Reduction of RAD23A extends lifespan and mitigates pathology in TDP-43 mice

Abstract

Protein misfolding and aggregation are cardinal features of neurodegenerative disease (NDD) and they contribute to pathophysiology by both loss-of-function (LOF) and gain-of-function (GOF) mechanisms. This is well exemplified by TDP-43 which aggregates and mislocalizes in several NDDs. The depletion of nuclear TDP-43 leads to reduction in its normal function in RNA metabolism and the cytoplasmic accumulation of TDP-43 leads to aberrant protein homeostasis. A modifier screen found that loss of rad23 suppressed TDP-43 pathology in invertebrate and tissue culture models. Here we show in a mouse model of TDP-43 pathology that genetic or antisense oligonucleotide (ASO)-mediated reduction in rad23a confers benefits on survival and behavior, histological hallmarks of disease and reduction of mislocalized and aggregated TDP-43. This results in improved function of the ubiquitin-proteasome system (UPS) and correction of transcriptomic alterations evoked by pathologic TDP-43. RAD23A-dependent remodeling of the insoluble proteome appears to be a key event driving pathology in this model. As TDP-43 pathology is prevalent in both familial and sporadic NDD, targeting RAD23A may have therapeutic potential.

Figure 1.. Effect of RAD23A KD on TAR4/4 mice: survival and behavior.

(A) Quantification of the ASO effect on RAD23A mRNA level. Group differences were found by ANOVA F_(2,6)= 83.63, p<0.0001, post hoc analysis: ***, p<0.001 (B) Quantification of the ASO effect on RAD23B mRNA level. Group differences were found by ANOVA F_(2,6)= 3.26, p<0.01, post hoc analysis: ***, p,0.001 (C) Western blots of cortex from mice that received intracerebroventricular (ICV) administration at postnatal day zero (P0) of an antisense oligonucleotide (ASO) with a scrambled sequence (SCR) or targeting rat RAD23A. ASO to RAD23A leads to a reliable and reproducible reduction in abundance of RAD23A, but not RAD23B, protein. N=4 shown. (D) Quantification of the ASO effects at the protein level. ***, p<0.001. (E) Survival of TAR4/4 male + female mice after administration of ASO targeting RAD23A or the SCR control. There is an ~9 day extension of lifespan in animals receiving the ASO to RAD23A. The lifespan curves were compared by the log-rank test (<0.0001) and the effect size was estimated by the Cox proportional hazards model, where HR indicated the hazard ratio (HR=10.15). (F) Progressive rise in the gait impairment score in the TAR4/4 mice treated with SCR ASO was blunted by ASO targeting RAD23A. **, p<0.01 and ***, p<0.001. (G) Progressive rise in the tremor score in the TAR4/4 mice treated with SCR ASO was blunted by ASO targeting RAD23A. (H) Progressive rise in the kyphosis score in the TAR4/4 mice treated with SCR ASO was blunted by ASO targeting RAD23A. ***, p<0.001. (I) Progressive rise in the hindlimb clasping score in the TAR4/4 mice treated with SCR ASO was blunted by ASO targeting RAD23A. ***, p<0.001. ***. Note, hindlimb clasping cannot be reliably assessed prior to P18. (J) There is a reduction of RAD23A (but not RAD23B) protein levels in mice heterozygous for the RAD23A null allele (RAD23A^−/+) and no RAD23A expression in the homozygotes (RAD23A^−/−). (K) Quantification of genotype effect on RAD23A and RAD23B protein levels. Group differences were found by ANOVA for RAD23A - F_(2,33)= 306.2, p<0.0001, post hoc analysis: *, p<0.05, ***, p<0.001. Group differences were found by ANOVA for RAD23B - F_(2,33)= 3.90, p<0.03, post hoc analysis: *, p<0.05, ***, p<0.001. (L) Survival of TAR4/4 mice in the background of WT, reduced or absent RAD23A levels. There is an ~5 day extension of lifespan in TAR4/4;RAD23A^−/+ animals in comparison with TAR4/4;RAD23A^+/+ or TAR4/4;RAD23A^−/− animals (HR=9.4, p<0.001 TAR4/4;RAD23A^−/+ versus TAR4/4;RAD23A^+/+). Number of animals in experimental groups (TAR4/4;RAD23A^+/+, TAR4/4;RAD23A^−/+ and TAR4/4;RAD23A^−/−) was 18,27 and 16 respectively. (M) Western blots of cortex from mice that received SCR ASO or ASO targeting RAD23B leads to a reliable and reproducible reduction in abundance of RAD23B but not RAD23A protein. N=4 shown. (N) Quantification of the ASO effects at the protein level. ***, p<0.001. (O) Survival of TAR4/4 mice treated with SCR or RAD23B targeting ASO. There is an ~5 day shortening of lifespan of animals associated with RAD23B protein KD extension (HR=4.3, p<0.001).

Figure 2.. Effect of RAD23A on TAR4/4 mice: histology and cell type markers.

(A) Immunohistology for a marker of neuronal nuclei (anti-NeuN) and astrocytes (anti-glial fibrillary acidic protein, GFAP) in WT and TAR4/4 mice who received ICV injection at P0 of SCR sequence or RAD23A targeted ASO. To study primary and secondary motor cortex, coronal tissue slices correspond to Bregma ~1.3 mm (assuming interaural position at 2.48 mm). At this level, a portion of hippocampus (e.g., CA1, CA2, CA3 and dentate gyrus) is visible ventrally. The motor cortex of WT animals treated with ASO SCR or targeting RAD23A display indistinguishable thickness and neuronal density. The same is true for GFAP immunoreactivity. The motor cortex of TAR4/4 animals treated with the ASO SCR is less thick and displays a reduced density of neurons particularly in layers VI and V. These layers contain a prominent population of GFAP(+) astrocytes. Astrocytes also populate more superficial layers (particularly layer II). Blow up of astrocyte, inset. The motor cortex of TAR4/4 animals treated with the ASO to RAD23A blunt these pathological findings with partial restoration of cortical thickness and density and reduced astrocytosis. Calibration bar of high lower inset = 15 microns. Calibration bar from low power octet = 50 microns (B – D) Quantitative PCR for GFAP (B), IBA1 (C), and CD68 (D) in WT animals and TAR4/4 animals that received ICV injection of ASO at P0 that are SCR sequence or target RAD23A. (B) SCR treated TAR4/4 animals have an elevated level of GFAP transcripts in comparison with WT animals and this is blunted in TAR4/4 animals treated with ASO targeting RAD23A. This is seen in both the cerebral cortex and spinal cord (p < 0.05). (C) There are no group difference in the cortical IBA1 transcripts. In the spinal cord, IBA1 transcripts are elevated in the SCR treated TAR4/4 animals and this is blunted in TAR4/4 animals treated with ASO targeting RAD23A (p < 0.05). SCR treated TAR4/4 animals have a modestly elevated level of CD68 transcripts in comparison with WT animals and this is blunted in TAR4/4 animals treated with ASO targeting RAD23A. This is seen in both the cerebral cortex and spinal cord (p > 0.05). (E) Full length (FL) caspase 3 is detectable in duplicate WT and TAR4/4 animals and its level of abundance is higher in the TAR4/4 animals regardless of ASO administered (denoted by red >> symbol). Only in the TAR4/4 animals is the cleaved caspase 3 species detectable (denoted by red >) and the level of abundance is higher in the SCR ASO treated animals in comparison with the RAD23A ASO treated animals. (F) Quantification of the abundance of FL caspase 3 reveals a ~30% increase in the TAR4/4 compared with WT mice (p<0.05). (G) Quantification of the abundance of cleaved caspase 3 reveals a ~50% decrease in the TAR4/4 RAD23A ASO treated compared with SCR ASO treated mice (p < 0.0001).

Figure 3. Effects on hTDP-43 on the mouse transcriptome are modified by KD of RAD23A.

(A) The volcano plots represent the differentially expressed genes (DEGs) in the pairwise comparisons between TAR4/4;RAD23A^SCR versus WT;RAD23A^SCR. Genes with an adjusted P-value of less than 0.05 were considered as DEGs. Upregulated and downregulated DEGs shown in red and blue, respectively. Genes with a positive log2(foldchange) are considered upregulated, while those with a negative log2(foldchange) are considered downregulated. The horizontal line represents the adjusted P-value cut-off (0.05). (B) The volcano plots represent the DEGs in the pairwise comparisons between TAR4/4;RAD23A^SCR versus TAR4/4;RAD23^KD. Statistical considerations as in (A). (C) Hierarchical clustering and heatmap generation was performed using 6121 genes which show a reversed expression patterns between TAR4/4 and TAR4/4 ASOs(RAD23A). All four conditions, WT and TAR4/4 mice treated with SCR ASO or ASO targeting RAD23A (10 replicates). The KD of RAD23A in the TAR4/4 mice leads to a shift in the transcriptomes towards the WT. The color key indicates the row Z-score (red: upregulation after treatment with RAD23A ASOs; blue: downregulation with RAD23A treatment). (D) Principal Component Analysis (PCA) was performed on 6121 genes that displayed reversed expression patterns between TAR4/4 SCR ASO versus TAR4/4 RAD23A targeting ASO. All four experimental groups and all their replicates were included in the PCA. FactoMineR and Factoextra R-packages were used for performing the PCA analysis. (E) Gene ontology terms revealing reversible enrichment patterns between the experimental groups. Pairwise comparisons of the Group I comparisons (Up regulated and down regulated transcripts are displayed separately); similarly Group II comparisons. The circle size corresponds to the number of genes, and the blue-to-red colour gradient represents the P-value.

Figure 4.. Effect of RAD23A on TAR4/4 mice: subcellular distribution and solubility of TDP-43.

(A) Western blots for human TDP-43 (hTDP-43) or both mouse and human TDP-43 (m+hTDP-43) of cortex from WT and TAR4/4 mice treated with ASOs. hTDP-43 is only detectable in the cortex of TAR4/4 mice and knockdown of RAD23A reduces the abundance of TDP-43 when blotting for hTDP-43 or m+hTDP-43. (B) Quantification of the ASO effects on the hTDP-43 protein level. Group differences were found by ANOVA, F_(2,15)= 70.5, p<0.0001, post hoc analysis: *, p<0.05, ***, p<0.001. **, p<0.01, ***, p<0.001 (C) Quantification of the ASO effects on the m+hTDP-43 protein level. Group differences were found by ANOVA, F_(2,15)= 51.6, p<0.0001, post hoc analysis:**, p<0.01, ***, p<0.001 (D) Transcript for hTDP-43 is present only in TAR4/4 mice; there is no different in this transcript when comparing SCR ASO versus RAD23A targeting ASO in either the cortex (p=0.14) or spinal cord (p=0.69) (E) Nuclear and cytosolic fractionation of cortex from WT and TAR4/4 was performed on mice that received the SCR sequence or RAD23A targeting ASO. After normalization, the ratio of nuclear to cytosolic hTDP43 was obtained on a per experimental animal basis. There is a statistically significant increase in the nuclear/cytosolic ratio in the RAD23A ASO treated animals in comparison with the SCR ASO treate animals. ***, p<0.001 (F) Brain lysates were subjected to sarkospin protocol and separated into sarkosyl soluble (sark. sol.) and sarkosyl insoluble (sark. insol.) fractions. Human TDP-43 is enriched in the sark. sol. fraction of TAR4/4 cortex (in comparison with the sark. insol. fraction) and RAD23A ASO reduces sark. sol. and insoluble hTDP-43 in comparison with SCR treated animals. This is also true when blotting for m+hTDP-43. Actin and histone H3 are enriched in the sark. sol. fraction and not present in the sark. insol. fraction, confirming the effectiveness of the sarkospin protocol. (G) Lysates of cortex from WT and TAR4/4 animals probed for m+hTDP-43 reveal more total TDP-43 in the TAR4/4 mice. In comparison with SCR ASO, the ASO targeting RAD23A reduces full length (FL) TDP-43 (as seen in panels A and F) upon standard exposure time on the LICOR scanner. With long exposure, the 25 kDa C-terminal fragment (CTF) is detectable in TAR4/4 brain lysates and in comparison SCR ASO treated animals, the ASO targeting RAD23A reduces the abundance of the CTF. (H_i) Quantification of the ASO effects on hTDP-43 in the sark. sol. fraction. Group differences were found by ANOVA, F_(3,32)= 228.3, p<0.0001, post hoc analysis: ***, p<0.001, ****, p<0.0001. (H_ii) Quantification of the ASO effects on hTDP-43 in the sark. insol. fraction. Group differences were found by ANOVA, F_(3,32)= 221.8, p<0.0001, post hoc analysis: ***, p<0.001, ****, p<0.0001. (I_i) Quantification of the ASO effects on m+hTDP-43 in the sark. sol. fraction. Group differences were found by ANOVA, F_(3,20)= 21.03, p<0.0001, post hoc analysis: **, p<0.01, ***, p<0.001. (I_ii) Quantification of the ASO effects on m+hTDP-43 in the sark. insol. fraction. Group differences were found by ANOVA, F_(3,20)= 20.31, p<0.0001, post hoc analysis: **, p<0.01, ***, p<0.001. (J_i) Quantification of the ASO effects on FL m+hTDP-43. Group differences were found by ANOVA, F_(3,20)= 128.5, p<0.0001, post hoc analysis: ***, p<0.001. (J_ii) Quantification of the ASO effects on CTF. Group differences were found by ANOVA, F_(3,20)= 64.0, p<0.0001, post hoc analysis: ***, p<0.001.

Figure 5.. Effect of RAD23A on TAR4/4 mice: proteasome level, localization and activity.

(A)Total ubiquitin levels in the soluble and insoluble fraction of WT and TAR4/4 mice treated with ASOs. The smear of ubiquitinated substrates in WT animals (insoluble and soluble fractions) is unaffected by ASO treatment. In contrast, the overall burden of total ubiquitinated substrates in the TAR4/4 mice is higher than in WT mice (in both insoluble and soluble fractions). KD of RAD23A reduces the abundance of total ubiquitinated substrates (in comparison with SCR control) in both fractions, but more significantly in the insoluble fraction. (B and C) Quantification of the ASO effects on total ubiquitin levels in the insoluble (B) and soluble (C) fraction of WT and TAR4/4 brain lysates. Group differences were found by ANOVA F_(3,12)= 70.86, p<0.0001, post hoc analysis: ***, p<0.001 **, p<0.01, ***, p<0.001 (B) and ANOVA F_(3,54)= 20.65, p<0.0001, post hoc analysis: ***, p<0.001 ***, p<0.001 (C). (D) Proteasome activity was assessed by measuring the hydrolysis of fluorogenic suc-LLVY-AMC (12.5 μM) by purified human proteasomes (10 nM) in the absence and presence of sarkosyl-soluble or -insoluble fractions from whole-cell lysates of TDP43-Q331K-overexpressing HEK293 cells. Time-dependent increase in activity of purified proteasomes (black circles) was completely inhibited by incubation with MG132 (black squares). Incubation of proteasomes with the sarkosyl soluble fraction of TDP-43 expressing HEK293 cells led to a modest reduction in proteasome activity (blue triangles) that was not statistically significant. Incubation of proteasomes with the sarkosyl insoluble fraction of TDP-43 expressing HEK293 cells led to a robust reduction in proteasome activity (red triangles) that was statistically significant. (E) Quantification of proteasome activity at the 60 minute time point reveals groups differences by ANOVA (F_(3,8)=72.8, p<0.0001). The post hoc analysis shows that the proteasome activity in the sarkosyl-insoluble treated samples was statistically significantly reduced in comparison with proteasomes untreated with cell lysates (***, p<0.0001) or treated with sarkosyl-soluble lysates (**, p<0.001). Exact p values and confidence intervals are summarized in Supplementary Figure 7. (F) Affinity purified 26S proteasomes contain PSMD4, PSMA4, PSMB6 and RAD23A (by immunoblots) while 20S proteasome lack RAD23A. (G_i) Whole cell lysate (WCL) from the brain of WT and TAR4/4 animals treated with SCR or RAD23A targeting ASO were assayed for proteasomal protease activity using the LLVY peptide. There is reduced protease activity in the TAR4/4 SCR ASO treated mice in comparison with the other experimental groups. Group differences were found by ANOVA F_(3,53)= 30.13, p<0.0001, post hoc analysis: ***, p<0.001 *, P,0.05, **, p<0.01, ***, p<0.001. (G_ii) Quantification of brain lysate protease activity using the RPPGFSAFK peptide. Group differences were found by ANOVA F_(3,44)= 21.0, p<0.0001, post hoc analysis: ***, p<0.001 **, p<0.01. (G_iii) Quantification of brain lysate protease activity using the KKVAPYPME peptide. Group differences were found by ANOVA F_(3,54)= 20.65, p<0.0001, post hoc analysis: ***, p<0.001 *, P,0.05, **, p<0.01, ***, p<0.001. (H) Immunoblots for proteasome components (e.g., PSMA3, PSMC3 and PSMC5) in the soluble and insoluble fractions of WT and TAR4/4 brain lysates treated with SCR or RAD23A targeting ASO. In WT animals proteasome components are predominantly found in the soluble fraction and ASO treatments do not affect their abundance. In contrast, in the TAR4/4 mice, the abundance of the proteasome subunits in the insoluble fraction is increased (in comparison with WT). ASO targeting rad23 reduces (in comparison with SCR control) the partitioning of proteasome subunits into the insoluble fraction. (I) Quantification of the PSMA3 abundance in the insoluble fraction. Group differences were found by ANOVA F_(3,12)= 32.92, p<0.0001, post hoc analysis: ***, p<0.001 **, p<0.01, ***, p<0.001. (J) Quantification of the PSMC3 abundance in the insoluble fraction. Group differences were found by ANOVA F_(3,12)= 42.15, p<0.0001, post hoc analysis: ***, p<0.001 **, p<0.01, ***, p<0.001. (K) Quantification of the PSMC5 abundance in the insoluble fraction. Group differences were found by ANOVA F_(3,12)= 19.65, p<0.0001, post hoc analysis: ***, p<0.001 **, p<0.01, ***, p<0.001. (L) Immunoblots of immunoprecipitated material from cells expressing epitope tagged A315T hTDP-43, RAD23A or ubiquitin. The results from IP with antibodies to the epitope tagged protein were compared with IP using a control IgG from the same species. Cell were treated with MG132 or vehicle (dimethyl sulfoxide - DMSO), or MG132. Input (10% of initial material) reveals the presence of the transgene and all the interrogated proteins. IP’ed TDP-43 (with anti-FLAG) pulls down TDP-43 but no RAD23A (probing with either anti-V5 or -RAD23A). MG132 leads to a high molecular weight smear in the anti-ubiquitin blot (using anti-HA) which likely represents ubiquitinated TDP-43. IP’ed RAD23A (with anti-V5) pulls down RAD23A (using anti-V5 or -RAD23A. With a long exposure there is a faint TDP-43 band in the (blotting with anti-FLAG or -TDP-43) in the MG132 treated cells. The molecular weight of this band corresponds to non-ubiquitinated TDP-43. IP’ed ubiquitin (with anti-HA) pulls down ubiquitin and TDP-43 but not RAD23A and MG132 increases the ubiquitin and TDP-43 signal.

Figure 6. RAD23A controls the composition of the insoluble protein fraction evoked by hTDP-43 expression in mice.

(A) Cartoon of experiment. TAR4/4 animals on standard show (i.e., ¹⁴N unlabeled proteins) received the ASO targeting RAD23A or the scrambled sequence ASO. Motor cortex was homogenized and protein concentration determined. In parallel motor cortex the ¹⁵N fully-labeled mouse was homogenized and protein content determined. A 1:1 mix of ¹⁴N with ¹⁵N homogenates was created for each experimental sample and then soluble/insoluble fractions generated by centrifugation. The insoluble fraction was then recovered and subjected to the LC-MS analysis. (B) Annotated representative raw MS1 spectra across the specified m/z ranges display distinct peptides from myelin basic protein and heterogeneous nuclear ribonucleoprotein C. The spectra highlight ¹⁵N peptide peaks in green, ¹⁴N peptide peaks in orange, and other peaks in gray. (C) Biological replicates cluster according to treatments in the tSNE (t-distributed stochastic neighbor embedding) plot. (D) Volcano plot comparing the insoluble protein fraction components of mice receiving RAD23A knockdown versus scrambled sequence. Peptide-based ANOVA test. Pie chart illustrating the differences between treatments. (E) GO cellular component (CC) terms of proteins significantly enriched in either experimental groups. X-Axis, GO terms, Y-axis, counts (greater values indicate a larger number of proteins within the GO term). Blue circles – significantly reduced enrichment in RAD23A ASO treated versus SCR ASO treated, Pink circles – significantly enhanced enrichment in RAD23A ASO treated versus SCR ASO treated. Circle size reflects −Log10 adjusted p-values. (F) Protein-protein interaction (PPI) network of proteins with significantly reduced enrichment in RAD23A ASO treated samples compared to SCR ASO treated samples. (G) PPI network of proteins with significantly increased enrichment in RAD23A ASO treated samples compared to SCR ASO treated samples.