Knockdown of the survival motor neuron (Smn) protein in zebrafish causes defects in motor axon outgrowth and pathfinding

Zebrafish is an excellent organism for studying human disease (Dodd et al., 2000) due to its well-characterized embryonic development and the ability to perform both forward and reverse genetics (for review see Grunwald and Eisen, 2002). Protein knockdown technology has facilitated analysis of zebrafish forms of muscular dystrophy (Parsons et al., 2002) and cystic kidney disease (Liu et al., 2002). Due to its well-characterized nervous system and relatively simple neuromuscular organization, zebrafish are well suited for analysis of neuromuscular diseases. Taking advantage of these attributes, we have used protein knockdown in zebrafish to model spinal muscular atrophy (SMA).

SMA, an autosomal recessive disorder, is the leading hereditary cause of infant mortality (Roberts et al., 1970) and is characterized by loss of α motoneurons in the spinal cord (Crawford and Pardo, 1996; Melki, 1997). The disease results from low levels of the protein encoded by the survival motor neuron (SMN) gene. Although SMN is expressed in all cell types, motoneurons are specifically affected in SMA, indicating their sensitivity to low SMN levels (Coovert et al., 1997; Lefebvre et al., 1997; Monani et al., 2000). Humans have two copies of the SMN gene, SMN1 and SMN2, which differ in a single base change in a splice enhancer site for exon 7 of SMN2 (Lorson et al., 1999; Monani et al., 1999; Cartegni and Krainer, 2002). Thus, SMN1 produces a majority of full-length transcript, whereas SMN2 generates mostly transcripts lacking exon 7, although some full-length transcript is produced (Lefebvre et al., 1995). SMN protein lacking exon 7 does not oligomerize effectively (Lorson and Androphy, 1998) and appears to be unstable and rapidly degraded (Lorson and Androphy, 2000). Thus, mutations in SMN1, but retention of the SMN2 gene, results in reduced protein levels and ultimately SMA (Lefebvre et al., 1995, 1997; Coovert et al., 1997).

The 38-kD SMN protein is ubiquitously expressed and localizes to both the cytoplasm and nucleus (Liu and Dreyfuss, 1996; Coovert et al., 1997; Lefebvre et al., 1997). In the nucleus, SMN localizes to structures termed gems, which overlap or are in close proximity to coiled bodies (Liu and Dreyfuss, 1996; Young et al., 2000a). It has been termed the master RNA assembler and, in particular, has been shown to be important in assembly of snRNP particles (for review see Terns and Terns, 2001). SMN also binds to the hn-RNP-R, which is involved in RNA editing and mRNA transport (Rossoll et al., 2002). Recent data shows that hn-RNP-R colocalizes with SMN in distal axons of embryonic motoneurons (Jablonka et al., 2001; Rossoll et al., 2002). SMN also has been shown to localize in the growth cones and branch points of developing neurons (Jablonka et al., 2001; Fan and Simard, 2002; Zhang et al., 2003). Ultimately, however, the function of SMN in relation to SMA pathology and etiology remains unclear.

To further analyze SMN function, animal models of SMA have been generated (Schrank et al., 1997; Hsieh-Li et al., 2000; Monani et al., 2000; Cifuentes-Diaz et al., 2002; Monani et al., 2003). In contrast to humans, mice have only one Smn gene, which is equivalent to human SMN1 (DiDonato et al., 1997; Viollet et al., 1997). Complete loss of this gene results in an embryonic lethal phenotype (Schrank et al., 1997). Introduction of one or two copies of human SMN2 rescues the embryonic lethal phenotype and results in mice with severe SMA (Hsieh-Li et al., 2000; Monani et al., 2000), whereas 8–16 copies of SMN2 completely rescue the SMA phenotype (Monani et al., 2000). Although both severe and mild SMA mice ultimately exhibit motoneuron cell body reduction (Monani et al., 2000, 2003), no early morphological or biochemical abnormality of the motoneurons has been reported.

A model of SMA in zebrafish has the potential to elucidate the effect of decreased Smn levels on motoneuron development in vivo. At 24 h, there are three well-characterized primary motoneurons per spinal cord hemisegment that innervate either the dorsal, rostral, or ventral region of each myotome (Eisen et al., 1986; for review see Beattie, 2000; Fig. S1). Over the next few days, each of the primary motor axons are joined by 20–30 secondary motor axons, which form three distinct nerves that innervate the three myotome regions (Myers et al., 1986; Pike et al., 1992). As these axons extend into defined myotome regions, they can be followed in living embryos; thus, perturbations in the organization of these neurons or their axons can be readily detected, followed during development, and quantitated (for review see Beattie, 2000).

We have used antisense morpholino technology to model the effects of low levels of smn in zebrafish. Reducing Smn protein levels in the developing embryo results in motor axon–specific truncations and branches, independent of motoneuron cell death. Moreover, by decreasing Smn levels in single motoneurons, we show that these defects are due to a cell-autonomous function of Smn in motoneurons. These are the first reported morphological abnormalities of motoneuron development in response to low levels of Smn. These data reveal that one of the earliest consequences of Smn protein reduction is severely compromised motor axon outgrowth, indicative of an essential role for Smn in motoneuron development.

The zebrafish smn gene was mapped to linkage group 5 on the LN 54 radiation hybrid panel (Hukriede et al., 1999). Southern blot and radiation hybrid mapping (unpublished data) indicated that there is only a single smn gene in zebrafish. RNA in situ hybridization showed that, like its mammalian counterpart, zebrafish smn appears to be expressed in all cell types based on diffuse and ubiquitous staining at 22, 27, 36, and 48 h after fertilization (h; Fig. S2).

Morpholino antisense oligonucleotide “knockdown” technology was performed in zebrafish to decrease the levels of the Smn protein and mimic SMA. Morpholino oligonucleotides (MOs) inhibit the translation of their target mRNA, thereby causing reduced levels of target protein (Nasevicius and Ekker, 2000). Control or smn MO was injected into embryos at the one- to four-cell stage, and then embryos were allowed to develop until the desired time. Western blot analysis showed that smn MO–injected embryos exhibited a 61% decrease in Smn protein at 36 h (Fig. 1, A and B). Approximately 78% (n = 580) of embryos injected with 6 ng of smn MO survived compared with 91% (n = 139) survival when injected with the same concentration of control MO. Embryos died between late gastrulation and early somitogenesis, suggesting essential Smn function during zebrafish development. smn MO–injected embryos that survived did not exhibit gross abnormalities.

Initial examination focused on a subset of primary motoneurons. Cell bodies of caudal primary (CaP) and variable primary (VaP) motoneurons are located in the middle of each spinal cord hemisegment and express the LIM gene islet2 (Appel et al., 1995). Using islet2 RNA in situ hybridization, we found that these motoneurons were present in both control MO (n = 6) and smn MO–injected embryos (n = 42; unpublished data). Using the znp1 mAb, we examined CaP motor axons. At 27 h, CaP axon projections have a stereotyped morphology and project into ventral muscle, as shown in a control MO–injected embryo (Fig. 1 C). smn MO–injected embryos, however, showed motor axon defects that could be grouped into two categories based on axon length and axon branching. Approximately 35.3% of smn MO–injected embryo sides had at least one truncated motor axon that did not fully project into the ventral muscle (Table I; Fig. 1 D). Affected sides in smn MO–injected embryos had an average of 1.90 truncated axons (Table II). Axon branching was also observed in smn MO–injected embryos. At 27 h, 14.3% of embryo sides had at least one branched axon (Table I; Fig. 1 E), and each affected side had an average of 2.31 branches (Table II). These defects were typically found between hemisegments 5 and 17, corresponding to middle through posterior trunk. It was rare that axon defects were seen along the proximal portion of the axon, between the ventral root and the first intermediate target at the nascent horizontal myoseptum (Fig. 1, white arrowhead).

By 36 h, secondary motor axons also extend into the ventral muscle and fasciculate to form a nerve (Fig. 1 F). At 36 h in smn MO–injected embryos, 61.5% of the sides had at least one truncated axon (Table I; Fig. 1 G), compared with 35.3% at 27 h. Affected sides had an average of 2.81 truncated nerves, which is consistent with what was observed at 27 h (Table II). However, at 36 h, there were more sides with increased numbers of truncated nerves. 44.6% of the sides in smn MO–injected embryos had at least one branched nerve (Table I; Fig. 1 H) in contrast to only 23.6% at 27 h. The average number of branches per side remained constant at 2.5 (Table II).

It has been shown in human patients and mouse models that the severity of SMA is dependent on the amount of SMN protein (Monani et al., 2000). To investigate if motor axon defects would become more severe when Smn protein was further reduced, 9 ng of MO was injected into embryos. Western blot analysis showed a 77% knockdown of protein (Fig. 2, A and B). In contrast to 89% survival when control MO was injected (n = 164), at this higher dose of smn MO, only 45% (n = 433) of the embryos survived, further suggesting the essentiality of Smn in zebrafish development.

Injecting higher doses of smn MO resulted in a dramatic increase in motor axon branching. At 27 h, 77.3% of sides from smn MO embryos had at least one branched motor axon (Table I; Fig. 2 E). The average number of motor axon branches per side was also increased at higher doses of MO (Table II). Branching defects were also significantly increased at 36 h; 76.6% of injected embryo sides had at least one branched motor nerve, compared with 44.6% of sides with the lower dose of smn MO (Table I; Fig. 2 H). These data indicate that the motor axon branching defect is more severe upon further knockdown of Smn.

Motor axon truncations were less evident at higher smn MO doses; in fact at 27 h, there is no significant difference when compared with control MO injections, where 5.6% of sides had truncated motor axons (Table I, Fig. 2 D). Also, all of the smn MO defective sides only had one truncated motor axon (Table II). At 36 h, a similar trend was apparent. 46.9% of 9 ng smn MO–injected embryo sides had at least one truncated motor nerve, which is slightly reduced compared with lower doses of MO (Table I; Fig. 2 G). The average number of truncated motor nerves per side is 2.47, slightly less than at lower doses of MO (Table II). These data suggest that further reduction of Smn protein levels results in more severe motor nerve branching defects and less severe truncation defects. To confirm that reduction in Smn results in motor axon/nerve defects, we used an additional, nonoverlapping smn MO and observed similar defects (Fig. S3), although this 5′ UTR MO was less efficient than the original MO described (Table SI).

To determine the dynamics of the axon defects observed in smn MO–injected embryos, we analyzed motor nerves over several days using gata2–GFP transgenic zebrafish (Meng et al., 1997). In this transgenic line, secondary motoneuron cell bodies and ventrally projecting axons express GFP starting at ∼33 h. 9 ng of smn MO or control MO was injected into gata2–GFP embryos, and ventral motor nerves were visualized from 36 to 74 h. smn MO was still effective in knocking down Smn protein at later time points, as evidenced by Western blot analysis performed at 74 h, which shows a 88% reduction in Smn protein (Fig. 3, A and B).

13 individual motor nerves were identified in smn MO–injected embryos (n = 13) at 36 h and were followed until 74 h. At 36 h, each of the 13 nerves was truncated at or near the ventral edge of the notochord (Fig. 3 D), a second intermediate target for motor axons (Beattie et al., 2000; Fig. S1). When the same motor nerves were examined at 52 h, 92% (12/13) of the nerves remained truncated but had begun to branch (Fig. 3 F). Upon examination at 76 h, 85% (11/13) of the nerves were truncated and branched (Fig. 3 H). The remaining motor nerves (2/13) that did reach the ventral muscle projected into the preceding hemisegment, nearly touching the motor nerve in the adjacent hemisegment; although these motor nerves were not truncated or branched, their pathfinding was clearly erroneous. In contrast, all of the nerves in control MO–injected embryos (n = 10) exhibited the stereotypic ventral nerve projections (Fig. 3, C, E, and G). These data show that defective motor nerves are dynamic, first becoming truncated and then branching at the site of truncation. Thus, nerve truncations are still occurring at higher doses of smn MO, but the extensive branching obscures this phenotype.

To determine whether other spinal motor axons were affected by decreased levels of Smn protein, we analyzed dorsal motor nerves in islet1–GFP transgenic zebrafish. Among other neuronal types, this transgenic line has GFP-expressing secondary motoneuron cell bodies and dorsally projecting axons (Fig. 4 A; Higashijima et al., 2000). 9 ng of smn MO or control MO was injected into islet1–GFP transgenic embryos, and dorsal nerves were analyzed at ∼74 h. Like ventral motor nerves, dorsally projecting motor nerves exhibited truncation and branching defects as a result of smn knockdown (Fig. 4, B and C). Branching is also more prominent at higher doses of smn MO in these nerves. 69.1% (n = 178) of sides of smn MO–injected embryos had at least one branched dorsal projecting motor nerve, whereas control MO–injected embryos had only 8.0% of sides with a branched dorsal nerve (n = 50). In contrast, 32.6% (n = 178) of sides of smn MO–injected embryos had at least one truncated nerve, whereas control MO–injected embryos had no truncated dorsal projecting nerves (n = 50). These defects are consistent with those seen in ventral projecting motor nerves, suggesting that dorsal projecting motoneurons are also affected by reduction of Smn protein.

To confirm the specificity of the MO function, RNA rescue experiments were performed using human SMN mRNA. One- to four-cell gata2–GFP embryos were injected with a mixture of smn MO (9 ng) and hSMN mRNA. When compared with smn MO alone, coinjection of hSMN and smn MO resulted in rescue of both the branching and truncation nerve defects (Table III). The partial rescue observed was probably due to the mosaic nature of the mRNA and MO injections, as reported by others (McClintock et al., 2002). The specificity of the RNA rescue was confirmed by the lack of rescue when hSMNΔ7 mRNA, a common mutation and the major form of mRNA produced by SMN2, was coinjected with smn MO (Table III; Lefebvre et al., 1995). The percentage of sides with branched motor nerves is slightly reduced compared with MO alone, likely due to the slight activity of the hSMNΔ7 protein (Cifuentes-Diaz et al., 2002; unpublished data). These rescue experiments confirm that the motor nerve defects observed when smn MO is injected are a result of the reduction of Smn protein and further validate this system as a model for SMA.

To determine whether the motor axon defects observed upon smn knockdown were caused by neuronal apoptosis, a TUNEL assay was performed. Unlike higher vertebrates, naturally occurring motoneuron cell death in the zebrafish ventral spinal cord has not been reported (Cole and Ross, 2001). Embryos were injected with 9 ng control MO or smn MO and then analyzed at 27 (n = 10), 36 (n = 10), and 50 h (n = 15). None of the injected embryos had any TUNEL-positive cells in the ventral spinal cord (Fig. 5, A and B). TUNEL-positive cells could occasionally be seen in the dorsal spinal cord, corresponding to the Rohon-Beard sensory neurons (Fig. 5 B; arrowhead), which have previously been shown to undergo apoptosis at this age (Cole and Ross, 2001). To further confirm that motoneuron numbers were not reduced by means other than apoptosis, the diameter of the spinal motor pool was measured. Embryos injected with control or smn MO were labeled with anti-neurolin mAb (zn-5), which labels the cell surface of secondary motoneurons in the spinal cord (Fashena and Westerfield, 1999). To analyze the motor pool size, the diameter was measured both dorso-ventrally and medio-laterally on both sides of the midline (Fig. 5, C and D). There was no significant difference in motor pool diameter on either axis between control MO– and smn MO–injected embryos (Fig. 5 E). These data indicate that the motor axon defects observed upon Smn reduction are not caused by motoneuron death and further support the idea that Smn is needed for motor axon development.

In human and mouse models of SMA, the affect of decreased levels of SMN is specific for motoneurons. To determine if the defects resulting from Smn knockdown were also motoneuron specific, we examined populations of interneurons and sensory neurons using the acetylated tubulin mAb. Many zebrafish hindbrain interneurons have stereotyped axonal projections that extend into the spinal cord (Mendelson, 1986; Metcalfe et al., 1986). We saw no defects in the medial longitudinal fascicle or the lateral longitudinal fascicle, both of which contain hindbrain and midbrain axons descending into the spinal cord (Fig. 6, A and B). To look for more subtle defects, we examined the single commissural axon of the Mauthner neuron present in hindbrain rhombomere 4. Using the 3A10 mAb, we found that in smn MO–injected embryos, the Mauthner axon projected across the midline and down the spinal cord in a manner indistinguishable from control MO–injected embryos (Fig. 6, C and D). Rohon-Beard sensory neurons, present in the dorsal spinal cord (Lamborghini, 1980; Metcalfe et al., 1990), send axons into the peripheral muscle and skin. These cells and their axons were also unaffected by Smn knockdown (Fig. 6, E and F). Lastly, we looked at the lateral line, a group of sensory neurons that send axons the length of the developing embryo. We observed no difference in length or morphology of the lateral line axons in control MO– or smn MO–injected embryos (Fig. 6, H and G). Thus, interneurons and sensory neurons are unaffected by decreased levels of Smn protein and support the finding that motoneurons are uniquely sensitive to decreased Smn levels.

To determine whether the motor axon defects observed were due to defects in muscle development, we examined the development and patterning of the myotomes. Overall myotome morphology was analyzed using live gata2–GFP transgenic embryos injected with either control MO or smn MO. Nomarski images were taken of muscle segments containing defective GFP-expressing nerves. No difference was seen in the muscle morphology between control MO– and smn MO–injected embryos (Fig. 7, A and B). No defects were seen in early muscle patterning in smn MO–injected embryos compared with control MO, as assayed by myoD expression, a gene expressed in the posterior region of the developing somite (Weinberg et al., 1996; Fig. 7, C and D). Lastly, we asked whether muscle cells were correctly specified by characterizing slow and fast muscle in smn MO–injected embryos. Using the slow muscle mAb F59 (Crow and Stockdale, 1986; Devoto et al., 1996) and the fast muscle mAb F310 (Crow and Stockdale, 1986; Zeller et al., 2002), we found no difference between muscle specification and organization upon depletion of Smn (Fig. 7, E–H). These data suggest that low levels of Smn do not adversely affect muscle organization, specification, or development.

As motor axons in smn MO embryos displayed aberrant axon outgrowth, we asked whether these axons were still capable of forming synapses on the muscle fibers. To address this, we analyzed acetylcholine receptor (AChR) clusters in smn MO embryos at 3 d. Using rhodamine-conjugated α-bungarotoxin, we found that all axons in smn MO and control MO embryos colocalized with AChR clusters (Fig. 8, A–F), suggesting that the axons are innervating muscle fibers but lack the stereotyped pattern of innervation (Fig. 8, D–F, arrows). Consistent with this observation, we found no overt movement defects or paralysis in these embryos (unpublished data).

As Smn is expressed in all tissues, it is not clear where its function is required for normal motor axon development. To test whether Smn functions cell autonomously with respect to motoneurons, we reduced Smn levels in single motoneurons in otherwise WT embryos. Optical clarity of zebrafish embryos allows us to identify single primary motoneurons in living embryos and iontophorese these cells with vital fluorescent dyes (Eisen et al., 1989; Beattie et al., 2000). smn or control MO was iontophoresed with rhodamine dextran dye into CaP cell bodies at 19.5–20.5 h; CaP axons were then imaged at 43 h. 100 and 50 μM solutions were found to cause significant cell death when iontophoresed into motoneurons, whereas a solution containing 1 μM of smn MO only caused 45.5% of the CaP motoneurons to die.

CaP motoneurons iontophoresed with 1 μM control MO exhibited no axon defects (n = 10; Fig. 9 A). In contrast, CaP motoneurons, iontophoresed with smn MO when their growth cones were at the first intermediate target, were defective 76.9% of the time (n = 13; Fig. 9, B, C, and D). The defects observed, axon truncations and branches, were consistent with defects seen when the entire embryo was injected with smn MO. CaP motoneurons were also iontophoresed with smn MO when their growth cones were further ventral at the second intermediate target. Only 62.5% of these motor axons (n = 8; Fig. 9 D) were defective, suggesting that the longer the Smn protein is reduced during outgrowth, the more likely pathfinding errors, such as truncations and branches, will occur. The incomplete penetrance of these defects may suggest variability in the phenotype, as is seen when the entire embryo is injected with smn MO.

To show the specificity of these defects, smn MO and fluorescent rhodamine dextran were also iontophoresed into ventral longitudinal descending (VeLD) neurons, spinal interneurons with axons that descend into the tail region by 43 h (Kuwada et al., 1990; Hale et al., 2001). All VeLD interneurons iontophoresed with smn MO (n = 4) had axons that traveled into the tail region and showed no other defects, such as branching (unpublished data). This further indicates that knockdown of Smn protein causes motoneuron-specific defects. Taken together, these data show that Smn functions cell autonomously in motoneurons and plays a critical and specific role in motor axon outgrowth and pathfinding.

Taking advantage of the relatively simple neuromuscular organization of the zebrafish embryo, we have shown that Smn has an essential function in motor axon development. The defects observed in motor axons upon knocking down Smn are independent of cell death and are specific for motoneurons. The data we present indicate that the earliest defects in SMA are in the developing motor axons and suggest that this defect in axon outgrowth and pathfinding may eventually lead to motoneuron death and SMA.

Embryos with decreased Smn levels show defects in motor axon outgrowth and pathfinding during the first three days of development. Defects were rarely seen in the proximal axon extending from the spinal cord to the first intermediate target, suggesting that initiation of axon outgrowth and pathfinding along the common pathway are unaffected by low Smn levels. Secondary motoneurons also displayed branching and truncation defects that continued up to 3 d. As primary and secondary motor axons fasciculate to form nerves, we cannot be certain that the defects observed at 36–72 h are defasciculation of the nerve or true branching of the axons. As branching in single CaP axons is seen, this suggests that true branching does occur. Although the motor axon projections are aberrant, these axons still colocalized with AChR, indicating that functional synapses had formed.

The motor axon branching and truncation defects were initially discovered upon injection of 6 ng of smn MO into zebrafish embryos. When a higher dose of MO, 9 ng, was used, there was a significant increase in the amount of motor axon branching and an apparent decrease in truncations. Analysis of embryos with GFP-expressing motor nerves, however, revealed that truncations were still occurring but were followed by branching, thus obscuring the truncation phenotype. These data indicate that the less Smn protein translated within the neuron, the greater the likelihood that the axon will branch. Ventral nerve truncations were consistently occurring at the second intermediate target. It is believed that intermediate targets are important for growth cones to proceed to their final targets (Tessier-Lavigne and Goodman, 1996). It is possible that in the presence of low Smn, growth cones stall at these locations, suggesting that they still try to assimilate cues. At very low levels of Smn, the growth cone may not even be responding to the intermediate target and branches indiscriminately.

Due to the distribution of SMN protein, it has been difficult to conclusively determine whether the primary defect resides in motoneurons or muscle. In smn MO–injected zebrafish embryos, the motoneurons are defective, but muscle specification, early patterning, and muscle development are normal, indicating that decreasing Smn is critical in motoneurons. We specifically reduced Smn in single motoneurons without affecting Smn levels in other neurons or muscle. Exclusively decreasing Smn in motoneurons recapitulated the motor axon defect seen when Smn was decreased throughout the embryo, indicating that Smn functions cell autonomously. We cannot, however, rule out the possibility that Smn also functions in muscle to affect motor axons.

Recent mouse and human studies suggest that motor axons may be affected before motoneuron loss. Mice exhibiting neuronal depletion of SMN (Smn^F7/Smn^D7, NSE-Cre⁺) exhibited a dramatic decrease in motor axons but only a modest decrease in motoneuron cell bodies at postnatal day 15 (Cifuentes-Diaz et al., 2002). Due to the severe reduction in full-length SMN and the relatively late developmental timing of this reduction, however, it is difficult to determine the disease relevance of these defects. Severe SMA mice (Smn^−/− with one to two copies of SMN2; Hsieh-Li et al., 2000; Monani et al., 2000) show moderate motoneuron cell body loss, whereas mild SMA mice (A2G missense) show no motoneuron cell body loss several days after birth; mild SMA mice do, however, exhibit a mild reduction in both motoneuron cell bodies and motor axon roots at 3.5 mo (Monani et al., 2003). Motor unit number estimation (MUNE) studies in humans show that presymptomatic individuals exhibit high motoneuron counts that dramatically decrease with onset of the disease (Bromberg and Swoboda, 2002). These results indicate that motoneuron cell body loss is a late feature of the disease. Our data are consistent with these findings and support the idea that axons are affected before cell death.

SMN has also been shown to localize in developing motoneurons. Rat SMN colocalizes with cytoskeletal elements in the spinal motor axons during nervous system maturation (Pagliardini et al., 2000). In primary motoneuron cultures from E14 mice embryos, SMN is enriched in branch points and growth cones (Jablonka et al., 2001). SMN has also been found in growth cones of P19 neurons (Fan and Simard, 2002). Interestingly, SMN interacting protein 1 (SIP1), a protein important in spliceosomal snRNP biogenesis (Fischer et al., 1997), does not coenrich in axon branch points, indicating the possibility of a unique SMN complex and function at these locations (Jablonka et al., 2001). Taken together with our findings of abnormal motor axon branching when Smn is depleted, the data suggest that Smn plays an important function in motor axon development and maintenance.

Although a mechanism by which SMN functions in motoneuron development remains unclear, recent data do give clues. SMN has been shown to interact with profilin (Giesemann et al., 1999), a protein involved in actin polymerization and growth cone motility (Gutsche-Perelroizen et al., 1999; Wills et al., 1999; Kim et al., 2001), processes important for axon outgrowth. Data have also implicated an interaction between SMN and motoneuron-enriched RNA processing molecules, hn-RNP-Q and hn-RNP-R (Rossoll et al., 2002). These proteins play a role in RNA transport, editing, and translation. RNA transport into the axon followed by localized protein translation has been shown to be important for growth cones to proceed past intermediate targets (Brittis et al., 2002). It is possible that SMN in axons and growth cones may be associated with RNP complex assembly associated with transport and/or translation of mRNA. Recent experiments have shown that SMN localizes with cytoskeletal filaments and is actively transported down axons (Zhang et al., 2003), further indicating that SMN may be involved in transport and the resulting translation of important axon guidance cues. It is possible that if SMN is reduced, motor axons cannot respond to cues at intermediate targets by turning on local translation, resulting in axon defects. Our finding that GFP-expressing motor nerves in smn MO–injected embryos are truncated at the second intermediate target is consistent with this idea. The validity of this model, however, will depend on elucidating a direct relationship between Smn and localized protein translation machinery in motor axon growth cones.

Adult zebrafish and embryos were maintained as previously described (Westerfield, 1995), allowed to develop at ∼28.5°C, and staged by hours or days after fertilization (Kimmel et al., 1995). All fish used in this study were on the *AB background unless otherwise specified.

Consistent with guidelines for different species, the mouse, human, or rat protein is SMN and the gene is SMN (human) or Smn (mouse). In fish, the protein is Smn, and the gene is smn (http://zfin.org/zf_info/nomen.html).

Zebrafish smn primers (5′-GTGATGATTCTGACATTTGG-3′ and 5′-CCATCCTCACCTTTCAAAGC-3′) were used to map the gene on the LN54 radiation hybrid panel (Hukriede et al., 1999). smn maps to linkage group 5, 3.25 cM from marker fb39c12. Southern blots were performed as previously described (Monani et al., 2000). The smn probe was generated by PCR with the following primer sequences: 5′-GTGATGATTCTGACATTTGG-3′ and 5′-GTCTTCAGAGCATCTTCATCC-3′. The zebrafish islet2 gene was used as a control.

Zebrafish smn cDNA clone (GenBank/EMBL/DDBJ accession nos. Y17256, AA494875, AA494767, and AF083557) was used to make sense and antisense digoxigenin-labeled riboprobes of 1016 bp (Bertrandy et al., 1999). The sense (Sp6) and antisense (T7) smn riboprobes were synthesized from plasmid linearized with XhoI and HindIII, respectively. Islet2 and MyoD riboprobes were synthesized as previously described (Appel et al., 1995; Weinberg et al., 1996). Whole-mount in situ hybridization protocol (1277073; Roche) was performed as previously described (Thisse et al., 1993).

An antisense MO was designed against the 5′ start sequence of the smn gene (Gene Tools, Inc.); 5′-CGACATCTTCTGCACCATTGGC-3′. An additional nonoverlapping MO was also designed ∼20 bp upstream of the original MO (designated 5′ UTR MO); 5′-TTTAAATATTTCCCAAGTCCAACGT-3′. Two- to four-cell *AB, gata2–GFP (Meng et al., 1997), or islet1–GFP (Higashijima et al., 2000) embryos were injected with ∼6 or 9 ng of MO in Danieau's solution with phenol red dye according to protocol (Nasevicius and Ekker, 2000). A standard control (Gene Tools, Inc.) MO (5′-CCTCTTACCTCAGTTACAATTTATA-3′) was used at 6 or 9 ng for control injections. Synthetic capped human SMN mRNA was produced using mMESSAGE mMACHINE kit (Ambion) according to the manufacturer's instructions. mRNA was produced from previously described plasmids (Le et al., 2000) linearized with Xho1. Approximately 300–500 pg of mRNA was coinjected with 9 ng of smn MO.

Immunohistochemistry and imaging were performed essentially as previously described (Beattie et al., 2000). The following mAbs were used: znp1 (1:100; Melancon et al., 1997), anti–acetylated tubulin (1:250; T-6793; Sigma-Aldrich), anti–slow twitch myosin F59 (1:10; Crow and Stockdale, 1986; Devoto et al., 1996), anti–fast twitch myosin F310 (1:10; Crow and Stockdale, 1986; Zeller et al., 2002), 3A10 (1:10; Hatta, 1992), and anti–neurolin zn-5 (1:75; Fashena and Westerfield, 1999). FITC (IgG2A) and TRITC (IgG1) isotype-specific conjugate secondary Abs (1080–02 and 1070–03, respectively; Southern Biotechnology Associates, Inc.) were used for fluorescent detection of znp-1 (IgG2A) and F59/F310 (IgG1) mAbs. Cross-sectional analysis was performed by embedding embryos in 1.5% agar/5% sucrose and sectioning on a crytostat at 16 μm. All immunofluorescent images were analyzed using a confocal microscope and photographed using digital imagery (Nikon Optiphot 2; Bio-Rad Laboratories, Inc. MRC 1024) unless otherwise specified.

MO-injected embryos were manually dechorionated and deyolked in a slurry of physiologic Ringers (Westerfield, 1995), Ringers ice chips, and protease inhibitors (Complete Mini; 1836170; Roche). Deyolked embryo samples were prepared and Western blots were performed as previously described (Monani et al., 2003). Smn (1:1,000; MANSMA7 or MANSMA21; Young et al., 2000b), anti–synaptic vesicle (SV2) protein (1:200; Buckley and Kelly, 1985), and anti–Hu 16A11 (1:500; A-21271; Molecular Probes; Marusich et al., 1994) mAb were visualized using the ECL Detection Kit (RPN 2109; Amersham Biosciences). Quantification of bands was performed using a densitometer (Shimadzu, Inc.).

Live gata2–GFP MO-injected transgenic zebrafish (Meng et al., 1997) were anesthetized in tricaine (A-5040; Sigma-Aldrich) at 36, 52, and 74 h and mounted on a glass coverslip. Individual motor nerves were visualized and imaged using a Photometrics SPOT camera. Images were compiled and edited using Adobe Photoshop^® software. MO-injected islet1–GFP embryos (Higashijima et al., 2000) were fixed at 72 h in 4% paraformaldehyde for 2 h at room temperature.

TUNEL assay was performed according to the manufacturer's protocol on staged embryos (Cole and Ross, 2001). Digoxigenin-labeled dUTP was used to label fragmenting DNA ends (Terminal Transferase Assay; 220582; Roche), and then embryos were sectioned on a crytostat at 16 μm.

MO-injected embryos were staged and fixed in 4% paraformaldehyde for 2 h at room temperature and water soaked for 3–6 h. After collagenase (C-9891; Sigma-Aldrich) treatment, embryos were then incubated in Alexa^®594-conjugated α-bungarotoxin (10 μg/ml; B13423; Molecular Probes) for 30 min, essentially as previously described (Ono et al., 2001). Embryos were then Ab stained using znp-1 and a FITC-conjugated secondary Ab.

Glass capillary microelectrodes were backfilled with a solution containing 2.5% rhodamine dextran (3,000 MW; Molecular Probes) and smn or control MO. 19.5–20.5-h embryos were anesthetized in tricaine and mounted in agar on a microslide as previously described (Eisen et al., 1989; Beattie et al., 2000). Individual CaP motoneurons or VeLD interneurons were impaled electrically by oscillating the electrode tip as previously described (Eisen et al., 1989). Solution was added to the cell by iontophoresis. Images were taken with a Photometrics SPOT camera and compiled and edited in Adobe Photoshop^®.

The supplemental material is available. Fig. S1 diagrams primary and secondary motor axon pathfinding in the developing zebrafish. Fig. S2 shows the expression pattern of smn during the development of zebrafish. Fig. S3 shows that an additional 5′ UTR smn MO results in similar motor axon defects. Table SI shows the percentage of sides with motor axon/nerve defects using additional 5′ UTR smn MO.

We thank Shuo Lin (University of California, Los Angeles, CA) and Michael Granato (University of Pennsylvania, Philadelphia, PA) for gata2–GFP fish; Hitoshi Okamoto (RIKEN Brain Science Institute, Saitama, Japan), Shin-ichi Higashijima (State University of New York at Stony Brook, Stony Brook, NY), and Anad Chandrasekhar (University of Missouri, Columbia, MO) for islet1–GFP fish; Frank Stockdale (Stanford University, Stanford, CA) for F59 and F310 mAb; Thanh Le (The Ohio State University) for SMN constructs; and Jiyan Ma for critical reading of the manuscript. The 3A10 mAb, developed by Thomas M. Jessell and Jane Dodd, and the SV2 mAb, developed by Kathleen M. Buckley, were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the National Institute of Child Health and Human Development and maintained by The University of Iowa Department of Biological Sciences.

U.R. Monani is the recipient of a Development Grant from the Muscular Dystrophy Association of America. This research is supported by National Institutes of Health grant RO1NS41649.