|Home | About | Journals | Submit | Contact Us | Français|
The inherited motor neuron disease spinal muscular atrophy (SMA) is caused by deficient expression of survival motor neuron (SMN) protein and results in severe muscle weakness. In SMA mice, synaptic dysfunction of both neuromuscular junctions (NMJs) and central sensorimotor synapses precedes motor neuron cell death. To address whether this synaptic dysfunction is due to SMN deficiency in motor neurons, muscle, or both, we generated three lines of conditional SMA mice with tissue-specific increases in SMN expression. All three lines of mice showed increased survival, weights, and improved motor behavior. While increased SMN expression in motor neurons prevented synaptic dysfunction at the NMJ and restored motor neuron somal synapses, increased SMN expression in muscle did not affect synaptic function although it did improve myofiber size. Together these data indicate that both peripheral and central synaptic integrity are dependent on motor neurons in SMA, but SMN may have variable roles in the maintenance of these different synapses. At the NMJ, it functions at the presynaptic terminal in a cell-autonomous fashion, but may be necessary for retrograde trophic signaling to presynaptic inputs onto motor neurons. Importantly, SMN also appears to function in muscle growth and/or maintenance independent of motor neurons. Our data suggest that SMN plays distinct roles in muscle, NMJs, and motor neuron somal synapses and that restored function of SMN at all three sites will be necessary for full recovery of muscle power.
The motor neuron disease spinal muscular atrophy (SMA) causes severe muscle weakness frequently leading to mortality in infancy (Dubowitz, 1995; Crawford, 2003). SMA is caused by homozygous mutation of the survival motor neuron 1 (SMN1) gene (Lefebvre et al., 1995), but all patients retain at least one copy of the SMN2 gene. Although a minority of transcripts derived from SMN2 is spliced to incorporate exon 7, most lack exon 7 and code for an unstable, truncated protein (Lorson et al., 1998; Lorson et al., 1999; Monani et al., 1999). Thus, insufficient expression of full length SMN protein causes SMA with disease severity correlating inversely with SMN2 copy number (Feldkotter et al., 2002). The ubiquitously-expressed SMN protein mediates the assembly of small nuclear ribonuclear proteins (snRNPs), but may also enable axonal mRNA transport and local mRNA processing (Burghes and Beattie, 2009).
SMA mouse models recapitulate key aspects of severe SMA including severe weakness and early death (Le et al., 2005; Park et al., 2010a). Nonetheless, SMA mice show surprisingly limited muscle denervation and motor neuron loss overall, although particular muscle groups and motor neuron subsets are more vulnerable than others (Kong et al., 2009; Mentis et al., 2011; Ling et al., 2012). Early symptomatic stages of disease are characterized by morphological and functional abnormalities of synapses, affecting both neuromuscular junctions (NMJs) (Kariya et al., 2008; Murray et al., 2008; Kong et al., 2009; Ling et al., 2010; Ruiz et al., 2010; Dachs et al., 2011; Lee et al., 2011) and synaptic inputs to motor neurons in the spinal cord (Ling et al., 2010; Mentis et al., 2011). These abnormalities are associated with impaired maturation of NMJ endplates and myofibers (Kong et al., 2009; Lee et al., 2011) as well as loss of vesicular transporter 1 positive (VGluT1+) inputs to motor neurons (Ling et al., 2010; Park et al., 2010b; Mentis et al., 2011). Synaptic abnormalities precede motor neuron death and this period of synaptic dysfunction may define a window of disease reversibility and therapeutic responsiveness in SMA. Nonetheless, it remains unknown whether these abnormalities are dependent on reduced SMN expression in pre- or postsynaptic sides of peripheral and/or central synapses and this has important implications for sites of therapeutics delivery in SMA.
In order to understand the cellular determinants of synaptic function in SMA, we generated three lines of SMA mice expressing increased levels of SMN principally in either motor neurons or muscle. We show that NMJ synaptic function and motor neuron somal synapses are substantially improved by increased SMN expression in motor neurons, but not by increased SMN expression in muscle suggesting that synaptic function and connectivity are largely determined by motor neurons in SMA. SMN also appears to have a role in growth of muscle that is independent of its effect on motor neuron synaptic function. Our findings suggest that SMN has variable roles at distinct sites within the motor unit.
All experiments were carried out in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, and were approved by Johns Hopkins University, Wright State University, and Jackson Laboratory Animal Care Committees. ChATCre (Jackson Laboratory Stock #006410), MyoDiCre (Jackson Laboratory Stock #014140) (Kanisicak et al., 2009), and Myf5Cre (Jackson Laboratory Stock #007893) (Tallquist et al., 2000) were bred to a lacZ-Cre reporter strain, B6.129S4-Gt(ROSA)26Sortm1Sor/J (The Jackson Laboratory Stock # 003474) (Soriano, 1999) in order to analyze Cre expression patterns in 15.5 days post conception embryos (E15.5) and postnatal day 7 (P7) pups. MyoDiCre is an optimized Cre recombinase that has high recombination efficiency in myoblasts and satellite precursor cells of the embryo (Kanisicak et al., 2009; Yamamoto et al., 2009).
SMA mice expressing a Smn Cre-inducible allele (SmnRes) (Lutz et al., 2011) and 2 copies of SMN2 and SMNΔ7 alleles (SmnRes/SMN2+/+/ SMNΔ7+/+) were bred to ChATCre, MyoDiCre, or Myf5Creexpressing mice. Cre− SMA mice were those carrying the inducible Smn allele in a homozygous state (SmnRes/Res/SMN2+/+/ SMNΔ7+/+) in the absence of Cre, Cre+ SMA mice were those those carrying the inducible Smn allele in a homozygous state (SmnRes/Res/SMN2+/+/ SMNΔ7+/+) in the presence of Cre, heterozygous (Het) mice were those carrying the inducible Smn allele in a heterozygous state (SmnRes/+/SMN2+/+/ SMNΔ7+/+) in the presence of Cre, and wild type (WT) mice were those homozygous for the wild type allele (Smn+/+/SMN2+/+/ SMNΔ7+/+) in the absence of Cre.
Mice were genotyped by polymerase chain reaction (PCR) of tail DNA as previously reported (Avila et al., 2007). To detect the uninverted SmnRes allele, the following primers were used: MUT-F: 5'GGCAGTTTTAGACTCATCATGTATCTG 3'and MUT-R: 5' ACTTATGGAGATCCCTCGAGATAAC 3' yielding a 103 base pair product. To detect the inverted form of the SmnRes allele the following primers were used INV-F: 5'GTTTTAGACTCATCATGTATCTG 3' and INV-R: 5' GTGTGAGTGAACAATTCAAGCC 3' yielding a 190 base pair inverted product and/or a 143 base pair endogenous product. To determine zygosity when necessary the following primers were used to detect the murine endogenous Smn1 locus: WT-F: 5' TGGGAGTCCATCCATCCTAAGTC 3' and WT-R: 5' GCTAAGAAAATGACAATTGCACATTTG 3' yielding a 143 base pair product. To detect the presence of the Cre allele in the ChatCre mice the following primers were used: ChAT Ires WT-F: 5' GTTTGCAGAAGCGGTGGG 3', ChAT Ires WT Rev: 5' AGATAGATA ATGAGAGGCTC 3', and ChAT Cre Rev: 5' CCTTCTATCGCCTTCTTGACG 3'. To detect the Cre allele in the Myf5Cre mice the following primers were used: Myf5 WT-F: 5' CGTAGACGC CTGAAGAAGGTCAACCA 3', Myf5 WT-Rev: 5'CACATTAGAAAACCTGCCAAC ACC 3', and Myf5 Cre-Rev: 5' ACGAAGTTATTAGGTCCCTCGAC 3'. To detect Cre in MyoDiCre mice the following primers were used: MyoD WT-F: 5'CTAGGCCACAGAATTGAAAGATCT 3', MyoD WT-Rev: 5' GTAGGTGGAAATTCTAGCATCATCC 3', MyoD Cre-F: 5'GCGGATCCGAATTCGAAGTTCC 3', and MyoD Cre-Rev: 5' TGGGTC TCCAAAGCGACTCC 3'.
Daily weights were measured starting from the day of birth (P1). The average value of two trials of righting time was determined starting on P1 with a maximal time of 30 seconds per trial as previously described (Avila et al., 2007). The hind-limb suspension test was performed between P2 and P12 by suspending the mouse from the hind-limbs on the edge of a 50 ml conical tube as previously described by El-Khodor et al. (El-Khodor et al., 2008). Average position score, total number of pulls, and total latency to fall time (maximum 30 seconds) during 2 trials were determined. A composite score of this test was used (Heier et al., 2010). Ambulation index was performed between P13 and P23. The total time upright, numbers of 5.5 × 5.5 cm grid squares crossed, and number of rearings onto the hindlimbs were determined during two 60 second trials.
RNA was isolated from tissues (whole spinal cord, quadriceps muscle, or pooled lumbar level 3–5 dorsal root ganglia (DRGs) using TRIzol reagent (Invitrogen) and converted to cDNA as previously described (Avila et al., 2007). Primers (SMN6m8h) to amplify the shortened cDNA product arising from the uninverted and inverted hybrid rescue Smn allele were: Exon 6m8h Fwd: 5' GGCTACCACACTGGCTACTATATGG 3', Exon8hrev: 5'GCTTCACATTCCAGATCTGTCT3', and Exon8hprobe: 6FamCATAGAGCAGCT CTA AATGACACCACTAAAGAAtamra. Primers (SMN67m8h) used to amplify the full length cDNA product arising from the inverted Smn allele were: Exon67fwd: 5'GCTACTATATGGGTTTCAGACAAAATAAAA3', Exon8hrev: 5'GCTTCACATTCCAGATCTGTCT3', and Exon8hprobe: 6FamCATAGAGCAGCTCTA AATGACACCACTAAAGAAtamra. Primers localized to the mouse Smn exons 1 and 2 region (SMN12m) were used as an endogenous control: Exon1mfwd: 5'GCTCCGAGCAGGAAGATACG3', Exon2mrev: 5'CAATGCTGTATCATCCCAAATGTC3', and Exon1mprobe: 6FamCTG TTC CGG CGT GGC ACC Gtamra. Reactions were run in triplicate using the ABI Prism 7900 Sequence Detector System as previously described (Avila et al., 2007).
Spinal cord or quadriceps muscle tissues were homogenized in RIPA buffer (Sigma) with protease inhibitors (Roche) and sonicated. Approximately 30 μg of spinal cord or 60 μg of muscle total protein lysate was resolved on 12% Tris-Glycine gels and transferred to PVDF membranes (Invitrogen). Primary antibodies included mouse anti-SMN (1:2500; BD Transduction Laboratories), mouse anti-β-actin clone AC-15 (1:10,000, Sigma) and mouse anti-glyceraldehyde 3-phosphate dehydrogenase (GAPDH) clone 6C–5 (1:5000; Abcam) antibodies followed by a donkey anti-mouse-HRP conjugated secondary antibody (1:10000; GE Healthcare). Densitometry measures of full length SMN protein levels were performed using NIH Image software.
For assessing patterns of Cre expression, pregnant females were euthanized, the uterus dissected, and extra-embryonic tissues removed in cold PBS. Embryos were individually frozen in O.C.T. compound on dry ice. A small piece of tissue was collected from each embryo prior to freezing to confirm the genotype. For P7 tissues, mice were euthanized and brain, eye, heart, lungs, skeletal muscle (iliopsoas, tibialis anterior, and quadriceps), skin, kidney, spleen, pancreas, liver, reproductive organs, bladder, intestine, brown fat, femur, and spinal cord (cervical, thoracic, and lumbar regions) were collected and frozen in O.C.T. compound on dry ice immediately after collection. Frozen embryos and tissues were cryo-sectioned, stained with X-gal overnight at 37°C, and counter-stained with Nuclear Fast Red. LacZ expression in skeletal muscle was quantified utilizing CellProfiler imaging software. Images of stained muscle sections from 2–3 mice per strain were analyzed for color differences; in this case blue (lacZ/β-gal) and red (Fast Red). The software analyzed each tissue section to measure total area of tissue in pixels as well as the specific area expressing homogeneous β-gal staining.
For muscle and myofiber quantification, dissected muscle tissues were flash frozen in liquid nitrogen. 30 μm cryostat sections were cut and stained with hematoxylin and eosin (H&E). Myofiber morphology and diameter were assessed using a Zeiss AxioImager Z1 microscope. Myofiber diameters were determined by measuring the longest aspect of the shortest axis using Zeiss AxioVision 4.6 software.
Mice were transcardially perfused with 4% paraformaldehyde (PFA) and post fixed in PFA for 24hrs. Spinal lumbar segment 1 (L1) was dissected from the spinal cord and serial transverse sections (70 μm thick) cut using a Leica Vibratome. Sections were stained using polyclonal goat anti-ChAT (1:100; Millipore), chicken anti-β-galactosidase (1:2000; Abcam), or anti-rabbit VGluT1 (1:1000; Synaptic systems) antibodies, followed by the appropriate secondary antibodies (1:200; Jackson ImmunoResearch). Images were taken at 1 μm intervals using a Zeiss LSM Meta 510 confocal microscope. Only ChAT+ motor neurons containing a visible nucleus were counted to avoid double counting motor neurons from adjoining sections.
NMJs were examined in serial muscle sections from the tibialis anterior (TA) and splenius capitis muscles. Mice were transcardially perfused with 4% PFA and muscles were isolated, post-fixed in 4% PFA for 24 hours, cryoprotected in 30% sucrose for 24hrs, and cut longitudinally on a cryostat at 60 μm thickness. Tissue sections were labeled using monoclonal mouse anti-SMI 312 (1:1000; Covance) and polyclonal rabbit-anti-synaptophysin (1:500; Invitrogen) antibodies, followed by FITC goat anti-mouse IgG1 (1:200; Jackson Immunoresearch Laboratories), Alexa Fluor 633-conjugated goat anti-rabbit secondary antibody (1:1000; Invitrogen) as well as α-Bungarotoxin Alexafluor-555 (1:500; Invitrogen). Images were obtained using a Zeiss AxioImager Z1 microscope. Z-stack projections were made from serial scanning every 1 μm to reconstruct the NMJ. Innervation status was examined by classifying individual NMJs into one of three groups: < 25%, 25–75%, or >75% occupancy of an α-bungarotoxin-labeled endplate by synaptophysin-stained presynaptic terminal. Presynaptic neurofilament (NF) accumulation was defined as NF occupying ≥ ¼ of the post-synaptic terminal area. NMJ size was determined by measuring the area of α-bungarotoxin staining using Zeiss AxioVision 4.6 software.
Mice were transcardially perfused and post-fixed for 24hrs with 3% glutaraldehyde/4% PFA and spinal cords then dissected. The L1 segment was rinsed with 0.1 M Sorensen's buffer, post-fixed using 2% osmium tetroxide, dehydrated in serial alcohol dilutions, embedded in propylene oxide and EMbed 812 plastic (Electron Microscopy Sciences), and placed in an oven to harden into capsule form. Thick sections (1 μm) were cut on an ultramicrotome and multiple sections were placed on Superfrost slides, dried, stained with toluidine blue, and viewed by light microscopy. Thin sections were cut at 60–90 nm, placed on Formvar grids, and viewed using Tecnai 12 (FEI) and Libra 120 (Zeiss) electron microscopes. Motor neuron somata in the ventral horn area containing a visible central nucleolus and rich cytoplasm were identified and all synapses contacting the surface of these somata then imaged. At least three animals were used for each group from the ChATCre SMA line and at least four motor neurons per animal were counted for the number of boutons, number of synapses (defined as boutons juxtraposed to an electron dense postsynaptic density (PSD)), PSD length, and synaptic density per synapse at 23,000 × magnification or higher using the TEM software. Readily releasable pools were defined as those vesicles within 100 nm distance from the PSD area.
Physiology was performed on P10–P14 mouse NMJs from the TA muscle as previously described (Kong et al., 2009). Briefly, muscle was perfused with Ringer solution containing (in millimoles per liter) NaCl, 118: KCl, 3.5; CaCl2, 2; MgSO4, 0.7; NaHCO3, 26.2; NaH2PO4, 1.7; glucose, 5.5 (pH 7.3–7.4, 20–22°C) equilibrated with 95% O2 and 5% CO2. All NMJs were imaged by staining with 4-Di-2-ASP and impaled within 100 μm of the endplate. Muscle fibers were crushed away from the endplate band and voltage clamped to −45 mV. Quantal content was determined directly by dividing evoked endplate current (EPC) amplitude by the average miniature endplate current (MEPC) amplitude for a given NMJ. Repetitive stimulation was given by applying a 50 Hz train of 10 pulses.
All data are expressed as mean ± standard error of the mean. Morphological and biochemical data were analyzed using Excel and Statistica software. Statistical significance was determined using either Student's t-test or a two-way analysis of variance (ANOVA). Physiological data were analyzed using Student's t-test or a nested ANOVA using Systat (Cranes Software).
In order to generate SMA mice with increased SMN expression selectively in motor neurons or muscle, we utilized SMA mice that express a Cre-inducible Smn allele (SmnRes) (Lutz et al., 2011) together with mice expressing Cre recombinase under the control of either the ChAT promoter, the MyoD promoter (Kanisicak et al., 2009), or the Myf5 promoter (Tallquist et al., 2000). ChAT is expressed in motor neurons in the ventral horn of the spinal cord starting at E12.5 (Phelps et al., 1991). Subsequently, ChAT slowly increases in expression until birth and then more rapidly increases postnatally until adulthood (Lonnerberg et al., 1995). MyoD and Myf5 are muscle-specific transcription factors expressed at high levels in muscle precursor cells and myofibers starting at E9.75 and E8, respectively (Ott et al., 1991; Kanisicak et al., 2009).
We examined the spatial and temporal pattern of Cre expression by breeding each Cre line to a Cre-dependent lacZ reporter line and by performing X-gal staining to detect β-gal expression in E15.5 whole embryos and dissected P7 tissues (Figure 1A, B). As has been previously shown at multiple embryonic and postnatal time points (see http://cre.jax.org/Chat/Chat-CreNanozoomer.html), β-gal expression in P7 ChATCre mice was restricted to large neurons in the ventral horn of the spinal cord (Figure 1B) and some neurons of the brainstem and cortex (not shown). Expression was absent from muscle (Figure 1B), heart, liver, other internal organs, bone, brown fat, and skin (data not shown). In MyoDiCre embryos, β-gal expression was evident in skeletal muscles including those around the vertebrae, ribs, and sternum and in the diaphragm and limb muscles (Figure 1A) as previously reported, but not in the neural tube (Kanisicak et al., 2009; Yamamoto et al., 2009). In P7 mice, robust staining was present in myofibers, but not in spinal cord (Figure 1B), brain, heart, or other internal organs (data not shown). Embryonic Myf5Cre mice showed a similar pattern of muscle staining to MyoDiCre mice, but also some foci of staining in the neural tube (Figure 1A). At P7, robust staining was evident in muscle, but low levels of staining were also evident in the ventral horn of the spinal cord (Figure 1B). Robust staining was also seen in dorsal root ganglion (DRG) neurons (Figure 1B), brown fat, and skin (data not shown). Expression of Cre in both the neural tube and DRG in this line of Myf5Cre mice has been previously reported (Gensch et al., 2008). High power examination of P7 quadriceps muscle cross sections revealed strong, diffuse lacZ staining of many myofibers that occupied 31.8% ± 0.069 and 39.1% ± 0.036 of the total muscle cross sectional area in MyoDiCre and Myf5Cre mice, respectively. In addition, punctate lacz staining was evident in most other myofibers.
In order to further investigate whether visible lacZ expression in the spinal cord was present specifically in motor neurons, we performed IHC for lacZ and ChAT on P7 spinal cords. This confirmed β-gal expression specifically in ChAT-positive ventral horn motor neurons in ChATCre and Myf5Cre mice, but not in MyoDiCre mice (Figure 1C). Together, these data indicate that Cre expression is limited to motor neurons in the spinal cord in ChATCre mice and to muscle in MyoDiCre mice, but is present in muscle, motor neurons, and DRG neurons in Myf5Cre mice.
The Cre-inducible Smn allele (SmnRes) contains a switch cassette flanked by loxP sites in opposing orientation, that was introduced into the endogenous mouse gene (Lutz et al., 2011). The switch cassette contains human exon 7 sequence in the correct orientation and the mouse exon 7 sequence in the opposite and thus translationally silent orientation. In the absence of Cre, this hybrid allele consists of upstream mouse Smn sequence (exons 1–6) together with human SMN2 exons 7 and 8 sequence and produces predominantly truncated SMN transcripts (SMN 68) (Figure 2A). Following Cre-mediated genomic recombination, human SMN2 exon 7 sequence is replaced by mouse Smn exon 7 sequence and predominantly full length SMN transcripts (SMN 678) are produced (Figure 2A). SMA mice expressing this allele together with Cre recombinase expressed either embryonically or early postnatally in all tissues have increased SMN expression and marked amelioration of the SMA phenotype (Lutz et al., 2011). After crossing SmnRes SMA mice to each of the tissue-specific Cre expressing lines of mice, we evaluated the expression patterns of truncated and full-length hybrid SMN transcripts in spinal cord and muscle tissues isolated at P10 by qRT-PCR. Primers were designed that are specific for mouse exon 6 and human exon 8 (SMN6m8h) and they detected both truncated (SMN 68) and full-length hybrid SMN transcripts (SMN 678). Primers specific for mouse exon 6, mouse exon7, and human exon 8 (SMN67m8h) detected only full-length hybrid transcripts (SMN 678) (Figure 2A). Primers specific for the mouse exons 1–2 (SMN12m) were used as an endogenous control (Figure 2A) and all data were normalized to the transcript levels present in a Myf5Cre+ Het mouse muscle (this calibrator sample was set to 1). The SMN 6m8h primers detected SMN 68 and/or SMN 678 hybrid transcript expression in both spinal cord and muscle in Cre− and Cre+ SMA mice homozygous for SmnRes (Figure 2A). Heterozygous mice expressing one copy of SmnRes showed approximately half the expression of the hybrid transcripts and WT mice, which do not have the SmnRes allele, showed no expression (Figure 2B). The SMN67m8h primers that amplify only SMN678 showed that this transcript is expressed in muscle only in MyoDiCre+ and Myf5Cre+ Het and MyoDiCre+ and Myf5Cre+ SMA mice (Figure 2C). Comparably lower levels of SMN 678 transcript were evident in spinal cord of ChATCre+ Het and ChATCre+ SMA mice. This relatively low level of expression detected in spinal cord is expected given that SMN rescue is specific to motor neurons, which constitute only a small fraction of total spinal cord tissue. A very low level of SMN 678 was also detectable in Myf5Cre+ SMA spinal cord, but not in MyoDiCre+ SMA spinal cord.
We also examined P10 tissues for corresponding changes in SMN protein expression (Figure 2D, E). Not unexpectedly, given the modest increase of SMN 678 transcript expression, there was no appreciable increase in full-length SMN protein expression in the spinal cord of ChATCre+ SMA mice by western blot. In contrast, MyoDiCre+ and Myf5Cre+ SMA mice both showed significant increases in full-length SMN protein expression in muscle by 491% and 365%, respectively, but no changes were observed in spinal cords (Figures 2D, E).
Given the appearance of lacZ expression in the DRG of Myf5Cre reporter mice (Figure 1B), we also examined all three lines of Cre+ SMA mice for SMN hybrid transcript expression levels in P10 lumbar DRGs (Figure 2F, G). SMN 678 was detectable only in Myf5Cre+ SMA mice not ChATCre+ or MyoDiCre+ SMA mice (Figure 2F).
In order to verify expression of full-length hybrid SMN transcript at an early postnatal time point, we repeated the spinal cord and muscle qRT-PCR analysis on spinal cord and muscle tissues isolated at P2 (Figure 2H, I). These data indicate that SMN 678 hybrid transcripts are present in all three lines of mice at this early stage in the expected tissue-specific patterns.
We next examined the phenotypic outcomes of SMA mice resulting from tissue-specific increases in full-length SMN expression (Figure 3). Cre+ SMA mice showed increased survival compared to Cre− SMA mice in all 3 lines. Median survival increased from 15 to 23 days in ChATCre+ SMA mice (53%, Log rank p<0.0001), from 13 to 19 days in MyoDiCre+ SMA mice (46 %, Log rank p< 0.0001), and from 15 to 21 days in Myf5Cre+ SMA mice (40 %, Log rank p<0.0004) (Figure 3A). Weights were also significantly improved with 64.7%, 34.6%, and 78.9% increases in maximal weights achieved in ChATCre+, MyoDiCre+, and Myf5Cre+ SMA mice, respectively (Figure 3B). Motor behavior assays showed reduced righting time latencies (Figure 4A) in all three lines and increased tube test scores in muscle-rescued lines (Figure 4B). Because ChATCre+ SMA mice showed a particular improvement in motor behavior after P10 including the ability to rear and run, which was not seen in the MyoDiCre+ and Myf5Cre+ SMA mice, we measured features of ambulation in ChATCre+ SMA and WT mice between P13–P23 (see Methods). All ambulatory indices were similar in WT and ChATCre+ SMA mice until approximately P20 further indicating a marked improvement in motor function in ChATCre+ SMA mice (Figure 4C). Together these data indicate that increased SMN expression in motor neurons or muscle can partially ameliorate the SMA disease phenotype.
One pathological hallmark of SMA is motor neuron loss, with recent studies suggesting that particular motor neuron populations may be especially vulnerable to SMN deficiency. We have previously documented a progressive loss of motor neurons in the L1 segment (~40% by P4 and 60% by P13) of SMA mice, but little loss of lateral motor neurons in the L5 segment even at disease endstage (Mentis et al., 2011). Given these findings, we examined the total number of motor neurons in the vulnerable L1 segment at P10 (Figure 5). In all lines of mice, there was an approximate 60% reduction in motor neuron number in Cre− SMA mice compared to WT mice (Figure 5A, B). ChATCre+ SMA mice showed a 64% increase in motor neuron number compared to Cre− SMA mice (p=0.013), but no increase in motor neuron number was seen in MyoDiCre+ or Myf5Cre+ SMA mice (Figure 5B). Another pathological feature of SMA mice is small muscles and myofibers (Kong et al., 2009; Lee et al., 2011). We examined the total muscle cross sectional area and myofiber diameter in P10 TA muscles (Figure 6A, B) as we have previously demonstrated reduced myofiber diameter by P9 in this muscle (Kong et al., 2009). In ChATCre+ SMA mice, myofiber diameter was modestly increased by 11% compared to Cre− SMA mice, but total muscle cross sectional area was not significantly increased (Figure 6A, B). MyoDiCre+ and Myf5Cre+ SMA mice showed a 38% and 67% increase in muscle area, respectively and both showed a 36% increase in myofiber diameter (Figure 6A, B).
We and others have previously shown that early stages of SMA disease pathogenesis are characterized by functional and structural abnormalities of NMJs. In order to determine whether these abnormalities arise from SMN deficiency in motor neurons alone or result from the combination of SMN deficiency at both pre- and post synaptic sides of the NMJ, we examined the physiology of NMJs in the TA muscle as we have previously documented these NMJs to be fully innervated, but to have abnormalities of neuromuscular transmission (Kong et al., 2009). Specifically, SMA NMJs have reduced evoked endplate current (EPC) amplitudes (Kong et al., 2009), which are determined by both the number of synaptic vesicles released following nerve stimulation (quantal content) and the amplitude of the muscle response to the transmitter released from a single vesicle (quantal amplitude). In this study, we confirmed our previous observation that EPC amplitude was reduced by ~60% (p < 0.01) in Cre− SMA compared to WT mice in all three mouse lines (as WT and Cre− SMA values were highly similar across the three lines, they were combined) (Figure 7A, B). The reduction in EPC amplitude was due to a 40% reduction in quantal content (p < 0.01) and a 30% reduction in quantal amplitude (p < 0.01), as measured by MEPC amplitude (Figure 7B). Previously, we also found that a reduction in the probability of synaptic vesicle release as shown by increased facilitation during repetitive stimulation was a likely contributor to the reduced quantal content in SMA NMJs (Kong et al., 2009). A similar increase in facilitation during repetitive stimulation was again observed in Cre− SMA compared to WT mice in the current study in all three lines of mice (Figure 7B). Finally, we previously found an increase in MEPC and EPC time constants in SMA mice that was likely due to prolonged postsynaptic expression of embryonic acetylcholine receptors (AChRs) (Kong et al., 2009). In the three lines of conditional mice, there was a trend towards prolongation of the time constant in Cre− SMA compared to WT mice, but this increase was not statistically significant (data not shown).
We next examined whether increased expression of SMN in motor neurons or muscle rescued the abnormalities of neuromuscular transmission. Increased expression of SMN in motor neurons in ChATCre+ SMA mice restored quantal content, MEPC (quantal) amplitude, EPC amplitude, and the facilitation response to repetitive stimulation to WT levels (Figure 7B). In contrast, MyoDiCre+ SMA mice showed no improvement in any parameter of NMJ function (Figue 7B). Myf5Cre+ SMA mice had improvements of EPC amplitude and quantal content, but not of MEPC (quantal) amplitude or the response to repetitive stimulation (Figure 7B).
Consistent with the restored function of NMJs in ChATCre+ SMA mice, the NMJs in the TA muscle also showed improved structural characteristics with reduced NF accumulation in presynaptic terminals and increased postsynaptic endplate size (Figure 7D) equivalent to WT levels. In contrast, MyoDiCre+ SMA mice showed no improvement in either parameter (Figure 7D) and Myf5Cre+ mice showed no change in NF accumulation, but did show increased endplate size (Figure 7D). We also examined NMJs in the splenius capitis muscle as this muscle was recently shown to be susceptible to denervation in SMA mice (Ling et al., 2012). As expected, Cre− SMA mice showed a reduced number of highly innervated NMJs (as defined by >75% occupancy of an endplate by the presynaptic terminal) compared to WT mice in all three lines (Figure 7C, D). Both ChATCre+ and Myf5Cre+ SMA mice showed an increased percentage of highly innervated and a decreased percentage of partially innervated NMJs (defined by 25–75% occupancy of an endplate by the presynaptic terminal), whereas MyoDiCre+ SMA mice showed no change in innervation patterns (Figure 7D). NF accumulation was also decreased in the splenius capitis muscle in ChATCre+, but not MyoDiCre+ or Myf5Cre+ SMA mice (data not shown). Together these data indicate essentially complete restoration of NMJ structure and function in the motor neuron-rescued ChATCre+ SMA mice, a partial improvement in Myf5Cre+ SMA mice, and no change in the muscle-rescued MyoDiCre+ SMA mice.
In addition to abnormalities of NMJ synapses, recent studies have also highlighted disruptions of central motor neuron synaptic inputs in SMA mice suggesting that spinal circuitry disconnections could exacerbate or even precipitate motor neuron dysfunction in SMA. In order to examine whether these synaptic disruptions are due to retrograde effects on motor neuron afferents from SMN-deficient motor neurons or due to intrinsic abnormalities of these motor neuron afferents themselves, we examined VGluT1+ sensorimotor synapses on L1 segment motor neuron somata in all three lines of SMA mice because we have previously shown these synapses to be structurally and functionally abnormal in SMA mice (Mentis et al., 2011). In all three lines of mice, we confirmed our previous observation that there was an approximately 60–70% reduction of VGluT1+ synapses at P10 on L1 motor neuron cell bodies in Cre− SMA compared to WT mice (Figure 8A, B). ChATCre+ SMA mice showed a 124% increase in the number of VGluT1+ inputs compared to Cre− SMA mice, but there was no change in MyoDiCre+ or Myf5Cre+ SMA mice (Figure 8A, B). We further examined motor neuron somal synaptic inputs by EM in the ChATCre line. We chose somata of approximately equivalent sizes (average perimeter length: WT=140.0 ± 5.9 μm, Cre− SMA=125.6±5.4 μm, and Cre+ SMA=124.0±10.7 μm). The total number of boutons present per soma was not statistically different between WT, Cre− SMA, and Cre+ SMA mice (WT=42.1±4.5, Cre− SMA =31.7±4.1, and Cre+ SMA =43.1±6.2), but synapse number as defined by the presence of a bouton apposing a PSD was reduced by 40% in Cre− SMA compared to WT mice and was completely restored in ChATCre+ SMA mice (Figure 8C, D). PSD length showed a trend toward a decrease in Cre− and Cre+ SMA mice compared to WT mice (WT=0.41±0.03 μm, Cre− SMA=0.35±0.01 μm, and Cre+ SMA=0.34±0.03 μm), but this did not reach statistical significance. We also observed a significant decrease in the density of synaptic vesicles in the presynaptic terminals of Cre− SMA motor neuron somal synapses, both when considering the density across the entire bouton as well as the density near the active zone, defined as vesicles within 100 nm of the PSD (Figure 8C, D). The density of synaptic vesicles was restored in ChATCre+ SMA mice (Figure 8C, D). Together these data indicate that increased SMN expression in motor neurons not only restores NMJ synapses, but also prevents loss of central motor neuron synaptic input.
Abnormalities of NMJ and central motor neuron synapses are principal cellular consequences of SMN deficiency during early, symptomatic stages of SMA. Here, we demonstrate that this synaptic connectivity is determined by the presence of SMN in motor neurons and is not influenced by SMN expression in muscle. SMN-deficient motor neurons thus determine synaptic integrity at the NMJ via a role restricted to the presynaptic terminal, but may influence central synapse connectivity via retrograde effects on somal synaptic inputs. Although selective SMN expression in muscle does not influence synaptic function and connectivity at either peripheral or central synapses, it can increase muscle growth.
Motor neurons and muscle interacting at NMJ synapses are dependent on each other for chemical, electrical, and trophic signals. Drosophila SMA models show NMJ abnormalities together with disorganized motor nerves and severe muscle atrophy (Chan et al., 2003; Rajendra et al., 2007; Chang et al., 2008) resulting from SMN deficiency at both pre- and postsynaptic sides of the NMJ (Chan et al., 2003; Chang et al., 2008). Recently, retrograde FGF signaling from muscle to nerve was demonstrated to rescue these NMJ defects (Sen et al., 2011), further illustrating the interdependency of motor nerve and muscle in these models.
SMA mice also have physiological and structural abnormalities of both presynaptic terminals and postsynaptic structures of the NMJ. Previous studies have reported reductions in quantal content (the number of vesicles released in response to nerve stimulation) at SMA NMJs that may be due to reduced probability of release (demonstrated by increased facilitation during repetitive stimulation) (Kariya et al., 2008; Murray et al., 2008; Kong et al., 2009; Ling et al., 2010; Ruiz et al., 2010). However, prolonged postsynaptic decay time constants and slowed maturation of the postsynaptic apparatus, including a delay in switching of acetylcholine receptor subunit expression, have also been described (Biondi et al., 2008; Kariya et al., 2008; Murray et al., 2008; Kong et al., 2009; Ling et al., 2010; Ruiz et al., 2010; Dachs et al., 2011).
In the current study, increased expression of SMN in motor neurons restored the presynaptic properties of quantal content and probability of synaptic vesicle release. In contrast, expression of SMN in muscle had no impact on these presynaptic functions. We also measured EPC and MEPC amplitudes (i.e. quantal amplitude) in this study and found them to be significantly reduced in SMA compared to WT mice. While our findings of reduced EPC and MEPC amplitude may seem to differ from other studies (Kariya et al., 2008; Murray et al., 2008; Ling et al., 2010; Ruiz et al., 2010), the likely explanation is our use of a two-electrode voltage clamp technique rather than the more common single electrode measurement of synaptic voltage. This approach measures EPC and MEPC amplitudes without the confound of reduced muscle size, which increases muscle input resistance such that smaller amplitude EPCs and MEPCs of a given size could trigger endplate potentials (EPPs) and MEPPs of increased amplitude. Quantal amplitude is determined by the number of AChRs on the muscle fiber (a postsynaptic property) and the amount of transmitter contained in a vesicle (a presynaptic property). Thus rescue of MEPC (quantal) amplitude could be due to improvement of either a presynaptic or a postsynaptic defect in the NMJ. The most parsimonious explanation is that it is due to a presynaptic effect on the amount of transmitter contained in each synaptic vesicle, but we cannot rule out an anterograde trans-synaptic trophic effect on the number of AChRs at the NMJ.
The restoration of NMJ function was accompanied by improved NMJ structure with reduced NF accumulation in presynaptic terminals, increased endplate size, and reduced denervation of the splenius capitis muscle. Our data indicate that the principal role of SMN at the NMJ is in presynaptic terminal function with little role in retrograde trophic support from muscle. These observations are consistent with the finding that selective reduction of SMN in motor neurons is sufficient to cause NMJ pathology (Park et al., 2010b). Interestingly, Myf5Cre+ SMA mice with low levels of increased SMN expression in motor neurons and high levels in muscle showed a partial improvement in NMJ physiology. This likely reflects the presence of SMN in motor neurons and indicates that even very low levels of SMN in motor neurons can impact NMJ function. Nonetheless, we cannot eliminate the possibility that low levels of SMN in motor neurons combined with high levels of SMN in muscle can together partially improve NMJ structure and function.
We also demonstrate that synaptic number and morphology of spinal motor neuron synaptic inputs are restored by increased expression of SMN in motor neurons, but not by expression in muscle. Previously, we demonstrated that VGluT1+ sensorimotor synapses are functionally impaired and that this is associated with their retraction prior to motor neuron death (Mentis et al., 2011). These abnormalities are associated with motor neuron hyperexcitability and are severe in L1 segment motor neurons (Mentis et al., 2011). Other motor neuron synapse types are also abnormal with preferential loss of excitatory versus inhibitory synapses (Ling et al., 2010). In the current study, the number of VGluT1+ somal synapses as well as the ultrastructural morphology of all somal synapses were significantly improved by increased expression of SMN in motor neurons indicating that these abnormalities are determined by retrograde effects of SMN-deficient motor neurons on presynaptic terminals. Increased expression of SMN in DRG neurons in Myf5Cre+ SMA mice whose central projections form the VGluT1+ synapses did not prevent synaptic loss, further emphasizing the role of motor neurons in governing this process. Consistent with these results, during the revision of this manuscript Gogliotti et al. demonstrated that early restoration of SMN expression in motor neurons driven by the Hb-9 promoter resulted in increased VGLuT1+ motor neuron synapse number and corrected motor neuron hyperexcitability (Gogliotti et al., 2012). Ling et al. have suggested that this loss of motor neuron somal synapses represents a form of synaptic stripping (Ling et al., 2010), the active removal of central synapses by microglia. Synaptic stripping has been best described after nerve transaction, but it also occurs during motor neuron degenerative disease when it may be mediated in part by nitric oxide (Moreno-Lopez et al., 2011).
Increased SMN in muscle resulted in complete rescue of myofiber growth accompanied by an increase in median survival and motor behavior. Interestingly, these benefits occurred independently of measurable improvement in NMJ and central synaptic structure. A previous study showed that increased expression of SMN controlled by the human skeletal actin (HSA) promoter did not result in a significant improvement in survival or motor behavior (Gavrilina et al., 2008). Although HSA is expressed in mature myofibers, it is not expressed in satellite cells or myoblasts. Satellite cells are critical for muscle growth during development and are also activated during muscle regeneration (Cerletti et al., 2008; Kang and Krauss, 2010). Several previous studies have suggested that SMN-deficient satellite cells and/or myoblasts may contribute to an impairment of SMA muscle development (Fidzianska et al., 1990; Braun et al., 1995; Nicole et al., 2003; Shafey et al., 2005; Dachs et al., 2011; Mutsaers et al., 2011) . In the current study, we chose the MyoD and Myf5 promoters to drive Cre because these two proteins are expressed in essentially all muscle progenitors beginning in early embryogenesis, providing a means for widespread SMN expression in skeletal muscle. Our data support the conclusion that SMN plays a role in myofiber growth and development, although it is not yet clear whether embryonic or postnatal expression is responsible for rescue of the SMA muscle phenotype. Given that the ChATCre+ SMA mice displayed only a partial increase in myofiber size, full restoration of muscle growth in severe forms of SMA may require targeting of SMN to muscle in addition to motor neurons. Further studies are needed to understand the specific molecular and cellular functions of SMN in muscle as previous attempts to activate muscle growth independent of SMN induction, including myostatin inhibition and insulin growth factor-1 treatment, have had limited success (Rose et al., 2008; Sumner et al., 2009; Bosch-Marce et al., 2011; Rindt et al., 2011).
One question that arises from this study given the restoration of synaptic integrity in ChATCre+ SMA mice is why their survival is not more substantially increased. Widespread genetic restoration of SMN expression can markedly improve the disease phenotype in these mice if achieved within the first postnatal week (Le et al., 2011; Lutz et al., 2011). Gene therapy and antisense oligonucleotides strategies to increase SMN expression have also resulted in remarkable therapeutic effects (Foust et al., 2010; Passini et al., 2010; Dominguez et al., 2011; Passini et al., 2011; Porensky et al., 2011) with a recent study suggesting significantly enhanced efficacy with systemic compared to CNS-directed therapy (Hua et al., 2011). This later study highlights the unresolved question of which other tissue types contribute to SMA pathogenesis. Studies have increasingly reported abnormalities of other organ systems including the autonomic nervous system, heart, liver, and brain in SMA mice (Hua et al., 2011; Sleigh et al., 2011; Gogliotti et al., 2012). Further studies are needed to define the relative roles of these tissues in SMA pathogenesis and more importantly, their relevancy to human patients. Our data suggest that restoration of SMN in both motor neurons and muscle will be necessary for full functional recovery of the motor unit.
CJS was supported by NINDS grant R01NS062869, Howard Hughes Medical Institute Physician Scientist Award and the Spinal muscular atrophy research team (SMART). MMR was supported by NINDS grant P01NS057228. DJG was supported by NIAMS grant R01AR052777. Funding for the work at Jackson Laboratory was supported by the SMA Foundation. The authors would like to thank Francisco Alvarez for advice regarding EM, John Thorndyke, Heloisa Carvalho, Lauren Wu, Diana Villanueva, Emily Bergbower, Kristen Klepac and Emmanuel Ohuabunwa for technical help, and Carol Cooke and Michael McCaffery for EM assistance.
The authors have no conflicts of interest.