Abstract
Spinal interneurons shape motor neuron activity. Gata3+ V2b neurons are a major inhibitory spinal population. These neurons are present at multiple spinal levels in mice, suggesting an important function in motor control. In zebrafish, our previous work showed that V2b neurons are evenly distributed along the spinal cord, where they act to slow down locomotion. However, the timing of V2b activity during locomotion, their postsynaptic targets other than motor neurons, and their recruitment across different behaviors remain unknown. In this study, we address these questions using larval zebrafish. First, via optogenetic mapping of output in the rostrocaudal axis, we demonstrate that V2b neurons robustly inhibit motor neurons and other major spinal populations, including V2a, V1, commissural neurons, and other V2b neurons. V2b inhibition is patterned along the rostrocaudal axis, providing long-range inhibition to motor and V2a neurons but more localized inhibition of V1 neurons. Next, by recording V2b activity during different visually and electrically evoked movements, we show that V2b neurons are specifically recruited for forward swims and turns, but not for fast escape movements. Furthermore, a subset of V2b neurons also exhibited short-latency sensory-evoked activity preceding motor initiation. Finally, we show that V2b inhibition occurs in phase with the leading edge of the motor burst, in contrast to V1 inhibition which occurs in phase with the falling edge of the motor burst. Taken together, these data show that in axial motor networks, V2b neurons act via multiple targets to produce in phase, leading inhibition during locomotion.
Significance Statement
Spinal interneurons are critical for executing and regulating movements. However, it has been challenging to understand their functions and interconnections because the spinal cord circuit is complex, with many long-range connections that are challenging to map. Using optogenetics in the larval zebrafish, we mapped the connectivity and activity of an inhibitory spinal population: V2b neurons. We show that V2b neurons not only inhibit motor neurons but also other major excitatory and inhibitory populations. With electrophysiology and calcium imaging, we recorded V2b activity during different behaviors and found that V2b neurons inhibit their targets on the rising phase of motor bursts, preferentially during slow locomotion. These results suggest that V2b neurons have a distinctive role in motor control.
Introduction
In vertebrates, locomotor movements are executed by intrinsic networks in the spinal cord, comprising excitatory and inhibitory interneurons that connect with each other and with motor neurons to orchestrate different movements (Stepien and Arber, 2008; Goulding, 2009). Across vertebrates, neuronal genetic identity, morphology, and neurotransmitter expression remain remarkably conserved, making it convenient to study and apply knowledge across species (Grillner, 2006; Goulding, 2009; Grillner and Jessell, 2009).
Spinal V2b neurons are a major inhibitory population marked by expression of the Gata3 transcription factor across vertebrates (Karunaratne et al., 2002; Lundfald et al., 2007; Callahan et al., 2019). V2b neurons have primarily been implicated in limb movements, where together with V1 neurons, they enforce flexor-extensor alternation (Zhang et al., 2014; Britz et al., 2015). The V2b population, however, is present not only at limb levels but all along the rostrocaudal extent of spinal cord in both mice (Al-Mosawie et al., 2007; Francius et al., 2013) and zebrafish (Callahan et al., 2019), suggesting an ancestral role in locomotor coordination. In axial circuits, evidence suggests that V2b neurons can serve as locomotor brakes, with V2b activation reducing swimming frequency and V2b suppression increasing swimming frequency (Callahan et al., 2019).
In mice, V2b neurons project ipsilateral descending axons spanning several segments (Flynn et al., 2017) and preferentially contact extensor motor neurons (Britz et al., 2015). V2b neurons form a subset of long descending propriospinal neurons connecting cervical and lumbar segments (Flynn et al., 2017) and are thought to facilitate rostrocaudal coordination through interlimb communication (Ruder et al., 2016). Zebrafish V2b neurons also project ipsilateral descending axons (Callahan et al., 2019) and contact motor neurons. Other postsynaptic partners of V2b neurons, including long-range targets, remain unknown in both mice and fish, despite the importance of long-range connectivity for movement. Previously, we demonstrated that spinal V1 neurons inhibit different postsynaptic targets locally and long range (Sengupta et al., 2021). It remains unclear whether these variations in rostrocaudal connectivity are specific to V1 neurons or a more general property of spinal circuits.
Though several studies have manipulated V2b neurons to examine the consequences for behavior, there are to date no recordings of the relationship between V2b neuronal activity and behavior. Recent studies in axial motor networks have confirmed that excitatory V2a neurons comprise subpopulations that are involved in either rhythm (slow vs fast) or pattern (bilaterally symmetric vs asymmetric) control (Menelaou and McLean, 2019; Agha et al., 2024). Whether V2b neurons also have dedicated roles in pattern or rhythm generation remains to be seen.
Though the spinal cord has traditionally been segregated into dorsal “sensory” circuits and ventral “motor” circuits, the reality is more complicated. In mammals, Ia afferents carrying proprioceptive information form anatomical synaptic connections with ventral horn V1 and V2b neurons, and loss of V1/V2b neurons eliminates the reciprocal flexor/extensor inhibitory input to motor neurons (Zhang et al., 2014). However, fish do not have proprioceptive muscle spindles, and it is unknown whether ventral horn populations in fish receive sensory input from the periphery.
In this study, we analyzed the longitudinal connectivity and recruitment pattern of V2b neurons in axial motor networks of larval zebrafish. Using optogenetic assisted circuit mapping, we demonstrate that V2b neurons inhibit motor neurons and other cardinal excitatory and inhibitory spinal populations, with structured output in the rostrocaudal axis. We show that some V2b neurons are preferentially active at slower speeds, for both symmetric and asymmetric movements. Furthermore, a subset of V2b neurons also exhibited spiking with a short latency to a sensory stimulus, implying the presence of direct sensory afferent input. Finally, V2b spiking exhibited a phase lead in the swim cycle relative to the motor burst, in contrast to V1 neurons which exhibit a phase lag. Combined, these results indicate that V2b neurons regulate speed in axial motor circuits, and their reciprocal connectivity with V1 neurons forms a motif that may recur in limb control circuits.
Materials and Methods
Animals
Adult zebrafish (Danio rerio) were maintained at 28.5°C with a 14/10 h light/dark cycle in the Washington University Zebrafish Facility following standard care procedures. Larval zebrafish, 4–6 days post fertilization (dpf), were used for experiments and kept in Petri dishes in system water or housed with system water flow. Animals older than 5 dpf were fed rotifers or Gemma dry food daily. Zebrafish do not differentiate sexually until ∼3 weeks old; therefore, it was not possible to identify males and females in the 4–6 dpf stage and experiments were carried out on animals that will eventually be assigned either sex. All procedures described in this work adhere to NIH guidelines and received approval by the Washington University Institutional Animal Care and Use Committee.
Transgenic fish lines
For all connectivity experiments, the stable transgenic line Tg(Gata3:Gal4;UAS:CatCh)stl602 (ZDB-ALT-201209–12; Callahan et al., 2019) generated by Tol2 mediated transgenesis previously in our lab was used. For targeting V2a and V1 neurons, the Tg(vsx2:loxP-DsRed-loxP-GFP)nns3Tg (ZDB-ALT-061204-4; Kimura et al., 2006) and Tg(En1: LoxP-dsRed-LoxP:DTA) nns55Tg (ZDB-ALT-191030-2; Kimura and Higashijima, 2019) lines, respectively, were crossed to Tg(Gata3:Gal4;UAS:CatCh) to generate double transgenics. Secondary motor neurons were targeted in part using the Tg(mnx:pTagRFP)stl603 line created in the lab. For recruitment studies in Figures 6 and 7, V2b neurons were targeted in the stable Tg (Gata3:LoxP-dsRed-LoxP:GFP)nns53Tg (ZDB-ALT-190724-4; Callahan et al., 2019). Calcium imaging was performed in double transgenic nacre larvae obtained by crossing Tg (Gata3:LoxP-dsRed-LoxP:GFP)nns53Tg to Tg(elavl3:Hsa.H2B-GCaMP6s)jf5Tg (ZDB-TGCONSTRCT-141023-1; Zhu et al., 2023).
Electrophysiology
A total of 4–6 dpf larvae were immobilized with 0.1% α-bungarotoxin and fixed to a Sylgard lined Petri dish with custom-sharpened tungsten pins. One muscle segment overlaying the spinal cord was removed at the midbody level (Segments 9–13). The larva was then transferred to a microscope (Nikon Eclipse E600FN) equipped with epifluorescence and immersion objectives (60×, 1.0 NA). The bath solution consisted of the following (in mM): 134 NaCl, 2.9 KCl, 1.2 MgCl2, 10 HEPES, 10 glucose, 2.1 CaCl2. Osmolarity was adjusted to ∼295 mOsm and pH to 7.5. For recording V2b spiking during swims, a combination of cell-attached and whole-cell patch-clamp recordings were obtained from V2b neurons. Patch pipettes (7–15 MΩ) were either filled with extracellular saline (cell-attached) or patch internal (whole-cell) solution and targeted to a V2b neuron. After formation of a gigaohm seal, extracellular spikes were recorded in cell-attached mode. To obtain whole-cell recordings, brief suction pulses were used to break into the cell. Spiking was recorded in current-clamp mode using the following patch internal solution (in mM): 125 K gluconate, 2 MgCl2, 4 KCl, 10 HEPES, 10 EGTA, and 4 Na2ATP. Ventral root recordings were obtained 2–3 segments caudal to the recorded V2b, using suction electrodes (diameters, 20–50 µm).
For mapping connectivity to target neurons, larvae were immobilized and dissected as before. Whole-cell patch-clamp recordings were made from neurons in voltage-clamp mode using the following patch internal solution (in mM): 122 cesium methanesulfonate, 1 tetraethylammonium-Cl, 3 MgCl2, 1 QX-314 Cl, 10 HEPES, 10 EGTA, and 4 Na2ATP. APV (10 µM) and NBQX (10 µM) were added to the bath to block glutamatergic transmission. For all patch internal solutions, pH was adjusted to 7.5 and osmolarity to 290 mOsm. Additionally, for identifying primary motor neurons, commissural neurons, and sensory populations, sulforhodamine 0.02% was included in the patch internal to visualize morphology of recorded cells post hoc. Recordings were acquired using a MultiClamp 700B amplifier and Digidata 1550 (Molecular Devices). Signals were filtered at 2 kHz and digitized at 100 or 50 kHz. For IPSCs, cells were voltage clamped at +0.3 mV (after correction for liquid junction potential of 14.7 mV).
Escape/swims were elicited by a brief electric stimulus to the tip of the head (20–50 V for 0.2 s for escapes and 10–20 V for 0.2–1 ms for swims; Liao and Fetcho, 2008). For optomotor responses, visual stimuli of blue and white moving gratings (spatial width of 1 cm) were projected on a screen ∼1 cm in front of the larva and moved at 1 cm/s (Ahrens et al., 2012).
Optogenetic stimulation
A Polygon 400 or 1000 Digital Micromirror Device (Mightex) was used to deliver optical stimulation. The projected optical pattern consisted of a 4 × 4 grid of 16 squares. Each square in the grid approximately measured 20 µm × 20 µm. One full stimulus pattern consisted of an ordered sequence of each of the 16 squares sequentially. The 16th square in most cases spilled out to neighboring segments or out of the spinal cord and hence was not included in the data. For each small square, illumination consisted of a 20 ms light pulse (470 nm) at 50% intensity (4.6–5.2 µW under 60×, 1.0 NA). The sequence was triggered using a TTL pulse from the Digidata to synchronize the stimulation with electrophysiology. The objective was carefully positioned over a single spinal segment prior to stimulus delivery; for each new segment, the stage was manually translated and repositioned. V2b spiking reliability was measured by delivering multiple trials to a selected square that had evoked spiking on the first trial. From analysis showing that each V2b neuron could be triggered to spike by illumination in ∼3 out of 15 grid squares (on average) and given that the total number of V2b neurons per hemisegment is ∼10 (quantitation of CatCh expression), we estimate that each grid square likely activates ∼2 V2b neurons ipsilateral to the recording (10 × 3/15). For the high-frequency stimulation in Figure 2, a single square was illuminated with a 20 Hz train of five 20 ms pulses.
Calcium imaging and behavior
A total of 5–6 dpf larvae were embedded dorsal side up in 2.5% low-melting point agarose in the center of a 35-mm-diameter Petri dish. A small portion of the tail was freed from agarose for recording tail movements. Calcium signals in the midbody (10–15) segments of the spinal cord were recorded using a Nikon A1RHD25 MP two-photon confocal microscope housed in the Washington University Center for Cellular Imaging (WUCCI). Excitation was provided via a tunable Titanium Sapphire laser at an excitation wavelength of 930–940 nm. Images were captured using a 25× water immersion objective and acquired at 29 Hz. After acquiring calcium activity, confocal images of GCaMP and Ds-Red signals were separately obtained to identify V2b neurons post hoc. Tail movements were captured using a Grasshopper 3 camera (FLIR, Edmund Optics). The tail was illuminated with IR LEDs and an infrared filter was added to the camera to collect IR light. Visual stimuli were projected using a mini projector (Optoma ML750ST). Custom MATLAB codes were used to project visual stimuli and synchronize calcium imaging with tail movement recordings (https://github.com/rob-the-bot/StimulusTrigger). Visual stimuli consisted of red and black converging gratings (spatial width of 1 cm), projected on a screen ∼1 cm in front of the larva, and moving at 1 cm/s (Ahrens et al., 2012). Trials were 10 s long, spaced with 5 s of stationary gratings.
Analysis
Electrophysiology data were imported into Igor Pro 6.37 (WaveMetrics) using NeuroMatic (Rothman and Silver, 2018). Spikes and IPSCs were analyzed using custom code in Igor and MATLAB (Sengupta et al., 2021). For analyzing motor recordings from the ventral nerve (Fig. 6), the raw swim signal (Fig. 6B, gray traces) was converted to the standard deviation signal over a sliding window of 10 ms (Fig. 6B, blue traces). Peaks from this SD signal, corresponding to the midpoint of motor bouts, were calculated using custom MATLAB code. V2b spike times were normalized relative to the interbout duration. Histograms of these spike times were normalized for each cell and then pooled together (from 23 cells) for the summary plot shown in Figure 7A. For connectivity experiments (Figs. 1–5), data were analyzed as previously reported (Sengupta et al., 2021). Briefly, charge transfer for the evoked response was calculated by integrating the current in a 50 ms window from the onset of the optical stimulus (Evoked) and subtracting this from Control 1, a similar integral over a 50 ms window before the optical stimulus, to account for spontaneous activity. To calculate noise values for statistical comparison with evoked IPSCs, a similar integral for a different 50 ms window at the end of the recording (Control 2) was subtracted from Control 1. Both the charge transfer of the evoked response and background noise were summed across the stimuli for each segment:
Peak amplitudes of IPSCs were calculated as the maximum value of the charge transfer trace. Conductances were calculated as peak amplitude/driving force (75 mV). Input resistance was measured by an average of small hyperpolarizing pulses.
Calcium imaging data was analyzed using Suite2p and MATLAB as described previously (Peron et al., 2015; Arriaga and Han, 2017). V2b neurons were identified using ROI registration in ImageJ (FIJI). Analyzed GCaMP images were extracted from Suite2p and registered to confocal images of Ds-Red cells. Overlapping ROIs were manually selected and their calcium activity analyzed with tail movements.
Tail movement videos were analyzed using ZebraZoom (Mirat et al., 2013) to extract x–y coordinates. These were then analyzed using MATLAB to determine symmetric forward swim movements or asymmetric turns to the right or left. Frames corresponding to the start of forward swims were extracted and used to define time periods for analyzing calcium activity.
Details of statistical tests and p values are reported in the figure legends. Custom analysis code is available on the Bagnall lab GitHub site (https://github.com/bagnall-lab/V2b-connectivity-and-behavior).
Results
Optogenetic assisted mapping of V2b connectivity
To map V2b output connectivity along the rostrocaudal axis, we utilized the high-throughput technique of optogenetic assisted circuit mapping (Petreanu et al., 2007). We used the transgenic line Tg(Gata3:Gal4;UAS:CatCh), in which V2b neurons expressed a calcium-permeable variant of channelrhodopsin, CatCh (Kleinlogel et al., 2011; Fig. 1A, schematic). We first calibrated the efficacy and specificity of the optical stimulus for activating V2b neurons restricted to a single segment. A 4 × 4 square grid pattern was projected approximately over a single spinal segment (Fig. 1B). Each square in this grid measured 20 × 20 µm and was illuminated in a sequential order with a 20 ms light pulse. V2b neurons recorded in current-clamp mode exhibited a mean resting membrane potential of −74.3 ± 2.37 mV. Direct illumination on the soma (black dot) or nearby (on average, 3.1 grid squares per neuron) elicited robust spiking in V2b neurons (Fig. 1C, red traces). V2b neurons project descending axons in the R-C axis, extending an average of seven segments (Callahan et al., 2019). We tested whether the optical stimulus could elicit antidromic spikes when the axon was illuminated by translating the stimulus to neighboring rostral and caudal segments while recording V2b somatic responses (Fig. 1D). Optical stimulation outside of Segment 0 (segment in which V2b is being recorded) did not evoke any appreciable activity (Fig. 1D,E; N = 10 neurons). V2b neurons also showed high reliability of spiking to repeated presentations of the optical stimulus, with a mean reliability of 75% ± 12.1 (Fig. 1F; N = 10 neurons). The latency from optical stimulus onset to the first evoked spike in the V2b neuron ranged from ∼4 to 40 ms (Fig. 1G). Together, these data show that the grid optical stimulus activated V2b neurons in a spatially restricted manner and therefore can be utilized for mapping V2b connectivity to different targets along the R-C axis.
Calibration of V2b activation by patterned optical stimulus. A, Schematic of the experimental setup showing targeted intracellular recording and optical stimulation in Tg(Gata3:Gal4,UAS:CatCh) animals. B, Schematic of the patterned optical stimulus. A 4 × 4 grid was overlaid on approximately one spinal segment and each small square in the grid (blue square) was optically stimulated in an ordered sequence with a 20 ms light pulse at 50% power. Position of the recorded cell is outlined in a gray dotted circle. Arrowheads show CSF-cNs also labeled in this line. Scale bar is 10 µm. C, Illustration of the analysis. Intracellular recordings elicited from optical stimulation in each grid square (left). Spiking is denoted in red. Same data shown as a heat map and superimposed on the optical stimulus grid (right). The position of the recorded cell is indicated with a black circle. D, V2b responses evoked by optical stimuli in segments rostral or caudal to the recorded neuron. Representative traces of activity (top) and spike count (bottom) of the same V2b neuron while the optical stimulation was moved along the rostrocaudal axis. Red traces indicate spiking. E, Quantification of spiking in V2b neurons as the optical stimulus is presented along the rostrocaudal axis. N = 10 neurons. F, Reliability of spiking in these neurons with multiple trials of the same optical stimulus. N = 10 neurons. G, Latency of the first spike from onset of the optical stimulus in V2b neurons. N = 10 neurons. In plots E and F, gray bar indicates median values and the red circle marks the representative neuron shown in D.
V2b neurons robustly inhibit motor neurons both locally and long range
In mice, V2b neurons in lumbar segments make anatomical contacts to motor neurons, preferentially those controlling extensors (Britz et al., 2015). In larval zebrafish, physiological studies show that V2b neurons inhibit both fast, primary motor neurons and slow, secondary motor neurons (Callahan et al., 2019), but it is not known if this connectivity is only local or extends over many segments. To investigate this, we recorded intracellularly from primary motor neurons in the Tg(Gata3:Gal4;UAS:CatCh) line (Fig. 2A). Primary motor neurons were identified in bright field with their characteristic large, laterally positioned soma and post hoc dye fill showing extensive muscle innervation (Menelaou and McLean, 2012). Cells were recorded using a cesium internal, held at 0 mV, and bathed in glutamate receptor blockers (NBQX, APV) to isolate inhibitory post synaptic currents (IPSCs). The grid optical stimulus was delivered at a single segment each time and translated rostrally for seven segments to cover the full descending axon length of V2b neurons.
Motor neurons receive local and long-range inputs from V2b neurons. A, Schematic of the experimental design showing intracellular recordings from primary (brown) and secondary (orange) motor neurons paired with optical stimulation of CatCh+ V2b neurons (green) along the rostrocaudal axis. B, Representative overlay of 15 traces of IPSCs recorded in a primary motor neuron (left) and a secondary motor neuron (right) during illumination of Segments 1, 3, and 7 rostral to the recorded neuron soma. Each trace is the evoked response by stimulation of a different square in the grid within one segment. Colored trace represents the mean. The optical stimulus is shown as a blue bar. C, Box plots showing total charge transfer per segment recorded in primary motor neurons (left) and secondary motor neurons (right). Red asterisks mark segments that were significantly different from noise. For pMNs, p values for Segments −2 to 7 = 0.123, 0.005, 0.004, 0.002, 0.0024, 0.000018, 0.00002, 0.009, respectively (Wilcoxon sign rank test, N = 9–25 pMNs for each data point). For sMNs, p values for Segments −2 to 7 = 0.03, 0.15, 0.006, 0.008, 0.002, 0.003, 0.002, 0.08, respectively (Wilcoxon sign rank test, N = 7–11 sMNs for each data point). D, Left, Representative overlay of 10 traces of IPSCs recorded in primary motor neurons (pMNs) during a 20 Hz optical stimulus train (5 pulses, 20 ms) delivered to one small square in Segment 1. Right, Average latency of the first IPSC with the train stimulus. Error bars represent standard error of mean. N = 5 neurons for pMNs and 2 neurons for sMNs. E, F. Comparison of the percent of optical stimulus grid squares in each segment that evoked IPSCs (E) and the peak conductance of IPSCs (F) in primary and secondary motor neurons. Here and in subsequent figures, circles represent median values and error bars indicate the 25th and 75th percentiles. N = 9–25 pMNs and 11 sMNs.
Primary motor neurons recorded in this configuration exhibited IPSCs when V2b neurons were optically activated nearby or at long range (Fig. 2B, left). To quantify these inputs, we calculated the charge transfer of evoked IPSCs (Fig. 2C, inset) for each segment relative to noise (see Materials and Methods). Primary motor neurons received significant V2b-mediated inhibition even up to seven segments distant from the site of stimulation (N = 10–25 neurons; Wilcoxon sign rank test, p < 0.05). We next performed similar mapping from V2b neurons to secondary motor neurons, which were identified either genetically in the double transgenic Tg(Gata3:Gal4;UAS:CatCh; mnx1:ptag:RFP) line or by post hoc dye label. Secondary motor neurons also received local and long-range inhibition from V2b neurons (Fig. 2B, right). Charge transfer at Segments 1 through 5 was significantly above noise levels (Fig. 2C, right; N = 7–11 neurons; Wilcoxon sign rank test, p < 0.05) but decreased to insignificant at seven segments away.
V2b neurons in zebrafish have purely descending axons and therefore are not expected to inhibit neurons in the rostral direction. Consistent with this expectation, we saw little to no evoked IPSCs during delivery of optical stimuli caudal to the recorded segment (Segments −1, −2; Fig. 2C). Because our transgenic line also labels CSF-cNs (Fig. 1B, gray arrowheads), which have purely ascending axons (Böhm et al., 2016; Knafo and Wyart, 2018) and are known to selectively inhibit one of the four primary motor neurons (Hubbard et al., 2016), this low-frequency connectivity likely results from activation of CSF-cNs.
To verify that evoked IPSCs were monosynaptically generated, we delivered a train of optical stimuli (five 20 ms pulses at 20 Hz). IPSCs followed the train with consistent latency and low jitter, consistent with monosynaptic connectivity (Fig. 2D).
Finally, we asked if V2b connectivity to motor neurons showed differences in number and/or strength of connections with distance. As a proxy for number of connections, we quantified the fraction of small squares in the 4 × 4 grid stimulus in each segment that successfully evoked IPSCs. We estimate that ∼2 V2b neurons are activated by each small grid square stimulus (see Materials and Methods). Because V2b neurons are evenly distributed along the R-C axis (Callahan et al., 2019) and our patterned stimulus tiles one full segment, the number of squares evoking IPSCs can be used as a rough proxy for the number of monosynaptically connected V2b neurons to the target. To quantify IPSC amplitude, we calculated the average maximum conductance for each segment. Interestingly, connectivity rates remained high along the longitudinal axis, with 63% of stimuli evoking IPSCs onto primary motor neurons even at seven segments away (Fig. 2E; N = 10–25 pMNs). However, the amplitude of these V2b contacts onto primary motor neurons tapered off, exhibiting a 60% decline from Segment 1 to Segment 7 (Fig. 2F). In contrast, for secondary motor neurons, both the number of stimuli eliciting V2b IPSCs (80% reduction) and the strength of connections (73% reduction) declined from Segment 1 to Segment 7 (Fig. 2E,F; N = 7–11 sMNs). Taken together, these data show that V2b neurons monosynaptically inhibit both primary and secondary motor neurons locally and at long distances. The strength of V2b-mediated inhibition progressively diminishes with distance but extends further for primary than for secondary motor neurons.
V2b neurons inhibit Chx10+ V2a neurons both locally and long range
Spinal V2a neurons are a major source of excitatory drive to motor neurons in both mice (Crone et al., 2008; Hayashi et al., 2018) and zebrafish (Kimura et al., 2013; Ampatzis et al., 2014; Song et al., 2018; Menelaou and McLean, 2019). Therefore, we examined V2b-mediated inhibition of the V2a population along the longitudinal axis. V2a neurons were targeted using the double transgenic Tg(Chx10:lox-dsred-lox:GFP);(Gata3:Gal4;UAS:CatCh) line (Fig. 3A, left). Stimulation of V2b neurons both locally and long-range elicited inhibitory synaptic inputs in V2a neurons (Fig. 3B, left). Charge transfer of inhibitory currents was significantly higher than noise for optical stimuli up to seven segments from the recording site (Fig. 3C, left; N = 7–10 neurons; Wilcoxon sign rank test, p < 0.05). V2a neurons did not receive any appreciable evoked synaptic input when caudal segments were illuminated (Fig. 3C, left, Segment −1, −2). As with primary motor neurons, the inferred number of stimuli eliciting IPSCs was maintained along the rostrocaudal axis (26% decrease from Segment 1 to Segment 7; Fig. 3D; N = 7–10 neurons) but the strength of V2b inhibitory inputs diminished gradually at long range (56% reduction from Segment 1 to Segment 7; Fig. 3E). Overall, these data identify V2a population as a major novel target of V2b neurons. Furthermore, we show that V2b neurons inhibit the V2a population both locally and long-range, with a gradual reduction in strength at long distances.
V2b neurons inhibit V2a and V1 neurons, with inhibition onto V1 neurons more spatially delimited. A, Schematic of the experimental design showing intracellular recordings from V2a (left) and V1 (right) neurons paired with optical stimulation of CatCh+ V2b neurons (green) along the rostrocaudal axis. B, Representative overlay of 15 traces of IPSCs recorded in a single V2a neuron (left) and V1 neuron (right) during illumination of Segments 1, 3, and 7 rostral to the recorded neuron soma. Colored trace represents the mean. Duration of the optical stimulus is shown as a blue bar. C, D, Box plots showing total charge transfer per segment recorded in V2a neurons (C) and V1 neurons (D). Red asterisks mark segments that were significantly different from noise. For V2as, p values for Segments −2 to 7 = 0.125, 0.0625, 0.0031, 0.0078, 0.0078, 0.0039, 0.0031, 0.0313, respectively (Wilcoxon sign rank test, N = 7–10 V2a neurons for each data point). For V1s, p values for Segments −2 to 7 = 0.1562, 0.3125, 0.0126, 0.0078, 0.0027, 0.0195, 0.0976, 0.0781, respectively (Wilcoxon sign rank test, N = 9 V1 neurons for each data point). E, F, Comparison of the percent of squares in the optical stimulus grid that evoked IPSCs (E) and the peak conductance of IPSCs (F) in V2a (cyan) and V1 (magenta) neurons. N = 10 V2a and 9 V1 neurons.
V2b neurons inhibit spinal V1 neurons locally
V1 neurons are a major inhibitory population in the spinal cord, marked by expression of Engrailed1 in mice (Gosgnach et al., 2006), Xenopus (Li et al., 2004), and zebrafish (Higashijima et al., 2004). These neurons share a special relationship with V2b neurons for limb control (Britz et al., 2015). In lumbar networks that control the hind limbs, V1/V2b neurons reciprocally inhibit antagonistic motor pools, driving alternation of flexors and extensors. Ia inhibitory interneurons for flexors and extensors have been shown to directly inhibit each other in several species (Pratt and Jordan, 1987a; Jankowska, 1992; Wilson et al., 2010a), including humans (Baldissera et al., 1987). While Ia inhibitory neurons belong to the V1/V2b populations (Alvarez et al., 2005; Zhang et al., 2014), direct evidence of reciprocal inhibition between these two genetically defined populations has not been found. To determine if V2b neurons inhibit the V1 population, we recorded from V1 neurons identified in the Tg(eng1b:lox-dsred-lox:DTA) line (Fig. 3A, right). Activation of V2b neurons evoked significant IPSCs in V1 neurons locally, at Segments 0–3, but not long range at Segments 5–7 (Fig. 3B, right, 3D; N = 8–9 neurons, Wilcoxon sign rank test). Both the fraction of squares evoking IPSCs (Fig. 3E) and the amplitude of evoked IPSCs (Fig. 3F) were maximal at 2–3 segments from the recording site and diminished in either direction. Therefore, the structure of V2b-mediated inhibition varies along the longitudinal axis to distinct downstream targets. These data are the first demonstration that V2b neurons inhibit the V1 population.
V2b connectivity to other ventral spinal populations
We next extended this map to include other spinal populations in the ventral horn. Commissural neurons in the spinal cord are a major functionally relevant group that helps secure left–right alternation in mice (Lanuza et al., 2004; Talpalar et al., 2013; Haque et al., 2018). They have also been implicated in rostrocaudal coordination in both mice (Ruder et al., 2016) and zebrafish (Kawano et al., 2022). In zebrafish, V3 commissural neurons additionally have been shown to modulate motor strength (Böhm et al., 2022; Wiggin et al., 2022). Ventral horn commissural neurons comprise more than one genetically defined class (dI6, V0, and V3 neurons; Andersson et al., 2012; Satou et al., 2012, 2020; Danner et al., 2019; Kishore et al., 2020), and each class also exhibits morphological variability. To first determine whether V2b neurons target any commissural neurons, we performed blind recordings of neurons and classified them post hoc based on their morphology from dye fills as commissural, bifurcating, and ascending neurons, likely belonging to the V0/dI6 classes. V3 neurons are unlikely to be included in this group because they are located at the extreme ventral edge of the spinal cord, an area we did not target for recordings (Böhm et al., 2022; Wiggin et al., 2022). We collectively refer to our morphologically defined commissural neurons as commissural neurons (CNs), as we cannot exclude possible origin from dorsal horn/sensory populations. Commissural neurons received modest and variable local inputs from V2b neurons (Fig. 4A; N = 5–7 neurons) that were significant only at Segment 0 (Wilcoxon sign rank test, p < 0.05). These results suggest that V2b neurons predominantly target ipsilateral pathways.
V2b neurons minimally inhibit other spinal populations. A, Top, Schematic of recordings from ventral derived commissural neurons (CNs) during optical excitation of V2b neurons. Middle, Representative overlay of 15 traces of IPSCs recorded in a commissural neuron during illumination of Segments 1, 3, and 7 rostral to the recorded neuron soma. Colored trace represents mean. Duration of the optical stimulus is shown as a blue bar. Bottom, Box plot showing total charge transfer per segment (inset, illustration) recorded in commissural neurons. For CNs, p values for Segments −2 to 7 = 0.625, 0.4375, 0.0468, 0.0781, 0.0625, 0.2187, 0.4867, 0.375, respectively (Wilcoxon sign rank test, N = 7 neurons for each data point). B–D, Same as in A for V2b inhibition to other V2bs (B), CoPA neurons (C), and DoLA neurons (D). For V2b neurons, p values for Segments −2 to 7 = 0.5, 0.125, 0.0156, 0.0156, 0.1093, 0.5468, 0.2968, 0.8125, respectively (Wilcoxon sign rank test, N = 9 neurons for each data point). For CoPA neurons, p values for Segments −2 to 7 = 0.622, 0.2661, 0.0922, 0.522,0.1015, 0.2333, 0.5771, 0.1762, respectively (Wilcoxon sign rank test, N = 12 neurons for each data point). For DoLA neurons, p values for Segments 0 to 7 = 1, 0.25, 1, 0.75, 0.25, 0.5, respectively (Wilcoxon sign rank test, N = 3 neurons for each data point).
V2b neurons have been shown to inhibit each other (Callahan et al., 2019), but the rostrocaudal extent and amplitude of this inhibition is unknown. We observed V2b inhibition onto other V2b neurons exclusively locally (Fig. 4B; N = 5–9 neurons; Wilcoxon sign rank test, p < 0.05), not at long range.
V2b neurons do not inhibit two dorsal horn sensory populations
Finally, we wanted to test if V2b neurons target dorsal horn sensory neurons. In zebrafish, the Commissural Primary Ascending (CoPA) neurons, likely homologous to mammalian dI5 neurons (Wells et al., 2011; Sengupta et al., 2021), are glutamatergic neurons activated during the tactile reflex (Knogler and Drapeau, 2014). During swims, CoPAs are gated by inhibition from V1 neurons (Higashijima et al., 2004; Sengupta et al., 2021) and possibly others (Knogler and Drapeau, 2014). CoPA neurons are readily identifiable by their distinct axonal and dendritic morphology in post hoc dye fills. Figure 4C shows representative traces (middle) and summary data (bottom) of V2b inputs to CoPA neurons. CoPAs did not receive any appreciable inhibition from V2b neurons (N = 11–12 neurons; Wilcoxon sign rank test, p > 0.05) suggesting that V2b neurons are not involved in sensory gating of CoPA neurons during swims. Interestingly, CSF-cNs are known to contact CoPAs (Hubbard et al., 2016). However, we did not capture this inhibition (Fig. 6C, Segments −1, −2), probably because CSF-cN inhibition to CoPA neurons is prominent slightly more distally, three segments away (Hubbard et al., 2016). We also examined responses in dorsal longitudinal (DoLA) neurons, a sensory target likely homologous to mammalian dI4 neurons (Todd and McKenzie, 1989; Wells et al., 2011). DoLAs are GABAergic and exhibit a characteristic axonal morphology but much less is known about their function. V2b neurons did not show any appreciable inputs to DoLA neurons (Fig. 4D; N = 3 neurons; Wilcoxon Sign Rank Test, p > 0.05). We conclude that inhibition of these sensory targets may not be a primary function of V2b neurons.
To compare V2b connectivity patterns across different cell types, we plotted a heat map showing median charge transfer for each population along the R-C axis (Fig. 5). For each target neuron, charge transfer was first normalized to its total cellular conductance (inverse of input resistance) to allow a direct comparison across targets. This map clearly reveals two key points: (1) extensive V2b-mediated inhibition of motor neurons and V2a neurons that slowly tapers over distance and (2) selectively localized V2b-mediated inhibition onto V1, other V2b, and commissural neurons. Taken together, this map shows that V2b inhibition is spatially structured in the rostrocaudal axis with distinct targets locally versus long range.
Summary of V2b connectivity. A, Heat map showing normalized charge transfer for each postsynaptic target population along the rostrocaudal axis. The charge transfer per segment for each target population was first normalized to its measured intrinsic neuronal conductance (inverse of Rin). Median values of charge transfer for each target cell population, normalized to the largest measured value across the populations, are shown. Values for Segments 4 and 6 were interpolated as averages of the two neighboring segments.
V2b recruitment during locomotor behaviors
This broad connectivity of V2b neurons to different motor and premotor populations ideally places them to influence network output through multiple targets. To investigate the effects of V2b inhibition on motor networks, we next asked when V2b neurons are recruited. Previous studies show that loss of V2b neurons increased tail beat frequency during swimming, suggesting their involvement in speed control (Callahan et al., 2019). To probe whether V2b neurons only influence locomotor speed or also contribute to specific motor patterns, we decided to study recruitment of V2b neurons during naturalistic behaviors covering a range of locomotor speeds and patterns. We recorded activity of V2b neurons in the transgenic Tg(Gata3:lox-dsred-lox:GFP) line while simultaneously monitoring fictive motor activity from ventral nerve responses in paralyzed zebrafish larvae at 4–6 dpf (Fig. 6A). Across all the behaviors tested, approximately half of V2b neurons (20 out of 43 neurons, 46.5%) did not exhibit any spiking associated with fictive locomotion (Fig. 6B). To test whether neuronal birthdate could explain the observed variability in V2b activity, we analyzed the dorsoventral (D-V) positions of all recorded V2b neuron somata. In zebrafish axial networks, D-V positions of neuron somata have been shown to be indicators of their birthdate and recruitment at different locomotor speeds (McLean et al., 2007; Fetcho and McLean, 2010). Our analysis did not reveal any difference in D-V positions between the active and nonactive V2b neurons (data not shown). Spiking and nonspiking V2b neurons also did not differ significantly in their neurotransmitter phenotype and both mixed and glycinergic neurons were found to be spiking/nonspiking.
V2b neurons are recruited for locomotion and sensory stimuli. A, Schematic of the experimental setup showing simultaneous intracellular recording from single V2b neurons and extracellular recording of the motor nerve. B, Representative traces showing motor activity and corresponding V2b neuron activity recorded in cell-attached mode for two different neurons. Blue traces indicate the extracted motor signal from the ventral root (see Materials and Methods). Peaks in this signal corresponded to the midpoint of motor bursts. Inset, The first two motor bursts are enlarged to show timing of V2b spiking relative to the motor burst. C, Representative images showing tail movements for turns, swims, and escapes, respectively, in head-embedded animals. In separate experiments, ventral nerve recordings and V2b activity were simultaneously recorded while evoking turns, swims, or escapes. D, Box plot showing percentage of motor bursts associated with V2b spikes for visually evoked (left) and electrically evoked (right) movements. [p (swim, ipsi turns, contra turns) = 0.6177, Kruskal–Wallis test, N = 6 neurons, p (swim, escape) = 0.011, rank sum test, N = 7 neurons]. E, Representative traces showing sensory-evoked V2b activity prior to swims (left), escapes (middle), and no motor activity (right). F, Scatterplot showing latency of V2b spikes to the electrical stimulus across eight neurons. Black open circles mark the cells shown in E.
For the 23/43 V2b neurons that exhibited spiking during fictive locomotion, we characterized their recruitment further. Among these active V2b neurons, spiking rates were similar between cel-attached and whole-cell recordings, and hence these were pooled for analysis. To elicit different motor patterns, we delivered optomotor stimuli, in which moving visual gratings evoke forward swim movements, characterized by short motor busts on both sides, or turns to the left or right, characterized by longer motor bursts on one side (Fig. 6C). To elicit movements at different frequencies, we delivered an electrical stimulus to the head to drive escapes (>40 Hz) and swims (<40 Hz; Fig. 6C). We compared V2b recruitment in these behaviors by computing recruitment reliability, defined as the percentage of ventral nerve bursts with associated V2b spikes for each behavior.
In 6/23 cases, we recorded V2b neurons during forward swims and turning behaviors elicited by a visual stimulus. We found that V2b neurons were similarly recruited for forward swims and turns, with V2b spiking occurring in 29.0 ± 12.4% of ventral root bursts during forward swims, in 24.6 ± 12.2% of ipsilateral turns, and in 15.0 ± 14.1% of contralateral turns (N = 6 neurons; Kruskal–Wallis test, p > 0.05). In an additional 7/23 cases, we recorded V2b neurons during swims and high-frequency escapes elicited by an electric stimulus to the head. V2b neurons did not exhibit spiking during high-frequency escape bouts but were active during slow swims recorded within the same experiment (Fig. 6D, right; N = 7 neurons, Wilcoxon rank sum test, p < 0.01). In some of these experiments (2/7 V2b neurons), we were also able to elicit slow movements (<10 Hz) that resembled retreats (also known as struggles). Retreats are characterized by prolonged, high-amplitude motor bursts on both sides that progress from the tail toward the head. V2b neurons were recruited for these movements (data not shown), but because our single-ventral root monitoring did not allow us to definitively identify them as retreats, we did not include them in our analysis. In 10/23 cases, we obtained recordings during spontaneous swims; these data are analyzed for spike timing, below. These results taken together indicate that approximately half of V2b neurons are recruited for slow and medium speed forward swimming and turns, but not for escape behaviors.
Interestingly, although V2b neurons did not spike during the motor bursts associated with the escape response, we noticed V2b spiking prior to the start of the motor signal, at short latency after the electrical stimulation. This observation suggested that V2b neurons can be activated by sensory input itself, rather than being dependent on the motor control circuit. If so, we would expect that sensory-evoked V2b spiking would occur irrespective of the type of motor activity that follows and that the V2b action potential would occur at a short latency to the sensory stimulus. Indeed, we found V2b neurons that exhibited spiking in response to the sensory stimulus, prior to any motor signal associated with escapes or swims (N = 8 neurons; Fig. 6E, left and middle). In some cases, V2b spiking was elicited by the sensory stimulus even when there was no motor activity subsequently (Fig. 6E, right). Furthermore, these V2b spikes occurred at a short latency (4.2 ± 0.3 ms) to the electric stimulus (Fig. 6F), suggesting these responses could be directly evoked by sensory inputs. This sensory-evoked activity for V2b neurons has not been described previously.
Timing of V2b activity during fictive swimming
Prior work has shown that motor neurons receive substantial amounts of inhibition in phase with their peak activity (Berg et al., 2007; Kishore et al., 2014; Agha et al., 2024). V1 neurons contribute inhibition coincident with the late portion of the motor burst (Kimura and Higashijima, 2019), but the identity of the source of early in-phase inhibition is not known. To examine the temporal relationship of V2b activity to motor activity, we pooled the data from all 23 active V2b neurons and constructed a histogram of V2b spike times normalized to the middle of the motor burst in each swim cycle (Fig. 7A). We observed that most V2b spikes occurred at a phase of 0.7–1, coinciding with the rising phase of motor neuron spiking (N = 23 neurons). This was also evident from representative traces in Figure 6B, where V2b neuron spiking preceded the peak of the motor bursts. These results contrast with the reported recruitment of V1 neurons during the falling phase of the motor burst in 3 dpf zebrafish larvae (Kimura and Higashijima, 2019). To test whether V1 recruitment occurs similarly late in 4–6 dpf larvae, we measured the phase of V1 spiking in Tg(eng1b:lox-dsred-lox:DTA) larvae. Using a similar analysis, we found that indeed, for V1 neurons, most of their spikes occurred at a phase of 0–0.2, coinciding with the falling phase of the motor burst (Fig. 7B; N = 6 neurons), consistent with activity in the falling phase of motor activity. Furthermore, to assess phase relationships for individual neurons, we analyzed the average phase of each V2b and V1 neuron by constructing polar plots (Fig. 7C). Each dot in this plot represents the average phase of a single V2b or V1 neuron, and the population averages are shown as vectors. On average, V2b neurons are active approximately one-fourth cycle earlier than V1 neurons (p < 0.01, Watson–Williams test for circular statistics). These data indicate that, unlike V1 neurons, V2b neurons supply inhibition early, coinciding with the rising phase of the motor burst. These timing differences between these two ipsilateral inhibitory populations may relate to their observed differences in contributions to motor control (Callahan et al., 2019; Kimura and Higashijima, 2019).
V2b neurons are active earlier in the motor burst than V1 neurons. A, Histogram of V2b spike timing with respect to phase of motor activity. Inset, Illustration for timing and phase relation between successive motor bouts. N = 1,044 spikes recorded over 794 swim cycles from 23 neurons. B, Same as in E but for spiking of V1 neurons. N = 111 spikes recorded over 112 swim cycles from 6 neurons. C, Polar plot showing mean phase for V2b (black) and V1 (magenta) neurons and their population means (arrows; p = 2.92 × 10−5, Watson–Williams test for circular statistics, N = 23 V2b neurons and 6 V1 neurons).
Inactivity in V2b neurons is not due to lack of sensory feedback
In our experiments above, we found that approximately half of recorded V2b neurons were not active during any of the behaviors tested while some V2b neurons exhibited sensory-evoked activity. To investigate whether inactivity in V2b neurons was due to insufficient sensory feedback in the paralyzed preparation, we repeated similar experiments in unparalyzed larvae using calcium imaging of V2b neurons. A total of 4–6 dpf double transgenic Tg(elavl3:Hsa.H2B-GCaMP6s; Gata3:lox-dsred-lox:GFP) larvae were embedded in agarose for imaging spinal segments (5–10) using two-photon microscopy (Fig. 8A). A small portion of the tip of the tail was freed and tail movements recorded using a high-speed video camera. A visual stimulus was projected on a screen in front of the larvae to evoke swims. V2b neurons were identified post hoc and calcium activity during motor episodes was analyzed (Fig. 8B). Even with sensory feedback intact, the majority of V2b neurons did not show changes in calcium activity during motor episodes, in agreement with our physiology results from paralyzed animals (Fig. 8C). Five out of 20 neurons exhibited significantly higher activity after the start of swimming compared with baseline (Fig. 8C; N = 20 neurons from 5 larvae; Wilcoxon sign rank test, p < 0.01). These data indicate that a subgroup of V2b neurons are not recruited by typical movements and may influence motor control in more nuanced ways.
Calcium imaging in V2b neurons. A, Top, Schematic of the experimental setup showing simultaneous calcium imaging in V2b neurons and high-speed video recording of tail movements during optomotor responses. B, Representative anatomical scan of elavl3:GCaMP6s and Gata3:DsRed labels (top) and neurons identified for analysis in Suite2p (bottom). Scale bar, 15 µm. Different colors in the Suite2p image indicate distinct ROIs detected and analyzed. C, Heat map showing V2b neuron activity during swims. Traces are aligned to the onset of movement (red dashed line). N = 20 neurons (rows) recorded over 3–6 swim trials each from 5 larvae. For each cell, 2 s of the dF / F signal before and after the start of movement were compared. Red asterisks mark neurons that responded significantly to locomotion; four exhibited increases in activity, and one (Cell 6) was suppressed (Wilcoxon sign rank test). p = 0.0003 for Cell 6, 0.0000012 for Cell 11, 0.000005 for Cell 16, 0.008 for Cell 18, and 0.0004 for Cell 20.
Discussion
In this study we show that V2b neurons inhibit motor neurons and other prominent ventral horn spinal populations, including V1 and V2a neurons. This inhibition is spatially organized in the rostrocaudal axis, supplying long-range, gradually tapering inhibition to motor neurons and V2a neurons, but local inhibition to V1 neurons. Furthermore, V2b neurons are recruited for bilaterally symmetric and asymmetric movements at slow locomotor speeds and exhibit sensory-evoked activity independent of motor output. During fictive locomotion, V2b spiking occurs in-phase with the leading edge of each motor burst, but V2b activity may also be elicited directly by sensory stimuli. In conjunction with previous work, this pattern and timing of V2b inhibition suggest that these neurons may have a unique function in axial motor control, facilitating slow locomotor speeds by supplying early coincident inhibition to motor neurons.
Speed regulation versus pattern control
Behaviorally, movements can be defined based on the pattern of muscles activated (bilaterally symmetric or asymmetric) and their rhythm (slow or fast speed). Prior studies on limbed locomotion using modeling as well as experimental evidence have suggested that the different facets of behaviors, like the pattern (bilaterally symmetric or asymmetric) or rhythm (slow or fast), can be manipulated independently and therefore are controlled by different neuronal populations (Lafreniere-Roula and McCrea, 2005; McCrea and Rybak, 2008; Rybak et al., 2015). Recent work in zebrafish supported this theory by showing that the two subpopulations of excitatory V2a neurons, with descending and bifurcating axons, contribute differentially to motor rhythm and pattern, respectively (Menelaou and McLean, 2019). Bifurcating V2a neurons exhibited targeted connectivity (Menelaou and McLean, 2019; Agha et al., 2024) and higher recruitment for specific motor patterns, such as ipsilateral turns (Jay et al., 2023). In contrast, descending V2a neurons exhibited more general connectivity (Menelaou and McLean, 2019) and were direction agnostic, appropriate for their role in rhythm control (Jay et al., 2023). Consistent with this two-layer model of rhythm and pattern control, our data suggests that V2b neurons are important regulators of rhythm. V2b neurons, which have exclusively descending axons, exhibited general connectivity to interneurons and were recruited for slow swim speeds, irrespective of the motor pattern (symmetric and asymmetric). Moreover, previous studies from our lab have shown that optogenetic silencing of V2b neurons resulted in faster tail beat frequency (Callahan et al., 2019). Together, these data suggest that V2b inhibition works to facilitate slow locomotor speeds in axial motor networks.
Local reciprocal inhibition of V1 neurons
We previously demonstrated that V1 neurons inhibit the V2b population (Sengupta et al., 2021). Together with the V2b connectivity presented here, we conclude that V1 and V2b neurons reciprocally inhibit each other. Because our mapping was done at a population level, it does not address whether single V1/V2b pairs show reciprocal inhibition. Nonetheless, this circuit motif closely resembles reciprocal inhibition underlying flexor-extensor modules. Mutual inhibition between Ia inhibitory interneurons has been reported in several species (Baldissera et al., 1987; Pratt and Jordan, 1987b; Jankowska, 1992; Wilson et al., 2010b) and is thought to be a crucial component for enforcing flexor-extensor alternation (McCrea and Rybak, 2008). Though Ia inhibitory interneurons are composed of a mixture of V1 and V2b populations, V1/V2b reciprocal innervation has so far not been systematically analyzed. We find that V1 and V2b neurons inhibit each other locally, up to three segments away. Specifically, V2b neurons inhibit V1 neurons located 2–3 segments caudally (this study), whereas V1 neurons inhibit V2b neurons located 2–3 segments rostrally (Sengupta et al., 2021).
Interestingly, this organization is consistent with the rostrocaudal structure of hindlimb flexor-extensor circuits. V1 and V2b neurons exhibit biased connectivity to flexor and extensor motor neurons, respectively (Zhang et al., 2014; Britz et al., 2015). As the peak of flexor motor neuron distribution is in L2 (segment lumbar 2), while the peak of extensor motor neuron distribution is in L5, reciprocal inhibition between V1 and V2b neurons is predicted to also occur within ∼3 spinal segments in limbed networks, similar to our mapped connectivity in zebrafish (Kudo et al., 1987; Britz et al., 2015). The origin of limb networks is debated. Some evidence suggests that tetrapod locomotion evolved from undulatory swimming and hence networks for limb control originated through gradual modification of circuits regulating axial musculature (Grillner, 2006; Jung and Dasen, 2015; Machado et al., 2015; Jung et al., 2018). Our results showing that V1/V2b reciprocally inhibitory circuits exist in axial networks may provide evidence of an ancestral circuit motif that could have been co-opted for flexor-extensor control.
Complementary roles of V1/V2b inhibition in speed control
What is the role of V1 and V2b reciprocal inhibition in axial control? Genetic knock-out of V1 neurons in zebrafish slowed down locomotion (Kimura and Higashijima, 2019) whereas optogenetic suppression of V2b neurons resulted in faster locomotor speeds (Callahan et al., 2019). These data suggest, in axial networks, V1 and V2b neurons regulate locomotor speed in opposing directions. Our current and prior findings support these complementary functions in speed control. Both V1 and V2b neurons inhibit slow and fast motor neurons. However, V2b inhibition onto primary motor neurons (average charge transfer over the first three segments, 11.4 fC) is stronger than V1 inhibition onto the same targets (3.8 fC; Sengupta et al., 2021). In contrast, V1 inhibition onto secondary motor neurons (13.3 fC) is stronger than V2b inhibition onto the same targets (10.7 fC). These results are consistent with a proposed model in which V1 neurons shut down slow MNs, facilitating fast locomotor speeds (Kimura and Higashijima, 2019). This biased connectivity along with V1/V2b reciprocal inhibition suggests this motif could have been utilized for implementing different speeds of locomotion in axial networks.
In addition to biases in connectivity, the complementary functions of V1/V2b neurons are supported by their distinct timing of activity during motor bursts. V2b neurons supply on-cycle inhibition early in the motor burst, while V1 neurons are active late in the motor burst (Fig. 6F,G; Kimura and Higashijima, 2019). Interestingly, this timing of V2b activity is similar to the timing of descending V2a neurons, which are thought to regulate locomotor rhythm (Jay et al., 2023; Agha et al., 2024). Inhibition coincident with excitation is widespread in the cerebral cortex, where it can influence multiple network parameters like synaptic gain, dynamic range, sharpening of sensory tuning, and spike synchrony (Isaacson and Scanziani, 2011). Motor neurons have been shown to receive inhibition during the motor burst, with an effect on synaptic gain (Berg et al., 2007; Vestergaard and Berg, 2015), but the source of this inhibition is unknown. Our data suggests that V2b neurons are the source of early coincident inhibition onto motor neurons. We hypothesize that V2b neurons might promote slower locomotion by two potential mechanisms. First, V2b neurons could slow locomotor speed by acting through V2a neurons, either directly inhibiting V2a spiking or shunting V2a excitation of motor neurons. Second, V2b neurons could be acting via V1 neurons to disinhibit motor neurons and increase motor burst width, consequently slowing down locomotor speed. Future analysis of motor neuron spiking during locomotion with or without V2b neurons will help tease out the mechanism of V2b function.
Future questions about V2b neurons
Recent evidence from transcriptional profiling as well as anatomical and functional properties has shown significant heterogeneity within each spinal interneuron population, in some cases set up by a sequential developmental program (McLean and Fetcho, 2009; Sengupta and Bagnall, 2023; Deska-Gauthier et al., 2024; Worthy et al., 2024a; Roome et al., 2025). We also observed heterogeneity within the V2b population. Approximately half of the V2b neurons in our dataset did not show any recruitment during escapes, forward swims, or turns. The fact that V2b neurons were also only intermittently active during calcium imaging experiments suggests that the inactivity of V2b neurons was not due to a lack of sensory feedback to these circuits. It remains unclear what circumstances recruit the nonactive V2b neurons. Furthermore, a subset of V2b neurons appear to exhibit sensory-evoked activity, occurring prior to and apparently independent of any motor output (Fig. 6). Recent work characterizing sensory input to spinal cord neurons has identified specific mapping from particular afferent types to dorsal horn interneurons (Gatto et al., 2021; Gradwell et al., 2024; Koch et al., 2025) but comparatively less attention has been paid to potential sensory inputs to canonically “motor” elements of the ventral horn. V2b neurons receive anatomical contacts from proprioceptive afferents in mouse (Zhang et al., 2014), and our study is the first to show putative physiological synaptic connections from somatosensory afferents to V2b neurons. It will be important to understand whether these are indeed direct monosynaptic inputs, and if so how those participate in locomotor control.
Overall, this study revealed functional heterogeneity among V2b neurons, and this heterogeneity is likely to increase even further in limbed animals, where diversification of interneuron pools seems to play a role in more complex motor control (Trevisan et al., 2024; Vijatovic et al., 2024; Worthy et al., 2024b). In this study, we mapped V2b output and activity. In future it will be helpful to map inputs to V2b neurons to fully understand the importance of different V2b subsets and their functions.
Footnotes
This work was supported by R01NS130483 (M.W.B), a McKnight Scholar Award (M.W.B.), and the McDonnell Center for Cellular and Molecular Neurobiology Postdoctoral Fellowship 2021 (M.S.). All microscopy experiments were performed in part, through the use of Washington University Center for Cellular Imaging (WUCCI) supported by Washington University School of Medicine, The Children’s Discovery Institute of Washington University and St. Louis Children’s Hospital (CDI-CORE-2015-505 and CDI-CORE-2019-813) and the Foundation for Barnes-Jewish Hospital (3770 and 4642).
The authors declare no competing financial interests.
- Correspondence should be addressed to Mohini Sengupta at mohini.sengupta{at}slu.edu or Martha W. Bagnall at bagnall{at}wustl.edu.
