Intra-amniotic antisense oligonucleotide treatment improves phenotypes in preclinical models of spinal muscular atrophy

Abstract

Neurological disorders with onset before or at birth are a leading cause of morbidity and mortality in infants and children. Prenatal treatment has the potential to reduce or prevent irreversible neuronal loss and facilitate normal neurodevelopment. We hypothesized that antisense oligonucleotides (ASOs) delivered to the amniotic fluid by intra-amniotic (IA) injection could safely distribute to the fetal central nervous system (CNS) and provide therapeutic benefit in the motor neuron disease spinal muscular atrophy (SMA), caused by mutations of the survival of motor neuron 1 gene (SMN1) leading to deficiency of SMN protein. Although the splice-switching ASO nusinersen ameliorates SMA when delivered postnatally, substantial deficits can remain in severely affected infants. Here, IA injection of ASOs to two mouse models of severe SMA increased SMN expression in the CNS. In SMAΔ7 mice, which manifest pathology in utero, prenatal treatment improved motor neuron numbers, motor axon development, motor behavioral tests, and survival when compared to mice treated postnatally (between P1 and P3). To assess the feasibility of prenatal treatment in a large animal model, ASOs were delivered mid-gestation to fetal sheep by IA or intracranial injection. ASOs delivered by IA injection distributed to the spinal cord at therapeutic concentrations and to multiple peripheral tissues without evidence of significant toxicity to the fetus or mother. These data demonstrated that IA delivery of ASOs holds potential as a minimally invasive approach for prenatal treatment of SMA and possibly other severe, early-onset neurological disorders.

One-sentence summary:

Antisense oligonucleotides delivered in utero can treat mouse models of SMA and can feasibly be administered to fetal lambs at mid-gestation.

Introduction

Early-onset neurological disorders, including neurodevelopmental disorders and malformations, as well as metabolic, neurodegenerative, and neuromuscular diseases, are often caused by single gene mutations (1–3). They are characterized by disease pathology that begins prenatally or neonatally, interrupting normal developmental processes and often resulting in severe, irreversible damage (4). Prenatally administered DNA or RNA-targeted therapeutics (genetic medicines) have the potential to address the proximal cause of these disorders at a time that maximizes therapeutic benefit (5). Potential advantages of prenatal delivery include improved drug biodistribution given the small fetal mass, the selective permeability of the fetal blood-brain-barrier, and the tendency of the fetal immune system to develop tolerance towards foreign antigens (6). Nonetheless, there are currently no clinical trials of in utero genetic therapies for early onset neurological disorders because it is unknown which therapeutics can be safely and effectively delivered to the fetal central nervous system (CNS).

Spinal muscular atrophy (SMA) is an early-onset motor neuron (MN) disease caused by recessive, loss-of-function mutations of the survival of motor neuron 1 gene (SMN1), and reduced expression of SMN protein (7). Mutations in SMN1 are partially compensated for by retention of a paralogous but alternatively spliced gene, SMN2 (8). SMN2 copy number varies, with patients generally carrying between 1 and 4 copies, and inversely correlates with disease severity (9). About 60% of patients with SMA have type I SMA with 2 copies of SMN2 and develop rapidly progressive muscle weakness soon after birth (10). In the absence of treatment, there is no acquisition of motor milestones and death occurs by age 2 years. Patients with type 0 SMA typically have 1 copy of SMN2 and have profound weakness and arthrogryposis evident at or before birth with very early lethality (11). Three therapeutic agents have been approved by regulatory authorities that can increase SMN in patients with SMA: i) onasemnogene abeparvovec, an intravenously-administered AAV9 (adeno-associated virus 9)-based gene therapy that delivers SMN1 cDNA (12), ii) nusinersen, an intrathecally-administered antisense oligonucleotide (ASO) that modifies splicing of SMN2 mRNA (13), and iii) risdiplam, an orally bioavailable small molecule that also alters SMN2 mRNA splicing (14). All three agents are markedly more effective when administered as soon as possible after birth (13, 15, 16), which has led to population-wide, neonatal screening for SMA to enable earlier treatment in the United States, Canada, Taiwan, and increasingly, in Europe. Despite these advances, many patients with severe SMA retain substantial clinical deficits even after early postnatal treatment (17, 18) likely due to motor neuron developmental and degenerative pathologies that begin in utero (19, 20).

Prenatal therapeutics are rapidly evolving and include different therapeutic modalities that can be delivered by distinct routes and have the potential to improve outcomes in early-onset genetic conditions (21). Umbilical vein injections have been commonly used for in utero blood transfusions (22) and, more recently, to deliver in utero enzyme replacement therapy (23). Intra-amniotic (IA) fluid delivery represents a promising alternative administration route. The amniotic fluid is commonly sampled during diagnostic amniocentesis with a low fetal complication rate (24). Recombinant protein and fluid replacement to the amniotic fluid is being assessed as a treatment for X-linked hypohidrotic ectodermal dysplasia (XLHED) (NCT04980638) (25) and renal agenesis (NCT03101891) (26), respectively. As amniotic fluid is ingested by the fetus, it diffuses through the gastrointestinal and respiratory tracts to the systemic circulation and may allow exposure to the immature CNS (27). ASOs are an appealing modality for IA delivery because of their specificity for their target pre-mRNA/mRNA, which minimizes off-target effects, and because their long half-lives in the CNS enable single dose administration. Previous studies have used IA delivery of ASOs in mouse models of Usher syndrome (28, 29) and Angelman syndrome (30). To date, there have been no studies examining the pharmacokinetics and safety of ASOs delivered in utero in a large animal model. Fetal sheep have been used extensively to explore human fetal therapies (31–34) given their similar size and physiology compared to humans, and the low incidence of preterm delivery (35, 36).

In this study, we examined the safety, biodistribution, and efficacy of prenatal, IA-delivered SMN2 splice-switching ASOs in both small and large animal models. We showed that single dose, IA delivery increases SMN expression in the CNS and ameliorates disease phenotypes in two severe mouse models of SMA. IA delivery of ASOs in fetal sheep was well-tolerated and resulted in CNS and peripheral tissue biodistribution. Taken together, these findings suggest the possibility of a minimally invasive approach for the prenatal treatment of SMA and potentially other early-onset neurological disorders.

RESULTS

Single-dose, IA administration of ASOs increases full-length SMN2 mRNA and SMN protein abundance in SMA mice

The ASO nusinersen increases SMN expression by binding an intronic splice silencing site, ISSN-1, in SMN2 pre-mRNAs, which sterically hinders recruitment of the splice repressor hnRNPA1 and enables exon 7 inclusion (37). To assess the feasibility of delivering a therapeutic ASO by the IA route, we first used an established severe SMA mouse model (“Taiwanese” mice, Smn^−/−SMN2^+/0, JAX#005058) and a nusinersen-analogue ASO of 20 nucleotides, which has a 2’-O-methoxyethyl modification with a phosphorothioate backbone [ASO-10-29 (38, 39)] (Table 1, Fig. 1). Because the gestational time of mice is 18–22 days (40), we treated SMA fetuses once at embryonic day 13 (E13) or E16 with ASO via IA injection at a dose of 110 µg/embryo, 250 µg/embryo, or 360 µg/embryo (fig. S1) using an injection procedure similar to those previously described (28). A dose of 250 µg delivered at E16 was most well-tolerated (90% survival) and therefore used in subsequent experiments (Fig. 1A). To assess the efficacy of intra-amniotically delivered ASO, we compared survival of “Taiwanese” mice treated with a single IA dose of ASO to mice treated prenatally with vehicle, postnatally with subcutaneous (SC) injection of ASO (30 µg) at P1 and P2, or both prenatal and postnatal treatment (IA+SC) (Fig. 1, A and B). As expected, vehicle-treated SMA mice had a median survival of 10.5 days, whereas all ASO-treated mouse cohorts demonstrated robust increases of survival (Fig. 1B). SMA mice treated with IA+SC dosing showed increased survival compared to postnatal SC dosing alone (median survival = 180 days IA + SC vs. 179 days IA vs. 97.5 days SC) (Fig. 1B). Weight gain was equivalent amongst the treated SMA cohorts (fig. S2A).

Table 1.

Length, sequence and backbone modifications of ASOs used in this study.

Aso Name	Length	Sequence and Design	Chemical Composition
ASO 10–29	20-mer	AsTsTsCsAsCsTsTsTsCs AsTsAsAsTsGsCsTsGsG	2’-O-methoxyethyl-modified oligonucleotide with phosphorothioate backbone. All C are 5–methyl–C.
Nusinersen	18-mer	TsCsAsCsTsTsTsCsAsTs AsAsTsGsCsTsGsG	2’-O-methoxyethyl with phosphorothioate modifications. All C are 5–methyl–C.
ASO 449323	18-mer	TsCoAsCoTsToTsCoAsT oAsAoTsGoCsToGsG	2’-O- methoxyethyl with mixed phosphodiester and phosphorothioate modifications. All C are 5–methyl–C. Cy3–labeled ASO was manufactured with a Cy3 phosphoramidite on the 5’-end.

Figure 1. Prenatal IA administration of ASO leads to widespread SMN expression in ‘Taiwanese’ SMA mice.

A. Schematic showing the mating scheme and timeline of ASO treatment and analyses. Image created with www.biorender.com. B. Probability of survival by group. *p<0.05 by logrank (Mantel-Cox) test. All groups were significantly different (****, p<0.0001) from vehicle controls. C. Radioactive RT-PCR gel showing full-length (FL) vs. truncated (Δ7) SMN mRNA in the indicated tissues of P10 mice. Percent exon 7 inclusion is shown (Incl (%)). D. Quantification of data shown in (C), presented as percent exon 7 inclusion (Incl %) [included/(included + skipped)] in select tissues of indicated experimental groups. *p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001 by one-way ANOVA followed by uncorrected Fisher’s (least significant difference) test. E. Western blot showing SMN and β-tubulin protein abundance in tissues of P10 mice. F. SMN to β-tubulin protein ratios in the indicated tissues, quantified from (E). *p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001 by one-way ANOVA followed by uncorrected Fisher’s LSD test.

To determine whether the survival benefit seen with ASO treatment was associated with improved SMN2 splicing (inclusion of exon 7) and SMN protein concentrations, spinal cord, brain, skeletal muscle, heart, kidney, and liver tissues were collected from a separate cohort of treated SMA mice at P10. Radioactive reverse-transcriptase polymerase chain-reaction (RT-PCR) demonstrated increased exon 7 inclusion (Incl, %) and full-length (FL) SMN2 mRNA in CNS tissues as well as in peripheral tissues in all treated cohorts when compared to untreated SMA mice (Fig. 1, C and D). IA and IA+SC dosing also showed full-length SMN2 expression equivalent to or superior to SC dosing alone, particularly in the brain, where IA+SC delivery resulted in a 1.5-fold increase (p=0.002) in full-length SMN2 expression compared to SC dosing alone (Fig. 1, C and D). SMN protein abundance determined by Western blot also demonstrated increases in the tissues of treated SMA mice compared to untreated mice (Fig. 1, E and F). In the spinal cord, we detected a 2.4-fold increase in SMN protein in the IA group (p=<0.0001), a 1.6-fold increase in the SC group (p=0.006) and a 2.2-fold increase in the IA+SC group (p=<0.0001) compared to untreated mice, whereas in brain tissues, there was a 3.7-fold increase in SMN protein expression in both IA and IA+SC groups (p=0.005 and 0.004, respectively), with no significant differences between untreated and SC-treated mice (p=0.06).

We next assessed functional outcome measures in untreated SMA mice versus those receiving IA, SC, or IA+SC ASO. Whereas untreated SMA mice showed increased time to right and reduced time to fall during the first two postnatal weeks of life, the treated cohorts showed improved times to right and to fall (figs. S2,B and C). Hindlimb scores were also improved in the treated cohorts compared to untreated mice during the first two postnatal weeks (fig. S2D). Lastly, forelimb grip strength at P30 was equivalent between IA, SC, and IA+SC groups (fig. S2E). Together these data indicated that a single, prenatal IA ASO treatment increased SMN expression in both CNS and systemic tissues and ameliorated features of disease in this severe SMA mouse model.

Prenatal IA ASO administration enhances therapeutic efficacy in SMAΔ7 mice

Unlike Taiwanese SMA mice, the SMAΔ7 mouse model of SMA (JAX#005025) shows impairments of motor axon growth beginning in utero (20). SMAΔ7 mice also show motor behavioral deficits at birth and neonatal motor neuron degeneration that further recapitulates features of type I or 0 human SMA phenotypes (20). SMAΔ7 mice harbor human SMN2Δ7 cDNA transgenes and a 35.5kb human SMN2 transgene (SMN2^+/+) on a null mSmn background (SMNΔ7^+/+;SMN2^+/+;mSmn^+/−). Litters included SMA (mSmn^−/−), heterozygous ( mSmn^+/−), and wild type (WT) mice (mSmn^+/+). In this model, we assessed a different ASO (#449323, Table 1) targeting the same sequence and length as nusinersen, but with 2’-O-methoxyethyl and mixed phosphodiester and phosphorothioate modifications to the backbone, which improves potency and tolerability (41, 42). No ASO 10–29 injections were performed in this model. Mice injected at E14.5 IA with ASO at a dose of 350 µg/embryo (Fig. 2A) showed a similar average litter survival rate: 66% ± 25 (31 litters, 204 pups) compared to vehicle IA-treated mice: 76% ± 18 survival in (24 litters, 193 pups) (p=0.91). Those mice surviving to birth after vehicle IA injection showed similar survival to untreated mice, suggesting that the IA injection procedure did not affect postnatal outcomes (fig. S3A).

Figure 2. Prenatal IA administration of ASO increases full length SMN2 mRNA expression in the CNS of SMAΔ7 mice.

A. Schematic showing the mating scheme and timeline of ASO delivery and functional analyses after birth. Image created with www.biorender.com. B. Representative immunohistochemistry images of ASO 449323 (brown, top panel) and immunofluorescence images of ASO 449323–Cy3 (yellow, bottom panel; nuclei are stained blue with DAPI) in the brainstem and cervical, thoracic, and lumbar spinal cord of mice injected at E14.5 (IA injection) and harvested at E18.5. White arrowheads indicate motor neurons. Inset in the upper right corner shows a low-magnification view of each tissue section. Scale bars = top row 250 µm (insets) and 50 µm (magnified view); bottom row 2.5 mm (insets) and 100 µm (magnified view). C and D. ASO concentration (µg/g) as measured by electrochemiluminescent assay (C) and full length SMN2 (FL-SMN2) mRNA expression assessed by RT-qPCR (D) in the brain, liver, lung, kidney, and gut of mice injected with 350 µg ASO at E14.5 and harvested at E18.5. Each symbol represents one fetus. Symbols denote different genotypes. **p<0.01 and ***p<0.001 by two-way ANOVA followed by uncorrected Fisher’s LSD test. E. Full-length SMN2 mRNA expression measured by RT-qPCR in spinal cords and brains of treated and untreated SMAΔ7 mice harvested at P10. **p<0.01 by two-way ANOVA followed by uncorrected Fisher’s LSD test.

To assess the biodistribution of the ASO after IA delivery, litters were injected at E14.5 with 350 µg/fetus and mice were harvested at E18.5 for immunohistochemistry (IHC) of the drug backbone. ASO immunoreactivity was not detected in any vehicle-treated mice (Table S1). In ASO-treated SMA, heterozygous, and WT mice, ASO was detected in the CNS, particularly in neurons of the brainstem and to a lesser extent in neurons of lower spinal levels (Fig. 2B, top row). Drug was also observed in multiple systemic tissues, particularly in the respiratory epithelium of the nose, epithelial lining of the small intestine, in liver and kidney (Table S2) and in the placenta in both cytotrophoblasts (fetal side) and syncytiotrophoblasts (maternal side) (Table S2). To better characterize cellular drug uptake, we also treated E14.5 mice with ASO (#449323) tagged with the small molecular weight fluorescent tag Cy3 (similar to previous reports (43)). At E18.5, 5 of 11 mice from 2 litters showed robust uptake in the brainstem and spinal cord (Fig. 2B, bottom row). ASO fluorescence was evident in neurons of the brainstem as well as in motor neurons in the ventral horn of the spinal cord at multiple levels, similar to what our group has previously observed in spinal cord tissues at autopsy from patients with SMA treated with ASOs (44). ASO fluorescence was also observed in other areas of the developing brain as well as in multiple peripheral tissues including the kidney, lung, and intestines (fig. S4).

ASO concentrations were quantified in E18.5 bulk brain, kidney, liver, lung, and intestine tissues using a highly sensitive, sequence-specific, dual-probe electrochemiluminescent (ECL) assay that detects full-length ASO and the n-1 metabolite. ASO was present in all examined organs of treated mice (Fig. 2C). SMN2 mRNA expression was assessed in these same tissues by RT-qPCR and showed increases in the kidney (p=0.002), liver (p=0.004), and lung (p=<0.001) in ASO-treated compared to vehicle-treated mice. (Figs. 2, C and D). Although full-length SMN2 mRNA concentrations were not significantly increased in the brains of IA-treated mice at E18.5 (p=0.70), they were increased in both the brain and spinal cord at P10 (p=0.007 and p=0.001, respectively, Fig. 2E).

We next assessed disease outcomes of SMAΔ7 mice treated with ASO, either prenatally, postnatally, or a combination of the two. WT mice treated with IA ASO 350 µg at E14.5 and SMA mice treated with vehicle IA at E14.5 plus vehicle SC at P1 were used as controls (Fig. 3, A to C). As expected, treated WT mice showed robust survival, weight gain, and rapid acquisition of time to right, whereas vehicle-treated SMA mice showed reduced survival (median=18 days), poor weight gain, and persistent impairments of righting time (Figs. 3, A to C). Treatment of SMA mice with a single dose of ASO 350 µg IA significantly improved survival (p=<0.001), weight (p=0.001), and righting time (p=<0.001) compared to SMA mice treated with vehicle. IA treatment also improved outcomes compared to SMA mice treated with a single dose of ASO 350 µg given SC at P2 (Fig. 3, A to C). Because human SMA treatment would likely incorporate prenatal and postnatal dosing of ASO, we also examined outcomes after treating SMAΔ7 mice prenatally with ASO 350 µg IA at E14.5 along with a postnatal dose of 350 µg SC at P1, compared to SMAΔ7 mice treated postnatally with ASO 350 µg SC at P1+P3. Those mice receiving prenatal treatment showed improved weight gain and righting time compared to SMAΔ7 mice receiving treatment at P2 (Fig. 3, B and C). In addition, long-term assessments of inverted hang testing performed during adult ages showed improved performance in those SMAΔ7 mice receiving prenatal + postnatal doses of ASO compared to those mice receiving two postnatal doses (Fig. 3D). Although SMA mice receiving a single prenatal dose of ASO initially showed improved performance at P60 compared to mice treated postnatally at P1 and P3, this finding did not persist with increasing age, likely due to waning drug concentrations at older ages after a single prenatal dose because drug half-life is shorter in neonatal mice compared to adult mice (45, 46). Of note, IA-treated SMA mice also showed enhanced survival and weight gain compared to SMA mice dosed postnatally with ASO by intracerebroventricular (ICV) injection (60 µg at P2) (Fig. S3 B and C).

Figure 3. Prenatal IA administration of ASO improves survival, motor behaviors, and motor neuron pathology in SMAΔ7 mice.

A to D. Plots displaying probability of survival (A), mouse weights (B), time to righting (C) and inverted hang test scores (D) for each experimental group. *p<0.05, **p<0.01, ***p<0.001 and ****p<0.0001 by log rank (Mantel-Cox) test for survival analysis (A), mixed effect analysis (B to D). # in P2 single treated mice (yellow line) indicates 2/13 mice were treated on P1, 11/13 were treated on P2. E. Representative immunofluorescence images of choline acetyltransferase (ChAT, red) expression in L1 spinal cord motor neurons of indicated groups at P14. Scale bar= 50 µm. F. Quantification of ChAT⁺ motor neurons in the entire L1 spinal cord of indicated experimental groups. **p<0.01 by one-way ANOVA followed by uncorrected Fisher’s LSD test. G. Representative single higher-magnification EM images of L1 ventral roots in prenatally and postnatally treated SMA mice at P14. Scale bar= 2 µm. H. Plot of number of axons of different categories in SMA mice treated with ASO by IA injection prenatally (pink) or by SC injection postnatally on P1+P3 (teal) (n = 5 prenatal, n= 3 postnatal) at P14. *p<0.05 by two-way ANOVA with uncorrected Fisher’s LSD test. I. Scatter plot of L1 ventral root axon g-ratios as a function of axon diameter in prenatally (pink, n = 5) and postnatally (teal, n = 3) SMA-treated mice (1095 axons measured for prenatal, 441 axons measured for postnatal). ****p<0.0001, determined by simple linear regression with slope comparison (F test).

To determine whether prenatal IA ASO treatment was associated with improvements in motor neuron pathology, we assessed motor neuron number in the lumbar spinal cord level 1 (L1), where we have previously described severe motor neuron and motor axon pathology (20, 47). We compared the total L1 segment motor neuron number at P14 in ASO-treated WT mice, vehicle-treated SMA mice, SMA mice treated prenatally with IA ASO 350 µg, and SMA mice treated postnatally SC with ASO 350 µg at P1 and P3. As expected, vehicle-treated SMA mice showed a ~50% reduction in motor neuron number compared to WT mice. Postnatal ASO treatment did not increase L1 motor neuron number above SMA-vehicle treated values, whereas prenatal treatment restored SMA motor neuron numbers to WT values (Fig. 3, E and F). We also assessed the developmental morphology of the L1 motor axons at P14 (20, 48) and noted an increase in the number of myelinated motor axons in prenatally (E14.5) compared to postnatally-treated (P1+P3) mice (Fig. 3, G and H). Axons of prenatally treated mice demonstrated increased myelin thickness evident by a reduced G-ratio (0.55 ± 0.004 (IA treatment) vs. 0.65 ± 0.005 (postnatal treatment), p<0.0001) (Fig. 3, G and I).

Prenatal intracranial administration of nusinersen achieves concentrations predicted to be therapeutic in the spinal cords of fetal lambs

To establish the clinical feasibility of this approach, we performed experiments in a fetal lamb model. Because prenatal ASO administration has not yet been reported in a large animal model, we first sought to establish a benchmark for pharmacokinetics and biodistribution, using an ASO (nusinersen) that is established for postnatal use (Fig. 4). Sheep do not have an SMN2 gene, so it was not possible to measure on-target pharmacology. We injected 8 fetal lambs with different doses of nusinersen [0.7 mg (n=3), 1.5 mg (n=2), and 2.1 mg (n=3)] at a median age of E72.5 (E72-77), equivalent to mid-gestation in fetal lambs (term = 150 days) (Fig. 4A). Fetuses were injected intracranially (IC) into the cisterna magna (a feasible surrogate for the postnatal intrathecal injection route) (Fig. 4B), after which pregnancies were allowed to continue until the time of harvest, about three weeks after the procedure (median E96, E93-E99). Nine untreated fetuses were negative controls, of which three received a sham IC surgery with an injection of artificial cerebrospinal fluid. Route of injection, injection group, delivery route, and dose for all fetuses are shown in Table S3. There was no preterm labor seen during the experimental window and only two fetuses did not survive to harvest, one in the sham group and one in the high-dose (2.1 mg) group, which were found dead at the time of harvest (Fig. 4C). Histopathological analysis of fetal tissues did not reveal a clear cause of demise.

Figure 4. Nusinersen achieves therapeutic concentrations in the CNS of fetal lambs when administered by IC injection.

A. Schematic of the experimental design: fetal lambs were injected with nusinersen at E72–77 and harvested at E93-E99. Image created with www.biorender.com B. Representative picture of an ultrasound-guided intracranial (IC) injection into the cisterna magna of a fetal lamb (left), and an ultrasound of a fetal lamb’s cisterna magna (right). C. Plot showing the percentage of fetal lambs in each litter that survived to harvest. Each dot represents the overall survival rate of a single pregnancy. Control (n=3); low dose (0.7 mg, n=2); medium dose (1.5 mg, n=2); high dose (2.1 mg, n=3); sham IC surgery (n=2). D to F. ASO concentrations (µg/g) in different levels of the spinal cord (D), brain (E), and peripheral organs (F) of fetuses after prenatal IC injection of nusinersen. Each dot represents one fetus. The green shaded area represents the estimated therapeutic range (1–10µg ASO/g tissue). G. ASO concentrations (µg/g) in maternal liver and uterus after fetal IC injection of nusinersen. Each dot represents one ewe.

To quantify the amount of ASO present in tissues, we again used the ECL assay. Based on a previous report (45), we aimed for a target drug concentration between 1–10 µg ASO per gram of tissue for CNS tissues (Fig. 4, D and E). Even at the lowest dose (0.7 mg), nusinersen achieved therapeutic concentrations in all sections of the spinal cord, with above-target values at the intermediate and high doses (Fig. 4D). Similar findings were seen in the rest of the examined CNS regions, although fetuses injected with the low dose trended towards sub-therapeutic values (Fig. 4E). Because of our previous experience with this model, in which prenatal IC administration can result in systemic biodistribution (31), we also examined the concentration of nusinersen in non-CNS tissues and some maternal organs (Fig. 4, F and G). Although therapeutic concentrations can vary from organ to organ (45), we detected ASO in multiple systemic tissues after IC injection, with the highest values in the kidney (consistent with its site of elimination) and liver (Fig. 4F). Maternal exposure to the ASO was minimal, with only very low values detected in the uterus (Fig. 4G).

We evaluated blood chemistries (including liver and kidney function tests, complete blood counts, and coagulation panels) and detailed autopsies of the injected lambs as part of a safety analysis (fig. S5) and detected a decrease in platelet counts in fetal lambs exposed to IC ASO, compared to controls, although the values remained in or slightly below the normal range defined for adult sheep (250–750K platelets/µl) (fig. S5A). We also observed changes in total bilirubin concentration (fig. S5A), which tended to correlate with histopathologic changes in the liver (pigment accumulation and bile canalicular stasis), although such changes were sometimes also observed in sham-operated animals, suggesting a response to surgery (figs. S5, B and C). In dams, there were some altered laboratory parameters after fetal ASO exposure (fig. S6), but the changes observed were largely within the normal range reported for this species or were consistent with a response to surgery.

IA ASO delivery results in delivery to the CNS in fetal sheep

We next performed IA injections in lambs using the same ASO used in the SMAΔ7 mouse model (449323, Table 1) (Fig. 5). Fetuses were injected intra-amniotically at a median age of E77.5 (E72-80) with a wide dose range of ASO (137 mg (n=2); 500 mg (n=5); 1000 mg (n=4)) (Fig. 5, A and B). Three additional animals were injected IC at a dose of 30 mg/fetus as a positive control for CNS delivery of this ASO (see Table S3). Fetuses were harvested about three weeks after injection at a median age of E98.5 (E93-E104). There was no preterm labor seen during our experimental window and fetal survival to harvest was 100% for all groups (Fig. 5C).

Figure 5. ASOs penetrate spinal cord tissues of fetal lambs when administered by IA injection.

A. Schematic of the experimental design: fetal lambs were injected with ASO #449323 at E72–80 and harvested at E93-E101. Image created with www.biorender.com. B. Representative picture of IA injection of a fetal lamb. C. Plot showing percentage of fetal lambs in each litter that survived to harvest. Each dot represents the overall survival rate of a single pregnancy. (Control (n=3); IC (n=2); IA (n=7)). D. ASO concentrations (µg/g) in different levels of the spinal cord after prenatal ASO administration. Each dot represents one fetus. Arrows denote animals represented in Fig. 5E. The green shaded area represents the estimated therapeutic range (1–10 µg ASO/g tissue). E. Representative IHC images of ASO (brown) in the cervical spinal cord tissues of indicated experimental groups. Scale bar = 50 µm. F. ASO concentrations (µg/g) in different CNS tissues of indicated experimental groups. Each dot represents one fetus. The green shaded area represents the estimated therapeutic range (1–10 µg ASO/g tissue).

We measured ASO concentrations in the spinal cord (Fig. 5D). IC-injected animals had ASO concentrations above the target range, except for one outlier animal (2272A), which we suspect was due to a technically suboptimal administration. IA-injected animals had sub-therapeutic ASO concentrations at the lowest dose (137 mg), whereas lambs injected with 500 mg and 1000 mg achieved the therapeutic range, albeit with variability seen with the 500 mg dose. A single IA administration achieved therapeutic ASO concentrations in the cervical and thoracic spinal cords of 6/9 animals and the lumbar spinal cords of 5/9 animals injected at the higher two doses. Consistent with these results, IHC examination of the spinal cords (Fig. 5E) of these animals revealed no ASO in spinal cord neurons at the 137 mg dose but detectable ASO signal in spinal cord neurons of 500 mg- and 1000 mg-injected animals. As with nusinersen, IC injections of ASO#449323 resulted in robust drug concentrations in multiple brain regions (Fig. 5F). IA injection also achieved ASO distribution into the brain at the two higher doses, particularly in the cerebellum and olfactory bulb, although overall uptake in the brain was more variable than the spinal cord (Fig. 5F). Furthermore, ASO was detected in many regions of the brain by IHC (meninges, cortex, cerebellum, subventricular zone, white matter tracts) even in animals with quantitative ASO values below target range (fig. S7).

IA ASO delivery results in widespread biodistribution in fetal lambs and is well-tolerated

To potentially address the non-neurological manifestations of SMA (49), we evaluated ASO uptake in peripheral tissues after IA administration. ASO concentrations were at or above those seen in CNS tissues in all organs examined (Fig. 6A, compared to CNS distribution in Fig. 5F), as was seen in the SMNΔ7 mice (Fig. 2C) and has previously been reported in fetal mice (30). As expected, fetuses injected with 500 or 1000 mg of ASO had higher values than those injected with 137 mg, with the highest values seen in the gastrointestinal tract and kidney, likely due to ASO absorption and elimination sites, respectively. Like nusinersen, IC injections of ASO 449323 also distributed beyond the CNS and into peripheral tissues (fig. S8). IHC staining confirmed the presence of ASO in the lung, liver, and kidney (Fig. 6B).

Figure 6. Biodistribution of ASO in peripheral tissues after prenatal IA administration.

A. ASO concentrations (µg/g) in different peripheral organs of fetal lambs after prenatal IA administration of different doses of ASO. Each dot represents one fetus. B. Representative IHC images of ASO (brown) in lung, liver, and kidney from fetal lambs in the indicated experimental groups. C. ASO concentrations (µg/g) in maternal liver and uterus after IA ASO administration to the fetus. Each dot represents one ewe.

Maternal exposure to ASO may be higher after IA administration, given the close contact between the placenta and fetal membranes with the maternal uterus and circulation. Consistent with this hypothesis, concentrations of ASO in the uterus and liver of the dams were higher than that seen with IC injection and comparable to those in the fetus (Fig. 6C). However, detailed evaluation of maternal chemistries and histology did not yield any concerning changes in exposed dams compared to unexposed or sham-operated controls (fig. S9). Detailed analysis of fetal chemistries demonstrated mild alterations in bilirubin and activated partial thromboplastin time, without a dose-response or relationship to pathologic alterations (fig. S10A to C), similar to changes seen in sham-injected animals in the previous set of experiments. There was no reduction in the platelet count, an adverse effect that has previously been reported following systemic ASO administration (50) (fig. S10A). Histopathological changes were limited to mild acute tubular necrosis (which was also seen in uninjected controls) and bile canalicular stasis (fig. S10D).

DISCUSSION

Here we sought to assess the biodistribution and preliminary safety and efficacy of ASOs administered by IA injection for the treatment of early-onset neurological disorders, using SMA as a candidate disease. Our data showed that: (i) ASOs distribute to the CNS after IA injection in both fetal mice and fetal lambs; (ii) prenatal IA ASO treatment increases SMN expression in multiple tissues and mitigates motor neuron pathology, functional behavioral deficits, and early lethality in severe SMA mice; and (iii) IA injection in the fetal lamb model achieves levels of ASO anticipated to be in the therapeutic range in the CNS, albeit only in a subset of experimental animals. Collectively, these findings support the safety and feasibility of IA injection to deliver ASOs to the CNS, although further development is needed prior to clinical applications for prenatal treatment of SMA and potentially other early-onset neurological diseases.

SMA represents a compelling candidate for in utero intervention because of the prenatal onset of motor neuron pathology in severe patients (19, 20, 51), the ready availability of prenatal diagnostics following recommended carrier screening (52, 53) and prior clinical experience showing that earlier administration markedly improves long-term outcomes (13, 15, 16, 54). Although ongoing population-wide, neonatal screening for SMA available in some countries improves treatment initiation, this timing is often not sufficiently early to completely prevent reverse disease manifestations in severe patients. For example, there is particular concern regarding continued truncal and bulbar muscle weakness despite treatment, as well as emerging phenotypes such as cognitive dysfunction (10, 55, 56). One factor that might limit postnatal treatment efficacy is a requirement for high expression of SMN during embryonic periods (44) when motor neuron axons are growing rapidly in radial diameter and maturing synaptic connections with their target muscles. The histopathology of patients with severe SMA and mouse models of disease is characterized by stunted radial growth of motor neuron axons as well as slowed maturation of both central and neuromuscular junction synapses (47, 57) that is followed by irreversible neuronal loss that likely begins in late gestation (19, 20) and continues rapidly neonatally (51, 58, 59). In utero treatment therefore has the potential to increase SMN abundance at a time when it is most needed, potentially reinstituting normal motor neuron developmental programs and preventing very early neuronal loss.

Previous preclinical studies have explored the feasibility of available SMA treatment modalities administered in utero. For example, one study assessed AAV9-SMN delivered by ICV injection at E14-15 in SMAΔ7 mice (60). When SMA mice survived gestation, treatment was associated with extensions of postnatal survival and improved weight gain; however, it was not demonstrated whether fetal therapy was superior to treatment begun postnatally. Although prenatal gene therapy with AAV9-SMN could be a viable option, a recent prenatal lamb study from our group using AAV9-GFP demonstrated several points of caution, including as an association with intrauterine growth restriction and maternal exposure to the virus (31). Alternatively, the orally bioavailable splice-switching small-molecule risdiplam could be an appealing candidate for in utero therapy because it crosses the placenta and thus can be delivered to the mother with exposure in the fetus. We have previously demonstrated the feasibility of in utero treatment of SMAΔ7 mice by treating the pregnant dam daily with a risdiplam analogue starting at E9.5 or E13.5 (20). SMA mice treated in utero showed postnatal improvements of motor axon radial growth, myelin acquisition, and enhanced motor neuron survival. Recently, a single human fetus diagnosed with type I SMA in utero was treated with risdiplam starting at 32 weeks (via maternal administration) and this resulted in increased SMN protein expression postnatally and absence of disease manifestations by 30 months of age, although there were also unexplained congenital anomalies and global developmental delay (61). Although promising, concern remains regarding the safety of in utero risdiplam because it can alter splicing of other pre-mRNAs including some with a role in cell cycle arrest (62). In contrast, ASOs target and bind specific pre-mRNAs and RNAs via Watson-Crick base pairing and thus have less risk of off-target mRNA effects. To date, nusinersen has been dosed to over 14,000 patients (63) and has demonstrated a well-established favorable safety profile (64).-While the postnatal delivery route is intrathecal, the minimally invasive approach of IA injection would be ideal for prenatal applications if sufficient CNS penetration can be achieved.

Utilizing the IA injection route, we demonstrated that injection with two distinct ASOs delivered once during the last third of gestation to two different mouse models of SMA can increase CNS tissue SMN expression and substantially improve disease outcomes. Our results, combined with our recent report of ASO delivery in fetal mice (30), suggest several potential routes of ASO entry into the CNS. High concentrations of drug in the respiratory tract, lung, and gastrointestinal tract of both fetal mice and fetal lambs suggest exposure through ingestion; similarly, high values in the placenta suggest hematogenous spread. An alternative route of CNS penetration could be the nose-brain pathway whereby drug penetrates the CNS via the olfactory mucosa (65, 66) supported by our finding of ASO in the olfactory bulb and in the cribriform plate in our previous study using a different ASO (30).

The main limitation of this study is that in the sheep model, the levels of ASO in the brain was lower with IA injections compared to IC injections; therapeutic levels were achieved in the spinal cords of only a subset of the animals after IA injection. It is possible that these results could be improved through earlier injection of ASO (which could increase brain delivery), or modifying the ASO backbone (for example, by using a new generation ASO that is currently in postnatal clinical trials (NCT05575011)), or dose titration. Another consideration is that the ASO may recirculate in the amniotic cavity due to renal excretion of the drug: urine is expelled to the amniotic fluid and ongoing swallowing by the fetus could enable “redosing” of ASO with more effective achievement of therapeutic values, possibly akin to the loading doses of ASOs that are administered during initial postnatal dosing. It is important to understand whether the concentration in the CNS would increase over time. Conversely, there could be further accumulation in the kidneys over gestation, resulting in renal toxicity (we did not observe dose-dependent changes in treated animals compared to controls, but this study was not powered to be a comprehensive safety analysis). Overall, given the decreased technical risk of IA injection over ICV injection, it is important to continue to develop this approach for a future clinical application.

The ideal clinical translation of this approach would be in patients with type 0 and type I SMA, in which a single IA dose could be administered under ultrasound guidance after prenatal diagnosis and non-directive counseling. IA injection is already being utilized in clinical trials to deliver proteins and fluids during the 3^rd trimester (NCT04980638, NCT03101891) (25, 26). Although couples who are aware of carrier status may opt for avoidance of affected pregnancies through in vitro fertilization and preimplantation genetic diagnosis, we recognize that many patients are diagnosed in the fetal period, when a therapeutic intervention could improve outcomes. For an IA approach, this could be done as early as 15 weeks, consistent with the timing in which amniocenteses are safe and recommended, (67) although treatment later in gestation would also be safe and potentially effective at reversing existing SMA pathology. (61) If an IC approach was to be pursued, an intraventricular injection of ASO would be technically feasible starting from late in the second trimester or early in the third trimester. Based on the usual half-life of ASOs, it is unlikely that a second dose would be necessary, but postnatal dosing could commence after birth or be followed by treatment with another modality.

On a final note, the molecular, cellular, and phenotypic improvements observed after IA delivery of splice-switching ASOs in two SMA mouse models, in combination with the spinal cord biodistribution of ASO in fetal lambs are encouraging. The observed distribution of ASO in systemic tissues could also be beneficial for addressing extra-neurological manifestations of SMA (49) and may be relevant for other diseases primarily affecting organ systems such as the lungs, heart, or GI tract (68, 69). Together, these results suggest that IA delivery of genetic medicines could potentially be used as a minimally invasive approach to treat SMA and other severe congenital disorders before birth, to prevent or reduce irreversible organ damage.

METHODS

Study design:

The objective of this study was to perform a preliminary assessment of the safety, feasibility, and efficacy of prenatal ASO administration. For this purpose, we performed in vivo experiments in mice and sheep. Experiments in mice included two different mouse models of SMA using two different ASOs. These experiments were done in compliance with regulatory protocols as established in the sections below. Assessments included quantitative and qualitative techniques. Sample sizes were determined by previous experience with these models from our groups, preliminary data, and power calculations. (20) Animal inclusion was decided by genotyping PCR with primers recommended by Jackson Laboratory for JAX stocks JAX #005058 and #005025. Both female and male mice were used for these experiments. Researchers were blinded when acquiring data and performing quantification. Sheep experiments were done in wild-type sheep using two different ASOs. All experiments were done in compliance with regulatory protocols as established in the sections below. Assessments included quantitative and qualitative techniques. Sample size was determined based on our previous experience with this model. (31) Fetus sex was not used as an exclusion criteria and ewes were randomly assigned to each experimental group. Researchers were not blinded to experimental groups during the experiments or the analysis. Specific Materials and Methods are described in the sections below. Additional Materials and Methods can be found in the Supplementary Materials file.

ASOs:

The ASOs (Table 1) were synthesized and purified as described previously (70). For ASO 449323–Cy3, ASO containing free amine was dissolved in 0.1 M Borate pH 8.5 solution at a concentration of 100 mg/ mL. Three equivalents of NHS-Cy3 (Lumiprobe, CAS# 2632339-91-2) was dissolved in anhydrous dimethylformamide at a concentration of 100 mg/ mL. Solutions were mixed and shaken for 30 minutes. ASO was purified via a Strong Anion Exchange gradient and desalted via C18. For injections, ASO was resuspended in PBS at room temperature and stored at −20°C until injection and resuspended in D-PBS at room temperature and stored at – 80°C for experiments.

Animal husbandry:

Mice experiments were performed using severe SMA “Taiwanese” mice (JAX#005058) and SMAΔ7 mice (JAX Stock #005025). Breeding and treatment of mice were performed in accordance with the National Institute of Health guidelines and protocols were approved by Institutional Animal Care and Use Committees of Cold Spring Harbor Laboratory (#23-20-17-14-3) and Johns Hopkins University School of Medicine (#MO24M187). All animals are checked daily. All animals are group-housed routinely on arrival or at weaning, whenever possible; some exceptions might be pregnant or surgical females, mothers with litters, stud males, fighting animals, or for scientific or health reasons. Breeding mice receive nesting material that enhances their environments. Unless scientifically justified, same-sex animals are housed in compatible groups. All rodents are housed on direct bedding. Mice receive either tap water, autoclaved tap water, or RO water in bottles with sipper tubes that have been autoclaved in-house. Structural enrichment opportunities are provided in the form of privacy areas with biohuts. Nonstructural enrichment opportunities are provided in the form of nesting materials and food treats. All animal rooms are on 12:12 light/dark cycle. All sheep handling complied with and was approved by the Institutional Animal Care and Use Committee (IACUC) of the University of California, San Francisco (#AN202125-00A) and University of California, Davis (#23659).

Breeding and perinatal ASO administration:

Genotyping was performed as previously described (71, 72). Both the Taiwanese and SMAΔ7 models contain copies of the murine Smn gene and the human SMN2 transgene. The SMAΔ7 model also contains copies of the gene encoding the cDNA of SMN lacking exon 7 (SMNΔ7). Embryonic development in these lines of mice is ~18.5 days and for all experiments the day of birth was considered P1. During timed mating, pregnancy was confirmed by visualization of sperm plug (E0.5) and observation of consistent weight gain of greater than 0.5 gram per day. At specified fetal ages, pregnant dams were anesthetized with 2.5% isoflurane, given buprenorphine at 0.1 mg/kg for pain control, and placed on a heating pad. The dam’s lower abdomen was shaved and disinfected with 3 rounds of 70% ethanol and betadine, and a 2 cm vertical incision was made. Embryos were extracted using blunt tipped forceps, placed on a sterilized heated gauze, and consistently moisturized with warm saline. Injections were made into the amniotic fluid of each amniotic sac with a pulled glass micropipette (0.17 mm inner diameter, 0.5 mm outer diameter) using UMP3 UltraMicroPump III. Embryos were given indicated doses of Cy3–conjugated ASO in 5 µL PBS, unconjugated ASO in 5 µL PBS with 0.025% Fast Green FCF dye or PBS alone with dye. Injected embryos were expeditiously reinserted into the womb, the dam was sutured and monitored for 1–2 hours post-procedure. Postnatal treatment of ASO was administered at indicated doses by subcutaneous injection (interscapular) of intracerebroventricularly. Sample size was determined by previous experience, preliminary data, and power calculations. The mice were randomized to treatment group.

Sheep injections:

After induction of general anesthesia and confirmation of pregnancy by ultrasound, a midline laparotomy was performed, followed by a hysterotomy of the uterine horn to expose the fetus. For intra-amniotic injections, an ultrasound was used to identify a pocket of amniotic fluid where it would be safe to perform the injection. A 27G butterfly needle was inserted through the uterine wall and into the amniotic cavity. A sample of amniotic fluid was taken prior to injection. ASO was administered in a total volume of 5 mL, followed by gentle uterine massage to ensure mixing with the amniotic fluid, after which a second amniotic fluid sample was taken. For intracranial injections, sterile ultrasound (Wisonic Piloter Vet Ultrasound Diagnostic System) was used to visualize the cisterna magna in the sagittal plane. The junction between the occipital bone and the upper cervical spine was identified and marked on the skin. Under direct ultrasound visualization, a 25G needle was advanced through a midline entry point in the skin, usually with a slight rostral angulation so the subarachnoid space was reached by passing immediately inferior to the margin of the occipital bone. A total volume of 0.4 mL was injected. These injections were a surrogate for intracerebroventricular injections, which were not surgically feasible given the small size of the fetal ventricles at the injection timepoint. However, cisterna magna injections spread across the CSF. The uterus was closed after amniotic fluid replacement with saline and antibiotics (penicillin-gentamycin). The abdomen was closed in layers and the ewe was recovered per protocol. In some pregnancies, uninjected littermates served as controls while in others, time-dated pregnancies without in-utero manipulation served as pure controls. At harvest, the ewe was anesthetized and a laparotomy and hysterotomy were performed to deliver the fetus(es). After delivery, neonatal blood was drawn from the umbilical vein, the lamb was euthanized with 100 mg/kg sodium pentobarbital (IV) and perfused with PBS. Following extraction of the fetuses, maternal tissues were harvested, and the ewe was euthanized with 0.5 ml/kg potassium chloride (IV). All tissues were placed in either 4% paraformaldehyde or 10% neutral buffered formalin for fixation over 48 hours, after which they were transferred to 30% sucrose and 70% ethanol, respectively, prior to embedding. Amniotic fluid was stored at 4 °C after collection. Fluid was centrifuged at 1320 g for 10 min at room temperature. Supernatant and amniocyte pellet were collected and frozen at −80 °C.

Statistical Analysis:

Results were graphed and analyzed using statistical software GraphPad Prism version 9.5.1 (GraphPad Software). All measurements were taken from distinct animal samples and no repeated measurements were taken. Parametric data (as determined by normality analysis in GraphPad Prism) were analyzed via unpaired and paired t-tests, mixed effect tests and one or two-way analysis of variance (ANOVA) tests with uncorrected post-hoc Fisher’s LSD test. Non-parametric data were analyzed using Kruskal Wallis and Wilcoxon-rank sum tests. Survival data were analyzed using logrank (Mantel-Cox) test. Electron microscopy axonal G-ratio data were analyzed using linear regression and slope comparison (F test). Maternal sheep laboratory values were analyzed via unpaired and two-tailed paired t-tests. Fetal laboratory values and pathology scores were described by ASO injection status and dose using median (1^st−3^rd quartile [Q1–Q3]) and frequencies. Differences by ASO injection status were tested using Wilcoxon rank-sum tests for continuous variables and Fisher’s exact tests for categorical variables. To test for trends in dose response Spearman correlation coefficients were used for laboratory values and pathology scores, and Cochran-Armitage tests were used for pathology scores dichotomized as non-zero or zero. Significance was set at 0.05.

Supplementary Material

LIST OF SUPPLEMENTARY MATERIALS

Materials and Methods

Figs. S1 to S9

Tables S1 to S3

References (73 to 80)

Data file S1

MDAR checklist

Data and Materials Availability Statement:

All data associated with this study are present in the paper or the supplementary materials.