Neuronal ELAVL proteins utilize AUF-1 as a co-partner to induce neuron-specific alternative splicing of APP

Abstract

Aβ peptide that accumulates in Alzheimer's disease brain, derives from proteolytic processing of the amyloid precursor protein (APP) that exists in three main isoforms derived by alternative splicing. The isoform APP695, lacking exons 7 and 8, is predominately expressed in neurons and abnormal neuronal splicing of APP has been observed in the brain of patients with Alzheimer's disease. Herein, we demonstrate that expression of the neuronal members of the ELAVL protein family (nELAVLs) correlate with APP695 levels in vitro and in vivo. Moreover, we provide evidence that nELAVLs regulate the production of APP695; by using a series of reporters we show that concurrent binding of nELAVLs to sequences located both upstream and downstream of exon 7 is required for its skipping, whereas nELAVL-binding to a highly conserved U-rich sequence upstream of exon 8, is sufficient for its exclusion. Finally, we report that nELAVLs block APP exon 7 or 8 definition by reducing the binding of the essential splicing factor U2AF65, an effect facilitated by the concurrent binding of AUF-1. Our study provides new insights into the regulation of APP pre-mRNA processing, supports the role for nELAVLs as neuron-specific splicing regulators and reveals a novel function of AUF1 in alternative splicing.

Figure 1. Expression of nELAVLs correlates with the APP695-specific pre-mRNA processing.

(A) High-throughput sequencing data from 17 human samples were analyzed for the association between APP isoforms and Hu, AUF-1 and TIA-1 mRNA expression. Color intensity and circle size indicate the strength of the correlation. Note that only APP695 is correlated with nELAVL expression. (B) Schematic representation of APP AS in neuronal and non-neuronal cells and localization of primers used in this study (arrows, F: forward, R: reverse) Boxes indicate exons and bold lines introns. RT-PCR was carried out using total RNA isolated from five cell lines and primary cortical neurons with primers specific for human or mouse ELAVL1, ELAVL2, ELAVL3, ELAVL4, and primers for human or mouse APP that allow the simultaneous detection of all AS events of exons 7 and 8. The indicated amplification bands resulting from AS of APP exons 7 and/or 8 were identified by their respective length. Quantification of the results was performed by scanning densitometry and the percentage of exclusion of both exons 7 and 8 is indicated below each lane. (C) Equal amounts of total protein from lysates of five cell lines and cortical neurons were analyzed on SDS-PAGE and immunoblotted with antibodies specific for ELAVLs, APP and GAPDH as a loading control. Note that similar to cortical neurons, significant levels of APP695 were observed only in the cells lines expressing nELAVLs. (D) Human SH-SY5Y and mouse CAD neuroblastoma cells were differentiated into a neuronal-like phenotype. Equal amounts of total protein from lysates of the above untreated and differentiated cells were analyzed on SDS-PAGE and immunoblotted with antibodies specific for the neuronal markers β-ΙΙΙ tubulin and SAP97 as well as for APP, ELAVLs and finally GAPDH as a loading control. Note that differentiated SH-SY5Y and CAD cells displayed a concurrent upregulation of nELAVLs and APP695.

Figure 2. nELAVLs promote the exclusion of exons 7 and 8 from the endogenous APP pre-mRNA.

(A) Schematic representation of exons 6 to 9 in the APP cDNA and localization of primers used for the simultaneous detection of all transcript variants (arrows, F: forward, R: reverse). PCR product size indicative of each transcript is depicted below. (B–F) Human SK-N-SH cells were transfected with the pCAGGS expression vector bearing either no insert or one ELAVL; the effect of ELAVLs on the inclusion of APP exons 7 and 8 was assayed two days later. (B) Splicing pathways were determined by RT-PCR using the primers shown in panel A. Amplification bands resulting from AS of APP exons 7 and/or 8 are indicated. Quantification of the results was performed by scanning densitometry and the percentage (mean ± standard deviation) of each transcript variant is depicted below each lane. (*P < 0.05, **P < 0.01) (C–E) Total RNA from the transfected SK-N-SH cells was also used in RT-qPCR experiments with primers specific for APP770 (C, arrows), APP695 (D, arrows) and APP751 (F, arrows). Total APP cDNA was used for normalization. Bars in graphs correspond to mean ± standard deviation of three independent experiments (**P < 0.01, ***P < 0.001). (F) Equal amounts of total protein from lysates of the same transfected SK-N-SH cells were analyzed on SDS-PAGE, immunoblotted with an antibody against APP, ELAVLs and GAPDH, as a loading control. Note that overexpression of ELAVL2, ELAVL3 or ELAVL4 promoted APP695-specific AS. (G) Mouse Neuro-2a cells were transfected with an expression vector carrying a shRNA specific for eGFP (control) Elavl2, Elavl3 or Elavl4 mRNA. The effect of reduced nELAVL expression on the AS pattern of App exons 7 and/or 8 was assayed by RT-qPCR using primers specific for App770, App751 and App695. Total App cDNA was used for normalization. Note that Neuro-2a cells with reduced levels of ELAVL2, ELAVL3 or ELAVL4 displayed significant changes in the relative expression of App770, App751 and App695, indicative of suppressed exclusion of exons 7 and 8 from the App mRNA. Bars in graphs correspond to mean ± standard deviation of four independent experiments (*P < 0.5, **P < 0.01, ***P < 0.001).

Figure 3. Association of nELAVs with the APP pre-mRNA.

(A) Localization of primers used for the detection of APP/App pre-mRNAs (arrows; APPI/AppI: Forward I and Reverse I; APPII/AppII: Forward II and Reverse II). Boxes indicate exons and bold lines introns. (B–D) Nuclear extracts prepared from human SK-N-SH (B), human SH-SY5Y (C) and mouse Neuro-2a (D) cells were immunoprecipitated with a mouse anti-ELAVL antibody or mouse IgGs as a control. RNA was isolated from immunoprecipitates, as well as their supernatants and analyzed by semi-quantitative and quantitative RT-PCR, respectively, using specific primers against human or mouse APP pre-mRNA (arrows) and intronic GAPDH. Minus RT lanes are included as controls. Note that APP pre-mRNA was detectable in the immunoprecipitate only when lysates from cells expressing nELAVLs were used. Bars in graphs depict mean ± standard deviation of three independent experiments (*P < 0.5, **P < 0.01).

Figure 4. nELAVLs promote APP exon 7 and/or 8 skipping from APPE78 minigene transcripts.

(A) Normalized nELAVL binding map in human APPE78 region. IGV Browser snapshot depicting the nELAVL binding profiles in all 17 human brain samples surrounding exons 7 and 8. Note that there is a cluster of nELAVL binding sites spanning the intronic region in-between exons 7 and 8, while three of the identified peaks are found in close proximity to exon 8. nELAVL peaks, are located on chromosome 21: 25,996,827–26,000,394 bp. (B) Schematic representation of human APPE78 and mouse AppE78 minigenes. Boxes indicate exons (AE: artificial exon; pA: poly-A signal; pCMV: CMV promoter; SA: splice acceptor, SD: splice donor). (C) Human SK-N-SH cells were co-transfected with the human or the mouse APPE78 minigene along with the pCAGGS expression vector bearing either no insert (empty) or one of the three nELAVLs. Splicing pathways were determined by RT-PCR using primers specific for artificial exons I and II. Amplification bands resulting from AS of APP exons 7 and/or 8 are indicated. Quantification of the results was performed by scanning densitometry and the percentage (mean ± standard deviation) of each transcript variant is depicted below each lane. Note that apart from transcripts lacking both APP exons, significant levels of transcripts lacking only exon 8 were also observed upon nELAVL-overexpression.

Figure 5. Identification of a U-rich element important for nELAVL-mediated APP exon 8 skipping.

(A,B) Schematic diagrams depicting the regions of human (A) and mouse (B) APP locus used for the generation of APPE8 and AppE8 minigenes, respectively. The gray box represents exon 8, bold lines its flanking intronic sequences, whose length is depicted, and triangles segments encoding for U-rich regions in the pre-mRNA. The 47 nt upstream sequence of wild-type and mutant APPE8.4 and AppE8.5 is also shown. (C,D) Human SK-N-SH cells were co-transfected with human (C) or mouse APPE8 (D) minigenes and the pCAGGS expression vector bearing no insert (empty) or ELAVL4. Splicing pathways were determined by RT-PCR using primers specific for the artificial exons of the minigene vector. Amplification bands resulting from exon 8 skipping are indicated. Quantification of the results was performed by scanning densitometry and the percentage of exon 8 exclusion (mean ± standard deviation) is displayed in graphs (**P < 0.01, ***P < 0.001). Note that ELAVL4 efficiently promoted exclusion of exon 8 from all wild-type artificial transcripts, but this effect was compromised in the case of mutant ones. (E,F) Biotinylated RNA segments transcribed from wild-type and mutant APPE8.4 (E) and AppE8.5 (F) minigenes were tested for nELAVL-binding after incubation with lysates of Neuro-2a cells and pull-down using streptavidin beads. The presence of nELAVLs was assayed by immunoblotting. Note that nELAVLs strongly associated with the wild-type transcripts, but this association was weaker when the mutant transcripts were used.

Figure 6. nELAVL-binding to sequences located both upstream and downstream of exon 7 is required for nELAVL-mediated APP exon 7 exclusion.

(A,B) Schematic representation of human (A) and mouse (B) APP genomic regions used for the generation of APPE7 and AppE7 minigenes, respectively. The gray box corresponds to exon 7, bold lines to its flanking intronic sequences, whose length is indicated and triangles to segments encoding for U-rich sequences in the pre-mRNA. (C,D) Human SK-N-SH cells were co-transfected with either APPE7 (C) or AppE7 (D) minigenes and the pCAGGS expression vector bearing ELAVL4. Splicing pathways were determined by RT-PCR using primers specific for the artificial exons of the minigene vector. Amplification bands of transcripts lacking exon 7 are shown. Quantification of the results was performed by scanning densitometry. Bars in graphs depict the percentage of exon 7 skipping (mean ± standard deviation). Note, that ELAVL4 promoted exon 7 exclusion only from transcripts containing U-rich sequences both upstream and downstream of this exon (E,F) Biotinylated RNA probes transcribed from the indicated human (E) and mouse (F) APPE7 minigenes were tested for nELAVL-binding after incubation with lysates of Neuro-2a cells and pull-down using streptavidin beads. The presence of nELAVLs was assayed by immunoblotting. Note, that nELAVLs strongly associated with both flanking intronic regions of APP exon 7.

Figure 7. ELAVL4 protein interferes with the binding of essential splicing factor U2AF65 upstream of APP exons 7 and 8.

Human SK-N-SH cells were co-transfected with two expression plasmids carrying either no insert (empty) or ELAVL4 in ratios depicted above each lane. (A) In order to ensure the presence of different ELAVL4 protein levels among the three conditions, equal amounts of total protein from lysates of the transfected SK-N-SH cells (input) were analyzed on SDS-PAGE and immunoblotted with antibodies against ELAVLs and β-TUBULIN (TUBB) as a loading control; membranes were also probed against U2AF65, to verify that ELAVL4 overexpression does not alter U2AF65 expression. (B) Whole cell lysates from the same transfected cells were used in a series of pull down assay using streptavidin beads and biotinylated RNA probes transcribed from the indicated human minigenes. The presence of U2AF65 and ELAVLs in the precipitates was assayed by immunoblotting. Quantification of the results was performed by scanning densitometry and the percentage (mean ± standard deviation) change of U2AF65-binding upon ELAVL4 overexpression is shown below each lane. Note, that ELAVL4-binding to these transcripts resulted in a reduction of U2AF65 binding (***P < 0.001).

Figure 8. AUF-1 facilitates ELAVL4-mediated APP695-specific AS.

Biotinylated RNA probes transcribed from human (A) and mouse (B) APP minigenes were tested for AUF-1 binding. Note, that AUF-1 isoforms p42 and/or p45 bind to all transcripts shown to interact with nELAVLs. (C) SK-N-SH cells were transfected with the DNA3.1 expression vector bearing either no insert (empty) or p42 AUF-1, or with the pENTR/U6 vector carrying a shRNA specific for LacZ or all AUF-1 mRNAs. Inclusion of APP exons 7 and 8 was assayed by RT-qPCR with primers specific for APP770, APP695 and APP751. Total APP cDNA was used for normalization. Bars in graphs correspond to mean ± standard deviation of three independent experiments. (D,E) SK-N-SH cells were co-transfected with the human (D) or the mouse (E) APPE78 minigene along with two expression vectors, one bearing ELAVL4 and the other one carrying no insert (empty) or p42 AUF-1, in ratios depicted. Splicing pathways were determined by RT-PCR, as described in Fig. 5. Quantification of the results was performed by scanning densitometry. Note, that AUF-1 enhances ELAVL4-mediated exclusion of APP exons 7 and 8. (F) Neuro2a cells were transfected with the pSilencer vector bearing either no insert (empty) or a shRNA targeting all Auf-1 transcripts. Inclusion of APP exons 7 and 8 was assayed by RT-qPCR, as described above (**P < 0.01). Note, that in the presence of nELAVLs, reduction of AUF-1 levels results in reduced APP695-specific AS. (G) SK-N-SH cells were co-transfected with two out of three expression plasmids carrying no insert (empty), ELAVL4 or p42 AUF-1 in ratios depicted. Whole cell lysates from these cells were used in a series of pull down assay using streptavidin beads and APP7.4 biotinylated riboprobe. The presence of ELAVLs, AUF-1 and U2AF65 in precipitates was assayed by immunoblotting. Quantification of the results was performed by scanning densitometry and the percentage (mean ± standard deviation) change of U2AF65-binding is shown below each lane. Note that concurrent binding of both proteins in the transcript resulted in a significant reduction of U2AF65 binding. (*P < 0.05, **P < 0.01).