Crossed activation of thoracic trunk motoneurons by medullary reticulospinal neurons

Abstract

Activation of contralateral muscles by supraspinal neurons, or crossed activation, is critical for bilateral coordination. Studies in mammals have focused on the neural circuits that mediate cross activation of limb muscles, but the neural circuits involved in crossed activation of trunk muscles are still poorly understood. In this study, we characterized functional connections between reticulospinal (RS) neurons in the medial and lateral regions of the medullary reticular formation (medMRF and latMRF) and contralateral trunk motoneurons (MNs) in the thoracic cord (T7 and T10 segments). To do this, we combined electrical microstimulation of the medMRF and latMRF and calcium imaging from single cells in an ex vivo brain stem-spinal cord preparation of neonatal mice. Our findings substantiate two spatially distinct RS pathways to contralateral trunk MNs. Both pathways originate in the latMRF and are midline crossing, one at the level of the spinal cord via excitatory descending commissural interneurons (reticulo-commissural pathway) and the other at the level of the brain stem (crossed RS pathway). Activation of these RS pathways may enable different patterns of bilateral trunk coordination. Possible implications for recovery of trunk function after stroke or spinal cord injury are discussed.

NEW & NOTEWORTHY We identify two spatially distinct reticulospinal pathways for crossed activation of trunk motoneurons. Both pathways cross the midline, one at the level of the brain stem and the other at the level of the spinal cord via excitatory commissural interneurons. Jointly, these pathways provide new opportunities for repair interventions aimed at recovering trunk functions after stroke or spinal cord injury.

INTRODUCTION

Activation of contralateral muscles by supraspinal neurons or crossed activation is critical for bilateral coordination. Studies in mammals have focused on the neural circuits that mediate crossed activation of limb muscles, but the neural circuits involved in crossed activation of trunk muscles are still poorly understood. The aim of the present study is to determine how reticulospinal (RS) neurons cross activate trunk motoneurons (MNs).

RS neurons have strong direct and indirect excitatory connections to ipsilateral trunk MNs (Davidson and Buford 2006; Drew and Rossignol 1990; Galea et al. 2010; Sivertsen et al. 2014; Szokol et al. 2008, 2011; Wilson et al. 1970). RS neurons may also have excitatory connections to contralateral trunk MNs, but these are probably mostly indirect (Femano et al. 1984; Peterson et al. 1979; Sivertsen et al. 2016; Szokol et al. 2008, 2011). The information about RS connections to contralateral trunk MNs is scanty and much comes from studying trunk MNs in the upper lumbar cord rather than the thoracic cord, where most trunk MNs reside (Brink et al. 1979; Holstege et al. 1987; Lipski and Martin-Body 1987; Smith and Hollyday 1983; Takahashi et al. 2010; Tani et al. 1994). On these grounds, and given the critical importance of crossed trunk activation for bilateral trunk coordination and movement, further investigation on RS activation of contralateral trunk MNs is warranted.

In a previous study on functional connections between RS neurons and lumbar MNs, we showed that stimulation of the RS neurons in the medullary reticular formation (MRF) recruits trunk and limb MNs according to a specific spatial organization (Szokol et al. 2008). Stimulation of the medial MRF (medMRF) activates predominantly hindlimb MNs, and stimulation of the lateral MRF (latMRF) activates predominantly trunk MNs. Since thoracic segments do not contain limb MNs, it is unclear whether RS in the latMRF would serve the contralateral trunk MNs in the thoracic cord the same way as the contralateral trunk MNs in the lumbar cord. Therefore, the first goal of the present study is to determine the extent to which RS neurons in the medMRF and the latMRF activate contralateral trunk MNs in the thoracic segments.

In both the cervical and lumbar regions of the spinal cord, RS activation of contralateral MNs is thought to be mediated, at least in part, by commissural interneurons (Bannatyne et al. 2003; Jankowska et al. 2003; Mitchell et al. 2016; Soteropoulos et al. 2013; Szokol et al. 2011). We recently showed that at least one subset of commissural interneurons, those with descending axons (dCINs), is present in the thoracic cord (Kasumacic et al. 2015). Therefore, the second goal of the present study is to determine the extent to which RS neurons in the medMRF and the latMRF activate these thoracic dCINs.

The present work substantiates two spatially distinct RS pathways to contralateral trunk MNs. Both RS pathways originate in the latMRF and involve crossing of the midline. One RS pathway crosses the midline at the level of the spinal cord via excitatory dCINs; the other crosses the midline at level of the brain stem. In principle, bilateral activation of either or both RS pathways could be used to produce various patterns of bilateral trunk coordination, including those observed during locomotor activity (Jean-Xavier and Perreault 2018). Possible implications for recovery of trunk function, following strokes that interrupt corticofugal projections to RS neurons or spinal cord injuries that interrupt RS projections to spinal neurons, are discussed. Some of this work has appeared in abstract form (LaPallo and Perreault 2016).

METHODS

Animals.

Experiments were performed on neonatal [postnatal day 0–4 (P0–P4)] wild-type (ICR/Ha, n = 31, and C57BL/6, n = 11) and genetically modified mice (n = 17; all from Jackson Laboratories). vGlut2-ires-Cre mice [stock no. 016963 (Vong et al. 2011)] were crossed with conditional knock-in mice harboring a floxed copy of either tdTomato [Ai9, stock no. 007909 (Madisen et al. 2010)] or GCaMP3 fluorescent reporter [Ai38, stock no. 014538 (Zariwala et al. 2012)]. Newborn offspring were recognized by their red (tdTomato) or green (GCaMP3) fluorescence. All animal procedures were approved by the Emory University Institutional Animal Care and Use Committee and were performed in accord with the Guide for the Care and Use of Laboratory Animals of the Institute of Laboratory Animal Resources, Commission on Life Sciences, National Research Council (National Academy Press, Washington, DC, 2010).

Brain stem-spinal cord preparations.

The procedures used in this study have been described in detail previously (Szokol and Perreault 2009). Under isoflurane (4%) anesthesia, pups were decerebrated by transecting the brain just rostral to the superior colliculus (precollicular/postmammillary decerebration). Preparations were then placed in a dissection chamber filled with ice-cold oxygenated (95% O₂− 5% CO₂) glycerol-based dissecting solution containing (in mM) 250 glycerol, 2 KCl, 11 d-glucose, 0.15 CaCl₂, 2 MgSO₄, 1.2 NaH₂PO₄, 5 HEPES, and 25 NaHCO₃. After a craniotomy and a laminectomy, the brain stem and the spinal cord were dissected out and the dura was removed.

Retrograde labeling.

Retrograde labeling of thoracic MNs and dCINs was performed by applying calcium green-1-conjugated dextran amine (CGDA; 3,000 MW, catalog no. C6775; Invitrogen) to the cut end of their axons (Glover 1995). For MNs in thoracic segments T7 and T10, the axons were cut close to the corresponding ventral root exit and retrograde transport continued in the dark at room temperature for 2 h. For dCINs in thoracic segment T7, the cut was restricted to the ventral and ventrolateral funiculi between T8/T9 segments and retrograde transport continued for 5 h. In VGLUT2/GCaMP3 mice, dCINs were retrogradely labeled with rhodamine-dextran amine (RDA; 3,000 MW, catalog no. D3308; Molecular Probes).

Calcium imaging.

Preparations containing CGDA- or GCaMP3-labeled neurons were transferred to the recording chamber, where they were positioned with the ventral side up. To image dCINs, an additional oblique transection (45° angle) was performed before transfer (see Fig. 6A). Once in the recording chamber, the preparations were superfused continuously with room temperature 23°C oxygenated artificial cerebrospinal fluid (aCSF) containing (in mM) 128 NaCl, 3 KCl, 11 d-glucose, 2.5 CaCl₂, 1 MgSO₄, 1.2 NaH₂PO₄, 5 HEPES, and 25 NaHCO₃. Single-cell calcium imaging was achieved using the ×40 water-immersion objective (LUMPLFLN, ×40, 0.8 NA; Olympus USA) of an epifluorescence microscope (BX51; Olympus USA) which we equipped with a 100-W halogen lamp driven by a direct current (DC) power supply (PAN35-20A; Kikusui Electronics Corporation, Japan) and excitation and emission filters (type EX: HQ480/40x; type BS: Q505lp; type EM: HQ535/50m). Fluorescence images were captured using a sCMOS camera (PCO.edge; PCO, Canada) mounted on a video zoom adapter set at ×0.5. Image streams (16 bit, 480 frames) were stored at 4 frames/s (binning 2 × 2, gain 1) using the acquisition and image analysis software Metamorph (v7.7; Universal Imaging Corporation, Molecular Devices). To synchronize electrical stimulation and optical recordings, a digital pulse from a digitizer (Digidata 1320A; Molecular Devices) was sent to both the stimulator and the charge-coupled device (CCD) camera. The timing markers for the stimulation and the gating pulse from the CCD camera were recorded by the digitizer at 500 Hz.

Fig. 6.

T7 commissural interneurons with descending axons (dCINs), as an heterogenous population, show no clear predominant activation by lateral medullary reticular formation (latMRF). A: obliquely cut brain stem-spinal cord preparation used to record calcium responses from individual T7 dCINs during latMRF and medial medullary reticular formation (medMRF) stimulation (2 times threshold current T, 5 s at 10 Hz). B: examples of waveform recordings showing the responses to medMRF and latMRF stimulation in 5 ipsilateral (ipsi) T7 dCINs. C: graph showing normalized differences between medMRF- and latMRF-evoked responses. Only 58% of the dCINs responded more strongly to latMRF than medMRF stimulation. Single dCIN responses (green spheres) and population mean response (black spheres) within individual experiments are aligned vertically. Box and whisker plots show grand averages across all experiments, with SE and SD.

CGDA and GCaMP3 have similar Ca²⁺ affinity with a dissociation constant K_d of 540 and 345 nM, respectively (Chen et al. 2013; Haugland 2002) and therefore can be expected to report similar fractions of neurons detected as responsive. To assess photobleaching during our recording sessions, we measured changes in baseline fluorescence at three different levels of low epi-illumination while recording either L2 MNs expressing ChAT-GCaMP3 [n = 3 animals, ChAT-ires-Cre, stock no. 006410 (Rossi et al. 2011)] or L2 MNs labeled with CGDA (n = 3 animals, ChAT/GCaMP3 littermate controls). Baseline fluorescence was measured at the beginning and the end of the 140-s-long recording sessions, and decreases in fluorescence ranged from 0.37% to 0.51% for GCaMP3+ MNs (20 MNs) and from 0.23% to 0.75% for CGDA-labeled MNs (22 MNs; see also Szokol and Perreault 2009). Because decreases in fluorescence never surpassed 1% for both indicators, we consider photobleaching to be negligible under our recording conditions.

Electrical stimulation of the MRF and post hoc histology.

Electrical stimulation was delivered with a monopolar tungsten microelectrode (parylene coated, shaft diameter 0.25 mm, tip diameter 1–2 µm, impedance 0.1 MΩ at 1 kHz) coated with 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI solution saturated in absolute alcohol; Molecular Probes). The electrode was mounted on a hydraulic microdrive (MO-103; Narishige, Japan) and connected to an eight-channel digital stimulator (DS8000; WPI) coupled to an isolation unit (ISOFlex; A.M.P.I., Jerusalem, Israel). The electrode approached the brain stem with an angle of ~90°, and final placement was done under the microscope, using surface illumination. The tip of the electrode was first placed at ~500 µm rostral to the point where the vertebral arteries converge to form the basilar artery and then moved laterally (250–650 µm from the midline) to target either the medial or the lateral region of the MRF (medMRF or latMRF). The medMRF has been previously defined as the region that produces larger responses in hindlimb than axial lumbar MNs, and the latMRF as the region that produces larger responses in axial lumbar than hindlimb MNs (Szokol et al. 2008). The border between medMRF and latMRF is therefore functionally defined, and in P0–P4 animals, it is located around 400 µm from the midline.

For each descent, we photographed the electrode’s entry point on the brain stem. We then lowered the electrode until a calcium response was observed in contralateral MNs or ipsilateral dCINs. During this initial search for effective stimulation sites, we used a current strength of 200 µA. At the first effective stimulation site, we then determined the minimal current needed to evoke a detectable increase in fluorescence (threshold current T). In the present study, T was 64 ± 35 µA, on average. During the experiment, stimulation was delivered at 2T in the form of trains (5-s train duration, 200-µs pulse, at 10 Hz). Stimulation trains were delivered twice (at 10 and 70 s for experiments with CGDA and at 40 and 100 s for experiments with GCaMP3) during each 140-s-long optical recording session. At the end of the experiments, an electrolytic lesion (60–70 µA for 3 s, cathodal followed by anodal DC) was made at the terminal site of the final electrode track. Brain stems were then removed, fixed (4% paraformaldehyde), cryoprotected (30% sucrose), embedded in OCT, frozen, cryostat cut (serial, 50-µm parasagittal sections), and dry mounted on microscope slides. The sections containing the DiI-labeled electrode tracks were photographed using a Microfire CCD camera (Optronics) mounted on a Nikon Eclipse E800 epifluorescence microscope equipped with a motorized stage (Ludl Electronics) and then stained with methylene blue [10 s in 0.3% (wt/vol) solution, catalog no. M9140; Sigma Aldrich], glycerol mounted, coverslipped, and photographed again. Fluorescence and bright-field images were overlaid to confirm position of the electrode within the ventral MRF (Fig. 1A). Post hoc histological confirmation of stimulation sites was obtained in 33/52 animals. In the remaining animals, photographs of the electrode tracks entry points (see above) provided mediolateral positions. Recovered simulation sites were plotted on three reference sections of the brain stem from a P0 mouse [bregma −6.63 mm, −6.99 mm, and −7.23 mm (Paxinos et al. 2007); Fig. 1, B–D]. For the remainder of stimulation sites, we plotted the entry points of electrode tracks (see linear plots below each reference section). To account for the difference in brain stem size between the different age groups, the following conversion factors were applied to the coordinates before plotting: P0 = 1.0, P1 = 0.98, P2 = 0.91, P3 = 0.88, and P4 = 0.85.

Fig. 1.

Stimulation sites are largely confined to the gigantocellularis reticular (Gi) nucleus. A: overlay of fluorescence and transmission light images of a parasagittal brain stem section that contained the track left by the lipophilic fluorescent dye coating of the stimulating electrode. Fluorescence image was taken before methylene blue staining. Arrow points to electrolytic lesion made at the end of experiment. Scale, 500 µm. B–D: spatial distribution of 91 stimulation sites (2 times threshold current T) plotted on 3 standard transverse brain stem sections from a postnatal day 0 mouse (modified from Paxinos et al. 2007). The 3 sections are from different anteroposterior levels as indicated by their distance from the bregma (Br); sections in B–D gather stimulation sites from Br 6.39–6.63 mm, Br 6.75–6.99 mm, and Br 7.11–7.47 mm, respectively. The location was not recovered for an additional 17 sites. For those, we plotted the mediolateral position of the electrode’s entry point (linear plots below each section) obtained from photographs (see methods). All stimulation sites were assigned an open or closed symbol according to whether they were in register with the medial and (medMRF) or lateral regions of the medullary reticular formation (latMRF) as defined in Szokol et al. (2008), whereas the type of symbol (square or circle) indicates whether the recordings were obtained from contralateral (contra) or ipsilateral (ipsi) motoneurons (MNs). 4V, 4th ventricle; 5Sol, trigeminal-solitary transition zone; 7N, facial nucleus; 10N, dorsal motor nucleus of vagus; 12N, hypoglossal nucleus; A1, A1 noradrenaline cell group; AmbC, ambiguus nucleus, compact part, AmbSC, ambiguus nucleus, subcompact part, C1, C1 adrenaline cell group; DPGi, dorsal paragigantocellular nucleus; dC, dorsal cochlear nucleus; Ecu, external cuneate nucleus; GiV, gigantocellular reticular nucleus, ventral part; icp, inferior cerebellar peduncle; IO, inferior olive; IS, inferior salivatory nucleus; Irt, intermediate reticular nucleus; LPGi, lateral paragigantocellular nucleus; LRV4, lateral recess of the 4th ventricle; LVe, lateral vestibular nucleus; LRt, lateral reticular nucleus; LRtPC, lateral reticular nucleus parvicellular; MVe, medial vestibular nucleus; MVeMC, medial vestibular nucleus magnocellularis; MVePC, medial vestibular nucleus parvicellularis; PCRt, parvicellular reticular nucleus; Pr, prepositus nucleus; PrBo, pre-Bötzinger complex; RMg, raphe magnus nucleus; Rpa, raphe pallidus nucleus; Ro, Roller nucleus; Rob, raphe obscurus nucleus; rs, rubrospinal tract; RVLM, rostral ventrolateral medulla; py; pyramidal tract; Sol, nucleus of solitary tract; sol.tr., solitary tract; Sp5C, spinal trigeminal nucleus (caudal); Sp5I, spinal trigeminal nucleus (interpolar); sp5, spinal trigeminal tract; SpVe, spinal vestibular nucleus; VCP, ventral cochlear nucleus (posterior); X, nucleus X.

Spinal lesions.

In about half of the experiments, calcium recordings from MNs or dCINs were obtained both before and after longitudinal lesion along the T7–T10 midline (see Fig. 3; n = 6 animals) or transverse hemisection at C1–C2 (see Fig. 4; n = 23 animals). Longitudinal midline lesions required the recording chamber to be moved under a stereomicroscope (Leica M80), whereas transverse hemisections were performed directly under the epifluorescence microscope. In both cases, the preparation remained pinned down to the Sylgard bottom of the recording chamber as we performed the lesion using fine microdissection scissors. The completeness of each lesion was verified with post hoc histology using methylene blue staining (described above).

Fig. 3.

T7–T10 midline lesion reduces crossed activation of thoracic motoneurons (MNs) by lateral medullary reticular formation (latMRF). A: simplified connectivity diagram of the putative synaptic connections between latMRF reticulospinal (RS) neurons, ipsilateral (ipsi) T7 descending commissural interneurons (dCINs), and contralateral (contra) T7 and T10 MNs tested by a longitudinal lesion of the midline that spanned from T7 to T10 (thick vertical black bar). The crossing connections by ipsi T7 dCINs that are rendered ineffective by the lesion have been grayed out. In this diagram, connections can be mono- or polysynaptic. B: graph showing normalized difference in latMRF-evoked responses before and after the lesion in 37 contra T7 MNs and 20 contra T10 MNs. A positive value indicates an increase in response after the lesion, a negative value indicates a decrease in response, and the zero line indicates no change in response. Single-cell values (gray spheres) and population mean response (black spheres) for each experiment are aligned vertically. T7–T10 midline lesion significantly reduced the latMRF-responses in all experiments. Box and whisker plots show grand averages across all experiments, with SE and SD.

Fig. 4.

Uncrossed and crossed reticulospinal (RS) projections contribute to crossed activation of thoracic motoneurons (MNs). A and B, left: simplified connectivity diagrams of the putative lateral medullary reticular formation (latMRF) RS neurons, ipsilateral (ipsi) T7 descending commissural interneurons (dCINs), and contralateral (contra) T7 and T10 MNs tested by an ipsilateral or contralateral cervical hemisection, respectively. Connections rendered ineffective by the hemisections have been grayed out. Right: graphs showing the changes in responses in ipsi T7 dCINs, contra T7 MNs, and contra T10 MNs after corresponding hemisection (n = 25 animals; 109 ipsi T7 dCINs, 113 contra T7 MNs, 109 contra T10 MNs). Changes in responses are expressed as normalized difference in response before and after the lesion. Hemisection on the same side as the stimulation decreased responses in all 3 neuronal populations (Wilcoxon test; ipsi T7 dCINs, P < 0.04 in 5/5 experiments; contra T7 MNs, P < 0.002 in 4/4 experiments; and contra T10 MNs, P < 0.05 in 4/5 experiments). Hemisection on the side opposite to the stimulation did not significantly affect the responses in ipsi T7 dCINs but decreased the responses in contra T7 and T10 MNs (Wilcoxon test; ipsi T7 dCINs, no significant difference in 3/3 experiments; contra T7 MNs, P < 0.002 in 3/4 experiments; and contra T10 MNs, P < 0.009 in 5/6 experiments). Single dCIN responses (green spheres), single MN response (gray spheres), and population mean response (black spheres) within individual experiments are aligned vertically. Box and whisker plots show grand averages across all experiments, with SE and SD.

Calcium signal analysis.

Using the image analysis Metamorph software (v7.7; Universal Imaging Corporation, Molecular Devices), fluorescence intensity was measured by extracting the changes in fluorescence intensity over time within regions of interest (ROIs) that we positioned manually over the soma of CGDA- or GCaMP3-labeled cells. Fluorescence intensity in the ROIs was averaged across all the pixels. These data were converted into text files and imported to Excel, where they were expressed as waveforms. Analysis was performed using a macro written using Visual Basic for Applications (VBA) programming language. For each ROI, the changes in fluorescence (ΔF) measured are reported as percent changes from an average baseline level F0 [(F − F0)/F0]. Response detection limit was set at 2 SD over the average baseline level, and response magnitude was defined as the area under the waveform during the entire duration of a train of stimulation (see hatched areas in Fig. 2, C and D). A mean response magnitude per cell was obtained by averaging the response to two trains of stimulation.

Fig. 2.

Predominant crossed activation of thoracic motoneurons (MNs) by lateral medullary reticular formation (latMRF). A and B: contralateral (contra) MNs in T7 and T10 segments were preloaded with calcium green-1-conjugated dextran amine (CaGDA; green arrows) 2 h before their responses were mapped to medial medullary reticular formation (medMRF) and latMRF stimulation (2 times threshold current T, 5 s at 10 Hz; black arrows). C and D: waveform recordings showing changes in Ca²⁺ fluorescence (∆F/F, in %) in 8 contra T7 MNs and 8 contra T10 MNs in response to medMRF and latMRF stimulation (vertical gray shading). Response magnitude was defined as the area under the waveform during the entire duration of a train of stimulation (hatched area; see methods). E: graph showing normalized differences between medMRF and latMRF-evoked responses (see text) for 30 contra T7 MNs and 83 contra T10 MNs. Single MN responses (gray spheres) and population mean response (black spheres) within individual experiments are aligned vertically. Box and whisker plots show grand averages across all experiments, with SE and SD.

dCIN segmental projections.

The axonal projection of T7 dCINs to lower thoracic segment T10 was assessed in ICR mice by applying sequentially two distinct color tracers to their cut axons (see Fig. 5). The first tracer, a 1:1 combination of biotin-conjugated dextran amine and fluorescein isothiocyanate, was applied after the axons were cut at the border of T9/T10 segments. The second tracer, rhodamine-dextran amine (RDA), was applied 4 h later after the axons were cut at the border between T8/T9 segments.

Fig. 5.

T7 commissural interneurons with descending axons (dCINs) project different distances, and many have axons reaching the T10 segment. A: schematic describing the dual retrograde labeling procedure used to investigate the extent of T7 dCIN axonal projections (n = 3 animals). Two fluorescent tracers of different colors were applied sequentially to the ventral and ventrolateral funiculi (VF+VLF); the first tracer, rhodamine-dextran amine (RDA) was applied at T9/T10, and a few hours later, the second tracer, biotin-conjugated dextran amine (BDA)-fluorescein isothiocyanate was applied at T8/T9. B: low-magnification confocal image showing a transverse section from the T7 segment in one of the experiments. Downward arrow on left indicates dCINs, and downward arrow and asterisk on right indicate extent of axonal labeling in VF+VLF. B1–B3: high-magnification confocal images showing BDA-labeled dCINs (green), RDA-labeled dCINs (red), and double-labeled dCINs (yellow in merge image with DAPI staining). The green-only T7 dCINs projected to T9 but not T10 (open arrowheads), whereas both the red and yellow T7 dCINs projected to T10 (solid arrowheads). Scale bar, 200 μm. C: T7 dCINs were counted in three 200-µm zones (Z1–Z3) that partitioned the ventral horn equally along its dorsoventral axis (see methods). The relative proportions of T7 dCINs whose axons reached T9 only (green in B) and T7 dCINs whose axons reached T10 (both red and yellow in B) are displayed in a horizontal stacked bar graph for each zone and each experiment.

Cell counts.

Thoracic segments containing retrogradely labeled T7 dCINs were removed, fixed, and cryostat cut (serial, 15-µm transverse sections). Tissue sections were immunoprocessed through exposure to primary antibodies (rabbit anti-TMRDA, 1:200, Invitrogen or rabbit anti-DsRed, 1:1,000, Clontech) and fluorophore-conjugated secondary antibodies (Cy3-conjugated goat anti-rabbit, 1:200, Invitrogen and Cy2-conjugated streptavidin, 1:200, Jackson ImmunoResearch). Sections were mounted with an anti-fade mounting medium containing the nuclear marker 4,6-diamidino-2-phenylindole (DAPI; Vectashield) and coverslipped for imaging. Confocal images (Olympus IX81 inverted microscope, FluoView 1000) were taken from four to six sections taken throughout the T7 segment of each animal using a ×20/0.75 objective. Image stacks were exported to Neurolucida (version 11.0; Microbrightfield, Inc.), and only T7 dCINs that displayed a clearly DAPI-stained nucleus were counted. The distribution of the T7 dCINs within the ventral half of the spinal cord was analyzed by counting T7 dCINs in three adjacent horizontal zones of equal dorsoventral extent (200 µm) drawn over the ventral horn (zones 1–3 in Figs. 5C and 7C). Mean cell counts in each zone are expressed as relative proportions and displayed in horizontal bar graphs (see Figs. 5C and 7C).

Fig. 7.

Glutamatergic commissural interneurons with descending axons (dCINs) are present in the thoracic cord and are predominantly activated by lateral medullary reticular formation (latMRF). A: experimental strategy used to assess the presence of glutamatergic dCINs in the T7 segment of vGluT2/tdTomato mice (n = 4). After retrograde labeling of T7 dCINs with the green fluorescent tracer biotin (BDA)-conjugated dextran amine-fluorescein isothiocyanate, VGLUT2-positive (VGLUT2+) dCINs could be recognized using fluorescence colocalization (green and red). B: low-magnification confocal image shows the spatial distribution of the labeled dCINs (left) and the extent of the axonal labeling in the ventral and ventrolateral funiculi (VF+VLF; right). Scale bar, 200 μm. B1–B3: high-magnification confocal images showing VGLUT2+ cells (red), BDA-labeled dCINs (green), and double-labeled VGLUT2+ dCINs (yellow in merge image with DAPI staining). Fluorescence colocalization (open arrowheads) was used to discriminate VGLUT2+ from VGLUT2− dCINs (green only). Scale bar, 50 μm. C: horizontal bar graph showing the relative proportion of VGLUT2+ dCINs along the dorsoventral axis of the ventral horn in each experiment (see methods). Note that in 2 of the experiments, T7 VGLUT2+ dCINs were found in zones 2 and 3 but not zone 1. Z1–Z3, zones 1–3. D: experimental strategy used to label VGLUT2+ T7 dCINs in VGLUT2/GCaMP3 mice and assess their relative responsiveness to medial (medMRF) and latMRF stimulation (n = 9). RDA, rhodamine-dextran amine. E: graph showing normalized differences between medMRF- and latMRF-evoked responses in VGLUT2+ dCINs. Single dCIN responses (yellow circles) and population mean response (black spheres) within individual experiments are aligned vertically. Box and whisker plots show grand averages across all experiments, with SE and SD.

Statistics.

Specific tests of significance are indicated in results. Significance level α was set at 5%.

RESULTS

Calcium responses evoked by unilateral electrical stimulation of the MRF were recorded in more than 300 contralateral thoracic MNs (contra MNs) and 200 ipsilateral thoracic dCINs (ipsi dCINs) from 52 spinally intact brain stem-spinal cord preparations. In about half of the preparations (28/52), neurons were recorded before and after partial lesion of the spinal cord (thoracic cord midline lesion and ipsilateral or contralateral cervical hemisection). Sample size n refers to number of animals.

Spatial distribution of the MRF stimulation sites.

The locations of most MRF stimulation sites were confirmed histologically using DiI tracks and electrolytic lesions (see methods, Fig. 1A; 91/108 sites). In Fig. 1, B–D, each recovered stimulation site was assigned a specific symbol according to whether their mediolateral position was in register with the medMRF or the latMRF (Szokol et al. 2008, 2011). We found that most of the stimulation sites were located in the gigantocellularis reticular nucleus (Gi; 89%, 81/91) and its ventral part (Giv; 2%, 2/91). A few sites were close to the Gi border but fell in the intermediate reticular nucleus (IRT; 8%, 7/91) or lateral paragigantocellular nucleus (LPGi; 1%, 1/91). This spatial distribution matches well the Gi area, where medullary RS neurons originate in adult and neonatal mice (Szokol et al. 2008; VanderHorst and Ulfhake 2006), indicating that Gi RS neurons are the most likely candidates for the responses reported below.

Predominant activation of contralateral trunk MNs by latMRF.

The MRF of the neonatal mouse is organized so that the medMRF and latMRF activate lumbar trunk and limb MNs with differential predominance (Szokol et al. 2008, 2011). The thoracic cord does not contain any limb MNs, and therefore it is not known how this mediolateral organization of the MRF would serve thoracic trunk MNs. In the present study, we compared the effect of stimulating the medMRF and latMRF on contra thoracic MNs. We recorded from 30 contra MNs in T7 (n = 4) and 83 contra MNs in T10 (n = 6; Fig. 2, A and B) and found that latMRF stimulation produced larger responses than medMRF in most of these MNs (Fig. 2, C and D). To quantify this, we established an estimator of relative effectiveness (Fig. 2E). For each contra MN, we normalized the response evoked from the latMRF and the response evoked from the medMRF by the sum of the two responses (latMRF + medMRF) and then subtracted the resulting values (normalized latMRF − normalized medMRF). The estimator has a value of +1 when a MN is recruited by latMRF only and a value of −1 when a MN is recruited by medMRF only. A value between 0 and +1 indicates a predominant activation by the latMRF, whereas a value between 0 and −1 indicates a predominant activation by medMRF. As illustrated in Fig. 2E, most of the contra MNs recorded in T7 (97%) and T10 (81%) showed predominant activation by latMRF (Wilcoxon test: contra T7 MNs, P < 0.05 in 3/4 experiments; contra T10 MNs, P < 0.01 in 5/6 experiments).

Collectively, these data indicate that the latMRF is a predominant source of RS inputs to contra trunk MNs in the thoracic cord. Together with our previous findings on trunk MNs in the lumbar cord (Szokol et al. 2008, 2011), these data support a prime role for the latMRF in crossed activation of trunk MNs throughout the spinal cord.

Contribution from axons crossing the midline at the level of the spinal cord.

Activation of contralateral trunk MNs by RS inputs must be mediated by axons crossing the midline. However, these axons may cross the midline at many levels of the neuraxis. To investigate the contribution of axons crossing the midline at the level of the spinal cord, we compared latMRF-evoked responses in contra T7 and T10 MNs before and after a longitudinal lesion of the midline that disrupted all commissural connections from T7 to T10 segments (Fig. 3A). Adapting our estimator of relative effectiveness (see Fig. 2E) to compare latMRF responses before and after lesion, we show that most contra T7 MNs and all contra T10 MNs recorded displayed reduced responses after T7–T10 midline lesion (Fig. 3B; n = 6, 37 contra T7 MNs and 20 contra T10 MNs). The average decrease across experiments was around 51% for contra T7 MNs and 90% for contra T10 MNs (Wilcoxon test, P < 0.05 in 6/6 experiments). The larger decrease in response in contra T10 MNs is attributed to the larger number of interrupted crossing axons compared with contra T7 MNs. Altogether, these data suggest an important contribution from axons crossing the midline at level of the spinal cord.

Contribution from thoracic dCINs and RS axons crossing the midline at the level of the brain stem.

Our next step was to investigate potential sources of midline crossing axons for the mediation of latMRF responses in contra trunk MNs. We first investigated the possibility of a contribution from spinal interneurons with midline crossing axons, commonly referred to as commissural interneurons (CINs). CINs have been reported in many species, and in the mouse, four classes of CINs have been defined, based on the spinal trajectory of their crossing axons (Nissen et al. 2005). We know that at least one class of CINs, those with descending axons (dCINs), is present in the thoracic cord (Kasumacic et al. 2015). Therefore, we tested this class specifically.

We tested whether latMRF could activate dCINs, via either uncrossed or crossed latMRF projections. Both projections are illustrated in the diagrams of Fig. 4, with the principal RS projection (uncrossed) shown as a thicker blue line to reflect the fact that ipsilaterally projecting medullary RS neurons in the mouse are about three times more numerous than contralaterally projecting RS neurons (Liang et al. 2011; Szokol et al. 2008; VanderHorst and Ulfhake 2006). We used the oblique cut preparation (shown in Fig. 6A) to record from ipsilateral T7 dCINs (ipsi T7 dCINs). We examined the effect of latMRF stimulation before and after hemisection of the ipsilateral or contralateral cervical cord, which eliminated uncrossed RS projections (Fig. 4A) and crossed RS projections, respectively (Fig. 4B). The effects of these hemisections on latMRF responses in contra T7 and T10 MNs were also examined.

Hemisection of the ipsilateral cervical cord substantially decreased latMRF-evoked responses in ipsi T7 dCINs (Fig. 4A; n = 5, 72 T7 ipsi dCINs) and contra T7 and T10 MNs (n = 9, 42 contra T7 MNs and 41 contra T10 MNs). The magnitude of the decrement in ipsi T7 dCINs was similar to that in contra T7 and T10 MNs (Fig. 4A; grand average of 70% vs. 62% and 79%), compatible with ipsi dCINs contributing to activation of contralateral trunk MNs by uncrossed RS projections. Supplemental tracing experiments indicate a substantial number of T7 dCINs projecting to T10 segments, and therefore many T7 dCINs have the appropriate axonal projection to mediate crossed activation of T10 MNs (Fig. 5; n = 3). The residual response in ipsi T7 dCINs after removal of the uncrossed RS projection may be due to several mechanisms. One possibility would be an activation of the ipsi T7 dCINs by a double-crossed pathway. This pathway may be kept under inhibition by the uncrossed RS projection. Such a scenario could theoretically rely on reciprocal connections between CINs on the two sides of the cord (Birinyi et al. 2003).

In contrast, hemisection of the contralateral cervical cord did not significantly affect latMRF-evoked responses in ipsi T7 dCINs (Fig. 4B; n = 3, 37 ipsilateral dCINs). It did, however, decrease latMRF-evoked responses in contra T7 and T10 MNs (n = 8; 71 contra T7 MNs and 68 contra T10 MNs). The magnitudes of the decrements in contra T7 and T10 MNs were similar to those after ipsilateral hemisections (Fig. 4B; grand average of 66% and 76%, respectively), suggesting that both crossed and uncrossed RS inputs contribute to contra trunk MNs.

Altogether, these experiments reveal two spatially distinct medullary RS projections to contra trunk MNs: an uncrossed RS projection that would activate contra trunk MNs via the recruitment of ipsi thoracic dCINs, and a crossed RS projection that would activate contra trunk MNs either directly and/or through the recruitment of yet unidentified spinal interneurons.

Predominant activation of glutamatergic thoracic dCINs by latMRF.

If ipsi dCINs indeed contribute to the activation of contra trunk MNs by uncrossed latMRF projections, we would expect them to display a similar predominant activation by latMRF as the contra trunk MNs (see Fig. 2). However, when we compared the effects of medMRF and latMRF stimulation on individual ipsilateral T7 dCINs (Fig. 6B; n = 6, 55 ipsi T7 dCINs), we found that only 58% of the ipsi T7 dCINs were predominantly activated by latMRF (Fig. 6C; Wilcoxon test, P < 0.05 in only 1/6 experiments). A confounding factor that might have obscured a clearer predominant activation is the dCINs heterogeneity with respect to transmitter phenotypes. In the lumbar cord, both excitatory and inhibitory dCINs have been reported (Butt and Kiehn 2003; Quinlan and Kiehn 2007).

To assess the presence of excitatory dCINs in thoracic segments, we established a VGLUT2/tdTomato mouse line (Fig. 7A). The expression pattern of tdTomato in these mice is very similar to the expression pattern reported in the lumbar segments in a different VGLUT2 mouse strain (Borgius et al. 2010; Hägglund et al. 2010), with the strongest expression in the most superficial laminae of the dorsal horn, where neuron density is greatest (Fig. 7B). Retrograde labeling of T7 dCINs in VGLUT2/tdTomato mice suggest that VGLUT2-positive (VGLUT2+) dCINs make up a little less than 20% of the dCIN population in T7 (n = 4; Fig. 7C) and they cluster in the most ventral regions (in 2 of the experiments, T7 VGLUT2+ dCINs were found in zones 2 and 3 but not zone 1). Despite their relatively low numbers, we sought to record from VGLUT2+ dCINs and assess their responsiveness to medMRF and latMRF stimulation. Therefore, we established a VGLUT2/GCaMP3 mouse line (Fig. 7D). We found that 75% of the VGLUT2+ T7 dCINs were predominantly activated by latMRF (n = 9; Fig. 7E). Collectively, our data thus show that glutamatergic dCINs are present in the thoracic cord and that, as contra trunk MNs, they are predominantly activated by latMRF. These findings provide additional support to the idea that excitatory thoracic dCINs contribute to crossed activation of trunk MNs by medullary RS neurons.

DISCUSSION

The present study reveals functional connections between RS neurons in the latMRF and contralateral MNs in the thoracic cord of the neonatal mouse, consistent with the idea that the latMRF is a region of prime importance for the control of axial musculature in mammals (Szokol et al. 2008, 2011). We found two spatially distinct RS pathways arising from this region, both involving midline crossing: one at the level of the spinal cord, via excitatory dCINs (reticulo-commissural pathway), and the other at the level of the brain stem (crossed RS pathway). Among the populations of spinal INs that may serve the reticulo-commissural pathway, we identify VGLUT2+ dCINs as prime candidates. The work provides a framework for understanding how the mammalian medullary RS system may achieve bilateral coordination of trunk muscles during postural adjustments and movements. The reticulo-commissural pathway, for instance, is likely relevant during locomotor movement given that interrupting the thoracolumbar commissural system disrupts the RS system’s ability to initiate locomotion (Cowley et al. 2008, 2009; Oueghlani et al. 2018).

Calcium imaging and detection of inhibition.

Somatic calcium signals in spinal neurons detect calcium influx associated with action potentials (Hinckley et al. 2015; Kasumacic et al. 2010; Lev-Tov and O’Donovan 1995). As a proxy for spiking, we used somatic calcium signals to detect RS-evoked activity in contra MNs and ipsi dCINs. However, one limitation is that calcium imaging cannot assess whether inhibitory synaptic inputs have contributed to RS-evoked activity. This is because calcium imaging can only detect hyperpolarization if it decreases neuronal firing (Ali and Kwan 2020; Nelson et al. 2003). In other words, in the absence of spontaneous firing activity, calcium signals cannot readily determine the extent to which inhibitory inputs contributed to activity. However, by using the same strategy as we employed in the present study for VGLUT2+ dCINs (genetically encoded calcium indicators in transgenic animals), it should be possible in future studies to determine whether GABAergic or glycinergic dCINs are preferentially recruited by latMRF.

Candidate RS neurons.

Neuronal activation with electrical microstimulation is the result of two main factors: current strength and proximity to electrode tip (Histed et al. 2009; Ranck 1975). The effective current spread from an electrode tip can be expressed as the square root of the current divided by the excitability constant, $\sqrt{\left(I/K\right)}$ (where K is expressed in µA/mm²). Because ~90% of the stimulation sites were located within the Gi nucleus, we can use the average K value of 859 µA/mm² for RS neurons in the Gi nucleus of rodents (Hentall et al. 1984). With an average current of 128 µA (2T stimulation, 200-µs cathodal pulse), we therefore estimate current spread to be <400 µm. Although this implies a prevailing activation of medullary RS neurons for most of our stimulation sites, for the sites that were located just outside the border of the Gi, activation of other descending neuronal populations must be considered. In rodents, there are only few groups of neurons outside the Gi that project to the lower half of the thoracic cord. These include the respiratory RS neurons of the rostral and caudal ventral respiratory groups (Bévengut et al. 2008; Smith et al. 2013), the sympathoexcitatory RS neurons of the LPGi/rostral ventrolateral medulla region (Kasumacic et al. 2012; Xiang et al. 2014), the trigeminospinal neurons (Ruggiero et al. 1981), and the lateral vestibulospinal tract (LVST) neurons (Kasumacic et al. 2010). At this point, we cannot rule out completely a contribution from respiratory and sympathoexcitatory RS neurons, but we think that an activation of the trigeminospinal and LVST neurons is unlikely because both populations lie well outside the estimated zone of potential current spread (see above). Additionally, we have previously shown that the response pattern that latMRF stimulation elicits in trunk and limb MNs of the lumbar cord is not altered by interruption of the trigeminospinal pathway (Szokol et al. 2008) and is opposite to that evoked by activation of the LVST pathway (Kasumacic et al. 2010; Szokol et al. 2008). However, a more definite conclusion must await an assessment of the axon trajectories of the trigeminal and lateral vestibulospinal neurons in the brain stem of the mouse.

The work of determining the transcription factors that act in combination to define glutamatergic neurons in the MRF (Cepeda-Nieto et al. 2005; Gray 2013) and of establishing how these molecular determinants relate to the various subsets of RS neurons is on its way (Bouvier et al. 2015; Bretzner and Brownstone 2013; Perreault and Giorgi 2019; Stornetta et al. 2006; Wu et al. 2017). The molecular identity of the RS neurons responsible for crossed activation of thoracic MNs is not known, but once available, it should open the door to their selective activation/silencing and provide valuable insights regarding their specific contribution to bilateral trunk coordination during various types of movement.

Genetic identity of VGLUT2+ dCINs.

Of the 10 cardinal classes of molecularly defined spinal interneurons (Gosgnach et al. 2017; Lu et al. 2015; Ziskind-Conhaim and Hochman 2017), 4 are located ventrally in the spinal cord (V0–V3). Based on two of the identification criteria used in the present study, transmitter phenotype and axonal projection, it is likely that the VGLUT2+ dCINs we investigated belong to V0v and/or V3 classes (Moran-Rivard et al. 2001; Pierani et al. 2001; Talpalar et al. 2013; Zhang et al. 2008). V0v and V3 classes are distributed along the entire rostrocaudal extent of the spinal cord (Francius et al. 2013). Both V0v (L1/L2 segments) and V3 (T11–T13 segments) have been shown to contain dCINs (Blacklaws et al. 2015; Griener et al. 2015). The relative proportion of dCINs with respect to the entire CIN population is not known for the V0v class, but it may be substantial based on studies in zebrafish (Björnfors and El Manira 2016; McLean et al. 2007; Satou et al. 2012). With regard to the V3 class, dCINs represent the second-largest subpopulation, reaching as much as 33% of the CIN population (Blacklaws et al. 2015). An important next step in establishing correspondence between the VGLUT2+ dCINs studied presently and V0v or V3 dCINs will be to determine the responsiveness of V0v and V3 dCINs to medullary RS inputs.

Multiple descending pathways for crossed activation of trunk muscles.

Activation of contralateral trunk MNs by medullary RS pathways (this study) and LVST pathway (Kasumacic et al. 2010) suggests considerable multiplicity of descending control. The purpose of this multiplicity is not well understood, but we can speculate that it could be used to differentially regulate specific aspects of contralateral trunk MN activation (e.g., timing, magnitude, etc.), support movement-specific bilateral activation patterns (e.g., whole body movement, locomotion, ventilation), and/or compensate motor deficits following strokes or spinal cord injuries. We did not investigate these aspects of descending control, and resolving this will likely require more information about the extent of the convergence between the RS and LVST pathways both at the motoneuronal level (epaxial and hypaxial MNs; Agalliu et al. 2009; Shirasaki and Pfaff 2002) and at the premotoneuronal level. With regard to the latter, our data indicate that thoracic dCINs are a major premotoneuronal target of the uncrossed latMRF projection and the uncrossed LVST projection (Kasumacic et al. 2015), but it remains to be determined if these projections excite the same or separate subsets of thoracic dCINs. Based on a previous study in the lumbar cord of the cat (Krutki et al. 2003), one may predict that a significant proportion of dCINs will be coexcited by RS and LVST inputs.

Possible role in recovery of trunk motor functions.

Trunk muscles act both as prime mover muscles and as postural muscles (Thorstensson et al. 1985). Their coordination is essential for the initiation and progression of trunk and limb movements both normally and after injury (Moraud et al. 2018; Morris et al. 2012; Nugent and Milner 2017; Rath et al. 2018; Saeys et al. 2012; Schepens et al. 2008; Stamenkovic and Stapley 2016; Van Criekinge et al. 2017). The reticulo-commissural pathway and the crossed RS pathway identified in this study may offer new opportunities for targeted interventions and improvement of these trunk motor functions after strokes and spinal cord injuries. Unilateral injuries that would impair only one of the RS pathways would still allow the spared RS pathway to secure activation of contralateral trunk MNs. Injuries that would interrupt both RS pathways, on the other hand, may still allow activation of contralateral trunk MNs through activation of the thoracic glutamatergic dCINs. In this context, it is noteworthy that CINs have been shown to be very plastic (Chédotal 2014; Xu and Sakiyama-Elbert 2015), spontaneously regenerating and remaking functional synaptic connections even after injury of their axon (Fenrich et al. 2007; Fenrich and Rose 2009). Thus the medullary RS systems described in this report offer new prospects for rehabilitation strategies relevant to trunk motor functions recovery after injury.

GRANTS

This research was supported by the Emory University Integrated Cellular Imaging Microscopy Core, the Craig H. Neilsen Foundation, and National Institute of Neurological Disorders and Stroke Grant R01 NS085387.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

M-C.P. conceived and designed research; B.K.L. and A.G. performed experiments; B.K.L., A.G., and M-C.P. analyzed data; B.K.L., A.G., and M-C.P. interpreted results of experiments; B.K.L., A.G., and M-C.P. prepared figures; B.K.L. and M-C.P. drafted manuscript; A.G. and M-C.P. edited and revised manuscript; B.K.L., A.G., and M-C.P. approved final version of manuscript.