
By using our quantitative RT-PCR assays, we next searched for a murine cell culture model that mimics the in vivo nonneuronal and neuronal L-APP expression pattern. For this, we chose the following murine cell lines: NIH3T3 (fibroblastoma), AtT20 (pituitary gland), N2a (neuroblastoma), and P19 (embryonic carcinoma). Analysis of the APP splicing pattern by phosphorimaging revealed that APP ex15 isoforms were expressed by all cell lines tested and constituted between 43 (NIH3T3) and 99% (P19) of the total APP message (Fig. 1D, lanes 19 -22; Table I). Conversely, L-APP constituted 14 or 57% in the non-neuronal cell lines AtT20 or NIH3T3, respectively, whereas only minor amounts of these APP isoforms were expressed in the two cell lines N2a (Ͻ5%) and P19 (Ͻ1%). Thus, the P19 cell line appeared to be most suitable for studying exon 15 splicing in a neuronal context since the very low or absent L-APP expression is a characteristic feature of neurons. The NIH3T3 cell line was chosen for the following experiments as representative for cells with the non-neuronal L-APP splicing pattern.

A Minigene Spanning APP Exons 14 -16 Mimics the Alternative Splicing Pattern of APP Exon 15 in NIH3T3 and P19
Cells-To examine the regulation of the alternative splicing of APP exon 15 in more detail, a minigene construct encompassing the exons 14 -16 of the murine APP gene was generated. For this, a mouse genomic library was screened with a fragment of the human APP cDNA (exons 14 -17). A hybridizing clone termed L12 contained a genomic fragment of ϳ18 kb including the exons 14 -17 of the murine APP gene ( Fig. 2A). Since cis-acting elements involved in the regulation of alternative splicing are often located in intronic sequences proximal to an alternatively used exon or in regions of upstream and downstream exons of neighboring sequences (21)(22)(23)(24), the genomic regions adjacent to exons 14 -16 were determined and compared with the corresponding regions in the human APP gene (Fig. 2B). The sequence alignment revealed a high degree of conservation in the intronic areas upstream and downstream of exon 15 (80 or 60% similarity, respectively), suggesting the presence of functionally important regulatory cis-elements.
To generate an APP minigene vector a 12.5-kb fragment  -14), two primary brain cell cultures (lanes 16 and 17), and four murine cell lines (lanes 19 -22) are shown. 2 g of total RNA of each tissue sample and the complete total RNA preparation of primary cell cultures were subjected to a reverse transcription reaction, and resulting cDNAs were amplified for 24 cycles in a PCR reaction using the primer pair sm1805 and am2225 ( 33 P-labeled). To check the specificity of amplified PCR signals, 2 g of total RNA of one tissue or the half of a RNA preparation from a primary cell culture sample were reverse-transcribed without SuperScript and subjected to the RT-PCR assay (lanes 15 and 18).

TABLE I
Quantification of APP ex15 and L-APP mRNA isoforms in peripheral tissues, neuronal tissues, primary cultured brain cells, and cell lines of mouse Radiolabeled RT-PCR products were analyzed by using a Phosphor-Imager. Each band was corrected by a corresponding background value. Relative amounts of both mRNA isoforms were determined in percent using the following calculations: %L-APP ϭ (APP ex15 ϩ L-APP)/100 ϫ L-APP; %APP ex15 ϭ (APP ex15 ϩ L-APP)/100 ϫ APP ex15 . spanning exons 14 -16 was inserted into the multiple cloning site of the mammalian expression vector pZeoSV as described under "Materials and Methods" (Fig. 2A). To examine whether the minigene-derived mRNA can be efficiently expressed in cells and whether the minigene-encoded exon 15 undergoes alternative splicing, the APP minigene MG1 was ectopically expressed in NIH3T3 (non-neuronal splicing system) and P19 cells (neuronal splicing system). The minigene splicing pattern was monitored by a quantitative RT-PCR assay, using vectorspecific primers in a PCR reaction with 26 cycles ( Fig. 2A). Concomitantly, the endogenously produced APP transcripts were detected by using the quantitative RT-PCR assay described in Fig. 1A. By using both primer pairs in two separate RT-PCRs, we thus were able to monitor and discriminate between the endogenously and the ectopically expressed APP isoforms. As shown in Fig. 3A, in addition to the endogenous APP mRNA fragments, two minigene-derived splice products were visible as bands of 447 and 501 bp, and these bands were not present in cDNA preparations from "mock"-transfected control cells (Fig. 3A, lanes 1 and 4 versus 3 and 6). In analogy to the APP mRNA isoforms generated from the endogenous APP gene, we refer to the 447-bp fragment as MG-L-APP (minigenederived APP mRNA isoforms without exon 15) and to the 501-bp fragment as MG-APP ex15 (minigene-derived APP mRNA isoforms with exon 15). The identity of the bands representing MG-L-APP and MG-APP ex15 was confirmed by comparison of their electrophoretical mobilities with those of corresponding bands derived by PCR amplification of a plasmid standard (Fig. 3A, lanes 1 and 4 versus 7) and by sequencing (data not shown). A quantitative analysis of the minigenederived PCR products demonstrated that exon 15 inclusion occurred to 41 Ϯ 3% (n ϭ 11) in NIH3T3 cells and to 86 Ϯ 5% (n ϭ 11) in P19 cells (Fig. 3B). The ectopic expression of the APP minigene did not influence the processing of the endogenous APP mRNA as revealed by a quantitative analysis of endogenous APP mRNA levels in transfected and non-transfected cells (Fig. 3B). This has also been confirmed for NIH3T3 and P19 cells stably transfected with the APP minigene (data not shown . Only restriction sites involved in cloning of APP minigenes are indicated (compare "Materials and Methods") and are not necessarily unique. The approximate scale is given by the 1-kb bar on the top. In MG1, the genomic fragment spanning exons 14 -16 (ϳ12.5 kb), marked by the connecting solid lines between L12 and MG1, was set under the control of a SV40 early promotor (SV40 prom, black box) and an SV40 polyadenylation signal (SV40 pA, black box) of the mammalian expression vector pZeo SV (Invitrogen). Vector sequences are not drawn in scale. Small black arrows show the localization of the minigene-specific primer pair spZEO1 and apZEO2 used in the quantitative RT-PCR assay to determine the splicing pattern of the minigene-derived mRNA molecules. Below, expected minigene splicing pattern in non-neuronal and neuronal cells is diagrammed (vector-specific sequences, black; constitutive exons, white; and alternative exon, gray), showing the two amplified PCR products either skipping (MG-L-APP) or including exon 15 (MG-APP ex15 ). B, alignment of genomic sequences flanking the mouse and human exon 15 (exonic sequence, capital letters; intronic sequence, small letters; exon 15, gray box; identical nucleotides, vertical bars). The depicted sequences were aligned by using the method of J. Hein (Lasergene 2.0). Constitutive splicing signals: 3Ј-SS, 3Ј-splice site; 5Ј-SS; 5Ј-splice site; *, putative branch points sharing four or more nucleotides with branch site consensus sequence. The intronic sequences of murine APP gene has been submitted to the GenBank TM Data Bank with accession numbers AF199003, AF199004, and AF199005. 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside (X-gal) confirmed similar transfection rates in NIH3T3 and P19 cells (data not shown). In summary, the genomic APP sequence spanning exons 14 -16 contained all cis-acting elements necessary for cell type-specific recognition of exon 15 in the minigene pre-mRNA molecules. Since the removal of a 2.8-kb fragment in intron 15 of the APP minigene did not affect the regulation of exon 15 splicing, we used this shortened minigene version, denoted as MG2, as parental construct in all subsequent experiments.
Identification of an Inhibitory cis-Element, Located 69 nt Upstream of the Exon 15 3Ј-splice site, Denoted as Upstream Control Region (UCR)-Alignment of human and mouse intronic sequences flanking exon 15 of the APP gene revealed the highest level of conservation in the 3Ј-terminal region of intron 14. This was reflected by 80% nucleotide sequence similarity within ϳ500 bp upstream of exon 15 3Ј-splice site (3Ј-SS) and suggested the presence of conserved cis-acting elements. To examine this more precisely, deletions increasing in length (between 65 and 371 bp) were introduced in this intronic area, and the splicing pattern of these mutagenized APP minigenes was determined in NIH3T3 and P19 cells (Fig. 4A, A1-A6). Removal of up to 303 bp led to a moderate increase of the relative MG-APP ex15 level between 56 Ϯ 3 and 65 Ϯ 2% in NIH3T3 cells (Fig. 4A, A1-A4). In contrast, the largest deletions present in minigenes A5 and A6 caused a drastic increase of the relative MG-APP ex15 amounts to 90 Ϯ 4 (n ϭ 5) and 84 Ϯ 2% (n ϭ 6), respectively, compared with 52 Ϯ 6% (n ϭ 6) MG-APP ex15 in control cells expressing the intact APP-minigene MG2 (Fig. 4A). None of the 3Ј-terminal deletions in intron 14 significantly influenced the recognition of exon 15 in P19 cells (Fig. 4A). To confirm the specificity of this effect, the potential regulatory region located 69 -135 nt upstream of exon 15 3'-SS was removed (Fig. 4B, ⌬UCR), or the orientation of this area was turned around (Fig. 4B, RCU). Expression of both constructs in NIH3T3 cells resulted in significantly elevated MG-APP ex15 levels (Fig. 4B, ⌬UCR, 79 Ϯ 3%, n ϭ 16, lane 2; RCU, 85 Ϯ 1%, n ϭ 7, lane 1) in comparison to wild typetransfected control cells (Fig. 4B, MG2, 52 Ϯ 5%, n ϭ 15, lane  3). In contrast, destruction or misorientation of this potential regulatory region did not give rise to a markedly altered minigene splicing pattern in the neuronal splicing system (Fig. 4B,  n ϭ 15, lane 6). From this analysis, we concluded that the intronic sequence located in the region Ϫ69 to Ϫ135 relative to the intron 14/exon 15 border contains cisacting element(s) controlling the recognition of the alternatively spliced exon in a cell type-specific manner. Therefore, this region of 67 nt was denoted as UCR.
Defining a Potential Regulatory Region Downstream of Exon 15, Named DCR-Since the sequence downstream of exon 15 is also conserved (63% similarity between human and mouse sequence within ϳ500 bp downstream of exon 15), a deletion scan of the 5Ј-terminal portion of intron 15, extending from the 5Ј-splice site (5Ј-SS) of exon 15 to the NcoI restriction site of intron 15, was performed (Fig. 5A). For this purpose, a set of deletion constructs was generated by removing stepwise fragments of ϳ400 to 1300 bp length in the 5Ј-portion of intron 15 (Fig. 5A, B1-B8). Splicing pattern analysis of these minigenes in NIH3T3 cells showed that the removal of large parts in the 5Ј-terminal sequences of intron 15 only caused a moderate increase of the relative MG-APP ex15 amount between 60 Ϯ 3 and 63 Ϯ 3% in comparison to 52 Ϯ 2% MG-APP ex15 wild type-transfected cells (Fig. 5A, B1-B6 versus MG2). In contrast, deletions of approximately 1.1 or 1.2 kb, respectively, led to an activation of exon 15 splicing in NIH3T3 cells that was mirrored by a strong increase of the relative MG-APP ex15 level to 74 Ϯ 2% (n ϭ 5) and to 81 Ϯ 2% (n ϭ 5) in comparison to 52 Ϯ 4% (n ϭ 6) in NIH3T3 cells expressing the wild type APPminigene (Fig. 5A, B7 and B8 versus MG2). None of the truncated APP minigenes influenced the inclusion of exon 15 in the minigene-derived mRNA transcripts of P19 cells (Fig. 5A). One possible explanation of this result is that the enhancement of exon 15 recognition observed in NIH3T3 cells transfected with the constructs B7 and B8 is due to the removal of a splice repressor sequence. Alternatively, the activating effect could solely depend on the size of the deleted intronic area. To discriminate between these two possibilities, the potential regulatory region in intron 15 (position ϩ21 to ϩ158) was deleted (Fig. 5B, ⌬DCR), or the removed intronic fragment was reintroduced in the opposite direction (Fig. 5B, RCD). These mutations significantly improved the recognition of exon 15 in NIH3T3 cells (Fig. 5B, lanes 1 and 2 versus 3) resulting in an increase of the relative MG-APP ex15 levels to 78 Ϯ 8% (n ϭ 13, ⌬DCR) and to 73 Ϯ 3% (n ϭ 6, RCD) compared with 50 Ϯ 8% MG-APP ex15 (n ϭ 13) in control cells transfected with the wild type minigene MG2. In P19 cells only the deletion present in the ⌬DCR construct slightly increased the relative MG-APP ex15 amount from 93 Ϯ 4% (n ϭ 11) to 98 Ϯ 2% (n ϭ 11), whereas the usage of exon 15 was not enhanced in cells transfected with the RCD vector (Fig. 5B, lanes 4 and 5 versus 6). In summary, these data suggested the presence of regulatory sequence elements located in intron 15 (position ϩ21 to ϩ158) that block the inclusion of exon 15 in NIH3T3 cells. This region was therefore designated DCR.
Both cis-Elements Act in a Distance-dependent Manner Regulating the Recognition of Exon 15-From the deletion analysis of the intronic regions proximal to exon 15, we concluded that these sequences contain cis-acting elements (UCR and DCR, respectively), both having an inhibitory influence on the exon 15 inclusion in non-neuronal cells. To test the functional relationship between these cis-elements and exon 15, we increased the distance between exon 15 and UCR from 68 to 418 nt and between exon 15 and DCR from 20 to 711 nt (Fig. 6A, UCR1 and DCR1). In both cases an increase of the distance between exon 15 and the corresponding cis-element resulted in an enhanced inclusion of exon 15 in the minigene mRNA of NIH3T3 cells. The observed effect was associated by an increase of the relative MG-APP ex15 amount up to 72 Ϯ 3 (n ϭ 4, UCR1) and 82 Ϯ 2% (n ϭ 8, DCR1) in comparison to 52 Ϯ 5% of MG-APP ex15 (n ϭ 26, MG2) in wild type-transfected control cells (Fig. 6, A and B, lanes 1 and 3 versus 6). However, the spacing between UCR and exon 15 did not affect the usage of the differential exon in P19 cells (Fig. 6B, lane 7). Enlarging the distance between DCR and exon 15 slightly improved the usage of exon 15 in P19 cells (Fig. 6B, DCR1, lane 9) similar to that observed in cells expressing a DCR-deficient minigene (Fig. 5B,  ⌬DCR, lane 5). Thus, these results suggested that the proxim- ity of both regulatory elements to exon 15 is one of the essential parameters determining their inhibitory function.
Many intronic cis-acting elements are characterized by their redundant architecture, which is reflected by the presence of multiple binding sites for certain trans-acting factors forming a regulatory protein complex (25)(26)(27). To examine this, the sequences of UCR and DCR were duplicated (Fig. 6A, dUCR and  dDCR). The duplication of UCR enhanced the blockage on exon 15 recognition in NIH3T3 cells (Fig. 6B, lane 2), reflected by a decrease of the relative MG-APP ex15 to 27 Ϯ 1% (Fig. 6A, n ϭ 6) compared with 53 Ϯ 5% MG-APP ex15 (n ϭ 26) in control cells expressing wild type minigene MG2 (Fig. 6, A and B, lane 6). Surprisingly, in P19 cells the dUCR construct slightly repressed the usage of exon 15 too (Fig. 6B, lane 8), reducing the MG-APP ex15 value to 84 Ϯ 7% (Fig. 6A, n ϭ 6). In contrary, duplication of DCR did not measurably inhibit the usage of exon 15 neither in NIH3T3 nor in P19 cells (Fig. 6, A and B,  lanes 4 and 10).
To address the question whether both cis-elements contained all sequence information to control the exon 15 recognition in non-neuronal cells, minigene M2 carrying a deletion of UCR and DCR was analyzed in NIH3T3 and P19 cells. In NIH3T3 cells exon 15 of the M2 pre-mRNA molecules was almost constitutively included into the mRNA (Fig. 6B, lane 5), reflected by a strong increase of the relative MG-APP ex15 portion to 91 Ϯ 3% (n ϭ 6) compared with 53 Ϯ 5% MG-APP ex15 (n ϭ 26) in control cells transfected with an APP minigene containing both regulatory regions (Fig. 6, A and B, lane 6). As expected, in P19 cells the removal of both cis-elements caused a slight activation of exon 15 inclusion producing 98 Ϯ 1% MG-APP ex15 (Fig. 6, A  and B, lane 12), as it was observed by deleting DCR alone (Fig.  5B, ⌬DCR). Shifting the splicing pattern toward the neuronal splice site recognition in non-neuronal cells by removing UCR and DCR suggested that both elements are key players mediating the blockage of exon 15 usage in NIH3T3 cells.
The Combination of Multiple Sequences Is Involved in the Inhibitory Function of UCR and DCR-The inhibitory effect of UCR duplication on the recognition of exon 15 described in the previous section implicated a redundant architecture of this sequence element. UCR from both the human and the mouse APP gene displays a high pyrimidine content of ϳ75% (Fig. 7A). The majority of the pyrimidine residues are organized in three stretches, each consisting of between 8 and 14 successive pyrimidines. We therefore referred to these sequence elements as polypyrimidine-rich sequences PPSI, PPSII, and PPSIII (Fig.  7A). Furthermore, PPSII and PPSIII are flanked at one or both ends, respectively, by a tetranucleotide motif UGCU. To examine the role of PPSI, PPSII, and PPSIII in the inhibition of exon 15 inclusion in non-neuronal cells, point mutations were introduced into each element as depicted in Fig. 7A, and the expression pattern of the corresponding constructs was examined in NIH3T3 and P19 cells. All point mutations in PPSI, PPSII, and PPSIII (Fig. 7A, PM25, PM26, and PM23, respectively) weakened the inhibitory function of UCR in non-neuronal cells that was reflected by an increase of the relative MG-APP ex15 amount in NIH3T3 cells between 67 Ϯ 2 and 74 Ϯ 4% in comparison to control cells expressing the parental APP construct MG2 (49 Ϯ 4%). In contrast, these mutations did not significantly affect the usage of exon 15 in P19 cells (Fig. 7A). The strength of this effect was in the range of that observed for the constructs ⌬UCR (79 Ϯ 3%) and RCU (80 Ϯ 3%), both completely destroying this regulatory element (Fig. 4B). This indicates that PPSI, PPSII, and PPSIII are involved in the repression of exon 15 recognition in non-neuronal cells.
Similar analysis was performed for the other cis-regulatory region DCR (Fig. 7B). In contrast to UCR, the pyrimidine content of human and mouse DCR is more balanced making up ϳ60% of all residues. Approximately 50% of the pyrimidines are clustered in four (murine DCR) or five (human DCR) short stretches between 5 and 16 nt in length, located at both ends of the DCR. Two of these pyrimidine-rich sequences contain a short conserved repeat GCCUCU(C)(U) and a truncated version GCUCUC (Fig. 7B). The involvement of these elements in the inhibitory function of DCR was tested by mutating them to the complementary sequences shown in Fig. 7B. Point mutations present in the constructs PM13, PM14, and PM15 had no significant effect on exon 15 usage neither in NIH3T3 nor in P19 cells (Fig. 7B). Since the function of an intronic cis-element is often mediated by a combination of different sequence motifs (27), in minigenes PM34.1 and PM35 parts of the mutated sequences in PM13 or PM15 were combined with the mutation of construct PM14, respectively (Fig. 7B). These double mutations drastically impaired the inhibitory function of DCR in NIH3T3 cells increasing the relative amount of APP ex15 to 78 Ϯ 4% (n ϭ 4) for PM34.1 and 74 Ϯ 1% (n ϭ 4) for PM35 compared with 53 Ϯ 10% (n ϭ 12) for the MG2-transfected control cells. In P19 cells, these mutations caused no significant alteration of the splicing pattern (Fig. 7B). Therefore, we concluded that at least two of these RNA motifs in the DCR sequence are necessary to inhibit the usage of exon 15 in NIH3T3 cells.
UCR and DCR Regulate the Inclusion of a Hybrid Exon in a Constitutive Splicing Background-The results reported in the preceding sections showed that the identified cis-elements block the recognition of exon 15 in NIH3T3 cells. To examine the ability of UCR and DCR to regulate the cell type-specific inclusion of a chimeric exon, both elements were grafted into the constitutive splicing background of the human APLP1 gene, a member of the APP gene family (28 -30). For this purpose, a 553-bp genomic segment of the APLP1 gene, encompassing the constitutively spliced exons 15-17, was amplified by PCR from genomic DNA and subcloned into the pZeoSV vector, generating APLP1 minigene MG4. To minimize possible effects of potential cis-acting elements in exon 16 (144 bp length) of APLP1 and to reduce the size of the exon to that of the APP exon 15 (54 bp), the minigene construct E16⌬90 was created by deleting 90 bp of the central part of APLP1 exon 16 (Fig. 8A). Since P19 cells did not accept the minigene-derived pre-mRNA molecules as splicing substrates (data not shown), for this particular study the neuroblastoma cell line N2a was used as a neuronal splicing system. After transient transfection of MG4 and E16⌬90 into NIH3T3 and N2a cells, the splicing pattern of the minigene-derived mRNA molecules was monitored by qualitative RT-PCR using minigene-specific primers spZEO1 and apZEO2 (Fig. 2A). The identity of each PCR product was confirmed by sequencing (data not shown). As expected, APLP1 exon 16 was constitutively included in chimeric mRNA molecules from both cell lines (Fig. 8B, lanes 1 and 2). However, removing 90 bp of the internal sequence in exon 16 (exon ⌬90) yielded a very different result, reflected by the generation of two splice products of 341-and 287-bp length in both cell lines (Fig. 8B, lanes 3 and 4). The faster migrating band corresponds to mRNA species lacking the exon ⌬90. To test whether UCR and DCR could block the inclusion of exon ⌬90 in NIH3T3 cells, the first 10 nt of exon ⌬90 with the flanking intronic sequences were exchanged against the corresponding regions of exon 15 with UCR and DCR (Fig. 8A, C1). Expression of this chimeric construct in NIH3T3 showed that the recognition of the chimeric exon was impaired, producing mainly splicing variants without this exon (Fig. 8B, lane 5). In contrast, the genomic APP portions of minigene C1 did not block the usage of this exon in N2a cells (Fig. 8B, lane 6). To confirm whether UCR and DCR were responsible for blockage of the chimeric exon in NIH3T3 cells, both cis-elements were deleted (Fig. 8A, minigene C2). The removal of both regulatory regions neutralized the observed blockage of chimeric exon in NIH3T3 cells (Fig. 8B, lane 7). Furthermore, this deletion led to a strong activation of the chimeric exon in NIH3T3 cells, resulting in its constitutive usage (Fig. 8B, lanes 7 and 8). These results demonstrated first that UCR and DCR regulated the recognition of the chimeric exon in a cell type-specific manner, indicated by its inhibition in NIH3T3 cells, whereas in N2a cells the repression of the chimeric exon was not enhanced; and second that these effects could be reversed in NIH3T3 cells by deleting UCR and DCR. DISCUSSION Alternative splicing plays a fundamental role in the regulation of eukaryotic gene expression, generating different mRNA isoforms from a single pre-mRNA precursor. In most cases, the usage of an alternatively spliced exon is controlled in a tissue-, developmental stage-, or sex-specific manner (reviewed in Refs. 31 and 32).
Cell Type-specific Recognition of Exon 15 Is Conserved between Mouse and Rat-Consistent with previously published data on the distribution of L-APP transcripts in rat (11), our expression analysis revealed the ubiquitous occurrence of L-APP in mouse tissues. On average, L-APP constituted ϳ55% of the total APP message in non-neuronal organs of mouse, and the measured relative L-APP levels were between 2 and 20% higher than those determined in rat. In both rodent species only low levels of L-APP were detected in brain (5-6%). The major source of brain L-APP appear to be cells of non-neuronal origin, since L-APP was undetectable by our sensitive RT-PCR assay in RNA preparations from primary cultured mouse neurons (Fig. 1C). In contrast, L-APP consisted of ϳ60% total APP transcripts in primary cultured microglial cells, in agreement with the previously reported relative abundance of L-APP in cultures of rat microglia and astrocytes (11). Taken together these data strongly indicate that the cell type-specific recognition of APP exon 15 is conserved between both rodent species. Therefore, the mouse model is a suitable system to analyze the molecular mechanism controlling the cell type-specific usage of APP exon 15 in non-neuronal cells and neurons.
Two Intronic Regions Flanking Exon 15 Repress Its Recognition in Non-neuronal Cells-The principal aim of the present study was to identify cis-acting elements in the APP gene involved in alternative splicing regulation of exon 15. Therefore, we established an APP minigene expression system, since this approach was very useful in the study of several alternatively spliced genes (33)(34)(35)(36). For this, the endogenous APP ex15 / L-APP splicing pattern of different murine cell lines was analyzed by quantitative RT-PCR. Determination of endogenous APP ex15 /L-APP mRNA levels revealed that P19 and N2a cells almost exclusively express APP ex15 isoforms, an APP splicing pattern also observed in neurons (Fig. 1, C and D). In contrast, L-APP accounted for 12 or 57% of the total APP message in AtT20 and NIH3T3 cells, respectively, representing the nonneuronal splicing type (Fig. 1, C and D). To study the APP exon 15 splicing regulation, we therefore chose P19 cells as neuronal splicing system and the NIH3T3 cells as representative for splicing pattern found in most non-neuronal cells. Transient expression of a 12.5-kb genomic APP fragment spanning exons 14 -16 in this cell culture system showed that it contained all cis-acting elements required to regulate the usage of exon 15, since the minigene-derived mRNA molecules displayed a splicing pattern nearly indistinguishable from that of the endogenous APP gene (Fig. 3, A and B).
Sequence conservation within intronic regions proximal to exon 15 of the human and mouse APP gene suggested the presence of cis-acting elements that might be involved in splicing regulation (Fig. 2B). Deletion scan analysis of the intronic region upstream and downstream of exon 15, using our APP expression system, led to the identification of two regulatory elements (Figs. 4 and 5) as follows. (i) The UCR is located 69 nt Both cis-elements mediated a splice repressor function, blocking the recognition of exon 15 in non-neuronal cells by acting in a sequence-and distance-dependent manner (Figs. 4B, 5B, and 6). Furthermore, UCR and DCR are the key players for regulating the exclusion of exon 15 in the non-neuronal splicing system, since their complete removal impaired the cell-specific recognition of exon 15 in non-neuronal cells (Fig. 6,  M2). The cell type-specific splice repressor function of UCR and DCR is underlined by the fact that the recognition of a chimeric exon in a constitutive splicing background is cell-specifically blocked in non-neuronal cells by the presence of both cis-elements, whereas in the neuronal splicing system these regulatory elements do not negatively influence the usage of this exon (Fig. 8, C1 and C2). In summary these experimental data clearly showed that these regulatory elements represent cell type-specific acting sequences controlling exon 15 exclusion in the non-neuronal splicing system we used.
One common feature of many intronic regulatory sequences, e.g. the neuron-specific N1 enhancer of c-src gene (25,27), is their redundant and complex architecture. This is reflected by the presence of multiple binding sites for certain factors, the latter being part of a protein complex that regulates splice site selection (26,37). Hence, the enhanced repression of exon 15 observed in non-neuronal cells after duplicating UCR can be interpreted as a reflection to its redundant architecture. Moreover, duplication of UCR interferes with its cell type-specific function, since the recognition of exon 15 was slightly impaired in neuronal cells, too (Fig. 6, dUCR). This could be caused by cooperative interactions between different binding sites of the two duplicated cis-elements in neuronal cells causing a basal level of inhibition, similar to the activation of a heterologous exon by the duplicated neuron-specific N1 enhancer in nonneuronal cells (27). Alternatively, the increased number of potential binding sites could titrate for limited trans-acting factor(s) that neutralize(s) the negative function of UCR in neuronal cells. The latter case is very unlikely, since a titration experiment with different dUCR plasmid DNA amounts (transfection of 50 ng to 5 g by liposome-mediated transfection) showed no enhanced function of UCR in P19 cells. 2 In contrast, duplication of DCR did not strengthen its negative effect neither in P19 nor in NIH3T3 cells (Fig. 6, dDCR). This result appears surprising, but in the context of the complex architecture of splicing regulatory sequences it is possible that DCR, spanning position ϩ21 to ϩ158, contains only a part of the target sequences for mediating its repressor function. Thus, a duplicated DCR did not amplify its inhibitory influence on exon 15 recognition, since cooperative interactions between certain components were still lacking. A similar effect has been reported for the neuron-specific intronic enhancer of the N1 exon; duplication of the enhancer core sequence (position ϩ38 to ϩ70) moderately activated the differential exon in neuronal cells, whereas a multimerization of a larger area (position ϩ38 to ϩ142) strongly increased the amount of mRNA transcripts including N1 exon (25,27).
Models for the Regulation of the Alternatively Spliced Exon 15-The recognition of an alternatively spliced exon in a threeexon system, like the APP minigene, is determined by the balance of splice site strength between the flanking exons and the internal exon, leading to the inclusion and/or exclusion of the internal exon. Common features of the pre-mRNA molecule like exon length (34,38), intron size (39), identity with splice site consensus (40,41), and RNA secondary structures (42,43) influence the splice site strength. A shift between both splicing pathways can be mediated by changing the strength of involved splice sites, mainly that of the alternatively spliced exon. In this context splicing enhancer and repressor elements located in intronic and exonic sequences play an important role in modulating the splice site strength of an alternatively spliced exon in a tissue-, sex-, or development-specific manner. In the case of the APP minigene our presented data suggest that UCR and DCR are key players controlling the cell-specific recognition of exon 15 in the non-neuronal splicing system. In principal, the inhibitory function of both cis-elements can be mediated by two different mechanisms as follows. (i) UCR and DCR are involved in the formation of an RNA secondary structure, masking the splice sites of exon 15 for its recognition by the spliceosomal components, as it was observed for exon 6B of the chicken ␤-tropomyosin pre-mRNA (42). (ii) UCR and DCR contain multiple binding sequences for assembling a proteinaceous repressor complex in non-neuronal cells.
We propose that both regulatory regions serve as target sequences for the formation of a protein repressor complex, since RNA secondary structure predictions identified no potential stem loop structures that can be involved in splice site selection.
The genomic localization, presence of pyrimidine-rich sequence elements, and inhibitory function of UCR and DCR resembled the situation of the intronic elements URE and DRE flanking exon 3 of ␣-tropomyosin. These regulatory elements repress the recognition of ␣-tropomyosin exon 3 in smooth muscle cells (44). Key mediator for the inhibition of exon 3 in smooth muscle cells is the "polypyrimidine tract binding protein" (PTB) that binds to pyrimidine-rich elements within DRE (45)(46)(47). Regarding its high pyrimidine content of almost 80%, the UCR sequence could serve as a PTB-binding target, too. Comparison of the UCR sequence with the PTB binding consensus ((U/G)C(A/Y)GCCUG(Y/G)UGCYYYYCYYYYG(Y/G)-CCC) (46) identified three potential PTB-binding sites sharing 17-18 nucleotides with the PTB consensus, whereas the pyrimidine-rich sequences (PPSI to III) of UCR form the internal pyrimidine core of the potential PTB-binding sites: PPSI (UCA-CAAAGGUUCUUUUUUUUCCCAGC), PPSII (CCAG-CUG-CU-CCUUUUUCAUGACUU), and PPSIII (UAUUGCAAAU-GCU-UUCCCUUGCUCU) (where Y is pyrimidine; identical nucleotides with PTB consensus are in boldface letters; pyrimidine core is underlined; dash indicates missing nucleotide). A possible involvement of PTB in the regulation of APP exon 15 splicing is strengthened by the fact that the destruction of PPSI, -II, or -III weakened the inhibitory function of UCR (Fig.  7A). The second regulatory region, DCR, contained the conserved RNA motifs GCCUCU(C)(U) two times and a truncated version GCUCUCU which shows sequence homology to the intronic CUCUCU element located upstream of c-src gene N1 exon. This short element blocks the inclusion of the neuronspecific N1 exon in non-neuronal cells (48) by binding PTB (49). The mutation analysis of these elements in DCR indicated that this could also be the case in exon 15 splicing, since the destruction of two of these RNA motifs impaired the cell-specific exclusion of exon 15 in non-neuronal cells (Fig. 7B). Therefore, PTB is an attractive candidate that mediates the repressor function of UCR and DCR in the APP exon 15 splicing system.
Since the presence or absence of the region encoded by exon 15 influences the proteolytic processing of APP to A␤ peptide (18) and neurons do not produce L-APP, one could speculate that APP ex15 and L-APP might play a different role in the progression of Alzheimer's disease. If this is the case the identified regulatory sequences could serve as a target to modulate 2 C. Bergsdorf and K. Beyreuther, unpublished data. neuronal APP splicing and may be beneficial in preventing or treating Alzheimer's disease.