A faithful internal representation of walking movements in the Drosophila visual system

Journal name:
Nature Neuroscience
Volume:
20,
Pages:
72–81
Year published:
DOI:
doi:10.1038/nn.4435
Received
Accepted
Published online

Abstract

The integration of sensorimotor signals to internally estimate self-movement is critical for spatial perception and motor control. However, which neural circuits accurately track body motion and how these circuits control movement remain unknown. We found that a population of Drosophila neurons that were sensitive to visual flow patterns typically generated during locomotion, the horizontal system (HS) cells, encoded unambiguous quantitative information about the fly's walking behavior independently of vision. Angular and translational velocity signals were integrated with a behavioral-state signal and generated direction-selective and speed-sensitive graded changes in the membrane potential of these non-spiking cells. The nonvisual direction selectivity of HS cells cooperated with their visual selectivity only when the visual input matched that expected from the fly's movements, thereby revealing a circuit for internally monitoring voluntary walking. Furthermore, given that HS cells promoted leg-based turning, the activity of these cells could be used to control forward walking.

At a glance

Figures

  1. Leg movement modulates the membrane potential dynamics of LPTCs.
    Figure 1: Leg movement modulates the membrane potential dynamics of LPTCs.

    (a) Schematics of the experimental set-up (left) and of the behavioral recording system (right). (b) Left, a head-fixed fly on a ball (side view). Right, schematic of the fly's optic lobe (horizontal-plane view). Information from the retina is processed through the lamina, medulla, lobula and lobula plate (highlighted in gray). (c) Left, a confocal image of the GMR81G07 fly shows HS cells (HSN, HS-North; HSE, HS-Equatorial; HSS, HS-South) marked with GFP. Scale bar represents 50 μm. Right, a segment of simultaneous recording from an HSE cell (membrane potential, Vm; indicated is the mean Vm at quiescence) and the fly's behavior (yaw and pitch velocities of ball rotation). (d) Left, a confocal image of the GMR39E01 fly shows VS cells (VS1–5 cells) marked with GFP. Scale bar represents 50 μm. Right, data are presented as in c for a VS1 cell. (e) Data are presented as in c for an HSE cell in a blind fly (norpA background). (f) Top, changes in Vm in behaving versus quiescent segments in VS cells (magenta), and in HS cells in WT (black) and blind flies (gray). Bottom, s.d. of Vm of the cells in behaving versus quiescent segments. (g) SNR of HS (blind flies, gray; WT, black) and VS cells (magenta). The SNR in HS cells (WT) was significantly larger than that in VS cells (HS cells, mean ± s.d. = 3.5 ± 0.6; VS cells, mean ± s.d. = 1.8 ± 0.6; ***P < 0.0001, Z = 4.15, Wilcoxon's rank-sum test). The SNR between HS cells in WT versus blind flies was not significantly different (mean ± s.d. = 4.3 ± 1.5, P = 0.26, Z = 1.13, Wilcoxon's rank-sum test). N = 19 cells (HS, WT flies), N = 9 cells (HS, blind flies), N = 10 cells (VS, WT flies).

  2. A behavioral-state signal in HS cells precedes the onset of, and continues after, the offset of behavior.
    Figure 2: A behavioral-state signal in HS cells precedes the onset of, and continues after, the offset of behavior.

    (a) Vm triggered at the onset of postural adjustments of the example cell from Figure 1c in walking bouts (left) or activity bouts (right). The black trace highlights the mean across individual bouts (gray traces). (b) Population mean Vm (mean ± s.e.m., black line and gray shade, respectively) triggered at the onset of (left) or at the offset of (right) walking bouts. (c) Data are presented as in b but for activity bouts. (d) Left, distribution of the 10% rise time of the Vm (Online Methods) of walking (black) or activity (gray) bouts. Right, distribution of the 90% decay of the Vm (Online Methods) of walking (black) or activity (gray) bouts. Arrowheads indicate the mean values of the distributions. For walking onset: −71 ± 46 ms, (mean ± s.d., N = 19 cells, P < 0.001, Z = −3.82, Wilcoxon signed rank test). For activity onset: −69 ± 42 ms (mean ± s.d., N = 11 cells, P < 0.001, Wilcoxon signed-rank test). These distributions are not significantly different from each other, P = 0.80 (Z = 0.26, Wilcoxon rank-sum test). For walking offset: 69.00 ± 109 ms (mean ± s.d., N = 16 cells, P = 0.008 (Z = 2.64, Wilcoxon signed-rank test). We could only fit a sigmoid function in two cells for activity offset; thus, we did not run statistical tests for this data.

  3. The dynamics of HS cells during walking bouts in darkness correlates with the angular and forward velocities of the fly.
    Figure 3: The dynamics of HS cells during walking bouts in darkness correlates with the angular and forward velocities of the fly.

    (a) Left, definition of the fly's velocity components. Right, example of a 1-min simultaneous recording of the Vm of a right-side HSN cell (top trace) and the fly's walking (middle trace, angular velocity; bottom trace, forward velocity). Dotted line in Vm represents the mean value of Vm at quiescence. (b) Cross-covariance analysis (mean ± s.d., solid line and shaded area, respectively) between the fly's angular (left) or forward (right) velocity and the Vm of the cell. CovV,Vm (τ) = coefficient of cross-covariance as a function of a lag, τ. Negative lags indicate that the fly's behavior preceded the Vm modulation (body first). Peak coefficient's lags (τpeak) and number of walking bouts >1.5 s (n) are indicated. The dark gray line in all cross-covariance plots shows the bootstrapped 95% confidence level of the analysis (Online Methods). (c,d) Data are presented as in a and b, but for a left-side HSE cell. (e) Cross-covariance analysis for the population of right- (gray) and left-side (black) HS cells for the fly's angular velocity (left) or forward velocity (right) (N = 8 cells, n = 477 bouts for left cells; N = 11 cells, n = 623 bouts for right cells). (f) Distributions of τpeak for the angular velocity versus Vm (black bar), or forward velocity versus Vm (white bar). τpeaks were significantly different between angular (mean ± s.d. = −119.8 ± 34.7 ms; N = 19 cells) and forward (mean ± s.d. = −19.8 ± 45.6 ms; N = 15 cells) walking modulations (P < 0.0001, Z = 4.91, Wilcoxon's rank-sum test). (g) Left, population velocity tuning map of left-side HS cells (color-coded, with respect to quiescence) across the population of left side HS cells (N = 8). Right, data are presented as on the left, but for the population of right-side HS cells (N = 11 cells). Normalized data points (on a log scale, total number of data points per bin / maximal number of data points) are shown next to each speed-tuning map.

  4. The behavioral-state signal arrives to HS cells through upstream visual circuits.
    Figure 4: The behavioral-state signal arrives to HS cells through upstream visual circuits.

    (a) Example of a 1-min simultaneous recording of the Vm of a right-side HSN cell (top), the angular (middle) and forward velocity (bottom) of a fly with silenced T4/T5-cells. Dotted line indicates the mean value of Vm in quiescence. Note the appearance of inhibitory potentials associated with NDmotor. (b) Left, cross-covariance between the Vm of the example cell and the angular (top), or forward velocity (bottom) of the fly (mean ± s.d., solid line and shaded area, respectively). The gray line indicates the bootstrapped 95% confidence level of the analysis. Right, mean population cross-covariance analysis (mean ± s.e.m., solid line and shaded area, respectively) for the left- (black) and the right-side (gray) HS cells recorded in flies with silenced T4/T5 cells, for the angular (top) and the forward (bottom) fly velocities. (c) Left, grand mean (± s.e.m.) of the HS cells' Vm triggered at the onset of postural adjustments for isolated walking bouts in flies with silenced T4/T5-cells (magenta) and the corresponding genetic controls (black, T4/T5-Gal4; gray, UAS-Kir). Right, the distributions of the 10% rise-time of Vm. Arrowheads indicate the mean values. For T4/T5-silenced flies: 48.7 ± 78.6 ms (mean ± s.d., N = 9). For T4/T5-GAL4 control flies: −22.0 ± 28.4 ms (mean ± s.d., N = 9). For UAS-Kir control flies: −22.5 ± 33.1 ms (mean ± s.d., N = 8). The distribution between parental controls and the T4/T5-silenced flies was different (P < 0.05, H = 8.02, Kruskal-Wallis multiple comparisons, followed by the Tukey-Kramer post hoc test). (d) Walking velocity tuning of the example cell shown in a (top, left), and of the HS cell population (top, right) of flies with silenced T4/T5 cells or their genetic controls (bottom). Color code: ΔVm, change in Vm with respect to quiescent segments. Note the unmasked Vm hyperpolarization associated with NDmotor in flies with silenced T4/T5 cells.

  5. HS cells are tuned by a weighted sum of three movement-related components.
    Figure 5: HS cells are tuned by a weighted sum of three movement-related components.

    (a) Schematic of the model. (b) Linear filters for angular (Va, red) and forward velocities (Vf, blue) and the behavioral state (BS, green). Top, right-side cells. Bottom, left-side cells. (c) Top, the observed (black) and predicted Vm for an example cell obtained with a single-component model (BS, orange). Second row, data are presented as in the top for a three-component model (Va + Vf + BS, magenta) and for a two-component model (Va + BS, green). Bottom three rows, the corresponding behavioral inputs. Arrowheads highlight a segment with low angular velocity. (d) Cross-correlation coefficients between the observed and predicted Vm for the three-component model, for matched and mismatched pairs (Online Methods; lines indicate mean values; ** indicate matched pairs, mean ± s.d. = 0.69 ± 0.09; mismatched pairs, mean ± s.d. = 0.01 ± 0.02; P < 0.001, Z = 3.82, Wilcoxon signed-rank test, N = 19 cells). (e) Walking velocity tuning predicted from the three-component model (left) and the observed one from the population of recordings (right, N = 19 cells). (f) The difference between the predicted and the observed velocity tuning maps z-scored. 94% of the data points lie within the Z = 1 contour, indicating the high similarity between the velocity maps.

  6. HS cells encode positive combinations of visual and angular velocities.
    Figure 6: HS cells encode positive combinations of visual and angular velocities.

    (a) Top, schematic of recording conditions. Bottom, example traces of simultaneous recordings of walking and HS cell activity for a closed-loop (middle) or an open-loop trial (bottom, replay). (b) Population mean observed change in Vm (normalized) of HS cells as a function of the estimated visual and motor modulations. θ, interaction angle between visual and motor modulations (Online Methods). (c) The mean change in Vm projected onto a 2D visual and angular velocities space during replay trials. PD/ND, preferred/nulled direction. For clarity of display, CW visual velocity > 0 (PDvision). Expected (not expected) is the expected (not expected) direction of the visual velocity produced by the turn direction of the fly if the trial occurred under closed-loop, natural conditions. (d) Left, correlation analysis from random-forest decoders between the prediction of different visuomotor interactions (the sum of visual and angular velocity signals, Vv + Va; or their difference, Vv – Va), and their observed values. The decoder performed better when predicting the positive combination of the visual and the turning velocities than when predicting their negative combination (V – M, mean ± s.d. = 0.23 ± 0.13; V + M, mean ± s.d. = 0.36 ± 0.11; P < 0.001, Wilcoxon's signed-rank test; N =13). Right, cell classification from the correlation analysis, color coded: (corrcoef(Vv + Va) – corrcoef(Vv – Va))/(corrcoef(Vv + Va) + corrcoef(Vv – Va)).

  7. HS cells control turning in the walking fly.
    Figure 7: HS cells control turning in the walking fly.

    (a) The ATP-gated cation channel P2X2 was misexpressed in HS cells. Application of a single pulse of ATP (10 mM, 30-ms pulse) at the terminal site of HS cells induced an expected depolarization in HS cells (n = 5 trials, mean ± s.d.). Dotted line indicates the time of ATP application. Inset, smaller timescale. (b) Left, schematic of the experiment. Application of pulses of ATP to either the right-side (middle) or left-side (right) brain hemispheres of an example fly induced ipsilateral turning. (c) Gray traces indicate mean responses in single flies. Black traces indicate mean ± s.e.m. across all tested flies (N = 9) expressing P2X2 in HS cells. Orange traces indicate mean ± s.e.m. across all tested flies (N = 7) that did not express P2X2, but that did receive pulses of ATP. Bottom, ATP-induced activity in HS cells controls turning independent on the ongoing direction of walking. Red, flies that were turning to the ipsilateral side of the stimulation before ATP application. Blue, flies that were turning contralateral to the side of activation before ATP application. (d) Proposed functions of HS cells during walking. Left, locomotor control. An intended straight course (1) can be modulated by asymmetries in the walking apparatus (2), which could then be compensated for by a calibrated signal through the congruent interaction of visual and locomotor signals (3). Middle, similarly, after the initiation of a voluntary turn, where the head precedes the body turn, compensatory calibrated head realignment can occur to keep the gaze stable. NMN: neck motor neuron. Right, HS cells represent information about self-movement by encoding the translational velocity (odometer) and angular velocity of the fly (compass-related information). From the integration of the activity dynamics of the bilateral network, downstream circuits could obtain information about the executed fly's walking movements.

  8. Visual responses of the recorded LPTCs
    Supplementary Fig. 1: Visual responses of the recorded LPTCs

    (a) The mean±SD (n=3 trials) of the direction-selective (DS) responses (i.e., subtracting the null direction, ND, from the preferred direction, PD) of the HSE cell shown in Fig. 1c, to a wide-field horizontal (left) or vertical (right) moving grating. The light gray background indicates the stimulus period. (b) The (time) average DS response amplitude during stimulation (grey dots), and the mean value (black lines). (c) Triple immunostaining for the GFP signal (green) driven by the R81G07Gal4 line, for the biocytin signal (red) from the recorded HS cell, and for the bruchpilot protein (brp, blue), staining for synaptic neuropile. The dash line delineates the border of the lobula plate. (d–f) Same as (a–c), but for the VS1 cell shown in Fig. 1d. (g–i) Same as (a–c), but for the HSE cell in a norpA mutant background (blind fly) shown in Fig. 1e. (j) DS visual responses from HS cells in T4/T5-Gal4 (left) and UAS-Kir (middle) control flies, or from HS cells in flies with T4/T5 cells silenced by the expression of Kir. Shown is the mean±SEM (N is indicated in each corresponding figure).

  9. Classification of tethered behavior on the ball
    Supplementary Fig. 2: Classification of tethered behavior on the ball

    (a) Swing-stance periods (black or white, respectively) during a walking bout as defined by videography. Note that at steady forward walking (from 0.5 to 4 s), the fly displays tripod gate (see zoom-in, N=6 flies). (b) Examples of a subset of the first ten principal components (PCs) of the side-view fly video recordings (Fig. 1a,b). Asterisks indicate the PCs whose scores are shown in (c). (c) Time-varying scores for different PCs. A zoom-in into PC5’s scores reveals a periodic behavior of a principal component associated with leg movement. (d) A wavelet transformation of PC5 scores further indicates the bandwidth of the signal, matching the mean cycle period obtained from the swing-phase analysis. (e) The PCs scores and their wavelet transform (c,d) were used to train JAABA (top schematic) to classify walking and other possible behaviors of the fly on the ball (bottom color-coded image). (f) Example of a 1-minute trial displaying the fly’s forward (grey top trace), and angular (grey bottom trace) velocity signals during walking (pink background) and non-walking (blue background) segments. For angular velocity, CCW turns are positive sign values. For forward velocity, negative sign indicates backward walking. Arrowheads show activity segments (i.e., non-walking but non-stationary segments).

  10. Walking-specific responses during locomotion in HS cells
    Supplementary Fig. 3: Walking-specific responses during locomotion in HS cells

    Cross-covariance analysis per cell between the fly’s angular velocity (or forward velocity) and the Vm of HS cells, in walking or activity bouts. The associated distributions indicate the % of activity time spent in walking (“walking ratio”). Note that the example cells included in Figure 3 are not presented here.

  11. Walking direction and speed sensitivity in HS cells
    Supplementary Fig. 4: Walking direction and speed sensitivity in HS cells

    (a) Change in Vm of left-side HS cells with respect to quiescence (ΔVm, color-coded) as a function of the forward and angular velocities of each individual fly. (b) Same as in (a), but for right-side HS cells. Scale bar is the same as for the maps in (a). (c) Population ΔVm (color-coded) as a function of the forward and angular acceleration of the fly. (d) Examples of the Vm of HS cells (orange), the forward (Vf, black), and angular velocities (Va, gray) of the fly in segments of walking bouts where the fly’s angular velocity did not exceed ±35°/s (gray shadow). (e) Left, weighted grand mean±SEM of the average angular velocity of the fly for the selected segments with low angular velocity as a function of the averaged Vf of the fly. The weight is given as a function of number of segments where the fly showed |Va| < 35°/s (mean = 240, range: 50-499 bouts). Right, weighted grand mean±SEM of the average Vm as a function of the average Vf of the fly. (f) Mean correlation coefficient between the fly’s Vf and the Vm of HS cells per cell, and between Vf and Va per cell. Significant correlations are red, non-significant correlations are blue (traces were shuffled 20 times by concatenated walking bouts, Wilcoxon’s signed-rank test).

  12. Movement-related modulations in VS cells
    Supplementary Fig. 5: Movement-related modulations in VS cells

    (a) Vm of a VS1 cell triggered at the onset of postural adjustments in walking (left), or activity (right) bouts. n = number of bouts. Lighter traces show triggered data from individual bouts. (b) Left, the mean Vm of the population of VS cells (mean±SEM) triggered at the onset of walking (black) or activity (gray) bouts. Right, the mean Vm triggered at the offset of walking (black) or activity (gray) bouts. n = total number of bouts. For walking onset bouts, N=10 cells. For activity onset analysis, N=6 cells. For walking offset analysis, N=7 cells. For activity offset analysis, N=4 cells. (c) Distribution of the 10% rise time of the Vm (see Methods) relative to the onset of walking bouts. Arrowhead indicates the median value. (d) ΔVm (color-coded) as a function of the forward and angular velocities of each individual fly. (e) ΔVm (color-coded) is plotted as a function of the forward and angular velocities of the fly for the population of recorded VS cells, right and left cells were pooled together (see Methods). Scale bar is the same as for the maps in (d).

  13. Walking-specific signals are independent of vision, are not coupled to the movement of antennae or halteres, and do not depend on the synaptic activity of leg mechanosensory cells.
    Supplementary Fig. 6: Walking-specific signals are independent of vision, are not coupled to the movement of antennae or halteres, and do not depend on the synaptic activity of leg mechanosensory cells.

    (a) Walking velocity tuning for the recorded HS cell shown in Fig. 1e. PDmotor and NDmotor are the cell’s preferred and null turning direction of the fly (see main text). (b) Walking velocity tuning map for the population of HS cells recorded from blind (norpA) flies (N=9 cells). (c) Walking velocity tuning maps for the population of HS cells recorded from flies with antennae and halteres waxed (N=11 cells, right and left-side HS cells were pooled, see Methods). (d) Left, walking velocity tuning maps for the population of HS cells recorded from flies with leg mechanosensory cells expressing tetanus toxin to disrupt their synaptic activity (N=8 cells). Right, walking velocity maps of parental control flies with functional synaptic activity in leg mechanosensory cells (N=7 cells). (e) Top, paths of freely walking experimental (red, flies with leg mechanosensory cells expressing tetanus toxin, TNT) or control flies (black). Bottom, zoomed-in view to highlight the straightness of the paths of the walking flies. (f) Probability distributions (mean±SD) of the straightness of walking paths in experimental (red) and control (black) flies (see Methods). (g) Mean path straightness for experimental (red) and control (black) flies. For straightness analysis: 13 control flies, mean±SD: 0.87±0.03; 15 experimental flies, mean±SD: 0.71±0.07, ***,P< 0.0001, Z=4.38, Wilcoxon’s rank-sum test. (h) Probability distributions (mean±SD) of walking speed in experimental (red) and control (black) flies. For walking velocity analysis: 13 control flies, mean±SD: 11.5±1.9 mm/s; 15 experimental flies, mean±SD: 9.6±1.5 mm/s, P<0.01, Z=2.63, Wilcoxon’s rank-sum test.

  14. Correlation between head and body movements during tethered walking in darkness
    Supplementary Fig. 7: Correlation between head and body movements during tethered walking in darkness

    (a) Top: Schematic of head-tracking set-up. Middle: example frame from a top-view camera. Bottom: example frame from a side-view camera. Contrast has been enhanced for clarity. (b) Example traces from one trial comparing the head yaw movement and the body’s turn. Top, angular velocity of the fly. Middle, head yaw angle. Bottom, head yaw velocity. Arrows indicate head-yaw angle offsets. (c) Cross-covariance between angular velocity and head angle for all walking bouts in darkness for a single fly (n=390 bouts from 126 trials). Thick black trace, mean; shadow, SD. The magenta line in all cross-covariance plots indicates the bootstrapped 95% confidence level of the analysis (see Methods). Light gray trace, example from (b). (d) Grand mean cross-covariance. The head precedes the body by 50 ms on average (mean±SEM =48.3±24.0 ms, N=9 flies). (e) Instantaneous head yaw position as a function of the angular and the forward velocities of the fly. (f) Idem as in (c) but the head yaw velocity was compared with the angular velocity of the fly. (g) Same as in (d) but with the head yaw velocity. (h) Instantaneous head yaw velocity as a function of the forward and angular velocities of the fly. (i) Cross-covariance analysis between the head pitch angle and the forward velocity of the fly. Grand mean±SEM (341±143ms, N=6 flies).

  15. Model performance analysis
    Supplementary Fig. 8: Model performance analysis

    (a) Predicted velocity tuning maps for different single-component models. (b) Mean cross-correlation coefficients per cell between observed and predicted Vm for different models (black, matched pairs; red, mismatched pairs; see Methods, lines: mean values, **: P<0.001, Z>3.78, N=19 cells, Wilcoxon signed-rank test). Right-most column: cross-correlation coefficients between the observed and the predicted yaw head angle. (c) Performance of the BS and the Va+Vf+BS models in each cell. Indicated are the example cells shown in Fig. 3a, and in (e). (d) Performance of the Va+Vf+BS and the Va+BS models for each cell. Mean (±SD) correlation coefficients between the predicted and observed HS dynamics (see Methods). Red and blue: the difference in the magnitude of correlation coefficients between the two models cannot be explained by the reshuffling procedure (traces were shuffled 20 times by concatenated walking bouts, Wilcoxon’s signed-rank test, P<0.005), whereas the gray pairs can (P >0.02, Wilcoxon‘s signed-rank test). (e) Example of a cell with the lowest three-compartment model performance. Note that the HS cells’ dynamics are still well described by the three-component model. (f) Top, the predicted velocity-tuning map for the example cell in Figure 5c scaled in mm/s units in both axes. θ: the angle between the forward velocity axis and the membrane potential change (ΔVm) gradient (see Methods). Bottom, distribution of θ for each cell; the mean value is indicated in black. The θ value for the observed population velocity map (N=19 cells) is indicated in red. (g) The observed (black) and predicted (magenta) head yaw angles, estimated from the three-component model fitting the head yaw angle instead of the Vm. (h) Power spectrum analysis of the observed, the predicted, and the difference between the two for Vm (left), or yaw head angle (right). Note that the largest difference for the head angle prediction is on the DC component, i.e., the offset of the head position (arrows in (g)).

  16. Decoding walking velocities from the bilateral activity of HS cells
    Supplementary Fig. 9: Decoding walking velocities from the bilateral activity of HS cells

    (a) Recordings from HS cells on one side of the brain were combined with predicted HS dynamics of the other right side using the walking behavior of the fly as input for the BS+Va+Vf model (Figure 5). The dynamics of the bilateral HS cells are labeled as modeled + recorded. (b) The mean angular (left) or forward (middle) velocities of the fly and normalized data points (right) were plotted as a function of the negative (VmLeft-VmRight) or positive (VmLeft+VmRight) combination of the bilateral dynamics of modeled + recorded cells. (c) Using the bilateral modeled + recorded dynamics, or the unilateral recorded cell dynamics, we applied a linear decoder (see Methods) to predict the angular velocity (Va) of the fly. Left, top two plots, example of decoding using bilateral HS cells’ dynamics. Bottom two plots, examples of decoding using the recorded cell dynamics only, i.e., the unilateral HS cells’ activity. Red trace, estimated; black trace observed. Right, covariance coefficients (Cov Coef) between the predicted and observed velocities using bilateral or unilateral HS cells’ dynamics. Walking bouts and the cell dynamics were reshuffled 20 times to obtain a population of Cov Coefs per fly and the mean value was used. Red, Cov Coefs between matched behavior and Vm (bilateral, mean±SD=0.85±0.06; unilateral, mean±SD=0.48±0.14,); gray, Cov Coefs between mismatched behavior and Vm. The difference in the magnitude of Cov Coefs between bilateral and unilateral decoding is significant (P<0.0005, Z=3.82, Wilcoxon signed-rank test, N=19 cells). (d) Decoding of the forward velocity (Vf) of the fly. Same as in (c) Cov Coefs between matched behavior and Vm for bilateral model: mean±SD=0.50±0.13; for the unilateral model: mean±SD=0.29±0.12. The difference in the magnitude of Cov Coefs between bilateral and unilateral decoding is significant (Wilcoxon’s signed-rank test P< 0.005, N=19 cells). (e) Transfer functions for the decoding of angular velocity (Va). Gray, individual fly’s filters, black, mean filter. Note that the magnitude of the mean filters obtained with the recorded data is similar to those obtained with the modeled data. (f) Same for the decoding of forward velocity (Vf).

  17. Visuomotor interactions in HS cells
    Supplementary Fig. 10: Visuomotor interactions in HS cells

    (a) Top: Velocity-tuning maps under replay visual stimulation per cell. Bottom, visual and angular velocity maps per cell. (b) Walking velocity tuning map across the population of right-side HS cells under replay trials (N=13). (c) Estimate of the visual modulation of HS cells’ activity under replay conditions in quiescent segments: the visual stimulus is scaled by the velocity tuning curve of HS cells (see Methods) to obtain an effective visual stimulus (red trace). This effective stimulus is convolved with a response kernel (see Methods) to estimate the visual-induced activity in HS cells (blue trace). For comparison, the observed visual responses under identical conditions is shown (black trace). The delay of the kernel was obtained by cross-correlation analysis between HS cell responses and visual stimuli in quiescent segments. (d) Distribution of θ (Fig. 6) for each cell for fits with R2>0.7 (top, 9/13 cells, mean±SD=37±9°). Black: mean value; red, θ for the population map shown in Figure 6b. (e) Predictions from ideal random forest decoders (see Methods) of the sum of the visual velocity (Vv) and the fly‘s angular velocity (Va, red traces), or the difference between the two (blue traces). The input signals for the decoder were either the estimated Vm dynamics from the visual stimulus and the three-component walking of the fly (top row), or the estimated Vm dynamics from Vv and the fly’s Va (bottom). c=correlation coefficient between the predicted and the observed values of the sum or difference between Vv and Va. (f) Cross-correlation coefficients between the predicted and the observed velocities for each recorded cell, for the two different inputs to the ideal decoders. Bars: the mean values across cells. Color code as in (e).

Videos

  1. Simultaneous recordings of physiology and behavior during an isolated walking bout.
    Video 1: Simultaneous recordings of physiology and behavior during an isolated walking bout.
    Left, side view of a walking fly in darkness. Right top: the angular velocity (“rotational speed”, blue), and forward velocity (“forward speed”, red) of the walking fly. Right bottom, the change of the membrane potential of an HS cell relative to quiescence during ongoing walking.
  2. Detection of postural adjustments before the onset of tethered behavior.
    Video 2: Detection of postural adjustments before the onset of tethered behavior.
    Left, original side view of a fly. Right, thresholded pixel-change side view of the fly. Movie is slow-down 5X. Red square indicates the onset of postural adjustments associated with the start of walking (see Methods).
  3. Head movement during tethered walking in darkness.
    Video 3: Head movement during tethered walking in darkness.
    Top left: top view of the head to detect head yaw movements. Bottom left: the head yaw angle (magenta) and angular velocity (red) of the walking fly. Top right: side view of the fly to detect head pitch movements. Bottom right: the head pitch position (green) and forward velocity (blue) of the walking fly.

References

  1. Franklin, D.W. & Wolpert, D.M. Computational mechanisms of sensorimotor control. Neuron 72, 425442 (2011).
  2. Whitlock, J.R., Sutherland, R.J., Witter, M.P., Moser, M.-B. & Moser, E.I. Navigating from hippocampus to parietal cortex. Proc. Natl. Acad. Sci. USA 105, 1475514762 (2008).
  3. Lappe, M., Bremmer, F. & van den Berg, A.V. Perception of self-motion from visual flow. Trends Cogn. Sci. 3, 329336 (1999).
  4. Warren, W.H. Jr., Kay, B.A., Zosh, W.D., Duchon, A.P. & Sahuc, S. Optic flow is used to control human walking. Nat. Neurosci. 4, 213216 (2001).
  5. Pfeffer, S.E. & Wittlinger, M. Optic flow odometry operates independently of stride integration in carried ants. Science 353, 11551157 (2016).
  6. Morgan, B. & Frost, B.J. Visual response characteristics of neurons in nucleus of basal optic root of pigeons. Exp. Brain Res. 42, 181188 (1981).
  7. Britto, L.R., Natal, C.L. & Marcondes, A.M. The accessory optic system in pigeons: receptive field properties of identified neurons. Brain Res. 206, 149154 (1981).
  8. Grasse, K.L. & Cynader, M.S. Electrophysiology of medial terminal nucleus of accessory optic system in the cat. J. Neurophysiol. 48, 490504 (1982).
  9. Hausen, K. Motion sensitive interneurons in the optomotor system of the fly. Biol. Cybern. 45, 143156 (1982).
  10. Kano, M.-S., Kano, M. & Maekawa, K. Receptive field organization of climbing fiber afferents responding to optokinetic stimulation in the cerebellar nodulus and flocculus of the pigmented rabbit. Exp. Brain Res. 82, 499512 (1990).
  11. Duffy, C.J. & Wurtz, R.H. Sensitivity of MST neurons to optic flow stimuli. I. A continuum of response selectivity to large-field stimuli. J. Neurophysiol. 65, 13291345 (1991).
  12. Kubo, F. et al. Functional architecture of an optic flow-responsive area that drives horizontal eye movements in zebrafish. Neuron 81, 13441359 (2014).
  13. Schnell, B. et al. Processing of horizontal optic flow in three visual interneurons of the Drosophila brain. J. Neurophysiol. 103, 16461657 (2010).
  14. Koenderink, J.J. Optic flow. Vision Res. 26, 161179 (1986).
  15. Bradley, D.C., Maxwell, M., Andersen, R.A., Banks, M.S. & Shenoy, K.V. Mechanisms of heading perception in primate visual cortex. Science 273, 15441547 (1996).
  16. Gu, Y., Angelaki, D.E. & Deangelis, G.C. Neural correlates of multisensory cue integration in macaque MSTd. Nat. Neurosci. 11, 12011210 (2008).
  17. Britten, K.H. & van Wezel, R.J. Electrical microstimulation of cortical area MST biases heading perception in monkeys. Nat. Neurosci. 1, 5963 (1998).
  18. Saleem, A.B., Ayaz, A., Jeffery, K.J., Harris, K.D. & Carandini, M. Integration of visual motion and locomotion in mouse visual cortex. Nat. Neurosci. 16, 18641869 (2013).
  19. Keller, G.B., Bonhoeffer, T. & Hübener, M. Sensorimotor mismatch signals in primary visual cortex of the behaving mouse. Neuron 74, 809815 (2012).
  20. Roth, M.M. et al. Thalamic nuclei convey diverse contextual information to layer 1 of visual cortex. Nat. Neurosci. 19, 299307 (2016).
  21. Borst, A., Haag, J. & Reiff, D.F. Fly motion vision. Annu. Rev. Neurosci. 33, 4970 (2010).
  22. Scott, E.K., Raabe, T. & Luo, L. Structure of the vertical and horizontal system neurons of the lobula plate in Drosophila. J. Comp. Neurol. 454, 470481 (2002).
  23. Haikala, V., Joesch, M., Borst, A. & Mauss, A.S. Optogenetic control of fly optomotor responses. J. Neurosci. 33, 1392713934 (2013).
  24. Kim, A.J., Fitzgerald, J.K. & Maimon, G. Cellular evidence for efference copy in Drosophila visuomotor processing. Nat. Neurosci. 18, 12471255 (2015).
  25. Chiappe, M.E., Seelig, J.D., Reiser, M.B. & Jayaraman, V. Walking modulates speed sensitivity in Drosophila motion vision. Curr. Biol. 20, 14701475 (2010).
  26. Greenspan, R.J. & Ferveur, J.F. Courtship in Drosophila. Annu. Rev. Genet. 34, 205232 (2000).
  27. Bloomquist, B.T. et al. Isolation of a putative phospholipase C gene of Drosophila, norpA, and its role in phototransduction. Cell 54, 723733 (1988).
  28. Kabra, M., Robie, A.A., Rivera-Alba, M., Branson, S. & Branson, K. JAABA: interactive machine learning for automatic annotation of animal behavior. Nat. Methods 10, 6467 (2013).
  29. Kamikouchi, A. et al. The neural basis of Drosophila gravity-sensing and hearing. Nature 458, 165171 (2009).
  30. Fraenkel, G. & Pringle, J.W.S. Halteres of flies as gyroscopic organs of equilibrium. Nature 141, 919920 (1938).
  31. Mendes, C.S., Bartos, I., Akay, T., Márka, S. & Mann, R.S. Quantification of gait parameters in freely walking wild type and sensory deprived Drosophila melanogaster. eLife 2, e00231 (2013).
  32. Tuthill, J.C. & Wilson, R.I. Parallel transformation of tactile signals in central circuits of Drosophila. Cell 164, 10461059 (2016).
  33. Preuss, T. & Hengstenberg, R. Structure and kinematics of the prosternal organs and their influence on head position in the blowfly, Calliphora erythrocephala. J. Comp. Physiol. A. 171, 483493 (1992).
  34. Schilstra, C. & Hateren, J.H. Blowfly flight and optic flow. I. Thorax kinematics and flight dynamics. J. Exp. Biol. 202, 14811490 (1999).
  35. Blaj, G. & van Hateren, J.H. Saccadic head and thorax movements in freely walking blowflies. J. Comp. Physiol. A Neuroethol. Sens. Neural Behav. Physiol. 190, 861868 (2004).
  36. Schnell, B., Raghu, S.V., Nern, A. & Borst, A. Columnar cells necessary for motion responses of wide-field visual interneurons in Drosophila. J. Comp. Physiol. A Neuroethol. Sens. Neural Behav. Physiol. 198, 389395 (2012).
  37. Haag, J. & Borst, A. Active membrane properties and signal encoding in graded potential neurons. J. Neurosci. 18, 79727986 (1998).
  38. Petreanu, L. et al. Activity in motor-sensory projections reveals distributed coding in somatosensation. Nature 489, 299303 (2012).
  39. Pabst, M. & Kral, K. Proprioceptive contribution to distance estimation by motion parallax in a praying mantid. J. Exp. Biol. 201, 14831491 (1998).
  40. Mischiati, M. et al. Internal models direct dragonfly interception steering. Nature 517, 333338 (2015).
  41. Kress, D. & Egelhaaf, M. Gaze characteristics of freely walking blowflies Calliphora vicina in a goal-directed task. J. Exp. Biol. 217, 32093220 (2014).
  42. Schnell, B., Weir, P.T., Roth, E., Fairhall, A.L. & Dickinson, M.H. Cellular mechanisms for integral feedback in visually guided behavior. Proc. Natl. Acad. Sci. USA 111, 57005705 (2014).
  43. Haag, J., Wertz, A. & Borst, A. Central gating of fly optomotor response. Proc. Natl. Acad. Sci. USA 107, 2010420109 (2010).
  44. Huston, S.J. & Krapp, H.G. Visuomotor transformation in the fly gaze stabilization system. PLoS Biol. 6, e173 (2008).
  45. Maimon, G., Straw, A.D. & Dickinson, M.H. Active flight increases the gain of visual motion processing in Drosophila. Nat. Neurosci. 13, 393399 (2010).
  46. Suver, M.P., Mamiya, A. & Dickinson, M.H. Octopamine neurons mediate flight-induced modulation of visual processing in Drosophila. Curr. Biol. 22, 22942302 (2012).
  47. Muijres, F.T., Elzinga, M.J., Iwasaki, N.A. & Dickinson, M.H. Body saccades of Drosophila consist of stereotyped banked turns. J. Exp. Biol. 218, 864875 (2015).
  48. Schuster, S., Strauss, R. & Götz, K.G. Virtual-reality techniques resolve the visual cues used by fruit flies to evaluate object distances. Curr. Biol. 12, 15911594 (2002).
  49. Pelah, A. & Barlow, H.B. Visual illusion from running. Nature 381, 282283 (1996).
  50. von Reyn, C.R. et al. A spike-timing mechanism for action selection. Nat. Neurosci. 17, 962970 (2014).
  51. Hughes, C.L. & Thomas, J.B. A sensory feedback circuit coordinates muscle activity in Drosophila. Mol. Cell. Neurosci. 35, 383396 (2007).
  52. Giorgianni, M.W. & Mann, R.S. Establishment of medial fates along the proximodistal axis of the Drosophila leg through direct activation of dachshund by Distalless. Dev. Cell 20, 455468 (2011).
  53. Stockinger, P., Kvitsiani, D., Rotkopf, S., Tirián, L. & Dickson, B.J. Neural circuitry that governs Drosophila male courtship behavior. Cell 121, 795807 (2005).
  54. Jenett, A. et al. A GAL4-driver line resource for Drosophila neurobiology. Cell Rep. 2, 9911001 (2012).
  55. Seelig, J.D. et al. Two-photon calcium imaging from head-fixed Drosophila during optomotor walking behavior. Nat. Methods 7, 535540 (2010).
  56. Wilson, R.I., Turner, G.C. & Laurent, G. Transformation of olfactory representations in the Drosophila antennal lobe. Science 303, 366370 (2004).
  57. Suter, B.A. et al. Ephus: multipurpose data acquisition software for neuroscience experiments. Front. Neural Circuits 4, 100 (2010).
  58. Reiser, M.B. & Dickinson, M.H. A modular display system for insect behavioral neuroscience. J. Neurosci. Methods 167, 127139 (2008).
  59. Martin, J.R. A portrait of locomotor behavior in Drosophila determined by a video-tracking paradigm. Behav. Process. 67, 207219 (2004).
  60. Wosnitza, A., Bockemühl, T., Dübbert, M., Scholz, H. & Büschges, A. Inter-leg coordination in the control of walking speed in Drosophila. J. Exp. Biol. 216, 480491 (2013).
  61. Kabsch, W. A solution for the best rotation to relate two sets of vectors. Acta Crystallogr. 32, 922923 (1976).
  62. Breiman, L. Random forests. Machine Learning 45, 532 (2001).

Download references

Author information

Affiliations

  1. Champalimaud Neuroscience Programme, Champalimaud Center for the Unknown, Lisbon, Portugal.

    • Terufumi Fujiwara,
    • Tomás L Cruz,
    • James P Bohnslav &
    • M Eugenia Chiappe
  2. Program in Neuroscience, Department of Neurobiology, Harvard University, Boston, Massachusetts, USA.

    • James P Bohnslav

Contributions

T.F. and M.E.C. designed the study. T.F. performed the electrophysiological experiments. T.L.C. designed the analysis of behavior and constructed the model. J.P.B. performed the tethered walking behaviors. All of the authors analyzed the data, and T.F., T.L.C. and M.E.C. wrote the manuscript.

Competing financial interests

The authors declare no competing financial interests.

Corresponding author

Correspondence to:

Author details

Supplementary information

Supplementary Figures

  1. Supplementary Figure 1: Visual responses of the recorded LPTCs (451 KB)

    (a) The mean±SD (n=3 trials) of the direction-selective (DS) responses (i.e., subtracting the null direction, ND, from the preferred direction, PD) of the HSE cell shown in Fig. 1c, to a wide-field horizontal (left) or vertical (right) moving grating. The light gray background indicates the stimulus period. (b) The (time) average DS response amplitude during stimulation (grey dots), and the mean value (black lines). (c) Triple immunostaining for the GFP signal (green) driven by the R81G07Gal4 line, for the biocytin signal (red) from the recorded HS cell, and for the bruchpilot protein (brp, blue), staining for synaptic neuropile. The dash line delineates the border of the lobula plate. (d–f) Same as (a–c), but for the VS1 cell shown in Fig. 1d. (g–i) Same as (a–c), but for the HSE cell in a norpA mutant background (blind fly) shown in Fig. 1e. (j) DS visual responses from HS cells in T4/T5-Gal4 (left) and UAS-Kir (middle) control flies, or from HS cells in flies with T4/T5 cells silenced by the expression of Kir. Shown is the mean±SEM (N is indicated in each corresponding figure).

  2. Supplementary Figure 2: Classification of tethered behavior on the ball (321 KB)

    (a) Swing-stance periods (black or white, respectively) during a walking bout as defined by videography. Note that at steady forward walking (from 0.5 to 4 s), the fly displays tripod gate (see zoom-in, N=6 flies). (b) Examples of a subset of the first ten principal components (PCs) of the side-view fly video recordings (Fig. 1a,b). Asterisks indicate the PCs whose scores are shown in (c). (c) Time-varying scores for different PCs. A zoom-in into PC5’s scores reveals a periodic behavior of a principal component associated with leg movement. (d) A wavelet transformation of PC5 scores further indicates the bandwidth of the signal, matching the mean cycle period obtained from the swing-phase analysis. (e) The PCs scores and their wavelet transform (c,d) were used to train JAABA (top schematic) to classify walking and other possible behaviors of the fly on the ball (bottom color-coded image). (f) Example of a 1-minute trial displaying the fly’s forward (grey top trace), and angular (grey bottom trace) velocity signals during walking (pink background) and non-walking (blue background) segments. For angular velocity, CCW turns are positive sign values. For forward velocity, negative sign indicates backward walking. Arrowheads show activity segments (i.e., non-walking but non-stationary segments).

  3. Supplementary Figure 3: Walking-specific responses during locomotion in HS cells (471 KB)

    Cross-covariance analysis per cell between the fly’s angular velocity (or forward velocity) and the Vm of HS cells, in walking or activity bouts. The associated distributions indicate the % of activity time spent in walking (“walking ratio”). Note that the example cells included in Figure 3 are not presented here.

  4. Supplementary Figure 4: Walking direction and speed sensitivity in HS cells (484 KB)

    (a) Change in Vm of left-side HS cells with respect to quiescence (ΔVm, color-coded) as a function of the forward and angular velocities of each individual fly. (b) Same as in (a), but for right-side HS cells. Scale bar is the same as for the maps in (a). (c) Population ΔVm (color-coded) as a function of the forward and angular acceleration of the fly. (d) Examples of the Vm of HS cells (orange), the forward (Vf, black), and angular velocities (Va, gray) of the fly in segments of walking bouts where the fly’s angular velocity did not exceed ±35°/s (gray shadow). (e) Left, weighted grand mean±SEM of the average angular velocity of the fly for the selected segments with low angular velocity as a function of the averaged Vf of the fly. The weight is given as a function of number of segments where the fly showed |Va| < 35°/s (mean = 240, range: 50-499 bouts). Right, weighted grand mean±SEM of the average Vm as a function of the average Vf of the fly. (f) Mean correlation coefficient between the fly’s Vf and the Vm of HS cells per cell, and between Vf and Va per cell. Significant correlations are red, non-significant correlations are blue (traces were shuffled 20 times by concatenated walking bouts, Wilcoxon’s signed-rank test).

  5. Supplementary Figure 5: Movement-related modulations in VS cells (299 KB)

    (a) Vm of a VS1 cell triggered at the onset of postural adjustments in walking (left), or activity (right) bouts. n = number of bouts. Lighter traces show triggered data from individual bouts. (b) Left, the mean Vm of the population of VS cells (mean±SEM) triggered at the onset of walking (black) or activity (gray) bouts. Right, the mean Vm triggered at the offset of walking (black) or activity (gray) bouts. n = total number of bouts. For walking onset bouts, N=10 cells. For activity onset analysis, N=6 cells. For walking offset analysis, N=7 cells. For activity offset analysis, N=4 cells. (c) Distribution of the 10% rise time of the Vm (see Methods) relative to the onset of walking bouts. Arrowhead indicates the median value. (d) ΔVm (color-coded) as a function of the forward and angular velocities of each individual fly. (e) ΔVm (color-coded) is plotted as a function of the forward and angular velocities of the fly for the population of recorded VS cells, right and left cells were pooled together (see Methods). Scale bar is the same as for the maps in (d).

  6. Supplementary Figure 6: Walking-specific signals are independent of vision, are not coupled to the movement of antennae or halteres, and do not depend on the synaptic activity of leg mechanosensory cells. (339 KB)

    (a) Walking velocity tuning for the recorded HS cell shown in Fig. 1e. PDmotor and NDmotor are the cell’s preferred and null turning direction of the fly (see main text). (b) Walking velocity tuning map for the population of HS cells recorded from blind (norpA) flies (N=9 cells). (c) Walking velocity tuning maps for the population of HS cells recorded from flies with antennae and halteres waxed (N=11 cells, right and left-side HS cells were pooled, see Methods). (d) Left, walking velocity tuning maps for the population of HS cells recorded from flies with leg mechanosensory cells expressing tetanus toxin to disrupt their synaptic activity (N=8 cells). Right, walking velocity maps of parental control flies with functional synaptic activity in leg mechanosensory cells (N=7 cells). (e) Top, paths of freely walking experimental (red, flies with leg mechanosensory cells expressing tetanus toxin, TNT) or control flies (black). Bottom, zoomed-in view to highlight the straightness of the paths of the walking flies. (f) Probability distributions (mean±SD) of the straightness of walking paths in experimental (red) and control (black) flies (see Methods). (g) Mean path straightness for experimental (red) and control (black) flies. For straightness analysis: 13 control flies, mean±SD: 0.87±0.03; 15 experimental flies, mean±SD: 0.71±0.07, ***,P< 0.0001, Z=4.38, Wilcoxon’s rank-sum test. (h) Probability distributions (mean±SD) of walking speed in experimental (red) and control (black) flies. For walking velocity analysis: 13 control flies, mean±SD: 11.5±1.9 mm/s; 15 experimental flies, mean±SD: 9.6±1.5 mm/s, P<0.01, Z=2.63, Wilcoxon’s rank-sum test.

  7. Supplementary Figure 7: Correlation between head and body movements during tethered walking in darkness (324 KB)

    (a) Top: Schematic of head-tracking set-up. Middle: example frame from a top-view camera. Bottom: example frame from a side-view camera. Contrast has been enhanced for clarity. (b) Example traces from one trial comparing the head yaw movement and the body’s turn. Top, angular velocity of the fly. Middle, head yaw angle. Bottom, head yaw velocity. Arrows indicate head-yaw angle offsets. (c) Cross-covariance between angular velocity and head angle for all walking bouts in darkness for a single fly (n=390 bouts from 126 trials). Thick black trace, mean; shadow, SD. The magenta line in all cross-covariance plots indicates the bootstrapped 95% confidence level of the analysis (see Methods). Light gray trace, example from (b). (d) Grand mean cross-covariance. The head precedes the body by 50 ms on average (mean±SEM =48.3±24.0 ms, N=9 flies). (e) Instantaneous head yaw position as a function of the angular and the forward velocities of the fly. (f) Idem as in (c) but the head yaw velocity was compared with the angular velocity of the fly. (g) Same as in (d) but with the head yaw velocity. (h) Instantaneous head yaw velocity as a function of the forward and angular velocities of the fly. (i) Cross-covariance analysis between the head pitch angle and the forward velocity of the fly. Grand mean±SEM (341±143ms, N=6 flies).

  8. Supplementary Figure 8: Model performance analysis (410 KB)

    (a) Predicted velocity tuning maps for different single-component models. (b) Mean cross-correlation coefficients per cell between observed and predicted Vm for different models (black, matched pairs; red, mismatched pairs; see Methods, lines: mean values, **: P<0.001, Z>3.78, N=19 cells, Wilcoxon signed-rank test). Right-most column: cross-correlation coefficients between the observed and the predicted yaw head angle. (c) Performance of the BS and the Va+Vf+BS models in each cell. Indicated are the example cells shown in Fig. 3a, and in (e). (d) Performance of the Va+Vf+BS and the Va+BS models for each cell. Mean (±SD) correlation coefficients between the predicted and observed HS dynamics (see Methods). Red and blue: the difference in the magnitude of correlation coefficients between the two models cannot be explained by the reshuffling procedure (traces were shuffled 20 times by concatenated walking bouts, Wilcoxon’s signed-rank test, P<0.005), whereas the gray pairs can (P >0.02, Wilcoxon‘s signed-rank test). (e) Example of a cell with the lowest three-compartment model performance. Note that the HS cells’ dynamics are still well described by the three-component model. (f) Top, the predicted velocity-tuning map for the example cell in Figure 5c scaled in mm/s units in both axes. θ: the angle between the forward velocity axis and the membrane potential change (ΔVm) gradient (see Methods). Bottom, distribution of θ for each cell; the mean value is indicated in black. The θ value for the observed population velocity map (N=19 cells) is indicated in red. (g) The observed (black) and predicted (magenta) head yaw angles, estimated from the three-component model fitting the head yaw angle instead of the Vm. (h) Power spectrum analysis of the observed, the predicted, and the difference between the two for Vm (left), or yaw head angle (right). Note that the largest difference for the head angle prediction is on the DC component, i.e., the offset of the head position (arrows in (g)).

  9. Supplementary Figure 9: Decoding walking velocities from the bilateral activity of HS cells (378 KB)

    (a) Recordings from HS cells on one side of the brain were combined with predicted HS dynamics of the other right side using the walking behavior of the fly as input for the BS+Va+Vf model (Figure 5). The dynamics of the bilateral HS cells are labeled as modeled + recorded. (b) The mean angular (left) or forward (middle) velocities of the fly and normalized data points (right) were plotted as a function of the negative (VmLeft-VmRight) or positive (VmLeft+VmRight) combination of the bilateral dynamics of modeled + recorded cells. (c) Using the bilateral modeled + recorded dynamics, or the unilateral recorded cell dynamics, we applied a linear decoder (see Methods) to predict the angular velocity (Va) of the fly. Left, top two plots, example of decoding using bilateral HS cells’ dynamics. Bottom two plots, examples of decoding using the recorded cell dynamics only, i.e., the unilateral HS cells’ activity. Red trace, estimated; black trace observed. Right, covariance coefficients (Cov Coef) between the predicted and observed velocities using bilateral or unilateral HS cells’ dynamics. Walking bouts and the cell dynamics were reshuffled 20 times to obtain a population of Cov Coefs per fly and the mean value was used. Red, Cov Coefs between matched behavior and Vm (bilateral, mean±SD=0.85±0.06; unilateral, mean±SD=0.48±0.14,); gray, Cov Coefs between mismatched behavior and Vm. The difference in the magnitude of Cov Coefs between bilateral and unilateral decoding is significant (P<0.0005, Z=3.82, Wilcoxon signed-rank test, N=19 cells). (d) Decoding of the forward velocity (Vf) of the fly. Same as in (c) Cov Coefs between matched behavior and Vm for bilateral model: mean±SD=0.50±0.13; for the unilateral model: mean±SD=0.29±0.12. The difference in the magnitude of Cov Coefs between bilateral and unilateral decoding is significant (Wilcoxon’s signed-rank test P< 0.005, N=19 cells). (e) Transfer functions for the decoding of angular velocity (Va). Gray, individual fly’s filters, black, mean filter. Note that the magnitude of the mean filters obtained with the recorded data is similar to those obtained with the modeled data. (f) Same for the decoding of forward velocity (Vf).

  10. Supplementary Figure 10: Visuomotor interactions in HS cells (560 KB)

    (a) Top: Velocity-tuning maps under replay visual stimulation per cell. Bottom, visual and angular velocity maps per cell. (b) Walking velocity tuning map across the population of right-side HS cells under replay trials (N=13). (c) Estimate of the visual modulation of HS cells’ activity under replay conditions in quiescent segments: the visual stimulus is scaled by the velocity tuning curve of HS cells (see Methods) to obtain an effective visual stimulus (red trace). This effective stimulus is convolved with a response kernel (see Methods) to estimate the visual-induced activity in HS cells (blue trace). For comparison, the observed visual responses under identical conditions is shown (black trace). The delay of the kernel was obtained by cross-correlation analysis between HS cell responses and visual stimuli in quiescent segments. (d) Distribution of θ (Fig. 6) for each cell for fits with R2>0.7 (top, 9/13 cells, mean±SD=37±9°). Black: mean value; red, θ for the population map shown in Figure 6b. (e) Predictions from ideal random forest decoders (see Methods) of the sum of the visual velocity (Vv) and the fly‘s angular velocity (Va, red traces), or the difference between the two (blue traces). The input signals for the decoder were either the estimated Vm dynamics from the visual stimulus and the three-component walking of the fly (top row), or the estimated Vm dynamics from Vv and the fly’s Va (bottom). c=correlation coefficient between the predicted and the observed values of the sum or difference between Vv and Va. (f) Cross-correlation coefficients between the predicted and the observed velocities for each recorded cell, for the two different inputs to the ideal decoders. Bars: the mean values across cells. Color code as in (e).

Video

  1. Video 1: Simultaneous recordings of physiology and behavior during an isolated walking bout. (2.1 MB, Download)
    Left, side view of a walking fly in darkness. Right top: the angular velocity (“rotational speed”, blue), and forward velocity (“forward speed”, red) of the walking fly. Right bottom, the change of the membrane potential of an HS cell relative to quiescence during ongoing walking.
  2. Video 2: Detection of postural adjustments before the onset of tethered behavior. (394 KB, Download)
    Left, original side view of a fly. Right, thresholded pixel-change side view of the fly. Movie is slow-down 5X. Red square indicates the onset of postural adjustments associated with the start of walking (see Methods).
  3. Video 3: Head movement during tethered walking in darkness. (3.6 MB, Download)
    Top left: top view of the head to detect head yaw movements. Bottom left: the head yaw angle (magenta) and angular velocity (red) of the walking fly. Top right: side view of the fly to detect head pitch movements. Bottom right: the head pitch position (green) and forward velocity (blue) of the walking fly.

PDF files

  1. Supplementary Text and Figures (2,430 KB)

    Supplementary Figures 1–10

  2. Supplementary Methods Checklist (443 KB)

Additional data