Genetic Identification of Spinal Interneurons PDF

Neuron, Vol. 42, 375–386, May 13, 2004, Copyright ©2004 by Cell Press Genetic Identification of Spinal Interneurons that Coordinate Left-Right Locomotor Activity Necessary for Walking Movements Guillermo M. Lanuza,1,3 Simon Gosgnach,1,3 that directs swimming movements c...

Neuron, Vol. 42, 375–386, May 13, 2004, Copyright ©2004 by Cell Press Genetic Identification of Spinal Interneurons that Coordinate Left-Right Locomotor Activity Necessary for Walking Movements Guillermo M. Lanuza,1,3 Simon Gosgnach,1,3 that directs swimming movements comprises multiple Alessandra Pierani,2,4 Thomas M. Jessell,2 burst generators organized into “half centers” located and Martyn Goulding1,* on each side of the spinal cord (Arshavsky et al., 1993; 1 Molecular Neurobiology Laboratory Orlovsky et al., 1999). These half centers are mutually The Salk Institute for Biological Studies inhibitory and connected by commissural interneurons 10010 North Torrey Pines Road (Arshavsky et al., 1993; Soffe et al., 1984; Cohen and La Jolla, California 92037 Harris-Warrick, 1984). 2 Howard Hughes Medical Institute Much less is known about the neuronal organization Department of Biochemistry and Molecular of the locomotor CPG in walking mammals, primarily Biophysics because of the difficulty in identifying and manipulating Center for Neurobiology and Behavior its intrinsic interneuronal components. Many of the neu- Columbia University rons that contribute to the mammalian locomotor CPG New York, New York 10032 are thought to be located in the ventromedial spinal cord (termed lamina VIII), a region abundant in commissural interneurons that project to the vicinity of contralateral Summary motor neurons (Harrison et al., 1986; Jankowska and Noga, 1990; Butt et al., 2002; Stokke et al., 2002; Birinyi The sequential stepping of left and right limbs is a et al., 2003; Butt and Kiehn, 2003). Some lamina VIII fundamental motor behavior that underlies walking commissural interneurons form monosynaptic connec- movements. This relatively simple locomotor behavior tions with motor neurons (Harrison et al., 1986; Jankow- is generated by the rhythmic activity of motor neurons ska and Noga, 1990; Butt et al., 2002; Stokke et al., under the control of spinal neural networks known 2002; Birinyi et al., 2003; Butt and Kiehn, 2003) and are as central pattern generators (CPGs) that comprise rhythmically active during locomotion (Butt et al., 2002; multiple interneuron cell types. Little, however, is Butt and Kiehn, 2003), suggesting they have roles in known about the identity and contribution of defined coordinating bilateral flexor and extensor locomotor ac- interneuronal populations to mammalian locomotor tivity. Consistent with this view, severing the ventral behaviors. We show a discrete subset of commissural commissure results in a loss in coordinated motor neu- spinal interneurons, whose fate is controlled by the ron activity in the left and right halves of the spinal cord activity of the homeobox gene Dbx1, has a critical (Kjaerulff and Kiehn, 1996; Cowley and Schmidt, 1997; role in controlling the left-right alternation of motor Kato, 1990). neurons innervating hindlimb muscles. Dbx1 mutant A more precise analysis of the logic by which CPG mice lacking these ventral interneurons exhibit an in- circuits control locomotor behavior in mammals requires creased incidence of cobursting between left and right the ability to identify and genetically manipulate the flexor/extensor motor neurons during drug-induced function of individual interneuronal populations and to locomotion. Together, these findings identify Dbx1- examine the consequences for locomotion. Recent ge- dependent interneurons as key components of the spi- netic studies have shown that inactivation of the recep- nal locomotor circuits that control stepping move- tor tyrosine kinase EphA4 results in defects in spinal ments in mammals. locomotor function (Kullander et al., 2003); however, identifying the neurons affected by this mutation has Introduction been difficult, due to widespread EphA4 expression in the spinal cord (Leighton et al., 2001; Dottori et al., 1998). The simple repetitive movements that underlie locomo- One potential strategy for selectively manipulating de- tion in nonsessile animals are generated by localized fined sets of CPG interneurons has emerged from recent neural networks known as central pattern generators findings that interneuron subtypes and their progenitors (CPGs) (Arshavsky et al., 1993; Pearson, 1993; Orlovsky in the developing spinal cord can be distinguished by et al., 1999). These CPG circuits provide an informative the restricted expression of homeodomain transcription model system for studying how neuronal networks gen- factors (Jessell, 2000; Lee and Pfaff, 2001; Goulding et erate simple behaviors. The local interneuron circuits al., 2002). These molecularly defined spinal interneuron that contribute to the vertebrate locomotor CPG reside subtypes appear to be related to functional neuronal in the spinal cord and generate the elemental patterns subclasses in the developing and mature spinal cord of motor activity that underlie swimming and walking (Goulding et al., 2002; Wenner et al., 2000; Sapir et al., movements (Orlovsky et al., 1999; Graham-Brown, 1914; 2004). Moreover, targeted inactivation of these homeo- Grillner and Zangger, 1979; Kjaerulff and Kiehn, 1996; box genes in mice results in the respecification of cell Cowley and Schmidt, 1997). In primitive vertebrates, fate, leading to the elimination of specific interneuron such as the lamprey and Xenopus, the locomotor CPG populations, while others are produced in greater num- bers (Burrill et al., 1997; Ericson et al., 1997; Briscoe et *Correspondence: [email protected] al., 1999; Pierani et al., 2001; Moran-Rivard et al., 2001). 3 These authors contributed equally to this work. Thus, the genetic elimination of defined interneuron sub- 4 Present address: CNRS-UMR 8542 Paris, France. sets through homeobox gene inactivation may permit Neuron 376 an analysis of their normal contribution to CPG circuitry interneurons that contribute to crossed locomotor path- and locomotor behavior. ways (Harrison et al., 1986). A more dorsolateral popula- Within the developing spinal cord, several distinct tion of !-gal" cells was also detected in E18.5/P0 classes of commissural inteneurons settle in lamina VIII Dbx1lacZ/" mice (Figures 1A and 1B). These cells had (Briscoe et al., 1999; Pierani et al., 2001; Moran-Rivard small nuclei and lacked NeuN expression (Figure 1A, et al., 2001; Gross et al., 2002; Muller et al., 2002; G.M.L. Int.), and many expressed GFAP (data not shown), indi- and M.G., unpublished data). Two of these interneuron cating that they are astrocytes. Thus, as with other ven- classes, which we term V0V and V0D neurons, differenti- tral spinal progenitor populations (Zhou et al., 2001), ate from a discrete progenitor domain, which is marked Dbx1 progenitors differentiate into both neuronal and by expression of the homeodomain transcription factor glial cell types: initially generating V0 interneurons that Dbx1 (Pierani et al., 2001). V0V interneurons derive from settle in and around lamina VIII, and later giving rise the ventral half of the Dbx1 progenitor domain and tran- to astrocytes. siently express the homeodomain protein Evx1 (Moran- We have addressed the fate of V0V and V0D interneu- Rivard et al., 2001), whereas V0D interneurons, which rons in Dbx1 mutant mice. Previously, we have shown lack Evx1 expression, derive from the dorsal Dbx1 do- that Evx1" V0V neurons are missing in Dbx1 mutant em- main (Pierani et al., 2001; G.M.L. and S.G., unpublished bryos, a reflection of their early respecification into ipsi- data). Both sets of Dbx1-derived interneurons exhibit laterally projecting inhibitory “V1-like” interneurons strikingly similar migratory patterns and initial commis- (Pierani et al., 2001), which express the homeodomain sural axonal trajectories (Pierani et al., 2001; Moran- protein En1 and settle close to motor neurons in lamina Rivard et al., 2001), suggesting that they are closely IX (see Supplemental Figure S1 at http://www.neuron. related both by provenance and function. org/cgi/content/full/42/3/375/DC1). In addition, the num- In this study, we have combined genetic and physio- ber of !-gal"/GFAP" cells in E18.5 Dbx1 mutants was logical approaches in the mouse to examine the role increased by !35% (data not shown), suggesting that of Dbx1-derived interneurons in spinal CPG circuitry Dbx1 progenitors destined to form V0V neurons convert and locomotion function. Elimination of Evx1 function to both V1 interneurons and astrocytes. causes a selective loss of V0V interneurons, leaving V0D The fate of V0D interneurons in Dbx1 mutant mice has, interneurons intact, whereas eliminating Dbx1 function however, remained unclear. In Dbx1 mutants analyzed results in the loss of both V0V and V0D interneurons (this at E18.5, !40% of the normal number of !-gal" neurons study; Pierani et al., 2001; Moran-Rivard et al., 2001). persisted in lamina VIII (Figures 1C and 1D), raising the We find that Evx1 mutant mice exhibit normal locomotor issue of whether these are unchanged V0D interneurons activity, whereas Dbx1 mutant mice exhibit profound or a distinct interneuronal population. dI6 commissural changes in locomotor coordination. Dbx1 mutant mice neurons, which are derived from progenitors immedi- exhibit frequent episodes of cocontraction of left and ately dorsal to the Dbx1 progenitor domain that gener- right flexor-extensor motor neurons, whereas phasic ip- ates V0D interneurons, transiently express the homeodo- silateral L2/L5 flexor-extensor activity persists normally. main transcription factor Lbx1 (Gross et al., 2002) and Our results provide genetic evidence that Dbx1-derived migrate along a similar ventromedial pathway to that of interneurons have a critical role in establishing alternat- V0D interneurons (Pierani et al., 2001; Moran-Rivard et ing left-right motor activity during locomotion and thus al., 2001; Gross et al., 2002). Furthermore, dI6 interneu- are critical components of the CPG circuitry that directs rons, like V0D interneurons, develop from Pax7", Dbx2" progenitors, but differ only in their lack of Dbx1 expres- locomotor behavior. More generally, our findings illus- sion (Pierani et al., 2001; Gross et al., 2002; Muller et trate how developmental insights into the transcriptional al., 2002). We therefore considered the possibility that control of interneuronal subtype specification can be V0D interneurons are respecified as dI6 interneurons in applied to the dissection of local neuronal circuits that the absence of Dbx1 function. Dbx1lacZ/" embryos showed regulate locomotor behavior in the mammalian spinal no overlap in the expression of Lbx1 and !-gal (Figures cord. 1E and 1F); however, in Dbx1lacZ/lacZ embryos, #80% of !-gal" neurons in lamina VIII expressed Lbx1 (Figures Results 1G and 1I), producing a !25% increase in the total number of Lbx1" dI6-like interneurons (Figure 1K). Dbx1-Dependent Specification of V0 Interneurons These !-gal" neurons also expressed Pax2 (Figure 1J), V0V and V0D commissural interneurons are generated consistent with their differentiation as dI6 interneurons between E10 and E13.5 and represent the major neu- (Gross et al., 2002). Many double-labeled !-gal"/Lbx1" ronal progeny of a discrete set of spinal cord progenitors neurons were detected in the dorsal Dbx1" subventricu- that express the homeodomain protein Dbx1 (Figure 1; lar zone of Dbx1lacZ/lacZ embryos (Figure 1H), demonstra- Pierani et al., 2001; Moran-Rivard et al., 2001). To moni- ting that prospective “V0D” neurons acquire a dI6 molec- tor V0 interneurons at later developmental stages ame- ular character soon after their generation. nable to physiological analysis, we made use of a We next examined the selectivity of ventral interneu- Dbx1lacZ knockin mouse line in which !-galactosidase ron loss in Dbx1 mutants. Sim1" V3 interneurons, (!-gal) expression persists in Dbx1-derived interneurons BarH1" dI1 interneurons, and Brn3a"/Lmx1b" dI5 inter- throughout embryogenesis. In E18.5/P0 Dbx1lacZ/" em- neurons, which are all commissural interneuron cell bryos, !-gal" cells expressing the neuronal marker types, were generated in normal numbers and occupied NeuN were restricted largely to lamina VIII of the spinal appropriate settling positions in the spinal cord of Dbx1 cord (Figure 1A, VIII), the location of commissural mutants (Supplemental Figure S2 at http://www.neuron. V0 Interneurons Coordinate Left-Right Alternation 377 Figure 1. Loss of V0 Interneurons in Dbx1 Mutant Mice (A) Lumbar P0 Dbx1lacZ/" spinal cord (sc) stained with antibodies to !-galactosidase (!-gal) and the neuronal marker NeuN. Most of the Dbx1- derived cells in lamina VIII (VIII) express NeuN, whereas those in the intermediate sc (Int.) have smaller nuclei and are negative for NeuN. N.B.: the ratio of !-gal" neurons to total NeuN" ventral cells is unchanged from E13 (14%) to E18.5 (15%), indicating !-gal is an excellent lineage marker for these neurons. (B and C) !-gal expression in the lumbar sc of an E18.5 Dbx1lacZ/" (B) and a Dbx1lacZ/lacZ (C) embryo. Note the decrease in lamina VIII !-gal" cells and the concomitant increase in !-gal cells in the intermediate region of the mutant sc in (C). (D) Comparison of Dbx1-derived (!-gal) cell numbers in the E18.5 lumbar sc of Dbx1lacZ/" heterozygote (solid) and Dbx1lacZ/lacZ mutant (hatched) embryos (left). There is a marked reduction in the number of Dbx1-derived cells in lamina VIII (right). (E–L) Respecification of V0D interneurons in the Dbx1 mutant sc. (E and F) Cross-sections through an E11.5 Dbx1lacZ/" sc showing V0 (!-gal", green) and dI6 (Lbx1", red) neurons form distinct populations that migrate ventrally toward the floor plate. (F) Double labeled cells (arrowhead) are only rarely observed in the spinal cord of Dbx1lacZ/" heterozy- gotes. (G–J) Cross-sections through an E11.5 Dbx1lacZ/lacZ sc showing V0D neurons that arise in the dorsal p0 domain express Lbx1. Ventrally migrating Dbx1-derived neurons express Lbx1 (H and I) and Pax2 (J). (K) !-gal, Lbx1, and Evx1 cell numbers in lamina VIII. (L) Summary of changes in V0 cell fate in the Dbx1 and Evx1 mutant spinal cord. org/cgi/content/full/42/3/375/DC1). Moreover, TAG1" axons contralaterally in the absence of Dbx1 function. axons were detected in the ventral commissure of wild- Together with previous studies (Pierani et al., 2001), type and Dbx1lacZ/lacZ embryos (Supplemental Figure S2), these data provide evidence for the selective loss of indicating that dorsal commissural neurons still project V0 interneurons in the Dbx1 mutant spinal cord. These Neuron 378 findings also indicate that prospective V0V and V0D interneuron progenitors generate V1 and dI6 interneu- rons, respectively. V0 Commissural Interneurons Form Inhibitory Connections with Motor Neurons Both sets of V0 interneurons extend axons contralater- ally (Pierani et al., 2001; Moran-Rivard et al., 2001); how- ever, their postsynaptic targets have not been defined. To assess whether V0 interneurons contact motor neu- rons, we examined whether introduction of a transynap- tic tracer into motor neurons resulted in the rapid label- ing of V0 interneurons. A GFP-expressing Bartha strain of Pseudorabies Virus (PRV152), which is known to be transported transynaptically (Smith et al., 2000; Kerman et al., 2003), was injected unilaterally into various hind- limb muscles of P1 Dbx1lacZ/" mice. Following injections into the gastrocnemius muscle, ipsilateral motor neu- rons in spinal segments L3 and L4, the location of the gastrocnemius motor pool, were selectively labeled with GFP at 24 hr (Figure 2B). At 36 hr, GFP-labeled, !-gal" V0 interneurons were present in the contralateral, but not ipsilateral, ventral lumbar spinal cord (Figure 2E, data not shown), consistent with their identity as com- missural interneurons. By 48 hr, increased numbers of labeled V0 interneurons were detected contralaterally, along with other sets of !-gal$ ventral commissural interneurons (Figures 2C and 2F). A number of !-gal$ ipsilateral interneurons were also labeled with GFP, demonstrating widespread transynaptic labeling of pre- motor neuron cell types by PRV152 (data not shown). The short latency (36 hr) between virally transduced GFP expression in motor neurons and the appearance Figure 2. V0 Interneurons Synapse with Motor Neurons of GFP in V0 interneurons (Figure 2E) argues that many (A) Summary of the PRV labeling protocol. V0 interneurons form direct commissural connections (B) PRV152-labeled motor neurons (MN, green) 24 hr after injecting with contralateral motor neurons. Injection of PRV152 the gastrocnemius. virus into either the extensor-related biceps femoris (Fig- (C and D) PRV-labeled V0 interneurons (yellow, arrowheads) 48 hr ure 2D) or the flexor-related semitendinosus muscles after injecting the gastrocnemius (GS) and biceps femoris (BF) muscles. (data not shown) also resulted in the retrograde labeling (E and F) Schematic showing lamina VIII commissural neurons in of contralateral !-gal" V0 interneurons, indicating that L3 that are (yellow % V0, green % non-V0) transynaptically labeled V0 interneurons innervate multiple motor neuron pools from GS motor neurons at 36 and 48 hr. in the mouse hindlimb. We observed that !70% of the contralateral lamina VIII neurons labeled with GFP after muscle injection of PRV152 lacked !-gal expression (see lamina VIII expressed VIAAT (Figure 3A). However, the Figures 2D, arrow, 2E, and 2F), demonstrating that V0 presence of !-gal" V0 neurons in lamina VIII that lacked neurons are not the only lamina VIII neurons that form VIAAT expression raised the possibility that not all V0 commissural connections with motor neurons. neurons express inhibitory transmitters. In support of A role for inhibitory commissural interneurons in left- this, !30% of !-gal" neurons in lamina VIII expressed right alternation during swimming and walking move- VGlut2 (Figure 3B), a marker of glutamatergic interneu- ments has been noted in a variety of vertebrate species rons (Bai et al., 2001; Takamori et al., 2001). These find- (Arshavsky et al., 1993; Cohen and Harris-Warrick, 1984; ings nevertheless suggest that the majority of V0 Butt and Kiehn, 2003; Buchanan, 1982; Soffe and Rob- interneurons release inhibitory amino acid neurotrans- erts, 1982). However, the origin and identity of these mitters. inhibitory commissural interneurons has not been deter- mined. To assess whether V0 interneurons constitute V0 Interneurons Are Activated an inhibitory class of lamina VIII neurons, we analyzed during Fictive Locomotion the distribution of the vesicular inhibitory amino acid We next asked whether !-gal" V0 neurons within lamina transporter (VIAAT) that delineates both GABAergic and VIII are activated during locomotor-like tasks and are glycinergic neurons (Sagne et al., 1997). It is presently therefore likely to contribute to the spinal circuits con- not known whether these inhibitory V0 interneurons uti- trolling locomotion. We reasoned that if V0 interneurons lize GABA, glycine, or both as neurotransmitters. In are integral components of the spinal CPG, we might E18.5 Dbx1lacZ/" mice, !70% of !-gal" V0 neurons in observe changes in c-fos expression in these neurons V0 Interneurons Coordinate Left-Right Alternation 379 Figure 3. Locomotor-Dependent Induction of c-fos in V0 Interneurons (A) Cross-section though E18.5 Dbx1lacZ/" spi- nal cord showing VIAAT expression (purple) in many !-gal" (brown) V0 interneurons (ar- rowheads mark double labeled cells). (B) Parallel section showing VGlut2 expres- sion (purple) in a subset of V0 interneurons (brown). (C) Number of !-gal"/VIAAT" and !-gal"/ VGlut2" cells per 20 &m hemichord sections from E18.5 Dbx1lacZ/" spinal cords (n % 15 sec- tions). (D and E) c-fos expression (red) in lamina VIII of a control E18.5 Dbx1lacZ/" spinal cord be- fore (D) and after (E) inducing locomotion. (F) Number of c-fos" neurons in lamina VIII before (loco$) or after (loco") locomotion. Cell counts are for 20 &m hemichord sections, n % 5 sections. (G and H) High-level expression of c-fos (red) in V0 interneurons (green, double labeled cells are yellow, arrowhead); N.B.: the c-fos protein is located in the nuclei of these neu- rons (H, arrowhead). Asterisk in (G) indicates c-fos (high) non-V0 interneuron. Low-level c-fos expression is seen in many V0 interneu- rons (arrows). (I) Number of !-gal"/c-fos (high) neurons in lamina VIII per 20 &m hemichord section (n % 12 sections). following a locomotor-like task (Barajon et al., 1992; ure 4A; Kudo and Yamada, 1987; Smith and Feldman, Jasmin et al., 1994; Carr et al., 1995). We examined 1987). We first assessed whether V0V interneurons have c-fos expression in isolated spinal cord preparations an essential role in coordinating left-right alternation by before and after inducing fictive locomotion with NMDA examining fictive locomotor activity in the Evx1 mutant and 5-HT. In control “nonlocomoting” spinal cords (n % spinal cord. Evx1 mutant mice exhibit a 65%–70% re- 3), few, if any, V0 interneurons expressed c-fos (Figures duction in the number of V0V interneurons, and the resid- 3D and 3F). In contrast, induction of locomotor-like mo- ual V0V-like interneurons fail to project axons interseg- tor neuron activity in isolated Dbx1lacZ/" spinal cords (n % mentally on the contralateral side of the spinal cord 4) resulted in a marked increase in c-fos expression in (Moran-Rivard et al., 2001). In both E18.5/P0 wild-type V0 neurons (Figures 3E–3I). Low levels of c-fos were (Figure 4B) and Evx1$/" (data not shown) mice, applica- observed throughout lamina VIII, in scattered !-gal" V0 tion of NMDA (5 &M) and 5-HT (5–15 &M) induced rhyth- interneurons (Figures 3G and 3H, arrows), but a subset mic “walking” activity in the isolated mouse spinal cord. of V0 interneurons in the ventral part of lamina VIII close This activity was characterized by strictly alternating to the white matter typically exhibited a high level of left-right flexor- or left-right extensor-related ENG motor c-fos induction (Figures 3G and 3H, arrowheads). The activity, as recorded from the ipsilateral L2 and L5 ven- activity-induced upregulation of c-fos protein expres- tral roots, respectively (Kjaerulff and Kiehn, 1996; Smith sion in V0 interneurons supports the idea that these and Feldman, 1987). This pattern shares many similari- neurons are among those recruited during locomotor ties with walking movements in adult rodents (Gruner behaviors. et al., 1980). E18.5/P0 Evx1 mutant spinal cords (n % 8) also exhib- ited an invariant alternating pattern of motor activity Left-Right Alternation in Motor Output between the left L2 (lL2) and right L2 (rL2) flexor-related Is Unaffected in Evx1 Mutants nerves (Figure 4C). Analysis of 940 episodes of L2 motor Since commissural connections in the ventral spinal bursting did not reveal any episodes of cobursting be- cord are necessary for left-right alternation (Kjaerulff and tween the left and right L2 ventral roots (see Table 1). Nor Kiehn, 1996; Cowley and Schmidt, 1997; Kato, 1990), we did we observe cobursting of left and right L5 extensor- examined the impact of the loss of V0 interneurons on related motor neurons (data not shown). Evx1 mutant locomotor behavior in an in vitro spinal cord preparation spinal cords also showed appropriate alternation of that exhibits NMDA-induced locomotor-like activity (Fig- flexor/extensor-related motor activity at hindlimb levels Neuron 380 Figure 4. Dbx1 Mutant Mice Display Syn- chronous Activity between Right and Left L2 Ventral Roots during Locomotor-like Activity (A) Schematic of recording setup with suction electrodes recording ENG activity from left and right L2 (lL2, rL2) and left and right L5 (lL5, rL5) ventral roots. Application of NMDA (5 &M) and 5-HT (10 &M) results in a locomo- tor-like pattern of activity. (B–E) ENG recordings from lL2, rL2, and rL5 ventral roots (left). Appropriate left-right al- ternation is observed in wild-type (B) and Evx1tm/tm (C) spinal cords. Dbx1lacZ/lacZ mutant (D) and hemisected (E, T8-caudal extent of spinal cord) spinal cords exhibit periods of synchronous activity in the lL2 and rL2 ventral roots (middle). In all instances, there is appro- priate alternation between lL5 and both L2 ventral roots (right). All circular plots show coupling of rL2 (cL2) and lL5 (iL5) bursts with respect to lL2 bursts over a 5 min period of activity. Points located near 0.5 represent al- ternation, while those located close to 1 rep- resent synchronicity. r values are shown for each polar plot. (Figure 4C, cf. lL2 and lL5). Furthermore, adult Evx1 Dbx1 Mutant Mice Exhibit Marked Changes mutant animals did not exhibit gait changes in locomotor in Left-Right Motor Coordination behavioral tests (G.M.L. and S.G., unpublished data; We next asked whether the elimination of V0D as well Moran-Rivard et al., 2001). Thus, neither the selective as V0V interneurons, a situation achieved in Dbx1 mutant depletion of V0V interneurons nor the concommitant mice, results in defects in locomotor behavior. Once overproduction of V1 neurons is sufficient to impair mo- again we used the in vitro spinal cord preparation, since tor coordination. Dbx1 mutant mice die at birth (Pierani et al., 2001). Strik- Table 1. Averaged Values for Step Cycle, Burst Period, and Cobursting in the In Vitro Spinal Cord Locomotor Preparation Step Period (s) Burst Period (s) In Phase Bursting (%) r Value (' SD) Wild-type (n % 8) 3.41 '.31 1.87 '.47 0 0.88 '.03 Evx1tm/tm (n % 8) 3.28 '.97 1.69 '.24 0 0.88 '.05 Dbx1lacZ/" (n % 7) 3.65 '.46 1.66 '.66 0 0.85 '.05 Dbx1lacZ/lacZ (n % 12) 3.78 '.58 1.84 '.69 22 0.52 '.15 Dbx1lacZ/lacZ " sarcosine (n % 9) 3.91 '.77 1.90 '.88 10 0.71 '.10 Dbx1lacZ/lacZ " nip. acid (n % 5) 5.19 '.47 1.94 '.18 12 0.56 '.07 V0 Interneurons Coordinate Left-Right Alternation 381 of (.25 or #.75 with respect to lL2, indicative of synchro- nous activation of the lL2 and rL2 ventral roots (Figure 4D, middle panel). The loss of left-right alternation ob- served in Dbx1 mutants resembles in part that reported after partial hemisection of the lower thoracic/lumbar spinal cord (Figure 4E; Cowley and Schmidt, 1997; Soffe et al., 1984). Despite the loss of left-right motor coordination, alter- nating L2 flexor-related and L5 extensor-related motor activity was maintained in the Dbx1 mutant spinal cord (Figure 4D, right panel, r % 0.89). Thus, V0 interneurons are not necessary for flexor-extensor alternation. No other overt differences in rhythmic motor activity were evident in the Dbx1 mutant spinal cord. The step cycle periods in wild-type and Dbx1 mutant spinal cord did not differ significantly (Table 1; wt, 3.41 s versus Dbx1$/$, 3.82 s), nor did we observe any significant lengthening in the burst phase of the step cycle (Table 1; wt, 1.87 s versus Dbx1$/$, 1.86 s) that might account for the overlap in left-right L2 motor activity. In contrast, complete hemi- section of the spinal cord resulted in a substantial lengthening of the step cycle (12.81 s, hemisected ver- sus 3.41 s, wt) (Figure 4E; Kjaerulff and Kiehn, 1996). The Dbx1-dependent elimination of V0 interneurons therefore produces a more selective disruption in loco- motor coordination than does ventral commissure tran- section. The impaired motor behavior of Dbx1 mutant mice was also accompanied by frequent episodes of en- hanced ventral root activity during the inhibition phase Figure 5. Synchronous Bursting of Contralateral Ventral Roots Often Has a Gradual Onset of the step cycle (Figure 5). These “minibursts” typically (A) Top, ENG traces showing motor neuron activity in the left and accompanied periods in which there was a progressive right L2 ventral roots during locomotor-like activity in a Dbx1lacZ/lacZ shift in the relative phasing of left-right motor neuron mouse. Bottom, rectified integrated ventral root recordings of the activity. The rectified trace in Figure 5 illustrates a series traces shown at top. of minibursts that precede each major burst in rL2 during (B) Phase values of rL2 (with respect to lL2) for 100 consecutive the inhibition phase of the step cycle (see arrow in Figure bursts. In many instances, the onset of cobursting between contra- 5A for an example). This episodic miniburst activity dur- lateral ventral roots was gradual, until complete synchronicity oc- curred (asterisk in A). During periods of synchronous activity be- ing the inhibition phase was typically accompanied by tween contralateral L2 ventral roots, small minibursts (indicated by a slow change in the relative phasing of the left and an arrow in A) would often occur during the inhibition phase of right spinal cord halves (Figure 5A, boxes) and led even- the step cycle. These minibursts were absent during periods of tually to the synchronous firing of left and right L2 ventral appropriate left-right alternation. Boxes indicate periods of signifi- roots (Figure 5A, asterisk). Strikingly, such minibursts in cant miniburst activity. the inhibition phase of the step cycle were not evident during periods of normal alternation, indicating a close ingly, isolated spinal cords from E18.5 Dbx1lacZ/lacZ mice correlation between supernumerary miniburst activity exhibited frequent episodes of overlap in the bursting and the loss of left-right alternation. During periods of of left and right L2 flexor-related motor neurons during stable drug-induced locomotion, Dbx1 mutants typically periods of drug-induced locomotion (Figure 4D, left underwent multiple episodes of phase shifting and syn- panel, arrow). Similar episodes of L5 ventral root co- chronous left-right motor activity (Figure 5B). bursting also occurred (data not shown). Episodes of synchronous bursting in Dbx1 mutants were inter- Enhanced Inhibitory Transmission Can Restore spersed with periods of normal alternation, suggesting Left-Right Alternation in Dbx1 Mutants a drift in the phasing of motor activity between the left Most V0 interneurons missing in Dbx1 mutants are inhib- and right halves of the spinal cord (Figure 5). Using polar itory commissural interneurons (Figure 3A), raising the plots to map the onset of the step cycle in rL2 with possibility that V0 interneurons normally function by in- respect to lL2 (Kjaerulff and Kiehn, 1996; Butt et al., hibiting contralateral motor neurons during the inhibition 2002), we typically observed phasing centered around phase of the step cycle. We reasoned, therefore, that 0.5 of the step cycle and r values close to 1 in wild- enhancing inhibitory transmission might prevent the ab- type and heterozygous Dbx1lacZ/" spinal cord (Figure 4B, normal activation of motor neurons during the inhibition middle panel), indicating normal left-right alternation. In phase of the cycle, thereby restoring the normal left- contrast, the onset of the contralateral step cycle in right alternation of motor outputs. We tested this idea Dbx1 mutant mice was highly variable (mean r value % by applying the transmitter reuptake inhibitors sarcosine 0.52), with many (22%) rL2 bursts exhibiting phase shifts (100 &M) or nipecotic acid (120 &M) to potentiate gly- Neuron 382 Figure 6. Effect of Inhibitory Neurotransmit- ter Agonists on Locomotor-like Activity in the Dbx1 Mutant Spinal Cord Left: Locomotor-like activity recorded from lL2, rL2, lL5 ventral roots following drug appli- cation. (A) Typical episode of locomotor like activity in a Dbx1 mutant cord showing occasional episodes of synchronous activity. (B) ENG activity in the same cord preparation after addition of 100 &M sarcosine. (C) Synchronous bursting is reduced upon addition of 120 &M nipecotic acid. Middle and right: Polar plots (5 min duration) indicating the phasing of bursts in rL2 and lL5 compared to lL2. r values are shown for each polar plot. cinergic and GABAergic transmission, respectively. In networks. Since transcription factors serve as determi- the Dbx1 mutant spinal cord, application of either sar- nants of interneuronal identity throughout the CNS, this cosine or nipecotic acid significantly lowered the inci- strategy may permit a more effective analysis of the dence of left-right cobursting during NMDA- and 5-HT- local circuits that control diverse mammalian behaviors. elicited locomotion (Figure 6, Table 1) and at the same time reduced the number of minibursts (data not shown). Assignment of Locomotor Defects to the Loss These findings are consistent with the idea that the loss of V0 Interneurons of left-right motor coordination observed in Dbx1 mutant In mice lacking Dbx1, many of the neural progenitor cells spinal cord results from an impairment in inhibitory in- that normally give rise to V0 neurons instead generate puts between commissural neurons and motor neurons. V1 and dI6 interneurons. The respecification of ventral interneurons evident in Dbx1 mutants therefore raises Discussion the issue of whether the aberrant Dbx1 mutant locomo- tor phenotype is caused by the loss of V0 interneurons The mammalian central nervous system exhibits a mod- or by the generation of supernumerary dI6 and V1 ular organization, in which localized networks of inter- interneurons. Several lines of evidence support the view neurons play critical roles in controlling neural function. that the loss of V0 interneurons is the primary cause of Dissecting the contribution of individual classes of in- the locomotor phenotype evident in Dbx1 mutants. terneurons to defined behaviors has, however, proved While some prospective V0V interneurons give rise to problematic. Using mouse genetics, we have identified En1" V1 interneurons in the absence of Dbx1 function, a discrete population of spinal CPG interneurons neces- the net increase in V1 interneuron number in older sary for walking movements. Mice lacking the homeodo- animals is small ((10% at E18.5; Supplemental Figure main transcription factor Dbx1 show a selective de- S1 at http://www.neuron.org/cgi/content/full/42/3/375/ pletion of V0 commissural interneurons and exhibit DC1). More tellingly, a similar switch from V0V to V1 a specific locomotor phenotype characterized by in- interneuronal fate occurs in Evx1 mutant mice, in the creased cobursting of left and right flexor or extensor absence of any overt locomotor defects (Figure 4; motor neuron populations. These findings establish a Moran-Rivard et al., 2001). In addition, V1 interneurons role for Dbx1 in specifying the identity of interneurons are absent in Pax6 mutants (Burrill et al., 1997; Ericson critical to the organization of spinal locomotor circuits et al., 1997), yet the left-right alternation of locomotor that control stepping movements. Our findings also indi- output is unaffected (S.G., S. Butt, O. Kiehn, and M.G., cate how the genetic manipulation of transcription fac- unpublished data). Consequently, the minor change in tors can be used to assess the behavioral function of V1 interneuron generation in Dbx1 mutants is unlikely specific classes of interneurons embedded within neural to cause the degradation of left-right locomotor phasing. V0 Interneurons Coordinate Left-Right Alternation 383 In principle, the !25% enhancement in the generation mic motor activity in the Dbx1 mutant spinal cord (Table of dI6 neurons in Dbx1 mutant mice might contribute to 1), suggesting that V0 interneurons are involved selec- the observed locomotor defects. However, dI6 commis- tively in the coordination of left-right motor alternation. sural interneurons, like the missing V0 interneurons, are The view that the locomotor defects detected in Dbx1 predominantly inhibitory (G.M.L. and S.G., unpublished mutants result from the loss of inhibitory commissural data), and accordingly, #85% of the residual !-gal" inputs to contralateral motor neurons is supported by “dI6” interneurons in the Dbx1 mutant spinal cord ex- the finding that inhibitors of glycine and GABA reuptake press VIAAT, but not VGlut2 (Supplemental Figure S3 at reduce the incidence of cobursting in the Dbx1 mutant http://www.neuron.org/cgi/content/full/42/3/375/ spinal cord (Figure 6, Table 1). One possibility is that DC1). These findings are not easily reconciled with the these agents restore normal motor function by increas- fact that the Dbx1 mutant phenotype appears to reflect ing the inhibitory drive from dI6 interneurons, which are a loss of inhibitory inputs to contralateral motor neurons. still present in the Dbx1 mutant spinal cord, thereby Moreover, if the overproduction of dI6 interneurons was permitting these neurons to compensate for the loss of the underlying cause of the Dbx1 locomotor phenotype, V0 interneurons. This interpretation is consistent with the addition of drugs that enhance inhibitory transmis- surgical studies, in which a decrease in excitatory loco- sion would be expected to increase, rather than de- motor drive restores the reduction in left-right coupling crease, the penetrance of the mutant locomotor pheno- elicited by partial transection of the ventral commissure type. Indeed, the modest increase in dI6 interneurons (Kjaerulff and Kiehn, 1996). Our genetic analysis, to- observed in Dbx1 mutants is more likely to underlie the gether with lesion and pharmacological studies, there- incomplete disruption of left-right alternation than to fore suggest a model in which V0 interneurons represent represent the underlying cause of the locomotor de- an important class of commissural interneurons that fects. Thus, the locomotor defects in Dbx1 mutants are contribute to the net threshold of contralateral inhibitory most easily explained by the loss of V0 interneurons, input needed to maintain appropriate left-right motor al- rather than by the overproduction of V1 or dI6 inter- ternation. neurons. The finding that a minority of V0 interneurons release The respective contributions of individual V0 interneu- excitatory transmitters raises the additional question of ronal subpopulations to the left-right alternation of mo- how this subset of V0 interneurons regulates contralat- tor output remains to be determined. V0D interneurons eral motor activity. In the neonate rat spinal cord, some could have a more critical role than V0V interneurons in lamina VIII excitatory commissural neurons form con- the inhibition of contralateral motor neurons, thereby nections with contralaterally located inhibitory interneu- accounting for the detection of a locomotor phenotype rons that, in turn, innervate motor neurons (Butt and in Dbx1 but not Evx1 mutants. Alternatively, the two sets Kiehn, 2003). This raises the possibility that the excit- of V0 neurons might have similar and additive roles in atory subset of V0 interneurons may inhibit motor output mediating contralateral motor neuron inhibition, in which through a disynaptic pathway. case the elimination of both neuronal populations might be necessary to produce a detectable locomotor pheno- Molecular Genetic Dissection of Spinal CPG type. Nevertheless, the persistence of dI6 inhibitory Interneuronal Circuitry commissural interneurons in Dbx1 mutants does not Recent studies in neonate rats have identified distinct compensate completely for the loss of V0 neurons, rais- populations of commissural interneurons that are rhyth- ing the possibility of functional selectivity in the actions mically active during locomotion (Butt et al., 2002; Butt of molecularly distinct populations of commissural in- and Kiehn, 2003). The relationship of these physiologi- hibitory interneurons. cally defined neurons to V0 interneurons is not known. Nevertheless, our observation that the elimination of A Selective Locomotor Defect in the Absence a discrete set of inhibitory commissural interneurons of Inputs from V0 Interneurons produces a selective motor behavioral phenotype pro- Our results indicate that V0 interneurons are key com- vides strong evidence that homeodomain transcription ponents of the commissural pathways that maintain factors such as Dbx1 and Evx1 specify the identity of proper alternation between the limbs during locomotion. functional interneuronal types that contribute to spinal Two observations, however, suggest that additional, V0- locomotor circuits. independent, commissural pathways are involved in Genetic perturbations of cell surface receptor signal- controlling left-right motor alternation. First, viral tracing ing between spinal cord cells can also influence locomo- studies show that V0 interneurons represent only a mi- tor function (Kullander et al., 2003). In particular, defec- nority of the lamina VIII commissural interneurons that tive EphA4 kinase signaling in the mouse spinal cord synapse onto motor neurons. Second, surgical transec- results in the synchronization of left-right hindlimb motor tion of the ventral commissure results in the complete activity (Kullander et al., 2003; Dottori et al., 1998). The uncoupling of alternating motor activity between the left expression of EphA4 in multiple spinal neuron cell types and right halves of the spinal cord (this study; Kjaerulff (Leighton et al., 2001; Dottori et al., 1998) has, however, and Kiehn, 1996; Cowley and Schmidt, 1997). Impor- made it difficult to ascribe this locomotor phenotype tantly, transection of the ventral commissure also cau- to an identifiable interneuronal subtype. Moreover, the ses a marked slowing in locomotor rhythm (Figure 4; EphA4 mutant locomotor defect most likely reflects a Kjaerulff and Kiehn, 1996), suggesting that commissural gain-of-function phenotype, as it has been proposed neurons may play an additional role in determining the that ipsilaterally projecting excitatory interneurons form duration of the step cycle. There is no slowing of rhyth- ectopic connections with contralateral motor neurons Neuron 384 in the absence of EphA4 function (Kullander et al., 2003). pettes. Animals were sacrificed 24–48 hr after injecting them. Spinal Thus, it remains unclear whether neurons misrouted in cords were dissected out in ice-cold PBS before fixation in 4% paraformaldehyde-PBS at 4)C. Sections from the lumbar spinal cord EphA4 mutant mice normally participate in the neural were stained with antibodies to GFP and !-gal to visualize trans- circuits that control left-right alternation. synaptically labeled V0 interneurons. The emergence of defined locomotor deficits in Dbx1 and EphA4 mutant mice nevertheless demonstrates that the genetic manipulation of spinal interneuron fate and Electrophysiology Electrophysiological experiments were performed on embryonic connectivity can provide useful insights into spinal CPG (E18.5) or early postnatal (P0) mice. Animals were anesthetized with circuitry and its link to locomotor behavior. The remark- halothane, decapitated, and eviscerated. Spinal cords were dis- able selectivity of transcription factor expression by sected out in ice-cold Ringers solution (Kjaerulff and Kiehn, 1996) subsets of spinal cord interneurons (Jessell, 2000; Lee and pinned, ventral side up, in a recording chamber constantly and Pfaff, 2001; Goulding et al., 2002) offers a particu- perfused with oxygenated Ringer’s solution composed of 111 mM larly powerful and systematic way of dissecting local NaCl, 3.08 mM KCl, 11 mM glucose, 25 mM NaHCO3, 1.18 mM KH2PO4, 1.25 mM MgSO4, 2.52 mM CaCl2. All recordings were made interneuronal function (Kiehn and Kullander, 2004), ei- at room temperature (20)C). ther through the strategy of cell fate switching illustrated in this study or through the neuronal subtype-restricted expression of toxins and membrane proteins that kill ENG Recordings neurons or silence their activity (Lee et al., 2000; Baines The second and fifth lumbar ventral roots on the right and left (i.e., et al., 2001). Our findings also highlight the potential for rL2, lL2, rL5, lL5) were placed in suction electrodes. Electroneuro- gram (ENG) recordings were amplified, bandpass filtered (100 Hz–1 using transcription factors as genetic tools to function- kHz), digitized, collected, and stored on a PC using the Axoscope ally analyze the neuronal circuits controling other impor- software (Axon Instruments). Rhythmic locomotor activity was in- tant mammalian behaviors such as respiration or visual duced in the ventral roots by adding N-methyl-D-aspartic acid perception. Finally, the identification of interneurons (NMDA, 5 &M) and 5-hydroxytryptamine (5-HT, 5–15 &M) to the with critical roles in locomotion should aid in the tar- perfusing Ringer’s solution. The effects of sarcosine (100–150 &M) geted design of more effective therapies for the recovery and nipecotic acid (100–150 &M) were investigated by adding these drugs to the NMDA/5-HT Ringer’s solution. of coordinated motor function after spinal cord injury. Experimental Procedures Analysis of Locomotor Activity Circular statistics (Zar, 1974) were used to determine the coupling Animals strength between opposing L2 and L5 ventral roots. Left L2 (lL2) The generation of Dbx1lacZ and Evx1taumyc mutant mice has been de- bursts occurring over a continuous 5 min interval (1/3 of the total scribed previously (Pierani et al., 2001; Moran-Rivard et al., 2001). recording time) were selected, and their phase values were calcu- Embryos were obtained from timed matings with the morning of the lated in reference to either the onsets of each rL2 or lL5 burst. Phase vaginal plug designated as E0.5. Genotyping of mice was performed values were determined by dividing the latency between the onset by PCR as previously described (Pierani et al., 2001; Moran-Rivard of the first lL2 burst and the following burst in rL2 (or lL5) by the et al., 2001). step cycle period (time between the reference lL2 burst and the next lL2 burst). Locomotor steps in which the lL2 and rL2 roots were In Situ Hybridization and Immunohistochemistry completely out of phase (i.e., appropriate left-right alternation) had In situ hybridization was performed as described previously (Gross phase values of !0.5. Those completely in phase (cobursting) had et al., 2002). The following in situ probes were used: mouse VIAAT, phase values of !1. The r values are a measure of the concentration VGlut2, Sim1, and BarhL1 (see Supplemental Figure S2 at http:// of phase values around the mean value for alternation (0.5). An r www.neuron.org/cgi/content/full/42/3/375/DC1 for references). For value of 1 indicates all the phase values are 0.5, whereas an r antibody in situ double staining, the proteinase K treatment step value of 0 indicates the phase values are distributed randomly. In was omitted and sections were incubated with an antibody to !-gal experiments investigating the effects of sarcosine or nipecotic acid (Cappel) after developing the DIG in situ reaction. on step cycle phasing, no measurements were made until at least Immunostaining on frozen spinal sections was performed as pre- 10 min after drug application to ensure sufficient time for the drug viously described (Gross et al., 2002; Moran-Rivard et al., 2001). 20 to wash into the preparation. Addition of either drug to wild-type &m serial sections were cut and incubated with primary antibodies. spinal cord preparations had no affect on left-right or extensor- Primary antibodies were detected using species-specific secondary flexor alternation (data not shown). antibodies conjugated with Cy3 or Cy2 (Jackson Laboratories). Im- Measurements of step cycle period (defined as the interval be- ages were captured using a Zeiss LSM510 confocal microscope tween onset of burst n and burst n " 1) and burst duration (defined and assembled using Photoshop. as time between onset of burst n and offset of burst n) were deter- mined by analysis of lL2 or rL2 activity using the DATAPAC software c-fos Analysis (Run Technologies). Averages of step cycle period and burst dura- Spinal cords isolated from P0 Dbx1lacZ/" heterozygous embryos were tion were determined from all locomotor bursts that occurred once placed in a recording chamber and perfused with Ringers (see be- a stable pattern of locomotor-like activity had been established. low). Locomotion (monitored with extracellular suction electrodes) was induced by the addition of 5 &m NMDA and 5–15 &m serotonin. After 1 hr of stable activity, spinal cords were perfused with ice- Neuronal Classification cold 4% paraformaldehyde-PBS. Cryostat sections taken from the V0 neurons are defined as the entire complement of neurons gener- lumbar cord were stained with antibodies to c-fos (Oncogene) and ated from the Dbx1 progenitor domain. V0V neurons derive from the !-gal (Gross et al., 2002). In control experiments, NMDA and 5-HT ventral (Pax7$) half of the Dbx1 progenitor domain and correspond were omitted. to the Evx1/2" set of V0 neurons described previously in Pierani et al. (2001). V0D neurons derive from the dorsal (Pax7") half of the PRV Tracing Dbx1 progenitor domain and correspond to D5 interneurons (Pierani Injections of PRV into hindlimb muscles were performed according et al., 2001). In the absence of additional markers for V0 interneu- to Kerman et al. (2003). Typically 1–2 &l of Bartha PRV152 viral stock rons, we cannot exclude that V0 neurons are divisible into further (2 * 108 infectious units per &l) was pressure injected sterotaxically functionally distinct classes of commissural interneurons. Each of into a single identified hindlimb muscle at P1 using glass micropi- these classes is likely to be disrupted in Dbx1 mutants. V0 Interneurons Coordinate Left-Right Alternation 385 Acknowledgments patrick, T., Bartlett, P.F., Murphy, M., Kontgen, F., and Boyd, A.W. (1998). EphA4 (Sek1) receptor tyrosine kinase is required for the We especially thank Drs. Ole Kiehn and Simon Butt for introducing development of the corticospinal tract. Proc. Natl. Acad. Sci. USA us to the intact spinal cord preparation and their invaluable advice 95, 13248–13253. on electrophysiology and circular statistics. We thank Kris Cowley Ericson, J., Rashbass, P., Schedl, A., Brenner-Morton, S., Kawa- for useful advice on polar plots. We would also like to thank Lynn kami, A., van Heyningen, V., Jessell, T.M., and Briscoe, J. (1997). Enquist for PRV stocks and advice on their use. We thank Chen- Pax6 controls progenitor cell identity and neuronal fate in response Min Fan for the Sim1 probe, Qiufu Ma for the mouse VIAAT probe, to graded Shh signaling. Cell 90, 169–180. Mirella Dottori for in situ probes to mouse VGlut2, and Eric Turner Goulding, M., Lanuza, G., Sapir, T., and Narayan, S. (2002). The for the Brn3a antibody. We thank Ole Kiehn, Simon Butt, Greg formation of sensorimotor circuits. Curr. Opin. Neurobiol. 12, Lemke, John Thomas, and Silvia Arber for critical comments on the 508–515. manuscript. This research is supported by grants from the National Institutes of Health (NIH), the Christopher Reeve Paralysis Founda- Graham-Brown, T. (1914). On the fundamental activity of the nervous tion, and the Human Frontiers Science Program (HFSP) to M.G. centres: together with an analysis of the conditioning of rythmic G.M.L. was supported by an HFSP postdoctoral fellowship. T.M.J. activity in progression, and a theory of the evolution of function in is supported by grants from the HFSP, NIH, and The Leila and Harold the nervous system. J. Physiol. 48, 18–41. Mathers Foundation and is a Howard Hughes Medical Institute In- Grillner, S., and Zangger, P. (1979). On the central generation of vestigator. A.P. is supported by the Association pour le Reserche locomotion in the low spinal cat. Exp. Brain Res. 34, 241–261. sur la Cancer. Gross, M.K., Dottori, M., and Goulding, M. (2002). Lbx1 specifies somatosensory association interneurons in the dorsal spinal cord. Received: February 26, 2004 Neuron 34, 535–549. Revised: March 29, 2004 Gruner, J.A., Altman, J., and Spivak, N. (1980). Effects of arrested Accepted: April 1, 2004 cerebellar development on locomotion in the rat. Exp. Brain Res. Published: May 12, 2004 40, 361–373. Harrison, P.J., Jankowska, E., and Zytnicki, D. (1986). Lamina VIII References interneurones interposed in crossed reflex pathways in the cat. J. Physiol. 371, 147–166. Arshavsky, Y.I., Orlovsky, G.N., Panchin, Y.V., Roberts, A., and Soffe, Jankowska, E., and Noga, B.R. (1990). Contralaterally projecting S.R. (1993). Neuronal control of swimming locomotion: analysis of lamina VIII interneurons in the middle lumbar segments of the cat. the pteropod mollusc Clione and embryos of the amphibian Xeno- Brain Res. 535, 327–330. pus. Trends Neurosci. 16, 227–233. Jasmin, L., Gogas, K.R., Ahlgren, S.C., Levine, J.D., and Basbaum, Bai, L., Xu, H., Collins, J.F., and Ghishan, F.K. (2001). Molecular and A.I. (1994). Walking evokes a distinctive pattern of Fos-like immuno- functional analysis of a novel neuronal vesicular glutamate trans- reactivity in the caudal brainstem and spinal cord of the rat. Neuro- porter. J. Biol. Chem. 276, 36764–36769. science 58, 275–286. Baines, R.A., Uhler, J.P., Thompson, A., Sweeney, S.T., and Bate, Jessell, T.M. (2000). Neuronal specification in the spinal cord: induc- M. (2001). Altered electrical properties in Drosophila neurons devel- tive signals and transcriptional codes. Nat. Rev. Genet. 1, 20–29. oping without synaptic transmission. J. Neurosci. 21, 1523–1531. Kato, M. (1990). Chronically isolated lumbar half spinal cord gener- Barajon, I., Gossard, J.P., and Hultborn, H. (1992). Induction of fos ates locomotor activities in the ipsilateral hindlimb of the cat. Neu- expression by activity in the spinal rhythm generator for scratching. rosci. Res. 9, 22–34. Brain Res. 588, 168–172. Kerman, I.A., Enquist, L.W., Watson, S.J., and Yates, B.J. (2003). Birinyi, A., Viszokay, K., Weber, I., Kiehn, O., and Antal, M. (2003). Brainstem substrates of sympatho-motor circuitry identified using Synaptic targets of commissural interneurons in the lumbar spinal trans-synaptic tracing with pseudorabies virus recombinants. J. cord of neonatal rats. J. Comp. Neurol. 461, 429–440. Neurosci. 23, 4657–4666. Briscoe, J., Sussel, L., Serup, P., Hartigan-O’Connor, D., Jessell, Kjaerulff, O., and Kiehn, O. (1996). Distribution of networks generat- T.M., Rubenstein, J.L., and Ericson, J. (1999). Homeobox gene ing and coordinating locomotor activity in the neonatal rat spinal Nkx2.2 and specification of neuronal identity by graded Sonic cord in vitro: a lesion study. J. Neurosci. 16, 5777–5794. hedgehog signalling. Nature 398, 622–627. Kiehn, O., and Kullander, K. (2004). Central pattern generators de- Buchanan, J.T. (1982). Identification of neurons with contralateral, ciphered by molecular genetics. Neuron 41, 317–321. caudal axons in the lamprey spinal cord: synaptic interactions and Kudo, N., and Yamada, T. (1987). N-methyl-D,L-aspartate-induced morphology. J. Neurophysiol. 47, 961–975. locomotor activity in a spinal cord-hindlimb muscles preparation of Burrill, J.D., Moran, L., Goulding, M., and Saueressig, H. (1997). the newborn rat studied in vitro. Neurosci. Lett. 75, 43–48. PAX2 is expressed in multiple spinal cord interneurons, including a Kullander, K., Butt, S.J., Lebret, J.M., Lundfald, L., Restrepo, C.E., population of EN1" interneurons that require PAX6 for their develop- Rydstrom, A., Klein, R., and Kiehn, O. (2003). Role of EphA4 and ment. Development 124, 4493–4503. EphrinB3 in local neuronal circuits that control walking. Science Butt, S.J., and Kiehn, O. (2003). Functional identification of interneu- 299, 1889–1892. rons responsible for left-right coordination of hindlimbs in mammals. Lee, S.K., and Pfaff, S.L. (2001). Transcriptional networks regulating Neuron 38, 953–963. neuronal identity in the developing spinal cord. Nat. Neurosci. 4, Butt, S.J., Harris-Warrick, R.M., and Kiehn, O. (2002). Firing proper- 1183–1191. ties of identified interneuron populations in the mammalian hindlimb Lee, K.J., Dietrich, P., and Jessell, T.M. (2000). Genetic ablation central pattern generator. J. Neurosci. 22, 9961–9971. reveals that the roof plate is essential for dorsal interneuron specifi- Carr, P.A., Huang, A., Noga, B.R., and Jordan, L.M. (1995). Cyto- cation. Nature 403, 734–740. chemical characteristics of cat spinal neurons activated during fic- Leighton, P.A., Mitchell, K.J., Goodrich, L.V., Lu, X., Pinson, K., tive locomotion. Brain Res. Bull. 37, 213–218. Scherz, P., Skarnes, W.C., and Tessier-Lavigne, M. (2001). Defining Cohen, A.H., and Harris-Warrick, R.M. (1984). Strychnine eliminates brain wiring patterns and mechanisms through gene trapping in alternating motor output during fictive locomotion in the lamprey. mice. Nature 410, 174–179. Brain Res. 293, 164–167. Moran-Rivard, L., Kagawa, T., Saueressig, H., Gross, M.K., Burrill, Cowley, K.C., and Schmidt, B.J. (1997). Regional distribution of the J., and Goulding, M. (2001). Evx1 is a postmitotic determinant of V0 locomotor pattern-generating network in the neonatal rat spinal interneuron identity in the spinal cord. Neuron 29, 385–399. cord. J. Neurophysiol. 77, 247–259. Muller, T., Brohmann, H., Pierani, A., Heppenstall, A.P., Lewin, G.R., Dottori, M., Hartley, L., Galea, M., Paxinos, G., Polizzotto, M., Kil- Jessell, T.M., and Birchmeier, C. (2002). The homeodomain factor Neuron 386 Lbx1 distinguishes two major programs of neuronal differentiation in the dorsal spinal cord. Neuron 34, 551–562. Orlovsky, G.N., Deliagina, T.G., and Grillner, S. (1999). Neural Control of Locomotion (Oxford, UK: Oxford University Press). Pearson, K.G. (1993). Common principles of motor control in verte- brates and invertebrates. Annu. Rev. Neurosci. 16, 265–297. Pierani, A., Moran-Rivard, L., Sunshine, M.J., Littman, D.R., Goulding, M., and Jessell, T.M. (2001). Control of interneuron fate in the developing spinal cord by the progenitor homeodomain protein Dbx1. Neuron 29, 367–384. Sagne, C., El Mestikawy, S., Isambert, M.F., Hamon, M., Henry, J.P., Giros, B., and Gasnier, B. (1997). Cloning of a functional vesicular GABA and glycine transporter by screening of genome databases. FEBS Lett. 417, 177–183. Sapir, T., Geiman, E., Wang, Z., Velasquez, T., Yoshihara, Y., Frank, E., Alvarez, F., and Goulding, M. (2004). Pax6 and En1 regulate two distinct aspects of Renshaw cell development. J. Neurosci. 24, 1255–1264. Smith, J.C., and Feldman, J.L. (1987). In vitro brainstem-spinal cord preparations for study of motor systems for mammalian respiration and locomotion. J. Neurosci. Methods 21, 321–333. Smith, B.N., Banfield, B.W., Smeraski, C.A., Wilcox, C.L., Dudek, F.E., Enquist, L.W., and Pickard, G.E. (2000). Pseudorabies virus expressing enhanced green fluorescent protein: a tool for in vitro electrophysiological analysis of transsynaptically labeled neurons in identified central nervous system circuits. Proc. Natl. Acad. Sci. USA 97, 9264–9269. Soffe, S.R., and Roberts, A. (1982). Tonic and phasic synaptic input to spinal cord motoneurons during fictive locomotion in frog em- bryos. J. Neurophysiol. 48, 1279–1288. Soffe, S.R., Clarke, J.D.W., and Roberts, A. (1984). Activity of com- missural inteneurons in the spinal cord of Xenopus embryos. J. Neurophysiol. 51, 1257–1267. Stokke, M.F., Nissen, U.V., Glover, J.C., and Kiehn, O. (2002). Projec- tion patterns of commissural interneurons in the lumbar spinal cord of the neonatal rat. J. Comp. Neurol. 446, 349–359. Takamori, S., Rhee, J.S., Rosenmund, C., and Jahn, R. (2001). Identi- fication of differentiation–associated brain-specific phosphate transporter as a second vesicular glutamate transporter (VGlut2). J. Neurosci. 21, RC182. Wenner, P., O’Donovan, M.J., and Matise, M.P. (2000). Topographi- cal and physiological characterization of interneurons that express engrailed-1 in the embryonic chick spinal cord. J. Neurophysiol. 84, 2651–2657. Zar, J.H. (1974). Biostatistical Analysis (Engelwood Cliffs, NJ: Pren- tice Hall). Zhou, Q., Choi, G., and Anderson, D.J. (2001). The bHLH transcrip- tion factor Olig2 promotes oligodendrocyte differentiation in collab- oration with Nkx2.2. Neuron 31, 793–809.