Netrin signaling is important to guide migrating neurons and axons in many systems. Experiments with vertebrate CNS explants suggested netrin is bifunctional, attracting some axons and repelling others. Netrin1-expressing cells attracted spinal commissural axons and repelled trochlear cranial nerve axons in these experiments. Subsequent genetic studies demonstrated that multiple axon types, including those of the spinal commissural neurons, are attracted to netrin in vivo; however, an in vivo role for netrin signaling in trochlear nerve repulsion has not been observed. Here, we demonstrate that mice with a null mutation in the netrin receptor Unc5c on the inbred C57BL/6J (B6) genetic background have ventral/ipsilateral trochlear nerve misprojections. These misprojections are attenuated on a hybrid B6 × SJL background. In addition, B6.Unc5c−/− mice die as neonates of apparent respiratory distress and have incomplete phrenic nerve innervation of the diaphragm muscle. Neither the trochlear nerve misprojections nor the phrenic nerve phenotype was observed in B6 embryos lacking the netrin receptors DCC or Neogenin1, or the ligand netrin1, indicating these signaling molecules are dispensable for guidance of these axons. Like the trochlear nerve, the phrenic nerve phenotype is modified in a B6 × SJL hybrid background. To identify these modifier loci, we performed genome scans of the hybrid Unc5c−/− mice and found a major SJL-derived suppressor locus on Chromosome 17. Our results provide the first evidence that genes involved in netrin signaling are necessary for proper mammalian spinal motor axon development and trochlear axon guidance. In addition, they demonstrate the importance of modifier genes in vertebrate axonal guidance.
Motor neuron axons find their target muscle using a variety of molecular cues that result in either attraction toward their targets or repulsion away from nontarget cells (Schneider and Granato, 2003). Peripheral axon guidance has been studied extensively in invertebrates. In Caenorhabditis elegans, the ligand UNC-6 (netrins in vertebrates) is both a chemoattractant and chemorepellent. Cells expressing the receptor UNC-40 (DCC and neogenin in vertebrates) are attracted to UNC-6 (Chan et al., 1996), whereas cells expressing UNC-5 (Unc5A-D in vertebrates) are repelled (Hedgecock et al., 1987, 1990; Wadsworth and Hedgecock, 1996; Wadsworth et al., 1996). Studies on Drosophila netrin-signaling mutations demonstrate that specific motor axon populations misproject after exiting the CNS (Kolodziej et al., 1996; Mitchell et al., 1996; Labrador et al., 2005). Netrin signaling regulates cell migration and axon pathfinding of several cell types in vertebrates (Barallobre et al., 2005), but a role for netrin signaling has not been demonstrated in vivo in motor axons.
In vertebrates, cranial and spinal motor neurons lie in a ventral column, lateral to the netrin1-expressing floor plate. Spinal motor axons initially project ventrally away from the midline. Other axons, including several cranial nerves, extend dorsally, growing through the neuroepithelium before exiting the CNS near the midpoint of the alar plate. Explant studies indicate that netrin1 contributes to the guidance of motor axons; however, subpopulations of these axons differ in their netrin response in vitro. Netrin repels dorsally projecting cranial motor axons from the trochlear, trigeminal, facial, glossopharyngeal, and vagus nuclei (cranial nerves IV, V, VII, IX, and X, respectively), but no netrin response is elicited in ventrally projecting spinal motor axons (Colamarino and Tessier-Lavigne, 1995; Shirasaki et al., 1996; Varela-Echavarria et al., 1997). Additional signals have been identified that may repel dorsally projecting cranial motor axons from the ventral brainstem (Varela-Echavarria et al., 1997; Chen et al., 2000; Giger et al., 2000), possibly explaining the normal projection of these axons in netrin1 mutant mice.
Mice homozygous for the rostral cerebellar malformation (rcm) mutation, a null mutation in the netrin receptor Unc5c, have numerous CNS phenotypes, including cerebellar hypoplasia, granule and Purkinje cell mismigration, and defects in corticospinal tract (CST) pathfinding (Ackerman et al., 1997; Przyborski et al., 1998; Goldowitz et al., 2000; Finger et al., 2002). Here, we show that mutation of Unc5c also disrupts the development of specific motor axons. Furthermore, the genetic background of the Unc5c mutant mice modulates this phenotype. Unc5c−/− mice in a hybrid genetic background live a normal life span; however, mutant mice on a C57BL/6J (B6) inbred background die perinatally of apparent respiratory distress. A misprojection of axons in the trochlear nerve was observed in early embryonic ages, and a paucity of motor innervation of the diaphragm by the developing phrenic nerve was seen in later-stage embryos. Netrin1 and the UNC-40-related receptors DCC and neogenin are dispensible for guidance of the trochlear and phrenic nerves, as demonstrated by the lack of motor axon defects in mice lacking these components of the netrin-signaling pathway. In addition, the SJL/J genetic background has a major modifier locus that suppresses the phrenic phenotype of B6.Unc5c mutant embryos.
Materials and Methods
B6.cg-Unc5crcmTg(Ucp)1.23Kz/Slac mice (hereafter referred as B6.Unc5c) were generated by repeated backcrossing of Unc5c+/− animals to C57BL/6J (B6) for at least 28 generations. B6.129-Ntn1Gt(pGT1.8TM)629Wcs (B6.Ntn1), B6.129S2-Dcctm1Wbg (B6.Dcc), and B6.129-Neo1 (B6.Neo1) gene-trap mice were crossed to B6 for a minimum of 24, 20, and 5 generations, respectively. In some experiments, motor axons were visualized by crossing the enhanced yellow fluorescence protein (EYFP) motor-neuron-reporter line B6.cg-(Thy1-YFP)16Jrs/J (Feng et al., 2000) (backcrossed to B6 for >10 generations) to Unc5c+/− mice. Dcc−/− and Thy1-YFP mice were identified as reported previously (Feng et al., 2000; Finger et al., 2002). Ntn1−/− were genotyped as described previously in embryonic day 18.5 (E18.5) embryos (Finger et al., 2002) and by quantitative PCR for the Neo gene in DNA prepared from the yolk sac of E11.5 embryos. The insertion site of the B-geo/PLAP secretory gene-trap vector (Leighton et al., 2001; Mitchell et al., 2001) into intron 19 of the neogenin gene was identified by Southern and PCR analysis, allowing design of PCR primers for genotype analysis. The primer pair WtF (CTGTCTGGGAGAAAAAGATTGC) and WtR (GGGAGGGAAAAAGGGAGTG) was used to identify the wild-type allele, and MutF (GGGTCTCTTTGTCAGGGTCAC) and MutR (GCTCTTTTCAGTAGGCTCTCTGC) amplified the mutant allele.
Unc5c−/− embryos were generated by timed matings of B6.Unc5c+/− or F1 Unc5c+/− mice from a B6.Unc5c+/− × SJL/J cross. Mating was inferred from the presence of a vaginal plug (designated E0.5), at which time females were segregated from males until embryos were collected.
For whole-mount nerve and muscle immunostaining or EYFP visualization, tissue was dissected and fixed for 2 h in cold 2% paraformaldehyde/PBS. Staining was performed as described previously (Gautam et al., 1996). Nerves and synaptic vesicles were visualized using mouse anti-neurofilament antibodies (165 kDa, 1:20, clone 2H3; Developmental Studies Hybridoma Bank, Iowa City, IA) and mouse anti-SV2 monoclonal antibodies (1:20; Developmental Studies Hybridoma Bank), except in Figure 5, in which rabbit anti-neurofilament antibodies (200 kDa, 1:300; Sigma, St. Louis, MO) and rabbit anti-synaptophysin (1:300; Zymed, San Francisco, CA) were used. Staining was visualized with goat anti-mouse IgG1 (Molecular Probes, Eugene, OR) or Alexa 488-conjugated goat anti-rabbit IgG (see Fig. 5) secondary antibodies. Acetylcholine receptors (AchRs) were visualized by staining with tetramethyl-rhodamine-conjugated α-bungarotoxin (1:1000; Molecular Probes). Samples were examined on a Nikon (Tokyo, Japan) fluorescence microscope, and images were collected using a SPOT 9000 camera or a Leica (Nussloch, Germany) NT confocal microscope.
Peripherin staining was done on serial cryostat sections through the midbrain and hindbrain of E18.5 mice. Approximately 80 serial 15 μm sections were collected from each sample to ensure the inclusion of the trochlear nucleus. Staining was performed using standard immunofluorescence techniques, using anti-peripherin antibodies (Novus Biologicals, Littleton, CO) at a 1:400 dilution, followed by Cy3-conjugated donkey anti-rabbit antibodies.
Whole-mount embryo staining for neurofilament was performed as described previously (Kitsukawa et al., 1997). After signal visualization with DAB, embryos were dehydrated through a methanol series and placed into methyl salicylate to clear the tissue. Images were obtained on a Leitz/Wild (Wetzlar, Germany) stereomicroscope with a Leica FX300 camera.
In situ hybridization. Digoxygenin-labeled Dcc, Neo1, and Chat riboprobes were generated by reverse transcription-PCR (RT-PCR) from total brain cDNA using the following primer sets: ChatF (GACTTACCTAAGTTGCCAGTGC), ChatR (CCTTAGCTGGTCATTGGTGTC), Neo1F (AGCTATGAACCAGATGAGCTG), Neo1R (CTCCCTTGGCTTCTGATATGTG), Dcc-F (ACCAAGGTCGTCATGGAGATG), and Dcc-R (TGCTGCCAGGGACCGTTCC). To facilitate riboprobe production, a T3 promoter 5′ overhang (GCAATTAACCCTCACTAAAGGG) was added to each forward (5′) primer, and a T7 promoter 5′ overhang (GTATATCGACTCACTATAGGG) was added to each reverse (3′) primer. Antisense and sense probes were generated by in vitro transcription with T7 or T3 polymerase, respectively, directly from the PCR product. Unc5c riboprobes were generated as described previously (Ackerman et al., 1997). Serial sections through the cervical spinal cord of fresh-frozen E14.5 embryos were collected using a cryostat, and in situ hybridization was performed as described previously (Burgess et al., 2004). After hybridization, sections were incubated with peroxidase-conjugated anti-digoxygenin antibodies (1:2000; Roche Products, Welwyn Garden City, UK). Signals were amplified with the Tyramide Signal Amplification Plus kit (TSA-Plus; PerkinElmer, Wellesley, MA) as per the manufacturer’s protocol, using a 1:50 dilution of tyramide substrate in reaction buffer. Signals were developed for 1 h. Parallel control experiments using sense-strand probes did not reveal any signals above background.
To determine axon numbers, left and right phrenic nerves were isolated from E18.5 or P0 mice in 2% glutaraldehyde and 2% PFA in 0.1 m cacodylate buffer. Nerves were postfixed for 4 h in the same fixative before processing for plastic embedding and transmission electron microscopy (TEM). Sections (0.5 μm) were also stained with toluidine blue and examined by light microscopy. TEM images were collected at 5000× magnification on a JEOL (Peabody, MA) 1230 microscope.
DNA for genome scan analysis was prepared from tail or liver samples using a PureGene DNA isolation kit (Gentra Systems, Minneapolis, MN). PCRs using fluorochrome-labeled microsatellite simple sequence length polymorphism markers were run using an ABI 3700, and the resulting data were analyzed using Genescan and Genotyper software (Applied Biosystems, Piscataway, NJ).
Quantitative trait loci (QTL) main effects were determined as described previously (Sen and Churchill, 2001) using the R/qtl version 0.98-57 program (Broman et al., 2003) (http://www.biostat.jhsph.edu/∼kbroman/software/). Multiple imputations are used in this method to generate genotypes on a regular grid of genome-wide locations conditional on the observed marker data. The logarithm of the odds (LOD) score for each marker location was calculated by averaging the LOD scores over imputations. A whole-genome scan was performed at 2 centimorgan (cM) spacing using 128 imputations for the phenotype considering single QTL for each chromosome. One thousand permutations of the phenotype values were used to determine 5% and 1% threshold values (Churchill and Doerge, 1994), which correspond the red and green lines in the mainscan plot.
The Unc5c mutation is lethal on a C57BL/6 background
The Unc5c gene was identified by molecular characterization of mice with cerebellar defects attributable to insertion of an unrelated transgene into the Unc5c locus (Ackerman et al., 1997). Insertion of this transgene was accompanied by the deletion of Unc5c exons and complete loss of Unc5c transcripts in mice homozygous for the transgene. Despite their severe ataxia, mutant mice have a normal life span on the original C57BL/6J × SJL/J segregating background.
A dramatic decrease in the number of mutant animals surviving until weaning was found when this Unc5c mutant allele was transferred to a C57BL/6J (B6) inbred background. This increase in mortality was evident by the third backcross generation (N3), and by the N10 generation (in which >99% of unlinked loci should be derived from B6), no Unc5c−/− mice were observed at 3 weeks of age in the 68 mice examined (Table 1). In contrast, 11 of 59 mice (19%) were homozygous for the Unc5c mutation from an intercross of F1 B6/SJL heterozygotes. This increase in lethality of the Unc5c mutants on an inbred B6 background suggested that loci from the B6 genetic background are capable of modifying the mutant phenotype.
Trochlear nerve defects in B6.Unc5c−/− mice
To look for additional neurodevelopmental defects that may arise in the B6 background to shorten the life span, we performed whole-mount neurofilament immunohistochemistry on E11.5 embryos. Grossly, axon outgrowth and projections appeared normal, even in B6.Unc5c−/− mice (Fig. 1A,B). However, striking misprojections of the trochlear nerve were seen in mutant embryos (Fig. 1B).
In control mice, the trochlear nerve projects dorsally from the trochlear nucleus located in the caudal midbrain/rostral region of rhombomere 1 and elongates along the midbrain/hindbrain boundary (Fig. 2A–D). At the dorsal midline, it crosses to exit the brainstem before turning rostroventrally to innervate the superior oblique muscle of the contralateral eye. In B6.Unc5c−/− mice, the trochlear nerve initially projects ventrally rather than dorsally and exits the CNS ipsilaterally rather than decussating. The phenotype was variable from embryo to embryo, and not all axons of the trochlear nerve misproject (Table 2); however, every B6.Unc5c−/− mouse examined had clear abnormalities in at least one trochlear nerve (n = 6). The misprojection is most evident in cases such as Figure 2E–H, where the right side of the embryo has two trochlear nerves (Fig. 2E), one arising from the left trochlear nucleus and decussating as it should (Fig. 2F), whereas the nerve arising from the right trochlear nucleus exits ventrally on the ipsilateral side. As a result, there is no trochlear nerve on the left side of the embryo (Fig. 2G). In other cases, both the left and right nerve make almost entirely ventral projections (Fig. 2I–L), or nerves on both sides of the embryos each take both dorsal and ventral trajectories (Fig. 2M–P).
Because Unc5c is known to guide cell migration, we examined the trochlear nucleus in B6.Unc5c−/− mice to determine whether the misprojection of the trochlear axons was accompanied by mislocalization of the cell bodies. The trochlear nucleus lies near the midbrain/hindbrain junction behind the occulomotor nucleus and at the rostral border of the mesencephalic nucleus of cranial nerve V. The cell bodies of the trochlear nucleus are near the midline, just ventral to the cerebral aqueduct, and can be identified by staining with peripherin (Pattyn et al., 1997; Coppola et al., 2005). Analysis of E18.5 wild-type and Unc5c−/− embryos (n = 7 control, n = 3 mutant) revealed the nucleus to be in the correct location, with no evidence of ectopic cells (Fig. 3A,B), indicating that the misprojecting trochlear axons in mutant embryos are unlikely to be secondary to abnormal migration of trochlear neurons.
Consistent with our hypothesis that the increased phenotypic severity of the Unc5c−/− mice on the B6 inbred background affects neurodevelopment, hybrid B6 × SJL F2 mice homozygous for the Unc5c mutation showed trochlear misprojections in only two of five homozygotes examined (Fig. 3C–H). No misprojection of trochlear axons could be seen in the other three Unc5c−/− animals, suggesting that the SJL background has genes that rescue this guidance phenotype.
Other netrin signaling molecules are dispensible for trochlear guidance
The conticospinal tract misprojections in Unc5c−/− mice phenotypes partially overlap those of mice with mutations in other netrin signaling molecules, including Dcc and netrin1 (Finger et al., 2002). However, although netrin1 (Ntn1) has been shown to be a chemorepellent for trochlear axons in vitro, no trochlear phenotypes were observed in Ntn1−/− mice in vivo (Serafini et al., 1996). Because of the dependency of this phenotype on the B6 genetic background in Unc5c−/− mice, we examined guidance of the trochlear axons in B6.Ntn1−/− embryos (Fig. 4A,B) (n = 6). In vitro assays have suggested UNC5 receptors act as coreceptors for DCC, modulating its netrin response from attraction to repulsion. Thus, we also examined this projection in B6.Dcc−/− embryos and embryos defective for the other UNC40 family member, neogenin1 (B6.Neo1−/−) (Fig. 4C,D) (n = 5 and 4, respectively). No trochlear nerve defects were observed in mice with any of these mutations, although Neo1−/− mice lacked the ophthalmic branch of the developing trigeminal nerve (Fig. 4C). Although Ntn1, Dcc, and Neo1 may be involved in trochlear axon guidance in modulatory or redundant roles, our results show that none of these genes are solely essential for the process in vivo.
Phrenic axons defects in B6.Unc5c mice
Although misprojection of the trochlear nerve is unlikely to account for neonatal lethality, these results suggested that other neurodevelopmental phenotypes may arise in Unc5c−/− embryos on the B6 background. We therefore looked for additional phenotypes in older embryos that may explain the shortened life span of Unc5c mutants.
We noted that some neonates in litters from the B6.Unc5c+/− parents appeared to be gasping rather than breathing efficiently at birth, and these mice became cyanotic and died within a few hours of birth. Genotype analysis indicated these mice were homozygous for the Unc5c mutation. Histological analysis revealed the gastrointestinal tracts were distended, consistent with air gasping; however, the gross morphology of all other organs, including the palate and lungs, appeared normal (data not shown).
Neonatal respiratory distress and subsequent lethality can be associated with defects in innervation of the diaphragm and other respiratory muscles (Sanes and Lichtman, 2001; Burgess et al., 2004). To determine whether such defects occur in B6.Unc5c mutant mice, diaphragm innervation was analyzed in E18.5 mutant and littermate control embryos. To visualize the phrenic nerves, diaphragms were stained with antibodies against neurofilament and either SV2 or synaptophysin, and postsynaptic AChRs on the muscle fibers were identified by rhodamine-conjugated α-bungarotoxin.
The phrenic nerves arise bilaterally from motor neurons in the cervical spinal cord (C3–C6) (see supplemental Fig. S1 for schematic, available at www.jneurosci.org as supplemental material). Motor axons originating from this region of the spinal cord enter the brachial plexus with most axons projecting laterally to innervate the forelimbs. However, axons that will become the phrenic nerve continue caudally through the thoracic cavity eventually contacting the diaphragm near the dorsoventral midpoint. At this initial contact point, the phrenic nerve branches into three primary trunks, with two branches extending across the muscle to innervate the ventral and dorsal diaphragm. Another axon bundle extends dorsally and medially to innervate the crus of the diaphragm. Short terminal branches from each nerve trunk synapse with the muscle forming the motor endplate band. Although the innervation pattern is not precisely identical, this process is very similar on both the right and left sides of the body.
In each of the B6.Unc5c−/− embryos examined (n > 25), the motor innervation pattern was severely abnormal with typically 30–75% of the muscle uninnervated (Fig. 5A,B). Regions of the diaphragm regions lacking innervation were never observed in Unc5c+/− or Unc5c+/+ littermates. Although all B6.Unc5c−/− embryos had large regions of the unninervated diaphragm, the exact presentation was variable and ranged from a complete lack of innervation (as visualized from the lack of both nerve and endplate staining) on either the left or the right side of the diaphragm, to a partial lack of innervation on both sides of the muscle. As in the trochlear nerve, whether the left side, the right side, or both sides were affected appeared stochastic. Clustering of synaptic proteins occurred on the innervated diaphragm, suggesting axons that reach the muscle form synaptic contacts. This is consistent with the apparently normal movements of mutant neonatal mice, which suggest many neuromuscular synapses are functional. However, the reduced number of axons in the mutant phrenic nerve resulted in a compensatory denser pattern of innervation than in control mice (Fig. 5C,D). In contrast to the diaphragm, motor innervation of the intercostal muscles of the ribcage, also involved in respiration, appeared normal in Unc5c−/− embryos (supplemental Fig. S2, available at www.jneurosci.org as supplemental material). Interestingly, sensory neuron innervation around the lateral edge of the diaphragm also appears relatively normal, although the sensory projections extend slightly more medially in muscle regions lacking motor innervation (Fig. 5A,B, asterisks).
Proximal to its point of contact with the muscle, the diameter of the phrenic nerve was markedly reduced, suggesting it contained fewer axons than the wild-type nerve. This was evident by immunofluorescence analysis with anti-neurofilament antibody (Fig. 6A–C′), and the decrease in nerve caliber qualitatively correlated with the severity of diaphragm innervation defects. Axon counts in the mutant animals were significantly reduced compared with wild-type and heterozygous controls (p < 0.05; two-tailed t test), as shown by transmission electron microscopic analysis of phrenic nerves (Fig. 6D–F, Table 3). The bilateral asymmetry of nerve defects in B6.Unc5c−/− mice was reflected in the variable axon numbers from left and right nerves, as shown in Table 3.
The state of postsynaptic differentiation and AChR clustering in the uninnervated portions of the E18.5 diaphragm are consistent with muscle that has never been contacted by a motor nerve (Lin et al., 2001; Yang et al., 2001). Furthermore, histological examination of the uninnervated muscle demonstrated normal muscle layering and thickness, suggesting there is no defect in myogenesis (data not shown). Finally, the aneural clustering of AChRs suggests the muscle is fully differentiated, because these clusters depend on expression of MuSK, a gene with expression that is induced in differentiated myotubes (Valenzuela et al., 1995). Thus, phrenic axonal-guidance phenotypes do not appear secondary to defects in muscle differentiation.
Diaphragm innervation defects in B6.Unc5c mutant mice could be attributable to either motor axon-pathfinding defects during early motor development or defects in the stabilization of neuromuscular synapses that could lead to secondary loss of phrenic nerve/muscle contact, as observed in mice defective for Erbb2 (Morris et al., 1999; Lin et al., 2000). To differentiate between these possibilities, diaphragms were isolated at E13.5, shortly after the phrenic nerve normally contacts the diaphragm, and axons were visualized by direct fluorescence of transgenic expression of EYFP in motor neurons (Fig. 7). Abnormalities in diaphragm innervation, similar to those observed in E18.5 Unc5c−/− embryos, were also observed in mutant embryos at E13.5. Thus, it is likely that phrenic axons fail to reach the diaphragm, rather than retracting after contact with the muscle.
In the trochlear nerve, we determined that the axons were misprojecting ventrally and ipsilaterally from the CNS. To determine where motor axons projections are disrupted in B6.Unc5c mutant embryos, we used the B6.Unc5c−/−; Thy1-YFP16 mice to visualize motor neuron cell bodies and axons at early stages of development. Fluorescence stereomicroscope visualization of dissected E13.5 embryos did not reveal straying axons between the brachial plexus and the diaphragm in Unc5c mutant embryos, indicating that there is not a gross defect in the fasciculation of the nerve distal to the brachial plexus (data not shown). Nor were obviously misprojecting axons seen in the ventral horn of the cervical spinal cord at E13. Similarly, the ventral roots exiting the cervical spinal cord did not show evidence of defasciculation (supplemental Fig. S3, available at www.jneurosci.org as supplemental material). Thus, the trajectory(s) that this subset of motor axons takes in Unc5c−/− embryos is unclear. Based on our observations of the trochlear nerve, we would predict they misproject from the CNS, possibly fasciculating with other motor axons unrelated to the phrenic nerve.
Other netrin signaling components do not affect phrenic nerve development
Although disruption of netrin 1 and its receptors, DCC and neogenin, cause perinatal death (Serafini et al., 1996; Fazeli et al., 1997; Srinivasan et al., 2003), a role in the guidance of motor axons has not been reported previously for these molecules, and no effects of these mutations on trochlear axon guidance were observed. Similarly, these genes did not alter the guidance of the trochlear nerve, even in B6 congenic animals. To determine whether other netrin receptors may also be involved in phrenic motor axon guidance, we examined the expression of Unc5c, Dcc, and Neo1 in the embryonic cervical spinal cord by in situ hybridization. Although these genes have distinct expression patterns in the developing spinal cord, Unc5c, Dcc, and Neo1 transcripts were all detected in Chat-positive motor neurons of the ventral horn of the spinal cord (supplemental Fig. S4, available at www.jneurosci.org as supplemental material).
The expression patterns of Dcc and Neo1 were consistent with a potential role in phrenic motor axon guidance. The phrenic nerve was of normal caliber, and the diaphragm was fully innervated, with no unusual routing of the axons in B6.Dcc−/− or B6.Neo1−/− mutant embryos (n = 8 and n = 5, respectively) (Fig. 8B,C). In addition, we also examined the phenotype of mice deficient for the Ntn1 gene (Fig. 8D). Like Dcc and Neo1 mutant embryos, B6.Ntn1−/− embryos did not show an altered pattern of motor innervation in the diaphragm (n = 9) (Fig. 8D), demonstrating that, as in guidance of trochlear motor axons, these genes are not essential for phrenic nerve innervation of the diaphragm.
SJL mice carry a suppressor of the diaphragm phenotype on Chromosome 17
The complete penetrance of trochlear misprojections and diaphragm innervation defects in Unc5c mutant embryos on the B6 background, and attenuation of these defects in the original mixed B6/SJL hybrid background, suggested the presence of strain-specific loci that modify netrin signaling during axon outgrowth. Because the trochlear nerve is best examined in E11–E13 mice and the phrenic nerve is best examined during late embryogenesis, it is difficult to characterize both phenotypes in a single embryo. Thus, we chose to examine the diaphragm innervation phenotype for a modifier screen, because it is more likely to contribute to the lethality of the mice.
To examine the effect of genetic background on phrenic nerve development, we analyzed diaphragms from F2 B6;SJL-Unc5c−/− embryos. Mice heterozygous for the Unc5c mutation from a B6 congenic colony were mated to SJL mice to create F1 Unc5c+/− mice. E18.5 embryos were obtained from timed matings of F1 heterozygous mice, of which 23% (50 of 217) were homozygous for the Unc5c mutation. Three classes of diaphragm phenotypes were noted among Unc5c−/− embryos. Twelve of the 50 Unc5c−/− embryos (24%) had diaphragm defects as severe as those seen in B6.Unc5c−/− embryos, with large regions (30–75%) of the uninnervated diaphragm and a marked decrease in the caliber of the phrenic nerve (Fig. 9A). Conversely, 23 (46%) of the mutant embryos had no discernable defects in diaphragm innervation and were indistinguishable from controls (Fig. 9B). Diaphragms from the remaining 15 mice (30%) had an intermediate phenotype with small (10–20%) regions of uninnervated muscle (Fig. 9C). These results suggest that introgression of the B6 background genetically predisposes Unc5c−/− embryos to defects in phrenic nerve development and further suggest that SJL-derived loci can suppress these defects in mutant embryos.
To identify the chromosomal location of modifier loci, a genome scan was performed on DNA from the 50 F2 Unc5c−/− embryos (Fig. 9D). Genomic DNA from individual embryos was analyzed with 70 microsatellite markers polymorphic between B6 and SJL and distributed at ∼15–20 cM intervals across the genome. Genome-wide significance thresholds were determined by permutation testing. Significant linkage (LOD, >5.77; p < 0.01) was observed on proximal Chromosome 17 at D17Mit16. LOD scores for markers on other chromosomes were not significant (p > 0.05). Analysis of Unc5c−/− embryos with additional Chromosome 17 microsatellite markers (Fig. 9E) revealed a peak LOD score at D17Mit16 (7.7 cM) with lower scores at flanking markers.
Examination of individual embryos revealed that embryos with severely disrupted diaphragm innervation most often had the B6/B6 genotype at the D17Mit16 locus (75%; 9 of 12), whereas the genotype of unaffected embryos was skewed toward SJL/SJL or SJL/B6 (96%; 22 of 23). These results significantly deviate from the expected Mendelian genotype ratio of 1:2:1 (χ2 = 15.25; p > 0.004) and are consistent with our hypothesis that a dominant, SJL-derived locus on Chromosome 17 mitigates the B6.Unc5c phrenic nerve phenotype. D17Mit16 genotypes of the 13 intermediate embryos appeared more random (5 B6/B6, 6 B6/SJL, 2 SJL/SJL), suggesting that SJL-mediated suppression of this phenotype is not complete and may involve complex interactions with additional modifier loci on other chromosomes.
We demonstrate that Unc5c mutant embryos on the B6 background have fully penetrant abnormalities in both cranial and spinal motor axon development, defects not previously associated with disruption of vertebrate netrin signaling. In particular, UNC5C is essential for proper projection of the trochlear nerve and proper diaphragm innervation by the phrenic nerve. Unc5c is clearly necessary for trochlear nerve guidance, as demonstrated by misprojection of axons that results in an aberrant CNS exit point. However, the exact nature of the phrenic nerve defects is less clear. Before the contact with the diaphragm, a striking reduction in the number of phrenic axons is observed. We did not observe defects in Unc5c−/− motor axons as they leave the ventral horn of the cervical spinal cord or defasciulation of the ventral roots, suggesting that, instead, mutant phrenic axons may defasiculate from the motor axon bundle distal to the ventral root, but before the brachial plexus. Alternatively, phrenic axons in mutant embryos may fail to navigate choice points associated with the turn at the brachial plexus, where most axons extend laterally to innervate the limb and only a small portion continue caudally through the thoracic cavity eventually contacting the diaphragm. Finally, it is possible that the phrenic nerve defect is attributable to abnormalities in differentiation or migration of phrenic motor neurons. However, the small number of axons that comprise the phrenic nerve, the lack of specific markers for these axons and/or the motor neurons from which they arise, and the stochastic laterality of the motor axon defects make it difficult to precisely determine the exact location of these defects in Unc5c mutant embryos.
UNC5 receptors can serve as the netrin coreceptor with UNC40 receptors when expressed ectopically, mediating long-range repulsion of axons from netrin sources (Hong et al., 1999; Keleman and Dickson, 2001). Although the Neo1 mutation appears to specifically disrupt the ophthalmic branch of the trigeminal nerve, a sensory projection that innervates the skin and deeper tissues including the dura mater, we demonstrate that the trochlear and phrenic nerves are normal in embryos homozygous for mutations in the vertebrate homologs of the unc-40 family, even when these mutations are placed on the B6 background. Both Dcc and Neo1 are expressed in the developing spinal motor column and on motor axons (supplemental Fig. S4, available at www.jneurosci.org as supplemental material) (Keino Masu et al., 1996). However, determination of whether phrenic motor neurons coexpress the UNC-5- and UNC-40-related receptors awaits the development of markers for this subset of motor neurons. Coexpression of DCC and Neo1 would suggest that these family members might be functionally redundant in this pathway. The results of this analysis will be particularly interesting, given recent findings that suggest Drosophila motor axon guidance phenotypes that are dependent on Unc-5 but not the DCC homolog frazzled, or vice versa, may be attributable to transcriptional regulation of different receptor combinations (Labrador et al., 2005).
The ventral rather than dorsal projection of the trochlear nerve in Unc5c−/− embryos is consistent with in vitro data indicating that the outgrowth of these axons is repelled by netrin1 (Colamarino and Tessier-Lavigne, 1995; Shirasaki et al., 1996; Varela-Echavarria et al., 1997). In wild-type embryos, these axons project dorsally and extend toward and across the midbrain–hindbrain boundary, crossing the midline and finally exiting the CNS dorsally to innervate the contralateral superior oblique muscle of the eye. Although defective trochlear nerves were observed in every mutant embryo examined, these projections are variable. In some cases, the nerve was missing. In other embryos, nerves entirely projected in the rostroventral direction, whereas in the remaining embryos, nerves have two apparently independent projections, one defective and one normal. These bifurcating projections may reflect independent guidance mechanisms for the axons arising from the midbrain trochlear neurons and those axons with cell bodies that are located in rhombomere 1 (Irving et al., 2002). Alternatively, these projections could be attributable to defasculation of later-migrating axons from the pioneer axons. In any case, axons with defective trajectories still reach the eye, suggesting that these axons retain intrinsic target recognition capability.
Although the initial dorsal migration of trochlear axons is thought to rely on chemorepellents, we cannot rule out the possibility that Unc5c−/− trochlear axons fail to respond to an attractant expressed by the isthmic organizer of the midbrain/hindbrain, a region shown to attract trochlear axons in vitro (Irving et al., 2002). The abnormal migration of cerebellar granule cells into this region in Unc5c−/− mice suggests that the UNC5C receptor responds to cues expressed in this region (Ackerman et al., 1997; Przyborski et al., 1998). A role for UNC5C in axon attraction is consistent with our findings that a major subset of CST axons in the Unc5c−/− mouse fail to cross the midline (Finger et al., 2002). Similarly, nonallelic noncomplementation was observed between the Unc5c and Dcc null mutations, suggesting that these molecules may be partially redundant in mediating either attraction or repulsion of CST guidance at the midline.
Whether the netrins are necessary for UNC5C-mediated guidance of the trochlear and phrenic nerve is unclear. We, and others (Colamarino and Tessier-Lavigne, 1995), have demonstrated that netrin1 is dispensable for the guidance of trochlear (and phrenic) motor axons. This result is in contrast to the phenotypes of C. elegans unc-5 mutants that consistently overlap with phenotypes of the mutations in the netrin-encoding unc-6 gene (Hedgecock et al., 1990). Like the trochlear defect, cerebellar neuronal migration defects (Bloch-Gallego et al., 1999; J. H. Finger and S. L. Ackerman, unpublished observation) and some CST abnormalities (Finger et al., 2002) observed in Unc5c−/− mice are also netrin1 independent. Furthermore, the lack of phrenic axon defects in Ntn1−/− embryos is consistent with results from in vitro experiments analyzing axonal outgrowth in rat embryo tissue explants cocultured with cells expressing netrin1 (Varela-Echavarria et al., 1997).
Although netrin1 is the best characterized netrin family member in mammals, other netrin family members exist, including the closely related netrin3 (Puschel, 1999; Wang et al., 1999) and netrin4 (β-netrin) (Koch et al., 2000; Yin et al., 2000). Formally, any of these ligands could be sufficient or redundant with netrin1 in signaling to UNC5C in migrating motor axons, but netrin3, which has both axonal outgrowth and repellent activities, has a high affinity to the vertebrate UNC receptors (Wang et al., 1999). Netrin3 is expressed in both motor and sensory neurons and their axons in the embryonic mouse spinal cord and by developing myoblasts in the embryo, suggesting it may play a role in guiding peripheral motor axons to their muscle targets (Wang et al., 1999; Seaman and Cooper, 2001).
Interestingly, the netrin3 gene is located on proximal Chromosome 17 (Wang et al., 1999) near the genomic region we have identified as having a major influence on Unc5c-mediated phrenic nerve guidance, making netrin3 a candidate for the modulator of the Unc5c phrenic axon phenotype. Quantitative RT-PCR failed to reveal differences in mRNA expression levels from the spinal cord and hindquarter muscle in C57BL/6J and SJL embryos, but two nucleotide changes predicted to alter amino acids were found: leucine 23 is changed to proline, and valine 215 is changed to methionine in SJL, but whether these changes affect function of this ligand is not known.
Although netrin3 is an interesting candidate for a modifier of this phenotype, it is important to remember that the rcm mutation is a null allele for Unc5c. Thus, changes in receptor signaling efficacy, such as that associated with the clr-1 protein tyrosine phosphatase in C. elegans, are not mechanistically possible (Chang et al., 2004). Therefore, we anticipate that the modifier locus we have identified will be a component of a partially redundant or parallel signaling pathway, which differs in function or expression between B6 and SJL. Although netrin3, or any other modifier gene, could be influencing the signaling of other netrin receptors, it is also possible that components of another signaling pathway, such as the neuropilin–semaphorin III/D pathway, which has been shown to act as a chemorepulsive signal for this nerve (Chen et al., 2000; Giger et al., 2000; Sahay et al., 2003; Watanabe et al., 2004), are disrupted in B6 mice.
Finally, although the trochlear nerve defects in Unc5c mutant embryos can be also rescued by the SJL genetic background, we do not know whether the same modifier locus influences the trochlear nerve. The generation of B6 congenic mice carrying the identified portion of Chromosome 17 from SJL will allow analysis of the contribution of this region to the suppression of both the trochlear and phrenic axon defects in Unc5c−/− embryos.
Our results demonstrate that genetic background has a profound impact on neonatal survival and motor axon guidance in Unc5c mutant mice. Genetic screens of mutagenized C. elegans have identified several loci that suppress unc5-dependent dorsal migrations in touch neurons ectopically expressing UNC5 (Colavita and Culotti, 1998). The strain dependence of the Unc5c phrenic nerve phenotype indicates that modifier loci also exist in vertebrates, with a major dominant locus influencing phrenic motor axon guidance on proximal Chromosome 17. The identification of such modifier loci will be a powerful way to isolate genes that modulate or interact with UNC5C/netrin signaling in motor axons.
This work was supported by grants from the Edward Mallinckrodt Jr. Foundation (R.W.B.) and the National Institutes of Health (NS35900 to S.L.A.). S.L.A. is an investigator of the Howard Hughes Medical Institute. We thank the scientific services at The Jackson Laboratory, in particular Doug McMinimy in the Allele Typing Service; Pete Finger, Greg Martin, and Dr. Jim Denegre in the Bioimaging Service; and Weidong Zhang in the Computational Biology Service. We thank Dr. Steven Rauch for assistance in quantitative PCR, Dr. Tom Gridley for comments on this manuscript, and Dr. Wayne Frankel for statistical assistance.
- Correspondence should be addressed to Dr. Susan L. Ackerman, The Jackson Laboratory, 600 Main Street, Bar Harbor, ME 04609. Email: