Metformin improves RAN protein pathology, alternative splicing, and behavioral phenotypes in SCA8 mice
Lisa EL Romano, Setsuki Tsukagoshi, Emily E Davey-Osuch, Ramadan Ajredini, Kamat Manasi, Tala VR Ortiz, Eduardo Rijos, Nathan J Bourgon, S Elaine Ames, Timothy J Garrett, John D Cleary, Eric T Wang, Laura PW Ranum
TL;DR
Metformin improves symptoms and pathology in SCA8 mice, suggesting potential for treating CAG•CTG expansion diseases.
Contribution
Metformin reduces RAN proteins and neuroinflammation in SCA8 mice, offering a novel therapeutic approach.
Findings
Metformin improves ambulatory performance in SCA8 mice using behavioral tests.
Metformin reduces RAN protein levels and splicing abnormalities without altering RNA levels.
Metformin decreases neuroinflammation and glial activation in SCA8 mice.
Abstract
Spinocerebellar ataxia type 8, a debilitating neurological disease with no effective treatment, is caused by a CAG•CTG expansion mutation. In SCA8 mice, metformin decreases repeat-associated non-AUG proteins and neuroinflammation, improves behavior, and partially corrects splicing abnormalities. Spinocerebellar ataxia type 8 (SCA8) is a member of a group of dominantly inherited, debilitating neurological diseases caused by CAG•CTG expansions for which there are no effective treatments. RAN translation, which was discovered in SCA8, has previously been shown to occur across CAG and CUG expansion transcripts, making treatments for SCA8 potentially relevant to a broad group of diseases, including SCA1, SCA2, SCA3, SCA6, SCA7, SCA12, Huntington’s disease, and myotonic dystrophy type 1. In addition, CUG and CAG expansion transcripts have been reported to cause RNA gain-of-function effects.…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure S1
Figure 2
Figure S2
Figure 3
Figure S3
Figure S4
Figure 4- —HHS | National Institutes of Health (NIH)http://dx.doi.org/10.13039/100000002
- —U.S. Department of Defense (DOD)http://dx.doi.org/10.13039/100000005
- —National Ataxia Foundation (NAF)http://dx.doi.org/10.13039/100002243
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsGenetic Neurodegenerative Diseases · RNA Research and Splicing · Amyotrophic Lateral Sclerosis Research
Introduction
Spinocerebellar ataxia type 8 (SCA8), a dominantly inherited form of ataxia with reduced disease penetrance, is caused by a CAG•CTG repeat expansion mutation located in the overlapping ATXN8/ATXN8OS genes (1). This expansion mutation, traditionally thought to be noncoding, is bidirectionally transcribed and produces both CAG and CUG expansion transcripts (2, 3). Similar to myotonic dystrophy type 1 (DM1), SCA8 CUG expansion transcripts form RNA foci and cause RNA gain-of-function alternative splicing abnormalities (3). Additional characterization of the SCA8 expansion mutation led to the discovery of RAN (repeat-associated non-AUG) translation, a process in which repeat expansion RNAs can produce mutant expansion proteins in all three reading frames without AUG- or AUG-like initiation codons (4). In SCA8 BAC mice and SCA8 human autopsy tissue, polyGln, polyAla, and polySer RAN protein aggregates accumulate in affected tissues, including the cerebellum and brainstem (2, 4, 5). RAN proteins have also been shown to accumulate in the CAG•CTG expansion disorders DM1 and Huntington’s disease (HD) and a growing number of diseases caused by intronic repeat expansion mutations, including C9orf72 amyotrophic lateral sclerosis and frontotemporal dementia (C9orf72 ALS/FTD) (6, 7, 8), spinocerebellar ataxias (SCA1, SCA2, SCA3, SCA6, and SCA7) (9), cerebellar ataxia neuropathy vestibular and areflexia syndrome (CANVAS) (10), and Alzheimer’s disease (AD) (11).
In 2020, Zu et al (12) showed that RAN translation across multiple repeat motifs, including CAG expansions, is regulated by the double-stranded RNA-dependent protein kinase (PKR) pathway and that PKR inhibition with the dominant-negative mutant PKR-K296R or the FDA-approved type 2 diabetes drug metformin reduces RAN proteins. Metformin was also shown to improve molecular, neuropathological, and behavioral phenotypes in C9orf72 ALS/FTD mice (12). More generally, metformin has been shown to have beneficial effects on aging and a number of neurodegenerative disorders (13, 14, 15, 16, 17, 18).
There is strong support that metformin improves phenotypes in DM1 and HD, two diseases caused by CAG•CTG expansion mutations. In DM1 patients, metformin was shown to improve mobility and gait (19). In HD, metformin improved disease phenotypes in three different mouse models (20, 21, 22) and cognitive function in patients (23). Although these studies support the potential utility of metformin for DM1 and HD, the molecular effects of metformin on relevant CAG•CTG expansion pathways remain unclear. Here, we test the therapeutic potential of metformin in SCA8 mice and its molecular effects on relevant CAG•CTG expansion pathways including RAN translation and RNA gain-of-function effects.
Results
Metformin ameliorates behavioral phenotypes in SCA8 mice
Based on the beneficial effects of metformin in DM1 and HD, we tested the hypothesis that metformin improves disease phenotypes caused by the CAG•CTG expansion mutation in SCA8 BAC transgenic mice.
From 4 to 7 wk, SCA8 mice and their nontransgenic (NT) littermates were given water with or without 2 mg/ml metformin (Fig 1A). At 8 wk of age, the dose was increased to 5 mg/ml, a concentration previously shown to result in plasma levels of ∼10 μM, which are comparable to conventional human doses of 20 mg/kg day used in diabetic patients (24, 25).
*Metformin treatment improves behavioral performance of SCA8 mice.(A) Schematic of experimental timeline. (B) Quantification of latency to fall of 32-wk-old mice treated with or without metformin. (C, D) Quantification of brake (C), step angle (D) as examples of two corrected forelimb DigiGait parameters. (E, F) Quantification of % brake stride (E) and stride length (F) as examples of two corrected hindlimb DigiGait parameters. (G, H, I) Quantification of ambulatory distance (G), ambulatory speed (H), and resting time (I) as three examples of improved open-field parameters. Error bars = SEM, n = 15 mice/group. Ordinary one-way ANOVA statistical test with Šídák’s multiple comparisons test, ***P < 0.0001, ***P < 0.001, **P < 0.01, P < 0.05.
To assess the effects of metformin on ambulatory function, SCA8 and NT littermates treated with or without metformin underwent rotarod testing at 8 and 32 wk before and after the onset of overt phenotypes, respectively (n ≥ 14 animals/group). As expected, no differences in performance were found between presymptomatic SCA8- and NT-treated and untreated groups at 8 wk of age (Fig S1A). At 32 wk of age, untreated SCA8 mice exhibited a shorter latency to fall than their NT littermates. SCA8 mice treated with metformin showed marked improvement with increased latency to fall (398 ± 38 s) compared with untreated SCA8 mice (214 ± 33 s, P = 0.0045). Rotarod performance of metformin-treated SCA8 mice was not significantly different from metformin-treated (462 ± 40 s) or untreated NT littermates (430 ± 36 s) (Fig 1B).
Metformin treatment improves behavioral performance of SCA8 mice.(A) Quantification of latency to fall of 8-wk-old mice (predisease) treated with or without metformin. (B, C) Quantification of brake (B) and step angle (C) performance of 16-wk-old mice (predisease) as examples of two parameters that are impaired in SCA8 mice after onset of the disease. (D) Fractions of total parameters of DigiGait forelimb behavioral test that are significantly different in SCA8-NT. The graph represents P-value comparisons of 14 DigiGait parameters among SCA8-NT and SCA8-SCA8met cohorts. (E) Fractions of total parameters of DigiGait hindlimb behavioral test that are significantly different in SCA8-NT. The graph represents P-value comparisons of 11 DigiGait parameters among SCA8-NT or SCA8-SCA8met cohorts. (F) Fractions of total parameters of open-field behavioral test that are significantly different in SCA8-NT. The graph represents P-value comparisons of 11 DigiGait parameters among SCA8-NT or SCA8-SCA8met cohorts. In all the graphs, yellow boxes define regions of significance compared with SCA8-NT and SCA8-SCA8met. Red datapoints indicate parameters with P ≤ 0.05 that are rescued and so within the yellow box. Black datapoints define parameters that are significantly different in SCA8-NT but are not rescued by metformin (n ≥ 14/group).
DigiGait analyses were used to quantify specific features of gait dynamics, posture, and forelimb–hindlimb stepping symmetry. No differences in performance were found at 16 wk between asymptomatic SCA8- and NT-treated or untreated groups (Fig S1B and C). At 32 wk of age, 14 forelimb DigiGait parameters differed between SCA8 and NT animals. Of these, 8 of 14 (57%) DigiGait abnormalities improved in metformin-treated SCA8 mice (Fig S1D). Examples of improved parameters, such as brake (P = 0.003) and step angle (P = 0.0022), are shown in Fig 1C and D, respectively. Similarly, metformin rescued 9 of 11 hindlimb abnormalities (Fig S1E), including % brake stride (P = 0.0058, Fig 1E) and stride length (P = 0.0018, Fig 1F).
Open-field studies done at 52 wk identified nine parameters that differed between untreated SCA8 and NT cohorts. Of these, six of nine parameters were improved in metformin-treated versus untreated SCA8 mice (Fig S1F). Improved phenotypes include ambulatory distance (P = 0.0231, Fig 1G), ambulatory speed (P = 0.0302, Fig 1H), and resting time (P = 0.0134, Fig 1I). Taken together, these data demonstrate that metformin increases ambulatory function and improves various aspects of motor function in SCA8 mice.
Metformin decreases RAN protein levels without changing CAG•CTG repeat length or RNA levels
To explore the effects of metformin on RAN proteins, we measured the levels of RAN proteins in affected brain regions from SCA8 mice. Immunohistochemistry (IHC) analyses using our previously reported C-terminal polySer antibody (5) showed that metformin treatment decreased the number of polySer aggregates in SCA8 mouse cerebellum (P = 0.0013, Fig 2A–C). Similarly, IHC showed that metformin reduced the number (P = 0.0001, Fig 2D–F) and area (P = 0.0005, Fig S2A) of polySer aggregates in the brainstem, an aggregate-rich region in SCA8 mice. Aggregate characteristics (number, size, and area) were quantified using macro analysis of deconvoluted images (Fig S2B) in ImageJ.
*Metformin treatment reduced polySer and polyGln aggregates in SCA8 mouse brains.(A) Mouse cerebellum schematic showing the area of polySer aggregate accumulation. (B) Quantification of polySer aggregate number in the cerebellum. (C) Immunohistochemistry (IHC) of cerebellum regions showing reduced polySer aggregates in SCA8 mice treated with metformin. (D) Mouse brainstem schematic showing area of polySer aggregate accumulation. (E) Quantification of polySer aggregate number in the brainstem. (F) IHC of brainstem regions showing reduced polySer aggregates in SCA8 mice treated with metformin. (G) Mouse brainstem schematic showing area of polyGln aggregate accumulation. (H) Quantification of the number of polyGln aggregates in the brainstem. (I) IHC of brainstem regions showing reduced polyGln aggregates in SCA8 mice treated with metformin. (J) Mouse cerebellum schematic showing the area of polyGln aggregate accumulation used for quantification (lobules VIa, VII, VIII, IX). (K) Quantification of the number of polyGln aggregates. (L) IHC of cerebellum regions showing reduced polyGln aggregates in SCA8 mice treated with metformin. Black arrows indicate polyGln aggregates in Purkinje cells. Error bars = SEM, n > or = 5 mice/group. Quantification was performed using macro analysis in ImageJ software. Statistics were performed with one-way ANOVA for multiple comparisons with the SCA8 groups shown, ***P < 0.001, **P < 0.01, P < 0.05.
*Metformin treatment reduced RAN protein aggregate total area in SCA8 mouse brain.(A) Quantification of polySer shows reduced polySer aggregate total area in the brainstem of SCA8 mice treated with metformin. (B) Representative picture of automated quantification for polySer comparing SCA8 with NT mice. (C) Quantification of polyGln shows reduced polyGln aggregate total area in the brainstem of SCA8 mice treated with metformin. (D) Representative picture of automated quantification for polyGln comparing SCA8 with NT mice. (E) Bar graph showing relative mRNA levels of ATXN8 and ATXN8OS with and without metformin. FPKM values for each transcript are normalized to FPKM value for ATXN8OS. FPKM, fragments per kilobase of transcript per million mapped reads. (F, G) DNA fragment analysis of cerebellum (F) and brainstem (G). (H) Metformin levels measured by mass spectrometry in the brainstem and cerebellum of treated mice relative to those of untreated mice. Error bars = SEM, n > 3 mice/group. Statistical one-way ANOVA test, ****P < 0.0001, ***P < 0.001, P < 0.05.
SCA8 polyGln proteins have been shown to be expressed by both canonical translation using the AUG initiation codon immediately upstream of the CAG repeat and RAN translation (4). Metformin-treated SCA8 mice showed reduced levels of polyGln aggregates (P = 0.0076, Fig 2G–I), and polyGln aggregate area in SCA8 mouse brainstem (P = 0.0360, Fig S2C). Because of the larger size of polyGln aggregates, quantification was performed using different ImageJ macros specific to polyGln aggregate size (Fig S2D). In SCA8 mouse cerebellum, we quantified the number of aggregates in lobules VIa, VII, VIII, and IX (Fig 2J). Similar to the brainstem, the cerebellum showed a reduced number of polyGln aggregates in metformin-treated SCA8 mice (P = 0.0357, Fig 2K and L).
To rule out the possibility that the decreases in RAN protein levels were not a downstream consequence of decreased ATXN8 and ATXN8OS transcripts, RNA sequencing (RNA-seq) was performed on cerebellum and brainstem from SCA8-treated and untreated mice. ATXN8 and ATXN8OS transcripts were quantified by RNA-seq as previously described (5). These data show that ATXN8 transcripts are expressed ∼sixfold higher than ATXN8OS transcripts in SCA8 mouse brain (Fig S2E) and that metformin treatment does not alter the relative levels of sense or antisense transcripts. DNA fragment analyses showed the GGGGCC repeat lengths were unaltered by metformin treatment in cerebellum and brainstem (Fig S2F and G). Mass spectrometry was used to confirm that all animals in the treatment group had detectable levels of metformin in their brain lysates (Fig S2H). Taken together, these data demonstrate that metformin decreases RAN protein aggregates without affecting RNA levels or repeat length.
Metformin decreases neuroinflammation and microglial activation
In 2018, we reported oligodendrocyte abnormalities and reactive astrogliosis in polySer‐positive regions in SCA8 mouse brains (5). GFAP staining in polyGln- and polySer-rich areas of the brainstem and cerebellum showed reduced astrocytosis in SCA8 animals treated with metformin versus water, respectively (P = 0.0013, P = 0.0167, Figs 3A–C and S3A). GFAP staining in metformin-treated SCA8 mice was similar to that of NT littermate controls.
*Metformin treatment improved neuroinflammation and reduced microglial activation in SCA8 mice.(A) Metformin-treated SCA8 mice (60 wk) show reduced reactive astrogliosis (GFAP) and reduced activated microglia (Iba1) in the regions with prominent polyGln and polySer accumulation. (B, C, D, E) Quantification was performed using the immunohistochemistry profiler plugin in ImageJ software for % GFAP-positive staining in brainstem (B) and cerebellum (C), and % positive Iba1 microglial staining in brainstem (D) and cerebellum (E). Error bars = SEM, n > 3 mice/group. Statistical one-way ANOVA test, **P < 0.01, P < 0.05.
*Metformin treatment reduced microglial activation in SCA8 mice.(A) Representative immunohistochemistry panel showing reactive astrogliosis (GFAP) and activated microglia (Iba1) in the regions with prominent polyGln and polySer accumulation. (B) Representation of skeleton 3D analysis using the ImageJ plugin. (C, D, E) Quantification outcomes of skeleton 3D analysis showing total number of skeletons (C), total number of Iba1 branches (D), and average length of Iba1 branches (E). Error bars = SEM, n > 3 mice/group. Statistical one-way ANOVA test, **P < 0.01, P < 0.05.
IHC staining of Iba1 in the brainstem showed increased microglia in SCA8 compared with NT littermates (P = 0.0479, Fig 3A and D). In SCA8 mice treated with metformin, microglial levels were reduced to normal levels (P = 0.0286, Fig 3D). Similarly, Iba1 staining in the cerebellum was decreased in animals treated with metformin (P = 0.0243, Fig 3E). To determine whether metformin treatment influences microglial morphology, including ramification and branching associated with activated microglia, we analyzed our images using the ImageJ Skeleton analysis plugin (26). This analysis allows us to measure the number of Iba1-positive reactive microglia, the number of branches, branch length, and morphological changes (Fig S3B). Metformin-treated SCA8 animals showed reduced levels of Iba1 microglia compared with untreated SCA8 mice (P = 0.0105, Fig S3C). Similarly, in SCA8 mice metformin treatment decreased the number (P = 0.0049, Fig S3D) and length (P = 0.0130, Fig S3E) of microglial branches. Taken together, these results demonstrate that metformin improves astrogliosis and microglial abnormalities in RAN-positive regions of SCA8 mouse brains.
Metformin partially rescues SCA8 splicing abnormalities
Similar to DM1 (27), SCA8 CUG expansion RNAs form foci that sequester MBNL/Mbnl proteins and result in alternative splicing changes (3). To better understand the splicing dysregulation in SCA8, we performed RNA-seq using cerebellar and brainstem tissue from SCA8 and NT mice. All libraries met standard quality metrics in FastQC (version 0.12.0) (28) and were sequenced to depths of 38–53 million input reads, providing sufficient coverage for the analysis of alternative splicing (AS) changes. The percent spliced in (PSI) values were estimated using replicate multivariate analysis of transcript splicing (rMATS) (29). Significant mis-splicing events were defined as those with a >10% change in mean PSI in either direction and a false discovery rate (FDR) of <0.05. We identified 554 differentially spliced exons in SCA8 versus NT littermate controls. Of these, there were 266 skipped exons (SE), 104 alternative 5′ splice sites (A5SS), 107 alternative 3′ splice sites (A3SS), 58 retained introns (RI), and 19 mutually exclusive exon (MXE) events (Fig S4A). The heatmap (Fig 4A) and scatterplot (Fig S4B) show global profiling of significantly dysregulated AS events in SCA8 compared with NT controls. Metascape gene ontology (GO) enrichment analyses (30) were used to identify disease-relevant pathways. The most enriched GO terms fall into three main categories: (1) regulation of mRNA regulation and splicing; (2) DNA damage and repair; and (3) synaptic signaling (Fig 4B). Transcripts mis-spliced in SCA8 include those expressed from a number of genes previously implicated in other neurological diseases, including Tardbp (31, 32, 33), Adarb1 (34), Rad52 (35), Nfib (36), Unc13b (37, 38), Baiap2 (39), and Clcc1 (40, 41) (Fig S4C–E).
Dysregulation of alternative splicing in the SCA8 mouse model.(A) Significantly mis-spliced skipped exon (SE), retained intron (RI), mutually exclusive exon (MXE), alternative 5′ splice site (A5SS), and alternative 3′ splice site (A3SS) events in SCA8 versus NT mice, number of each event shown on bar, |ΔPSI| > 0.1, FDR < |0.05|. (B) Scatterplot of mean PSI for NT versus SCA8 mice for all splicing events measured; 554 mis-splicing events were detected as significantly mis-spliced (|ΔPSI| > 0.1, FDR < 0.05) and are highlighted in purple. Mis-splicing events in light pink represent the example reported in the following figure. (C, D, E) Percent spliced in (PSI) for select mis-splicing events in NT and SCA8 mice. (F) Comprehensive heatmap showing PSI values for all 554 significantly dysregulated events in SCA8 compared with NT and their trend in metformin-treated SCA8 mice (samples in which there are fewer than five reads for a given event are represented as a gray area in the heatmap).
Metformin partially rescued dysregulation of alternative splicing in the SCA8 mouse model–affected brain regions.(A) Heatmap showing 554 significantly regulated splicing events in SCA8 when compared to NT, using FDR < 0.05, ΔPSI > |0.1|. (B) Enrichment of gene ontology (GO) terms identified using Metascape analysis of the 554 skipped exon events dysregulated in SCA8 mice. Broad functional categories of terms are indicated by bar color. (C) Venn diagram showing the overlapping mis-splicing events between SCA8 water versus NT water when compared to SCA8 water versus metformin-treated SCA8 mice. (D) Heatmap showing the overlapping 114 significant mis-splicing events. (E, F, G) Percent spliced in (PSI) for transcripts belonging to specific GO terms, including DNA damage and repair (green) (E), synaptic signal (orange) (F), and the mRNA processing category (yellow) (G).
Metformin has previously been shown to affect splicing regulation in cancer (42, 43) and DM1 (44). To understand the effects of metformin on AS in SCA8, we compared the 554 significant mis-splicing events between SCA8 versus NT animals with dysregulated AS events in SCA8 versus metformin-treated SCA8 mice. The Venn diagram in Fig 4C shows that 114 of the AS events in SCA8 versus NT animals are also changed and potentially corrected in metformin-treated SCA8 mice versus water-treated SCA8 animals. The heatmap in Fig 4D shows that metformin-treated SCA8 mice exhibit significant rescue of ∼20% of the total (554) mis-splicing events. These events met read (5×), FDR (0.05), and PSI (0.1) thresholds. The heatmap shown in Fig S4F shows all 554 significant events between SCA8 and NT animals and, for metformin-treated SCA8 mice, includes all data that met read depth criteria (colored) and data that did not meet 5× read depth criteria (gray). Examples of dysregulated splicing events that are corrected by metformin include transcripts from genes involved in DNA damage and repair, such as Pms2 (45) and Ppp2r5c (46, 47) (Fig 4E), synaptic signaling including Cacna1b (48), Kcnq2 (49), and Tbccd1 (Fig 4F), and RNA processing Celf4 (50, 51) (Fig 4G).
Taken together, these data show that in addition to RAN protein accumulation (4, 5), SCA8 mice also exhibit widespread AS dysregulation and that metformin substantially improves both RNA dysregulation and SCA8 RAN protein levels.
Discussion
Here, we show that metformin improves multiple facets of disease in SCA8 BAC transgenic mice, including: (1) increased ambulatory function measured by DigiGait, rotarod, and open-field analyses; (2) reduced RAN protein aggregate levels; (3) decreased neuroinflammation, including reduced astrogliosis and microglial activation; and (4) rescued alternative splicing abnormalities. These data identify metformin as a promising drug that mitigates key behavioral and molecular phenotypes driven by the SCA8 repeat expansion mutation. These preclinical data provide support for testing the safety and efficacy of the FDA-approved drug metformin in SCA8, with potential, relevance to the broader group of CAG•CTG expansion disorders.
Metformin, a commonly prescribed drug for type 2 diabetes, has been reported to have benefits on aging (15, 52) and neurodegenerative disorders (15, 16). Metformin treatment was recently shown to decrease RAN protein levels and improve behavior in a BAC mouse model of C9orf72 ALS/FTD, a disease caused by a hexanucleotide GGGGCC repeat expansion (12). This study led to an open-label clinical trial (NCT04220021) to test the safety and therapeutic potential of metformin for C9orf72 ALS patients. In the context of CAG•CTG repeat expansion disorders, metformin has been shown to improve disease phenotypes in HD mouse models (20, 21, 22), HD patients (23), and DM1 patients (19). In a separate study using patient-derived DM1 myoblasts, 4 of 20 alternative splicing events were rescued by metformin treatment (44). Although these HD and DM1 studies examined behavior and selected RNA splicing abnormalities, both mutations have previously been shown to express RAN proteins (4, 53), raising the question of whether the beneficial effects of metformin in these diseases result from unexamined RAN protein and/or RNA effects. Our data showing that metformin reduces SCA8 RAN protein levels and RNA splicing abnormalities strongly suggest that metformin will also improve these repeat-driven pathologies across a broad range of CAG•CTG diseases, including SCA1, SCA2, SCA3, SCA6, SCA7, SCA12, DM1, HD, and FECD.
Although sense and antisense transcripts and RAN proteins have now been reported for more than a dozen repeat expansion diseases (54, 55, 56), most therapeutic strategies (e.g., antisense oligonucleotides [ASOs]) for these disorders have been focused on targeting only the sense transcripts. Unfortunately, ASO strategies have repeatedly failed for multiple expansion diseases, including C9orf72 ALS (57), HD (58), SCA2 (59), and SCA3 (60). Although it is not yet clear why each of these trials failed, the answer may be that knocking down the levels of sense RNAs leads to unintended consequences. Previous studies in SCA7 (61) and HD (62) have shown that knockdown of sense expansion transcripts leads to the up-regulation of the corresponding antisense expansion transcript. A decrease in the levels of sense transcripts with siRNAs or ASOs may disrupt sense/antisense transcript balance, leading to unintended increases in antisense transcript levels and toxicity. An advantage of metformin treatment, as shown in this study, is that it improved both RAN and RNA splicing phenotypes without changing the levels or the balance of sense or antisense expansion RNAs. In addition, metformin has a well-established safety record, making it a relatively safe approach for reducing RAN proteins. In patients, known adverse events of metformin range from common, mild gastrointestinal issues to rare, but serious, lactic acidosis, which results in severe lethargy. In our mouse study, we did not observe signs of gastrointestinal distress or lethargy in animals treated with metformin.
Our data also show that metformin reduces astrogliosis and microglial activation at sites of RAN protein pathology. Metformin has been previously shown to decrease neuroinflammation and its associated effects across multiple different diseases, including Alzheimer’s disease (AD), Parkinson’s disease (PD), multiple sclerosis (MS), and HD (15). Although to date, PD and MS have not been associated with repeat expansion mutations, RAN protein aggregates have been reported in HD (53) and a recent study shows RAN protein aggregates also accumulate in a substantial fraction of AD autopsy brains (11). Our SCA8 data raise the possibility that decreasing RAN protein aggregate burden through the use of metformin reduces neuroinflammation in SCA8. Alternatively, metformin may reduce neuroinflammation by activating the AMP-activated protein kinase (AMPK) pathway (13, 63, 64). In C9orf72 ALS/FTD cells and mice, reducing the levels of pPKR using either a dominant-negative PKR (K296R) or metformin decreases RAN protein levels (12). Future research aimed at blocking the PKR pathway or targeting SCA8 RAN proteins will lead to a better molecular understanding of how metformin improves disease in SCA8 mice.
CAG•CTG trinucleotide repeat expansion mutations underlie SCA8, DM1, HD, and multiple SCAs. In DM1, in addition to RAN translation, RNA gain-of-function of CUG expansion transcripts has been shown to play an important role in DM1 through the sequestration of the muscleblind-like (MBNL) family of proteins (27) and global disruption of alternative splicing. In one DM1 study, metformin was shown to have beneficial effects on selected alternative splicing abnormalities through a mechanism in which activation of AMPK down-regulates the levels of the cold-shock RNA-binding motif protein 3 (RBM3) (44).
Alternative splicing in SCA8 (3) and other SCAs (65) has been reported but has not been widely characterized. Here, we demonstrate substantial dysregulation of alternative splicing in SCA8 BAC transgenic mice with alterations in transcripts important for synaptic signaling, RNA processing/splicing, and DNA damage and repair. Moreover, we show that metformin treatment rescued ∼20% of these dysregulated alternative splicing events including transcripts linked to mRNA metabolism and synaptic signaling. Additional work will be needed to understand how metformin improves SCA8 mis-splicing abnormalities and whether metformin has direct or indirect effects on one or more RNA splicing factors.
In summary, our work shows that metformin treatment improves behavior, reduces RAN protein aggregates, decreases neuroinflammation, and corrects ∼20% of mis-spliced transcripts in SCA8 BAC transgenic mice without altering sense or antisense RNA levels. A distinct advantage of metformin over strategies that target and degrade sense expansion RNAs is that reducing RAN protein levels using metformin is predicted to lower both sense and antisense pathogenic proteins. Metformin’s RAN protein-lowering and anti-inflammatory properties, together with its long track record as a safe and affordable drug, make metformin an excellent candidate for clinical trials for SCA8 and potentially other CAG•CTG expansion disorders.
Materials and Methods
Study design
In this study, we sought to determine whether metformin can prevent and/or reverse the behavioral and pathological phenotypes of SCA8 transgenic mice. Because RAN proteins start to accumulate early in the disease, we designed our study by treating female and male SCA8 BAC transgenic mice beginning at the age of 4 wk, well before the development of behavioral and pathological phenotypes. During the treatment period of 60 wk, animals were assessed for behavioral defects at 8, 16, 32, and 52 wk. Final take-downs were done at 60 wk of age.
Mouse model
Mice used in this study were housed and treated in accordance with the NIH Guide for the Care and Use of Laboratory Animals. The Animal Care and Use Committee approved all animal studies at the University of Florida. Previously described SCA8 BAC transgenic lines on the FVB background (BAC EXP2, 2878) were used (2).
Metformin administration
Male SCA8-BAC mice were bred with female FVB/N mice obtained from Jackson Laboratory to generate cohorts of mice for metformin treatment. Hemizygous mice carrying the SCA8 BAC transgene were genotyped by PCR as previously described (2). Female and male SCA8-positive mice were separated into different cohorts, along with their NT littermates, and were treated with or without metformin in the drinking water. SCA8 and NT control mice in the treatment group received 2 mg/ml from 4 through 7 wk of age. At 8 wk of age, they received 5 mg/ml metformin in drinking water for up to 60 wk of age. Metformin hydrochloride was purchased from Thermo Fisher Scientific (M1566, 1115-70-4, C4H11N5·HCl).
Blinding procedure
Researchers handling the animals were blinded during the following experiments: (1) in vivo efficacy study, behavior, and IHC analyses; and (2) IHC and histological staining; quantification was performed using ImageJ by a researcher who did not perform the staining.
Rotarod analysis
Rotarod training was performed at 8 and 32 wk of age using an accelerating rotarod (Ugo Basile) as previously described (5). Three trials were run per day for 4 d, and the averages of the three trials on day 4 are presented. One-way ANOVA was performed, followed by post hoc multiple comparisons analysis to assess differences in rotarod performance between groups (NT untreated, SCA8 untreated, NT metformin-treated, and SCA8 metformin-treated).
Gait analysis
DigiGait analysis was performed at 16 and 32 wk of age on animals from all experimental groups. Digital video images of the underside of the mouse were collected with a high-speed video camera from below the transparent belt of a motorized treadmill (DigiGait Imaging System, Mouse Specifics). Each mouse was allowed to explore the treadmill compartment with the motor speed set to 14 cm/s for 1 min, and then, the motor speed was increased to 24 cm/s for video recording. Only video recordings in which the mouse walked straight ahead with a constant relative position with respect to the camera were used for analysis. Data from each paw were analyzed with DigiGait automated gait analysis software (Mouse Specifics). All data analyses were performed in a blinded fashion. One-way ANOVA was performed, followed by post hoc multiple comparisons analysis as described above.
Open-field analysis
Open-field analysis was performed at 52 wk by testing mouse behavior during a 30-min session in a completely dark open chamber (17″ × 17″) (Med Associates). Approximately 2 h before the start of analysis, mice were placed in the behavior room to allow for acclimation to the room. Mice were then placed in the center of the darkened activity-monitoring chamber. The trace path and center time were recorded and analyzed with Activity Monitor (MED Associates, Inc.) software. All analyses were performed in a blinded fashion. One-way ANOVA was performed, followed by post hoc multiple comparisons analysis as described above.
Tissue processing for histopathological and downstream analysis
For histological analysis, animals were anesthetized using 100 mg of ketamine and 10 mg of xylazine per kg of body weight and perfused through the ascending aorta with 15 ml of isotonic saline. Half of the brain was collected for histopathological analysis and fixed in 10% buffered formalin. The other half of the brain was harvested, dissected, and snap-frozen for total subsequent RNA isolation or protein isolation.
Histology, IHC, and imaging
For the detection of polySer RAN protein in fixed brain tissue, animals were perfused transcardially with 1× PBS. The brain was dissected and fixed in 10% formalin for 24 h and later removed into 70% ethanol. After histological processing and paraffin embedding, 5-μm sagittal sections were cut using a microtome. Sections were deparaffinized in xylenes (2 × 15 min) and rehydrated through an alcohol gradient (100%, 100%, 95%, 80%; 10 min each step). Sections were then treated with the following antigen retrieval steps: first, 1 μg/ml proteinase K treatment in 1 mM CaCl_2_, 50 mM Tris buffer (pH = 7.6) for 30 min at 37°C; second, steam in 10 mM EDTA (pH = 6.5) for 30 min using a steamer; and third, 95% formic acid treatment for 5 min. Endogenous peroxidase was blocked in 3% H_2_O_2_ methanol for 12 min. To block nonspecific binding, a nonserum block (Biocare Medical) was applied for 15 min. Primary antisera were diluted in 1:10 nonserum block/water at the concentrations indicated below and incubated at 4°C overnight. Rabbit Linking Reagent (streptavidin) was applied for 30 min at RT. Secondary antibodies were Biotin–Avidin/Streptavidin labeled using ABC reagent (Vector Laboratories, Inc.). Each step was followed by 3 × 5-min washes in PBS. Antibody detection was performed using Vector Red Substrate Kit (Vector Laboratories, Inc.). Slides were washed in running water (10 min) and counterstained using Hematoxylin QS (Vector Laboratories) at RT (2 min). After washing in water for 5 min, slides were dehydrated (ethanol 80%, 95%, 100%, xylene) and mounted using Cytoseal 60 (Electron Microscopy Sciences). For the detection of polyGln, the antigen retrieval steps are different, and they include 95% formic acid treatment for 5 min and steaming in 10 mM citrate buffer (pH = 6.8) for 30 min using a steamer. Antibody detection was performed using a 3,3′-diaminobenzidine (DAB) peroxidase substrate kit (Vector Laboratories). Slides were washed in running water (10 min) and counterstained using Hematoxylin QS (Vector Laboratories) at RT (2 min). For the detection of astrocytes and microglia, the protocol is similar to the one described above, except the antigen retrieval steps are different. For the detection of GFAP, slides are incubated in 10 mM citrate buffer (pH = 6.8) for 30 min. For the detection of Iba1, slides are steamed in 10 mM citrate buffer (pH = 6.8) for 30 min, and then they are incubated for 5 min in 95% formic acid for 5 min. Primary antibodies/sera were used at the following conditions in mouse tissue: polySerCT RAN proteins (5) (Project #1306, rabbit #F3672, 1:10,000; NEP), polyGln 1C2 (mouse, 1:15,000; Millipore), astrocytes GFAP (Ab7260-GFAP, 1:5,000; Abcam), and microglia Iba1 (Ab 1:5,000; Abcam). Images were captured with an Olympus BX51 light microscope using either a 20× or a 40× objective, depending on the downstream analysis.
Tissue for RNA isolation and RNA-seq library preparation
Mouse tissue from cerebellum and brainstem was lysed and homogenized with 800 μl TRIzol reagent (15596018; Thermo Fisher Scientific) in 2-ml tubes prefilled with 1.5-mm high-impact zirconium beads (D1032-15; Benchmark Scientific). Tubes were agitated at 10,000g for 1 min using a Bead Ruptor 12 (19-050A; OMNI International). Then, they were placed on ice for 1 min, for a total of four repetitions for complete tissue lysis. RNA was then extracted following the Direct-zol RNA Miniprep kit protocol, with DNase I treatment (R2070; Zymo Research). RNA concentration was measured using a NanoDrop One spectrophotometer (Thermo Fisher Scientific, ND-ONE-W). RNA integrity was assessed by measuring the RNA integrity number using capillary electrophoresis (M5310AA; Agilent) paired with the RNA kit (DNF-471-0500; Agilent). Samples with an RQN > 8 were further processed. RNA-seq libraries were constructed using NEBNext Ultra II Directional RNA Library Prep Kit for Illumina, using ribosomal RNA depletion followed by strand-specific RNA-seq preparation. Samples were amplified with PCR for 9–11 cycles and sequenced using the Illumina NextSeq 2000. A total of 300 ng RNA was used as input for library preparation. rRNA was depleted from samples (E7405; NEB), and libraries were prepared for paired-end Illumina sequencing (E7760L; NEB) with the following adjustments to the manufacturer’s protocol: 1:40 adaptor dilution, universal and indexing primers at 1 μM final concentration, and 10 cycles for PCR enrichment of adaptor-ligated DNA. Library quality was assessed via capillary gel electrophoresis (M5310AA; Agilent) paired with the High Sensitivity Next Generation Sequencing Fragment Kit (DNF-474-0500; Agilent). Library concentrations were determined using Library Quant Kit for Illumina (E7630L; NEB) following the manufacturer’s protocol. Sequencing was performed on the Illumina NextSeq 2000 platform (20038897; Illumina) using the onboard denature and dilute protocol, paired with P3 reagents for 200 cycles (20040560; Illumina). Sequencing output was demultiplexed and converted to FASTQ format via BaseSpace BCL Convert (version 2.3.0).
Read mapping, splicing isoform quantitation, and gene ontology enrichment analysis
FASTQ file quality and total number of reads were assessed using FASTQC (version 0.11.9), and datasets with an average read depth >35 million paired-end reads were accepted for this study. FASTQ files were aligned to the GRCm39/mm39 reference genome, which was modified to include the human ATXN8OS coding sequence, using STAR. To compare expression levels of ATXN8 and ATXN8OS, a human reference genome containing ATXN8OS and 10-kb upstream and downstream sequence was indexed using HiSat2 as described before (5) and raw RNA-seq reads were aligned to this custom reference genome. Fragment per kilobase million (FPKM) of ATXN8 and ATXN8OS was calculated by normalizing the read count mapped to gene‐specific regions of ATXN8 and ATXN8OS to the length. The difference was calculated by dividing both FPKM values by ATXN8OS FPKM value. This analysis was done on four biological replicates. Using BAM files as input, PSI values were quantified with rMATS-turbo (version 4.2.0) (29) to identify dysregulated exons. AS events were considered to be dysregulated according to the FDR (≤0.05) and absolute value of the inclusion-level difference (|Δ PSI| ≥ 0.1) when comparing SCA8 mice to NT mice. rMATS raw files are provided in Tables S1 and S2. Heatmaps were generated using the ComplexHeatmap (2.10.0) R package. Gene ontology analysis was performed using Metascape (version v3.5.20230101) (30) with default express analysis.
Table S1. Alternative splicing analysis comparing SCA8 water and NT water.
Table S2. Alternative splicing analysis comparing SCA8 metformin and SCA8 water.
Mass spectrometry (UHPLC-HRMS/MS)
Cerebellum and brainstem protein lysates from mice were quantified and diluted to make a 250 μg/ml solution in RIPA. 50 μl of each sample was transferred to a labeled Eppendorf tube. Metformin calibration curve: Calibration curve was prepared by spiking metformin standard in albumin at levels—0.25, 1, 10, 25, 50, 100, 200, 400, and 800 ng/ml. QCs were prepared at 15, 150, and 600 ng/ml levels. Metformin-d6 (1 μl of 500 ng/ml concentration) was added as internal standard to each calibration level, QCs, and samples. Extraction was performed by adding 200 μl of acetonitrile to precipitate proteins. The samples were then centrifuged, and the supernatant was transferred to LC vials for analysis by UHPLC-HRMS/MS. Instrumentation: Samples were analyzed on a Thermo Q Exactive Orbitrap mass spectrometer with Dionex UHPLC by positive heated electrospray ionization, with the following settings: mass resolution of 35,000 for full scan and 17,500 for MS/MS; full scan range—m/z 70–1,000 and MS/MS—m/z 130.1087 (metformin) and 136.1463 (metformin-d6) at NCE 55; mobile phase A—95:5—10 mM ammonium formate in water with 0.1% formic acid: acetonitrile; mobile phase B—95:5—acetonitrile: water with 0.1% formic acid; Column—Agilent HILIC Plus RRHD 1.8 μm; 2.1 × 150 mm; flow rate—0.4 ml/min; injection volume—2 μl.
DNA fragment analysis PCR for SCA8 mice using 3730xl analyzer
To examine the ATXN8 repeat length, genomic samples from mouse cerebellum and brainstem were analyzed by a PCR consisting of 1X Phire reaction buffer, 0.5 mM dNTPs, forward primer (0.5 μM), reverse primer (0.5 μM), DNA 20–50 ng/reaction, and 1U Phire Hot Start II DNA polymerase (Thermo Fisher Scientific). PCR cycling conditions were 98°C for 3 min, followed by 30 cycles of 98°C for 20 s, 62°C for 20 s, 72°C for 1 min 15 s, and a final extension at 72°C for 3 min. Primer sequences are SCA8-1F-FAM: 56-FAM/TTTGAGAAAGGCTTGTGAGGA, SCA8-1R: TCTGTTGGCTGAAGCCCTAT. FAM-labeled PCR products were mixed with GeneScan 1200 LIZ dye Size Standard (437995; Applied Biosystems) and analyzed on an ABI3730xl DNA analyzer (Applied Biosystems), and the data were analyzed using GeneMarkers software (version 1.75, SoftGenetics).
Statistical analysis
Sample sizes needed for behavioral, histological, and biochemical experiments were based on previously published papers using this model (1, 5) (e.g., n > 14 animals/group for RAN protein staining comparisons; n > 14 animals/group for DigiGait, rotarod, and open-field analyses; n > 3 animals/group for neuroinflammation comparison; n = 4 animal/group for RNA-seq, splicing, and transcript analyses). For quantification of polySer and polyGln aggregates, we use ImageJ scripts/macros explicitly designed for each protein. For GFAP and Iba1, we used IHC profiler (66) to quantify the total positive signal, and the skeleton plugin was used to quantify specifically Iba1 (26). Differences between groups were determined by a multiple comparisons ANOVA test in GraphPad software. Alternative splicing statistical tests comparing WT and SCA8 and treated SCA8 were performed concurrently with rMATS-turbo with default settings.
Supplementary Material
Reviewer comments
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1Koob MD, Moseley ML, Schut LJ, Benzow KA, Bird TD, Day JW, Ranum LP (1999) An untranslated CTG expansion causes a novel form of spinocerebellar ataxia (SCA 8). Nat Genet 21: 379–384. 10.1038/771010192387 · doi ↗ · pubmed ↗
- 2Moseley ML, Zu T, Ikeda Y, Gao W, Mosemiller AK, Daughters RS, Chen G, Weatherspoon MR, Clark HB, Ebner TJ, (2006) Bidirectional expression of CUG and CAG expansion transcripts and intranuclear polyglutamine inclusions in spinocerebellar ataxia type 8. Nat Genet 38: 758–769. 10.1038/ng 182716804541 · doi ↗ · pubmed ↗
- 3Daughters RS, Tuttle DL, Gao W, Ikeda Y, Moseley ML, Ebner TJ, Swanson MS, Ranum LPW (2009) RNA gain-of-function in spinocerebellar ataxia type 8. P Lo S Genet 5: e 1000600. 10.1371/journal.pgen.100060019680539 PMC 2719092 · doi ↗ · pubmed ↗
- 4Zu T, Gibbens B, Doty NS, Gomes-Pereira M, Huguet A, Stone MD, Margolis J, Peterson M, Markowski TW, Ingram MAC, (2011) Non-ATG-initiated translation directed by microsatellite expansions. Proc Natl Acad Sci U S A 108: 260–265. 10.1073/pnas.101334310821173221 PMC 3017129 · doi ↗ · pubmed ↗
- 5Ayhan F, Perez BA, Shorrock HK, Zu T, Banez-Coronel M, Reid T, Furuya H, Clark HB, Troncoso JC, Ross CA, (2018) SCA 8 RAN poly Ser protein preferentially accumulates in white matter regions and is regulated by e IF 3F. EMBO J 37: e 99023. 10.15252/embj.20189902330206144 PMC 6166133 · doi ↗ · pubmed ↗
- 6Ash PE, Bieniek KF, Gendron TF, Caulfield T, Lin WL, Dejesus-Hernandez M, van Blitterswijk MM, Jansen-West K, Paul JW 3rd, Rademakers R, (2013) Unconventional translation of C 9ORF 72 GGGGCC expansion generates insoluble polypeptides specific to c 9FTD/ALS. Neuron 77: 639–646. 10.1016/j.neuron.2013.02.00423415312 PMC 3593233 · doi ↗ · pubmed ↗
- 7Mori K, Weng SM, Arzberger T, May S, Rentzsch K, Kremmer E, Schmid B, Kretzschmar HA, Cruts M, Van Broeckhoven C, (2013) The C 9orf 72 GGGGCC repeat is translated into aggregating dipeptide-repeat proteins in FTLD/ALS. Science 339: 1335–1338. 10.1126/science.123292723393093 · doi ↗ · pubmed ↗
- 8Zu T, Liu Y, Bañez-Coronel M, Reid T, Pletnikova O, Lewis J, Miller TM, Harms MB, Falchook AE, Subramony SH, (2013) RAN proteins and RNA foci from antisense transcripts in C 9ORF 72 ALS and frontotemporal dementia. Proc Natl Acad Sci U S A 110: E 4968–E 4977. 10.1073/pnas.131543811024248382 PMC 3870665 · doi ↗ · pubmed ↗