Intrathecal administration of a novel siRNA modality extends survival and improves motor function in the SOD1^G93A ALS mouse model

Abstract

Antisense oligonucleotides (ASOs) were the first modality to pioneer targeted gene knockdown in the treatment of amyotrophic lateral sclerosis (ALS) caused by mutant superoxide dismutase 1 (SOD1). RNA interference (RNAi) is another mechanism of gene silencing in which short interfering RNAs (siRNAs) effectively degrade complementary transcripts. However, delivery to extrahepatic tissues like the CNS has been a bottleneck in the clinical development of RNAi. Herein, we identify potent siRNA duplexes for the knockdown of human SOD1 in which medicinal chemistry and conjugation to an accessory oligonucleotide (ACO) enable activity in CNS tissues. Local delivery via intracerebroventricular or intrathecal injection into SOD1^G93A mice delayed disease progression and extended animal survival with superior efficacy compared with an ASO resembling tofersen in sequence and chemistry. Treatment also prevented disease-related declines in motor function, including improvements in animal mobility, muscle strength, and coordination. The ACO itself does not target any specific complementary nucleic acid sequence; rather, it imparts benefits conducive to bioavailability and delivery through its chemistry. The complete conjugate (i.e., siRNA-ACO) represents a novel modality for delivery of duplex RNA (e.g., siRNA) to the CNS that is currently being tested in the clinic for treatment of ALS.

Graphical abstract

Duan and colleagues reported that siRNA-ACO is a novel modality enabling RNAi in CNS tissue after ICV or IT injection. Targeted knockdown of SOD1 mRNA via siRNA-ACO in transgenic ALS (i.e., SOD1^G93A) mice provides superior efficacy compared to with an ASO identical to tofersen in sequence and chemistry.

Introduction

Amyotrophic lateral sclerosis (ALS) is a devastating neurodegenerative disease caused by the dysfunction and loss of motor neurons in the CNS. Disease progression is generally quick with a median survival at approximately 3–5 years after diagnosis.¹ Early symptoms typically include muscle cramps, twitching, weakness, and stiffness. Patients inevitably begin to experience problems with movement and speech, which eventually manifest into assisted breathing, paralysis, and inevitable death.

More than 50 genes have been linked to ALS, in which approximately 20% of all genetically defined cases are associated with mutation to the superoxide dismutase 1 (SOD1) gene.² While SOD1 normally functions to protect cells from reactive oxygen species by catalyzing superoxide anions, loss of its activity has been shown to be independent of neuronal cell death.^3,4 Rather, mutant SOD1 confers a toxic gain of function caused by intracellular aggregation of a misfolded protein that leads to motor neuron degradation.⁵ As such, reduction of SOD1 in CNS tissue has been the preferred therapeutic strategy for the treatment of ALS.⁶ For instance, tofersen (BIIB067) is an antisense oligonucleotide (ASO) with a gapmer architecture approved for clinical use that has shown therapeutic benefit in the treatment of ALS by suppressing mutant SOD1 mRNA levels via the RNase H mechanism.⁷

RNA interference (RNAi) is another mechanism of gene silencing that can provide potent knockdown of targeted mRNA transcripts.⁸ However, delivery to extrahepatic tissues (e.g., CNS) has limited its therapeutic development to broader indications. RNAi canonically requires double-stranded RNAs referred to as short interfering RNAs (siRNAs) to guide the RNA-induced silencing complex to complementary transcripts and suppress expression via mRNA cleavage.⁹ Disparate tolerances to chemical modification advantageous for ASO delivery has been one reason hindering drug development of RNAi for CNS indications.^10,11,12 Only recently have new technologies emerged demonstrating siRNA delivery to CNS tissues with drug-like properties.^13,14

We have developed another approach for delivering duplex RNA (e.g., siRNA) to the CNS by conjugating a single-stranded accessory oligonucleotide (ACO) to siRNAs, creating a novel modality termed siRNA-ACO, a platform technology called smart chemistry-aided delivery (SCAD). The ACO does not target any specific complementary nucleic acid sequence; rather, it imparts benefits conducive to bioavailability and delivery through its physiochemical composition (i.e., chemical modification and structure) typically incompatible with canonical siRNA duplexes. Herein, we identify potent siRNAs via high-throughput screening that knocked down human SOD1 in which medicinal chemistry and ACO conjugation enable durable activity for at least 8 weeks following injection into the cerebrospinal fluid (CSF) of SOD1^G93A mice. Both intracerebroventricular (ICV) and intrathecal (IT) administration delays disease progression and extends animal survival in a dose-dependent manner. Furthermore, siRNA-ACO treatment prevented loss of motor function, as well as demonstrated superior efficacy in comparison with an ASO compound identical to tofersen in both sequence and chemistry.

Results

Development of siRNA drug candidates for knockdown of human SOD1

Sequence comprising the open reading frame (ORF) of human SOD1 transcript served as the template for siRNA design. A total of 268 siRNA duplexes were designed and synthesized at 21 nucleotides in length with no more than four repetitive nucleotides in a row and GC content between 35% and 65%. Knockdown activity of each siRNA was assessed in 293A cells using high-throughput RT-qPCR at both 0.1 and 10 nM concentrations. Data were ranked according to the mean knockdown activity in which 121 siRNAs reduced SOD1 by more than 90% at 10 nM (Figure 1A). As an indicator of potency, 69 and 15 siRNAs reduced SOD1 levels at 0.1 nM treatments by more than 50% and 75%, respectively. The overall top 25 performing siRNAs were subjected to an additional round of screening at 6 concentrations (i.e., 0.0064–20 nM) in 293A cells to demonstrate dose dependent activity (Figure S1). Propidium iodide (PI) was also integrated into sample collection as an indicator of untoward cytotoxicity. Shown in Figure 1B are data for only the top 5 siRNAs (i.e., siSOD1-063, 047, 104, 005, and 258) with the most potent knockdown activity in absence of overt cytotoxicity (i.e., <20% reduction in PI staining).

Figure 1.

siRNA screen for SOD1 knockdown in vitro

(A) 293A cells were transfected with each siRNA duplex (268 in total) in duplicate at 10 or 0.1 nM concentrations for 24 h. SOD1 expression levels were quantified via RT-qPCR using gene specific primer sets. TBP was amplified as an internal reference used to normalize expression data. Shown are the expression values of SOD1 mRNA of each experimental replicate relative to Mock treatments (dotted line). Mock samples were transfected in absence oligonucleotide. Individual siRNAs are labeled on the x axis. (B) Knockdown activity and cell viability of the top performing siRNAs was quantified in 293A cells at six escalating concentrations (i.e., 0.0064, 0.032, 0.16, 0.8, 4, and 20 nM) via RT-qPCR and PI staining, respectively. Shown are the results for both SOD1 knockdown and cytotoxicity of the top five performing siRNAs (i.e., siSOD1-063, 047, 104, 005, and 258) relative to mock treatments (dotted line). (C) Dose-response curves were generated in SK-N-AS cells for each of the top 5 siRNAs at eight treatment concentrations (i.e., 0.00006, 0.0002, 0.001, 0.004, 0.016, 0.063, 0.25, and 1 nM) via RT-qPCR. Data represent mean ± SD from two experimental replicates. (D) SK-N-AS cells were transfected at 0.1 nM with four different chemically modified variants of each siRNA (i.e., M1, M2, M3, or M4) or non-specific siRNA control (siCon) for 24 h. Knockdown activity was assessed via RT-qPCR relative to mock treatment. (E) Dose-response curves were generated in SK-N-AS cells for each of the M3 modified variants (i.e., siSOD1-063M3, 047M3, 104M3, 005M3, and 258M3) via RT-qPCR. Data represent mean ± SD from three experimental replicates.

Dose-response curves were subsequently generated for each of the top 5 siRNA candidates to validate potency in model cell lines representative of neuronal disease, including SK-N-AS (Figure 1C) and T98G (Figure S2) cells. As summarized in Table S1, the in vitro potency of each duplex was in the low picomolar range for both cell lines. Untoward cytotoxicity was also evaluated 72 h after treatment at concentrations well above 200× the extrapolated median inhibition concentration (IC₅₀) values. As shown in Figures S3A and S3B, only siSOD1-047 and 005 had no detectable impact on apoptosis or cell number in either SK-N-AS or T98G cells, whereas the remaining candidates (i.e., siSOD1-063, 104, and 258) had dose-dependent responses with regard to caspase 3/7 activity in T98G cells that inversely correlated with cell viability. While SK-N-AS cells seemed to be more tolerant to treatment, a similar pattern was observed for cell viability.

Several medicinal chemistry patterns referred to as M1, M2, M3, or M4 representing different duplex lengths (i.e., 20, 21, 22, and 23 bp, respectively) composed of phosphorothioate (PS) backbone modifications at select positions with 2′-O-methylation (2′Ome) or 2′-fluoro (2′F) substitutions at every nucleotide were applied to each lead candidate and screened for target mRNA knockdown activity. As shown in Figure 1D, all M3 variants (i.e., siSOD1-063M3, 047M3, 104M3, 005M3, and 258M3) generally had better knockdown activity compared with the other chemically modified siRNAs at 0.1-nM treatment concentrations in SK-N-AS cells. A near identical pattern was also observed in T98G cells (Figure S4). To further characterize potency, dose-response curves were generated for all M3-modified duplexes in both SK-N-AS (Figure 1E) and T98G (Figure S5) cell lines. As summarized in Table S1, the in vitro potency was generally well retained after chemical modification.

ACO conjugation was developed to impart self-delivery properties similar to ASOs by sharing medicinal chemistry normally not tolerated by canonical siRNAs. As such, the passenger strand of each M3 variant (i.e., siSOD1-063M3, 047M3, 104M3, 005M3, and 258M3) was synthesized covalently linked to a 14-nucleotide ACO (referred to AC1) via a triethylene glycol (TEG) linker (L9) that possessed PS backbone substitutions and 2′-O-methoxyethyl (2′MOE) modifications at every position within the ACO (Figure 2A). Gel shift assays demonstrate that AC1 conjugation promotes novel interactions with factors in human serum compared with only chemically modified duplexes (Figure S6). AC1 was intentionally designed to lack significant homology to any known transcript using chemistry typically incompatible with ASO-mediated RNase H activity. As shown in Figure 2B, AC1 treatment alone had no detectable knockdown of SOD1 at 0.25- and 2.5-nM concentrations in vitro, whereas activity was only perceived when conjugated to siRNA. Additional dose response analysis noted an approximately 10× loss in siRNA potency as a consequence of AC1 conjugation (e.g., siSOD1-005M3-AC1) compared with only chemically modified duplex (e.g., siSOD1-005M3) (Figure 2C). However, potency was restored upon modification with 5′-(E)-vinylphosphonate (5′VP) at the 5′ terminus of the guide strand (e.g., siSOD1-005M3-AC1^VP) (Figure 2D). In comparison with an ASO identical to tofersen in both sequence and chemistry (i.e., ASO^SOD1), SOD1 knockdown with either siRNA-ACO variant (i.e., siSOD1-005M3-AC1 or siSOD1-005M3-AC1^VP) was still more potent than ASO^SOD1, regardless of 5′VP modification (Figure 2D).

Figure 2.

siRNA-ACO knockdown activity in vitro

(A) Depicted is a visual representation of the siRNA-ACO structure in which a 14-nt ACO (referred to AC1) was conjugated to the 3′-terminus of the passenger strand via TEG linker (L9). (B) 293A and T98G cells were transfected at 0.25 or 2.5 nM concentrations with exemplary siRNA-ACO (i.e., siSOD1-005M3-AC1) or AC1 only for 24 h. Mock samples were transfected in absence oligonucleotide. Treatment with siCon served as a negative control for knockdown activity. SOD1 expression levels were quantified via RT-qPCR using gene specific primer sets. TBP was amplified as an internal reference used to normalize expression data. Shown are the mean expression values ± SD of SOD1 mRNA for each experiment relative to Mock treatments (dotted line). (C) Dose-response curves were generated in T98G cells comparing siRNA knockdown activity with (siSOD1-005M3-AC1) or without (siSOD1-005M3) AC1 conjugation at 10 treatment concentrations (i.e., 0.0003, 0.0009, 0.0027, 0.0082, 0.024, 0.074, 0.22, 0.67, 2, and 6 nM) via RT-qPCR. (D) Dose-response curves were generated comparing knockdown activity of siRNA-ACOs with (siSOD1-005M3-AC1^VP) or without (siSOD1-005M3-AC1) 5′VP modification to ASO^SOD1.

siRNA-ACO conjugates were subsequently synthesized using M3 chemistry and 5′VP modification (i.e., siSOD1-063M3-AC1^VP, 047M3-AC1^VP, 104M3-AC1^VP, 005M3-AC1^VP, and 258M3-AC1^VP) for downstream screening in vivo. Prior to treatment, dose-response curves were generated in both SK-N-AS and T98G cells to validate activity of each siRNA-ACO candidate (Figures S7A and S7B). As summarized in Table S1, in vitro potencies were generally well conserved in comparison with their non-conjugate forms. Apoptosis and cell viability were also quantified 72 h after treatments to measure any changes in cytotoxicity as a consequence of chemical modification and ACO conjugation. As shown in Figures S8A and S8B, siSOD1-047M3-AC1^VP and 005M3-AC1^VP remained generally unaffected, with nominal impact on cell health in both SK-N-AS and T98G cells, whereas siSOD1-104M3-AC1^VP retained signs of untoward cytotoxicity and the remaining candidates (i.e., siSOD1-063M3-AC1^VP and 258M3-AC1^VP) noted an improvement in in vitro safety (i.e., reduction in caspase 3/7 activity) compared with their non-modified forms (Figures S3A and S3B).

In vivo selection of siRNA-ACO drug candidates

Mice hemizygous for SOD1^G93A transgene express a mutant form of human SOD1 and exhibit disease phenotypes similar to ALS, including progressive loss of motor function and abbreviated life span via neuronal degradation.^15,16 To test knockdown activity in vivo, SOD1^G93A mice were treated via ICV injection at a 10 nM/dose of each siRNA-ACO conjugate (i.e., siSOD1-063M3-AC1^VP, 047M3-AC1^VP, 104M3-AC1^VP, 005M3-AC1^VP, and 258M3-AC1^VP). All siRNA-ACOs were formulated in artificial CSF (aCSF) in which treatment alone served as a vehicle control to establish baseline expression, while siCON1-AC1^VP functioned as a negative control for siRNA-ACO activity. Mice were also treated with ASO^SOD1 at approximately 3× molar excess of siRNA-ACO (i.e., 28 nM/dose) as a comparative control for human SOD1 (hSOD1) knockdown. Tissues from the CNS (e.g., frontal cortex, cerebellum, cerebrum, and spinal cord) and periphery (e.g., liver) were harvested on day 14 after treatment in which hSOD1 expression levels were quantified via RT-qPCR. As shown in Figure 3A, both siSOD1-047M3-AC1^VP and siSOD1-005M3-AC1^VP had more potent knockdown activity in the CNS tissues compared with the other siRNA-ACO candidates. Activity was well retained within the CNS in which drainage to peripheral tissue (e.g., liver) did not produce comparable activity. Knockdown via ASO^SOD1 produced results similar to siSOD1-258M3-AC1^VP, yet substantially less than the lead candidates (i.e., siSOD1-047M3-AC1^VP and 005M3-AC1^VP).

Figure 3.

siRNA-ACO activity in CNS tissue of SOD1^G93A mice

(A) Adult SOD1^G93A mice were treated via ICV injection with each siRNA-ACO drug candidate (i.e., siSOD1-063M3-AC1^VP, 047M3-AC1^VP, 104M3-AC1^VP, 005M3-AC1^VP, and 258M3-AC1^VP) at 10 nM. Animals administered ASO^SOD1 were dosed at 28 nM. Treatment with aCSF alone was used as a vehicle control to establish baseline expression, while a non-specific siRNA-ACO (i.e., siCON1-AC1^VP) served as a negative control for knockdown activity. hSOD1 expression was quantified via RT-qPCR using gene specific primer sets in tissues from the CNS (i.e., frontal cortex, cerebellum, cerebrum, and spinal cord) and periphery (e.g., liver) on day 14 after treatment. All tissues of the cerebral cortex excluding sample taken from the frontal cortex are represented in data labeled cerebrum. Mouse Tbp (mTbp) was amplified as an internal reference to normalize expression data. Shown are the mean expression values ± SD (n = 3–4 mice/group) of hSOD1 relative to aCSF treatment. The dotted gray line represents 80% knockdown relative to baseline (dashed line). (B) Knockdown activity was quantified in tissues from the brain (i.e., frontal cortex, cerebellum, and cerebrum), spinal cord (i.e., cervical, thoracic, and lumbar), and periphery (i.e., liver) via RT-qPCR on day 14 following treatment via ICV injection at a fixed molecular dose (20 nM) with siRNA-ACO candidates (i.e., siSOD1-047M3-AC1^VP and 005M3-AC1^VP) or their non-conjugated derivatives (i.e., siSOD1-047M3^VP and 005M3^VP). Treatment with siCON2-AC1^VP served as a negative control for knockdown activity. Shown are the mean expression values ± SD (n = 2–6 mice/group) of hSOD1 relative to aCSF treatment. (C) Knockdown durability was assessed via RT-qPCR in the indicated CNS tissues at 2 and 8 weeks after a single 20 mg/kg dose of siSOD1-005M3-AC1 (1.0 μmol/kg), siSOD1-005M3-AC1^VP (1.0 μmol/kg), or ASO^SOD1 (2.8 μmol/kg) via ICV injection. Treatment with aCSF alone served as a vehicle control. Data represent mean ± SD (n = 3–4 mice/group) relative to SOD1 levels pre-treatment (i.e., 0 weeks).

SOD1^G93A mice were also treated via ICV injection at equimolar quantities (i.e., 10 nM/dose) with siRNA-ACOs (i.e., siSOD1-047M3-AC1^VP or siSOD1-005M3-AC1^VP) in comparison with non-conjugate controls (i.e., siSOD1-047M3^VP or siSOD1-005M3^VP) to demonstrate the effect AC1 conjugation imparts on knockdown activity in vivo. As shown in Figure 3B, both siRNA-ACO duplexes provided greater knockdown activity compared with their non-conjugate cognates in all tissues of the brain (i.e., frontal cortex, cerebellum, and cerebrum) and spinal cord (i.e., cervical, thoracic, and lumbar spine) with minimal activity in the periphery (e.g., liver). Knockdown durability was also characterized for siSOD1-005M3-AC1^VP relative to its non-5′VP control (i.e., siSOD1-005M3-AC1) in comparison with ASO^SOD1. As shown in Figure 3C, knockdown via siSOD1-005M3-AC1^VP was generally well sustained out to 8 weeks in all CNS tissues, whereas activity of siSOD1-005M3-AC1 and ASO^SOD1 began to return to baseline. Taken together, the combination of both AC1 conjugation and 5′VP modification provided the siRNAs with enhanced activity needed for a durable response in vivo.

Single-dose ICV injection of siRNA-ACO delays disease progression and prolongs survival in SOD1^G93A mice

Adult SOD1^G93A mice were treated via ICV injection with lead siRNA-ACO candidates (i.e., siSOD1-047M3-AC1^VP or siSOD1-005M3-AC1^VP) at 50, 100, 200, or 400 μg/dose on post-natal day (PND) 85 or 60, respectively. Drug concentrations and hSOD1 expression levels were quantified in a subset of animals on day 14 after treatment in CNS tissue (e.g., cerebellum, cerebrum, and spinal cord). As shown in Figures 4A and 4B, both activity and tissue accumulation of siSOD1-047M3-AC1^VP and siSOD1-005M3-AC1^VP were dose dependent, in which SOD1 knockdown inversely correlated with increasing concentrations of siRNA-ACO within the CNS tissues. Drug concentrations projected to elicit a median effective dose (ED₅₀) response in each tissue are summarized in Table S2.

Figure 4.

Dose-dependent relationship between knockdown activity and tissue accumulation of siRNA-ACO

Adult hSOD1^G93A mice were treated with siSOD1-047M3-AC1^VP (A) or siSOD1-005M3-AC1^VP (B) at the indicated doses (i.e., 50, 100, 200, or 400 μg) via ICV injection. Knockdown activity of hSOD1 was quantified via RT-qPCR in select CNS tissues (i.e., cerebellum, cerebrum, and spinal cord) on day 14 following treatment. All tissues of the cerebral cortex are represented in data labeled cerebrum. Animals given aCSF alone represent expression levels at baseline and drug quantities detectable in absence of siRNA-ACO treatment (0 μg). Knockdown activity is shown as percent (%) inhibition of SOD1 relative to baseline (0 μg). Drug concentrations are shown as siRNA-ACO quantities relative to tissue sample mass (μg/g). Data represent mean ± SD (n = 3–4 mice/group).

In the remaining animals, changes in body weight were plotted to monitor growth rate and disease progression. As shown in Figure 5A, all groups treated with either siSOD1-047M3-AC1^VP or siSOD1-005M3-AC1^VP continued to gain weight in comparison with aCSF treatment. In addition, disease-related weight loss was delayed in a dose-dependent manner as noted by time needed for growth rates to return to starting weight (dotted line). Disease progression was confirmed in each animal when a 10% loss in peak body weight was recorded. Plotting data via Kaplan-Meier curves indicated when animals in each treatment group transitioned to progressive disease (Figure 5B). Data were collected until animals inevitably succumbed to their disease in which survival curves were also generated (Figure 5C). To summarize, both siSOD1-047M3-AC1^VP (Table S3) and siSOD1-005M3-AC1^VP (Table S4) treatment delayed disease progression and extended animal survival in which the highest dose (i.e., 400 μg) prolonged life by 70 and 111.5 days compared with vehicle controls, respectively.

Figure 5.

Single dose treatment via ICV injection with siRNA-ACO delays disease progression and prolongs survival

(A) Adult hSOD1^G93A mice were treated with siSOD1-047M3-AC1^VP or siSOD1-005M3-AC1^VP at the indicated doses (i.e., 50, 100, 200, or 400 μg) via ICV injection on PND 85 or PND 60, respectively. Growth rates (i.e., percent change in body weight) relative to first day of treatment (dotted line) were plotted to monitor disease progression. Data are plotted as the mean ± SD. (B) Data are plotted as percent animals in each treatment group at peak body weight. (C) Animal survival is plotted as percentage of surviving animals in each treatment group. Animal numbers (n) are indicated in each graph.

Pathogenic mutation in target sites of siRNA-ACO impact lead selection

In silico analysis of pathogenic single-nucleotide polymorphisms (SNPs) located within the target site of siSOD1-005M3-AC1^VP revealed five total SNPs in which four were in the region complementary to its “seed” sequence (Figure S9A). Mismatches in this region are known to inhibit siRNA activity, which could eliminate a projected approximately 8.9%–12.4% of global ALS^SOD1 patients from its treatment pool.^17,18 Conversely, siSOD1-047M3-AC1^VP has only two reported pathogenic SNPs within its target site (i.e., P.E22G and P.F21C) comprising approximately 4.04% and 2.70% of the global ALS^SOD1 population, respectively (Figure S9B). Luciferase reporter constructs (i.e., pLuc^SOD1, pLuc^P.E22G, and pLuc^P.F21C) containing either consensus sequence perfectly complementary to siSOD1-047M3-AC1^VP guide strand or one of pathogenic mutations (i.e., P.E22G and P.F21C) were co-transfected into 239A cells along with siRNA-ACO. Knockdown of luciferase activity was specific to siSOD1-047M3-AC1^VP, as transfection with a scramble control (i.e., siCON2-AC1^VP) did not reduce reporter expression (Figure S9C). Dose-response data indicated that the P.E22G mutation does not have any significant impact on siSOD1-047M3-AC1^VP knockdown activity/potency compared with consensus target site sequence, while P.F21C was partially resilient to treatment, demonstrating incomplete knockdown and inferior potency (Figure S9D).

IT injection of siRNA-ACO provides therapeutic efficacy against ALS in SOD1^G93A mice

Based on perceived patient populations, siSOD1-047M3-AC1^VP was selected for further in vivo analysis. Male and female SOD1^G93A mice were treated via IT injection with two sequential doses of siSOD1-047M3-AC1^VP on PNDs 68 and 100 at 75, 150, or 300 μg/dose. A non-specific siRNA-ACO (i.e., siCON3-AC1^VP) served as a negative control for therapeutic efficacy. Animal weight was monitored with comparison with wild-type animals. All siSOD1-047M3-AC1^VP treatments provided a similar benefit to male mice, whereas weight gain in females seemed to be more noticeably dose dependent (Figure 6A). Both disease progression (i.e., 10% loss in peak body weight) and animal survival were also plotted in which siSOD1-047M3-AC1^VP treatments delayed advance disease and extended survival in both male and female mice (Figures 6B and 6C). Table S5 summarizes the median days for disease progression and survival after siRNA-ACO treatment via IT injection in male and female populations. Overall, siCON3-AC1^VP provided no therapeutic benefit in comparison with vehicle control, while equivalent doses of siSOD1-047M3-AC1^VP (i.e., 150 μg) extended animal survival in males and females by 61 and 29 days, respectively.

Figure 6.

siRNA-ACO delays disease progression and prolongs survival via IT injection

(A) Male and female adult hSOD1^G93A mice were treated twice with siSOD1-047M3-AC1^VP at the indicated doses (i.e., 75, 150, or 300 μg) via IT injection on PND68 and PND100. Non-specific control (i.e., siCON3-AC1^VP) and ASO^SOD1 were dosed at 150 μg/injection. Treatment with aCSF served as vehicle control. Body weight in grams was plotted compared with background animals (WT) to monitor disease progression. Data are plotted as the mean ± SD. (B) Data are plotted as percent animals in each treatment group at peak body weight. (C) Animal survival is plotted as percentage of surviving animals in each treatment group. Animal numbers (n) are indicated in each graph.

Neuromuscular performance was also evaluated in groups composed of both male and female mice. Distance traveled via open field roaming (Figure 7A), grip strength (Figure 7B), and rotarod test (Figure 7C) all showed siSOD1-047M3-AC1^VP treatment greatly improved motor function in SOD1^G93A mice, which was generally well sustained until the end of study. Comparing rotarod performance of each individual animal at an early time point prior to measurable weight loss (i.e., PND 90) with their respective terminal time point further indicates siSOD1-047M3-AC1^VP treatment retained or improved neuromuscular performance in a majority of animals compared with controls or ASO^SOD1 (Figures S10 and S11).

Figure 7.

siRNA-ACO treatment improves motor function in SOD1^G93A mice

Male and female hSOD1^G93A mice treated twice with the indicated test articles were subject to open field tests, total distance traveled and rotarod test. Roughly 50% of mice of each gender were assigned to open filed and traveled distance test, and another 50% to rotarod test. (A) Open field tests during the daytime. Total distance each animal moved was autonomously recorded in centimeters (cm) over the course of 15 min. Data are plotted as the mean ± SD distance traveled for each treatment group. (B) Grip strength was assessed in triplicate and the average value was recorded for each animal. Data is plotted in grams (g) as mean ± SD grip strength for each treatment group. (C) Animal fatigue and coordination was assessed by rotarod test for 5 min. Experiments were performed in triplicate in which the longest latency time to fall was recorded in seconds (s) for each animal. (D) Motor function was scored using the ALS TDI NS scale for all animals prior to open field, rotarod, and/or grip strength tests. Mean NS ± SD is shown at each treatment group at the indicated time points.

Motor function was also scored for all animals prior to open field, rotarod, and/or grip strength tests using the ALS Therapy Development Institute (TDI) neuroscore (NS) system. As shown in Figure 7D, all mice developed abnormal splay (i.e., NS2) by approximately PND 130 in both aCSF and siCON3-AC1^VP control groups that continued to increase in severity over time (e.g., ≥NS3). Conversely, mean scores for all siSOD1-047M3-AC1^VP doses never surpassed NS2 at any time point during the course of the study, particularly for the highest dose group (i.e., 300 μg), in which the mean NS remained predominantly flat at around NS1. ASO^SOD1 treatment also demonstrated an improved NS compared with vehicle and siCON3-AC1^VP controls. However, by PND 150, NS began to increase at a slope similar to controls despite ASO^SOD1 having approximately 3× molecular excess than that of siSOD1-047M3-AC1^VP at the 150-μg dose.

Discussion

The first clinical study targeting SOD1 for the treatment of ALS was an ASO called ISIS 333611.¹⁹ Due to limitations in its potency, results were presented as proof of concept for clinical development of ASO delivery via IT injection. It was not until the advent of tofersen (BIIB067) that an ASO with improved knockdown activity returned to the clinic for testing in ALS patients.⁶ In 2023, tofersen was granted accelerated approval for clinical use, a decision based on the findings of the an efficacy, safety, tolerability, pharmacokinetics and pharmacodynamics study of BIIB067 (Tofersen) in adults with inherited ALS with confimred SOD1 mutation (VALOR) trial. This pivotal trial demonstrated tofersen’s ability to reduce plasma neurofilament light, a blood-based biomarker indicative of axonal damage and neurodegeneration. Additionally, a modest decrease in SOD1 levels in the CSF of ALS patients was observed, suggesting that an enhanced knockdown might improve therapeutic outcomes. While the VALOR trial confirmed tofersen’s effectiveness in lowering these biomarkers, it did not show a significant improvement in clinical endpoints within the 28-week study period. However, the trial provided crucial insights, particularly that patients who started tofersen treatment early experienced significantly slower disease progression and enhanced quality of life compared with those with delayed treatment initiation, highlighting the importance of timely intervention.²⁰

RNAi remains the preferred tool for gene silencing, in part, by being a more potent modality for many targets.^8,21 However, development for indications like ALS have been historically hampered by poor delivery and limited biodistribution of siRNA in target tissues. Only recently have new platforms emerged enabling delivery to the CNS, including the SCAD technology disclosed within this study. By screening siRNAs and implementing chemical modification, we identified siRNA drug candidates with low picomolar potency for knockdown of SOD1 in model cell lines. Conjugation to AC1 enabled broad knockdown activity in animal tissues throughout the brain and spinal cord, while maintaining potency comparatively better than a second-generation ASO identical to tofersen in sequence and chemistry (i.e., ASO^SOD1). Further modification with 5′VP provided enhanced knockdown durability and demonstrated therapeutic efficacy (i.e., delayed disease progression, prolonged survival, and improved motor function) in SOD1^G93A mice at molar equivalent doses approximately 3× lower than ASO^SOD1.

An added benefit of siRNA-ACO is the simplicity of its manufacturing. Synthesis including conjugation is compatible with classic oligonucleotide solid-phase chemistry in which the entire process is performed on CPG support. Even the TEG linker is sourced as a spacer 9 phosphoramidite (CAS# 146668-73-7). Other than 5′VP, the disclosed siRNA-ACO sequences were composed entirely of standard amidites (e.g., PS, 2′Ome, 2′F, and 2′MOE) with known pre-clinical and clinical safety data. In addition, good manufacturing practice-compliant manufacturing of siRNA-ACO to scale has already been validated via contract manufacturing.

Although ACO conjugation was developed to impart self-delivery properties similar to ASOs, the precise mechanism by which it facilitates siRNA delivery has not been fully elucidated. In vitro assays with human serum revealed a supershift in gel migration only for the siRNA-ACOs indicative of protein binding. This is reminiscent of the protein interactions facilitated by PS chemistry that have been reported to influence aspects of ASO performance in vivo, including biodistribution and cellular uptake.²² Since the ACO component within the siRNA-ACO conjugate is chemically analogous to an ASO, the mechanisms for delivery are likely similar to each other in some respects.

Introducing features indicative of ASOs to help improve cellular uptake of siRNAs in vivo, including chemical modification (e.g., increased PS content) and/or structural asymmetries (e.g., exaggerated single-stranded tails), have been explored before (e.g., sd-rxRNA).²³ PS-driven cellular uptake and tissue retention are also supported by divalent siRNA architecture carrying approximately 40% PS in each strand resulting in a cooperative interaction between two partially PS-modified siRNAs. This concept was validated by elimination of PS content completely abolished efficacy.¹³ However, this is, to the best of our knowledge, the first example in which conjugation of a short non-targeting oligonucleotide onto a conventional siRNA scaffold enables robust activity in the CNS. In comparison, ACO composition is readily amendable to changes in sequence, length, and/or chemistries for further optimization as canonical ASO function via complementarity to a cognate sequence is irrelevant for siRNA-ACO knockdown activity.

Our current lead candidate (i.e., siSOD1-047M3-AC1^VP) represents our first iteration of the siRNA-ACO technology for RNAi delivery to the CNS. With future studies focused on safety pharmacology and toxicology, we aim to clinically validate this delivery platform using ALS as an exemplary indication. However, the prevalence of pathogenic mutations located within the siRNA target site of the SOD1 transcript must be considered for estimating actual patient population size. For instance, of the two pathogenic SNPs located in the target sequence of siSOD1-047M3-AC1^VP, only P.F21C was partially resilient to knockdown. P.F21C has been linked to patients of Chinese and Korean descent who present with familial ALS and comprise approximately 2.7% of the total ALS^SOD1 population.^18,24,25,26 As such, P.F21C may serve as a genetic marker for drug efficacy and/or patient exclusion in clinical study design.

Our study corroborates established gender disparities in ALS progression and treatment response. As observed in Figure 6, female mice displayed delayed disease onset and extended survival duration within treatment groups, consistent with the published literature. Prior research indicates that female SOD1G93A mice exhibit slower disease progression compared with males, characterized by later symptom onset, prolonged survival, and better response to treatments.^27,28,29 This difference may be attributed to sex-specific mechanisms, including mitochondrial activity and sex hormones.^30,31,32 By elucidating these sex-specific factors, future research can pave the way for personalized treatment strategies tailored to individual patients.

In summary, this study demonstrates the potential of siRNA-ACO, a novel platform technology for delivering siRNA to the CNS for the treatment of ALS. The lead candidate, siSOD1-047M3-AC1^VP, displayed potent and durable knockdown of mutant SOD1 in brain and spinal cord tissues, resulting in delayed disease progression, extended survival, and improved motor function in SOD1^G93A mice. While siRNA-ACO shares features like ease of manufacturing and readily adaptable chemical composition with ASOs, its superior potency in silencing target genes offers a significant advantage. Further optimization and safety assessments in humans are necessary, but siRNA-ACO holds significant promise as a transformative therapeutic strategy for ALS and potentially other neurodegenerative diseases.

Materials and methods

High-throughput screening of siRNAs targeting hSOD1

ORF from hSOD1 cDNA sequence (GenBank: NM_000454.5) served as the template for siRNA design via an in-house algorithm. A total of 268 duplexes were synthesized at 19-nucleotides in length without medicinal chemistry (Table S6). Plating and transfection of 293A cells was performed in 96-well plates, each containing 32 siRNAs and 8 quality control treatments at 2 concentrations (i.e., 0.1 and 10 nM) in duplicate. Cells were cultured for 24 h and lysis was automated via the Fluent System 780 liquid handling system (Tecan) using an optimized formula containing PI based on the Cell Lysis (CL) buffer for one-step RT-qPCR as previous described.²⁶ Integration of PI in sample preparation served to monitor variation in cell number (e.g., untoward cytotoxicity) by staining total nucleic acid content in crude lysates. Staining was quantified via optical density (OD) at 535 nm excitation and 615 nm emission wavelengths on the Infinite M200 Pro microplate reader (Tecan). Samples were subsequently transferred to 384-well plates for analysis by RT-qPCR on the 480 Real-Time PCR system (Roche) using the One-Step TB Green PrimeScript RT-PCR Kit II (Takara). Preparation of PCR reactions was automated by the Echo 525 Acoustic Liquid Handler (Beckman Coulter). A secondary screen was subsequently performed only on the top 30 performing siRNAs at 6 concentrations (i.e., 0.0064–20 nM). All samples were amplified in triplicate.

siRNA synthesis

Oligonucleotide sequences were synthesized in-house at Ractigen Therapeutics on solid-phase support using an HJ-12 synthesizer (Highgene-Tech Automation) and subsequently purified via RP-HPLC using an acetonitrile gradient over an UniPS column (NanoMicro Technology). Each sequence was reconstituted in sterile water via buffer exchange. Equal molar quantities of each strand were annealed into their corresponding duplexes by briefly heating strand mixtures and cooling to room temperature. Resolution of a single band via gel electrophoresis at predicted molecular weights was used to qualify duplex formation. Electrospray ionization mass spectrometry was used to confirm duplex identity, while overall purity was analyzed via size exclusion chromatography-high performance liquid chromatography using a XBridge Protein BEH SEC 125 A column (Waters Corporation). Endotoxin levels in each batch were quantified using the endpoint Chromogenic Endotoxin Quant Kit (Bioendo) via proenzyme factor C. All control duplexes and chemically modified sequences are listed in Table S7.

Cell culture and transfection

We maintained 293A cells (Cobioer; Cat# CBP60436) and SK-N-AS (Procell; Cat# CL-0621) cells in DMEM medium supplemented with 10% fetal bovine serum (FBS), penicillin (100 U/mL), and streptomycin (100 mg/mL). T98G cells (Cobioer; Cat# CBP60301) were maintained in MEM supplemented with 10% FBS, 1% NEAA, sodium pyruvate (1 mM), penicillin (100 U/mL) and streptomycin (100 μg/mL). All cell lines were cultured in a humidified atmosphere of 5% CO₂ at 37°C. Transfections were carried out using Lipofectamine RNAiMax (Thermo Fisher Scientific) in growth media without antibiotics according to the manufacture’s protocol.

Gene expression analysis via RT-qPCR

Animal tissue frozen in RNALater (Sigma-Aldrich) was homogenized in Total RNA Isolation Reagent (Biosharp) using a Bioprep-24 Homogenizer (Allsheng). Chloroform was added to the homogenate in which the aqueous phase was removed and mixed with isopropanol. Total RNA was extracted from the tissue prep using the RNeasy RNA kit (Qiagen) according to the manufacture’s protocol. RNA from cell culture was extracted using the Auto-Pure 96A (Allsheng) nucleic acid extraction system. Reverse transcription (RT) reactions were performed with 1 μg total RNA using the PrimeScript RT kit with gDNA Eraser (Takara). The resulting cDNA was amplified in triplicate on the 480 Real-Time PCR system (Roche) using SYBR Premix Ex Taq II (Takara) in conjunction with primer sets specific to hSOD1 and an internal control for either human (i.e., TBP) or mouse (i.e., mTbp) samples. Melting curves were performed after amplification to confirm primer specificity. Averaged cycle threshold (Ct) values for each sample were used to calculate relative gene expression via the ΔΔCt method. Primer sequences are available in Table S7.

Caspase 3/7 activity assay

Caspase 3/7 activity was quantified in cell culture by using the Caspase-Glo 3/7 assay system (Promega). Briefly, a luminogenic substrate was added directly to culture media and incubated for 20 min at 37°C. Luminescence was subsequently measured on an Infinite M200 Pro microplate reader (Tecan). Relative Caspase 3/7 activity was calculated by subtracting background signal of blank from the luminescence values in each well and normalizing data to non-treated (mock) controls.

Cell viability assay

In vitro cell viability was measured using the CCK-8 assay (Dojindo) according to the manufacture’s protocol. Briefly, fresh media containing WST-8 substrate was added to each well of the tissue culture plate and incubated for at least 1 h at 37°C. Absorbance was measured at 450 nm on an Infinite M200 Pro microplate reader (Tecan). Relative viability was calculated by subtracting background absorbance of the blank control from the OD values in each well and normalizing data to non-treated (mock) controls.

Gel shift assay

siRNA or siRNA-ACO duplexes were incubated at 37°C in 50% human serum at a final concentration of 3 μM for ≤1 h. Samples were subsequently mixed with 10× Loading Buffer (Takara) and resolved on agarose gels using GelRed Nucleic Acid Gel Stain (BiotiumUSA) to visualize shifts in gel migration. Gel images were captured on the ChemiDoc XRS+ Imager System (Biorad).

Luciferase reporter constructs and knockdown assessment

Target sequence containing either P.E22G or P.F21C mutant SNPs were cloned into the multiple cloning site of luciferase reporter vector pmirGLO (Promega) between the NheI and SalI restriction enzyme sites downstream of the firefly luciferase gene (luc2) to generate constructs pLuc^P.E22G and pLuc^P.F21C, respectively. A control reporter construct (i.e., pLuc^SOD1) was also created containing consensus hSOD1 sequence perfectly complementary to siSOD1-047M3-AC1^VP guide strand. All constructs were subcloned in DH5a bacteria (Tolobio) and colonies were selected for DNA sequencing to confirm insertion of target sequence. Exemplary colonies were scaled up for plasmid isolation via midiPrep (Qiagen). We plated 293A cells in 96-well cell culture plates at 30,000 cells/well in absence of antibiotics. Cells were co-transfected with one of reporter plasmids (e.g., pLuc^P.E22G, pLuc^P.F21C, or pLuc^SOD1) at 100 ng/well in combination with siSOD1-047M3-AC1^VP or scramble control at the indicated concentrations using 0.3 μL of Lipofectamine 2000 (Thermo Fisher Scientific). Wells treated in absence of test article (0 nM) served as non-treated controls. Cells were cultured for 24 h and luciferase activity was quantified using the Dual-Glo Luciferase Assay System (Promega) according to the manufacture’s protocol. Briefly, cells were lysed in 50 μL Passive lysis buffer (Promega) in which 20 μL of lysate was mixed with 20 μL Dual-Glo Luciferase Reagent and incubated for 10 min at room temperature. Luminescence was subsequently measured on an Infinite 200 Pro microplate reader (Tecan) to quantify luciferase activity. Following measurements, 20 μL of Dual-Glo Stop & Glo Reagent (Promega) was added to each well and incubated for an additional 10 min at room temperature. Luminescence was again measured to quantify Renilla activity, which served to normalize luciferase reporter results. Data were calculated as the ratio of reporter luminescence to Renilla luminescence relative to the ratio of non-treated controls. Percent knockdown was calculated as 1 − (Ratio_siRNA/Ratio_non-treated) × 100.

Animal handling and grouping

Parental transgenic hSOD1^G93A mice (strain ID #004435) were purchased from The Jackson Laboratory and imported into China via Nantong University. Mice were delivered to the animal facility at 6 weeks of age and subsequently bred domestically at Nantong University, which supplied the animals for this study. All animal procedures were approved by the Institutional Animal Care and Use Committee at Nantong University. Formulations for animal treatments were prepared fresh prior to use by dissolving allotments of lyophilized oligonucleotide into aCSF to create stock solutions for dilution to the intended treatment concentrations. Animals were randomly allocated into study groups based on body weight and sex. Any animals in poor health or with obvious abnormalities were omitted from the experiments. Randomization was analyzed using ordinary one-way ANOVA via GraphPad Prism version 8.3.0 Windows (GraphPad Software). Female hSOD1^G93A mice typical weighed approximately 20%–25% less than their male liter mates.

ICV injection

Avertin (1.2%) was prepared fresh and sterilized via 0.2-micron filter. Mice were dosed at 0.30–0.35 mL per 10 g body weight via intraperitoneal injection in a stereotaxic apparatus to rapidly induce anesthesia for up to 30 min. An approximate 11.5-mm incision was made in the animal’s scalp and a 25G needle attached to a Hamilton syringe containing the appropriate siRNA formulation was placed at bregma level. The needle was moved to the appropriate anterior/posterior and medial/lateral coordinates (0.2 mm anterior/posterior and 1 mm to the right medial/lateral). A total of 10 μL was injected into the lateral ventricle at an approximate rate of 1 μL/s. After treatment, the needle was slowly withdrawn and the wound sutured close.

IT injection

Anesthesia was administered via 3.0% isoflurane in an induction chamber for a continuous 10 min. Hair was shaved around the injection site at the base of the tail and cleaned with 75% ethanol. The space between the L5-L6 spinous processes was identified and a 30G needle attached to a microliter syringe containing the appropriate drug formulations was slowly inserted into the intradural space until a tail flick was observed. The needle position was subsequently secured in which 10 μL total volume of solution was injected over the course of 1 min.

Quantification of siRNA-ACO in animal tissues

Tissue lysate was prepared in lysis buffer (0.5% CA-630, 1 mM EDTA, 150 mM NaCl) using a Bioprep-24 Homogenizer (Allsheng). Samples were subsequently heated to 95°C to inactivate sample proteins. Serial dilution of non-treated lysate spiked with siRNA-ACO was used to generate 8-point standard curves. RT reactions were performed using the PrimeScript RT reagent kit (Takara) in conjugation with custom stem-loop primers specific to siRNA guide strands. Each sample was amplified in triplicate on the 480 Real-Time PCR system (Roche) using SYBR Premix Ex Taq II (Takara) reaction mix with primer sets specific to guide strand cDNA. Melting curves were performed after amplification to confirm primer specificity. Absolute quantities of siRNA were extrapolated by linear regression using the appropriate standard curves. Tissue concentrations were calculated as the ratio of absolute siRNA mass (ng) relative to the total weight (g) of tissue sample prepped for lysis.

Clinical observation and endpoint criteria

Animals were observed after injection for up to 4 h and daily thereafter until endpoint. Body weight was determined before test substance administration and at recorded intervals thereafter. Animals with weight loss of more than 20% relative to their initial mass at day of treatment or a NS of NS4 met endpoint criteria.

Neurological scoring

For animals treated via IT injection, mice were evaluated for signs of motor deficit using the ALS TDI NS system, which was developed to provide an unbiased assessment of disease progression based on hindlimb dysfunction common to SOD1^G93A mice.³³ NS was assigned based on the following 4-point scale: 0 if no signs of motor dysfunction (i.e., pre-symptomatic), 1 if hindlimb tremors are evident when suspended by tail (i.e., first symptoms), 2 if gait abnormalities are present (i.e., onset of paresis), 3 if dragging at least 1 hindlimb (i.e., partial paralysis), and 4 if inability to right itself within 10 s (i.e., endpoint paralysis).

Open field test

Each mouse was placed in the corner of an open field apparatus (50 cm length × 50 cm width × 50 cm height) during daylight hours and allowed to freely roam for 15 min. An overhead camera recorded the travel path of each animal. Video footage was analyzed by automated tracking software Samart 3.0 (Bioseb) to calculate total distance traveled.

Rotarod analysis

Animals were trained for 3 days prior to data acquisition. Mice were placed on a motionless rotarod apparatus (XinRuan Information Technology) with a swivel bar 60 mm in diameter. Rotational speed was accelerated from 0 to 30 rpm over the course of 300 s. Latency time was recorded as the amount of time it took for each animal to fall off the swivel bar. Each animal was tested in triplicate, in which the longest value represented latency time.

Grip strength test

Mice were lowered onto a grid plate in which its forepaws and hind paws were allowed to grasp the grid. The tail was gently pulled and maximal muscle strength was measured on the XR501 grip strength meter (XinRuan Information Technology) in units of mass until the animal relinquished its grasp. Each animal was tested in triplicate in which the mean value represented grip strength.

Statistical analysis

Data analytics were performed using GraphPad Prism version 8.3.0 Windows. Dose-response curves and IC₅₀ values were extrapolated using non-linear regression via four-parameter concentration-inhibition model. Where specified, mean values were compared using Tukey’s multiple comparison test to determine statistical differences between different dose-response curves. Drug quantities in tissue in relationship to knockdown activity including extrapolation of ED₅₀ values were performed using non-linear regression via three-parameter concentration-response model. Time-stratified data (e.g., peak weight analysis and animal survival) was plotted via Kaplan-Meier graphs in which statistical significance was verified using the Mantel-Cox test.

Data and code availability

The authors confirm that the data supporting the findings of this study are available within the article and its supplementary materials.