Minor snRNA gene delivery improves the loss of proprioceptive synapses on SMA motor neurons

Spinal muscular atrophy (SMA) is an inherited neuromuscular disorder caused by reduced expression of the survival motor neuron (SMN) protein. SMN has key functions in multiple RNA pathways, including the biogenesis of small nuclear ribonucleoproteins (snRNPs) that are essential components of both major (U2-dependent) and minor (U12-dependent) spliceosomes. Here we investigated the specific contribution of U12 splicing dysfunction to SMA pathology through selective restoration of this RNA pathway in mouse models of varying phenotypic severity. We show that viral-mediated delivery of minor snRNA genes specifically improves select U12 splicing defects induced by SMN deficiency in cultured mammalian cells as well as in the spinal cord and dorsal root ganglia of SMA mice without increasing SMN expression. This approach resulted in a moderate amelioration of several parameters of the disease phenotype in SMA mice including survival, weight gain and motor function. Importantly, minor snRNA gene delivery improved aberrant splicing of the U12 intron-containing gene Stasimon and rescued the severe loss of proprioceptive sensory synapses on SMA motor neurons, which are early signatures of motor circuit dysfunction in mouse models. Taken together, these findings establish the direct contribution of U12 splicing dysfunction to synaptic deafferentation and motor circuit pathology in SMA. Research Neuroscience


Introduction
Removal of introns from precursor mRNAs (pre-mRNAs) is a finely regulated nuclear process carried out by the spliceosome, a highly dynamic RNA-protein machine (1). There are two functionally distinct spliceosomes comprising different subsets of small nuclear ribonucleoproteins (snRNPs) that are dedicated to the excision of different types of introns based on their specific splicing consensus sequences (2). The majority of introns referred to as U2-type introns are excised by the major (U2-dependent) spliceosome that includes U1, U2, U4/U6, and U5 snRNPs. Additionally, a small fraction of introns referred to as U12-type introns are processed through the low-abundance minor (U12-dependent) spliceosome formed by U11, U12, U4atac/U6atac and U5 snRNPs (2). U5 is the only snRNP common to both spliceosomes and minor snRNPs are about one hundred-fold less abundant than major snRNPs (3), reflecting the fact that U12 introns account for less than 1% of all introns (4,5).
The accurate removal of U12 introns is essential for organism development (6)(7)(8); and its disruption by mutations in components of the minor spliceosome has recently been associated with human disease (9). Prominent examples are autosomal recessive developmental disorders caused by loss of function mutations in the U4atac snRNA gene including microcephalic osteodysplastic primordial dwarfism type I (10, 11), Roifman syndrome (12) and Lowry Wood syndrome (13), all of which are characterized by postnatal growth retardation and microcephaly among other clinical features. Interestingly, U12-type introns and their relative positions are evolutionarily conserved and not randomly distributed across the genome (14)(15)(16)(17). Rather they are enriched in information processing genes implicated in DNA and RNA metabolism as well as vesicular transport and voltage-gated ion channels that have especially important functions in neurons. Accordingly, disturbance of U12 splicing has also been associated with neurodevelopmental and neurological deficits (9), including early-onset cerebellar ataxia caused by homozygous mutations in the U12 snRNA gene (18). The first link between U12 splicing dysfunction and neurodegenerative disease came from studies of spinal muscular atrophy (SMA) (19)(20)(21), and further evidence implicating this RNA pathway in motor neuron disease has been accumulating from studies of amyotrophic lateral sclerosis (ALS) (22-25). To date, however, conclusive evidence that U12 splicing defects contribute to the etiology of these neurodegenerative diseases has been missing and ideally requires demonstration that selective functional restoration of this RNA pathway leads to phenotypic improvement in animal models.
Here we sought to determine whether U12 splicing dysfunction contributes to the SMA phenotype in mouse models with varying disease severity. SMA is an autosomal recessive neurodegenerative disorder characterized by loss of motor neurons, atrophy of voluntary muscles and motor impairment (26, 27). SMA is caused by a ubiquitous reduction in the survival motor neuron (SMN) protein due to homozygous deletions or mutations of the Survival Motor Neuron 1 (SMN1) gene with preservation of a nearly identical gene called SMN2 that produces low levels of functional SMN (28). SMN is essential for the biogenesis of Sm-class snRNPs of the major and minor spliceosomes (29, 30). Specifically, SMN mediates the assembly of a heptameric ring of Sm proteins on a conserved sequence of each snRNA to form the protein core required for the biogenesis, stability and function of snRNPs (31, 32). SMN-dependent impairment of snRNP assembly leads to a preferential decrease of minor snRNPs of the U12 spliceosome in SMA mice (19,33,34) and accumulation of U12 splicing defects has been documented in a variety of cellular and animal models of SMN deficiency (20,21,35,36). Importantly, we previously showed that dysregulation of a U12 intron-containing gene we called Stasimon (also known as Tmem41b) is responsible for specific neuronal phenotypes induced by SMN deficiency in Drosophila and zebrafish models of SMA (21). Furthermore, we recently demonstrated the direct contribution of Stasimon to motor circuit pathology in a severe mouse model of SMA (37). These and other studies support the notion that snRNP dysfunction leads to perturbation of RNA splicing of select genes that in turn contribute to SMA (21,33,(37)(38)(39)(40)(41), but a specific pathogenic role of U12 splicing dysfunction in mouse models of the disease has not been directly demonstrated.
Here, we devised a means to functionally enhance the U12 splicing pathway in the context of SMN deficiency through viral-mediated delivery of minor snRNA genes. This approach was effective in selectively correcting U12 splicing defects without increasing SMN levels or influencing other RNA processing events regulated by SMN. Furthermore, enhancing U12 splicing ameliorated disease phenotypes in SMA mice leading to prolonged survival and improved motor function. Importantly, minor snRNA gene delivery improved missplicing of the Stasimon gene and robustly rescued the loss of proprioceptive sensory synapses on motor neurons, which represent early signatures of motor dysfunction in SMA mice (21,37,42,43). These findings highlight a direct pathogenic role of U12 splicing dysregulation in mouse models of SMA and reveal the functional requirement of minor snRNPs for the maintenance of synaptic integrity in mammalian sensory-motor circuits.

Correction of SMN-dependent U12 splicing defects in mammalian cells by viral-mediated expression of minor snRNAs.
To investigate the role of U12 splicing dysfunction in SMA, we constructed lentiviral and selfcomplementary adeno-associated virus serotype 9 (AAV9) vectors that contained expression cassettes for human U11, U12 and U4atac snRNAs under the control of the human U2 promoter (44) ( Figure 1A). For initial validation of vector-derived minor snRNA overexpression, HeLa and 293T cells as well as fibroblasts from a type I SMA patient (GM03813) (45) were transduced with the U11/U12/U4atac-expressing lentivirus and RNA was extracted 72 hours later for RT-PCR analysis ( Figure 1B). The 5S ribosomal RNA (rRNA) as well as U1 and U2 major snRNAs were utilized as internal controls. We found a robust increase in the expression of human U11, U12 and U4atac snRNAs above the endogenous levels in each of the human cell types examined ( Figure 1B). Moreover, the expression of endogenous U1, U2, and 5S rRNA was unchanged, indicating that the overexpression of minor snRNAs was specific and did not alter the levels of major snRNAs. To determine the incorporation of vector-derived minor snRNAs into snRNPs, we transiently transfected 293T cells with the U11/U12/U4atac lentiviral construct and performed coimmunoprecipitation experiments with anti-SmB antibodies under stringent conditions followed by RT-qPCR analysis ( Figure 1C-D). The increased amounts of immunoprecipitated snRNAs from U11/U12/U4atac transfected cells relative to mock-transfected controls are consistent with proper formation of the Sm core on the overexpressed minor snRNAs.
Next, we sought to determine the effects of minor snRNA overexpression on U12 splicing in a setting of SMN deficiency. To do so, we employed previously characterized mouse NIH3T3-Smn RNAi fibroblasts in which doxycycline (Dox) inducible RNAi knockdown of endogenous Smn causes snRNA reduction and U12 splicing impairment (21). Following transduction with the U11/U12/U4atac lentivirus or mock treatment, NIH3T3-SmnRNAi cells were cultured for 5 days in the presence of Dox to induce knockdown of Smn. RT-qPCR analysis showed that the efficient reduction of Smn mRNA expression upon Dox treatment was not altered by transduction with the U11/U12/U4atac lentivirus (Figure 2A). In contrast, the reduction in the steady-state levels of U11, U12 and U4atac snRNAs induced by SMN deficiency in NIH3T3 was corrected by lentiviral-mediated expression of minor snRNAs, with U11 and U12 being similar to, and U4atac exceeding, the levels found in control cells with normal Smn ( Figure 2B).
Having established a means to restore minor snRNA levels, we then investigated their potential effects in correcting U12 splicing dysfunction in SMN-deficient NIH3T3 cells. To do so, we analyzed representative U12-dependent mRNA processing events that we have previously demonstrated to be dysregulated in response to SMN deficiency in NIH3T3 cells (21). These include increased retention of the U12 intron in the Tetraspanin 31 (Tspan31) mRNA, accumulation of an aberrantly spliced Stasimon mRNA due to activation of a cryptic 5' splice site and an alternatively spliced Chloride Voltage-Gated Channel 7 (Clcn7) mRNA in which both exons flanking the U12 intron are skipped due to inefficient U12 splicing ( Figure 2C). Consistent with our previous study (21), RT-qPCR analysis of these target mRNAs confirmed U12 splicing dysfunction in SMN-deficient NIH3T3 cells relative to control cells ( Figure 2C). Importantly, all these splicing defects were strongly corrected by transduction of the U11/U12/U4atac lentivirus in SMNdeficient NIH3T3 cells ( Figure 2C).
Together, these results indicate that viral-mediated gene delivery is an effective means to counteract the reduction in the levels of U11, U12 and U4atac minor snRNAs and improve U12 splicing dysfunction induced by SMN deficiency in mammalian cells.

AAV9-mediated minor snRNA gene delivery improves Stasimon U12 splicing in SMA mice
We sought to test our viral-based approach of snRNA gene delivery in vivo in order to investigate the contribution of U12 splicing dysfunction in the SMNΔ7 mouse model of severe SMA. These mice are knockout for the mouse Smn gene and harbor two copies of the human SMN2 gene as well as multiple copies of a human SMN∆7 transgene (46). SMNΔ7 SMA mice display reduced weight gain, severe motor dysfunction and death by the second postnatal week. We employed intracerebroventricular (ICV) injection of ~1×10 11 viral genome particles (vgp) of either AAV9-U11/U12 or AAV9-U11/U12/U4atac in SMNΔ7 SMA mice at postnatal day 1 (P1) as previously described (47)(48)(49). We then isolated spinal cords from AAV9-injected SMA mice as well as untreated SMA mice and unaffected heterozygous mice at P9 and measured the levels of U11, U12 and U4atac snRNAs by RT-qPCR. Given that primers that would specifically and selectively detect individual snRNAs from one species but not the other could not be designed; we relied on primers that detect both endogenous (mouse) and viral-expressed (human) snRNAs with identical efficiency (Supplementary Table 1). Consistent with previous studies (19,33,34); we found a reduction in the levels of U11, U12 and U4atac in the spinal cord of untreated SMA relative to unaffected controls. However, we did not find an increase in the levels of these snRNAs in SMA injected with AAV9-U11/U12 or AAV9-U11/U12/U4atac ( Figure 3A). While seemingly surprising, the lack of an increase at whole tissue levels likely reflects the tropism of AAV9 which is highly effective in transducing motor We next investigated whether injection of AAV9-U11/U12 or AAV9-U11/U12/U4atac could correct defective U12 splicing of Stasimon mRNA in the spinal cord of SMA mice. Consistent with previous study (21), an aberrantly spliced Stasimon mRNA isoform accumulates due to impaired excision of the U12 intron in the spinal cord of untreated SMA mice relative to unaffected control mice ( Figure 3B). This splicing defect was significantly corrected in SMA mice injected with AAV9-U11/U12 or AAV9-U11/U12/U4atac as well as upon injection of AAV9-SMN ( Figure 3B), which was used as a positive control. To expand our analysis to other U12 splicing events dysregulated in SMA, we investigated three genes (Myh9, Myo10 and Rasgrp3) for which increased U12 intron retention was previously validated in the spinal cord of a different mouse model of SMA (36). Interestingly, while we confirmed defective splicing of the U12 introns of Myh9, Myo10 and Rasgrp3 mRNAs in the spinal cord of SMNΔ7 SMA mice relative to unaffected controls, these defects were not corrected by treatment with either AAV9-U11/U12/U4atac or AAV9-SMN (Supplementary Figure S2B). These unexpected results suggest that these U12 splicing changes occur in spinal cells that are not efficiently transduced by AAV9 and that dysregulation of these genes does not significantly contribute to SMA pathology.
Given that Stasimon missplicing is prominent in SMA proprioceptive neurons that reside in the dorsal root ganglia (DRGs) (21), we analyzed lumbar DRGs isolated from SMA mice treated with AAV9-U11/U12 and AAV9-SMN relative to untreated controls (Supplementary Figure 2). Previous studies have documented highly efficient transduction of DRG neurons including proprioceptive sensory neurons (43,51). We found here that the strong accumulation of aberrantly spliced Stasimon mRNA in SMA DRGs was markedly reduced by minor snRNA gene delivery and fully corrected by SMN restoration (Supplementary To determine the selectivity of the effects, we also monitored well-established mRNA processing events in other SMN-regulated RNA pathways that are disrupted in SMA but unrelated to U12 splicing. Specifically, we analyzed the U7-dependent 3'-end processing of histone H1c mRNA (52) and the expression of Cdkn1a mRNA that is mediated by p53 activation through dysregulation of Mdm2 and Mdm4 alternative splicing (41,53). The accumulation of both 3'-end extended H1c mRNA and Cdkn1a mRNA induced by SMN deficiency in the spinal cord of SMA mice was not corrected by AAV9-U11/U12 or AAV9-U11/U12/U4atac, while it was normalized by treatment with AAV9-SMN as expected ( Figure 3B), indicating specificity of the effects of minor snRNA gene delivery in improving U12 splicing of Stasimon in SMA mice.
Lastly, it was important to establish that SMN levels were not elevated following AAV9-mediated minor snRNA gene delivery in SMA mice. Therefore, we harvested disease-relevant tissues from treated mice at P9 and performed Western blot analysis. Compared to untreated SMA samples, SMN protein expression in brain, spinal cord and skeletal muscle (gastrocnemius) was not increased in AAV9-treated relative to untreated SMA mice ( Figure 4). These results demonstrate that treatment with either AAV9-U11/U12 or AAV9-U11/U12/U4atac vectors does not elevate SMN protein levels in SMA mice, indicating that the improvement in U12 splicing is independent of SMN induction.

Minor snRNAs improve motor function and extend survival in severe and intermediate SMA mouse models
We next investigated the phenotypic effects of AAV9-U11/U12 and AAV9-U11/U12/U4atac in SMNΔ7 SMA mice following a single ICV injection of ~1×10 11 vgp at P1. To assess any potential toxicity associated with this treatment, similar injections were also performed in unaffected, heterozygous littermates. Moreover, AAV9 expressing full-length human SMN was used as a positive control that induces robust rescue of the SMA phenotype as reported previously (54)(55)(56). Compared to the untreated SMA mice with a median survival of 11 days ( Figure 5A), the introduction of U11 and U12 snRNAs extended life span to a median of 17 days, with the longest-lived animals reaching 20 days ( Figure 5A).
Similar results were observed with the AAV9-U11/U12/U4atac vector, where the median survival was extended to 16 days and a maximal life span of 20 days ( Figure 5A). Treatment with each of the vectors also increased total body weight compared to untreated SMA animals over a time course analysis and when measured as the percent weight gain relative to birth weight ( Figure 5B and 5C).
Motor behavior was assessed by measuring the righting reflex from a prone position, an assay used in this model as a reliable measurement of motor function (48,57). This analysis started at P6 and motor performance was recorded daily for each group continuously throughout the life span. As expected, unaffected controls performed much better than untreated SMNΔ7 SMA mice and the AAV9-SMN treated SMA animals were able to right as quickly as the unaffected animals by P17 ( Figure 5D). Treated SMA mice in the U11/U12 and U11/U12/U4atac cohorts showed improved righting reflexes and were able to right themselves faster starting at P12, at which time the untreated SMA animals were no longer able to stand ( Figure 5D). These results demonstrate that delivery of minor snRNAs significantly improves important parameters of disease in a severe mouse model of SMA. Although the overall phenotypic benefit is relatively moderate, a complete rescue would not be expected because there are several other pathways regulated by SMN that contribute to SMA pathology that are not targeted for correction with our approach (37,39,41,52).
To assess the role of minor snRNAs in a milder disease context, we next investigated their effects in Smn 2B/mice, which are a well-described intermediate model of SMA harboring one Smn knockout allele and one hypomorphic Smn allele in which exon 7 is mostly skipped due to the mutation of an exonic splicing enhancer (58). These SMA mice have a delayed onset of neuromuscular pathology and motor dysfunction appears in the third postnatal week (59). First, we analyzed the expression of minor snRNAs, which was not previously described in this mouse model, and found significant reduction in the levels of U11 and U12 but not U4atac in the spinal cord of Smn 2B/mice relative to WT controls at P18 (Supplementary Figure   S4A). We then analyzed the same representative SMN-dependent mRNA processing events that are dysregulated in SMN∆7 SMA mice ( Figure 3B) and found that Stasimon's U12 splicing, histone H1c  Figure S4B). Nevertheless, the disruption of Stasimon splicing is conserved across models and provides evidence for defective U12 splicing in Smn 2B/mice. Next, we injected Smn 2B/mice with AAV9 carrying either U11/U12 or U11/U12/U4atac minor snRNAs (~1×10 11 vgp) at P2 by ICV delivery. The median survival for untreated Smn 2B/mice was 31 days, whereas the median survival of Smn 2B/mice treated with U11/U12 or U11/U12/U4atac snRNAs was substantially extended to 39 and 41 days, respectively ( Figure 5E). Moreover, the introduction of these minor snRNAs improved the average weight gain when compared with the untreated SMA animals from 40 days of age onward ( Figure 5F).
Together, these results indicated that minor snRNAs contribute to the disease phenotype in a milder SMA mouse model.

Minor snRNAs do not prevent neuromuscular pathology in severe SMA mice.
To identify the cellular basis for the improvement of AAV9-mediated delivery of minor snRNAs on motor behavior of SMA mice, we sought to investigate the effects on several morphological parameters of the motor circuit that are strongly affected by SMN deficiency and contribute to motor dysfunction in the SMN∆7 severe model of SMA (60,61). First, we focused on the degeneration of vulnerable SMA motor neurons, a hallmark of the disease in both patients and mouse models. To do so, we delivered AAV9-U11/U12 and AAV9-U11/U12/U4atac in SMA mice by ICV injection at P1 and quantified the total number of vulnerable motor neurons that reside in the medial motor column (MMC) of the lumbar L5 segment of the spinal cord at P11 using ChAT immunohistochemistry and confocal microscopy as previously described (41) (Figure 6). Consistent with previous studies (41,42,53), we found that approximately 40% of L5 MMC motor neurons are lost in untreated SMA mice relative to unaffected controls at P11 ( Figure 6A).
Moreover, we found that neither AAV9-U11/U12 nor AAV9-U11/U12/U4atac was able to prevent motor neuron loss in SMA mice ( Figure 6B). We recently showed that Stasimon dysfunction contributes to the degeneration of SMA motor neurons by promoting p38MAPK-mediated phosphorylation of p53 (37). It is possible that a partial degree of correction of Stasimon's splicing by minor snRNA expression in motor neurons was insufficient to prevent induction of the neurodegenerative process.
Neuromuscular junction (NMJ) pathology has been well characterized in the SMNΔ7 mouse model of SMA, demonstrating that there are subsets of muscle groups particularly vulnerable to SMN deficiency (62)(63)(64). In order to examine the effects of our treatments on NMJ phenotypes, mice of each experimental group were perfused on P12 and the longissimus capitis, a vulnerable muscle implicated in the control of head stability and movement, was isolated and analyzed ( Figure 6C). We found that the longissimus capitis from SMA mice injected with AAV9-U11/U12 or AAV9-U11/U12/U4atac showed no significant improvement in the proportion of denervated and partially innervated NMJs when compared to the same muscle from untreated SMA controls ( Figure 6D). Furthermore, we examined additional aspects of NMJ development by analyzing endplate morphology in the sternocleidomastoid muscle, which previous studies showed to be less mature in SMA mice than in the healthy unaffected animals (65). During normal neuromuscular development, NMJs of healthy animals become progressively more complex, increase in size, and develop from a plaque-like structure to a perforated, pretzel-like shape, whereas the NMJs in SMA mice remain undersized and plaque-like with reduced numbers of acetylcholine receptors (62,63). We found that at P12 about 85% of the NMJs in SMA mice treated with AAV9-U11/U12 or AAV9-U11/U12/U4atac were developmentally delayed and displayed a plaque-like morphology (defined by the lack of perforations in the AChR clusters) similar to untreated SMA mice. In contrast, only about 20% of the NMJs in the healthy cohorts lacked perforations (Supplementary Figure 1). Overall, our analysis of NMJ morphology and endplate maturity showed that minor snRNA gene delivery failed to ameliorate neuromuscular pathology in SMA mice.

Minor snRNAs rescue the loss of proprioceptive synapses on motor neurons of SMA mice
Afferent innervation of motor neurons is critical for motor control and this element of the motor circuit is severely disrupted in SMA (60,61). In particular, the loss of VGluT1 + synapses on SMA motor neurons has been well-characterized both morphologically and functionally in SMNΔ7 mice (42,43) (62,66,67).
Additionally, it is well established that all the VGluT1 + synapses onto somata and dendrites of motor neurons are exclusively derived from proprioceptive afferent neurons (68)(69)(70). In order to determine the effect of minor snRNAs on the loss of these excitatory sensory synapses, we quantified VGluT1 + proprioceptive boutons juxtaposed to the somata of ChAT + motor neurons located in the L3-L5 lumbar segments of the spinal cord of SMA mice at P12 ( Figure 7A). Consistent with previous studies, we found that untreated SMA animals had a strong reduction in the number of VGluT1 + synapses onto motor neuron somata compared with unaffected controls ( Figure 7B). Remarkably, we observed a robust rescue in the number of proprioceptive synapses impinging on motor neurons of SMA mice treated with AAV9-U11/U12 or AAV9-U11/U12/U4atac compared to untreated SMA controls ( Figure 7A and 7B). Furthermore, our results indicated that AAV9-mediated expression of either U11/U12 or U11/U12/U4atac snRNAs was equally effective in preventing the stripping of proprioceptive SMA synapses ( Figure 7B). Collectively, these results implicate minor snRNA dysfunction in the process underlying the loss of sensory synapses on motor neurons of SMA mice.

Discussion
It is increasingly evident that SMN has multiple functions in RNA metabolism and plays a central role in gene regulation (71,72). However, the specific contribution of individual SMN-dependent RNA pathways to SMA pathology is only beginning to emerge. Identifying which RNA pathways and downstream genes -among many potentially dysregulated events -are directly relevant to the disease phenotype is critical to elucidate molecular mechanisms and may help uncover therapeutic approaches distinct from SMN upregulation. This study aimed to address these issues by investigating the contribution of reduced assembly of minor snRNPs and dysregulated U12 splicing to SMA pathology in mouse models of the disease, which provide the best recapitulation of the human condition both genetically and phenotypically (26, 60). We utilized AAV9-mediated gene delivery of minor snRNAs in SMA mice to demonstrate functional improvement in the splicing of U12 introns as well as moderate but significant phenotypic benefit, including extension of survival and enhanced motor function independent of any alteration of SMN levels. Importantly, these effects were associated with a robust rescue of the loss of proprioceptive sensory synapses impinging on SMA motor neurons, an early signature of motor dysfunction in mouse models of the disease (42,43,62). Taken together, these findings reveal that U12 splicing dysfunction contributes directly to the loss of synapses and altered sensory-motor connectivity induced by SMN deficiency, providing new insight into the RNA-mediated mechanisms of SMA pathogenesis.
This work was based on the premise that phenotypic benefit from selective restoration of individual RNA pathways regulated by SMN in a mouse model of the disease would provide conclusive evidence of their involvement in the etiology of SMA. Additionally, correction of molecular events that are directly dependent on the biological function(s) of SMN would provide confidence that the most proximal steps in the pathogenic cascade are being targeted. To date, however, none of the RNA pathways implicated in SMA pathology have been evaluated for their specific contribution to the disease phenotype in mouse models, partly due to the lack of adequate tools for uncoupling distinct SMN functions and testing their respective role in disease pathogenesis. Here, we addressed this issue by focusing on the well-established activity of SMN in snRNP assembly and in particular on the biogenesis of minor snRNPs of the U12 spliceosome, which are prominently reduced in SMA mouse models (19,33,34). We developed a means to selectively enhance U12 splicing in the setting of SMN deficiency through viral-mediated minor snRNA gene delivery. Our approach is based on the reasoning that overexpression of minor snRNAs, which are about one hundred-fold less abundant than major snRNAs (2, 3), could tilt the equilibrium in favor of their increased incorporation into functional snRNPs even under conditions in which SMN-mediated assembly is severely reduced. As proof-of-concept for the validity of this approach, we show specific elevation of minor snRNA expression levels following lentiviral transduction of U11, U12 and U4atac genes in various cultured mammalian cells with either normal or reduced SMN levels (Figures 1 and 2B). Using a previously established cell model system in NIH3T3 fibroblasts (21), we further show that this approach was effective not only in counteracting the reduction in minor snRNAs but also in correcting U12-dependent splicing deficits induced by SMN deficiency (Figure 2C), thereby validating minor snRNA gene delivery as a suitable approach to selectively improve the functionality of this SMN-dependent RNA pathway.
Building on the in vitro results with mammalian cells, we went on to show that AAV9-mediated delivery of minor snRNA genes in SMA mice improves defective U12 splicing of the Stasimon mRNA in the spinal cord and lumbar DRGs of severe SMA mice ( Figure 3B and Supplementary Figure 2), which we previously associated with sensory-motor circuit dysfunction in several animal model of SMA (21,37).
The effects are specific and unrelated to SMN induction because other SMN-regulated RNA events such as histone mRNA 3'-end processing (52) and p53-mediated Cdkn1a mRNA upregulation (41,53) were not corrected ( Figure 3B), and SMN levels did not change ( Figure 4). Interestingly, the AAV9 vector expressing U11 and U12 was as effective as the one harboring U11, U12 and U4atac on Stasimon pre-mRNA splicing in vivo, suggesting that the SMN-dependent reduction of U11 and U12 snRNAs have a greater functional impact than U4atac. U11 and U12 snRNPs associate with U12 introns as a preformed di-snRNP complex that simultaneously binds the 5' splice site and branch point sequence (73) and contribute to exon definition (15). Defects in these processes due to reduced availability of U11/U12 di-snRNP are consistent with the observed minor splicing abnormalities induced by SMN deficiency resulting from disruption of the initial recognition of U12 introns.
Our results point to a direct contribution of U12 splicing dysfunction to SMA pathology in mouse models of the disease with varying degree of severity as we found that minor snRNA gene delivery ameliorated disease phenotypes of severe and milder SMA mice ( Figure 5). There is a modest but significant extension of survival and weight gain from birth to peak as well as a transient improvement of motor function in severe SMA mice. Increased survival and enhanced weight gain are also found in a milder model of SMA. Consistent with these results, we show that SMN deficiency decreases the levels of U11 and U12 snRNAs as well as disrupts U12 splicing of Stasimon mRNA in the spinal cord of milder SMA mice (Supplementary Figure 3), providing direct evidence for dysregulation of this RNA pathway that was previously investigated only in severe mouse models of SMA (19,21,35,36,39). Moreover, overall similar phenotypic benefit was observed in SMA mice injected with either AAV9-U11/U12 or AAV9-U11/U12/U4atac, as is the case for Stasimon splicing correction. In agreement with other studies in which specific aspects of SMA pathology have been corrected independent of SMN upregulation (41,53,74), it is not surprising that the degree of phenotypic benefit associated with minor snRNA delivery is relatively limited and wears out as disease pathogenesis progresses towards the end stage. This likely reflects the multifactorial nature of SMA pathology (75,76) and the involvement of other SMN-dependent RNA pathways beyond U12 splicing (71,72). Additionally, our approach relies on the low levels of functional SMN present in SMA that likely limits the amount of minor snRNPs that can assemble following AAV9 treatment. Therefore, it is also possible that our results underestimate the potential contribution of U12 splicing dysfunction to SMA pathology due to incomplete correction of this RNA pathway as compared to what is accomplished with SMN restoration.
We highlight a functional requirement of minor snRNPs for the maintenance of synaptic integrity in mammalian sensory-motor circuits, the disruption of which contributes to SMA. Previous studies have characterized several defects induced by SMN deficiency in the motor circuit of mouse models that contribute to neuromuscular pathology and are clinically relevant features of SMA (60,61), including dysfunction and loss of specific synapses as well as death of select motor neuron pools. Here we looked at their potential link with U12 splicing dysfunction and found that it specifically contributed to disrupt sensory-motor synaptic connectivity in severe SMA mice (Figures 6 and 7). The reduction of VGluT1 + proprioceptive synapses on spinal motor neurons is an early pathogenic event in SMA mice (42,62).
Moreover, dysfunction and loss of sensory synapses is a cell autonomous process caused by SMN deficiency within proprioceptive neurons that reduces the excitatory drive on SMA motor neurons (43), resulting in motor neuron hyper-excitability, altered firing and compromised muscle contraction.
Remarkably, we found that minor snRNA gene delivery in SMA mice robustly rescued the number of VGluT1 + synapses on the somata of lumbar motor neurons relative to untreated SMA mice, thereby linking U12 splicing dysfunction to motor neuron deafferentation (Figure 7). It is plausible that preservation of proprioceptive synapses contributes to benefits in motor function of severe SMA mice and is mediated by the effects of minor snRNAs on correcting splicing of Stasimon -an ER-resident transmembrane protein implicated in the regulation of autophagy (77-80) -in proprioceptive DRG neurons. This interpretation is consistent with early studies in which we identified Stasimon as a U12 intron-containing gene essential for sensory-motor circuit function that contributes to SMN-dependent neuronal phenotypes in a Drosophila model of SMA (21) as well as our recent results with AAV9-mediated Stasimon restoration in SMA mice (37). Additionally, the dysregulation of Stasimon splicing that we report here in milder SMA mice in which the loss of VGluT1 + synapses onto motor neurons has previously been documented (81) suggests that Stasimon dysfunction may contribute to deafferentation across mouse models of the disease. In contrast to minor snRNA gene delivery, however, AAV9-Stasimon overexpression does not increase survival or weight gain in severe SMA mice (82). It is therefore likely that dysregulation of U12 intron-containing genes other than Stasimon might be involved in distinct aspects of SMA pathology as recently suggested (36). Interestingly, we tested a handful of other U12 intron-containing genes dysregulated in SMA mice but found that, in contrast to Stasimon mRNA, they were neither corrected by minor snRNA gene delivery nor by AAV9-SMN. These results suggest that, while Stasimon is misspliced in disease-relevant neurons (21), these other changes occur in other spinal cells that are not efficiently transduced by AAV9 and unlikely contribute to SMA pathology. Future studies are needed to reveal the identity and potential role in disease of other U12 intron-containing genes regulated by SMN.
In summary, beyond mutations genetically linking minor snRNPs to human disease (9), U12 splicing dysfunction has been implicated in the pathogenesis of motor neuron diseases such as SMA (19-21, 35, 36) and ALS (22-24), but direct evidence for a functional involvement of this pathway in the disease process has been missing. Through selective enhancement of minor snRNA function, our study supports the conclusion that defects in U12 splicing play a role in SMA pathogenesis and expands the range of human disorders associated with dysregulation of this RNA processing pathway.

Animal studies
Research animals were housed and treated in accordance with the guidelines of the Animal Care and Use Committees at the University of Missouri and at Columbia University following the regulations established by the National Institute of Health's Guide for the Care and Use of Laboratory Animals (83).
Additional information regarding Methods can be found in the Supplemental section.        Tissues for Western blot analysis were collected from brain, spinal cord and the gastrocnemius muscle.       proprioceptive synapses compared to the untreated SMA mice. One-way ANOVA was applied where significance is represented by *** = P ≤ 0.001.