Analysis of zebrafish sidetracked mutants reveals a novel role for Plexin A3 in intraspinal motor axon guidance

In recent years, many of the basic principles of axonal guidance have been defined at the molecular and cellular levels (reviewed in Dickson, 2002). Work in vertebrates and in invertebrates has shown that a small group of conserved guidance molecules, including ephrins, netrins, semaphorins and slits, via their cognate receptors, steer axons towards their synaptic targets. It is also clear now that axonal trajectories are broken down into segments, and that individual guidance decisions along these segments are controlled independently (i.e. using divergent guidance cue/receptor pairs). Although this enables a relatively small number of guidance molecules to generate the intrinsic patterns of neural connectivity, one of the remaining challenges is to `reconstruct' for each neural cell type all the cue/receptor pairs that control each decision along its entire axonal trajectory.

For motor axon trajectories, one of the first guidance decisions is their longitudinal migration towards a segmental spinal cord exit point. Although the spinal cord is externally unsegmented, motor axons exit and sensory afferents enter through distinct, segmentally organized zones along the anteroposterior axis - the motor exit points (MEPs) and the dorsal root entry zone (DREZ), respectively. During development, growth cones from motoneurons located within a spinal segment navigate longitudinally towards a shared segmental exit point. In rodents and birds, these segmental exit points are marked by Krox20 (Egr2)-positive cells - the boundary cap cells(Golding and Cohen, 1997; Niederlander and Lumsden,1996). Elegant ablation studies have shown that boundary cap cells are important to confine motoneuron somata to the spinal cord, but that they appear to be dispensable for guiding motor growth cones(Vermeren et al., 2003). In contrast to later steps of motor axon guidance, the cellular and molecular mechanisms by which motor axons navigate towards the segmental exit point,recognize these zones and subsequently turn ventrally to exit, are poorly understood.

In the zebrafish, three distinct and identifiable motor axons pioneer into the periphery through a common, mid-segmental exit point(Fig. 1A)(Eisen et al., 1986). Because Krox20-positive boundary cap cells have not been identified in zebrafish,caudal primary (CaP) motoneurons located just dorsal of the exit points can be used as a landmark for the exit zone (Fig. 1A). CaP growth cones do not migrate longitudinally, but extend ventrally into the periphery. By contrast, the growth cones of middle primary(MiP) and rostral primary (RoP), the somata of which are located further rostrally, navigate within the spinal cord towards the segmental exit point(Fig. 1A). Once they exit the spinal cord, CaP, MiP and RoP growth cones pioneer a common path towards a somitic choice point, at which they diverge onto cell-type-specific trajectories (Fig. 1A)(Eisen et al., 1986; Myers et al., 1986). Because each of these primary motoneuron subtypes can be identified and visualized by their stereotyped axonal trajectories and cell body position, they present an excellent system in which to study motor axon guidance, including their intraspinal navigation, at single-cell resolution(Beattie et al., 2002).

In this manuscript, we identify the guidance receptor Plexin A3 to play a major role in intraspinal motor axon guidance (see below). Plexins are transmembrane receptors containing a Semaphorin (Sema) domain followed by a cysteine-rich Met-related domain and a unique and conserved Sex-Plexin domain(SP) in the cytoplasmatic region(Tamagnone et al., 1999). Based on sequence similarity, vertebrate plexins are divided into four subfamilies (A-D), and, in mammals, plexin A3 has been shown to mediate axonal repulsion by Sema3A and Sema3F in sympathetic and sensory neurons(Yaron et al., 2005; Cheng et al., 2001). Most of the secreted class 3 semaphorins, including Sema3A and Sema3F, do not bind plexins directly, but signal via a receptor complex composed of one of the four A plexins and the obligate co-receptors neuropilin 1 or neuropilin 2(Chen et al., 2000; Giger et al., 2000). While the neuropilins bind the ligand, the plexins transduce the signal through their cytoplasmic SP domain. Although members of the plexin A subfamily have been shown to control the development of several sensory and CNS trajectories,little is known about the specific role that they play during vertebrate motor axon guidance.

In an antibody-based genetic screen, we had previously isolated mutants of the sidetracked (set) gene, based on the observation that motor axons exit from the spinal cord at ectopic positions(Birely et al., 2005). Here, we show that the sidetracked gene plays a crucial role early during guidance, when pioneering motor growth cones navigate towards and then through the segmental exit point. We positionally cloned the sidetracked gene and showed that it encodes Plexin A3. Using single-cell labeling, time-lapse microscopy and chimera analysis we characterized this previously unknown function of Plexin A3. During intraspinal migration of motor growth cones,Plexin A3 acts cell autonomously to direct subsets of growth cones, those further away, towards the segmental exit point. We propose that this is due to repulsive Plexin A3 signaling, possibly triggered by semaphorins secreted from the posterior somite. Thus, our results provide the first molecular genetic insights into the earliest pathfinding events of spinal motor axons, when they orient and path-find inside the spinal cord towards the exit point.

All experiments, except where indicated, were performed with the sidetracked^p55emcf allele(Birely et al., 2005). The Tg(Hb9:GFP) transgenic line(Flanagan-Steet et al., 2005)primarily labels motoneurons, but also labels ventral interneurons, presumably ventral longitudinal descending spinal interneurons (VeLDs).

Antibody stainings were performed as previously described(Zeller et al., 2002). The following primary antibodies were used: znp-1 [1:200;(Trevarrow et al., 1990);Antibody Facility, University of Oregon]; and SV2 (1:50, Developmental Studies Hybridoma Bank, University of Iowa). Antibodies were visualized with corresponding Alexa-Fluor-488, -546 or -594 conjugated secondary antibodies(1:500; Molecular Probes, Eugene, OR). Embryos were imaged using confocal microscopy. Colorimetric in situ hybridizations were performed according to Odenthal and Nusslein-Volhard (Odenthal and Nusslein-Volhard, 1998). Images were processed using Adobe Photoshop and Adobe Illustrator.

Chimeric embryos were generated and analyzed as previously reported(Zeller and Granato, 1999). Donor cells were from Tg(Hb9:GFP) transgenic embryos.

Wild-type and mutant embryos bearing the Tg(Hb9:GFP) transgene were anesthetized in 0.02% tricaine and embedded in 1.5% low-melting-point agarose. Time-lapse analysis was carried out on a confocal microscope. At 3-minute intervals, z-stacks of ∼15 μm were captured and then flattened by maximum projection.

Embryos were injected with Hb9:GFP plasmid at the one-cell stage and allowed to develop to 24 hours post fertilization (hpf). Fixed embryos were stained with anti-GFP and SV2 antibodies, mounted and confocal images taken. For scoring, we measured the distance between outlying segmental exit points (a), and between the soma and the endogenous exit point (b). We then calculated the ratio (a:b) and found that, for wild-type MiP (0.19-0.45) and RoP (0.22-0.57), the ratio is between 0.19 and 0.57, whereas, for CaP:VaP,this ratio is always less than 0.19. We then applied this to sidetracked mutants and considered all soma with a ratio between 0.19 and 0.57 as presumptive MiP/RoP neurons.

To clone plexin A3, sidetracked^p13umal mutant embryos were first identified using SV2 antibody staining, their genomic DNA was isolated, and the entire coding sequence of ENSDARG00000007172 and ENSDARESTT00000003054, which contains the 5′ region including the translational start, was amplified with gene-specific primers. The amplification products were cloned, sequenced and analyzed using Sequencher software.

We had previously shown that, in sidetracked mutants, motor axons displayed two prominent phenotypes: branching at various points along the axon and exiting from the spinal cord at ectopic locations(Fig. 1B,C)(Birely et al., 2005). To identify which of the three subpopulations (CaP, MiP, RoP) exit ectopically,we labeled individual motoneurons by injecting the motoneuronal Hb9:GFP construct into one-cell-stage embryos(Flanagan-Steet et al., 2005). This results in embryos with a number of stochastically labeled motoneurons. Analysis based on cell body position and axonal trajectory of wild-type siblings at 24 hours post fertilization (hpf) revealed the stereotypic somata position and axonal trajectories for RoP, MiP and CaP, demonstrating that all three primary motoneurons are labeled by this method(Fig. 1D-F; see Materials and methods for scoring of individual motoneurons). In addition, this also labeled a fourth, variable primary motoneuron, VaP, which is present in less than 50%of the hemisegments, and which is also located dorsal to the exit zone,adjacent to CaP (Eisen et al.,1990).

By contrast, analysis of sidetracked mutants revealed that motoneuron somata located immediately above the endogenous exit point (i.e. presumptive CaP/VaP neurons) were unaffected (14/15; Fig. 1G), whereas approximately 50% (17/32) of motoneurons with cell bodies located rostrally to the exit point (i.e. presumptive MiPs and RoPs) displayed dramatic pathfinding defects. In mutants, presumptive RoP/MiP axons projected towards the exit point but bypassed it (6/32, Fig. 1H,I),or, instead of projecting caudally towards the exit point, projected rostrally(4/32, Fig. 1J), or exited ectopically at the position of their cell body (7/32, Fig. 1K,L). Thus, single-cell labeling demonstrates that the sidetracked phenotype is not simply caused by axonal overgrowth, but rather that sidetracked plays a pivotal role in intraspinal guidance.

We next used time-lapse microscopy to examine the behavior of sidetracked growth cones in intact, live embryos. In Tg(Hb9:GFP);sidetracked siblings, motor axons exited at segmental exit points and navigated into the periphery (Fig. 2A and see Movie 1 in the supplementary material). In Tg(Hb9:GFP); sidetracked mutants CaP and VaP axons navigate like their wild-type counterparts through segmental exit points, whereas presumptive RoP/MiP axons exit at ectopic positions, taking novel paths into the periphery (Fig. 2B and see Movie 2 in the supplementary material). By contrast, presumptive VeLD interneurons, axons of which extend caudally, navigate properly in sidetracked mutants (Fig. 1A,B, arrowheads), suggesting that intraspinal guidance in general is unaffected.

Although we show that, in sidetracked mutants, presumptive VeLD and primary commissural ascending (CoPA) interneuron axon projections are indistinguishable from those in wild type(Fig. 2 and data not shown), we cannot exclude the possibility that sidetracked also plays a role in the guidance of other spinal cord neurons. However, it is unlikely that the observed phenotypes are caused by mis-specified or misplaced motoneurons. We previously showed that, in sidetracked mutants, anteroposterior polarity of the overlying somites, and the number and position of each of the three primary motoneuron populations - CaP, MiP and RoP - are unaffected [Fig. 8 in Birely et al. (Birely et al.,2005)]. Moreover, cell transplantations have shown that, when placed in ectopic locations, motoneurons either develop axonal trajectories appropriate for their original soma positions or appropriate for their new soma position, but never exit the spinal cord at ectopic positions(Eisen, 1991). Thus, we conclude that sidetracked primarily guides growth cones of the distantly located RoP and MiP motoneurons towards the segmental exit point.

Finally, we found that sidetracked is also crucial for axonal morphogenesis once motor growth cones navigate the periphery. In contrast to Hb9:GFP-labeled wild-type axons, sidetracked motor axons already displayed excessive branching along the entire shaft at 24 hpf. This is unlikely to be a mere consequence of inappropriate axonal pathfinding, because sidetracked motor axons that exited through the endogenous exit point also displayed this phenotype (Fig. 2C,D). In sidetracked mutants, exuberant branching became even more pronounced at 48 hpf (Fig. 2E,F). In wild-type embryos, such extensive side branches have only developed by around 72 hpf, and invade the myotome to form distributed and myoseptal neuromuscular synapses(Downes and Granato, 2004; Liu and Westerfield, 1990),suggesting that sidetracked functions to restrict motor axons from extending into non-synaptic muscle territories precociously. Thus, sidetracked motor axons exit at ectopic spinal cord locations and also form excessive branches that invade inappropriate territories, consistent with the idea that both phenotypes are caused by defective axonal repulsion.

We first used bulk segregant analysis to locate the sidetrackedmutation on chromosome 8 (Fig. 3A). We then refined the map position by meiotic recombination mapping, defining a crucial interval between marker fb82f05.x1 (1/420 recombinants) and marker Z43564b (2/598 recombinants). On the ENSEMBL genome assembly, these two markers delineate a 480 kb interval, in which many predicted and three known protein coding genes are located: sodium and chloride dependent creatine transporter 1 (ENSDARG00000043646),calcium/calmodulin-dependent protein kinase type 1b (ENSDARG00000043648) and plexin A3 (ENSDARG00000007172)(Fig. 3A). The conceptually translated protein (XM_690717) is 61% identical to zebrafish Plexin A4(NP_001004495) but 73% identical to human and mouse Plexin A3, suggesting that XM_690717 encodes zebrafish Plexin A3. Sequencing of the plexin A3coding region from sidetracked mutants revealed a single base-pair change (T2312A), resulting in a nonsense mutation, truncating the protein at position 662 in the second PSI (Plexin-Semaphorin-Integrin) domain(Fig. 3B,C). Because this truncates the protein before the transmembrane domain and the signal transducing Sex-Plexin domain, the sidetracked^p13umalallele probably represents a null allele. Although it is possible that the truncated protein is expressed, it is unlikely that it retains biological activity because it lacks both the transmembrane and the signal transducing domains. Thus, given the absence of a second plexin A3 in the zebrafish genome, sidetracked mutants probably lack all Plexin A3-mediated semaphorin signaling.

During the time period of intraspinal motoneuron guidance (18-24 hpf), plexin A3 expression is detectable at low levels throughout the spinal cord, and in each spinal segment is enriched in two ventral neurons positioned at the level of the somite boundary(Fig. 3D,E). The dorsoventral and anteroposterior position within each segment is consistent with the position of MiP and RoP motoneurons (Fig. 1A,E,F) (Feldner et al.,2005). To determine whether sidetracked indeed functions in these two motoneurons, we generated chimeric embryos in which labeled wild-type blastula cells were transplanted into age-matched sidetracked hosts. Analysis of 26 hpf chimeric embryos revealed that MiP and RoP motoneurons derived from wild-type donors developed wild-type-like axonal trajectories in sidetracked mutants (n=6/6 RoPs, 6/6 MiPs, Fig. 4A). Conversely,mutant-derived motoneurons developed sidetracked-like axonal trajectories when transplanted into wild-type hosts (n=4/6 RoPs; 4/5 MiPs; Fig. 4B). Thus, Plexin A3 acts cell autonomously in MiP and RoP, suggesting a role as a receptor that is crucial for intraspinal guidance.

How does Plexin A3 control intraspinal motor axon guidance? In most circumstances in which it has been examined, type A plexins transduce axonal repulsion (reviewed in Negishi et al.,2005; Tamagnone et al.,1999). One attractive model is that Plexin A3-sensitive RoP and MiP growth cones avoid a repellent that diffuses anteriorly and posteriorly from the posterior part of each somite(Fig. 4E,F). Such a simple model would explain why wild-type RoP and MiP growth cones migrate caudally to a mid-segmental point, where repulsion would be minimal. It would also account for the various pathfinding defects observed in sidetracked mutants:RoP and MiP, now insensitive to the repellent, bypass the endogenous exit point, or turn rostrally towards the next anterior exit point(Fig. 1H,J), or just exit the spinal cord below their soma (Fig. 1K,L).

We next examined the distribution of known plexin ligands at the posterior somite compartment. Plexin A family members function in a complex with neuropilins as receptors for semaphorins, in particular for Sema3 family members (reviewed in Negishi et al.,2005). In the zebrafish, a number of Sema3 family members have been identified, including Sema3Aa (Yee et al., 1999), Sema3Ab (Roos et al., 1999), Sema3C (Yu and Moens, 2005), Sema3D (Halloran et al., 1999), Sema3Fa, Sema3Fb, Sema3Ga, Sema3Gb(Yu and Moens, 2005) and Sema3H (Stevens and Halloran,2005). Of these, Sema3Aa has been reported to be expressed in the posterior region of each somite, and knockdown of sema3Aa and nrp1a expressed in motoneurons affects axon guidance in the periphery(Feldner et al., 2005; Sato-Maeda et al., 2006). We therefore examined two other semaphorin ligands, Sema3Fa and Sema3Ab. In sensory and sympathetic neurons, the Plexin A3-Neuropilin 2 complex preferentially transduces the activities of Sema3F semaphorins(Yaron et al., 2005). Whereas sema3Fa is expressed diffusely in ventral somites(Fig. 4C), Sema3Ab is localized in a discrete band in the posterior part of each somite(Fig. 4D)(Roos et al., 1999), more consistent with our model. However, systematic studies to examine the expression patterns of all semaphorin genes, as well as additional functional studies, will be required to identify the precise mechanisms by which Plexin A3 mediates intraspinal axonal guidance.

Independent of the model, our studies, as well as those by Tanaka et al. in this issue (Tanaka et al.,2007), provide the first genetic demonstration that Plexin A3 signaling plays a crucial role during intraspinal motor axon guidance. Although previous studies have demonstrated a role for class 3 semaphorins and neuropilins in later stages of motor axon guidance [i.e. in the periphery(Huber et al., 2003)], our results clearly implicate a plexin signaling pathway in the very early steps,when growth cones orient towards the segmental exit point. The finding that sidetracked encodes Plexin A3 provides two novel insights. First, our results show that motor axons can exit the spinal cord at seemingly random positions. Previous studies had suggested that sites of axonal entry or exit from the spinal cord are restricted and prefigured by the presence of specialized neural crest derivatives - the boundary cap cells(Niederlander and Lumsden,1996). Our results suggest that the neural tube is competent to form exit points along its entire length, but that the precise locations of segmental exit points are determined by the growth cone.

Second, intraspinal guidance towards spinal exit points is governed, at least in part, by repulsive guidance. In the chick, elegant rotation experiments have shown that inverting rhombomeres in their anteroposterior orientation does not alter the growth of motor axons towards their anterior-lying exit points, which suggested that exit points represent a chemoattractive, intermediate target guiding motor axons(Guthrie and Lumsden, 1992). Our results clearly demonstrate that intraspinal guidance requires Plexin A3 signaling, which propagates the repulsive activities of semaphorins. Our model is consistent with recent results obtained by Feldner et al using plexin A3 as well as semaphorin 3A morpholinos(Feldner et al., 2007), as well as by Tanaka et al. (Tanaka et al.,2007). Although this does not exclude the co-existence of a chemoattractive mechanism, our results present compelling genetic evidence that a particular guidance mechanism - repulsion - guides the intraspinal migration of motor growth cones.

Our studies also reveal a later role for Plexin A3 in restricting axonal branching. In sidetracked (plexin A3) mutants, branches invaded the myotome prior to the time period when wild-type motor axons do(Fig. 3F). This phenotype is somewhat reminiscent of the behavior that mossy fibers display in plexin A4 mutant mice (Suto et al.,2007). There, it has been proposed that plexin A4 prevents mossy fibers from invading the entire CA3 region, thereby restricting them to a narrow zone (Suto et al.,2007). Thus, similar to the situation in the CNS, sidetracked (plexin A3) signaling in the periphery appears to be crucial in order to prevent precocious spreading of motor axonal branches into future synaptic muscle fields, possibly coordinating pre- and postsynaptic development.

We would like to thank V. Schneider and J. Birely for help with the in situ hybridizations; C. Moens (HHMI, WashU) for providing sema3Fa; and D. Gilmour(EMBL) for comments on the manuscript. We would also like to thank H. Tanaka and H. Okamoto for communicating unpublished results. This work was supported by grants from the American Heart Association (K.A.P.), the National Science Foundation (M.G.) and the National Institute of Health (M.G.).