Motor neuron cell bodies are actively positioned by Slit/Robo repulsion and Netrin/DCC attraction

Abstract

Motor neurons differentiate from a ventral column of progenitors and settle in static clusters, the motor nuclei, next to the floor plate. Within these cell clusters, motor neurons receive afferent input and project their axons out to muscle targets. The molecular mechanisms that position motor neurons in the neural tube remain poorly understood. The floor plate produces several types of guidance cues with well-known roles in attracting and repelling axons, including the Slit family of chemorepellents via their Robo receptors, and Netrin1 via its DCC attractive receptor. In the present study we found that Islet1⁺ motor neuron cell bodies invaded the floor plate of Robo1/2 double mutant mouse embryos or Slit 1/2/3 triple mutants. Misplaced neurons were born in their normal progenitor column, but then migrated tangentially into the ventral midline. Robo1 and 2 receptor expression in motor neurons was confirmed by reporter gene staining and anti-Robo antibody labeling. Mis-positioned motor neurons projected their axons longitudinally within the floor plate, and failed to reach their normal exit points. To test for potential counteracting ventral attractive signals, we examined Netrin-1 and DCC mutants, and found that motor neurons shifted dorsally in the hindbrain and spinal cord, suggesting that Netrin-1/DCC signaling normally attracts motor neurons closer to the floor plate. Our results show that motor neurons are actively migrating cells, and are normally trapped in a static position by Slit/Robo repulsion and Netrin-1/DCC attraction.

Introduction

Motor neurons are a prototypical example of a cell type induced at a specific position within a morphogen gradient. Because of their position near the floor plate, neural progenitors in a narrow column are exposed to a specific concentration of Sonic hedgehog morphogen and respond by expressing a cascade of transcriptional regulators that specifies their differentiation as motor neurons and clustering into motor nuclei (Briscoe and Ericson, 2001; Briscoe et al., 1999; Osterfield et al., 2003). The position of motor neuron clusters is important for the function of motor circuits, because motor neuron position influences where their axons exit the CNS toward their peripheral muscle targets (Bravo-Ambrosio and Kaprielian, 2011), which synaptic inputs they receive from sensory and other neurons (Chang and Balice-Gordon, 2000; Fritzsch et al., 1993; Song and Pfaff, 2005), and whether neighboring neurons become electrically coupled within motor pools (Brenowitz et al., 1983; Kiehn and Tresch, 2002). Motor neuron cell bodies can shift locally within their nucleus to settle into topographic order (Leber and Sanes, 1995), with these shifts regulated by Reelin and cadherin signaling (Demireva et al., 2011; Palmesino et al., 2010; Price et al., 2002; Yip et al., 2000). Therefore the functional organization of motor nuclei relies on motor neurons being precisely induced in the correct position and remaining in their local nucleus.

However, a few motor neuron populations in the brain stem do undergo substantial migrations out of their local motor nuclei. A well-known example is the facial branchiomotor population which migrates posteriorly to traverse hindbrain segments alongside the floor plate (Auclair et al., 1996; Bingham et al., 2010; Chandrasekhar, 2004; Fritzsch and Nichols, 1993; Glasco et al., 2012). Likewise, a subset of oculomotor neuron cell bodies migrates across the ventral midline of the midbrain to the contralateral nucleus (Chilton and Guthrie, 2004; Fritzsch et al., 1995; Puelles-Lopez et al., 1975). Interestingly, Semaphorin or receptor mutations cause motor neuron cell bodies to translocate out of the spinal cord (Bron et al., 2007; Vermeren et al., 2003), which suggests broader capacity of motor neurons to migrate.

Because motor neuron cell bodies cluster close to the floor plate, guidance molecules from the floor plate could regulate the position of these cells. The Slit family of repellents, and their Robo receptors, guide migration of diverse neuron types (Kidd et al., 1999; Wu et al., 1999). However, tests of Slit/Robo functions in motor neuron position and axon guidance have given mixed results. Expression of dominant negative Robo receptors in chick hindbrain motor neurons, predicted to interfere with Slit responses, caused some motor axons and motor neuron cell bodies to shift into the floor plate, consistent with Slit/Robo repulsion (Hammond et al., 2005). However, contrasting with the chick embryo experiments, mouse embryos mutant for Robo1, Robo2, or with combined Slit1 and 2 mutations, resulted in motor axons growing into the floor plate, but not motor neuron cell bodies (Hammond et al., 2005), which was further supported by a separate analysis of Robo1/2 combined mutant mouse embryos (Bai et al., 2011). In contrast, Slit2 promotes the axon outgrowth in culture of a spinal cord accessory motor axon population (Bravo-Ambrosio et al., 2012). The discrepancies between these results leave unresolved the in vivo role of Slit/Robo signals in guiding motor neurons in hindbrain or spinal cord. As for counteracting attractive signals from the floor plate, Netrin-1 and its attractive receptor DCC are required in many neuronal populations for axon guidance and neuron migration (Culotti and Merz, 1998; Kawasaki et al., 2006; Serafini et al., 1996; Yee et al., 1999). Surprisingly, a repulsive function of Netrin-1 for motor neurons has been proposed, because cranial motor axons in vitro are repelled by Netrin-1 (Colamarino and Tessier-Lavigne, 1995; Murray et al., 2010; Varela-Echavarría et al., 1997), and this repulsive effect can be blocked by antibodies against the repulsive Netrin-1 receptor Unc5a (Murray et al., 2010). Similar to Slit perturbations, mouse Netrin-1 mutations caused a subset of cranial motor axons to grow into the floor plate, but consistent misplacement of motor neuron cell bodies into the floor plate was observed only in chick embryos following expression of a dominant negative Unc5a receptor (Murray et al., 2010). In contrast, cultures of spinal cord motor axons are normally unresponsive to Netrin-1, although a switch to attraction can be caused by perturbations of Presenilin and its proteolytic processing of the DCC receptor (Bai et al., 2011). Together, the above studies suggest that floor plate signals are important for guiding motor axons, and also in some cases for positioning their cell bodies. However, the discrepant results between genetic strategies, i.e. knockout mice, and misexpression/dominant negative strategies, i.e. in chick, leave unresolved the in vivo relevance of which floor plate guidance cues might set the position of motor neuron cell bodies, and whether these roles are positive or negative.

The present study is based on our unexpected observation that neuronal cell bodies were present in the hindbrain floor plate of early Slit and Robo mutant mouse embryos (Kim et al., 2011). This observation led us to focus on the earliest stages of motor neuron development in the brain stem and spinal cord to reevaluate the in vivo roles of Slit/Robo and Netrin1/DCC signals in positioning motor neuron cell bodies. Through a range of observations on mutant mouse embryos, we found that the normal position of motor neurons is due to migratory reactions to repulsive Slit and attractive Netrin signals.

Materials and Methods

Mouse embryos

Mice were maintained and the experiments carried out in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, by protocols approved by the University of Nevada, Reno Institutional Animal Care and Use Committee. Embryonic day 9.5 (E 9.5), E10 and E10.5 embryos were obtained via uterine dissection. Wild type CD-1 mice (6–8 weeks old) were purchased from Charles River Laboratories (Wilmington, MA USA). The Robo, Slit, Netrin-1, and DCC mutant strains were gifts of Marc Tessier-Lavigne, Genentech, and Frederic Charron, ICMR, Montreal CA. Robo and Slit PCR genotyping were performed as previously described (Grieshammer et al., 2004; Plump et al., 2002). Netrin-1 and DCC genotyping were performed as previously described (Fazeli et al., 1997; Serafini et al., 1996). The Islet-1^MN-GFP-F strain was a gift of Samuel Pfaff (Lewcock et al., 2007), Salk Institute, and was crossed into CD1 and Robo1/2 mutant backgrounds.

Immunohistochemistry

For whole-mount immunolabeling, embryos were fixed in 4% PFA overnight and prepared by dissecting out the neural tube and washing for several hours in PBS containing 10% fetal bovine serum and 1% Triton X-100 (PBST). Primary antibodies used were rabbit anti-βIII-tubulin (Covance. 1:1000), rabbit anti-Robo1 and Robo2 (kind gift of Elke Stein, Yale, 1:10,000), mouse anti-Islet-1 (DSHB, 1:100), goat anti-DCC (Santa Cruz, 1:250), mouse anti-NKX6.1 (DSHB, 1:100), rabbit anti-Olig2 (Millipore, 1:1000), rabbit anti-Phox2b (Pattyn et al., 1997) (1:2000), and mouse anti-4C7 (DSHB, 1:1). Primary antibodies were applied in PBST for 2–3 days. Our previous characterization of the Robo1 and Robo2 antisera showed specific labeling by immunofluorescence that was lost in homozygous mutants (Kim et al., 2011). After washing in PBST overnight, secondary antibodies (Jackson Immuno Laboratories) were applied in PBST for 2–3 days, followed by overnight washes. For cryostat section immunolabeling, embryos were embedded in a solution of 15%/sucrose and 7.5% gelatin, frozen, and then sectioned at 16 μm using a cryostat (Leica). To melt gelatin off of tissue sections, slides were placed in warm (37–45°C) 0.1 M phosphate buffer for a couple of minutes. Sections were washed for 30 min to an hour in PBST (0.1% Triton X-100). Primary antibody was applied, and then slides incubated in a humidified chamber for 4 hours to overnight. After washing several hours in PBST, secondary antibodies were applied for 2 hours, followed by several washes. In situ hybridization for Slit1 was carried out using probes previously described on cryosections (Farmer et al., 2008), followed by Islet1 antibody labeling.

Motor axon tracing with diI

To trace motor axons from mis-positioned motor neurons, embryos were fixed in 4% PFA overnight. Embryos were dissected using fine forceps to remove the skin and mesenchyme, just over the target site. Small crystals were inserted using a fine tungsten needle under a dissecting microscope. The embryos were placed at 37°C for 1–2 days in 4% PFA to allow the dye to diffuse, and then examined by fluorescence microscopy.

Quantification of motor neuron position defects

The number of Islet-1+ cell bodies in the floor plate was counted under a fluorescence microscope from cryostat sections at the levels of the r1, r4, and brachial spinal cord from Robo and Slit mutant and control embryos on E9.5 and E10.5. The number of total Islet-1+ cells in sections was counted and then the percentage of mis-positioned motor neurons was calculated.

The distance between nIV was used to quantify position of motor nuclei in Netrin-1 or DCC embryos. The measurements were made on Islet-1 labeled whole mount embryos on E10.5. TIFF images of Islet-1 labeled embryos were imported into Image J (Meijering et al., 2004). Varying sizes of embryos were normalized by comparing to the embryo width at the midbrain-hindbrain boundary.

To measure the position of r4 and spinal motor neurons in Netrin-1 or DCC embryos, the distance from the midline to the lower edge of the motor column and the circumferential distance between the ventral midline and the lower edge of the motor column was quantified from r4 and brachial spinal cord sections on E10.5, respectively, with varying stages of embryos normalized by the distance between the dorsal-ventral midlines.

Data are expressed as means + S.E.M, and differences tested for significance using student t-tests. Data are considered significantly different from the control values were when p <0.05.

Results

Islet-1⁺ motor neurons enter the floor plate when Robos are missing

The floor plate in the hindbrain and spinal cord is normally devoid of neuron cell bodies. Unexpectedly, we previously observed that βIII-tubulin⁺ cell bodies are located within the ventral midline of the hindbrain in E10.5 Robo1^−/−;2^−/− mutants (Kim et al., 2011).

Since motor neurons are generated early and located adjacent to the floor plate, we used a motor neuron marker, the transcription factor Islet-1(Ericson et al., 1992), to attempt to identify the mis-positioned cell bodies (Fig 1, Table 1). We found that more than 90% of βIII-tubulin⁺ cell bodies located within the hindbrain ventral midline of Robo1^−/−;2^−/− mutants are Islet-1⁺ (n=4 embryos, n>12 sections from each), suggesting that Islet-1⁺ motor neurons were specifically sensitive to loss of Robo. (The HB9 antibody was also used to verify motor neuron identity in posterior sections; data not shown; however, this antibody does not label motor neurons in anterior hindbrain). In Robo1^−/−;2^−/− mutant embryos, Islet-1⁺ cells entered into the floor plate on E9.5 and E10.5 (Fig 1B, D). To verify the location of the ectopic motor neurons, Islet-1 antibody labels were combined with Slit1 in situ hybridization. Whereas in wild type, motor neurons were positioned just lateral to the Slit1⁺ floor plate, the subset of ectopic motor neurons were clearly located within the Slit1⁺ floor plate in Robo mutants (Fig 1E, F). Nearly 30% of the Islet-1⁺ cells were mis-located on E9.5 in the hindbrain (Fig 1I). By E10.5, when motor neuron numbers increase rapidly, 10% of motor neurons were located in the floor plate in r1 and r4 (Fig 1I). In addition, motor neurons in the spinal cord were mis-located on E9.5 and E10.5. However, the largest number of mis-positioned neurons in the floor plate were in r1 and in r4 (Fig 1I). We note that it is not clear whether a random subset of Isl1⁺ motor neurons, or a specific subset, migrate into the floor plate. Further studies are needed to reveal which subtypes of motor neurons are restricted by Slit/Robo signals at early stages.

Figure 1. Islet-1⁺ motor neurons enter the floor plate in Robo1/2 double mutants.

Islet-1 labeling on cryosections of E9.5 and E10.5 embryos. A–D. Islet-1 labeling in Robo1^+/+;2^+/+ and Robo1^−/−;2^−/− E9.5 and E10.5 embryos (n=4 embryos, n=12 sections for each genotype of E9.5, n=8 embryos, n=24 sections for each genotype of E10.5) showing Robo1^−/−;2^−/− embryos had Islet-1⁺ cell bodies in the floor plate (A, B in r1, C,D in r4). E, F. Islet-1 labeling combined with in situ for Slit1 of Robo1^+/+;2^+/+ and Robo1^−/−;2^−/− r1 E10.5 embryos. In wild type, motor neurons stay just outside of the Slit1-expressing floor plate, but in Robo mutants a subset shift into the Slit⁺ floor plate. G, H. Schematics of locations of Islet-1⁺ cell bodies in Robo1^+/+;2^+/+ and Robo1^−/−;2^−/− embryos. I. Summary graphs show percentage of motor neurons in the floor plate of Robo1^−/−;2^−/− embryos. The percentage of motor neurons in the floor plate was higher for the embryos quantified on E9.5 than E10.5 with higher percentages in the hindbrain than in the spinal cord (*, p<0.05; **, p<0.01). Scale bars: A–D, 50 μm; C–F, 100 μm. MN: motor neuron, FP: floor plate.

Table 1.

Summary of ectopic MNs in the floor plate in Robo mutants

		R1+/+;2+/+			R1−/−;2−/−			R1+/−;2−/−			R1−/−;2+/−
		Total number of MNs	Number of MNs in FP	% of MNs in FP	Total number of MNs	Number of MNs in FP	% of MNs in FP	Total number of MNs	Number of MNs in FP	% of MNs in FP	Total number of MNs	Number of MNs in FP	% of MNs in FP
10.5 n=8	r1	112.8 + 4.1	0	0	111.8 + 3.1	10.8 + 1.6	9.6 + 0.7	108.2 + 3.7	5 + 1.5	4.6 + 1.5	115 + 3.8	0.8 + 0.3	0.6 + 0.2
	r4	109.3 + 2.8	0	0	109 + 5.3	11.8 + 0.9	10.8 + 1	114.3 + 4.4	5.8 + 0.9	5 + 1.8	119.3 + 4.2	0.6 + 0.2	0.5 + 0.2
	SC	219.3 + 6.5	0	0	206.5 + 9.5	1 + 0.3	0.5 + 0.5	205 + 3.7	0.6 + 0.2	0.3 + 0.2	204.3 + 6.8	0	0
9.5 n=4	r1	19.3 + 4	0	0	15.3 + 2.5	4.3 + 0.8	27.8 + 2.3
	r4	22 + 6.2	0	0	17 + 1.4	4.8 + 0.5	27.9 + 2.5
	SC	22.7 + 3.1	0	0	20.8 + 1.5	1 + 0.4	4.8 + 0.5

To follow the fate of the mis-located Islet-1⁺ motor neurons, we found that Islet-1⁺ cells were no longer visible in the floor plate by E12.5 (data not shown), implying that the cells die, turn off the Islet1 marker, or migrate out of the floor plate. These findings suggest that Robo receptors are required to set the position of motor neuron cell bodies by keeping them out of the floor plate at an early embryonic stage.

Motor neurons are generated from normal motor progenitor columns, then migrate tangentially into the floor plate

The appearance of motor neurons in the floor plate in Robo1^−/−;2^−/− mutants suggests either that motor neurons are generated in their normal lateral position, then migrate into the floor plate, or are derived from ectopic progenitors abnormally differentiating within the floor plate. To investigate how motor neurons come to reside in the floor plate, we tested whether motor neuron progenitors shifted in Robo mutants by labeling with the motor progenitor column markers, NKX6.1 and Olig2 (Novitch et al., 2001). Since Olig2 is not expressed in anterior regions, we used NKX6.1 antibody to label progenitor cells located in r1. NKX6.1 was normally expressed by both progenitor cells and motor neurons derived from them (Fig 2A). In Robo1^−/−;2^−/− mutants, NKX6.1⁺ progenitor columns were located at their normal positions, while NKX6.1⁺ cells streamed ventrally along the mantle layer into the floor plate (Fig 2B). The progenitor population number appeared unchanged, because the number of NkX6.1+ cells in mutants was not significantly different from that in wild type embryos (357.3 + 16.7 in wild-type, n=3 embryos, n=9 sections, vs. 373.7 + 21.2 in mutant, n=3 embryos, n=9 sections). One caveat in using the NKX6.1 marker is that it could not completely eliminate the possibility that the weakly NKX6.1⁺ progenitors may migrate into the floor plate in Robo1^−/−;2^−/− mutants (Müller et al., 2003). However, a similar marker for more posterior regions, Olig2, clearly revealed that it was normally expressed on motor neuron progenitor cells located in r4 and posteriorly through the spinal cord, and double labeling showed the appearance of Islet-1⁺ motor neurons within the pial surface of the Olig2 progenitor columns (Fig 2C). In Robo1^−/−;2^−/− mutants on E9.5, Olig2⁺ progenitor cells were located at their normal position, lateral to the floor plate, while superficial Islet-1⁺ motor neurons entered into the floor plate (Fig 2D). The number of Olig2⁺ cells in mutants was also not significantly different from wild type embryos (82.7 + 6.4 in wild type, n=3 embryos, n=9 sections vs. 88.4 + 7.9 in mutant, n=3 embryos, n=9 sections). These results show that progenitor columns form in Robo1^−/−;2^−/− mutants, but as motor neurons differentiate, a subset migrates into the floor plate (Fig 2I).

Figure 2. Motor neurons are generated from normal progenitor columns, then migrate tangentially into the floor plate in Robo mutants.

Motor progenitor and neuron markers in Robo1^+/+;2^+/+ and Robo1^−/−;2^−/− E9.5 and E10.5 embryos. A, B. Nkx6.1 labeling of r1 progenitors and motor neurons in E10.5 Robo1^+/+;2^+/+ (n=3 embryos, n=9 sections) and Robo1^−/−;2^−/− (n=3 embryos, n=9 sections) embryos. Nkx6.1⁺ trochlear motor neurons (arrows) entered the floor plate from normal progenitor columns in Robo1^−/−;2^−/− embryos. C, D. Olig2 and Islet-1 double labeling of E9.5 Robo1^+/+;2^+/+ (n=3 embryos, n=9 sections) and Robo1^−/−;2^−/− embryos (n=3 embryos, n=9 sections) shows that motor neurons (arrowheads) in r4 entered the floor plate from normal Olig2⁺ progenitor columns. E–G. Images of early stage ectopic migrating motor neurons. Cryosections of Isl1^MN-GFP-F Robo1^+/+;2^+/+ and Robo1^−/−;2^−/− E9.5 embryos (n=3 embryos, n=9 sections for each genotype). Mis-positioned motor neuron cell bodies had a bipolar shape (*), with axon-like processes leading into the midline in Robo1^−/−;2^−/− embryos. Arrowheads show motor axons in the floor plate. Note that the motor neuron processes precede the arrival of any commissural axons at the floor plate. G. Schematic of tissue organization in Robo1^+/+;2^+/+ (left side) and Robo1^−/−;2^−/− (right side), showing motor neurons (MNs) are generated from NKX6.1⁺ or Olig2⁺ progenitor cells (PC), and then migrate tangentially into the floor plate (FP) in Robo1^−/−;2^−/− mutants. Scale bars: A–D, 100 μm; C, D, 50 μm; E–H, 25 μm. PC: progenitor cell.

The Islet-1 marker showed that motor neuron cell bodies moved into the floor plate. However, to observe how more directly motor neuron cell bodies migrate, we used the motor neuron-specific transgenic reporter Isl1-GFP (Lewcock et al., 2007) in a Robo1/2 mutant background. Consistent with the our previous study in which no βIII-tubulin⁺ neuron cell bodies can be found in the anterior hindbrain floor plate of wild type embryos (Kim et al., 2011), no Isl1-GFP⁺ motor neuron cell bodies or axonal processes were found in the floor plate of wild type embryos (Fig 2E, G). In Robo1/2 mutants, the Isl1-GFP⁺ marker clearly showed motor neuron cell bodies and their extending axonal processes at the earliest stage of migration, E9.5 in the hindbrain and spinal cord. The average number of axons in the floor plate of the hindbrain and spinal cord were similar (5.3 + 0.8/section and 3.5 + 0.9/section, respectively)(Fig 2F, H). Interestingly, many of the motor neurons in the floor plate had bipolar morphology (Fig 2F, H).

These early neurons moving into the floor plate appeared to lack projections toward the motor exit points, which was tested directly below. An important observation is that the first ectopic motor neurons appear to pioneer an independent pathway into the floor plate, because the ectopic motor neurons enter the floor plate before any commissural axons (which begin to appear by E10.5) or any other βIII-tubulin⁺ neuron fibers. These findings suggest that, when Robos are missing, motor neurons migrate tangentially into the floor plate.

Robos are expressed by trochlear and spinal cord motor neurons

Since Robo receptors mediate the repulsive effect of Slits produced by the floor plate, we next attempted to verify previous reports of Robo expression by spinal cord and posterior hindbrain neurons (Bai et al., 2011; Brose et al., 1999; Hammond et al., 2005), using reporter inserts available for the Robo1 and Robo2 mutant alleles. Because the Robo1 mutant allele has a β-geo (β-gal/neo) insertion and Robo2 has a LacZ-tau insertion (Long et al., 2004), we labeled with anti-P-galactosidase on Robo1^+/− and Robo2^+/− heterozygote embryos. Both trochlear and spinal cord motor neurons were β-galactosidase⁺, showing Robo1 and Robo2 expression (Fig 3A–D). In addition, mis-positioned motor neurons in Robo1^+/−;2^−/− were β-galactosidase⁺, demonstrating that the migrating cells expressed Robo receptors (Fig 3B). As an additional way to more easily view the cell surface of motor neurons, we performed immunostaining of dissociated motor neurons with anti-Robo1 and anti-Robo2 antibodies, which clearly showed that Robo receptors were expressed on the cell membrane and internal membrane structures of motor neuron cell bodies (Fig 3E, H). Together, the expression pattern of Robo1 and Robo2 receptors on trochlear and spinal cord motor neurons is consistent with the genetic requirement for repulsive Slit/Robo signals to set their position.

Figure 3. Robo1 and 2 are expressed by trochlear and spinal cord motor neurons.

A, D. β-galactosidase and Islet-1 double labeling of cryosections of E10.5 Robo1^lacZ/+ (n=3 embryos, n=9 sections) and Robo2^lacZ/+ (n=3 embryos, n=9 sections) embryos, showing that trochlear and spinal cord motor neurons express Robo1 and Robo2 receptors. E–J. Anti-Robo1 and anti-Robo2 antibody labeling on dissociated cells (n=3 embryos) from ventral hindbrain show that Robo receptors are expressed by Islet-1⁺ cell bodies. Scale bars: A–D, 50 μm; E–J, 10 μm.

Motor neurons migrate into the floor plate in Robo2 mutants

Because both Robo1 and 2 were expressed by motor neurons, we wanted to test whether both Robos were involved in controlling migration. We previously found that a single wild type Robo2 allele is sufficient to prevent the appearance of neuronal cell bodies in the floor plate (Kim et al., 2011). To test whether a single wild type Robo2 allele could function to specifically prevent motor neuron migration, we generated mutants which carried a single wild type allele of either Robo1 or Robo2 in a homozygous mutant background for the other Robo gene, i.e. Robo1^+/−;2^−/− and Robo1^−/−;2^+/− (Fig 4; Table 1). On E9.5 and E10.5, Islet-1⁺ cells were seen in the floor plate of embryos with a single functional allele of Robo1, i.e. Robo1^+/−;2^−/− (Fig 4A, C). Motor neurons in embryos with a single functional allele of Robo2, i.e. Robo1^−/−;2^+/−, however, were restricted to their normal positions (Fig 4B, D). These observations suggest that Robo2 is the primary mediator of Slit repulsion and prevents abnormal migration of motor neurons into the floor plate (Fig 4G). However, the percentage of motor neurons in the floor plate is significantly less in Robo1^+/−;2^−/− mutants compared to Robo1^−/−;2^−/− mutants, suggesting that Robo1 gene function does have a minor role, when Robo2 is absent, in keeping motor neurons away from the floor plate.

Figure 4. Motor neurons migrate into the floor plate in Robo2 mutants.

Islet-1 labeling on cryosections of Robo1^+/−;2^−/− and Robo1^−/−;2^+/− embryos (n=4 embryos, n=12 sections for each genotype of E9.5, n=8 embryos, n=24 sections for each genotype of E10.5). A–D. Islet-1 labeling in single allele mutant embryos, Robo1^+/−;2^−/− and Robo1^−/−;2^+/−, showing both on E9.5 and E10.5 embryos, motor neurons entered the floor plate in Robo1^+/−;2^−/−. However, Robo1^−/−;2^+/− embryos did not have motor neurons in the floor plate (A–D in r1). E, F. Schematics of motor neurons in the neural tube of single allele mutants. G. Summary graphs showing the percentage of motor neurons in the floor plate of Robo1^+/−;2^−/− and Robo1^−/−;2^+/− E10.5 embryos (**, p<0.01; N.S., not significant). Scale bars: A, B, 25 μm; C, D, 25 μm.

Motor neurons enter the floor plate in Slit2 mutants

We next tested whether Slit mutants have the same phenotype as Robo mutants. Previous studies show that all three Slit genes are expressed in the floor plate of the hindbrain and the spinal cord, and also in motor neurons themselves, with Slit2 expression prominent in motor neurons (Brose et al., 1999; Farmer et al., 2008).

A subpopulation of Islet-1⁺ cells entered the floor plate of Slit1^−/−;2^−/−;3^−/− triple mutants (Fig 5B, F), similar to Robo1^−/−;2^−/− double mutants. We then tested which Slit was involved in the abnormal migration (Fig 5; quantification summarized in Table 2). Since we found that Slit1 was not required for positioning of motor neurons using Slit1^+/−;2^−/− and Slit1^−/−;2^+/− embryos (data not shown), we generated embryos with only single functional alleles of either Slit2 or Slit3 in a homozygous mutant background for Slit1. In Slit1^−/−;2^+/−;3^−/− embryos (Fig 5D, H), mis-positioned cell bodies were not seen in the floor plate, however, Slit1^−/−;2^−/−;3^+/− embryos had Islet-1⁺ cells in the floor plate (Fig 5C, G). Furthermore, in Slit1^−/−;2^+/−;3^+/− (Fig 5A, E), one Slit2 wild type allele restored proper motor neuron position (Fig 5I, Table 2). All of the lines tested were in Slit1 null backgrounds, which were viable and had wild type motor neuron positions. These findings suggest that Slit2 is the main floor plate repellent that maintains motor neurons in their normal position. We note that these global knockouts remove Slits from both floor plate and motor neurons, and so it is possible that motor-neuron specific conditional knockouts, which were not available for this study, might reveal a role for the motor neuron-expressed Slits. However, the similar ectopic motor neuron shifts in Slit1/2/3 triple mutants and Robo1/2 mutants are consistent with a critical function of floor plate Slit preventing Robo-expressing motor neurons from migrating into the floor plate.

Figure 5. Motor neurons enter the floor plate in Slit2 mutants.

Islet-1 labeling on cryosections of Slit mutant embryos (n=3 embryos, n=9 sections for each genotype of E10.5) A–H. Islet-1 labeling on cryosections of Slit1^−/−;2^−/−;3^−/−, Slit1^−/−;2^−/−;3^+/−, Slit1^−/−;2^+/−;3^+/−, and Slit1^−/−;2^+/−;3^−/− show that trochlear and spinal motor neurons entered into the floor plate of Slit1^−/−;2^−/−;3^−/− and Slit1^−/−;2^−/−;3^+/− (A–D in r1, E–F in spinal cord). However, Slit1^−/−;2^+/−;3^+/− and Slit1^−/−;2^+/−;3^−/− did not have motor neurons in the floor plate. Yellow arrows show cell bodies in the floor plate. I. Summary graphs showing the percentage of motor neurons in the floor plate of Slit1/2/3 E10.5 embryos (**, p<0.01; N.S., not significant). Scale bars: A–D, 50 μm; E–H, 50 μm.

Table 2.

Summary of ectopic MNs in the floor plate in Slit mutants

		Slit1−/−:2+/−;3+/−			Slit1−/−:2−/−;3−/−			Slit1−/−:2−/−;3+/−			Slit1−/−:2+/−;3−/−
		Total number of MNs	Number of MNs in FP	% of MNs in FP	Total number of MNs	Number of MNs in FP	% of MNs in FP	Total number of MNs	Number of MNs in FP	% of MNs in FP	Total number of MNs	Number of MNs in FP	% of MNs in FP
n=3	r1	119 + 4	0	0	118 + 5.3	14.7 + 0.9	12.5 + 2.3	114 + 4.2	8.7 + 0.9	7.6 + 1.5	120.3 + 4.6	0	0
	SC	160 + 15.3	0	0	153.3 + 23.2	1.3 + 0.2	0.9 + 0.7	170 + 16.1	0.83 + 0.4	0.5 + 0.4	163.3 + 16.9	0	0

Mis-positioned motor neurons project axons into the floor plate, instead of to exit points

Because a significant number of motor neurons with bipolar morphology were found in the floor plate when Robo receptors were missing (Fig 2F,G), we next asked whether motor neurons located in the floor plate were still able to project axons to their normal exit points.

First, the identity and projection pattern of these midline cell bodies was verified using the Isl1-GFP reporter in a Robo1/2 mutant background (Fig 6). Every embryo examined (n=3/3) showed that Isl1-GPF⁺ motor neurons projected axons longitudinally within the floor plate (Fig 6B–D). Every spinal cord section of Robo1^−/−;2^−/−::Islet-1^MN-GFP-F embryos (n=3/3) showed that motor axons crossed the midline and fasciculated motor axons are located in the floor plate (Fig 6F). However, no Isl1-GFP⁺ motor neuron cell bodies and axonal processes were found in the floor plate of wild type embryos (Fig 6A, E). Furthermore, an important observation is that trigeminal (nV) and branchial facial (nVII) motor neurons also migrated into the floor plate, and their Isl1-GFP+ motor axons also deviated into the floor plate and bundled in the midline with fewer axons projecting to their exit points (Fig 6B). Double labeling with the branchiomotor (bm) progenitor column marker, NKX2.2 and the bm marker, Phox2b, confirmed that bm neurons were generated from normal progenitor columns, then migrated into the floor plate in Robo mutants (Suppl. Fig1). These observations suggest that, in addition to positioning somatic motor neurons (sm, such as nIV), Slit/Robo signals may also position other cranial motor neuron classes, such as branchiomotor and visceral motor neurons, an issue that will be addressed in future experiments.

Figure 6. Axons from mis-positioned motor neurons project into the floor plate instead of to their exit points.

A–D. Open-book preparations of Robo1^+/+;2^+/+::Islet-1^MN-GFP-F and Robo1^−/−;2^−/−::Islet-1^MN-GFP-F embryos (n=3 embryos for each genotype of E10.5) show that mis-positioned motor neurons projected axons longitudinally within the hindbrain floor plate. E, F. Spinal cord sections of Robo1^+/+;2^+/+::Islet-1^MN-GFP-F and Robo1^−/−;2^−/−::Islet-1^MN-GFP-F embryos (n=3 embryos, n=9 tissues for each genotype of E10.5) show that motor axons cross the midline of Robo1^−/−;2^−/−::Islet-1^MN-GFP-F embryos (arrow in F). An arrowhead is showing fasciculated motor axons in the floor plate of Robo1^−/−;2^−/−::Islet-1^MN-GFP-F embryos. Dashed line in C marks the midline. Yellow arrows in D show cell bodies in the floor plate. r1, r2, r3 and r4, first, second, third, and forth rhombomeres; nIV, trochlear nucleus; nV, trigeminal nucleus; nVII, facial nucleus. Scale bars: A, B, D, E, 50 μm; C, F, 25 μm.

To trace motor neuron exit points, retrograde diI labeling was used (Fig 7). First, a diI crystal was placed into the dorsal midbrain-hindbrain boundary to target the projections of trochlear motor axons en route to their dorsal exit point in wild type and Robo1^+/−;2^−/− embryos (Fig 7A, D). This tracing strategy labeled only cell bodies located in a normal position in both genotypes, but failed to label any cell bodies in the floor plate. Based on the Isl1-GFP⁺ midline fibers, we then labeled from the midline in posterior r1 to trace motor axons back to their cell bodies, which were located in the floor plate only in Robo mutant embryos but not in wild type (Fig 7B, E). Every embryo examined (n=5/5) showed axons from the mis-positioned cell bodies projected within the floor plate instead of finding their normal exit point in Robo mutants (Fig 7E). However, there were no axon projections within the floor plate in wild type embryos (Fig 7B; summarized in Fig 7G, H). This observation is consistent with previous observations that a subset of motor axons grow into the floor plate in Slit or Robo mutants (Bai et al., 2011; Hammond et al., 2005). However, we also observed motor neuron cell bodies in the floor plate, likely because we examined a range of embryonic stages, particularly earlier stages. Thus, in the absence of Slit/Robo repulsion, a subset of motor neurons projected axons into and longitudinally within the floor plate, while their cell bodies followed their axons to end up in the floor plate on E10.5.

Figure 7. Axon tracing shows neuron cell bodies in the hindbrain floor plate.

A–F. Dil tracing for trochlear (nIV) axon trajectories on hindbrain whole mount preparations of E10.5 Robo1^+/+;2^+/+ (n=5) and Robo1^+/−;2^−/− (n=5) embryos. A, D. Retrograde diI did not label nIV neurons in the floor plate. B, E. Midline labeling on hindbrain whole mount preparations of control Robo1^+/+;2^+/+ (n=5) was not able to retrogradely label any neurons in the floor plate. However, midline labels of Robo1^+/−;2^−/− (n=5) embryos retrogradely labeled neuron cell bodies in the floor plate of Robo1^+/−;2^−/− embryos. G, H. Schematics of diI labeled trochlear axon trajectories, showing that mis-positioned neurons projected axons into the floor plate. Dashed lines in A, B, D, and E mark the midline. An arrowhead in E shows mis-positioned axon fibers projecting longitudinally within the floor plate, consistent with motor axon fibers in the midline of Robo1^−/−;2^−/−::Islet-1^MN-GFP-F in Fig 6. Yellow arrows in F show cell bodies in the floor plate. Scale bars: A, B, D, E, 100 μm; C, F, 25 μm. MLF: medial longitudinal fasciculus, CA: commissural axon, ML: midline, MB: midbrain.

Motor neurons shift dorsally in Netrin-1 or DCC mutant mice

The directed movement of motor neurons into the floor plate in the absence of repulsive Slit/Robo signals suggests that motor neurons are attracted to the floor plate. Since Netrin-1 is an attractive guidance cue primarily produced by the floor plate and motor neurons express high levels of the attractive receptor DCC (Keino-Masu et al., 1996), we tested whether attractive Netrin-1/DCC signals are required to maintain motor neurons in their normal position.

First, we verified that DCC was expressed on trochlear and spinal cord motor neurons on E10.5 by double labeling with Islet-1 antibody. DCC labeling was clearly seen on Islet-1⁺ trochlear and spinal cord motor neuron cell bodies and axons (Suppl. Fig 2), which agrees with mRNA expression in rat embryos (Barrett and Guthrie, 2001). To functionally test Netrin-1 and DCC, we examined embryos mutant for either Netrin-1 or DCC. Islet-1⁺ trochlear and spinal cord motor neurons shifted away from the floor plate when Netrin-1 or DCC was missing compared to their littermate controls, as assayed in both whole mounts and sections (Fig 8A–H). To quantify this dorsal shift, the distance between the lateral clusters of trochlear motor neurons, and the distance from the midline at the forelimb level in the spinal cord was measured, then normalized for variations in embryo size/stage (see Methods). The distance from the midline was significantly increased in Netrin-1 and DCC mutants compared to Netrin-1 heterozygotes and DCC heterozygotes, respectively (Fig 8I, J). The circumferential distance between the ventral midline and the lower edge of the motor column was also significantly increased in Netrin-1 (17.9 + 2.2%) and DCC (13.5 + 3.8%) mutants compared to their controls. We also found no significant difference between wild type and heterozygous mutants (not shown), suggesting that one Netrin-1 and DCC wild-type allele restored proper motor neuron position. In addition, about 5% of motor neurons in the spinal cord shifted dorsally at the upper edge of the motor column in both Netrin-1 and DCC mutants; motor neurons in this ectopic dorsal position were not seen in control embryos (Fig 8F, H). To investigate how motor neurons come to reside dorsally, we tested whether motor neuron progenitors shifted in Netrin-1 mutants by labeling with the motor progenitor column marker, Olig2. In Netrin-1 mutants on E10.5, Olig2⁺ progenitor cells in the r4 were located at their normal position, while the upper edge of the Islet-1⁺ motor column was shifted dorsally (13.3 + 3.3%) compared to their littermate controls (Fig 8L). The dorsal shifting was more obvious in E10 Netrin-1 mutants in which the upper edge of the Islet-1⁺ motor column was shifted dorsally (26.9 + 7.1%) compared to their littermate controls (Fig 8N). Interestingly, double labeling with βIII-tubulin and Islet-1on spinal cord sections of E10.5 showed that a ventral population of non-Islet1-expressing neurons, potentiallyV₃ neurons, located between the floor plate and the Islet-1⁺ motor neurons, which also moved away from the floor plate in Netrin-1 mutants (Suppl. Fig 3A). This raises the possibility that other type of neurons could use floor plate signals to find their normal position. Further studies are needed to reveal whether a conserved positioning mechanism exists amongst other ventral neurons. To test whether floor plate identity and morphology are maintained in Netrin-1 mutants, we labeled with the 4C7 antibody against the floor plate transcription factor HNF3b/FoxA2 on hindbrain and spinal cord sections (Suppl. Fig 4). Both by examining the tissue morphology, and the 4C7 pattern, floor plate morphology was not changed by the loss of Netrin-1 within the early embryonic hindbrain or spinal cord, so the shifts in neuron positions were not due to morphological tissue changes in Netrin1 mutants. Together, these findings suggest that ventral-ward attractive Netrin-1/DCC signals are required for setting the position of the motor nuclei, and for keeping motor neurons in their specific motor column in the neural tube.

Figure 8. Motor neurons shift dorsally in Netrin-1 or DCC mutants.

A–H. Analysis of motor neuron position in Netrin-1 and DCC mutants. Islet-1 labeling on whole mounts and cryosections show that trochlear (r1) and spinal cord motor neurons shifted away from the floor plate in Netrin-1 and DCC E10.5 mutants. For consistent comparisons, spinal cord sections from the same brachial (forelimb) level were compared between control and mutant embryos. White lines in A, B, C, and D mark the midline. Yellow lines show the distance between trochlear motor neurons (A–D) and the distance from the midline in the spinal cord (E–H). F, H. Some spinal cord motor neurons moved dorsally (arrows) outside of the motor columns in Netrin-1 and DCC mutants. I, J. Summary graphs show that trochlear and spinal cord motor neurons shifted dorsally significant distances in Netrin-1 (n=5 embryos, n=15 sections) and DCC (n=4 embryos, n=12 sections) mutants than controls (Netrin-1^+/− (n=6 embryos, n=18 sections), DCC^+/− (n=5 embryos, n=15 sections)). (*, p<0.05; **, p<0.01). K, L. Olig2 labeling of progenitors and Islet-1 labeling of motor neurons on r4 sections of E10.5 Netrin-1^+/− (n=3 embryos, n=9 sections) and Netrin-1^−/− (n=3 embryos, n=9 sections). The sections were taken from similar branchial sections, and aligned at the ventral midline (horizontal yellow line). Islet-1⁺ motor columns (dorsal and ventral borders marked with white lines) shifted dorsally from normal progenitor columns (arrows) in Netrin-1^−/− embryos. M, N. Olig2 labeling of progenitors and Islet-1 labeling of motor neurons on spinal cord sections of Netrin-1^+/− (n=3 embryos, n=9 sections) and Netrin-1^−/− (n=3 embryos, n=9 sections) E10 embryos. Islet-1⁺ motor columns (white lines) shifted dorsally from normal progenitor columns (green) in Netrin-1^−/− embryos. Scale bars: A–B, 100 μm; E–H, 50 μm; K, L, 50 μm; M, N, 50 μm. ML: midline.

Discussion

The formation of motor neurons as clusters in discrete ventral positions in the developing neural tube is a clearly defined case of a specific cell type induced at a specific position within a morphogen gradient (Ericson et al., 1997). Motor neurons form at a specific ventral-lateral position within the neuroepithelium because their progenitors are located close to the floor plate, and thus are exposed to a specific concentration of Shh morphogen, which sets off a cascade of transcription factors that ultimately specifies motor neuron identity for the cells at this position. However, our results imply that the clustering of motor neurons to form motor nuclei is not simply the outcome of where they are induced, but these neuron populations depend on floor-plate derived migratory cues to stay within their nucleus. This positioning mechanism acts as an important developmental fine-tuning, a backup positioning system to correct errors in motor neuron specification or migration.

Motor neurons have a widespread ability to migrate

Although motor neurons generally undergo only local shifts within their nuclei, a few types of motor neurons do migrate out of their motor nucleus. Brain stem motor neuron populations such as facial, oculomotor, and acoustic neurons all undergo substantial migrations, implying that these exceptional neuron populations may simply be released to express the general migratory ability of motor neurons. Other extreme examples include long migrations that place trochlear motor neurons into the cerebellum in lamprey (Fritzsch and Northcutt, 1993) or facial motor neurons into the spinal cord in mouse (Fritzsch, 1998). From an evolutionary perspective, we suggest that these examples of migratory cranial motor neurons have not gained a special migratory ability, but instead the majority of motor neurons have become trapped in relatively stationary positions. The specific examples of midline crossing of oculomotor and acoustic neurons likely involve regulated switching of response to midline attractants and repellents, a hypothesis to be tested in future experiments.

Motor neurons are also kept within the bounds of motor nuclei by boundary cap cells. These neural crest-derived cells sit at motor axon exit points, and express Semaphorin signals that prevent motor neuron cell bodies from translocating out of the neural tube (Bron et al., 2007; Mauti et al., 2007; Vermeren et al., 2003). The ectopic migration in these situans suggest that motor neurons also respond to boundary cap cues, which act as a selective barrier to motor neuron cell bodies while letting their axons pass through the pial surface to the periphery. These examples of migrating motor neurons emphasize that motor neurons have a general, but normally cryptic, migratory ability.

Cellular mechanisms involved in motor neuron migration

Our findings show that motor neurons are migratory early in their differentiation, as Isl1⁺ cells are readily seen near and within the floor plate in Robo mutants (Fig 1B, D). In contrast, their progenitors retain their position in normal Olig2⁺ columns (Fig 2D). This is consistent with a time course of mutant development in which motor neurons are born in their correct column, followed by anomalous ventral migration. Furthermore, the mis-positioned motor neurons appear to repolarize, adopt a bipolar shape with a leading process, and enter into the ventral midline (Fig 2E, F). These changes in cell shape and behavior suggest that Robo loss triggers changes in internal regulators of motor neuron migration. Our preferred mechanism is that Slit/Robo loss primarily causes motor axons to project into the floor plate, instead of to their exit points, and consequently their cell bodies simply follow the mis-guided axons. This would be similar to the ectopic emigration when boundary cap signals are disrupted. Alternatively, it is possible that Robos function independently in axon and cell bodies, acting as separate sites of Slit-mediated repulsion.

Our evidence suggests that Slit/Robo signals act directly on motor neurons to regulate their migration. Because ectopic motor neurons are observed at very early E9.5 stages of development that precede other neuron migration or potential substrates such as commissural axons, and because the motor neurons normally express both Robo1 and 2, our observations suggest that the abnormal migration phenotypes are a direct and cell autonomous requirement for Slit/Robo signals on the motor neurons. However, a strict test of the cell autonomous function of Robo in motor neurons was not possible for this study because conditional alleles were not available. It is also interesting to note that, as is typical for the axon or cell migration phenotypes in other mutants and axon systems, only some motor neurons undergo ectopic migrations. The migrating neurons could represent a random but substantial proportion that fail to be guided by other redundant cues, or the migrating neurons could be specified subpopulations that are selectively sensitive to Slit/Robo loss. Although markers are not yet available to distinguish such subpopulations, they appear to be widespread through the brain stem and spinal cord.

The consequences of abnormal motor neuron migration might include a partial depletion of motor pools, abnormal motor synapses in the CNS, or abnormal exits. However, our observations suggest that mis-positioned motor neurons disappear at later embryonic stages, possibly because they die, migrate away, or turn off motor neuron markers. Additionally, examining the functional consequences of mis-positioned motor neurons is not possible because the Robo or Slit knockout alleles are perinatal lethal as homozygotes. Future experiments using motor neuron-specific Robo knockouts could overcome the lethality of these mutants and allow tests for functional defects in specific motor systems.

Slit and Netrin signals control motor neuron migration and position

Many reports suggest that axon guidance molecules such as the Slits and Netrin-1 are also important for directing neuronal migration (Cariboni et al., 2012; Kawasaki et al., 2006; Yee et al., 1999). Indeed, Slit/Robo signals play roles in tangential migration in many systems. For instance, Slit/Robo repulsion guides neuroblasts to migrate a long distance from the subventricular zone to the olfactory bulb (Wu et al., 1999). In the mouse hindbrain, in vitro and in vivo studies show that migrating precerebellar neurons use Slit/Robo signals to find their proper position (Causeret et al., 2002; Causeret et al., 2004). These systems appear quite analogous to the floor plate guidance of the tangential migration of motor neurons. Slit/Robo signals appear to control neuronal migration by regulating cell polarity, mediated by local Ca²⁺ transients, redistribution of active RhoA, and/or centrosome positioning (Higginbotham et al., 2006; Xu et al., 2004). Motor neurons are clearly mis-oriented toward the floor plate in Robo mutants, consistent with a role for Slit signals in influencing neuron polarity.

The migration of motor neurons into the midline in Slit and Robo mutants, and the migration of motor cell bodies in explants exposed to ectopic Slit, is clearly consistent with a ventral repellent role for Slit/Robo signaling, and supports prior studies of Slit repulsion of motor axons (Bai et al, 2011; Colamarino and Tessier-Lavigne, 1995; Varela-Echavarría et al., 1997). It is possible that the failure to observe midline motor neurons in two previously reported analyses of Slit/Robo mutant mice (Bai et al., 2011; Hammond et al., 2005) was because those analyses were at relatively late stages, E1 1.5–12.5, days later than our earliest observations. The Hammond et al. analysis also relied primarily on retrograde labeling from motor exit points (Hammond et al., 2005), which would miss most mis-migrating motor neurons because their axons project into the floor plate instead of to their exit points. We suggest that midline motor neurons move in early, but disappear later, likely undergoing cell death because a failure to gain survival signals from the periphery or because of abnormal types or levels of floor plate signals. Our observations of midline motor neurons is, however, consistent with the results from mis-expression of dominant negative Robos in chick embryos (Hammond et al., 2005).

In contrast, the function of Netrin1 for the in vivo guidance of motor neurons and axons remains unclear because of conflicting evidence from various approaches. Several studies showed clear repulsion of cultured motor axons by Netrin1 (Colamarino and Tessier-Lavigne, 1995; Murray et al., 2010; Varela-Echavarría et al, 1997), suggesting that Netrin1 is a key component of floor plate repulsive signals. However, in another case, cultured spinal cord motor axons showed little effect of Netrin1, suggesting that Netrin1 responses are normally silenced (Bai et al., 2011). Together, these results suggest that Netrin1 has varied functions, in which specific regional subtypes of motor neurons differ in their axonal responses, at least in culture, depending on which sets of attractive and repellent Netrin receptors are expressed (Murray et al, 2010). On the other hand, we observe the opposite effect in vivo, with a clear dorsal shift of motor neuron cell bodies even in the case of trochlear motor neurons which have axons with long dorsal trajectories that have been proposed to be caused by repulsion by ventral Netrin1 (Colamarino and Tessier-Lavigne, 1995). Motor neuron cell body shifts could be the result of intrinsic differences in Netrin1 responses in cell bodies versus axons, such as Netrin1 attraction of cell bodies and repulsion of axons, although this potential cellular compartmentalization of motor neuron responses seems unlikely on the mechanistic level.

The mis-positioning phenotype in various regions of knockout embryos suggests that floor plate signals keep motor neurons in specific motor columns. In fact, Slit/Robo repulsion also regulates cell adhesion during migration (Ypsilanti et al., 2010), such as cooperation of N-cadherin, with Slit1/Robo2, in ganglion assembly (Shiau and Bronner-Fraser, 2009) and migration of precerebellar neurons (Taniguchi et al., 2006). Together, floor plate signals could regulate neuron adhesion within their motor column, such as shown for cadherins (Demireva et al., 2011; Price et al., 2002). We suggest that motor neuron adhesion increases rapidly during early development, so that the position of motor neuron columns becomes fixed and independent of floor plate signals. However, motor columns do undergo a secondary reorganization into sub-clusters consisting of motor pools, a process regulated by Reelin and cadherin signaling (Demireva et al, 2011; Palmesino et al, 2010; Price et al, 2002; Yip et al., 2000). This regulated shifting of cell bodies may also involve floor-plate derived cues to push or pull cells into position.

Motor neurons integrate opposing Slit and Netrin signals

The opposing shifts between Slit/Robo and Netrin-1/DCC mutants suggest a positioning mechanism in which actively migrating motor neurons are held in balance between these repellent and attractive signals (Fig 9). An interesting feature of this mechanism is that the balance relies on simultaneous responses to opposing Slit and Netrin-1 signals, likely along with other unknown positioning signals. However, this type of balance would not be consistent with the silencing mechanism for commissural axons, in which Slit/Robo signals silence Netrin/DCC attraction in cultured Xenopus axons (Stein and Tessier-Lavigne, 2001).

Figure 9. Summary: Floor plate-derived guidance cues are required to keep motor neurons in their normal position.

After progenitor cells (PC) shift out to the mantle zone (gray arrows) and begin to differentiate as motor neurons (MN), they are kept in motor columns by Slit/Robo repulsion (red) and Netrin-1/DCC attraction (green).

In spinal motor axons, prevention of floor plate attraction was recently proposed to rely both on the conventional Slit silencing mechanism and a novel activity of the Presenilin (PS1) secretase complex that cleaves DCC receptors to further interfere with attraction to Netrin (Bai et al., 2011). That study found that PS1 mutant motor axons have increased attraction to Netrin1 in vitro, and in vivo project into the floor plate in a Netrin1-dependent manner. These motor axon projections are similar to our observations in Slit or Robo mutants, which we interpret as an unbalanced attraction to the floor plate. Surprisingly, despite their increased Netrin attraction, PS1 mutant motor columns were observed to be disorganized and tend to shift away from the floor plate, and their study did not report motor neuron cell bodies within the floor plate in either PS1 or Robo1/2 mutants. One way to reconcile these conflicting results is that their study was limited to motor axons in the spinal cord. From our study of the earliest stages of motor neuron differentiation in the brain stem and spinal cord, the dorsal shift of motor neuron cell bodies in Netrin-1 or DCC mutants suggests that silencing is, at most, a weak in vivo factor in motor neurons because the position of the cell bodies is dependent on Netrin-1/DCC signaling even while they are exposed and respond to high levels of Slits. There is a growing realization that Netrin and Slit signals can interact in ways more diverse than silencing, including Slits acting through a co-receptor complex to switch on Netrin attraction for thalamocortical axons (Bielle et al., 2011; Leyva-Díaz et al., 2014), and the opposite effect of Netrin attenuation of Slit repulsion (Fothergill et al., 2013). An example that may be more similar to the mechanism that we have uncovered for motor neurons is the parallel independent Netrin and Slit guidance of fly commissural axons (Garbe and Bashaw, 2007). In another recent study, we have shown that simultaneous Netrin-1/DCC and Slit/Robo signals guide pioneer longitudinal axons through the mouse hindbrain, providing further evidence of neurons that balance Netrin1 attraction and Slit repulsion (Kim et al., 2014).

Overall, our results suggest that motor neurons are trapped in a static position by a balance of repulsive Slit/Robo and attractive Netrin/DCC signals. Further research into the mechanisms of how motor neurons integrate Slit repulsion and Netrin attraction will be needed to more fully understand the positioning mechanisms of motor neurons.

Highlights.

• Motor neurons migrate into the hindbrain and spinal cord floor plate in Robo and Slit mutants.

• Mis-positioned motor neurons project axons longitudinally within the floor plate.

• Motor neurons shifted away from the floor plate in Netrin1 and DCC mutants.

• Motor neuron position is set by Slit/Robo repulsion and Netrin/DCC attraction