Spinal Muscular Atrophy (SMA) is a motor-neuron disease and the leading genetic cause of infant mortality; it is caused by loss-of-function mutations in the Survival motor neuron 1 (SMN1) gene 1 . Humans have a paralog, SMN2, whose exon 7 is predominantly skipped 2 ; the limited amount of functional, full-length SMN it expresses cannot fully compensate for the lack of SMN1. SMN is important for spliceosomal snRNP biogenesis 3 , but downstream splicing targets involved in pathogenesis remain elusive. There is no effective treatment for SMA, but SMN restoration in spinal-cord motor neurons is believed to be necessary and sufficient for therapy 4 . Non-CNS pathologies, including cardiovascular defects, were recently reported in severe SMA patients and mouse models 5–8 , reflecting autonomic dysfunction or direct effects in cardiac tissues. Here we compared the effects of systemic versus CNS restoration of SMN in a severe mouse model with 10-day survival 9,10 . We used an antisense oligonucleotide (ASO-10-27) that effectively corrects SMN2 splicing and restores SMN expression in motor neurons after intracerebroventricular (ICV) injection 11,12 . Surprisingly, systemic administration to neonates robustly rescued severe SMA mice, much more effectively than ICV administration; subcutaneous injections extended the median lifespan by 25-fold. We further demonstrate decreased expression of hepatic Igfals in neonatal SMA mice, leading to a pronounced reduction of circulating insulin-like growth factor 1 (IGF1); ASO treatment restored IGF1 to normal levels. These results suggest an important role of the liver in SMA pathogenesis, underscoring the importance of SMN in peripheral tissues, and demonstrate the efficacy of a promising drug candidate. To compare the effectiveness of ASO 10-27 delivered centrally versus systemically, we injected 20 μg of ASO-10-27 ICV at postnatal day 1 (P1) to increase SMN in CNS tissues, or we injected the ASO on two separate days subcutaneously (SC) at 50 μg per g of body weight (μg/g), between P0 and P3 (2 doses). These doses were based on our previous studies with this ASO 11,13 . We also evaluated combined ICV and SC injections, and repeated SC injections (Supplementary Table 1). Control heterozygous mice (Smn +/−; SMN2 +/0) that received ICV and/or SC ASO injections had normal survival and behavior. Severe SMA mice (Smn −/−; SMN2 +/0) that received ICV and/or SC saline- injections survived 1–2 weeks, with a median survival of ~10 days, similar to untreated mice (Fig. 1a; Supplementary Figs. 1a,2a, Video 1). ASO delivered only into the CNS efficiently corrected SMN2 exon 7 splicing in the spinal cord and led to a striking increase in SMN protein levels, but modestly extended the median survival to 16 days, with a single pup surviving for one month (Fig. 1a–c; Supplementary Fig. 2b–d). In marked contrast, systemic treatment with two SC injections resulted in a median survival of 108 days. Combining ICV and SC ASO injections further increased the median survival to 173 days; and additional SC injections at P5-P7, after the initial SC injections at P0-P3, extended the median survival to 137 days (Fig. 1d). Treated mice varied in size from comparable to heterozygous littermates to runt; the average weight was low, and the tails were much shorter (Supplementary Figs. 3,4). The surviving runts slowly gained weight, reaching ~18 g at ~3 months. Most rescued SMA mice could run and climb normally; however, their tail and ears developed necrosis, and were gradually lost, resembling the phenotype of type III SMA mice (Supplementary Fig. 3e, f). Additional ASO delivered either by ICV injection at P1 or repeat SC injections at P5-P7 delayed necrosis (Supplementary Fig. 3g). To further characterize the effects of the ASO administered systemically, we carried out a dose-response study with 0 (SC0), 40 (SC40), 80 (SC80) and 160 (SC160) μg/g/injection, given twice between P0 and P3. Systemic treatment with the ASO resulted in a dose-dependent increase in survival (Fig. 1e), with the median survival increasing from 10 to 84, 170, and 248 days, respectively. At the highest dose tested, ASO-10-27 given systemically resulted in long-term survival comparable to the best results achieved by adeno-associated-virus (AAV) expression of the SMN protein in a slightly less severe mouse model 14–16 . Remarkably, 2/14 mice in the SC160 group, and 2/18 in the SC80 group are still alive and active after >500 days. A similar survival benefit was achieved by intraperitoneal (IP) dosing (Supplementary Fig. 5). There was no significant difference in weight gain among the three SC-dosing groups; however, SC160 mice had significantly longer tails (Supplementary Fig. 6a–e). We also observed dose-dependent rescue of ear and tail necrosis, and dose-dependent delays in the development of cataracts and rectal prolapse (Supplementary Fig. 6f–h). Administration of the ASO on days P5 and P7 resulted in a modest increase in survival, compared to earlier treatments on P0 and P3, emphasizing the importance of early-postnatal therapeutic intervention (Fig. 1d). To examine SMN2 splicing changes in various tissues after SC ASO injection, we performed RT-PCR on RNA samples from P7 mice, and detected a dose-dependent increase in exon 7 inclusion in spinal cord, brain, liver, heart, kidney, and skeletal muscle, with the strongest effect occurring in the liver and the weakest in the kidneys. In contrast, ICV administration of the ASO resulted in a much more robust change in exon 7 inclusion in brain and spinal cord tissues, but very limited effects in peripheral tissues (Fig. 2a,b; Supplementary Fig. 7, 8). Immunoblotting of the spinal cord, liver, and heart tissue samples from mice treated by SC administration demonstrated a corresponding increase in full-length SMN protein (Fig. 2c; Supplementary Fig. 7a). Exon 7 inclusion in liver significantly decreased after P30 (Fig. 2d,e; Supplementary Fig. 9), consistent with a measured ASO half-life of 22 days in liver, (data not shown). These data suggest that transiently increasing SMN expression in peripheral tissues during the first few weeks of life has a profound effect on long-term survival. The SMN2-splicing changes were consistent with the ASO distribution assayed by immunohistochemistry, with the apparent exception of the kidney, but in this case most of the ASO was not internalized into cells (Supplementary Fig. 10, 11). We also observed some ASO accumulating in spinal-cord motor neurons (Supplementary Fig. 10, 11). The limited ASO distribution and the moderate SMN2-splicing changes in the CNS after systemic administration likely reflect incomplete closure of the BBB in neonates 17 , and/or ASO retrograde transport. However, we detected strong cytoplasmic SMN staining and/or a pronounced increase in gem number in spinal-cord motor neurons after ICV injection of 20 μg ASO, but not after two SC injections of ASO at 80 μg/g/injection between P0 and P3, a dosage that markedly rescued the severe SMA mice (Supplementary Fig. 12, 13). Therefore, the effect on splicing in the CNS after systemic administration probably contributes to the extended survival, which is consistent with the combined ICV and SC treatment giving even better survival than SC administration alone (Fig. 1d); yet, the striking effects of systemic administration on survival in this severe mouse model cannot be solely explained by a direct effect on SMN2 splicing in the CNS. The rescue of severe SMA mice by systemic administration of, e.g., histone deacetylase inhibitors or AAV vectors, has been attributed to their ability to cross the BBB 10,15 . However, our data indicate that SMN restoration in peripheral tissues, in combination with partial restoration in the CNS, can achieve efficient rescue of severe SMA mice. Histological examination of tissues/organs associated with SMA in mice treated systemically with 160 μg/g of ASO-10-27 and sacrificed at P9, revealed striking improvements, consistent with the markedly increased survival of mice in the SC160 group. The α-motor-neuron counts in spinal cord were comparable to the control heterozygous littermates, and the mean area of muscle fiber cross-sections was >80% of that in heterozygotes (Fig. 3a,b; Supplementary Fig. 14a). Likewise, the heart weight, and the thickness of the inter-ventricular septum and left ventricular wall were similar in ASO-treated mice and heterozygous littermates (Fig. 3c,d; Supplementary Fig. 14b). Finally, staining of neuromuscular junctions (NMJ) showed that NMJ integrity was similar to that in heterozygous littermates (Fig. 3e). Most systemic-ASO-treated mice showed no overt signs of motor dysfunction (Supplementary Video 2, Supplementary Table 2). We employed three tests to evaluate their behavior and motor function. The first was a Rotarod test, which requires limb-muscle strength, as well as balance and coordination. Three-month-old mice in the SC80 and SC160 groups could stay on the rotating rod for ~12 sec, i.e., shorter than heterozygotes, but longer than mice in group SC40. Some treated mice passed a 30-sec acceleration-profile test that many heterozygotes failed (Fig. 3f,g; Supplementary Video 3). Considering that SMA is a neuromuscular disease, this performance represents a remarkable phenotypic improvement. The second test evaluated muscle strength in mice from the SC160 group at 5 and 9 months. At both ages, the forelimb grip strength of treated SMA mice was ~80% that of heterozygous mice (Fig 3h). The final test employed HomeCageScan, a video-based platform for automated high-resolution behavior analysis 18 . Treated SMA mice performed various behaviors similarly to heterozygous mice, except for rearing, suggesting some hindlimb weakness (Supplementary Fig. 15). Two reasons prompted us to examine the growth-hormone (GH)/IGF1 axis. First, all severe SMA mice are small 9,10,19 , reflecting growth retardation. Second, the major effect of SC ASO injection on SMN2 splicing is in the liver, which contributes ~75% of circulating IGF1 20 . Moreover, liver-derived IGF1 is sufficient to support normal postnatal growth in Igf1-null mice 20 . IGF1 is a potent neurotrophic factor 21 , and is also involved in cardiac development and function 22 . ELISA of P6-P9 serum samples showed undetectable or greatly reduced IGF1 levels, compared to heterozygote controls, and SC ASO administration restored IGF1 to normal levels (Fig. 4a). RT-PCR showed that hepatic Igf1 mRNA was not reduced in SMA mice, and increased from P1 to P5 in both heterozygote and SMA mice (Fig. 4b). IGF-binding-protein acid labile subunit (IGFALS), which is postnatally stimulated by GH, binds to IGF1 and IGFBP3 to form a stable ternary complex, extending the half-life of IGF1 from 10 min to >12 h 23 . Inactivation of Igfals results in low circulating IGF1 and IGFBP3, as well as impaired postnatal growth 23 . RT-PCR revealed a marked reduction in Igfals mRNA in the liver of both P1 and P5 SMA mice; moreover, ASO administration rescued Igfals expression (Fig. 4c–d). We conclude that the striking reduction in serum IGF1 levels in SMA mice is likely caused by decreased Igfals expression, which correlates with SMN deficiency and SMA progression. Because Igfals expression is decreased at P1, when the pups are still healthy, we propose that the early deficiency in circulating IGF1 may be one of the factors that contribute to the pathogenesis of severe SMA mice (Supplementary Fig. 1b). Though in several mouse mutants, an impaired GH/IGF-1 axis results in increased lifespan 24 , a severe lack of IGF1 may contribute to SMA progression in concert with other defective factors. Consistent with our hypothesis, two recent studies showed that a local increase of IGF1 in either spinal cord or muscle increases survival of severe SMA mice 25,26 . Indeed, disruption of the IGF1 system is a common feature of neurodegenerative diseases, including Alzheimer’s and amyotrophic lateral sclerosis (ALS) 27 . IGF1-null mice also display some phenotypic similarity to SMA mice, such as small size, severe and generalized muscle dystrophy, including of the diaphragm and heart, with most of them dying at birth 28 . Moreover, dysregulation of the IGF1 receptor and its downstream signaling pathway has been observed in type I SMA patients 29 . However, in light of the inconsistent results of IGF1 therapy for ALS between humans and mice 21,30 , it will be crucial to determine to what extent SMA mouse models accurately mimic human SMA. This will further refine our understanding of the mouse models, and influence therapeutics development and clinical treatments for SMA. METHODS SUMMARY ASO-10-27 was synthesized as described 11 and dissolved in 0.9 % saline. The severe SMA mouse model was generated from an SMA type III mouse model, as described 10 . All mouse protocols were in accordance with Cold Spring Harbor Laboratory’s Institutional Animal Care and Use Committee guidelines. Treated mice were provided with additional gel food. The procedures for neonatal ICV injection, tissue-sample collection, RT-PCR, and Western blotting, and the human-specific anti-SMN antibody (SMN-KH) were described previously 11 . Primers for gene expression analysis are shown in Supplementary Table 3. Serum IGF1 was analyzed with a Quantikine Mouse/Rat IGF1 Immunoassay kit (R&D Systems). Mouse spinal cord, quadriceps, and heart were fixed and H&E stained as described 12 . α-motor neurons in serial 10–20 μm cross-sections of lumbar L1-L2 spinal-cord were counted. Muscle-fiber cross-sectional area was calculated with Axivision LE velocity. NMJ in toto staining was performed as described 12 . For the Rotarod (AccuScan Instruments) test, a four-phase profile was used: phase 1, from 1 to 10 rpm in 7.5 s; phase 2, from 10 to 0 rpm in 7.5 s; phase 3, from 0 to 10 rpm in 7.5 s in the opposite direction; phase 4, from 10 to 0 rpm in 7.5 s. A grip-strength meter (Columbus Instruments) was used for the gripping test. Mice were allowed to grasp a triangular bar with their forelimbs, and were pulled back horizontally. The test was repeated 5 times for each mouse, and the highest value was recorded as the grip force for that animal. Statistical significance was analyzed by two-tailed Student’s t-tests. Kaplan-Meier survival data were analyzed with Mantel-Cox tests using GraphPad Prism. Supplementary Material 1
