Potent Mixed Backbone Antisense Oligonucleotide Safety Suppressed Expression of Mutant C9ORF72 Transcripts and Polypeptides: First in Human Pilot Study


 Expansions of a G4C2 repeat in the C9ORF72 gene are the most common genetic cause of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD), two devastating adult-onset neurodegenerative disorders. Proposed disease mechanisms include a gain of toxic functions of the G4C2 repeats, implying that selective reduction in levels of the repeat-containing transcripts would represent a treatment strategy for this disorder. In the present study, using C9-ALS/FTD patient derived cells and C9ORF72 BAC transgenic mice, we have generated and optimized antisense oligonucleotides (ASOs) that selectively blunt expression of G4C2 repeat containing transcripts in both the sense and anti-sense strands of C9ORF72 and effectively suppress tissue levels of polyGP dipeptides. In a single patient harboring mutant C9ORF72 with the G4C2 repeat expressions, repeated dosing by intrathecal delivery of the optimal ASO was well tolerated, leading to significant reductions in levels of CSF polyGP.

Antisense oligonucleotides (ASOs) can drive therapeutic effects by mechanisms that include splice-modulation 26 or, if the ASO contains DNA, activation of endogenous RNase H 27,28 to degrade the target RNA. The broad bioavailability of ASOs in the CNS, including both neurons and glial cells 29 has prompted development of ASOs as therapy for dominantly transmitted genetic disorders of the central nervous system (e.g. Huntington's disease and ALS caused by mutations in the SOD1 gene) [30][31][32][33] .
Here, we report development of antisense oligonucleotides targeting the sense and antisense strands of C9ORF72 to treat ALS and FTD caused by the HRE in the V1 and V3 isoforms. Using different C9 related model systems, including patient-derived samples and two C9BAC transgenic mouse models 34,35 , we have generated ASOs that specifically reduce levels of the transcripts harboring the HRE as well as their DPR products, with minimal effects on the most abundant V2 isoform which does not contain the HRE. We show that modification of a subset of the phosphodiester internucleoside linkages significantly improves the tolerability without impairing the potency of the tested ASOs. We demonstrate that in a single patient harboring mutant C9ORF72 with the G4C2 repeat expressions, repeated intrathecal dosing of the optimal ASO was well tolerated and led to significant reductions in levels of CSF polyGP.

2708
G4C2 targeting ASO reduces the C9ORF72 repeat containing transcripts in patient derived fibroblasts and C9BAC mouse derived neurons 21 4 Because haploinsufficiency of C9ORF72 is thought to be adverse, we developed ASOs that target only the 5'end of transcripts V1 and V3 that bear the G4C2 repeat expansion, sparing transcript V2.
As it is not fully clear whether the repeat containing intron is retained or spliced out, we focused our effort on ASO sequences targeting the intron-repeat junction (Fig. 1a). Others have previously tested ASOs against this target region in patient derived samples in vitro with success 36,37 .
We developed a dual luciferase screen for ASOs that suppress expression of transcripts V1,V2 and V3 (Fig. S1a), and used this assay to narrow our focus to five ASOs. All ASOs were designed as gapmers to elicit RNase H-mediated target degradation, and all linkages were fully phosphorothioate (PS) modified. Sugars of the 5' and 3' regions of the ASOs were modified with either locked nucleic acid (LNA) or 2'-O-methoxyethyl (MOE) substitutions (Fig. 1b). All cytosine residues were 5-methylated to reduce immunogenicity 38 .
We treated primary C9-ALS/FTD patient-derived fibroblasts (with >1000 repeats, Fig. S1b, and showing visible nuclear C9ORF72 foci, Fig. S1c) with ASOs 1-5 at a dose of 100 nM by lipid transfection. After 72 hours, we assayed the level of V1-V3 repeat containing transcripts by qRT-PCR. Relative to untreated cells (UT) or a non-targeting control (NTC), all five ASOs reduced V1-V3 expression to almost undetectable levels (Fig. 1c). Silencing was dose-dependent in patientderived fibroblasts (Fig. S1d) and in HEK293 cells expressing the dual luciferase reporter (Fig.

S1e).
A hallmark of C9-ALS/FTD is the presence of repeat-containing RNA foci 1,2,17,36,39-41 . Three days after ASO treatment, we used FISH to show that the number of cells with foci was markedly reduced from 80% in the untreated condition to 20-40% in the treated conditions. Moreover, fewer foci per cell were detected, showing that all five ASOs were potent inhibitors of G4C2 RNA foci ( Fig. 1d-e).
Each of the five ASOs was also active by gymnotic (lipid-free) delivery to neurons. Primary cortical neurons were derived from E15.5 C9BAC embryos and treated with 1µM ASO at 5 days in vitro (DIV). 15 days after treatment (20DIV), expression of human V1V3 repeat containing transcript was significantly reduced (from 40% reduction with ASO1 to 80% with ASO3) in all treated conditions as compared to the non-targeting control or untreated condition (Fig. 1f).
From these in vitro experiments, we conclude that in various cell models (HEK 293 cells expressing a C9 intron1 reporter assay, C9-ALS/FTD patient-derived fibroblasts, and C9BAC 5 mouse-derived cortical neurons), ASOs 1-5 (Fig. 1b) all potently inhibit expression of V1 and V3 repeat containing transcripts while sparing V2, and all are efficiently taken up by neurons without the need for a transfection reagent.

ASOs selectively reduce the C9ORF72 repeat-containing transcripts and peptides after CNS infusion in C9BAC mice
We next evaluated the properties of these ASOs in vivo in wild-type (WT) and C9ORF72 transgenic mice. ASOs can be delivered to the brain tissue and spinal cord through the surrounding cerebral spinal fluid (CSF) via an intracerebroventricular (ICV) bolus injection or osmotic pump infusion 42 . We first assessed ASO tolerability in wild-type C57BL/6 mice. Each animal received a single ICV bolus dose of one of our five ASOs. Each treatment group consisted of 2-4 mice.
While these WT mice tolerated 30 nmol of MOE-modified ASOs 4-5, more than 5 nmol of LNAmodified ASOs 1-3 were lethal. Mice injected with ASOs 1, 2 and 4 had severe seizure-like phenotypes upon recovery from anesthesia or did not survive 24 hours post-injection, while mice treated with ASOs 3 and 5 (that share the same nucleotide sequence but differ by their sugar modification) remained alert and responsive to stimuli with no obvious detrimental effects up to one week after injection (Fig. 2a).
Encouraged by the safety profile of ASOs 3 and 5 in WT mice, we then compared their tolerability and efficacy in C9BAC transgenic mice via ICV administration. C9BAC transgenic mice generated in our laboratory express approximately 600 G4C2 repeat motifs within a truncated human C9ORF72 gene (from exons 1-6). Although these mice do not develop a motor phenotype, they fully recapitulate the distinct disease hallmarks including repeat containing RNA foci and DPR 42 and thus are a suitable C9-ALS/FTD mouse model for assessing efficacy of ASOs 3 and 5 activities in vivo. In our C9BAC mice, we were not able to safely perform intracerebroventricular infusions with more than 10 nmol of any LNA-modified compounds using bolus injections. To overcome this limitation, we next used osmotic pumps, which allowed infusion of a higher total amount of LNA modified ASO (ASO3) and permitted comparison of its potency with the MOEmodified version (ASO5). Doses ranging from 2.5 to 20 nmol per day of each ASO were continuously infused over 10 days into the right lateral ventricle of age-matched heterozygous C9BAC mice through a cannula using an implanted Alzet osmotic pump (Fig. 2b). Five to seven animals were used per group condition and animals infused with PBS or a non-targeting ASO were used as controls. Brains of animals sacrificed two weeks after the 10-day infusion demonstrated 6 widespread ASO distribution throughout the brain with associated neuronal uptake (Fig. 2c). The cortex and spinal regions of animals treated with ASO3 and ASO5 demonstrated potent, dosedependent reduction in V1 and V3 repeat-containing transcripts in both the cortex and spinal cord regions (Fig. 2d, f) as compared to PBS-infused animals; no such reduction was seen in the nontargeting control animals. Importantly, despite their impact on V1 and V3, neither ASO3 nor ASO5 produced any substantial reduction of the level of the V2 transcript (and hence the total C9ORF72 transcript variants) (Fig. 2e, g). Poly-GP DPR was also reduced in the cortex of mice treated with both ASO3 and 5 (Fig. 2h).
With chronic infusion, the LNA and MOE-modified ASOs produced no adverse behavioral side effects throughout the course of ASO administration; all animals remained healthy until they were sacrificed at 21 days. Routine clinical blood chemistry and liver and kidney morphology after H&E staining revealed no gross abnormalities (data not shown). Body weight monitoring from treatment onset to the time of sacrifice revealed a loss of up to 10% as compared to initial body weight before treatment; this was likely related to the pump implantation surgery as it also occurred in mice infused with PBS ( Fig. 2i). At the 100 nmol dose, mice receiving ASO3 showed more weight loss (18%). We therefore focused on ASO5 for further optimization.

Minimizing the PS content of ASO5 improves its safety profile without affecting its biological activity after a single CNS administration in C9BAC mice
When delivered by bolus injection, ASO5 distributed broadly throughout the mouse CNS within three weeks (Fig. S2). This finding is consistent with previous reports showing excellent ASO tissue distribution after bolus infusion (and improved pharmacodynamics activity in liver and brain tissues as compared to chronic administration) 42,43 . We found that mice injected with a 30 nmol dose of ASO5 survived and thrived. However, none survived a bolus injection of 50 nmol (Fig.   3a). This low MTD notably limits the dosing scheme and narrows the therapeutic window. To achieve a better balance safety-efficacy, we therefore sought to improve the efficacy and tolerability of ASO5.
ASO5 has a fully modified phosphorothioate (PS) backbone (Fig. 3b,c). Despite its advantages (improved resistance to nucleases and improved cellular uptake), the PS modification also shows increased protein binding that can lead to toxicity 44,45 . Because CSF has relatively low nuclease 7 activity 46,47 , we sought to reduce the number of PS linkages to generate a less toxic ASO without compromising activity in the nervous system. [48][49][50] Our first attempt consisted of introducing a PS linkage only every other nucleotide. Injection of 50 nmol of this "skipped PS" ASO was tolerated but still was not satisfactory, as mice showed abnormal behavior (slow movement, less reactive to stimuli) in the first 24 hours after treatment. We therefore reduced the PS content within the 5' and 3' MOE-modified "wings" of the ASO, maintaining PS modification throughout the DNA gap.
ASO5-1 and ASO5-2 differed only in the presence or absence of a PS linkage between nucleotides 5-6 and 13-14. Both modification patterns showed a higher MTD than ASO5 (Fig. 3a). To explore the mechanism of the active motor phenotypes seen here and to follow up on the improvement observed by mixed backbone patterns, we explored the effect of sugar and phosphate modifications on acute in vivo tolerability 51 . Both mixed-backbone ASOs maintained silencing of the repeatcontaining transcripts in C9-ALS/FTD fibroblasts (Fig. 3d, e), with no significant difference in potency between ASO5, ASO5-1 and ASO5-2.
We administered 30 nmol of each ASO via a bolus ICV injection in C9BAC mice and analyzed efficacy at 8 weeks (Fig. 3f). Each treatment was well tolerated with no adverse side effects or significant weight loss, or changes in blood chemistry ( Fig. 3g; Fig. S3b,i). Mice treated with 30nmol of ASO5 and ASO5-2 had a significantly reduced level of V1 and V3 transcripts in cortex and spinal cord as compared to the PBS treated group (Fig. 3h,j) with minimal effect on total transcripts (Fig. 3i,k) showing that absence of PS inter-nucleotide linkages between two MOE modified nucleotides did not impair biological activity in vivo. The silencing was specific and dose dependent (Fig. S3g). By contrast, ASO5-1, which lacked the PS linkage at the junction between a MOE modified and an unmodified DNA nucleotide, achieved only ~ 25% knock down of V1-V3 at a 30nmol dose (Fig. S3c,e). ASO5-2 treatment also reduced polyGP levels in cortex and spinal cord (Fig 3l,m). Based on its efficacy and tolerability profile, we selected ASO5-2 as our lead compound.

Sustained and potent effect of mixed-backbone ASO5-2 in the CNS of two C9BAC mouse models.
We further evaluated ASO5-2 in two ways. First, we obtained a full in vivo dose response of ASO5-2, which revealed an IC50 of 4.75nmol (Fig. 4a) in the cortex of C9BAC mice three weeks 8 after treatment. At the same timepoint, ASO5-2 also significantly reduced levels of polyGP in a dose-dependent manner (Fig. 4b).
We also evaluated the duration of effect of ASO5-2. Our initial studies showed that absence of PS linkages flanked by MOE nucleotides (as in ASO5-2) did not impair biological activity in vitro and in vivo three weeks after treatment. However, since PS linkages also protect ASOs from nuclease degradation, we wondered if the duration of effect of ASO5-2 would be comparable to its fully PS modified parent ASO5. To address this question, we injected C9BAC transgenic mice with 30nmol of ASO5-2 and analyzed the level of V1-V3 transcripts 24 h or 3, 8 or 20 weeks after injection (Fig. 4c). No effect was observed on the V1-V3 target RNA 24 hours after injection (Fig.   4d). However, a significant, specific dose-dependent reduction of ~80% of V1V3 transcripts but not total C9 transcripts was observed 3 weeks after injection and, importantly, this was sustained up to 20 weeks in the cortex (Fig. 4d). Analogously, we also observed a sustained reduction in the levels of the polyGP proteins, approaching 90% reduction even at 20 weeks after the single injection of 30 nmol ASO5-2 ( Fig. 4e).

No significant body weight loss or behavioral adverse events (as defined in the methods section)
were detected in animals treated with ASO5-2 or the PBS control (Fig. 4f). Likewise, no change in liver, kidney and spleen weight or morphology was observed (Fig. 4g,h). Finally, to further assay the tolerability of ASO5-2 treatment, we analyzed the coordinated motor functions of mice treated with ASO5-2 or injected with vehicle PBS. Seven mice per group were tested in a blinded manner on their rotarod performance weekly after treatment for 19 weeks. No motor deficit was observed in the treated group, underlining the tolerability of ASO treatment (Fig. 4i).
From these observations we concluded that ASO5-2 can be safely administered to transgenic C9BAC mice via intracerebroventricular delivery and that this durably suppresses the offending V1 and V3 transcripts but not the V2 transcript of C9ORF72, as well as the toxic polyGP dipeptides.

ASO5-2 is non-toxic in large animals
Encouraged by safety and efficacy data in the C9BAC mice, we next obtained additional preclinical toxicity profiles in large animals focusing on ASO5-2 as our lead ASO. As a pilot behavioral study, we performed an intrathecal injection of ASO5-2 at 2 mg/kg in four sheep. For intrathecal injection in sheep, it was necessary to thread a microcatheter up though the intrathecal 9 space and deliver the ASO directly into the cisterna magna (see Methods) 51 . Intracisternal contrast injection and cone beam computed tomography confirmed the correct catheter position prior to ASO injection. At one month, the sheep showed no neurological abnormalities.
We then purchased a batch of GMP-grade ASO5-2 (ChemGenes). The primary sequence of the GMP ASO was confirmed by mass spectrometry (Fig. S4).
We engaged an outside laboratory (Charles River Laboratories) to conduct a GLP safety study of ASO5-2 in cynomolgus monkeys. Twenty-eight normal monkeys averaging 2. were observed out to 90 days at either dose; the necropsies did not show pathological findings attributable to ASO5-2 (data not shown).  Table 1) but was otherwise benign. As summarized in Fig. 5, on August 29, 2019, he was treated with intrathecal ASO5-2 at 0.5 mg/kg, followed by 1.0 mg/kg two weeks later. On January 26, 2020, he received a third dose at 1.5 mg/kg, increased to 2.0 mg/kg on February 13; 2.0 mg/kg was subsequently administered four more times (Fig. 5). He experienced no medically or neurologically adverse effects from these interventions; his laboratory safety studies were unremarkable (Supplementary Table). In multiple CSF evaluations (Table 1), his CSF cell counts, protein and glucose levels were largely unremarkable. At the time of the 4 th and 5 th CSF analyses, he demonstrated very mild pleocytosis (but not in both CSF tubes on a given day). He also showed a progressive increase in CSF phosphorylated neurofilament heavy and neurofilament light chains (Table 1), which peaked after the 4 th 2.0 mg/kg dose of ASO5-2 and then partially subsided. The patient's CSF polyGP levels in two samples prior to dosing were in the range 0.01-0.03 ng/ml. After sequential doses of 2.0 mg/kg, the relative CSF polyGP levels dropped by approximately 80%, correlating with increasing CNS levels of ASO5-2 (Fig. 5, Table 1). During the months of treatment, the patient's ALS functional rating score (ALSFRS-R) was largely stable. Initially it was 38 (48 is normal), decreasing to a nadir of 33, but then improving to 38 concomitantly with the higher doses of ASO5-2.

DISCUSSION 803
Several years after the identification of the G4C2 repeat expansion in the gene C9ORF72 as the most common genetic cause of ALS and FTD, multiple investigations have defined potential disease mechanisms. Partial loss of function of C9ORF72 likely contributes to the neurotoxicity of hexanucleotide expansions but does not by itself recapitulate motor neuron degeneration 8,9,52 .
In several lines of mice, forced expression of mutant human C9ORF72 in mice also fails to reproduce frank motor neuron pathology but does replicate important molecular signatures, including intranuclear RNA foci and cytoplasmic polydipeptides. These are generated selectively by the two of the three C9ORF72 transcripts, V1 and V3, that harbor hexanucleotide expansions.
Together, these observations suggest that an effective therapeutic strategy will be to suppress expression of V1 and V3 to reduce levels of repeat containing RNA and polydipeptides while maintaining baseline levels of V2, which produces the full-length C9orf72 protein and thereby avoids haploinsufficiency. Here, we report that ASOs targeting the intron-repeat 5'junction selectively reduce C9ORF72 repeat containing transcripts V1 and V3 in a dose dependent manner after ICV administration in C9BAC transgenic mice, without reducing the overall level of C9ORF72 V2 transcripts. This intervention significantly reduces numbers of intranuclear RNA foci and depresses level of polyGP dipeptides, which are predicted to be produced from both sense 11 and anti-sense strands of the hexanucleotide repeat. Moreover, in a single patient this also reduced polyGP dipeptide levels by approximately 75%.
This report confirms earlier reports from in vitro studies that ASOs can suppress V1 and V3 transcripts in C9-ALS/FTD 36 and extends our understanding of this therapeutic approach in several important ways. Importantly, we have examined the impact of modifications of the backbone on both the potency and toxicity of the ASOs. Introduction of PS linkages only at the ASO termini and throughout the central DNA region, substantially increased the acute tolerability in the CNS without impairing its potent biological activity. In the course of these studies of ribose and backbone modifications, we compared the toxicities of ICV-bolus vs chronically administered ASOs. It appears that, for comparable doses, delivery by bolus is more effective than by pump. These observations prompted our study of repeated dosing of ASO5-2 via intrathecal delivery in a single patient. This was well tolerated as gauged by clinical examination and the safety profile in blood and CSF (Table 1). With repeated dosing of ASO5-2, we observed a progressive but transient elevation of two CSF biomarkers, phosphorylated neurofilament and neurofilament light chains. We ascribe this to some degree of nerve root irritation provoked by the intrathecal ASO or to some component of the disease not influenced by ASO5-2. This phenomenon is described in the literature in humans but fortunately has not be associated with major neurological consequences 50 . In our case, these biomarkers subsided somewhat in the later CSFs, as the interval between dosing was extended, suggesting a dose-response relation. The patient's clinical metrics improved across the course of therapy, despite the increment in the CSF neurofilament biomarkers. 12 Most importantly, our study provides proof-of-concept that ASO therapy in a human can effectively and safely suppress levels of the C9ORF72 transcript that harbor expansions (V1 and V3) without significantly affecting the predominant V2 transcript. That is, this intervention specifically targets not only the mutant allele but the miscreant transcripts generated by that allele.
In our models in vitro and in vivo, and in an individual with a C9ORF72 expansion, there was significant reduction in levels of polyGP dipeptide. Suppression of CSF levels of a C9ORF72 polydipeptide has not previously been reported in humans. There is controversy regarding some aspects of the polydipeptide pathology in C9ORF72 ALS. Early studies in drosophila suggested that toxicity arises not from the expanded RNA transcripts but from dipeptides, and particularly those containing arginine 19 . Other reports have also pointed to directly neurotoxic effects of these peptides 23,53-59 , although directly toxic effects of the expanded RNA transcript have also been recently described 60 . Our findings strongly encourage the view that suppressing the expression of the mutant alleles of C9ORF72 may be clinically beneficial, regardless of whether the primary benefit is mediated by reduction in mutant transcripts or polydipeptide levels. Larger clinical trials now underway to assess the clinical impact of therapies that silence C9ORF72 will likely illuminate this hypothesis.

Design of Mouse and Human Studies
The goal of this study was to develop an antisense oligonucleotide (ASO) as a potential clinical therapeutic candidate to treat C9-ALS/FTD. Our experimental approach combined C9 patient derived samples to evaluate selective efficacy in vitro and C9BAC transgenic mice to assess safety, efficacy and duration of effect in vivo. Per experiment, all mice were aged-matched and randomly assigned to control or experimental group. Samples and data were blindly collected, processed and quantified. Molecular and physiological readouts include expression level of V1V3 repeat containing transcripts, all transcripts and peptides, weight of body and organs and motor functions.
All outliers were included in data analysis. Exploratory experiments were performed on at least five mice per genotype. Sample size was calculated using the G-power analysis method based on previously defined effect size and standard deviation measuring variability within the sample. 13 The project described in this publication was supported by the University of Massachusetts CTSA award number UL1TR001453 from the National Center for Advancing Translational Sciences of the National Institutes of Health. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
The flow chart for the clinical study is presented in Figure 5. Digestion was halted by addition of 10% FBS/DMEM. Cells were triturated, resuspended in neurobasal media supplemented with Glutamax (ThermoScientific), 2% penicillin/streptomyocin and B27 supplement (ThermoScientific) and seeded at 0.5 × 10E6 cells/well in 6-well plates precoated with poly-ornithine (Sigma). Neurons were treated with ASO at the indicated dose five days after culture and collected 15 days after treatment.

C9ORF72 Bac Transgenic Mice
C9BAC mice were generated as previously described 34 and backcrossed to C57BL/6. All experimental protocols and procedures were approved by the University of Massachusetts Medical School Institutional Animal Care and Use Committee.

Stereotaxic Pump Implantation and Bolus Injection of ASO in the Mouse Brain.
For intracerebroventricular (ICV) infusion of ASO or PBS vehicle through a micro-osmotic pump (Alzet pump model 1007D attached to Alzet brain infusion kit 3), wild-type C57BL/6 or C9BAC transgenic mice were anesthetized and maintained on 2.5% isoflurane via a nose cone under a stereotaxic frame. Implantation procedure was performed as previously described 62 , with a 3mm cannula implantation 0.2mm posterior and 1.0mm lateral to the right of Bregma. 15 For ICV bolus injection mice were anesthetized with isoflurane and placed into a stereotaxic frame. 10µL of sterile PBS or ASO was injected into the right lateral ventricle using the following coordinates: 0.2 mm posterior and 1.0 mm lateral to the right from the Bregma and lowered to a depth of 3 mm.

Mouse Behavior Monitoring
Over the course of treatment and immediately prior to sacrifice, each animal was blindly weighed and evaluated weekly by a trained observer for adverse events, defined as any behavior not typical in a naïve matched control animal, including, but not limited to: limb clasping, tremors, abnormal respiration, paralysis, spasticity, impaired reflex, hyperactivity and lethargy.

Rotarod
Coordinated motor functions were assessed in control and treated mice using the rotarod test as previously described 34 . Briefly, mice were tested weekly beginning two weeks prior to ASO/vehicle administration and ending at the week of sacrifice. Each animal was given three trials on a 4-40rpm accelerating rotarod for 5 minutes with a one minute inter trial interval.
Latencies to fall for each animal was automatically recorded by a computer and plotted as a mean +/-SEM.

Blood Biochemistry
Whole blood samples were collected after cardiac puncture (terminal procedure). Blood biochemistry was performed using the VetScan Comprehensive diagnostic profile (Abaxis, Union City, CA).

Southern Blot
Southern blot was performed on 10 µg genomic DNA isolated using Gentra Puregene Tissue kit (Qiagen). DNA was digested overnight with AluI and DdeI at 37°C and separated by electrophoresis on a 0.6% agarose gel, transferred to a positively charged nylon membrane (Roche Applied Science), cross-linked by UV, and hybridized overnight at 55°C with a digoxigeninlabeled G2C4 DNA probe in hybridization mix buffer (EasyHyb, Roche). The digoxigenin-labeled probe was detected with anti-digoxigenin antibody and CDP-Star reagent as recommended by the manufacturer (Roche). 16 Total RNA was isolated from snap frozen cortex or spinal cord tissue using Trizol (ThermoScientific) and subsequently treated with DNase I (Qiagen). One µg of total RNA was reverse transcribed into cDNA using random hexamers and MultiScribe reverse transcriptase (ThermoScientific) following the manufacturer's instructions. Quantitative PCR was performed on a StepOnePlus Real-Time PCR system using SYBR Green Master Mix (Applied Biosystems) and 0.2 µM of forward and reverse primers as described in 54 . Ct values for each sample and gene were normalized to GAPDH. The 2exp (−ΔΔCt) method was used to determine the relative expression of each target gene.

Fluorescence in situ Hybridization (FISH)
FISH was performed as previously described 54  Response values corresponding to the intensity of emitted light upon electrochemical stimulation of the assay plate using the MSD QUICKPLEX SQ120 were acquired and background corrected using the average response from lysates obtained from wild-type C57Bl/6 brain extract.

Detection of pNFH and NFL
Neurofilament light chain (NFL) and phosphorylated neurofilament heavy chain (pNFH) were detected in CSF samples using sandwich immunoassays and the MSD QUICKPLEX SQ120. For pNFH measures, the Iron Horse Diagnostics clinically validated assay was used with a monoclonal anti-pNFH antibody as capture and sulfo-tagged polyclonal anti-pNFH antibody for detection.
Purified pNFH was used to generate a calibration curve. All assays were performed in triplicate with a coefficient of variation (CV) < 5% for all samples. The lower limits of quantification of the pNFH assay is 1.2 pg/ml in CSF. For NFL measures, a R-PLEX human NFL MSD assay kit was used (MSD catalog # F217X) following manufacturer specifications. All assays were performed in triplicate with a CV < 5% for all samples. The lower limits of quantification of the NFL assay is 2.0 pg/ml.

Immunohistochemistry
Brains were rapidly removed from euthanized animals. The contralateral hemibrain was post-fixed in 10% formalin. Paraffin-embedded or cryoprotected blocks were cut in 10µm thick sagittal sections. Slides were permeabilized with Triton 0.1% for 10 min. Non-specific antibody binding was blocked by incubation with 10% goat serum in PBS/Tween0.01% for one hour. Primary antibodies were diluted in blocking solution and sections were incubated overnight at 4°C. After three washes in PBS/Tween 0.01%, sections were incubated with Alexa fluor-488 or -546 conjugated secondary antibodies diluted in PBS for one hour at room temperature.
Autofluorescence was quenched by slide immersion in 0.5% Sudan BlackB in 70% ethanol and cell nuclei were stained with DAPI. Primary antibodies used: mouse anti-NeuN (1:500, Millipore), rabbit anti-P/S ASO (1:500, in house). Briefly, a rabbit polyclonal antibody raised in house by inoculating two female New Zealand White rabbits with a fully-PS-modified, KLH-conjugated ASO. Boosts and bleeds were carried out at regular intervals over one year, and antisera were used for histology (Moazami and Watts, in preparation).

Haematoxylin and eosin (HE) staining
8 to 10 µm thick sections of mouse liver and kidney were cut from formalin-fixed, paraffin embedded blocks. Standard HE staining was performed.
Mass spectrometric characterization of C9orf72 antisense oligonucleotide 18 The sequence and modification pattern of the clinical C9orf72 antisense oligonucleotide (ASO) was confirmed by mass spectrometric (MS) analysis performed by nanoflow electrospray ionization (i.e., nanospray) on a Fourier transform ion cyclotron resonance (FTICR) mass spectrometer (vide infra for conditions). The multiply charged ions observed in negative ion mode ( Figure S4a) provided a monoisotopic molecular mass of 6461.1894 u, which matched very closely the expected value of 6461.1888 Da calculated from sequence. In subsequent experiments, the [M -4H] 4molecular ion at 1615 m/z was submitted to tandem mass spectrometry (MS 2 ) to achieve sequence confirmation (vide infra). Upon isolation in the mass selective quadrupole (Q) and activation by collisions with Ar in the collision cell (q), the precursor ion exhibited typical fragment series corresponding to the central region of the construct, which confirmed the sequence spanning from A7 to C13 ( Figure S4b). Distinctive signals were additionally detected at 1179 and 1971 m/z, which were assigned to fragments corresponding to the entire G1:T6 and G14:C18 moieties, respectively. In turn, each of these first-generation fragments were individually isolated and activated in the FTICR cell to complete MS 3 determinations ( Figure S4c-d). These experiments provided abundant second-generation fragments that confirmed the sequences of the construct's terminal regions.
All analyses were carried out on a Bruker (Billerica, MA) 12T solariX FTICR equipped with a 12-tesla superconducting magnet and a home-built ion source that enabled static nanospray operations. No sample desalting or chromatographic steps were employed. All samples consisted of a 4 µM solution of C9orf72 antisense oligonucleotide in 150 mM ammonium acetate (pH 7.4) and 10% volume of 2-propanol. In each analysis, a 5 mL aliquot was loaded onto a quartz emitter prepared in-house. A stainless steel wire was inserted from the back end to provide the voltage necessary to achieve a stable spray. The instrument was calibrated by using a 1 mg/mL CsI solution that provided an accuracy of 87 ppb. Detection was accomplished in broadband mode, which afforded a typical 230,000 resolution. The first activation step was carried out in the collision cell (q) of the instrument, which was flooded with a low pressure of Ar and subjected to an 18.5 V activation voltage. The second activation step was carried out in the FTICR cell by irradiating selected ions with a frequency that was 250 Hz off-resonance and 0.65-1.29% power. The data were processed by using the DataAnalysis package provided by 19 Bruker. The results were interpreted by using the Mongo Oligo Mass Calculator v2.08 available at https://mods.rna.albany.edu/masspec/Mongo-Oligo.

Statistical Analysis
All data were graphed as mean ± SEM and analyzed using GraphPad Prism Software (Version7).
Tests between two groups used the two-tailed student-t test. Tests between multiple groups used one-way analysis of variance (ANOVA) corrected with Bonferroni multiple comparison post-hoc test. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001, ns not significant         G4C2 targeting LNA and MOE modi ed ASOs selectively reduces the C9ORF72 repeat containing transcripts and peptides after CNS infusion in C9BAC mice (a) ASO sequences and their in vivo response in 4-6-month-old wild-type mice after ICV bolus injection: bold green, LNA; bold blue, MOE; italic black, DNA. (b) Schematic of experimental design in heterozygous C9BAC mice. Vehicle control (PBS), a nontargeting control ASO (NTC) and 25 to 200nmol of ASO3 and ASO5 were infused into the right lateral ventricle of 5-6-monthold C9BAC mice over 10 days. Brain and spinal cord were harvested and dissected two weeks after pump removal for RNA and DPR analysis (spinal cord and ipsilateral brain hemisphere), and ASO staining (contralateral brain hemisphere). (c) ASO3 and ASO5 (green) are taken up in neurons (red) -nuclei counterstaining in blue. (d-g) Expression of V1-V3 repeat containing transcripts (d, f) and all transcripts (e, g) in cortex and spinal cord quanti ed by qRT-PCR in mice infused with PBS (dark grey), NTC ASO (light grey), ASO3 (green) or ASO5 (blue) at the indicated dose. For each dose level, n= 5-7, except NTC group (n=3). (h) Relative expression of polyGP in the cortex of mice treated with ASO3 (green) and ASO5 (blue) assayed by sandwich immunoassay. Data are represented as mean ± SEM. (i) % of body weight loss at end point relative to before treatment. Figure 3 Reducing the phosphorothioate content of ASO5 improves its safety pro le without reducing its biological activity after CNS administration in C9BAC mice (a) Chemistry and maximum tolerated dose