Abstract
The limited information about how descending inputs from the brain and sensory inputs from the periphery use spinal cord interneurons (INs) is a major barrier to understanding how these inputs may contribute to motor functions under normal and pathologic conditions. Commissural interneurons (CINs) are a heterogeneous population of spinal INs that has been implicated in crossed motor responses and bilateral motor coordination (ability to use the right and left side of the body in a coordinated manner) and, therefore, are likely involved in many types of movement (e.g., dynamic posture stabilization, jumping, kicking, walking). In this study, we incorporate mouse genetics, anatomy, electrophysiology, and single-cell calcium imaging to investigate how a subset of CINs, those with descending axons called dCINs, are recruited by descending reticulospinal and segmental sensory signals independently and in combination. We focus on two groups of dCINs set apart by their principal neurotransmitter (glutamate and GABA) and identified as VGluT2+ dCINs and GAD2+ dCINs. We show that VGluT2+ and GAD2+ dCINs are both extensively recruited by reticulospinal and sensory input alone but that VGluT2+ and GAD2+ dCINs integrate these inputs differently. Critically, we find that when recruitment depends on the combined action of reticulospinal and sensory inputs (subthreshold inputs), VGluT2+ dCINs, but not GAD2+ dCINs, are recruited. This difference in the integrative capacity of VGluT2+ and GAD2+ dCINs represents a circuit mechanism that the reticulospinal and segmental sensory systems may avail themselves of to regulate motor behaviors both normally and after injury.
SIGNIFICANCE STATEMENT The way supraspinal and peripheral sensory inputs use spinal cord interneurons is fundamental to defining how motor functions are supported both in health and disease. This study, which focuses on dCINs, a heterogeneous population of spinal interneurons critical for crossed motor responses and bilateral motor coordination, shows that both glutamatergic (excitatory) and GABAergic (inhibitory) dCINs can be recruited by supraspinal (reticulospinal) or peripheral sensory inputs. Additionally, the study demonstrates that in conditions where the recruitment of dCINs depends on the combined action of reticulospinal and sensory inputs, only excitatory dCINs are recruited. The study uncovers a circuit mechanism that the reticulospinal and segmental sensory systems may avail themselves of to regulate motor behaviors both normally and after injury.
- calcium imaging
- excitation-inhibition
- lumbar spinal interneurons
- motor coordination
- neonatal mouse
- reticulospinal
Introduction
A primary but underexplored operational feature of the spinal cord interneurons (INs) involved in motor control is their integrative capacity, which allows them to combine supraspinal and sensory signals (Sherrington, 1906; Baldissera et al., 1981; Jankowska, 1992, 2001, 2016, 2022). Determining how supraspinal and sensory inputs alone and in combination recruit different classes of INs may provide information about the circuit mechanisms available to regulate motor behaviors both normally and after injury.
One major class of INs in motor control are the commissural interneurons (CINs) with midline-crossing axons (Jankowska, 2008; Kiehn et al., 2008; Maxwell and Soteropoulos, 2020). In the lumbar spinal cord of rodents, CINs are divided into various subpopulations according to (1) pattern of intraspinal axonal projection [intrasegmental or short-range CINs and multisegmental ascending CINs, descending CINs (dCINs), or bifurcating or ascending and descending adCINs; Nissen et al., 2005], (2) transmitter phenotype (glutamate, GABA, glycine, or acetylcholine; Restrepo et al., 2009; Zagoraiou et al., 2009), (3) responsiveness to neuromodulators (Zhong et al., 2006a, b; Sharples et al., 2020) and (4) genetic developmental profile (V0, V3, and dI6; Briscoe et al., 1999; Moran-Rivard et al., 2001; Pierani et al., 2001; Lanuza et al., 2004; Andersson et al., 2012; Haque et al., 2018; Deska-Gauthier et al., 2020). This heterogeneity likely provides the CNS with a robust system for the multiplicity of bilateral motor coordination patterns observed in moving mammals (Talpalar et al., 2013; Laflamme and Akay, 2018; Mrachacz-Kersting et al., 2018; Zelenin et al., 2021).
dCINs in the layer (L)2 segment of the spinal cord of rodents are particularly interesting because they have been shown to have direct and indirect excitatory and inhibitory connections to contralateral motoneurons at rest and during fictive motor activity (Butt and Kiehn, 2003; Quinlan and Kiehn, 2007). Additionally, we have found that L2 dCINs can be recruited by supraspinal inputs, including reticulospinal and vestibulospinal, and those not recruited by these descending inputs can be recruited by peripheral sensory inputs (Szokol et al., 2011; Kasumacic et al., 2015; LaPallo et al., 2019). However, we still do not know whether supraspinal or sensory recruitment of dCINs differs across transmitter phenotypes or if sensory activation of dCINs can be facilitated by supraspinal inputs or vice versa. The goal of this study was to establish and compare the recruitment of excitatory and inhibitory dCINs by reticulospinal and segmental sensory afferent inputs, both independently and in combination. For the excitatory phenotype, we focused on the glutamatergic dCINs that express the vesicular glutamatergic transporter 2 (VGluT2). For the inhibitory phenotype, we focused on the GABAergic dCINs that express glutamic acid decarboxylase 2 (GAD2).
Immunohistochemistry to determine the relative proportion and spatial distribution of VGluT2+ and GAD2+ dCINs in the L2 segment reveals that VGluT2+ dCINs are twice as numerous as GAD2+ dCINs and that their area of highest density is located more dorsally than that of GAD2+ dCINs. Electrophysiology combined with single-cell calcium imaging to assess the recruitment of L2 VGluT2+ and GAD2+ dCINs shows that both dCIN subpopulations can be recruited by reticulospinal or segmental sensory input alone. We also show that when recruitment depends on the combined action of reticulospinal and sensory inputs (subthreshold inputs), only VGluT2+ dCINs are recruited. These results expand our understanding of how reticulospinal and segmental sensory input alone or in combination may drive transmitter-specific subpopulations of dCINs to support crossed motor responses and bilateral motor coordination. Some of these findings have been previously reported in abstract format (Giorgi and Perreault, 2018).
Materials and Methods
Mouse lines
The following strains of mice were used: Vglut2-ires-Cre (stock #016963, The Jackson Laboratory; Vong et al., 2011), Gad2-ires-Cre (stock #010802, The Jackson Laboratory; Taniguchi et al., 2011), RCL-GCaMP3 (Ai38; stock #014538, The Jackson Laboratory; Zariwala et al., 2012), RCL-GCaMP6f (Ai95; stock #024105, The Jackson Laboratory; Madisen et al., 2015), RCL-tdTomato (Ai9; stock #007909, The Jackson Laboratory; Madisen et al., 2010). For calcium recording experiments, Vglut2-ires-Cre and Gad2-ires-Cre mice were cross-bred with RCL-GCaMP3 or RCL-GCaMP6f mice. For dCINs cell counts, VGlut2-ires-Cre and Gad2-ires-Cre were cross-bred with RCL-tdTomato and RCL-GCaMP6f mice. All procedures were approved by the Emory University Institutional Animal Care and Use Committee and performed according to the National Institutes of Health Guide for the Care and Use of Laboratory Animals (eighth edition).
Ex vivo brainstem–spinal cord
Details about the surgical procedures for isolating the brainstem–spinal cord from postnatal day (P)0 to P4 animals have been described previously (Szokol and Perreault, 2009). Briefly, pups of either sex were anesthetized with isoflurane 4% (McKesson), decerebrated, eviscerated, gonad assessed, and fixed to a Sylgard-coated polystyrene Petri dish (diameter 15 mm) superfused with ice-cold oxygenated (95% O2, 5% CO2) glycerol-containing dissection solution containing the following (in mm): 250 glycerol, 2 KCl, 11 d-glucose, 0.15 CaCl2, 2 MgSO4, 1.2 NaH2PO4, 5 HEPES, and 25 NaHCO3). After craniotomy, laminectomy, and dura mater removal, the brainstem–spinal cord preparation with dorsal and ventral roots attached was gently dissected out.
Backfills of dCINs
Detailed accounts of retrograde labeling of dCINs and visualization have been previously reported (LaPallo et al., 2019). Briefly, L2 dCINs were backfilled from their cut axons by inserting crystals of fluorescent tetramethyl rhodamine-conjugated dextran (RDA; catalog #D3308, Invitrogen) or a biotin and fluorescein dextran 1:1 mixture (FITC) into a unilateral transverse cut of the ventral and ventrolateral funiculi made at L3–L4. The dissection solution was then replaced with oxygenated, room temperature artificial CSF (aCSF) containing the following (in mm): 128 NaCl, 3 KCl, 11 d-glucose, 2.5 CaCl2, 1 MgSO4, 1.2 NaH2PO4, 5 HEPES, and 25 NaHCO3, and retrograde transport continued in the dark for 3 h. Preparations destined for calcium imaging experiments were further subjected to an oblique cut at the level of the L2 segment and then transferred to the recording chamber. Preparations destined for dCIN cell count experiments were directly processed for immunohistochemistry (see below).
Calcium imaging
The brainstem–spinal cords containing the retrogradely labeled VGluT2+/GAD2+ GCAMP3/GCAMP6f-expressing dCINs were positioned with the ventral side up (Szokol and Perreault, 2009). Imaging was performed using an epifluorescence microscope (LUMPLFLN 40XW, Olympus, 0.8 NA, BX51) equipped with a 100 W halogen lamp driven by a DC power supply (PAN 35-20A, Kikusui Electronics), excitation (EX), beamsplitter (BS) and emission (EM) filters (EX, HQ480/40×; BS, Q505lp; EM, HQ535/50m; EX, D535/25×; BS, Q565lp; EM, 610/55m), and a sCMOS camera (pco.edge) mounted on a video zoom (0.5×, Olympus). A digital transistor-to-transistor logic (TTL) pulse was sent from a digitizer (Axon Digidata 1440, Molecular Devices) via an eight-channel digital stimulator (DS8000, World Precision Instruments) to gate the camera and initiate optical recording. Image streams for response magnitude measurements were acquired at 10 fps for 140 s or 56 s (1400 or 560 frames, 16 bits, binning 2). Image streams for latency measurements were acquired at 100 fps for 4 s (400 frames, 16 bits, binning 4). All recordings were acquired on the computer hard disk using MetaMorph (version 7.8.1.0, Molecular Devices).
Both GCAMP3 and GCAMP6f indicators enable the detection of single action potentials in neuronal somata (Tian et al., 2009; Chen et al., 2013; Dana et al., 2019). Studies in neuronal cultures also show that GCAMP3 and GCAMP6f indicators have similar Ca2+ affinity (Kd of 345 nm compared with 375 nm) but that GCAMP3 has slower kinetics and lower dynamic range (Chen et al., 2013). The lower dynamic range of GCAMP3 could potentially translate into a lower signal-to-noise ratio and lower detection of Ca2+ transients. To assess this in our ex vivo brainstem–spinal cord preparation, we compared the fraction of responsive dCINs reported by each indicator during paired medullary reticular formation (MRF)–dorsal root (DR) stimulation. We found that the fractions of responsive dCINs reported by GCAMP3 and GCAMP6f were similar [GCAMP3, n = 13, median fraction of 44% (17–50%) compared with GCAMP6, n = 13, median fraction of 50% (40– 100%; Wilcoxon, p = 0.264], and therefore the two datasets were pooled. Similar results have been reported for astrocytes in vivo (Ye et al., 2017).
MRF stimulation
Discrete focal electrical stimulation was delivered to the medial MRF (medMRF) and lateral MRF (latMRF) regions of the medullary reticular formation to activate reticulospinal neurons. The MRF stimulations consisted of five 200 µs pulses delivered at 10 Hz (for response magnitude measurements) or a single 200 µs pulse (for response latency measurements). The electrode was positioned with a 45° angle using 4× objective and surface illumination, entering the ventral surface of the brainstem 500 μm rostral to the point of convergence of vertebral arteries and basilar artery and 250 μm (medMRF) or 625 μm (latMRF) lateral to the midline. The entry point of each descent was photographed, and the last stimulation site was marked with an electrolytic lesion (90 µA DC, 3 s duration, cathodal followed by anodal). The initial search for an effective stimulation site was done using a current strength of 200 μA. At the first effective stimulation site, the minimal current needed to recruit at least one VGluT2+ or GAD2+ L2 dCIN was assessed. This recruitment threshold, which we termed 1TCa2+ in this study, was 60 μA (50–100 μA) for the medMRF and 60 μA (40–85 μA) for the latMRF.
Stimulation was delivered using a monopolar tungsten microelectrode (parylene coated, shaft diameter 0.25 mm, tip diameter 1–2 µm, impedance 0.1 MΩ at 1 kHz) which we coated with fluorescent dye 1,1′-dioctadecyl-3,3,3′,3′-tetramethyl-indocarbocyanine perchlorate (DiI; catalog #D3911, Invitrogen) for post hoc histologic processing (see below). The electrode was mounted on a hydraulic microdrive (MO-103, Narishige) and connected to an eight-channel digital stimulator (DS8000, World Precision Instruments) coupled to an isolation unit (ISOFlex, AMPI). The timing marker of the brainstem stimulation was synchronized to the TTL gating pulse of the camera, and both markers were acquired at 500 Hz (Axon Digidata 1440, Molecular Devices).
At the end of the experiments, brainstems were immersed overnight in 4% paraformaldehyde, cryoprotected (30% sucrose), embedded in optimal cutting temperature compound, frozen, cryostat cut (50 µm parasagittal sections), and dry mounted on microscope slides. Histologic confirmation of the MRF stimulation sites was a two-step process. First, sections containing the DiI-labeled electrode tracks were photographed using a MicroFire charge-coupled device camera (Optronics) mounted on a Nikon Eclipse E800 epifluorescence microscope equipped with a motorized stage (Ludl Electronics). Then the same sections were stained with methylene blue (10 s in 0.3% wt/vol solution; catalog #M9140, Sigma-Aldrich), glycerol mounted, coverslipped, and photographed again using bright field microscopy. Overlays of the fluorescence and bright field images were used to confirm stimulation sites (LaPallo et al., 2019). Recovered stimulation sites were plotted on reference sections of the brainstem from a P0 mouse (Paxinos, 2007). To account for the difference in brainstem size between the different age groups, a conversion factor was applied to the coordinates before plotting (P0 = 1.00, P1 = 0.97, P2 = 0.92, P3 = 0.90, P4 = 0.89).
DR stimulation and afferent volley recording
To activate peripheral sensory afferents, electrical stimulation consisting of a single pulse (200 μs duration) was delivered to the L2 DR using a fire-polished borosilicate glass suction electrode filled with the same aCSF as in the recording chamber. After gently pulling the cut end of the L2 DR into the mouth of the electrode, negative pressure was applied until a tight fit between the electrode and the root was obtained. The electrode was connected to the same eight-channel digital stimulator as the MRF stimulating electrode (see above) but coupled to a different isolation unit (ISOFlex, AMPI). The recruitment threshold 1TCa2+ for L2 DR stimulation, which was set the same way as for MRF stimulation, was 12 μA (6–30 μA).
In a subset of experiments, we positioned a second suction electrode at the L2 DR entry zone to record the incoming afferent volley. These recordings were performed both in normal aCSF and in aCSF containing the sodium channel blocker tetrodotoxin (TTX; 1 μm; Tocris Bioscience). The afferent volley signals were recorded with a differential amplifier with 10 K gain, 30 Hz high-pass filtering, 1 kHz low-pass filtering (DPA-2FS, NPI Electronic), sampled at 5 kHz (Axon Digidata 1440, Molecular Devices) and saved for off-line analysis (Clampfit, Molecular Devices).
Pairing MRF and DR stimulation at strength subthreshold for dCIN recruitment
To study the integration of descending and sensory signals in dCINs, we paired MRF and DR stimulation at different intervals ranging from −200 to +200 ms. Both MRF and DR stimulations were set at strength just below the threshold for dCIN recruitment (0.9TCa2+). Paired stimulation was repeated twice at each interval, and the responses were averaged. Recording sessions with unpaired MRF and DR stimulation at the beginning, middle, and end of each experiment ensured that the MRF and DR thresholds remained stable.
In a subset of experiments, we also assessed the effects of mephenesin (Spectrum Chemical) on dCIN responses to paired stimulation. Mephenesin, a drug that depresses early inward action potential currents (Klee and Faber, 1974), only slightly increases the firing threshold of neurons when used at a low concentration of 1 mm. Although this leaves monosynaptic transmission little or not affected, it strongly attenuates or blocks polysynaptic transmission (compounding effect across a chain of synaptically connected neurons). Accordingly, mephenesin has been used to separate the polysynaptic component from the monosynaptic component of evoked responses in many systems and a variety of preparations (Ziskind-Conhaim, 1990; Lev-Tov and Pinco, 1992; Floeter and Lev-Tov, 1993; Vinay et al., 1995; Juvin and Morin, 2005; Szokol et al., 2011; Kasumacic et al., 2015; Hsu et al., 2023). For these experiments, the recording chamber was divided into a brainstem and a spinal cord compartment using a plastic wall placed just above the cervical segment C1–C2. The plastic wall was sealed with petroleum jelly, and the tightness of the seal was verified by adding phenol red to one of the compartments at the beginning of the experiment. Calcium recordings started 20 min after adding mephenesin to one or both compartments (Table 1).
Analyses
Evoked calcium responses
Circular regions of interest (ROIs) were positioned over VGluT2+ or GAD2+ dCINs cell bodies using the image analysis software MetaMorph (version 7.8.2, Molecular Devices). ROI selection was made according to labeling intensity and availability in a selected focus plane (on average, about five VGluT2+ dCINs and three GAD2+ dCINs). Fluorescence intensity in each ROI was extracted by averaging across all pixels within each frame. These fluorescence time series were converted into text files and exported to Microsoft Excel, where they were expressed as waveforms and analyzed using a custom macro written using Visual Basic for Applications. Changes in fluorescence (ΔF) were reported as percentage changes from the average baseline level F0 [(F-F0)/F0]. F0 was calculated from an epoch immediately preceding the onset of the stimulation (5 s epoch for response magnitude measurements and 1 s epoch for response latency measurements). The limit for detecting responses during stimulation was set at 2 SDs above F0 (see Fig. 2B, horizontal dashed line). Response magnitudes were calculated by integrating the changes in fluorescence for a period of 500 ms after the onset of stimulation. The portions of the calcium transients that composed the calcium transient decays were therefore not included. For each cell, a mean response magnitude was obtained by averaging the response to two stimulations. Finally, onset response latencies were defined as the time between the onset of stimulation and the first data point of the response above the 1SD detection limit. See Table 2 for additional information about statistics.
Cell counts
Retrograde labeling of dCIN labeling was performed in P4/P5 offspring of either sex. In VGluT2: tdTomato pups, dCINs were labeled with a mixture (1:1) of biotin and FITC, whereas in GAD2:GCAMP6f pups, dCINs were labeled with RDA (see above). The L2 segment containing the retrogradely labeled dCINs was removed from each animal, fixed, and cryostat cut (serial, 30 µm transverse sections). Tissue sections were immunoprocessed through exposure to primary antibodies and appropriate fluorophore-conjugated secondary antibodies (Table 2). Sections were mounted with an antifade mounting medium containing the nuclear marker 4′, 6′-diamidino-2-phenylindole (DAPI; VECTASHIELD, Vector Laboratories), and coverslipped for imaging. Five to six hemisections taken throughout the L2 segment of each animal were imaged with confocal microscopy at 20× magnification (z-step, 1.0 μm, NA 0.75) using automatic tiling followed by manual stitching (stitches of six tiles/image). Image stacks were exported to Neurolucida software (version 11.0, MBF Bioscience), and only dCINs that displayed a DAPI-stained nucleus were marked and counted. The number of VGluT2+ or GAD2+ dCINs and dCINs was assessed in each animal to obtain the proportion of VGluT2+/GAD2+ dCINs relative to the total of dCINs. Spatial and density distributions of VGluT2+ and GAD2+ populations were displayed as 2D contour plots using a modification of the MATLAB script by T.A. Machado (Bikoff et al., 2016; see Fig. 7D,E).
Study design and statistics
Immunohistochemistry
Results
Half of the L2 dCIN population is VGluT2+
Glutamatergic neurons constitute ∼26% of the CIN population in the lumbar spinal cord of the mouse (Restrepo et al., 2009). How these glutamatergic CINs are distributed across intrasegmental CINs and intersegmental CINs subclasses (ascending, descending, and bifurcating) is unknown. Here, we specifically assessed the proportion of glutamatergic neurons in the intersegmental dCIN subclass of the L2 segment. To do this we retrogradely labeled dCINs in VGluT2-tdTomato animals and used dual fluorescence to identify VGluT2+ dCINs (n = 4; Fig. 1A). We found that VGluT2+ dCINs clustered in the most dorsal region of the L2 dCIN spatial domain (Stokke et al., 2002; Nissen et al., 2005) and constituted 52% (48–53%) of the L2 dCIN population (Fig. 1B,C). This suggests that glutamatergic neurons are twice as numerous in the dCIN subclass than in the general CIN population (see below, Discussion).
Glutamatergic dCINs make up half of the dCIN population in L2. A, Experimental strategy to label glutamatergic dCINs in P4 VGluT2:tdTomato mice. After labeling L2 dCINs with a retrograde green fluorescent tracer, tdTomato-expressing VGluT2+ dCINs were recognized by the colocalization of green and red fluorescence. B, Confocal image showing retrogradely labeled axons in the ventral funiculi on the side of the tracer application (bottom left, upward white arrow) and the spatial distribution of retrogradely labeled dCINs cell bodies on the side opposite the tracer application (dashed box). B1–3, Higher-magnification confocal images showing VGluT2+ INs (red), dCINs (green), and double-labeled VGluT2+ dCINs (merge image, yellow). Small white arrows indicate cells that colocalized red and green fluorescence. C, Pie chart displaying the proportion of VGluT2+expressing cells within the L2 dCIN population (median, interquartile range; n = 4 hemicords).
VGluT2+ dCINs in L2 are recruited extensively by medullary reticulospinal and segmental inputs
Next, we investigated whether L2 VGluT2+ dCINs could be recruited by ipsilateral reticulospinal inputs or ipsilateral segmental sensory afferent inputs. To be able to record calcium responses specifically from L2 VGluT2+ dCINs in these experiments, we retrogradely labeled L2 dCINs in VGluT2:GCAMP3/6f animals (Fig. 2A).
L2 VGluT2+ dCINs are extensively recruited by medullary reticulospinal inputs. A, To discriminate between GCaMP-expressing L2 VGluT2+ dCINs (yellow) from other GCaMP-expressing L2 VGluT2+ spinal cord interneurons (green) during calcium imaging experiments in VGluT2:GCaMP3/6f mice, L2 dCINs were retrogradely labeled with RDA (red). B, Examples of calcium responses recorded in three different VGluT2+ dCINs from the same experiment during 2TCa2+ stimulation of the medMRF and latMRF (5 × 200 µs pulses at 10 Hz, vertical gray lines). MedMRF and latMRF stimulations were repeated twice during each recording, and the evoked responses (overlapping gray waveforms) were averaged (yellow waveform). The pairs of horizontal lines before stimulation indicate the average baseline (solid line) and 2 SD (dashed line) levels (see Methods). C, Doughnut charts displaying incidence of responsive VGluT2+ dCINs (yellow) and nonresponsive VGluT2+ dCINs (black) to 2TCa2+ medMRF and latMRF stimulation. The number of animals (n) is indicated in each chart. D, Normalized differences in magnitudes and onset latencies of VGluT2+ dCIN responses to 2TCa2+ medMRF and 2TCa2+ latMRF stimulation. Differences in responses to medMRF and latMRF stimulation in individual VGluT2+ dCINs were normalized to the sum of these responses. Single VGluT2+ dCIN values (small open circles) and population median (solid black circles) from individual experiments are grouped vertically. Box and whiskers plots display interquartile range IQR1–IQR3 (boxes), median (horizontal black lines in boxes), and maximum and minimum values (whiskers) across all animals.
Having previously defined two sources of reticulospinal inputs based on their location in the MRF (medMRF vs latMRF) and their predominant influence on lateral and medial motoneurons (Szokol et al., 2008), we tested both sources of medullary reticulospinal inputs. On average, stimulation of the medMRF (n = 15) or latMRF (n = 21) at threshold for detectable somatic calcium response (1TCa2+) recruited a little less than half of the recorded VGluT2+ dCINs (medMRF, 47 ± 28%; latMRF, 48 ± 35%; data not shown). Increasing these stimulations to 2TCa2+ enhanced recruitment to 80 ± 26% and 78 ± 34%, respectively (Fig. 2B,C). In 14 experiments (total of 58 cells), medMRF and latMRF were stimulated in the same preparation, allowing us to determine the extent to which individual VGluT2+ dCINs could be activated by both MRF inputs. Although some VGluT2+ dCINs responded exclusively to medMRF or latMRF, ∼70% responded to both. Interestingly, within this population, response magnitudes were larger with latMRF stimulation than with medMRF stimulation (Fig. 2; Wilcoxon, p = 0.03; small effect size, r = 0.24). Considering somatic calcium transients as a proxy for spiking activity, the larger response magnitudes by latMRF stimulation suggest more action potentials and more robust recruitment. Consistent with this idea, a comparison of onset latencies of responses evoked by single-pulse stimulation revealed that VGluT2+ dCINs were recruited by latMRF at significantly shorter onset latencies than medMRF [Fig. 2D; Wilcoxon, p = 0.0012, n = 5 for a total of 17 cells recorded at 100 Hz, median onset latency of 80 ms (50–160 ms) for latMRF vs 230 ms (110–260 ms) for medMRF].
We then sought to investigate sensory recruitment of L2 VGluT2+ dCINs by single pulse stimulation of the L2 DR (n = 23). First, we wanted to relate the current intensity of DR stimulation at and around the recruitment threshold for VGluT2+ dCINs (called 1TCa2+ in this study) to the recruitment of specific groups of sensory afferent fibers. To do this, we stimulated the L2 DR with graded current intensities while we simultaneously recorded both calcium responses in VGluT2+ dCINs (n = 3/6) and sensory afferent volleys in L2 DR (n = 6/6; Fig. 3A). We found that the minimum current intensity at which L2 DR stimulation recruited L2 VGluT2+ dCINs (1TCa2+) was 8 µA (6–9 µA) and near the minimum current intensity at which L2 DR stimulation recruited L2 DR afferent fibers [1TDRrec, 5 µA (5–7 µA); Wilcoxon, p > 0.5, n = 3]. In the example of Fig. 3B, stimulation of the L2 DR at 5 µA, just below the recruitment threshold for L2 VGluT2+ dCINs (0.6TCa2+), evoked a tiny afferent volley in the L2 DR. This small volley had a short onset latency and disappeared on TTX application and therefore was attributed to the activation of the most excitable afferent fibers in L2 (putatively from the low-threshold, fastest conducting cutaneous group A and muscle group I–III with the possible exclusion of muscle group II; Vincent et al., 2017). Increasing stimulation of the L2 DR to 16 µA or twice the recruitment threshold for L2 VGluT2+ dCINs (2TCa2+) increased the magnitude of this volley and evoked an additional, later occurring afferent volley. We attribute this later occurring volley to the activation of the less excitable afferent fibers in L2 (putatively from the high-threshold, slower conducting cutaneous group C and muscle group IV).
L2 VGluT2 dCINs are extensively recruited by segmental sensory inputs. A, Schematic representation of the experimental setup showing how we stimulated the distal segment of L2 DR with graded current strength (single 200 μs pulse) while we recorded calcium responses in L2 VGluT2+ dCINs (n = 3). In these experiments and an additional three experiments, we recorded afferent volleys from the proximal segment of L2 DR to establish correspondence between the recruitment of VGluT2+ dCINs and the recruitment of sensory afferent fibers. Afferent volley recordings were repeated after the application of the Na+ channel blocker TTX (1 mm) to eliminate afferent volley responses and isolate the stimulus artifact. Scale bar, 100 µm. B, A typical example of current values required to recruit VGluT2+ dCINs and sensory afferent fibers displayed with corresponding afferent volleys recordings, demonstrating the relationship among L2 DR current strength of stimulation, recruitment of L2 sensory afferent fibers, and recruitment of L2 VGluT2+ dCINs. At 7 µA, the stimulation, which was just below the recruitment threshold for VGluT2+ dCINs (0.9TCa2+), had reached 1.4 times the recruitment threshold (TDR rec) for the most excitable sensory afferent fibers (first volley component marked A/I–III). At 16 µA, the stimulation had reached 2TCa2+ for VGluT2+ dCINs and 3.2TDR rec for sensory afferent fibers, effectively recruiting another group of fibers (second volley component marked group C/IV). Recordings with and without TTX are overlapped (dotted and solid traces, respectively). C, Examples of calcium responses recorded in three different VGluT2+ dCINs from the same experiment during 1TCa2+ stimulation of the L2 DR (1 × 200 µs pulse). D, Doughnut chart displaying incidence of responsive VGluT2+ dCINs (blue) versus nonresponsive (black) VGluT2+ dCINs to 1 Ca2+ DR stimulation. E, Doughnut charts showing the proportions of medMRF- and latMRF-responsive VGluT2+ dCINs that also responded to 1TCa2+ DR stimulation (mixed yellow and blue pattern) and those that did not (yellow only). Figure 2 provides other details.
Based on these results, we focused on sensory recruitment of L2 VGluT2+ dCINs by afferent fibers from the fastest conducting groups of afferents and limited our stimulation of the L2 DR to 1TCa2+ for the remainder of the study. We found that such stimulation recruited on average 77 ± 25% of the VGluT2+ dCINs (Fig. 3C,D). As L2 DR stimulation was delivered in many of the same experiments where the medMRF or the latMRF was stimulated (n = 14/15 and n = 19/21, respectively), we also quantified sensory recruitment in each MRF-responsive VGluT2+ dCIN population. We found that both medMRF- and latMRF-responsive VGluT2+ dCIN populations were extensively recruited also by 1TCa2+ DR stimulation (medMRF 82 ± 19% or latMRF 77 ± 25%; Fig. 3E).
Altogether, our findings indicate that L2 VGluT2+ dCINs are extensively recruited by medullary reticulospinal inputs alone and segmental sensory afferent inputs alone and that ∼80% are recruited by both sources of inputs.
Pairing subthreshold reticulospinal and sensory stimulations facilitates the recruitment of L2 VGluT2+ dCINs
Convergence of descending and peripheral sensory inputs onto VGluT2+ dCINs, either through monosynaptic or polysynaptic pathways, opens opportunities for interactions such as summation. Summation could allow cooperation and be beneficial in circumstances where reticulospinal neurons or sensory afferents are less active (e.g., during specific phases of movement) or rendered incapable of recruiting VGluT2+ dCINs on their own (e.g., after spinal cord injury). To investigate whether reticulospinal and segmental sensory inputs could cooperate to recruit VGluT2+ dCINs, we paired MRF and DR stimulations after adjusting their respective stimulation strength just below the threshold for recruitment of VGluT2+ dCINs, that is, 0.9TCa2+ [Fig. 4A,B; medMRF (0.9TCa2+) plus DR (0.9TCa2+), n = 22 and latMRF (0.9TCa2+) plus DR (0.9TCa2+), n = 28]. Based on the extensive recruitment of VGluT2+ dCINs observed during 1TCa2+ DR stimulation, we surmised that pairing subthreshold DR stimulation with subthreshold MRF stimulation would still lead to extensive recruitment of VGluT2+ dCINs. We initially set the pairing interval between stimulations to zero milliseconds (synchronous). Surprisingly, medMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing either did not recruit any, or only a few, VGluT2+ dCINs per experiment (n = 21; Fig. 4C,D), resulting in a median fraction of VGluT2+ dCINs recruited of only 33% (0−60%). By contrast, latMRF (0.9TCa2+) plus DR(0.9TCa2+) pairing often recruited all VGluT2+ dCINs, resulting in a median fraction of VGluT2+ dCINs recruited of 100% (50–100%; n = 29; Fig. 4C,D). The greater recruitment during latMRF plus DR pairing was statistically significant (Student's t test, p = 0.0003). We observed similar greater recruitment by latMRF (0.9TCa2+) plus DR (0.9TCa2+) at the −5 ms pairing interval [Student's t test, p = 0.04, n = 3 for medMRF (0.9TCa2+) plus DR (0.9TCa2+); n = 5 for latMRF (0.9TCa2+) plus DR (0.9TCa2+)] but not at other pairing intervals [−200 ms to +200 ms, all p values > 0.05, n = 3–6 experiments/interval for medMRF (0.9TCa2+) plus DR (0.9TCa2+); n = 3–8 experiments/interval for latMRF (0.9TCa2+) plus DR (0.9TCa2+)]. To further document the difference between medMRF (0.9TCa2+) plus DR(0.9TCa2+) and latMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing, we also compared the magnitudes of the responses in individual recruited VGluT2+ dCINs as a function of the interval between the stimulations (Fig. 5A,B). We tested intervals from −200 ms to +200 ms and found that calcium responses in recruited VGluT2+ dCINs were larger during latMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing than during medMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing (Fig. 5C; Wilcoxon, p = 0.004, all intervals combined, medMRF plus DR pairing, n = 10; latMRF plus DR pairing, n = 16). Altogether, the data suggest that latMRF plus DR pairing not only recruits more VGluT2+ dCINs than medMRF plus DR pairing but also produces larger calcium responses, and therefore presumably more action potentials, in the VGluT2+ dCINs recruited.
Pairing of subthreshold MRF and DR stimulations demonstrates that medullary reticulospinal and sensory inputs can cooperate to recruit VGluT2+ dCINs. A, Calcium recording arrangement used for pairing of subthreshold (0.9TCa2+) MRF and DR stimulations displayed together with a paradigm for such pairing with hypothesized effects on cell recruitment and calcium response. Right, The gray horizontal dotted line indicates the putative threshold for action potential, and the shaded green area defines the measured calcium response magnitude. B, Control recordings showing the expected lack of calcium response in the three VGluT2+ dCINs during subthreshold (0.9TCa2+) stimulation of medMRF (5 pulses), latMRF (5 pulses), and DR (1 pulse). C, Recordings from the same three VGluT2+ dCINs during paired stimulations. D, Graph comparing the fraction of recruited VGluT2+ dCINs per experiment (individual circles) during medMRF (0.9TCa2+) plus DR (0.9TCa2+) and latMRF (0.9TCa2+) plus DR (0.9TCa2+) stimulations at pairing interval of 0 ms. The fraction of VGluT2+ dCINs recruited was much lower during medMRF (0.9TCa2+) plus DR (0.9TCa2+) than during latMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing.
Pairing subthreshold latMRF and DR stimulations produces larger responses in VGluT2+ dCINs than pairing subthreshold medMRF and DR stimulations. A, B, Graphs displaying response magnitudes in each VGluT2+ dCINs recruited as a function of pairing interval for medMRF (0.9TCa2+) plus DR (0.9TCa2+) and latMRF (0.9TCa2+) plus DR (0.9TCa2+). C, Averaged response magnitude across all animals at each pairing interval for medMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing (light green) and latMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing (dark green). Error bars indicate SEs.
The summation of reticulospinal and sensory signals leading to the recruitment of VGluT2+ dCINs during latMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing likely occurs at the level of the spinal cord. However, an involvement of the brainstem through the engagement of the reticulospinal network by spinoreticular pathways is also possible (Fig. 6A; Rossi and Brodal, 1957; Shimamura and Livingston, 1963; Eccles et al., 1975; Menétrey et al., 1980; Chaouch et al., 1983; Shimamura and Kogure, 1983; Drew et al., 1996; Antonino-Green et al., 2002). To evaluate these possibilities, we partitioned the recording chamber into a brainstem and a spinal cord compartment and repeated latMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing before and after adding 1 mm mephenesin, a drug that reduces transmission along polysynaptic pathways more than monosynaptic pathways (see above, Materials and Methods). To block transmission along the spinoreticular pathways, mephenesin was first applied to the brainstem compartment (Fig. 6A1). This did not significantly alter signal summation and recruitment of VGluT2+ dCINs during subthreshold latMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing (Fig. 6B; Kruskal–Wallis, Steel post hoc comparison against aCSF control, p = 0.09, n = 17 cells in 6 animals). However, when mephenesin was also added to the spinal cord compartment (Fig. 6A2), the recruitment of VGluT2+ dCINs was markedly reduced (Fig. 6B; Kruskal–Wallis, Steel post hoc comparison against aCSF control, p = 0.0002, n = 12 cells in 4 animals). Successful washing out of the mephenesin from the spinal cord compartment (Fig. 6A3) led to full recovery of the responses in VGluT2+ dCINs (Fig. 6B; n = 7 cells in 1 animal). Altogether, the data support a substantial contribution from the spinal cord but not the brainstem, and because the response that remained after applying mephenesin to the spinal cord (presumed to be monosynaptic) was small, also a predominant contribution from polysynaptic connections.
Effect of mephenesin on VGluT2+ dCIN responses during subthreshold latMRF plus DR pairing. A, Left, Schematic of putative connections in the brainstem and spinal cord that may be activated during paired stimulation. Smallest size interneurons indicate putative polysynapticity and may be excitatory (open-end terminal) or inhibitory (flat-end terminal). Right, Experimental strategy for testing the contribution of polysynaptic connections with 1 mm mephenesin applied to the compartmentalized brainstem spinal cord preparation, first to the brainstem compartment (1) to assess any potential contribution from ascending spinoreticular polysynaptic pathways and then to the spinal cord compartment (2) to assess the contribution of sensory and reticulospinal polysynaptic pathways. Mephenesin washout from the spinal cord compartment (3). B, Graph displays the normalized response magnitudes of individual VGluT2+ dCINs in each of the three conditions depicted in A (normalized to control responses before mephenesin). Values from individual VGluT2+ dCINs are shown as small open circles, and population response from individual experiments are shown as solid black circles. The boxes and whiskers plots summarize the data across all experiments for each condition; ***p ≤ 0.001; Kruskal–Wallis test followed by Steel post hoc comparison against control.
In summary, the results obtained by pairing subthreshold MRF-DR stimulations suggest that descending reticulospinal and segmental sensory inputs can effectively cooperate to recruit VGluT2+ dCINs in conditions where each source of input would be either less active or unable to recruit VGluT2+ dCINs on its own. The cooperation was found to be more effective between latMRF and sensory inputs than between medMRF and sensory inputs, providing further support to the idea of more direct and robust connections between latMRF reticulospinal neurons and VGluT2+ dCINs. Finally, split-bath experiments indicate that spinoreticular pathways contribute little, if at all, to the summation of reticulospinal and sensory signals that underlies the recruitment of VGluT2+ dCINs during subthreshold latMRF-DR pairing. This enables us to postulate the existence of shared spinal neurons between the reticulospinal and segmental sensory inputs. The identity of these neurons is currently unknown, but spinal interneurons presynaptic to VGluT2+ dCINs, and to a much lesser extent, the VGluT2+ dCINs themselves, are good candidates.
GAD2+ dCINs are also recruited extensively by reticulospinal and sensory inputs, but unlike VGluT2 dCINs, their recruitment is not facilitated by pairing subthreshold reticulospinal and sensory stimulations
To determine whether the principles governing the recruitment of L2 VGluT2+ dCINs by reticulospinal and segmental sensory inputs applied to other L2 dCIN populations, we set out to investigate GAD2 (or GAD65) expressing GABAergic dCINs. We first assessed the relative proportion of GAD2+ dCINs within the L2 dCIN population using retrograde labeling and dual fluorescence (n = 3; Fig. 7A). We found that GAD2+ dCINs were less numerous than VGluT2+ dCINs and represented about one-fourth of the L2 dCIN population [26% (25–27%); Fig. 7B,C]. GAD2+ dCINs also clustered more ventrally than VGluT2+ dCINs (Fig. 7D,E).
GABAergic dCINs make up a quarter of the dCIN population in L2. A, Experimental strategy to label GABAergic dCINs in P4 Gad2:GCAMP6f mice. After labeling of L2 dCINs with a retrograde red fluorescent tracer, GDP-expressing GAD2 positive (GAD2+) dCINs were recognized by colocalization of red and green fluorescence. B, Confocal image showing retrogradely labeled axons in the ventral funiculi on the opposite side (bottom left, white arrow) and spatial distribution of the retrogradely labeled dCINs cell bodies on the side opposite to tracer application (dashed box). B1–3, Higher-magnification confocal images showing GAD2+ INs (green), dCINs (red), and double-labeled GAD2+ dCINs (merge image, yellow). Small white arrows indicate cells that colocalized green and red fluorescence. C, Pie chart displaying the proportion of GAD2+-expressing cells within the L2 dCIN population (median, interquartile range; n = 3 hemicords). D, Graph comparing the spatial distribution of VGluT2+ dCINs (left) and GAD2+ dCINs (middle) in the L2 segment. The mediolateral (ML) and dorsoventral (DV) coordinates of the cell bodies were computed from six sections per animal and plotted on the template of the L2 segment of the Allen Annotated Reference P4 Spinal Cord Atlas (http://mousespinal.brain-map.org/). The median ML and DV coordinates and interquartile ranges were ML 225 mm (217–244 mm) and DV −108 mm (−118 to −93 mm) for VGluT2+ dCINs, and ML 225 mm (224–225 mm) and DV −143 mm (−160 to −131 mm) for GAD2+ dCINs. Right, Overlapping density contour plots. E, Density profile distributions displayed as heat maps show that despite the extensive overlap the area of highest density lies more dorsally for VGluT2+ dCINs than for GAD2+ dCINs (p = 0.0017, 2D Kolmogorov–Smirnov test). Orange- to red-filled contours have the highest density of cells, and blue-filled contours have the lowest density. To facilitate comparison, the two maps are displayed using the same scale.
Next, using calcium imaging from retrogradely labeled GCAMP-expressing GAD2+ dCINs (Fig. 8A), we assessed whether individual L2 GAD2+ dCINs could be recruited either by reticulospinal or segmental sensory afferent inputs. Stimulation of the medMRF or latMRF at 1TCa2+ recruited about two-thirds of the recorded GAD2+ dCINs (n = 5 medMRF, 63 ± 34%; n = 5 latMRF, 70 ± 41%; data not shown) and increasing the stimulation to 2TCa2+ enhanced recruitment to 90 ± 22% and 93 ± 15%, respectively (Fig. 8B,D). Stimulation of the L2 DR at 1TCa2+ recruited 89 ± 24% of the GAD2+ dCINs (n = 9; Fig. 8C,E). Concerning sensory recruitment of MRF-responsive GAD2+ dCINs subpopulations, we found that both medMRF- and latMRF-responsive GAD2+ dCIN populations were extensively recruited by 1TCa2+ DR stimulation (medMRF 75 ± 32% or latMRF 100 ± 0%, respectively; Fig. 8F). Together, these results indicate that like VGluT2+ dCINs medullary reticulospinal inputs and segmental sensory afferent inputs extensively recruit GAD2+ dCINs. However, in stark contrast to VGluT2+ dCINs, when we paired subthreshold medMRF or latMRF stimulation with subthreshold DR stimulation, we found that neither medMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing (n = 4) nor latMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing (n = 5) effectively recruited GAD2+ dCINs (Fig. 9B,C). The lack of recruitment during subthreshold stimulation pairing was observed at all pairing intervals tested, that is, independently of whether medMRF or latMRF stimulation preceded or followed DR stimulation (Fig. 9D). The current intensities used during subthreshold stimulation pairing experiments on GAD2+ dCINs were similar to those used on VGluT2+ dCINs and thus cannot account for the observed difference in integrative capacity between the two populations [medMRF 1TCa2+, Wilcoxon, p = 0.25, 45 µA (42–80 µA for GAD2+ dCINs) vs 65 µA (54–77 µA for VGluT2+ dCINs); latMRF 1TCa2+, Wilcoxon, p = 0.34, 40 µA (33–67 µA for GAD2 dCINs) vs 61 µA (40–72 µA for VGluT2 dCINs); DR 1TCa2+, Wilcoxon, p = 0.12, 6 µA (5–9 µA for GAD2 dCINs) vs 10 µA (6–13 µA for VGluT2 dCINs)].
L2 GAD2 dCINs are extensively recruited by reticulospinal and segmental sensory inputs. A, To discriminate GCaMP-expressing L2 GAD2+ dCINs (yellow) from other GCaMP-expressing L2 GAD2+ spinal cord interneurons (green) during calcium imaging experiments in VGluT2:GCaMP3/6 mice, L2 dCINs were retrogradely labeled with RDA (red). B, C, Examples of calcium responses recorded in two different GAD2+ dCINs from the same experiment during 2TCa2+ stimulation of the medMRF and latMRF (5 × 200 µs pulses at 10 Hz) and 1TCa2+ stimulation of the L2 DR (1 × 200 µs pulse). D, Doughnut charts displaying incidence of responsive GAD2+ dCINs (yellow) and nonresponsive GAD2+ dCINs (black) to 2TCa2+ medMRF and latMRF stimulation. E, Doughnut chart displaying incidence of responsive GAD2+ dCINs (blue) and nonresponsive GAD2+ dCINs (black) to 1TCa2+ DR stimulation. F, Doughnut charts showing the proportions of medMRF- and latMRF-responsive GAD2+ dCINs that also responded to 1TCa2+ DR stimulation (mixed yellow and blue pattern) and those that did not (yellow only). Figure 2 provides other details.
L2 GAD2+ dCINs are not recruited by pairing subthreshold MRF-DR stimulations. A, Control recordings showing the expected absence of calcium responses in two VGluT2+ dCINs during subthreshold (0.9TCa2+) stimulation of medMRF (5 pulses), latMRF (5 pulses), and DR (1 pulse). B, Recordings showing that the same GAD2+ dCINs remain nonresponsive during both medMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing and latMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing. C, Doughnut charts showing a general lack of GAD2+ dCINs recruitment both during medMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing and latMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing (pairing interval of 0 ms). D, Averaged response magnitude (across all animals) at each pairing interval tested for medMRF (0.9TCa2+) plus DR (0.9TCa2+) pairing (light green) and latMRF(0.9TCa2+) plus DR (0.9TCa2+) pairing (dark green). We used the same scale as in Figure 5D to facilitate comparison with VGluT2+ dCINs. Inset, Values between −50 ms and +50 ms on expanded x and y scales.
To summarize, the results indicate that GAD2+ dCINs constitute a smaller subpopulation of L2 dCINs than VGluT2+ dCINs, but like VGluT2+ dCINs they are extensively recruited by reticulospinal and sensory inputs. Yet, the data also suggest that GAD2+ dCINs do not have the same capacity to integrate subthreshold reticulospinal and sensory inputs as VGluT2+ dCINs.
Discussion
To interrogate how reticulospinal and peripheral inputs recruit spinal cord interneurons, we investigated two populations of spinal cord interneurons implicated in crossed motor responses and bilateral motor coordination—VGluT2+ and GAD2+ dCINs in the L2 segment of the spinal cord. Combining mouse genetics, anatomy, electrophysiology, and single-cell calcium imaging, we found that although both VGluT2+ and GAD2+ dCINs are recruited by medullary reticulospinal or segmental sensory input alone, they integrate these inputs differently. Critically, we show that when both reticulospinal and sensory inputs are just below the threshold for dCIN recruitment, VGluT2+ dCINs, but not GAD2+ dCINs, have the capability to integrate these inputs.
A limitation of the present study is the use of electrical stimulation for sensory afferents activation. Although electrical stimulation produced clearly distinguishable low- and high-threshold components in the afferent volley, we were unable to activate high-threshold afferents without activating low-threshold afferents. Therefore, other approaches, such as optogenetics-based strategies that selectively activate specific groups of afferents (Montgomery et al., 2016), may be necessary to elucidate the contribution of high-threshold afferents to the recruitment of dCINs.
Glutamatergic dCINs may outnumber other transmitter phenotype dCINs in the L2 segment
The L2 dCIN population may comprise transmitter phenotypes other than the VGluT2+ glutamatergic and the GAD2+ gabaergic dCINs studied here, and include GAD1+ gabaergic dCINs (Henry and Hohmann, 2012; but see Dougherty et al., 2009), glycinergic dCINs (Coulon et al., 2011; Haque et al., 2018), and perhaps also cholinergic dCINs (Stepien et al., 2010). However, we think that these other transmitter phenotypes likely represent only a small group, at least at early postnatal ages. This is based on the finding that altogether glutamatergic and GABAergic dCINs already account for ∼78% of the L2 dCINs (52% VGluT2+ dCINs and 26% GAD2+ dCINs), leaving only 22% for all other transmitter phenotypes. With a transmitter phenotype ratio glutamatergic/GABAergic/others of ∼2:1:1, glutamatergic dCINs may prevail as the largest transmitter phenotype group within the L2 dCIN population. Whether such a prevalence of glutamatergic dCINs would persist into adulthood is unclear, and we cannot exclude the possibility that some, or all, of the GAD2+ dCINs have embarked on a developmental path to become neurons that will release glycine only. Such a development path has been described for GAD+ neurons in laminae VIII and IX of the ventral spinal cord (Sunagawa et al., 2017; Shimizu-Okabe et al., 2022). Glutamatergic dCINs, which likely belong to the V0v and/or V3 transcription factor-defined classes of ventral interneurons (Zhang et al., 2008; Talpalar et al., 2013; Blacklaws et al., 2015; Griener et al., 2015; Osseward et al., 2021), have been reported in many vertebrates, including cat (Bannatyne et al., 2009), rat (Liu et al., 2010), lamprey (Mahmood et al., 2009), tadpole (Li et al., 2007), and zebrafish (Higashijima et al., 2004; Satou et al., 2012; Björnfors and El Manira, 2016; Wiggin et al., 2022). However, save for the present study and the adult lamprey, little is known of the relative proportion of glutamatergic dCINs versus other transmitter phenotype dCINs in these species.
The significance of having a high fraction of glutamatergic dCINs in the L2 segment is not clear because the size of a neuronal population does not necessarily equate to the size of its effect on the network. Nevertheless, one hypothesis is that the size of the glutamatergic dCIN population correlates with the number of motoneurons it aims to control. Accordingly, a high number of VGluT2+ dCINs in L2 segment may be necessary to control both axial and hindlimb motoneurons in the lumbar segments to fulfill the need for bilateral coordination during a variety of movements including swimming, forward and backward locomotion, scratching, and body righting (Borowska et al., 2013; Zelenin et al., 2021; Zhang et al., 2022). Supporting this idea, we previously found that VGluT2+ dCINs in the T7 segment account for only 20% of the dCIN population (LaPallo et al., 2019). This may be sufficient to control axial motoneurons (no hindlimb motoneurons in thoracic segments). Support for this hypothesis will require further physiological analysis of the output connectivity of VGluT2+ dCINs to motoneurons innervating axial and hindlimb muscles.
Reticulospinal-responsive glutamatergic dCINs
Reticulospinal-responsive VGluT2+ dCINs were recruited by both medMRF and latMRF reticulospinal inputs, but latMRF inputs evoked shorter latency and larger calcium responses than medMRF inputs (both alone and when paired with subthreshold segmental sensory inputs). One theoretical explanation for this difference may be more robust synaptic connections between latMRF reticulospinal neurons and VGluT2+ dCINs than between medMRF reticulospinal neurons and VGluT2+ dCINs. However, other factors such as a higher number of latMRF reticulospinal neurons targeting VGluT2+ dCINs may also contribute. Acquiring the specific genetic signature of the medMRF and latMRF reticulospinal neurons may help distinguish these possibilities because it would enable targeted delivery of transgene (e.g., channelrhodopsin-2) either via transgenic mouse lines or recombinant adeno-associated virus (rAAV) carrier systems, although the long time required for robust transgene expression with rAAV carrier systems may make this approach less compatible with experiments in newborns.
A majority of the L2 reticulospinal-responsive VGluT2+ dCINs were also recruited by segmental sensory inputs, suggesting that reticulospinal-responsive VGluT2+ dCINs would be well suited for the regulation of crossed motor response and bilateral motor coordination by sensory feedback. However, as previously mentioned, an important next step toward defining their more precise roles will be to assess how VGlut2+ dCINs connect to the different functional groups of motoneurons (trunk vs extensor vs flexor motoneurons). In this context, approaches such as nontoxic rabies viral vectors (Chatterjee et al., 2019) and laser glutamate uncaging combined with whole-cell patch-clamp recording (Perreault, 2012; Chopek et al., 2018) may be helpful. L2 excitatory dCINs have previously been shown to have direct and indirect synaptic connections to presumed hindlimb motoneurons both at rest and during drug-induced fictive locomotor activity (Butt and Kiehn, 2003; Quinlan and Kiehn, 2007). Whether these include reticulospinal-responsive VGluT2+ dCINs has not been tested directly, but we think that it is highly likely because the great majority of the L2 VGluT + 2 dCINs in this study were responsive to reticulospinal inputs (Fig. 2).
Differential integration of central descending and peripheral sensory inputs by VGluT2+ and GAD2+ dCINs
Little is known about when the individual inputs to dCINs operate at suprathreshold or subthreshold levels during motor activity in vivo and, therefore, when these inputs might be cooperating to recruit dCINs. In this study, we demonstrated that medullary reticulospinal inputs alone and sensory inputs alone can recruit many of the same VGluT2+ dCINs or GAD2+ dCINs. But when these inputs are set up to co-operate at subthreshold levels (pairing experiments), VGluT2+ dCINs, but not GAD2+ dCINs, are recruited. In other words, GAD2+ dCINs cannot be recruited by subthreshold reticulospinal inputs even with the help of subthreshold sensory inputs and vice versa. Identifying the underlying circuit mechanism responsible for the lack of input summation in GAD2+ dCINs will require further in-depth investigations. One hypothetical circuit that may explain the lack of summation in GAD2+ dCINs is feedforward lateral inhibition (Fig. 10). A circuit motif that would include feedforward lateral inhibition may account for (1) the activation of individual GAD2+ dCINs by either reticulospinal or sensory input alone, (2) the lack of summation in GAD2+ dCINs during subthreshold reticulospinal and sensory input pairing, and (3) the reciprocity of the phenomenon (lack of summation when reticulospinal input preceded sensory input but also when sensory input preceded reticulospinal input). Feedforward lateral inhibition is a canonical circuit that has been used to describe competitive selection in different brain regions, for example, in the superior colliculus (Mysore and Knudsen, 2012; Mysore and Kothari, 2020).
Hypothetical circuit model for activation of GAD2+ dCINs. Working hypothesis in the form of a connectivity diagram depicting how feedforward lateral inhibition may account for the recruitment of GAD2+ dCINs by MRF stimulation (reticulospinal input), DR stimulation (sensory inputs), and the lack of recruitment during subthreshold paired stimulations. The number of synapses between neurons is unknown. The vertical dashed line represents the midline.
Conclusion
We found that dCINs, which are considered part of the spinal network for crossed motor responses and bilateral motor coordination, are not a homogeneous group and comprise at least two distinct dCINs subpopulations, VGluT2+ and GAD2+ dCINs. We also found that although both populations can be recruited by reticulospinal and sensory inputs, the two populations differ with respect to how they integrate these inputs. This study expands our understanding of how the glutamatergic and GABAergic components of the lumbar dCINs network may be driven by reticulospinal or peripheral inputs and provides a platform for future studies aimed at elucidating specific contributions during motor activity, particularly during the phases of movement when reticulospinal or sensory neurons may be less active (Perreault et al., 1993, 1999; Matsuyama and Drew, 2000a, b; Prentice and Drew, 2001; Oueghlani et al., 2018; Frigon et al., 2021), or when reticulospinal or sensory drives are weakened after injury (Horstman et al., 2019; Engmann et al., 2020; Hough et al., 2021; Wang et al., 2021).
Footnotes
This work was supported by Craig H. Neilsen Foundation Grant 313238, National Institutes of Health–National Institute of Neurological Disorders and Stroke Grant R01 NS085387, and National Science Foundation Division of Biological Infrastructure Grant 2015317. We thank Renee Shaw for help with cryostat sectioning, histology, immunohistochemistry, and animal care; Brandon K. LaPallo for the provision of two retrogradely labeled preparations; and the Emory University Integrated Cellular Imaging Core for the use of a confocal microscope.
The authors declare no competing financial interests.
- Correspondence should be addressed to Marie-Claude Perreault at m-c.perreault{at}emory.edu
