TDP-43 regulates the alternative splicing of hnRNP A1 to yield an aggregation-prone variant in amyotrophic lateral sclerosis

Abstract

See Fratta and Isaacs (doi:10.1093/brain/awy091) for a scientific commentary on this article.The RNA binding proteins TDP-43 (encoded by TARDBP) and hnRNP A1 (HNRNPA1) are each mutated in certain amyotrophic lateral sclerosis cases and are often mislocalized in cytoplasmic aggregates within motor neurons of affected patients. Cytoplasmic inclusions of TDP-43, which are accompanied by a depletion of nuclear TDP-43, are observed in most amyotrophic lateral sclerosis cases and nearly half of frontotemporal dementia cases. Here, we report that TDP-43 binds HNRNPA1 pre-mRNA and modulates its splicing, and that depletion of nuclear TDP-43 results in increased inclusion of a cassette exon in the HNRNPA1 transcript, and consequently elevated protein levels of an isoform containing an elongated prion-like domain, referred to as hnRNP A1B. Combined in vivo and in vitro approaches demonstrated greater fibrillization propensity for hnRNP A1B, which drives protein aggregation and is toxic to cells. Moreover, amyotrophic lateral sclerosis patients with documented TDP-43 pathology showed neuronal hnRNP A1B cytoplasmic accumulation, indicating that TDP-43 mislocalization may contribute to neuronal vulnerability and loss via altered HNRNPA1 pre-mRNA splicing and function. Given that TDP-43 and hnRNP A1 each bind, and thus modulate, a third of the transcriptome, our data suggest a much broader disruption in RNA metabolism than previously considered.

Figure 1

Loss of nuclear TDP-43 increases hnRNP A1B protein levels. (A–C) siRNA-treated cells were (A) immunoblotted for TDP-43, Actin and hnRNP A1/A1B (9H10) and (B) quantified by densitometry. (C) The level of hnRNP A1B was also assessed by immunofluorescence using a custom hnRNP A1B specific antibody and the mean intensity was quantified. (D and E) Cells were depleted of endogenous TDP-43 and transfected with mock or Flag-TDP-43^ΔNLSsiRES cDNA (ΔNLS). Levels of hnRNP A1 and hnRNP A1B were assessed by western blot and quantified by densitometry. (F and G) Nuclear and cytoplasmic fractions from siRNA-treated cells were immunoblotted for TDP-43, hnRNP A1/A1B (9H10), Tubulin and HuR and quantified by densitometry. For all, the mean ± SEM of three to five independent experiments is plotted. *P < 0.05, **P < 0.02. Scale bar = 25 µm.

Figure 2

TDP-43 binds and modulates HNRNPA1 mRNA. (A) Quantitative PCR with a probe set recognizing both hnRNP A1 and A1B (exon 1–2) or only hnRNP A1B (exon 7b-8) transcripts. Data was normalized to ACTN. (B) Quantitative PCR for hnRNP A1B transcripts, normalized to 18S, following actinomycin D treatment. (C) Luciferase assay using HeLa cells first treated with siRNA and then co-transfected with the reporter vectors containing the luciferase gene located downstream of the HNRNPA1 3′ UTR. GAPDH 3′ UTR and a random sequence (R03) serve as negative controls. GRN is included as positive control. (D) Luciferase assay using HeLa cells first treated with siRNA and then co-transfected with the reporter vectors containing the luciferase gene upstream of the promoter. The GAPDH promoter and random sequence (R01) serve as negative controls. (E) TDP-43 protein and its associated transcripts were co-immunoprecipitated from HeLa cells. TDP-43 bound mRNAs were extracted from the precipitates, reverse transcribed and amplified for HNRNPA1, TARDBP, CAMK2A and HSPA4. FLAG IP serves as a control. Dotted lines indicate that the samples are not contiguous lanes. For all, the mean ± SEM of three to four independent experiments is plotted. *P < 0.05, **P < 0.02.

Figure 3

TDP-43 influences HNRNPA1 pre-mRNA splice site selection in vitro. (A) RBPmap results demonstrate potential TDP-43 binding sites (red lines) in the introns upstream and downstream of exon 7B. (B) RT-PCR for endogenous hnRNP A1- and A1B-encoding transcripts from HeLa cells treated with control or TDP-43 siRNA. (C) Structure of the model HNRNPA1 pre-mRNA/mini-gene. The two 5′ splice sites (ss) of exon 7 and exon 7B compete for a unique 3′ splice sites (adenovirus). (D) Western blot for TDP-43 in varying volumes of HeLa nuclear extracts depleted or not of endogenous TDP-43 (left). Different amounts of recombinant TDP-43 protein were also loaded and detected simultaneously (right). (E and F) Spliced products were amplified by RT-PCR and separated on native PAGE gels. Oligo (TG)₁₃ competes away the activity of endogenous TDP-43. The percentage of proximal splicing was quantified and expressed as mean ± SEM of three independent experiments, **P < 0.02, ***P < 0.005. (G and H) Addition of recombinant TDP-43 protein restores splicing activity in extracts prepared from cells depleted of TDP-43 by stable shRNA expression. The percentage of proximal splicing was quantified and expressed as mean ± SEM of four independent experiments. **P < 0.02.

Figure 4

Fibrillization propensity of hnRNP A1B-derived peptides. (A) HnRNP A1 and hnRNP A1B were analysed using ZipperDB. Enlarged views of ZipperDB output illustrating the peptides generated by exon 7B inclusion and selected for further study. (B and C) Thioflavin T (ThT) incorporation assay indicates fibrillization of the peptides generated by (B) exon 7B inclusion or the ALS-causing mutation D262V. (C) Transmission electron micrographs demonstrating the fibrillization of the peptides. Data are representative of three to six experiments. Scale bar = 500 nm.

Figure 5

The inclusion of exon 7B promotes aggregation in vivo. (A and B) HeLa cells were transfected with the indicated Myc-tagged cDNA and immunostained for Myc. Scale bar = 10 µm. Arrows indicate cytoplasmic aggregates. (B) Quantification of the percentage of cells with cytoplasmic aggregates and (C) quantification of the size of the aggregates. The mean ± SEM of 3–4 independent experiments is plotted, a minimum of 100 cells per genotype were analysed per experiment. *P < 0.05, ***P < 0.015. (D) HEK293FT cells were transfected with the indicated cDNA. Aggregation was assayed by filter retardation using non-binding cellulose acetate membranes. (E) Whole cell lysates were immunoblotted for Myc and actin. Data are representative of three independent experiments.

Figure 6

HnRNP A1B expression causes cell death. (A) Whole lysates of CB3 cells, which lack endogenous hnRNP A1 or were modified to re-express one of the isoforms, were immunoblotted for hnRNP A1/A1B (9H10) and actin. (B) Viability of CB3 cell lines was assayed by trypan blue dye exclusion. The mean ± SEM of three independent experiments is plotted. *P < 0.05 **P < 0.002. (C) Aggregation was assayed by filter retardation using non-binding cellulose acetate membranes and immunoblotted for hnRNP A1/A1B (9H10). Data are representative of three independent experiments.

Figure 7

HnRNP A1B is expressed in the CNS and forms cytoplasmic aggregates in ALS patient neurons. (A) Rat tissue lysates were immunoblotted for hnRNP A1/A1B (9H10), GAPDH and actin. (B) Low magnification images of hnRNPA1 and hnRNP A1B, detected with antibody 4B10 (left) or a custom anti-hnRNP A1B antibody (right). Low magnification scale bar = 100 µm. Higher magnifications are from boxed regions in lower magnification images in B, scale bar = 10 µm. (C) Confocal images of spinal motor neurons from two ALS cases stained for hnRNP A1B and TDP-43. Scale bar = 5 μm.