## Abstract

Spinal muscular atrophy (SMA) is a progressive neurodegenerative disease affecting lower motor neurons. SMA is caused by mutations in the Survival Motor Neuron 1 (SMN1) gene, which result in reduced levels of functional SMN protein. Biochemical studies have linked the ubiquitously expressed SMN protein to the assembly of pre-mRNA processing U snRNPs, raising the possibility that aberrant splicing is a major defect in SMA. Accordingly, several transcripts affected upon SMN deficiency have been reported. A second function for SMN in axonal mRNA transport has also been proposed that may likewise contribute to the SMA phenotype. The underlying etiology of SMA, however, is still not fully understood. Here, we have used a combination of genomics and live Ca2+ imaging to investigate the consequences of SMN deficiency in a zebrafish model of SMA. In a transcriptome analyses of SMN-deficient zebrafish, we identified neurexin2a (nrxn2a) as strongly down-regulated and displaying changes in alternative splicing patterns. Importantly, the knock-down of two distinct nrxn2a isoforms phenocopies SMN-deficient fish and results in a significant reduction of motor axon excitability. Interestingly, we observed altered expression and splicing of Nrxn2 also in motor neurons from the Smn−/−;SMN2+/+ mouse model of SMA, suggesting conservation of nrxn2 regulation by SMN in mammals. We propose that SMN deficiency affects splicing and abundance of nrxn2a. This may explain the pre-synaptic defects at neuromuscular endplates in SMA pathophysiology.

## INTRODUCTION

Spinal muscular atrophy (SMA) is a common autosomal recessive neurodegenerative disease characterized by degeneration of lower α-motor neurons (MN) in the spinal cord (1). The loss of MNs leads to progressive muscle weakness, paralysis and eventual death, particularly in infants (2). SMA patients display homozygous loss of the Survival of Motor Neuron 1 (SMN1) gene (1) but retain varying numbers of a second copy SMN2. This second allele is almost identical to SMN1 except for a silent C to T substitution in exon 7 (1,3), which leads to increased exon 7 skipping (4,5). As a consequence, the major protein product translated from SMN2 mRNA is an unstable protein (SMNΔ7), whereas only small amounts of functional SMN protein are generated. Hence, SMN2 only partially compensates for the loss of SMN1 and its copy number inversely correlates with the severity of the manifested phenotype (6–9). This demonstrates that SMA disease severity depends on the levels of functional SMN protein present (8,10). However, how reduction of ubiquitously expressed SMN leads to an MN-specific degeneration in SMA remains unclear.

The functional characterization of SMN allowed to put forward hypotheses for the etiology of SMA (11). The first hypothesis builds on SMN's role in the assembly of RNA–protein complexes of the pre-mRNA processing system (U snRNPs). These complexes form essential parts of the major and minor spliceosome (12–15) and hence establish a link to the production of mature mRNA molecules (16,17). It has been proposed that reduced U snRNP assembly could affect splicing of transcripts important for MN function (18). Consistent with this notion, introduction of purified U snRNPs can rescue axonal phenotypes observed upon SMN knock-down in a zebrafish model of SMA (19). Also, snRNP levels are strongly reduced in severe SMA mice (Smn−/−;SMN2+/+), mildly reduced in mild SMA carrier mice (Smn+/−;SMN2+/+) and restored to normal levels in phenotypically rescued SMN2tg mice (SMN2+/−;SMN2(566)+/−;Smn−/−) (20). This suggests that snRNP assembly controlled by SMN is critical for the development of the SMA phenotype. Recent reports furthermore showed that reduced levels of SMN preferentially affect splicing of minor spliceosome-dependent U12 intron containing transcripts (18,21,22). Interestingly, it was reported that one affected transcript, Stasimon, is not required in MNs, but in sensory and interneurons suggesting non-cell autonomous mechanisms causative for motor circuit defects in a Drosophila model of SMA (21,23). This suggests that certain transcripts may be more susceptible to aberrant splicing in cell types critical for MN function.

An alternative hypothesis involving MN-specific functions of SMN independent of U snRNP assembly has been proposed (11). SMN is expressed in MNs and can be found in growth cones in vitro (24,25), notably without co-localization of Sm proteins (26). SMN forms complexes with hnRNP R and other members of the hnRNP family (25). In association with hnRNP R, it contributes to translocation of β-actin mRNA and possibly other mRNAs along motor axons to growth cones (27). SMA mice show reduced hnRNP R and β-actin levels in motor axons suggesting a possible role of SMN in axonal transport of β-actin mRNA required for local translation at growth cones (27). SMA has been proposed to be a neuromuscular junction (NMJ) synaptopathy, where NMJ denervation precedes neurodegeneration (28–30). Pre-synaptic defects reported at the NMJ of SMA mice include abnormal synaptic transmission, reduced quantal content (28,29) and altered pre-synaptic intraterminal Ca2+ levels during repetitive stimulation (31). Defective clustering of N-type voltage-gated calcium channels (VGCCs) has also been observed in growth cones of primary SMA mouse MNs in culture and these MNs exhibited reduced spontaneous excitability and reduced local Ca2+ transients at their axon terminals (32). Importantly, however, there has been no evidence of a mechanistic link between aberrant splicing of transcripts downstream of SMN and the reduced pre-synaptic axon excitability observed in SMA.

Here, we used transcriptome analysis in a zebrafish model for SMA to identify neurexin 2 (nrxn2), encoding a pre-synaptic membrane protein important for neurotransmitter release as a new mRNA target in SMN deficiency. In both zebrafish and mouse models of SMA, nrxn2 exhibits changes in expression and alternative splicing patterns upon SMN knock-down. Importantly, reduced levels of nrxn2a phenocopy defects of reduced motor axon excitability observed after SMN knock-down. We propose that aberrant splicing of nrxn2a caused by SMN deficiency is involved in mediating pre-synaptic MN defects in SMA.

## RESULTS

### Transcriptome analysis identifies nrxn2a as target for SMN in zebrafish

SMN forms a complex with Gemin2 during assembly of spliceosomal U snRNPs (12,14,15,33). Knock-down of either or both proteins leads to reduced snRNP levels in vitro and in vivo (18,34,35). This is thought to result in aberrant pre-mRNA splicing with subsequent degradation of wrongly spliced transcripts by nonsense-mediated mRNA decay (NMD) (36,37). To identify down-regulated transcripts under SMN- and Gemin2-deficient conditions on a global scale, microarrays were performed using cDNA from SMN Morpholino (MO), Gemin2 MO- and Control MO-injected zebrafish embryos at 48 hours post fertilization (hpf). Transcripts down-regulated after knock-down of both SMN and Gemin2 were identified and further validated. Six biological replicates were performed for SMN MO and seven biological replicates performed for Gemin2 MO. The number of annotated transcripts with significant fold changes was 1545 and 1178 for SMN and Gemin2, respectively; 737 significantly dysregulated transcripts overlapped between the SMN MO and Gemin2 MO arrays. This overlapping pool of transcripts was analyzed for potential candidates based on the criteria of expression in MNs or in neurons, involved in synaptic transmission, related to NMJ, or related to calcium channels/cation channels. After database searches (PubMed, ZFIN, OMIM) for known biological functions, 33 potential candidate genes (Table 1) were shortlisted for further validation by qRT–PCR (primers listed in Supplementary Material, Table S2); 19 out of 33 (76.0%) genes displayed transcriptional changes consistent with the microarray results. Genes that exhibited significant down-regulation after SMN MO knock-down in both microarray and qRT–PCR were shortlisted for functional validation. Among these candidates, we identified neurexin2a (nrxn2a), which belongs to a family of neuron-specific cell surface proteins present at the pre-synaptic terminal (38). These bind to postsynaptic membrane proteins such as neuroligin and dystroglycan (39–41) and act as pre-synaptic cell adhesion molecules (39,42). α-Nrxns functionally couple voltage-gated Ca2+ channels to synaptic vesicle exocytosis (43,44) and are essential for Ca2+ -triggered neurotransmitter release at the NMJ synapse (45). In zebrafish, the nrxn family consists of nrxn1a, nrxn1b, nrxn2a, nrxn2b, nrxn3a and nrxn3b, as orthologs of mammalian Nrxn1, Nrxn2 and Nrxn3 genes, respectively, with a high degree of sequence homology and protein domain conservation (46).

Table 1.

Expression level changes of genes shortlisted from SMN MO and Gemin2 MO microarrays

SMN MO

Gemin2 MO

S/N ID Unigene ID Gene Fold change Reg Fold change Reg
BI879720 Dr.82643 syn2b 5.959534 Down 4.0850115 Down
BG728481 Dr.83149 plek2 5.745815 Down 4.9594765 Down
CD606547 Dr.85326 utro 5.0761156 Down 4.0662274 Down
AI601801 Dr.113730 latro 4.8633604 Down 3.6505525 Down
BC092962 Dr.87327 chrna5 4.836111 Down 3.1569462 Down
BM025904 Dr.81240 sdnah9 4.541099 Down 3.2342565 Down
AY436322 Dr.90867 gfra2 4.4383264 Down 4.418084 Down
AB087185 Dr.84843 chrnb3b 4.05756 Down 3.6565442 Down
CD606734 Dr.77108 ucp4 4.0055313 Down 4.7451315 Down
10 BC090478 Dr.24793 scfd2 3.8695138 Down 4.2816424 Down
11 AI522541 Dr.75084 desm 3.7772496 Down 3.1825519 Down
12 BI882735 Dr.62387 nrxn2a 3.1741772 Down 2.4237323 Down
15 AF268051 Dr.82556 igf1 2.9402876 Down 2.2010093 Down
17 AY292647 Dr.89527 p2rx7 2.9139013 Down 2.0392005 Down
19 CK706171 Dr.85547 syngr1 2.9053807 Down 2.3551302 Down
20 AW421031 Dr.5764 ephb4a 2.853329 Down 3.3008952 Down
21 AB070208 Dr.85655 rapsn 2.756902 Down 3.1649163 Down
22 BC096497 Dr.75984 snx6 2.657225 Down 2.8307834 Down
23 AY445046 Dr.113407 hlxbl9b 2.6524372 Down 1.8215746 Down
24 AY292649 Dr.89525 p2rx1 2.6456316 Down 2.1711686 Down
25 CK705430 Dr.83915 zplxdc2 2.3160033 Down 4.339301 Down
26 BC074062 Dr.28930 bgn 2.2892895 Down 2.8513803 Down
27 NM_001003981 Dr.88108 ptn 2.2319345 Down 2.4857042 Down
29 BE605580 Dr.81813 scdic2 2.03932 Down 2.3242311 Down
30 AL730893 Dr.151468 nes 2.0210247 Down 2.016073 Down
31 BC092953 Dr.84216 cript 2.01504 Down 2.366853 Down
14 AI444359 Dr.76915 cyth1 3.1739771 Up 2.7500966 Up
16 BC096810 Dr.84313 elp3 2.9377375 Up 2.1909373 Up
18 AA495105 Dr.11940 ptprk 2.9059074 Up 2.5899403 Up
28 U66568 Dr.266 mef2a 2.0419617 Up 2.138967 Up
32 AF201451 Dr.75485 cxcr4b 1.8897021 Up 1.5939336 Up
33 AJ005693 Dr.132860 stat3 1.579189 Up 1.3764645 Up
SMN MO

Gemin2 MO

S/N ID Unigene ID Gene Fold change Reg Fold change Reg
BI879720 Dr.82643 syn2b 5.959534 Down 4.0850115 Down
BG728481 Dr.83149 plek2 5.745815 Down 4.9594765 Down
CD606547 Dr.85326 utro 5.0761156 Down 4.0662274 Down
AI601801 Dr.113730 latro 4.8633604 Down 3.6505525 Down
BC092962 Dr.87327 chrna5 4.836111 Down 3.1569462 Down
BM025904 Dr.81240 sdnah9 4.541099 Down 3.2342565 Down
AY436322 Dr.90867 gfra2 4.4383264 Down 4.418084 Down
AB087185 Dr.84843 chrnb3b 4.05756 Down 3.6565442 Down
CD606734 Dr.77108 ucp4 4.0055313 Down 4.7451315 Down
10 BC090478 Dr.24793 scfd2 3.8695138 Down 4.2816424 Down
11 AI522541 Dr.75084 desm 3.7772496 Down 3.1825519 Down
12 BI882735 Dr.62387 nrxn2a 3.1741772 Down 2.4237323 Down
15 AF268051 Dr.82556 igf1 2.9402876 Down 2.2010093 Down
17 AY292647 Dr.89527 p2rx7 2.9139013 Down 2.0392005 Down
19 CK706171 Dr.85547 syngr1 2.9053807 Down 2.3551302 Down
20 AW421031 Dr.5764 ephb4a 2.853329 Down 3.3008952 Down
21 AB070208 Dr.85655 rapsn 2.756902 Down 3.1649163 Down
22 BC096497 Dr.75984 snx6 2.657225 Down 2.8307834 Down
23 AY445046 Dr.113407 hlxbl9b 2.6524372 Down 1.8215746 Down
24 AY292649 Dr.89525 p2rx1 2.6456316 Down 2.1711686 Down
25 CK705430 Dr.83915 zplxdc2 2.3160033 Down 4.339301 Down
26 BC074062 Dr.28930 bgn 2.2892895 Down 2.8513803 Down
27 NM_001003981 Dr.88108 ptn 2.2319345 Down 2.4857042 Down
29 BE605580 Dr.81813 scdic2 2.03932 Down 2.3242311 Down
30 AL730893 Dr.151468 nes 2.0210247 Down 2.016073 Down
31 BC092953 Dr.84216 cript 2.01504 Down 2.366853 Down
14 AI444359 Dr.76915 cyth1 3.1739771 Up 2.7500966 Up
16 BC096810 Dr.84313 elp3 2.9377375 Up 2.1909373 Up
18 AA495105 Dr.11940 ptprk 2.9059074 Up 2.5899403 Up
28 U66568 Dr.266 mef2a 2.0419617 Up 2.138967 Up
32 AF201451 Dr.75485 cxcr4b 1.8897021 Up 1.5939336 Up
33 AJ005693 Dr.132860 stat3 1.579189 Up 1.3764645 Up

In microarrays, nrxn2a was down-regulated by 3.17-fold (P = 0.005) in SMN MO (Table 1) and by 2.43-fold (P = 0.011) in Gemin2 MO-injected embryos (Table 1). Zebrafish nrxn2a has two isoforms generated by alternative promoter usage, a long α-isoform (nrxn2aa) and a short β-isoform (nrxn2ab) (Fig. 1A and B), which are orthologs of Nrxn2α and Nrxn2β in mammals, respectively (46). Nrxn2aa encodes a deduced N-terminal signal peptide and three modular repeats of two laminin–neurexin–sex hormone (LNS) domains flanking an epidermal growth factor (EGF)-like domain that comprise its extracellular domain, followed by a predicted O-glycosylation region, a transmembrane domain and a conserved cytoplasmic domain (Fig. 1C), equivalent to the mammalian situation (48). Nrxn2ab has a similar deduced C-terminal protein organization but a smaller extracellular domain with an atypical N-terminal signal peptide and a short unique amino acid sequence (Fig. 1D), as in mammals (48,49). As the probes used in our microarrays were designed against the 3′ ends of transcripts, it was not possible to discriminate the two nrxn2a isoforms containing identical 3′ termini by microarray analysis. Thus, to confirm down-regulation and to assess if there was any isoform-specific difference in expression levels, qRT–PCR primers specific for each nrxn2a isoform were designed (Supplementary Material, Table S2). qRT–PCR analysis showed that both nrxn2aa and nrxn2ab isoforms were significantly down-regulated in SMN MO (P–value <0.0001) compared with Control MO-injected embryos (Fig. 1E and F; Table 2), which was consistent with the significant down-regulation observed in the microarray analysis (Table 1). nrxn2aa was down-regulated to a greater extent (3.03-fold, P < 0.0001; Fig. 1E, Table 2) compared with nrxn2ab (1.87-fold, P < 0.0001; Fig. 1F, Table 2). This suggests an isoform-specific difference of nrxn2a transcript levels after knock-down of SMN.

Table 2.

Expression level changes of shortlisted genes in SMN morphants after real-time qRT–PCR

Gene P-value Real-time RT–PCR fold change Reg Microarray fold change
nrxn2aa <0.0001 3.03030303 Down 3.1741772
ucp4 <0.0001 2.777777778 Down 4.0055313
latro <0.0001 2.638522427 Down 4.8633604
hlxbl9b <0.0001 2.624671916 Down 2.6524372
syn2b <0.0001 2.469135802 Down 5.959534
cxcr4b <0.0001 2.450980392 Down 1.8897021
bgn <0.0001 2.044989775 Down 2.2892895
nes <0.0001 1.941747573 Down 2.0210247
nrxn2ab <0.0001 1.872659176 Down 3.1741772
chrna5 <0.0001 1.706484642 Down 4.836111
ptn <0.0001 1.647446458 Down 2.2319345
syngr1 <0.0001 1.633986928 Down 2.9053807
cript <0.0001 1.618122977 Down 2.01504
mef2a <0.0001 1.577287066 Down 2.0419617
stat3 <0.0001 1.57 Up 1.579189
igf1 <0.0001 1.567398119 Down 2.9402876
snx6 <0.0001 1.543209877 Down 2.657225
scdic2 <0.0001 1.538461538 Down 2.03932
p2rx1 <0.0001 1.524390244 Down 2.6456316
desm <0.0001 1.389 Up 3.7772496
scfd2 <0.0001 1.379310345 Down 3.8695138
gfra2 0.001 1.298 Up 4.4383264
rapsn 0.072 1.248439451 Down 2.756902
chrnb3b 0.0018 1.232 Up 4.05756
plek2 0.0309 1.189060642 Down 5.745815
ptprk 0.1216 1.153402537 Down 2.9059074
elp3 0.0338 1.14416476 Down 2.9377375
ephb4a 0.2338 1.117 Up 2.853329
utro 0.0945 1.109 Up 5.0761156
cyth1 0.2705 1.050420168 Down 3.1739771
p2rx7 0.6214 1.031991744 Down 2.9139013
zplxdc2 0.8092 1.015 Up 2.3160033
sdnah9 0.9882 1.001001001 Down 4.541099
Gene P-value Real-time RT–PCR fold change Reg Microarray fold change
nrxn2aa <0.0001 3.03030303 Down 3.1741772
ucp4 <0.0001 2.777777778 Down 4.0055313
latro <0.0001 2.638522427 Down 4.8633604
hlxbl9b <0.0001 2.624671916 Down 2.6524372
syn2b <0.0001 2.469135802 Down 5.959534
cxcr4b <0.0001 2.450980392 Down 1.8897021
bgn <0.0001 2.044989775 Down 2.2892895
nes <0.0001 1.941747573 Down 2.0210247
nrxn2ab <0.0001 1.872659176 Down 3.1741772
chrna5 <0.0001 1.706484642 Down 4.836111
ptn <0.0001 1.647446458 Down 2.2319345
syngr1 <0.0001 1.633986928 Down 2.9053807
cript <0.0001 1.618122977 Down 2.01504
mef2a <0.0001 1.577287066 Down 2.0419617
stat3 <0.0001 1.57 Up 1.579189
igf1 <0.0001 1.567398119 Down 2.9402876
snx6 <0.0001 1.543209877 Down 2.657225
scdic2 <0.0001 1.538461538 Down 2.03932
p2rx1 <0.0001 1.524390244 Down 2.6456316
desm <0.0001 1.389 Up 3.7772496
scfd2 <0.0001 1.379310345 Down 3.8695138
gfra2 0.001 1.298 Up 4.4383264
rapsn 0.072 1.248439451 Down 2.756902
chrnb3b 0.0018 1.232 Up 4.05756
plek2 0.0309 1.189060642 Down 5.745815
ptprk 0.1216 1.153402537 Down 2.9059074
elp3 0.0338 1.14416476 Down 2.9377375
ephb4a 0.2338 1.117 Up 2.853329
utro 0.0945 1.109 Up 5.0761156
cyth1 0.2705 1.050420168 Down 3.1739771
p2rx7 0.6214 1.031991744 Down 2.9139013
zplxdc2 0.8092 1.015 Up 2.3160033
sdnah9 0.9882 1.001001001 Down 4.541099
Figure 1.

Isoform-specific differences in coding sequence, protein domains and down-regulation by SMN knock-down between nrxn2aa and nrxn2ab. (A) The Nrxn2aa isoform contains 24 exons comprising 5013 bp, with coding sequence for 1670 amino acids. (B) The Nrxn2ab isoform contains 7 exons comprising 4333 bp, with coding sequence for 642 amino acids. Exon 1 of nrxn2ab corresponds to a region of intron 18 of nrxn2aa. (C) nrxn2aa has a larger extracellular domain than nrxn2ab and consists of three modular repeats of LNS-EGF-LNS domain and a common C terminal structure with nrxn2ab (47). (D) nrxn2ab consists of a unique signal peptide sequence followed by a similar structure of the sixth LNS domain, O-Glycosylated region, transmembrane domain and intracellular PDZ domain found in both isoforms. (E and F) Validation by real-time qRT–PCR shows that nrxn2aa is down-regulated to a greater extent than nrxn2ab after knock-down of SMN (schematic drawings not to scale).

Figure 1.

Isoform-specific differences in coding sequence, protein domains and down-regulation by SMN knock-down between nrxn2aa and nrxn2ab. (A) The Nrxn2aa isoform contains 24 exons comprising 5013 bp, with coding sequence for 1670 amino acids. (B) The Nrxn2ab isoform contains 7 exons comprising 4333 bp, with coding sequence for 642 amino acids. Exon 1 of nrxn2ab corresponds to a region of intron 18 of nrxn2aa. (C) nrxn2aa has a larger extracellular domain than nrxn2ab and consists of three modular repeats of LNS-EGF-LNS domain and a common C terminal structure with nrxn2ab (47). (D) nrxn2ab consists of a unique signal peptide sequence followed by a similar structure of the sixth LNS domain, O-Glycosylated region, transmembrane domain and intracellular PDZ domain found in both isoforms. (E and F) Validation by real-time qRT–PCR shows that nrxn2aa is down-regulated to a greater extent than nrxn2ab after knock-down of SMN (schematic drawings not to scale).

Neurexin transcripts have been reported to be highly alternatively spliced with thousands of possible isoforms having different functions and synaptic coupling properties (48,50–52). The splicing patterns of nrxn2a are known to be altered when neurons are depolarized, particularly at alternative splice site 1 and alternative splice site 3 (SS3) (51). It was shown in rat neuronal cultures that the exclusion of a particular exon at SS3 depends on the levels of Ca2+ present (51). We, therefore, tested whether there was any similar change in splicing patterns at SS3 in nrxn2aa transcripts under SMN-deficient conditions in zebrafish (Fig. 2A). There was a significantly stronger down-regulation of PCR products including exon 12 (45.96% ± 6.338% SEM), compared with the down-regulation of products excluding exon 12 in SMN MO (62.42% ± 17.26% SEM) when normalized to the control MO situation (Fig. 2B and C). Exon 12 skipping at SS3 in nrxn2aa was further validated using qPCR with primers to detect the presence of exon 12 and the overall transcript levels of nrxn2aa. By this, normalized intensity ratios of exon 12/total nrxn2aa were obtained. This provided an accurate assessment of the extent of alternative splicing in the presence of global reduction of nrxn2aa in SMN MO. We observed a significant reduction (P < 0.05) in normalized intensity ratio in SMN MO (0.739 ± 0.0965 SEM, n = 4 biological replicates) compared with Gemin2 MO (1.192 ± 0.140 SEM, n = 4 biological replicates) or Control MO (Fig. 2D). Together, this strongly suggests that there is elevated exon 12 skipping of nrxn2aa under SMN-deficient conditions, which indicates that reduced levels of SMN are able to affect the splicing patterns of nrxn2aa. Surprisingly, a Gemin2 knock-down resulted only in weak, but not significant changes of exon 12 skipping at SS3 of nrxn2aa (Fig. 2B and C).

Figure 2.

Knock-down of SMN leads to increased exon 12 skipping in nrxn2aa. (A) Design of primers at SS3 alternative splice site in nrxn2aa. (B) The alternative splicing assay using RT–PCR shows reduction of PCR products containing exon 12 (upper bands) in SMN MO-injected embryos compared with Control MO injected embryos. gapdh was used as normalization control. (C) There was a statistically significant reduction of the ratio of transcripts containing exon 12 in SMN MO-injected embryos (45.96% ± 6.338) relative to control MO-injected embryos (100%), but no significant change in Gemin2 MO-injected embryos (78.90% ± 6.266). There was also no significant change in ratios of transcripts without exon 12 in SMN MO-injected embryos (62.42% ± 17.26) or in Gemin 2 MO-injected embryos (62.92% ± 8.507). (D) qPCR validation of increased exon 12 skipping by calculating normalized intensity ratio of exon 12/total nrxn2aa. There was a statistically significant reduction of exon12/total nrxn2aa in SMN MO (0.739 ± 0.0965 SEM., n = 4 biological replicates) compared with Gemin2 MO (1.192 ± 0.140 SEM, n = 4 biological replicates) or Control MO.

Figure 2.

Knock-down of SMN leads to increased exon 12 skipping in nrxn2aa. (A) Design of primers at SS3 alternative splice site in nrxn2aa. (B) The alternative splicing assay using RT–PCR shows reduction of PCR products containing exon 12 (upper bands) in SMN MO-injected embryos compared with Control MO injected embryos. gapdh was used as normalization control. (C) There was a statistically significant reduction of the ratio of transcripts containing exon 12 in SMN MO-injected embryos (45.96% ± 6.338) relative to control MO-injected embryos (100%), but no significant change in Gemin2 MO-injected embryos (78.90% ± 6.266). There was also no significant change in ratios of transcripts without exon 12 in SMN MO-injected embryos (62.42% ± 17.26) or in Gemin 2 MO-injected embryos (62.92% ± 8.507). (D) qPCR validation of increased exon 12 skipping by calculating normalized intensity ratio of exon 12/total nrxn2aa. There was a statistically significant reduction of exon12/total nrxn2aa in SMN MO (0.739 ± 0.0965 SEM., n = 4 biological replicates) compared with Gemin2 MO (1.192 ± 0.140 SEM, n = 4 biological replicates) or Control MO.

Taken together, transcriptome analysis in a zebrafish SMA model identified nrxn2a as a strongly down-regulated and aberrantly spliced target opening the possibility that nrxn2a could mediate the MN defects observed after SMN knock-down.

### Expression of nrxn2a isoforms in the zebrafish nervous system

We next tested, whether both nrxn2a isoforms are expressed in zebrafish MNs to test their relevance for a MN-related SMA phenotype. No expression patterns for nrxn2a have so far been reported in zebrafish, but nrxn1a is expressed in the zebrafish central nervous system and in endothelial cells (53). We designed isoform-specific riboprobes and found that both isoforms are expressed robustly in various regions of the central nervous system (CNS) including telencephalon, mesencephalon, hindbrain and spinal cord (Fig. 3A–H, Supplementary Material, Fig. S1A–D). Expression in the brain was developmentally regulated with broad expression at 20 hpf (Supplementary Material, Fig. S1A and B) and 24 hpf (Fig. 3A–F) but more restricted patterns at 31 hpf (Fig. 3G and H) and 48 hpf (Supplementary Material, Fig. S1C and D). The expression of both nrxn2a isoforms in the spinal cord appeared to be strongest at ∼24 hpf (Fig. 3C,D) but was reduced from 31 hpf onwards (Fig. 3G and H). The high levels of expression in neurons within the spinal cord at 24 hpf is coincident with the timing of NMJ synaptogenesis (46,54). This opens the possibility that nrxn2a isoforms are involved in the formation or development of NMJ synapses.

Figure 3.

Expression pattern of nrxn2aa and nrxn2ab in zebrafish displays isoform-specific differences and expression in motor neurons. (AD) From whole mount RNA in situ hybridization, nrxn2aa and nrxn2ab are expressed in telencephalon (tel), midbrain (mb), hindbrain (hb) and spinal cord (sc) with differences in the pattern of expression in the brain and spinal cord at 24 hpf. (A,C,E) nrxn2aa is restricted to the anterior portion of the spinal cord. (B,D,F) Expression of nrxn2ab is stronger and more uniform in the spinal cord compared with nrxn2aa. (F) The two rows of signal in the spinal cord are in the region of sensory neurons and motor neurons (inset). (G and H) Expression of nrxn2aa and nrxn2ab is developmentally regulated and shows specific expression patterns in brain regions at 31 hpf. (I) Fluorescent in situ hybridization of nrxn2ab at 24 hpf enables single cell imaging at higher resolution. (JL) Fluorescent in situ hybridization coupled with Isl1 immunohistochemistry shows expression of nrxn2ab RNA in Rohon beard (RB) sensory neurons and motor neurons (MN) in the spinal cord (asterisks).

Figure 3.

Expression pattern of nrxn2aa and nrxn2ab in zebrafish displays isoform-specific differences and expression in motor neurons. (AD) From whole mount RNA in situ hybridization, nrxn2aa and nrxn2ab are expressed in telencephalon (tel), midbrain (mb), hindbrain (hb) and spinal cord (sc) with differences in the pattern of expression in the brain and spinal cord at 24 hpf. (A,C,E) nrxn2aa is restricted to the anterior portion of the spinal cord. (B,D,F) Expression of nrxn2ab is stronger and more uniform in the spinal cord compared with nrxn2aa. (F) The two rows of signal in the spinal cord are in the region of sensory neurons and motor neurons (inset). (G and H) Expression of nrxn2aa and nrxn2ab is developmentally regulated and shows specific expression patterns in brain regions at 31 hpf. (I) Fluorescent in situ hybridization of nrxn2ab at 24 hpf enables single cell imaging at higher resolution. (JL) Fluorescent in situ hybridization coupled with Isl1 immunohistochemistry shows expression of nrxn2ab RNA in Rohon beard (RB) sensory neurons and motor neurons (MN) in the spinal cord (asterisks).

We observed spatial and temporal differences in expression patterns between the two isoforms at 20 (Supplementary Material, Fig. S1A and B), 24 (Fig. 3A and B), 31 (Fig. 3G and H) and 48 hpf (Supplementary Material, Fig. S1C and SD). nrxn2aa is strongly expressed in the telencephalic and midbrain regions (Fig. 3A and C), and spinal cord expression appeared to be restricted to the anterior region at 24 hpf (Fig. 3C and E). nrxn2ab is more strongly expressed in the spinal cord at 24 hpf (Fig. 3B and D) and does not appear to be restricted to the anterior regions but displayed a more uniform expression in the spinal cord (Fig. 3F). The isoform-specific expression patterns suggest a difference in regulation and function of both isoforms during development. We found that both nrxn2aa and nrxn2ab are expressed in regions of the spinal cord containing MN and Rohon Beard sensory neurons (RB) (Fig. 3F, inset). To confirm expression in MNs, fluorescent in situ hybridization of nrxn2aa (Supplementary Material, Fig. S1E) and nrxn2ab (Fig. 3I and J) was performed together with whole mount immunostaining using an α-isl1 antibody, which labels MN and RB sensory neurons, as well as some ventral interneurons (Fig. 3K). Both nrxn2aa and nrxn2ab showed co-localization with isl1 in MN and RB (Fig. 3L, Supplementary Material, Fig. S1E, asterisks). This demonstrates that both nrxn2a isoforms are expressed in MNs in zebrafish and likely in other subtypes of neurons as well.

### Knock-down of nrxn2a results in motor axon defects and phenocopies SMN defects

Next, we assessed the effects of nrxn2a knock-down on MNs by using splice-blocking antisense Morpholino Oligos (MOs). As controls, mismatch MOs (MM) were designed containing five C-G base pair replacements, having identical length, Tm and C-G content, but being unable to target the original splice junction (55) (Supplementary Material, Table 1). Distinct MOs were designed to knock-down each splice isoform specifically (Supplementary Material, Fig. S2A and B). For nrxn2aa, the MOs were designed against exon 1–intron 1 and intron 1–exon 2 splice junctions, and named nrxn2a MO1 and nrxn2a MO2, respectively (Supplementary Material, Fig. S2A). Owing to use of a different promoter and ATG start codon, the sequence of exon 1 of nrxn2ab corresponds to a portion of intron 18 sequence found in nrxn2aa (Fig. 1A). For nrxn2ab, the MOs were designed against exon 1–intron 1 and the intron 1–exon 2 splice junctions, and named nrxn2a MO3 and nrxn2a MO4, respectively (Supplementary Material, Fig. S2B). By targeting the exon 1–intron 1 boundary of nrxn2ab, MO3 specifically affects pre-mRNA splicing of nrxn2ab, but not that of nrxn2aa (Supplementary Material, Fig. S2B). MO4 is able to bind both nrxn2aa and nrxn2ab transcripts, thus blocking splicing of both isoforms. The efficiency of splice blocking was confirmed by RT–PCR (Supplementary Material, Fig. S2C).

Earlier reports have shown that the knock-down of SMN in zebrafish leads to truncation and branching of caudal primary (CaP) motor axons (19,56). We repeated the SMN knock-down using the published MO sequence (Supplementary Material, Table 1). As expected, we observed a significant increase (P < 0.0001) in the percentage of embryos with CaP axon defects such as branching and truncations in SMN MO-injected embryos (68.18 ± 4.488% SEM, n = 535 embryos) compared with uninjected controls (15.19 ± 1.957% SEM, n = 293 embryos) or SMN MM-injected embryos (26.98 ± 8.192% SEM, n = 134 embryos) (Supplementary Material, Fig. S3B and E). These defects could be rescued by co-injection of full-length SMN mRNA (Supplementary Material, Fig. S3D and F). We next tested whether a knock-down of nrxn2aa and nrxn2ab could induce similar defects and phenocopy the effects of SMN deficiency. The knock-down of both nrxn2a isoforms exhibited similar CaP axon truncations and branching defects (Fig. 4B–D and F–H) as found in SMN MO-injected embryos (Supplementary Material, Fig. S3B). As the α-znp1 antibody used in this study specifically labels synaptotagmin 2 on motor axons and not the entire CaP axon (47), it is possible that the truncation and branching defects observed by znp-1 immunohistochemistry were restricted to the pre-synaptic terminals, whereas the remaining parts of the axons were still intact. To test this, a HB9:mCherry transgenic line was generated to label primary MNs with a cytoplasmic mCherry fluorescent reporter. Confocal imaging of live HB9:mCherry embryos injected with MO3+4 confirmed that the knock-down of Nrxn2a results in truncation and bifurcation defects of CaP axons (Fig. 4F–H). Identical phenotypes were observed upon injection of individual MOs (Fig. 4F and I) or in combination (Fig. 4G and H), which suggests that knock-down of each individual isoform produces similar phenotypes. The mismatch morpholino MM4 was used as a control and showed no significant difference in the percentage of CaP axon defects between MM4-injected embryos (13.37 ± 2.707% SEM, n = 150 embryos) and uninjected controls (14.82 ± 2.093% SEM, n = 524 embryos), while the increase of CaP axon defects was significant in MO4-injected embryos (70.52 ± 4.187% SEM, n = 316 embryos) (Fig. 4I and M). The overall morphology of the nrxn2a MO-injected embryos as well as somite and spinal cord development appeared normal (Fig. 4O) and the observed defects did not appear to be due to developmental delay or general deformation.

Figure 4.

Knock-down of nrxn2a isoforms produces similar truncation and branching defects of CaP axons as found upon SMN knock-down with rescue by nrxn2ab (A) Uninjected controls have CaP axons with normal extensions toward ventral targets as shown by immunohistochemistry with α-znp1. (BD) Knock-down of nrxn2aa, nrxn2ab or both isoforms results in truncations (arrowheads) and branching defects of CaP axons (numbers in brackets indicate MO concentration in mg/ml). (FH) Similar truncation (arrowheads) and branching (arrow) defects of CaP axons were observed upon injection of (F) individual nrxn2a MOs or (G and H) in combination, as shown by live imaging of HB9:mCherry transgenic line. (J) Injection of mismatch morpholino nrxn2a MM4 did not cause CaP axon phenotypes. (K) Over-expression of full-length nrxn2ab mRNA (50 ng/μl) is able to rescue the phenotypes caused by knock-down of nrxn2a. (L) Full-length nrxn2ab mRNA alone did not cause any CaP axon phenotypes. (M) Summary graph showing significant increase in percentage of embryos with CaP axon defects in MO4-injected embryos (n = 316 embryos) compared with uninjected controls (n = 524 embryos) and MM4-injected embryos (n = 150 embryos). (N) Over-expression of nrxn2ab mRNA rescues the phenotype in nrxn2a MO4 injected. There is a significant reduction in the percentage of embryos with CaP axon defects in nrxn2a MO4+nrxn2ab mRNA coinjected embryos (n = 95 embryos) compared with MO4-injected embryos (n = 86 embryos). There was no significant difference in the percentage of embryos with phenotypes between coinjected embryos, mRNA-injected embryos (n = 92 embryos) or uninjected controls (n = 83 embryos). (O and P) Development of somites, spinal cord and overall morphology was normal in nrxn2a MO-injected embryos. (QT) Truncated axon (arrowhead) in nrxn2a MO-injected embryos can still extend ventrally after pausing at choice point. Images from timelapse video of 29 hpf to 31 hpf nrxn2a MO4-injected HB9:mCherry. (UY) Rescue of CaP axon phenotypes in SMN MO by co-injection of nrxn2ab mRNA (50 ng/μl). Axon phenotype in non-injected control embryo (U), SMN MO (V), SMN MO and nrxn2ab mRNA co-injected (W) as well as nrxn2ab mRNA-injected embryo (X). There was a significant reduction of axon defects in co-injected embryos (y; n = 78 embryos) compared with SMN MO-injected embryos (n = 94 embryos). There was no significant difference between co-injected embryos and uninjected controls (35.55 ± 7.78% SEM, n = 94 embryos) or nrxn2ab mRNA-injected embryos (35.85 ± 7.80% SEM, n = 83 embryos). Data were obtained from at least three-independent biological replicates.

Figure 4.

Knock-down of nrxn2a isoforms produces similar truncation and branching defects of CaP axons as found upon SMN knock-down with rescue by nrxn2ab (A) Uninjected controls have CaP axons with normal extensions toward ventral targets as shown by immunohistochemistry with α-znp1. (BD) Knock-down of nrxn2aa, nrxn2ab or both isoforms results in truncations (arrowheads) and branching defects of CaP axons (numbers in brackets indicate MO concentration in mg/ml). (FH) Similar truncation (arrowheads) and branching (arrow) defects of CaP axons were observed upon injection of (F) individual nrxn2a MOs or (G and H) in combination, as shown by live imaging of HB9:mCherry transgenic line. (J) Injection of mismatch morpholino nrxn2a MM4 did not cause CaP axon phenotypes. (K) Over-expression of full-length nrxn2ab mRNA (50 ng/μl) is able to rescue the phenotypes caused by knock-down of nrxn2a. (L) Full-length nrxn2ab mRNA alone did not cause any CaP axon phenotypes. (M) Summary graph showing significant increase in percentage of embryos with CaP axon defects in MO4-injected embryos (n = 316 embryos) compared with uninjected controls (n = 524 embryos) and MM4-injected embryos (n = 150 embryos). (N) Over-expression of nrxn2ab mRNA rescues the phenotype in nrxn2a MO4 injected. There is a significant reduction in the percentage of embryos with CaP axon defects in nrxn2a MO4+nrxn2ab mRNA coinjected embryos (n = 95 embryos) compared with MO4-injected embryos (n = 86 embryos). There was no significant difference in the percentage of embryos with phenotypes between coinjected embryos, mRNA-injected embryos (n = 92 embryos) or uninjected controls (n = 83 embryos). (O and P) Development of somites, spinal cord and overall morphology was normal in nrxn2a MO-injected embryos. (QT) Truncated axon (arrowhead) in nrxn2a MO-injected embryos can still extend ventrally after pausing at choice point. Images from timelapse video of 29 hpf to 31 hpf nrxn2a MO4-injected HB9:mCherry. (UY) Rescue of CaP axon phenotypes in SMN MO by co-injection of nrxn2ab mRNA (50 ng/μl). Axon phenotype in non-injected control embryo (U), SMN MO (V), SMN MO and nrxn2ab mRNA co-injected (W) as well as nrxn2ab mRNA-injected embryo (X). There was a significant reduction of axon defects in co-injected embryos (y; n = 78 embryos) compared with SMN MO-injected embryos (n = 94 embryos). There was no significant difference between co-injected embryos and uninjected controls (35.55 ± 7.78% SEM, n = 94 embryos) or nrxn2ab mRNA-injected embryos (35.85 ± 7.80% SEM, n = 83 embryos). Data were obtained from at least three-independent biological replicates.

To ensure that the CaP axon phenotypes are specific for the knock-down of nrxn2a, we tested whether over-expression of full-length nrxn2ab mRNA could rescue axonal phenotypes in nrxn2a morphants. Co-injection of nrxn2a MO4 with full-length nrxn2ab mRNA (Fig. 4K) reduced the percentage of embryos with CaP axon defects (27.04 ± 7.549% SEM, n = 95 embryos) when compared with embryos injected with MO4 alone (59.90 ± 5.696% SEM, n = 86 embryos) (Fig. 4N). The CaP axons in the rescued embryos had normal extensions and did not appear to have any morphological defects (Fig. 4K). The injection of full-length nrxn2ab mRNA alone (Fig. 4L) did not result in any significant appearance (P > 0.05) of CaP axon defects (27.92 ± 5.276% SEM, n = 92 embryos) when compared with uninjected controls (15.35 ± 5.764% SEM, n = 83 embryos) (Fig. 4I and N). This suggests that over-expression of nrxn2ab was able to rescue the phenotypes of the nrxn2a knock-down. As an additional control, we also tested the hypothetical possibility that nrxn2a defects can be rescued by SMN over-expression. As expected, we found no rescue of CaP axon phenotypes in nrxn2a morphants with simultaneously overexpressed SMN consistent with the idea that nrxn2a acts downstream of SMN (Supplementary Material, Fig. 3G, arrowheads, arrow). There was no significant difference in the percentage of CaP axon defects in embryos co-injected with SMN mRNA and nrxn2a MO4 (83.39 ± 9.089% SEM, n = 104 embryos) when compared with nrxn2a MO4-injected embryos (72.33 ± 1.024% SEM, n = 91 embryos) (Supplementary Material, Fig. S3H). We next asked whether the injection of nrxn2a mRNA could rescue CaP axon phenotypes observed in SMN MO embryos. Owing to inherent limitations in producing long mRNA transcripts in vitro with efficient 5′ capping, it was not possible to obtain functional nrxn2aa mRNA (5013 nt) in sufficient amounts. Therefore, we chose the shorter full-length nrxn2ab mRNA (1926 nt) for the rescue experiment. Notably, co-injection of SMN MO with nrxn2ab mRNA significantly reduced the percentage of embryos with CaP axon defects (51.39 ± 6.05% SEM, n = 78 embryos) compared with SMN MO-injected embryos (75.43 ± 2.18% SEM, n = 94 embryos) (Fig. 4U–Y). The rescued embryos had normal CaP axon extensions that were similar to uninjected controls (Fig. 4U and W). Taken together, this partial rescue of SMN morphants by nrxn2ab mRNA injection strongly suggests that nrxn2a acts as a downstream mediator of SMN.

### Knock-down of nrxn2a results in delayed axon outgrowth and reduced Ca2+ influx at pre-synaptic axon terminals in zebrafish

It has been reported that truncated CaP axons in SMN morphants retain their ability to extend beyond the choice point and eventually grow ventrally toward the second intermediate target but remain truncated and eventually branch at later stages (56). To check if there were similar effects after nrxn2a knock-down, time lapse live imaging of HB9:mCherry transgenic line was employed and showed that the truncated axons of nrxn2a MO-injected embryos were still able to extend past the choice point and were able to grow ventrally toward the target and remain truncated (Fig. 4Q–T, arrowhead). This indicated that the nrxn2a knock-down did not result in complete and irreversible axon outgrowth defects but suggests that the truncated CaP axons pause at the choice point for a longer duration than normal, before resuming its axonal growth activity. The first NMJ synapses are formed at the choice point between 20 and 24 hpf, after which the axons resume their growth toward ventral targets (54,57). The delay in axon outgrowth observed both in SMN and nrxn2a morphants opens the possibility that there are defects in NMJ synapse establishment or stability, leading to defective synaptic transmission and causing the CaP axons to pause for a longer period under SMN- and nrxn2a-deficient conditions.

Abnormal synaptic transmission as well as a significant reduction in quantal content has been reported in SMA mouse models (28,29). Consistent with this, we showed an approximate 2-fold reduction of evoked Ca2+ influx into the pre-synaptic motor axon terminal upon glutamate stimulation in a zebrafish model of SMA in vivo (See and Winkler; HMG under revision). On the other hand, α-neurexins are required for neurotransmitter release at the NMJ (45) and perform an essential role of functionally coupling Ca2+ channels to the exocytosis of synaptic vesicles (43). We therefore tested whether the knock-down of nrxn2a can phenocopy the SMN defect and results in similar reduced evoked Ca2+ influx into the pre-synaptic axon terminal and reduced motor axon excitability. For this, live Ca2+ imaging was employed on nrxn2a MO4 and control MM4 injected as well as uninjected embryos of the HB9:D3cpv/MNsensor line (See and Winkler; HMG under revision).

After knock-down of both nrxn2a isoforms, we observed a significant reduction (P < 0.0001) of the average magnitude of evoked Ca2+ influx into the pre-synaptic axon terminal after glutamate stimulation in nrxn2a MO4-injected embryos (20.86 ± 1.592% SEM, n = 31 recordings) (Fig. 5B), compared with uninjected controls (41.03 ± 6.017% SEM, n = 13 recordings) (Fig. 5A) or MM4-injected embryos (34.37 ± 3.859% SEM, n = 18 recordings) (Fig. 5C–E). There was an approximate 2-fold reduction in the average magnitude of evoked Ca2+ influx in nrxn2a MO4-injected embryos (Fig. 5D and E), which was similar to the situation observed in pre-synaptic axon terminals under SMN-deficient conditions (See and Winkler; HMG under revision). To test if the defect is specific for the pre-synaptic axon terminal, Ca2+ imaging was performed on the cell body of nrxn2a MO4 injected as well as uninjected embryos. We observed no significant difference between the average magnitude of Ca2+ response in the cell body of nrxn2 MO4 injected (10.620 ± 1.190 SEM, n = 9 recordings) compared with uninjected embryos (11.771% ± 1.206 SEM, n = 12 recordings) (Fig. 5F). This suggests that the defect in Ca2+ response is restricted to the pre-synaptic axon terminal, which is strikingly consistent with phenotypes observed in SMN deficiency (See and Winkler; HMG under revision). Taken together, this strongly suggests that knock-down of nrxn2a can phenocopy the knock-down of SMN and leads to reduced excitability of the pre-synaptic axon terminal in vivo.

Figure 5.

Reduced nrxn2a levels result in reduction of evoked Ca2+ influx into pre-synaptic axon terminal. (A) Addition of K+ Glutamate causes transient Ca2+ influx (asterisk) into pre-synaptic axon terminal in uninjected controls. (B) Knock-down of nrxn2a results in reduced magnitude of evoked Ca2+ influx into pre-synaptic axon terminal in MO4-injected embryos. (C) MM4-injected embryos display normal evoked Ca2+ influx (asterisk) into pre-synaptic axon terminal. (D) Representative graphs of %ΔR/Ro at pre-synaptic axon terminal in uninjected controls, MO4-injected embryos and MM4-injected embryos. (E) Summary graph of average magnitude of evoked Ca2+ influx into pre-synaptic axon terminal after nrxn2a knock-down. Significant reduction of %ΔR/Ro at the pre-synaptic axon terminal in MO4-injected embryos (n = 31 recordings) compared with uninjected controls (n = 13 recordings) or MM4-injected controls (n = 18 recordings). (F) Normal evoked Ca2+ response in MN cell body of MO4-injected embryos. There was no statistically significant difference in %ΔR/Ro between MO4 injected (n = 9 recordings) and uninjected control embryos (n = 12 recordings).

Figure 5.

Reduced nrxn2a levels result in reduction of evoked Ca2+ influx into pre-synaptic axon terminal. (A) Addition of K+ Glutamate causes transient Ca2+ influx (asterisk) into pre-synaptic axon terminal in uninjected controls. (B) Knock-down of nrxn2a results in reduced magnitude of evoked Ca2+ influx into pre-synaptic axon terminal in MO4-injected embryos. (C) MM4-injected embryos display normal evoked Ca2+ influx (asterisk) into pre-synaptic axon terminal. (D) Representative graphs of %ΔR/Ro at pre-synaptic axon terminal in uninjected controls, MO4-injected embryos and MM4-injected embryos. (E) Summary graph of average magnitude of evoked Ca2+ influx into pre-synaptic axon terminal after nrxn2a knock-down. Significant reduction of %ΔR/Ro at the pre-synaptic axon terminal in MO4-injected embryos (n = 31 recordings) compared with uninjected controls (n = 13 recordings) or MM4-injected controls (n = 18 recordings). (F) Normal evoked Ca2+ response in MN cell body of MO4-injected embryos. There was no statistically significant difference in %ΔR/Ro between MO4 injected (n = 9 recordings) and uninjected control embryos (n = 12 recordings).

### Nrxn2a shows altered expression in a SMA mouse model

To investigate whether SMN deficiency also leads to alterations in the expression of neurexin isoforms in a mammalian model of SMA, we investigated Nrxn2 expression in Smn−/−; SMN2 mice. These mice express two copies of the human SMN2 gene on a mouse Smn null background and thus resemble the most severe form of SMA in humans (7). Affected mice show severe signs of paralysis at birth, and only few of these animals survive up to 5 days after birth. Interestingly, the number of MN cell bodies in the spinal cord is only mildly reduced at pre-final stages of disease indicating that the severe paralysis found in this mouse model is not caused by increased rates of cell death but by alterations in axons and neurotransmission at the neuromuscular endplate. Isolated MNs from Smn−/−/SMN2 mouse embryos survive in culture similarly as wild-type MNs in the presence of neurotrophic factors, but they show significantly shorter axons and smaller growth cones after 7 days in culture (Fig. 6A) (27). Furthermore, abnormalities in spontaneous excitability are observed that predominantly affect the axons (32). This corresponds to altered clustering of VGCCs in axonal growth cones (32).

In a first approach, we investigated Nrxn2 expression in MNs of Smn−/−/SMN2 mice that were isolated at embryonic day 13.5 (E13.5) and cultured for 7 days in vitro. These isolated SMA MNs (Smn−/−/SMN2) did not show any alterations in Nrxn2α or Nrxn2β expression (Nrxn2α: 104.0 ± 8.8%; t = 0.453, df = 3, P = 0.6811; Nrxn2β: 98.7 ± 5.4%; t = 0.229, df = 3, P = 0.8330; n = 4 independent experiments, one sample t-test) when compared with control mice (Smn+/+/SMN2; control: 100%) as revealed by qRT–PCR (Fig. 6B). We then tested for any changes in the expression of Nrxn2 mRNA in SMA MNs in vivo. This approach takes into account that the expression of Nrxn2 could be influenced by factors from contacting cells, i.e. sensory neurons that project to spinal MNs, or contact with muscle cells. Recent reports suggested that SMN deficiency has a major effect on sensory neurons and interneurons that project to MNs, and therefore, this effect could contribute to altered Nrxn2 expression in MNs (23). For this purpose, we isolated the lumbar spinal cords from E14, E18 and 2 days old (P2) Smn−/−/SMN2 mice and investigated the level of Nrxn2 mRNA expression by qRT–PCR. No difference in the expression of Nrxn2α or Nrxn2β was observed at E14 (Nrxn2α: 109.9 ± 17.8%; t = 0.556, df = 2, P = 0.6338; Nrxn2β: 112.9% ± 40.8%; t = 0.3157, df = 2, P = 0.7821; n = 3) or E18 (Nrxn2α: 94.6 ± 19.7%; t = 0.270, df = 2, P = 0.8125; Nrxn2β: 92.6 ± 6.5%; t = 1.123, df = 2, P = 0.3780; n = 3), but the P2 SMA mouse spinal cord showed a significant decrease in the expression of Nrxn2α (Nrxn2α: 72.7 ± 3.7%; t = 7.170, df = 2, P = 0.0189; Nrxn2β: 97.8 ± 14%; t = 0.150, df = 2, P = 0.8939; n = 3). As control, expression levels of Nrxn1α and Nrxn3α were investigated at P2 by qRT–PCR revealing that Nrxn1α and Nrxn3α are unaltered (Nrxn1α: 91.4 ± 2.1%; t = 3.999, df = 2, P = 0.0572; Nrxn3α: 96.8 ± 0.5%; t = 5.301, df = 2, P = 0.0338; n = 3). This indicates that the effect is specific for Nrxn2 mRNA (Fig. 6C–E). Defects in axon elongation and excitability in Smn−/−/SMN2 MNs are normally not observed within the first 3 or 4 days of culture, when MNs are isolated from embryonic day 13.5 or E14 Smn−/−/SMN2 mouse spinal cords (27). These defects develop only between Day 4 and 7 in culture, and thus possibly reflect a later developmental stage when MNs have already contacted skeletal muscle in vivo and when the pre-synaptic terminals differentiate to form functional neuromuscular endplates. Therefore, we investigated the expression of Nrxn2α and Nrxn2β in MNs at P2, a stage corresponding to MNs isolated at E14 and cultured for an additional 7 or 8 days. Interestingly, in these Smn−/−/SMN2 spinal MNs microdissected at P2, a >40% reduction of Nrxn2α expression was observed (Nrxn2α: 51.6 ± 1.5%; t = 31.86, df = 2, P = 0.0010; n = 3) (Fig. 6F–G), which is consistent with the effect observed in zebrafish (Fig. 1). Conversely, Nrxn2β expression was not altered (Nrxn2β: 164.1 ± 37.8%; t = 1.696, df = 2, P = 0.2320; n = 3). These data indicate that Nrxn2 expression is altered in both zebrafish and mouse SMA models albeit with possible species-specific differences.

Figure 6.

Altered expression of Neurexin isoforms in SMA mouse. (A) Motor neurons from wild-type and Smn−/−; SMN2 E13.5 embryos cultured for 7 days in vitro showing reduced axon outgrowth in Smn−/−; SMN2. (BE) Fold change in Nrxn2α and Nrxn2β expression by qRT–PCR using cDNA from (B) isolated E13.5 wild-type and Smn−/−; SMN2 mouse motor neurons cultured 7 days in vitro, (C) lumbar spinal cord of E14 mouse embryos, (D) lumbar spinal cord of E18 mouse embryos, (E) lumbar spinal cord of postnatal day 2 mice. Expression of Nrxn1α and Nrxn3α transcripts is also shown. (F) Cryosections of the P2 lumbar spinal cord stained with cresyl violet showing motor neurons before laser microdissection (encircled by red line) and after microdissection. (G) Fold change in Nrxn2 expression by qRT–PCR using cDNA from laser microdissected motor neurons of postnatal day 2 mice. (H and I) Analysis of exon11 skipping at SS3 in Nrxn2α transcripts using semi-quantitative RT–PCR. The alternative splicing assay from postnatal day 2 spinal cord cDNA shows increased reduction of PCR products including exon 11 (upper band) as compared with reduction in transcripts excluding exon11 (lower band) in Smn−/−; SMN2 when compared with wild-type. (J) qRT–PCR confirms increased skipping of exon 11 relative to total Nrxn2α transcripts in Smn−/−; SMN2 (n = 2) postnatal day 2 spinal cord. Gapdh was used as normalization control. Data shown represent mean ± SEM., n = 3 or more independent experiments in each case unless otherwise stated.

Figure 6.

Altered expression of Neurexin isoforms in SMA mouse. (A) Motor neurons from wild-type and Smn−/−; SMN2 E13.5 embryos cultured for 7 days in vitro showing reduced axon outgrowth in Smn−/−; SMN2. (BE) Fold change in Nrxn2α and Nrxn2β expression by qRT–PCR using cDNA from (B) isolated E13.5 wild-type and Smn−/−; SMN2 mouse motor neurons cultured 7 days in vitro, (C) lumbar spinal cord of E14 mouse embryos, (D) lumbar spinal cord of E18 mouse embryos, (E) lumbar spinal cord of postnatal day 2 mice. Expression of Nrxn1α and Nrxn3α transcripts is also shown. (F) Cryosections of the P2 lumbar spinal cord stained with cresyl violet showing motor neurons before laser microdissection (encircled by red line) and after microdissection. (G) Fold change in Nrxn2 expression by qRT–PCR using cDNA from laser microdissected motor neurons of postnatal day 2 mice. (H and I) Analysis of exon11 skipping at SS3 in Nrxn2α transcripts using semi-quantitative RT–PCR. The alternative splicing assay from postnatal day 2 spinal cord cDNA shows increased reduction of PCR products including exon 11 (upper band) as compared with reduction in transcripts excluding exon11 (lower band) in Smn−/−; SMN2 when compared with wild-type. (J) qRT–PCR confirms increased skipping of exon 11 relative to total Nrxn2α transcripts in Smn−/−; SMN2 (n = 2) postnatal day 2 spinal cord. Gapdh was used as normalization control. Data shown represent mean ± SEM., n = 3 or more independent experiments in each case unless otherwise stated.

As mentioned before alterations in the splicing pattern of neurexins have been reported in several cases (51,58–60). We therefore tested whether in analogy to the situation of elevated nrxn2aa exon 12 skipping in SMN-deficient zebrafish (Fig. 2), the corresponding exon 11 of mouse Nrxn2α is similarly affected in Smn−/−/SMN2 mice. Therefore, we designed primers flanking exon 11 and isolated cDNA from P2 Smn−/−/SMN2 spinal cords. We observed a reduction in both exon 11 including and exon 11 excluding PCR products. Interestingly, however, the levels of exon 11 including transcripts were significantly reduced to a larger extent than those of exon11 excluding transcripts (Fig. 6H and I; Nrxn2α exon11 including transcript: 35.7 ± 14.8%; t = 4.327, df = 3, P = 0.0228; Nrxn2α exon11 excluding transcript: 61.6 ± 18.7%; t = 2.044, df = 3, P = 0.1336.; n = 4). This was confirmed by qRT–PCR using primers to detect Nrxn2α exon11 levels relative to total Nrxn2α transcripts. A significant reduction in exon11 containing transcripts was observed relative to total Nrxn2α in Smn−/−; SMN2 postnatal day 2 spinal cords (Fig. 6J; 77.76±0.8%; t = 27.68, df = 1, P = 0.0230; n = 2). Together, this suggests altered splicing of Nrxn2α also in the spinal cord of SMN-deficient mice.

## DISCUSSION

Two main hypotheses have been discussed in the past to explain how reduced levels of ubiquitously expressed SMN lead to MN-specific degeneration observed in SMA. Several studies proposed MN-specific functions for SMN (11,26,27). On the other hand, a critical role of SMN for U snRNP assembly has been established, and it was proposed that MNs are particularly vulnerable to reduced splicing efficiency (12–14,16,61). This is supported by findings that the minor spliceosome and splicing of transcripts containing minor introns are significantly affected by SMN deficiency (21). Most recently, splicing of Stasimon mRNA containing a minor intron was shown to be affected in a Drosophila model for SMA (21). These authors showed that Stasimon is not required in MNs, but in sensory- and interneurons implying a non-cell autonomous role for SMN in neural circuit function. In the present study, we found that the transcript for nrxn2a is aberrantly spliced after SMN knock-down in zebrafish. Interestingly, nrxn2 does not contain minor introns and there was no obvious enrichment of minor intron containing transcripts in our approach. From the overlapping pool of transcripts identified, the percentage of annotated transcripts containing minor introns that were significantly deregulated was only 0.68%. This suggests that mis-splicing of transcripts processed by the major spliceosome can also contribute to MN defects under SMN deficiency. The nrxn2a mRNA is present in MNs as well as other neuronal cells, and displayed reduced expression and altered splicing patterns upon SMN deficiency in both zebrafish and mouse. Using live Ca2+ imaging in zebrafish, we show that a knock-down of nrxn2a leads to reduced excitability of motor axons. This establishes nrxn2a, encoding a pre-synaptic axonal membrane protein, as a novel downstream target of SMN.

### Identification of novel candidates implicated in SMA phenotype

We used transcriptome analyses in zebrafish to identify transcripts that are down-regulated in embryos with reduced SMN and Gemin2 activities. Whole embryos rather than isolated MNs were used in this approach to take into consideration transcripts expressed in non-MN cell types that could potentially mediate non-cell autonomous effects. We chose an unbiased approach by extending the search beyond U12 intron containing transcripts, and screened for genes down-regulated under both SMN- and Gemin2-deficient conditions. SMN and Gemin2 interact closely during U snRNP assembly and a deficiency in either factor results in MN defects (19). We identified a total of 737 transcripts that were down-regulated in both types of morphants. This number was narrowed down by shortlisting transcripts for which neuronal expression or function has been described previously. Besides nrxn2a, we also identified syn2b and hlxb9lb (Table 1; and data not shown). Both genes are highly expressed in MNs and strongly down-regulated under SMN- and Gemin2-deficient conditions. Interestingly, knock-down of syn2b and hlxb9lb resulted in similar motor axon truncations and branching phenotypes as observed upon SMN knock-down (Supplementary Material, Fig. S5; and data not shown). This suggests that the approach we have chosen is suitable for identifying genes with important functions in motor axons. Future studies will be required to investigate whether other candidate genes identified in our approach are aberrantly spliced in SMN morphants.

nrxn2a was strongly down-regulated in SMN morphants, with clear differences in the extent of down-regulation for two analyzed isoforms, nrxn2aa and nrxn2ab (Table 2). Nrxn genes encode for neuronal pre-synaptic membrane proteins (48,50), which play important roles in mediating cell adhesion at the synapse (42), synapse formation and development (62,63) and in triggering neurotransmitter release at the NMJ synapse (43,45). All classes of neurons co-express multiple Nrxns in different permutations and ratios, and isoform-specific expression and alternative splicing are highly regulated (50). In zebrafish, we identified isoform-specific differences in the expression patterns of nrxn2aa and nrxn2ab. This suggests that nrxn2aa and nrxn2ab have different regulation and possibly different functions. It has been reported that the α and β isoforms of Nrxn have different expression patterns in rat neurons and are expressed independently of each other, due to the use of two independent promoters (50). We found that the expression of nrxn2ab is stronger and more uniform than the expression of nrxn2aa in the spinal cord. This is consistent with the finding that the rat Nrxn2b isoform exhibited a more uniform expression pattern compared with the Nrxn2a isoform, which appeared to be highly expressed only in selective cell types (50).

### Changes in alternative splicing patterns of nrxn2a upon SMN depletion

We observed a change in alternative splicing patterns of nrxn2aa at alternative splice site SS3 when SMN levels were reduced in either zebrafish or mouse. The observed exon 12 skipping at SS3 of nrxn2aa in zebrafish, respectively, exon 11 in mouse is consistent with the skipping of exon 11 of similar length at SS3 of its ortholog Nrxn2a in rat (51). Notably, the increase in exon skipping at SS3 has been reported to be Ca2+ dependent and modulated by depolarization of neurons, suggesting that the alternative splicing pattern of nrxn2aa could be dependent on the level of neuronal activity (51). Accordingly, these splicing changes could be a consequence and not necessarily the cause of the observed reduction in evoked Ca2+ influx under SMN-deficient conditions. This would be consistent with the observation that altered Nrxn2a splicing in the mouse is only detectable postnatally, when the frequency of MN firing increases in controls. It thus remains to be shown whether a SMN deficiency affects splicing of Nrxn2a directly or indirectly via reduced Ca2+ influx, or a combination of both.

Neuronal activity-dependent splicing at SS4 of Nrxn1 by SAM68 is known to regulate inclusion or exclusion of exon 20, which displays different binding affinities to postsynaptic partner neuroligin 1B (58,59). It is possible that the exclusion of exon 12 at SS3 of nrxn2aa is modulated by similar mechanisms mediated through other proteins such as neuron-specific polypyrimidine tract binding protein, which was shown to play a role in nrxn2a alternative splicing at SS4 (60). The SS3 alternative splice site is found within the fourth extracellular LNS domain of nrxn2aa, which is important for binding to postsynaptic partners at the synapse (39,42). The essential role of Nrxn in triggering neurotransmitter release at synapses depends on its extracellular structure (44). It is therefore possible that the exclusion of zebrafish exon 12, respectively, mouse exon 11 at SS3 changes the binding properties and thus modulates synaptic coupling properties (51). We thus hypothesize that reduced SMN levels affect nrxn2aa alternative splicing patterns, which consequently compromises the levels of intracellular Ca2+ within the pre-synaptic axon terminal and neuronal activity at the synapse.

Importantly, we observed that co-injection of nrxn2ab mRNA partially rescues the MN defects caused by SMN deficiency (Fig. 4). Surprisingly, the knock-down of Gemin2, which interacts with SMN during U snRNP assembly (14,15), did not change alternative splicing of nrxn2aa at SS3 with similar efficiencies. This suggests that the particular pattern of alternative splicing observed in nrxn2aa is specific for the knock-down of SMN. These findings uncover a direct link between nrxn2a mis-splicing in SMN-deficient models and MN defects. Technical limitations prevented us to test the interesting possibility that full-length nrxn2aa alone or in combination with nrxn2ab can further improve the rescue efficiency. However, we believe that a complete rescue of SMN deficiency by nrxn2a alone is unlikely as other mis-splicing events affecting additional targets are expected to contribute to the SMN-deficient phenotype, as has been recently demonstrated for Stasimon (21).

### Defective Ca2+ influx in zebrafish nrxn2a morphants, Nrxnα knock-out mice and SMA mouse models

We found that the knock-down of nrxn2a does not lead to complete axon outgrowth defects. Instead, truncated axons are still able to extend ventrally toward their muscle target, however, with a significant delay. The axons pause at the choice point for a longer duration than normal axons. This phenotype is similar to that observed after SMN knock-down in zebrafish, where the truncated CaP axons are also able to extend toward their ventral target after pausing at the choice point and before eventually becoming branched at later time points (56). Albeit speculative, it is tempting to suggest that the delayed outgrowth is attributed to defects in the early stages of maintenance of the NMJ synapses.

In SMA mouse models, electrophysiological recordings of tissue slices from sacrificed animals provided evidence for pre-synaptic defects in MNs, such as abnormal synaptic transmission (28,29) and intraterminal Ca2+ level alterations during repetitive stimulation (31). Furthermore, Ca2+ imaging demonstrated the reduced excitability of motor axons in a zebrafish model of SMA in vivo (See and Winkler; HMG under revision). In the present study, we report similar changes in the excitability of the pre-synaptic axon terminal under nrxn2a-deficient conditions. In mice, there are striking similarities in phenotypes between SMA mouse models and the triple Nrxnα knock-out (Nrxnα KO) mouse (43,44). Both mouse models display impaired synaptic transmission, reduced evoked transmitter release, and reduced quantal content at NMJ synapses (28,29,31,44,45). In addition, the Nrxnα KO mice exhibited low postnatal survival rates and displayed breathing difficulties such as irregular respiratory rhythms in either single, double or triple knock-out mice (43). Abnormal breathing patterns such as smaller ventilation volume, increased breath duration and increased apnea frequency were also observed in SMA mouse models (64). The compromised synaptic transmission and neurotransmitter release would likely affect the release of synaptic vesicles critical for neurotransmission involved in breathing and locomotion that could lead to the respiratory failures observed. Of note, in human patients of type I SMA, progressive respiratory failure often leads to death within the first 2 years (1). Taken together, it is plausible to hypothesize a possible link between SMN deficiency and affected levels of Nrxn2 activity in SMA.

In support of this, we found a strong down-regulation of Nrxn2α in MNs isolated from SMA mice (Smn−/−/SMN2) when compared with controls (Smn+/+/SMN2). This demonstrates that the expression of Nrxn2 is affected in MNs in a mammalian model of SMA, consistent with findings from the zebrafish SMA model. We note that altered expression and splicing changes of Nrxn2α are only observed at later stages, when MNs are differentiated and form functional synapses with skeletal muscle, but not at early embryonic stages. This defect only occurs after neurons have progressed in their differentiation program toward maturation of neuromuscular synapses and have received afferent synaptic inputs from the spinal cord. This progressive change in alternative expression of Nrxn2 isoforms is coincident with the onset of the early symptomatic stage at postnatal day 2 in SMA mice, suggesting that there could be a correlation between the extent of severity with the extent of altered expression patterns. Interestingly, we found a different regulation of α- and β-isoforms in zebrafish and mouse. In zebrafish, both isoforms were down-regulated under SMN-deficient conditions. Possible explanations for this down-regulation could be either aberrant splicing and subsequent NMD, or indirect effects on the regulation of the two promoters driving transcription of the α- and β- isoforms. In the mouse SMA model, we, however, did not observe any significant change in the level of Nrxn2β transcripts. Future studies need to show whether the down-regulation of Nrxn2α transcripts in SMA mice is caused by aberrant splicing and NMD or represents an indirect effect.

In conclusion, we propose that Nrxn2α acts as a novel downstream candidate of SMN involved in the pathomechanism of SMA. We show that SMN deficiency leads to altered splicing and down-regulation of Nrxn2α in MNs and possibly other cell types. Furthermore, we demonstrate that a reduction of nrxn2 levels in zebrafish affects motor axon outgrowth and excitability and phenocopies a SMN knock-down. Further validation of our microarray screen will likely reveal more candidates, furthering our understanding of the pathomechanisms of SMA.

## MATERIALS AND METHODS

### Maintenance of zebrafish and mice

All zebrafish were raised and maintained in compliance with approved protocols and National University of Singapore (NUS) Institute of Animal Care and Use Committee (IACUC) guidelines (protocol numbers 075/07; 082/10; BR19/10). HB9:mCherry and HB9:D3cpv/MN transgenic fish were generated and maintained as described (See and Winkler; HMG under revision). The HB9 promoter was kindly provided by Dirk Meyer (University of Innsbruck). Smn−/−/SMN2 mice (low copy) used in this study are described in (7) and were bred on a FVB/NCrl background.

### Microarrays

Microarrays were performed as previously described (65) with six-independent batches of samples each for SMN MO injected, Control MO injected and seven independent batches for Gemin2 MO injected used as biological replicates. Briefly, 10 μg total RNA extracted from SMN MO, Gemin2 MO- and Control MO-injected whole zebrafish embryos at 48 hpf was used for reverse transcription (RT) reaction to produce amino allyl labeled cDNA (aa-cDNA). Purification of neutralized aa-cDNA was performed by centrifugation in Microcon YM30 column, before fluorescent labeling with mono-functional NHS-ester Cy3 and Cy5 dyes (Amersham Biosciences, UK). Fluorescent aa-cDNA was quenched and QIAquick PCR purification kit (Qiagen) was used for purification. Arrays were hybridized overnight at 42°C using Micro Array User Interface (MAUI) hybridization stations (BioMicro). Scanning of the arrays was performed using the GenePix Pro 4.0 image analysis software on a GenePix 4000B microarray scanner (Axon Instruments, CA, USA). 16 bit TIFF files generated from the scans were used for gridding to separate the spots. Microarray data were analyzed using the GeneSpring GX v10.0 Mac software. Date with fold changes of >2.0 and P values <0.05 were considered for statistical analyses. False discovery rate control using Benjamini–Hochberg was applied as multiple test correction to minimize possibility of false positives. Hierarchical clustering analysis of the changes in expression levels showed consistency across the biological repeats. Microarray data have been uploaded to NCBI GEO database with accession number (GSE47001).

### RNA in situ hybridization and immunohistochemistry

A 719 bp coding sequence present in nrxn2a exon 15 to exon 17, which is unique for the nrxn2aa isoform (transcript ENSDART00000087657) was amplified with Phusion polymerase (Finnzymes) and cloned into TOPO® BLUNT pCRII vector (Invitrogen). For nrxn2ab, a 836 bp region present in the 5′ untranslated region and exon 1, which is specific for the nrxn2ab isoform (transcript ENSDART00000087660) was used. Non-fluorescent and fluorescent in situ hybridization was done as described (66) with minor modifications. Prior to rehydration, embryos were incubated in 3% H2O2/MeOH for 20 min to inactivate endogenous peroxidase activity. Hybridization was performed in 50% formamide, 5× SSC, 50 ng/ml heparin, 0.5 mg/ml Torula RNA (Sigma), 0.1% Tween20 and 9 mM citric acid (pH 6.0–6.5). The Tyramide Signal Amplification TSA™ PLUS Fluorescein and Tetramethylrhodamine kit (PerkinElmer) was used for detection. For double labeling of RNA and protein, fluorescent in situ hybridization was performed first, followed directly by whole mount immunohistochemistry. The following primary antibodies were used: rabbit α-GFP polyclonal (1:500, Abcam, USA), mouse α-Islet1/2 monoclonal (39.45, 1:500, DSHB, USA), mouse α-synaptotagmin2 monoclonal (znp1, 1:100, ZIRC, USA). As secondary antibodies, α-rabbit Alexa488 (1:1000), α-mouse Alexa633 (1:500), α-mouse Alexa488 (1:200; all Molecular Probes, Life Technologies, USA) were used.

### Morpholino knock-down and rescue experiments

MOs were obtained from Genetools, USA. Mismatch control MOs (MM) contain five C-G base substitutions. MOs were injected at a concentration of 3.125 mg/ml into one-cell stage embryos, unless stated otherwise. RT–PCR with primers flanking the targeted splice sites was used to detect changes in splicing patterns after MO injection. Increase in product size suggested intron retention, a decrease suggested exon skipping and use of an alternative splice donor/acceptor resulted in a different band size depending on the splice site in question. PCR products were sequenced to confirm fragment identity. ImageJ was used to quantify and compare relative changes in normalized expression levels.

For rescue, full-length cDNAs for zebrafish smn and nrxn2ab were amplified with Phusion TAQ polymerase (Finnzymes) and cloned into TOPO blunt pCRII (Invitrogen), before subcloning into pCS2P+. Capped mRNAs were in vitro transcribed using the mMESSAGE mMACHINE Sp6 kit (Ambion) as described previously (67). As the SMN MO targets the ATG start codon, the forward primer was designed (Supplementary Material, Table S2) to introduce mutations in the third codon position (‘wobble bases’) and prevent binding of the SMN MO to the exogenous SMN mRNA. The cloned full-length SMN mRNA sequence contained the entire coding sequence of transcript ENSDART000000028099 with wobble base mutations in its 5′ region. Full-length nrxn2ab cDNA was cloned, according to transcript ENSDART00000132583 without modifications as the mature mRNA is not affected by the used nrxn2a splice MOs. Assessment of rescue was done by the quantification of percentage of embryos displaying CaP axon phenotypes in at least three-independent biological replicates.

### Mouse MN cell culture

Lumbar spinal MNs were isolated from E13.5 Smn−/−/SMN2 mouse embryos as previously described (68) with minor modifications. The spinal cord was dissected and collected in HBSS, then trypsinized (0.1% Trypsin in Ca2+–Mg2+-free buffer) for 15 min at 37°C, and MNs were enriched by panning using an antibody against the p75NTR receptor. Enriched MNs from each embryo were plated in a single well of 24-well dish (Nunc) precoated with polyornithine and laminin (Invitrogen). Neurons were grown in neurobasal medium with 2% B27 supplement (Invitrogen), containing 500 µM GlutaMAX (Invitrogen), 2% horse serum (Linaris) and the neurotrophic factor BDNF (5 ng/ml) at 37°C and 5% CO2. Culture medium was changed on Day 1 and then every second day. Genotyping was performed from corresponding embryos after dissection of spinal cords.

### Laser capture microdissection

Laser capture microdissection (Leica DM6000B laser microdissection system) was applied to isolate MN cell bodies from the spinal cord of 2-day-old postnatal mice. The lumbar spinal cord were embedded in optimum cutting temperature compound (Tissue-Tek) and immediately immersed in isopentane, chilled in liquid nitrogen for rapid freezing. About 100–150 cross-sections of 15 μm thickness were prepared from each spinal cord on a Leica cryostat, transferred to 0.9 µm POL membranes (Leica) and stained in Cresyl Violet solution. A total of 1500–2000 MN cell bodies were dissected from each spinal cord and collected in RNA lysis buffer. The total RNA was purified from the samples (69) and real-time RT–PCR was performed. To compare expression levels of Neurexin2α in wild-type versus SMN MNs, GAPDH expression served as denominator and relative expression was calculated.

### Real-time qRT–PCR

20–30 zebrafish embryos per sample at 48 hpf were stabilized in 100 μl of RNAlater RNA stabilization reagent (Qiagen) and kept at −80°C prior to RNA extraction using the RNeasy kit (Qiagen). All RNA samples were subjected to DNase I digestion to eliminate traces of genomic DNA. Samples were then subjected to a second round of RNA cleanup. RNA was reversed transcribed using the Revertaid first strand cDNA synthesis kit (Fermentas).

For the zebrafish transcripts, qRT–PCR reactions were performed in MicroAmp® 96-well plates (ABI) with a Applied Biosystems 7000 thermal cycler (ABI) using the ABI Prism 7000 SDS software (ABI) with the following thermal cycler conditions. Samples were heated at 50°C for 2 min, at 95°C for 10 min, followed by 40 cycles of 95°C at 15 s and 60°C at 1 min per cycle. Post-run dissociation melting curves were checked for the presence of single dissociation peaks to confirm specificity. Non-template control and minus RT (-RT) controls were included to exclude genomic DNA contamination. The housekeeping gene gapdh was used for normalization and at least three technical and three biological replicates were analyzed. Comparison of gene expression levels in SMN MO relative to Control MO were performed using the relative quantification mode in the Prism7000 software and filtered based on fold change >1.5 and P–value <0.05.

To characterize alternative splicing patterns of zebrafish nrxn2aa by semi-quantitative RT–PCR, primers were designed to bind to exon 11 and exon 14 at the SS3 splice site in nrxn2aa (Supplementary Material, Table S2). The four possible splice products contain exons 11,12,13,14 (344 bp), exons 11,12,14 (335 bp), exons 11,13,14 (317 bp) and exons 11,14 (308 bp). The exons 11,12,14 product (344 bp) and exons 11,14 product (308 bp) were cloned and sequenced. A 3% TAE/agarose gel was used for better resolution.

For mice, lumbar spinal cords were dissected from mouse embryos or 2-day-old postnatal mice and immediately frozen in liquid Nitrogen. QIAshredder (Qiagen) was used to homogenize the spinal cords. RNA was isolated using the RNeasy Plus-Mini-Kit (Qiagen) and reverse transcribed using the Invitrogen Superscript III reverse transcriptase kit according to the manufacturer's instruction. qPCRs were run on a Lightcycler 1.5 (Roche) using FastStart DNA master SYBR green1 reagents. Relative expression levels were calculated with offline analysis and control for efficiency. Intron-spanning primers were selected with the Oligo 6.0 software (MedProbe) and PCR conditions, primer concentration and MgCl2 concentration were optimized. PCR products were analyzed by gel electrophoresis, melting curve analysis. To characterize alternative splicing patterns, primers were designed to bind to exon 10 and exon 13 at the SS3 splice site in mouse Nrxn2α. The four possible splice products contain exons 10,11,12,13 (121 bp), exons 10, 11,13 (112 bp), exons 10,12,13 (94 bp) and exons 10,13 (85 bp). A 3% TAE/agarose gel was used for better resolution. Used primers are listed in Supplementary Material, Table 2.

The real-time qRT–PCR cycle conditions for the mouse transcripts were as follows: kinetic cycle conditions in four segments: Nrxn1α (95°C, 1 s, 55°C, 5 s, 72°C, 7 s, 79°C, 5 s), 2 mM MgCl2, 30 pmol primer. Nrxn2α (95°C, 1 s, 60°C, 5 s, 72°C, 9 s, 86°C, 5 s), 2 mM MgCl2, 30 pmol primer. Nrxn3α (95°C, 1 s, 60°C, 5 s, 72°C, 8 s, 84°C, 5 s), 2 mM MgCl2, 30 pmol primer. Nrxn2β (95°C, 1 s, 60°C, 5 s, 72°C, 7 s, 86°C, 5 s), 2 mM MgCl2, 40 pmol. Nrxn2α exon11 (qPCR) (95°C, 1 s, 62°C, 5 s, 72°C, 9 s, 81°C, 5 s), 3 mM MgCl2, 30 pmol primer. Nrxn2α exon11 (PCR) (95°C, 1 s, 60°C, 5 s, 72°C, 6 s, 85°C, 5 s), 3 mM MgCl2, 30 pmol primer. GAPDH (95°C, 0 s, 59°C, 5 s, 72°C, 6 s, 83°C, 5 s) 3 mM MgCl2, 30 pmol primer. HMBS (95°C, 1 s, 59°C, 5 s, 72°C, 6 s, 85°C, 5 s), 2 mM MgCl2, 30 pmol primer. HPRT1 (95°C, 1 s, 58°C, 5 s, 72°C, 6 s, 80°C, 5 s), 3 mM MgCl2, 30 pmol primer.

### Live Ca2+ imaging

Live Ca2+ imaging in HB9:D3cpv/MN embryos was performed as described (See and Winkler; HMG under revision). Briefly, 31 hpf live zebrafish embryos were manually dechorionated and anesthetized in 0.05% tricaine solution (Sigma Aldrich, USA) before embedding in 1.5% low melting point agarose (Bio-rad, USA) on a coverslip bottom petridish (Iwaki, Japan). For image acquisition, an UltraVIEW VOX live imaging system (Perkin Elmer, USA) consisting of a spinning disk confocal system on an inverted microscope IX 81 (Olympus, Japan) and a high sensitivity C9100-13 EM-CCD camera (Hamamatsu, Japan) with a 40×/1.15 water objective (Olympus, Japan) was used; 512 × 512 pixel images were typically acquired at ∼2.5 fps over a 30 s time interval with exposure time of 50 ms, gain setting of three and sensitivity setting of 190; ∼6 mW of laser excitation at 440 nm from a solid state diode laser with diode module was used and emission filters of 587/W125 and 485/W60 were used for collection of FRET and CFP emission respectively; 1 ml of 1 mM l-glutamic acid potassium salt monohydrate, K+ Glutamate, (G1501, Sigma, USA) was carefully added to elicit the evoked Ca2+ response without introducing movement artifacts or focus drift. The Volocity 5.3.1 (Mac) software was used for image acquisition and images exported were analyzed with ImageJ 1.44d (Mac) using the plugins MultiStackReg v1.45 and Ratio Plus as described (70). To eliminate possible imaging artifacts due to noise or registration issues, background correction was performed using rolling ball algorithm. Registration correction was performed using MultiStackReg to align the FRET and CFP images before the generation of FRET/CFP ratio images using Ratio Plus. A lookup table (16_colors) was applied to generate pseudo colored images to facilitate visualization of evoked changes. The area of the Regions of Interest (ROIs) used in the analysis of the pre-synaptic axon terminal was 4.241 μm2, which was kept constant for comparison across samples. The background corrected FRET/CFP ratio (R) was determined from the final FRET/CFP ratio images generated and the following formula was applied in the calculation of the relative change in R, (%ΔR/Ro) (71). %ΔR/Ro was calculated using the difference in magnitude of R from the average of the baseline period (Ro), which is defined as the interval before the addition of stimulant at 5 s.

$$\% \Delta \hbox{R}/\hbox{R}_{\rm o} = \displaystyle{{R - R_{\rm o} } \over {\hbox{R}_{\rm o} }} \times 100\% .$$

Unless otherwise stated, statistical analyses were performed using GraphPad Prism 5.0a (Mac). For comparisons between three or more groups, one-way ANOVA was performed followed by Tukey's Multiple Comparison post hoc test to check for statistically significant differences in mean between the different groups. For comparisons between two groups, unpaired Student's t-test was performed. All data are presented as means ± SEM, with P values displayed as appropriate (P > 0.05n.s., P < 0.05*, P < 0.01**, P < 0.0001***).

## FUNDING

This project was supported by MOE/AcRF (R-154-000-478-112) and MOH/NMRC (CBRG12nov097) grants to C.W., the Hermann und Lilly Schilling Stiftung im Stifterverband der Deutschen Industrie to M.S. and DFG Fi573-8/1 to U.F. K.S. is supported by an Agency for Science, Technology and Research (A*STAR) Graduate Scholarship (AGS), P.Y. by the Graduate School of Life Sciences, University of Wuerzburg, and M.G. and H.V. by National University of Singapore (NUS) Faculty of Science Graduate Scholarships.

## ACKNOWLEDGEMENTS

We thank the NUS-DBS confocal unit and Centre for BioImaging Sciences (CBIS) for support. We are grateful to Dr Thorsten Wohland, Dr Philip W. Ingham and Dr Soong Tuck Wah for discussions and Dr Robert Blum for helpful support in primer design.

Conflict of Interest statement: The authors declare that they have no conflict of interest.

## REFERENCES

1
Lefebvre
S.
Burglen
L.
Reboullet
S.
Clermont
O.
Burlet
P.
Viollet
L.
Benichou
B.
Cruaud
C.
Millasseau
P.
Zeviani
M.
, et al.  .
Identification and characterization of a spinal muscular atrophy-determining gene
Cell
,
1995
, vol.
80
(pg.
155
-
165
)
2
Werdnig
G.
Two early infantile hereditary cases of progressive muscular atrophy simulating dystrophy, but on a neural basis. 1891
Arch. Neurol.
,
1971
, vol.
25
(pg.
276
-
278
)
3
Monani
U.R.
Lorson
C.L.
Parsons
D.W.
Prior
T.W.
Androphy
E.J.
Burghes
A.H.
McPherson
J.D.
A single nucleotide difference that alters splicing patterns distinguishes the SMA gene SMN1 from the copy gene SMN2
Hum. Mol. Genet.
,
1999
, vol.
8
(pg.
1177
-
1183
)
4
Kashima
T.
Manley
J.L.
A negative element in SMN2 exon 7 inhibits splicing in spinal muscular atrophy
Nat. Genet.
,
2003
, vol.
34
(pg.
460
-
463
)
5
Kashima
T.
Rao
N.
David
C.J.
Manley
J.L.
hnRNP A1 functions with specificity in repression of SMN2 exon 7 splicing
Hum. Mol. Genet.
,
2007
, vol.
16
(pg.
3149
-
3159
)
6
Prior
T.W.
Swoboda
K.J.
Scott
H.D.
Hejmanowski
A.Q.
Homozygous SMN1 deletions in unaffected family members and modification of the phenotype by SMN2
Am. J. Med. Genet. A
,
2004
, vol.
130A
(pg.
307
-
310
)
7
Monani
U.R.
Sendtner
M.
Coovert
D.D.
Parsons
D.W.
Andreassi
C.
Le
T.T.
Jablonka
S.
Schrank
B.
Rossoll
W.
Prior
T.W.
, et al.  .
The human centromeric survival motor neuron gene (SMN2) rescues embryonic lethality in Smn(−/−) mice and results in a mouse with spinal muscular atrophy
Hum. Mol. Genet.
,
2000
, vol.
9
(pg.
333
-
339
)
8
Lefebvre
S.
Burlet
P.
Liu
Q.
Bertrandy
S.
Clermont
O.
Munnich
A.
Dreyfuss
G.
Melki
J.
Correlation between severity and SMN protein level in spinal muscular atrophy
Nat Genet.
,
1997
, vol.
16
(pg.
265
-
269
)
9
McAndrew
P.E.
Parsons
D.W.
Simard
L.R.
Rochette
C.
Ray
P.N.
Mendell
J.R.
Prior
T.W.
Burghes
A.H.
Identification of proximal spinal muscular atrophy carriers and patients by analysis of SMNT and SMNC gene copy number
Am. J. Hum. Genet.
,
1997
, vol.
60
(pg.
1411
-
1422
)
10
Coovert
D.D.
Le
T.T.
McAndrew
P.E.
Strasswimmer
J.
Crawford
T.O.
Mendell
J.R.
Coulson
S.E.
Androphy
E.J.
Prior
T.W.
Burghes
A.H.
The survival motor neuron protein in spinal muscular atrophy
Hum. Mol. Genet.
,
1997
, vol.
6
(pg.
1205
-
1214
)
11
Burghes
A.H.
Beattie
C.E.
Spinal muscular atrophy: why do low levels of survival motor neuron protein make motor neurons sick? Nat
Rev. Neurosci.
,
2009
, vol.
10
(pg.
597
-
609
)
12
Pellizzoni
L.
Yong
J.
Dreyfuss
G.
Essential role for the SMN complex in the specificity of snRNP assembly
Science
,
2002
, vol.
298
(pg.
1775
-
1779
)
13
Meister
G.
Buhler
D.
Pillai
R.
Lottspeich
F.
Fischer
U.
A multiprotein complex mediates the ATP-dependent assembly of spliceosomal U snRNPs
Nat. Cell Biol.
,
2001
, vol.
3
(pg.
945
-
949
)
14
Fischer
U.
Liu
Q.
Dreyfuss
G.
The SMN-SIP1 complex has an essential role in spliceosomal snRNP biogenesis
Cell
,
1997
, vol.
90
(pg.
1023
-
1029
)
15
Liu
Q.
Fischer
U.
Wang
F.
Dreyfuss
G.
The spinal muscular atrophy disease gene product, SMN, and its associated protein SIP1 are in a complex with spliceosomal snRNP proteins
Cell
,
1997
, vol.
90
(pg.
1013
-
1021
)
16
Pellizzoni
L.
Kataoka
N.
Charroux
B.
Dreyfuss
G.
A novel function for SMN, the spinal muscular atrophy disease gene product, in pre-mRNA splicing
Cell
,
1998
, vol.
95
(pg.
615
-
624
)
17
Brow
D.A.
Annu. Rev. Genet.
,
2002
, vol.
36
(pg.
333
-
360
)
18
Zhang
Z.
Lotti
F.
Dittmar
K.
Younis
I.
Wan
L.
Kasim
M.
Dreyfuss
G.
SMN deficiency causes tissue-specific perturbations in the repertoire of snRNAs and widespread defects in splicing
Cell
,
2008
, vol.
133
(pg.
585
-
600
)
19
Winkler
C.
Eggert
C.
D.
Meister
G.
Giegerich
M.
Wedlich
D.
Laggerbauer
B.
Fischer
U.
Reduced U snRNP assembly causes motor axon degeneration in an animal model for spinal muscular atrophy
Genes Dev.
,
2005
, vol.
19
(pg.
2320
-
2330
)
20
Gabanella
F.
Butchbach
M.E.
Saieva
L.
Carissimi
C.
Burghes
A.H.
Pellizzoni
L.
Ribonucleoprotein assembly defects correlate with spinal muscular atrophy severity and preferentially affect a subset of spliceosomal snRNPs
PLoS One
,
2007
, vol.
2
pg.
e921

21
Lotti
F.
Imlach
W.L.
Saieva
L.
Beck
E.S.
Hao le
T.
Li
D.K.
Jiao
W.
Mentis
G.Z.
Beattie
C.E.
McCabe
B.D.
, et al.  .
An SMN-dependent U12 splicing event essential for motor circuit function
Cell
,
2012
, vol.
151
(pg.
440
-
454
)
22
Boulisfane
N.
Choleza
M.
Rage
F.
Neel
H.
Soret
J.
Bordonne
R.
Impaired minor tri-snRNP assembly generates differential splicing defects of U12-type introns in lymphoblasts derived from a type I SMA patient
Hum. Mol. Genet.
,
2011
, vol.
20
(pg.
641
-
648
)
23
Imlach
W.L.
Beck
E.S.
Choi
B.J.
Lotti
F.
Pellizzoni
L.
McCabe
B.D.
SMN Is Required for Sensory-Motor Circuit Function in Drosophila
Cell
,
2012
, vol.
151
(pg.
427
-
439
)
24
Fan
L.
Simard
L.R.
Survival motor neuron (SMN) protein: role in neurite outgrowth and neuromuscular maturation during neuronal differentiation and development
Hum. Mol. Genet.
,
2002
, vol.
11
(pg.
1605
-
1614
)
25
Rossoll
W.
Kroning
A.K.
Ohndorf
U.M.
Steegborn
C.
Jablonka
S.
Sendtner
M.
Specific interaction of Smn, the spinal muscular atrophy determining gene product, with hnRNP-R and gry-rbp/hnRNP-Q: a role for Smn in RNA processing in motor axons?
Hum. Mol. Genet.
,
2002
, vol.
11
(pg.
93
-
105
)
26
Jablonka
S.
Bandilla
M.
Wiese
S.
Buhler
D.
Wirth
B.
Sendtner
M.
Fischer
U.
Co-regulation of survival of motor neuron (SMN) protein and its interactor SIP1 during development and in spinal muscular atrophy
Hum. Mol. Genet.
,
2001
, vol.
10
(pg.
497
-
505
)
27
Rossoll
W.
Jablonka
S.
Andreassi
C.
Kroning
A.K.
Karle
K.
Monani
U.R.
Sendtner
M.
Smn, the spinal muscular atrophy-determining gene product, modulates axon growth and localization of beta-actin mRNA in growth cones of motoneurons
J. Cell Biol.
,
2003
, vol.
163
(pg.
801
-
812
)
28
Kong
L.
Wang
X.
Choe
D.W.
Polley
M.
Burnett
B.G.
Bosch-Marce
M.
Griffin
J.W.
Rich
M.M.
Sumner
C.J.
Impaired synaptic vesicle release and immaturity of neuromuscular junctions in spinal muscular atrophy mice
J. Neurosci.
,
2009
, vol.
29
(pg.
842
-
851
)
29
Kariya
S.
Park
G.H.
Maeno-Hikichi
Y.
Leykekhman
O.
Lutz
C.
Arkovitz
M.S.
Landmesser
L.T.
Monani
U.R.
Reduced SMN protein impairs maturation of the neuromuscular junctions in mouse models of spinal muscular atrophy
Hum. Mol. Genet.
,
2008
, vol.
17
(pg.
2552
-
2569
)
30
McGovern
V.L.
Gavrilina
T.O.
Beattie
C.E.
Burghes
A.H.
Embryonic motor axon development in the severe SMA mouse
Hum. Mol. Genet.
,
2008
, vol.
17
(pg.
2900
-
2909
)
31
Ruiz
R.
Casanas
J.J.
Torres-Benito
L.
Cano
R.
Tabares
L.
Altered intracellular Ca2+ homeostasis in nerve terminals of severe spinal muscular atrophy mice
J. Neurosci.
,
2010
, vol.
30
(pg.
849
-
857
)
32
Jablonka
S.
Beck
M.
Lechner
B.D.
Mayer
C.
Sendtner
M.
Defective Ca2+ channel clustering in axon terminals disturbs excitability in motoneurons in spinal muscular atrophy
J. Cell Biol.
,
2007
, vol.
179
(pg.
139
-
149
)
33
Gubitz
A.K.
Feng
W.
Dreyfuss
G.
The SMN complex
Exp. Cell Res.
,
2004
, vol.
296
(pg.
51
-
56
)
34
Campion
Y.
Neel
H.
Gostan
T.
Soret
J.
Bordonne
R.
Specific splicing defects in S. pombe carrying a degron allele of the Survival of Motor Neuron gene
EMBO J
,
2010
, vol.
29
(pg.
1817
-
1829
)
35
Shpargel
K.B.
Matera
A.G.
Gemin proteins are required for efficient assembly of Sm-class ribonucleoproteins
,
2005
, vol.
102
(pg.
17372
-
17377
)
36
Wittkopp
N.
Huntzinger
E.
Weiler
C.
Sauliere
J.
Schmidt
S.
Sonawane
M.
Izaurralde
E.
Nonsense-mediated mRNA decay effectors are essential for zebrafish embryonic development and survival
Mol. Cell Biol.
,
2009
, vol.
29
(pg.
3517
-
3528
)
37
Behm-Ansmant
I.
Kashima
I.
Rehwinkel
J.
Sauliere
J.
Wittkopp
N.
Izaurralde
E.
mRNA quality control: an ancient machinery recognizes and degrades mRNAs with nonsense codons
FEBS Lett
,
2007
, vol.
581
(pg.
2845
-
2853
)
38
Petrenko
A.G.
Perin
M.S.
Davletov
B.A.
Ushkaryov
Y.A.
Geppert
M.
Sudhof
T.C.
Binding of synaptotagmin to the alpha-latrotoxin receptor implicates both in synaptic vesicle exocytosis
Nature
,
1991
, vol.
353
(pg.
65
-
68
)
39
Nguyen
T.
Sudhof
T.C.
Binding properties of neuroligin 1 and neurexin 1beta reveal function as heterophilic cell adhesion molecules
J. Biol. Chem.
,
1997
, vol.
272
(pg.
26032
-
26039
)
40
Sugita
S.
Saito
F.
Tang
J.
Satz
J.
Campbell
K.
Sudhof
T.C.
A stoichiometric complex of neurexins and dystroglycan in brain
J. Cell Biol.
,
2001
, vol.
154
(pg.
435
-
445
)
41
Ichtchenko
K.
Hata
Y.
Nguyen
T.
Ullrich
B.
Missler
M.
Moomaw
C.
Sudhof
T.C.
Neuroligin 1: a splice site-specific ligand for beta-neurexins
Cell
,
1995
, vol.
81
(pg.
435
-
443
)
42
Ushkaryov
Y.A.
Petrenko
A.G.
Geppert
M.
Sudhof
T.C.
Neurexins: synaptic cell surface proteins related to the alpha-latrotoxin receptor and laminin
Science
,
1992
, vol.
257
(pg.
50
-
56
)
43
Missler
M.
Zhang
W.
Rohlmann
A.
Kattenstroth
G.
Hammer
R.E.
Gottmann
K.
Sudhof
T.C.
Alpha-neurexins couple Ca2+ channels to synaptic vesicle exocytosis
Nature
,
2003
, vol.
423
(pg.
939
-
948
)
44
Zhang
W.
Rohlmann
A.
Sargsyan
V.
Aramuni
G.
Hammer
R.E.
Sudhof
T.C.
Missler
M.
Extracellular domains of alpha-neurexins participate in regulating synaptic transmission by selectively affecting N- and P/Q-type Ca2+ channels
J. Neurosci.
,
2005
, vol.
25
(pg.
4330
-
4342
)
45
Sons
M.S.
Busche
N.
Strenzke
N.
Moser
T.
Ernsberger
U.
Mooren
F.C.
Zhang
W.
M.
Steffens
H.
Schomburg
E.D.
, et al.  .
alpha-Neurexins are required for efficient transmitter release and synaptic homeostasis at the mouse neuromuscular junction
Neuroscience
,
2006
, vol.
138
(pg.
433
-
446
)
46
Rissone
A.
Monopoli
M.
Beltrame
M.
Bussolino
F.
Cotelli
F.
Arese
M.
Comparative genome analysis of the neurexin gene family in Danio rerio: insights into their functions and evolution
Mol. Biol. Evol.
,
2007
, vol.
24
(pg.
236
-
252
)
47
Fox
M.A.
Sanes
J.R.
Synaptotagmin I and II are present in distinct subsets of central synapses
J. Comp. Neurol.
,
2007
, vol.
503
(pg.
280
-
296
)
48
Missler
M.
Sudhof
T.C.
Neurexins: three genes and 1001 products
Trends Genet.
,
1998
, vol.
14
(pg.
20
-
26
)
49
Ushkaryov
Y.A.
Hata
Y.
Ichtchenko
K.
Moomaw
C.
Afendis
S.
Slaughter
C.A.
Sudhof
T.C.
Conserved domain structure of beta-neurexins. Unusual cleaved signal sequences in receptor-like neuronal cell-surface proteins
J. Biol. Chem.
,
1994
, vol.
269
(pg.
11987
-
11992
)
50
Ullrich
B.
Ushkaryov
Y.A.
Sudhof
T.C.
Cartography of neurexins: more than 1000 isoforms generated by alternative splicing and expressed in distinct subsets of neurons
Neuron
,
1995
, vol.
14
(pg.
497
-
507
)
51
Rozic-Kotliroff
G.
Zisapel
N.
Ca2+ -dependent splicing of neurexin IIalpha
Biochem. Biophys. Res. Commun.
,
2007
, vol.
352
(pg.
226
-
230
)
52
Li
Q.
Lee
J.A.
Black
D.L.
Neuronal regulation of alternative pre-mRNA splicing
Nat. Rev. Neurosci.
,
2007
, vol.
8
(pg.
819
-
831
)
53
Rissone
A.
Foglia
E.
Sangiorgio
L.
Cermenati
S.
Nicoli
S.
Cimbro
S.
Beltrame
M.
Bussolino
F.
Cotelli
F.
Arese
M.
The synaptic proteins beta-neurexin and neuroligin synergize with extracellular matrix-binding vascular endothelial growth factor a during zebrafish vascular development
Arterioscler Thromb Vasc. Biol.
,
2012
, vol.
32
(pg.
1563
-
1572
)
54
Myers
P.Z.
Eisen
J.S.
Westerfield
M.
Development and axonal outgrowth of identified motoneurons in the zebrafish
J. Neurosci.
,
1986
, vol.
6
(pg.
2278
-
2289
)
55
Eisen
J.S.
Smith
J.C.
Controlling morpholino experiments: don't stop making antisense
Development
,
2008
, vol.
135
(pg.
1735
-
1743
)
56
McWhorter
M.L.
Monani
U.R.
Burghes
A.H.
Beattie
C.E.
Knockdown of the survival motor neuron (Smn) protein in zebrafish causes defects in motor axon outgrowth and pathfinding
J. Cell Biol.
,
2003
, vol.
162
(pg.
919
-
931
)
57
Eisen
J.S.
Myers
P.Z.
Westerfield
M.
Pathway selection by growth cones of identified motoneurones in live zebra fish embryos
Nature
,
1986
, vol.
320
(pg.
269
-
271
)
58
Iijima
T.
Wu
K.
Witte
H.
Hanno-Iijima
Y.
Glatter
T.
Richard
S.
Scheiffele
P.
SAM68 regulates neuronal activity-dependent alternative splicing of neurexin-1
Cell
,
2011
, vol.
147
(pg.
1601
-
1614
)
59
Ehrmann
I.
Dalgliesh
C.
Liu
Y.
Danilenko
M.
Crosier
M.
Overman
L.
Arthur
H.M.
Lindsay
S.
Clowry
G.J.
Venables
J.P.
, et al.  .
The tissue-specific RNA binding protein T-STAR controls regional splicing patterns of neurexin pre-mRNAs in the brain
PLoS Genet.
,
2013
, vol.
9
pg.
e1003474

60
Resnick
M.
Segall
A.
Gabriela Rozic-Kotliroff
G.R.
Lupowitz
Z.
Zisapel
N.
Alternative splicing of neurexins: a role for neuronal polypyrimidine tract binding protein
Neurosci. Lett
,
2008
, vol.
439
(pg.
235
-
240
)
61
Chari
A.
Golas
M.M.
Klingenhager
M.
Neuenkirchen
N.
Sander
B.
Englbrecht
C.
Sickmann
A.
Stark
H.
Fischer
U.
An assembly chaperone collaborates with the SMN complex to generate spliceosomal SnRNPs
Cell
,
2008
, vol.
135
(pg.
497
-
509
)
62
Graf
E.R.
Zhang
X.
Jin
S.X.
Linhoff
M.W.
Craig
A.M.
Neurexins induce differentiation of GABA and glutamate postsynaptic specializations via neuroligins
Cell
,
2004
, vol.
119
(pg.
1013
-
1026
)
63
Li
J.
Ashley
J.
Budnik
V.
Bhat
M.A.
Crucial role of Drosophila neurexin in proper active zone apposition to postsynaptic densities, synaptic growth, and synaptic transmission
Neuron
,
2007
, vol.
55
(pg.
741
-
755
)
64
Michaud
M.
Arnoux
T.
Bielli
S.
Durand
E.
Rotrou
Y.
Jablonka
S.
Robert
F.
Giraudon-Paoli
M.
Riessland
M.
Mattei
M.G.
, et al.  .
Neuromuscular defects and breathing disorders in a new mouse model of spinal muscular atrophy
Neurobiol. Dis.
,
2010
, vol.
38
(pg.
125
-
135
)
65
Mathavan
S.
Lee
S.G.
Mak
A.
Miller
L.D.
Murthy
K.R.
Govindarajan
K.R.
Tong
Y.
Wu
Y.L.
Lam
S.H.
Yang
H.
, et al.  .
Transcriptome analysis of zebrafish embryogenesis using microarrays
PLoS Genet.
,
2005
, vol.
1
(pg.
260
-
276
)
66
Welten
M.C.
de Haan
S.B.
van den Boogert
N.
Noordermeer
J.N.
Lamers
G.E.
Spaink
H.P.
Meijer
A.H.
Verbeek
F.J.
ZebraFISH: fluorescent in situ hybridization protocol and three-dimensional imaging of gene expression patterns
Zebrafish
,
2006
, vol.
3
(pg.
465
-
476
)
67
Schafer
M.
Rembold
M.
Wittbrodt
J.
Schartl
M.
Winkler
C.
Medial floor plate formation in zebrafish consists of two phases and requires trunk-derived Midkine-a
Genes Dev.
,
2005
, vol.
19
(pg.
897
-
902
)
68
Wiese
S.
Herrmann
T.
Drepper
C.
Jablonka
S.
Funk
N.
Klausmeyer
A.
Rogers
M.L.
Rush
R.
Sendtner
M.
Isolation and enrichment of embryonic mouse motoneurons from the lumbar spinal cord of individual mouse embryos
Nat. Protoc.
,
2010
, vol.
5
(pg.
31
-
38
)
69
Grundemann
J.
Schlaudraff
F.
Haeckel
O.
Liss
B.
Elevated alpha-synuclein mRNA levels in individual UV-laser-microdissected dopaminergic substantia nigra neurons in idiopathic Parkinson's disease
Nucleic Acids Res.
,
2008
, vol.
36
pg.
e38

70
Kardash
E.
Bandemer
J.
Raz
E.
Imaging protein activity in live embryos using fluorescence resonance energy transfer biosensors
Nat. Protoc.
,
2011
, vol.
6
(pg.
1835
-
1846
)
71
Hendel
T.
Mank
M.
Schnell
B.
Griesbeck
O.
Borst
A.
Reiff
D.F.
Fluorescence changes of genetic calcium indicators and OGB-1 correlated with neural activity and calcium in vivo and in vitro
J. Neurosci.
,
2008
, vol.
28
(pg.
7399
-
7411
)

## Author notes

Present address: Genome Institute of Singapore, 138672 Singapore.