Hypogonadotropic hypogonadism and peripheral neuropathy inEbf2-null mice

Olf/Ebf genes (Garel et al.,1997; Hagman et al.,1993; Malgaretti et al.,1997; Wang and Reed,1993; Wang et al.,1997) (reviewed by Dubois and Vincent, 2001) (herein referred to as Ebf genes, in agreement with the accepted mouse nomenclature) encode transcription factors strikingly conserved in evolution from pseudocoelomates to chordates. These transcription factors contain an HLH dimerization domain and a characteristic DNA binding domain featuring a C-C-H-H zinc-finger motif. In vitro, these proteins bind an imperfect palindromic DNA consensus sequence, both as homo- and heterodimers. Albeit originally identified for their roles in the immune system(Hagman et al., 1993), Ebf genes have been implicated in various aspects of neural development(Dubois et al., 1998;Garel et al., 1999;Malgaretti et al., 1997;Pozzoli et al., 2001;Prasad et al., 1998) and neuronal function (Kudrycki et al.,1993). Specifically, they have been found to play roles in ventral nerve cord fasciculation in C. elegans(Prasad et al., 1998) and axon navigation (Garel et al.,1999), as well as neuronal migration(Garel et al., 2000) and differentiation (Dubois et al.,1998; Pozzoli et al.,2001; Prasad et al.,1998) in various organisms. Several in vivo studies have disclosed genetic interactions between Ebf genes and some of the most important regulators of neural development, including hedgehog(Vervoort et al., 1999) and Notch (Crozatier and Vincent,1999). All the above data suggest a potential role for Ebf genes in short-range dorsoventral patterning and morphogenesis of the embryonic nervous system.

Three mammalian Ebf genes have been isolated so far: Ebf1, Ebf2and Ebf3. An additional member of the family, O/E4, was isolated more recently (Wang et al.,2002) and its expression in the embryonic neural tube has not been mapped to date. In the embryonic mouse nervous system, Ebf1, Ebf2 andEbf3 are transcribed in partially overlapping territories(Garel et al., 1997), but display significant differences in their distributions(Garel et al., 1997;Malgaretti et al., 1997),suggesting that each may play roles in the morphogenesis of different head territories. As one example, Ebf1 is the only family member expressed in the striatal anlage (Garel et al.,1997). Second, noticeable differences are observed in the distribution of Ebf transcripts across the wall of the neural tube (NT):Ebf1 and Ebf3, which map to mouse chromosomes 11 and 7,respectively (Garel et al.,1997), are co-expressed at most sites in the mantle layer of the developing NT, whereas Ebf2, which maps to mouse chromosome 14, is expressed by younger postmitotic neural cells located in the subventricular layer of the prospective spinal cord (Garel et al., 1997; Malgaretti et al., 1997). In the developing nervous system, the functions ofEbf1 and Ebf3 appear to be somewhat redundant, in thatEbf1^-/- mutants feature defects in neuronal differentiation (Garel et al.,1999) and short-range neuronal migration(Garel et al., 2000) only in territories in which Ebf3 is not expressed. However, no genetic evidence is available regarding the specific role of Ebf2 in the context of vertebrate neural development. Again, overexpression studies conducted in Xenopus, where an Ebf1 ortholog has not been isolated to date, indicate that Xebf2/Xcoe2(Dubois et al., 1998;Pozzoli et al., 2001) is expressed earlier than Xebf3(Pozzoli et al., 2001), and suggest that the two genes may play distinct roles in the context of primary neurogenesis (Pozzoli et al.,2001). In particular, gain-of-function experiments suggest thatXebf2 and Xebf3 differ in their sensitivity to lateral inhibition (Pozzoli et al.,2001). In fact, Notch activation can suppress the ability of NeuroD (Neurod1 — Mouse Genome Informatics) to promote Xebf2 but not Xebf3 transcription(Pozzoli et al., 2001). Finally, lateral inhibition interferes with the ability of overexpressedXebf2, but not Xebf3, to activate transcription of early(N-tubulin) and late (NF-M) neuronal differentiation markers(Pozzoli et al., 2001). These results indicate that Notch activation interferes in a cell-autonomous fashion with Ebf2 function, not just with its transcription.

The features and interactions common to the entire Ebf gene family in phylogeny, as well as the specific features of the Ebf2 homolog inXenopus neurogenesis prompted us to study the contribution ofEbf2 (Malgaretti et al.,1997) to neural development and function, using a genetic approach in the mouse. Ebf2^-/- mice were generated by gene targeting, and revealed a key role for Ebf2 in the embryonic migration of gonadotropin-releasing hormone (GnRH)-synthesizing neurons from the vomeronasal organ to the hypothalamus. In addition, our results implicated this gene in peripheral nerve morphogenesis.

All experiments described in this paper were conducted in agreement with the stipulations of the San Raffaele Scientific Institute Animal Care and Use Committee.

Eleven Ebf2 genomic clones were isolated by screening of a 129/SvJ mouse genomic library (10⁶ pfu) with an Ebf2-specific intronic probe. the probe was derived by restriction of a PCR product obtained with a forward primer from exon 1 and a reverse primer from exon 3 (F:5′-CTG GGT GCC GAG ATG GAT T; R: 5′-TGT TGG TCT TCT CAT TGC CTT). We used a modified pPNT vector containing a Neo cassette flanked byLoxP sites. The construct contains a PGK-tk cassette downstream of the 3′ arm of homology. After homologous recombination,5.5 kb of genomic sequence were deleted from the Ebf2 gene, including the putative translation initiation site and the first five exons.

ES cells (TBV2 line) were grown in Dulbecco's Modified Eagles Medium (DMEM)(Gibco), 15% fetal calf serum (Gibco), 10^-4 Mβ-mercaptoethanol (Gibco), 2 mM L-Glutamine and 1000 U/ml LIF (Heiko), on an embryonic fibroblast feeder layer previously inactivated with Mitomycin C. Electroporation, positive and negative selection were performed as described(Joyner, 1993). Resistant colonies were picked after 8-10 days of selection. Genomic DNA was extracted from expanded clones, digested with HindIII and analyzed by Southern blotting at the 3′ end of the recombinant locus. Homologous recombinant clones were analyzed at the 5′ end by EcoRI digest and Southern analysis. Out of 1200 ES clones screened, three scored positive for homologous recombination and were propagated.

The targeted ES clones were injected into blastocysts derived from C57BL/6J females. The chimeric embryos were then transferred into the uteri of 2.5 day pseudopregnant foster mothers. Chimeric males with 40-100% agouti color were test-bred by crossing with wild-type C57BL/6J females and germline transmission was identified by the presence of agouti offspring. Heterozygous mice were initially identified by Southern blotting with the 3′ probe. Subsequent genotyping was carried out by PCR amplification withlacZ-specific and Ebf2-specific primers.

RNA was extracted by the caesium chloride method from E13 embryos and from postnatal and adult sciatic nerves. Reverse transcription was conducted with oligo-dT and random hexamers, as described previously(Ausubel et al., 1995). A primer pair (F: 5′-TGG AGA ATG ACA AAG AGC AAG; R: 5′-GGG TTT CCC GCT GTT TTC AAA) specific for the cDNA region encoding the Ebf2 DNA-binding domain was used for PCR amplification, according to standard protocols.Gapdh primers used for normalization were 5′-CGC ATC TTC TTG TGC AGT G (forward) and 5′-GTT CAG CTC TGG GAT GAC (reverse).

The mutation was transferred onto different genetic backgrounds by three distinct procedures: first, we incrossed the transmitting chimeras with 129/SvJ females to transfer the mutation onto a 129/SvJ background and obtain pure coisogenic mutants. Second, we intercrossed(129/SvJ×C57BL/6J)F₁ mice to obtain F₂ homozygotes of heterogeneous genetic backgrounds. Third, we backcrossed(129/SvJ×C57BL/6J)F₁ hybrids with C57BL6/J mice.

The KO construct contains a neomycin minigene(Joyner, 1993). This cassette is flanked by loxP sites. In order to exclude a possible effect of the Neo minigene on the mutant phenotype, we crossedEbf2^+/- mice with transgenic mice expressing theCre recombinase under control of the β-actin promoter. By intercrossing Cre+ Ebf2^+/- transheterozygotes, we obtained Neo-negative, Ebf2^-/- mice, that wereCre-negative, formally excluding mosaicism for the Neocassette. Phenotypic and pathological abnormalities observed in those mice were indistinguishable from those scored in Ebf2^-/-,Neo-positive mice (data not shown).

For whole-mount lacZ staining, whole tissues were removed from mice anesthetized and transcardially perfused with 0.9% NaCl followed by 4%paraformaldehyde (PFA). Dissected organs were washed in wash buffer (0.01%sodium deoxycolate, 0.02% Nonidet P40, 2 mM magnesium chloride in 1×phoshate buffer) and stained for 2-6 hours at 37°C in staining solution(50 mg X-gal, 0.106 g potassium ferrocyanide and 0.082 g potassium ferricyanide were dissolved in 50 ml of wash buffer).

Embryos and mice were respectively fixed by immersion, or anesthetized and transcardially perfused with 0.9% NaCl followed by 4% PFA. Brain and testes were postfixed overnight in 4% PFA at 4°C, dehydrated in ethanol, embedded in paraffin wax and cut into 5-7 μm section using a microtome. Sections were counterstained with Cresyl Violet (Sigma) or Hematoxylin And Eosin(Sigma). For immunohistochemistry, tissues were cryoprotected in 30% sucrose,1×PBS (overnight), included in OCT (Bioptica) and stored at -80°C,before sectioning in a cryotome (5-15 μm). Immunohistochemical analysis was conducted on cryosections with the following antibodies: polyclonal anti GnRH(LR1,1:2000, courtesy of R. Benoit, Montreal), polyclonal anti-peripherin(1:1000, Chemicon) and anti Tag-1 adhesion molecule (courtesy of A. J. W. Furley, Sheffield). Sections were immunostained as suggested (Vector Laboratories), counterstained with Methyl Green (Sigma), dehydrated and mounted with DPX (BDH-Merck).

For dual immunofluorescence, E15 embryo cryosections were treated for 10 minutes with 0.1 M glycine, preincubated in 15% goat serum, 0.2% triton X-100,1×PBS and incubated overnight at 4°C with the two primary antibodies(polyclonal anti GnRH, 1:1000, monoclonal anti-β-galactosidase, 1:500 Promega). The sections were washed 6 times for 10 minutes in 0.45 M NaCl, 0.3%triton X-100, 20 mM phosphate buffer, rinsed in 1×PBS and incubated for 1 hour at room temperature with the two secondary antibodies (TRITC anti-mouse, 1:150, Jackson ImmunoResearch Laboratories and FITC anti rabbit,1:150, Sigma). Controls consisted of replacing either the first primary Ab or the second primary Ab with goat serum. Control sections indicated no cross-reactivity between the first and the second Ab.

Sciatic nerve specimens were fixed with 2% glutaraldehyde (2-3 hours),postfixed in 1% osmium tetroxide, following ethanol dehydration, and finally embedded in Epon/araldite. Electron microscopy analysis was performed on ultrathin transverse sections made on a Reichert ultramicrotome, stained with uranyl acetate and lead citrate. Slides were examined under a Zeiss electron microscope.

Adult male mice were injected intraperitoneally with 100 mg/kg of bromodeoxyuridine (BrdU, Sigma) and sacrificed 2 hours later. Anti BrdU immunohistochemistry was performed as described(Garel et al., 1997).

Mice were anesthetized with trichloroethanol, 0.02 ml/g body weight, and placed under a heating lamp in order to avoid hypothermia. Electrophysiological tests were carried out as described(Zielasek et al., 1996). Motor responses were acquired with a Medelec Sapphire 4 Me electromyograph (Medelec,Woking, UK). Statistical analysis was performed by using the Student'st-test for unpaired data.

In order to produce a gene targeting construct and inactivate theEbf2 gene (Fig. 1), we screened a genomic phage library (129/SvJ) using a probe corresponding to intron 1 of the gene. We isolated two genomic DNA clones spanning 4 kb upstream of the transcription initiation site through exon 6, located 12.5 kb further downstream (Fig. 1A),and generated a targeting construct in which a neomycin resistance minigene(Pgk-Neo), preceded by a promoterless lacZ cDNA, was used to replace a 5.2 kb stretch encompassing the translation initiation site and the first five exons. This region of the gene encodes the majority of the Ebf2 DNA-binding domain (Malgaretti et al.,1997; Wang et al.,1997). We electroporated embryonic stem (ES) clones with theNotI-linearized recombinant plasmid. By standard molecular genetic techniques, we isolated three homologous recombinant ES cell clones that we expanded and used to inject blastocysts. High percentage chimeric males were mated with C57BL/6J females, and F₁ recombinant progeny were selected based on agouti coat color and genotyping at the Ebf2 locus. By intercrossing heterozygous mutants[(129/SvJ^Ebf2-lacZ×C57BL/6J)F₁] we obtained a percentage of homozygous mutant F₂ mice consistent with expected Mendelian frequencies, suggesting that outbredEbf2^-/- mice do not die as embryos.

RT-PCR experiments conducted on total RNA extracted from wild-type and homozygous mutant E13 embryos provided evidence that homozygous mutants produce no Ebf2 transcript 3′ to the lacZ insertion(Fig. 1D). To maximize sensitivity, a similar experiment was conducted on cDNA extracted from the peripheral nerve, with Southern transfer of RT-PCR products followed by hybridization with an Ebf2 probe, yielding concordant results (not shown).

The Ebf2 gene is expressed at numerous sites in the nervous system, both before and after birth (Garel et al., 1997; Malgaretti et al., 1997). Taking advantage of the promoterless lacZgene (encoding a cytoplasmic β-galactosidase) integrated into theEbf2 locus, we analyzed Ebf2 expression by histochemicallacZ staining. The staining distribution in early development is in full agreement with the results of previous expression studies(Garel et al., 1997;Malgaretti et al., 1997). The present paper will focus on expression sites corresponding to the major morphological changes observed in the mutants.

Fig. 1E shows the phenotypic features of two 20-day-old Ebf2^-/- males next to a wild-type male littermate. At birth, Ebf2^-/- mice were hardly distinguishable from Ebf2^+/- andEbf2^+/+ ones after gross examination. The first obvious abnormalities became detectable starting at 5 days after birth (P5).Ebf2^-/- mice were small compared with theirEbf2^+/+ and Ebf2^+/- littermates:homozygotes weighed less than half the bodyweight of their littermates by P30(Ebf2^+/+, 16.46±1.84 g, n=9;Ebf2^+/-, 15.34±2.45 g, n=12;Ebf2^-/-, 6.87±1.78 g, n=7;Ebf2^+/+ versus Ebf2^-/-,P<0.0001; Ebf2^+/- versusEbf2^-/-, P<0.0001; Ebf2^+/+versus Ebf2^+/-, not significant). Studies are in progress to define the physiological basis of this growth retardation.

Phenotypic and pathological changes described herein have been scored in mice derived through at least five generations of backcrossing on the C57BL6/J background, in order to minimize the effects of genetic background heterogeneity. Once they reached adulthood, Ebf2^-/- males and females failed to reproduce when mated to wild-type breeders of the opposite gender. In addition, Ebf2^-/- mice were mildly uncoordinated and walked with an unsteady, waddling gait. Most noticeably,they exhibited a `hunchback' posture both at rest and, more prominently, while walking.

The reproductive failure observed in our mutants suggested several possible explanations, including an alteration in the hypothalamus-pituitary gonadotropic axis, which governs sexual maturation(Wilson et al., 1998). In this respect, we investigated the development of gonadotropin releasing hormone GnRH-neurons; these neurons are initially located in the vomeronasal organ anlage of the nasal placode and subsequently migrate medially and dorsally through the nasal mesenchyme and cribriform plate of the ethmoid bone along the vomeronasal and terminal nerve fibers, then penetrate into the rostral forebrain and move caudally into the septohypothalamic region, where they extend axons towards the median eminence (ME) and infundibulum of the pituitary gland (reviewed by Wray,2001). In agreement with previous reports(Wang et al., 1997),lacZ staining in heterozygous midgestation embryos revealedEbf2 expression in the olfactory epithelium, and particularly in the vomeronasal organ (Fig. 2A,B). In Ebf2^-/- embryos, olfactory and vomeronasal fibers could be detected at all correct locations in the mutant forebrain at embryonic day 15, and stained positive for peripherin(Wray et al., 1994)(Fig. 2D) and for the Tag-1 adhesion molecule (Wolfer et al.,1994) (not shown) in wild-type and mutant mice alike. Furthermore,by histological analysis, we examined olfactory bulb sections obtained from P30 wild-type and null mutant mice. No abnormalities were observed in the cytoarchitecture of this region in Ebf2^-/- mice(Fig. 2F).

By dual immunofluorescence, we determined that Ebf2 is expressed in migrating GnRH-neurons as the onset of their migration, i.e. E11 in the mouse (Schwanzel-Fukuda and Pfaff,1989) (Fig. 3).Fig. 3 shows colocalization(Fig. 3C) of GnRH(Fig. 3A) andβ-galactosidase (B) immunostaining in migrating E15 GnRH-neurons in the null mutant's nasal mesenchyme. We used β-galactosidase-like staining as an unambiguous marker of Ebf2 expression, as polyclonal anti-Ebf2 antibodies available to us crossreact with other Ebf proteins. At embryonic day 15 (E15), wild-type (wt) GnRH-neurons were mostly located dorsal to the cribriform plate and were navigating caudally towards the preoptic region;conversely, in E15 mutant embryos, most GnRH-neurons were still detected in the nasal mesenchyme (Fig. 4B,Fig. 3A).Ebf2^-/- neurons were very tightly clustered and in some cases did not show any obvious leading nor trailing processes typical of migrating GnRH-neurons. At birth (P0), GnRH-positive fibers were evident in the ME of the hypothalamus (Fig. 4C), whereas no fibers were detectable in the ME of the mutants(Fig. 4D). Finally, in P30 mutant brains, hardly any GnRH-positive neurons were detected either in the septohypothalamic region (Fig. 4F) or in any region spanning the olfactory bulbs through the anterior commissure. The defect in GnRH-neuron development did not appear to be secondary to abnormal growth or navigation of the olfactory fibers that support their neurophilic migration. In fact these fibers are normally present in null mutants (Fig. 2D).

Based on the above observations, we investigated the eventual fate of GnRH-neurons in Ebf2^-/- mice. As mentioned earlier, at E15 most neurons were abnormally retained in the nasal cavity ofEbf2^-/- embryos, unlike in their wild-type counterparts. Hoechst counterstaining of these neurons revealed that in most cases their nuclei had indented borders and fuzzy to undiscernible nucleoli, suggesting that neurons that failed to migrate eventually degenerated (not shown). Neuronal degeneration appeared to be secondary to defective migration, as numerous neurons were still found in the nasal mesenchyme 5.5 days after their birth in the vomeronasal organ. Although migration of most GnRH-neurons was arrested in the nasal mesenchyme or just dorsal to the cribriform plate, a subset of GnRH-immunoreactive neurons were found in the forebrain at birth, at the caudal boundary of the olfactory bulb(Fig. 4G).

Consistent with the observed defect in GnRH-neuron development, inbredEbf2^-/- mice were hypogonadic and failed to reproduce or to exhibit any mating behavior when caged with wild-type C57BL6/J breeders. This statement refers to a cumulative period of over 14 months of random matings (n=3 homozygous mutant males; n=3 homozygous mutant females). At birth, homozygous mutants displayed completely formed testes,that were similar in size to their littermates. Conversely, inspection of post-pubertal (P45, P60) male gonads showed a dramatic disparity in their volume in Ebf2^-/- and Ebf2^+/+ males(Ebf2^+/+: 3.29±0.14 mm, n=4;Ebf2^-/-: 2.22±0.05 mm, n=4;P=0.0004). Histological examination of Ebf2^-/-male gonads indicated testis hypoplasia without Leydig cell hyperplasia. This was consistent with hypogonadotropic hypogonadism and argued strongly against an intrinsic defect of the seminipherous tubule. Testes displayed a reduction of the interstitial component, similar to what has been observed in hypogonadal (hpg) mice (Cattanach et al.,1977), which carry an intragenic deletion in theGnrh-gene (Mason et al.,1986). While Sertoli cells from normal andEbf2^-/- mice were morphologically indistinguishable,mutant seminipherous tubules contained fewer dividing spermatogonia(Fig. 5D) compared with wild-type tubules, as assayed by BrdU incorporation. Likewise, mutant epidydimes had a reduced diameter and contained very few sperm cells(Fig. 5F), when compared with control wild-type epidydimes.

The defects observed in the development of GnRH-synthesizing neurons were not the only morphogenetic abnormalities observed in the CNS ofEbf2^-/- mice. In fact, null mutants exhibited substantial defects in cerebellar morphogenesis that are beyond the scope of the present study, and are being actively investigated by our group (L. C. and S. Z.,unpublished). In addition, major defects were also observed in theEbf2^-/- peripheral nervous system.

In a systematic analysis of Ebf2 gene expression, we foundEbf2 expression in the embryonic and postnatal spinal cord, dorsal root ganglia and peripheral nerve (Fig. 6). A detailed analysis of lacZ expression in the mutant mice provided us with new information in addition to that stemming from previous studies. At E12.5, Ebf2 was expressed in dorsal root ganglia(DRG), as reported previously (Garel et al., 1997; Malgaretti et al.,1997). However, its expression (blue signal inFig. 6A) was clearly confined to presumptive satellite cells negative for the heavy chain neurofilament(NF-H). Likewise, dorsal root fibers positive for NF-H were negative for cytoplasmic lacZ, while presumptive peripheral glial progenitors intercalated to those fibers were positive (blue signal inFig. 6B). In agreement with this, E16.5 immature glial progenitors positive for the low-affinity nerve growth factor receptor p75 were also positive for lacZ staining(Fig. 6C). In P15 sciatic nerves from null mice, Ebf2 was expressed in both myelin forming(msc) (arrows in Fig. 6D,E) and non-myelin forming Schwann cells (nmsc) (arrowhead inFig. 6D), while in heterozygous mice Ebf2 expression was detected only in nmsc (arrows inFig. 6F). In the P30 spinal cord, Ebf2 was expressed in dorsal interneurons (lamina 2), in the commissural gray matter and in other as yet undefined neural cells localized in dorsal, and commissural areas of the spinal cord (blue signal inFig. 6G). Finally, inFig. 6H a high magnification of ventral region of the spinal cord revealed colocalization of Ebf2 (in blue) with choline acetyltransferase, a motoneuron marker (brown). Unlike in postnatal development, Ebf2 was not obviously expressed in islet 1-positive motoneuron precursors at E10(Garel et al., 1997) or at subsequent midgestation embryonic stages (L. C. and G. G. C.,unpublished).

In keeping with the evidence of Ebf2 expression throughout glial cell differentiation and in postnatal motoneurons, Ebf2-null sciatic nerves showed various abnormalities in their postnatal development. At gross examination, sciatic nerve trunks came apart spontaneously during surgical manipulation. Histological analysis of Toluidine Blue-stained semi-thin sections revealed that axons, especially large caliber ones, were considerably hypomyelinated in the Ebf2^-/- sciatic nerve(Fig. 7). Likewise,ultrastructural analysis of Ebf2^-/- nerves revealed several abnormalities. As previously described(Webster et al., 1973), by P15 wild-type sciatic nerves display a complete segregation of large axons (≥1μm across) from smaller ones. The former establish a one-to-one relationship with MSC and become myelinated; the latter are fasciculated together by each NMSC and wrapped by cytoplasmic lamellae, remaining unmyelinated (Fig. 8A). Conversely, examination of mutant nerves revealed many unsorted medium to large axons (1-4 μm) that failed to be myelinated. Even in adult nerves(P30), these were fasciculated into anomalous bundles next to smaller axons(Fig. 8B). Occasionally,unsorted axons still wrapped by NMSC cytoplasm became abnormally myelinated(Fig. 8C). Longitudinal sections clearly demonstrated that in some cases myelinated and unmyelinated segments belonged to consecutive internodes of the same axon, constituting a case of segmental dysmyelination (Fig. 8D). In addition, some myelinated fibers featured abnormally thin myelin sheaths (Fig. 8E). With nerve maturation, signs of axonopathy became apparent: starting at about P15,it became possible to observe features of axonal damage, such as accumulations of vesicular, membrane-bound material in the axon(Fig. 8F).

Based on the morphological abnormalities observed in mutant nerves, we conducted electrophysiological tests on adult, age- and gender-matched homozygous mutants (n=9) and controls (n=8). We stimulated the sciatic nerve at the ankle and at the ischiatic notch; the compound motor action potential (cMAP) was recorded from the paw muscles with a pair of needle electrodes to measure motor nerve conduction velocity (NCV). The mean values of NCV, compound motor axon potential (cMAP) latencies and amplitude,and F-wave latencies are summarized inTable 1. The main finding emerging from our studies was a striking reduction (by 40%) of NCV in the knockout group when compared with the wild-type group. All members of the knockout group scored lower in terms of NCV than controls. The mean amplitude of motor responses was also significantly decreased in comparison with controls, although cMAP amplitude data should be interpreted cautiously when dealing with needle recordings. Likewise, although the mean value of F-wave latency was found to be significantly higher in mutant nerves, this finding may simply reflect the overall decrease in NCV. Actually, the finding of measurable F-wave recordings in the mutant group ruled out a conduction block at the level of proximal nerve segments or spinal roots.

Both gain-of-function (Dubois et al.,1998; Pozzoli et al.,2001) and loss-of-function(Garel et al., 1999;Prasad et al., 1998) evidence obtained in various model organisms implicates Ebf genes in key stages of neural development. Although the three known Ebf genes are widely co-expressed in the developing neural tube (Garel et al., 1997), recent evidence has led to the proposal of distinct developmental roles for vertebrate Ebf family members, and specifically differential control of their expression and transactivating functions(Pozzoli et al., 2001).

Our data are consistent with the notion that proteins belonging to the Ebf network play some redundant roles. Overall, Ebf2^-/- mice complete embryonic development successfully, albeit displaying subtle abnormalities that are beyond the scope of the present study. However, our evidence also indicates that some specific morphogenetic processes clearly require Ebf2, despite the fact that other Ebf genes are co-expressed with it. For example, although Ebf1 is coexpressed with Ebf2in migrating GnRH-neurons (Wray,2001), their migration is defective in Ebf2 mutants. Likewise, although both Ebf1 and Ebf3 are co-expressed withEbf2 in the postnatal nerve (L. C. and G. G. C., unpublished), nerve organization and conduction are clearly altered in Ebf2^-/-mice.

The study of Ebf2^-/- embryos indicates thatEbf2 is not required for the birth of GnRH-neurons, or that redundant mechanisms are in place to offset the lack of Ebf2 in this function. However, our mutants show a clear impairment in GnRH-neuron migration from the olfactory epithelium to the diencephalon. In particular, GnRH-neurons exit the nasal mesenchyme later than in wild-type mice, and show signs of degeneration. Those mutant neurons that manage to reach dorsal to the cribriform plate, fail to deflect caudally into the ventral forebrain. Instead, they move dorsally through the rostral telencephalon, possibly along the dorsal branch of the vomeronasal nerve. Altered migration of GnRH-neurons has been observed in other animal models. For example, it has been reported that enzymatic removal of polysialic acid from NCAM-rich vomeronasal nerves at E12 significantly inhibits the migration of GnRH-neurons, without affecting the vomeronasal tract itself (Yoshida et al.,1999). The same authors found that NCAM180 knockout mutants show a shift in the migratory route of GnRH cells, resulting in an excess number of these neurons in the accessory olfactory bulb. Likewise, Dcc (deleted in colorectal cancer, which encodes a netrin-1 receptor) knockout mice revealed GnRH neurons that were misrouted into the cerebral cortex(Schwarting et al., 2001). However, these mutants could not be studied at later stages, because of postnatal lethality. In both genetic mutants cited above, the impairment of GnRH-neuron migration is accompanied by malformation of the vomeronasal nerve. Conversely, no alterations have been observed in Ebf2^-/-vomeronasal nerve axons, either by histology or using specific surface markers(Wolfer et al., 1994;Wray et al., 1994). Likewise,no changes have been found in the olfactory bulbs of null mutants(Fig. 2F). In other mutants(Yoshida et al., 1999), a significant number of GnRH-neurons appear to follow a normal migratory route into the diencephalon and to send projections to the ME, in keeping with the existence of different GnRH-neuron subpopulations or heterogeneity of the migratory pathway (Skynner et al.,1999; Tobet et al.,1996; Wray et al.,1994). In Ebf2^-/- mice we observed a virtually complete lack of immunoreactive fibers in the ME, and a complete lack of immunoreactivity throughout the septohypothalamic region; this evidence, along with the observed expression of Ebf2 in GnRH neurons, strongly supports a fundamental role for the Ebf2 protein in GnRH-neuron development. Previous studies have shown that mutation of another Ebf family member(Ebf1) (Lin and Grosschedl,1995) results in a subpopulation of facial branchiomotor neurons adopting a rhombomere 6-specific short-range migration pattern in rhombomere 5, in the embryonic hindbrain (Garel et al., 2000). Not only are our findings consistent with those observations, but they clearly implicate Ebf2 as a primary factor in the long range developmental migration of an entire population of neuroendocrine neurons. As the Ebf2 gene is expressed both in GnRH and vomeronasal neurons, our results allow no conclusions to be drawn as to whether the observed migration defect is cell-autonomous, or secondary to the lack of axonal cues. Transgenic approaches should make it possible to solve this riddle, shedding further light on the role of Ebf2 in neuronal migration.

To the authors' knowledge this is the first report of a single genetic defect that directly affects GnRH-neuron migration with no gross alterations of their migration substrate. In this respect, our findings differ substantially from those reported in a fetus with Kallmann Syndrome (KS),where the defect in GnRH-neuron migration is accompanied by atrophy of the olfactory bulb (Schwanzel-Fukuda et al.,1989). Several genes have been proposed as potential candidates for other forms of KS or for idiopathic hypogonadotropic hypogonadism (HH). These include GnRH receptors, the LH and FSH genes, andDAX1, etc. (Layman,1999). The evidence reported in the present paper indicates the human ortholog of Ebf2 as a solid functional candidate for genetic studies of HH. Curiously, by computer analysis of the human genome sequence,we determined that the EBF2 gene maps within 450 kb of theGNRH in human chromosomal band 8p21.

In agreement with the observed CNS defects, examination of post-pubertal mutant mice shows clear evidence of testis hypoplasia in the absence of Leydig cell hyperplasia. The presence of testes, albeit hypoplastic, in these animals excludes an essential role for Ebf2 in the prenatal stages of gonadal development, and is coherent with the observation that, during fetal development, differentiation and proliferation of both Sertoli and Leydig cells are independent of endogenous gonadotropins(Baker and O'Shaughnessy, 2001;Griswold, 1993;Lejeune et al., 1998). Not surprisingly, testicular abnormalities seen in Ebf2^-/-mice are superimposable with those observed in hypogonadal (hpg) mice, which carry a null mutation of the Gnrh gene, and, therefore, fail to produce gonadotropins (Mason et al.,1986).

The second most striking abnormality in Ebf2^-/- mutants resides in the peripheral nerve. Ebf2 is expressed in peripheral glial cell progenitors starting at E12.5, and remains expressed throughout birth in immature Schwann cells, and through adulthood in nmsc. The defect in the developmental downregulation of Ebf2 transcription observed in msc from null mutant mice suggests an important role for this gene in late embryonic and postnatal development of those cells. In addition to Schwann cells, Ebf2 is also expressed in postnatal motoneurons. The sciatic nerve of Ebf2 mutants features glial and axonal defects. These include incomplete axon sorting, which results in defective axonal fasciculation. The notion of an evolutionarily conserved role for Ebf family members in axon fasciculation is supported by the abnormal organization observed in the ventral nerve cord of unc-3 mutant nematodes, that carry a mutation within an Ebf homolog(Prasad et al., 1998).

In parallel to a clear defect in axonal sorting,Ebf2^-/- nerves feature signs of segmental dysmyelination and hypomyelination. In fact, the finding of large unsorted axons in transverse nerve sections and that of segmentally unmyelinated axons in longitudinal sections may be linked features of a common defect, whereby large axons fail to establish contacts with msc throughout their length, and as a consequence, are myelinated discontinuously. However, persistent expression of the lacZ reporter gene in Ebf2^-/- msc may also suggest a direct role for Ebf2 in msc terminal differentiation (which may, in turn, explain the finding of hypomyelinated axons in the adult nerve).

Consistent with the observed signs of dysmyelination, electrophysiological tests reveal a 40% decrease in NCV in mutant nerves, suggesting that currents may leak across the unmyelinated membrane, hampering an efficient action potential propagation. Again, in keeping with the segmental nature of the defect, our conduction studies do not disclose any electrophysiological hallmarks of widespread demyelination, such as a temporal dispersion and polyphasia of motor responses, or a significant proximal-to-distal amplitude decrement. Although segmental dysmyelination is expected to slow down conduction, altered expression or defective clustering of Na⁺ or K⁺ channels in the nodal and juxtaparanodal regions, respectively,may offer an alternative or additional explanation. Further studies are required to address this point.

In conclusion, through a genetic approach, we have implicated theEbf2 gene in the pathogenesis of two phenotypic defects of potential relevance in medical genetics, namely hypogonadotropic hypogonadism and peripheral neuropathy with segmental dysmyelination. In addition, our data provide evidence for a pivotal role of Ebf2 in the formation of the neuroendocrine axis, which supports pubertal development, and in several morphogenetic events required for peripheral nerve maturation. Molecular and biochemical studies are now required to dissect the genetic circuits involved in those processes.

We thank F. Zanoni, and C. Mocchetti for their important contribution to this project; A. Ballabio for granting us access to the microinjection facility at the Telethon Institute of Genetics and Medicine; and Giulio Cossu and Monica Vetter for critical reading of the manuscript. The financial support of grants 1216 and E.501 by the Italian Telethon Foundation to G. G. C. was essential for our work and is most gratefully acknowledged. R. M. is funded by Telethon grant E.523. A. C. was the recipient of an EMBO short-term fellowship.