Functional contribution of mesencephalic locomotor region nuclei to locomotor recovery after spinal cord injury

Summary Spinal cord injury (SCI) results in a disruption of information between the brain and the spinal circuit. Electrical stimulation of the mesencephalic locomotor region (MLR) can promote locomotor recovery in acute and chronic SCI rodent models. Although clinical trials are currently under way, there is still debate about the organization of this supraspinal center and which anatomic correlate of the MLR should be targeted to promote recovery. Combining kinematics, electromyographic recordings, anatomic analysis, and mouse genetics, our study reveals that glutamatergic neurons of the cuneiform nucleus contribute to locomotor recovery by enhancing motor efficacy in hindlimb muscles, and by increasing locomotor rhythm and speed on a treadmill, over ground, and during swimming in chronic SCI mice. In contrast, glutamatergic neurons of the pedunculopontine nucleus slow down locomotion. Therefore, our study identifies the cuneiform nucleus and its glutamatergic neurons as a therapeutical target to improve locomotor recovery in patients living with SCI.

Correspondence frederic.bretzner.1@ulaval.ca In brief Roussel et al. characterize the distinct contribution of midbrain neurons to locomotor recovery after spinal cord injury (SCI). Although both glutamatergic cuneiform nucleus (CnF) and pedunculopontine nucleus contribute to spontaneous motor recovery, only activation of glutamatergic CnF initiates locomotion and improves stepping ability after chronic SCI.

INTRODUCTION
Although the spinal cord contains all the circuitry necessary for locomotion, people with spinal cord injury (SCI) are unable to walk due to the absence of commands from the brain. Motor recovery can be partially achieved by rehabilitative training and neuromodulatory therapies intended to promote the descending motor command from the brain to the spinal cord after SCI. [1][2][3][4][5][6][7] Recently, deep brain stimulation of the mesencephalic locomotor region (MLR), a supraspinal locomotor center, has been shown to improve locomotor functions in rats with chronic but incomplete SCI with even a few spared axonal fibers. [6][7][8] Interestingly, these functional changes come with an extensive reorganization in the brainstem region after SCI, 9 thus supporting the important contribution of the MLR to locomotor recovery after incomplete SCI.
The anatomic correlate of this functional region has been initially identified as the cuneiform nucleus (CnF), a cluster of glutamatergic neurons, and the pedunculopontine nucleus (PPN), a cluster of glutamatergic and cholinergic neurons. Despite a growing body of evidence from mouse genetics studies, [10][11][12][13][14][15][16][17][18] there is still debate about the exact anatomic correlate of this supraspinal locomotor center. Previously, using a headrestrained mouse on an air-lifted ball, 19,20 it was shown that optogenetic stimulation of glutamatergic neurons of the MLR (including the CnF and PPN) can generate locomotion in contrast to cholinergic or GABAergic MLR neurons. More recently, using smaller volumes of adeno-associated virus to circumscribe photostimulation to a nucleus of interest, it was shown that glutama-tergic CnF neurons can initiate locomotion in freely behaving mice, 10,11,15 whereas glutamatergic PPN neurons exhibit higher variability in generating locomotion. 10,11,[14][15][16] Furthermore, glutamatergic CnF neurons accelerate locomotor rhythm and speed during ongoing locomotion, whereas glutamatergic PPN neurons only prolong the stance phase, contributing to postural adjustments and slowing locomotor rhythm in normal conditions. 10,15 Although cholinergic PPN neurons were initially reported to increase speed during head-restrained locomotion, 20 more recent studies suggest that they do not actually modulate locomotor speed in freely behaving mice. 10,11 Although deep brain stimulation (DBS) in the PPN 7 or the CnF 8 can improve locomotor recovery in animal models of SCI, there are still questions about which neuronal population is the most efficient. With ongoing clinical trials aiming to assess DBS in the vicinity of the MLR of patients with incomplete SCI, 21 it is now urgent to gain a better understanding of how these distinct neuronal populations of the midbrain can contribute to and promote functional locomotor recovery after SCI.
We hypothesize here that glutamatergic neurons of the CnF and PPN contribute to spontaneous locomotor recovery after SCI and that selective activation of glutamatergic CnF, but not glutamatergic PPN neurons, can further improve locomotor functions. Combining detailed kinematics, electromyographic (EMG) recordings, anatomic analyses, and mouse genetics, we found that glutamatergic CnF neurons modulate locomotor pattern and rhythm and enhance motor efficacy in ipsilesional hindlimb muscles after SCI, in contrast to glutamatergic neurons of the PPN. As a therapeutical approach, we also found that 1-s (C) Locomotor score of the ipsilesional (left) hindlimb before and after SCI (n = 15 mice, Friedman test [p < 0.0001] with Dunn's multiple comparisons test, ****p < 0.0001; ***p < 0.001). (D) Stick diagrams of the ipsilesional hindlimb during the swing and stance phase (arrows indicate movement direction), joint angles, and gait diagrams (bars represent the stance). (E) Cumulative distribution function (CDF) of paw movement in reference to the iliac crest before and after SCI (n = 10 mice). (F) Forward and backward paw placement in reference to the iliac crest (Pre-SCI vs. Week 1-7, n = 10 mice, one-way ANOVA [ p < 0.0001] with Dunnett's multiple comparisons test, ****p < 0.0001, *p < 0.05). trains of photostimulation delivered above glutamatergic CnF neurons promote functional recovery of initiation and locomotion after SCI, whereas glutamatergic PPN neurons slow down locomotion, thus identifying the CnF and its glutamatergic neuronal population as a neurosurgical target to promote functional locomotor recovery in patients with SCI.

RESULTS
Mice exhibit asymmetrical locomotor pattern following incomplete SCI Adult mice underwent a lateral hemisection at the low thoracic level, abolishing supraspinal inputs from the brain on the left side of the lumbar spinal cord controlling hindlimb locomotion. Using kinematic and electromyographic recordings, we assessed spontaneous locomotor recovery after SCI (Figures 1 and S1). As previously reported in other animal species, 22-28 mice initially displayed a transient paralysis on the side of the SCI over the first week post-SCI ( Figures 1C, 1D, 1G and 1H) with a decrease in toe movement and forward foot placement ( Figures 1E and 1F) and in the motor activity of the flexor, which were co-activated with otherwise weak extensors ( Figures 1G-1J). This likely contributed to hind-paw dragging during swing and a very short stance. Although this limb displayed some locomotor-like movements, there was loss of weight support and no plantar stepping. Eventually, within a few weeks after SCI, the ipsilesional limb (i.e., on the lesion side) increased its stance duration and extensor activity, in addition to exhibiting a better coordination between its flexor and extensor locomotor activities ( Figures 1I and 1J). This in turn reduced hind-paw dragging during swing and improved plantar stepping ability ( Figures 1D and  1G), thus contributing to locomotor recovery of the ipsilesional limb over time.
No anatomic reorganization of medullar projecting MLR neurons after SCI Given that the medullary reticular formation relays MLR inputs during locomotion 29-31 and that SCI leads to an extensive reorganization of projections between midbrain and brainstem nuclei after SCI, 9 we hypothesized an asymmetrical reorganization of the connectivity between glutamatergic neurons of MLR nuclei and their postsynaptic medullary targets following incomplete SCI. To test this hypothesis, a retrograde tracer Fast Blue was injected stereotaxically in the contralesional medullary reticular formation of adult transgenic VGlut2-Cre mice 7 weeks after SCI or sham surgery ( Figure 2A; mice with Fast Blue injections leaking in the ipsilesional brainstem were excluded from our analysis, Figure S2). We specifically targeted the gigantocellular reticular nucleus, the alpha and ventral portion of the gigantocellular reticular nucleus, and the lateral paragigantocellular nucleus, which are important to the motor command. [32][33][34][35] Combining immunohistochemistry and stereological techniques ( Figure 2B), we identified and quantified the neurotransmitter phenotype (e.g., glutamatergic, cholinergic, or both glutamatergic/cholinergic) of retrogradely labeled neurons in ipsilesional (left) and contralesional (right) CnF and PPN nuclei.
Using the most ventral part of the fourth ventricle as a reference, our 3D and 2D reconstructions illustrate a bilateral and symmetrical organization of CnF and PPN nuclei according to their neurotransmitter phenotype (e.g., glutamatergic, cholinergic, or double) and their unilateral projection in the contralesional medullary reticular formation ( Figures 2C-2E). We found a high density of medullary-projecting glutamatergic neurons in both left and right CnF and PPN, but very few cholinergic cells within the PPN ( Figures 2C-2E), suggesting a bilateral organization of brainstem-projecting MLR populations in both sham and SCI mice. As previously reported, 36-39 we also found a few double glutamatergic/cholinergic neurons in the PPN ( Figure S2). Nevertheless, in contrast to a previous SCI study, 9 the cell density profile and count of glutamatergic and cholinergic neurons of CnF or PPN nuclei of SCI animals were not significantly different from those of sham mice ( Figures 2D-2G), suggesting that the organization of glutamatergic and cholinergic MLR neurons projecting to brainstem locomotor circuits was maintained after SCI.
Genetic deletion of glutamatergic neurons of the contralesional CnF or PPN impairs spontaneous locomotor recovery in chronic SCI mice Given the functional role of the MLR to locomotor recovery after SCI, 6,8,9 we next evaluated the requirement of glutamatergic neurons of specific MLR nuclei to functional recovery (Figures 3, S3, and S4). To test this requirement, locomotor functions were assessed before and after conditional genetic ablation of glutamatergic neurons of the contralesional (right) CnF or PPN in chronic mice that had spontaneously recovered locomotor functions 8 weeks after left SCI ( Figure 3A). We ensured that both CnF and PPN groups exhibited a similar behavioral recovery 8 weeks after SCI, 40 prior to genetic ablation of their glutamatergic neurons, and both groups showed the same extent of SCI on postmortem analysis ( Figure 3B). As previously reported, 41,42 the cre-dependent virus (AAV-mCherry-flex-DTA) allowed us to visualize the extent of the injection site by mCherry expression in all neurons and to restrict the genetic ablation to Cre-expressing neurons ( Figures S3A-S3E). In comparison with their contralateral control nuclei, there was a significant decrease of about 50% in NeuN-expressing neurons with no difference in the num-   Figures 3G-3I), however, genetic ablation of glutamatergic CnF neurons significantly decreased the angular excursion of the hip in 60% of mice, the knee in 80% of mice, and the ankle in 80% of mice, whereas ablation of glutamatergic PPN neurons significantly decreased the angular excursion of the hip in 20% of mice, the knee in 80% of mice, and the ankle in 50% of mice ( Figure 3H). Genetic deletion of glutamatergic CnF neurons also decreased the swim cycle frequency and power stroke in 80% of CnF mice ( Figure 3H) and swimming speed in all CnF mice ( Figure 3J), whereas fewer mice were impaired upon genetic ablation of glutamatergic PPN neurons. These findings argue that glutamatergic neurons of the contralesional CnF are more important than their PPN counterparts to spontaneous locomotor recovery after SCI.
Glutamatergic CnF neurons contribute to spontaneous locomotor recovery after SCI Having shown that glutamatergic neurons of the contralesional (right) CnF or PPN are important to spontaneous locomotor recovery after left SCI, we next assessed their functional contribution by evaluating changes in motor efficacy throughout the step cycle upon photostimulation ( Figures 4A and S5 for the extent of the cre-lox recombination and location of optical fibers). The extent of lesion size and changes over time in locomotor score were similar for both CnF and PPN groups ( Figure 4B). 40 To assess changes in motor efficacy as a proxy of dynamic changes in polysynaptic connectivity between the MLR and the motoneuronal pools, the percentage of failure, excitatory, and inhibitory phase-dependent EMG responses were measured in the ipsilesional flexor and extensor muscles over a 50-ms time window upon photostimulation of 10 ms pulse duration delivered during locomotion at steady and comfortable speed before and after SCI ( Figure 4C). An increase in the number of motor spikes indicated excitatory motor responses, whereas a decrease indicated inhibitory motor responses, and an absence of change indicated a failure. We quantified changes over time in the proportion of motor responses in the flexor muscle during the swing phase ( Figures 4D-4F) and in the extensor during the stance phase ( Figure S7) as function of the locomotor score of the ipsilesional hindlimb. There was a high failure rate in motor responses in both flexor and extensor muscles 1 week after SCI ( Figures 4D and S6B), which returned eventually toward pre-injury levels over time while the animals resumed spontaneous locomotor recovery. Indeed, changes in the failure rate in both muscles correlated negatively with the locomotor score of the ipsilesional hindlimb ( Figures 4D and S6B): the lower the failure rate, the higher the locomotor score, thus suggesting a transient interruption of the descending motor drive after SCI.
Regarding the flexor muscle during the swing phase ( Figures 4E  and 4F), changes in the proportion of excitatory motor responses correlated significantly only with changes in locomotor score upon photostimulation of glutamatergic CnF neurons ( Figure 4E). In contrast, changes in the proportion of inhibitory motor responses correlated only with changes in the locomotor score upon photostimulation of glutamatergic PPN neurons ( Figure 4F), suggesting that glutamatergic CnF neurons by their excitatory drive contribute more efficiently than the PPN to motor recovery of the flexor muscle during the swing phase after SCI.
Regarding the extensor muscle during the stance phase ( Figures S6C and S6D), correlations between changes in the proportion of inhibitory motor responses and locomotor scores were positive upon activation of either the CnF or PPN ( Figure S6D). However, changes in the proportion of excitatory responses correlated significantly and positively with changes in the locomotor score upon photostimulation of glutamatergic CnF neurons, whereas the correlation was negative upon photostimulation of glutamatergic PPN neurons ( Figure S6C), thus supporting overall a higher excitatory efficiency of the CnF over the PPN in spontaneous recovery of the extensor muscle during the stance phase after SCI.
Overall, changes in the density of motor spikes in excitatory and inhibitory motor responses were positively correlated with changes in locomotor score in both flexor and extensor muscles upon photostimulation of the CnF or PPN ( Figure S7). However, changes in the amplitude of motor spikes were only correlated with the locomotor score in excitatory responses of the flexor evoked upon glutamatergic CnF neurons ( Figure S8). Taken together, these changes in the excitatory motor drive and behavior suggest that glutamatergic CnF neurons by their action on some hindlimb muscles might contribute more efficiently than PPN neurons to spontaneous recovery of stepping ability after SCI.

Glutamatergic neurons of the CnF promote initiation of locomotion in chronic SCI mice
Having shown that glutamatergic neurons of both the CnF and PPN are necessary and contribute to some extent to spontaneous recovery after SCI, we next investigated whether stimulation of one of these neuronal populations would be more efficient in promoting initiation of locomotion after chronic SCI (Figures 5 and S9). As illustrated by their trajectory in open field (Figures 5A), 1-s trains of photostimulation delivered above glutamatergic neurons of the right CnF generated consistent episodes of locomotion (Figures 5B and 5D: 100% of trials at 20 Hz and 84% at 50 Hz) with long-distance displacement ( Figures 5A and 5D). As illustrated by body direction (Figure 5C), the first 500 ms of stimulation generated already straight locomotion ( Figure 5C) at very short latency ( Figure 5E). In contrast, 1-s trains of photostimulation delivered above right glutamatergic PPN neurons generated inconsistent and unreliable bouts of locomotion (Figures 5I-5K: 23% of trials at 20 Hz and 34% at 50Hz to evoke locomotion) and always occurred with a long latency after the end of the photostimulation (legend continued on next page) 3% of trials at 20 Hz and 5% at 50 Hz) but systematically evoked head rotation ipsilaterally to the stimulation without body displacement, as illustrated by shorter arrows ( Figure 5J, 100% of trials). Although electrical stimulation in the PPN has recently been shown to induce stress, 7 we did not find any signs of discomfort upon optogenetic stimulation ( Figures S9C-S9F). In summary, sole activation of glutamatergic neurons of the CnF generates straight locomotion in chronic SCI mice.
Activation of glutamatergic CnF neurons improves posture and recovery of basic and voluntary stepping in chronic SCI mice Having shown that glutamatergic neurons of the CnF initiate episodes of locomotion in animals at rest, in contrast to the PPN ( Figure 5), we next investigated how activation of these neurons can modulate posture and voluntary locomotion in a corridor.
One-second train of photostimulation of glutamatergic neurons of the right CnF increased the posture (i.e., iliac crest height), step height, and step frequency, and enhanced the locomotor speed of chronic SCI mice during ongoing locomotion ( Figures 6A-6C), whereas neurons of the PPN decreased locomotion and postural tone ( Figures 6D and 6E). Similar observations were also reported during treadmill locomotion at a comfortable steady speed ( Figure 7B for the CnF and Figure S11B for the PPN). Moreover, changes in postural tone increased linearly as a function of changes in locomotor speed: photostimulation of glutamatergic CnF neurons increased speed and posture, whereas glutamatergic PPN neurons decreased these parameters ( Figure S10), thus suggesting that these two neuronal populations could modulate gait and posture. Patients with SCI exhibit higher variability in their intralimb coordination that impedes their complete recovery. 43,44 We therefore hypothesize that by decreasing this variability, photostimulation of the glutamatergic neurons of the CnF might improve stepping ability. To further investigate that possibility, we analyzed kinematic changes of the intralimb coordination during treadmill locomotion at steady and comfortable speed of chronic SCI mice (Figure 7). As a proxy of variability, we measured the coefficient of variation of joints' angular excursion, which was especially high during the swing phase of locomotion prior to stimulation (black traces in Figure 7C). Photoactivation of the glutamatergic CnF neurons significantly decreased the variability in hindlimb joints (blue traces in Figure 7C), with the strongest changes in the hip and ankle followed by the knee (blue traces in Figures 7C, 60% of trials in the hip, 50% in the ankle, and 25% in the knee). Interestingly, this decreased variability during the swing phase co-varied according to the coefficient of variation of the hip. Indeed, when the coefficient of variation of the hip was significantly reduced, the variability of the ankle joint decreased significantly in 87% of trials and that of the knee in 47% of trials, thus supporting smoother and steadier stepping. Furthermore, analyses of EMG recordings also showed that activation of glutamatergic neurons of the CnF decreased the burst duration of flexor and extensor muscles of both ipsi-and contralesional hindlimbs ( Figures 7D and 7E for the ipsilesional hindlimb and Figure S11A for the contralesional hindlimb), thus contributing to an increased locomotor speed. Stimulation also increased the burst amplitude of the ankle dorsiflexor activity (e.g., tibialis anterior), contributing to increasing the step height and toe clearance during the swing phase of locomotion. Taken together, these results suggest that stimulation of glutamatergic neurons of the CnF can efficiently modulate the spatiotemporal recruitment of muscles, smoothen and stabilize the intralimb joint coordination of the ipsilesional hindlimb, and enhance speed, thus improving overall locomotor recovery after chronic SCI.
Unloading the body weight improves locomotor functions driven by glutamatergic CnF neurons in chronic SCI mice Although sensory feedback participates in functional locomotor recovery after SCI, unloading the body weight (even partially) is often combined with physical training to promote motor recovery in incomplete SCI patients and animal models. 45,46 Therefore, we tested whether 1-s trains of photostimulation delivered above glutamatergic neurons of either the contralesional (right) CnF or PPN can promote locomotor function while unloading the animal's weight during swimming in chronic SCI (Figures S12 and S13).
Seven weeks after SCI, activation of glutamatergic CnF neurons increased swimming speed and frequency (Figures S12A-S12D) and improved the trajectory of the hindlimb joints ( Figure S12E), in addition to increasing the angular excursion of the hip in 60% of mice and of the knee and ankle in 80% of mice ( Figure S12F). Furthermore, activation of the CnF significantly decreased the variability of hindlimb joints, thus improving fluidity of the movement during the power stroke and return stroke of the ipsilesional hindlimb in more than 80% of trials in the hip, knee, and ankle joints ( Figure S12G). In contrast, (F) Bilateral fore-and hindlimb coupling for all mice (each circle represents a mouse, anchored on the right forelimb and hindlimb, respectively). (G) Example from one mouse of the mean and SD of hip, knee, and ankle joint angles of the ipsilesional hindlimb during swimming before and after genetic ablation of glutamatergic CnF neurons. CDF of the angular excursion of hindlimb joints (n = swim cycles, Mann-Whitney test or unpaired t test according to the normality of the distribution).
(H) Percentage of mice with significant decrease (detoriation), increase (improvement), or absence of change in the angular excursion, swim cycle frequency, and power stroke (each mouse was compared with its pre-DTA level using a Mann-Whitney test or unpaired t test according to the normality of the distribution).   Figure S13). Taken together, these results show that activation of glutamatergic CnF neurons promotes more efficient and smoother intralimb coordination of the ipsilesional hindlimb, thus improving functional locomotor recovery of the ipsilesional hindlimb of chronic SCI mice.

DISCUSSION
Our results reveal new concepts that fundamentally alter the understanding of the distinct contribution of MLR nuclei to functional locomotor recovery following SCI and their potential as DBS targets to promote rehabilitation in patients with chronic SCI: (1)  The PPN is a suboptimal target for reliably generating locomotion Using animal models, the PPN has been initially identified as an anatomic correlate of the MLR based on electrical stimulation and postmortem reconstructions. [47][48][49] Given that Parkinson's disease (PD) patients typically display difficulties in initiating and executing locomotor movements, with rigidity, tremor, and postural instability, 50,51 DBS of the PPN has been investigated in advanced parkinsonian patients with motor complications who are refractory to pharmacological treatments. Although PPN stimulation improves gait and postural adjustments in some parkinsonian patients, 52,53 the functional outcomes have been extremely variable across clinical studies. 54 Such variability, also reported in several animal models of PD, 55,56 has recently raised questions about the efficacy of the stimulation parameters and the anatomic correlates of the MLR. 57,58 Although there is still no consensus regarding the most appropriate stimulation parameters, as little is known about the neural mechanisms activated by DBS in the vicinity of the PPN, 59-61 recent optogenetic studies of glutamatergic neurons of the PPN have shown functional discrepancies in animal models with locomotor initiation, deceleration, head rotation, and even anxious-like behaviors, 10,11,17,18 that could reflect the complexity of this nucleus. Indeed, in addition to the multiple pathways running through it, the PPN exhibits an extreme divergence in its presynaptic inputs and postsynaptic projections. 11,62 As shown by recent optogenetic studies, neuronal populations within the PPN can exhibit distinct functional effects according to their postsynaptic projection. If little is known about medullary-projecting PPN neurons, which will be the population most likely involved in locomotion, there is evidence that substantia nigra-projecting PPN neurons generate grooming and handling, 14 striatal-projecting PPN axons generate head rotation, 63 and spinally projecting PPN neurons induce locomotor arrest and rearing, 14 thus arguing that the PPN is a complex neurological structure with a wide range of functional motor outcomes.
Glutamatergic neurons of the CnF enable stepping ability after SCI without discomfort Recently, electrical stimulation of the PPN has been shown to initiate bipedal locomotion of chronic SCI rats contused at the  7 However, electrical PPN stimulation also appears to induce pain as reported by a grimace test in these chronic SCI animals. However, using optogenetic tools in the mouse, we found that photoactivation of glutamatergic neurons of the PPN did not induce any signs of discomfort in chronic SCI mice. Although we cannot exclude some discrepancies between both animal species and SCI models, electrical stimulation is not as specific as optogenetic activation in recruiting different neuronal populations, axons of passage, and unwanted neural structures in the vicinity of the electrode. By their caudal location to the PPN, electrical stimulation of the PPN might have also recruited the Kö lliker/parabrachial nuclei involved in grimaces and aversive responses during pain. 64,65 In contrast, optogenetic stimulation of glutamatergic neurons, especially in the CnF, resulted in almost no or very little apparent discomfort in chronic SCI mice.
Electrical stimulation of the CnF has also been recently shown to promote quadrupedal locomotion of chronic rats with a severe thoracic section sparing only the ventromedial funiculus. 8 Using a less severe thoracic SCI, we also found that optogenetic photostimulation of glutamatergic neurons of the CnF with 1-s trains of 20 Hz increased postural tone and initiated locomotion in chronic SCI mice, with functionally better stepping ability and smoothness Although stimulation of glutamatergic neurons of the PPN usually failed to evoke locomotion with our standard stimulation protocol at 20 Hz, increasing the stimulation frequency up to 50 Hz induced systematic upper body rotation ipsilaterally to the stimulation followed by very slow locomotor movements. These locomotor bouts, however, were not reliable and occurred only at very long latency after the end of the stimulation before and after SCI, as previously shown in intact mice. 11 Interestingly, using trains of 40 Hz for 10 s, glutamatergic neurons of the PPN have been recently shown to initiate locomotion at long latency in intact and akinetic mouse models following a pharmacological block of dopaminergic transmission. 16 Similar results have also been reported upon photostimulation at 20 Hz for 10 s of glutamatergic neurons of the CnF in intact mice and after severe acute dopaminergic depletion. 12,13 Although we cannot exclude the possibility that long photostimulation for 10 s could also generate locomotion after SCI, stimulations of 20 Hz for 1 s in the CnF were sufficient to initiate locomotion and improve inter-and intralimb coordination after chronic SCI. Overall, our findings suggest that the CnF and especially glutamatergic neurons of the CnF are a better target than the PPN for promoting functional recovery of locomotor initiation in chronic SCI.
Glutamatergic CnF neurons are necessary and sufficient to improve functional recovery during locomotion Genetic anterograde and retrograde tracing studies have shown a direct connection between glutamatergic neurons of both the CnF and PPN nuclei and the medulla. 11,14 Pharmacological, lesion, cooling, and genetic deletion studies have previously shown that reticulospinal pathways of the medullary reticular formation relay supraspinal MLR inputs to the spinal locomotor circuit. [29][30][31]35 Not surprisingly, reticulospinal pathways appear to contribute to functional locomotor recovery following SCI, 9,[66][67][68] presumably by relaying MLR and/or cortical inputs. Although anatomic reorganization has been reported between medullary-projecting midbrain neurons following cervical SCI 9 and axonal midbrain sprouting in the medulla after thoracic SCI, 8 we found no anatomic changes in medullary-projecting CnF or PPN glutamatergic neurons after thoracic SCI. This discrepancy in retrogradely labeled neurons could reflect some changes in other genetically unidentified neurons, or more likely differences in SCI models. However, despite the absence of anatomic changes, we found functional behavioral and motor improvements when stimulating glutamatergic CnF neurons. Indeed, genetic deletion of glutamatergic neurons of the CnF or PPN impaired spontaneous locomotor recovery during walking and swimming, but deficits were stronger upon genetic ablation of CnF neurons than those of the PPN. Similarly, motor efficacy measurements also revealed that excitatory motor responses evoked in ipsilesional flexor and extensor muscles by glutamatergic neurons of the CnF correlated robustly with changes in locomotor recovery. This was not the case with the PPN, suggesting that glutamatergic neurons of the CnF contribute more efficiently than those of the PPN to spontaneous recovery of locomotor functions following SCI.
Glutamatergic CnF neurons promote functional locomotor recovery in chronic SCI mice As previously shown before SCI, 10 activation of glutamatergic CnF neurons increased postural tone as well as locomotor pattern and rhythm in chronic SCI mice during treadmill locomotion, as well as during voluntary locomotion while walking in a corridor or swimming in a pool, whereas activation of glutamatergic PPN neurons usually slowed locomotor rhythm, eventually inducing locomotor arrests. Interestingly, similar decelerations and stops have also been recently reported upon photostimulation of PPN neurons projecting to the spinal cord in intact animals. 14 Although far from complete, there is growing understanding of the neural brainstem networks underlying locomotion. 69,70 Recent optogenetic studies have shown that photostimulation of glutamatergic neurons of the LPGi initiates locomotion and increases locomotor rhythm, 35 presumably by relaying glutamatergic inputs of the CnF. In contrast, photoactivation of glutamatergic or glutamatergic V2a (Lhx3/Chx10) expressing neurons of the gigantocellular reticular nucleus (GI ) reset locomotor rhythm and induce locomotor arrests, 33,34,71,72 presumably by relaying glutamatergic inputs of the PPN. Further studies are needed to genetically dissect the neural mechanisms of these medullary nuclei and their reticulospinal projections to motor recovery after SCI. Overall, our results support the hypothesis that glutamatergic CnF neurons are more efficient than those of the PPN in improving stepping ability of the ipsilesional hindlimb, in addition to enhancing gait and posture, in chronic SCI mice during stereotyped and voluntary locomotion.

Limitations of the study
With a current clinical trial assessing MLR DBS in patients with incomplete SCI, 21,73 our findings reveal that the CnF offers a better and more reliable neurological target in comparison with the PPN for promoting recovery of motor and locomotor functions, strengthening the importance of evaluating the role of the CnF and especially glutamatergic neurons of the CnF in patients suffering from SCI. Our results also highlight the importance of continuing genetic dissection of functional microcircuits within the PPN in biomedical research. Although optogenetic and optical technologies are still in their infancy regarding first-in-human clinical trials targeting more accessible neural structures to according to the evolution of coefficients of variation of the hip angle (middle). Pie chart showing that ankle and knee stabilization occurred mainly when the hip was stabilized (right). Article ll OPEN ACCESS restore sensory loss, 74,75 advances are needed for control of deep brain neuronal populations to promote recovery of gait and posture in humans suffering from SCI or neurodegenerative diseases, such as PD or amyotrophic lateral sclerosis. In summary, our current findings in an animal model of SCI suggest that DBS of the CnF or optogenetic activation of glutamatergic CnF neurons should be further investigated in chronic SCI patients.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following: