1. Neuroscience
Download icon

Mechanosensory neurons control the timing of spinal microcircuit selection during locomotion

  1. Steven Knafo
  2. Kevin Fidelin
  3. Andrew Prendergast
  4. Po-En Brian Tseng
  5. Alexandre Parrin
  6. Charles Dickey
  7. Urs Lucas Böhm
  8. Sophie Nunes Figueiredo
  9. Olivier Thouvenin
  10. Hugues Pascal-Moussellard
  11. Claire Wyart Is a corresponding author
  1. Sorbonne Universités, UPMC Univ Paris 06, Inserm, CNRS, Institut du Cerveau et la Moelle épinière, Hôpital Pitié-Salpêtrière, France
  2. Hôpital Pitié-Salpêtrière, France
Research Article
Cited
0
Views
1,482
Comments
0
Cite as: eLife 2017;6:e25260 doi: 10.7554/eLife.25260

Abstract

Despite numerous physiological studies about reflexes in the spinal cord, the contribution of mechanosensory feedback to active locomotion and the nature of underlying spinal circuits remains elusive. Here we investigate how mechanosensory feedback shapes active locomotion in a genetic model organism exhibiting simple locomotion—the zebrafish larva. We show that mechanosensory feedback enhances the recruitment of motor pools during active locomotion. Furthermore, we demonstrate that inputs from mechanosensory neurons increase locomotor speed by prolonging fast swimming at the expense of slow swimming during stereotyped acoustic escape responses. This effect could be mediated by distinct mechanosensory neurons. In the spinal cord, we show that connections compatible with monosynaptic inputs from mechanosensory Rohon-Beard neurons onto ipsilateral V2a interneurons selectively recruited at high speed can contribute to the observed enhancement of speed. Altogether, our study reveals the basic principles and a circuit diagram enabling speed modulation by mechanosensory feedback in the vertebrate spinal cord.

https://doi.org/10.7554/eLife.25260.001

Introduction

The generation of complex motor behaviors at a given speed relies on the sequential selection and orderly activation of groups of muscles by the nervous system (Bellardita and Kiehn, 2015). At the level of locomotor circuits in the spinal cord, premotor excitatory V2a interneurons and motor neurons are recruited in a frequency-dependent manner (Crone et al., 2008; Kimura et al., 2006; McLean et al., 2007), a feature thought to underlie speed control in zebrafish (McLean et al., 2008). Recently, circuit-mapping experiments revealed preferential connectivity profiles between identified groups of V2a interneurons and motor neurons suggesting that distinct microcircuit modules could regulate the mobilization of individual muscles during locomotion (Ampatzis et al., 2014; Bagnall and McLean, 2014). Genetic ablation of V2a interneurons in mammals is associated with deficits in left-right coordination and gait transitions suggesting that V2a interneurons play a key role in controlling speed across vertebrates (Crone et al., 2008). The nature of spinal circuits involved in motor pattern generation and speed control has mainly been studied during fictive locomotion when sensory feedback is absent (Kiehn, 2016; McLean and Dougherty, 2015). However, less is known about the cellular and circuit mechanisms regulating the timing of microcircuit selection in the spinal cord during movement, which incorporate proprioceptive feedback from muscle contraction.

Peripheral mechanosensory feedback plays a critical role in setting the timing of muscle activation during movement (Windhorst, 2007). In particular, stimulation of afferent sensory neurons in decerebrate and paralyzed cats can reorganize the pattern of motor output during fictive locomotion, either by entraining or resetting the locomotor rhythm, depending on the nature of active spinal circuits (Conway et al., 1987; Guertin et al., 1995; Jankowska, 1992; McCrea, 2001; Rossignol et al., 2006). In addition, altering the development of peripheral sensory neurons leads to dramatic perturbations of motor sequences during walking and swimming in mice, suggesting that mechanosensory circuits can directly control the activity of spinal central pattern generators responsible for rhythmic motor outputs (Akay et al., 2014; Takeoka et al., 2014). However, the molecular identity of spinal neurons mediating these effects in mammals remains elusive.

To understand the contribution of mechanosensory feedback to speed control, we investigated the role of spinal mechanosensory neurons during acoustic escape responses in the zebrafish larva, a genetic model system with tractable locomotor behaviors that can be analyzed using fine kinematic analysis. We developed a bioluminescence-based assay to monitor the activity of populations of spinal motor neurons in moving animals and combined this approach with calcium imaging and quantitative kinematic analysis of locomotor behaviors to compare the dynamics of motor activity in active and fictive swimming. We demonstrate that glutamatergic mechanosensory neurons enhance the recruitment of spinal motor neurons during motion by increasing locomotor frequency, and consequently, the speed of locomotion. In particular, we show that mechanosensory feedback selectively enhances the fast regime over the slow regime during escape responses, thereby promoting the transition between fast and slow locomotion. Using optogenetics for circuit connectivity mapping, we identified a functional microcircuit module formed by mechanosensory Rohon-Beard neurons onto ipsilateral glutamatergic V2a interneurons selectively active during fast locomotion. Altogether our results reveal how spinal microcircuits controlling locomotor frequency can be selected by mechanosensory pathways to enhance speed during active locomotion.

Results

Monitoring the recruitment of spinal motor neurons in moving animals

In order to investigate the contribution of sensorimotor circuits during locomotor activity in zebrafish larvae, we used the bioluminescent calcium sensor GFP-Aequorin to monitor calcium activity emitted by identified neuronal populations in freely moving animals (Baubet et al., 2000; Naumann et al., 2010; Shimomura et al., 1962). To maximize expression of GFP-Aequorin in zebrafish motor circuits, we codon-optimized the Aequorin sequence and targeted its expression to spinal motor neurons. We recorded bioluminescence signals emitted from spinal motor neurons during active behaviors elicited by an acoustic stimulus in Tg(mnx1:gal4;UAS:GFP-aequorin-opt) zebrafish larvae at four days post-fertilization (dpf) (Figure 1A–B, see Materials and methods). To remove the concern of collecting unspecific signals following large changes in intracellular calcium concentration in muscle fibers during tail contractions, we systematically determined both in live animals and after GFP immunostainings that there was no expression in non-neuronal tissues such as muscles (Figure 1B, n = 5 larvae, Video 1). Acoustic stimulations triggered stereotypical escape responses characterized by an asymmetrical initial C-bend followed by fast swimming. In rare cases, possibly due to coelenterazine application, acoustic stimulations led to stereotypical slow swims characterized by symmetrical bends with undulations propagating from rostral to caudal along the tail (7.4% of trials, 21 out of 283 stimulations; see Mirat et al. [2013]) (Figure 1C–D, Videos 23, n = 10 larvae, see Budick and O'Malley, 2000). Each locomotor behavior was associated with a specific bioluminescence signal (Figure 1D, Videos 23). While time-to-peak and time decay of bioluminescence signals were invariant across maneuvers (data not shown), stereotypical escape responses generated larger bioluminescence amplitudes than slow swims (Figure 1D–E, mean bioluminescence amplitude = 30.8 + /- 1.4 photons/10 ms for escapes and 4.1 ± 1.6 photons/10 ms for slow swimming, p<0.001). Bioluminescence signal amplitude correlated with the maximal angle of the tail bend (Figure 1F, R = 0.4, p<0.001), suggesting that spinal motor neuron recruitment is increased during behaviors with larger tail bends and faster locomotor frequencies. However, monitoring global bioluminescence signals lacks the single cell resolution required to explain whether these increased bioluminescence signals result from the recruitment of additional motor neurons during fast locomotion or from increased signals in the same pool of active cells. To decipher between these possibilities we recorded the activity of pools of motor neurons using the genetically-encoded calcium indicator GCaMP6f during spontaneous fictive slow swimming and elicited fictive escape responses in Tg(mnx1:gal4;UAS:GCaMP6f,cryaa:mCherry) (Figure 2A–B, Böhm et al., 2016). This strategy allowed us to monitor dynamics of both individual neurons and populations of neurons during locomotor behaviors previously probed using bioluminescence. At 4 dpf, fictive locomotor burst frequencies ranged between 20–30 Hz for slow swims (left panel of Figure 2C, see Masino and Fetcho [2005]) and 20–80 Hz for escapes (right panel of Figure 2C). During slow swimming, a small fraction of ventral motor neurons was active (23.2%, left panel of Figure 2D, Figure 2E, Video 4). In contrast, the majority of motor neurons, including large dorsal motor neurons, was recruited during escapes (88.4%, right panel of Figure 2D, Figure 2E, and Video 5). Dorsalmost motor neurons showed larger calcium transients than ventral ones (Figure 2D), and the averaged ΔF/F across all recruited motor neurons was higher during escapes compared to slow swims (Figure 2F, mean ΔF/F/active cell = 91.1 + /- 4.7% during escape responses and 25.9 ± 3.8% during slow swimming, p<0.001, n = 10 larvae). These data revealed that the generation of fast escape responses requires the recruitment of more spinal motor neurons compared to slow swimming and that the amplitude of calcium signals are higher in cells active during escapes compared to spontaneous swimming. Therefore, we conclude that high bioluminescence signal amplitude during escape responses compared to slow swims likely results from the combination of two effects: a larger pool of active cells driven by an increase in swimming frequency and larger calcium transients per active neuron.

Amplitude of bioluminescence signals in spinal motor neurons correlates with the type of locomotor maneuver during active swimming.

(A) Bioluminescence signals emitted from spinal motor neurons in Tg(mnx1:gal4;UAS:GFP-aequorin-opt) zebrafish larvae at 4 dpf were recorded using a photomultiplier tube under infrared illumination during active behaviors elicited by an acoustic stimulus. (B) Live fluorescent image (upper panel) and immunohistochemistry for GFP (lower panel) in a 4 dpf Tg(mnx1:gal4;UAS:GFP-aequorin-opt) zebrafish larva showing selective expression in spinal motor neurons (arrowhead: dorsal primary, arrow: ventral secondary motor neurons), and strictly no expression in muscle fibers (n = 5). (C) Motor behaviors elicited by acoustic stimuli. Superimposed traces illustrate the amplitude of tail contractions over time, for each behavior. Traces are color-coded according to the delay from stimulus onset. Automated categorization classified maneuvers into escapes (n = 245/283) or swims (n = 21/283) (n = 10 larvae and 300 trials). (D) Example traces of typical bioluminescence signals and kinematic parameters observed for each category. (E) Mean bioluminescence amplitude was higher for escapes (28.4 ± 0.9 photons/10 ms; normalized amplitude per larva = 0.41 + /- 0.18) than swims (3.9 ± 1.4 photons/10 ms, p<0.001; normalized amplitude = 0.06 + /- 0.02, p<0.001). (F) Correlation between bioluminescence signal amplitude and maximum tail angle amplitude (R = 0.4, p<0.001).

https://doi.org/10.7554/eLife.25260.002
Calcium imaging in paralyzed larvae confirms a larger recruitment of spinal motor neurons for fast escapes than for slow swimming.

(A) Calcium transients emitted from spinal motor neurons in Tg(mnx1:gal4;UAS:GCaMP6f,cryaa:mCherry) zebrafish larvae at 4 dpf were recorded simultaneously with ventral nerve root recordings (VNR) during fictive behaviors elicited by a water puff to the otic vesicle. (B) Expression pattern in a Tg(mnx1:gal4;UAS:GCaMP6f,cryaa:mCherry) larva at 4 dpf (R is rostral, V is ventral; scale bar is 50 µm). (C) VNR for each fictive behavior illustrates typical spontaneous fictive slow swims and induced fictive escapes: burst frequencies ranged between 20–30 Hz for slow swims (left panel) and 20–80 Hz for escapes (right panel). (D) GCaMP6f signals from individual motor neurons during spontaneous slow swimming (4 swims, left panel) and during an evoked escape response (right panel). For each recording, out-of-focus light was estimated within the spinal cord (black trace) and used as a criterion to determine the active vs inactive status of each cell. Pie charts represent the proportion of active cells in each behavior: 16/69 cells across 27 swims versus 61/69 cells across 12 escapes (n = 3 larvae). (E) Dorso-ventral (D–V) position of cells recruited during each maneuver shows dorsal motor neurons only active during escapes (the dorso-ventral axis within the spinal cord is normalized to 0 at the ventral limit and 1 the dorsal limit; mean D-V position for escapes = 0.41 + /- 0.02 versus 0.25 ± 0.01, p<0.001, n = 78 cells in n = 3 larvae). (F) Mean ΔF/F amplitude was higher during escapes compared to spontaneous swims across larvae (91.2 ± 4.7% versus 25.9 ± 3.8%, p<0.001, 12 escapes and 27 swims in n = 3 larvae).

https://doi.org/10.7554/eLife.25260.003
Video 1
Expression of GFP in a 4 dpf Tg(mnx1:gal4; UAS:GFP-aequorin-opt) larva revealed by immunohistochemistry (one example shown out of the n = 5 larvae where IHC was performed).
https://doi.org/10.7554/eLife.25260.004
Video 2
Bioluminescence signal emitted by a 4 dpf Tg(mnx1:gal4; UAS:GFP-aequorin-opt) free-tailed transgenic larva and corresponding kinematics during escape.
https://doi.org/10.7554/eLife.25260.005
Video 3
Bioluminescence signal emitted by a 4 dpf Tg(mnx1:gal4; UAS:GFP-aequorin-opt) free-tailed transgenic larva and corresponding kinematics during swimming.
https://doi.org/10.7554/eLife.25260.006
Calcium imaging of spinal motor neurons in a 4 dpf Tg(mnx1:gal4; UAS:GCaMP6f) larva during fictive slow swimming.

Animals were paralyzed with bungarotoxin injections. A small fraction of ventral motor neurons is active. Recordings were acquired at 20 Hz. Movies were sped up by 2.5 X.

https://doi.org/10.7554/eLife.25260.007
Video 5
Calcium imaging of spinal motor neurons in a 4 dpf Tg(mnx1:gal4; UAS:GCaMP6f) larva during a fictive escape response.

Animals were paralyzed with bungarotoxin injections. A large portion of both ventral and dorsal motor neurons is recruited. Recordings were acquired at 20 Hz. Movies were sped up by 2.5 X.

https://doi.org/10.7554/eLife.25260.008

Mechanosensory feedback enhances the recruitment of spinal motor neurons

To test whether mechanosensory feedback modulates the global recruitment of spinal motor neurons during locomotion, we compared motor dynamics in active and fictive escape responses by recording bioluminescence before and after blocking muscle contraction following the bath application of cholinergic blocker pancuronium bromide (Figure 3A, Panier et al., 2013). Across all Tg(mnx1:gal4;UAS:GFP-aequorin-opt) larvae, mean bioluminescence amplitude was markedly decreased in paralyzed compared to motile animals (Figure 3B–C, signal amplitude before paralysis = 26.6 + /- 0.9 photons/10 ms; signal amplitude after paralysis = 11.1 + /- 0.4 photons/10 ms, p<0.001, n = 10 larvae). Within each larva, the mean normalized bioluminescence amplitude was decreased seven-fold (0.056 ± 0.08 versus 0.36 ± 0.18, p<0.001). Since this effect could be partly explained by the inhibition of acetylcholine receptors located on targets of reticulospinal neurons, such as the Mauthner cell (Koyama et al., 2011), we conducted the same experiments in immotile relaxed (cacnb1ts25) mutant zebrafish (Granato et al., 1996; Zhou et al., 2006). The mean bioluminescence amplitude was markedly decreased in immotile Tg(cacnb1ts25/ts25; mnx1:gal4; UAS:GFP-aequorin-opt) larvae compared to motile Tg(mnx1:gal4, UAS:GFP-aequorin-opt) control siblings (Figure 3D–E, mean signal amplitude in motile siblings = 37.6 + /- 1.4 photons/10 ms; mean signal amplitude in immotile mutants = 9.8 + /- 0.4 photons/10 ms, p<0.001, n = 3 larvae). Altogether these results demonstrate that the recruitment of spinal motor neurons is enhanced during active locomotion, indicating that mechanosensory feedback pathways recruited during muscle contraction contributes to increasing spinal motor output. However, from these measurements it is unclear whether the decrease in signal amplitude observed in Aequorin-expressing larvae when muscle contraction is impaired is indicative of a reduction in amplitude or frequency of motor output during acoustic escapes responses.

Mechanosensory feedback enhances the recruitment of spinal motor neurons during active locomotion.

(A) To compare the recruitment of spinal motor neurons when mechanosensory feedback was present (‘active locomotion’) or suppressed (‘fictive locomotion’), we conducted bioluminescence assays in 4 dpf Tg(mnx1:gal4;UAS:GFP-aequorin-opt) zebrafish larvae before and after paralysis using pancuronium bromide, and in immotile 4 dpf Tg(mnx1:gal4;UAS:GFP-aequorin-opt;cacnb1ts25/ts25) mutants compared to their motile siblings. (B) Bioluminescence signals from motor neurons revealed a marked decrease in bioluminescence amplitude after paralysis. (C) Quantification of the change in mean bioluminescence amplitude (before paralysis: 26.6 ± 0.9 photons/10 ms; after paralysis: 11.1 ± 0.4 photons/10 ms, n = 10 larvae in each group, 30 trials per larva, p<0.001). (D) Similarly, averaged bioluminescence signals from motor neurons were markedly decreased in immotile Tg(mnx1:gal4;UAS:GFP-aequorin-opt;cacnb1ts25/ts25,) mutant larvae when compared with motile siblings. (E) Mean bioluminescence amplitude in motile siblings (37.6 ± 1.4 photons/10 ms) compared to immotile mutants (9.8 ± 0.4 photons/10 ms, n = 300 trials in 10 larvae for each group, p<0.001).

https://doi.org/10.7554/eLife.25260.009

Silencing mechanosensory neurons decreases the speed of locomotion

To decipher between these possibilities, we probed the contribution of mechanosensory neurons during active swimming through fine kinematic analysis of locomotor behaviors in freely moving animals. We expressed the zebrafish-optimized Botulinum toxin under the control of the isl2b promoter (Auer et al., 2015; Böhm et al., 2016; Sternberg et al., 2016) (see Materials and methods and Video 6) in order to block synaptic release in glutamatergic mechanosensory neurons, namely Rohon-Beard neurons (RB) and dorsal root ganglia (DRG) in the spinal cord and trigeminal ganglia (Figure 4A, note the bundle of RB and DRG axons ascending in the spinal cord in Figure 4B, (Won et al., 2011)). We performed quantitative kinematic analysis of locomotor behaviors in freely-swimming Tg(isl2b:gal4,cmlc2:eGFP;UAS:BoTxBLC-GFP) larvae and found alteration of several kinematic parameters compared to control siblings (Figure 4C–D, n = 304 larvae, 890 escapes). Distance traveled (Figure 4E1) and total number of oscillations (Figure 4E2) were unaffected in animals deprived of mechanosensory feedback. However, we noted that tail-beat frequency (TBF) (Figure 4E3) and, accordingly, speed (Figure 4E4) were markedly reduced in BoTxBLC-GFP+ animals. Coincident with the reduction in TBF and stability of number of oscillations, the escape duration increased in larvae deprived of mechanosensory feedback (Figure 4E5, from 168.1 ± 2.4 ms to 182.0 ± 3 ms). While the amplitude of the first C-bend was not affected in BoTxBLC-GFP+ animals (Figure 4—figure supplement 1), the amplitudes of subsequent tail bends were increased. This was especially pronounced on the ipsilateral side, suggesting a modulation of motor circuits recruited after the response onset. However, from this global analysis it is unclear if the effects on locomotor frequency and speed by mechanosensory feedback reflect a specific modulation of fast and/or slow swimming.

Figure 4 with 1 supplement see all
Silencing mechanosensory neurons decreases speed of locomotion.

(A) In vivo fluorescence image and (B) immunohistochemistry for GFP in 4 dpf Tg(isl2b:gal4,cmlc2:eGFP;UAS:GFP-aequorin-opt) double transgenic zebrafish larva show selective expression of GFP-aequorin in mechanosensory neurons: trigeminal ganglia (arrow), Rohon-Beard spinal neurons and their ascending axons (arrowhead) and dorsal root ganglia (*), but no expression in muscle fibers (n = 4). (C–D) Superimposed images from the raw high-speed camera recording (left), the corresponding automated tracking (middle, pink triangles represent the head and green lines the tail of the fish) and tail angle over time extracted from the tracking analysis (right; stimulus is represented by the black vertical line, duration of the event is the black horizontal line) of a typical escape elicited by an acoustic stimulus in a freely-swimming control sibling larva at 6 dpf (C), and in a Tg(isl2b:gal4,cmlc2:eGFP;UAS:BoTxBLC-GFP) larva (D). (E1) Distance travelled was unchanged in BoTxBLC+ and control larvae (10.2 ± 0.2 mm versus 10.6 ± 0.2 mm, p=0.1). (E2) Number of oscillations was unchanged in BoTxBLC+ and control larvae (6.0 ± 0.2 versus n = 6.1 + /- 0.2 oscillations, p=0.6). (E3) BoTxBLC+ larvae showed a decreased tail-beat frequency (TBF, 34.2 ± 0.4 Hz versus 37.8 ± 0.3 Hz, p<0.001). (E4) BoTxBLC+ larvae showed a decreased speed of escape responses (56.9 ± 1.1 versus 64.1 ± 0.9 mm/s, p<0.001). (E5) Escape duration was increased in BoTxBLC+ larvae compared to control siblings (182.0 ± 3.0 ms versus 168.1 ± 2.4 ms, p<0.001). (For all parameters: control group: n = 176 larvae from 8 clutches, n = 561 escapes; BoTxBLC+ group: n = 128 larvae from 8 clutches, n = 329 escapes).

https://doi.org/10.7554/eLife.25260.010
Video 6
Expression of GFP in a 4 dpf Tg(isl2b:gal4,cmcl2:eGFP;UAS:GFP-aequorin-opt) larva revealed by immunohistochemistry (one example shown out of the n = 4 larvae where IHC was performed).
https://doi.org/10.7554/eLife.25260.012

To determine if mechanosensory feedback selectively modulates fast versus slow locomotion in our silencing experiments, we separated the fast from the slow locomotor regime in elicited escape responses by applying a cutoff at 30 Hz, a threshold that defines the recruitment of fast motor neurons during swimming (McLean et al., 2008; Severi et al., 2014) (Figure 5A–B). This analysis revealed a decrease of the distance travelled, duration, number of cycles, speed, and tail-beat frequency of the fast component in BoTxBLC-GFP+ compared to wild type siblings (Figure 5C1–C5). In contrast, there was an increase of the distance travelled, duration and number of cycles of the slow component in BoTxBLC-GFP+ while the tail-beat frequency and speed were not altered (Figure 5D1–D5). Remarkably, the total number of cycles remains constant in BoTxBLC-GFP+ compared to wild type siblings (Figure 4E2), indicating that the decrease of fast swimming frequencies observed in BoTxBLC-GFP+ larvae leads to a decrease of the fast component duration and induces an increase of the slow component duration. Therefore, the reduction of speed and locomotor frequency observed in mechanosensory-deprived animal is specifically restricted to the fast component but not the slow one (Figure 5A–B). Altogether, our results demonstrate that mechanosensory feedback enhances speed of locomotion by promoting the transition between fast and slow locomotion during escape responses. Due to the fact that the Gal4/UAS combinatorial expression system labels cells in a variegated manner in zebrafish, we only observed BoTxBLC-GFP in half the RB neurons although the isl2b promoter drives expression in all RB neurons, dorsal root ganglia and trigeminal neurons. This observation suggests that the effects of silencing mechanosensory feedback are likely underestimated in our behavior experiments.

Silencing mechanosensory neurons accelerates the transition between fast and slow locomotion during acoustic escape responses.

(A–B) We separately analyzed the fast component of the escape response (cycles with TBF >30 Hz) and the slow component (cycles with TBF ≤30 Hz) as shown for control (n = 176 larvae from 8 clutches, n = 561 escapes) and BoTxBLC+ larvae (n = 128 larvae from 8 clutches, n = 329 escapes). (C1–C5) Within the fast regime of the escape, silencing mechanosensory feedback reduces the distance travelled in BoTxBLC+ larvae compared to control siblings (C1, 5.9 ± 0.2 versus 7.4 ± 0.2 mm, p<0.001), duration (C2, 60.0 ± 2.1 versus 67.8 ± 1.9 ms, p=0.005), number of oscillations (C3, 2.9 ± 0.1 versus 3.6 ± 0.1 oscillations, p<0.001), speed (C4, 104.3 ± 1.9 versus 112.6 ± 1.6 mm/s, p=0.001), and TBF (C5, 45.8 ± 0.4 versus 48.4 ± 0.3 Hz, p<0.001). (D1–D5) Within the slow component of the escape, silencing mechanosensory feedback in BoTxBLC+ larvae increases the distance travelled (D1, 3.5 ± 0.1 versus 2.4 ± 0.1 mm, p<0.001), duration (D2, 140.9 ± 3.4 versus 97.1 ± 3.2 ms, p<0.001, p<0.001) and number of oscillations (D3, 2.8 ± 0.1 versus 1.9 ± 0.1, p<0.001) but has no effect on speed (D4, 24.9 ± 0.4 versus 24.6 ± 0.4 mm/s, p=0.5) and tail-beat frequency (D5, 25.3 ± 0.1 versus 25.4 ± 0.1 Hz, p=0.5). For all parameters: n = 304 larvae, number of fast components = 1013; number of slow components = 1265.

https://doi.org/10.7554/eLife.25260.013

Mechanosensory neurons synapse onto V2a interneurons involved in fast locomotion

Given the selective effect on tail-beat frequency and speed during the fast component of escape responses following the silencing of mechanosensory feedback, we reasoned that mechanosensory neurons may be able to selectively recruit spinal interneurons involved in fast locomotion. Previous studies in zebrafish larvae showed that spinal V2a interneurons are recruited in a topographic and frequency-dependent manner during fictive locomotion, with dorsalmost V2a interneurons solely active during fast locomotion (McLean et al., 2008, 2007). Ablation, silencing and optogenetic manipulations revealed that these neurons constitute an important class of excitatory interneurons driving the recruitment of motor neurons—especially during fast locomotion (Ampatzis et al., 2014; Bagnall and McLean, 2014; Crone et al., 2008; Kimura et al., 2006, 2013; Ljunggren et al., 2014; McLean et al., 2007). Interestingly, we noticed patterns of perisomatic innervation of dorsal V2a cell bodies by axonal projections originating from one cell type targeted by the isl2b driver, Rohon-Beard cells (Figure 6A–D, mean soma position of V2a = 0.72 + - 0.04, n = 12 cells in 4 larvae).

Axons of isl2b+ Rohon-Beard project onto dorsalmost chx10+ V2a interneurons in the spinal cord.

(A) Z-projection stack of the spinal cord imaged from the lateral side in a 4 dpf Tg(isl2b:gal4,cmcl2:eGFP;UAS:GFP-aequorin;chx10:DsRed) triple transgenic larva. White dashed lines delineate ventral and dorsal limits of the spinal cord; these limits define the dorso-ventral (D/V) axis from 0 to 1. The axon bundle from isl2b+ Dorsal Root Ganglia (DRG) and Rohon-Beard (RB) neurons contact the soma of dorsal chx10+ V2a interneurons. The white square is enlarged in the right panel to stress the anatomical connections (arrow). (B) Dendrites of dorsal V2a interneurons are targeted by axonal processes from isl2b+ sensory neurons (arrows). (C) D/V positions of V2a neurons receiving putative inputs from isl2b+ cells (mean D/V position = 0.72 ± 0.04. N = 12 cells in n = 4 larvae). (D) Profile of expression of 4 dpf Tg(isl2b:gal4,cmcl2:eGFP;chx10:DsRed) injected with the Tg(UAS:CoChR-tdTomato) construct reveals projections from single Rohon Beard cells onto chx10+ dorsal V2a interneurons. Scale bars in (A) and (D) are 50 µm and is 10 µm in (B). For each panel rostral side (R) is on the left and ventral side (V) is at the bottom.

https://doi.org/10.7554/eLife.25260.014

In order to test whether Rohon-Beard neurons form functional synapses onto V2a interneurons, we performed optogenetic-mediated mapping of synaptic connectivity in the spinal cord (Figure 6D, Figure 7D). Rohon-Beard cells are low input resistance neurons and required current injections ranging from 200 to 700 pA to generate single spikes (Figure 7B, n = 6 cells, 250 ms injection steps). In comparison, dorsalmost V2a neurons only required from 15 to 50 pA to fire action potentials (Figure 7H, n = 5 cells). As a consequence, RBs might not be amenable to optogenetic stimulations at larval stages using first generation opsins such as ChR2-H134R (Boyden et al., 2005). To circumvent this potential issue we expressed the novel opsin CoChR-tdTomato (Klapoetke et al., 2014) in Rohon Beard cells and tested their responses to blue light stimulation (Figure 6D, Figure 7A). We found that 1 to 5 ms blue light pulses were sufficient to elicit photocurrents driving single action potentials in Rohon-Beard neurons at 4 and 5 dpf (Figure 7C and G, mean time to peak from the onset of the light pulse = 2.80 + /- 0.29 ms, n = 6 cells). We then recorded from target V2a interneurons while photostimulating RB neurons in Tg(isl2b:gal4,cmcl2:eGFP;UAS:CoChR-tdTomato;chx10:DsRed) larvae (Figure 7D–F). We found that RB stimulation led to short latency excitatory post-synaptic currents (EPSCs) in target V2a interneurons (Figure 7D–E and G, mean delay from the onset of light pulse = 3.71 + /- 0.19 ms, n = 8 cells in 8 larvae). To assess whether RB-V2A ipsilateral connection might be monosynaptic, we analyzed the lag between RB spikes time-to-peak and V2a EPSCs onset and found that it was below 1 ms (Figure 7G, 0.9 ms), a value compatible with monosynaptic connection (see Li et al. [2007a]) where cutoff is set at 3 ms for monosynaptic connections). Assuming a synaptic delay of 0.5 ms, the estimated conduction velocity along the axon of RB neuron would be 0.75 m/s, a value compatible with previous measurement in the field (Li et al., 2007a, 2004a). EPSC amplitude induced by RB stimulation in V2a interneurons were on average 28.14 ± 3.9 pA, leading to estimated depolarization of 11.71 ± 2.59 mV on target V2a (n = 8). In some cases, RB inputs were therefore sufficient to elicit action potentials in V2a interneurons tested in current clamp mode (Figure 7F). Given the glutamatergic nature of RB neurons, we incubated the preparation with blockers of glutamatergic neurotransmission and found that it abolished EPSCs in target V2a (Figure 7I, n = 2 cells). Altogether, these results indicate a monosynaptic connection between RB neurons and ipsilateral dorsalmost V2a interneurons and demonstrate the ability of RB neurons to recruit selectively a subpopulation of V2a interneurons controlling fast swimming in the dorsal spinal cord. This direct pathway from RB neurons onto ipsilateral ‘fast’ V2a neurons may act in concert with afferences from dorsal root ganglia and trigeminal neurons to enhance speed during active locomotion.

Rohon-Beard neurons selectively synapse onto ipsilateral dorsalmost V2a interneurons.

(A) Schematic of the optogenetic calibration of CoChR-mediated stimulation in RB neurons. (B) Current-injection steps recording reveals the low-input resistance nature of Rohon Beard neurons, n = 7 cells). (C) 5 ms blue light pulse stimulations reliably elicit single action potentials in CoChR-expressing RB cells (n = 3 cells). (D) Schematic of whole cell recording of V2a interneurons with optogenetic stimulation of RB neurons. (E) Voltage clamp recording of a target V2a interneuron receiving monosynaptic inputs from a RB neuron. (F) Current clamp recording of a target V2a interneuron firing an action potential following the stimulation of a RB neuron. (G) Timing of RB spike time-to-peak (2.8 ms ±0.3 ms after the onset of the blue light pulse, n = 103 stimulations recorded in 6 cells) and onset of EPSCs in target V2a interneurons (3.7 ms ±0.2 ms after the onset of the blue light pulse, n = 78 EPSCs in 8 cells). The average lag between RB spiking and V2a EPSC is 0.9 ms. The blue bar represents the duration of the blue light pulse. (H) Membrane properties of recorded V2a interneurons in the dorsal cord, n = 5 cells). (I) Blockade of glutamatergic neurotransmission following the application of 10 µM of CNQX and D-AP5 suppresses elicited post-synaptic currents in a V2a interneuron (n = 2 cells).

https://doi.org/10.7554/eLife.25260.015

Discussion

Understanding the contribution of peripheral mechanosensory feedback during locomotion is an important question in the field of motor control. Although the state-dependent nature of spinal modulation by afferent muscle fibers has been appreciated for more than a century in limbed vertebrates, little is currently known on the role of mechanosensory feedback during ongoing locomotion. The reasons for this lie in the difficulty of recording and manipulating neural circuits in moving animals and in the fact that mechanosensory feedback is absent in fictive preparations. In this study, we took advantage of the genetic and optical accessibility of the zebrafish larva to unravel mechanisms of circuit and behavior modulation by mechanosensory feedback using a combination of optical methods for circuit interrogation during animal motion with electrophysiology and high-throughput kinematic analysis of locomotor behaviors. Our results demonstrate that mechanosensory feedback enhances swimming speed by providing excitation to spinal circuits controlling the frequency of motor output during fast locomotion, thereby controlling the timing of spinal microcircuit activation and the transition between fast and slow locomotion. As it allows zebrafish larvae to travel the same distance in a shorter time (about 15 ms), mechanosensory feedback also bears ecological implications for fitness. By speeding up escapes during a predator attack, mechanosensory feedback may help enhancing survival.

To probe the contribution of mechanosensory feedback during active locomotion in zebrafish, we used the bioluminescent GFP-Aequorin sensor in motor neurons. We found that the amplitude of Aequorin signals from motor neurons could be used as a signature for distinguishing locomotor maneuvers including slow swimming and fast escape responses that have specific ranges of swimming frequencies and distinct profiles of motor neurons recruitment (Liao and Fetcho, 2008; Masino and Fetcho, 2005; McLean et al., 2008). We found that eliminating mechanical sensory feedback originating from muscle contraction led to a decrease of bioluminescence signals from motor neurons during acoustic escape responses. Our observations suggest that mechanosensory feedback enhances the recruitment of motor circuits. However, the effect observed could be driven by a decrease in swimming frequency and /or a decrease in the number of active spinal motor neurons. The poor temporal resolution of GFP-Aequorin prevented us from resolving these possibilities.

Speed modulation by mechanosensory feedback

To determine how the deprivation of mechanosensory feedback affects motor neuron recruitment, we analyzed kinematic parameters of animals deprived of excitatory synaptic transmission in glutamatergic mechanosensory neurons using the genetically encoded Botulinum toxin. The perturbation of major kinematic parameters such as left/right coordination or locomotor frequency during fictive locomotion usually requires the ablation of entire classes of spinal neurons (Crone et al., 2008; Gosgnach et al., 2006; Talpalar et al., 2013; Zhang et al., 2008). To identify the consequence of the selective silencing of mechanosensory feedback in the zebrafish spinal cord, we implemented a high-throughput kinematic analysis of acoustic escape responses in larvae deprived of neurotransmission in isl2b+ mechanosensory neurons versus their control siblings. Our analysis revealed that mechanosensory feedback increases the swimming frequency and corresponding locomotor speed without affecting the total number of oscillations during an escape. Inputs from mechanosensory neurons first contribute to reaching the maximum range of fast swimming frequencies and second set the transition time from fast to slow swimming frequencies. Without mechanosensory inputs, the initial swimming frequency is reduced and the transition from fast to slow swimming occurs earlier, which leads to an increase of locomotor event duration for the same number of oscillations and distance travelled, and therefore an overall slower escape response.

Contrary to mammals, there are no muscle spindles relaying mechanical information on muscle contraction to spinal circuits in fish. Instead, glutamatergic stretch receptor neurons that can sense body elongation in lamprey (Grillner et al., 1981; Rovainen, 1974) project onto spinal sensory neurons, motor neurons and interneurons (Di Prisco et al., 1990) and can entrain the locomotor central pattern generators (CPGs, [Grillner et al., 1981; McClellan and Sigvardt, 1988]). Recently, GABAergic cerebrospinal-fluid contacting neurons have been shown to detect body curvature following muscle contraction in both zebrafish (Böhm et al., 2016; Grillner et al., 1984; Jalalvand et al., 2016). Interestingly, the activation of each of these pathways, either during natural mechanical oscillations of the tail or optogenetic stimulations, is sufficient to trigger locomotor activity demonstrating that mechanosensory feedback can provide inputs to locomotor central pattern generator circuits in the spinal cord (Böhm et al., 2016; Fidelin et al., 2015; McClellan and Grillner, 1983; Wyart et al., 2009). Yet, the molecular identity of CPG neurons receiving inputs from mechanosensory neurons has remained unclear.

Circuit architecture of mechanosensory neurons in zebrafish

Previous studies have described a remarkably simple and conserved disynaptic cutaneous skin reflex relying on spinal mechanosensory neurons referred to as Rohon Beard (RB) neurons in Xenopus and zebrafish and ‘dorsal cells’ in lamprey (Buchanan and Cohen, 1982; Christenson et al., 1988; Clarke et al., 1984; Easley-Neal et al., 2013; Li et al., 2003; Roberts et al., 1983). In these species, spinal mechanosensory neurons respond to skin stimulation at rest by firing a single spike and project onto glutamatergic commissural interneurons (referred to as dIc, CoPA and giant interneurons respectively [Buchanan and Cohen, 1982; Easley-Neal et al., 2013; Li et al., 2003; Pietri et al., 2009; Sillar and Roberts, 1988]), which drive contralateral motor neuron activation and subsequent swimming away from the stimulus (Buchanan and Cohen, 1982; Christenson et al., 1988; Li et al., 2003, 2004b). During swimming, excitation of commissural sensory interneurons from sensory neurons was shown to be filtered out by phase-locked inhibition (Knogler and Drapeau, 2014; Sillar and Roberts, 1988), ensuring that the recruitment of RB neurons potentially arising from movement-related activation would not trigger the disynaptic cutaneous reflex.

Here we discovered a novel sensory disynaptic pathway originating from spinal sensory neurons. We demonstrate that mechanosensory Rohon-Beard neurons synapse onto dorsalmost V2a interneurons previously shown to drive fast swimming in zebrafish (Ampatzis et al., 2014; McLean et al., 2008, 2007). Furthermore, we show that in some cases, this monosynaptic connection can provide suprathreshold inputs to V2a interneurons recruited during fast locomotion. Given that the isl2b transgenic line used in this study also labels dorsal root ganglia and trigeminal neurons, it is possible that the modulation of speed results from synergistic inputs from these three neuronal populations (Buhl et al., 2012). Dorsal root ganglia might in particular provide glutamatergic sensory inputs that could modulate locomotor frequency as well. Yet, we identified here a direct RB-V2a pathway that could contribute to the enhancement of speed triggered by mechanosensory feedback during the escape. The contribution of RB neurons to the enhancement of speed by mechanosensory feedback is consistent with previous experiments showing that stimulation of RB neurons can reset or increase locomotor frequency in Xenopus tadoples (for example,[Sillar and Roberts, 1992]). The role of mechanosensory feedback to increase swimming speed is consistent with former observations in lamprey showing an entrainment of locomotion by mechanical stimulation of the spinal cord (Grillner et al., 1981). Therefore, the direct connection we demonstrated here from spinal mechanosensory neurons onto a subtype of ipsilateral V2a interneurons recruited during fast swimming could underlie the effect we observed on swimming speed, along with other potential pathways involving dorsal root ganglia and trigeminal neurons.

RB neurons have complex processes branching under the skin as well as ascending and descending in the spinal cord. One critical experiment in the future will consist of recording the activity of RB neurons during muscle contraction in freely swimming animals. Recording reliable calcium transients associated with mechanical stimulation has been possible by combining fluorescent calcium sensors in one wavelength with another fluorescent protein enabling to control for cell position (Böhm et al., 2016). Nonetheless measuring activation of RB neurons during mechanical stimulation remains challenging due to the fact that RB neurons usually fire single action potentials ( et al., 2017; Roberts et al., 2012; Winlove and Roberts, 2012) and most calcium indicator fail to report single spike without extensive averaging (see Chen et al. [2013]; Tian et al. [2009]). Further studies will identify the connectivity diagrams of each of these mechanosensory cell types and devise genetic strategies to silence these neurons individually to probe their contribution to speed modulation during active locomotion. While local reflex pathways have been characterized at rest, further studies will decipher the mechanisms underlying the diversity of behavioral effects due to mechanosensory neuron recruitment: escapes for single touch at rest (Buchanan and Cohen, 1982; Douglass et al., 2008; Roberts et al., 1983), struggling for continuous stimulation (Li et al., 2007b) and enhancement of speed during active locomotion (this study).

Conclusion

Altogether, our results about the role of mechanosensory feedback in speed modulation can explain the general observation made in various invertebrate as well as vertebrate species that locomotion is faster in active compared to fictive conditions. In invertebrates, there is evidence in several species for fictive locomotor patterns being typically twice slower than real locomotion (sea slugs, (Morton and Chiel, 1993), 1993; crayfish, (Bacqué-Cazenave et al., 2015); and fruit fly, (Mendes et al., 2013). The leech is a particularly interesting example as it can perform both swimming and crawling. The effect of denervation is a systematic slowdown of locomotor activity, although it is much larger in crawling (twofolds) (Baader and Kristan, 1992; Eisenhart et al., 2000; Kristan and Calabrese, 1976). In lamprey where the same animal was used to compare active and fictive locomotion, the frequency of fictive swimming appears 5–10 times slower than the frequency of active swimming recorded through electromyography (Wallén and Williams, 1984). The general slowdown of locomotor frequencies when peripheral sensory feedback is removed is commonly observed in fictive locomotion in mice as well (Bellardita and Kiehn, 2015; Talpalar et al., 2013). New preparations to manipulate and record from spinal neurons in moving mice (Hayes et al., 2009; Hochman et al., 2012) will enable in the future to investigate the dynamic role of mechanosensory feedback during active locomotion in mammals.

Our study explains how mechanosensory neurons enhance speed during active locomotion and reveal one possible spinal circuit that could mediate this effect. Furthermore, our work demonstrates that microcircuit selection during locomotor behaviors is not only set by descending commands from supra-spinal centers but could rather result from a combination of inputs from descending and local spinal sensorimotor circuits recruited while the movement is performed.

Material and methods

Zebrafish care, generation and characterization of transgenic lines

All procedures were approved by the Institutional Ethics Committee at the Institut du Cerveau et de la Moelle épinière (ICM), Paris, France, the Ethical Committee Charles Darwin and received subsequent approval from the EEC (2010/63/EU). Adult AB and and Tüpfel long fin (TL) strains of Danio rerio were maintained and raised on a 14/10 hr light cycle and water was maintained at 28.5°C, conductivity at 500 μS and pH at 7.4. Embryos were raised in blue water (3 g of Instant Ocean salts and 2 mL of methylene blue at 1% in 10 L of osmosed water) at 28.5°C during the first 24 hr before screening for GFP expression. The Tg(mnx1:gal4)icm11 line was based on the injection of the mnx1 construct provided by Dr. Thomas Auer and Dr. Filippo Del Bene (Institut Curie, Paris, France, see Table 1). The original sequence for GFP-Aequorin provided by Dr. Ludovic Tricoire (Université Pierre et Marie Curie, Paris, France) was subsequently codon-optimized for expression in zebrafish and subcloned into the PT2 14XUAS plasmid provided by Pr. Koichi Kawakami (National Institute of Genetics, Mishima, Japan). Injection of the Tg(UAS:GFP-aequorin-opt) construct in Tg(mnx1:gal4)icm23 larvae allowed the generation of the Tg(mnx1:gal4;UAS:GFP-aequorin-opt)icm09 line with selective expression of GFP-Aequorin in spinal motor neurons: more prominently primary dorsal motor neurons but also intermediate and ventral secondary motor neurons (Figure 1B) without any expression in the muscles and only very limited expression in the brain and hindbrain (note we cannot exclude that a minority a spinal ventral neurons distinct from motor neurons might be targeted). The transgenic line Tg(UAS:GCaMP6f,cryaa:mCherry)icm06 was generated by subcloning GCaMP6f (Chen et al., 2013) into pDONR221 and then assembled into the final expression vector in a three-fragment Gateway reaction using p5E-14XUAS, pME-GCaMP6f, p3E-poly(A) and pDest-CryAA:mCherry. Given the potential variability in expression patterns of Gal4 lines with different UAS, we characterized the Tg(mnx1:gal4;UAS:GCaMP6f,cryaa:mCherry) line and observed a similar expression in motor neurons as in the Tg(mnx1:gal4;UAS:GFP-aequorin-opt)icm09 line. In both cases, we observed the targeting of spinal neurons exiting the spinal cord in the ventral root, and therefore referred to as motor neurons. However, we cannot exclude that the mnx1 promoter may target other spinal neurons, distinct from motor neurons. relaxed mutants (cacnb1ts25/ts25) (Granato et al., 1996) were provided by Pr. Paul Brehm (Vollum Institute, Oregon Health and Science University, Portland, USA). In homozygous cacnb1ts25/ts25 mutants, a mutation of the skeletal muscle dihydropyridine receptor β1a subunit interferes with the calcium release and mutant larvae are immotile (Granato et al., 1996). The Tg(isl2b:gal4,cmlc2:eGFP) line (Auer et al., 2015) driving expression in Dorsal Root Ganglia (DRG), trigeminal, Rohon-Beard neurons, a subset of Retinal Ganglion Cells (RGCs), and blood vessels was provided by Dr. Thomas Auer and Dr. Filippo Del Bene (Institut Curie, Paris, France) (see Table 1).

Table 1

Stable transgenic lines used or generated in this study.

https://doi.org/10.7554/eLife.25260.016
NameOriginal reference
Tg(mnx1:gal4)icm23(Böhm et al., 2016)
Tg(isl2b:gal4,cmlc2:eGFP)(Auer et al., 2015)
Tg(UAS:GFP-aequorin-opt)icm09This paper
Tg(UAS:GCaMP6f,cryaa:mCherry)icm06(Böhm et al., 2016)
Tg(chx10:loxP:DsRed:loxP:GFP)(Kimura et al., 2006)
relaxed (cacnb1ts25)(Granato et al., 1996)
Tg(UAS:BoTxLCB-GFP)icm21(Auer et al., 2015; Böhm et al., 2016; Sternberg et al., 2016)

Immunohistochemistry

Larvae were fixed in 4% PFA for 4 hr at 4°C followed by 3 × 5 min washes in PBS. Larvae were blocked for 1 hr in blocking solution (10% NGS, 1% DMSO, 0.5% Triton X100 in 0.1 M PBS). Larvae were incubated with the primary antibody over night at RT (1% NGS, 1% DMSO, 0.5% Triton X100 in 0.1% PBS). After washing three times for 5 min in 0.1 M PBST with 0.5% Triton X100 (PBST), larvae were incubated in the dark with the secondary antibody in PBST. After washing again three times for 5 min in PBST, larvae were mounted on a slide with mounting medium and were imaged on an upright confocal microscope (FV-1000, Olympus, Tokyo, Japan). The following antibodies were used: anti-GFP (Abcam ab 13970, dilution 1:500), Alexa Fluor 488 goat anti-chicken IgG (A11039, dilution1:1000, Invitrogen, Carlsbad, CA, USA). Immunostaining specificity was established by omitting the primary antibody.

Monitoring neuronal activity with bioluminescence

Embryos were dechorionated and soaked at 26°C in 100 µL of blue water with a final concentration of 60 µM of coelenterazine-h (Biotium, Fremont, CA, USA). Coelenterazine-h was renewed at 2 dpf. All experiments were performed at 4 dpf. In all experiments, one larva was head-embedded in 1.5% low-melting point agarose with the tail free to move in a circular (2 cm diameter) 3D-printed arena (Sculpteo, Villejuif, France). The arena was then placed in a lightproof box (Figure 1A) and attached to a small speaker (2 Ohm). Each trial consisted of a 500 ms baseline followed by a 10 ms acoustic stimulus and 1990 ms subsequent recording. Assays consisted of 30 trials with 1 min inter trial intervals to reduce habituation. Sinusoidal stimuli (5 cycles, 500 Hz) were delivered through a wave generator (33210-A, Agilent, Santa Clara, CA, USA) and audio amplifier (LP2020A, Lepai). Intensity was adjusted to the lowest value that reliably elicited an escape response (between 0.5 and 5 Vpp). The same larvae used for the active assay were subsequently paralyzed by bath application of pancuronium bromide (P1918, Sigma-Aldrich, Saint-Louis, MO, USA) at 0.6 mg/mL final concentration and stimulation intensity was adjusted to the lowest value that elicited a bioluminescent signal. For Figure 3, cacnb1ts25/ts25 and control siblings were tested alternatively on the same day and compared to each other. In non-moving animals (that is, paralyzed or cacnb1 mutants), the intensity was progressively increased until stimuli elicited bioluminescence signals. A higher intensity of the acoustic stimulus was often needed after addition of pancuronium bromide, possibly due to modulation of cholinergic arousal brain circuitry (Yokogawa et al., 2007). As negative controls, bioluminescence assays of wild type animals or Tg(mnx1:gal4;UAS:GFP) (Zelenchuk and Brusés, 2011) where motor neurons express GFP only revealed no signals (n = 3 wild type larvae with 30 trials each, n = 5 Tg(mnx1:gal4;UAS:GFP) larvae with 30 trials each). Animals deprived of GFP-Aequorin did not produce any signals above baseline noise level during escape responses. Infrared light illumination for monitoring larval behavior was provided by an 850 nm LED (Effisharp, Effilux, France) mounted with 2 long-pass 780 and 810 filters (Asahi ZIL0780 and Asahi XIL0810, respectively) and a diffuser (DG10-120B, Thorlabs, Newton, NJ, USA). Video acquisition was performed at 1000 Hz using a high-speed infrared sensitive camera (Eosens MC1362, Mikrotron, Unterschleissheim, Germany; objective Nikkor 50 mm f/1.8D, Nikon, Tokyo, Japan) at 320 × 320 pixels resolution controlled by the software Hiris (RD Vision, Saint-Maur-des-Fossés, France). Photons were counted with a photomultiplier tube (PMT, H7360-02, Hamamatsu, Japan) located under the larva arena and sent to an acquisition card (NI PCI 6602, National Instruments, Cos Cob, CT, USA). A band-pass filter (525 nm / 50 nm, ref. 489038–8002, Carl Zeiss, Thornwood, NY, USA) and a short-pass filter (670 nm, XVS0670, Asahi, Japan) were placed between the larva and the PMT. A custom application-programming interface developed in collaboration with R&D Vision synchronized the video acquisition with the photon count and the stimulus delivery using 30 trials batched TTL chronogram (EG Chrono, RD Vision, France).

Kinematics and bioluminescence analysis

The base and tip of the tail were manually determined for each larva and the tail was subsequently automatically tracked with a custom Matlab algorithm (R2012b, Mathworks, Natick, MA, USA). The tail angle was computed for each frame and filtered using median filtering (window size = 10). The start of the movement was determined as the first frame followed by 3 with a differential tail angle value above 0.08. The end was determined as the end of the 20 frames with a differential tail angle value below 0.1 degree. Local minimal and maximal values of the tail angle occurred at least 2 ms apart and 1 degree above the 5 ms preceding value. Automated movement categorization was determined as follows: ‘escapes’ for all movements starting with an asymmetrical C-bend and number of cycles ≥ 1; ‘slow swims’ for all movements with symmetrical bends of amplitude <25° and number of cycles > 1.

Photons were counted with a temporal resolution of 1 ms and then binned every 10 ms. The signal was filtered using a running average with a window size of 10, giving a typical signal-to-noise ratio (SNR) for active movements of 50 to 1. Noise was extrapolated from a linear fit of the cumulative photon count before the stimulus and subtracted from the signal. The start and end of the bioluminescent signal were computed as the first time point followed by three points with a differential value above 0.4 photons/10 ms and below 0.2 photons/10 ms, respectively. The time-to-peak was calculated between the start and the peak of the bioluminescent signal while the decay coefficient was derived from the one-term exponential fit between the peak and the end of the signal.

Calcium imaging of spinal motor neurons and ventral nerve root recordings (VNR)

4 dpf Tg(mnx1:gal4; UAS:GCaMP6f,cryaa:mCherry) double transgenic larvae were screened for dense labeling and good expression of GCaMP6f in spinal motor neurons under a dissecting microscope equipped with an epifluorescence lamp (Leica, Wetzlar, Germany). Larvae were anaesthetized in 0.02% Tricaine-Methiodide (MS-222, Sigma-Aldrich) diluted in fish facility water and mounted on their lateral side in 1.5% low-melting point agarose in glass-bottom dishes filled with external solution ([NaCl] = 134 mM, [KCl] = 2.9 mM, [MgCl2] = 1.2 mM, [HEPES] = 10 mM, [glucose] = 10 mM and [CaCl2] = 2.1 mM; adjusted to pH 7.7–7.8 with NaOH and osmolarity 290 mOsm). Larvae were immobilized by injecting 0.1–0.3 nL of 0.5 mM α-Bungarotoxin (Tocris, Bristol, UK) in the ventral axial musculature. A portion of agar was removed using a razor blade in order to expose 2 to 3 segments. To achieve a strong signal-to-noise ratio during fictive locomotion recordings, the skin overlying these segments was removed using suction glass pipettes. Zebrafish larvae were imaged using a custom spinning disk microscope (Intelligent Imaging Innovation, Denver, CO, USA) equipped with a set of water-immersion objectives (Zeiss 20X, 40X, NA = 1). Recordings were acquired using Slidebook software at 20 Hz using a 488 nm laser. Gain and binning were optimized to maximize signal to noise ratio. Z-projection stacks showed full pattern of expression using Fiji (Schindelin et al., 2012). Positions of cells along the D-V axis were computed using Fiji and Matlab (Mathworks, USA). Calcium signals were extracted online using custom scripts. Regions of interest (ROIs) were manually designed and calcium signals time series were extracted as the mean fluorescence from individual ROIs at each time point of the recording. We observed that out-of-focus signals varied between animals, from dorsal to ventral spinal cord regions in a behavior-dependent manner. To estimate the contribution of out-of-focus signals we systematically picked two background ROIs, one placed below the ventral limit of the spinal cord to capture out-of-focus signals at the level of ventral motor neurons during slow swimming, the second in the dorsalmost part of the spinal cord to capture out-of-focus signals in the dorsal spinal cord during the escape. We estimated the maximum out-of-focus signals observed during each behavior and used this value as a threshold for discriminating active from silent motor neurons. Thin-walled, borosilicate glass capillaries (Sutter Instruments, Novato, CA, USA) were pulled and fire-polished from a Flaming/Brown pipette puller (Sutter Instruments, Novato) to obtain peripheral nerve recording micropipettes. Pipettes were filled with external solution and positioned next to the preparation using motorized micromanipulators under the microscope. Light suction was applied when the pipette reached the muscle region located at the vicinity of intermyotomal junctions, ventral to the axial musculature midline. VNR signals were acquired at 10 kHz in current clamp IC = 0 mode using a MultiClamp 700A amplifier (Molecular Devices–Axon Instruments, Sunnyvale, CA, USA), a Digidata series 1322A digitizer (Axon Instruments) and pClamp 8.2 software (Axon instruments). Recordings were considered for analysis when the background noise did not exceed 0.05 mV amplitude and signal to noise ratio for fictive locomotor events detection was above three. VNR recordings were analyzed offline and aligned to calcium imaging data using custom-made Matlab scripts.

Fluorescence-guided whole-cell recordings and pharmacology

4 dpf larvae were pinned down in Sylgard-coated, glass-bottom dishes filled with external solution with thin tungsten pins through the notochord. Skin was removed from segments 5–6 to the end of the tail using sharp forceps. After removing the skin, one to two segments were dissected using glass suction pipettes. Patch pipettes (1B150F-4, WPI, Sarasota, FL, USA) were designed to reach a tip resistance of 8–12 MOhms and were filled with potassium-containing internal solution (concentrations in mM: K-gluconate 115, KCl 15, MgCl2 2, Mg-ATP 4, HEPES free acid 10, EGTA 0.5, 290 mOsm, adjusted to pH 7.2 with KOH and supplemented with Alexa 647 at 4 mM). Cells were held at around −60 mV. We calculated the liquid junction potential in our experiments (−9 mV) but did not correct for it since it did not affect the outcome of our experiments. NMDA receptor antagonist (D-AP5, Tocris) and AMPA receptor antagonist (CNQX, Tocris) were bath applied at 10 µM final concentrations. Kinetic parameters of light-evoked currents and EPSCs were extracted and analyzed using custom-made Matlab scripts.

Behavioral analysis of freely moving BoTxBLC larvae

Tg(isl2b:gal4,cmlc2:eGFP;UAS:BoTxBLC-GFP) larvae were screened at 2 dpf for expression. At 6 dpf, larvae were tested 4 by 4: each larva was positioned in a separate dish (2 cm diameter) and illuminated from below, freely moving. Each swim arena contained 400 μL of external solution to minimize movement in the z-axis. Escapes were elicited by delivering a 500 Hz stimulus for 1 ms via 20W speakers. Each trial consisted of a 200 ms pretrial window followed by the stimulus and 800 ms subsequent recording. Five trials were performed in succession with 2 min long intertrial resting periods. Behavior was recorded at 650 fps with a high-speed camera (acA2000-340km, Basler, Ahrensburg, Germany) and analyzed using a tracking algorithm (ZebraZoom, [Mirat et al., 2013]) and a custom Matlab script. Extracted kinematic variables are listed and defined as follows: duration (the amount of time during movement), distance (travelled distance), speed (distance/duration), number of oscillations (1/2 of the number of peaks in tail angle over time), and mean tail beat frequency (TBF, inverse of the duration between two subsequent peaks in the tail angle). Visual inspection of the unsmoothed tail angle was used to exclude trials in which erroneous tracking occurred. Trials fulfilling at least one of the following criteria were programmatically excluded: (i) Detected movement prior to the stimulus. In these cases, spontaneous pretrigger movement lingering into the post-trigger recording window can occlude escape-type movements. (ii) First peak in tail-angle occurring more than 20 frames after the stimulus. These no longer represent low-latency escapes and most likely represent a different category of movement altogether (Budick and O'Malley, 2000; Liu and Fetcho, 1999). (iii) First peak in tail-angle lower than 60 degrees. Bouts lacking an initial C-bend are not escapes and are most likely slow swim bouts or other movement types (Budick and O'Malley, 2000). Hemi-periods were calculated as the interval between two consecutive peaks in the tail angle, and subsequently used as a determinate to extract fast and slow components of the escape with a cutoff of 650/60 frames (30 Hz). Peaks which alternated between the cutoff value after the first appearance of a slow peak were excluded.

Statistical analysis

SPSS 20 (IBM, Armonk, NY, USA) was used to perform all statistical analyses. Comparisons of bioluminescence signals parameters was conducted using a t-test for paired samples for repeated measures within subjects (that is, active versus paralyzed data). Mixed linear model analysis with repeated measures using an auto-regressive covariance structure was performed to compare the bioluminescence amplitude between movement categories in active assays, and between moving and immotile larvae in active versus fictive assays. The same model was used to analyze behavioral parameters in order to account for repeated measures in within-fish (trial number) and between-group comparisons (genotype). Bioluminescence time decay coefficients were included if the goodness of fit r-square value was >0.95. A Pearson test was used to assess correlations for parametric data. Statistical significance is represented in the graphs as *** for p<0.001, ** for p<0.01, * for p<0.05, corrected for multiple comparisons when needed. All data are provided in the figures and text as means +/- standard error of the mean (SEM).

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
    Activities of identified interneurons, motoneurons, and muscle fibers during fictive swimming in the lamprey and effects of reticulospinal and dorsal cell stimulation
    1. JT Buchanan
    2. AH Cohen
    (1982)
    Journal of Neurophysiology 47:948–960.
  11. 11
    Locomotor repertoire of the larval zebrafish: swimming, turning and prey capture
    1. SA Budick
    2. DM O'Malley
    (2000)
    The Journal of Experimental Biology 203:2565–2579.
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
    Genes controlling and mediating locomotion behavior of the zebrafish embryo and larva
    1. M Granato
    2. FJ van Eeden
    3. U Schach
    4. T Trowe
    5. M Brand
    6. M Furutani-Seiki
    7. P Haffter
    8. M Hammerschmidt
    9. CP Heisenberg
    10. YJ Jiang
    11. DA Kane
    12. RN Kelsh
    13. MC Mullins
    14. J Odenthal
    15. C Nüsslein-Volhard
    (1996)
    Development 123:399–413.
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
    Rhythmic swimming activity in neurones of the isolated nerve cord of the leech
    1. WB Kristan
    2. RL Calabrese
    (1976)
    The Journal of Experimental Biology 65:643–668.
  40. 40
    The spinal interneurons and properties of glutamatergic synapses in a primitive vertebrate cutaneous flexion reflex
    1. WC Li
    2. SR Soffe
    3. A Roberts
    (2003)
    Journal of Neuroscience 23:9068–9077.
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
    Features of entrainment of spinal pattern generators for locomotor activity in the lamprey spinal cord
    1. AD McClellan
    2. KA Sigvardt
    (1988)
    Journal of Neuroscience 8:133–145.
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
    Initiation and control of swimming in amphibian embryos
    1. A Roberts
    2. SR Soffe
    3. JD Clarke
    4. N Dale
    (1983)
    Symposia of the Society for Experimental Biology 37:261–284.
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
    The role of premotor interneurons in phase-dependent modulation of a cutaneous reflex during swimming in xenopus laevis embryos
    1. KT Sillar
    2. A Roberts
    (1992)
    Journal of Neuroscience 12:1647–1657.
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79
  80. 80
  81. 81
  82. 82
  83. 83

Decision letter

  1. Ronald L Calabrese
    Reviewing Editor; Emory University, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Mechanosensory Neurons Control the Timing of Spinal Microcircuit Selection during Locomotion" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom, Ronald L Calabrese (Reviewer #1), is a member of our Board or Reviewing Editors and the evaluation has been overseen by Didier Stainier as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Mark A Masino (Reviewer #2); David McLean (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This is an elegant exploration of how sensory feedback affects escape and slow swimming in zebra fish larvae. Combing genetic, imaging, electrophysiological, optogenetic and kinematic analysis, the authors show that mechanosensory feedback enhances the recruitment of motor pools and controls the speed and duration of escape swimming. They then show that Rohon Beard neurons (spinal cord mechanoreceptors) make anatomical contacts and functional synaptic connections with V2a interneurons that are known to control speed of escape swimming. The experiments are carefully done with many thoughtful controls. The data shown is sufficient and analyzed with appropriate statistics. This work reveals microcircuits that underlie speed modulation by mechanosensory feedback in the vertebrate spinal cord and will thus be of wide interest.

Essential revisions:

There is no direct causal evidence that Rohon Beard neurons do indeed mediate the speed changes associated with active swimming, because they are not selectively silenced. Moreover, there is no direct demonstration that they are activated by the animals own movements during active swimming. These weaknesses can be mitigated in two ways.

1) An experiment where calcium imaging is used to monitor transients in RB neurons in partially embedded versus fully embedded and/or paralytic fish. This would demonstrate whether the RB cells participate in reafference.

2) More extensive analysis of the connectivity data between RB neurons and V2a neurons, including increases in the number of independent observations. Latency repeatability and jitter should be critically evaluated both for the EPSC and for the spike response of V2a neurons to RB cell photoactivation. We would like to be reassured that the nature of the connection is truly distinct from the one revealed by prior work, and is thus capable of influencing V2a activity during swimming.

The detailed reviews of two of the reviewers are included to help in revision but the focus of revision should be the two major points above.

Reviewer #1

1) The paper is well written within text but important details are missing from the legends that allow easy interpretation of the data. Some figures are too dense as if composed for a journal with restrictions on figure numbers.

All legends must be upgraded so that a reader can easily understand the data. Let me focus on Figure 1 to illustrate what must be done. What is going on in panels A and G? What are the superimposed lines in panel C (I know this but I am just saying be complete)? What is the color code in panel H and how does it relate to panel I. In panel I, I am really confused please explain the traces. Why are there multiple bumps in the left traces and what does the superposition mean in each set of traces? Please write figure legends that allow the data to be easily understood. Figure 3 is also a problem. What do the arrow and the arrowhead point out in panel A? What are we looking at in the left most image of panel C and in the middle image of panel C and left image of panel D? What are the different colors showing? In all graphs like panel E1 the mean and SEM are just too difficult to see especially amid the blue points; make bold.

Split Panels G to K of Figure 1 into a separate figure. Consider splitting Figure 5 also.

2) I am concerned about the morphological data shown in Figure 1B. Was this observation made in a number of larvae? In Figure 5A and B, was this observation made in a number of larvae. The legend of panel C implies that it was but you must be explicit. In Figure 5G, was this observation made in a number of larvae?

3) There is no direct causal evidence that Rohon Beard cells do indeed mediate the speed changes associated with active swimming, because they are not selectively silenced or optogenetically activated. This needs to be more explicitly acknowledged.

4) Given the central effects of pancuronium bromide why wasn't α-bungarotoxin used to immobilize the larvae in the experiments of Figure 2? I am also concerned that you did not control for stimulus intensity in these experiments. In active fish and during pancuronium paralysis, is the size of the aequorin response dependent of stimulus amplitude? If it is, then comparisons of amplitudes across these two groups is suspect because different stimulus amplitudes were used.

Reviewer #3:

1) If RB neurons are encoding proprioceptive information in addition to the cutaneous exteroceptive information they are normally associated with, then activation of RB neurons during acoustic stimuli needs to be confirmed. The prediction is that RB neurons would respond to skin stimulation and acoustic stimulation when the tail can move, but not to acoustic stimuli when the tail is immotile. Without a better idea of whether RB neurons are in fact responding to self generated movements, then it is difficult to attribute the behavioral deficit to them and justify the ensuing circuit busting. For example, it could be that there are neurons in the DRG that are instead responsible.

2) There has been quite a bit of work in Xenopus tadpoles studying RB circuitry involved in cutaneous skin reflexes (e.g., Li et al., 2003, J Neurosci.; Li et al., 2004, J Neurophysiol). Abrupt stimulation of RB afferents generates a reflexive response away from the stimulus. Also, as I understand it, they demonstrate that drive from RB cells that potentially arises from movement-related stimulation would largely be filtered out by inhibition (Sillar and Roberts, 1988, Nature). However, RB neurons can reset or increase swimming rhythms when they are stimulated (e.g., Sillar and Roberts, 1992, J Neurosci), so there is clearly the potential for RB neurons to act as they are proposed to here. I think it would be worth explaining more how the findings here fit into previous work in tadpoles, as well as studies of dorsal cells (the RB equivalents) in lampreys (Buchanan and Cohen, 1982, J Neurophysiol; Christenson etal.,1988, Brain Res). I think the role of the RB-V2a connection would be worth considering in the context of struggling as well (Li et al., 2007, J Neurosci).

3) The optogenetic-electrophysiology data as currently presented don't provide sufficient temporal resolution to determine if the RB-V2a connection is monosynaptic or not. If it is, this would be novel, since studies of RB circuits in Xenopus have only revealed indirect connections from RBs to V2a-like neurons (with extremely short latencies though). In fact, the responses have multiple components which would be more consistent with a polysynaptic connection (especially since RB neurons tend to fire once). Given this is not the main goal of the work, I would suggest perhaps stepping back from saying it is a direct connection and instead focus on the excitatory nature of that connection, monosynaptic or not. Of course, this could all be moot if the RBs are not even active during body bending.

4) For the botulinum experiments it is largely assumed that the sensory populations are silenced, but it would be better if there was some confirmation in this line of fish. Patch recordings would be ideal, but if not possible a mention of whether the fish fail to respond to touch or not would suffice.

5) Since the RB driven response is specific to high frequency swimming commensurate with large deflections of the tail, I was expecting that the controls for the aequorin experiments would be zebrafish that are fully embedded so the tail can't move. Is it known if there is still less motor neuron recruitment in this condition comparable to toxin and mutants?

6) Work in lampreys studying the role of edge-cells seems appropriate to discuss more, since the function appears to be very similar to that proposed here and a lot of that work is performed while the tail is moving. Also the recent development of a rodent preparation is another precedence for neural recordings in moving animals (Hayes etal., 2009, J Neurophysiol; Hochman et al., 2012, Front. Biosci).

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Mechanosensory Neurons Control the Timing of Spinal Microcircuit Selection during Locomotion" for further consideration at eLife. Your revised article has been favorably evaluated by Didier Stainier (Senior editor), a Reviewing editor, and two reviewers.

The manuscript has been greatly improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

A few more changes are necessary that do not require re-review by the expert reviewers.

1) There was concern that the authors overstated the case for monosynapticity in the Introduction and Results, subsection “Mechanosensory neurons synapse onto V2a interneurons involved in fast locomotion, given that experiments with a high concentration of divalent ions – or the like – to eliminate (reduce the probability of) disynaptic events were not performed. We suggest that the text starting in subsection “Mechanosensory neurons synapse onto V2a interneurons involved in fast locomotion be modified something like "To assess whether RB-V2A ipsilateral connection might be monosynaptic, we analyzed the lag between RB spikes time-to-peak and V2a EPSCs onset and found that it was below 1 ms (Figure 7G, 0.9 ms), a value compatible with monosynaptic connection (see (Li et al., 2007a) where cutoff is set at 3 ms for monosynaptic connections)." The Introduction can be correspondingly edited.

2) Quoting one of the expert reviewers "There is an underlying assumption that all UAS lines (GFP-aequorin-opt and GCaMP6f;cryaa:mCherry) will produce the same expression patterns when paired with this Gal4 (mnx1:gal4) line. Just because the icm23 line (mnx1:gal4-UAS:preEGFP) was characterized by Bohm et al., this does not mean that Gal4 will work the same with different versions of UAS (UAS:GFP-aequorin-opt or GCaMP6f;cryaa:mCherry). Have these lines been characterized? If so, references to that work must be made to show MN specificity/variegated expression patterns (if any)."

Please make an effort to give any references that are relevant and provide any characterizations performed at least in Material and methods.

https://doi.org/10.7554/eLife.25260.021

Author response

Essential revisions:

There is no direct causal evidence that Rohon Beard neurons do indeed mediate the speed changes associated with active swimming, because they are not selectively silenced. Moreover, there is no direct demonstration that they are activated by the animals own movements during active swimming.

We agree with the reviewers that an ideal strategy would consist in selectively targeting and silencing Rohon-Beard neurons during active locomotion. Yet, we searched extensively the literature and existing databases and could not find any transgenic line targeting RBs without trigeminal neurons & dorsal root ganglia (see Kucenas et al.,2006 Neuroscience 138:641–652; Scott et al.,2007 Nature Methods 4:323 as well as Koichi Kawakami’s and Harry Burgess’s enhancer trap libraries).

These weaknesses can be mitigated in two ways.

1) An experiment where calcium imaging is used to monitor transients in RB neurons in partially embedded versus fully embedded and/or paralytic fish. This would demonstrate whether the RB cells participate in reafference.

We thank the reviewers for suggesting these experiments. The technical challenge of reliably detecting calcium transients in RB neurons lies in the fact that these cells usually fire single spikes in response to current steps in zebrafish larva (see Figure 7B for one example and Author response image 1). This is also the case in the tadpole, as previously shown by the group of Alan Roberts using current steps and gentle touch (Winlove and Roberts, 2012, EJN 36:2926-40). In particular, Alan Roberts and collaborators found that touch led to few spikes while stronger potentially damaging stimuli led to multiple spikes (Roberts et al.,2011 Dev. Neurobiol. 575-584).

Nonetheless, to assess the recruitment of RB neurons during swimming, we performed three independent sets of experiments.

1) First, we drove expression of genetically encoded calcium indicators GCaMPs in isl2b+ neurons in order to monitor the recruitment of RB neurons during active escape responses.

With GCaMP6f (Kd ~ 250 nM in solution, Chen et al., Nature 2013 and Loren Looger personal communication), we did not reach enough expression in RB neurons to perform calcium imaging experiments. On the contrary, with GCaMP5G (Kd = 373nM in solution, Tian et al., Nature Methods 2009 and Loren Looger personal communication), expression of the sensor was satisfactory (see Author response image 2). Yet, we observed calcium transients during escape responses only in a minority (2 out of 10) of fish expressing GCaMP5G in RB neurons (see examples from one responding fish depicted in Author response image 2). As such, RB responses were not representative of the 10 tested fish so we did not include them in the main manuscript.

Author response image 1
Responses to RB neurons to electrical and optogenetic stimulations consists in single spikes (n = 5).

RB neurons respond to a long current step by a single action potential (top row). RB neurons in Tg(isl2b:gal4) expressing the construct UAS:CoChR-tdTomato respond to 5ms light pulses by single action potentials (bottom row).

https://doi.org/10.7554/eLife.25260.017
Author response image 2
Calcium transients recorded in RB neurons from a non representative zebrafish larva in Tg(isl2b:gal4;cry:GFP,UAS: mRFP; UAS:GCaMP5G) triple transgenic animals.

(A,B) Expression of mRFP (A) and GCaMP5G (B) is restricted to RB neurons in this field of view. Successive snapshots show signals before the stimulus (0s), during tail motion (3 s), immediately after the motion artifact when the cells are back in the plane (3.5 s) and long after tail motion (9 s). (C) Calcium transients estimated as relative △F/F = △F (green) – △F (red) is depicted for four trials before (top) and after (bottom) application of the paralytic agent Pancuronium Bromide (PB) in the bath.

https://doi.org/10.7554/eLife.25260.018

One issue is that even the ultra-sensitive genetically encoded calcium sensors GCaMP with Kd ~ 300nM are not sensitive enough to detect single spikes in vivowithout intensive averaging across trials and across neurons (check for example the averaging performed in Tian et al., 2009 Nature Methods 6:875-81 for GCaMP5F and in Chen, Wardill et al.,2013 Nature 499: 295- 300 for GCaMP6f). Therefore by using these sensors in vivo, we can easily miss calcium transients associated with single spikes in RB neurons in response to tail contractions.

2) In order to optimize the imaging of calcium in RB neurons, we therefore retrogradely labeled RB neurons by injections in the spinal cord of Calcium Green Dextrans (Kd ~ 190-270 nM in solution, see Paredes et al., 2008 Methods 46:143-151). We attempted to detect single spikes in 6 RB neurons using an assay where the spinal cord is mechanically stimulated with a glass probe (see Böhm et al.,Nature Communications 2016). Skin stimulations are known to reliably activate RB neurons, which in turn fire single action potentials. However, the Calcium Green Dextran approach led to variation of fluorescence associated with RB neuron firing in a minority of neurons (1 out of 6 tested RB neurons using the mechanical probe).

3) As fluorescent calcium sensors failed to show transients in single cells, we took advantage of the bioluminescence assay to monitor the activity of isl2b+neurons during acoustic escape responses before and after blockade of muscle contraction. GFP-Aequorin bioluminescence signals were detected during active tail contraction (see Author response image 3) and were completely abolished during escape responses when muscles were paralyzed (see Author response image 3 C and D). These data therefore suggest that isl2b+neurons are recruited during active locomotion. Yet, since the isl2b promoter targets multiple cell types (RB neurons, trigeminal, DRGs, along with blood vessels and retinal ganglion cells), this approach cannot prove that RB neurons are the main neurons recruited during active swimming.

Conclusion

Despite our efforts, the three approaches used to test whether RB neurons were recruited during active motion are not conclusive. This could be due to the difficulty of monitoring calcium transients associated with single spikes in moving animals. Consequently, we edited the manuscript:

a) to emphasize that behavioral effects observed using BoTxLC-GFP could be due to silencing of either RB neurons, trigeminal neurons or dorsal root ganglia (Abstract and Discussion);

b) to mention that we could not measure calcium transients in response to bending in RB neurons reliably (see Discussion).

Author response image 3
Probing the recruitment of Rohon-Beard neurons expressing GFP-Aequorin during movement.

(A, B) Lateral view showing the expression of GFP-Aequorin in 4 dpf Tg(isl2b:gal4;UAS:GFP- Aequorin-opt) larvae throughout the animal (A) and in the spinal cord (B). The arrowhead points to one Rohon-Beard cell in the dorsal spinal cord. The white dashed lines delineate the ventral and dorsal limits of the spinal cord. As previously stated for Tg(mnx1:gal4; UAS:GFP-Aequorin-opt) larvae, we verified on 5 animals that expression never targeted muscle fibers in these double transgenic animals. Ventral is bottom, rostral is left. (C) Bioluminescence signals emitted from 10 isl2b+following an acoustic stimulus during motion (blue) or after blockade of muscle contraction using pancuronium bromide in the bath (red). (D) Quantification of the amplitude of bioluminescence signals across fish in active and paralyzed larvae (4.53 +/- 0.21 versus 2.20 +/- 0.10 photons / 10 ms, n = 10 fish, n = 600 trials, p < 0.001).

https://doi.org/10.7554/eLife.25260.019

2) More extensive analysis of the connectivity data between RB neurons and V2a neurons, including increases in the number of independent observations. Latency repeatability and jitter should be critically evaluated both for the EPSC and for the spike response of V2a neurons to RB cell photoactivation.

We thank reviewers for pointing at this critical aspect. We analyzed the connectivity of 4 additional RB-V2a pairs, reaching a total of 8 out of 8 V2a interneurons in 8 different larvae receiving synaptic currents from ipsilateral RB neurons.

To demonstrate the monosynaptic nature of the RB-V2a connection, we first analyzed the time- to-peak of light-evoked action potentials in CoChR+ RB neurons (5-ms blue light stimulations, n = 103 stimulations recorded in 6 cells) as well as the delay of ESPCs in chx10+V2a neurons following RB stimulations (78 EPSCs recorded in 8 cells). Our analysis shows that RB fire single action potentials in average 2.8 ms ± 0.3 ms after the onset of the blue light pulse (see time-to- peak in Figure 7C and G). EPSCs in V2a neurons were detected on average 3.7 ms ± 0.2 ms after the onset of the blue light pulse (see EPSC onset on Figure 7E, G). This means that the lag between RB spiking and V2a EPSC was 0.9 ms on average, which is compatible with a monosynaptic connection from RB neurons onto V2a interneurons.

As a reference, in Xenopus tadpole, Wen-Chang Li in the Roberts lab used for double patch clamp recordings the criterion of 3 ms to distinguish mono- from polysynaptic currents (Li et al., 2007 Neural Development 2:17). In their conditions, monosynaptic connections identified by double patch clamp recordings were associated with delays ranging from 0.94 to 1.78 ms (mean, ± 0.10 ms, see Li et al., 2004 PNAS 101, 15488-493). From our study here on the connections from RB neurons onto ipsilateral V2a interneurons, we can estimate that the average axon distance to travel was approximately ~ 300 µm. Assuming a synaptic delay of 0.5 ms, the calculated conduction velocity was 0.75 m / s – a value compatible with previous reports in Xenopus tadpoles (Li et al.,2004 PNAS 101, 15488-493; Li et al.,2007 Neural Development 2:17).

We also added in the text a description of the EPSCs amplitude recorded in V2a interneurons, as well as the estimated depolarization induced in V2a cells (see lines subsection “Mechanosensory neurons synapse onto V2a interneurons involved in fast locomotion” in Results).

Our new results and analysis from physiology of the RB-V2a synapse are now detailed in the main text, subsection “Mechanosensory neurons synapse onto V2a interneurons involved in fast locomotion”. We added in Figure 7G the distribution of time-to-peaks for each RB cells and the EPSC delays for each V2a neurons. For each cell, we plotted the data’s standard deviation to illustrate the jitter.

Note that the jitter of EPSCs recorded in V2a neurons reflects both the variability in spike timing of the RB neurons in response to the optical stimulation as well as the variability in synaptic transmission from RB to V2a neurons. We therefore did not feel comfortable in using the jitter as an argument for monosynaptic connection, as it is used for double patch clamp recording where the precise timing of the presynaptic neuron is precisely known.

We would like to be reassured that the nature of the connection is truly distinct from the one revealed by prior work, and is thus capable of influencing V2a activity during swimming.

Previous work from the Roberts lab in Xenopus tadpole showed connections from RB neurons onto commissural excitatory interneurons (dIc and ecINs, see Li et al.,2007 Neural Development 2:17; Roberts et al.,2011 Developmental Neurobiology 575-584), which enable a delayed excitation onto contralateral dINs and motor neurons.

Here, we show for the first time that the dorsalmost chx10+ V2a interneurons recruited during fast locomotion as shown previously (McLean et al.,2007 Nature 1;446(7131):71-5, McLean et al.,2008 Nature Neuroscience 11(12):1419-29) are receiving direct monosynaptic inputs from ipsilateral RB neurons. These connections from ipsilateral RB onto dINs were rarely observed by the group of Roberts (see Li et al.,2007 Neural Development 2:17), most likely because the synapses from RB onto ipsilateral dorsalmost V2a neurons are located dorsally at the midline, which is altered by sectioning for the open book tadpole preparation.

As described above, we have strong evidence that the projection we discovered using anatomy and optogenetics from RB neurons onto ipsilateral V2a dorsalmost interneurons is monosynaptic. Of course, we cannot exclude that there may be polysynaptic inputs onto V2a interneurons in addition to this monosynaptic component.

This result has important implication for the sensory modulation of motor circuits controlling speed in the spinal cord. As stated now in the Results section, EPSC amplitude induced by RB stimulation in V2a interneurons were on average 28.14 ± 3.9 pA, leading to estimated depolarization of 11.71 ± 2.59 mV on the V2a target (n = 8). The RB to V2a synapse is therefore relevant for determining spiking in V2a. For some V2a tested in current clamp mode, RB inputs were sufficient to elicit an action potential in V2a neurons (see example in Figure 7F).

By this mean, RB neuron recruitment could therefore selectively feedback to interneurons driving high speeds and projecting onto a large pool of motor neurons downstream on the ipsilateral side. This point is now discussed in the Discussion, in light of results in Xenopus and lamprey.

Two important remarks on the relevance of this connection:

a) As mentioned previously, we cannot exclude that the behavioral effects we have described when silencing the output of isl2b+ neurons with BotxLC-GFP could be due to connections from other sensory neurons targeted by the Tg(isl2b:gal4) line -- we therefore added to the revised manuscript a sentence to explain alternative interpretations of the behavior modulation via dorsal root ganglia and/or trigeminal neurons.

b) Interestingly, the connection from RB neurons onto ipsilateral V2a neurons is very selective for dorsal most V2a cells. When RB neurons are recruited at rest by a touch stimuli, an escape is triggered by initiation of a C bend on the contralateral side. This selective connection from RB onto dorsalmost V2a interneurons most likely is not sufficient to induce subsequent motor activity on the ipsilateral side at rest. However, this connection on the contrary could be modulatory and impact speed if RB neurons are recruited during active locomotion to modulate dorsalmost V2a spiking.

The detailed reviews of two of the reviewers are included to help in revision but the focus of revision should be the two major points above.

Reviewer #1

1) The paper is well written within text but important details are missing from the legends that allow easy interpretation of the data. Some figures are too dense as if composed for a journal with restrictions on figure numbers.

All legends must be upgraded so that a reader can easily understand the data. Let me focus on Figure 1 to illustrate what must be done. What is going on in panels A and G? What are the superimposed lines in panel C (I know this but I am just saying be complete)? What is the color code in panel H and how does it relate to panel I. In panel I, I am really confused please explain the traces. Why are there multiple bumps in the left traces and what does the superposition mean in each set of traces? Please write figure legends that allow the data to be easily understood. Figure 3 is also a problem. What do the arrow and the arrowhead point out in panel A? What are we looking at in the left most image of panel C and in the middle image of panel C and left image of panel D? What are the different colors showing? In all graphs like panel E1 the mean and SEM are just too difficult to see especially amid the blue points; make bold.

Split Panels G to K of Figure 1 into a separate figure. Consider splitting Figure 5 also.

We agree with the reviewer. Accordingly, we rewrote the legends to be more explicit. We also split Figure 1 into two figures, new Figure 1 and new Figure 2 by separating the GFP-Aequorin from the GCaMP data, as well as Figure 5, becoming new Figure 6 and new Figure 7 by separating the anatomy from the physiology.

2) I am concerned about the morphological data shown in Figure 1B. Was this observation made in a number of larvae? In Figure 5A and B, was this observation made in a number of larvae. The legend of panel C implies that it was but you must be explicit. In Figure 5G, was this observation made in a number of larvae?

Since bioluminescence signals originating from single neurons are very small, we estimated activity of neuronal populations monitored by counting photons in a 1 ms bin on a single PMT (and the plotted them as number of photons per bin of 10ms. With this approach, single cell resolution is out-of-reach and specificity of the GFP-Aequorin expression pattern to the cells of interest is critical. Consequently, in the Tg(mnx1:gal4;UAS:GFP-Aequorin-opt) transgenic larvae, we had already previously verified that there was strictly no muscle expression by performing anti-GFP immunochemistry and complete larvae imaging on multiple larvae (n = 5, precision added in the main text, subsection “Mechanosensory neurons synapse onto V2a interneurons involved in fast locomotion” and in the Figure 1 legend, Video 1. In all animals tested (5 out of 5), there was strictly no muscle expression. We also added in the revised version the number of larvae analyzed for the old Figure 5A to B (i.e. new Figure 6A toB).

Similarly, note that the connectivity between RB neurons and V2a interneurons has been confirmed using anatomy and electrophysiology in all animals tested. For the anatomy in Figure 6A to B, we observed projections from RB neurons onto ipsilateral dorsalmost V2a neurons in 12 cells from 4 larvae. For the electrophysiology in Figure 7, we confirmed a monosynaptic connection in 8 out of from 8 recorded dorsal V2a neurons from 8 larvae that were selected based on the RB axonal projection observed onto V2a somata. Stimulating CoChR+ RB neurons led to reliable EPSCs with a short delay in all V2a neurons tested. Since the result was unambiguous and observed in all cells recorded based on anatomy, we estimated that these numbers were sufficient to conclude that a monosynaptic connection exists from RB sensory neurons onto dorsalmost V2a interneurons.

3) There is no direct causal evidence that Rohon Beard cells do indeed mediate the speed changes associated with active swimming, because they are not selectively silenced or optogenetically activated. This needs to be more explicitly acknowledged.

We agree with the reviewer. We had originally mentioned this limitation in the manuscript. As we cannot exclude that the behavioral effects we have described when silencing the output of isl2b+ neurons with BotxLC-GFP could be due to connections from other sensory neurons targeted by the Tg(isl2b:gal4) line (see comments and references above), we added to the revised manuscript alternative interpretations of the behavior modulation via dorsal root ganglia or trigeminal neurons (see Abstract, Results Discussion).

4) Given the central effects of pancuronium bromide why wasn't α-bungarotoxin used to immobilize the larvae in the experiments of Figure 2?

Both α-bungarotoxin and pancuronium bromide paralyze animals by blocking the α 7 subunit of ionotropic cholinergic receptors. The concern that such drugs act on cholinergic receptors in the brain as well as on the periphery at the larval stage is as much valid for α- bungarotoxin than pancuronium bromide bath application (Prof. Yoichi Oda, personal communication).

Author response image 4
Amplitude of bioluminescence signals originating from motor neurons does not increase with the intensity of voltage applied to induce acoustic stimuli.

(A) As shown here for 10 fish tested in control conditions and with pancuronium bromide, larger voltages actually led to decreased amplitude of the bioluminescence signals before paralysis (mean bioluminescence amplitudes across fish = 28.3 +/- 1.7; 25.7 +/- 4.6; 14.9 +/- 4.7 photons /10ms for stimuli voltage of respectively 1, 2 or 4 V, p = 0.03) and no correlation after paralysis. Furthermore, one can note that the stimulus voltage used in paralyzed conditions was equal or higher than in control conditions. (B) As shown here for 10 control fish and 10 immotile mutants, larger voltages did not lead to larger amplitude of the bioluminescence signals: there was no correlation in control siblings while immotile mutants showed smaller signals for larger stimuli (mean bioluminescence amplitudes across fish = 13.9 +/- 3.6; 10.5 +/- 6.4; 7.9 +/- 7.3; 5.0 =:- 2.0 photons /10ms for respective stimuli voltage of 2, 3, 5 or 8 V; p = 0.001). Furthermore, one can note that the stimulus voltage used in immotile mutants (mean = 3.9 +/- 0.1 V) was higher than in control conditions (mean = 2.4 +/- 0.01 V, p < 0.001).

https://doi.org/10.7554/eLife.25260.020

I am also concerned that you did not control for stimulus intensity in these experiments. In active fish and during pancuronium paralysis, is the size of the aequorin response dependent of stimulus amplitude? If it is, then comparisons of amplitudes across these two groups is suspect because different stimulus amplitudes were used.

We thank the reviewer for pointing at this potentially confounding factor. In free-tailed animals, intensity of the auditory stimulus was gradually increased until it elicited an escape response in actively moving animals while in paralyzed animals, the intensity of the auditory stimulus was gradually increased until it elicited a bioluminescence signal. In immotile mutants, the intensity was set to the lowest value eliciting a bioluminescence signal. We checked that the amplitude of the bioluminescence signal was not correlated with the intensity of the acoustic stimulation (Author response image 4). In the active versus paralyzed experiment, larger movements, and therefore signals, were actually obtained for smaller stimuli, whereas there was no correlation after paralysis (see Author response image 4, legend). Note also that the voltage controlling the acoustic stimulus was higher in the paralyzed and immotile conditions compared to controls.

Reviewer #3:

1) If RB neurons are encoding proprioceptive information in addition to the cutaneous exteroceptive information they are normally associated with, then activation of RB neurons during acoustic stimuli needs to be confirmed. The prediction is that RB neurons would respond to skin stimulation and acoustic stimulation when the tail can move, but not to acoustic stimuli when the tail is immotile. Without a better idea of whether RB neurons are in fact responding to self generated movements, then it is difficult to attribute the behavioral deficit to them and justify the ensuing circuit busting. For example, it could be that there are neurons in the DRG that are instead responsible.

We agree with the reviewer. As explained above, we tried to perform calcium imaging with single cell resolution on RBs. Due to technical constraint, GCaMPs did not appear sensitive enough for detecting calcium transients associated with single spikes. Nonetheless, we performed GFP- Aequorin experiments using the Isl2b driver and observed signals only when the tail is motile, suggesting mechanosensory neurons (Rohon Beard, Trigeminal or DRG) are recruited during active locomotion. These experiments indicate that at least one cell type among RB, TG and DRG is recruited during active acoustic escapes --- but we cannot pin it down to one or a combination within the three cell types.

Nonetheless, we would like to point out that the ipsilateral connection RB => V2a neurons that we described here – motivated by the effect we saw on speed in the behavior dataset – is actually very interesting by itself:

a) it was never described by Li and Roberts, most likely for technical limitations associated with the open book preparation;

b) the RB-V2a synapse was associated with large EPSCs that could modulate spike timing in the V2a target (“EPSC amplitude induced by RB stimulation in V2a interneurons were on average 28.14 ± 3.9 pA, leading to estimated depolarization of 11.71 ± 2.59 mV on the V2a target (n = 8)”); c) this circuit may enable a speed selective modulation by projecting onto a subset of V2a neurons (the dorsal most ones), as the anatomy of RB neurons is not compatible with projections onto the dendrites of more ventral V2a neurons.

2) There has been quite a bit of work in Xenopus tadpoles studying RB circuitry involved in cutaneous skin reflexes (e.g., Li et al., 2003, J Neurosci.; Li et al., 2004, J Neurophysiol). Abrupt stimulation of RB afferents generates a reflexive response away from the stimulus. Also, as I understand it, they demonstrate that drive from RB cells that potentially arises from movement-related stimulation would largely be filtered out by inhibition (Sillar and Roberts, 1988, Nature). However, RB neurons can reset or increase swimming rhythms when they are stimulated (e.g., Sillar and Roberts, 1992, J Neurosci), so there is clearly the potential for RB neurons to act as they are proposed to here. I think it would be worth explaining more how the findings here fit into previous work in tadpoles, as well as studies of dorsal cells (the RB equivalents) in lampreys (Buchanan and Cohen, 1982, J Neurophysiol; Christenson etal.,1988, Brain Res). I think the role of the RB-V2a connection would be worth considering in the context of struggling as well (Li et al., 2007, J Neurosci).

We agree with the reviewer that this discussion is important and was missing from the original submission. We revised the Discussion accordingly, see new paragraph in Discussion of revised manuscript, see below:

“Previous studies have described a remarkably simple and conserved disynaptic cutaneous skin reflex relying on spinal mechanosensory neurons referred to as Rohon Beard neurons in Xenopus and zebrafish and “dorsal cells” in lamprey (Buchanan and Cohen, 1982; Christenson et al., 1988; Clarke et al., 1984; Easley-Neal et al., 2013; Li et al., 2003; Roberts et al., 1983). […]. While local reflex pathways have been characterized at rest, further studies will decipher the mechanisms underlying the diversity of behavioral effects of stimulating mechanosensory neurons: escapes for single touch at rest (Buchanan and Cohen, 1982; Douglass et al., 2008; Roberts et al., 1983), struggling for continuous stimulation (Li et al., 2007b) and enhancement of speed during active locomotion (this study).”

3) The optogenetic-electrophysiology data as currently presented don't provide sufficient temporal resolution to determine if the RB-V2a connection is monosynaptic or not. If it is, this would be novel, since studies of RB circuits in Xenopus have only revealed indirect connections from RBs to V2a-like neurons (with extremely short latencies though). In fact, the responses have multiple components which would be more consistent with a polysynaptic connection (especially since RB neurons tend to fire once). Given this is not the main goal of the work, I would suggest perhaps stepping back from saying it is a direct connection and instead focus on the excitatory nature of that connection, monosynaptic or not. Of course, this could all be moot if the RBs are not even active during body bending.

We thank the reviewer for pointing at this critical aspect. As stated above, we analyzed the synaptic connectivity from RB neurons in 4 additional V2a interneurons, reaching a total of 8 cells.

To demonstrate the monosynaptic nature of the RB-V2a connection we analyzed the time-to- peak of elicited action potentials in CoChR-tdtomato+ RB neurons in response to 5 ms blue light stimulations (103 stimulations in 6 cells) as well as the delay of ESPCs in chx10+V2a neurons following RB stimulations (78 EPSCs in 8 cells).

Our analysis shows that RB fire single action potentials in average 2.80 ± 0.29 ms after the onset of the blue light pulse. EPSCs in V2a neurons were detected 3.71 ± 0.19 ms after the onset of the blue light pulse. This means that, in average, there is a lag of 0.9 ms between the peak of RB spikes and V2a EPSCs, which is compatible with a monosynaptic component. See details on the jitter for each cell in Figure 7G, subsection “Mechanosensory neurons synapse onto V2a interneurons involved in fast locomotion” and Discussion above.

4) For the botulinum experiments it is largely assumed that the sensory populations are silenced, but it would be better if there was some confirmation in this line of fish. Patch recordings would be ideal, but if not possible a mention of whether the fish fail to respond to touch or not would suffice.

We previously showed evidence that BoTxBLC-GFP was effective at silencing synaptic release in motor neurons, GABAergic sensory neurons as well as glutamatergic V2a neurons (Sternberg et al., Current Biology 2016). In one case, we could perform the proper demonstration that silencing of GABA release was fully effective by recording the postsynaptic target while stimulating the presynaptic GABAergic sensory neurons using optogenetics. This demonstration was achieved in triple transgenic animals Tg(pkd2l1:gal4;UAS:BoTxBLC-GFP; UAS:ChR- mCherry) where the targets of GABAergic sensory neurons, CaP motor neurons, is obvious due to the basket structure formed by the axons of the GABAergic cells (see Hubbard et al., Current Biology 2016; Sternberg et al., Current Biology 2016).

However, to demonstrate full blockade of synaptic release from mechanosensory neurons, we would need to have the opsin CoChR-tdTomato combined with the BotxBLC-GFP in RB neurons expressing Gal4, and a fluorescent tag in one of RB targets, such as CoPA or dorsalmost V2a neurons. This experiment would require to inject the construct (UAS: CoChR-tdTomato) in triple transgenic Tg(isl2b:gal4;UAS:BoTxLC-GFP; Chx10:DsRed) eggs. Unfortunately, we did not raise triple transgenic animals with this combination of transgenes and could not therefore attempt this experiment within few months.

In our silencing experiments, we have indications that in the Tg(isl2b:gal4;UAS:BoTxLC-GFP) double transgenic larvae only a fraction of mechanosensory neurons are silenced:

a) the Tg(isl2b:gal4;UAS:BoTxBLC-GFP) transgenic larvae do not show GFP in all RB neurons: due to variegation of the UAS/Gal4, only a fraction (between 1/3rd to 2/3rd depending on the fish) of the total number of RB neurons are labeled;

b) 2 dpf Tg(isl2b:gal4;UAS:BoTxBLC-GFP) transgenic larvae larvae upon touch stimulation respond in about half the cases – which is compatible with our point a). We added a comment on the effectiveness of our silencing method in the manuscript’s Result section, see below:

“Due to the fact that UAS/Gal4 labels cells in a variegated manner, we only observed BotxLCB- GFP in half the RB neurons so although the isl2b promoter drives expression in all RB neurons, dorsal root ganglia and trigeminal neurons. This suggests that the effect of silencing mechanosensory feedback in our behavior experiments are underestimated in our experiments.”

5) Since the RB driven response is specific to high frequency swimming commensurate with large deflections of the tail, I was expecting that the controls for the aequorin experiments would be zebrafish that are fully embedded so the tail can't move. Is it known if there is still less motor neuron recruitment in this condition comparable to toxin and mutants?

From our previous observations during the study of GABAergic sensory neurons in the spinal cord (Böhm et al., Nature Communications 2016), non paralyzed larvae that are embedded in agar perform very strong local contractions locally, which look like transient “accordion” where the body is firmly contracting against the agar wall. During these local contractions of the tail confined in agar, there is most likely stimulation of RB neurons via friction against the agar. We therefore did not consider that this experimental paradigm would constitute a good negative control.

6) Work in lampreys studying the role of edge-cells seems appropriate to discuss more, since the function appears to be very similar to that proposed here and a lot of that work is performed while the tail is moving. Also the recent development of a rodent preparation is another precedence for neural recordings in moving animals (Hayes etal., 2009, J Neurophysiol; Hochman et al., 2012, Front. Biosci).

We agree and added references to the extensive work performed in lamprey on edge cells in preparations where the tail is free in particular, see Discussion.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

1) There was concern that the authors overstated the case for monosynapticity in the Introduction and Results, subsection “Mechanosensory neurons synapse onto V2a interneurons involved in fast locomotion, given that experiments with a high concentration of divalent ions – or the like – to eliminate (reduce the probability of) disynaptic events were not performed. We suggest that the text starting in subsection “Mechanosensory neurons synapse onto V2a interneurons involved in fast locomotion be modified something like "To assess whether RB-V2A ipsilateral connection might be monosynaptic, we analyzed the lag between RB spikes time-to-peak and V2a EPSCs onset and found that it was below 1 ms (Figure 7G, 0.9 ms), a value compatible with monosynaptic connection (see (Li et al., 2007a) where cutoff is set at 3 ms for monosynaptic connections)." The Introduction can be correspondingly edited.

We edited the corresponding lines as suggested:

Summary: In the spinal cord, we show that connections compatible with monosynaptic inputs from mechanosensory Rohon-Beard neurons onto ipsilateral V2a interneurons selectively recruited at high speed can contribute to the observed enhancement of speed.

Subsection “Mechanosensory neurons synapse onto V2a interneurons involved in fast locomotion”: To assess whether RB-V2A ipsilateral connection might be monosynaptic, we analyzed the lag between RB spikes time-to-peak and V2a EPSCs onset and found that it was below 1 ms (Figure 7G, 0.9 ms), a value compatible with monosynaptic connection (see (Li et al., 2007a) where cutoff is set at 3 ms for monosynaptic connections).

2) Quoting one of the expert reviewers "There is an underlying assumption that all UAS lines (GFP-aequorin-opt and GCaMP6f;cryaa:mCherry) will produce the same expression patterns when paired with this Gal4 (mnx1:gal4) line. Just because the icm23 line (mnx1:gal4-UAS:preEGFP) was characterized by Bohm et al., this does not mean that Gal4 will work the same with different versions of UAS (UAS:GFP-aequorin-opt or GCaMP6f;cryaa:mCherry). Have these lines been characterized? If so, references to that work must be made to show MN specificity/variegated expression patterns (if any)."

So far, we only observed motor neurons when we used the mnx1 promoter to drive expression of GCaMP5G (Sternberg et al., Current Biology 2016) or Gal4 in a UAS:RFP line. However, we cannot exclude that other spinal neurons, distinct from motor neurons, may be targeted by this promoter. Therefore we added in the Material and methods the following lines:

Given the potential variability in expression patterns of Gal4 lines with different UAS, we characterized the Tg(mnx1:gal4;UAS:GCaMP6f,cryaa:mCherry) line and observed a similar expression in motor neurons as in the Tg(mnx1:gal4;UAS:GFP-aequorin-opt)icm09 line. In both cases, we observed the targeting of spinal neurons exiting the spinal cord in the ventral root, and therefore referred to as motor neurons. However, we cannot exclude that the mnx1 promoter may target other spinal neurons, distinct from motor neurons.

https://doi.org/10.7554/eLife.25260.022

Article and author information

Author details

  1. Steven Knafo

    1. Institut du Cerveau et de la Moelle épinière (I.C.M.), Sorbonne Universités, UPMC Univ Paris 06, Inserm, CNRS, Institut du Cerveau et la Moelle épinière, Hôpital Pitié-Salpêtrière, Paris, France
    2. AP-HP, Hôpital Pitié-Salpêtrière, Paris, France
    Contribution
    SK, Data curation, Formal analysis, Investigation, Writing—original draft, Writing—review and editing
    Contributed equally with
    Kevin Fidelin
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon 0000-0002-8733-0429
  2. Kevin Fidelin

    Institut du Cerveau et de la Moelle épinière (I.C.M.), Sorbonne Universités, UPMC Univ Paris 06, Inserm, CNRS, Institut du Cerveau et la Moelle épinière, Hôpital Pitié-Salpêtrière, Paris, France
    Contribution
    KF, Data curation, Formal analysis, Investigation, Writing—original draft, Writing—review and editing
    Contributed equally with
    Steven Knafo
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon 0000-0001-7676-1410
  3. Andrew Prendergast

    Institut du Cerveau et de la Moelle épinière (I.C.M.), Sorbonne Universités, UPMC Univ Paris 06, Inserm, CNRS, Institut du Cerveau et la Moelle épinière, Hôpital Pitié-Salpêtrière, Paris, France
    Contribution
    APr, Data curation, Formal analysis, Investigation, Writing—original draft, Writing—review and editing
    Competing interests
    The authors declare that no competing interests exist.
  4. Po-En Brian Tseng

    Institut du Cerveau et de la Moelle épinière (I.C.M.), Sorbonne Universités, UPMC Univ Paris 06, Inserm, CNRS, Institut du Cerveau et la Moelle épinière, Hôpital Pitié-Salpêtrière, Paris, France
    Contribution
    P-EBT, Data curation, Formal analysis, Investigation
    Competing interests
    The authors declare that no competing interests exist.
  5. Alexandre Parrin

    Institut du Cerveau et de la Moelle épinière (I.C.M.), Sorbonne Universités, UPMC Univ Paris 06, Inserm, CNRS, Institut du Cerveau et la Moelle épinière, Hôpital Pitié-Salpêtrière, Paris, France
    Contribution
    APa, Data curation, Formal analysis
    Competing interests
    The authors declare that no competing interests exist.
  6. Charles Dickey

    Institut du Cerveau et de la Moelle épinière (I.C.M.), Sorbonne Universités, UPMC Univ Paris 06, Inserm, CNRS, Institut du Cerveau et la Moelle épinière, Hôpital Pitié-Salpêtrière, Paris, France
    Contribution
    CD, Data curation, Formal analysis
    Competing interests
    The authors declare that no competing interests exist.
  7. Urs Lucas Böhm

    Institut du Cerveau et de la Moelle épinière (I.C.M.), Sorbonne Universités, UPMC Univ Paris 06, Inserm, CNRS, Institut du Cerveau et la Moelle épinière, Hôpital Pitié-Salpêtrière, Paris, France
    Contribution
    ULB, Data curation
    Competing interests
    The authors declare that no competing interests exist.
  8. Sophie Nunes Figueiredo

    Institut du Cerveau et de la Moelle épinière (I.C.M.), Sorbonne Universités, UPMC Univ Paris 06, Inserm, CNRS, Institut du Cerveau et la Moelle épinière, Hôpital Pitié-Salpêtrière, Paris, France
    Contribution
    SNF, Data curation
    Competing interests
    The authors declare that no competing interests exist.
  9. Olivier Thouvenin

    Institut du Cerveau et de la Moelle épinière (I.C.M.), Sorbonne Universités, UPMC Univ Paris 06, Inserm, CNRS, Institut du Cerveau et la Moelle épinière, Hôpital Pitié-Salpêtrière, Paris, France
    Contribution
    OT, Data curation
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon 0000-0003-4853-7555
  10. Hugues Pascal-Moussellard

    1. Institut du Cerveau et de la Moelle épinière (I.C.M.), Sorbonne Universités, UPMC Univ Paris 06, Inserm, CNRS, Institut du Cerveau et la Moelle épinière, Hôpital Pitié-Salpêtrière, Paris, France
    2. AP-HP, Hôpital Pitié-Salpêtrière, Paris, France
    Contribution
    HP-M, Conceptualization
    Competing interests
    The authors declare that no competing interests exist.
  11. Claire Wyart

    Institut du Cerveau et de la Moelle épinière (I.C.M.), Sorbonne Universités, UPMC Univ Paris 06, Inserm, CNRS, Institut du Cerveau et la Moelle épinière, Hôpital Pitié-Salpêtrière, Paris, France
    Contribution
    CW, Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Project administration, Writing—original draft, Writing—review and editing
    For correspondence
    claire.wyart@icm-institute.org
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon 0000-0002-1668-4975

Funding

European Research Council (311673)

  • Claire Wyart

National Institutes of Health (5U01NS090501)

  • Claire Wyart

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

Acknowledgements

We would like to thank Dr. Ludovic Tricoire (University Pierre et Marie Curie, France) for providing the original GFP-Aequorin sequence, Dr. Michael Granato (University of Pennsylvania, USA), Dr. Paul Brehm (Vollum Institute, USA) for providing the relaxed (cacnb1ts25) mutants, Dr. Maximilliano Suster and Prof. Koichi Kawakami (NIG, Japan) for sharing the Tg(UAS:BoTxBLC-GFP) transgenic line, Dr. Tom Auer and Dr. Filippo Del Bene (Institut Curie, France) for providing the mnx1 and isl2b constructs and the Tg(isl2b:gal4,cmlc2:eGFP) line, and Dr. Minoru Koyama (Janelia Farm, USA) for the UAS:CoChR-tdTomato construct. We thank Dr. Andrew Straw (University of Freiburg, Germany), Dr. Björn Brembs (University of Regensburg, Germany), and Dr. William Kristan (UCSD, USA) for their valuable feedback. We are indebted to RD Vision for developing the custom API to synchronize photon collection and video acquisition. We thank Bogdan Buzurin, Monica Dicu and Natalia Maties for fish care. SK received a PhD fellowship from Inserm and Assistance Publique – Hôpitaux de Paris, KF from University Pierre et Marie Curie, and UB from Ecole des Neurosciences de Paris (ENP). AP received a postdoctoral fellowship from the Mairie de Paris Research in Paris program and from the EMBO. This work received financial support from the Institut du Cerveau et de la Moelle épinière with the French program ‘Investissements d’avenir’ ANR-10-IAIHU-06, the ENP Chair d’excellence, the Fondation Bettencourt Schueller (FBS), the City of Paris Emergence program, the ATIP/Avenir junior program from INSERM and CNRS, the NIH Brain Initiative grant no. 5U01NS090501 and the European Research Council (ERC) starter grant ‘OptoLoco’ #311673. 

Ethics

Animal experimentation: All procedures were approved by the Institutional Ethics Committee at the Institut du Cerveau et de la Moelle épinière (ICM), Paris, France, the Ethical Committee Charles Darwin and received subsequent approval from the EEC (2010/63/EU).

Reviewing Editor

  1. Ronald L Calabrese, Reviewing Editor, Emory University, United States

Publication history

  1. Received: January 19, 2017
  2. Accepted: June 17, 2017
  3. Accepted Manuscript published: June 17, 2017 (version 1)
  4. Accepted Manuscript updated: June 19, 2017 (version 2)
  5. Version of Record published: July 6, 2017 (version 3)

Copyright

© 2017, Knafo et al.

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

Metrics

  • 1,482
    Page views
  • 313
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: PubMed Central, Scopus, Crossref.

Comments

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)