RNA Toxicity and Perturbation of rRNA Processing in Spinocerebellar Ataxia Type 2.

BACKGROUND: Spinocerebellar ataxia type 2 (SCA2) is a neurodegenerative disease caused by expansion of a CAG repeat in Ataxin-2 (ATXN2) gene. The mutant ATXN2 protein with a polyglutamine tract is known to be toxic and contributes to the SCA2 pathogenesis.
OBJECTIVE: Here, we tested the hypothesis that the mutant ATXN2 transcript with an expanded CAG repeat (expATXN2) is also toxic and contributes to SCA2 pathogenesis.
METHODS: The toxic effect of expATXN2 transcripts on SK-N-MC neuroblastoma cells and primary mouse cortical neurons was evaluated by caspase 3/7 activity and nuclear condensation assay, respectively. RNA immunoprecipitation assay was performed to identify RNA binding proteins (RBPs) that bind to expATXN2 RNA. Quantitative PCR was used to examine if ribosomal RNA (rRNA) processing is disrupted in SCA2 and Huntington's disease (HD) human brain tissue.
RESULTS: expATXN2 RNA induces neuronal cell death, and aberrantly interacts with RBPs involved in RNA metabolism. One of the RBPs, transducin β-like protein 3 (TBL3), involved in rRNA processing, binds to both expATXN2 and expanded huntingtin (expHTT) RNA in vitro. rRNA processing is disrupted in both SCA2 and HD human brain tissue.
CONCLUSION: These findings provide the first evidence of a contributory role of expATXN2 transcripts in SCA2 pathogenesis, and further support the role of expHTT transcripts in HD pathogenesis. The disruption of rRNA processing, mediated by aberrant interaction of RBPs with expATXN2 and expHTT transcripts, suggest a point of convergence in the pathogeneses of repeat expansion diseases with potential therapeutic implications.

Spinocerebellar ataxia type 2 (SCA2) is an autosomal dominant disorder caused by a CAG repeat expansion in the first exon of the Ataxin‐2 (ATXN2) gene located on chromosome 12q24. The repeat is in‐frame to encode polyglutamine (polyQ). The signs and symptoms of SCA2 include progressive deterioration in balance and coordination, neuropathies, nystagmus and slow saccadic eye movements, slurred speech, and cognitive impairment. , , , SCA2 is the second most common form of autosomal dominant ataxia, with a prevalence of 1–2 cases/105 inhabitants, varying somewhat by ethnicity and geographic location. , , , , The highest prevalence of the SCA2 mutation occurs in Cuba (6.57 cases/105 inhabitants) and is likely a consequence of a founder effect. SCA2 neuropathology is characterized by a significant loss of cerebellar Purkinje neurons, a less prominent loss of cerebellar granule cells ; marked neuronal loss in the inferior olive, pontocerebellar nuclei, and substantia nigra; degeneration of the thalamus and pons, and thinning of the cerebellar cortex without changes in neuronal density. , , , The normal ATXN2 allele contains 15 to 32 CAG triplets, whereas the disease allele typically has 33 to 64 triplets. The most common disease allele has 37 triplets, and neonatal onset SCA2 cases with over 200 CAG repeats have been reported. Similar to other CAG repeat diseases, the repeat length in SCA2 is inversely correlated to age of onset. , Recently, intermediate CAG expansion in ATXN2 has been associated with a higher risk for amyotrophic lateral sclerosis (ALS). Current evidence indicates that neurotoxicity of ATXN2 protein, which is involved in multiple cellular pathways, including messenger RNA (mRNA) maturation, translation, and endocytosis, is central to SCA2 pathogenesis. , This is supported by data from several SCA2 cell and mouse models expressing mutant ATXN2 protein. , ,

However, multiple laboratories, including ours, have demonstrated an important neurotoxic role for mutant RNA transcripts in CAG/CTG repeat expansion diseases, including myotonic dystrophy type 1 (DM1), , Huntington's disease (HD), , Huntington's disease‐like 2 (HDL2), SCA3, , , and SCA8. RNA‐triggered pathogenic processes are thought to be, at least in part, mediated by aberrant interaction between expanded repeat‐containing RNA transcripts and RNA‐binding proteins (RBPs). , , The basic hypothesis is that expanded CAG/CUG repeats in transcripts form hairpin structures that sequester multiple RBPs and hence, prevent the RBPs from performing their normal function in cells. To add to the pathomechanistic complexity of CAG/CUG repeat diseases, antisense transcripts that span the CAG/CUG repeat regions are also expressed at the DM1 (CAG direction), HDL2 (CAG direction), , SCA7 (CUG direction), SCA8 (CUG direction), and HD (CUG direction) loci. We have recently described a transcript expressed antisense to ATXN2 at the SCA2 locus and provided evidence that this antisense ATXN2 (ATXN2‐AS) transcript contributes to SCA2 pathogenesis and potentially to ALS associated with an intermediate repeat expansion at the ATXN2 locus.

We hypothesized that, in addition to mutant ATXN2 protein and mutant ATXN2‐AS transcript, mutant sense ATXN2 RNA also contributes to SCA2 pathogenesis.

As predicted by this hypothesis, the data presented here demonstrate that sense expATXN2 transcripts are neurotoxic in cell models in the absence of expression of mutant ATXN2 protein, aberrantly interact with RBPs that are involved in ribosomal RNA (rRNA) processing, and lead to disruption of rRNA processing. We demonstrate a similar disruption of rRNA processing in HD patient brain tissue. Similar to findings in other repeat expansion diseases, SCA2 is, therefore, the fifth neurodegenerative CAG/CTG repeat expansion disease in which pathogenesis is likely a consequence of a combination of expression of mutant protein and bi‐directionally expressed mutant RNA.

Materials and Methods

A description of the materials and methods is provided in the Supporting Data.

Results

The Non‐Translatable Transcript Is Neurotoxic

To confirm the toxicity of expATXN2 transcripts, we cloned full‐length (FL) ATXN2 complementary DNA (cDNA) with 22, 58, or 104 CAG triplets into the 3′ untranslated (UTR) region of Renilla luciferase (Rluc) cDNA, thereby allowing expression of expATXN2 RNA transcripts, but preventing ATG‐initiated translation of the RNA into FL ATXN2 protein (Fig. 1A). Primers F1/R1 and F1/R2, spanning the junction of the fusion Rluc‐ATXN2 construct, were used to detect mRNA transcript that contains CAG repeat in SK‐N‐MC cells overexpressing Rluc‐ATXN2‐(CAG)n (Supplementary Fig. S1). Anti‐ATXN2 Western blotting suggested no detectable expression of FL ATXN2‐Q58 or ATXN2‐Q104 protein in SK‐N‐MC cells overexpressing Rluc‐ATXN2‐(CAG)58 or Rluc‐ATXN2‐(CAG)104, respectively (Supplementary Fig. S2). Further, no evidence of expression of an expanded polyglutamine tract was detected by Western blotting using the expanded polyglutamine‐specific antibody 1C2, confirming that the 3′ UTR cloning approach indeed eliminated detectable ATG‐initiated translation of the FL ATXN2 (Fig. 1B). As an additional test of the 3′ UTR approach, expression from EGFP ORF was eliminated when placed at the 3′ UTR of Rluc ORF (Supplementary Fig. S3). Caspase 3/7 activity assay showed that overexpression of Rluc‐ATXN2‐(CAG)58 or Rluc‐ATXN2‐(CAG)104 was significantly more toxic than Rluc‐ATXN2‐(CAG)22 in SK‐N‐MC cells (Fig. 1C). Comparable expression levels of overexpressed transcripts in SK‐N‐MC cells were confirmed by quantitative polymerase chain reaction (qPCR) using Rluc primers (Fig. 1D). However, hairpin‐forming expanded CAG repeats can also be translated in the absence of ATG start codon through the mechanism of repeat‐associated non‐ATG translation (RAN translation). To exclude the possibility that RAN translation of protein fragments with expanded amino acid tracts leads to neurotoxicity in our SK‐N‐MC model system, we cloned an ATXN2 fragment containing a CAG repeat expansion, multiple upstream stop codons, and a total of ~150 bp of ATXN2 sequence flanking the repeat (thereby excluding all ATGs) into a vector with tags for each of the three open reading frames (Supplementary Fig. S4A). It was previously shown that HEK293 cells support the RAN translation of expATXN2 transcript, however, there were no detectable protein fragments from any of the three reading frames when overexpressed in SK‐N‐MC cells (Supplementary Fig. S4 B–E), indicating that SK‐N‐MC cells do not support the RAN translation of expATXN2 transcripts, consistent with our previous results indicating that RAN translation is cell line‐specific. Our results, therefore, confirm that expression of FL expATXN2 transcript is sufficient to trigger neurotoxicity in SK‐N‐MC cells, even when the transcripts are not translated into proteins. Consistent with these observations in neuroblastoma cells, overexpression of Rluc‐ATXN2‐(CAG)104 triggers neurotoxicity in primary mouse cortical neurons, as measured by nuclear condensation assay (Fig. 1E). Rluc‐ATXN2‐(CAG)58 was not toxic in this assay, perhaps reflective of the short time frame of the experiment (nuclear condensation is a later stage event, whereas caspase 3/7 activation occurs at an early stage in cell death), differences in the levels of transcript expression in primary neurons compared to neuroblastoma cell lines, or different sensitivity to transcript‐induced toxicity in primary neurons and SK‐N‐MC cells.

FIG. 1

Non‐translatable full length (FL) expATXN2 transcript is neurotoxic to SK‐N‐MC cells. (A) Schematic presentation of the non‐translatable FL ATXN2 cell model. FL ATXN2 cDNA was cloned into 3′ UTR region of Renilla luciferase (Rluc) cDNA to prevent its translation. Primer locations were indicated. (B) Rluc‐ATXN2‐(CAG)n constructs do not express canonically translated polyglutamine (polyQ), as confirmed by immunoblotting with polyQ‐specific 1C2 antibody. β‐Actin was used as a loading control. A representative blot was shown. (C) At 72 hours after overexpression, both Rlu‐ATXN2‐(CAG)58 and Rluc‐ATXN2‐(CAG)104 transcripts are toxic to neuronal‐like SK‐N‐MC cells, as determined by caspase 3/7 activity assay. The caspase 3/7 activity in Rluc‐ATXN2‐(CAG)22 transfected SK‐N‐MC cells was normalized to 100. Data were expressed as mean ± SEM from 4 independent samples per condition (n = 4); **P < 0.01 by Kruskal–Wallis test and Dunn's multiple comparison test. (D) Comparable expression levels of exogenous Rluc‐ATXN2‐(CAG)n transcripts were confirmed by qPCR. ACTB transcript was used as an internal control. Locations of qPCR primers for Rluc were indicated in (A). The Rluc/ACTB ratio in Rluc‐ATXN2‐(CAG)22 transfected SK‐N‐MC cells was normalized to 1. Data were expressed as mean ± SEM from 4 independent samples per condition (n = 4); Kruskal–Wallis test. (E) At 48 hours after overexpression, Rluc‐ATXN2‐(CAG)104 is toxic to primary mouse cortical neurons, as determined by a nuclear condensation assay. (F) expATXN2 transcript toxicity depends on the repeat's ability to form toxic hairpin structures. ATXN2‐(CAG)104, but not the interrupted ATXN2‐(CAG/CAA)105, is toxic to primary mouse cortical neurons, as indicated by a nuclear condensation assay. Data are expressed as mean ± SEM from 4–8 independent samples per condition (n = 4–8). In each sample, 4000 neurons per condition were analyzed. *P < 0.05, **P < 0.01 by Kruskal–Wallis test and Dunn's multiple comparison test.

It has been suggested that the CAG repeats form stable hairpin structures, whereas CAA interruptions either break hairpin regularity or induce the formation of branched structures. To examine whether preventing the formation of hairpin structures in expATXN2 ameliorate its neurotoxicity, we replaced the pure CAG repeat region in the ATXN2‐(CAG)104 with a fragment of heavily interrupted CAG/CAA triplets to obtain the ATXN2‐(CAG/CAA)105 construct. Inserting interruptions abolished expATXN2 toxicity in primary mouse cortical neurons (Fig. 1F), suggesting that the secondary hairpin structure adopted by the pure CAG repeat may be critical for neurotoxicity.

Full‐Length Transcripts Form RNA Foci in SCA2 Cell and Mouse Models and in One Human SCA2 Brain

Repeat‐containing mutant transcripts form RNA foci in all CUG/CAG diseases, in which RNA neurotoxicity has been demonstrated to contribute to pathogenesis. , , We sought to detect similar foci in SCA2 models and human brain by fluorescence in situ hybridization (FISH) using a 20‐mer 2‐O‐methyl‐CUG riboprobe that binds to the CAG repeat region on the transcript. CAG RNA foci were absent in SK‐N‐MC neuroblastoma cells that overexpress GFP alone (Fig. 2A) or a FL ATXN2 construct modified to have only one CAG triplet (GFP‐ATXN2Q1, Fig. 2B) and were only rarely detected in cells overexpressing FL normal ATXN2 (nATXN2) transcripts with 22 triplets (GFP‐ATXN2Q22, Fig. 2C). Foci were much more abundant in cells overexpressing full‐length (FL) expanded ATXN2 (expATXN2) transcripts with 58 or 104 CAG triplets (GFP‐ATXN2Q58 or GFP‐ATXN2Q104, Fig. 2D,F), as quantified in Figure 2E. The foci are resistant to deoxyribonuclease (DNase) treatment and are degraded by Ribonuclease (RNase) treatment (Fig. 2G,H). This set of experiments demonstrates that expATXN2 transcripts form RNA foci, with a similar number of foci formed in cells transfected with GFP‐ATXN2Q58 and GFP‐ATXN2Q104. Furthermore, although not detected in wild‐type (WT) (Fig. 2I,J) mice, ATXN2 RNA foci are present in cerebellar Purkinje neurons of SCA2 transgenic mice (Fig. 2K,L), which express FL ATXN2 with 127 CAG triplets specifically in Purkinje neurons, whereas Purkinje cells were identified based on the size and morphology of Hoechst stained nuclei. Finally, of the five human postmortem brains available for this study, ATXN2 RNA foci were detected in cerebellar Purkinje cells in one brain (H1 case, Table 1) that had 38 triplets for the mutant allele (Fig. 2O,P) but not in the control human brains (Fig. 2M,N.) RNA foci may be only a hallmark for RNA toxicity. Whether RNA foci are toxic or not remains to be further determined.

FIG. 2

expATXN2 transcripts form RNA foci. (A–H) Exogenous expATXN2 transcripts form nuclear CAG RNA foci in SK‐N‐MC neuroblastoma cells. GFP‐ATXN2‐(CAG)58 or 104 (D and F) transcripts formed frequent foci, whereas RNA foci were occasionally detected in GFP‐ATXN2‐(CAG)22 expressing cells (C) and were absent in cells expressing GFP (A) or GFP‐ATXN2‐(CAG)1 (B). (E) The percentage of cells with foci. Data were expressed as mean ± SEM from 3 independent samples per condition (n = 3); *P < 0.05 by Kruskal–Wallis test and Dunn's multiple comparison test, compared to GFP‐ATXN2‐(CAG)22. (G,H) GFP‐ATXN2‐(CAG)104 RNA foci were resistant to DNase treatment (G) and degraded by RNase treatment (H). (I–L) expATXN2 transcript forms RNA foci in the cerebellar Purkinje cells of SCA2 transgenic (Tg) mice in which the expression of FL ATXN2‐Q127 cDNA is driven by a Purkinje cell specific Pcp2 promoter (K,L). RNA foci were not detected in wildtype control mice (I,J). (M–P) expATXN2 transcript forms RNA foci in cerebellar Purkinje cells of a human SCA2 brain (O,P), but not in human control cerebella (M,N). Scale bar, 5 μm. Arrows point to RNA foci and asterisks indicate the Purkinje cells. [Color figure can be viewed at wileyonlinelibrary.com]

TABLE 1

Human control and patient brain information

Disease	Case ID	ATXN2 alleles	HTT alleles	Age of death	Age of onset	Gender	PMI (hr)
Control	201	22/22	20/25	62	N/A	M	14
	219	22/22	17/18	35	N/A	F	8
	249	22/22	18/18	49	N/A	M	12
SCA2	K1	22/37	17/18	72	40	F	6
	H1	22/38	15/17	74	60	F	19
	K2	22/41	14/17	49	26	M	24
	K3	22/41	18/23	55	35	F	24
	M1	22/44	17/22	43	30	M	23
HD	791	22/22	36/43	56	38	M	10
	172	22/25	18/45	63	N/A	M	8.5
	372	22/22	15/47	62	34	M	18
	903	22/22	18/45	59	38	M	7

Transcripts Aberrantly Interact with RBPs

Next, we examined whether the neurotoxicity of expATXN2 transcript is mediated by aberrant expATXN2 RNA‐RBP interactions. We performed an in vitro biotinylated ATXN2 RNA pull‐down assay (Fig. 3A) and identified by mass spectrometry (MS) a total of 57 RBPs that preferentially bind to the expATXN2, compared to the nATXN2 transcript. Go analysis of functional annotation and STRING analysis of the expATXN2 RBPs are shown in Supplementary Figure S6. The list of expATXN2 RBPs is shown in Supplementary Table S1. Of the 57 expATXN2 RBPs, 40 are localized in the nucleus, with 20 of them in the nucleolus, suggesting that aberrant expATXN2‐RBP interactions may predominantly occur in the nucleus. Interestingly, among the 20 nucleolar RBPs, 7 of them contain WD40 repeat domains, of which five (PWP1, TBL3, WDR3, WDR36, and UTP18) (Supplementary Table S1 and Supplementary Fig. S6B) are components of the small subunit (SSU) processome for rRNA processing. We, therefore, became interested in the SSU processome components that were identified as expATXN2 RBPs. Of the five SSU components , that are potential expATXN2 interactors, we selected transducin β‐like protein 3 (TBL3) for further analysis because we were interested in RNA‐mediated disease mechanisms shared by both SCA2 and HD. By the same method, TBL3 appeared to interact with the expanded Huntingtin (expHTT) transcript as well (Fig. 3B,C) and has a relatively greater number of peptide hits and percentage of protein coverage compared with other SSU components identified by MS (Supplementary Table S1), although the number of peptide hits does not always imply stronger interaction.

FIG. 3

TBL3 aberrantly interacts with expATXN2 and expHTT transcripts. (A) Schematic illustration of the biotinylated RNA pull‐down procedure. (B) TBL3 interacts with biotinylated expATXN2 (with 58 or 104 CAG repeats) and expHTT (with 56 or 80 CAG repeats) RNAs in an in vitro biotinylated RNA pull‐down assay. The interaction of TBL3 is specific for expanded CAG repeats, because no binding between TBL3 and CUG repeats in JPH3‐(CUG)55, ATXN2‐AS‐(CUG)104 or HTT‐AS‐(CUG)80 was observed. (C) Interaction between TBL3 and expATXN2 or expHTT is dependent on the CAG repeat region. Incubation with (CTG)8C, but not control Morpholino (MO), abolished the binding of TBL3 to ATXN2‐(CAG)104 or HTT‐(CAG)80 transcript. SK‐N‐MC cell lysate was used as a positive control. n = 3 independent experiments and representative blots were shown. (E) Nitrocellulose filter‐binding analysis of MBP‐NTD‐TBL3 binding to ATXN2‐(CAG)22, 108, ATXN2‐(CAG/CAA)105, and ATXN2‐AS‐(CUG)110 RNAs. The circles represent the mean fraction RNA bound to MBP‐NTD‐TBL3 in the absence or presence of competitor tRNA (w tRNA), respectively. Schematic presentation of the transcripts with various repeats and lengths of flanking regions are shown. Mean and SEM are shown; n ≥ 3 independent trials. The fits through the data are from non‐linear regression analysis of the binding curves to a Scatchard plot. In the absence of competitor tRNA, the KD obtained for ATXN2‐(CAG)22, ATXN2‐(CAG)108, ATXN2‐(CAG/CAA)105, and ATXN2‐AS‐(CUG)110 were 350, 420 nM, 2.1 μM, and 650 nM, respectively. [Color figure can be viewed at wileyonlinelibrary.com]

TBL3 Binds to Expanded CAG Repeats In Vitro

We performed additional RNA pull down experiments and Western blots to confirm that TBL3 interacts with expATXN2 in vitro (Fig. 3B). TBL3 was not detected when RNA pull down was performed using beads that are not loaded with transcripts (beads only, Fig. 3B) or when beads loaded with exogenous GFP transcripts (Supplementary Fig. S5) were used, indicating that the interaction between TBL3 and expATXN2 transcript is specific. To test whether the interaction is disease‐specific, we also included expHTT transcripts associated with HD, the most prevalent and most studied CAG repeat disease. , , , Studies from multiple laboratories, including ours, support the idea that RNA neurotoxicity contributes to HD27, 28, 29, 48 We confirmed that TBL3 interacts in vitro with expanded CAG repeats flanked with either ATXN2‐ or HTT‐specific sequence (Fig. 3B), but not with expanded CUG repeats flanked with either antisense ATXN2 (ATXN2‐AS ), antisense HTT (HTT‐AS; expressed on HD locus), or junctophilin‐3 (JPH3) flanking sequence (Fig. 3B.) To further confirm that the interaction between TBL3 and expATXN2 and expHTT was dependent on the CAG repeat, we pre‐incubated expATXN2 and expHTT transcripts with (CTG) 8C Morpholino (MO), which we have previously established hybridizes to CAG repeat expansions. The pretreatment with (CTG)8C prevented TBL3 from binding to either transcript in vitro and provided further evidence that both expATXN2‐TBL3 and expHTT‐TBL3 interactions are dependent on the presence of an expanded CAG repeat (Fig. 3C). Taken together, these data indicate that TBL3 binds to expanded CAG repeats independent of flanking sequence.

To investigate whether TBL3 binds to expanded CAG repeats independently of other cellular proteins, we purified the TBL3 N‐terminal RNA binding domain as a fusion with maltose binding protein (MBP‐NTD‐TBL3) and measured its binding with expATXN2 transcripts using an in vitro nitrocellulose filter binding assay. The isolated TBL3 NTD associated with ATXN2 CAG RNA, with KD = 350 nM and 420 nM for ATXN2‐(CAG)22 and ATXN2‐(CAG)108, respectively (Fig. 3D). Although overall binding was weak, the yeast homolog of TBL3, Utp13, binds pre‐rRNA as a tetramer with other UtpB complex proteins. Therefore, the weak affinity of the isolated MBP‐NTD‐TLB3 for expATXN2 may be because of the absence of its normal binding partners. The in vitro binding reactions saturated ~20%–30% of refolded ATXN2 RNA, suggesting that a fraction of the ATXN2 RNA is unable to refold into a conformation that is competent to bind TBL3.

Despite its weak affinity for RNA, MBP‐NTD‐TBL3 bound ATXN2 CAG repeats more strongly than control RNAs, including the ATXN2‐AS‐(CUG)110 transcript (KD = 650 nM) and a CAG repeat containing CAA interruptions, ATXN2‐(CAG/CAA)105 (KD = 2.1 μM). The total saturation of the CAG repeat containing CAA interruptions was greater than that of the continuous CAG repeats, which we attribute to better refolding of the interrupted repeat‐RNA. Nevertheless, tighter binding to the continuous CAG repeats raised the possibility that TBL3 recognizes the hairpin structure of CAG repeat RNA. To test this idea, the filter binding assays were also carried out in the presence of a competitor yeast transfer RNA (tRNA), which is expected to be structured under our assay conditions. The tRNA competitor abolished the interaction between MBP‐NTD‐TBL3 and ATXN2 or ATXN2‐AS transcripts (Fig. 3D), consistent with the idea that TBL3 binding depends on the structures of ATXN2 CAG repeats.

The Effect of TBL3 Reduction on 45S pre‐rRNA Level and Processing

Depletion of UTP13, the yeast homolog of TBL3, increases the steady‐state level of unprocessed 35S pre‐rRNA in yeast. We hypothesized that, although the interaction between expATXN2 and TBL3 may not be direct and likely involves other proteins, sequestration of TBL3 in a complex that interacts with expATXN2 may disrupt its normal function and affect rRNA maturation. We, therefore, predicted that knockdown of TBL3 in cells, mimicking its sequestration, would increase the level of unprocessed 45S pre‐rRNA, the human counterpart of yeast 35S pre‐rRNA. Three individual small interfering RNAs (siRNAs) were used to knock down TBL3 in HEK293T cells to minimize the possibility of alternative mechanisms of TBL3 reductions through off‐target effects. Each siRNA reduced TBL3 protein level by 50%–80% at 72 hours post transfection (Fig. 4A,B). Next, we examined 45S pre‐rRNA levels by qPCR using primers against the 5′ external transcribed spacer as indicated in Figure 4C. Knockdown of TBL3 in HEK293T cells using each siRNA increased steady state 45S pre‐rRNA levels (normalized to ACTB, Fig. 4D), consistent with a previous study using stable short hairpin RNA (shRNA) transfection. As previously reviewed, , a complex sequence of cleavage steps is required to release the mature RNAs (18S, 5.8S, and 28S) from the precursor 45S pre‐rRNA. qPCR using primers against 18S rRNA would detect the mature 18S rRNA, unprocessed 45S pre‐rRNA, as well as any intermediate rRNAs containing 18S sequence. Similarly, qPCR using primers against 28S rRNA would detect the mature 28S rRNA, the 45S pre‐rRNA precursor, as well as intermediate rRNAs that contain the 28S sequence (Fig. 4C). Therefore, we used the ratios of 18S rRNA/45S pre‐rRNA and 28S rRNA/45S rRNA measured by qPCR as readouts for 18S rRNA maturation and 28S rRNA maturation, respectively. Depletion of UTP13 in yeast has been previously shown to decrease 18S rRNA maturation. , Consistently, we found that knockdown of TBL3 in HEK293T cells decreased the ratio of 18S rRNA to 45S pre‐rRNA (Fig. 4E), indicating that TBL3 may play a role in 18S rRNA maturation. In addition, 28S rRNA maturation was also decreased after TBL3 knockdown (Fig. 4F). We attempted to determine if overexpression of TBL3 has the opposite effect, however, forced expression of TBL3 by itself triggered toxicity and mislocalized the protein into nuclear aggregates (data not shown). Similarly, MBNL1, an RBP previously shown to interact with expanded CAG/CUG (expCAG/CUG) transcript also appeared to be toxic when overexpressed or knocked down, suggesting that expression of certain RBPs must be tightly controlled to maintain their normal function.

FIG. 4

rRNA processing is affected in TBL3 knockdown cells, as well as in SCA2 and HD brains. (A,B) siRNAs against TBL3 (siTBL3), but not control siRNAs (siCtl), efficiently knocked down TBL3 protein expression by 50%–80% in HEK293T cells after 72 hours. Representative blots were shown. TBL3/β‐Actin protein expression in siCtl‐treated cells was normalized to 100. Data were expressed as mean ± SEM from 3 independent samples per condition (n = 3). *P < 0.05 and **P < 0.01 by Kruskal–Wallis test and Dunn's multiple comparison test. (C) Locations for qPCR primers for the detection of 45S pre‐rRNA, 18S rRNA and 28S rRNA. (D–F) TBL3 knockdown increased 45S pre‐rRNA/ACTB ratio but decreased 18S rRNA/45S pre‐rRNA ratio, compared to siCtl‐treated cells. 45S pre‐rRNA/ACTB, 18S rRNA/45S pre‐rRNA, and 28S rRNA/45S pre‐rRNA ratios in siCtl‐treated cells were normalized to 1, respectively. Data were expressed as mean ± SEM from 3 independent samples per condition (n = 3) for (D–F). *P < 0.05 and **P < 0.01 by Kruskal–Wallis test and Dunn's multiple comparison test. (G–I) 45S pre‐rRNA/ACTB, 18S rRNA/45S pre‐rRNA, and 28S rRNA/45S pre‐rRNA ratios in human control, HD, and SCA2 postmortem cerebella. 45S pre‐rRNA/ACTB, 18S rRNA/45S pre‐rRNA, and 28S rRNA/45S pre‐rRNA ratios in Ctl group were normalized to 1, respectively. Data were expressed as mean ± SEM from n = 3 (Ctl), 4 (HD) and 5 (SCA2) individual patient cerebella samples. *P < 0.05 by Kruskal–Wallis test and Dunn's multiple comparison test.

45S pre‐rRNA Level and Processing Is Altered in SCA2 and HD Postmortem Tissue

Finally, we examined the expression of 45S pre‐rRNA in human postmortem SCA2 and HD cerebella. qPCR amplification suggested that there was a slight, although not statistically significant, increase of 45S pre‐rRNA level (normalized to ACTB) in both SCA2 and HD cerebella, compared with control (Fig. 4G). The qPCR results suggested that there was a decrease in both 18S rRNA maturation and 28S rRNA maturation in SCA2 and HD cerebella, compared with the controls (Fig. 4H,I), consistent with the trend observed with the knockdown of TBL3 (Fig. 4D–F). Only the decrease in HD samples, but not in SCA2 samples, reached statistical significance, under the caveat that the relatively lower statistical power in SCA2 samples may not allow the detection of small changes. Taken together, the data support the idea that aberrant RNA‐RBP interactions may affect the steady‐state level and the maturation of 45S pre‐rRNA in both SCA2 and HD.

Discussion

We have previously shown that antisense ATXN2‐AS transcripts contribute to SCA2 pathogenesis. Here, we provide the first evidence that sense expATXN2 transcripts is involved in SCA2 pathogenesis. First, in establishing its potential pathogenicity, we show that untranslatable FL ATXN2 transcript is neurotoxic (Fig. 1). This model is not suitable to test the contribution of sense ATXN2 transcript relative to the toxicity of ATXN2 protein, or antisense ATXN2‐AS RNA, because neither ATXN2 protein nor ATXN2‐AS RNA is present in this model. In the future, genome editing approaches can be used to establish human SCA2 induced pluripotent stem cells (iPSC) cell models that specifically model protein versus the RNA‐triggered mechanism of pathogenesis. SCA2 iPSCs can also be subjected to transcriptome, proteome, and RNA interactome analysis to identify additional pathways that are involved in RNA‐mediated aspects of SCA2 pathogenesis. Differentiation into neuronal types of greater and lesser selective vulnerability in SCA2 (eg, Purkinje cells, cortical excitatory neurons, etc.) could be used to determine cell‐type vulnerability to RNA and protein mediated neurotoxicity. Next, we show that the expATXN2 transcripts aggregate into nuclear RNA foci in SCA2 cell and transgenic mouse models, as well as in human SCA2 postmortem brain tissue. However, of five human SCA2 postmortem brains available for this study, expATXN2 RNA foci were only detected in case H1 that had the 22/38 ATXN2 CAG repeat lengths and the latest disease on‐set (Table 1). Interestingly, we have recently characterized a transcript that is expressed in the direction antisense to ATXN2 (ATXN2‐AS) and contains an expanded CUG repeat. CUG RNA foci containing this transcript were detected in SCA2 cases K3 and M1 (Table 1.) Given that SCA2 is associated with a relatively short repeat expansion, detection of foci may require a more sensitive assay. There are a number of alternative explanations for the absence of foci in the other SCA2 brains: (1) CAG RNA foci are highly toxic or appear in cells marked for early death such that Purkinje cells with foci may not have been present by the time of death; (2) CAG RNA foci are protective and associated with late onset disease and perhaps slower disease progression; (3) detectable foci were lost consequent to the process of brain collection or storage; (4) RNA foci are a byproduct of neurotoxic processes and have a neutral role in neurotoxicity; and (5) RNA foci are an epiphenomenon, present in some SCA2 brains because of an unknown genetic or environmental factor and with no relevance to disease. Recent work describing a transgenic BAC mouse model expressing expanded C9orf72 (expC9orf72) and exhibiting widespread RNA foci, but lacking behavioral abnormalities and neurodegeneration, even at advanced ages, suggests that RNA foci are not sufficient to trigger toxicity in ALS. A transgenic mouse model expressing non‐translatable FL ATXN2, which could be tracked in live cells in real time, might help determine the relevance of CAG RNA foci to disease pathogenesis.

Our data strongly suggest that the neurotoxicity of expATXN2 transcript involves aberrant expATXN2‐RBP interactions that perturb rRNA maturation. We initially focused on TBL3, a component of the SSU processome required for rRNA processing. Although our RNA pull‐down assay indicates that TBL3 interacts with expATXN2 RNA, this assay cannot be used to prove a direct interaction. Filter‐binding assays showed that recombinant TBL3 NTD can weakly interact with expATXN2 RNA, and preferentially binds the structures of the CAG repeats. (Fig. 3D). Other components of the SSU processome likely stabilize the interaction of TBL3 with the expATXN2 RNA in the cell. The yeast homolog of TBL3, Utp13, recognizes double‐stranded regions of the pre‐rRNA as a heterotetramer with other Utp proteins. Indeed, mass spectrometry analysis of expATXN2 interactors did identify other proteins from the SSU processome in our isolated complexes (Supplementary Table S1). One interesting possibility is that the multi‐dentate recognition of structured RNA by TBL3 and its binding partners, which is a normal feature of their function in pre‐rRNA processing, also contributes to the toxic aggregation of CAG repeat RNAs. Future experiments will be needed to determine which proteins are most important for neuronal toxicity.

Our results indicate that a subset of RBPs bind to both expanded ATXN2 and HTT transcripts. This is not surprising because it is well established that transcripts containing expanded CAG repeats form similar secondary structures in vitro , , and, hence, at least some of the downstream effects are likely to be shared between different CAG repeat diseases. Although SCA2 primarily affects cerebellum, HD is primarily characterized by atrophy of striatum and cerebral cortex. However, recent evidence indicates that cerebellum is also affected in HD , and, in fact, appears to degenerate early and independently from the striatal atrophy. This suggests that similar mechanism of pathogenesis may contribute to cerebellar pathology in both SCA2 and HD. Whether and to which degree mutant RNA‐triggered mechanisms contribute to this pathology remain to be further determined.

Interestingly, it was previously reported that expanded Ataxin‐3 (ATXN3) transcripts involved in spinocerebellar ataxia type 3 (SCA3) interact with nucleolin. In SCA3, this aberrant nucleolin‐ATXN3 interaction decreases 45S pre‐rRNA levels in cell and Drosophila models of SCA3. Moreover, aberrant interaction between the expC9orf72 transcripts and nucleolin may contribute to the decreased maturation of 28S, 18S, and 5.8S rRNAs from the precursor 45S pre‐rRNA in ALS patients associated with CCCCGG hexamer expansion in C9orf72 gene. Ribosomes are ribonucleoprotein complexes that consist of rRNA and ribosomal protein components. Perturbation of rRNA processing disrupts ribosome biogenesis, which in turn compromises protein translation and causes neuronal dysfunction. As a consequence of rRNA processing failure, ribosomal proteins are not assembled into ribosome ribonucleoprotein subunits. These unincorporated ribosomal proteins accumulate in the neuronal cytosol and prevent proteasomal degradation of p53, which induces nucleolar stress and neuronal cell death. Protein translation and nucleolar stress are therefore two possible mechanisms for further investigation. It is quite possible that a therapeutic agent that prevents aberrant RNA‐RBP interactions between toxic hairpin‐forming transcripts and RBPs may be at least partially effective across multiple diseases. Alternatively, similar therapies may target shared pathogenic pathways downstream of the toxic transcripts.

In summary, we provide the first evidence that the ATXN2 transcript with an expanded repeat may contribute to SCA2 pathogenesis with similar properties to transcript‐mediated toxicity in HD. The ATXN2 transcript with an expanded CAG repeat itself, or its protein interactors, may provide valuable therapeutic targets in the future.

Author Roles

D.D.R. and P.P.L conceived the study, oversaw the project, and designed the experiments; P.P.L., R.M., H.F., X.S., N.A., J.J., L.O.M., E.H., and D.D.R. carried out the experiments and analyzed data; H.Y.E.C., C.A.R., S.M.P., R.L.M., and S.W. provided fundamental reagents and intellectual contribution; P.P.L., R.M., R.L.M., S.W., and D.D.R. wrote the manuscript. All the authors had final approval of the submitted version.

Full Financial Disclosures for the Previous 12 Months

P.P.L. receives grant funding from the National Institute of Neurological Disorders and Stroke. H.Y.E.C. receives funding from Research Grants Council of Hong Kong and The Chinese University of Hong Kong, and consultancy from Codex Genetics Limited. C.A.R. is a consultant for Huntington Study Group, Annexon, Azevan, NeuExcell, Skyhawk, and Spark and receives research grants from CHDI, NIH, Prilenia, Roche, Sage, and Teva. R.L.M. receives grant funding from the National Institute of Health and the ABCD Charitable Trust, and provided legal consultation. S.M.P. receives licensing and royalties from the University of Utah. S.W. receives grant funding from the National Institute of General Medical Sciences, the National Institute of Neurological Disorders and Stroke, and the National Science Foundation. H.F is currently employed by Gensci Pharmaceuticals, China. X.S. is currently employed by Jacobio Pharmaceuticals, Inc., China. N.A. is currently employed by Neuropsychiatry Center for Therapeutic Innovation, France. L.O.M. is currently a graduate student in Ross University School of Veterinary Medicine. E.H. is currently an active‐duty officer in the United States Army, while studying medicine at the Uniformed Services University of the Health Sciences F. Edward Hébert School of Medicine. D.D.R. is currently employed by National Center for Advancing Translational Sciences, National Institute of Health. H.F., X.S., N.A., L.O.M., E.H., and D.D.R. contributed to this work before current positions.

Figure S1. RT‐PCR detection of Rluc‐ATXN2 fusion transcript containing the CAG repeat in SK‐N‐MC cells overexpressing Rluc‐ATXN2‐(CAG)n. Primer pairs F1/R1 and F1/R2, which span the junction of the fusion construct, were used. Primer locations were as indicated in Fig. 1A. Negative control (water) was included.

Click here for additional data file.

Figure S2. No detectable FL ATXN2‐Q58 or ATXN2‐Q104 protein in SK‐N‐MC cells overexpressing Rluc‐ATXN2‐(CAG)58 or Rluc‐ATXN2‐(CAG)104. Anti‐ATXN2 Western blotting was performed. SK‐N‐MC cells overexpressing FL ATXN2‐Q58 and ATXN2‐Q104 were used as positive controls, whereas untransfected SK‐N‐MC cells were used as a negative control. β‐actin was included as an internal loading control.

Click here for additional data file.

Figure S3. 3′ UTR approach eliminated ATG‐initiated expression of EGFP. (A) Schematic presentation of Rluc‐EGFP construct. ATXN2 ORF (Fig. 1A) was replaced by EGFP ORF at the 3′ UTR of Rluc gene. (B–D). No detectable EGFP signal in SK‐N‐MC cells overexpressing Rluc‐EGFP (B–C), whereas pEGFP‐N1 overexpressing SK‐N‐MC cells were used as a positive control (D–E). Scale bar, 50 μm. n = 3 independent experiments and representative images were shown. Scale bar, 50 μm.

Click here for additional data file.

Figure S4. Non‐ATG initiated (RAN) translation does not contribute to the toxicity of expATXN2 in SK‐N‐MC cells. (A) Schematic presentation of the ATXN2‐(CAG)n constructs with the 6X STOP cassette and three tags (Flag, HA, and myc) in three ORFs. (B–E) SK‐N‐MC cells were transfected with ATXN2‐(CAG)n constructs, and the presence of RAN translation products in polyQ, polyalanine (polyAla), or polyleucine (polyLeu) ORFs was assessed by Western blotting 72 hours post‐transfection. pcDNA3.1 empty vector was used as a negative control. eEF1A1‐Myc‐Flag and Akt‐HA plasmids were used as positive controls for antibodies used in the experiment. β‐actin was used as a loading control. Note that positive controls were purposefully under‐loaded n = 3 independent experiments and representative blots were shown.

Click here for additional data file.

Figure S5. TBL3 does not interact with exogenous GFP transcript in the RNA pulldown assay. RNA pulldown was performed as previously described for Fig. 3B,C. Anti‐TBL3 Western blotting of the RNA pulldown samples indicates that TBL3 does not bind to in vitro transcribed GFP RNA, with HTT‐(CAG)80 RNA was a positive control n = 3 independent experiments and a representative blot was shown.

Click here for additional data file.

Figure S6. expATXN2 RBPs grouped by functions. A total of 57 preferential expATXN2 RBPs that were not present in the bead only control and were (1) either identified as binding exclusively to expATXN2 (with 58 and/or 104 repeats); or (2) had at least twice the number of peptide hits in ATXN2‐(CAG)104 pull own compared to that in ATXN2‐(CAG)22, were included in the analysis. (A) Go analysis of functional annotation. Preferential expATXN2 interactors were submitted to Metascape (https://metascape.org/gp/index.html#/main/step1) for GO analysis of functional annotation. P value cut off was 0.05. ‐Log10(P) was shown. (B), STRING network analysis using the STRING web server shows major subnetworks and potential protein–protein interactions among the 57 preferential expATXN2 RBPs. Node colors represent different subnetworks based on k means clustering of 2. RBPs that are not connected within the network are not shown. The interaction score was set at 0.9 with the highest confidence. Two major clusters shown as pre‐mRNA splicing (red) and ribosome biogenesis (green). Note that five proteins (TBL3, UTP18, WDR3, WDR36, and PWP2; underlined) are components of the small subunit (SSU) processome for ribosomal RNA processing.

Click here for additional data file.

Table S1. A full list of RBPs that were identified as preferentially or specifically binding to expATXN2 from in vitro biotinylated RNA pull‐down assay followed by mass spectrometry (MS).

Click here for additional data file.