Identification of V2a neurons for recordings during ‘fictive’ swimming

A) Composite DIC/epifluorescence image of Tg[Chx10:GFP] larval zebrafish at 5 days post fertilization (5d). Scale bar, 0.5 mm.

B) Confocal image of the Tg[Chx10:GFP] spinal cord from segments 10-15 at midbody (purple box in A). Scale bar, 30 μm.

C) Pseudo-colored images of V2a somata based on high (H, white) and low (L, purple) GFP intensity. Left panel: sagittal section, right panel: coronal section. Dashed lines indicate boundaries of spinal cord, which are normalized to 0-1 in the dorso-ventral and medio-lateral axes. White arrowheads indicate large, lateral V2a somata with low levels of GFP expression. Scale bar, 20 μm.

D) Plots of numbers of low and high intensity V2a neurons per midbody segment (n = 5-6 segments from 3 fish). Left panel: Bar plot of average number (+/- SEM) of low and high GFP V2a neurons along the dorsoventral (DV) axis. There is no significant difference in the medians of the dorsoventral distributions of high and low GFP expressing V2a neurons (Wilcoxon rank-sum rest; W = 68479, p = 0.6975, n = 347 high and 401 low GFP neurons from 3 fish). Right panel: Box and whisker plot showing average number of high and low GFP neurons per segment (Wilcoxon rank-sum test, W = 88, p = 0.05269, n = 17 segments in 3 fish).

E) Two-dimensional contour density plots showing the mediolateral (ML) and dorsoventral (DV) distributions of V2a neurons expressing low or high levels of GFP. Density distributions of ML positions of V2a neurons are shown at the bottom of the contour plots, which are significantly different (***, Two-sample Kolmogorov-Smirnov test, D = 0.16452, p = 8.46×10-05, n = 347 high and 401 low GFP neurons from 3 fish).

F) Top panel: an example of an extracellular ventral rootlet recording of three consecutive swim episodes evoked by a mild electrical stimulus (s) to the tail fin (at arrowhead, artifact blanked). Scale bar, 500 ms. Bottom left panel: a cartoon of the locations of the stimulus and the recording electrodes. Bottom right panel: individual ventral rootlet bursts (vrb) on an expanded timescale. Black dots mark the center of each vrb used to calculate swim frequency in Hz (1/s), which is indicative of swimming speed. Scale bar, 20 ms.

G) Scatter plot of vrb frequency (Hz) versus the time in swim episode (ms). The swimming episode immediately following the stimulus is called Swim 1 (left), and any subsequent swimming episodies are called Swim 2+ (right).

Cell-type-specific electrophysiological properties linked to recruitment order

A) Composite DIC/epifluorescent images of post-hoc fills of V2a-Descending (V2a-D, top) and V2a-Bifurcating (V2a-B, bottom) neurons. Arrows indicate continuation of axons outside the field of view. Scale bar, 10 μm.

B) Left panel: scatter plot of soma positions of V2a-D and V2a-B neurons normalized to dorsoventral (DV) position, with 1 demarcating the dorsal and 0 the ventral boundaries of the spinal cord. Right panel: boxplots of dorso-ventral soma positions of V2a-D and V2a-B neurons classified as tonic, bursting, or chattering based on their step-firing responses in panels E and F. TD= tonic V2a-D, CB=chattering V2a-B, BB=bursting V2a-B, BD=bursting V2a-D.

C) Left panel: scatter plot of input resistance (GΩ) versus the rheobase (pA) of V2a-B (black) and V2a-D (blue) neurons. Right panel: Box plots of input resistance (GΩ) of BD, TD, BB, and CB V2a neurons.

D) Mean (+/– SD) instantaneous spike rates (Hz) between 1-2x rheobase for TD, CB, BD and BB neurons.

E) Top panel: step-firing responses of V2a-D neurons defined as tonic (TD). Bottom panel: step-firing responses of V2a-D neurons defined as bursting (BD). Grey dots depict the midpoint of slower membrane oscillations driving the bursting spiking behavior. Same scale as panel F.

F) Top panel: step-firing responses of V2a-B neurons defined as chattering (CB). Bottom panel: V2a-B neurons defined as bursting (BB). Grey dots depict the midpoint of slower membrane oscillations driving the bursting spiking behavior. Scale bars, 20 mV, 200 pA, 200 ms.

G) Top panels: Step-firing responses close to rheobase of TD, BD, CB, and BB neurons. Bottom panels: Swim firing responses evoked by a mild electrical shock shown on the same time scales. Black dots on ventral root bursts (grey) indicate swim frequency. Black arrows mark the stimulation artifact. Scale bars, 10 mV, 50 ms.

H) Scatter plot of median swim firing frequency (Hz) versus the input resistance (GΩ) of CB, TD, BB, and BD on logarithmic x- and y-scales. Shaded grey box indicates 30-40 Hz range reflecting transition between anguilliform and carangiform swim modes.

I) A comparison of median swim-firing and median step-firing frequencies between 1-2x rheobase (+/– SD) for TD, BD, and BB neurons. For TD neurons, step frequencies represent spike rates, while for BD and BB neurons they represent burst rates. CB neurons were too variable in step firing frequency to obtain reliable median values. Shaded grey box indicates 30-40 Hz range reflecting transition between anguilliform and carangiform swim modes.

Differences in recruitment patterns among the distinct V2a types

A) Left panel: Cell-attached (ca) recordings of descending V2a (V2a-D) neurons during fictive swimming evoked by a brief stimulus (at black arrowhead). Black dots on ventral root (vr) bursts (grey) indicate swim frequency. The top neuron fires immediately after the stimulus during high frequency swimming, while the bottom neuron fires near the end of the episode at lower frequencies. Right panel: As shown to the left, but for bifurcating V2a (V2a-B) neurons. Scale bar, 100 ms.

B) Top panel: spike timing of V2a-D (left) and V2a-B (right) neurons relative to the start of swimming ordered by the median of the distribution. Bottom panel: A scatterplot of vrb frequency (Hz) as a function of time for all episodes (Swim 1 and 2+) during which the V2a-D (left) or V2a-B (right) neurons were recruited.

C) Raster plots of ventral root burst frequencies (Hz) over which V2a-Ds were recruited ordered by the median vrb recruitment frequency (Hz). Shaded box indicates 30-40 Hz range reflecting transition between anguilliform and carangiform swim modes.

D) Raster plots of ventral root burst frequencies (Hz) over which V2a-Bs were recruited ordered by the median vrb recruitment frequency (Hz). Shaded box indicates 30-40 Hz range reflecting transition between anguilliform and carangiform swim modes.

Distinct types of slow and fast V2a neurons distinguished based on recruitment

A) Top panel: Cell-attached (ca) and ventral rootlet (vr) recordings during slow (<35 Hz) and fast (>35 Hz) swimming illustrate V2a-D neurons (blue) that fire exclusively at slow and fast speeds. Bottom panel: V2a-B neurons (black) fire more reliably at both slow and fast speeds. Scale bar, 25 ms.

B) Plots of the mean (+/– SD) dorsoventral (DV) positions of fast (FD) and slow (SD) V2a-Ds, and fast (FB) and slow (SB) V2a-Bs versus their mean (+/– SD) input resistance (Rin).

C) Spike timing of slow (S) and fast (F) V2a-D and V2a-B neurons relative to the center of the ventral root burst (vrb) depicted by the dashed vertical line. Tick marks represent individual spikes from multiple cycles and circles represent median spike timing.

D) Density plots of spike timing relative to the center burst for slow and fast V2a-B and V2a-D neurons illustrated in panel C.

E) Top panel: Box plots of the percentage of ventral root bursts with spikes (vrb spikes %) as a function of vrb frequency (Hz) for fast (FD, n = 9) and slow (SD, n = 8) V2a-D neurons. For box plots, the ventral root burst frequency was binned at 15 Hz intervals starting at 15 Hz, and the final interval of 45+Hz consisted of a 20 Hz range spanning from 45-65 Hz. Box plots are superimposed on trendlines from individual neurons fit to data binned at 5 Hz intervals, whose slopes define whether they are fast (positive) or slow (negative). Dashed line indicates transition between slow carangiform and fast anguilliform modes. Bottom panel: As above but for fast (FB, n = 16) and slow (SB, n = 6) V2a-B neurons.

F) Top panel: Box plots of vrb spikes % for slow and fast V2a-D neurons from panel E collapsed into single slow (<35 Hz) and (>35 Hz) bins for purposes of statistical analysis (Wilcoxon signed-rank test; FD, V = 1, p = 0.007812, n = 9; SD, V = 36, p = 0.007813, n = 8). Bottom panel: As above but for V2a-B neurons (Wilcoxon signed-rank test: FB, V = 0, p = 6.104e-05, n = 16; SB, V = 20, p = 0.0625, n = 6).

G) Top panel: Scatter plots of instantaneous spike rates versus vrb frequency (Hz) for V2a-D (left) and V2a-B (right) neurons color coded as fast (dark) and slow (light). FD, n = 9; SD, n = 8; FB, n = 16; SB; n = 6, Bottom panel: Scatter plots of the number of spikes per ventral root burst (#) as a function of vrb frequency (Hz), organized and color coded as above.

H) Top panel: Box plots of maximum instantaneous spike rates for fast (n = 5) and slow (n = 7) V2a-D and fast (n = 16) and slow (n = 5) V2a-B neurons. Note, only a subset of V2a neurons that fired two or more spikes per cycle were included in this analysis (Wilcoxon rank-sum test; FB-FD, W = 73, p = 0.00412,; FB-SD, W = 90, p = 0.02247). Bottom panel: Box plots of the maximum number of spikes/vrb (Wilcoxon rank-sum test; FB-FD, W = 132.5, p = 0.0005348,; FB-SD, W = 100.5, p = 0.024; FD-SB, W = 50, p = 0.00667; FD, n = 9; SD, n = 8; FB, n = 16; SB, n = 6).

Differences in levels of synaptic drive related to speed and V2a cell type

A) Voltage-clamp (vc) recordings of excitatory post-synaptic currents (EPSCs) received by fast (top, FD) and slow (bottom, SD) V2a-D neurons along with ventral root (vr) recordings during fictive swimming triggered by an electrical stimulus (black arrowheads, artifact blanked). Black dots indicate burst intervals for frequency measures. Grey shaded boxes indicate region expanded inset, illustrating the holding potential (dashed line, -75 mV) and individual EPSCs (tick marks). Scale bars, 25 pA, 100 ms (20 ms inset).

B) As in A, but for fast (top, FB) and slow (bottom, SB) V2a-B neurons.

C) Scatter plot of maximum EPSC amplitude (pA) as a function of input resistance (GΩ) on logarithmic x- and y-scales. Dashed logarithmic trendlines are included for illustrative purposes (Spearman’s rank correlation test, Rho = -0.8730769, p < 2.2 x 10-13, n = 39).

D) Voltage-clamp recordings of inhibitory post-synaptic currents (IPSCs) at a holding potential of 10 mV from fast (top, FD) and slow (bottom, SD) V2a-D neurons, organized as detailed in panel A. Scale bars, 100 pA, 100 ms (20 ms inset).

E) As in D, but fast (top, FB) and slow (bottom, SB) V2a-B neurons.

F) Scatter plot of maximum IPSC amplitude as a function of input resistance (GΩ) on logarithmic x- and y-scales (Spearman’s rank correlation test; Rho = -0.6332046, p = 5.041 x 10-05, n = 36).

G) Top panel: Box plots of the maximum EPSC per cycle as a percentage of the maximum current (max EPSC%) for fast (FD, n = 9) and slow (SD, n = 6) V2a-D neurons, organized into 4 bins based on frequency. Dashed line indicates transition between slow carangiform and fast anguilliform modes. Bottom panel: as above, but for fast (FB, n = 11) and slow (SB, n = 5) V2a-B neurons.

H) Top panel: Box plots of max EPSC% for fast and slow V2a-D neurons from panel G, collapsed into single slow (<35 Hz) and fast (>35 Hz) bins for purposes of statistical analysis (Wilcoxon signed-rank test; FD, V = 0, p = 0.003906, n = 9; SD, V = 21, p = 0.03125, n = 6). Bottom panel: As above but for V2a-B neurons (Wilcoxon signed-rank test; FB, V = 2, p = 0.001953, n = 11; SB, V = 15, p = 0.0625, n = 5). Boxed plots are superimposed on trendlines from individual neurons. Note, 1 cell each for V2a-D and V2a-B slow subtypes behaved more like fast subtypes (dashed lines).

I) Top panel: Box plots of the maximum IPSC per cycle as a percentage of the maximum current (max IPSC%) for fast (FD, n = 7) and slow (SD, n = 7) V2a-Ds organized as in panel G. Bottom panel: As above but for fast (FB, n = 12) and slow (SB, n = 6) V2a-B neurons.

J) Top panel: Box plots of max IPSC% for fast and slow V2a-D neurons from panel I, collapsed into single slow (<35 Hz) and fast (>35 Hz) bins for purposes of for statistical analysis (Wilcoxon signed-rank test; FD, V = 0, p = 0.01563, n = 7; SD, V = 1, p = 0.03125, n = 7). Bottom panel: as above, but for fast and slow V2a-B neurons (Wilcoxon signed-rank test; FB, V = 0, p= 0.0004883, n = 12; SB, V = 4, p = 0.2188, n = 6).

K) Box plots of maximum excitatory and inhibitory current for FD (Wilcoxon signed-rank test; V = 0, p = 0.00781, n = 9 for excitation, n = 8 for inhibition) and SD (Wilcoxon signed-rank test; V = 0, p = 0.01563, n = 8 for excitation, n = 7 for inhibition) neurons.

L) As in K but for FB (V = 1, p = 0.0001221, n = 16 for excitation, n = 15 for inhibition) and SB (V = 0, p = 0.03125, n = 6 for excitation, n = 6 for inhibition) neurons.

Differences in timing of synaptic drive related to speed and V2a cell type

A) Raw (grey) and low-pass filtered (black) traces of post-synaptic current (PSC), normalized to the peak (1) and trough (0) of phasic excitation (left) and inhibition (right). Corresponding ventral root (vr) bursts used to define in- and anti-phase (gray shaded boxes) currents are shown below. Block dots denote the peaks identified by the findpeaks() function in MATLAB. Note, excitation has been inverted to simplify comparisons to inhibition.

B) Left panel: Circular bar plots of spike phase for fast (FD, n = 9) and slow (SD, n = 8) V2a-D neurons. Right panel: Circular bar plots of phase normalized peaks in low pass filtered EPSCs (FD, n = 9; SD, n = 7) and IPSCs (FD, n = 7; SD, n = 7). Arrowhead indicates prominent in-phase IPSC component for slow V2a-Ds. Scale bars, 10% total distribution.

C) Left panel: Circular bar plots of spike phase for fast (FB, n = 16) and slow (SB, n = 6) V2a-B neurons. Right panel: Circular bar plots of phase normalized peaks in low pass filtered EPSCs (FB, n = 11; SB, n = 6) and IPSCs (FB, n = 12; SB, n = 6). Arrowhead indicates prominent in-phase IPSC component for fast V2a-bs. Scale bars, 10% total distribution.

D) Line plots from V2a-D neurons of averaged excitation (Ex; FD, n = 9; SD, n = 6), averaged inhibition (In; FD, n = 7; SD, n = 7), and spike densities (S; FD, n = 7; SD, n = 7) broken into four ventral root burst frequency (Hz) bins. Post synaptic current was normalized on a cycle-by-cycle basis, such that the minimum and maximum current per cycle equaled 0 and 1 respectively. Scale bar for spike densities, 50% total distribution.

E) As in panel D, but averaged excitation (Ex; FB, n = 11; SB, n = 5), inhibition (In; FB, n = 12; SB, n = 6), and spike densities (S; FB, n = 16; SB, n = 6) for V2a-B neurons.