EphA4 defines a class of excitatory locomotor-related interneurons

Results

Locomotor Pattern in ephA4^lacZ Mice. Previously, we have shown that the abnormal adult hopping gait in ephA4-KO mice can be elicited in the isolated neonatal spinal cord preparation, suggesting that intrinsic defects in the spinal cord locomotor CPG underlie this phenotype (16). In the isolated spinal cord, the abnormal hopping gait was seen as synchrony between bilaterally located, lumbar segment 2 (L2) predominantly flexor motor neurons as well as synchrony between caudally located, predominantly extensor L5 motor neurons. Normal alternating coordination was, however, maintained between ipsilaterally located L2 and L5 motor neurons. In the present experiments, where we used gene-trap ephA4^lacZ mice instead of the ephA4 ^neo knockouts, we first asked whether the isolated spinal cord from newborn animals in this genetic background also produced a hopping pattern of locomotor activity. This analysis was crucial to determine the putative function of EphA4-positive neurons.

Rhythmic locomotor-like activity (Fig. 1B) was elicited by perfusing the cord with a combination of NMDA (3–7 μM) and 5-HT (3–9 μM) in a total of 56 mice: 20 heterozygote ephA4^{lacZ /+} mice, 30 homozygote ephA4^{lacZ /-} mice, and 6 C57/BL6 WT mice used for reference. Averaging of the locomotor cycle relative to the iL5 burst over a 5- to 10-min period revealed the overall pattern of coordination (Fig. 1C), and in all cases the rhythmic iL2 DC oscillations were out-of-phase with the iL5. We used circular statistics (see refs. 16 and 20) to provide a quantitative measure of the strength of coupling between the cL2 and iL5 ventral roots. In all WT mice and the majority of heterozygote mice (17 of 20) these roots were synchronous (circular statistic vector points close to 0.0), corresponding to the coactivation of flexors and extensors on the left and right side of the spinal cord during normal locomotion (swing muscles and stance muscles on either side of the body). The remaining three heterozygotes had no significant vector between cL2 and iL5 coordination because of drift (or uncoupling) between the two sides of the cord. The majority (18 of 30) of homozygotes displayed a rabbit-like hopping gait (vector point close to 0.5; Fig. 1D) with alternation between cL2 and iL5 similar to what has been observed in straight ephA4-KO mice. Of the remaining 12 homozygotes, 4 exhibited significant synchronization between cL2 and iL5, indicative of a normal locomotor pattern not previously detected in straight ephA4-KO, whereas the remaining 8 showed drift.

Thus, it appears that the ephA4^{lacZ /-} homozygote mice exhibited a greater degree of variability in the pattern of bilateral coordination than the ephA4-KO mice (16). Because we were interested in the synaptic effect of EphA4-positive neurons on ipsilateral motor neurons, the only significance of this variability is the occurrence of drift between bilateral ventral roots in some preparations. To counter this, we always calculated the phase of intracellular spike activity recorded in the iL2 segment relative to the iL5 ventral root (see Fig. 1C), because the iL2 ventral root activity always showed strict alternation with that of the iL5 in all preparations (WT, heterozygote, and homozygote; also see ref. 16).

Locomotor-Related Activity in Ventral EphA4-Positive Neurons. We first determined the activity of EphA4-postive cells during locomotion by recording intracellularly from randomly sampled neurons in the area where the hindlimb locomotor CPG is located (23). Although the hindlimb locomotor CPGs are distributed over lumbar segments, we concentrated our recordings in the L2 segment, which is known to have a high locomotor capability compared with more caudal segments (18, 24). Using “blind patching” (see Materials and Methods) in the ipsilateral L2 segment, we recorded a total of 76 neurons from the most ventral half of the spinal cord in WT, heterozygotic (ephA4^{lacZ /+}), and homozygotic (ephA4^{lacZ /-}) mice. At rest, the average membrane potential of cells was -49 ± 6 mV (SD) (n = 76; not corrected for liquid junction potential), and input resistance (IR) was 427 ± 206 MΩ (SD) (n = 40); there was no statistical difference (t test; P > 0.05) between neurons grouped according to their different genotypes (data not shown). In the absence of ventral root bursts, spontaneous action potentials were observed in 51% of the recorded cells. Forty-five Lucifer yellow-filled neurons recorded from hetero- and homozygotic mice were successfully recovered after the experiments. Nineteen of these neurons were immunopositive for β-gal [n = 6 (ephA4^{lacZ /+}) and 13 (ephA4^{lacZ /-}); Fig. 2A] and were distributed throughout the area of the ventral horn sampled (Fig. 2B). The activity of the neurons during locomotion was determined during stable NMDA/5-HT-induced locomotor-like activity. Postrecording analysis of the cells according to whether or not they expressed β-gal revealed a degree of heterogeneity with regard to both their rhythmicity and phase of firing both in the EphA4-positive and EphA4-negative populations. Three of the 19 positive neurons were most likely motor neurons (Fig. 3A) given that they exhibited pronounced membrane oscillations, fired spikes exclusively in one phase of the locomotor cycle, and had significantly lower IRs (105 ± 22; n = 3) and hyperpolarized membrane potentials [MP < -50 to 55 mV at rest (also see ref. 25)], as compared with other positive cells (IR = 329 ± 138, n = 16; P = 0.005; MP > -50 mV). It is not surprising to find EphA4-positive motor neurons because EphA4 has previously been shown to be expressed in motor neurons and to have a prominent role in motor axon guidance (26). Twelve of the EphA4-positive cells were either defined as highly significantly rhythmic (hS) or significantly rhythmic (S) on the basis of circular statistics performed on 25 spikes (see Materials and Methods and ref. 20; Fig. 3A Middle and Right). Furthermore, no positive neurons were recorded that had r values, a measure of the degree of rhythmicity (see Materials and Methods), within the lowest two bins (r < 0.2), unlike with the EphA4-negative population (Fig. 3B). The average rhythmicity, as defined by the r value of positive neurons, did not differ significantly between heterozygotes (0.496 ± 0.203) and homozygotes (0.507 ± 0.184).

Plotting the preferred phase of firing for positive and negative neurons revealed a broad distribution, with neurons firing both in the ipsilateral L2 phase (gray shaded areas in Figs. 3 C and D) and ipsilateral L5 phase (unshaded hemispheres). However, the EphA4-positive interneurons tended to fire in the ipsilateral L2 phase [12 of 19 (9 of 13 cells in homozygotes, and 3 of 6 cells in heterozygotes); Figs. 3 C and D].

These experiments demonstrate that the majority of neurons expressing the EphA4 receptors and located in the ventral region of the spinal cord are rhythmic during drug-induced locomotor-like activity and therefore could be constituents of the hindlimb CPG.

Local Synaptic Effects of EphA4-Positive Neurons. To test whether any of the EphA4 interneurons projected onto motor neurons (MNs), we used spike-triggered averaging of the DC L2 ventral root recording (also see ref. 27). We have previously used this technique in the neonatal rat preparation to investigate the function of commissural neurons (CINs) in coordinating left–right activity (20). We decided to test the effect of the neurons on the local, ipsilateral MNs given that previous data suggests that these neurons are normally ipsilateral projecting and only in ephA4-null mice have aberrant crossed axon collaterals (4). Analysis of the effect on the local MNs would also allow us to maximize our data sample from the “blind” patching technique because the data could be analyzed irrespective of genotype. Here, we tested the synaptic effect of 36 neurons onto local, ipsilateral motor neurons both in the absence and the presence of locomotor-like activity. A fair number of the neurons exhibited spontaneous firing in the absence of ventral root activity (see above), and these spikes were used for the spike-triggered averaging of the DC iL2 ventral root recording (see Materials and Methods). In silent cells, firing was elicited by current injection (5–30 pA) to obtain spike-triggered averaging in the nonlocomotor state.

Three of the cells, which were classified as motor neurons on the basis of the criteria listed previously, exhibited large orthodromic spikes in the DC recordings (data not shown). Three types of postsynaptic response in the DC recordings were observed as a result of the interneuron spike-triggered averaging: excitatory (Exc, Fig. 4) and inhibitory (Inhib) responses that typically ranged from 4 to 20 μV in amplitude (compare with ref. 20), and no signal (NS).

Excitatory responses were elicited by 10 neurons, 6 of which were EphA4-positive. These responses occurred at latencies ranging from 3.1 to 11.2 ms (average of 7.6 ± 2.9 ms) and were sensitive to CNQX (10 μM) and AP-5 (20 μM) (Fig. 4A Top). Block of the excitatory response by CNQX and AP-5 resulted in the emergence of a longer latency inhibitory response (average latency of 45.6 ± 15.1 ms), which was insensitive to the glycine receptor blocker strychnine (n = 1; 0.3 μM) and/or the GABA_A receptor blocker bicuculline (n = 1; 2 μM) but was attenuated by incubation in low-Ca²⁺ Ringer's (n = 2). The nature of this potential is uncertain, but it might be mediated by activation of metabotropic glutamate receptors, which in turn activate a calcium-activated potassium conductance similar to what has been reported in ventral midbrain dopamine neurons (28). Twelve interneurons, of which only one was EphA4-positive, evoked inhibitory responses in the averaged DC trace at a wide range of latencies (5.2–55.8 ms; average of 26.5 ± 14.4 ms) both at rest and during locomotor-like activity. Longer latency responses were sensitive to bicuculline (n = 2; Fig. 4A, Bic) or CNQX/AP-5 (n = 2; Fig. 4A) except for one that, although sensitive to low-Ca²⁺ Ringer's, was not blocked by bicuculline or strychnine. These connections are presumably polysynaptic (cf. ref. 20). Of the shorter latency responses (n = 3), one was tested and found to be sensitive to bicuculline but not strychnine; the other two were sensitive to low Ca²⁺ Ringer's, but their specific neurotransmitter-type was not ascertained (Fig. 4A Middle). Eleven cells evoked no signal in the ipsilateral L2 ventral root with the number of spikes used for the averaging (similar numbers of spikes used for averaging as in cells that evoked excitation or inhibition), five of which were EphA4-positive.

Fig. 4B summarizes the results of these tests for EphA4-positive and EphA4-negative neurons by dividing the cells into groups according to their various synaptic effects on the ipsilateral motor neurons. The numbers of positive and negative neurons were approximately equal in the categories of excitatory and no-signal neurons, while the majority (10 of 11) of inhibitory interneurons were negative for EphA4. Analysis of the preferred phase of firing in the three categories of interneurons (Fig. 4C) revealed that excitatory EphA4-positive interneurons fired at the peak of ipsilateral motor neuron burst (≈0.25; Fig. 4C Left), whereas the majority of inhibitory EphA4-negative neurons fire out-of-phase with these motor neurons (Fig. 4C Right). All of the EphA4-positive cells that exerted local synaptic effects (n = 7) were recorded from homozygotic mice. Of the five neurons recorded in heterozygotes, two were putative motor neurons, and the remaining three had no local DC signal.

Discussion

Here, we demonstrate that the majority of interneurons expressing the EphA4 tyrosine kinase receptor in the ventral region of the mouse lumbar spinal cord are rhythmically active during locomotion. A subset of these neurons provide excitation of motor neurons on the same side, indicating that they are involved in generating the ipsilateral rhythmic motor activity. These findings correlate well with previous findings based on extracellular ventral root recordings that suggest that EphA4-ephrinB3 signaling is crucial to the correct development of the ipsilateral CPG, and specifically axon guidance of excitatory interneurons (16).

The Role of EphA4-Positive Interneurons in the Locomotor CPG. The majority of EphA4-positive interneurons recorded fired in a rhythmic manner during drug-induced locomotor-like activity. This result, taken with the fact that the null phenotype exhibits such a pronounced abnormal gait, would suggest that some EphA4-positive interneurons are components of the locomotor CPG. The activity of EphA4-positive interneurons during drug-induced locomotor-like activity was not uniform, with neurons firing both in- and out-of-phase with the ipsilateral L2 predominantly flexor motor burst. This distribution can in part be explained by the function of these neurons as determined by the spike-triggered averaging of the ipsilateral ventral root DC trace as discussed below.

All interneurons that were EphA4-positive and excitatory fired at the peak of the ipsilateral motor activity. These neurons constituted ≈50% of the EphA4-positive cells. We suggest that these neurons provide excitatory drive to local motor neurons during locomotion. Whether these excitatory interneurons form monosynaptic or polysynaptic projections onto ipsilateral motor neuron was not specifically tested here (see ref. 20) and cannot be determined based on the measured latencies (average of 7.6 ± 2.9 ms) of the spike-triggered averaged excitatory events. Previous studies have shown that synaptic latencies are long in the still immature spinal cord (29, 30). An additional factor that might slow synaptic latencies is the fact that all recordings are performed at room temperature. All EphA4-positive excitatory interneurons were found in homozygote animals. This result could indicate (i) that in addition to the proposed changes in left–right coordination there is also a reconfiguration of the ipsilateral CPG unit in the ephA4-nulls (for example, increased ipsilateral aberrant sprouting in ephA4-nulls), or (ii) that we only had a small sample of interneurons recovered from ephA4^{lacZ /+} animals, and we did not record from any of the excitatory cells. We favor the latter interpretation because previous data (4) and the fact that the null animals display precise bilateral coordination would suggest that these cells are normally ipsilateral-projecting and coordinate local, segmental motor neurons.

EphA4-positive neurons that have no local, ipsilateral effect (≈42% of the EphA4-positive cells) usually fired off-phase of the ipsilateral L2 flexor burst. Without the synaptic effects on motor neurons, it is hard to ascribe a definitive functional role to these cells. However, we can predict that they are ipsilateral because few commissural interneurons express the EphA4 receptor (16). So, either these neurons are inhibitory and project to flexor-related motor neurons located outside L2 or they are excitatory neurons projecting to extensor-related motor neurons (firing out-of-phase with the L2) caudal of L2. Only one EphA4-positive neuron exerted a local inhibitory effect with a preferred phase of firing at the peak of the ipsilateral L2 out-of-phase period. What is apparent from our recordings is that the interneuron populations that express the EphA4 receptor are not a homogeneous population but are likely to contribute to the overall coordination of the ipsilateral locomotor-drive. Therefore, it is perhaps unsurprising that in the absence of correct axon guidance in ephA4-null mouse there is a dramatic alteration in gait.

Excitatory Central Pattern Generator Interneurons. The basis for this article was the recent finding that aberrant local connections within the locomotor CPG can account for the hopping locomotor-like pattern exhibited by ephA4-KO and ephrinB3-KO mice (16). This led us to hypothesize that a group of EphA4 neurons are ipsilateral excitatory CPG interneurons, and that these neurons additionally have aberrantly crossing fibers in EphA4-KO and EphrinB3-KO mice, although it was not shown directly (also see refs. 12 and 31). In the present study, we directly show that a subpopulation of EphA4-positive neurons are rhythmically active and excitatory ipsilaterally projecting interneurons. Due to space constraints imposed by the headstages used to record intracellularly and the DC signal from the iL2 ventral root, it was not possible in the present experiments to simultaneously record form both the ipsilateral and contralateral roots in the same segment. Therefore, we did not test electrically as to whether or not the normally ipsilateral interneurons project contralaterally in the ephA4-null, as would have been predicted by Kullander et al. (16). Moreover, because we were unable to trace individual axons for long distances from Lucifer-filled cells, we could not determine in the present study whether excitatory EphA4-positive cells also had aberrant midline crossing fibers in epha4-null mice. Notwithstanding this, the significance of the present study is that we have shown that a subpopulation of EphA4-positive cells are rhythmically active and provide excitation of, and in-phase with, the ipsilateral motor neurons. Ipsilaterally projecting excitatory CPG neurons have been described in a number of aquatic vertebrates (32, 33). However, putative excitatory CPG interneurons have been previously only identified partly in mammals. Short-range projecting L7 interneurons that are antidromically activated from the nearby extensor motor nucleus and from group I afferents have been found to be rhythmically active during locomotion in the cat (34). However, the postsynaptic effects of these neurons were not determined directly but inferred to be excitatory from their being activated by group I afferents and their firing and projection pattern. Two-thirds of the group II L3-L4 cells in the cat that project caudally as recorded by Shefchyk et al. (35) were rhythmically active. Although it was suggested that these cells were excitatory, this was not shown directly for the rhythmically active cells but inferred from previous studies by Jankowska and colleagues (36), who showed that the group II midlumbar cells form a mixed population of excitatory and inhibitory cells. It is therefore difficult at the moment to say anything definitive about the excitatory cat midlumbar group II neurons during locomotion. Finally, recent experiments have shown that mouse spinal interneurons that are positive for the transcription factor HB9 and contain the vesicular glutamate transporter are rhythmically active during drug-induced locomotion (3, 4). The axons of these neurons project into the motor nucleus, where they appear to make contact with motor neurons or nearby cells. These findings suggest that the HB9-positive interneurons are also involved in generating synaptic drive of ipsilateral motor neurons during locomotion, although there was no direct demonstration of an excitatory effect on ipsilateral motor neurons. The EphA4-positive interneurons reported here fire in the right phase of the locomotor cycle to be at least partly responsible for the excitatory drive of ipsilateral motor neurons during locomotion. In this way, they are strong candidate neurons for defining a distinct group of identified excitatory CPGs neurons in mammals.

Clearly, EphA4-positive cells are not the only excitatory cells in the CPG, because we also found a number of last-order excitatory interneurons that are EphA4-negative. A number of excitatory EphA4-negative interneurons fired in a rhythmic manner at the extensor–flexor transitions (0.0 phase values), suggesting that they might have a slightly different role in the CPG than the EphA4-postive excitatory interneurons.

Conclusion

Our data show that the majority of EphA4-positive neurons are rhythmically active and that a subset of these are excitatory ipsilaterally projecting interneurons. Ipsilaterally projecting glutamatergic excitatory interneurons are important for generating the rhythmic motor neuron drive in the tadpole and lamprey swimming CPGs (37). We suggest that the rhythmically active excitatory EphA4-positive cells are strong candidate neurons for defining a group of identified excitatory interneurons in the mammalian CPGs.