Evx1 Is a Postmitotic Determinant of V0 Interneuron Identity in the Spinal Cord

Neuron, Vol. 29, 385–399, February, 2001, Copyright 2001 by Cell Press Evx1 Is a Postmitotic Determinant of V0 Interneuron Identity in the Spinal Cord

Laura Moran-Rivard,2,4 Tetsushi Kagawa,1,4,5 1992, 1994). More recently, characterization of tran- Harald Saueressig,1,6 Michael K. Gross,1 scription factor expression in subsets of postmitotic John Burrill,1,7 and Martyn Goulding1,3 neurons has led to the identification of four primary 1Molecular Neurobiology Laboratory populations of interneurons in the ventral embryonic The Salk Institute spinal cord (Burrill et al., 1997; Ericson et al., 1997; Ma- 10010 North Torrey Pines Road tise and Joyner, 1997; Briscoe et al., 1999; Pierani et La Jolla, California 92037 al., 1999; Zhou et al., 2000). These four classes are desig- 2 Biology Graduate Program nated V0, V1, V2, and V3 interneurons and are marked by University of California, San Diego the expression of Evx1/2, En1, Lim3/Chx10, and Sim1, 9500 Gilman Drive respectively. La Jolla, California 92093 Recent studies have given rise to a model for spinal cord patterning in which the spatially organized development of ventral and intermediate cell types in the Summary neural tube is regulated by retinoic acid (RA) and Sonic hedgehog (Shh) (Yamada et al., 1993; Ericson et al., Interneurons in the ventral spinal cord are essential for 1996, 1997; Pierani et al., 1999). Shh induces motor coordinated locomotion in vertebrates. During embryo- neurons and interneurons in a concentration-dependent genesis, the V0 and V1 classes of ventral interneurons manner by repressing the activity of Gli3 in ventral neural are defined by expression of the homeodomain tran- tube progenitors and by regulating the expression of scription factors Evx1/2 and En1, respectively. In this known patterning genes in dorsoventrally restricted study, we show that Evx1 V0 interneurons are locally subsets of neural progenitors (Marti et al., 1995; Ericson projecting intersegmental commissural neurons. In et al., 1996, 1997; Pierani et al., 1999; Litingtung and Evx1 mutant embryos, the majority of V0 interneurons Chiang, 2000). Many of these patterning genes encode fail to extend commissural axons. Instead, they adopt homeodomain transcription factors that function in- an En1-like ipsilateral axonal projection and ectopi- structively in a combinatorial manner to define five pro- cally express En1, indicating that V0 interneurons are genitor domains that each give rise to a single early transfated to a V1 identity. Conversely, misexpression class of ventral neuron (Briscoe et al., 2000). In addition, of Evx1 represses En1, suggesting that Evx1 may sup- cross-repressive interactions between these homeo- press the V1 interneuron differentiation program. Our domain transcription factors refine and sharpen the findings demonstrate that Evx1 is a postmitotic deter- boundaries between adjacent progenitor domains (Bris- minant of V0 interneuron identity and reveal a critical coe et al., 2000). Support for this model has come from the postmitotic phase for neuronal determination in the following findings. First, manipulating Shh levels both developing spinal cord. in vitro and in vivo regulates the generation of ventral cell types in a concentration-dependent manner (Chiang Introduction et al., 1996; Ericson et al., 1996, 1997), while the development of intermediate cell types requires RA signaling (Pierani et al., 1999). Second, loss-of-function studies The generation of spinal interneuron populations with demonstrate essential roles for early patterning genes distinct patterns of neuronal connectivity is critical for such as Pax6, Nkx2.2, and Nkx6.1 in specifying neuronal forming the reflex circuits that control the coordinated identity (Burrill et al., 1997; Ericson et al., 1997; Tanabe movement of muscles in vertebrates. Physiological et al., 1998; Briscoe et al., 1999; Sander et al., 2000). studies in the adult spinal cord reveal that the primary Finally, neural progenitors can be respecified when dor- function of interneurons located in the ventral spinal soventral patterning genes are ectopically expressed in cord is the control of locomotion and posture (Grillner, the ventricular zone (Briscoe et al., 2000; Sander et al., 1975; Allum et al., 1989). Many of these ventral interneur- 2000; Pierani et al., 2001 [this issue of Neuron]). To- ons synapse directly with motor neurons and either facil- gether, these and other studies demonstrate that the itate or inhibit their activity (Rekling et al., 2000). The restricted expression of transcription factors in dividing molecular mechanisms that specify distinct classes of progenitors plays an instructive role in determining cell spinal interneurons are poorly understood, largely be- fate in the neural tube. cause early attempts to identify embryonic interneurons While a number of studies have analyzed the role relied on neuroanatomical features that appear only of transcription factors in motor neuron development after neurons are specified (Silos-Santiago and Snider, (reviewed in Jessell, 2000), similar functional studies of interneurons have focused primarily on transcription fac- 3 To whom correspondence should be addressed (e-mail: goulding@ tors that are expressed in dividing progenitors (Briscoe et salk.edu). al., 2000; Sander et al., 2000; Zhou et al., 2000). Conse- 4 These authors contributed equally to this work. quently, it is not known whether transcription factors 5 Present address: National Institute for Physiological Sciences, that are expressed exclusively in postmitotic cells play Okasaki, Aichi 444-8585, Japan. 6 Present address: Evotec Neurosciences, D-22525 Hamburg, Germany. important roles in specifying neuronal identity. In the 7 Present address: Incyte Pharmaceuticals, Palo Alto, California aforementioned studies, changes in cell identity were 94304. largely documented by alterations in the expression of Neuron 386

transcription factors in subsets of postmitotic neurons. Results However, it is not known how these changes in gene expression relate to other aspects of neuronal identity Evx1 Cells Are a Discrete Population such as cell body position in the spinal cord, neurotrans- of Postmitotic Interneurons mitter phenotype, projection pattern, and target selec- Previous studies showing Evx1 is expressed in two bilat- tion. Lineage tracing studies in the spinal cord suggest eral stripes of cells in the ventral hindbrain and spinal that neuronal identity is determined in the ventricular cord did not attempt to analyze these cells in detail zone either just prior to or as progenitors exit the cell (Bastian and Gruss, 1990; Dolle et al., 1994). To further cycle (Leber et al., 1990; Leber and Sanes, 1995). These characterize these Evx1-expressing cells, an antibody studies, however, did not determine whether cells are specific for Evx1 was generated and its expression com- irreversibly committed to a particular neuronal fate upon pared with markers that identify specific populations of exiting the cell cycle, or whether there is an early postmi- early spinal cord neurons. The Evx1 neurons are gener- totic phase during which cell fate can be altered. The ated in a narrow band at the lateral edge of the ventricu- restricted expression of transcription factors in subsets lar zone just ventral to the sulcus limitans, during a of differentiating neurons suggests that they may func- period of rapid neurogenesis that occurs from E9.5 to tion as neural determinants postmitotically (Nornes et E13.5 (Figures 1A–1C; Bastian and Gruss, 1990). These al., 1990; Zhadanov et al., 1995; Burrill et al., 1997; Eric- Evx1 cells arise from dividing neural precursors that son et al., 1997; Matise and Joyner, 1997; Lee et al., express Pax6, Dbx1, and Dbx2, but not Pax3 or Pax7 1998). However, only V1 interneurons have been charac- (Figure 1D; Burrill et al., 1997; Pierani et al., 1999, 2001). terized in sufficient detail to enable examination of the Soon after being generated, Evx1 neurons migrate ven- role of En1 in V1 interneuron differentiation (Matise and tromedially toward the floor plate and settle in the ventral Joyner, 1997; Wenner et al., 1998; Saueressig et al., horns medial to motor neurons (Figures 1A–1C and 1J), 1999). This analysis revealed that En1 does not specify in a region corresponding to lamina VIII in the adult spinal cord. The migration route taken by the Evx1 V0 V1 interneuron identity and instead regulates late as- interneurons differs markedly from other neurons in the pects of V1 axon pathfinding (Saueressig et al., 1999). ventral spinal cord, including En1 V1 interneurons (Fig- Thus, it remains to be determined whether postmitotic ure 1J). Although En1 V1 interneurons are born immedi- transcription factors such as Evx1 contribute to the es- ately adjacent to the Evx1 cells, they migrate ventrolater- tablishment of distinct interneuron cell fates in the devel- ally and settle in the lateral ventral horns close to motor oping spinal cord. neurons (Matise and Joyner, 1997; Saueressig et al., Previous expression studies have shown that the ho- 1999). Chx10 V2 interneurons generated adjacent to mo- meodomain transcription factor Evx1 is expressed in tor neurons also migrate laterally in a manner similar to the spinal cord from E9.5 to E13 in two bilateral stripes the V1 interneurons (data not shown). just ventral to the sulcus limitans (Bastian and Gruss, Ϫ ϩ In both mouse and chick, Evx1 is expressed in Pax6 1990; Dolle et al., 1994; Burrill et al., 1997). These Evx1 cells at the lateral margin of the intermediate ventricular cells are derived from progenitors that express the ho- zone (Figure 1D), indicating they are postmitotic. To meodomain transcription factor Dbx1 (Pierani et al., confirm this, E10.5 mouse embryos were pulse labeled 1999, 2001). Past studies of Evx1 function did not ad- with BrdU and examined for Evx1 expression (Figure 1E). dress the role of Evx1 in the developing spinal cord Of 546 Evx1ϩ cells examined, 545 were unambiguously (Spyropoulos and Capecchi, 1994). In this study, we BrdU negative, demonstrating that Evx1 cells are not generated mice with a conditional Evx1 null allele that mitotically active. These findings argue that Evx1 ex- expresses the neuronal reporter protein tau-myc (Thor pression is initiated in a population of newly born in- and Thomas, 1997). These mice were used to character- terneurons where it may regulate postmitotic aspects ize the axonal morphology and migratory behavior of of cell identity. V0 interneurons and to examine the function of Evx1 in To determine whether Evx1 selectively marks a popu- V0 interneuron development. Our studies show that lation of differentiating neurons in the spinal cord, the Evx1 defines a ventral population of commissural in- expression of Evx1 was compared with other neuronal terneurons that project rostrally for one to four spinal subclasses as defined by transcription factor expres- cord segments in the ventral funiculus. These cells mi- sion. This analysis showed that Evx1 V0 interneurons grate along a ventromedial pathway and congregate in do not express Lbx1 (Figure 1F; Jagla et al., 1995), a the ventral horn medial to somatic motor neurons. In marker of dorsal interneurons, nor do they express the Evx1 mutant embryos, V0 interneurons no longer retain V1 and V2 interneuron markers En1 (Davis et al., 1991) their wild-type axon projection pattern. Instead, many and Chx10 (Liu et al., 1994), respectively (Figures 1G of these interneurons elaborate ipsilateral projections and 1H). In addition, none of the Evx1 cells located in characteristic of the V1 class of interneurons and mi- the ventral horn coexpress Isl1 (Ericson et al., 1992), grate to ectopic sites in the lateral ventral horn. These which labels differentiating motor neurons (Figure 1I). aberrant interneurons express En1, indicating that Evx1 Thus, Evx1 expression defines a population of ventral V0 interneurons transfate to V1 interneurons in the ab- interneurons in the spinal cord that is distinct from V1, sence of Evx1. This change in cell fate, coupled with V2, and V3 interneurons. the observed suppression of endogenous En1 by the misexpression of Evx1, demonstrates Evx1 functions as Generation of Conditional Evx1tau-myc Knockin Mice a postmitotic determinant of V0 interneuron identity in To further characterize Evx1 V0 interneurons and exam- the spinal cord. ine the role of Evx1 in their development, we generated Evx1 Is a Determinant of Ventral Interneuron Fate 387

Figure 1. Early Expression of Evx1 in V0 In- terneurons (A–C) Developmental time course of Evx1 expression at forelimb levels in the spinal cord as detected by an antibody specific for the N terminus of Evx1. (A) At E10.5, most Evx1 cells are located adjacent to their generation zone. (B and C) In E11.5 and E12.5 embryos, Evx1ϩ cells migrate ventrally and progressively accumulate in the medial ventral horn adjacent to the floor plate. (D) Section through E10.5 mouse spinal cord showing Evx1 (red) is not expressed in dividing Pax6ϩ cells (green). (E) Spinal cord section from an E10.5 mouse embryo pulsed with BrdU (green) for 2 hr and stained with an antibody to Evx1ϩ (red). (F–I) Sections from E10.5 mouse embryos demonstrating that Evx1 V0 interneurons do not coexpress markers for other classes of spinal neurons. Evx1ϩ interneurons (red) develop ventral to Lbx1 dorsal interneurons (green) (F), dorsal to En1ϩ V1 interneurons (green) (G), and dorsal to Chx10 V2 interneurons (green) (H). The ventromedial migratory Evx1ϩ neurons are also distinct from motor neurons that express Isl1 (green) (I). (J) Schematic representation of neuronal classes in the ventral spinal cord at E10.5 and the migratory routes of Evx1 (red) and En1 (blue) interneurons at E12.5. fp, floor plate.

a conditional knockin null allele of Evx1. Previously de- vent this reported embryonic lethal phenotype, we uti- scribed Evx1 knockout mice exhibit an early embryonic lized the Cre/loxP recombination system (Gu et al., 1994) lethal phenotype caused by the failure of Evx1Ϫ/Ϫ blasto- to generate a conditional null allele of Evx1. The Evx1 cysts to form extraembryonic tissues (Spyropoulos and targeting vector was generated by introducing loxP sites Capecchi, 1994). While this phenotype was assumed to into the second intron and the 3Ј untranslated region of result from the loss of low level Evx1 expression in the Evx1 (Figure 2A). An IRES-tau-myc neuronal reporter visceral endoderm of E5.0 egg cylinders, recent studies cassette (Thor and Thomas, 1997) was included in the by ourselves and others argue that this phenotype is targeting vector to enable visualization of the Evx1 V0 not due to the loss of Evx1 (see Discussion). To circum- interneurons. The targeted allele permits the normal ex- Neuron 388

Figure 2. Generation of a Floxed Evx1tau-myc Null Allele in Mice (A) Schematics of wild-type, targeted, and Cre-deleted Evx1 loci. The wild-type Evx1 locus is composed of three exons (I–III), with the homeodomain (hatched) interrupted by intron 2. Homologous recombination in ES cells was used to create the conditional targeted Evx1 allele, which contains loxP sites in the second intron and in the 3ЈUTR, a PGKneopA cassette upstream of the second loxP site, and an IRES-tau-myc reporter cassette downstream of the second loxP site. The Cre-deleted Evx1 allele was generated by crossing conditional Evx1tau-myc mice with a Protamine-Cre transgenic line. The recognition helix of the homeodomain and the pA sequence of the neo cassette are excised following Cre-induced recombination, thereby producing a null Evx1 allele that expresses the axonal marker tau-myc. (B) Southern blot analysis was used to screen ES cell clones and to genotype the progeny of Evx1tau-myc/Protamine-Cre matings. Both ES cell clones and genomic DNAs were digested with Kpn and probed with an internal 5Ј Sal1-Pst1 DNA probe, producing 4.9 kb, 7.0 kb, and 2.8 kb bands for the wild-type, targeted and Cre-deleted Evx1 alleles, respectively. Bgl, BglII; H, HindIII; K, KpnI; P, PstI; S, Sal1; Xb, XbaI.

pression of Evx1 but not tau-myc prior to Cre-mediated terneurons. In transverse sections from E11 Evx1tm/ϩ em- recombination. Three recombinant clones harboring the bryos, tau-myc faithfully recapitulates the endogenous Evx1tau-myc (Evxtm) allele were identified by Southern Blot pattern of Evx1 expression, with cell bodies and axons analysis using probes to detect homologous recombina- both showing extensive tau-myc staining (Figures 3A tion at the Evx1 locus (Figure 2B). One of these clones and 3B). Evx1 V0 interneurons extend axons along the (76) gave germline transmission of the floxed Evx1 allele same pathway as dorsal TAG1 commissural neurons when injected into blastocysts. whose axons course medial to motor neurons before Our initial strategy for restricting Cre-mediated re- crossing the midline (Dodd et al., 1988; Figure 3B, combination of the targeted Evx1 locus to V0 interneu- arrow). After reaching the contralateral side of the spinal rons utilized several independent strains of Dbx1-Cre cord, Evx1 V0 axons form a fasciculated bundle in the transgenic mice (T. K. and M. G., unpublished data). ventral funiculus (vf). In contrast, little or no tau-myc While a number of the lines were able to induce Cre- staining is found more laterally in the ventrolateral funic- mediated recombination, we often observed mosaic ulus (vlf), which contains ipsilaterally projecting axons tm tau-myc expression in the progeny of Dbx1-Cre/Evx1 (Vaughn and Grieshaber, 1973; Saueressig et al., 1999). crosses that may have been caused by low levels of Cre To determine whether the axons of Evx1 V0 interneu- protein, resulting in incomplete recombination. Although rons project locally or over multiple spinal cord seg- we obtained preliminary results from Dbx1-Cre inter- ments, and to ascertain whether Evx1 is expressed by crossed mice (Figures 3A and 3B), we subsequently a single class of interneurons, we performed retrograde utilized a line of transgenic mice that express Cre under labeling studies to examine the axonal projections of the control of the protamine promoter to induce germline Evx1 V0 interneurons. Fluorescein-conjugated dextran recombination (Figure 2B; O’Gorman et al., 1997). This approach was taken in response to reports that the (FD) was injected into one side of the ventral neural loss of Evx1 does not cause an early embryonic lethal tube at the level of the first thoracic segment, T1. Serial phenotype (D. Goldman and G. Martin, personal commu- sections spanning the injection site (up to 3 mm or ten nication). When crossed with the conditional Evx1tm spinal cord segments rostral and caudal to the injection mice, Protamine-Cre mice gave viable offspring that site) were stained with an antibody to Evx1 and analyzed were used to establish an Evx1tm/ϩ colony for subsequent by confocal microscopy (Figures 3C–3E). In each of the phenotypic analysis. injected embryos, we detected FD-labeled cells up to 2.5 mm (eight to nine segment equivalents) from the ϩ ϩ Evx1 Neurons Are Commissural Interneurons injection site. When the distribution of FD /Evx1 in- that Project Locally terneurons in wild-type embryos was examined (n ϭ 3 Evx1tm/ϩ heterozygous embryos were used to morpho- embryos), 94% were found to project commissurally, logically characterize the Evx1 V0 class of spinal in- and of these 95% had rostrally directed axons extending Evx1 Is a Determinant of Ventral Interneuron Fate 389

Figure 3. Evx1 V0 Interneurons Project Commissurally and Locally (A and B) Transverse sections of E11.5 Evx1tm/ϩ embryos derived from crosses of a Dbx1-Cre transgenic strain with the conditional Evx1tau-myc line. Stainings with antibodies to myc (red) and (B) TAG1 (green) demonstrate that Evx1 V0 interneurons are commissural and follow a similar axonal trajectory to TAG1ϩ dorsal commissural interneurons (arrows). Evx1tm/ϩ interneurons are generated just ventral to the sulcus limitans (arrowhead), recapitulating the endogenous Evx1 domain. Tau-mycϩ axons project in the ventral funiculus. (C and D) Cross sections of the ipsilateral and contralateral sides of E12.5 ventral spinal cords retrogradely labeled with fluorescein-conjugated dextran (green) and stained with an Evx1 antibody (red). Arrows indicate examples of double-labeled cells (yellow) on the (D) contralateral side, which are not present on the adjacent (C) ipsilateral side. (E) Summary schematic showing the axonal projections of Evx1 cells in the spinal cord. Counts represent the number of double-labeled cells from three embryos. The arrows indicate the distance (␮m) double-labeled cells project to the injection site. Contralaterally projecting neurons are marked in red and ipsilaterally projecting neurons are marked in blue. The length of one spinal cord segment is indicated. vf, ventral funiculus.

one to four spinal cord segments (160–1200 ␮m) after Evx1 Interneurons Exhibit Altered Migration crossing the floor plate (Figure 3E). The apparent lack and Axonal Projections in Evx1 Mutants of FDϩ/Evx1ϩ interneurons at distances greater than The restricted expression of Evx1 in newly born V0 in- 1200 ␮m from the injection site, together with the ob- terneurons suggests Evx1 may function as a determi- served peak of FDϩ/Evx1ϩ cells between 160 ␮m and nant of V0 identity. To test this hypothesis, we generated 480 ␮m suggests that the majority of Evx1 V0 interneu- homozygous Evx1tm/tm mutants. As suggested by recent ϩ rons are local intersegmental interneurons that do not findings discussed above, interbreedings of Evx1tm/ project supraspinally. While a small number of ipsilater- mice produced viable Evx1tm/tm adult progeny. While a ally projecting Evx1 interneurons were also detected in small number of adult homozygous mutant mice exhib- this analysis, their significance is not clear, as they were ited tail kinks, no other gross morphological or motor relatively rare and many were located close to the injec- defects were apparent in newborn or adult mice (data tion site. As such, these cells may represent background not shown). Use of the tau-myc reporter protein enabled labeling from damaged axons on the injected side, a a careful comparison of the morphology and axonal projections of Evx1 V0 interneurons in heterozygous and conclusion consistent with the absence of ipsilaterally mutant embryos. In Evx1tm/ϩ embryos, Evx1 V0 interneu- projecting Evx1 neurons in Evx1tm/ϩ heterozygous em- rons have commissural axons that cross the floor plate bryos (see below). and project in the contralateral ventral funiculus (vf) (Fig- The contralateral-rostral axon projection of Evx1 V0 ures 4A, 4D, 4G, and 4J), as was previously described interneurons is distinct from the En1 V1 and Chx10 V2 (Figure 3). In addition, the cell bodies of V0 interneurons interneuron classes, both of which project ipsilaterally migrate along the same route that their axons take to- (Saueressig et al., 1999). The Evx1 V0 interneuron pro- ward the floor plate, progressively accumulating in a jections also differ markedly from dorsal commissural region medial to motor neurons (Figures 4D, 4G, and interneurons that extend axons over five or more spinal 4J; see also Figure 1). segments (J. B. and M. G., unpublished data). Our retro- When spinal cords from homozygous Evx1tm/tm em- grade tracing studies, together with the characterization bryos were examined, a striking difference in the migra- ϩ of Evx1tm/ embryos, strongly suggest that Evx1 V0 in- tory behavior of Evx1 V0 interneurons was observed terneurons comprise a class of rostrally projecting com- from E11.5 onwards. In E10.5 Evx1tm/tm embryos, the missural interneurons that form local intersegmental majority of Evx1 interneurons are located adjacent to connections with neurons in the ventral horn. the ventricular zone, just ventral to the sulcus limitans Neuron 390

Figure 4. Evx1 V0 Interneurons Exhibit Altered Cell Migration and Axon Projections in Evx1 Mutants (A, B, D, E, G, H, and J–M) Cross sections through the spinal cords of E10.5–E12.5 Evx1tm/ϩ and Evx1tm/tm embryos stained with a myc antibody and an HRP-conjugated secondary antibody to reveal the morphology of Evx1 V0 interneurons. (C, F, and I) Quantitation of tau-mycϩ interneurons in E10.5–E12.5 Evx1tm/ϩ and Evx1tm/tm embryos. The ventral horn was divided into medial and lateral quadrants. Only cells located in the ventral horn away from the generation zone were included in the analysis of lateral versus medial cells. For total cell numbers, all Evx1ϩ cells were counted (asterisk). Cell counts represent the number of cells on one side of the spinal cord in a 20 ␮m section. (A, B, and C) At E10.5, (A) Evx1tm/ϩ and (B) Evx1tm/tm embryos have commissural axons extending across the ventral midline, and (C) there is no difference in the distribution of tau-mycϩ cells. (D, G, and J) In E11.5–12.5 Evx1tm/ϩ embryos, Evx1 interneurons migrate ventrally toward the floor plate. The Evx1tm/ϩ axons are commissural (arrow) and have established a longitudinal tract in the vf after crossing the floor plate. (E, H, and K) In E11.5–12.5 Evx1tm/tm embryos, a subset of Evx1 interneurons migrate ventrolaterally toward the lateral ventral horns. These aberrant Evx1 interneurons elaborate axons radially (arrow) toward the vlf, which is markedly increased in tau-myc staining. Tau-myc staining is greatly reduced in the vf of Evx1tm/tm embryos. (F and I) The number of ectopic lateral tau-mycϩ interneurons increases dramatically in E11.5–12.5 Evx1tm/tm embryos, while the size of the medial population remains constant. (L–M) A cross-section of an E11.5 Evx1tm/tm spinal cord reveals a tau-mycϩ neuron with a bifurcated axon extending the longer branch toward the lateral ventral horn (arrow). fp, floor plate; vf, ventral funiculus; vlf, ventrolateral funiculus; MNs, motor neurons. Evx1 Is a Determinant of Ventral Interneuron Fate 391

(Figure 4B). A detailed comparison of Evx1tm/ϩ and axons to the ventral midline (Figure 4K; data not shown), Evx1tm/tm embryos revealed no significant difference in at more dorsal locations we occasionally observed Evx1 the distribution of tau-mycϩ cells at this time (Figure interneurons with bifurcated axons (Figure 4L). In the 4C). However, from E11.5 onwards, large numbers of example shown, a tau-mycϩ neuron can be seen ex- Evx1 interneurons were observed migrating ventrolater- tending a major branch laterally (Figure 4M, arrow), while ally into the lateral ventral horns of Evx1tm/tm embryos the ventromedial offshoot appears truncated (Figure 4M, (Figures 4E, 4H, and 4K). This aberrant migratory behav- asterisk). These observations suggest that in Evx1tm/tm ior resulted in the accumulation of Evx1 interneurons in mutants, ectopic Evx1 interneurons may initiate a wild- the lateral ventral horn close to motor neurons (Figures type ventral migration before being redirected into a 4E, 4H, and 4K). Whereas in Evx1tm/ϩ embryos the num- lateral migratory pathway. Such a migratory behavior ber of medially located Evx1 interneurons increased in would account for the delayed appearance of Evx1 in- a linear manner from E10.5 to E12.5, similar numbers terneurons in the lateral ventral horn, although the mech- of medially located Evx1 interneurons were present in anism that allows a subset of Evx1 interneurons to elab- E10.5–E12.5 Evx1tm/tm mutant embryos (Figures 4C, 4F, orate axons that extend to the ventral midline remains and 4I). Instead, we observed a dramatic increase in the unclear. number of tau-mycϩ cells located in the lateral ventral horns of Evx1tm/tm mutants from E11.5 to E12.5. By E12.5, Evx1 Is Required for Evx2 Expression 65% (33/51 cells per 20 ␮m ventral horn section) of Evx1 in V0 Interneurons neurons that had migrated from their generation zone Evx1 and Evx2 are close homologs and are coexpressed were found in the lateral ventral horn (Figures 4C, 4F, in V0 interneurons (Dolle et al., 1994; Pierani et al., 1999; and 4I). No difference was found in the total number of data not shown). Thus, the observed delay in the appear- ϩ tm/ϩ tm/tm tau-myc cells in Evx1 and Evx1 embryos (Figures ance of ectopically located Evx1 interneurons in Evx1tm/tm 4C, 4F, and 4I), and TUNEL analysis did not show in- mutants, coupled with our finding that some Evx1 in- tm/tm creased cell death in Evx1 embryos (data not shown). terneurons still migrate ventromedially and have axons These results reveal that Evx1 is required for the sus- that project toward the ventral midline, may reflect a tained ventromedial migration of V0 interneurons but is transient compensation by Evx2 for the loss of Evx1. not essential for their survival. We therefore analyzed Evx2 expression in Evx1 mutant The dramatic change in the migratory behavior of Evx1 embryos, using a monoclonal antibody that recognizes tm/tm interneurons in Evx1 embryos was accompanied by the C termini of Evx1 and Evx2 (Pierani et al., 1999). We a corresponding shift in their axonal projections. Rather deduced that any positive staining by the Evx1/2 mono- than projecting commissurally, the ectopically posi- clonal antibody in Evx1tm/tm spinal cords would reflect Evx2 tioned Evx1 interneurons possessed axons that were expression, as the C terminus of Evx1 is truncated follow- directed radially toward the lateral edge of the ventral ing Cre induced recombination (Figure 2A). In E10.5 and horn (Figure 4K, arrow). We also observed an associated E12.5 wild-type embryos, we observed labeled cells in tm/tm lateral shift in funicular tau-myc staining in Evx1 em- the ventral horn near the sulcus limitans consistent with tm/tm bryos (c.f. Figures 4D and 4E), indicating that in Evx1 the expression of Evx1 and Evx2 in V0 interneurons mutants, a large fraction of Evx1 interneurons possess (Figures 5A and 5B, arrows). In contrast, Evx2 expres- axons that project longitudinally in the ventrolateral fusion was completely absent from the ventral spinal cord niculus (vlf). While Evx1 interneurons with commissural- of E10.5 and E12.5 Evx1tm/tm embryos (Figures 5D and like projections were still present in E11.5 and E12.5 5E, arrows). This loss of Evx2 expression in Evx1tm/tm Evx1tm/tm embryos (Figures 4E, 4H, and 4K), the sharp embryos was confirmed by in situ hybridization using a loss of tau-myc staining in the ventral funiculus indicates probe specific for mouse Evx2 (Figures 5C and 5F). that few, if any, of these axons project any distance in the contralateral ventral funiculus. The absence of tau- mycϩ commissural axons that cross the midline and Evx1 Interneurons Ectopically Express En1 tm/tm extend to the ventrolateral funiculus further suggests in Evx1 Mutants that these medially positioned commissural-like Evx1 Two mechanisms can be invoked to explain the aberrant interneurons fail to project longitudinally in Evx1 mutant migratory behavior and axonal projections of Evx1 in- embryos. We conclude from the above findings that terneurons in mutant embryos: (1) Evx1 interneurons Evx1 is required for the commissural projection of most transfate and adopt the identity of another neuronal V0 interneurons and the longitudinal projection of all V0 class or (2) Evx1 interneurons fail to activate a full V0 interneurons in the ventral funiculus. differentiation program and are therefore unable to mi- In observing the altered migratory behavior and axo- grate normally or extend axons any significant distance nal projections of Evx1 interneurons in mutant embryos, within the spinal cord. To distinguish between these two we noted that tau-mycϩ cells do not migrate laterally possibilities, we asked whether Evx1 V0 interneurons from their generation zone into the lateral ventral horn. in Evx1tm/tm embryos express markers characteristic of ϩ Instead, Evx1 interneurons were observed to first mi- other neuronal cell types. In E11.5 Evx1tm/ embryos, grate a short distance ventrally, then turn and enter the tau-myc is never coexpressed with En1, Chx10, Lbx1, lateral ventral horn (Figures 4K and 4L; data not shown). or Isl1 (Figures 6A–6D), supporting our earlier finding We were therefore interested in whether Evx1 interneur- that Evx1 expression is restricted to V0 interneurons ons that initially enter the V0 migration stream exhibit (Figure 1). In contrast, widespread coexpression of En1 changes in their axon morphology that might presage was detected in tau-mycϩ neurons in E11.5 Evx1tm/tm a change in their direction of migration. While all medial embryos, (Figure 6F, arrows, inset). Similar sections Evx1 interneurons located close to the floor plate extend showed no expression of Lbx1, Chx10, or Isl1 in tau- Neuron 392

Figure 5. Evx1 Is Necessary for Evx2 Expression in V0 Interneurons (A and B) Evx1/2 protein expression (red) was detected at E10.5 and E12.5 in wild-type mouse embryos (arrows). (D and E) Evx2 protein is not present at E10.5 or E12.5 in the ventral spinal cord of Evx1tm/tm mutant embryos (arrows). (C and F) In situ analysis of Evx2 expression at cervical spinal cord levels. At E11.5, Evx2 is expressed in V0 interneurons (arrow) and a population of dorsal interneurons (arrowhead) in the spinal cord of wild-type embryos (C), but not in the spinal cord of Evx1tm/tm embryos (F). mycϩ cells, arguing that these ectopically positioned have shown En1 does not specify V1 interneuron fate Evx1 neurons do not differentiate as dorsal interneurons, (Saueressig et al., 1999). V1 interneurons can however V2 interneurons, or motor neurons (Figures 6G–6I). To be distinguished from other ventral interneurons by their determine the proportion of ectopic Evx1 interneurons stereotypical ipsilateral axon projections (Saueressig et that coexpress En1, representative sections from E11.5 al., 1999), thereby allowing us to assess whether Evx1 mutant embryos were scanned at multiple focal planes. interneurons acquire a V1 fate. Using fluorescein-conju- ϩ When tau-myc cells with clearly visible nuclei were gated dextran to label neurons in the ventral spinal cord ϩ counted, 94% of the tau-myc neurons located in the and a myc antibody to detect Evx1 interneurons, we lateral ventral horn expressed En1, a definitive marker of observed a dramatic difference in the location of double- V1 interneurons. In contrast, En1 expression was never labeled cells in the heterozygous and homozygous detected in medially located Evx1 interneurons that have Evx1tm/tm embryos. In Evx1tm/ϩ embryos, mycϩ/FDϩ neu- axons projecting toward the floor plate. Our observation rons were located contralateral to the injection site (Fig- that nearly all ectopically positioned Evx1 interneurons ures 7A and 7B), recapitulating the projection pattern express En1 suggests that in the absence of Evx1, a of wild-type Evx1 interneurons (Figure 3). In contrast, in substantial fraction of the V0 interneurons differentiate ϩ ϩ Evx1tm/tm embryos the majority of the myc /FD neurons as V1-like En1ϩ interneurons. In embryos lacking Dbx1, were located ipsilateral to the injection site and elabo- a similar induction of En1 in V0 interneurons occurs rated axons that project rostrally for one to two seg- (Pierani et al., 2001). To test whether the ectopic induc- ments (Figures 7A and 7B). This projection phenotype tion of En1 in Evx1 mutants is due to changes in Dbx1 expression, we examined Dbx1 expression in Evx1tm/ϩ is identical to that of the En1 V1 interneurons (Saueressig and Evx1tm/tm embryos (Figures 6E and 6J). No change et al., 1999). Furthermore, the rostral projection of these in Dbx1 was observed. This suggests that the induction ipsilateral Evx1 interneurons is notably shorter (up to of En1 in V0 interneurons in Dbx1 mutant mice may be two segments) than that of wild-type contralateral Evx1 due to the loss of Evx1 (Pierani et al., 2001). interneurons (one to four segments), which is a hallmark of the V1 interneuron projection pattern. This dramatic In Evx1 Mutants, V0 Interneurons Project reduction in the number of contralateral projections that Ipsilaterally in a Manner Analogous occurs in Evx1tm/tm embryos is consistent with diminished to V1 Interneurons tau-myc staining in the ventral funiculus (Figure 4) and While it is significant that Evx1tm/tm embryos exhibit ec- argues that few, if any, Evx1 interneurons maintain their topic expression of En1, this finding is not sufficient to normal intersegmental commissural projections. In sum- show they adopt a V1 identity, since previous studies mary, the induction of En1 in Evx1 interneurons, together Evx1 Is a Determinant of Ventral Interneuron Fate 393

Figure 6. Evx1 Interneurons Ectopically Express En1 in Evx1tm Mutant Embryos (A–D) Cross sections of E12 Evx1tm/ϩ embryos stained with an antibody to myc (red) demonstrate that Evx1 is not coexpressed with (A) En1 (green), (B) Chx10 (green), (C) Isl1 (green), or (D) Lbx1. In E12 Evx1tm/tm embryos, Evx1 interneurons do not express (I) Lbx1, (G) Chx10, or (H) Isl1. (F) En1 is ectopically expressed by Evx1tm/tm interneurons located in the lateral ventral horn (n ϭ 51/54 tau-mycϩ cells). The arrows denote examples of double-labeled cells. The high magnification inset depicts two mycϩ/En1ϩ cells with nuclear En1 (green) staining and cytoplasmic/ axonal myc (red). (E and J) No change in the expression of Dbx1 (red) was observed in ventricular zone of E12 Evx1tm/tm mutant spinal cords (J) as compared to wild-type (E). A shift in the funicular expression of tau-myc can also be seen in Evx1 mutant spinal cords (c.f. [B] and [G]). fp, floor plate; vf, ventral funiculus; vlf, ventrolateral funiculus. with the changes in their axon trajectories and migration, sion (Burrill et al., 1997). No significant change in Pax6 provides compelling evidence that Evx1 regulates V0 expression was seen on the electroporated side (Figures interneuron fate, and when Evx1 is absent, a substantial 8E and 8F, n ϭ 6 embryos). Taken together, our data proportion of V0 interneurons differentiate as En1 V1 suggest that Evx1 may contribute to the specification interneurons (Figure 7C). of V0 commissural interneuron identity, in part by repressing the V1 interneuron differentiation program that Misexpression of Evx1 Can Repress Endogenous is associated with En1 expression (Figure 8I). Expression of En1 The ectopic expression of En1 in Evx1tm/tm neurons sug- Discussion gests that one function of Evx1 may be to repress V1 specific genes in commissural interneurons. To test this, The molecular mechanisms that generate interneuron we utilized the chick system, which is more amenable diversity in the spinal cord are poorly defined. In this to misexpression studies. Embryos electroporated at study, we utilized Evx1tau-myc knockin mice to characterize stage 11 with a single vector coexpressing Evx1 and V0 interneurons and examine the role that Evx1 plays GFP were analyzed at stage 23 for changes in gene in spinal interneuron subtype specification. We show expression. Control electroporations using a vector that that Evx1 expression delineates a population of locally expresses GFP alone had no effect on En1 expression projecting intersegmental commissural neurons that are (Figures 8G and 8H, arrowhead). However, ectopic ex- likely to participate in the spinal circuitry that controls pression of Evx1 dramatically downregulated En1 in the locomotion. Analysis of the Evx1 mutant phenotype ventral spinal cord, with abolishment of En1 expression demonstrates that Evx1 is required for the correct speci- corresponding to domains of high Evx1 expression (Fig- fication of V0 interneurons during development. In Evx1 ures 8A and 8B, arrowhead, n ϭ 6 embryos). In contrast, null embryos, the majority of V0 interneurons are respec- Isl1 expression was not disrupted in either motor neu- ified and adopt a V1 interneuron fate. Our studies also rons or dorsal interneurons (Figures 8C and 8D, n ϭ 5 show Evx1 functions as a postmitotic determinant, act- embryos), indicating that the repression of En1 by Evx1 ing downstream of Dbx1 to consolidate V0 interneuron is specific and does not result from a nonspecific repres- identity. These findings, when integrated with the ac- sor activity. We also examined Pax6 expression in elec- companying study showing Dbx1 controls the specifica- troporated embryos to test whether Evx1 interferes with tion of V0 interneuron progenitors (Pierani et al., 2001), genes that are expressed in dividing precursors, since define a transcription factor cascade that governs the the loss of Pax6 is known to downregulate En1 expres- generation of V0 interneurons. Our results are discussed Neuron 394

Figure 7. Evx1 V0 Interneurons Adopt an En1 V1-like Projection Pattern in Evx1tm Mutant Embryos (A) Sections from Evx1tm/ϩ and Evx1tm/tm embryos retrogradely labeled with fluorescein-dextran (green) and stained with a myc antibody (red) to localize double-labeled Evx1 interneurons (yellow) in the ventral spinal cord. In Evx1tm/ϩ embryos, mycϩ/dextranϩ cells were located contralateral to the injection site, as seen in wild-type embryos (Figure 3). In Evx1tm/tm embryos, mycϩ/dextranϩ cells were located ipsilateral to the injection site, indicating that Evx1 V0 interneurons are converted to a V1 identity following the loss of Evx1. (B) Quantitation of double-labeled cells in wild-type and Evx1tm/tm embryos. The majority of wild-type Evx1 interneurons are commissural (red) and project rostrally for one to four segments after crossing the floor plate. The majority of Evx1tm/tm interneurons are ipsilateral (blue) and project rostrally for only one to two segments, faithfully recapitulating the En1 projection phenotype. (C) Schematic summary of the V0 interneuron behavior in Evx1 mutants. In Evx1tm/ϩ embryos, Evx1 interneurons migrate ventromedially (arrows) toward the floorplate and elaborate a commissural axon that projects rostrally for one to four segments after crossing the floor plate (red). Evx1 Is a Determinant of Ventral Interneuron Fate 395

Figure 8. Misexpression of Evx1 Can Repress Endogenous Expression of En1 Cross sections of stage 23 chick spinal cords electroporated at stage 11 with a CS-Evx1/EGFP expression vector (A–F) or a control CS-EGFP expression vector (G–H). Multiple sections (12–15 per embryo) were examined for changes in gene expression in regions exhibiting high level EGFP expression and representative sections were chosen. Sections were stained with antibodies to EGFP (green), En1 ([B and H], red), Isl1 ([C and D], red) and Pax6 ([E and F], red). (A–B) Evx1 represses endogenous expression of En1. The arrowhead in (B) indicates a single remaining En1ϩ cell located dorsal to a patch of Evx1/EGFP expression. (C–F) Misexpression of Evx1 does not affect Isl1 expression in dorsal interneurons or motor neurons ([C and D], arrowheads), nor does it alter Pax6 expression in the ventricular zone (E and F). (G and H) EGFP alone does not affect endogenous expression of En1 (arrowhead). (I) Model for transcriptional regulation of V0 interneuron specification. Dbx1 activates Evx1 in newly differentiated V0 interneurons, which in turn represses En1 expression and the V1 interneuron differentiation program. Dashed lines mark the neural tube. below and are incorporated with the current model of early embryonic lethality. In contrast, our results show neuronal patterning in the developing spinal cord. that mice lacking exon 3 of Evx1 are viable and survive to adulthood. We believe the Evxtm allele is a complete Evx1 Null Mutants Are Viable null for the following reasons. First, Cre-mediated re- It has been previously reported that targeting of the Evx1 combination deletes the recognition helix of the Evx1 gene causes early embryonic lethality due to a loss of homeodomain, including Asn 51 (Kissinger et al., 1990), extraembryonic tissues (Spyropoulos and Capecchi, which is necessary for DNA binding. Second, the recom- 1994). This finding prompted us to use the Cre-loxP bined Evx1 allele lacks the conserved C-terminal domain recombination system to generate a conditional null al- that mediates transcriptional repression (Han and Man- lele of Evx1. In their study, Spyropoulos and Capecchi ley, 1993; Briata et al., 1997). Third, the truncated Evx1 (1994) targeted Evx1 by inserting a neomycin gene ex- protein is localized to the cytoplasm (data not shown), pression cassette into exon 2 and obtained a single ES thereby interfering with any residual function the Evx1 cell clone that transmitted through the germline. How- N-terminal fragment might retain. Finally, three addi- ever, mice derived from this cell line have a second tional strains of Evx1 null mice have been generated closely linked mutation (D. Spyropoulos, personal com- independently and, in all cases, produce viable homozy- munication) that is now thought to underlie the observed gous offspring (D. Goldman and G. Martin, personal

In Evx1tm/tm embryos, Evx1 V0 interneurons migrate ventrolaterally (arrows) toward the lateral ventral horns and project ipsilaterally and rostrally for one to two segments (blue), indicating that Evx1 V0 interneurons are clearly transfated to a V0 identity in Evx1tm mutants. VZ, ventricular zone; MNs, motor neurons. Neuron 396

communication; D. Spyropoulos, personal communica- and the migratory routes they follow in the mantle zone. tion). Consequently, Evx1 function is not necessary for We suggest that neuronal cell bodies are towed by their the differentiation of extraembryonic tissues in early em- growing axons, which are elaborated prior to the onset bryogenesis as previously proposed (Spyropoulos and of soma migration. Normally, Evx1 V0 interneurons mi- Capecchi, 1994). grate along a ventromedial pathway that follows the trajectory of their commissural axons toward the floor Evx1 Defines a Population of Intersegmental plate (Figure 4). These behaviors distinguish V0 in- Commissural Interneurons terneurons from V1 and V2 interneurons that project We have shown that Evx1 demarcates a class of ventral ipsilaterally and migrate along a ventrolateral pathway commissural interneurons that project axons rostrally (Saueressig et al., 1999). In contrast, a large proportion for one to four spinal cord segments in the ventral funicu- of the Evx1 interneurons (65%) in Evx1 mutant mice lus, where they terminate close to motor neurons (Figure adopt the features of V1 interneurons in that they project 3). While the function of Evx1 V0 interneurons in the axons ipsilaterally and migrate toward the lateral ventral adult spinal cord is not known, a comparison of their horn (Figures 4 and 7). In Figure 4, an example of an cell body position and projections with those of identi- Evx1 interneuron with a bifurcated axon whose longest fied adult spinal neurons suggests that they may func- branch projects laterally is shown, leading us to propose tion in the crossed reflex circuits that coordinate bilat- that these lateral axon collaterals may reorient the mi- eral motor neuron activity. During development, Evx1 gration of Evx1 interneurons. Accordingly, we observe V0 commissural interneurons progressively accumulate that respecified Evx1 interneurons do not immediately in the medial ventral horn, in a domain that will form enter the ventrolateral migration pathway and instead lamina VIII in the adult spinal cord (Figure 1). In the adult migrate a short distance ventrally toward the floor plate cat, interneurons that synapse directly with contralateral before turning laterally. We have also observed that the motor neurons are located primarily in laminae VIII (Har- cell bodies of V0 interneurons are more broadly distrib- rison et al., 1986). These interneurons receive monosyn- uted in the ventral horns of netrin-1 mutant mice, as aptic inputs from Ia and Ib sensory afferents and form a compared to wild-type mice, and this phenotype corre- disynaptic crossed reflex pathway that controls walking lates closely with the axon projections of commissural movements (Jankowska and Noga, 1990). neurons in netrin-1 mutants (L. M.-R. and M. G., unpub- V0 interneurons may also contribute to central pattern lished data; Serafini et al., 1996). These and other obser- generator (CPG) activity in the spinal cord. CPGs control vations argue that axon guidance plays a primary role the rhythmic firing of motor neurons that underlies be- in positioning the cell bodies of neurons within the spinal haviors, such as walking, that require left/right alterna- cord. tions of motor activity (Grillner, 1975; Delcomyn, 1980). Our analysis of the Evx1 mutant phenotype reveals Lesion studies in the neonatal rat demonstrate that the an additional function for Evx1 in dictating later aspects generation of left/right alternations depends on neurons of V0 interneuron identity. In Evx1 mutants, V0 interneur- in the ventral third of the spinal cord that have axons ons that retain a commissural projection fail to extend crossing in the ventral commissure (Kjaerulff and Kiehn, axons rostrally in the contralateral ventral funiculus (Fig- 1996, 1997). Based on the position of their soma, com- ure 7), demonstrating a possible tonic requirement for missural interneurons in the neonatal rat spinal cord can Evx1 in the elaboration of intersegmental projections. be classified into four groups: dorsomarginal, dorsal, This altered axonal morphology could also reflect a func- centromarginal, and ventral (Eide et al., 1999). The ven- tion for Evx2 in V0 interneuron pathfinding, since Evx2 tral population most likely corresponds to the interneur- expression in V0 interneurons is dependent on Evx1 ons identified by Kjaerulff and Kiehn (1996) and is lo- (Figure 5). In Drosophila, Even-skipped, the fly homolog cated at the same position V0 interneurons cluster in of Evx1, is required for the axons of motor neurons in the embryonic spinal cord. the intersegmental nerve (ISN) to project to and inner- Evx1tm/tm mice that survive to adulthood do not exhibit vate their dorsal muscle targets (Landgraf et al., 1999). gross motor deficits and changes in their gait as a result This study, together with our findings, suggests a con- of alterations in Evx1 interneuron connectivity. One rea- served role for the Evx/Even-skipped class of homeo- son for the observed absence of gross motor deficits domain transcription factors in regulating axon guid- in Evx1 mutants may be the presence of a class of Dbx1 derived interneurons located just dorsal to the Evx1 pop- ance, albeit in different neuronal cell types. While it is ulation that shares many of the features of V0 interneur- likely that Evx1/2 and Even-skipped regulate the expres- ons (Pierani et al., 2001). It is also possible that some sion of genes that mediate responsiveness to extracellu- Evx1 V0 interneurons with truncated commissural axons lar guidance cues, the identity of these gene targets, still synapse with contralateral motor neurons; however, and the extent to which they are conserved in verte- because these axons fail to project along the rostrocau- brates and flies, remains to be determined. dal axis, any such connections would be intrasegmental rather than intersegmental. Currently, more sensitive be- Some V0 Interneurons Do Not Require Evx1 havioral studies are being undertaken to search for sub- to Project Commissurally tle changes in locomotor reflexes. Our studies demonstrate that Evx1 is required in postmitotic neurons to fully enact the V0 interneuron differentia- Evx1 Coordinates Axon Pathfinding tion program that requires Dbx1 activity in V0 progenitors. and Cell Migration However, not all V0 interneurons respond identically to Our studies reveal a close correlation between the axo- the loss of Evx1. At E12.5, approximately one-third of nal projections of different spinal interneuron classes Evx1 interneurons in Evx1 mutants fail to upregulate En1 Evx1 Is a Determinant of Ventral Interneuron Fate 397

and subsequently continue to migrate toward the floor Our findings and those of Pierani et al. (2001) suggest plate. Although these neurons still have axons that reach that Dbx1 represses the V1 ipsilateral program that is the ventral midline, the axons fail to extend rostrally in the associated with En1 expression, possibly by upregulat- contralateral ventral funiculus and do not acquire a mature ing Evx1, and that a V1 identity may constitute the de- V0 morphology. It remains unclear why this subset of fault differentiation state of Dbx2ϩ progenitors in the Evx1 V0 interneurons can enact a partial V0 identity ventral spinal cord. This model is consistent with the program in the absence of Evx1 function. These V0-like finding that Evx1 can suppress the V1 interneuron interneurons with truncated axons may arise stochas- marker En1 in regions where V1 interneurons normally tically, or unidentified genes may be sufficient to dictate develop (Figure 8). Our finding that V0 interneurons can early aspects of V0 interneuron differentiation in this differentiate as V1 interneurons despite being patterned population. While the evidence that Evx2 does not com- appropriately in the ventricular zone, shows that there pensate for the loss of Evx1 is compelling, it is possible is a critical period after neuroblasts exit the cell cycle that transient low level expression of Evx2 prior to E10.5 during which their identity remains plastic. However, generates a class of early-born V0 interneurons that the ability of V0 interneurons to adopt the identity of a extend truncated commissural projections. different interneuron class is restricted, a likely reflection Our studies also show that in Evx1 mutant embryos, of the prior patterning of progenitors in the ventricular the majority of Evx1 interneurons migrate along a ventro- zone. Consequently, in Evx1 mutants, V0 interneurons lateral pathway, upregulate ectopic En1 expression and can transfate to V1 interneurons, but not to Lbx1ϩ dorsal adopt a V1-like interneuron identity. However, these interneurons or V2 interneurons. In summary, our find- Evx1 interneurons do not immediately enter a ventrolat- ings suggest that Evx1 is a critical regulatory link be- eral migratory pathway and instead migrate a short dis- tween the early patterning events that generate V0 pro- tance ventromedially from the generation zone (Figure genitors and their differentiation as V0 commissural 4). These findings suggest a model in which Dbx1,in interneurons. combination with other patterning genes, initiates the Our results showing V0 interneurons can adopt a V1 V0 differentiation program in the absence of Evx1/2 fate in Evx1 mutants, irrespective of their patterning in function. Therefore, V0 interneurons may not require the ventricular zone, demonstrate that neuronal precur- Evx1 for the earliest phase of commissural axon out- sors exit the cell cycle with a less well-established iden- growth and cell body migration. This model is consistent tity than previously recognized. Recently, Briscoe et al. with the observed bifurcated axon morphology of Evx1 (2000) proposed a three-step model for the generation interneurons (Figure 4) and with the delayed onset of of neuronal subtypes in the ventral neural tube. First, En1 expression in these cells, which presumably reflects Shh, secreted by the notochord and floor plate, estab- the point at which Dbx1 can no longer direct V0 in- lishes a dorsoventral gradient of Shh-dependent signal- terneuron differentiation and the Evx1 interneurons being that generates localized domains of homeodomain gin to switch their identity. protein expression in the ventricular zone. Second, this homeodomain code is further refined by cross-repres- Evx1 Is a Postmitotic Determinant of V0 sive interactions between neighboring homeodomain Interneuron Fate proteins, thereby establishing sharply restricted progen- Although differences exist in the spinal cord phenotypes itor domains within the ventricular zone. Third, cells that result from the inactivation of Dbx1 and Evx1,a ϩ within these discrete progenitor domains initiate unique similar increase in En1 interneurons coupled with a subtype-specific neuronal differentiation programs. While loss of V0 interneurons is seen in both mutants (this the study by Briscoe et al. (2000) elegantly demonstrates study; Pierani et al., 2001). V0 and V1 interneurons are the important role patterning genes play in defining neu- generated from Pax6ϩ/Dbx2ϩ progenitors that are ven- ronal progenitor populations, it did not address whether tral to the Pax3/Pax7 expression domain (Pierani et al., homeodomain proteins that are expressed in postmitotic 1999). This ventral zone is further subdivided into two neuronal subtypes also determine interneuron identity. progenitor domains by the expression of Dbx1: a dorsal Our study defines an additional step to the aforemen- Pax6ϩ/Dbx1ϩ/Dbx2ϩ domain that generates Evx1 V0 in- tioned model. In this fourth step, we propose the pres- terneurons, and a ventral Pax6ϩ/Dbx2ϩ/Dbx1Ϫ domain ence of a class of neural identity factors, as exemplified that gives rise to En1 V1 interneurons (Pierani et al., by Evx1, that are expressed in early postmitotic neuronal 1999). A second population of progenitors dorsal to the sulcus limitans also expresses Dbx1, and it appears that subtypes that direct subtype-specific neuronal differen- these cells do not differentiate as V1 interneurons in tiation programs in postmitotic neurons. We argue that Dbx1 mutant embryos. Instead, the number of LacZϩ/ the activation of these postmitotic neural determinants En1ϩ neurons in Dbx1-LacZ mutants roughly corre- in response to the combinatorial action of patterning sponds to the size of the Evx1 V0 population, indicating genes is a mechanism that ensures the uninterrupted that only those neurons that derive from ventral Dbx1 flow of positional information from neural progenitors precursors and express Evx1 activate ectopic En1 ex- to differentiating postmitotic neurons as they leave the pression (Pierani et al., 2001). In Dbx1 mutants, V0 in- ventricular zone. terneurons that migrate laterally and express En1 do not extend axons rostrally for one to two segments as Experimental Procedures is seen in Evx1 mutant embryos. Since Dbx1 is upstream Immunohistochemistry and In Situ Hybridization of Evx1, it is likely to control the expression of additional A GST fusion protein containing the first 192 amino acids of the genes, including genes that may direct the rostral Evx1 protein was used to raise polyclonal antibodies in both rabbits growth of axons (see Pierani et al., 2001). and rats. An Lbx1 rat polyclonal antibody was generated using a Neuron 398

GST fusion protein that includes 122 amino acids of the Lbx1 C 0.2% fast green) or a control CS-EGFP vector. Embryos were then terminus. The following commercially available antibodies were electroporated (BTX Electro Square Porator T820) unilaterally with used: En1 (mouse monoclonal, Developmental Studies Hybridoma five 25V pulses lasting 50 ms each. Following a 48 hr incubation Bank [DSHB]), Isl1 (mouse monoclonal, DSHB), GFP (rat polyclonal, at 38ЊC, stage 23 embryos were harvested and scored for GFP Molecular Probes), BrdU (mouse monoclonal, Calbiochem), Pax6 expression. Embryos exhibiting high GFP fluorescence were fixed (mouse monoclonal, DSHB) and TAG-1 (mouse monoclonal, DSHB). in 4% paraformaldehyde/PBS and processed for cryosectioning. The anti-Chx10 (guinea pig polyclonal) and anti-myc (rabbit poly- Sections double stained with antibodies to En1 or Evx1 (data not clonal) antibodies were kind gifts of Sam Pfaff. Anti-Evx1/2 (mouse shown) and GFP were analyzed by confocal microscopy. monoclonal) and anti-Dbx1 (rabbit polyclonal) were provided by Alessandra Pierani and Tom Jessell (Pierani et al., 1999). Immuno- Acknowledgments stainings were performed as described in Gross et al. (2000). Radio- active in situ hybridization was performed on E11.5 mouse cryosec- We would like to thank Mirella Dottori, Tamar Sapir, Guillermo La- tions as described (Goulding et al., 1993), using an 35S-labeled nuza, Greg Lemke, Josh Thaler, Sam Pfaff, and Tom Jessell for their antisense RNA Evx2 probe (Dolle et al., 1994). valuable comments on the manuscript. The anti-Chx10 and anti- myc antibodies were kind gifts from Sam Pfaff. Anti-Evx1/2 was BrdU Labeling provided by Alessandra Pierani and Tom Jessell. We thank Stephan E10.5 mouse embryos were pulsed for 2 hr in utero with bromodeox- Thor for the IRES-tau-myc neuronal reporter cassette, Tom Maynard yuridine (50 mg/ml in 0.9% saline injected i.p.). Embryos were har- for the CS-EGFP expression vector, and Yann Herault and Denis vested, fixed for 30 min in 4% paraformaldehyde/PBS, and incu- Deboule for the Evx2 in situ probe. Technical assistance was pro- bated in 20% sucrose/PBS at 4ЊC overnight. Embryos were then vided by Martin Nakatsu and Joseph Wu. In addition, we thank embedded in OCT (Tissue-Tec) and sectioned at 20 ␮m on a cryo- Steve O’Gorman for the Protamine-Cre transgenic mice. We would stat. Sections were stained with anti-Evx1 and then processed for also like to thank Gail Martin and Demetri Spyropoulos for communi- staining with anti-BrdU (4% paraformaldehyde/PBS 5 min; PBS 3 ϫ cating unpublished data about the Evx1 mutant phenotype. This 5 min; 2N HCl 20 min; PBS 3 ϫ 5 min; PBT [0.2% Triton X-100] 20 work was supported by grants to M. G. from the National Institutes min; 0.1 M Boric Acid [pH 8.5] 20 min; PBS 3 ϫ 10 min; 10% FBS/ of Health, the Pew Charitable Trust, and the Fritz Burns Foundation. PBT 10 min; anti-BrdU in 10% FBS/PBT overnight at 4ЊC). T. K. was supported by a Human Frontiers Fellowship, J. B. by an NRSA Fellowship, H. S. by a postdoctoral fellowship from the Retrograde Labeling with Fluorescein-Dextran Deutsche Forschungsgemeinschaft, and L. M.-R. by a Predoctoral E12.5 mouse embryos were harvested and dissected ventrally to NIH Training Fellowship. reveal the spinal cord. One side of the ventral spinal cord was cut at the level of T1, and a crystal of lysine fixable fluorescein- Received November 9, 2000; revised January 10, 2000. conjugated dextran (10 mg/ml, Molecular Probes) was applied to the cut site. Embryos were incubated in transport solution (5 M References

NaCl, 1 M KCl, 1 M CaCl2, 24 mM Na bicarbonate, 1 M MgCl2,1M Allum, J.H., Honegger, F., and Pfaltz, C.R. (1989). The role of stretch NaH2PO4, 11 mM glucose [pH 7.4]) with a supply of carbogen. After 5 hr, embryos were fixed for 45 min in 4% paraformaldehyde/PBS and vestibulo-spinal reflexes in the generation of human equilibrat- and processed for cryosectioning. Sections were stained with an ing reactions. Prog. Brain Res. 80, 399–409. antibody to Evx1 and a Cy3-conjugated secondary antibody. Con- Bastian, H., and Gruss, P. (1990). A murine even-skipped homo- secutive sections were analyzed using a Zeiss LSM 510 confocal logue, Evx 1, is expressed during early embryogenesis and neuro- microscope. The positions of Evx1ϩ/dextranϩ cells were determined genesis in a biphasic manner. EMBO J. 9, 1839–1852. Erratum: relative to the injection site. EMBO J. 10(12), 1991. Briata, P., Ilengo, C., Van DeWerken, R., and Corte, G. (1997). Map- Generation of Evx1tau-myc Mice ping of a potent transcriptional repression region of the human The Evx1tau-myc targeting vector (Figure 2A) was generated by intro- homeodomain protein EVX1. FEBS Lett. 402, 131–135. Ј ducing the PGKneopA and IRES-tau-myc cassettes into the 3 UTR Briscoe, J., Sussel, L., Serup, P., Hartigan-O’Connor, D., Jessell, of the Evx1 locus. LoxP sites were inserted into intron 2 and down- T.M., Rubenstein, J.L., and Ericson, J. (1999). Homeobox gene stream of the neo gene to create a conditional allele. An HSV-thymi- Nkx2.2 and specification of neuronal identity by graded Sonic Ј dine kinase expression cassette was inserted into the 5 UTR of Evx1 hedgehog signalling. Nature 398, 622–627. to aid elimination of nonhomologous recombinant clones following Briscoe, J., Pierani, A., Jessell, T.M., and Ericson, J. (2000). A homeo- electroporation. CCE ES cells were electroporated and selected domain protein code specifies progenitor cell identity and neuronal using standard techniques. ES cell clones were screened for homol- fate in the ventral neural tube. Cell 101, 435–445. ogous recombination (Figure 2B) using an internal 5Ј Sal-Pst DNA probe (Kpn1 genomic digest) and a 3Ј BglII-Xba probe (Xba genomic Burrill, J.D., Moran, L., Goulding, M.D., and Saueressig, H. (1997). digest). Two recombinant ES cell clones were microinjected into PAX2 is expressed in multiple spinal cord interneurons, including a C57Bl6 blastocysts and implanted into pseudopregnant ICR fosters population of EN1ϩ interneurons that require PAX6 for their devel- to produce chimeric offspring. Highly chimeric males were bred to opment. Development 124, 4493–4503. C57Bl6 females, and both lines exhibited germline transmission. To Chiang, C., Litingtung, Y., Lee, E., Young, K.E., Corden, J.L., West- generate heterozygous Evx1 null mice, Evx1ϩ/Ϫ males were crossed phal, H., and Beachy, P.A. (1996). Cyclopia and defective axial pat- with female Protamine-Cre transgenic mice (O’Gorman et al., 1997), terning in mice lacking Sonic hedgehog gene function. Nature 383, producing Evx1tm/Ϫ;Pro-Creϩ/Ϫ progeny. Double heterozygous males 407–413. tau-myc were bred to ICR females to establish the recombined Evx1 Davis, C.A., Holmyard, D.P., Millen, K.J., and Joyner, A.L. (1991). tau-myc colony. Homozygous Evx1 mice were generated by breeding Examining pattern formation in mouse, chicken and frog embryos tm/ϩ tm/ϩ tm/tm Evx1 males with Evx1 or Evx1 females. For harvesting with an En-specific antiserum. Development 111, 287–298. embryos, the plug date was designated as E0.5. Genotyping of Delcomyn, F. (1980). Neural basis of rhythmic behavior in animals. Evx1tau-myc mice and embryos was performed by Southern and PCR Science 210, 492–498. analyses. Tau primers were used in combination with Evx1 primers to genotype both heterozygous and homozygous Evx1tau-myc mice. Dodd, J., Morton, S.B., Karagogeos, D., Yamamoto, M., and Jessell, T.M. (1988). Spatial regulation of axonal glycoprotein expression on subsets of embryonic spinal neurons. Neuron 1, 105–116. In Ovo Electroporations A cDNA fragment encoding the full-length Evx1 protein was cloned Dolle, P., Fraulob, V., and Duboule, D. (1994). Developmental expres- into a CS-EGFP expression vector containing SV40-GFP (a kind gift sion of the mouse Evx-2 gene: relationship with the evolution of the of Dr. Tom Maynard). A picospritzer was used to inject the neural HOM/Hox complex. Dev. Suppl. 143–153. tubes of stage 11 chick embryos with CS-Evx1/EGFP (4 mg/ml with Eide, A.L., Glover, J., Kjaerulff, O., and Kiehn, O. (1999). Character- Evx1 Is a Determinant of Ventral Interneuron Fate 399

ization of commissural interneurons in the lumbar region of the Marti, E., Takada, R., Bumcrot, D.A., Sasaki, H., and McMahon, A.P. neonatal rat spinal cord. J. Comp. Neurol. 403, 332–345. (1995). Distribution of Sonic hedgehog peptides in the developing Ericson, J., Thor, S., Edlund, T., Jessell, T.M., and Yamada, T. (1992). chick and mouse embryo. Development 121, 2537–2547. Early stages of motor neuron differentiation revealed by expression Matise, M.P., and Joyner, A.L. (1997). Expression patterns of devel- of homeobox gene Islet-1. Science 256, 1555–1560. opmental control genes in normal and Engrailed-1 mutant mouse Ericson, J., Morton, S., Kawakami, A., Roelink, H., and Jessell, T.M. spinal cord reveal early diversity in developing interneurons. J. Neu- (1996). Two critical periods of Sonic Hedgehog signaling required rosci. 17, 7805–7816. for the specification of motor neuron identity. Cell 87, 661–673. Nornes, H.O., Dressler, G.R., Knapik, E.W., Deutsch, U., and Gruss, Ericson, J., Rashbass, P., Schedl, A., Brenner-Morton, S., Kawa- P. (1990). Spatially and temporally restricted expression of Pax2 kami, A., van Heyningen, V., Jessell, T.M., and Briscoe, J. (1997). during murine neurogenesis. Development 109, 797–809. Pax6 controls progenitor cell identity and neuronal fate in response O’Gorman, S., Dagenais, N.A., Qian, M., and Marchuk, Y. (1997). to graded Shh signaling. Cell 90, 169–180. Protamine-Cre recombinase transgenes efficiently recombine target Grillner, S. (1975). Locomotion in vertebrates: central mechanisms sequences in the male germ line of mice, but not in embryonic stem and reflex interaction. Physiol. Rev. 55, 247–304. cells. Proc. Natl. Acad. Sci. USA 94, 14602–14607. Gross, M.K., Moran-Rivard, L., Velasquez, T., Nakatsu, M.N., Jagla, Pierani, A., Brenner-Morton, S., Chiang, C., and Jessell, T.M. (1999). K., and Goulding, M. (2000). Lbx1 is required for muscle precursor A sonic hedgehog-independent, retinoid-activated pathway of neu- migration along a lateral pathway into the limb. Development 127, rogenesis in the ventral spinal cord. Cell 97, 903–915. 413–424. Pierani, A., Moran-Rivard, L., Sunshine, M.J., Littman, D.R., Gould- Goulding, M.D., Lumsden, A., and Gruss, P. (1993). Signals from the ing, M., and Jessell, T.M. (2001). Control of interneuron fate in the notochord and floor plate regulate the region-specific expression developing spinal cord by the progenitor homeodomain protein of two Pax genes in the developing spinal cord. Development 117, Dbx1. Neuron, 29, this issue, 367–384. 1001–1016. Rekling, J.C., Funk, G.D., Bayliss, D.A., Dong, X.W., and Feldman, Gu, H., Marth, J.D., Orban, P.C., Mossmann, H., and Rajewsky, K. J.L. (2000). Synaptic control of motoneuronal excitability. Physiol. (1994). Deletion of a DNA polymerase beta gene segment in T cells Rev. 80, 767–852. using cell type-specific gene targeting. Science 265, 103–106. Sander, M., Paydar, S., Ericson, J., Briscoe, J., Berber, E., German, Han, K., and Manley, J.L. (1993). Transcriptional repression by the M., Jessell, T.M., and Rubenstein, J.L. (2000). Ventral neural pat- Drosophila even-skipped protein: definition of a minimal repression terning by Nkx homeobox genes: Nkx6.1 controls somatic motor domain. Genes Dev. 7, 491–503. neuron and ventral interneuron fates. Genes Dev. 14, 2134–2139. Harrison, P.J., Jankowska, E., and Zytnicki, D. (1986). Lamina VIII Saueressig, H., Burrill, J., and Goulding, M. (1999). Engrailed-1 and interneurones interposed in crossed reflex pathways in the cat. J. Netrin-1 regulate axon pathfinding by association interneurons that Physiol. (Lond.) 371, 147–66. project to motor neurons. Development 126, 4201–4212. Jagla, K., Dolle, P., Mattei, M.G., Jagla, T., Schuhbaur, B., Dretzen, Serafini, T., Colamarino, S.A., Leonardo, E.D., Wang, H., Beddington, G., Bellard, F., and Bellard, M. (1995). Mouse Lbx1 and human LBX1 R., Skarnes, W.C., and Tessier-Lavigne, M. (1996). Netrin-1 is re- define a novel mammalian homeobox gene family related to the quired for commissural axon guidance in the developing vertebrate Drosophila lady bird genes. Mech. Dev. 53, 345–356. nervous system. Cell 87, 1001–1014. Jankowska, E., and Noga, B.R. (1990). Contralaterally projecting Silos-Santiago, I., and Snider, W.D. (1992). Development of commis- lamina VIII interneurones in middle lumbar segments in the cat. Brain sural neurons in the embryonic rat spinal cord. J. Comp. Neurol. Res. 535, 327–330. 325, 514–526. Jessell, T.M. (2000). Neuronal specification in the spinal cord: induc- Silos-Santiago, I., and Snider, W.D. (1994). Development of in- tive signal and transcriptional codes. Nat. Rev. Gen. 1, 20–29. terneurons with ipsilateral projections in embryonic rat spinal cord. J. Comp. Neurol. 342, 221–231. Kissinger, C.R., Liu, B.S., Martin-Blanco, E., Kornberg, T.B., and Pabo, C.O. (1990). Crystal structure of an engrailed homeodomain- Spyropoulos, D.D., and Capecchi, M.R. (1994). Targeted disruption DNA complex at 2.8 A˚ resolution: a framework for understanding of the even-skipped gene, evx1, causes early postimplantation le- homeodomain-DNA interactions. Cell 63, 579–590. thality of the mouse conceptus. Genes Dev. 8, 1949–1961. Kjaerulff, O., and Kiehn, O. (1996). Distribution of networks generat- Tanabe, Y., William, C., and Jessell, T.M. (1998). Specification of ing and coordinating locomotor activity in the neonatal rat spinal motor neuron identity by the MNR2 homeodomain protein. Cell 95, cord in vitro: a lesion study. J. Neurosci. 16, 5777–5794. 67–80. Kjaerulff, O., and Kiehn, O. (1997). Crossed rhythmic synaptic input Thor, S., and Thomas, J.B. (1997). The Drosophila islet gene governs to motoneurons during selective activation of the contralateral spi- axon pathfinding and neurotransmitter identity. Neuron 18, 397–409. nal locomotor network. J. Neurosci. 17, 9433–9447. Vaughn, J.E., and Grieshaber, J.A. (1973). A morphological investiga- Landgraf, M., Roy, S., Prokop, A., VijayRaghavan, K., and Bate, M. tion of an early reflex pathway in developing rat spinal cord. J. (1999). Even-skipped determines the dorsal growth of motor axons Comp. Neurol. 148, 177–209. in Drosophila. Neuron 22, 43–52. Wenner, P., Matise, M.P., Joyner, A., and O’Donovan, M.J. (1998). Leber, S.M., and Sanes, J.R. (1995). Migratory paths of neurons and Physiological and molecular characterization of interneurons in the glia in the embryonic chick spinal cord. J. Neurosci. 15, 1236–1248. developing spinal cord. Ann. NY Acad. Sci. 860, 425–427. Leber, S.M., Breedlove, S.M., and Sanes, J.R. (1990). Lineage, ar- Yamada, T., Pfaff, S.L., Edlund, T., and Jessell, T.M. (1993). Control rangement, and death of clonally related motoneurons in chick spi- of cell pattern in the neural tube: motor neuron induction by diffusible nal cord. J. Neurosci. 10, 2451–2462. factors from notochord and floor plate. Cell 73, 673–686. Lee, K.J., Mendelsohn, M., and Jessell, T.M. (1998). Neuronal pat- Zhadanov, A.B., Bertuzzi, S., Taira, M., Dawid, I.B., and Westphal, terning by BMPs: a requirement for GDF7 in the generation of a H. (1995). Expression pattern of the murine LIM class homeobox discrete class of commissural interneurons in the mouse spinal cord. gene Lhx3 in subsets of neural and neuroendocrine tissues. Dev. Genes Dev. 12, 3394–3407. Dyn. 202, 354–364. Litingtung, Y., and Chiang, C. (2000). Specification of ventral neuron Zhou, Y., Yamamoto, M., and Engel, J.D. (2000). GATA2 is required types is mediated by an antagonistic interaction between Shh and for the generation of V2 interneurons. Development 127, 3829–3838. Gli3. Nat. Neurosci. 10, 979–985. Liu, I.S., Chen, J.D., Ploder, L., Vidgen, D., van der Kooy, D., Kalnins, V.I., and McInnes, R.R. (1994). Developmental expression of a novel murine homeobox gene (Chx10): evidence for roles in determination of the neuroretina and inner nuclear layer. Neuron 13, 377–393.