INTRODUCTION

The development of the embryonic spinal cord is marked bythe generation of multiple neuronal cell types that arise atdifferent dorsoventral positions in the neural tube in responseto inductive signals that pattern the neural tube along itsdorsoventral (DV) axis (Tanabe and Jessell, 1996). Recentstudies have revealed that the patterning of neuronal cell typesby these inductive signals is regulated by transcription factorsthat are expressed in restricted populations of dividingneuronal precursors within the ventricular zone (Burrill et al.,1997; Ericson et al., 1997; Briscoe et al., 1999). Thesetranscription factors are thought to function by controllingthe expression of a second class of “cell-type-specific”transcription factors that are expressed in restricted populationsof early postmitotic neurons. Three cell-type-specifictranscription factors that are expressed in the embryonic spinalcord are the homeodomain proteins, EN1, CHX10 and EVX1,which mark three populations of ventral interneurons (Burrillet al., 1997; Ericson et al., 1997; Matise and Joyner, 1997).However, the role that these proteins play in the differentiationprogram of spinal interneurons has not been determined,primarily because the cell types in the embryonic spinal cordthat express these transcription factors are not known. Instead,studies of these and other cell-type-specific transcriptionfactors in the spinal cord have been largely restricted to usingthem as molecular markers to examine the patterning of earlyneuronal cell types in the embryonic neural tube (Pfaff et al.,

1996; Burrill et al., 1997; Ericson et al., 1997; Matise andJoyner, 1997; Lee et al., 1998; Briscoe et al., 1999).

EN1 is a prototypic cell-type-specific transcription factorthat is expressed in a restricted population of early postmitoticventral neurons that are located in two bilateral columns dorsalto motor neurons (Davis et al., 1991; Burrill et al., 1997; Matiseand Joyner, 1997). The expression of EN1 in these neurons iscontrolled by inductive signals that pattern the ventral neuraltube (Pfaff et al., 1996; Ericson et al., 1997). EN1 expressionin the ventral spinal cord is dependent on the activity of thePAX6 transcription factor which is expressed in ventralprogenitors that give rise to EN1 interneurons, and EN1 is nolonger expressed in the ventral spinal cord of Small eye (Pax6−) mutant embryos (Burrill et al., 1997; Ericson et al.,1997). The restricted expression of EN1 in early postmitoticneurons, together with the specific loss of these cells in Pax6mutant mice has led to the hypothesis that EN1 marks asubclass of ventral interneurons and that this interneuronsubclass may be specified in part by EN1 (Burrill et al., 1997;Matise and Joyner, 1997). However, the interneurons in theembryonic spinal cord that express EN1 have not beencharacterized in detail, nor has the function of En1 in theseneurons been determined.

In this study, we have used tau-lacZ as a neuronal tracer tocharacterize in detail the EN1-expressing neurons in theembryonic spinal cord and analyze the role En1 plays in theirdevelopment. We find that EN1 is expressed specifically byipsilaterally projecting association interneurons that extend

4201Development 126, 4201-4212 (1999)Printed in Great Britain © The Company of Biologists Limited 1999DEV1452

During early development, multiple classes of interneuronsare generated in the spinal cord including associationinterneurons that synapse with motor neurons and regulatetheir activity. Very little is known about the molecularmechanisms that generate these interneuron cell types, noris it known how axons from association interneurons areguided toward somatic motor neurons. By targeting theaxonal reporter gene tau-lacZ to the En1 locus, we show thecell-type-specific transcription factor Engrailed-1 (EN1)defines a population of association neurons that projectlocally to somatic motor neurons. These EN1 interneuronsare born early and their axons pioneer an ipsilaterallongitudinal projection in the ventral spinal cord. The EN1

interneurons extend axons in a stereotypic manner, firstventrally, then rostrally for one to two segments where theiraxons terminate close to motor neurons. We show that thegrowth of EN1 axons along a ventrolateral pathway towardmotor neurons is dependent on netrin-1 signaling. Inaddition, we demonstrate that En1 regulates pathfindingand fasciculation during the second phase of EN1 axongrowth in the ventrolateral funiculus (VLF); however,En1 is not required for the early specification of ventralinterneuron cell types in the embryonic spinal cord.

Key words: Engrailed-1, Netrin-1, Pathfinding, Associationinterneuron, Motor neuron, Mouse

SUMMARY

Engrailed-1 and Netrin-1 regulate axon pathfinding by association

interneurons that project to motor neurons

Harald Saueressig, John Burrill and Martyn Goulding*

Molecular Neurobiology Laboratory, The Salk Institute, 10010 North Torrey Pines Rd, La Jolla, CA 92037, USA*Author for correspondence (e-mail: goulding@salk.edu)

Accepted 19 July; published on WWW 7 September 1999

4202

locally toward motor neurons. Axons from the EN1interneurons project ventrally then rostrally in the VLF, wherethey establish local ipsilateral projections to motor neurons inthe embryonic spinal cord. The EN1 axons express DCC andrequire netrin-1 to extend axons ventrally toward motorneurons where they normally pioneer the formation of the VLF.Our studies demonstrate that En1 does not control early cellfate in ventral spinal interneurons, instead En1 controls lateaspects of EN1 interneuron differentiation where it is necessaryfor the organized growth of EN1 axons in the VLF towardmotor neurons.

MATERIALS AND METHODS

Generation and breeding of En1taulacZ “knock-in” miceThe En1tau-lacZ targeting vector is shown in Fig. 1. The vector wascreated by replacing sequences encoding theendogenous methionine initiation codon andfirst 72 amino acids of EN1 with a genecassette encoding the tau-lacZ gene and thePGKneopA gene flanked by loxP sites (seeFig. 1) similar to the strategy described byHanks et al. (1995). CCE ES cells werecultured on neomycin/gancyclovir-resistantLIF-secreting STO fibroblast cells. 2×107 EScells were electroporated with 25 µg of KpnI-linearized DNA using standard settings(Biorad Gene Pulser) and recombinant cloneswere selected in G418 (200 µg/ml active drug)and FIAU (0.2 µM) containing media. ES cellclones were screened for correctly targetedintegration of tau-lacZ using an external3′ genomic DNA probe (HindIII-digestedgenomic DNA probed with 700 bpEcoRI/HindIII DNA fragment). Positiveclones were confirmed by Southern blotanalysis using a 5′ genomic DNA probe(Hanks et al., 1995). Two ES cell clones(C7 and C145) were microinjected intoC57Bl6 blastocysts and reimplanted intopseudopregnant ICR fosters. Chimeric micewere bred with C57Bl6 females to obtaingermline transmission and the En1tlZ allelewas maintained on a mixed 129SV/C57Bl6background. Three chimeric mice transmittedthe En1tlZ allele through the germline andtheir offspring were used for these studies.Genotyping of En1tlZ embryos was performedby Southern analysis as described above usingyolk sac DNA. The En1tlZ/tlZ mutant embryoswere also identified by the loss of midbrainstructures and altered limb morphology. Theaxonal projection pattern of the EN1interneurons was examined in netrin-1-deficient embryos by crossing En1tlZ/+heterozygous mice with netrin-1+/− mice(Skarnes et al., 1995). En1tlZ/+; netrin-1−/−

mutant embryos were obtained by crossingEn1tlZ/+; netrin-1−/+ males with netrin-1−/+females. Genotyping was performed usingPCR primers specific for each gene.

Whole-mount X-Gal stainingMouse embryos were fixed in 0.2%glutaraldehyde in 100 mM phosphate buffer,

pH 7.3, 5 mM EGTA for 30 minutes at 4°C, then rinsed in the samebuffer without glutaraldehyde. X-gal staining was performed aspreviously described (Matise and Joyner, 1997).

Whole-mount antibody staining Mouse embryos were immersion fixed in PBS containing 4%paraformaldehyde (PFA), rinsed twice in phosphate-buffered salinePBS containing 0.5% Triton X-100 (PBST), blocked with PBST/10%NGS and incubated in primary antibody for 1-3 days. The followingantibodies were used in this study: anti-β-galactosidase (1:10,000;Organon Technikon-Cappel), anti-Engrailed (1:3, 4G11 monoclonal,Developmental Hybridoma Bank), anti-TAG1 (1:100; 4D7monoclonal, Developmental Hybridoma Bank), anti-DCC (1:200;Calbiochem), anti-GAD65 (1:1000; monoclonal, DevelopmentalHybridoma Bank), anti-TUJ1 (1:500; BAbCO), anti-Isl1 (1:10;monoclonal, Developmental Hybridoma Bank) and anti-LIM3(1:3000, S. Pfaff). After several washes with PBST, specimens wereincubated for several hours with HRP- or FITC/Cy3-conjugated

H. Saueressig, J. Burrill and M. Goulding

Fig. 1. Generation ofEn1tlZ knock-in mice.(A) Structure of themouse En1 locus. TheEn1taulacZ knock-intargeting vector isshown below.Abbreviations:

B, BamHI; C, ClaI; H, HindIII;R, EcoRI; X, XbaI; neo,PGKneopA G418 resistancecassette used for positiveselection flanked by loxPrecombination sites (triangles);T-lacZ, coding region of thetau-lacZ fusion gene; pA, SV40polyadenylation signal. DNAprobe: A 0.7 kb EcoR1-HindIIIfragment from the 3′ end of theEn1 gene was used to screen EScells for homologousrecombinants by Southernblotting. The mutated allele ofEn1 gives a predicted band of4.5 kb with this 3′ DNA probe

compared to the wild-type allele which gives a 7.5 kb band. (B) Southern blots of HindIIIdigested genomic DNA from 129Sv mice DNA, the CCE ES cell line and two ES cell linesthat have undergone homologous recombination at the En1 locus (C7 and C145). The 7.5 kbwild-type band is present in all four samples. A diagnostic 4.5 kb HindIII band is present inthe two ES cell lines, C7 and C145, where the En1 gene is correctly targeted. (C) Southernanalysis of genomic DNA from a litter of F2 embryos. (Both parents were En1tlZ/+ mice).(D) β-galactosidase (β-gal) expression in E12 En1tlZ knock-in mouse embryos. β-gal isexpressed in the spinal cord (arrow) and in the midbrain (arrowhead). Note the loss of themidbrain En1 expression domain in En1tlZ/tlZ embryos.

4203Pathfinding by EN1 spinal interneurons

secondary antibodies (Jackson ImmunoResearch) for DAB stainingreactions or immunofluorescence, respectively. After final washes,specimens were mounted in PBS/glycerol (1:9) or glycerol/DABCOfor microscopy.

Retrograde tracing experimentsSpinal cords from E11-E12 mouse embryos were dissected usingsharpened tungsten needles and then pinned out straight in Sylgard-coated Petri dishes. Fluorescein-conjugated Dextran (MolecularProbes) was injected into the ventral spinal cord in the area of theVLF and allowed for transport for several hours (see Burrill et al.,1997). After fixation in 4% PFA, whole-mount antibody-labeling wasperformed as described above with antibodies to β-galactosidase.Spinal cords were then separated into two halves and cleared inglycerol/DABCO before being mounted on coverslips for confocalmicroscopy. Double-labeled cells were scored and their positionrelative to the injection site measured.

Confocal microscopy Whole-mounted spinal cords and sections were analyzed using a Zeissconfocal microscope (LSM 510, Version 1.5). Confocal images wereassembled with Adobe Photoshop 4.0.

RESULTS

Generation of En1taulacZ knock-in miceTo determine the identity of the EN1 cells, we generated a“knock-in” mouse that expresses the axonal reporter proteintau-β-galactosidase (Callahan and Thomas, 1994) under thecontrol of En1 regulatory sequences. A tau-lacZ reporter genecassette was recombined into the first exon of En1 in ES cellsand these cells were then injected into blastocysts to generatechimeric mice (see Methods for details, Fig. 1A,B). En1tlZ/+heterozygous mice were analyzed at a number of differentdevelopmental ages, revealing a pattern of tau-β-galactosidase(tau-β-gal) activity that was indistinguishable from the

endogenous expression pattern of En1 (Figs 1D, 2A). Twostripes of tau-β-gal expression were observed in the ventralspinal cord from E9.5 until E16. Expression of tau-β-gal washighest from E9.5 to E13, declining in older embryos toundetectable levels at birth. The insertion of the tau-lacZ geneinto the first exon of En1 was predicted to result in an En1 nullallele and consistent with this, EN1 protein staining was absentin homozygous En1tlZ/tlZ embryos (data not shown).Homozygous En1tlZ/tlZ embryos also exhibited a truncation ofthe midbrain and limb defects (Fig. 1D), as had previously beendescribed for the En1 knockout mice (Wurst et al., 1994;Loomis et al., 1996). No morphological defects were found inEn1tlZ/+ heterozygous embryos demonstrating that the En1gene is haplosufficient.

EN1 expression defines a population of locallyprojecting ipsilateral interneurons in the spinal cordEn1tlZ/+ heterozygous embryos of various ages were stainedwith an antibody to β-galactosidase to characterize the EN1interneurons. Tau-β-gal was localized to the cell bodies and theaxonal processes of the EN1 interneurons, enabling theirmorphology to be visualized in detail (Fig. 2). In E10 and E12embryos, stained EN1 axons could be seen throughout theventral horn at spinal cord and hindbrain levels. In E10embryos, two prominent phases of early axon growth were

Fig. 2. Morphology of the EN1 spinal interneurons. Morphology ofEN1 interneurons in E10.5-12.5 mouse embryos. (A) Transverseconfocal section (2 µm) through an E10 spinal cord showingcolocalization (arrow) of engrailed-1 protein (nuclear, green) andtau-β-galactosidase (axonal, red) (B) Transverse section through thelumbar region of an E10 spinal cord showing the tau-β-galactosidasereporter protein localized to EN1 interneurons and their ventrallydirected axons. An undifferentiated EN1 cell with a neuroepithelial-like morphology (arrowhead) and a ventrally directed EN1 axon(arrow) are indicated. (C) Side view of a whole-mount spinal cordfrom an E10 En1tlZ chimeric embryo showing an EN1 interneuron(asterix) with a rostrally directed axon in the VLF at the level ofmotor neurons. (D) Side view of a whole-mount spinal cord from anE11 En1tlZ/+ embryo showing the morphology of a single EN1interneuron (asterisk). The axon of this interneuron has acharacteristic inverted L-shape and has a rostrally directed axons thatis ~200 µm long. Bar marking the ventral edge of the spinal cord =40 µm. (E) Transverse confocal section (2 µm) through the ventralspinal cord of an E12.5 En1tlZ/+ embryo showing retrogradelylabeled motor neurons (green) and EN1 axons (red) embryo. Thearrows mark axonal processes immediately adjacent to motorneurons. (F) Schematic outlining the typical axon morphology ofEN1 interneurons. EN1 interneurons exhibit two prominent phase ofaxon growth: (1) ventral growth toward the ventral horn and (2)longitudinal growth in a rostral direction in the VLF. Abbreviations:FP, floor plate; IN, interneuron; MN, motor neuron; VLF,ventrolateral funiculus.

4204

discernible. Soon after they are born, the EN1 interneuronsmigrate laterally into the mantle zone where they extend axonsalong a ventrolateral trajectory toward newly born motorneurons in the ventral horn (Fig. 2B). Upon reaching the samedorsoventral level as motor neurons, the EN1 axons turnrostrally where they grow for varying distances in thedeveloping ventrolateral funiculus, immediately lateral tomotor neurons. Since the spinal cord at E10 already containslarge numbers of EN1 interneurons making it difficult tovisualize individual axons, chimeric En1tlZ embryos weregenerated that contained fewer marked neurons in the spinalcord. In E10 chimeric embryos, individual EN1 interneuronswere seen extending axons along a ventrolateral route towardmotor neurons in the ventral horn (Fig. 2C). At E10.5,individual EN1 interneurons could be seen extending axons ina rostral direction for distances of up to 200 µm, immediatelyadjacent to differentiating motor neurons (Fig. 2D). In spinalcords from E10.5 En1tlZ/+ heterozygous embryos, these EN1axons formed a broad non-fasciculated tract immediatelylateral to motor neurons and in older E12-E13 embryos theEN1 axons were distributed in an arc encompassing all somaticmotor neurons (see Figs 7 and 8).

To examine whether axons from the EN1 interneuronsproject into the motor neuron column, fluorescein-conjugateddextran was injected into the forelimb muscles of E12.5En1tlZ/+ embryos to retrogradely label motor neurons, and anantibody to β-galactosidase was then used to visualize EN1axonal processes. Axons from EN1 interneurons were foundintermingled with motor neurons throughout the ventral horn(Fig. 2E). To determine whether these EN1 axons contactedmotor neurons, sections through E12.5 retrogradely labeledEn1tlZ/+ embryos were then imaged in 2 µm optical slices witha confocal microscope. Throughout these sections, axonalprocesses from EN1 interneurons were seen closely abuttingthe soma of motor neurons (Fig. 2E). However, it should benoted that, using this approach, we were not able to determinewhether the EN1 axons directly contact motor neurons.Furthermore, attempts to analyze synapse formation in olderembryos by immuno-EM have been hindered by reducedEn1tlZ expression at later times in development. Nevertheless,our observation that axons from EN1 interneurons terminateadjacent to motor neurons and that their processes closely abutmotor neurons supports our hypothesis that they formconnections with motor neurons. This conclusion is supportedby studies in the embryonic avian spinal cord showingRenshaw cells and a number of last order interneurons expressEN1 (Wenner et al., 1998). Interestingly, no axons fromEN1 interneurons were seen projecting to preganglionicsympathetic neurons in the intermediolateral column, arguingthe EN1 interneurons form connections specifically withsomatic motor neurons.

EN1 interneurons establish early ipsilateralprojections in the ventral spinal cord The EN1 interneurons develop early when few otherdifferentiated neurons are present in the embryonic spinal cordsuggesting that the EN1 interneurons pioneer ipsilateral rostralprojections in the ventral spinal cord. We were thereforeinterested in ascertaining whether the EN1 axons navigateindependently toward motor neurons or use other axons todirect their growth. Previous studies in the mouse and chick

have shown commissural neurons are among the first neuronsthat are born and differentiate in the embryonic spinal cord(Wentworth, 1984; Holley and Silver, 1987), suggestingcommissural axons might provide a scaffold for the ventralgrowth of the EN1 axons. Since TAG1 is expressed on theaxons of commissural neurons, an antibody to TAG1 (4D7)was used to compare the growth of commissural neurons withEN1 axon outgrowth. When the EN1 axons first beginprojecting at E10, little or no TAG1 expression was seen in thedorsal spinal cord that could be attributed to commissuralaxons (Fig. 3A). Although, TAG1 expression was observed inthe ventral horns in the vicinity of the EN1 axons, furtheranalysis showed this expression to be restricted to motorneurons and their axons. These axons exit the ventral spinalcord and are therefore unlikely to function as a substratum forEN1 axon extension. In older E10.5 embryos, TAG1commissural axons were seen projecting directly toward thefloor plate along a characteristic medial pathway (Fig. 3C),whereas the EN1 interneurons extended axons along a morelateral route toward motor neurons (Fig. 3B,C). Furthermore,our analysis of TAG1 and β-gal staining in En1tlZ/+ embryosshows the first EN1 axons reach the ventral horn and turnrostrally before any commissural axons have crossed the floorplate (data not shown).

To investigate whether other axons provide a substratum forgrowing EN1 axons, we then compared the onset and timingof EN1 axon outgrowth with the development of other early-born neurons in the spinal cord using an antibody to neuron-specific β-tubulin (Yaganuma et al., 1990; Easter et al., 1993;Burrill et al., 1997). En1tlZ/+ embryos were stained withantibodies to β-gal and β-tubulin (TuJ1) and the trajectories ofventrally directed EN1 axons were then imaged by confocalmicroscopy and compared with other axons marked by β-tubulin expression. None of the individual EN1 axonsexamined (n=38) were seen fasciculating with axons fromTuJ1+/EN1− interneurons or with other EN1 axons, arguingthat the first EN1 axons that grow toward the ventral horn doso independently (Fig. 3D, also see Fig. 2B). In addition, weexamined the early trajectory of isolated EN1 axons as theyturned rostrally within the ventral spinal cord (Fig. 3E,F). Onceagain, no other longitudinally aligned axons were seen in theimmediate vicinity of EN1 axons either during or after theyhad turned rostrally in the VLF (n=27). Consequently, whenthe EN1 axons enter the VLF, they do not grow together withother axons, nor do they join a pre-existing axon scaffold.Although a small number of EN1− longitudinal processes wereseen within the VLF at E10, in most cases these were wellseparated from the EN1 axons. Our finding that relatively fewlongitudinally projecting EN1− axons are present in the ventralspinal cord when EN1 interneurons begin extending axonsrostrally, leads us to conclude the EN1 interneurons helppioneer the formation of the VLF in the mammalian spinalcord.

Netrin-1 controls the ventral growth of ipsilateralinterneurons in the spinal cordThe EN1 interneurons project axons ventrally albeit by adifferent pathway than commissural neurons that extend axonsmedial to motor neurons en route to the floor plate. Whileprevious studies have demonstrated an essential role for netrin-1 in guiding commissural neurons toward the floor plate, its

H. Saueressig, J. Burrill and M. Goulding

4205Pathfinding by EN1 spinal interneurons

role in diverting the growth of axons from association neuronshas not been examined due to an inability to specifically markthese cells. We were therefore interested in determiningwhether ipsilaterally projecting association neurons alsorespond to netrin-1 to guide their axons ventrally toward motorneurons, since low levels of netrin-1 have been reported to bepresent throughout the ventral spinal cord (Kennedy et al.,1994; Serafini et al., 1996). Spinal cords from E10.5 En1tlZ/+heterozygous embryos were first examined to determinewhether the netrin receptor Deleted in Colorectal Cancer(DCC) is expressed on the axons of the EN1 interneurons (Fig.4). Both proteins were found co-localized in EN1 neurons andtheir axons (Fig. 4C, see arrow) at E10.5. Interestingly, DCCwas not expressed by immature neuroepithelial-like EN1 cellsthat lacked a ventrally directed axon (Fig. 4C, arrowhead), norwas it expressed at E12.5, when EN1 interneurons no longerextend axons ventrally toward the ventral horn (data notshown).

To test directly whether netrin-1 regulates the ventral growthof ipsilaterally projecting axons, En1tlZ/+ mice were crossedwith netrin-1 (net-1+/−) heterozygous mice (Skarnes et al.,1995). Analysis of the axonal morphology of EN1 interneuronsin second generation progeny from these crosses revealed adramatic difference in the projection pattern of the EN1interneurons in spinal cords from En1tlZ/+; net-1−/− mutantembryos. In E11 wild-type and net-1 heterozygous embryos,all the EN1 neurons project ventrally to the level of the motorneurons before turning rostrally in the VLF (Fig. 5A). Incontrast, the majority of these axons were either truncated orturned rostrally before reaching the VLF in homozygous net-1−/− embryos (Fig. 5B,C). While a few EN1 axons exhibitednear normal projection patterns in net-1−/− embryos (Fig. 5C,arrowheads), the vast majority of the ventrally projecting EN1axons possessed short ventral processes with an average axonlength of 60-70 µm compared with 140-150 µm in wild-typeembryos (Fig. 5D). A number of longitudinally directed axonswere also seen at the same dorsoventral level as the EN1 cellbodies (see arrow in Fig. 5E) and among these, we were ableto trace individual axons that projected in a rostral direction.This observation demonstrates netrin-1 is necessary for the firstphase of EN1 axon growth and that it seems to play no role inthe decision to project rostrally or caudally. Furthermore, thisresult suggests that rostrocaudal cues for association axons arepresent throughout the spinal cord and are not restricted to theVLF. We also observed small numbers of EN1 neurons withdorsally directed axons in netrin-1 mutant spinal cords (Fig.5F), further supporting our conclusion that netrin-1 is requiredfor the ventral growth of ipsilateral axons, since dorsallydirected EN1 axons were never observed in wild-type embryos.

En1 does not control early aspects of ventralinterneuron cell fate Previous functional studies of engrailed (en) in invertebrateshave revealed a role for en in controlling neuronal cell fate andneurotransmitter phenotype. In the grasshopper embryo, encontrols the differentiation of glia and neurons that arise fromthe midline neuroblast (Condron et al., 1994) and en is alsorequired for the development of serotonergic neurons from theNB 7-3 neuroblast in Drosophila (Lundell et al., 1996). Theexpression of EN1 in a discrete population of early postmitoticinterneurons prompted us to ask whether En1 specifies early

interneuron cell fate or neurotransmitter phenotype in theembryonic spinal cord. E12 En1 mutant embryos wereexamined to determine whether loss of the En1 gene causes theEN1 interneurons to assume an alternative cell fate, asevidenced by a change in their early axonal morphology. Theaxon trajectories of EN1 interneurons were examined byretrogradely labeling axons in the ventral spinal cord with afluorescein-conjugated dextran tracer. In En1tlZ/+ heterozygousembryos, retrogradely labeled EN1 interneurons were foundwith axons extending up to 500-600 µm or two spinal cordsegments in a rostral direction (Fig. 6A), and many of theseappeared to project intrasegmentally (<250 µm). While the vastmajority of EN1 interneurons labeled in this manner possesseda rostrally directed axon, a few caudally projecting EN1 spinalinterneurons (n=3/178) were also labeled. A carefulcomparison of retrogradely labeled EN1 interneurons inheterozygote (n=178) and En1 mutant spinal cords (n=179) didnot reveal any dramatic change in the projections of the EN1interneurons that would indicate a change in cell fate (Fig. 6A).Unlike the CHX10/LIM3 interneurons, the EN1 interneuronsdid not project over several spinal cord segments, nor did theyproject contralaterally like the EVX1 interneurons. A smalldifference, however, was seen in the axon length ofretrogradely labeled cells in En1 mutant spinal cords, and thisprimarily involved an increase in the number of neurons thatproject 200-300 µm together with an associated loss of EN1cells with short-projecting axons (<100 µm). Whether thisdifference is due to the pathfinding defects observed in En1mutant embryos or instead involves the loss of a population ofshort-projecting EN1 interneurons remains to be determined.

The above retrograde analysis of the EN1 projectionsconfirms our hypothesis that the EN1 interneurons projectlocally to motor neurons and suggests En1 does not controlearly aspects of cell fate in these cells. To further examinethis, we asked whether En1 controls the expression ofneurotransmitters such as GABA in these interneurons.Previous physiological studies in the spinal cord indicate thatmany of the locally projecting interneurons present in theventral spinal cord are GABAergic. To first ascertain whetherthe EN1 interneurons synthesize GABA, we examined theexpression of GAD65 in the embryonic spinal cord. Transversesections through the spinal cord of E11 En1tlZ/+ heterozygousembryo, when stained with antibodies to GAD65 and β-galactosidase, showed coexpression of GAD65 in all of theEN1 neurons as well as in a population of commissural neuronsimmediately adjacent to the EN1 interneurons (Fig. 6B). Toascertain whether EN1 is necessary for these interneurons todifferentiate into GABAergic neurons, GAD65 expression wasalso analyzed in spinal cords from E11 En1tlZ homozygousembryos. No difference was observed in the overall pattern ofGAD65 staining at E11 in the En1− mutants and, in particular,expression of GAD65 was unchanged in the EN1 interneurons(Fig. 6C). Consequently, En1 function is not required forestablishing the early GABAergic phenotype of the EN1interneurons.

EN1 interneurons exhibit fasciculation andpathfinding defects in En1 mutants Our morphological analysis of the EN1 interneurons showsthey project axons locally to the ventral horn where theyinnervate motor neurons within 1-2 segments of their cell

4206

bodies. Consistent with this, the EN1 axons upon entering theVLF do not fasciculate and are instead distributed throughoutthe funiculus in a parallel array of single axons that extendrostrally for short distances (Fig. 7). As early as E10.5, EN1interneurons with a typical inverted L-shape morphology wereseen at all levels of the spinal cord in En1tlZ/+ embryos. Whileaxons with roughly similar trajectories were also observed inEn1tlZ−/− embryos, clear differences could be seen betweentheir trajectories and those present in heterozygous En1tlZ/+embryos. In E10.5 En1tlZ−/− embryos, EN1 axons were oftenseen to turn dorsally soon after entering the VLF (Fig. 7B,inset). Furthermore, even at these early stages, axon bundlescontaining three or more EN1 axons were found throughoutthe VLF in E10.5 En1tlZ/tlZ mutant spinal cords (see arrows inFig. 7B), in sharp contrast toheterozygous En1tlZ/+ embryoswhere EN1 axons do not fasciculatewith each other (Fig. 7A). Whenspinal cords from older E12heterozygous embryos werescanned by confocal microscopy,sagittal sections through the VLFrevealed individual EN1 axonsaligned longitudinally in a parallelarray at the lateral edge of theventral horn (Fig. 7C). In contrast,many EN1 axons were bundledtogether into fascicles containing10 or more axons in En1tlZ/tlZ

homozygous embryos (see arrows

in Fig. 7F, c.f. 7E). In addition to these fasciculation defects,the EN1 axons also exhibited markedly disorganizedtrajectories within the VLF of E12 En1 mutants, giving thema matted appearance when viewed in sagittal sections (see Fig.7D).

The abnormal morphology of the EN1 axons in homozygousEn1 mutant embryos led us to ask whether these defects alterthe normal development of EN1 projections to somatic motorneurons. Motor neurons in E12.5 En1tlZ/+ and En1tlZ/tlZ

embryos were visualized using an antibody to ISL1 and thedistribution of EN1 axons relative to motor neurons wasanalyzed in sections stained with an antibody to β-gal. Intransverse sections through heterozygous En1tlZ/+ embryos,axons from EN1 interneurons were distributed throughout the

H. Saueressig, J. Burrill and M. Goulding

Fig. 3. Analysis of early axonoutgrowth by EN1interneurons. (A-C) Spinalcords from E10-10.5 En1tlZ/+heterozygous embryos stainedwith antibodies to β-gal (red)and TAG1 (green). (A) Dorsaland (B) ventral view of crosssection through an E10 spinalcord. At the time that EN1axons first begin projectingtoward the ventral horn only afew weakly stained TAG1+

commissural axons areobserved in the dorsal spinalcord. (C) Cross sectionthrough an E10.5 spinal cord.The EN1 axons (red, arrows)and commissural axons (green,arrowheads) project alongdifferent pathways in theventral spinal cord.(D-F) Spinal cords from E10En1tlZ/+ heterozygousembryos stained withantibodies to β-gal (red) and β-tubulin (green). (D,E) Sagittalconfocal section through an E10 spinal cord. The arrowheads in D and E mark isolated EN1 axons (orange) that are growing ventrally towardthe ventral horn. Note the absence of closely associated EN1− (green) axons. A more medially located commissural axon is marked with anarrow. (F) A 20 µm collapsed confocal Z-series showing EN1 axons turning in the ventral horn. The large arrowhead in F indicates an EN1axon that is turning and beginning to grow rostrally in the VLF. Very few longitudinally projecting EN1− axons are present in the VLF at thistime (small arrowhead).

Fig. 4. DCC is expressed in EN1 interneurons. A transverse section through ventral half of an E10.5mouse spinal cord stained with antibodies to DCC and β-gal. (A) DCC expression; (B) β-galexpression; (C) merged image showing DCC expression in EN1 axons in the VLF (arrow). DCC isnot expressed by undifferentiated EN1 cells that have not extended an axon (arrowhead).

4207Pathfinding by EN1 spinal interneurons

VLF in a punctate manner, forming a crescent-shaped axontract encompassing both the medial motor column (MMC, Fig.8A) and the lateral motor column (LMC, data not shown).When spinal cords from En1tlZ/tlZ homozygous embryos wereexamined, we observed a significant difference in thedistribution of EN1 axons within the VLF. Processes from EN1interneurons were markedly absent from ventromedial regionsof the VLF (Fig. 8B,D). In particular, few if any EN1 axonswere seen adjacent to the medial-most motor neurons in theMMCm. This loss of EN1 axons immediately adjacent to themedial MMCm was observed at all axial levels. In addition toreduced numbers of longitudinally projecting EN1+ axonsadjacent to the medial MMCm (Fig. 8D), spinal cords from En1mutant embryos also contained far fewer EN1 axons within theMMCm, when compared to age-matched controls (Fig. 8E).Although fewer EN1 axons were seen in the medial half of theMMCm, an increase was seen in the number of EN1 axons inthe lateral half of the MMCm suggesting the EN1 axons thatnormally grow toward the medial MMCm instead enter thelateral MMCm in the En1− mutants.

DISCUSSION

EN1 axons pioneer local ipsilateral projections tomotor neurons and are guided by netrin-1 The EN1 interneurons are among the first postmitoticinterneurons to develop in the embryonic spinal cord wherethey begin differentiating shortly after the first motor neuronsare born (Pfaff et al., 1986 and Fig. 3). We observe that theearliest EN1 axons grow independently of other neighboring

Fig. 6. Analysis of projection and neurotransmitter phenotype ofEN1 interneurons. (A) Double-labeled cells were identified byimmunostaining flourescein-dextran retrogradely filled cells with anantibody to β-galactosidase. Whole spinal cords were imaged with aZeiss confocal microscope and six independent single injection sitesat brachial level were analyzed for each genotype. Cell numbers andtheir relative distance in µm from the edge of the injection site areindicated. Positive numbers indicate cells with rostrally directedaxons. (B,C) Transverse sections through the spinal cords of an E11En1tlZ/+ heterozygote (B) and an E11 homozygous En1tlZ/tlZ mutantlittermate (C) showing coexpression of GAD65 (green) in EN1interneurons (see arrows). Also note the expression of GAD65 in theEN1 axons in the VLF and in commissural axons (arrowheads)Abbreviation. VLF, ventrolateral funiculus.

Fig. 5. Netrin-1 is a chemoattractant for ipsilateral axons in the spinal cord. (A) Sagittal confocal section through the ventral spinal cord of anE11 En1tlZ heterozygous embryo showing β-gal expression in axons that project to the VLF (arrow). (B,C) Sagittal confocal sections throughthe spinal cord of an E11 En1tlZ/+; net-1−/− embryo showing the widespread loss of EN1 axons that project ventrally. A few EN1 axons exhibita near normal ventral projection pattern (arrowheads). Many longitudinally directed axons are present at the same DV level as the EN1 cellbodies (arrows). (D) Sagittal section showing many truncated EN1 axons projecting into the ventral spinal cord of an E11 En1tlZ/+; net-1−/−

embryo. (E,F) Sagittal confocal sections from an E11 En1tlZ/+; net-1−/− spinal cord at the level of the EN1 interneuron cell bodies showingexamples of mis-routed EN1 axons projecting rostrally (E) and dorsally (F). Dashed lines mark the ventral edge of the spinal cord.

4208

axons toward motor neurons (Figs 2, 3) indicating they respondautonomously to guidance cues in the embryonic spinal cordto establish an ipsilateral projection pathway in the ventralspinal cord. The CHX10/LIM3 interneurons also differentiate

at the same time as the EN1 interneurons and have axons thatproject longitudinally in the VLF; however, our results indicatethese cells are not required for EN1 axon guidance, since EN1axons do not grow together with other axons at E10, including

H. Saueressig, J. Burrill and M. Goulding

4209Pathfinding by EN1 spinal interneurons

axons from CHX10/LIM3 interneurons (Fig. 3, data notshown). It is also unlikely that commissural neurons pioneerthe formation of the VLF, since the first EN1 axons that turnrostrally in the ventral spinal cord do so before anycommissural axons have crossed the floor plate (Fig. 3). Ourfindings do not support the presence of an early bornpopulation of ventral interneurons that help guide the EN1axons that are like the primitive longitudinal (PL) cells, whichhave been proposed to pioneer the formation of the VLF in thechick (Yaganuma et al., 1990), and such cells have notbeen found in the spinal cords of rodents (H. S. and M. G.,unpublished observations). In addition, our observation that theEN1 axons still project rostrally in netrin-1 mutant embryos(Fig. 5), albeit at more dorsal locations in the spinal cord,further argues against the presence of localized guidepost cellsthat direct the rostral growth of the EN1 axons. While the EN1interneurons are among the first cells to extend axonslongitudinally in the VLF, it is not clear whether they serve aspioneers for later developing interneurons as has beendescribed elsewhere (Caudy and Bentley, 1986; Bastiani andGoodman, 1986; Kuwada, 1986).

This study provides the first evidence that netrin-1 is essentialfor guiding axons from ipsilaterally projecting associationneurons along a circumferential pathway toward motor neurons.Netrin-1 was first identified as a chemoattractant forcommissural neurons in vertebrates (Kennedy et al., 1994;Serafini et al., 1994) and, in both vertebrates and in Drosophila,netrin-1 signaling is required for axons to cross the ventralmidline and establish the ventral commissure (Serafini et al.,1996; Kolodzeij et al., 1996; Mitchell et al., 1996). Our results,together with other studies that have analyzed pathfinding bycommissural neurons (Shirasaki et al., 1995; Serafini etal.,1996; Fazeli et al., 1997), demonstrate netrin-1 functions asa chemoattractant for multiple interneuron cell types in thevertebrate spinal cord. This is analogous to the describedactivity of the C. elegans netrin-homologue UNC6, whichtogether with its receptors UNC5 and UNC40 controls thecircumferential migration of multiple pioneer axons in C.elegans embryos (Hedgecock et al., 1990). We also find thatcommissural neurons and ipsilateral interneurons that arelocated close to each other in the ventral spinal cord extendaxons along different pathways in the ventral spinal cord in

response to netrin-signaling (T. Kagawa and M.G., unpublishedobservations). Consequently, while both cell types are capableof responding to similar levels of netrin-1 by extending axonsventrally, the pathways that are taken by these netrin-responsiveaxons differ significantly. Interestingly, EN1 expression alsoappears to overlap with that of Unc5H1, a mammalianhomologue of the UNC-5 receptor that mediateschemorepulsion by UNC-6 in C. elegans (Leonardo et al.,1997). However, signaling via the netrin/DCC/UNC5 pathwayalone appears unlikely to account fully for the circumferentialgrowth of the EN1 axons and their avoidance of the ventralmidline, since 2-3% of the EN1 axons exhibit near normaltrajectories in netrin-1 mutant embryos (Fig. 5) and no EN1axons are observed crossing the floor plate, even thoughcommissural axons are still able to do so in netrin-1 and DCCmutant embryos (Serafini et al., 1996; Fazeli et al., 1997).

The mechanisms that control this rostral turn and inhibit theEN1 axons from crossing the midline are not known. Recently,it has been demonstrated that the Drosophila protein,roundabout, and the C. elegans homologue, SAX3, arecomponents of a guidance system that regulates midlinecrossing of axons (Kidd et al., 1998; Zallen et al., 1998). Basedon the expression patterns of the mammalian roundabouthomologues, robo-1 and robo-2, and their ligands, slit-1 andslit-2 (Brose et al., 1999), it appears that this guidancemechanism is conserved in higher vertebrates and may repelEN1 axons from the midline. The presence of a parallelrepulsive signaling pathway during the initial ventral phase ofEN1 axon growth would also explain why EN1 interneuronstake a more lateral route before turning rostrally some distancefrom the floor plate. In particular, slit-2 is expressed at E10 inthe ventral horn and may prevent the EN1 axons fromprojecting directly toward the floor plate like commissuralneurons. None of the above mechanisms can account for thestereotypical rostral turning of EN1 interneurons suggestingadditional signals must control this decision. Several genesinvolved in rostral/caudal axon elongation have been describedin C. elegans and a similar independent guidance mechanismmay account for the rostral turning behavior of the EN1 axons(McIntire et al., 1992; Wightman et al., 1997). Our observationthat EN1 axons that fail to project ventrally in netrin-1 mutantembryos still turn rostrally (Fig. 5), argues that these cues forrostrocaudal guidance are distributed throughout the DV axisof the embryonic spinal cord.

EN1 expression defines a class of locally projectingspinal interneurons in the embryonic spinal cordOur studies showing EN1 identifies a population ofinterneurons with a unique axonal projection pattern and is oneof only four early ventral interneuron cell types in the ventralspinal cord, argues for an early limited diversification ofneurons during development. Whereas neurons that co-expressCHX10 and LIM3 also extend axons ipsilaterally in the VLF,their axons project for long distances (>3-5 segments) in bothcaudal and rostral directions, in contrast to the EN1interneurons that have short axon trajectories (J. B. and M. G.,unpublished results; K. Sharma and S. Pfaff, pers. comm.).This limited early diversification of ventral interneurons in theembryonic spinal cord appears to be a simple mechanism thatgenerates a primitive interneuron axonal scaffold that can thenbe used to “wire up” the spinal cord. Rather than initially

Fig. 7. EN1 axons are disorganized and fasciculate abnormally inEn1 mutant embryos. (A,B) Flattened sagittal Z-series of the ventralspinal cord of an E10.5 En1tlZ/+ embryo (A) and an E10.5 En1tlZ/tlZ

embryo (B) showing the early axon morphology of the EN1interneurons. In the inset are examples of hook-shaped axons thatturn dorsally in En1− embryos (arrowheads). Fasciculated bundles ofEN1 axons (arrows) are also present in En1− mutant spinal cords.(C,D) Sagittal confocal sections through the VLF of an E12 En1tlZ/+heterozygote embryo (C) and an E12 En1tlZ/tlZ homozygote mutantembryo (D). Note the marked difference in the trajectories of theEN1+ axons. Normally the axons are oriented longitudinally in theventrolateral funiculus (VLF in C); however, in homozygous En1−

embryos these axons are more randomly oriented. (E,F) Transversesections through the ventral spinal cord of an En1tlZ heterozygoteembryo (E) and a En1tlZ/tlZ homozygous embryo (F) at E12.5. Notethe fasciculated bundles of EN1 axons that are present inhomozygous En1tlZ/tlZ embryos (arrows in D and F) are absent inheterozygous embryos (arrow in C). The arrowhead in E indicatesthe region of the VLF that lacks EN1 axons in homozygous embryos.

4210

generating many different neuronal cell types, neurons in thespinal cord would be initially segregated into groups of cellswith similar axonal projections that then differentiate furtherto form the many physiologically distinct interneuron cell typesthat populate the adult spinal cord. According to this model,the CHX10/LIM3 interneurons would give rise to interneuronsubtypes that innervate motor neurons over multiple spinalcord segments and function in propriospinal reflexes pathways.In contrast, neurons derived from the EN1 interneurons woulddifferentiate into neurons that form local connections withmotor neurons, either intrasegmentally or between adjacentsegments, thereby generating local reflex circuits thatcoordinate motor neurons within a few spinal cord segments.

Although the mature morphologies and functionalcharacteristics of the EN1 interneurons have not beendescribed, studies in the E10 chick spinal cord indicateRenshaw cells and other last order inhibitory interneuronsexpress EN1 (Wenner et al., 1998). Renshaw cells are located

close to motor neurons in the ventral horn and synapse withthem, forming a circuit that mediates the recurrent inhibitionof motor neurons (Renshaw, 1946; Eccles et al., 1954). Theadult phenotype of Renshaw cells is consistent with theexpression of GAD65 in the EN1 interneurons, and with thedistribution and morphology of the EN1 interneurons (see Figs2, 6). Embryonic EN1 interneurons also share a number ofmorphological similarities with Ia inhibitory interneurons inthe adult cat spinal cord (Jankowska and Lindstrom, 1972). Iainhibitory interneurons are located dorsomedial to motorneurons, have short axonal processes in the VLF and makelocal projections to motor neurons, raising the possibility thatIa inhibitory interneurons also develop from embryonicinterneurons that express EN1.

En1 regulates late aspects of ventral interneuronphenotypeIn contrast to previous studies showing engrailed is requiredfor the differentiation of specific neural cell types ininvertebrates (Condron et al., 1994; Lundell et al., 1996), wefind no evidence that En1 controls early interneuron cell fatein the embryonic spinal cord. Previous studies show that theEN1 interneurons are still present in the spinal cords of En1−

mutant mice (Matise and Joyner, 1997); however, these studieswere unable to provide any detailed analysis of changes to cellfate. We observe that, in homozygous En1tlZ/tlZ null embryos,the EN1 interneurons do not adopt the morphologicalcharacteristics of other ventral interneurons, nor do theyectopically express markers found in other interneuron celltypes (data not shown). Furthermore, we find no evidence of achange in their neurotransmitter phenotype. GAD65expression is normal in the spinal cords of En1 null embryos(Fig. 6) and other neurotransmitter-specific enzymes, e.g.ChAT and TH, are not ectopically expressed in the EN1interneurons (data not shown). The lack of early cell fatechanges in the En1 mutant embryos contrasts with the markedalterations to cell fate that occur when members of the LIM-homeodomain gene family are inactivated in mice. In micelacking Lhx3/4, ventral motor neuron subtypes are lost orconverted to dorsal motor neuron subtypes (Sharma et al.,1998), while loss of Isl1 function causes a complete failure ofsomatic motor neuron differentiation (Pfaff et al., 1996). TheEN1 interneurons also fail to develop in Isl1 mutant mice;however, this is most likely due to the extensive cell death thatoccurs throughout the ventral spinal cord in these embryos. Ourfinding that En1 does not determine cell fate differs markedlyfrom the functions that have been described for transcriptionfactors that are expressed in the ventricular zone such as PAX6and NKX2.2, which have essential roles in determining the fateof the EN1 and SIM-1 ventral interneurons, respectively(Ericson et al., 1997; Briscoe et al., 1999). MNR2, which isalso transiently expressed in dividing motor neuron precursors,may also function as a motor neuron determination factor(Tanabe et al., 1998). Our study shows that, although EN1 isexpressed early in undifferentiated postmitotic neurons with aneuroepithelial morphology, it does not control an early step inthe differentiation program of these cells.

An important finding of this paper is that En1 controlsfasciculation and axon pathfinding in association interneuronsthat project to somatic motor neurons. Normally these EN1axons do not fasciculate with one another, a behavior that is

H. Saueressig, J. Burrill and M. Goulding

Fig. 8. EN1 axons do not project to the medial MMCm inhomozygous En1 mutant embryos. (A-D) Spinal cords from E12.5En1tlZ/+ heterozygous embryos stained with antibodies to β-gal (red)and either ISL1 (A,B; green) to mark all motor neurons, or LIM3(C,D; green) to mark motor neurons in the MMCm. The position ofthe MMCm is noted in A and B. Note the absence of EN1 axonsadjacent to the medial MMCm in En1tlZ/tlZ homozygous embryos(arrow in B). EN1 axons (arrows) are present in the medial MMCmin En1tlZ/+ heterozygous embryos, but not in En1tlZ/tlZ homozygousembryos. (E) EN1 axonal processes were counted in 10 µm sectionsthrough the thoracic spinal cord of E12.5 heterozygous andhomozygous En1tlZ embryos. The results represent compositenumbers for 16 thoracic spinal cord sections. Error bars are shown.

4211Pathfinding by EN1 spinal interneurons

observed in regions of the developing CNS where axons areactively establishing connections with other neurons. Evidencethat axon defasciculation is important for axon branching andconnectivity has come from studies in both invertebrates andvertebrates (Stoekli and Landmesser, 1995; Daston et al., 1996;Fambrough and Goodman, 1996; Tuttle et al., 1998).Association spinal interneurons, which include the EN1interneurons, establish synapses with multiple motor neurons“en passant” over the length of their axons (Vaughn andGreishaber, 1973). In both mouse and rat embryos,synaptogenesis involves the extensive invasion of motor neurondendrites into the VLF, and these dendrites intercalate with andform synapses with the axons of association neurons (Vaughnet al., 1977). Consequently, the defasciculated behavior of theEN1 axons is likely to be important for promoting contactsbetween axons and motor neuron dendrites, thereby facilitatingsynapse formation. The EN1 axons are abnormally bundled inEn1− mutant embryos forming fascicles that contain 10 or moreaxons (Fig. 7 and 8). As a result, synapse formation betweenthe EN1 association interneurons and motor neurons may besignificantly impaired in En1− mutant mice.

In addition to the fasciculation phenotype, the EN1 axonsfail to pathfind correctly toward somatic motor neurons in En1−

mutant embryos (Fig. 8) and, as a result, two differences in thetrajectories of the EN1 axons are seen. Firstly, the EN1 axonsmeander dorsoventrally within the VLF and, secondly, there isa marked loss of EN1 axons projecting to motor neuronslocated in the medial MMCm. While the effects of thesepathfinding defects on connectivity have not yet beendetermined, the aberrant trajectories of the EN1 axons in theVLF may have profound effects on the specificity of motorneuron innervation. During embryonic development, somaticmotor neurons are organized into pools of cells that occupydiscrete domains in the ventral horn (Landmesser, 1978; Smithand Hollyday, 1983; Gutman et al., 1993; Lin et al., 1998) andmany of the inhibitory local circuit interneurons that innervatespecific motor neuron pools do so in a topologically orderedmanner (Eccles et al., 1957; Jankowska and Roberts, 1972).The majority of the early synapses between associationneurons and motor neurons are either axiosomatic or arelocated on the proximal dendrites (Vaughn et al., 1977),indicating initial synapse formation is largely constrained bythe axonal trajectories of the association interneurons. TheEN1 axons project longitudinally in the VLF and, since motorneurons are also organized into columns that run longitudinallyin the spinal cord, this indicates that individual EN1 neuronsmay preferentially form synapses with pools of motor neuronsthat innervate the same or synergistic muscles. In view of thedorsoventral meandering of the EN1 axons that occurs in En1mutant spinal cords, this early organization of synapsesbetween the EN1 interneurons and pools of motor neurons islikely to be altered in En1 mutant embryos.

Our finding that EN1 axons no longer project to motorneurons in the medial MMCm in the En1− mutant spinal corddemonstrates En1 function is essential for EN1 interneurons topathfind their way to a subset of somatic motor neurons. Thesemotor neurons are located in the medial half of the MMCm andspecifically innervate the deep muscles of the back (Smith andHollyday, 1983). Furthermore, dendrites from these motorneurons extend into the ventromedial region of the VLF whereEN1 axons are no longer present in the En1 mutant embryos.

This suggests that EN1 may be part of a combinatorialtranscription factor code that specifies connectivity betweenlocally projecting interneurons and pools of somatic motorneurons. Alternatively, En1 may control the later emergence ofa subclass of interneurons that innervate motor neurons locatedin the medial half of the MMCm and, in En1 mutants, these aretransformed to interneurons that innervate more laterallylocated motor neurons.

We would like to thank Marc Tessier-Lavigne for the netrin-1mutant mice, Chris Callahan and John Thomas for the tau-lacZplasmid, Alexandra Joyner for the En1 targeting vector, SteveO’Gorman for ES cells, Yelena Marchuk for blastocyst injections andMarc Olivier and Shad Schidel for technical assistance. We would alsolike to thank Alexandra Joyner, Chris Kintner, Tom Jessell, EliseLamar, Greg Lemke, Horst Simon, Chuck Stevens, John Thomas,Stefan Thor, Detlef Weigel and Todd Zorrick for their advice andcomments on the manuscript. This work was supported by grants fromNIH (M. G.), the Pew Charitable Trusts (M. G.) and the Fritz BurnsFoundation (M. G.). Harald Saueressig was supported by a grant fromthe Deutsche Forschungsgemeinschaft (SA 652/2-1) and John Burrillwas supported by a NRSA Postdoctoral Fellowship.

REFERENCES

Bastiani, M. and Goodman, C. S. (1986). Guidance of neuronal growth conesin the grasshopper embryo. III. Recognition of specific glial pathways. J.Neurosci. 6, 3542-3551.

Brose, K., Bland, K. S., Wang, K. H., Arnott, D., Henzel, W., Goodman,C. S., Tessier-Lavigne, M. and Kidd, T. (1999). Slit proteins bind roboreceptors and have an evolutionarily conserved role in repulsive axonguidance. Cell 96, 795-806.

Briscoe, J., Sussel, L., Serup, P., Hartigan-O’Conner, D., Jessell, T. M.,Rubenstein, J. L. R. and Ericson, J. (1999). Homeobox gene Nkx2.2 andspecification of neuronal identity by graded Sonic hedgehog signallingNature 398, 622-627.

Burrill, J. B., Moran, L., Goulding, M. and Saueressig, H. (1997). PAX2is expressed in multiple spinal cord interneurons, including a population ofEN1+ interneurons that require PAX6 for their development. Development124, 4493-4503.

Callahan, C. A. and Thomas J. B. (1994). Tau-β-galactosidase, an axontargeted fusion protein. Proc. Natl. Acad. Sci. USA 91, 5972-5976.

Caudy, M. and Bentley, D. (1986). Pioneer growth cone steering along aseries of neuronal and non-neuronal cues of different affinities J. Neurosci.6, 1781-1795.

Condron, B. G., Patel, N. H. and Zinn, K. (1994). engrailed controlsglial/neuronal cell fate decisions at the midline of the central nervoussystem. Neuron 13, 541-554.

Daston, M. M., Bastmeyer, M., Rutishauser, U. and O’Leary, D. D. M.(1996). Spatially restricted increase in polysialic acid enhances corticospinalaxon branching related to target recognition and innervation. J. Neurosci.16, 5488-5497.

Davis, C. A., Holmyard, D. P., Millen, K. J. and Joyner, A. L. (1991).Examining pattern formation in mouse, chicken and frog embryos with anEn-specific antiserum. Development 111, 287-298.

Easter, S. S., Ross, L. S. and Frankfurter, A. (1993). Initial tract formationin the mouse brain. J. Neurosci. 13, 285-99

Eccles, J. C., Fatt, P. and Koketsu, K. (1954). Cholinergic and inhibitorysynapses in a path from motor-axon collaterals to motorneurones. J. Physiol.126, 524-562.

Eccles, J. C. (1957). The Physiology of Nerve Cells. London: OxfordUniversity Press.

Ericson, J., Rashbass, P., Schedl, A., Brenner-Morton, S., Kawakami, A.,Van Heyningen, V., Jessell, T. M. and Briscoe, J. (1997). Pax6 controlsprogenitor cell identity and neuronal fate in response to graded Shhsignaling. Cell 90, 169-180.

Fambrough, D. and Goodman, C. S. (1996). The Drosophila beaten pathgene encodes a novel secreted protein that regulates defasciculation at motoraxon choice points. Cell 87, 1049-1058

4212

Fazeli, A., Dickinson, S. L., Hermiston, M. L., Tighe, R. V., Steen, R. G.,Small, C. G., Stoeckli, E. T., Keino-Masu, K., Masu, M., Rayburn, H.,Simons, J., Bronson, R. T., Gordon, J. I., Tessier-Lavigne, M. andWeinberg, R. A. (1997). Phenotype of mice lacking functional Deleted inColorectal Cancer (DCC) gene. Nature 386, 796-804.

Gutman, C. R., Ajmera, M. K. and Hollyday, M. (1993). Organization ofmotor pools supplying axial muscles in the chicken. Brain Research 609,129-136.

Hanks, M., Wurst, W., Anson-Cartwright, L., Auerbach, A. B. and Joyner,A. L. (1995). Rescue of the En-1 mutant phenotype by replacement of En-1 with En-2. Science 269, 679-682.

Hedgecock, E., Culotti, J. G. and Hall, D. H. (1990). The unc-5, unc-6 andunc-40 genes guide circumferential migrations of pioneer axons andmesodermal cells on the epidermis in C. elegans. Neuron 4, 61-85.

Holley, J. A. and Silver, J. (1987). Growth pattern of pioneering chick spinalcord axons. Dev. Biol. 123, 375-388.

Jankowska, E. and Lindstrom, S. (1972). Morphology of interneuronsmediating Ia reciprocal inhibition of motor neurons in the spinal cord of thecat. J. Physiol. 226, 805-823.

Jankowska, E. and Roberts, W. J. (1972). An electrophysiologicaldemonstration of the axonal projectories of single spinal interneurons in thecat. J. Physiol. 226, 805-823.

Kennedy, T. E., Serafini, T., de la Torre, J. R. and Tessier-Lavigne, M.(1994). Netrins are diffusable chemotrophic factors for commissural axonsin the embryonic spinal cord. Cell 78, 425-435.

Kidd, T., Brose, K., Mitchell, K. J., Fetter, R. D., Tessier-Lavigne, M.,Goodman, C. S. and Tear, G. (1998). Roundabout controls axon crossingof the CNS midline and defines a novel subfamily of evolutionarilyconserved guidance receptors. Cell 92, 205-215.

Kolodzeij, P. A., Timpe, L. C., Mitchell, K. J., Fried, S. R., Goodman, C.S., Jan, L. Y. and Jan, Y. N. (1996). frazzled encodes a Drosphila memberof the DCC immunoglobulin subfamily and is required for CNS and motoraxon guidance. Cell 87, 197-204.

Kuwada, J. Y. (1986). Cell recognition by neuronal growth cones in a simplevertebrate embryo. Science 233, 740-746.

Landmesser, L. (1978). The distribution of motoneurones supplying the chickhindlimb muscles. J. Physiol. 284, 371-389.

Lee, K. J., Mendelsohn, M. and Jessell, T. M. (1998). Neuronal patterningby BMPs: a requirement for GDF7 in the generation of a discrete class ofcommissural interneurons in the mouse spinal cord. Genes Dev. 12, 3394-3407.

Leonardo, E. D., Hinck, L., Masu, M., Keino-Masu, K., Ackerman, S. L.and Tessier-Lavigne, M. (1997). Vertebrate homologues of C. elegansUNC-5 are candidate netrin receptors. Nature 386, 833-838.

Lin, J. H., Saito, T., Anderson, D. J., Lance-Jones, C., Jessell, T. M. andArber, S. (1998). Functionally related motor neuron pool and musclesensory afferent subtypes defined by coordinate ETS gene expression. Cell95, 393-407.

Loomis, C. A., Harris, E., Michaud, J., Wurst, W., Hanks, M. and Joyner,A. L. (1996). The mouse Engrailed-1 gene and ventral limb patterning.Nature 382, 360-363.

Lundell, M. J., Chu-LaGraff, Q., Doe, C. Q. and Hirsh, J. (1996). Theengrailed and huckebein genes are essential for development of serotoninneurons in the Drosophila CNS. Mol. Cell. Neuro. 7, 46-61.

McIntire, S. L., Garriga, G., White, J., Jacobson, D. and Horvitz, H. R.(1992). Genes necessary for directed axonal elongation or fasciculation inC. elegans.. Neuron 8, 307-322.

Matise, M. P. and Joyner, A. L. (1997). Expression patterns of developmentalcontrol genes in normal and Engrailed-1 mutant mouse spinal cord revealearly diversity in developing interneurons. J. Neurosci. 17, 7805-7816.

Mitchell, K. J., Doyle, J. L., Serafini, T., Kennedy, T. E., Tessier-Lavigne,M., Goodman, C. S. and Dickson, B. J. (1996). Genetic analysis of Netringenes in Drosophila: Netrins guide CNS commissural axons and peripheralmotor axons. Neuron 17, 203-215.

Pfaff, S. L., Mendelsohn, M., Stewart, C. L., Edlund, T. and Jessell, T. M.(1996). Requirement for LIM homeobox gene Isl1 in motor neurondifferentiation reveals a motor-neuron dependent step in interneurondifferentiation. Cell 84, 309-320.

Renshaw, B. (1946). Central effects of centripetal impulses in axons of spinalventral roots. J. Neurophysiol. 9, 191-204.

Serafini, T., Kennedy, T. E., Galko, M. J., Mirzayan, C., Jessell, T. M. andTessier-Lavigne, M. (1994). The netrins define a family of axon outgrowth-promoting proteins homologous to C.elegans UNC-6. Cell 78, 409-424.

Serafini, T., Colamarino, S. A., Leonardo, E. D., Wang, H., Beddington,R., Skarnes, W. C. and Tessier-Lavigne, M. (1996). Netrin-1 is requiredfor commissural axon guidance in the developing vertebrate nervous system.Cell 87, 1001-1014.

Sharma, K., Sheng, H. Z., Lettieri, K., Li, H., Karavanov, A., Potter, S.,Westphal, H. and Pfaff, S. L. (1998). LIM homeodomain factors Lhx3 andLhx4 assign subtype identies for motor neurons. Cell 95, 817-828.

Shirasaki, R., Tamada, A., Katsumata, R. and Murakami, F. (1995).Guidance of cerebellofugal axons in the rat embryo: directed growth towardthe floor plate and subsequent elongation along the longitudinal axis. Neuron14, 961-972.

Skarnes, W. C., Moss, J. E., Hurtley, S. M. and Beddington, R. S. P. (1995).Capturing genes encoding membrane and secreted proteins important formouse development. Proc. Natl. Acad. Sci. USA 92, 6592-6596.

Smith, C. L. and Hollyday, M. (1983). The development and postnatalorganization of motor nuclei in the rat thoracic spinal cord. J Comp. Neur.220, 16-28.

Stoekli, E. T. and Landmesser, L. T. (1995). Axonin-1, Nr-CAM and Ng-CAM play different roles in the in vivo guidance of chick commissuralneurons. Neuron 14, 1165-1179.

Tanabe, Y. and Jessell, T. M. (1996). Diversity and pattern in the developingspinal cord. Science 274, 1115-1123.

Tanabe, Y., Henderson, C. and Jessell, T. M. (1998). Specification of motorneuron identity by the MNR2 homeodomain protein. Cell 95, 67-80.

Tuttle, R., Braisted, J. E., Richards, L. J. and O’Leary, D. D. M. (1998).Retinal axon guidance by region-specific cues in the diencephalon.Development 125, 791-801.

Vaughn, J. E. and Grieshaber, J. A. (1973). A morphological investigationof an early reflex pathway in developing rat spinal cord. J. Comp. Neur. 148,177-210.

Vaughn, J. E., Sims, T. and Nakishima, M. (1977). A comparison of theearly development of axodendritic and axosomatic synapses uponembryonic mouse spinal motor neurons. J. Comp. Neurol. 175, 79-100.

Wenner, P., Matise, M., Joyner, A. and O’Donovan, M. J. (1998).Physiological, optical and molecular characterization of the avian Renshawcell and other last order interneurons in the embryonic spinal cord. Neurosci.Abstr. 24, 316.2.

Wentworth, L. (1984). The development of the cervical spinal cord of themouse embryo. II. A Golgi analysis of sensory, commissural and associationcell differentiation. J Comp. Neurol. 222, 96-115.

Wightman, B, Clark, S. G., Taskar, A. M., Forrester, W. C., Maricq, A. V.,Bargmann, C. I. and Garriga, G. (1996). The C. elegans gene vab-8 guidesposteriorly directed axon outgrowth and cell migration. Development 122,671-682.

Wurst, W., Auerbach, A. B. and Joyner, A. L. (1994). Multipledevelopmental defects in Engrailed-1 mutant mice: an early mid-hindbraindeletion and patterning defects in forelimbs and sternum. Development 120,2065-2075.

Yaganuma, H., Shiga, T., Homma, S., Ishihara, R. and Oppenheim, R. W.(1990). Identification of early developing axon projections from spinalinterneurons in the chick embryo with a neuron specific β-tubulin antibody.Development 108, 705-716.

Zallen, J. A., Yi, B. A. and Bargmann, C. I. (1998). The conservedimmunoglobulin superfamily member SAX3/Robo directs multiple aspectsof axon guidance in C. elegans. Cell 92, 217-227.

H. Saueressig, J. Burrill and M. Goulding