Experiments were performed on 326 new-born C57BL/6JRJ mice, aged postnatal day 1 (P1) to P3 and P8 to P12, without consideration of sex. All procedures were carried out in strict accordance with the guidelines of the European Committee Council Directive and approved by the local ethic committee of the University of Bordeaux and the French Agriculture and Forestry Ministry for handling animals (approval number 2016012716035720). All efforts were made to minimize animal suffering and reduce the number of animals used.

1-3 postnatal days mouse pups were anesthetized with 4% isoflurane until reflexes were lost and eviscerated after decapitation. Spinal cords were isolated via laminectomy, sectioned at the thoracic level 1 (T1) and placed ventral side up in the recording chamber. Dissection and recording procedures were performed under continuous superfusion with artificial cerebrospinal fluid (aCSF) containing the following (in mM): NaCl 130; KCl 3; CaCl₂ 2.50; MgSO₄ 1.3; NaH₂PO₄ 0.58; NaHCO₃ 25; glucose 10; equilibrated with 95% O₂, 5% CO₂ adjusted to pH 7.4 at room temperature (23°C). Recordings were performed after a recovery period of at least one hour.

Lumbar spinal cords from P1-P3 and P8-P12 mice were dissected free in an ice-cold sucrose-based saline solution containing the following (in mM): KCl 2, CaCl₂ 0,5, MgCl₂ 7, NaH₂PO₄ 1,15, NaHCO₃ 26, glucose 11 and sucrose 205, bubbled with 95% O₂, 5% CO₂ and sectioned into transverse slices (350 μm) with a vibroslicer (Leica, VT1000S). Slices were allowed to recover in oxygenated aCSF (see above) for at least 1 h at 30°C before being transferred to the recording chamber.

In the in vitro spinal cord preparation, motor output was recorded extracellularly using a custom-made differential AC amplifier from the right and left lumbar 2 (rL2, lL2, respectively), rL5 and lL5 ventral roots using glass suction electrodes filled with aCSF. Bouts of fictive locomotion were evoked by electrical stimulation (50 Hz, 2 s) of the VLF using a concentric bipolar electrode (Phymep) positioned between the T9 and T11 levels (Magnuson and Trinder, 1997). Neurograms were amplified (x 5000) using high impedance AC amplifiers (200 – 3000 Hz) built at the laboratory, digitized, acquired at 2 kHz and integrated using Axograph software (Axograph, Australia), for future analysis. Ten VLF stimulations were applied with an interval of 2 min. The stimulation series was started at the beginning of the perfusion of the mGluR agonists. In our experimental conditions, with a flow rate of 7 ml/min, the solution in the bath was completely replaced around 10 min after the beginning of the perfusion. To reduce the number of animals used, after a first series of VLF stimulations performed in aCSF, the three agonists were successively applied always in the same order (DHPG, LY354740 and then L-AP4) on each spinal cord preparation tested, with a wash-out period of at least one hour with aCSF between each drug perfusion. To allow comparison between experiments and drug applications, values of motor burst amplitudes and rhythm periods were normalized to the mean amplitude and period, respectively, measured during the first two responses obtained in a given series.

Whole-cell patch clamp recordings from putative MNs located throughout the lumbar spinal cord enlargement, identified by their relatively large soma (superior to 20 μm) and location in the lamina IX, were made under visual control with a Multiclamp 700B (Axon Instruments), using glass microelectrodes (3–6 MΩ) filled with the following (in mM): K Gluconate 120, KCl 20, MgCl₂ 0.1, EGTA 1, HEPES 10, CaCl₂ 0.1, GTP 0.1, cAMPc 0.2, Leupeptine 0.1, D-mannitol 77, Na₂-ATP 3, with a pH of 7.3. All experiments were conducted at room temperature (23°C). Voltage- or current- clamp recordings were obtained using a Multiclamp 700B amplifier (Axon Instruments, CA, United States). Data acquisition and analysis were performed using Axograph software. All intracellular recordings were made in high cation aCSF containing 7.5 mM CaCl₂ and 8 mM MgSO₄ in order to decrease polysynaptic transmission. Throughout recordings, GABAergic and glycinergic inputs were blocked with gabazine and strychnine (1 μM each), respectively.

Input membrane resistance was determined from voltage-current curves obtained from MNs held at −60 mV. After-hyperpolarization (AHP) parameters were computed from single action potential (AP) elicited by injecting a brief depolarizing current pulse (7 ms, 0.25 nA). MN firing behavior was studied using series of depolarizing current pulse injections with increasing amplitude. Then, the instantaneous frequency of firing (f-I) was calculated and fitted with linear functions to compute the slope of the linear part (first three points) of the function. MNs were held at −60 mV by bias current application ranging from −500 pA to + 150 pA in current clamp mode.

Synaptic responses and activity-dependent synaptic plasticity were evoked as previously described (Lenschow et al., 2016). A bipolar stimulating tungsten electrode was placed in the ventrolateral quadrant of the spinal cord slice to stimulate axons of the VLF. Excitatory postsynaptic currents (EPSCs) were evoked in MNs held at −60 mV in voltage-clamp mode by monophasic constant current pulses (100 μs) delivered through a stimulus isolation unit (ISO-flex, AMPI) to a bipolar stimulating tungsten electrode (Microprobes, tip separation 75 μm, stimulation intensities ranging from 10 to 60 μA). After a stable period of 10 min paired-pulse VLF stimulation (50 ms interval applied at 0.03 Hz), a 50 Hz high frequency stimulation (HFS) was applied for 2 s to VLF axons (VLF-HFS). During VLF-HFS, the intensity of the VLF stimulation was doubled compared to the baseline condition for eliciting EPSCs, and MNs were held in current-clamp mode to allow normal depolarization and firing. VLF-EPSC amplitudes were expressed as values normalized to mean control pre-HFS VLF-EPSC amplitude values. MNs were categorized into short-term depression (STD)-expressing motoneurons when VLF-EPSC amplitude exceeded 80% of the pre-HFS baseline control value 500 s after HFS and into long-term depression (LTD)-expressing motoneurons when depression lasted throughout the recording period (more than 30 min).

All drugs were prepared as stock solutions, aliquoted and frozen until use. The following pharmacological agents from Abcam (Cambridge, United Kingdom) were used: gabazine (GABA_A antagonist; 1 μM), a group I/II metabotropic glutamate receptor (mGluR) antagonist MCPG (methylene cyclopropyl glycine; 200 μM); a group I mGluR agonist DHPG (3,5-dihydroxy- phenylglycine; 5 μM); a group II mGluR agonist LY354740 [(1S,2S,5R,6S)-2-aminobicyclo[3.1.0]hexane-2,6-dicarboxylic acid; 0.5 μM] and a group III mGluR agonist L-AP4 (l-2-amino-4-phosphonobutylate; 1 μM). The concentration of the different mGluR agonists used in the present study was determined by performing dose-response curves to reach similar inhibitory effects on the glutamatergic synaptic transmission (data not shown, see section “RESULTS”). The glutamate transporter inhibitor TBOA (DL-threo-benzyloxyaspartic acid; 5 μM) was purchased from Tocris (Bristol, United Kingdom), and the glycine receptor antagonist strychnine (1 μM) from Sigma-Aldrich (St. Louis, MO, United States).

Animals were anesthetized with isoflurane and implanted with bipolar electrodes (coated nichrome wire, 50 μm; A-M Systems, Carlsborg, WA, United States) directly pinned into the Gastrocnemius or Tibialis Anterior muscles under binocular control. Due to the small size of the pups, only one muscle per hindlimb was recorded (i.e., one Gastrocnemius or one Tibialis Anterior). Animals were left to recover from anesthesia until spontaneous movement production and then placed into a tank filled with warm (ca. 35°C) water. Spontaneous swimming activity was recorded during approximately 10 s. Electromyographic signals were amplified (x 5000) using custom-made amplifiers, and filtered at 2 kHz. Data acquisition and analysis were performed using Axograph software. After three sessions of control swimming activity spaced 5 min apart, mice were randomly assigned to the following four treatment groups: vehicle (aCSF), DHPG (20 mg/kg), LY354740 (0.5 mg/kg), or L-AP4 (10 mg/kg). These concentrations were determined by an extrapolation from the concentrations used for intracellular recordings by a simple conversion based on mouse pup weight. The DHPG concentration has to be increased due to an absence of effects in vivo, while the LY354741 concentration was reduced due to a complete blockage of the swimming activity in pups (data not shown). Solutions were injected subcutaneously using an Hamilton syringe (Hamilton, Reno, NV, United States) at a volume of 0.3 μL for P3 and 1 μL for P8-P10 animals. Swimming activity was recorded 5, 10, 20 and 30 min after the injection. Between each session, mouse pups were placed on a heating pad.

The ventral lumbar part of the spinal cord or individual L2 and L5 ventral segments were isolated from isoflurane-anesthetized mouse pups and immediately frozen at −80°C until use. During all procedures, care was taken to avoid RNA degradation. Total RNA was extracted from samples using the RNA NucleoSpin^® RNA Plus extraction kit (Macherey-Nagel, Hoerdt, France). RNA quality and quantity were analyzed using Agilent RNA 2200 Screen Tape system (Agilent Technologies, Les Ulis, France) and DS-11 spectrometer (DeNovix, Wilmington, NC, United States) respectively.

To perform RT-qPCR from identified MNs innervating the Gastrocnemius or Tibialis Anterior muscles, MNs were backfilled with cholera toxin ß-subunit (CTB) conjugated with Alexa Fluor 594 or Alexa Fluor 488 (C34775 and C34777, respectively, Thermo Scientific, Waltham, MA, United States). Both muscles of both hindlimbs were injected in anesthetized P0 mice. After 12 h of migration, the lumbar spinal cord was dissected out from anesthetized animals, cryosectioned into longitudinal sections (20 μm) and collected on polyethylene-naphthalate membrane RNase-free microscope slides. Immediately after dehydration in successive baths of ethanol at increasing concentrations, labeled MN somas were dissected at 63x magnification using a PALM laser microdissection and capture system (P.A.L.M. Microlaser Technologies AG, Bernried, Germany). Gastrocnemius or Tibialis MNs were collected in separate tubes with adhesive caps for no longer than 30 min per slide to limit RNA degradation. The collected material was then treated with lysis buffer and stored at −80°C until RNA extraction. In the case of laser capture microdissection (LCM) samples, the ReliaPrep™ RNA Cell Miniprep System from Promega (Promega, La Farlede, France) was used for RNA extraction. The RNA integrity and quantity were checked by the Bioanalyser 2100 (Agilent Technologies, Massy, France) and the Nanodrop 1000 (Thermo Scientific, Waltham, MA, United States), respectively.

For all the samples used, total RNA was reverse-transcribed to cDNA with GoScript™ Reverse Transcription Kit (Promega) by standard protocols. Samples were distributed in duplicate using the Eppendorf epMotion 5073 automated pipetting. PCR was performed with the GoTaq Master Mix from Promega and the CFX384 real-time PCR detection system (Bio-Rad, Marnes-la-Coquette, France). Cycling parameters for the qPCR reaction included a 3 min hot start followed by 40 cycles of denaturation at 90°C for 10 s, annealing at 60°C for 30 s, and elongation at 72°C for 30 s. Sdha (succinate dehydrogenase) and actine ß (Actß) were used as internal controls. Two technical replicates were run for each RT-qPCR experiments and relative expression levels were calculated using the 2^–ΔΔCt method after normalization to the Sdha housekeeping gene. The primers sequences used (Table 1) were provided by Sigma Genosys (The Woodlands, TX, United States).

Statistical analysis was conducted on raw data using GraphPad Prism software. The data were checked for normal distribution. Wilcoxon matched pairs or Mann–Whitney tests were used to compare two series of data. Kruskal–Wallis one-way analysis of variance (ANOVA) was carried out to test for significant effects between the different drugs for unpaired observations. Two-way or RM two-way ANOVAs followed by Sidak’s multiple comparison tests were performed to evaluate drug and age effects. Chi-square tests were used to compare motoneuron proportions displaying STD, LTD or no change in the different conditions tested. In the text and figures, all data are expressed as means ± SEM. Statistical significance level was set at p < 0.05.

In the baseline control condition, paired-pulse stimulation (50 ms interval) applied to the VLF in slices triggered excitatory postsynaptic currents (VLF-EPSCs) in lumbar MNs held at −60 mV (Figure 1A). As previously shown by our group (Lenschow et al., 2016), VLF-EPSCs exhibit paired-pulse facilitation (PPF) with this stimulation paradigm (Figures 1A1–A3). In 1 week old rat pups, MCPG, a broad spectrum mGluR antagonist has been shown to block the late component of dorsal root response in lumbar MNs but failed to affect VLF-induced responses (Arvanian et al., 2005). We observed a similar lack of effect of 200 μM MCPG on both VLF-EPSC amplitude (Figures 1A1,A4) and the PPF ratio (VLF-EPSC2/VLF-EPSC1 = 1.3 ± 0.3 in control condition and 1.3 ± 0.3, n = 18 in the presence of MCPG, p = 0.5, Wilcoxon matched-pairs test) in P1-P3 mice. At excitatory glutamatergic synapses, mGluRs, in contrast to NMDA and AMPA receptors that are enriched in the postsynaptic density, are localized perisynaptically and have been shown to be activated when high concentrations of glutamate are reached in the synaptic cleft (Scheefhals and MacGillavry, 2018). We therefore examined the effects of enhancing endogenous glutamate level on VLF-EPSC amplitude with the glutamate transporter inhibitor TBOA. The bath-application of 5 μM TBOA on spinal cord slices significantly decreased the VLF-EPSC amplitudes (Figures 1A2,A4). However, this depressive effect was counteracted when MCPG was added to the TBOA-containing aCSF (Figures 1A3,A4). These results thus indicate that increasing the endogenous glutamate concentration not only activates mGluRs, but also that the latter might be acting as inhibitory autoreceptors at VLF-MN synapses.

It has been previously shown that the activation of extrasynaptic mGluRs by high synaptic concentrations of glutamate that spill over the synaptic terminals is a major determinant of synaptic plasticity expressed in supraspinal structures (Gonçalves-Ribeiro et al., 2019; Valtcheva and Venance, 2019). We thus investigated whether the activation of mGluRs at VLF-MN synapses during VLF high-frequency contributes to the ADSP expression at these synapses. As previously reported (Lenschow et al., 2016), a 2 s 50 Hz stimulation applied to VLF axons (VLF-HFS), that mimic the discharge of reticulospinal neurons during locomotion (Matsuyama and Drew, 2000a,b), leads to a long (more than 30 s; Figures 1B1,B2) or a short-term (Figures 1B3,B4) depression (LTD, STD respectively) of VLF-EPSC amplitudes at VLF-MN synapses in control conditions in P1-P3 mice. LTD and STD were almost equally distributed amongst the 31 recorded lumbar MNs (Figures 1B5). Of 19 MNs then exposed to the mGluR antagonist MCPG, LTD (Figures 1C1,C4) and STD (Figures 1C2,C4) were still expressed in 12 and 3 MNs, respectively, whereas in the remaining 4 MNs, VLF-HFS failed to trigger any changes in VLF-EPSC amplitudes (No plasticity, Figures 1C3,C4).

To amplify the glutamate release induced by HFS and to further activate potential perisynaptic mGluRs at VLF-MN synapses, VLF-HFS was applied in the presence of TBOA. As shown in Figure 1D, in these experimental conditions, it was still possible to induce LTD (Figures 1D1,D3) of VLF-EPSC amplitudes but the expression of STD was completely blocked in lumbar MNs (Figures 1D2,D3). The magnitudes of depression of VLF-EPSC amplitudes observed during LTD or STD in the presence of MCPG were not significantly different from those in the control condition: LTD: control −45 ± 6%, n = 14, MCPG −42.4 ± 4%, n = 12 and STD: control −59.3 ± 4%, n = 17 and MCPG −34.6 ± 8%, n = 3 (Table 2). In contrast, a significant increase in LTD magnitude was observed in the presence of TBOA (compare Figures 1B2,D1): control −45 ± 6%, n = 14, TBOA −62.8 ± 4%, n = 12, Mann-Whitney test, p = 0.04). Altogether these data indicate that mGluRs are activated during VLF-HFS and contribute to the expression of ADSP at VLF-MN synapses.

To assess the separate roles of the three different groups of mGluRs in the modulation of ADSP at VLF-MN synapses, we tested the impact of exposure to specific mGluR agonists, DHPG, LY354740, and L-AP4 for mGluRI, mGlurII, and mGluRIII, respectively (see section “DISCUSSION”). The agonists were applied to the slices throughout recording periods. 10 min after the onset of each agonist’s superfusion, a 10 min period of stable low frequency (0.03 Hz) VLF stimulation was applied prior to VLF-HFS application. In P1-P3 mice, as in the control condition (Figures 2A,G), in the presence of DHPG, VLF-HFS elicited both LTD and STD (Figures 2B,H) but also failed to induce changes in VLF-EPSC amplitudes in some of the lumbar MNs tested (Figures 2B,H). Similar diverse effects were observed in the presence of the mGluRIII agonist, L-AP4 (Figures 2D–J). LTD was expressed in 44% of the MNs tested, STD in 36.8% and No Plast in 15.8% in the presence of DHPG (Figure 2H), while with L-AP4, LTD-expressing MNs represented 50% of the neurons tested, STD-expressing MNs 25% and No Plast-expressing ones 25% (Figure 2J). When mGluRsII were activated with LY354740, LTD was observed in 44% of the MNs recorded (Figures 2C,I), but only one of the 25 tested MNs presented an STD (4%, Figure 2I) after the VLF-HFS. The 52% remaining neurons recorded in the presence of LY354740 exhibited no synaptic plasticity (Figures 2C,I), indicating that the mGluRII agonist was causing an almost complete and selective blockade of STD expression at VLF-MN synapses.

To assess for developmental changes, the same experiments were conducted on lumbar MNs recorded from P8-P10 mice. At this maturational stage, in control conditions, VLF-HFS again led to either LTD, STD or no plasticity (Figures 2E, G) as previously reported (Lenschow et al., 2016). When applied in the presence of each of the three mGluR agonists tested, VLF-HFS also elicited one of the same three possible responses - i.e., LTD, STD and No Plast- at VLF-MN synapses, regardless of the compound tested (Figures 2F,H,I,J). There were also no significant differences between the three different agonists in terms of the number of MNs expressing a particular response (Figures 2H,I,J; Chi square test, p = 0.45). Moreover, the magnitudes of depression of VLF-EPSC amplitudes attained during LTD or STD were not significantly different in control condition and in the presence of each of the three mGluR agonists (Table 3). Altogether these data therefore suggest that in contrast to P1-P3 MNs, mGluR activation does not influence the expression of ADSP induced by a HFS at VLF-MN synapses of P8-P10 mice.

In a next step, we tested whether mGluR activation was accompanied by a modulation of the synaptic inputs received by MNs. Under baseline conditions of VLF stimulation that resulted in PPF, DHPG had no effect on VLF-EPSC amplitudes in P1-P3 MNs (Figures 3A1,D1), but caused a significant decrease in this parameter in P8-P12 MNs (Figures 3A2,D1 and Table 4, VLF-EPSC ampl). The presence of the mGluRI agonist did not significantly modify the PPF ratio at both developmental stages tested (Figures 3A,E1 and Table 4, EPSC2/EPSC1 ratio). In contrast, LY354740 (Figure 3B) and L-AP4 (Figure 3C) induced a substantial decrease in VLF-EPSC amplitude in both P1-P3 and P8-P10 lumbar MNs (Figures 3D2,D3 and Table 4) that was consistently accompanied by an elevation in the PPF ratio (Figures 3E2,E3 and Table 4), suggesting a presynaptic locus of mGluR II and mGluR III modulation.

As previously described (Vinay et al., 2000; Smith and Brownstone, 2020), we observed that the input membrane resistance of MNs significantly decreased with age: mean 72 ± 3 MΩ, n = 86 from 65 P1-P3 pups versus 48 ± 3 MΩ, n = 67 from 59 P8-P12 mice, Mann-Whitney test, p < 0.0001. Regardless of age, the three mGluR agonists had no significant effect on the resting membrane potential of MNs when bath-applied in current clamp mode but significantly increased their input membrane resistance (Figure 4A and Table 4, Rin). We then sought whether mGluR activation changed the instantaneous frequency-current (f-I) relationship in lumbar MNs by applying depolarizing current pulses injected in series (Figure 4B). In contrast to previous findings (Vinay et al., 2000; Smith and Brownstone, 2020), we observed that the rheobase computed in control condition was significantly increased in P8-P10 MNs compared to P1-P3 MNs (P1-P3: 543 ± 30 pA, n = 68 and P8-P12: 866 ± 80 pA, n = 49; Mann-Whitney test, p = 0.0006). During this protocol, two different types of responses were observed when spike threshold was reached: 22 out of 68 (32%) P1-P3 MNs and 29 out of 51 (43%) P8-P12 MNs exhibited only a solitary spike at the beginning of the depolarizing pulse (Figure 4B1), whereas in the remaining cells, tonic impulse firing occurred throughout the depolarization (Figure 4B2). Neither firing pattern was modified in the presence of any of the three mGluR agonists tested (examples shown for DHPG and L-AP4 in Figures 4B1,B2, respectively; data not shown for LY354740). Unexpectedly, we observed no significant influence of the three mGluR agonists on the rheobase (Figure 4C, and Table 4, Rheobase) at P1-P3 nor on the slope of the f-I relationships in the two age ranges examined (Figure 4D and Table 4, f-I slope). In P8-P10 MNs, however, DHPG and L-AP4 (Figures 4C1,C3), but not LY354740 (Figure 4C2) significantly decreased the MN rheobase (Table 4, Rheobase).

We then examined whether the after-spike hyperpolarization (AHP), a major target of neuromodulators in MNs, was modified in the presence of the mGluR agonists. In control conditions, the AHP amplitude values were found not to be significantly different between P1-P3 and P8-P12 MNs (2.4 ± 0.2 mV, n = 82 and 2.7 ± 0.2 mV, n = 65, respectively; Mann-Whitney test, p = 0.3). As illustrated in Figures 5A1,A2, a significant decrease in AHP amplitudes was found only in the presence of DHPG in P8-P10 MNs, whereas for the other two mGluR agonists tested, no significant effects were observed in either age range (Figures 5A3,A4 and Table 4, AHP ampl). This finding is therefore consistent with the lack of effect of the agonists on the f-I relationships of MNs, which have been shown to be closely correlated with AHP amplitudes (Fulton and Walton, 1985).

To complete this analysis of the cellular basis of mGluR modulation in lumbar MNs, we examined the impact of the three mGluR agonists on the hyperpolarization-activated mixed cation current I_H, another important cellular target of neuromodulatory systems in the spinal cord (Deardorff et al., 2021). For this purpose, voltage steps of -60 mV from a holding membrane potential of -60 mV were applied and the instantaneous current evoked was measured immediately after the capacitive transient (filled diamond in Figure 5B1) and the steady-state current at the end of the hyperpolarizing pulse (open diamond in Figure 5B1) in control condition and in the presence of each mGluR agonist. As previously established, the difference between the steady state and the instantaneous current represents the I_H current (Bertrand and Cazalets, 1998; Kjaerulff and Kiehn, 2001; Tartas et al., 2010). The amplitude of I_H computed in P8-P12 MNs (−192 ± 19 pA, n = 67) was found to be significantly higher than in P1-P3 MNs (−96 ± 8 pA, n = 83, Mann-Whitney test, p < 0.0001, Figures 5B2–B4), and all three mGluR agonists significantly decreased I_H amplitude regardless of age (Figure 5B and Table 4).

Together these results show that mGluR modulation in lumbar MNs is complex, with specific effects being associated with a specific mGluR subtype and developmentally regulated.

At present, there are only few data available concerning the developmental expression of the various mGluR genes in the motor spinal cord (Berthele et al., 1999). To further explore these developmental changes, we looked for age-related modifications in expression levels of six mGluR genes (Grm) known to be present in the mouse spinal cord: Grm1, 2, 3, 4, 5, and 7 (Valerio et al., 1997; Berthele et al., 1999) by RT-qPCR analysis. In the whole lumbar ventral spinal cord of P1-P3 and P8-P12 mice, all 6 mGluR gene subtypes were found to be present in the two age ranges (n = 12 mice per age group), although at least four subtypes, mGluR5, mGluR2, mGluR3 and mGluR7, were significantly down-regulated during the second postnatal week (Figure 6A and Table 5, P8-P12 vs P1-P3).

Since the plasticity expressed at VLF-MN synapses is a function of the flexor or extensor function of lumbar MNs (Lenschow et al., 2016) and is modulated by mGluR activation, we next asked whether flexor and extensor MNs exhibit different and specific expression profiles of the mGluR genes. To this end, in a first series of experiments, we compared mRNA levels between the L2 and L5 ventral spinal segments (n = 12 animals per age group) which have been shown to contain predominantly flexor and extensor MNs, respectively (Cazalets et al., 1992; Kiehn and Kjaerulff, 1996). In P1-P3 animals, no significant differences in gene expression levels were found between the L2 and L5 segments (Figure 6B1 and Table 5, P1-P3 L5 vs L2). However, in P8-P12 mice, Grm3 and Grm7 were significantly more expressed in L5 compared to the L2 segment (Figure 6B2 and Table 5, P8-P12 L5 vs. L2).

In a next step, we quantified and compared mGluR gene expression levels in laser-microdissected MNs that had been retrogradely labeled from the ankle extensor muscle Gastrocnemius or from the ankle flexor Tibialis Anterior muscle (Figure 6C1). It should be noted that this analysis was restricted to MNs in P1 animals only, since Gastrocnemius MNs have been shown to split into LTD-MNs or No Plast-MNs after VLF-HFS in P8-P12 MNs and could therefore not be considered as an homogenous population in terms of ADSP expression (Lenschow et al., 2016). The mRNA from 1017 Tibialis MNs (from 15 P1 mice) and 783 Gastrocnemius MNs (from 5 P1 mice) were collected and purified (Rin = 8.7 and 7, respectively). With this experimental approach, all the genes tested were detectable in both MN subtypes with the exception of Grm4 and RT-qPCR analysis revealed that the level of Grm2 expression was significantly lower in Gastrocnemius MNs compared to Tibialis MNs (Figure 6C2 and Table 5, P1-P3 Gastroc vs Tibialis). This finding therefore indicated that extensor and flexor MNs exhibit slightly, albeit significantly, different mGluR expression profiles.

As the ADSP expressed at VLF-MN synapses is dependent on the flexor or extensor identity of MNs, we sought whether the effects of the mGluR agonists differ between MNs exhibiting different ADSP profiles at VLF-MN synapses after VLF-HFS. We found that the mGluR-mediated modulations of VLF-EPSC amplitude values, EPSC2/EPSC1 ratio, input membrane resistance and AHP amplitude were similar whatever the ADSP expressed by MNs and at the two developmental stages tested (data not shown).

In the in vitro spinal cord preparation of new-born rodents, bouts of fictive locomotion can be induced by 50 Hz stimulation of the VLF and recorded from lumbar ventral roots (Magnuson and Trinder, 1997). We then took advantage of this experimental protocol to decipher the functional impact of mGluR modulation at the motor network level under experimental conditions that should trigger ADSP at VLF-MN synapses. For this purpose, series of 10 stimulations (50 Hz, 2 s) of the VLF were applied every 2 min to induce episodes of fictive locomotion that were compared in the absence and presence of the mGluR agonists. To assess possible differential effects on extensor and flexor MNs, recordings were performed from the L2 and L5 ventral roots (Cazalets et al., 1992; Kiehn and Kjaerulff, 1996). In control aCSF alone, all preparations examined exhibited episodes of evoked fictive locomotion with stable period values (Figures 7A,F), numbers of bursts (6.2 ± 0.6 for the first VLF stimulation and 6 ± 0.5 for the last stimulation, n = 14 preparations) and L2 burst amplitudes (Figures 7A,G) across individual series of 10 VLF stimulations, although a small decline in the amplitude of L5 motor bursts occurred (Figures 7A,H). As indicated in the section “MATERIALS AND METHODS,” the three agonists were tested on each spinal cord preparation, with successive applications being separated by a one hour period of aCSF perfusion to allow a complete wash-out of the preceding applied agonist. After each drug trial, effective washout was reached as indicated by the similar mean periods and even larger motor burst amplitudes measured during the two first evoked locomotor episodes in each stimulation series (mean period: aCSF 1.3 ± 0.05 s, DHPG 1.3 ± 0.08 s, LY354740 1.3 ± 0.06 s, L-AP4 1.2 ± 0.05 s, n = 14, Kruskal-Wallis test, p = 0.63; L2 amplitude relative to that in aCSF: DHPG 1.7 ± 0.3, LY354740 1 ± 0.06, L-AP4 1.3 ± 0.2, n = 14, Kruskal-Wallis test, p = 0.06; L5 amplitude relative to that in aCSF: DHPG 1 ± 0.1, LY354740 1.5 ± 0.1, L-AP4: 1.5 ± 0.2, n = 14, Kruskal-Wallis test, p = 0.01, significantly different).

In the presence of DHPG, all preparations tested maintained the ability to express locomotor episodes throughout each stimulation series (Figure 7E). The amplitudes of L2 and L5 bursts remained stable during DHPG application (Figures 7B,G,H and Table 6, L2 and L5 amplitude) but the rhythm cycle period progressively increased until it became significantly different from that in the aCSF condition for the 9th and 10th VLF stimulations (Figures 7B,F and Table 6, period of the VLF-induced activity). In contrast, both LY354740 and L-AP4 led to an eventual suppression of the VLF-induced fictive locomotion before the 10^th stimulation in about 50% of the preparations tested (Figure 7E). In the remaining preparations that still expressed locomotor activity, L-AP4 induced a significant reduction of the rhythm period (Figures 7D,F and Table 6, period of the VLF-induced activity) as well as causing a strong decrease in both L2 and L5 motor burst amplitudes (Figures 7D,G,H and Table 6, L2 and L5 burst amplitude). In contrast, LY354740 only affected locomotor burst amplitudes, but did not significantly alter the cycle period of each evoked locomotor episode (Figures 7C,F,G,H and Table 6).

All three mGluR agonists tested have been shown to modulate neuronal processes through systemic routes of administration (Linden et al., 2004; Jin et al., 2012; Zhang et al., 2012). On this basis, together with the relative permeability of the blood brain barrier in newborn rodents (Saunders et al., 2012), we then tested the effects of subcutaneous injections of mGluR agonists on the locomotor behavior of P3 and P8-P10 pups by making EMG recordings from the Gastrocnemius and Tibialis Anterior muscles during episodes of spontaneous swimming behavior (see Materials and Methods). The subcutaneous injection of aCSF had no effect on the parameters of monitored swimming activity, either of P3 (Figures 8A,E1,F1 and Table 7) or P8-P10 animals (Figures 8C,E3,F3 and Table 7). In P3 mice, DHPG did not modify either swimming cycle periods or the amplitudes of motor bursts recorded from hindlimb muscles (Figures 8B,E2,F2 and Table 7). In P8-P10 animals, however, whereas DPHG was still ineffective in modulating swimming cycle period, a significant decrease in motor burst amplitudes compared to pre-injection values were observed in both recorded muscles (Figures 8D,E4,F4 and Table 7). In contrast, the mGluRII agonist LY354740 exerted similar effects when injected into both P3 or P8-P10 animals, leading to a progressive slowdown of spontaneous swimming activity coupled with a significant decrease in the amplitudes of motor bursts in both muscles recorded (Figures 9A,E1,F1, 9C,E3,F3, respectively and Table 7). Finally, the injection of L-AP4 significantly altered swim cycle period as well as the amplitude of both Gastrocnemius and Tibialis motor bursts in P3 pups (Figure 9B,E2,F2 and Table 7) but failed to induce significant changes in P8-P10 pups (Figures 9D,E4,F4 and Table 7).