Senataxin deficiency disrupts proteostasis through nucleolar ncRNA-driven protein aggregation

Abstract

Senataxin is an evolutionarily conserved RNA-DNA helicase involved in DNA repair and transcription termination that is associated with human neurodegenerative disorders. Here, we investigated whether Senataxin loss affects protein homeostasis based on previous work showing R-loop-driven accumulation of DNA damage and protein aggregates in human cells. We find that Senataxin loss results in the accumulation of insoluble proteins, including many factors known to be prone to aggregation in neurodegenerative disorders. These aggregates are located primarily in the nucleolus and are promoted by upregulation of non-coding RNAs expressed from the intergenic spacer region of ribosomal DNA. We also map sites of R-loop accumulation in human cells lacking Senataxin and find higher RNA-DNA hybrids within the ribosomal DNA, peri-centromeric regions, and other intergenic sites but not at annotated protein-coding genes. These findings indicate that Senataxin loss affects the solubility of the proteome through the regulation of transcription-dependent lesions in the nucleus and the nucleolus.

Figure 1.

Protein aggregation is increased in human cells with SETX deficiency. (A) Protein levels of SETX in human U2OS cells with depletion of endogenous SETX and expression of recombinant wild-type (WT), without N-terminal domain (∆N-term, 1 a.a.–665 a.a.) or without helicase domain (∆HD, 1931 a.a.–2456 a.a.) by Western blotting with anti-SETX antibody; β-actin monitored for normalization. (B) Flow cytometric analysis of protein aggregates in U2OS cells with control or SETX shRNA using Proteostat (Enzo). Proteasome inhibitor MG132 (10 μM) treated cells were employed as positive control. At least 10,000 cells were measured in each replicate; three replicates are shown with box plot, the center line indicates the median, the box bounds indicate the first and third quartiles. (C) Schematic diagram of SETX showing known features, including N-terminal protein interaction domain (yellow) and helicase domain (blue). P413L and L1976R are reported SETX mutations causing AOA2 disease. (D) Proteostat intensities per cell were quantified by FACS in U2OS cells as in B with SETX shRNA and expression of recombinant SETX WT, ∆N-term, or ∆HD as indicated. (E) Levels of SETX in U2OS cells with SETX shRNA and expression of recombinant WT, P413L or L1976R alleles by Western blotting; β-actin monitored for normalization. (F) Proteostat intensities as in B from U2OS cells with SETX depletion and expression of recombinant SETX WT, P413L, or L1976R as indicated. (G) Levels of SETX in control and SETX knock-out (KO) U2OS cells by Western blotting; Tubulin serving as an internal control. (H) Proteostat intensities as in B from wild-type or SETX KO U2OS cells compared to wild-type cells with MG132 (10 μM) treatment. (I) Degradation of a UbL-YFP-eRR model proteasome substrate was measured relative to an mCherry expression control. Ratio of YFP fluorescence per cell normalized by mCherry signal as measured in at least 10,000 cells per condition. YFP/mCherry fluorescence ratios per cell were quantified by FACS in U2OS cells with control or SETX shRNA and expression of recombinant SETX WT, or MG132 (10 μM) as indicated. (J) Proteasome activity was measured as in I, comparing wild-type and SETX KO cells to MG132 (10 μM)-treated cells. (K) Activation of the unfolded protein response and ER stress was measured with a CHOP-mCherry reporter (Oh et al., 2012) by FACS with at least 10,000 cells per condition. mCherry fluorescence levels per cell were quantified by FACS in U2OS cells with control or SETX shRNA and expression of recombinant SETX WT, or tunicamycin (1 μg/ml) as indicated. (L) ER stress in U2OS wild-type and SETX KO cells as in WT and SETX KO data also shown in Fig. S8. All P values are derived using a two-tailed t test assuming unequal variance, using the mean of the fluorescence values from three biological replicates (n = 3). ** indicates P < 0.005, *** indicates P < 0.0005. Box plots show all measurements from all three replicates. Source data are available for this figure: SourceData F1.

Figure 2.

Insoluble protein aggregates accumulate in the absence of SETX. (A) Detergent-resistant aggregates were isolated in U2OS cells with SETX shRNA-mediated depletion or mock treatment as previously described (Lee et al., 2021). Whole lysates and detergent-resistant aggregates were isolated and analyzed by Western blotting for PSMD2, TDP-43, and PSMD8. (B) Quantification of PSMD2, full-length TDP-43 (FL), truncated TDP-43 (TDP35), and PSMD8 abundance in aggregate fractions normalized by lysate levels from three independent experiments, normalized to levels in control cells. (C) SH-SY5Y cells with control or SETX shRNA were induced or not induced to differentiate as indicated. Western blotting was performed to measure the levels of PSMD2, TDP-43, and PSMD8 in lysates and aggregates. (D) Levels of PSMD2, full-length TDP-43 (FL), truncated TDP-43 (TDP35), and PSMD8 in aggregate fractions normalized by lysate levels were quantified and shown relative to control cells. (E) U2OS cells with control or SETX shRNA were treated with PARP inhibitor veliparib (10 μM) as indicated. Levels of PSMD2, TDP-43, and PSMD8 were monitored by Western blotting in lysates and aggregate fractions. (F) Levels of PSMD2, full-length TDP-43 (FL), truncated TDP-43 (TDP35), and PSMD8 in aggregate fractions normalized by lysate levels were quantified in three independent experiments; shown relative to control cells. (G) Detergent-resistant aggregates were isolated in U2OS cells with SETX shRNA-mediated depletion with N-Acetyl cysteine (NAC, 1 mM) as indicated. Whole lysates and detergent-resistant aggregates were isolated and analyzed by Western blotting for PSMD2 and PSMD8. (H) Quantification of PSMD2 and PSMD8 abundance in aggregate fractions normalized by lysate levels from three independent experiments, normalized to levels in control cells. All P values are derived using a two-tailed t test assuming unequal variance, using three biological replicates (n = 3). Error bars indicate standard error. *, **, ***, and **** indicate P < 0.05, 0.01, 0.001, and 0.0001, respectively; ns = not significant. Source data are available for this figure: SourceData F2.

Figure 3.

Global proteomics analysis of lysates and aggregates from SETX-depleted U2OS and differentiated SH-SY5Y cells shows similarities with neurodegeneration-associated aggregates. (A) Cell lysates from U2OS cells with mock treatment or SETX shRNA-mediated depletion (three biological replicates for each) were analyzed by mass spectrometry, identifying 1,579 proteins present in all samples. The abundance of each protein in control versus SETX-depleted cells is shown, with non-significant differences in grey and significant differences in red. Significance was determined using a two-tailed t test assuming unequal variance as well as Benjamini–Hochberg control of the false discovery rate at 0.05. See also Table S1. (B) Detergent-resistant aggregates were isolated from control and SETX-depleted cells and analyzed by mass spectrometry with lysate normalization. Red/grey indicators as in A. (C and D) Cell lysates and aggregate fractions were prepared from differentiated SH-SY5Y cells with control or SETX shRNA (three biological replicates for each) and analyzed as described in A and B. See also Table S1. (E) Levels of each aggregated protein, normalized by lysate values, were analyzed by hierarchical clustering (Pearson) comparing shControl samples (3) and shSETX samples (3) in U2OS cells. Only proteins with statistically significant differences between the groups are shown, which includes 405 proteins (of 1,580 total identified in all samples) in U2OS cells. Relationships between samples, as determined by the clustering output, are shown. (F) Levels of each aggregated protein, normalized by lysate values, were analyzed by hierarchical clustering (Pearson) comparing shControl samples (3) and shSETX samples (3) in differentiated SH-SY5Y cells as in E. 446 proteins shown (of 1,712 total identified in all samples) for SH-SY5Y cells. (G) Overlap between increased aggregated proteins identified in U2OS and in differentiated SH-SY5Y cells. See Table S2. (H) Gene Ontology terms and fold enrichment of proteins identified from overlap in G, number of genes and FDR values in parentheses. (I and J) Relative percentages of proteins identified from various disease (model) datasets in the non-aggregated protein group (NIA) and increased aggregated fractions caused by SETX deficiency in U2OS (I) or differentiated SH-SY5Y cells (J), sources for aggregated proteins as indicated (Dammer et al., 2012; Zuo et al., 2021; McCormack et al., 2019; Hales et al., 2016; Kepchia et al., 2020; Mahul-Mellier et al., 2020). (K and L) TANGO scores of proteins identified in NIA and aggregate fractions from U2OS (K) and differentiated SH-SY5Y cells (L). In I and J, The chi-squared test was used to evaluate the statistical significance of differences in distribution. In K and L, Welch’s t test was used to compute P values. (I–L) P values shown on each graph.

Figure S1.

Comparative analysis of protein aggregates isolated from SETX-depleted U2OS and differentiated SH-SY5Y cells. (A) Spearman correlation coefficients for shControl and shSETX pellet values normalized by lysates (three biological replicates each); 1,580 proteins present in all samples. See Table S1. (B) Enriched Gene Ontology groups are shown from analysis of significantly aggregated proteins in U2OS cells with the number of proteins and FDR values in parentheses. (C) Enriched Gene Ontology groups and fold enrichment of proteins identified in aggregates from SETX-depleted SH-SY5Y cells as in B, number of genes and FDR values in parentheses. (D and E) Aggregates and total lysates from three replicates of SETX shRNA depletion were analyzed by mass spectrometry. Aggregate values (normalized by lysate, log₂ transformed) for each protein (mean of three replicates) were plotted: (D) SETX-depleted cells (X axis) versus ATM shRNA-treated cells (Y axis) (Lee et al., 2021), 310 proteins, Spearman correlation coefficient = 0.51; (E) Mre11-depleted cells (X axis) versus ATM shRNA-treated cells (Y axis) (Lee et al., 2021), 1,424 proteins, Spearman correlation coefficient = 0.978. (F) SETX-depleted cells (X axis) versus insoluble protein log₂ fold change in ATM KO U2OS cells (Huiting et al., 2022) (Y axis), 1,044 proteins, Spearman correlation coefficient = 0.002. (G) SETX-depleted cells (X axis) versus insoluble protein log₂ fold change in CPT-treated U2OS cells (Huiting et al., 2022) (Y axis), 818 proteins, Spearman correlation coefficient = 0.01. For each comparison, only proteins with measurements in both datasets are plotted.

Figure 4.

SETX-deficient cells show higher R-loops and slower growth which are rescued by human RNaseH1 overexpression. (A) Laser micro-irradiation was performed in live U2OS cells stably expressing bacterial RNaseH (D20R-E48R)-mCherry as R-loop sensor. The circle indicates the site of laser damage. (B) RNaseH (D20R-E48R)-mCherry signal at laser damage sites at different time points was quantified from >8 control or SETX-depleted cells as in A. (C) Proliferation of U2OS cells with control or SETX shRNA treatment and induction of wild-type SETX, monitored by MTT assay. (D) Proliferation of U2OS cells with control or SETX shRNA treatment and induction of human RNaseH1, monitored by MTT assay as in C. (E) Slot blot analysis of genomic DNA isolated from U2OS cells with SETX knock-out (KO) and wild-type human RNaseH1 overexpression as indicated, using S9.6 and dsDNA antibodies to measure RNA-DNA hybrids and dsDNA, respectively. Recombinant RNaseH treatment confirms the specificity of the S9.6 antibody. (F) Levels of RNA-DNA hybrid signal (S9.6) were quantified in three replicates of E, normalized by dsDNA signal; shown relative to control cells. Error bars indicate standard deviation. (B–D) Error bars indicate standard error. *, **, ***, and **** indicate P < 0.05, 0.01, 0.001, and 0.0001, respectively, by two-tailed t test assuming unequal variance; ns = not significant. Source data are available for this figure: SourceData F4.

Figure 5.

R-loops are increased at intergenic SETX-binding sites. (A) Genome browser views of R-ChIP and input control signal in wild-type and SETX knock-out U2OS cells at representative genome locations. Two biological replicates were used and ChIP signal for each line is shown after the removal of input contribution as well as read depth normalization. (B) Normalized R-ChIP signal from wild-type and SETX knock-out cells at promoter sites (TSS) with standard error. TSS locations determined from the dominant transcript for each protein-coding gene (highest number of reads) using poly(A)-selected RNAseq data from wild-type U2OS cells. N = 13,341. (C) Visual representation of gene body, intergenic, or promoter locations of R-ChIP peaks called by MACS2 call peak for wild-type and SETX knock-out R-ChIP as well as SETX ChIP (Cohen et al., 2018). (D) Genome browser views of R-ChIP and input control signal in wild-type and SETX knock-out U2OS cells at representative intergenic genome locations. (E) Visualization of SETX KO R-ChIP signal at locations of SETX binding (called peaks from SETX ChIP no DSB treatment SETX_mOHT from Cohen et al. [2018]); “center” is center of SETX ChIP peak. (F) Quantification of R-ChIP signal from wild-type and SETX KO U2OS cells at locations of SETX binding in the absence of DNA damage; graph showed input-normalized, log₂ transformed ChIP signal. *** indicates P < 0.0001 by two-tailed Mann–Whitney test. (G) Accumulated SETX ChIP signal at locations within genes (left panel) and intergenic R-ChIP peak locations (right panel) in SETX knock-out cells.

Figure S2.

R-ChIP signal at annotated genes is lower in SETX-depleted cells and correlates with overall transcript level. (A and B) Wild-type or SETX KO U2OS cells expressing catalytically inactive RNaseH were used to perform R-ChIP with quantification by qPCR at the SNPRN and beta-actin loci (A) or two regions within the rDNA IGS (B). Three biological replicates were analyzed for each condition with Flag antibody (+Ab) as well as in the absence of antibody (−Ab). * indicates P < 0.05 or *P < 0.005 by two-sample unpaired t test. (C) R-ChIP signal from two biological replicates was quantified for each annotated gene in the promoter region (1,000 nt upstream of transcription start through 500 nt downstream) and plotted against transcript level for the gene in U2OS wild-type and SETX knock-out cell lines as indicated. Genes were grouped into bins according to transcript abundance in each cell line with the average R-ChIP signal and corresponding transcript abundance level shown.

Figure 6.

Protein aggregates caused by SETX depletion are reduced by human RNaseH1 overexpression or transcription inhibition. (A) Detergent-resistant aggregates were isolated in U2OS cells with shRNA-mediated depletion of SETX, and induction of human wild-type RNaseH1 expression as indicated. PSMD2, TDP-43, and PSMD8 in lysate and aggregate fractions are shown by Western blotting. (B) Three replicates of A were performed and quantified; levels of PSMD2, TDP-43, and PSMD8 in aggregate fractions normalized by lysate levels are shown relative to control cells. (C) Aggregation assays as in A with SETX knock-out and human wild-type RNaseH1 overexpression as indicated. (D) Three replicates of C were performed and quantified; levels of PSMD2, TDP-43, and PSMD8 in aggregate fractions normalized by lysate levels are shown relative to control cells. (E) Aggregation assays as in A with control or SETX shRNA and DRB (20 μM) added as indicated. (F) Three replicates of E were performed and quantified; levels of PSMD2, TDP-43, and PSMD8 in aggregate fractions normalized by lysate levels are shown relative to control cells. All P values are derived using a two-tailed t test assuming unequal variance, using three biological replicates (n = 3). Error bars indicate standard error. *, **, ***, and **** indicate P < 0.05, 0.01, 0.001, and 0.0001, respectively; ns = not significant. Source data are available for this figure: SourceData F6.

Figure S3.

TGIRT RNAseq analysis shows transcription differences in shSETX and SETX knock-out U2OS cell lines. RNA samples from control, SETX shRNA (shSETX), and SETX knock-out (KO) U2OS cells were analyzed by TGIRTseq. Differential gene expression is shown in a volcano plot with q value (y axis) and log fold change (x axis) for shSETX compared to control cells (left panels) and SETX KO compared to control cells (right panels). (A–F) Analysis is shown for protein-coding genes (A and B), non-coding RNAs (C and D), and transposable elements (E and F). Transcripts with >1 log₂ fold change difference between treatment and control groups and P values <0.001 are shown in red. See also Table S3.

Figure S4.

SETX ChIP and TGIRT RNAseq patterns at SETX knock-out R-ChIP locations. Normalized SETX ChIP data (Cohen et al., 2018) and TGIRT RNAseq data are shown at the locations of intergenic R-ChIP accumulation in SETX knock-out cells (center, narrowpeak MACS output).

Figure 7.

Aggregated proteins in SETX-depleted cells are localized primarily in nucleoli. (A) Cytosolic (CP), nucleoplasmic (NP), and nucleolar (No) fractions were isolated from U2OS cells with control or SETX shRNA by sucrose gradient, and each fraction was used to isolate detergent-resistant aggregates. Lysates and aggregates from different fractions were analyzed by Western blotting for PSMD2, PSMD8, β-actin (cytoplasmic markers), and fibrillarin (a nucleolar marker). (B) Three replicates of A were performed and quantified; levels of PSMD2 and PSMD8 in the aggregates of nucleolar fraction normalized by lysate levels are shown relative to control cells. (C) Localization of endogenous PSMD8 was performed by immunostaining with PSMD8 and fibrillarin antibodies in U2OS cells, with shRNA-mediated SETX depletion or mock treatment. Arrows indicate PSMD8 in nucleoli. (D) Quantification of control and SETX-depleted U2OS cells showing the percentage of cells with PSMD8-positive nucleoli in C. A total number of 165 cells were counted for the control group, and 90 cells were counted for SETX-depleted group. Error bars indicate standard error. *, **, ***, and **** indicate P < 0.05, 0.01, 0.001, and 0.0001, respectively, by two-tailed t test assuming unequal variance (B) and chi-square test (D), between control and SETX-deficient groups; ns = not significant. (E) Localization of endogenous SETX in U2OS cells was detected by immunostaining with SETX and fibrillarin antibodies. Source data are available for this figure: SourceData F7.

Figure 8.

IGS ncRNAs are increased in human cells with SETX deficiency and rescued by human RNaseH1 expression. (A) Schematic diagram of a single copy of human rDNA cassette (∼43 kb) with locations of primer pairs used for reverse transcription with quantitative polymerase chain reaction (RT–qPCR). (B) Total RNA was extracted from U2OS cells with control or SETX shRNA. IGS ncRNAs were analyzed by RT-qPCR with primers as indicated in A and quantified in three independent experiments, normalized to levels of 7SK control RNA. (C) Total RNA was isolated from SETX-depleted U2OS cells with control or human wild-type RNaseH1 overexpression, and the levels of IGS ncRNAs were measured and quantified as in B. (D) RNA extracted from U2OS control or SETX knock-out cells was probed by northern blotting for IGS18 and IGS22 ncRNAs, with 7SK as an internal control. (E) Total RNA was extracted from U2OS cells with SETX knock-out and human RNaseH1 overexpression. IGS ncRNAs were analyzed and quantified by RT-qPCR and normalized to 7SK control gene as in B. (F) Total RNA was extracted from U2OS control or SETX knock-out cells and measured by TGIRT-seq. The peaks of IGS ncRNAs were visualized by IGV software. Top panel: positive-strand reads; bottom panel: negative-strand reads. Locations of repeats related to Alu elements as well as long repeat 1 and 2, Sal X 3, the CDC7 pseudogene, and the 3ʹ end of the PAPAS lncRNA are shown. (G) R-ChIP accumulation in the IGS region in WT and SETX KO U2OS cells. Error bars in B, C, and E indicate standard deviation. *, **, ***, and **** indicate P < 0.05, 0.01, 0.001, and 0.0001, respectively, by two-tailed t test assuming unequal variance between control and SETX-depleted groups, or between SETX-depleted cells with control or human RNaseH1 overexpression. Source data are available for this figure: SourceData F8.

Figure S5.

Comparisons of IGS-derived non-coding RNA in SETX knock-out compared to control cells: regions surrounding qPCR amplicons. Normalized TGIRT RNAseq data are shown at the locations in the IGS region of the rDNA surrounding locations of qPCR amplicons shown in Fig. 7 A. Positive-strand and negative-strand reads are shown in the control cells (black) or SETX KO cells (red) as indicated. Plus: Sense strand of rDNA; Minus: Antisense strand of rDNA.

Figure S6.

Comparisons of IGS-derived non-coding RNA in SETX knock-out compared to control cells: view of entire IGS region. IGV view of the entire rDNA IGS showing positive- and negative-strand reads in the control and SETX knock-out cells. Reads mapped to the human ribosomal DNA complete repeat unit (Genebank U13369.1) were divided by their orientation (plus or minus). Annotations of the IGS and repeat elements were shown on the top. The arrow bar indicates the location where the 3′ part of the PAPAS (promoter and pre-rRNA antisense) extends to. Reads were down-sampled to 100 reads if the maximum depth is over 100. Plus: sense strand of rDNA; Minus: antisense strand of rDNA.

Figure S7.

Wild-type hRNaseH overexpression reduces R-loops in SETX KO cells at IGS rDNA and peri-centromeric locations. (A and B) hRNaseH was expressed in U2OS SETX KO cells and R-loops were measured by DRIP immunoprecipitation followed by qPCR for (A) IGS locations in the rDNA and (B) several peri-centromeric regions. Results include three biological replicates per condition; error bars indicate standard deviation. *P < 0.05 or **P < 0.005 by two-sample unpaired t test, comparing the SETX KO to SETX KO +hRNaseH.

Figure 9.

Protein aggregates in SETX-deficient cells are driven by RNA POL1 and non-coding RNAs from the rDNA loci. (A) Levels of IGS ncRNAs were measured in U2OS control or SETX knock-out cells with mock or BMH-21 (1 μM) treatment, as indicated, by RT-qPCR, normalized by 7SK RNA. Error bars indicate standard deviation. *, **, ***, and **** indicate P < 0.05, 0.01, 0.001, and 0.0001, respectively, by two-tailed t test assuming unequal variance between control cells with BMH-21 or mock treatment, and between SETX knock-out cells with BMH-21 or mock treatment. (B) Detergent-resistant aggregates were isolated from control or SETX knock-out U2OS cells with mock or BMH-21 (1 μM) treatment as indicated. Abundance of PSMD2, TDP-43, and PSMD8 in lysate and aggregate fractions are shown by Western blotting. (C) Levels of PSMD2, TDP-43, and PSMD8 in aggregate fractions were quantified in three independent experiments of B and normalized by lysate; shown relative to control cells. Error bars indicate standard error. *, **, ***, and **** indicate P < 0.05, 0.01, 0.001, and 0.0001 by two-tailed t test assuming unequal variance; ns = not significant. (D and E) Levels of IGS ncRNAs were measured in U2OS shSETX or SETX knock-out cells, respectively, with Adm Cas13d expression and either control vector (NT) or Cas13d gRNA expression by RT-qPCR, normalized by 7SK RNA. (F and G) Detergent-resistant aggregates were isolated from shSETX or SETX knock-out U2OS cells with Adm Cas13d expression and either control vector (NT) or Cas13d gRNA expression. PSMD2, TDP-43, and PSMD8 in lysate and aggregate fractions are shown by Western blotting. (H and I) Levels of PSMD2, TDP-43, and PSMD8 in aggregate fractions were quantified in three independent experiments of F and G, respectively, normalized by lysate; shown relative to control cells. Error bars and P values as in C. (J) Model showing non-coding RNA expression from the nucleolus in SETX-deficient cells with associated protein aggregates (nucleolar sequestration). Source data are available for this figure: SourceData F9.

Figure S8.

Cas13d expression with gRNAs specific for IGS ncRNA reduces ER stress in SETX KO U2OS cells. Activation of the unfolded protein response and ER stress was measured with a CHOP-mCherry reporter (Oh et al., 2012) by FACS with at least 10,000 cells per replicate, three replicates per condition. mCherry fluorescence levels per cell were quantified by FACS in U2OS cells in control cells or SETX KO cells with Cas13d and expression of IGS-specific gRNAs as indicated. Control and SETX KO data are also shown in Fig. 1 L. P values are derived using a two-tailed t test assuming unequal variance, using the mean of the fluorescence values from each biological replicate.** indicates P < 0.005, ** indicates P < 0.0005. Box plots show all measurements from all replicates.