To label dI2 neurons, axons, and terminals in the chick spinal cord, we used intersection between enhancers of two TFs expressed by dI2 – Ngn1 and Foxd3 – via the expression of two recombinases (Cre and FLPo) and double conditional reporters (Figure 1—figure supplement 1A). We have shown previously that the combination of these enhancers reliably labels dI2 neurons (Avraham et al., 2009; Hadas et al., 2014). The recombinases and the double conditional reporter plasmids were delivered via spatially restricted electroporation to the lumbar spinal cord at HH18. At E5, early postmitotic dI2 neurons migrate ventrally from the dorsolateral to the midlateral spinal cord (Figure 1A). As they migrate ventrally, at E6, dI2 neurons assume a midlateral position along the dorsoventral axis (Figure 1B). Subsequently, dI2 neurons migrate medially, and at E17, comparable to postnatal day 4 (P4) of mice, most of them (70.7 and 71.5% at the sciatic and the crural levels, respectively) occupy lamina VII (Figure 1C and F). At all rostrocaudal levels and embryonic stages, dI2 axons cross the floor plate (Figure 1A–D). After crossing, dI2 axons extend rostrally for a few segments in the ventral funiculus (VF) and subsequently turn into the lateral funiculus (LF) (Avraham et al., 2009; Figure 1C and D). Collaterals originating from the crossed VF and LF tracts invade the contralateral spinal cord (Figure 1C and D; white arrows).

Figure 1 with 3 supplements see all

Download asset Open asset

Figure 1—source data 1

Localization of dI2 neurons at the sciatic level.

The X/Y coordinates of dI2, large-diameter dI2, and small-diameter dI2 neurons (Figure 1F).

https://cdn.elifesciences.org/articles/62001/elife-62001-fig1-data1-v2.xlsx

Download elife-62001-fig1-data1-v2.xlsx

Figure 1—source data 2

Localization of excitatory and inhibitory dI2 neurons.

The X/Y coordinates of dI2 neurons expressing either vGlut2 or VIATT (Figure 1I).

https://cdn.elifesciences.org/articles/62001/elife-62001-fig1-data2-v2.xlsx

Download elife-62001-fig1-data2-v2.xlsx

A recent study suggested that dI2 neurons at early stages of development in mice (E9.5–E13.5, comparable to chick E4–8) can be divided into several subclasses based on their genetic signature and degree of maturation (Delile et al., 2019). To assess the diversity of dI2 neurons in the chick, the expression of dI2 TFs in dI2::GFP cells was analyzed at E5 before and during ventral migration, at E6 and E14. The early postmitotic dI2::GFP cells at E5 were a homogenous population defined by Foxd3⁺/Lhx1⁺/Pou4f1⁺/Pax2^- (Figure 1—figure supplement 2A, B, E). dI2 neurons that underwent ventral migration at E5, as well as at E6 and E14, express variable combinations of Lhx1, Pou4f1, and FoxP1/2/4 (Figure 1—figure supplement 2C–E). At E14, approximately 50% of dI2::GFP cells did not express any of the tested TFs (Figure 1—figure supplement 2E), suggesting that the early expression of TFs is required for cell fate acquisition, axon guidance, and target recognition, while their expression is not required after the establishment of the circuitry, as shown for other spinal INs (Bikoff et al., 2016). Interestingly, approximately 12% of ventrally migrating dI2 neurons (from E5 to E14) expressed Pax2 (Figure 1—figure supplement 2D and E). Pax2 is associated with a GABAergic inhibitory phenotype (Cheng et al., 2004), suggesting that a subpopulation of dI2 are inhibitory neurons. The distribution of excitatory and inhibitory dI2 neurons is also apparent at E17. In situ hybridization on cross-sections of the E17 dI2::GFP-labeled lumbar spinal cord using the vGlut2 probe revealed that 73% were vGlut2⁺, while the VIAAT probe, which labels GABAergic and glycinergic inhibitory neurons, measured 27% VIAAT⁺ dI2 neurons (Figure 1G–I; N = 308 neurons from two embryos). Similar percentages of Gad2- and Slc6a5-dI2-expressing cells were also found in mice (Delile et al., 2019). At E13–17 at the caudal lumbar level and at the level of the sciatic plexus, most dI2 neurons are located in the medial aspect of lamina VII. Approximately 91% of dI2 neurons are small-diameter neurons located in the lateral dorsal aspect of lamina VII, and 9% are large-diameter neurons. At the lumbar sciatic plexus level, large-diameter dI2 neurons are located mostly at the ventral aspect of lamina VII (Figure 1F) and at the level of the crural plexus in the ventral and dorsal aspect of lamina VII (Figure 1—figure supplement 3A). The location of the lumbar dI2 neurons, mainly lamina VII, is also apparent in posthatching chicks (P8, Figure 1—figure supplement 3C and D). Importantly, large-diameter dI2 neurons were apparent only at the lumbar level (Figure 1F, Figure 1—figure supplement 3A). The division of large- and small-diameter lumbar dI2 neurons was not reflected in the expression of the tested TFs or in a specific neurotransmitter phenotype; the inhibitory/excitatory ratios were 0.297 ± 0.13 and 0.343 ± 0.02, for large- and small-diameter dI2 neurons, respectively. Hence, dI2 neurons consist of several subpopulations, as has been shown in other spinal INs (Bikoff et al., 2016; Delile et al., 2019; Sweeney et al., 2018).

To study the supraspinal targets of dI2 neurons, axonal and synaptic reporters were expressed in lumbar dI2 neurons (Figure 2A). At stage HH18 (E3), dI2 enhancers were co-electroporated with the double conditional axonal reporter membrane-tethered Cherry and the synaptic reporter SV2-GFP (Figure 1—figure supplement 1A). Expression in the lumbar spinal cord was attained by using thin electrodes positioned near the lumbar segments. At E17, the stage in which the internal granule layer is formed in the chick cerebellum, the axons and synapses of dI2 neurons were studied. dI2 axons cross the spinal cord at the floor plate at the segmental level, ascend to the cerebellum, enter through the superior cerebellar peduncle, and cross back to the ipsilateral side of the cerebellum (Figure 2B). Synaptic boutons were noticeable in the granule layer at the ipsilateral and contralateral sides of the anterior cerebellar lobules (Figure 2C). Synaptic boutons were also present in the central cerebellar nuclei (Figure 2—figure supplement 1A).

Figure 2 with 1 supplement see all

Download asset Open asset

Figure 2—source data 1

Localization of precerebellar neurons and dI2 synapses at the sciatic level.

The X/Y coordinates of precerebellar neurons (Figure 2E) and dI2 synapses (Figure 2G) at the sciatic level.

https://cdn.elifesciences.org/articles/62001/elife-62001-fig2-data1-v2.xlsx

Download elife-62001-fig2-data1-v2.xlsx

The difference in soma size between dorsally and ventrally located dI2 neurons prompted us to test which dI2 neurons project to the cerebellum. dI2 neurons and precerebellar neurons were colabeled by genetic targeting of dI2 at early stages of embryogenesis (HH18), coupled with intracerebellar injection of replication-defective HSV-LacZ at E15 or PRV-Cherry (Figure 2A). Cholera toxin subunit B (CTB) was coinjected with PRV-Cherry to verify primary infection of precerebellar neurons. Spinal neurons retrogradely labeled from the cerebellum consist of double-crossed VSCT neurons and ipsilaterally projecting DSCT neurons. However, dI2 neurons, double labeled by genetic targeting and retrograde labeling from the cerebellum, are all VSCT neurons since dI2 neurons are commissural neurons. The soma distributions of precerebellar neurons, dI2 neurons, and dI2 synapses overlapped at the sciatic level (Figure 2E, G and H) and to a lesser extent at the crural level (Figure 2—figure supplement 1B–D). We found that large-diameter dI2 neurons were mostly colabeled, most of them in the ventral aspect of lamina VII (Figure 2D). Interestingly, many of the cherry⁺ and LacZ⁺ neurons were contacted by dI2 axons (Figure 2F), suggesting that dI2 neurons innervate precerebellar neurons.

Segmental crossing at the lumbar level and recrossing back to the ipsilateral side at the cerebellum are characteristic of the VSCT projection pattern. To measure the proportion of precerebellar neurons in dI2, we used sparse labeling of lumbar dI2 neurons coupled with whole-mount imaging of the E13 spinal cord from the sacral level to the cerebellum utilizing light-sheet microscopy and iDISCO imaging (Belle et al., 2014; Renier et al., 2014; Figure 3; Video 1, Video 2, Video 3, Video 4). 85 neurons were labeled at the lumbar level, 17 of which were large-diameter neurons. The axons of all dI2 neurons cross the midline. Longitudinally projecting axons were apparent at the contralateral VF and LF (Figure 3A–F; Video 1, Video 2, Video 3). In the gray matter, axonal ramifications originating from collateral branching were apparent along the entire extent of the spinal cord (Figure 3A–C; Video 1, Video 2, Video 3). Bifurcation of the longitudinal axons was also apparent. The bifurcating branch elongated rostrally for a few segments and subsequently turned transversely into the spinal cord (Figure 3—figure supplement 1A). We counted the numbers of longitudinally projecting axons at different levels (Figure 3A and D-H): 22 axons at LS1 (Figure 3D), 30 at T1 (Figure 3E), 21 at C9 (Figure 3F), 8 at the rostral brain stem (Figure 3G), and 7 entering the cerebellum via the superior cerebellar peduncle (Figure 3H). The elevated number of axons in T1 levels likely reflects longitudinal bifurcations. Considering that the spinal cord was observed at E13, a relatively early stage of development, it is likely that additional axons enter the cerebellum at later stages of development.

Figure 3 with 1 supplement see all

Download asset Open asset

Video 1

Download asset

Video 2

Download asset

Video 3

Download asset

Video 4

Download asset

To evaluate the complexity of the branching pattern, we reconstructed the axonal projection and branching pattern of two large-diameter dI2 neurons at the lumbar and brachial levels (Figure 3B and C). Numerous collaterals that penetrate the spinal cord along its entire length were evident. Importantly, VSCT dI2 neurons projected to spinal targets at the lumbar, thoracic, and brachial spinal levels and to the brain stem and cerebellum (Figure 3).

To measure the proportion of dI2 in VSCT neurons, we labeled VSCT axons with GFP and dI2 axons with Cherry (for experimental design, see Figure 3—figure supplement 1B and C). The number of axons expressing the reporters at the contralateral superior cerebellar peduncle was scored. 10% of the VSCT axons belonged to dI2 neurons (Figure 3—figure supplement 1D). Thus, the large-diameter dI2 neurons constitute 10% of the VSCT neurons, consistent with the anatomical observation that the VSCT comprises a heterogeneous population of INs (Jankowska and Hammar, 2013; Stecina et al., 2013).

To obtain the connectome of dI2 neurons, we employed enhancer-mediated synaptic labeling of presynaptic neurons coupled with soma labeling of postsynaptic neurons. We used three criteria for assessing synaptic contact: (1) the likelihood of connectivity was examined by spatial overlap of axonal terminals from the presumed presynaptic neurons and the somata of the postsynaptic neurons; (2) synaptic boutons were detected on the somatodendritic membrane of postsynaptic neuron; and (3) colabeling was observed between the presynaptic reporter and synaptotagmin (syn). We used confocal imaging and 3D reconstitution to score overlap (Figure 4—figure supplement 1A, B).

As a proof of concept, we tested the colabeling of dI2::SV2-GFP and syn in dI2 to contralateral pre-MN synapses. Of 144 genetically labeled boutons, 121 (84%) were syn⁺. The syn^- boutons were significantly smaller (Figure 4—figure supplement 1C). We set a volume threshold (0.07 µm³, Figure 4—figure supplement 1C), and small-volume SV2-GFP boutons were not considered synapses in the study. Using these criteria, we mapped the putative pre-dI2 and post-dI2 neurons.

To assess the synaptic input to dI2 neurons, we investigated their synaptic connectivity with the following: (1) dorsal root ganglion (DRG) neurons (Figure 4A). (2) Ipsilateral pre-MNs. General ipsilateral pre-MNs were labeled by injecting a PRV-Cherry virus into the ipsilateral hindlimb musculature (Hadas et al., 2014; Figure 4B). Two genetically defined classes of pre-MNs were examined: dI1i excitatory INs (Figure 4C, Figure 4—figure supplement 2) and the V1 inhibitory pre-MN population (Bikoff et al., 2016; Gosgnach et al., 2006; Figure 4D). (3) Reticulospinal tract neurons (Figure 4E). dI2, DRG, V1, and dI1 neurons were labeled using specific enhancers (Figure 1—figure supplement 1A).

Figure 4 with 3 supplements see all

Download asset Open asset

Figure 4—source data 1

Localizations of pre-dI2 terminals and synapses at the sciatic level.

The X/Y coordinates of dorsal root ganglion (DRG) axons and premotor neurons (pre-MNs) axons (Figure 4A and B). The X/Y coordinates of dI1, V1, and synapses and 5HT terminals at the sciatic level (Figure 4C–E).

https://cdn.elifesciences.org/articles/62001/elife-62001-fig4-data1-v2.xlsx

Download elife-62001-fig4-data1-v2.xlsx

A density profile of the axons of DRG neurons (Figure 4A, Figure 4—figure supplement 3A–D) was aligned with the density plots of the dl2 somata. Overlap between the axonal terminals of DRG neurons was evident (Figure 4A, Figure 4—figure supplement 3C and D). Contact between DRG axons and dI2 neurons was mainly apparent in the dorsal dI2 neurons, while the ventral dI2 neurons received little to no input from DRG neurons (2.8 ± 2.4 vs. 16.9 ± 11.3 contacts per neuron for ventral and dorsal dI2 neurons, respectively; p<1e-5; Figure 4A, Figure 4—figure supplement 3E). In contrast, large- and small-diameter dI2 neurons did not exhibit a significant difference in DRG axon contacts (10.4 ± 14.9 vs. 7.8 ± 5.7 contacts per neuron for large and small dI2 neurons, respectively; p=0.4) (Figure 4—figure supplement 3F).

The density plot of PRV-labeled pre-MNs overlapped with the density plots of the dl2 somata (Figure 4B), and the axonal terminals of pre-MNs were visible on dI2 somata and dendrites (Figure 4B′,B′′), suggesting that pre-MNs contacted dI2 neurons. To solidify the evidence for pre-MN/dI2 connectivity, we used synaptic reporters expressed in genetically identified dI1 and V1 pre-MNs. Excitatory dI1 synapses and inhibitory V1 synapses overlapped with the density plots of the dl2 somata (Figure 4C and D). Synaptic connections, evaluated by boutons found on dI2 dendrites and somata, were apparent from V1 and dI1i (Figure 4B–D, Figure 4—figure supplement 3G).

Serotonergic neurons are the main reticulospinal input to VSCT in cat (Hammar et al., 2004; Hammar and Maxwell, 2002). Serotonergic synapses were concentrated on motor neurons and were not observed on dI2 neurons (Figure 4E, Figure 4—figure supplement 3H). Double labeling of 5-HT and dI2 neurons did not reveal any synaptic input. The lack of synaptic serotonergic input may be related to the difference in species or may suggest that other, non-dI2 VSCT neurons located adjacent to motoneurons are contacted by the reticulospinal neurons. The analysis of synaptic inputs supports the concept that dI2 neurons constitute part of the VSCT. These cells receive input from sensory afferents and inhibitory and excitatory pre-MNs and project to the cerebellum.

Axon collaterals of dI2 invade the gray matter along the entire length of the spinal cord, as revealed by whole-mount staining of spinal cords electroporated with an alkaline phosphatase reporter (dI2::AP) (Figure 5A), cross-sections of dI2 neurons expressing membrane-tethered EGFP (Figure 5B), and light-sheet microscopy analysis (Figure 3—figure supplement 1; Video 1, Video 2, Video 3). The region innervated by dI2 collaterals (arrow in Figure 5B) overlaps with that of the V0 and V1 pre-MNs (Lai et al., 2016; Griener et al., 2015) as well as with that of the contralateral dI2 neurons (Figures 1 and 5B). To assess the potential spinal targets of dI2 neurons, we inspected the degree of overlap between dI2 synapses and dI2 somata (Figure 5C), ipsilateral pre-MNs (Figure 5D), and contralateral pre-MNs (Figure 5E). The alignment revealed an overlap of dI2 synapses with ipsilateral/contralateral pre-MNs and dI2 neurons (Figure 5C–E), supporting their potential connectivity. Labeling of dI2 synapses coupled with labeling of the above neuronal population showed dI2 synaptic boutons on pre-MNs and dI2 neurons at the lumbar level (Figure 5C–E, Figure 5—figure supplement 1A–C).

Figure 5 with 1 supplement see all

Download asset Open asset

Figure 5—source data 1

Localization of dI2 synapses on post-dI2 neurons at the sciatic and brachial levels.

The X/Y coordinates of dI2 synapses, ipsi premotor neurons (pre-MNs), commissural pre-MNs, and contralateral dI2 at the sciatic level; dI2 synapses and ipsi pre-MNs at the brachial level (Figure 5).

https://cdn.elifesciences.org/articles/62001/elife-62001-fig5-data1-v2.xlsx

Download elife-62001-fig5-data1-v2.xlsx

The pattern of dI2 collaterals along the entire rostrocaudal axis (Figure 3A–C, Figure 5A) suggests that dI2 neurons innervate contralateral pre-MNs and dI2 neurons at multiple levels. To test this hypothesis, labeling of lumbar dI2 neurons was coupled with labeling of brachial pre-MNs and dI2 somata by injecting PRV into the wing musculature or electroporating a reporter into brachial dI2 neurons, respectively (Figure 5F, Figure 5—figure supplement 1D). dI2 synapses overlapped with the putative targets, and synaptic boutons originating from lumbar-level dI2 neurons were apparent on dI2 neurons and on the contralateral and ipsilateral pre-MNs of the wings (Figure 5F, Figure 5—figure supplement 1D and E).

Neuronal and synaptic labeling experiments showed that lumbar dI2 neurons innervate the cerebellum, lumbar and brachial pre-MNs, and contralateral dI2 neurons. Hence, dI2 neurons may relay peripheral and intraspinal information to the cerebellum and to the contralateral lumbar and brachial motor control centers.

The synaptic input to dI2 neurons and their putative targets implicates them as relaying information about motor activity to the contralateral spinal cord and the cerebellum. Thus, we hypothesized that manipulation of their neuronal activity may affect the dynamic profile of stepping. To study the physiological role of dI2 neurons, we silenced their activity by expressing the tetanus toxin (TeTX) light chain gene, which blocks synaptic transmission (Yamamoto et al., 2003), in the bilateral lumbar dI2 neurons. EGFP was cotargeted in a 2/1 TeTX/EGFP ratio for post hoc analysis of the efficacy of electroporation (Supplementary file 1). Chicks expressing EGFP in dI2 neurons and chicks that did not undergo electroporation were used as controls. To maximize the number of targeted dI2 neurons, we combined genetic targeting with the Foxd3 enhancer and spatial placement of the electrodes at the dorsal lumbar spinal cord (Figure 1—figure supplement 1A). Embryos were electroporated at HH18. Upon hatching, chicks were trained for targeted overground locomotion. The gait parameters of four controls and 5 TeTX-treated chicks were measured while chicks were walking toward their imprinting trainer along a horizontal track (6–20 walking sessions, 5–8 strides each, per chick).

To test whether silencing of dI2 neurons impairs posthatching development and muscle strength, the chicks were weighed at P8, and their foot grip strength was evaluated on the same day. All chicks were of comparable weight (average – 144.7 ± 12.1 g; Supplementary file 1). As a functional measure of foot grip, we tested the ability of the chicks to maintain balance on a tilted mesh surface. TeTX-manipulated chicks and control chicks maintained balance on the tilted surface up to 63–70°, with no apparent statistically significant differences (Supplementary file 1). Thus, manipulation of dI2 neuronal activity did not impair the development or balance or muscle strength.

Analysis of overground locomotion in the control and TeTX-treated chicks revealed no significant differences in swing velocity or striding pattern. A 180° out-of-phase pattern was found during stepping in all the manipulated and control chicks (Figure 6—figure supplement 1A, Table 1). However, substantial differences were scored in stability parameters: TeTX chicks exhibited whole-body collapses during stepping (Figures 6B and C and 7A), a wide-base gait (Table 2), and variable limb movements (Figure 6A, D and E; Figure 7B and C; Figure 7—figure supplement 1; Table 3).

Figure 6 with 1 supplement see all

Download asset Open asset

Figure 6—source data 1

Analysis of knee height trajectories during the swing phase.

The knee height trajectories during the normalized swing in all the analyzed steps of all chicks (Figure 6D).

https://cdn.elifesciences.org/articles/62001/elife-62001-fig6-data1-v2.xlsx

Download elife-62001-fig6-data1-v2.xlsx

Figure 6—source data 2

Analysis of TMP angles during the swing phase.

The TMP angle trajectories during the normalized swing in all the analyzed steps of all chicks (Figure 6E).

https://cdn.elifesciences.org/articles/62001/elife-62001-fig6-data2-v2.xlsx

Download elife-62001-fig6-data2-v2.xlsx

Figure 6—source data 3

Statistical analysis of knee height trajectories and TMP angles.

Statistical analysis for the data presented in Figure 6.

https://cdn.elifesciences.org/articles/62001/elife-62001-fig6-data3-v2.docx

Download elife-62001-fig6-data3-v2.docx

Figure 7 with 2 supplements see all

Download asset Open asset

Figure 7—source data 1

Analysis of collapses.

The number of collapses of each chick (Figure 7A).

https://cdn.elifesciences.org/articles/62001/elife-62001-fig7-data1-v2.xlsx

Download elife-62001-fig7-data1-v2.xlsx

Using the intersection between genetic drivers and spatially restricted delivery of reporters to define lumbar and brachial neurons, we identified several targets of dI2 lumbar neurons. Lumbar dI2 neurons innervate contralateral lumbar dI2 neurons as well as commissural and noncommissural lumbar pre-MNs. This connectivity may influence the bilateral spinal output circuitry at the lumbar cord (e.g., Bras et al., 1988; Jankowska and Hammar, 2013). Moreover, the ascending axons of lumbar dI2 neurons give off gray matter collaterals innervating contralateral dI2 neurons and commissural and noncommissural pre-MNs throughout the brachial spinal cord (Figure 7D and E). Therefore, lumbar dI2 neurons may also contribute to the inter-enlargement coupling described between the segments of the spinal cord that move the legs and the wings (e.g., Valenzuela et al., 1990; Ruder et al., 2016 for forelimb to hindlimb coupling connectivity in mice).

We demonstrated that lumbar dI2 cells receive sensory innervation, premotor inhibitory and excitatory innervation, and innervation from contralateral lumbar dI2 cells (Figure 7D and E). Thus, lumbar dI2 neurons can provide the cerebellum and the contralateral pre-MNs with proprioceptive information, copies of motor commands delivered from the ipsilateral pre-MNs, and integrated information from contralateral dI2 neurons (Figure 7D and E).

Our wiring-decoding studies are based on the availability of enhancer elements that direct expression in specific spinal INs. The lack of enhancer elements for known pre-MNs, such as V2, precluded their analysis. Future experiments using identified regulatory elements that direct expression in two other pre-MN populations, the dI3 and V0 neurons (Avraham et al., 2010b; Gard et al., 2017), will reveal the extent of premotor information relayed by dI2 neurons.

Our study reveals two anatomically distinct subpopulations of dI2 neurons: precerebellar projection dI2 cells, which also innervate spinal targets along the entire extent of the spinal cord, and propriospinal dI2 cells, which innervate targets within the lumbar level. The laminar and medial/lateral positions of the two dI2 populations are similar, and we did not find subtype-specific expression of the known dI2 TFs. Contrary to our findings, a recent study (Osseward et al., 2021) has shown that tract and propriospinal INs in the mouse spinal cord differ in the localization of their somata along the mediolateral axis and their transcription of TFs. Tract neurons reside in the lateral spinal cord and express group-N TFs, while propriospinal neurons settle at the medial spinal cord and express group-Z TFs. Several reasons may explain the discrepancy between our results and those of Osseward et al. The analysis in mice (Osseward et al., 2021) was applied to numerous cardinal populations of neurons but excluded dI2 neurons. Hence, dI2 may represent an exception. In addition, our study revealed that precerebellar dI2 neurons share two wiring patterns: tract and propriospinal neurons. Applying single-cell RNA sequencing to dI2 neurons will reveal whether short- and long-range targeting in dI2 neurons are characterized by distinct transcriptomes or by a shared N + Z transcriptome.

The unclear genetic origin of physiologically equivalent lumbar VSCT neurons has prevented a better understanding of their role in hindlimb locomotion. Our wiring and neuronal-silencing studies implicated dI2 as a significant contributor to the regularity and stability of locomotion in P8 hatchlings. The kinematic analysis of TeTX-treated hatchlings revealed imbalanced locomotion with occasional collapses, increased stride variability, a wide-base gait, and variable limb movements during stepping.

The mechanisms accounting for impaired stepping following dI2 neuron silencing are still unknown. One of the possible mechanisms is that silencing dI2 neurons perturbs the delivery of peripheral and intrinsic feedback to the cerebellum, leading to unreliable updating of the motor output produced by the locomotor networks, thereby impairing bipedal stepping. Another possible mechanism is based on the similarity of the gait instabilities of TeTX-treated hatchlings to ataxic motor disorders. Mammalian VSCT neurons receive descending input from reticulospinal, rubrospinal, and vestibulospinal pathways (Bras et al., 1988; Jankowska and Hammar, 2013). Neurons from the lateral vestibular nucleus have been reported to innervate extensor motoneurons at the lumbar level, as well as INs residing at medial lamina VII (Murray et al., 2018), the location where dI2 neurons were found to reside in our study. Thus, the vestibulospinal tract may convey input directly to the ipsilateral motor neurons and indirectly to contralateral motor neurons through dI2 neurons that innervate contralateral pre-MNs.

The local spinal connections between dI2 neurons and the contralateral pre-MNs and contralateral dI2 neurons may serve an important component of coordinated limb movements. dI2 synapses were found on both ipsilaterally and contralaterally projecting pre-MNs, both within their segmental level and at the brachial level, which regulates the movement of the wings. Thus, dI2 neurons may affect the motor output of the contralateral and ipsilateral sides of the cord by contacting commissural pre-MNs. Specific targeting of dI2 subpopulations – the precerebellar versus the propriospinal dI2 cells – is necessary to determine the relative contribution of dI2 subpopulations to the impaired stepping phenotype. However, there is no available genetic technique for differentially targeting the two subpopulations. In addition, the fact that the precerebellar dI2 neurons also innervate the lumbar spinal cord precludes the use of retrogradely target-derived neuronal activity modifiers.

In summary, our mapping studies of dI2 neurons and their connectomes, followed by characterization of the effects of their silencing on bipedal stepping, offer new insights into the function of dI2 neurons in vertebrates. We suggest that lumbar dI2 neurons not only relay sensory and intrinsic spinal network information to the cerebellum but also act as active mediators of motor functions at the lumbar segments and at the wing-controlling brachial segments of the spinal cord. Further circuit-deciphering studies of the constituents of subpopulations of dI2 cells, their targets, and their descending inputs are required to extend our understanding of the function of dI2 subpopulations in motor control.

Fertilized white leghorn chicken eggs (Gil-Guy Farm, Israel) were incubated under standard conditions at 38°C. All experiments involving animals followed the designated policies of the Experiments in Animals Ethics Committee and were performed with its approval.

The programs for both 3D reconstruction and the density plot analysis were written in MATLAB. The density plots were generated based on cross-sectional images converted to a standard form. The background was subtracted, and the cells were filtered automatically based on their soma size or using a manual approach. Subsequently, two-dimensional kernel density estimation was obtained using the MATLAB function ‘kde2d.’ Finally, unless indicated otherwise, a contour plot was drawn for density values between 20% and 80% of the estimated density range, with six contour lines.

A DNA solution of 5 mg/mL was injected into the lumen of the neural tube at HH stage 17–18 (E2.75–E3). Electroporation was performed using 3 × 50 ms pulses at 25–30 V, applied across the embryo using a 0.5 mm tungsten wire and a BTX electroporator (ECM 830). Following electroporation, 150–300 μL of antibiotic solution containing 100 unit/mL penicillin in Hanks’ balanced salt solution (Biological Industry, Beit-Haemek) was added on top of the embryos. Embryos were incubated for 3–19 days prior to further treatment or analysis.

Embryos were fixed overnight at 4°C in 4% paraformaldehyde/0.1 M phosphate buffer, washed twice with phosphate buffered saline (PBS), incubated in 30% sucrose/PBS for 24 hr, and embedded in optimal cutting temperature (OCT) compound (Scigen, Grandad, USA). Sections with a thickness of 20 μm were cut on a cryostat. These sections were collected on Superfrost Plus slides and kept at −20°C. For 100 μm sections, spinal cords were isolated from the fixed embryos and subsequently embedded in warm 5% agar (in PBS), and 100 μm sections (E12–E17) were cut with a vibratome. Sections were collected in wells (free-floating technique) and processed for immunolabeling.

The following primary antibodies were used: rabbit polyclonal GFP antibody 1:1000 (Molecular Probes, Eugene, OR, USA), mouse anti-GFP 1:100, goat anti-GFP 1:300 (Abcam), rabbit anti-RFP 1:1000 (Acris), goat anti-ChAT antibody 1:300 (Cemicon, Temecula, CA, USA), mouse anti-synaptotagmin antibody 1:100 (ASV30), mouse anti-Lhx1/5 1:100 (4F2), mouse anti-FoxP4 1:50 (hybridoma bank, University of Iowa, Iowa City, USA), mouse anti-Brn3a 1:50 (Mercury), rabbit anti-Pax2 antibody 1:50 (Abcam), chicken anti-lacZ antibody 1:300 (Abcam), rabbit anti-Calbindin 1:200 (Swant), rabbit anti-VGLUT2 antibody (Synaptic Systems, Göttingen, Germany), goat anti-FoxP2:1000 (Abcam), anti-FoxP1:100 (ABR Synaptic), and rabbit anti-5-HT (Abcam). The following secondary antibodies were used: Alexa Fluor 488/647-conjugated AffiniPure donkey anti-mouse, anti-rabbit, and anti-goat (Jackson) and Rhodamine Red-X-conjugated donkey anti-mouse and anti-rabbit (Jackson). Images were taken under a microscope (Eclipse Ni; Nikon) with a digital camera (Zyla sCMOS; Andor) or captured using the integrated camera of a confocal microscope (FV1000; Olympus).

In situ hybridization was performed as previously described (Avraham et al., 2010a). The following probes were employed: Foxd3, vGlut2, and VIAAT probes were amplified from the cDNA of E6 chick embryos using the following primers. Foxd3: forward TCATCACCATGGCCATCCTG, reverse GCTGGGCTCGGATTTCACGAT. vGlut2: forward GGAAGATGGGAAGCCCATGG, and reverse GAAGTCGGCAATTTGTCCCC. VIAAT: forward CTGAACGTCACCAACGCCATCC, reverse GGGTAGGAGAGCAAGGCTTTG. The T7 RNA polymerase cis-binding sequence was added to the reverse primers.

The standard forms of the spinal cord (for the crural, sciatic, and brachial plexus levels) were computationally divided into polygons for the different laminae (Martin, 1979). The number of neurons or synapses inside each lamina border was scored using their coordinates.

Light-sheet microscopy dI2::mCherry was electroporated into the embryos at HH stage 17–18. Embryos were removed at E13, and the spinal cord and cerebellum were isolated. The tissue was cleared using the iDISCO technique as described (Renier et al., 2014). The mCherry-expressing neurons were stained by application of an anti-RFP antibody followed by Rhodamine Red-X-conjugated donkey secondary antibody. Each staining step included 3 days of incubation with the antibody and subsequent washing for 2 days. Then, the cleared tissue was divided into three segments: a lumbar spinal segment, a brachial spinal segment, and a segment including the brainstem and cerebellum. Each sample was placed in a quartz imaging chamber (LaVision BioTec) and scanned by a LaVision Ultramicroscope II light-sheet microscope operated by ImspectorPro software (LaVision BioTec). An Andor Neo sCMOS camera was used for 16-bit image acquisition. The imaging was performed at 2× magnification with a 0.5–1 µm step size and a green excitation filter (peak – 525 nm/width − 50 nm). Then, for 3D reconstruction and analysis of the samples, the resulting image z-stacks were converted to IMS format using Imaris File Converter (version 9.5). Because of the sample size, several z-stacks were required for full acquisition of each sample; they were stitched together into one z-stack by Imaris Stitcher. Then, the files were uploaded to Imaris (9.6 version) for advanced visualization and analysis. dI2 axons were tracked using the filament tracer feature in semiautomatic mode. After tracking, Imaris was used to generate videos and snapshots describing different features of the analyzed samples. Finally, text, arrows, and other symbols were added using Adobe AfterEffects software.

Validation of SV2-GFP reporter specificity was performed by using Imaris software. High-resolution confocal images of spinal cord sections, with clear SV2-GFP reporter expression and synaptotagmin (syn) immunolabeling, were used to quantify the degree of overlap of GFP⁺ terminals and syn⁺-labeled boutons. Both signals were three-dimensionally reconstructed, and we used the automatic quantification abilities of Imaris, further validated by additional manual counting, to quantify the number of GFP⁺ presynaptic terminals containing at least one syn⁺ bouton. In addition, the volume of GFP⁺ presynaptic terminals was documented to explore a possible dependence between terminal volume and syn⁺ bouton containment.

The treated embryos were fixed with 4% paraformaldehyde–PBS for 24 hr at 4°C and washed twice with PBS for 30 min at 4°C. The fixed embryos were incubated at 65°C in PBS for 8–16 hr to inactivate endogenous AP activity. The treated embryos were washed with 100 mM Tris–Cl (pH 9.5) containing 100 mM NaCl and 50 mM MgCl₂, and the residual placental alkaline phosphatase activity was visualized by incubating the embryos with NBT/BCIP (Roche) in the same buffer at 4°C for 24 hr. After extensively washing the embryos with PBS–5 mM EDTA, the spinal cord was imaged.

From the attenuated PRV Bartha strain, we used two isogenic recombinants that express enhanced GFP (PRV152) and monomeric red fluorescent protein (PRV614). The viruses were harvested from Vero cell cultures at titers of 4 × 10⁸, 7 × 10⁸, and 1 × 10⁹ plaque-forming units (PFU/mL). Viral stocks were stored at −80°C. Injections of 3 μL of PRV152 or PRV614 were made into the thigh, pectoralis, or distal wing musculature of E13 or E14 chick embryos using a Hamilton syringe (Hamilton; Reno, NV, USA) equipped with a 33-gauge needle. The embryos were incubated for 36–40 hr and sacrificed for analysis. For spinocerebellar projecting neuron labeling, we used a replication-defective HSV (TK^-) that contains a lacZ reporter. The virus was injected into the cerebellum of E12–15 embryos in ovo, and the embryos were incubated for another 40–48 hr. Alternatively, CTB conjugated to Alexa Fluor 647 (Thermo Fisher) was injected into the cerebellum of E12–15 embryos together with the virus for visualization of both precerebellular neurons and the upstream neurons.

Muscle strength was evaluated using the measurement of the slope at which the chicks fell from a mesh surface as it was gradually tilted up from the horizontal. This test was repeated for each chick at least three times, and the average falling angle was calculated.

Stride duration was measured as the time from right toe-off/foot-off to the next right toe-off (as a complete stride cycle for the right leg), and the ‘half-cycle’ duration was measured as the time from right toe-off to the time of left toe-off. The following formula was used to calculate the phase: ((LeftToeOff_1 - RightToeOff_1)/(RightToeOff_2 – RightToeOff_1))*360.

The embryos were bilaterally electroporated and then allowed to develop and hatch in a properly humidified and heated incubator. Afterwards, within 32 hr after hatching, the hatchling chicks were imprinted on the trainer. The P8 chicks were filmed in slow motion (240 fps) while freely walking (side and top views). The following parameters were scored: (1) weight; (2) foot grip strength; and (3) kinematic parameters during overground locomotion: (a) swing velocity, (b) swing and stance duration, (c) phase of footfalls, (d) heights of the knee and TMP joints, (e) angles of the TMP and ankle joints, (f) stride width (distance between feet during the double stance phase), and (g) landing angle.

Using semiautomated MATLAB-based tracking software (Hedrick, 2008), several points of interest were encoded. The leg joints as well as the eye and the tail were tracked. The position of these reference points was used for computational analysis using in-house MATLAB code for calculating different basic locomotion parameters (e.g., stick diagrams, velocity, joint trajectory, angles, range, and elevation), step patterns, and degrees of similarity. The landing angle was calculated as the angle between the imaginary line connecting the knee and the TMP joints and the ground, at the end of the swing phase. Dunnett’s test (Dunnett, 1955) was used to perform multiple comparisons of group means following one-way ANOVA. Circular statistics were used for analyses of angular data utilizing Oriana (KCS, version 4).

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

The authors thank Haya Falk for PRV purification; Alona Katzir, Cole Bendor, Mevaseret Avital, Sapir Shevah, Eitan Yisraeli, Ruth Segal, Fedaa Bazan, and Eden Kimchi for technical assistance; Nadav Yayon for assistance with the light sheet microscopy; and Michael O’Donovan for comments on the manuscript. This work was supported by grants to AK from the Israel Science Foundation (grant no. 1400/16), the US–Israel Binational Science Foundation (grant no. 2017/172), and the Avraham and Ida Baruch endowment fund.

All experiments involved with animals were conducted in accordance with the designated Experiments in Animals Ethic Committee policies and under its approval.

© 2021, Haimson et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.