Author response: The role of V3 neurons in speed-dependent interlimb coordination during locomotion in mice

Article Figures and data Abstract Editor's evaluation Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Speed-dependent interlimb coordination allows animals to maintain stable locomotion under different circumstances. The V3 neurons are known to be involved in interlimb coordination. We previously modeled the locomotor spinal circuitry controlling interlimb coordination (Danner et al., 2017). This model included the local V3 neurons that mediate mutual excitation between left and right rhythm generators (RGs). Here, our focus was on V3 neurons involved in ascending long propriospinal interactions (aLPNs). Using retrograde tracing, we revealed a subpopulation of lumbar V3 aLPNs with contralateral cervical projections. V3OFF mice, in which all V3 neurons were silenced, had a significantly reduced maximal locomotor speed, were unable to move using stable trot, gallop, or bound, and predominantly used a lateral-sequence walk. To reproduce this data and understand the functional roles of V3 aLPNs, we extended our previous model by incorporating diagonal V3 aLPNs mediating inputs from each lumbar RG to the contralateral cervical RG. The extended model reproduces our experimental results and suggests that locally projecting V3 neurons, mediating left–right interactions within lumbar and cervical cords, promote left–right synchronization necessary for gallop and bound, whereas the V3 aLPNs promote synchronization between diagonal fore and hind RGs necessary for trot. The model proposes the organization of spinal circuits available for future experimental testing. Editor's evaluation This article will interest neuroscientists who study how spinal circuits control locomotion. While the role of spinal interneurons in control of left–right and flexor–extensor alternations has been studied extensively, their role in hind–forelimb coordination has not been sufficiently studied. Zhang et al. study interlimb coordination by combining experimental data and computer simulation to shed light on how a population of spinal neurons may coordinate hind and fore limbs during locomotion at different speeds. https://doi.org/10.7554/eLife.73424.sa0 Decision letter eLife's review process Introduction Coordinated rhythmic movement of the limbs during locomotion in mammals is primarily controlled by neural circuitry within the spinal cord. This spinal circuitry includes rhythm-generating (RG) circuits (Graham Brown, 1911; Graham Brown, 1914) and multiple commissural, long propriospinal, premotor, and pattern formation neurons (Grillner, 2006; Kiehn, 2006; Kiehn, 2011; Kiehn, 2016; Rybak et al., 2006; Rybak et al., 2013; Rybak et al., 2015; Jankowska, 2008; McCrea and Rybak, 2008; Danner et al., 2016; Danner et al., 2017; Danner et al., 2019; Ausborn et al., 2021). It is commonly accepted that each limb is controlled by a separate spinal RG (Forssberg et al., 1980; Thibaudier et al., 2013; Frigon, 2017; Danner et al., 2019; Latash et al., 2020) and that RGs controlling left and right forelimbs and left and right hindlimbs are located on the corresponding sides of cervical and lumbar enlargements of the spinal cord, respectively (Kato, 1990; Ballion et al., 2001; Juvin et al., 2005; Juvin et al., 2012). The left and right lumbar and cervical circuits are connected through multiple types of local commissural interneurons (CINs), which coordinate left–right activities (Stein, 1976; Butt and Kiehn, 2003; Quinlan and Kiehn, 2007; Talpalar et al., 2013; Bellardita and Kiehn, 2015; Rybak et al., 2015; Shevtsova et al., 2015). In turn, descending long propriospinal neuron (dLPNs) and ascending long propriospinal neurons (aLPNs) mediate interactions between cervical and lumbar circuits (Juvin et al., 2005; Dutton et al., 2006; Reed et al., 2006; Brockett et al., 2013; Ruder et al., 2016; Flynn et al., 2017; Pocratsky et al., 2020). Diverse populations of CINs and LPNs are involved in coordination of limb movements, defining locomotor gait, and controlling speed and balance during locomotion (Danner et al., 2016; Danner et al., 2017; Kiehn, 2016; Ruder et al., 2016; Pocratsky et al., 2020). Experimental studies with genetic ablation, silencing, or activation of genetically identified neuron types, such as V0V, V0D, V1, V2a, V2b, V3, Shox2, and Hb9, allowed partial identification and/or suggestion of neuron-type-specific roles in spinal circuits and motor control, including locomotion (Lanuza et al., 2004; Gosgnach et al., 2006; Crone et al., 2008; Zhang et al., 2008; Goulding, 2009; Dougherty et al., 2013; Talpalar et al., 2013; Shevtsova et al., 2015; Bikoff et al., 2016; Kiehn, 2016; Caldeira et al., 2017; Ziskind-Conhaim and Hochman, 2017; Dougherty and Ha, 2019; Falgairolle and O’Donovan, 2019). However, so far these studies have mostly focused on lumbar circuits. Genetic identities have only begun to be ascribed to long propriospinal pathways connecting lumbar and cervical spinal segments (Ruder et al., 2016; Flynn et al., 2017). Removal of dLPNs resulted in transient periods of disordered left–right coordination in mice (Ruder et al., 2016); an effect that our previous computational model (Danner et al., 2017) attributed to excitatory diagonally projecting (commissural) V0V LPNs. Similarly, more recent experimental silencing of aLPNs was shown to affect left–right coordination in rats in certain locomotor contexts (Pocratsky et al., 2020); however, the specific populations involved in the effects are unknown. Although the excitatory V3 neurons are unlikely to have significant descending propriospinal projections (Ruder et al., 2016; Flynn et al., 2017), it remains unknown whether there is a subpopulation of V3 neurons that are aLPNs with involvement in fore–hind and/or left–right interlimb coordination. In this study, we specifically focused on the potential role of V3 neurons in long propriospinal interactions between lumbar and cervical circuits controlling interlimb coordination. These neurons are defined by postmitotic expression of the transcription factor single-minded homolog 1 (Sim1). They are excitatory neurons, and the majority of them project to the contralateral side of the spinal cord (Zhang et al., 2008). Results from initial V3 silencing experiments suggested that V3 neurons are involved in the control of locomotion, specifically robustness of the rhythm and left–right coordination (Zhang et al., 2008). Our prior modeling studies suggest that V3 commissural neurons are involved in promoting left–right synchronization during synchronous gaits (Rybak et al., 2013; Rybak et al., 2015; Shevtsova et al., 2015; Danner et al., 2016; Danner et al., 2017) by providing mutual excitation between the extensor half-centers of the left and right lumbar RGs (Danner et al., 2019). Yet, these studies did not provide an explanation for the unbalanced locomotion and variable left–right coordination after inactivation of V3 neurons (Zhang et al., 2008). Also, these studies did not account for heterogeneity of the V3 population that was shown to contain distinct subpopulations with different biophysical properties, laminar distributions, and connectivity (Borowska et al., 2013; Borowska et al., 2015; Blacklaws et al., 2015; Deska-Gauthier et al., 2020), which may underlie different functions. Here, we identified a subset of V3 neurons with cell bodies in the lumbar spinal cord that have direct excitatory projections to the contralateral side of the cervical enlargement. The recruitment of these neurons increased with locomotor speed. Mice with glutamatergic transmission conditionally knocked out in V3 neurons (V3OFF mice) had significantly reduced maximal speeds of locomotion. Moreover, even moving with relatively low and medium speeds, V3OFF mice lost the ability to trot stably, replacing this most typical mouse gait with a lateral-sequence walk. At higher locomotor speeds, V3OFF mice exhibited high step-to-step variability of left–right coordination, which could be a reason for the speed limitation observed in these animals. To determine potential connectivity and functions of local and long propriospinal V3 neurons in spinal locomotor circuits, we updated and extended our previous computational model of spinal circuits consisting of four RGs coupled by multiple CIN and LPN pathways (Danner et al., 2017). Based on the novel anatomical and in vitro electrophysiological data, we included in the model V3 aLPN populations that provided diagonal RG synchronization necessary for trot, in addition to the local V3 CINs involved in left–right synchronization necessary for gallop and bound. The updated model reproduced speed-dependent gait expression in wildtype (WT) mice as well as the experimentally detected changes in the maximal speed, interlimb coordination, and gait expression in V3OFF mice. Taken together, our results suggest different functional roles of the local V3 CINs and V3 aLPNs in interlimb coordination and speed-dependent gait expression during locomotion. Results Experimental studies Lumbar propriospinal V3 interneurons provide ascending excitatory drives to the contralateral cervical locomotor circuits Several subpopulations of V3 neurons have been found and characterized in the mouse spinal cord (Borowska et al., 2013; Chopek et al., 2018; Deska-Gauthier et al., 2020). Until now, however, it has not been shown whether any V3 neurons can serve as LPNs to connect the spinal circuits in lumbar and cervical regions for fore–hindlimb coordination. Previous studies indicated that there might be only a limited number of excitatory dLPNs projecting from cervical to lumbar regions (Ruder et al., 2016; Flynn et al., 2017). Therefore, we primarily focused on studying the aLPNs with projections from lumbar to cervical regions for potential overlap with the V3 neuronal population. To do so, we injected a retrograde tracer, cholera-toxin B (CTB), into the cervical C5 to C8 region of Sim1Cre/+; Rosa26floxstopTdTom (Sim1TdTom) mice (Figure 1A1). After 7 days, we harvested the lumbar spinal cords, and then identified and mapped the tdTomato (tdTom) fluorescent protein and CTB double-positive neurons in lumbar cross sections (Figure 1A2). CTB and tdTom double-positive neurons were indicative of V3 neurons with ascending projections to the cervical region (V3 aLPNs). We found that V3 aLPNs were almost exclusively located contralaterally to the cervical injection sites. Overall, 30% of V3 neurons (n = 3 mice) in L1–L3 were stained by CTB but only 12% of V3 neurons in L4 to L6 were CTB positive. Even though clusters of V3 neurons were distributed across ventral to deep dorsal horn (Zhang et al., 2008; Borowska et al., 2013; Borowska et al., 2015; Blacklaws et al., 2015), the highest density of V3 aLPNs was found in deep dorsal horn, lamina IV to VI, in rostral lumbar segments (Figure 1A3). A relatively small number of these neurons were located in the intermediate and ventral regions, lamina VII and VIII, in caudal lumbar segments (Figure 1A4). Figure 1 with 3 supplements see all Download asset Open asset Ascending V3 long propriospinal neurons identified in the lumbar spinal cord. (A1) Illustration of the experimental strategy to identify lumbar V3 ascending long propriospinal interactions (aLPNs). Cholera-toxin B (CTB; green) injected into the cervical region is picked up at axon terminals and retrogradely transported to cell bodies located in the lumbar region. (A2) Representative image of cross section of the lumbar spinal cord of Sim1tdTom mouse (far left panel). The ‘ipsilateral’ and ‘contralateral’ sides are relative to the injection side in the cervical region. Immunohistochemical staining with different antibodies to illustrate the neurons expressing TdTom (magenta) and CTB (green) (right panels). (A3, A4) Color-coded heat maps of the distribution pattern of V3 aLPNs in the rostral (A3) and caudal (A4) lumbar segments. Scale bars of cell numbers for heat maps are shown below. (B1) Schematic of the experimental strategy to identify lumbar V3 aLPN synapses onto the cervical locomotor region, AAV2/9-hSyn-eGFP is injected into the lumbar region. (B2) Left: a representative image of a cervical hemisection injected with AAV2/9-hSyn-eGFP in the lumbar region. The dashed lines indicate approximate Rexed’s laminae, and white squares indicate the approximate positions of the enlarged images. Middle (lamina VII) and right (lamina IX): representative images of the Vglut2-positive terminals (blue) of lumbar V3 aLPNs (magenta) expressing GFP (green) in cervical locomotor regions. Three random regions (~480 µm × 370 µm) in lamina IX and six regions (~370 µm × 370 µm) in lamina VII region from two cords were sampled. The Vglut2/GFP double-positive puncta (mean of 481.7 ± 47.4 in lamina IX; 143.2 ± 33.4 in lamina VII) and Vglut2/GFP/tdTom triple-positive puncta (12.7 ± 1.7 in lamina IX; 39.5 ± 9.3 in lamina VII) were quantified. (C1) Illustration of an isolated neonatal spinal cord placed in a split bath recording chamber. The suction electrodes for the electroneurogram (ENG) recordings were placed at ventral roots in the lumbar and cervical segments. The chamber is partitioned into two sides with a Vaseline wall (represented by yellow cloud in the figure). The photoactivation (blue light), indicated by the blue transparent circle, is on the ventral side of rostral lumbar, L1–L3, segments. (C2) Averaged traces of rectified ENG recordings at ventral roots of both sides of cervical (C5, C8) and L5 lumbar spinal segments of a P2 Sim1Ai32 mouse, before (Ringer, black), during (kynurenic acid [KA], yellow), and after washout (Washout, gray) of KA application. The time period of photostimulation is indicated by the blue shadowed area. (C3) The changes in integral ENG activity during light stimulation (n = 12) at the period before (dark gray circles), during (yellow circles), and after (light gray circles) KA application. Each circle represents the value of estimated parameter for one trace. Statistics for C3 can be found in Supplementary file 1. To identify lumbar V3 aLPN synapses on neurons within the cervical locomotor region, such as laminae VII/VIII and IX, we injected AAV2/9-hSyn-eGFP in the lumbar spinal cords of Sim1tdTom mice (Figure 1B1) and detected V3 neurons expressing GFP in the intermediate and deep dorsal regions of lumbar cord, as expected (Figure 1—figure supplement 1). In the cervical segments, C5–C8, we observed a broad distribution of GFP-positive axon terminals (Figure 1B2). We detected GFP/TdTom/Vglut2 (vesicular glutamate transporter 2) triple-positive terminals in the ventral interneuron (lamina VII/VIII) and motoneuron (lamina IX) regions (Figure 1B2), indicating that lumbar V3 aLPNs broadly innervate ventral cervical neurons. Interestingly, within the randomly sampled areas in these regions, tdTom-positive puncta are ~2.5% of total Vglut2/GFP puncta in the motoneuron region and ~28% in the interneuron regions (Figure 1B2), indicating that V3 aLPNs may strongly influence cervical RG circuits. To test whether these ascending V3 LPNs in the lumbar segment affect motor output in the cervical region, we employed Sim1Cre/+; Ai32 (Sim1Ai32) mice, which expressed channelrhodopsin 2 (ChR2) specifically in Sim1-positive V3 neurons. Isolated spinal cords from P2–3 Sim1Ai32 mice were placed in a perfusion chamber split into two compartments. This split chamber was constructed over the thoracic T6–T8 segments with petroleum jelly (Vaseline) walls (Figure 1C1). Suction electrodes for electroneurogram (ENG) recordings were placed on the lumbar and cervical ventral roots. We first applied photostimulation on the cervical region, which evoked strong activation at cervical ventral roots, while the lumbar roots were silent (Figure 1—figure supplement 2A1 and A2). These results confirm the lack of descending V3 projections, which is consistent with previous studies (Flynn et al., 2017; Ruder et al., 2016). Then, we applied photostimulation to the V3 neurons in the lumbar region. In this case, the stimulation evoked strong activity in all recorded lumbar and cervical ventral roots (Figure 1—figure supplement 2B1 and B2, see also Figure 1C2 and C3). Then, to test whether this excitation measured in the cervical roots was provided directly by ascending projections of lumbar V3 neurons, we blocked glutamatergic transmission selectively in the lumbar region with 2 mM of kynurenic acid (KA), which completely blocked NMDA and AMPA/kainate receptors (Hägglund et al., 2010). We found that in this case photostimulation of lumbar V3 neurons evoked small or no responses in lumbar ventral roots, while the motor responses in the cervical roots were still present (Figure 1C2 and C3, Figure 1—figure supplement 3). The lumbar responses reappeared after the drug was washed out. Such differential responses of lumbar and cervical roots to lumbar glutamatergic receptor blockade were consistent in all of our testing episodes with lumbar V3 photoactivation (Figure 1C3). Taken together with our anatomical findings, these results demonstrate that lumbar V3 aLPNs directly innervate the cervical spinal cord affecting cervical motor output. V3 aLPNs are active during locomotion To test whether V3 aLPNs were active during locomotion, we injected CTB in the cervical region of young adult (P35–40) Sim1tdTom mice. Seven days post-injection, we subjected the animals to treadmill locomotion at either 15 cm/s or 40 cm/s for an hour. The control group was left undisturbed in the cage for one hour. Then, after one hour of rest in the home cage, which maximizes c-Fos expression in neurons (Dai et al., 2005), we harvested the lumbar spinal cord and used immunohistochemical staining to detect the expression of c-Fos protein in V3 neurons (Figure 2A). We found that the number of triple-labeled (c-Fos/CTB/tdTom) V3 neurons significantly increased after treadmill locomotion at both speeds compared to animals that only rested, but the percentage of these triple-positive V3 aLPNs almost tripled at 40 cm/s compared to 15 cm/s (Figure 2B1–B3 and C). Figure 2 Download asset Open asset Ascending V3 long propriospinal neurons are recruited during locomotion. (A) Representative image of TdTom+ V3 neurons (magenta), cholera-toxin B (CTB) (green), and c-Fos (white) of a half cross section of the spinal cord from a Sim1Cre/+; Rosa26floxstopTdTom (Sim1tdTom) mouse. Enlarged images showing triple-positive cells. Scale bars: 200 μM and 20 μM. (B1–B3) Color-coded heat maps showing the distribution of c-Fos, CTB, and TdTom triple-positive V3 neurons in a cross section of lumbar spinal cord during resting (B1), 15 cm/s (B2), and 40 cm/s (B3) treadmill locomotion (nresting = 3, n15cm/s=3, n40cm/s = 3). (C) Histogram of the corresponding percentage of triple-positive cells in TdTom/CTB double-positive population in L1–L3 segments. *0.01<p<0.05. Statistics for (C) can be found in Supplementary file 1. Taken together, we conclude that there are subsets of lumbo-cervical projecting commissural V3 neurons providing direct excitatory drive to cervical locomotor networks, particularly at medium locomotor speeds. Elimination of V3 neurons in the mouse spinal cord limits locomotor speed To study the contribution of V3 neurons to the control of locomotion and interlimb coordination, we generated Sim1Cre/+; Slc17a6flox/flox (V3OFF) mice, in which the expression of Vglut2 was deleted in V3 neurons (Chopek et al., 2018). We then subjected the control (WT) and V3OFF mice to treadmill locomotion at different speeds. We found that the highest speed that the V3OFF mice could reach was around 35 cm/s (mean = 34.17 ± 6.69 cm/s; Figure 3), and only 3 out of 11 mice could reach 40 cm/s (the maximal speed included for the subsequent analysis). Under the same experimental conditions, WT mice could run with higher speeds, up to 75 ± 7.56 cm/s (Figure 3). This result indicates that V3 neurons are essential for high-speed locomotion in mice. Figure 3 Download asset Open asset Maximum speeds of wildtype (WT) and V3OFF mice on the treadmill. The box-and-whisker plots showing the highest speed of the individual WT mice (n = 7) and V3OFF mice (n = 11) on the treadmill. **p<0.01. Statistics can be found in Supplementary file 1. Elimination of V3 neurons changed speed-dependent interlimb coordination To understand the changes in interlimb coordination at low and medium speeds caused by V3 silencing, we calculated the phase differences (relative phases) between different pairs of limbs in both WT and V3OFF mice: homologous (left–right), between two forelimbs and two hindlimbs; homolateral, between left forelimb and hindlimb; and diagonal, between left hindlimb and right forelimb (Figure 4, Figure 4—figure supplement 1). We then compared the corresponding relative phases and their variability between WT and V3OFF mice on (Figure and at different treadmill speeds (Figure Figure 4—figure supplement 1). Figure with 1 supplement see all Download asset Open asset coordination in wildtype (WT) and V3OFF mice at different speeds. plots of hindlimb (A1) and forelimb (B1) left–right phase phase differences and diagonal phase differences in WT (blue) and V3OFF mice. for the forelimb left–right phase the left hindlimbs are used as the Each blue (WT) and in the indicates the value and robustness of the phase The circle is into The the distribution of phase differences of all at all speeds = = of of phases at individual speeds of V3OFF and WT (blue) mice. for of phase for of the variability parameter of the phase Statistics for and can be found in Supplementary file 1. The left–right hindlimb and forelimb phase differences did not significantly between WT and V3OFF mice across speeds (Figure and B2, Figure 4—figure supplement and were to in both WT and V3OFF mice. The at individual treadmill speeds no significant at 40 the forelimb and hindlimb left–right phase differences of V3OFF mice from of the WT mice (Figure 4—figure supplement and However, the variability of left–right phase differences in V3OFF mice significantly increased compared to WT mice at higher locomotor speeds (Figure and In to the left–right phase the of the (Figure and Figure 4—figure supplement and diagonal phase differences (Figure and Figure 4—figure supplement in V3OFF mice significantly from in WT mice. that these differences were significant across almost all speeds. At the highest speed the between the phase differences was that at speeds but still and the diagonal phase differences did not This indicates at at low to medium speeds, the V3OFF mice were unable to diagonal synchronization and Elimination of V3 neurons changes the gait from trot to at intermediate speeds and gait at higher speeds between the four limbs locomotor gaits 1976; 1980; Bellardita and Kiehn, 2015; et al., 2016). the changes in the and diagonal after functional of V3 neurons gait the and gait expression of V3OFF mice were in a speed-dependent compared to WT Figure and representative phases of all four limbs at different treadmill speeds in WT and V3OFF mice, respectively also Figure Figure with supplements see all Download asset Open asset of wildtype (WT) and V3OFF mice at different speeds. (A) Illustration of limb The of individual limb are by the phase of each is shown with and phase is the between two The is measured from the between the of of two of the same between the limbs was calculated as the between the of of a specific limb and the limb by the period or of the Representative phases of episodes of WT (B1) and V3OFF (B2) mice at low cm/s and and medium cm/s and 40 treadmill speeds. episodes to from the The indicates the gait lateral-sequence light and gray periods of overlap between phases of the The bars the phase of the At low treadmill speeds cm/s and WT mice used trot gait characterized by diagonal synchronization and left–right Figure however, these are speeds and the WT mouse exhibited a high step-to-step variability with that could be as different types of or even gait with only one diagonal between trot At the same low treadmill speeds cm/s and V3OFF mice used a lateral-sequence gait with phases the of a hindlimb is by that of the Figure The lateral-sequence of V3OFF mice was stable and consistent with variability between the different (Figure Figure Figure At higher treadmill speeds cm/s and 40 the pattern of V3OFF mice (Figure while the WT mice exhibited consistent trot with step-to-step variability and synchronization of the diagonal limbs (Figure The variable gait of the V3OFF mice (Figure was characterized by with left–right hindlimb and forelimb phase differences see episodes in Figure including and a high step-to-step variability (Figure 3, Figure To understand how V3 silencing speed-dependent gait expression across we calculated the and of each gait et al., at different speeds in WT and V3OFF mice (Figure Figure supplement 1). The was the percentage of of a certain gait within the total the of a certain gait was the of a of this gait to be by of the same the was calculated as the number of to the gait by the total number of all of gait in two Figure with 1 supplement see all Download asset Open asset of wildtype (WT) (A) and V3OFF mice at different speeds. The of circle indicates the relative of the indicated The of the circle is shown on the far The of the of the indicates the of the corresponding The of the of the indicates the of a The relative of these for each gait and the of between different gaits at different speeds are in Figure These together that gait expression of V3OFF mice was compared to the WT mice in a speed-dependent (Figure Figure supplement 1). WT mice used trot as the gait at all speeds and and increased with speed (Figure Figure supplement In the most and gait in V3OFF mice at low to medium speeds up to 35 cm/s was a lateral-sequence Figure Figure supplement Although V3OFF animals exhibited that could be as trot, the and of trot episodes were The in the gait between trot in WT mice and lateral-sequence or in V3OFF mice was with the changes in the and diagonal phase differences (Figure and WT mice have trot as a gait, which even more and with speed (Figure Figure supplement In in V3OFF mice the and of the lateral-sequence their gait, with speed, while the and of

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.