TDP-43 binds to RNA G-quadruplex structure and regulates mRNA stability and translation

Abstract

TDP-43 is a hallmark protein associated with neurodegenerative diseases. Recent studies revealed TDP-43 as an RNA G-quadruplex (rG4)-binding protein, impacting mRNA transport and function. However, our knowledge of the TDP-43-RNA secondary structure interaction and information on its specific rG4 targets are limited. Herein, we show that TDP-43 exhibits a preference for binding to the rG4 under K+ condition using high-throughput RNA bind-n-seq. Besides, we find that the loss of TDP-43 contributes to a transcriptome-wide decrease in mRNA structure using SHALiPE-seq technology. By analyzing the SHALiPE-seq data of TDP-43-binding sites, we demonstrate that the reduction in structuredness is likely due to the loss of TDP-43 binding to the RNA targets, especially in the 3'UTR. Importantly, our transcript-specific investigation reveals that TDP-43 binds to 3'UTR rG4 of SLC1A5 transcript, promoting its mRNA stability and translation. Removing the rG4 and incorporating BRACO-19 competition result in translation inhibition of SLC1A5, highlighting the importance of rG4 in gene regulation by TDP-43. Our findings not only offer new insights into the role of TDP-43 in regulating RNA structures such as rG4 but also contribute to a better understanding of its broader functions and provide potential targets for therapeutic strategies in TDP-43-related diseases.

Graphical Abstract

Figure 1.

RBNS showed that TDP-43 preferentially binds to potential G-quadruplex motifs. (A) Schematic illustration of RBNS. Different concentrations of purified His-tagged TDP-43 protein are incubated with a random RNA pool under either potassium (K⁺) or lithium (Li⁺) condition. TDP-43 is pulled down using Ni-NTA magnetic beads, and the associated RNA is enriched for library preparation and sequencing. Ni-NTA bead, anti-his tag coated bead. (B) Profile of top ten enriched sequence motifs across five different TDP-43 concentrations under K⁺ condition and Li⁺ condition. TDP-43 could enrich G-rich RNA sequences in the presence of K⁺ condition to a greater extent compared to UG-rich RNA sequences in the presence of Li⁺ condition. R-value, the frequency of 6-mer in TDP-43-bound reads divided by its frequency in input reads. (C) Putative G-quadruplex sequences in the TDP-43-associated RNA in the presence of K⁺ and Li⁺. The enrichment of TDP-43-bound rG4 sequence motifs is higher in K⁺ condition compared to Li⁺ condition. Bars, means of bootstrapped subsets. Error bars, standard deviation across subsets, Student’s t-test, ***: P value < 0.001. (D) Enrichments for G4 structure as predicted by computational RNA folding in TDP-43-associated RNA relative to input RNA in the presence of K⁺ and Li⁺. The structure-folded rG4s in the TDP-43-bound RNAs were more enriched in the K⁺ condition compared to the Li⁺ condition. Bars, means of bootstrapped subsets. Error bars, standard deviation across subsets, Student’s t-test, ***: P value < 0.001.

Figure 2.

SHALiPE-seq on HEK293T cells with or without TDP-43 knockdown yields RNA structurome data with high accuracy and reproducibility. (A) WB results of TDP-43 expression in control (Ctrl) and TDP-43 knockdown (TDP-43 kd) HEK293T cells. TDP-43 protein expression was downregulated in TDP-43 knockdown cells compared to control cells. Ctrl1/2 was samples transfected with two shRNAs that targeted firefly luciferase as negative control, and TDP-43 kd1/2/3 are samples that were transfected with three shRNAs (shTDP43-1/2/3) that targeted different region of CDS of TDP-43 gene TARDBP. GAPDH is used as an internal control. (B) Schematic illustration of SHALiPE-seq applied to control and TDP-43 knockdown cells. Briefly, ctrl and TDP-43 kd cells are collected for DMSO/NAI probing followed by total RNA extraction and polyadenylated RNA selection. The library preparation process includes fragmentation, 3′dephohosphorylation, 3′-adapter ligation, reverse transcription (RT), 5′-adapter ligation, and PCR amplification. The libraries are submitted to sequencing for data analysis. RNA structural difference is identified by discriminative signals of NAI reactivities in ctrl and TDP-43 kd cells. (C) The denaturing polyacrylamide gel image shows consistent modification of NAI reactivity on the cDNA of 5.8S rRNA from ctrl and TDP-43 kd cells. The NAI modification nucleotides appear in NAI-treated samples (lanes 6, 8, 10, and 12) but not in DMSO-treated samples (lanes 5, 7, 9, and 11). DMSO is used as the control for chemical probing. Lanes 1–4 are reference nucleotides generated by adding ddA, ddT, ddG, and ddC in the RT reactions. (D) Correlation between RT stops in two replicates with r values ranging from 0.82 to 0.91. r: Pearson correlation coefficient. (E) ROC curves showing the accuracy of 18S rRNA SHALiPE-seq reactivity of ctrl and TDP-43 kd samples against reference structure with AUC values of 0.750 and 0.764. Black point, the optimal cutoff value for the ROC curve.

Figure 3.

The effect of TDP-43 on mRNA secondary structure. (A) The average NAI reactivity of mRNAs between control and TDP-43 knockdown cells. TDP-43 kd cells showed higher NAI reactivity compared to ctrl cells. P-value, statistical significance calculated by Wilcoxon signed-rank test. The average NAI reactivity of transcript regions in the (B) 5′UTR, (C) CDS, and (D) 3′UTR in ctrl are higher than that in TDP-43 kd cells. The largest difference in NAI reactivity is observed in the 3′UTR. (E) The binned average NAI reactivity and Δ Reactivity (kd-ctrl) of mRNAs for ctrl and TDP-43 kd samples across the length of transcript regions. The NAI reactivity in ctrl samples is lower than TDP-43 kd samples, showing low TDP-43 leads to a loss in RNA structureness across the whole transcript region. 5′UTRs: 25 bins, CDS: 50 bins, 3′UTRs: 25 bins. The gray bars represent 95% confidence intervals of the average Δ Reactivity of each bin (paired two-sided Student’s t-test). (F) The number of DRR is calculated for each transcript region.

Figure 4.

The effect of TDP-43 on the localized structure of its binding targets. TDP-43-bound mRNAs in TDP-43 kd samples showed higher average reactivity than ctrl samples in the (A) full transcript, (B) 5′UTR, (C) CDS, and (D) 3′UTR. TDP-43-bound mRNAs were identified from published iCLIP data [48]. P-value, statistical significance calculated by Wilcoxon signed-rank test. (E) TDP-43-bound mRNAs showing increased NAI reactivity than unbound transcripts in TDP-43 kd samples compared to ctrl samples. (F) The binned average reactivity and Δ Reactivity (kd-ctrl) of TDP-43-bound mRNAs across the length of transcript regions. The difference in NAI reactivity is largest in the 3′UTR followed by CDS and 5′UTR. 5′UTRs: 25 bins, CDS: 50 bins, 3′UTRs: 25 bins. The grey bars represent 95% confidence intervals of the average Δ Reactivity of each bin (paired two-sided Student’s t-test). (G) The average NAI reactivities across TDP-43-binding sites in TDP-43 kd samples are higher than in ctrl samples. (H) The number of TDP-43-binding sites with an increase in NAI reactivity is greater than that of decrease in NAI reactivity in TDP-43 kd samples compared to ctrl samples. (I) Density of DRRs with TDP-43-binding sites and increased NAI reactivity (increased DRR) across transcript regions in TDP-43 kd cells compared to ctrl cells. Density was normalized by TDP-43-binding frequency.

Figure 5.

TDP-43 binds to a G-quadruplex structure in the 3′UTR of SLC1A5 RNA. (A) Proportional enrichment of rG4s versus non-G4 motifs in TDP-43-binding sites with DRRs relative to non-DRR sites. rG4s constitute a higher proportion than nonG4 in TDP-43-binding sites with DRRs in comparation to that of without DRRs. (B) RIP-qPCR for the mRNA candidates from panel (A) in the HEK293T cell lysate upon IgG and TDP-43 IP. SLC1A5 RNA was significantly highly enriched by TDP-43 compared to IgG control. *: 0.01 < P-value < 0.05, ***: P-value < 0.001. (C and D) CD signal for SLC1A5G4wt and SLC1A5G4mut, respectively. The CD profiles of SLC1A5 G4wt RNA and SLC1A5 G4mut RNA under K⁺ and Li⁺ conditions revealed the formation of SLC1A5 G4wt RNA in the K⁺ condition but not the Li⁺ condition. While SLC1A5 G4mut RNA cannot form G4 structure under the same condition. (E) Biotinylated pull-down assay with HEK293T cell lysate by SLC1A5 G4wt and G4mut RNA oligos. The TDP-43 protein enrichment was shown by WB. GAPDH was used as control protein. SLC1A5 G4mut is used as the negative control. Beads control: only streptavidin magnetic bead and cell lysate are incubated for pull-down. Ten micrograms input of cell lysate was shown as control reference. SLC1A5 G4mut was used as the negative control. (F) The MST-binding assay showed that TDP-43 binds to SLC1A5 G4wt RNA directly. G4mut is used as a negative control. (G) Confocal images of transfected SLC1A5 G4wt RNA oligo with FAM probes and endogenous TDP-43 in SH-SY5Y cells; scale bar: 25 μm. Nuclei are stained with Hoechst 33342. (H) Colocalization analysis of TDP-43 and SLC1A5 G4. Manders’ colocalization coefficient was calculated based on cell images in panel (G). The fraction of SLC1A5 G4wt overlapping TDP-43 is significantly higher than that of SLC1A5 G4mut. ***: P value < 0.001.

Figure 6.

TDP-43 reduction leads to SLC1A5 mRNA instability and translation inhibition. (A) Relative expression of SLC1A5 mRNA in control and TDP-43 knockdown HEK293T cells at different time points following actinomycin D treatment. GAPDH was used as an internal control for normalization. *: 0.01 < P-value < 0.05, **: 0.001 < P-value < 0.01. (B) Schematic illustration of dual luciferase reporter plasmids for mRNA stability detection. (C) Normalized mRNA was determined by quantifying the Renilla/Firefly mRNA abundance at different time points following actinomycin D treatment. n.s.: not significant. *: 0.01 < P-value < 0.05. **: 0.001 < P-value < 0.01. (D) MeRIP-qPCR Analysis of the SLC1A5 G4 region in control and TDP-43 knockdown cells Using IgG and m6A immunoprecipitation. *: 0.01 < P < 0.05. (E) WB results of SLC1A5 protein expression in ctrl and TDP-43 kd HEK293T cells. TDP-43 and GAPDH are used as controls. (F) Schematic illustration of dual luciferase reporter plasmids. The 3′UTR of SLC1A5 including G4wt, G4 removal, and UG removal was individually inserted into the dowstream of Renilla luciferase. (G) Normalized luciferase activity of cells transfected with plasmids of panel (D). Renilla gene activity is reduced by removing G4 sequence, while no significant changed are observed by removing UG repeats compared to G4wt. *: 0.01 < P-value < 0.05. n.s.: no significant difference with P-value > 0.05. (H) Saturation plot of BRACO-19 for its inhibition of SLC1A5 rG4–TDP-43 interaction. The IC₅₀ was calculated as 5.84 ± 0.49 μM. (I) Normalized luciferase activity of cells transfected with plasmids of SLC1A5 3′UTR G4wt and 15 μM BRACO-19. Renilla gene activity is reduced by BRACO-19 addition compared to untreatment (UT) ctrl. **: 0.001 < P-value < 0.01. (J) WB results of SLC1A5 protein expression in ctrl and TDP-43 kd HEK293T cells with or without BRACO-19 treatment. BRACO-19 treatment contributed to the SLC1A5 decrease in HEK293T cells. TDP-43 and GAPDH are used as controls.