A G-rich element forms a G-quadruplex and regulates BACE1 mRNA alternative splicing

Abstract

β-Site amyloid precursor protein (APP) cleaving enzyme 1 (BACE1) is the transmembrane aspartyl protease that catalyzes the first cleavage step in the proteolysis of the APP to the amyloid β-protein (Aβ), a process involved in the pathogenesis of Alzheimer disease. BACE1 pre-mRNA undergoes complex alternative splicing, the regulation of which is not well understood. We identified a G-rich sequence within exon 3 of BACE1 involved in controlling splice site selection. Mutation of the G-rich sequence decreased use of the normal 5' splice site of exon 3, which leads to full-length and proteolytically active BACE1, and increased use of an alternative splice site, which leads to a shorter, essentially inactive isoform. Nuclease protection assays, nuclear magnetic resonance, and circular dichroism spectroscopy revealed that this sequence folds into a G-quadruplex structure. Several proteins were identified as capable of binding to the G-rich sequence, and one of these, heterogeneous nuclear ribonucleoprotein H, was found to regulate BACE1 exon 3 alternative splicing and in a manner dependent on the G-rich sequence. Knockdown of heterogeneous nuclear ribonucleoprotein H led to a decrease in the full-length BACE1 mRNA isoform as well as a decrease in Aβ production from APP, suggesting new possibilities for therapeutic approaches to Alzheimer's disease.

Figure 1

Deletion of the G-rich sequence changes selection of the 5′ ss in BACE1. (A) Representation of the pre-mRNAs used in the in vivo splicing assay. Exon 5 as well as part of exon 4 were deleted to generate BACE1 3-trunc4 minigene. A 24-nucleotide sequence was deleted from the BACE1 3-trunc4 WT and BACE1 3-5 WT constructs to generate BACE1 3-trunc4 Delta G and BACE 1 3-5 Delta G respectively. (B) The strength of the 5′ ss in exon 3 were evaluated using the Shapiro and Senepathy scoring method. (C) Splicing products from in vivo splicing assays in HEK 293 cells were amplified by RT-PCR and fractionated on gel. Amplification of ribosomal protein L39 (RPL39) was used as a control in each reaction. The percentage of each isoform was calculated and represented in the graph. The experiment was done in triplicate and is representative of the three experiments. (D) BACE1 3-5 WT and BACE1 3-5 Delta G minigenes were transfected in triplicate. After RNA purification, splicing products were amplified and quantified by real-time PCR. Quantities were normalized to total amount of BACE1. Quantity of each isoform was compared between the WT and Delta G constructs. * corresponds to a statistical difference (p value <0.05).

Figure 2

The G-elements are important for the selection of the 5′ ss. (A) Schematic representation of the minigenes transfected in HEK 293 cells (top panel). The first, second and fourth GGG elements were mutated to GAG to generate m1, m2 and m4 constructs respectively. The GGGG motif was modified to GAAG (m3). Following RNA purification, splicing products were quantified by real-time PCR and normalized to total amount of BACE1 (bottom panel). Each experiment was done in triplicate. Quantity of each isoform was compared relative to the WT construct and * represents a p value <0.05. (B) A 46-nucleotide sequence was inserted into the central exon of DUP 5.1 HpaI minigene (DUP + G). A control sequence in which the GGGG motif was mutated to GAAG was also inserted (DUP + m3). Semi-quantitative PCR was carried out, and amplicons were loaded on gel (bottom panel). The percentage of skipped isoforms were calculated and represented in the graph. The experiment was performed in triplicate. Mean corresponding to the percentage of skipped isoform of DUP + G minigene was compared with the DUP construct and * corresponds to a p value <0.05.

Figure 3

The G-rich sequence can fold into a G-quadruplex structure. (A) 5′ biotinylated RNA used in the RNase T1protection assay and fifteen potential cleavage sites are represented by triangles (top panel). RNA was digested in KCl-, NaCl- or LiCl-containing buffer for 0.5, 1, 2 and 5 min. Lanes designated L and – represent an alkaline hydrolysis and a control reaction (without enzyme) respectively. The position of the residues G1 to G15 is represented. * corresponds to a degradation product. The lower portion of the film has been exposed a little bit longer in order to visualize smaller fragments. (B) Spectral changes of the 1D ¹H NMR imino proton region as a function of increasing the temperature. 0.6 mM RNA in 10 mM cacodylic acid pH 6.5 plus 150 mM KCl. (C) CD spectra of the WT 33-nt sequence in the presence of KCl, NaCl or LiCl. The parallel conformation proposed is represented on the left part of the panel.

Figure 4

Identification of a potential trans-acting factor that binds the G-rich sequence. (A) RNA oligonucleotide containing the G-elements was immobilized on adipic acid dihydrazide-agarose beads. The RNA-bound beads were incubated in the presence of HEK 293 nuclear extract and proteins were eluted with 100 mM KCl (two times), followed by 250, 500 and 1000 mM. As a control, mock treated beads (without RNA) were used to detect non-specific binding of proteins. Because of low abundance, two proteins could not be identified by MS analysis (*). (B) Pull-down of biotinylated RNA oligonucleotides (WT 33 nt and m3 33 nt) with streptavidin beads after incubation in nuclear extract. After a series of washing steps, proteins that remained bound were eluted by heating the mixture. Incubation of streptavidin beads only (without RNA) was used as control. Proteins were loaded on gel, and detection of hnRNP H protein was carried out by western blot. (C) Electrophoretic mobility shift assay using 1.5, 3 and 4 μM of recombinant his-hnRNP H protein. Binding of hnRNP H to the G-rich sequence was tested by visualization of RNA-protein complexes following addition of the recombinant protein to the WT 33 nt or m3 33 nt RNA.

Figure 5

hnRNP H activates production of the 501 isoform. (A) In vivo splicing assay using BACE1 3-5 WT minigene and siRNA against hnRNP H. The quantity of each isoform is indicated in the graph. * represents a statistical difference (p value <0.05) between the WT and WT + siH. Standard errors of the mean are too small to be visualized on the graph for isoform 476, 457 and 432. (B) RT-PCR amplification of the endogenous BACE1 after knocking down hnRNP H. The isoforms were quantified relative to the internal control RPL39. Quantity of each isoform corresponding to the siH condition was compared with the NT cells. * correspond to a statistical difference (p value <0.05) between the NT and siH. (C) Cell lines that stably express shRNA against hnRNP H (shH) and scrambled shRNA (shA) were created. * correspond to a statistical difference (p value <0.05) between shA and shH. (D) RT-PCR results after cotransfection of BACE1 3-trunc4 WT or BACE1 3-trunc4 Delta G minigenes with siRNA against hnRNP H. Isoforms 501, 457 and 127 were quantified. Statistical comparison was calculated between the indicated groups. Experiments (A, B and C) were performed in triplicate and Western blot was generated to monitored knockdown of hnRNP H.

Figure 6

Reduction of hnRNP H proteins decreases production of Aβ40. (A) HEK 293sw cell lines that stably express shRNA against hnRNP H (shH) and scrambled shRNA (shB) were created. Western blot densitometric analyses were performed to measure levels of hnRNP H relative to β-actin. Levels of hnRNP H were substantially reduced by shH (0.46) but only slightly reduced by shB (0.91) relative to NT (1.0). (B) To measure the amount of Aβ40 secreted, media was changed, and 16 h later samples were analysed by ELISA. Quantity of Aβ40 produced was compared with a non-treated cell line (NT) (Figure 6B). The experiment was performed in triplicate, and * corresponds to a statistical difference (p value <0.05).