Abstract
Cranial motor and sensory nerves arise stereotypically in the embryonic hindbrain, act as sensitive indicators of general and region-specific neuronal development, and are directly or indirectly affected in many human disorders, particularly craniofacial syndromes. The molecular genetic hierarchies that regulate cranial nerve development are mostly unknown. Here, we describe the first mouse genetic screen that has used direct immunohistochemical visualization methods to systematically identify genetic loci required for cranial nerve development. After screening 40 pedigrees, we recovered seven new neurodevelopmental mutations. Two mutations model human genetic syndromes. Mutation 7-1 causes facial nerve anomalies and a reduced lower jaw, and is located in a region of conserved synteny with an interval associated with the micrognathia and mental retardation of human cri-du-chat syndrome. Mutation 22-1 is in the Pax3 gene and, thus, models human Waardenburg syndrome. Three mutations cause global axon guidance deficits: one interferes with initial motor axon extension from the neural tube, another causes overall axon defasciculation, and the third affects general choice point selection. Another two mutations affect the oculomotor nerve specifically. Oculomotor nerve development, which is disrupted by six mutations, appears particularly sensitive to genetic perturbations. Phenotypic comparisons of these mutants identifies a “transition zone” that oculomotor axons enter after initial outgrowth and in which new factors govern additional progress. The number of interesting neurodevelopmental mutants revealed by this small-scale screen underscores the promise of similar focused genetic screens to contribute significantly to our understanding of cranial nerve development and human craniofacial syndromes.
- cranial nerve
- mutations
- mice
- axon guidance
- brain development
- genetics
- sensory neurons
- neural tube
- neural crest
- neural plate
- nerve
- motor neuron
- development
Introduction
The nerves that innervate the vertebrate head, throat, and heart arise stereotypically in the embryonic hindbrain and are affected by many human craniofacial syndromes (Cordes, 2001; Wilkie and Morriss-Kay, 2001). During early embryogenesis, each cranial sensory and motor nerve can be easily recognized by its origin, pathway, and target. Thus, examination of hindbrain neurodevelopment provides a path toward understanding both general and region-specific neurobiologic mechanisms. The embryonic hindbrain is transiently subdivided into seven to eight segments, the rhombomeres (r1–r8), along its anterior–posterior axis (Lumsden and Keynes, 1989). Cranial motor nerves arise within specific rhombomeres (as shown in supplemental Fig. 1, available at www.jneurosci.org as supplemental material) and can be subdivided into two general classes: the ventral motor nerves (vMNs) and the dorsal motor nerves (dMNs). Axons of vMNs have ventral exit points just like somatic motor neurons of the spinal cord. All other cranial motor neurons belong to the dMN class and project axons dorsally out of the neural tube. Sensory components of cranial ganglia arise from placodes or from neural crest cells. The distinctive appearance of cranial sensory and motor nerve processes provides a unique tool for studying the mechanisms that control vertebrate neurodevelopment and axon guidance.
Analyses of mouse mutants have begun to identify molecules required for specification and pathfinding of cranial nerves (Cordes, 2001). These insights have been gained either by fortuitous observations of cranial nerve anomalies in knock-out or classical mouse mutants or by deliberate pairing of gene expression analyses with reverse genetics. In the latter approach, molecules expressed in neuronal subsets are identified and the consequences of their loss and gain of function analyzed in the mouse and chick (Tsuchida et al., 1994; Lin et al., 1998; Briscoe et al., 2000; Liu et al., 2001). Pairing expression-based analyses directly with embryonic stem cell-based “genetrap” approaches has expedited progress further (Leighton et al., 2001; Mitchell et al., 2001). Most often, reverse genetic approaches have focused on molecules with previously known neurobiological roles in invertebrates. To identify molecules with previously unknown neurobiologic roles or less incriminating expression patterns, a forward genetic screen that systematically recovers mutations affecting cranial nerve development in the mouse should prove invaluable.
The potent mutagen N-ethyl-N-nitrosourea (ENU) and availability of the mouse genome sequence have made forward genetic screens and identification of the affected gene(s) highly feasible in the mouse (Takahashi et al., 1994; Cordes, 2005). Once treated with ENU, a single male mouse can produce 100–150 progeny (G1, “first generation”), each of which represents one mutagenized gamete. On average, the high efficiency of ENU mutagenesis results in a new mutation in any single locus in one of 500–1000 G1 animals. To screen for recessive mutations, each G1 male is used to create a three-generation pedigree, in which he is mated with three to six of his daughters (G2 females) (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). This approach has led to the recovery of mutations affecting processes ranging from sonic hedgehog signaling and telencephalic development to neonatal viability (Kasarskis et al., 1998; Zoltewicz et al., 1999; Anderson, 2000; Eggenschwiler et al., 2001; Herron et al., 2002; Garcia-Garcia and Anderson, 2003). Here, we have used ENU mutagenesis to generate and an immunohistochemical assay that stains neuronal processes to identify recessive mutations that affect cranial nerve development. We recovered mutations affecting processes ranging from global axon outgrowth to region-specific axon guidance. Even before identification of the causative genes, phenotypic comparison among mutants revealed that axon guidance of the oculomotor nerve is particularly sensitive to genetic perturbations. Finally, two of the mutants model known human genetic syndromes and, therefore, promise to provide novel insights into their etiology and treatment.
Materials and Methods
Mutagenesis and establishment of mouse strains. Twenty-five 8- to 10-week-old C57BL/6J male mice were treated with three weekly fractionated doses of 85 mg/kg ethyl-N-nitrosourea. Ten weeks after their final ENU treatment, male mice had regained their fertility and were bred to normal C57BL/6J females. In these experiments, no more than five G1 males were derived from the same G0 ENU-treated male. To generate pedigrees from 40 of the resulting G1 progeny males, each G1 male was bred to two normal C57BL/6J females and subsequently was bred with five to seven of his resulting G2 daughters. G3 embryos were harvested at 10.5 days postcoitus (dpc), and stained with anti-neurofilament antibody as described below. A G1 male was considered to carry a neurodevelopmental mutation if the identical phenotype was observed in multiple progeny from the same litter and in multiple progeny from other independent litters. To generate strains from G1 carrier males, the G1 male was bred to both C57BL/6J females for strain maintenance and potential behavioral analyses and to C3H/HeJ females for genetic mapping studies. Carrier C57/C3H hybrid males were identified by breeding.
Immunohistochemical staining of embryos. Immunohistochemical staining protocol was adapted from that described by Mark et al. (1993). The 10.5 dpc embryos were collected in PBS, fixed overnight in 4% paraformaldehyde in PBS at 4°C, washed three times in PBS containing 0.1% Tween-20 (PBT), and dehydrated with successive 15 min washes in 25, 50, 75% MeOH/PBT, and two washes of 100% MeOH. Litters were stored at –20°C until 20–50 litters had been collected and could be processed simultaneously. To block endogenous peroxidases, embryos were treated with a 4:1:1 MeOH:DMSO:H2O2 solution for 5–6 h at room temperature. Thereafter, the embryos were rehydrated in successive 15 min washes of 75, 50, 25% MeOH, and PBT at room temperature. To allow for better antibody penetration, the telencephalon was punctured. The 10.5 dpc embryos were treated with 10 μg/ml proteinase K (Sigma, St. Louis, MO) in PBT for 10–12 min at room temperature, and washed twice with freshly prepared PBT containing 2 mg/ml glycine. The embryos were refixed with fresh 0.2% glutaraldehyde/4% paraformaldehyde in PBT for 20 min and washed three times with PBT. Nonspecific antibody binding was blocked by washing the embryos in PBSMT (PBS plus 0.1% Triton plus 2% skim milk) for 1 h each at room temperature twice and a subsequent incubation of the embryos in 5% “secondary” sera for 2 h to overnight at 4°C in PBSMT. Embryos were incubated in a 1:1000 dilution of ascites containing 2H3 anti-neurofilament antibody (a monoclonal antibody raised against the 165 kDa neurofilament subunit and obtained from the Developmental Studies Hybridoma Bank, Iowa City, IA) (Dodd et al., 1988) in PBSMT at 4°C for 12–24 h with gentle rocking. After three rinses with PBSMT, the embryos were washed five times in PBSMT at 4°C for 1 h each. Thereafter, the embryos were incubated in PBSMT plus 5% goat serum for 1 h. To reduce background, the secondary antibody was preabsorbed by incubating 50 μl of goat anti-mouse secondary antibody (Vectastain) with 0.5 ml of PBSMT, 5% goat serum, and 3 mg/ml embryo powder for each 10 ml final antibody solution that was used for 2–6 h at 4°C. After centrifugation, the antibody-containing supernatant was further diluted to a final dilution of 1:200 of the original goat anti-mouse antibody. Embryos were incubated in PBSMT plus 5% goat serum for 1 h and were then incubated overnight at 4°C with the secondary antibody diluted in PBSMT. After three rinses with PBSMT, the embryos were washed five times in PBSMT at 4°C for 1 h each. Peroxidase staining was developed according to the Vectastain protocol (Vector Laboratories, Burlingame, CA). In a few instances, embryos were dehydrated through a methanol series and cleared in benzyl alcohol:benzyl benzoate (1:2).
Genetic mapping of recessive mutations. The genetic map position of mutations was determined by analysis of segregation of chromosome-specific markers on C3H inter-subspecific intercross panels by standard procedures. For the mapping of a recessive mutation, the progenitor (G1) male was bred to C3H/HeJ females. Five of the resulting G2 C57/C3H males were mated to three to four of their G2 sisters, and progeny was tested to identify “carrier” males, which were heterozygous for the mutation. The 10.5 dpc embryos were scored for the mutant phenotype by whole-mount immunohistochemistry with anti-neurofilament antibody. DNA was generated from all embryos from litters containing two or more embryos with the expected mutant phenotype. In previous studies, 45 of such DNA samples were genotyped for 70 MIT markers that show distinct C57BL/6J and C3H/HeJ specific variations (Kasarskis et al., 1998). Markers located near the mutation should show homozygosity for the C57-specific variant in mutant animals and be either homozygous for the C3H variant or heterozygous for the C3H and C57 variants in unaffected animals. We found that, in some instances, we could obtain an initial approximate genetic map position with fewer DNA samples. After determining the penetrance of a given mutation, as shown in supplemental Table 1 (available at www.jneurosci.org as supplemental material), we focused, if possible, on collecting affected embryos for the purposes of genotyping. Thus, for the 22-1 and 25-3 mutations, we initially genotyped 20 affected animals. For 26-1, 6 affected and 39 unaffected animals were genotyped. For the 7-1 mutation, a total of 12 affected and 33 unaffected animals were initially genotyped. In the text, megabase positions of markers are given according to the assembly 32 from the ENSEMBL database. The flanking markers for these mutations are as follows: D15MIT130— 7-1—D15MIT138; D1MIT132—22-1—D1MIT136; D9MIT121—26-1— telomere; and D4MIT162—25-3—D4MIT136. We have confirmed linkage, in all instances, by additional analyses of at least an additional 20 intercross progeny.
Results
Genetic screen design
To identify recessive neurodevelopmental mutations, we screened third-generation (G3) progeny from mutagenized C57BL/6J male mice using whole-mount immunohistochemistry with the 2H3 anti-neurofilament antibody. We performed these immunohistochemical analyses on 10.5 dpc G3 embryos for several important developmental and technical reasons: First, the antibody penetrates the embryonic tissues well at this stage, and, thus, neurons and axonal processes can be detected in their stereotypic positions allowing early innervation processes to be observed. Second, in contrast to 8.5–9.5 dpc timed matings, the variation in 10.5 dpc embryonic staging is minimal, and embryos can be harvested with ease. Third, embryos with more severe patterning defects often can survive up to 10.5 dpc, and thus mutations that affect overall neural patterning are still recoverable. Fourth, some genetic effects on axon migration may be missed unless mouse embryos are examined relatively early (10.5–12.5 dpc) in development, because early axon migration defects can in principle be corrected by remodeling and compensatory mechanisms by 15.5 dpc, as observed in Semaphorin 3A mouse mutants (White and Behar, 2000). In contrast to developmental gene expression-based assays, this immunohistochemical approach allows analysis of neuron development directly and makes no assumptions of what gene expression patterns may be the most important. Finally, processed embryos can be stored for over a year, thus preserving the experimental record.
After mutagenizing C57BL/6J male mice with ENU and mating these with C57BL/6J females, we randomly selected 40 first generation (G1) progeny males for additional breeding, bred each G1 male to five to seven of his daughters, and assayed the resulting G3 progeny (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). A G1 male was considered to carry a mutation if multiple embryos in a litter and in several subsequent litters showed the same phenotype. We chose the C57BL/6J strain for these experiments, because its genomic sequence is available in the public domain and it performs well in many behavioral assays. Also, by maintaining a uniform genetic background, we are able to assess phenotypic variability immediately. Our confidence of the initial phenotypic presentation informed our selection of which mutants to pursue. In this manner, we identified one mutation that specifically disrupts facial nerve development, one that affects sensory neuron development and localization of motor neuron cell bodies, one that disrupts ventral exit of motor neuron axons from the neural tube, and four mutations with axon guidance deficits in the periphery (supplemental Table 1, available at www.jneurosci.org as supplemental material).
Identification of a mutation with facial nerve deficits
We were most interested in identifying new mutations that affected only one or a subset of cranial nerves and might model aspects of human craniofacial syndromes. We identified one such mutation, 7-1. In mice homozygous for mutation 7-1, the facial and acoustic nerves consistently show abnormal “sprouting” or “branching” in 10.5 dpc embryos (Fig. 1). In 12.5–13.5 dpc 7-1 homozygotes, the lower jaw is severely reduced, the head appears smaller, and blood is largely absent. (Fig. 1E,F). Genetic mapping determined that 7-1 is located on chromosome 15 between D15MIT280 at 24 Mb and D15MIT111 at 32.2 Mb within a region of conserved synteny to that in which the gene(s) responsible for the craniofacial and mental retardation deficits of human cri-du-chat syndrome reside (supplemental Table 1, available at www.jneurosci.org as supplemental material).
Identification of a mutation with abnormal sensory ganglia development
In the mouse, previous genetic screens have used obvious defects in head morphology at 8.5 and 10.5 dpc to identify mutations required for dorsoventral patterning or telencephalic development (Kasarskis et al., 1998; Zoltewicz et al., 1999; Anderson, 2000; Eggenschwiler et al., 2001; Garcia-Garcia and Anderson, 2003). We also identified possible mutations that caused a variety of morphological defects. These included one mutation that caused severe deficits in cardiac development, two mutations that caused a dramatic overproliferation of the anterior CNS similar to that observed in presenilin 1, Casp9 (caspase 9), and Apaf1 (apoptosis protease-activating factor 1) mutants (Wong et al., 1997; Cecconi et al., 1998; Hakem et al., 1998; Kuida et al., 1998), two mutations (17-2 and 26-1) with abnormal telencephalic development, and one (22-1) with deficits in neural tube closure. After analyzing neuronal development in these with anti-neurofilament staining, the telencephalic mutation 26-1 and the neural tube closure mutation 22-1 consistently showed abnormalities in several aspects of cranial nerve development (Fig. 2). Because exencephaly and neuronal deficits of presumptive 17-2 mutants was extremely variable even on a pure C57BL/6J background, we chose not to pursue this mutation further.
In the 22-1 mutation, overall embryonic morphology appears to be fairly normal except for defects in dorsal neural tube closure, which were more prevalent in the anterior of the embryo (Fig. 2C,D). Immunostaining of sections with anti-neurofilament antibody revealed that sensory neurons appeared to be largely missing (Fig. 3B,D,E), and some motor neuron cell bodies were observed outside of the ventral neural tube. Motor neuron identity of these ectopic neurons was confirmed by Kreisler/MafB expression (data not shown). The pairing of these two deficits has been previously observed in animals, in which the neural crest has been physically or genetically ablated. General dorsal–ventral patterning, however, as assessed by region-specific expression of sonic hedgehog (Shh), Islet1 (Isl1), and Math1 appeared normal (Fig. 3H,K,Q). Genetic mapping located 22-1 near D1MIT132 at 43.7 ± 7 Mb on chromosome 1. Given that, aside from the neuronal defects seen in 22-1 homozygotes, 22-1 heterozygotes display a white belly spot and that 22-1 homozygotes also have a curly tail, it seemed likely that 22-1 might represent a mutation in the Pax3 gene located in this region (Mansouri and Gruss, 1998). Sequencing of Pax3 from 22-1 homozygotes identified an A-to-T transition that alters isoleucine at amino acid position 86, conserved in the paired box of all known paired domain proteins, to an asparagine (Fig. 4A). Such a nonconservative change would be predicted to have dramatic consequences on Pax3 function. Thus, the 22-1 phenotype is almost certainly caused by this mutation in Pax3. In the future, complementation analyses with existing Pax3 alleles can test whether this mutation is indeed responsible for the 22-1 phenotype and molecular biologic and phenotypic analyses can determine whether the 22-1 allele acts as a hypomorphic or null allele. Previously, mutations that cause amino acid substitutions in the paired box of the human PAX3 protein have been associated with the auditory–pigmentary type I Waardenburg syndrome (WS-I) (Fig. 4B) (Zlotogora et al., 1995; Fortin et al., 1997; Sotirova et al., 2000; Wollnik et al., 2003).
Identification of a mutation that affects global ventral motor axon outgrowth
Anti-neurofilament staining of embryos homozygous for mutation 26-1 revealed that, although the anterior telencephalon is absent, anterior–posterior patterning, as judged by the position of sensory ganglia, appeared somewhat normal up to the midbrain–hindbrain boundary and the optic cup was always present (Fig. 2E,F). From examination of whole-mount embryos, it was unclear whether motor neurons were entirely absent or whether they simply failed to extend their axons into the periphery. However, examination of anti-neurofilament-immunostained sections revealed that, along the entire spinal cord and hindbrain, motor neurons were present, but extended their axons primarily dorsally in a somewhat feathery, disorganized manner (Fig. 3C,F). In contrast, normally most motor axons, with the exception of those of the dMNs, extend ventrally out of the neural tube into the periphery (Fig. 3A). This phenotype was remarkably penetrant, because no ventral motor axon extension was observed along the entire embryonic axis of 26-1 mutants. In addition, neural crest-derived sensory nerves are also markedly reduced, and neural tube closure does not occur. Disruptions in sonic hedgehog signaling have been shown to cause deficits in both anterior patterning and motor neuron specification (Ericson et al., 1995; Marigo et al., 1996). However, sonic hedgehog expression appeared normal in the notochord and floorplate of 26-1 homozygotes, and the presence of the motor neuron specification marker Islet1 in domains adjacent to the floorplate indicated that sonic hedgehog signaling-dependent motor neuron specification occurred normally (Fig. 3I,L). In contrast, motor neuron-specific expression of the MafB(Kreisler) motor neuron differentiation marker was not observed in 10.5 dpc embryos. Because 26-1 homozygotes die shortly after 10.5 dpc, later events in neural development could not be assessed. The 26-1 mutation is situated on chromosome 9 near D9MIT151 at 121.5 Mb, a region in which no genes with known roles in motor neuron outgrowth or anterior patterning have been located. Thus, 26-1 probably affects a novel gene with important roles in anterior neural patterning, neural tube closure, and, most excitingly, the initial events governing motor neuron differentiation and/or ventral motor neuron axon extension.
Mutations with deficits in general axon guidance
Axonal pathfinding proceeds in a stepwise manner. Initially, axons fasciculate as they extend toward their first target. On reaching a choice point within their first target field, they reorient themselves and their growth cones turn toward their next target field. A variety of molecules, the majority of which identified to date are secreted or cell surface molecules, mediate the guidance toward a given target area and axon fasciculation. On reaching their final destination, axons defasciculate and synaptogenesis begins. Progress past the first choice point and toward the second target is a complex process that involves molecular cues that mark the choice point and the intrinsic ability of a neuron to reset itself and redirect its axon migration appropriately.
Despite the small size of this screen, we recovered representative mutations that affect axon fasciculation and choice point execution (Fig. 5). Progeny from male 8-1 showed defects in overall axon fasciculation (Fig. 5C,D). Similar abnormalities in axon fasciculation have been reported in animals homozygous for disruptions of the neuropilin/VEGF (vascular endothelial growth factor) and the Eph receptor/ephrin signaling pathways (Kitsukawa et al., 1995; McBride and Ruiz, 1998; Wang et al., 1998; Adams et al., 1999; Gerety et al., 1999; Huynh-Do et al., 1999; Kawasaki et al., 1999; Helbling et al., 2000). Thus, one of these pathways may be affected here as well.
We also recovered one mutation, 25-3, that appears to affect axon targeting and possibly axonal redirection at the first choice point (Fig. 5E,F). Embryos homozygous for the 25-3 mutation appear to exhibit deficits in choice point execution. In 10.5 dpc 25-3 embryos, the axons of the cranial nerves appear to exit the neural tube and fasciculate appropriately. However, as these axons approach their presumptive first choice point, they appear to meander in a zigzagging manner before proceeding to their second target field. This is particularly apparent in the trajectory of the oculomotor nerve and in the additional branches seen for the facial nerve (Fig. 5F).
Mutations with deficits in oculomotor neuron guidance
In our screen, we recovered two mutations that affect the axonal projections of the oculomotor nerve. In embryos from male 3-4, the axons of the oculomotor (III) nerve stream out from the neural tube in a detached and disorderly manner (Fig. 5G,H). This could represent either a defect in specification or guidance of the oculomotor nerve, because no other nerves appear to be overtly affected at 10.5 dpc. Finally, in progeny from mouse 3-3, the oculomotor nerve consistently showed delays in its migration toward its first target field (Fig. 5I,J). This may represent an extrinsic deficit in the target field or an intrinsic inability of the oculomotor nerve to respond to its guidance cues.
Discussion
Our screen is the first one in the mouse that has examined early neuronal development directly with immunohistochemical techniques. Other small-scale recessive screens have identified neurodevelopmental mouse mutations by screening neonates or 10.5 dpc embryos for gross external morphological abnormalities or by examining expression of a neuron-specific LacZ reporter in transgenic embryos (Kasarskis et al., 1998; Zoltewicz et al., 1999; Herron et al., 2002; Kile et al., 2003; Zarbalis et al., 2004). Our immunohistochemical approach has the advantage that it can be performed on any strain without previous generation of transgenic animals and examines neurodevelopment directly.
The recovery of seven neurodevelopmental mutations after screening only 40 G1 mice demonstrates that this strategy, if expanded, would be as productive as similar invertebrate genetic screens. Considering that the ENU dosage used should lead to identification of a recessive mutation in any given gene after screening 500–1500 G1 animals, the mouse genome contains ∼25,000–30,000 genes, and ∼20% of G1 mice carried neurodevelopmental mutations, ∼1–2% of randomly introduced recessive mutations should result in neurodevelopmental deficits scorable in 10.5 dpc embryos. Although G1 males carry mutations in multiple genes, successive outcrossing significantly reduces confounding effects of other mutations. Our results also predict that at least 1–2% of existing mouse mutations, whether generated by gene targeting or random mutagenesis, are likely to have early neurobiologic consequences. In the future, secondary screens, including other immunohistochemical assays, transgenic reporters, or behavioral assays, can assist in the additional classification of mutations. Thus, an expanded version of this screen would provide a valuable resource for the neurobiology community.
The mouse acts as an important intermediate for understanding human neurobiology. In this screen, we recovered two mutations that model human genetic diseases. One of these, mutation 7-1, models aspects of the human cri-du-chat deletion syndrome. Our identification of the 7-1 mutation, which is almost certainly caused by a mutation in a single gene, strongly suggests that loss of this one gene is responsible for the lower jaw deficits seen in some cri-du-chat patients. Cri-du-chat patients often exhibit difficulty swallowing and suckling, strabismus, and saliva control problems, which are likely caused by as-yet-undefined deficits in cranial nerve development. Thus, the 7-1 mutation will help elucidate specific cranial nerve deficits present in some cri-du-chat patients that heretofore may have been overshadowed by the more overwhelming symptoms (Church et al., 1995, 1997; Cornish and Munir, 1998; Mainardi et al., 2001).
Our identification of the I86N alteration in the Pax3 protein provides additional structure–function information for a protein with an already-established role in human type I and type III Waardenburg syndromes. In humans, WS-I is an auditory–pigmentary syndrome generally caused by heterozygous loss-of-function mutations in the PAX3 gene (Fig. 4B) (Vogan et al., 1993; Zlotogora et al., 1995; Fortin et al., 1997; Sotirova et al., 2000). Homozygosity or compound heterozygosity for PAX3 mutations has been documented in patients with Klein–Waardenburg syndrome (WS-III). On occasion, missense or specific deletion mutations in the paired domain of the PAX3 gene can cause WS-III (Fig. 4B) (Wollnik et al., 2003). Thus, additional analyses of the 22-1 phenotype will contribute further to our understanding of PAX3 structure and function in humans (Engelkamp and van Heyningen, 1996; Fortin et al., 1997).
The Shh signaling pathway has been shown to play essential roles in the anterior CNS patterning (Hooper and Scott, 2005), neuronal development (Roelink et al., 1994; Ericson et al., 1995), and guidance within the ventral neural tube (Charron et al., 2003). Thus, defective Shh signaling easily could account for the twin observations of deficits in anterior patterning and motor neuron development in 26-1 mutants. To our surprise, Shh signaling appeared normal in 26-1 mutants, but motor axons projected their axons dorsally rather than ventrally. At present, only two genes, the chemokine stromal cell-derived factor 1 (SDF1) (Cxcl12) and its receptor Cxcr4, have been shown to have specific roles in governing initial ventrally directed motor axon outgrowth (Lieberam et al., 2005). The 26-1 mutation is in neither of these genes. Although chemokines and their receptors have been shown to direct cell migration in the nervous system (Bagri et al., 2002; Lu et al., 2002; Stumm et al., 2003; Belmadani et al., 2005; Knaut et al., 2005; Miller and Tran, 2005; Sapede et al., 2005), so far none has been shown to affect telencephalic patterning. Nonetheless, it is attractive to speculate that 26-1 affects a gene with a key role in directing motor axon outgrowth by affecting the SDF1(Cxcl12)–Cxcr4 signaling pathway.
Aside from the two patterning mutations described, all remaining neurodevelopmental mutations appeared morphologically indistinguishable from their normal littermates, and only immunohistochemical staining revealed axon guidance abnormalities. The 10.5 dpc progeny from male 8-1 showed defects in overall axon fasciculation, but not in axonal convergence. Interestingly, in 13.5 dpc progeny from male 8-1, striking vascular abnormalities were also seen. An increasing number of molecules important for axon guidance, such as the neuropilins, semaphorins, and Eph receptors/ephrins play key roles in angiogenesis as well (Kitsukawa et al., 1995; McBride and Ruiz, 1998; Wang et al., 1998; Adams et al., 1999; Gerety et al., 1999; Huynh-Do et al., 1999; Kawasaki et al., 1999; Helbling et al., 2000; Park et al., 2004; Gu et al., 2005). Thus, it is likely, but remains to be explored, that these two phenotypes may be caused by the same mutation.
Mutation 25-3 causes axons to zigzag as they approach their first choice point and may affect reorienting of cranial nerves at their first choice point. Some molecular cues that direct axons to cross the vertebrate and invertebrate midline have been identified (Brose and Tessier-Lavigne, 2000), but we know almost nothing about those that act globally or in a region-specific manner to redirect axon extension in the periphery. Thus, mutation 25-3 should provide a unique foothold toward understanding axon targeting and choice point selection in mammals.
Interestingly, we found that mutations could be categorized by their effects on the oculomotor nerve (Fig. 6). The oculomotor nerve originates in the ventral midbrain, initially extends its axons toward the trigeminal (V) nerve, and then reorients its projections toward the optic cup (Fig. 6). Our ability to identify mutations that affect the oculomotor nerve trajectory may be a consequence of its long, solo pioneering path from the midbrain to its target. During its progress, no other major neurons obscure its visibility or can provide compensatory guidance cues. Thus, it is both more visible and probably more vulnerable to mutations than other cranial nerves.
Whereas 8-1 mutants showed general fasciculation deficits in the oculomotor and other nerves, three other mutations had more precise and biologically informative deficits in oculomotor nerve guidance. Mutation 3-4, in which the initial projections of the oculomotor nerve originate from the midbrain in a diffuse manner, may reflect anomalies in early midbrain development or deficits in oculomotor nerve differentiation or fasciculation. The oculomotor nerve deficits in mutations 3-3 and 25-3 suggest that cues present halfway between the midbrain and trigeminal nerve play a key role in oculomotor nerve guidance. In mutation 3-3, axonal extension of the oculomotor nerve stalls halfway between the trigeminal nerve and the midbrain, at a time when the trigeminal nerve has already extended its upper branch toward the optic cup. In the 25-3 mutation, oculomotor axons begin to zigzag halfway between the midbrain and trigeminal nerve, at roughly the same point at which the nerve stalls in 3-3 mutants. Thus, it seems plausible that guidance occurs in two phases here. This “halfway” point may mark the beginning of a “transition zone” and represent an interface between sequentially acting cues that keep the oculomotor axons on the straight and narrow and ultimately allow it to extend to its target area. Given these intriguing results, identification and analyses of additional mutations with specific deficits in oculomotor nerve guidance may provide one of the most fruitful avenues for exploring region-specific axon guidance in the mouse.
The availability of mouse genome sequence and advances in molecular genetic techniques will facilitate identification of the affected genes and additional analysis of their biological functions. So, genetic complementation experiments can be expedited by the ever-increasing number of targeted mutations (>2000) in the mouse, the availability of mutations generated by multiple large-scale gene-trapping efforts in embryonic stem cells, and the gene-driven screening of ENU-mutagenized embryonic stem cells or cryopreserved sperm to generate allelic series (Chen et al., 2000; Coghill et al., 2002). Meiotic recombination experiments can be augmented with region-specific deletions to refine critical intervals (Justice et al., 1997; You et al., 1997; Goodwin et al., 2001; Kile et al., 2003). Gene rescue experiments can be performed by generation of mice transgenic for bacterial artificial chromosomes, which can be elegantly modified using the Red/ET recombineering system, which uses the λ Redα/Redβ recombinase pair (Lee et al., 2001; Yu and Bradley, 2001; Testa et al., 2003; Valenzuela et al., 2003). Finally, with the advent of stem cell technologies and efforts toward neuronal regeneration, molecules that will be identified in genetic screens and our understanding of their function may ultimately lead to the ability to regenerate precisely any nerves damaged by accident or illness. The productivity of this small-scale screen and the information we have already obtained underscores the power of this approach. The future identification of the genes affected in the mutations identified here and the expansion of this screen and addition of other primary and secondary assays should prove tremendously productive and informative.
Footnotes
This work was funded by an EJLB Foundation Scholar Award to S.P.C., by a generous gift from Mr. Henry Bernick and Mrs. Esther Bernick, and by Grant 14312 from the Canadian Institute for Health Research. We thank Monica Justice for advice regarding mutagenesis of C57BL/6J mice, Joe Culotti for critical reading of this manuscript, Lucy Osborne and the Canadian Models for Human Disease genotyping core for initial determination of genetic map positions, Dragana Vukasovic for secretarial assistance, Karin Weiss and Igor Vukabradovic for help with animal husbandry, and Timur Urakov for help with 26-1 mutant analysis.
Correspondence should be addressed to Sabine P. Cordes, Samuel Lunenfeld Research Institute, Room 865, Mount Sinai Hospital, 600 University Avenue, Toronto, ON, Canada M5G 1X5. E-mail: cordes{at}mshri.on.ca.
DOI:10.1523/JNEUROSCI.3813-05.2005
Copyright © 2005 Society for Neuroscience 0270-6474/05/2511787-09$15.00/0