Proximal spinal muscular atrophy (SMA) is a common, frequently fatal neuromuscular disorder caused by mutations in the Survival of Motor Neuron 1 (SMN1) gene and, consequently, reduced levels of its translated product, the SMN protein (1–3). In humans, an almost identical copy gene, SMN2, is unable to compensate for loss of SMN1 owing to a synonymous, C→T nucleotide change in exon 7 that disrupts the splicing pattern of the homologue, rendering most of its transcripts devoid of the exon (4,5); relatively few SMN2 transcripts remain full-length (FL). The excision of exon 7 from FL-SMN does not adversely affect the transcript, however, the resulting SMNΔ7 protein is unstable and rapidly degraded. Accordingly, SMN2 is reported to contribute but a fraction (10–15%) to overall levels of the functional SMN protein complex. Still, the invariable presence of SMN2 in SMA patients ensures ubiquitous low levels of the FL-SMN protein. Moreover, the greater the number of copies of SMN2, the less severe is the SMA phenotype (6,7).
Rodents lack an SMN2 gene, and breeding mice heterozygous for the single murine Smn gene fails to produce viable Smn−/− offspring (8). The embryonic lethality associated with complete loss of murine Smn can be rescued by expressing one or more copies of the human SMN2 gene on the null background (9,10). What is more, SMN2;Smn−/− mice, or derivatives thereof, carrying 1–2 copies of the human transgene accurately model many aspects of the human SMA phenotype (11,12). Chief amongst these are the degeneration of the spinal motor neurons and the accompanying loss of muscle function. However, it is now clear that these signature features are preceded by an even earlier pathology of the distal motor unit. In particular, the post-synaptic specializations of mutant NMJs fail to properly mature, unable to expand in area or increase in complexity, sometimes appearing dimly stained—as if undergoing disassembly—when visualized for acetylcholine receptors (AChRs) (13–15). These post-synaptic defects are accompanied by prominent pre-synaptic abnormalities exemplified by nerve terminals swollen with neurofilament protein. The abnormal accumulation of neurofilaments in SMA terminals may be at least partly responsible for the inability to form the intricate arbors normally found at wild-type NMJs. The collective perturbations of the neuromuscular synapses, first discerned in SMA mice, were subsequently confirmed in human patients (13,16). However, it is uncertain why disrupting SMN, which is principally associated with the housekeeping functions of snRNP biogenesis and pre-mRNA splicing (17,18), would trigger such tissue-specific effects. The molecular mediators and the manner in which the NMJ defects evolve in SMA have thus, for the most part, remained elusive.
Plausible links between the canonical function of the SMN protein and NMJ phenotypes associated with a deficiency of the protein might be established through the identification of motor neuron or muscle-specific transcripts whose proper splicing is critical to the structure and/or function of the neuromuscular synapse. In this study, we explore the mediating effects of one such molecule, Agrin, which was reported mis-spliced in the motor neurons of SMA model mice (19). Agrin is expressed in motor neurons as well as muscle and best known for its role in organizing the NMJ by clustering post-synaptic acetylcholine receptors (20–22). This is accomplished chiefly through a motor neuron-derived isoform, Z+Agrin, which is defined by the presence of an 8, 11 or 19 (8 + 11) amino acid (Z) insert that greatly enhances the ability of the molecule to cluster AChRs relative to isoforms (Z−) devoid of it (22–24). The Z insert residues are encoded, respectively, by exons 32, 33 or both, and the incorporation of the exons into the mature transcript subject to alternative splicing. Z+Agrin is critically important to the formation as well as maintenance of the NMJ and accomplishes this through effects on both the pre- as well as post-synaptic specializations (21,22). We demonstrate that Z+Agrin is indeed depleted in the motor neurons of SMA model mice. Selective repletion of the protein in these cells mitigates the NMJ defects caused by SMN deficiency by enhancing post-synaptic development and reducing distal axonal pathology. This is effected without any rise in SMN levels, arrest of spinal motor neuron loss or an increase in the number of sensory inputs onto these cells. Nevertheless, the improvement in NMJ morphology translates into a modest albeit significant improvement in survival. We conclude that neuronally derived Agrin is an important mediator of the SMA phenotype and that improving NMJ function may be one means of altering the severity of the human disease.
Motor neuronal repletion of Z+Agrin modulates the phenotype of SMA model mice without altering SMN levels
Recent reports suggest that spinal motor neurons of severely affected SMA model mice express reduced levels of the Agrin isoform harboring the Z insert (Z+Agrin) (19). We initially sought to verify this result. Quantitative PCR analysis of the anterior horn cells (motor neurons) of the spinal cords of postnatal day 3 (PND3) wild-type and SMNΔ7 SMA mice, a commonly employed model of the severe form of the human disease (25) confirmed the earlier finding. Whereas Agrin transcripts containing the Z insert were present at significantly lower levels in SMA motor neurons than in control motor neurons, transcripts lacking the insert (Z−Agrin) were correspondingly increased in expression (Supplementary Material, Fig. S1). Accordingly, we decided to test the effects of restoring levels of Z+Agrin to the motor neurons of SMA mice. To do so, mice expressing a chicken Z+Agrin cDNA driven by the motor neuron-specific Hb9 promoter (26) were bred with SMA carriers, and mutants with (Hb9:Z+Agrin;SMN2;SMNΔ7;Smn−/−) or without (SMN2;SMNΔ7;Smn−/−) the Hb9:Z+Agrin transgene were generated.
Forced expression of Agrin, through inappropriate binding and activation, respectively, of the muscle-derived Lrp4 and MuSK receptors, might cause post-synaptic perturbations of the AChRs; over-expressing MuSK has been documented to trigger such effects (27). To ensure that this was not the case in the Hb9:Z+Agrin mice, and prior to assessing the effect of restoring Z+Agrin to SMA motor neurons, we carried out the following additional experiments. First, we examined the area of AChR clusters in muscle of 1-month-old Hb9:Z+Agrin mice and controls. Second, we examined the width of the endplate band in the diaphragms of PND0 Hb9:Z+Agrin transgenic animals and controls. We found that the expression of the Hb9:Z+Agrin transgene neither altered the morphology or area of the AChRs in the adult muscle nor perturbed nerve terminals or AChR clusters in the endplate bands of newborn mice (Supplementary Material, Fig. S2A–D). We also asked if the expression of the transgene declines as is sometimes the case when the regulatory elements of the Hb9 gene are used to drive motor neuronal gene expression, and investigated the possibility of ectopic chicken Z+Agrin expression in muscle. We found that when normalized to β-actin, the expression of the chicken Z+Agrin transgene did not change significantly between PND7 and PND21 (Supplementary Material, Fig. S2E). When normalized to a second commonly used housekeeping gene, Gapdh, we found a ∼25% drop in expression between PND7 and PND14, but no further decline between PND14 and PND21 (data not shown). As expected, there was no evidence of Hb9:Z+Agrin transgene expression in muscle (Supplementary Material, Fig. S2F).
Having ensured that the Hb9:Z+Agrin transgene does not cause intrinsic perturbations of the NMJs, we examined the effects of restoring Z+Agrin to the motor neurons of SMA model mice. Hereafter, SMA mice with the transgene are referred to as Z+Agrin-SMA mutants. Although SMA model mice expressing the Hb9:Z+Agrin transgene did not look overtly different from littermates without the transgene (Fig. 1A and B), we found that restoring Z+Agrin to mutants did mitigate disease severity by significantly delaying death caused by very low levels of the SMN protein. Thus, whereas SMA mutants quickly succumbed to disease with a mean survival of ∼9 days, littermates restored for Z+Agrin lived ∼40% longer (Fig. 1C). We further showed that the chicken Z+Agrin transgene was indeed expressed in the spinal cords of Z+Agrin-SMA mutants (Fig. 1D) and was expressed at ∼80% of levels of endogenous murine Z+Agrin (Fig. 1E). Total Z+Agrin in these mice is expected to equal, if not exceed by 1.3–1.5-fold, wild-type levels of the transcript. Similar results for total Z+Agrin in the Z+Agrin-SMA mutants were obtained when transcript levels were normalized to a different housekeeping gene, Gapdh (data not shown). To determine if Z+Agrin repletion had affected SMN levels, we examined protein concentrations in the spinal cords of the Z+Agrin-SMA mutants and control littermates. We found that SMN levels continued to be significantly lower in the spinal cords of these animals relative those in wild-type mice and no different from those in mutants without the Hb9:Z+Agrin transgene (Fig. 1F and G), suggesting that the mitigating effects of Z+Agrin repletion do not involve increases in SMN.