Role of EphA4 and EphrinB3 in Local Neuronal Circuits That Control Walking

See allHide authors and affiliations

Science  21 Mar 2003:
Vol. 299, Issue 5614, pp. 1889-1892
DOI: 10.1126/science.1079641


Local circuits in the spinal cord that generate locomotion are termed central pattern generators (CPGs). These provide coordinated bilateral control over the normal limb alternation that underlies walking. The molecules that organize the mammalian CPG are unknown. Isolated spinal cords from mice lacking either the EphA4 receptor or its ligand ephrinB3 have lost left-right limb alternation and instead exhibit synchrony. We identified EphA4-positive neurons as an excitatory component of the locomotor CPG. Our study shows that dramatic locomotor changes can occur as a consequence of local genetic rewiring and identifies genes required for the development of normal locomotor behavior.

Rhythmic movements such as locomotion and swimming require that muscles contract and relax in a complex repetitive pattern. Central pattern generators, or CPGs, are local spinal neuronal networks that generate and coordinate these rhythmic muscle activities (1, 2). In the fruit fly, it was recently shown that the CPG for peristaltic crawling develops in the complete absence of sensory input (3). In two nonmammalian vertebrate species, the lamprey and theXenopus tadpole, the critical neuronal components of the locomotor CPG have been identified (4, 5). In mammals, the CPGs controlling limb movements are located in the ventromedial part of the spinal cord (6). However, the neuronal organization is still poorly understood (2), and no molecules that contribute to CPG development have been identified. Because CPGs are important for spinal control of walking in humans (7), understanding their neuronal organization and molecular determination is essential in the ongoing effort to reestablish locomotor function in patients with spinal cord injury.

Interactions between ephrin ligands and their Eph receptor tyrosine kinases are crucial for organizing several neuronal systems (8, 9). The synchronous hindlimb locomotion defects observed in ephA4- andephrinB3-null mice (Fig. 1), and in mice carrying a kinase-inactive version of EphA4 (10), have been suggested to originate from a corticospinal tract (CST) axon-guidance defect, but defects in the CPG have not been discounted (11–13). To investigate the potential involvement of local spinal CPGs in the abnormal locomotive behavior, we examined locomotor activity in isolated spinal cords from newborn mice (14). In wild-type (WT) mice, the addition of serotonin and N-methyl-d-aspartate (NMDA) to spinal cord preparations produced rhythmic locomotor activity for several hours (fig. S1). The rhythmicity was characterized by alternation between left lumbar (L) segment 2 and right L2 ventral-root bursts, similar to left-right alternation between limbs (15,16). There was also alternation between L2 and L5 activity on the same side of the spinal cord, similar to the flexion and extension in one limb during locomotion (Fig. 1, A and D).

Figure 1

The lumbar segments of the spinal cords of neonatal ephA4- and ephrinB3-null mice exhibit synchronous left-right ventral root activity. (A toC) Images of WT mice displaying normal locomotor activity (A) orephA4-null mice (B) and ephrinB3-null mice (C) displaying a rabbitlike gait. (D to F) Recorded activity after application of NMDA and serotonin to the isolated spinal cord (a 4-μM solution of each drug) of WT mice (D),ephA4-null mice (E), and ephrinB3-null mice (F) in flexor (L2) and extensor ventral (L5) roots. r, right; l, left. (G to I) Circular phase diagrams derived from 20 locomotor cycles for the WT (G), ephA4-null (H), and ephrinB3-null (I) mice shown in (A) to (C), respectively, (J to L) Plots show the vector points of L2 pairings for all experiments conducted on WT mice (n = 5) (J, green squares); ephA4heterozygotes (n = 13) (K, black triangles) and homozygotes (n = 14) (K, blue circles); andephrinB3 homozygotes (n = 9) (L).

In isolated spinal cords from ephA4- andephrinB3-null mice, the normal rhythmic pattern was defective at an age (0 to 5 days) independent of CST influence (17). Bilateral ventral roots displayed an abnormal synchronous rhythm rather than the WT pattern of left-right alternation (Fig. 1, E and F). Ipsilateral L2 and L5 ventral roots showed a normal alternating pattern in mutant mice (Fig. 1, E and F), hence replicating the rabbitlike gait observed in the adults (Fig. 1, B and C). The pattern between pairs of roots can be expressed in circular phase plots for individual animals (Fig. 1, G to I), where the length of the vector is used to calculate whether there is a significant direction to the phase (6). Alternation is seen as a mean phase value near 0.5, whereas synchrony has a value near 0.0. In all WT mice (n = 5), the mean phases exhibited alternation (Fig. 1J), but in most (15 out of 23) of the ephA4- andephrinB3-null mice (Fig. 1, K and L), the mean phases were close to synchrony. The remaining null mice showed a drifting pattern with no significant direction of the vector. A WT-like alternating pattern was observed in 10 ephA4-heterozygous mice, whereas a drifting pattern was seen in three others (Fig. 1K, black triangles). The findings establish that changes in the CPG organization are sufficient to explain the observed rabbitlike gait pattern inephA4 and ephrinB3 mutants and make it unlikely that CST defects are responsible as previously hypothesized (11–13).

Normal left-right coordination is mediated by commissural interneurons (CINs) (18). A switch from left-right alternation to synchrony can be induced in rats by blocking the glycine- and/or γ-aminobutyric acid type A (GABAA)–dependent CIN inhibitory transmission (19). Therefore, the observed phenotype in the ephA4- and ephrinB3-null mice could be due to a change in the balance between excitation and inhibition in the CINs. Analysis of the distribution of CINs in L2, however, revealed similar patterns in WT and mutant mice (Fig. 2, A to D) (20); this result argues against the possibility that changes in the CIN system are responsible for the observed defects in the mutants.

Figure 2

Nerve fibers aberrantly cross the spinal cord midline in ephA4- and ephrinB3-null mice. (A) Schematic of a lumbar spinal cord illustrating the site of tracer application (red box) and the paths of labeled CIN projections (red) and unlabeled CIN projections (black). The position of cross sections shown in (B) to (D) is indicated between black arrows. (B to D) Unilateral spinal cord sections showing the distribution of CINs (white arrows) after contralateral caudal application of the tracer (14). No overt difference is detected between spinal cords from WT (n = 5) (B),ephA4-null (n = 15) (C), andephrinB3-null (n = 11) (D) mice. However, fibers extensively cross the midline (white arrowheads in enlargements) in both ephA4-null (C) and ephrinB3-null (D) mice. (E) Schematic showing the midline tracing paradigm. (F) Transverse section after midline tracer application. (G and H) β-Gal immunohistochemistry (green) combined with retrograde tracing (red). Rhodamine-dextran and anti–β-Gal signals were detected in adjacent cells in (G) but in the same cell in (H), white arrowhead. (I) Quantification of cells double-positive for EphA4 and midline tracer. There is a fivefold increase in double-labeled cells detected betweenephA4lacZ/+ mice (n = 6) (G) andephA4lacZ/– mice (n = 12) (H). (J and K) Position of double-positive cells (derived from six sections) in ephA4lacZ/+ and ephA4lacZ/– mice after midline labeling. Scale bars, 200 μm in (B) to (F), 20 μm in (G) and (H).

Numerous additional fibers were observed crossing the spinal cord midline to the contralateral side in theephA4-null mice (Fig. 2, C and D, arrows; fig. S2), possibly representing axons of neurons that no longer recognize ephrinB3 in the midline. Previous studies report that CST fibers aberrantly recross the cord midline, but crossing defects in the local spinal cord have not been described (11, 12). To determine whether these fibers originate from local EphA4-positive neurons, we usedephA4-null mice that express lacZ under the control of the ephA4 promotor (ephA4lacZ ) (21) in combination with a ventral midline tracing technique (Fig. 2, E and F) that only includes neuronal projections involved in CPG coordination (6, 14, 22). InephA4lacZ/+ control mice, very few traced neurons were positive for β-galactosidase (β-Gal), suggesting that local EphA4-positive fibers normally avoid crossing the midline (Fig. 2, G and I). In contrast, we found a fivefold increase of traced β-Gal–labeled neurons (Fig. 2H, arrow) inephA4lacZ/– mice, indicating that spinal cord neurons that normally express EphA4 cross the midline when they do not express functional EphA4 receptors (Fig. 2I). Plots showing the location of double-labeled cells in ephA4lacZ/– mice revealed a widespread pattern of cells in the ventral part of the cord (Fig. 2, J and K).

To examine the functional consequences of these aberrant projections, we used either the glycine/GABA uptake blockers sarcosine and nipecotic acid or the glycine/GABAA receptor antagonists strychnine and bicuculline to strengthen or weaken, respectively, the reciprocal inhibition (2). As is the case in rats (19), the weakening of inhibition by the addition of strychnine (0.3 μM) and/or bicuculline (10 μM) induced a switch from alternation to synchrony in heterozygous ephA4 mice (n = 3) (fig. S3). When inhibition was strengthened by the addition of sarcosine (n = 4) or nipecotic acid (n = 2), alternation was maintained, although the rhythmic cycle was prolonged (fig. S4). However, inephA4-null mutants, strengthening inhibition by the addition of sarcosine (n = 6) resulted in the reversal of synchronous locomotor activity to normal alternation (Fig. 3). Thus, normal locomotion could be restored by chemical reinforcement of the CIN inhibitory component to counteract a stronger aberrant excitatory innervation. We infer from this result that fibers that aberrantly cross the midline and connect to contralateral CPG components in ephA4- andephrinB3-null mice are predominantly excitatory in nature. Taken together, the data suggest that EphA4-positive fibers normally avoid crossing the midline and that at least some EphA4-positive cells constitute an ipsilateral excitatory element of the CPG.

Figure 3

Antagonists of inhibitory neurotransmitters convert the abnormal synchronous activity observed in ephA4-and ephrinB3-null mice to normal alternation. (A) Adding NMDA (7 μM) and serotonin (7 μM) results in synchronous rL2-lL2 activity in ephA4-null mice. (B) Alternating activity in the same mouse after 15 min of incubation with 100 μM sarcosine. (C and D) Averages of 100 normalized locomotor cycles (rectified and linear smoothed) with respect to rL2 activity (solid line) and lL2 (stippled line). (E) Plot of vector points (black circles) for null mutant locomotion (n = 6 mice), which exhibit either drift or significant synchrony. (F) The same preparations perfused with sarcosine display vectors (black circles) indicating alternation.

We performed in situ hybridization experiments (14) for vesicular glutamate transporters (VGLUTs), which are markers for glutamatergic neurons (23,24), combined with EphA4 immunostaining. VGLUT1-expressing cells were found to be restricted to a small population of neurons dorsal to the central canal in P20 lumbar spinal cord sections (Fig. 4A). In contrast, VGLUT2-positive cells were found dispersed in both ventral and dorsal parts of the spinal cord. (Fig. 4B). In early postnatal spinal cords, VGLUT2- and EphA4-positive cells are widely distributed but with somewhat different patterns (Fig. 4, C and D). By combining VGLUT2 in situ hybridization with β-Gal immunostaining in ephA4lacZ/+ mice, we identified EphA4, VGLUT2 double-positive neurons (Fig. 4, E to G) in the ventromedial area where the CPG is located (22). In this region, 51% of the VGLUT2-positive cells were also EphA4 positive. Conversely, 23% of the EphA4-positive cells were also VGLUT2 positive (Fig. 4H). These experiments confirm that EphA4-positive neurons can be excitatory in nature.

Figure 4

A subset of spinal cord EphA4-positive neurons are glutamatergic. A P20 WT spinal cord showing VGLUT1-positive [(A), black arrow] and VGLUT2-positive (B) neurons is shown. (C and D) P4 sections reveal widespread distributions of EphA4- and VGLUT2-positive neurons. The box in (D) marks the area where double-labeled neurons in (H) were counted. (E to G) Close-up of the ventromedial spinal cord showing VGLUT2 [(E), arrow] and EphA4-positive neurons as identified by β-Gal antibodies in ephA4lacZ/+ spinal cords [(F), arrow], shown superimposed in (G). (H) Counts of single- and double-labeled neurons (mean ± SEM) in the ventromedial quadrant. Ratios are expressed as the percentage of double-labeled neurons with respect to either the total number of EphA4-positive neurons (23%) or the total number of VGLUT2-positive neurons (51%). Scale bars, 200 μm in (A) to (D), 40 μm in (E) to (G).

This study identifies EphA4-positive neurons as a key component of the mammalian CPG. These CPG neurons aberrantly cross the midline in ephrinB3- and ephA4-null mutant mice to produce an abnormal synchronous coordination through reciprocal overexcitation of the CPG. We conclude that ephrinB3-induced EphA4 signaling repels some components of the mammalian spinal CPG and restricts their axonal projections to one side of the spinal cord. We postulate that in the normal CPG, ipsilateral neurons that express the EphA4 receptor are predominantly excitatory. Understanding the molecular components of the mammalian CPG could offer novel therapeutic approaches for the treatment of paraplegics whose locomotor performance can be improved by stimulating plasticity in spinal cord walking circuits (25, 26).

Supporting Online Material

Materials and Methods

Figs. S1 to S4


  • * To whom correspondence should be addressed. E-mail: klas.kullander{at} (K.K.); ole.kiehn{at}

  • These authors contributed equally to this work.


View Abstract

Navigate This Article