Restoration of SMN in Schwann cells reverses myelination defects and improves neuromuscular function in spinal muscular atrophy

Spinal muscular atrophy (SMA) is a neuromuscular disease caused by low levels of SMN protein, primarily affecting lower motor neurons. Recent evidence from SMA and related conditions suggests that glial cells can influence disease severity. Here, we investigated the role of glial cells in the peripheral nervous system by creating SMA mice selectively overexpressing SMN in myelinating Schwann cells (Smn−/−;SMN2tg/0;SMN1SC). Restoration of SMN protein levels restricted solely to Schwann cells reversed myelination defects, significantly improved neuromuscular function and ameliorated neuromuscular junction pathology in SMA mice. However, restoration of SMN in Schwann cells had no impact on motor neuron soma loss from the spinal cord or ongoing systemic and peripheral pathology. This study provides evidence for a defined, intrinsic contribution of glial cells to SMA disease pathogenesis and suggests that therapies designed to include Schwann cells in their target tissues are likely to be required in order to rescue myelination defects and associated disease symptoms.


Introduction
. SMN2 therefore produces considerably less full-length SMN protein than SMN1 and cannot fully compensate for its loss (4).
Motor neurons are particularly vulnerable to low levels of SMN (4). However, with the emergence of SMA animal models and a clearer appreciation of disease pathology in human patients, it has become apparent that reduced levels of SMN have additional effects on multiple other cell-types and tissues (7). For example, skeletal muscle cells in cultures from severe SMA patients show defective development and are disorganized (8,9); cardiac arrhythmias and abnormalities are one of the most common peripheral phenotypes reported in patients with severe SMA (10); while vascular (11)(12)(13)(14)(15), liver (16,17), metabolic (18) and consistent intestinal and lung (13) abnormalities have also been reported.
In addition, there is a growing evidence from studies of related motor neuron diseases, such as amyotrophic lateral sclerosis (ALS), that non-neuronal cells can play a defining role in regulating motor neuron degeneration (19). This may have particular relevance for SMA in light of recent studies showing that restoration of SMN in motor neurons had only modest effects on the SMA phenotype (20)(21)(22)(23).
An evidence of gliosis in SMA patients has been known for some time (24)(25)(26) and recent work has suggested that astrocytes in particular may contribute to SMA pathogenesis in the central nervous system (CNS). For example, morphological and cellular alterations in astrocytes occur prior to the loss of motor neurons (27). Similar evidence for astrogliosis has been presented from post-mortem patient spinal cords (28). Glial cells from the peripheral nervous system (PNS) have also been implicated in SMA, with changes observed in non-myelinating terminal Schwann cells in several different SMA mouse models (29)(30)(31). Similarly, in vivo and in vitro analyses of myelinating Schwann cells demonstrated intrinsic defects in SMA (evidenced by SMN-dependent failure of isolated SMA-derived Schwann cells to respond to myelination cues), leading to defects in myelination and generation of extracellular matrix in peripheral nerve (32,33).
Despite accumulating evidence for intrinsic pathology of Schwann cells in SMA, it remains unknown whether Schwann cells directly influence motor neuron pathology. Answering this question will be important, not only to better understand the contribution of cell autonomous and non-cell autonomous pathways to motor neuron degeneration in SMA, but also to identify critical targets for delivering effective therapies for the treatment of SMA.
FLAG-SMN (1 kDa larger than SMN) was expressed in peripheral nerve, but not in a range of other tissues (muscle, brain, heart, liver and spinal cord) in the GO40 line, indicating that FLAG-SMN expression was indeed limited to the PNS, as expected (Fig. 1D). PNS expression was confirmed by immunohistochemical analyses of sectioned and teased nerve fibres from sciatic nerve; an antibody to the 5 0 FLAG tag on the transgene demonstrated clear expression of FLAG-SMN in sciatic nerve costained with S100, a Schwann cell marker (Fig. 1E).

Over-expression of SMN in myelinating Schwann cells does not affect normal neuromuscular development
Next, we wanted to confirm that overexpression of SMN in myelinating Schwann cells was not detrimental to normal neuromuscular development. At the gross anatomical level, mice expressing SMN1 SC were indistinguishable from control animals, with similar survival (Fig. 4H) and breeding rates (data not shown). We confirmed that SMN was significantly overexpressed in SMN1 SC positive control mice (control þ SMN1 SC ) compared with wild-type mice (control) ($2000-fold) ( Fig. 2E and F). To investigate potential effects at the cellular level, we carried out an assessment of intercostal nerves from P7 mice expressing the SMN1 transgene. Measurement of axon thickness and axon plus myelin thickness allows a G-ratio to be calculated. G-Ratio measurements enable assessment of myelin thickness with increased G-ratios indicating a reduction in average myelin sheath thickness relative to the axon diameter. There was no qualitative or quantitative difference in myelin sheath thickness in mice expressing SMN1 SC ( Fig. 2A and B). There was also no change in the number of non-myelinated large diameter (>1 mm) axons ( Fig. 2C) or animal weight (Fig. 2D). Thus, supra-physiological levels of SMN protein are well tolerated by Schwann cells, with no overt impact on normal peripheral nerve myelination.
Following confirmation that the GO40 line expressed SMN as expected and that there were no detrimental effects because of overexpression of SMN in myelinating Schwann cells, we backcrossed these mice onto a FVB/N background until 98.4% homozygosity was reached. Experimental 'rescue' animals (mice lacking endogeneous Smn1, heterozygous for SMN2 and expressing SMN1 in Schwann cells; Smn À/À ;SMN2 tg/0 ;SMN1 SC or SMA þ SMN1 SC ) were then generated by crossing SMN1 SC mice with the severe 'Taiwanese' SMA line (Smn À/À ;SMN2 tg/tg or SMA) (34À36). We confirmed that SMN was significantly overexpressed in SMN1 SC positive mice ($2000-fold in control þ SMN1 SC ; $3000-fold in SMA þ SMN1 SC mice compared with non-SMN1 SC carrying control and SMA mice ( Fig. 2E and F).

Restoration of SMN to myelinating Schwann cells rescues myelin sheath defects in SMA mice
Previously, we reported that intercostal nerves from Taiwanese SMA mice had defective myelination, manifesting as a thinner myelin sheath and a larger number of large diameter unmyelinated axons at both mid-(P7) and late-(P11) symptomatic time points (33). To evaluate whether restoration of SMN expression in myelinating Schwann cells was sufficient to ameliorate these myelin defects we assessed intercostal nerve myelination in experimental rescue mice (SMA þ SMN1 SC ) at a mid-/late-symptomatic time point (P7). Intercostal nerves were selected for these experiments because of the findings from previous results showing that the most robust myelination defects observed in Taiwanese SMA are to be found in intercostal nerves (e.g. compared with sciatic nerve) (33).
Qualitatively, myelin thickness in SMA rescue mice appeared closer to that observed in control littermates than SMA mice (Fig. 3A). Quantitative analyses revealed a significant rescue of the myelin sheath pathology in SMA rescue mice with a significantly reduced average G-ratio (0.656 6 0.007) compared with SMA mice (0.796 6 0.012), restored to levels observed in rescue SMA mice at P7 compared with 'Taiwanese' SMA mice (SMA), with higher G-ratios across the range of axon calibers (D). *P < 0.05, ****P < 0.0001; all tests twotailed, unpaired t-tests; N ¼ 3 mice per genotype, n > 50 independent nerve fibre measurements per genotype. Scale bar ¼ 1.5 lm (A). controls (0.670 6 0.009) (Fig. 3C). Restoration of myelin thickness was observed across a full range of axonal calibres, with no evidence for reduced axon diameters in the SMA rescue mice (Fig. 3D). In addition, there was a significant reduction in the number of non-myelinated large diameter (>1 mm) axons in SMA rescue mice (Fig. 3B).
Taken together, these findings support the hypothesis that myelination defects in SMA result from direct, intrinsic deficiencies in Schwann cells, rather than occurring as a secondary consequence to pathology in neighbouring motor neurons (32,33).
Restoration of SMN to myelinating Schwann cells improves neuromuscular function and ameliorates neuromuscular junction pathology without affecting motor neuron soma loss or systemic pathology We next wanted to establish whether restoring SMN levels in Schwann cells could have any impact on neuromuscular function and/or motor neuron pathology. Initial tests of neuromuscular function [using an established 'righting test' (37)] revealed that SMA rescue mice (SMA þ SMN1 SC ) had a clear functional improvement at a mid-/late-symptomatic time-point (P7) compared with SMA mice (Fig. 4A). Although not maintained through to the end-stage of disease, SMA rescue mice also had modest, but significant, improvements in weight at P6, P7 and P8 (Fig. 4B).
Neuromuscular junction (NMJ) pathology is a major hallmark of motor neuron degeneration in SMA, with denervation and delayed motor endplate maturation being observed in many different mouse models of severe SMA (38,39). However, in our hands, the only robust NMJ pathology observed in the 'Taiwanese' mouse model of SMA is a reduced number of axonal inputs to NMJs, particularly notable in the levator auris longus (LAL) muscle (37). We therefore wanted to establish whether the improvements in neuromuscular function we observed in SMA rescue mice (SMA þ SMN1 SC ) correlated with an amelioration of this aspect of NMJ pathology. We found a significant increase in the number of axonal inputs to NMJs in the LAL of SMA rescue mice, with the average number of inputs restored to levels observed in controls (P8/9; Fig. 4C and D). Interestingly, these improvements in NMJ pathology occurred in the absence of any parallel improvement in skeletal muscle, with no rescue in the smaller muscle fibre diameters (from the LAL muscle) in the SMA rescue mice compared with SMA mice (P8/9; Fig. 4E).
Improvements in neuromuscular function and NMJ pathology observed in SMA rescue mice (SMA þ SMN1 SC ) occurred in the absence of any significant improvement in levels of motor neuron soma loss from the ventral horn of spinal cord (Fig. 4F and G). The number of motor neurons in the ventral horn of spinal cord was reduced by $25% in SMA mice compared with controls, as previously reported (40). However, similar reduced numbers of motor neurons were also observed in the SMA rescue mice (SMA þ SMN1 SC ).
Previous reports have suggested that restoration of SMN solely in motor neurons improves motor neuron pathology, but only modestly influences systemic SMA pathology (20)(21)(22)(23). We therefore wanted to establish whether restoring SMN in myelinating Schwann cells, with concurrent improvements in myelination, neuromuscular function and NMJ pathology, resulted in any improvements in systemic pathology. Restoration of SMN in myelinating Schwann cells did not lead to any increase in survival or improvements in the gross appearance of SMA mice (Fig. 4H, data not shown), indicating selective effects of targeting Schwann cells on glial/motor neuron pathology and resulting neuromuscular function.

Discussion
In this study we have demonstrated that selective restoration of SMN in Schwann cells can lead to marked improvements in selective aspects of disease pathogenesis in a mouse model of severe SMA. Elevated levels of SMN in Schwann cells were well tolerated and did not influence normal Schwann cell or neuromuscular development. Restoring SMN in Schwann cells was sufficient to reverse the myelination defects that occur in SMA, suggesting that these occur because of an intrinsic defect in the glia cells, rather than as a secondary consequence of degenerative events occurring in neighbouring motor neurons. Alongside improvements in myelination, elevated levels of SMN in Schwann cells improved neuromuscular function and aspects of NMJ pathology. This indicates that at least some of the pathological changes and functional defects occurring in the neuromuscular system during severe forms of SMA are a consequence of defective glial cells. Whether such defects are also occurring in less severe forms of the disease needs to be established, and should be aided by the recent development of milder mouse models of SMA (e.g. SMN2B/À mice).
One interesting observation made during the current study was that neuromuscular function and NMJ pathology were ameliorated in the absence of overt improvements in motor neuron loss or skeletal muscle fibre atrophy. This suggests that improvements in some aspects of neuromuscular pathology in SMA can be generated by interventions that do not directly target motor neurons and/or skeletal muscle. Thus, improving the health of myelinating Schwann cells alone would be predicted to have a significant impact on NMJ stability of neighbouring motor neurons, and also improve neuromuscular function in SMA patients (41). The precise mechanisms through which Schwann cells can influence morphological and functional aspects of neuromuscular connectivity in health and disease are only beginning to be uncovered. However, our findings in SMA are consistent with the recent demonstration that robust Schwann cell-motor neuron interactions are required in order to sculpt appropriate synaptic connectivity at the NMJ during development in neonatal mice (42).
Our findings provide significant additional experimental support for a model whereby SMA represents a complex, multifactorial disease, where disease symptoms cannot simply be ascribed to changes occurring solely in motor neurons and skeletal muscle fibres (7,43). Thus, therapeutic targeting of motor neurons alone is unlikely to be sufficient to ameliorate all aspects of SMA disease pathogenesis. For example, we previously reported significant amelioration of the neuromuscular phenotype, with no apparent influence on systemic pathology, in SMA mice following delivery of a b-catenin inhibitor (37). Moreover, viral gene therapy approaches used to restore SMN to astrocytes in the D7 model of SMA reported significant improvement in motor neuron-associated pathology, but again only produced a modest amelioration of systemic pathology and lifespan (28). These findings are also in agreement with several studies that together suggest that restoration of SMN solely within motor neurons (returning SMN levels to those seen in wild-type healthy controls) leads to unexpectedly minor phenotypic improvements (20)(21)(22)(23). Thus, therapies designed to restore SMN levels in human SMA patients would likely benefit from the ability to target Schwann cells (and other similar cell types) alongside motor neurons, if glial cell-associated defects are also to be successfully ameliorated.
In conclusion, whilst there is currently no cure for SMA, there is increasing evidence from pre-clinical trials that viral gene therapy approaches, and antisense oligonucleotide approaches can be both efficacious and translationally viable with respect to restoring SMN levels in vivo (44). Our findings indicate that therapies capable of targeting SMN-dependent pathology in glial cells, alongside other critical disease targets, such as motor neurons and skeletal muscle, are likely to be required in order to ameliorate the full range of neuromuscular pathology observed in SMA.

pMPZ-SMN construct generation
The pMpz-SMN1 transgene was created using a two-step cloning procedure from two starting plasmids: plasmid containing the full-length human SMN1 construct and the hybrid Mpz/P0-Cx32 construct that contains a 1.1 kb fragment containing the rat Mpz promoter fused to exons 1b and 2 and the intervening intron from the human Cx32 gene (34) (Fig. 1A). Briefly, SMN1 was amplified using two overlapping forward primers that introduced NheI and AscI sites, an ATG start site and a 5 0 FLAG tag (F1, 5 0 -GCTAGCTAGCAGGCGCGCCATGGATTACAAGGATGACGACGATA AG-3 0 ; F2, 5 0 -AAGGATGACGACGATAAGGGAGGTGCGATGAGCAG CGGCGGC-3 0 ) and a reverse primer that introduced BglII and AatII sites and a TGA stop site (R, 5 0 -GGCTAAGATCTTGACG TCAATTTAAGGAATGTGAGCACCTTCC-3 0 ). Amplifed products were digested using NheI and BglII (New England Biolabs), ligated to an intermediate plasmid (pGL4.13, Promega) using T4 DNA ligase (Promega), then confirmed by sequencing. The intermediate SMN1-containing plasmid and the Mpz-promoter plasmid were digested using AscI and AatII (New England Biolabs) and ligated using T4 DNA ligase. pMpz-SMN1 constructs were confirmed by restriction digestion and sequencing.

Pronuclear injection
Fifty micrograms of pMpz-SMN1 were digested with MluI and NotI (New England Biolabs), to release the 5.8 kb transgene cassette. Transgenic injections were performed as previously described (45).

Immunohistochemical analysis of sectioned and teased nerve fibres
Sciatic nerve was fixed in 4% paraformaldehyde (PFA) for 30 min then washed in phosphate buffered saline (PBS). Prior to sectioning, nerves were embedded longitudinally in a 1:1 solution of 30% sucrose and OCT embedding matrix (CellPath). Ten micrometres of sections were prepared on slides. For nerve tease, the perineurium was carefully removed and fixed sciatic nerve fibres were teased in 0.1 M PBS on 3-amino propyltriethoxysilane (TESPA) coated slides. Slides were blocked for 0.5 h with 5% bovine serum albumin (BSA) and 0.2% Triton X-100. All nerves were incubated with primary antibodies raised against S100 (mouse, 1:100, Abcam) and FLAG (rabbit, 1:1000, ThermoFisher), then incubated in a solution of secondary antibodies (goat anti-mouse Alexa Fluor 555 and donkey anti-rabbit Alexa Fluor 488; both 1:500, Invitrogen). Slides were coverslipped using Mowiol (Calbiochem) or FluorShield (Abcam).

Non-quantitative and quantitative fluorescent western blotting
Fresh tissue was dissected and frozen on dry ice and protein extracted in RIPA buffer (ThermoScientific) with protease inhibitor cocktail (Sigma). Non-quantitative western blots were performed using a primary antibody against SMN (mouse, 1:500; BD Transduction Laboratories). Odyssey secondary antibodies were added according to the manufacturers' instructions (goat antimouse IRDye 680). Quantitative blots were carried out using a primary antibody against tubulin (mouse, 1:10 000; Abcam). Blots were imaged using an Odyssey Infrared Imaging System (Li-COR, Biosciences) at a resolution of 169 lM. Each blot was scanned and measured in triplicate to minimize user variability.

Electron microscopy
Intercostal nerves were incubated for 48 h in 4% PFA: 2.5% glutaraldehyde at 4 C before post-fixation in 1% osmium tetroxide in 0.1 M phosphate buffer for 45 min. Following dehydration through an ascending series of ethanol solutions and propylene oxide, sections were embedded on glass slides in Durcupan resin. Regions to be used for the assessment of myelination were then cut out from a randomly selected section using a scalpel and glued onto a resin block for sectioning. Ultrathin sections (60 nm) were cut and collected on formvar-coated grids (Agar Scientific, UK), stained with uranyl acetate and lead citrate in an LKB Ultrostainer and then quantitatively assessed in a Philips CM12 transmission electron microscope equipped with a Gatan digital camera. Intercostal nerve fibres were measured using ImageJ. For each individual fibre, axon diameters and Gratios were calculated as previously described (46).

Assessment of SMA rescue mice
Righting reflex tests were performed to assess neuromuscular function, as previously described (37). Mice were weighed daily and Kaplan-Meier survival analyses performed as previously described (21). Muscle fibre diameter measurements were taken from phase-contrast micrographs of teased muscle fibre preparations using ImageJ software (47). NMJ pathology was assessed on whole-mount preparations of LAL muscles (39). Motor neuron cell body counts in the spinal cord were performed as previously described (40,48).

Statistical analysis
All data were collected into Microsoft Excel and analysed using GraphPad Prism software (statistical tests used for each comparison are detailed in the text). For all statistical analyses, P < 0.05 was considered statistically significant. All data are expressed as mean 6 SEM.