TDP-1, the Caenorhabditis elegans ortholog of TDP-43, limits the accumulation of double-stranded RNA

Abstract

Caenorhabditis elegans mutants deleted for TDP-1, an ortholog of the neurodegeneration-associated RNA-binding protein TDP-43, display only mild phenotypes. Nevertheless, transcriptome sequencing revealed that many RNAs were altered in accumulation and/or processing in the mutant. Analysis of these transcriptional abnormalities demonstrates that a primary function of TDP-1 is to limit formation or stability of double-stranded RNA. Specifically, we found that deletion of tdp-1: (1) preferentially alters the accumulation of RNAs with inherent double-stranded structure (dsRNA); (2) increases the accumulation of nuclear dsRNA foci; (3) enhances the frequency of adenosine-to-inosine RNA editing; and (4) dramatically increases the amount of transcripts immunoprecipitable with a dsRNA-specific antibody, including intronic sequences, RNAs with antisense overlap to another transcript, and transposons. We also show that TDP-43 knockdown in human cells results in accumulation of dsRNA, indicating that suppression of dsRNA is a conserved function of TDP-43 in mammals. Altered accumulation of structured RNA may account for some of the previously described molecular phenotypes (e.g., altered splicing) resulting from reduction of TDP-43 function.

Keywords: RNA editing; RNA structure; neurodegeneration; splicing.

Figure 1. TDP-1 maintains transcripts with potential dsRNA structure

1. Scatter plot comparing the log₂ of transcript RPKMs (Reads Per Kilobase of transcript per Million mapped reads) between wild-type and tdp-1(ok803) poly(A)-selected RNA-seq experiments. Only transcripts significantly (corrected P < 0.05) increased (red) or decreased (blue) compared to wild-type are shown. Scatter plot represents significantly different genes calculated from two independent biological replicates of both wild-type and tdp-1(ok803) poly(A) RNA-seq prepared from L4 animals.

2. Percentage increased and decreased transcripts in tdp-1(ok803) poly(A) RNA-seq that have antisense overlap with another spliced gene or have intronic sequences (> 1 kb) that contain inverted repeats. *P < 1 × 10⁻⁴ (hypergeometric distribution); NS, not significant relative to control gene sets.

Figure 2. TDP-1 limits the amount of nuclear dsRNA

1. Fixed, isolated intestinal tissue probed with anti-dsRNA antibody (J2). The J2 antibody recognizes dsRNA stretches of 40 bp or more in a sequence-independent manner. Intensely stained inclusions (red dots, indicated by arrows) were detected in intestinal nuclei (blue, DAPI counterstain) specifically in tdp-1(ok803) mutant worms (middle panel). J2-reactive inclusions were observed in 28% tdp-1(ok803) intestinal nuclei scored (30/107), but not detected in wild-type controls (0/122) or tdp-1(ok803) (0/105) fixed tissue pretreated with dsRNA nuclease (V1) before J2 staining. J2-reactive foci were still observed (arrows, bottom panel) in intestinal nuclei pretreated with the ssRNA-specific nuclease (T1) (29%, 12/41). Scale bar, 20 μm.

2. Quantification of anterior J2 inclusions using ImageJ software. tdp-1(ok803) worms had significantly more J2 inclusions (*P < 0.01, Student's t-test, error bars = SEM). Representative projection images used to generate this data are shown in Supplementary Fig S4.

Figure 3. TDP-1 limits levels of A-I RNA editing

1. A Fold change in % A-to-I editing in tdp-1(ok803) compared to wild-type RNA-seq is shown. For all regions, the average percent editing for each potentially edited nucleotide was calculated. Bars in graph represent the average percent editing across individual edited regions altered in tdp-1(ok803) (73 altered/154 analyzed). Genomic location of edited regions, actual change in percent editing and actual P-values are listed in Supplementary Table S4.

2. B, C Examples of hyper-edited intronic regions in tdp-1(ok803) mutant RNA-seq: intron 3 of ABC-transporter haf-6 (B) [ChrI: 1170900-1173858] and intron 9 of FOXO transcription factor daf-16 (C) [ChrI: 10774096-10774765]. Purple bars indicate annotated structural elements containing editing. Regions were divided into 10 equal windows and mean fraction edited for each window is plotted (y-axis). Connected points indicate editing in the same structural element. Percent editing in wild-type (blue), tdp-1(ok803) (red) and adr-2(gv42) (green) is shown. Scale bars (upper left, black), 100 bp.

Figure 4. Immunoprecipitation with J2 antibody selects for dsRNA in wild-type worm extracts

1. A The J2 anti-dsRNA antibody immunoprecipitates a subset of RNAs from wild-type extracts. Plotted are RPKMs_(log2) for two independent wild-type input RNA-seq datasets (left graph) and for a J2-IP RNA-seq dataset versus input RNA (right graph). 9,086 transcripts are shown. R² values for each plot are displayed in the graph.

2. B The fold change (in RPKM) over input (y-axis) of all significantly enriched repetitive transcripts (type indicated on x-axis) and all exons (corrected P < 0.001) in wild-type J2-IP RNA-seq. Graph shows both mean (black bars) and median (gray bars) fold enrichment/input.

3. C Coverage tracks (black bumps) of alignments to consensus sequences of representative transposons enriched in J2-IP RNA compared to total RNA. Coverage track height was set to the same value for both input RNA and J2-immunoprecipitated RNA, and height is proportional to abundance of each region.

4. D The percentage of total transcripts (gray bars) with potential dsRNA of three types (antisense overlap determined from Thierry-Mieg & Thierry-Mieg, , A-to-I RNA editing taken from Supplementary Table S4 and siRNA targets taken from Warf et al, 2012) and the percentage of each of those transcript types enriched by J2-IP (in three independent biological replicates; black bars, *P-value < 0.001 (hypergeometric distribution); n = number of genes analyzed in each structural category).

5. E–G Examples of transcripts enriched in the J2-IP RNA-seq. The gfi-1 locus (E) expresses an antisense transcript, as shown by both antisense (red) and sense (blue) reads. The 3′ UTR of lem-2 (F) undergoes dsRNA-specific A-to-I RNA editing. The editing pattern in the boxed region is shown. (G) Example of transcript not enriched by J2-IP (y57g7a.5). Black arrows indicate direction of transcription.

Figure 5. Double-stranded RNA transcripts are preferentially recovered in tdp-1(ok803) mutant extract

1. Graph comparing the log₂ RPKM (normalized to input) of tdp-1(ok803) J2-IP (y-axis) versus wild-type J2-IP (x-axis) for all genes significantly increased (red dots) and decreased (blue dots) in representation (P < 0.05, FDR < 0.1 for all changes). Plot derived from data shown in Supplementary Fig S10 using three independent biological replicates of wild-type and tdp-1(ok803) J2-IP.

2. The percentage (x-axis) of all expressed repetitive elements (type shown on y-axis) that are significantly increased (black bars) and decreased (gray bars) in tdp-1 mutant J2-IP compared to wild-type J2-IP (P < 0.05 for all changes). *P < 1 × 10⁻¹⁰ chi-square test. Plot derived from data shown in Supplementary Fig S12.

3. A representative example of a tandem repeat region that is increased for dsRNA structure/stability in tdp-1(ok803) J2-IP. Coverage tracks from both the J2-IP RNA-seq and input RNA-seq are shown; region displayed: chromosome I: 10,130,500–10,133,500.

4. Coverage tracks of col-119 gene (transcribed antisense with the intron of expressed gene C53B4.4) in poly(A) RNA-seq (top), input (total) RNA-seq (middle) and J2-IP RNA-seq (bottom) (note: scale of coverage tracks for J2-IP is normalized to input for display purposes). Gene models for both genes are shown below coverage tracks, the direction of transcription is indicated by black arrows). Scale bars (black line), 200 bp. See also Supplementary Fig S11.

Figure 6. TDP-1 directly limits the dsRNA structure or stability of intronic RNA

1. Log₂ fold change of tdp-1(ok803) J2-IP/wild-type J2-IP (x-axis) for all expressed intronic regions selected in the J2-IP graphed according to increasing number of reads that map to each region (x-axis). Abundance levels for introns were normalized to input (see Materials and Methods for details). All introns selected in the J2-IP that are significantly (P < 0.05, FDR < 0.1, three biologically independent replicates) increased (red dots) and decreased (blue dots) between tdp-1 mutant and wild-type J2-IP are shown. See Supplementary Table S9.

2. Example of a tdp-1(ok803) J2-IP enriched intronic region that exhibits altered splicing in tdp-1(ok803) polyA RNA-seq. Coverage tracks from poly(A) RNA-seq and J2-IP RNA-seq are displayed for pqn-41 (ChrIII: 1,931,953–1,934,515). A-to-I RNA editing patterns in the boxed regions for each transcript are also depicted. Arrows depict the direction of transcription, and scale bars (solid black line below gene model) are set to 250 bp.

3. ChIP-seq pattern for TDP-1 within the long, structured gene ppfr-1 (ChrI: 9,364,769-9,393,087). The location of significant TDP-1 peaks in both replicates as well as in the RNase control is shown by blue boxes below the ChIP-seq coverage tracks. The enrichment by anti-dsRNA immunoprecipitation (J2-IP) within regions bound by TDP-1 is displayed below significant peaks. Both wild-type (top) and tdp-1(ok803) (bottom) J2-IP tracks are shown to indicate excess RNA structure when TDP-1 is absent. The J2-IP enrichment for an intron not bound by TDP-1 is also shown. Arrows depict the direction of transcription, and scale bars (solid black line below gene model) are set to 250 bp.

Source data are available online for this figure.

Figure 7. TDP-1 maintains chemotaxis by limiting RNA interference

1. The chemotaxis index (y-axis) for wild-type and tdp-1(ok803) animals toward 1% butanone and 1% isoamyl alcohol are shown. Negative values signify repulsion. Error bars represent SEM. **P < 0.01, ***P < 0.001 (Student's t-test).

2. Rescue of chemotaxis toward 1% butanone (top two graphs) and 1% isoamyl alcohol (bottom two graphs) in tdp-1(ok803);rde-1(ne219) double-mutant animals. Assays were done in triplicates at least 3 independent times. Error bars represent SEM. ***P < 0.001 (Student's t-test).

3. Immunostaining of isolated worm nuclei in wild-type, tdp-1(ok803) mutants and tdp-1(ok803); rde-1(n2219) double mutants. Single plane images distally overlaid with DIC image. The presence of J2 foci in tdp-1(ok803); rde-1(n2219) mutants indicates that deletion of rde-1 does not suppress the accumulation of dsRNA in tdp-1 mutants.

Figure 8. Mammalian TDP-43 also functions to limit the accumulation of dsRNA

1. Cultured HeLa cells depleted for TDP-43 (green) with TARDBP siRNA display increased intensity of J2 anti-dsRNA labeling (red) within nuclear foci. Scale bars, 20 μm.

2. Densitometric quantification of the intensity of nuclear dsRNA foci. Graph shows the mean J2-dsRNA densitometric value, error bars = SEM. ***P < 0.0001 (Student's t-test). N = number of objects detected in each group.

3. Cultured M17 cells display increased nuclear and cytoplasmic dsRNA staining (red) upon knockdown of TDP-43. Scale bars, 20 μm.

4. Quantification of nuclear dsRNA in TDP-43 knockdown M17 cells. Graph shows the mean J2-dsRNA densitometric value, error bars = SEM. ***P < 0.0001 (Student's t-test). N = number of objects detected in each group.

5. Semi-quantitative RT–PCR of RNA precipitated from J2-IP on lysate from mock-treated M17 cells (left panel) and TDP-43 knockdown M17 cells (right panel). RT–PCR was done in triplicates (only one replicate shown).

6. Expression tracks from J2-IP RNA-seq in regions shown to have increased recovery in TDP-43 knockdown J2-IP compared to mock-treated controls. The gene model for each region is shown above the expression tracks. Evidence of A-to-I RNA editing in human sequencing read is also shown for HP1BP3. Herv-K expression tracks represent reads mapping to a consensus sequence of the Herv-K genome. All expression tracks are normalized to the number of reads in each sample, and the height of the track indicates reads depth. The fold enrichment by J2-IP in TDP-43 knockdown over control (normalized to total RNA-seq) for each region is shown to the right. The thin black line indicates scale, and scale is set to 500 bp. Red lines show the location of primer sequences relative to the gene. Expression tracks for NCAD1 (loading control) are shown in Supplementary Fig S20.

Source data are available online for this figure.

Figure 9. Human TDP-43 displays strand dissociation activity in vitro

1. Assay for strand displacement activity (adapted from Naeeni et al, 2012).

2. RNA strand dissociation activity was measured as a change in relative FRET signal (y-axis) over time (x-axis) of two complementary, fluorescently labeled (Cy5 and Cy3, respectively) RNA oligos (annealed for 480 s at 37°C) following the addition of tenfold molar excess of a competitor oligo complementary to the Cy5-labeled RNA molecule. RNA molecules were excited at 535 nm, and FRET signal was recorded at 680 nm. Initial FRET fluorescence was set to zero and then measured every 15 s for 480 s. Data represent results from six to seven independent assays. Error bars, SEM calculated between replicates at each 15 s time point. The assay was done in the presence of Bovine Serum Albumin (BSA) protein (blue line); recombinant human Lupus Antigen (Human La) (red line), and recombinant human Tar-DNA Binding protein 43 (TDP-43) (green line).