|Home | About | Journals | Submit | Contact Us | Français|
The β-amyloid peptide (Aβ) that accumulates in senile plaques in Alzheimer’s disease is formed by cleavage of the Amyloid Precursor Protein (APP). The APP gene has several intronic Alu elements inserted in either the sense or antisense orientation. In this study, we demonstrate that binding of SC35 and hnRNPA1 to Alu elements on either side of exon 7 in the transcribed pre-mRNA is involved in alternative splicing of APP exons 7 and 8. Neuronal cells transfected with the full-length form of APP secrete higher levels of Aβ than cells transfected with the APP695 isoform lacking exons 7 and 8. Finally, we show that treatment of neuronal cells with estradiol results in increased expression of APP695, SC35, hnRNPA1, and lowers the level of secreted Aβ. An understanding of the regulation of splicing of APP may lead to the identification of new targets for treating Alzheimer’s disease.
A major characteristic of Alzheimer’s disease (AD) is the presence of extracellular amyloid deposits in the brain, which predominantly contain the 4kDa β-amyloid peptide (Aβ).1,2 This 39 - 42aa peptide is derived from the β-amyloid precursor protein (APP). The APP gene contains 18 exons, and gives rise to several isoforms varying in size from 695 to 770 aa by alternative splicing of exons 7, 8, and 15.3-5 The two longer forms, APP770 (full-length) and APP751 (skipped exon 8) are ubiquitously expressed, while APP695 (skipped exons 7 and 8) is expressed predominantly in neuronal tissues. Exon 7 encodes a 56-aa domain with homology to the Kunitz serine protease inhibitors (KPI).6 KPI-containing APP has been shown to inhibit serine proteases involved in the coagulation cascade, leading to the suggestion that it may act as an intracerebral anticoagulant.7-9 It has been hypothesized that the production of Aβ from APP is regulated by a protease that is inhibited by this domain.10 It is therefore important to understand how the generation of these different isoforms is regulated.
Alternative splicing is a widespread mechanism of gene regulation that also serves to expand the proteome by several orders of magnitude.11 Studies of expressed sequence tag (EST) and mRNA datasets suggest that well over 50% of human genes are alternatively spliced.12,13 There are two types of exons: constitutive and alternative. Constitutive exons are included in all mRNAs expressed from a given gene, whereas alternative exons are included in only a fraction of mRNAs synthesized from a gene. Of the four major types of alternative splicing, exon skipping, alternative 3′ splice site, alternative 5′ splice site, and intron retention, exon skipping appears to occur most frequently.12 Alternative splicing is modulated by SC35, a serine-arginine (SR) rich protein, interacting with different proteins, including hnRNP family members.14,15
It has been shown previously that skipping of exon 15 of APP during glial differentiation of embryonal carcinoma cells is suppressed by U2 small nuclear ribonucleoprotein auxiliary factor (U2AF).16 The presence of cis-acting elements in an intronic region flanking exon 817, and a splicing enhancer downstream of exon 8 that interacts with a CUG-binding protein18 have been proposed as regulators of exon 8 skipping. However, no protein or DNA sequence that regulates skipping of exon 7, encoding the KPI domain, has been described so far. We report here a role for SC35, hnRNPA1 and Alu elements from introns 6, 7 and 8 in the control of alternative splicing of exons 7 and 8 of the APP gene.
Human Embryonal Kidney (HEK293) and malignant pluripotent embryonal carcinoma (NTERA-2) cell lines were obtained from the European Collection of Animal Cell Cultures (ECACC; Salisbury, UK). HEK293 and NTERA-2 were propagated in Minimum essential medium (Eagle) and Dulbecco’s modified Eagle’s medium (DMEM), respectively, both containing 10% heat-inactivated fetal serum, supplemented with glutamine, penicillin and streptomycin (Invitrogen, Paisley, UK). NTERA-2 cells were terminally differentiated to neurons (NT2N) by adding 0.01mM trans-retinoic acid to the medium as previously reported.19 The purity of neuronal cultures was monitored by microscopy, and found to vary between 94 and 97 percent in different experiments. Estradiol (17β-estradiol, E2) was purchased from Sigma-Aldrich (Dorset, UK) and was added to the cell medium of NT2N for 96 hours at final concentration of 10nM.
The Riboprobe System – T7 (Promega) was used for in vitro RNA preparation. Oligonucleotides GAGAATTCGCTGGTCTCGAACTCCTGACCTCAAGTGATCCCACCGAATTCGA, GAGAATTCGCTCCTGACCTCAAGTGATCCCACCGAATTCGA, GAGAATTCGGTCTCGAACTGAATTCGA and their complementary sequences were purchased from Invitrogen. The first sequence (probe 1) is the hnRNPA1 binding motif (Fig. 1b) in double stranded DNA (dsDNA).20 The second sequence (probe 2) comprises same binding motif, but with the B-box of the promoter for polymerase III deleted. In the third sequence (probe 3) HREs 1, 2 and 3 were deleted. Each oligonucleotide was incubated in PBS with its complementary sequence to obtain dsDNA. All sequences were designed to contain Eco RI restriction sites at both ends. After Eco RI digestion they were cloned in an Eco RI linearised pALTER-1 in vitro transcription vector. After Sac I digestion, T7 RNA polymerase was used to synthesize RNA, which was labelled by the addition of 50 μCi [α-32P]rCTP to the reaction mix. DNA template was removed by RQ1 DNase.
HnRNPA1 was cloned into pcDNA4/HisMax-TOPO expression vector (Invitrogen) as described previously.20 The vector was then transfected transiently into HEK293 cells by SuperFect transfection reagent (Qiagen Ltd., UK). The expressed recombinanta hnRNPA1 was purified by Xpress System Protein Purification (Invitrogen), and the N-terminal fusion tag was removed by Enterokinase Max (Invitrogen). Heat-denatured RNA probes 1-3 (80ng each) were incubated with 100, 200 or 300 ng of hnRNPA1 for 50 min at 22°C in binding buffer (20 mM Tris-HCl pH 7.4, 0.1 mM EDTA, 0.1% Triton X-100, 1% 2-mercaptoethanol, 2 mM MgCl2, 5% glycerol), containing either 0.1 mg/ml poly-d(IC) or 0.1 mg/ml poly-d(IC) combined with 1 mg/ml sonicated salmon sperm DNA. Labelled probes 1-3 and those from the incubation reactions were separated in a 15% native polyacrylamide gel (acrylamide:N,N’-methylenebisacrylamide = 60:1). The RNA signal was detected by autoradiography.
The APP minigene containing the full length antisense Alu in intron 7 was constructed by PCR using clone S491 (GeneBank™ accession number AP001440) as a template with forward and reverse primers specific for introns 6 and 8, respectively (Table 1). Both primers were designed to contain four additional nucleotides at their 5′ ends, which are necessary for splicing. The sequence amplified by this primer pair has a single sense oriented Alu element in each of the intron 6 and intron 8 segments. Transcription of this minigene results in the presence of a single SC35 binding site in intron 7 of the pre-mRNA (Fig. 1). A second minigene with a reversed SC35 binding site was constructed by PCR using three sets of primer pairs (Table 1). Each primer flanking the SC35 binding motif was designed to contain either Acc III or Kpn I restriction sites at the 5’ end. There are no other Acc III and Kpn I sites in the amplified fragments. These restriction sites were designed on primers amplifying the fragment that had to be reversed so as to ensure its insertion in the desired orientation only. After PCR, products were digested with Acc III and Kpn I, and ligated using T4 ligase. Minigenes similar to the first construct, but with reversed Alu elements in introns 6 or 8 were designed using the strategy described for the second minigene. Primer pairs used for each construct are given in Table 1. Platinum Taq DNA Polymerase (Invitrogen) was used as the amplifying enzyme. The three constructs were cloned in mammalian expression vector pcDNA4/HisMax-TOPO TA (Invitrogen) and checked by sequencing. Plasmids were introduced into HEK293 cells using the Superfect transfection kit (Qiagen). Splice variants were examined by RT-PCR using the plasmid-specific primers Xpress and BGH, which allowed discrimination from endogenously expressed APP-isoforms. RT-PCR products from expression of the minigenes were examined by sequencing.
SiRNA oligonucleotides used to knock down expression of SC35 (sense strand, 5′-AAUCCAGGUCGCGAUCGAAdTdT-3′), hnRNPA1 (sense strand, 5′-GCUCUUCAUUGGAGGGUUGdTdT-3′) and as a negative control (sense strand, 5′-CAGTCGCGTTTGCGACTGGdTdT-3′) were described previously.21,22 The specificity of these siRNAs for their targets was confirmed in a BLAST search (www.ncbi.nlm.nih.gov/BLAST). DsRNAs (Invitrogen) were cotransfected in HEK293 cells with the minigene containing the SC35 binding motif and Alu elements in introns 6 and 8, using Superfect transfection kit (Qiagen) according to the manufacturer’s recommendations. Western blot and RT-PCR analysis were performed 48 hours post-transfection.
Coding sequences of the three proteins were amplified by RT-PCR using total RNA from normal human brain (Ambion Europe Ltd., Huntington, UK). Primers for APP770 were (AAGCTAGCCACCATGCTGCCCGGTTT – forward and CGAAGCTTCTAGTTCTGCATCTGCTCAAAGA – reverse); for APP751 were (ATACGCGTGCCACCATGCTGCCCGGTTT – forward and CGGCTAGCCTAGTTCTGCATCTGCTCAAAGA – reverse); for APP695 were (AAGTCGACGCCACCATGCTGCCCGGTTT – forward and AAGCGGCCGCCTAGTTCTGCATCTGCTCAAAGA – reverse). All forward primers contain a Kozak sequence for improving mRNA translation. Products with the right size for each protein were gel-purified and cloned into pZeoSV2(+) (Invitrogen) by Nhe I and Hind III restriction sites (APP770), into pBI (BD Biosciences) multiple cloning site I by Mlu I and Nhe I restriction sites (APP751), and into pBI multiple cloning site II through Sal I and Not I sites (APP695). PCR products were examined by sequencing. Plasmids pZeoSV2(+) and pBI with cloned cDNAs were transiently co-transfected using the Superfect transfection kit into HEK293 Tet On cells (BD Biosciences). Expression of APP751 and APP695 from pBI was induced by 50 μg/ml doxycyclin. NT2N neurons were transfected with the pZeoSV2(+) vector containing either the full-length APP770 isoform or APP695.
Cytoplasmic RNA was purified from HEK293 cells or from HEK293 transfected with various plasmids, using a Qiagen RNeasy mini kit as instructed by the supplier. Briefly, for cytoplasmic RNA purification, buffer RLN (50 mM Tris, pH 8.0, 140 mM NaCl, 1.5 mM MgCl2, and 0.5% Nonidet P-40) was directly added to monolayer cells, and cells were lysed at 4°C for 5 min. Nuclei were removed by centrifugation, cytoplasmic RNAs were purified by column and controlled for contamination from nuclear RNA and DNA by RT-PCR employing a primer pair specific for intron 2 of human β-actin gene (data not shown). Only samples that did not show amplification were used in all analyses.
The soluble subcellular protein fractions were prepared from HEK293 Tet On cells and NT2N neurons as described previously.23 The fractions from non-transfected, transfected and E2 treated cells were run in a NuPAGE 4-12% Bis-Tris precast gel (Invitrogen). The proteins were then transferred onto a nitrocellulose membrane and the APP isoforms detected by anti-APP, N-terminal antibody developed in rabbit (Sigma-Aldrich) as described by the supplier. Anti-rabbit IgG-peroxidase (Sigma-Aldrich) was used as a secondary antibody and detected by enhanced chemiluminiscence (ECL, Amersham, Buckinghamshire, UK). To detect the secreted Aβ peptide in HEK293 Tet On cells and NT2N neurons (untransfected, transfected and E2 treated), 5ml medium from each flask was concentrated in Ultrafree centrifugal filter units (Millipore) up to a final volume of 20μl. Concentrated samples were run in a NuPAGE 4-12% Bis-Tris precast gel (Invitrogen) and transferred onto a nitrocellulose membrane. Aβ was detected by rabbit anti-Aβ (Sigma-Aldrich) using anti-rabbit IgG-peroxidase (Sigma-Aldrich) as a secondary antibody. The signal was detected by ECL, quantified densitometrically (Quantity one 4.3.0 software, BioRad), and triplicate meiasurements were evaluated statistically using Student’s t-test. Protein analysis of SC35 and hnRNPA1 was performed on whole cell lysates by mouse monoclonal antibodies against SC35 and hnRNPA1 (Sigma-Aldrich). As control we used mouse monoclonal anti-β-actin antibody (Sigma-Aldrich).
The APP gene contains several intronic Alu elements inserted in either the sense or antisense orientation (Fig. 1a). In this study, we tested the hypothesis that these intronic Alu elements are involved in the regulation of alternative splicing of APP exons 7 and 8. Alu elements have been broadly classified into three subfamilies: Old (Alu J), Middle (Alu S) and the Youngest (Alu Y) based on certain diagnostic site mutations.24 We previously identified a motif that binds hnRNPA1 in double stranded DNA and is homologous to sequences in AluSx (Fig. 1b).20 This motif contains a polymerase III promoter element (B box) and several hormone-response elements (HREs). Interestingly, when AluSx elements containing the hnRNPA1 binding motif are present in the antisense orientation, transcripts encoded by the complementary DNA strand contain the SC35 binding site identified previously (Fig. 1c).25
We first examined whether hnRNPA1 could bind the RNA sequence transcribed in vitro from the hnRNPA1 binding motif previously identified in Alu elements (Fig. 1b).20 Our results showed that specific binding of hnRNPA1 to this sequence in RNA does occur (Fig. 2, probe 1), even when the B-box is deleted from the DNA template (Fig.2, probe 2). However, the protein did not bind to RNA transcribed from the B-box itself (Fig.2, probe 3). Therefore, hnRNPA1 can bind pre-mRNAs within sequence transcribed from HREs in Alu elements inserted in the sense orientation.
Two artificial minigenes were then prepared. The first contained a fragment from intron 6 including an Alu element inserted in the sense orientation, exon 7, intron 7 including the single Alu element inserted in the antisense orientation, exon 8, and a fragment from intron 8 including two Alu elements inserted in the sense and antisense orientations (Fig. 3a). The second minigene contained the same regions but with an inverted SC35 binding site in the antisense Alu element in intron 7 (Fig. 3b). The vector contains the PCMV promoter 5′ from exon 7 and the BGH polyadenylation sequence 3′ from exon 8. These constructs were transiently transfected into the human embryonal kidney cell line, HEK293, which has previously been shown to support alternative splicing of APP exons 7 and 8.18 A day later, cytoplasmic RNA was isolated from both transfected cell populations. Splice variants were examined by RT-PCR using the plasmid specific primers Xpress forward and BGH reverse primers. The results showed unambiguously that while the minigene with the antisense Alu element in intron 7 supports expression of all four possible splice variants, inversion of the SC35 binding sequence results in expression of only the variant containing both exons 7 and 8 (Fig. 3a,b). These findings indicate that SC35 binding at this site might be essential for skipping of exons 7 and 8.
Experiments with the minigenes further show that although there are hnRNPA1 binding sites in introns 6 and 8, alternative processing of this pre-mRNA does not occur without the SC35 binding site in intron 7. Therefore, we next studied whether hnRNPA1 binding to the pre-mRNA is involved in skipping of APP exons 7 and 8. Another set of minigenes was thus constructed with inverted sense Alu elements from either intron 6 (Fig. 3c) or intron 8 (Fig. 3d). In both experiments only a major band from the full-length product and a minor product with a skipped exon 8 were detected by RT-PCR, while no skipping of exon 7 or exons 7 and 8 together were observed. Therefore, the sense Alu elements from introns 6 and 8 are required for alternative splicing of the APP minigene pre-mRNA. However, inversion of the antisense Alu element in intron 8 (Fig.3e) did not influence the alternative splicing of the minigene.
The role of SC35 and hnRNPA1 in skipping of exons 7 and 8 was then examined by siRNA knockdown of these proteins. HEK293 cells were transfected with the first APP minigene containing the antisense Alu element in intron 7 (Fig. 3a), plus siRNA specifically degrading SC35 mRNA. This led to a markedly reduced expression of the SC35 protein (Fig. 4a) and a significant reduction in the isoforms with skipped exons 7 and 8 (Fig. 4b). The same experiments performed with hnRNPA1 siRNA (Fig. 4a) led to a markedly reduced expression of the hnRNPA1 protein, and a similar reduction in the isoforms with skipped exons 7 and 8 (Fig. 4b). These results show that both hnRNPA1 and SC35 are required for alternative splicing of the minigene. The efficiency of SC35 and hnRNPA1 knockdown (Fig. 4c) was assessed from Western analysis shown in Fig. 4a. Only 17% and 9% of expression in cells transfected with the control siRNA was detected for the hnRNPA1 and SC35 siRNAs, respectively.
We then tested whether this mechanism for regulating alternative splicing could affect the secretion of Aβ. HEK293 Tet On cells were cotransfected with the full length isoform of APP, APP770, cloned under the control of a constitutively active promoter in the vector pZeoSV2, and with APP751 (lacking exon 8) and APP695 (lacking exons 7 and 8) both cloned in the inducible bi-directional promoter vector pBI. Two days later, soluble subcellular proteins were prepared either from doxycycline treated or untreated cells. Doxycycline induces expression of both APP751 and APP695, while without the antibiotic these genes are silent. Western blot analysis was performed using anti-APP N-terminal antibody, which recognises all three isoforms.
Transfected cells showed increased expression of APP770, APP751 and APP695 upon doxycycline induction and increased expression of only APP770 without doxycycline induction (Fig. 5a). There are three bands with increased intensity after doxycycline treatment detected by the antibody. The bands of 112 and 105kDa are most likely APP751 with different post-translational modifications, such as glycosylation and/or phosphorylation.26,27 The secreted Aβ peptide was then detected by anti-Aβ (Fig. 5b). When only APP770 is expressed, 6-fold more Aβ was secreted than when all three APP isoforms were expressed (Fig. 5c). Taken together, these data suggest that the Alu elements encoding the SC35 and hnRNPA1 binding motifs in introns 6 and 7 of the APP pre-mRNA are required to keep the secreted Aβ at a low level. Similar experiments were performed to demonstrate the role of Alu controlled splicing in Aβ secretion in NTERA-2 cells terminally differentiated to neurons (NT2N) (Fig. 6). APP770 transfected cells secreted twice as much Aβ as compared to untransfected neurons (Fig. 6b,c). However, when we transfected NT2N with the APP695 isoform that lacks exons 7 and 8, a 60% decrease in secreted Aβ was detected.
Estradiol (E2) was recently shown to regulate alternative splicing of the APP pre-mRNA in cortex of mouse brain.28 Therefore, we next investigated if this hormone has a similar effect in human neuronal NT2N cells. Our experiments show that E2 treatment upregulates APP695 and downregulates both APP770 and APP751 isoforms (Fig. 7a). Of particular interest, E2 treatment also led to a lower secretion of Aβ (Fig. 7b). Finally, we tested whether treatment of NT2N cells with E2 affects expression of SC35 and hnRNPA1 (Fig. 7c). Both proteins showed increased expression in E2 treated cells compared to untreated cells.
The proportion of KPI-containing forms of APP in soluble subcellular fractions of AD brain tissue is significantly elevated compared with normal brain tissue.29 This is in agreement with our findings that overexpression of APP770, a KPI-containing protein, leads to secretion of more Aβ by both HEK293 cells and NT2N neurons than overexpression of APP695 (Figs. (Figs.5,5, 6). KPI-containing isoforms of APP regulate extracellular cleavage of secreted APP by inhibiting the activity of a secreted APP-degrading protease30 and may thus be more amyloidogenic.31,32 Our findings therefore support an imbalance of APP isoforms as a possible mechanism for Aβ deposition. Furthermore, our data (Figs. (Figs.22--6)6) suggest that the Alu elements from introns 6, 7 and 8 of APP pre-mRNA, and particularly the SC35 and hnRNPA1 binding sites within them, play a crucial role in maintaining the correct ratio between the APP isoforms.
It should be noted that the pattern of Alu densities across chromosomes is highly similar to the pattern of gene densities.33 Although there are extensive data demonstrating the involvement of Alu in exonisation of intronic transposable elements leading to the generation of alternative exons,34 there are few investigations on the influence of intronic Alu elements on exon skipping.35-37
HnRNPA1 has recently been shown to interact with dsDNA through the HREs.38,39 This is in agreement with our earlier investigations20,40 showing binding of this protein to HREs within Alu elements. In this study, we show that hnRNPA1 binds to RNA transcribed from an Alu element inserted in the sense orientation (Fig. 2). This RNA contains a sequence (GCGGGA) that partially matches the previously reported high-affinity binding sequence for hnRNPA1 (UAGGGA).41 Although hnRNPA1 alone does not cause exon skipping (Fig. 3), knocking down expression of this protein by siRNA significantly reduces the amount of products with skipped exons 7 and 8 (Fig. 4). This observation is in agreement with recent experiments performed on hnRNPA1 deficient cells42 and demonstrates that this protein is an important component of the mechanism of alternative splicing of the APP pre-mRNA.
A model for the role of hnRNPA1 in alternative splicing of hnRNPA1 pre-mRNA itself has been proposed.43 Here, an interaction between hnRNPA1 molecules bound to the two flanking introns of exon 7B of the hnRNPA1 gene is suggested to form a loop. This brings the splice site of exon 7 closer to the splice site of exon 8, increasing the frequency of exon 7B skipping. A similar mechanism could be proposed for skipping of exons 7 and 8 in the APP pre-mRNA, as these exons are flanked by introns both containing Alu elements in the sense orientation encoding hnRNPA1 binding sites. This model is further supported by our experiments (Fig. 3) demonstrating a requirement for the hnRNPA1 binding site in the introns flanking exons 7 and 8 for optimal exon skipping. However, in addition to the proposed model, we found that the formation of such a loop by hnRNPA1 bound to introns 6 and 8 is not sufficient, but that binding of SC35 to its consensus binding motif in intron 7 is required for exon skipping.
It has been reported that insertion of an Alu element into intron 18 of the Factor VIII gene leads to skipping of exon 19 and results in a severe form of hemophilia A.36 Similar observations have been reported for patients with Dent’s disease37 and neurofibromatosis35 where insertion of intronic Alu element into the CLCN5 and NF1 genes, respectively, leads to the introduction of exonic splicing enhancers and subsequent exon skipping. These findings support an SC35/hnRNPA1/Alu mediated model for exon skipping, and suggest that it may be more common in humans than previously appreciated.
With the recognition of the importance of splicing defects in human disease has come a realization that regulated splicing reactions are potential therapeutic targets. Different approaches have been proposed so far, from conventional small-molecule drugs to RNA-based gene therapy.44 Our study suggests that SC35 and hnRNPA1 might be useful targets for reducing formation of senile plaques in AD brains by using different steroid hormones, for example estradiol (Fig. 7). Our data show that such a treatment of neuronal cells modulates alternative splicing in the APP pre-mRNA shifting the balance of the APP isoforms to APP695 that reduces secretion of the Aβ. This occurs together with increased expression of SC35 and hnRNPA1 splicing factors that regulate skipping of exons 7 and 8 during pre-mRNA maturation. These findings are in agreement with others45,46 demonstrating increased expression of SR proteins and hnRNPF, and stimulation of alternative splicing upon E2 treatment. Notably, hnRNPF/H and hnRNPA/B have been recently found to stimulate pre-mRNA splicing, both constitutive and alternative, through their binding sites in introns.47
We are grateful to Dr. A. Fujiyama, RIKEN-GSC, Japan for providing clone S491 used in amplification of the APP minigenes, and Dr. R. Killick, Institute of Psychiatry, Denmark Hill, London, and Dr. Sara Nakielny, Cancer Research UK London Research Instititute, for helpful discussions. This work was partly supported by Cancer Research UK.