Neuronal Control of Drosophila Walking Direction

See allHide authors and affiliations

Science  04 Apr 2014:
Vol. 344, Issue 6179, pp. 97-101
DOI: 10.1126/science.1249964

Backward or Forward

Although land animals generally walk forward, they readily switch to walking backward if they sense an obstruction or danger in the path ahead. Such a switch is likely to involve a neural signal sent from the brain down to local motor circuits, instructing these motor circuits to alter the phase at which specific leg muscles are activated. Bidaye et al. (p. 97; see the Perspective by Mann) identified such a neuron in Drosophila, which they call MDN (moonwalker descending neuron). Blocking synaptic transmission from MDN inhibited backward walking, and conversely artificially activating MDN caused flies to walk backward.


Most land animals normally walk forward but switch to backward walking upon sensing an obstacle or danger in the path ahead. A change in walking direction is likely to be triggered by descending “command” neurons from the brain that act upon local motor circuits to alter the timing of leg muscle activation. Here we identify descending neurons for backward walking in Drosophila—the MDN neurons. MDN activity is required for flies to walk backward when they encounter an impassable barrier and is sufficient to trigger backward walking under conditions in which flies would otherwise walk forward. We also identify ascending neurons, MAN, that promote persistent backward walking, possibly by inhibiting forward walking. These findings provide an initial glimpse into the circuits and logic that control walking direction in Drosophila.

Walking relies on the intrinsic rhythmic activity of local motor circuits within the central nervous system, called central pattern generators (CPGs). Such locomotor circuits have been documented in a number of species, including cats (1), rodents (2), crayfish (3), stick insects (4), and cockroaches (5). Proprioceptive feedback from leg mechanosensors ensures the accurate timing of each joint movement, but walking toward or away from specific targets requires descending signals from the brain. These descending inputs might act on CPGs to adjust the order, timing, or amplitude of individual leg movements (6, 7). The nature and identity of these descending commands are poorly understood.

We designed a thermogenetic screen in Drosophila to systematically identify any neurons that, upon activation, would change the fly’s walking direction. This screen used a UAS-trpA1 transgene to target expression of the thermosensitive cation channel TrpA1 (8) to arbitrary but stereotyped sets of neurons, as defined by the Vienna Tiles (VT) collection of GAL4 driver lines. VT-GAL4 UAS-trpA1 flies were placed in open 10-mm-diameter arenas and visually monitored for aberrant locomotion upon warming to ~30°C, the temperature at which the TrpA1 channel opens. From 3470 GAL4 lines screened, 4 were identified in which flies walked backward in warmed chambers but not at the control temperature of ~24°C. Backward walking was most pronounced with the line VT50660, which was selected for further analysis (movie S1). We called this the “moonwalker” line. These flies walk backward by reversing both the order and timing of individual leg movements (fig. S1 and supplementary materials).

To facilitate a quantitative analysis of backward and forward walking, we designed two assay systems. The first uses a ringlike chamber in which flies are confined to a narrow circular groove of 2.5-mm width and 16-mm diameter (Fig. 1A). This chamber restricts the fly’s lateral movement but leaves it unconstrained for forward or backward locomotion. We also developed computer software to automatically track the orientation and movement of individual flies (Fig. 1A). In these assays, VT50660-GAL4 UAS-trpA1 flies predominantly and persistently walked backward at 30°C (Fig. 1, B to E, table S1, and movie S2). The same flies tested at 22°C, or UAS-trpA1 control flies tested at either temperature, rarely walked backward.

Fig. 1 Activation and inhibition of backward walking.

(A) Ring assay for locomotion direction and tracking sample for a fly walking predominantly forward in an anticlockwise direction. The black line indicates the fly’s position in the ring (from 0° to 360°), and green and magenta indicate forward and backward walking, respectively. (B) Total distance walked in ring chamber during 5-min assay (n = 20 flies). Box-and-whisker plots in this and other panels show 10-, 25- 50-, 75-, and 90-percentiles. ***P < 0.0001 in comparison to UAS-trpA1 controls. (C to E) Representative tracking data for ring walking assays. Each column shows tracking data for one fly for 5 min. (F) Linear walking assay and tracking sample. (G) Total distance walked during 10-min assays in linear (top, n = 24) and ring (bottom, n = 20) chambers. ***P < 0.0001 in comparison to UAS-TNT controls. (H and I) Representative tracking data for linear walking assays. Each column shows tracking data for one fly for 10 min.

The second assay system uses a linear groove, 1.5 mm wide and 75 mm long, designed with the objective of inducing flies to walk backward (Fig. 1F). Stick insects walk backward upon mechanical stimulation of their antennae (9), suggesting that backward walking is used as a strategy to maneuver out of tight “dead ends.” Upon reaching the end of the linear assay chamber, flies are similarly confronted with a dead end, and the chamber is too narrow for them to easily turn around and walk forward again in the opposite direction. We anticipated that, trapped in this way, flies too might walk backward, and if so, we could determine whether backward walking requires the activity of moonwalker neurons.

Control flies indeed walked backward for several millimeters upon reaching the end of the linear chamber (Fig. 1, G and I, table S2, and movie S3). To test whether moonwalker neurons contribute to this backward walking, we combined VT50660-GAL4 with a UAS-TNT transgene, which encodes an inhibitor of synaptic transmission (10). In contrast to control flies, VT50660-GAL4 UAS-TNT flies rarely backed up when they reached the end of the chamber (Fig. 1, G and H, table S2, and movie S3). Rather, they typically stalled there, often for several minutes, before eventually squeezing around to walk forward again in the opposite direction. The same behavior was observed upon acute silencing in adult flies (fig. S2). In linear chambers, VT50660-GAL4 UAS-TNT flies also walked forward less than control flies (Fig. 1G), but this was due to the extended periods they spent stuck at the ends of the chamber. In circular chambers, in which forward walking is unimpeded, VT50660-GAL4 UAS-TNT walked forward at least as much as control flies (Fig. 1G and table S3). Thus, the activity of moonwalker neurons is essential for backward but not forward locomotion.

VT50660-GAL4 reproducibly labels seven distinct cell types in the central nervous system (Fig. 2, A to D, and fig. S3, A to E). We expressed an epitope-tagged TrpA1c-myc in random subsets of these cells, tested them for moonwalking at 30°C, and then dissected and stained the central nervous systems of individual flies to determine in which cells TrpA1c-myc was expressed (supplementary materials). Two cell types were bilaterally labeled in most of the moonwalkers (33 of 36) and in none of the non-moonwalkers (0 of 34; Fig. 2E and fig. S3F). One of these is a descending neuron, which we refer to as MDN (moonwalker descending neuron; Fig. 2, F and H); the other is an ascending neuron, MAN (moonwalker ascending neuron; Fig. 2, G and H). All other cell types were equally often labeled in both moonwalking and non-moonwalking flies (Fig. 2E). Moreover, of 17 flies in which only MDN or MAN was bilaterally labeled, but not both, only 3 were scored as moonwalkers (fig. S3F). We conclude from these data that sustained backward walking requires the coactivation of both MDN and MAN.

Fig. 2 Stochastic activation reveals moonwalker neurons.

(A) Brain (left) and VNC (right) of a VT50660-GAL4 UAS-mCD8-GFP male stained with antibodies against GFP (anti-GFP) (yellow) and the synaptic marker monoclonal antibody nc82 (blue). (B to D) Partial segmentations of single neurons representing each of seven neuronal classes that compose the VT50660 pattern, following nonrigid registration onto common reference templates. (E) Frequency of labeling for each cell type in moonwalker and nonmoonwalker flies. ***P < 0.0001, Fisher’s exact test. (F and G) Single MDN and MAN neurons stained with anti-GFP (yellow), with nc82 counterstain (blue), and registered onto a common reference template. Samples were obtained during stochastic labeling experiments with UAS>stop>mCD8-GFP. (H) Overlaid registered segmentations of single MDN and MAN cells.

To reproducibly and specifically target transgene expression to these two cell types, we used the split-GAL4 system (11, 12) to derive a set of four drivers that target MDN but not MAN (MDN-1-4), four that target MAN but not MDN (MAN-1-4), and two that target both (MDN+MAN-1-2; Fig. 3 and fig. S4, table S4, and supplementary materials). We first used these split-GAL4 reagents to further characterize the anatomy of MDN and MAN, by driving expression of the membrane marker mCD8-GFP (green fluorescent protein) (13), the dendritic marker DenMark (14), or the presynaptic marker syt-GFP (15). We typically observed two MDN cells per hemisphere, with their soma in the medial posterior protocerebrum, bilateral dendritic arborizations in the medial ventral protocerebrum and subesophageal ganglion (SOG), and axons that extend down to the contralateral thoracic ganglia (Fig. 3D). Presynaptic sites of MDN were observed in the SOG and in each of the leg neuropils. One MAN cell was observed on each side of the midline in the metathoracic (hindleg) ganglion, where its dendrites are also located. MAN sends an axon to the brain, with its predominant presynaptic sites in close proximity to MDN dendrites in the medial SOG (Fig. 3E). Some presynaptic sites for MAN were also observed within the nerve cord, in both the metathoracic ganglion and the wing neuropil.

Fig. 3 Split-GAL4 combinations for MDN and MAN.

(A to C) Brains (top) and ventral nerve cords (VNCs, bottom) of flies carrying the indicated split-GAL4 combination and UAS-mCD8-GFP, stained with anti-GFP (yellow) and nc82 (blue). Red and green arrowheads indicate MDN and MAN soma, respectively. SOG, subesophageal ganglion; T1 to T3, leg neuropils. (D and E) Brains (top) and VNCs (bottom) of flies in which the indicated split-GAL4 combination drives expression of UAS-syt-GFP and UAS-DenMark reporters, stained with anti-GFP (green, presynaptic), anti-DsRed (DenMark, red, postsynaptic), and nc82 (blue).

In functional assays, we found that both of the MDN+MAN combinations recapitulated the activation and silencing phenotypes we had previously observed with the original VT50660 moonwalker line (Fig. 4 and fig. S5, tables S5 to S8, and movies S6 and S7). Moreover, activation of MDN alone was sufficient to induce backward walking (Figs. 4A and fig. S5B, table S5, and movie 6), and synaptic silencing of MDN almost completely blocked backward walking (Fig. 4C and fig. S5E, table S7, and movie S7). Silencing of MDN did not disrupt forward walking (Fig. 4D and table S8). In contrast, activation (Fig. 4A and fig. S5C, table S5, and movie S6) and silencing (Fig. 4C and fig. S5F, table S7, and movie 7) of MAN had little, if any, effect on backward walking. We thus conclude that MDN activity alone is both necessary and sufficient to trigger backward walking.

Fig. 4 Distinct functions of MDN and MAN in the control of walking direction.

(A and B) Total walking distance (A) and mean distance per bout (B) in 5-min assays in ring chambers (n = 20 to 58). Colored dots show P values for comparisons to no GAL4 (–) and VT50660-GAL4 UAS-trpA1 (VT) controls. (C and D) Total walking distance in 10-min assays in linear chambers [(C), n = 16 to 60] and 10-min assays in ring chambers [(D), n = 18 to 66]. Colored dots show P values for comparisons to no GAL4 (–) and no UAS-TNT (TNT) controls.

Nonetheless, our stochastic activation experiments, in which we had selected the most persistent backward walkers, suggested that MAN activation does somehow facilitate sustained backward walking. Indeed, in the ring chambers, although flies in which MDN alone was activated walked backward for total distances similar to those in which both MDN and MAN were activated (Fig. 4A and table S5), backward walking was more sustained when both cell types were activated (Fig. 4B, table S6, and movie S6). Average bout distances for backward walking were <10 mm for MDN activation but >40 mm for MDN+MAN activation. These observations suggest that MAN might promote backward walking, either by inhibiting the natural tendency to walk forward or by maintaining the backward walking state induced upon activation of MDN. To discriminate between these possibilities, we examined the forward-walking behavior of flies in which MAN was either activated or silenced. Silencing MAN had no consistent effect on forward locomotion (Fig. 4D and tables S7 and S8), but activating MAN induced frequent stalling and hence reduced bout lengths (Fig. 4B and fig. S5C, and tables S5 and S6). These data favor a model in which MAN activity inhibits forward locomotion. Accordingly, activation of both MDN and MAN is necessary for the persistent backward walking that characterizes the “moonwalker” phenotype.

In summary, we have identified a descending neuron, MDN, that triggers backward walking in flies. Persistent backward walking is facilitated by MAN, an ascending neuron that appears to inhibit forward walking. MDN resembles the DBNc1-2 or DBNc2-2 neurons in crickets (16) and the DMIa-1 neuron in cockroaches (17), all of which respond with short latency (<8 ms) to mechanical stimulation of the antennae. MDN might similarly receive direct synaptic input from antennal or other head mechanoreceptors. Analogous circuits mediate backward crawling in Caenorhabditis elegans, with the AVA command interneurons receiving input from head mechanosensory neurons as well as upstream interneurons (18).

MDN activation reverses both the sequence and phase of leg movements, suggesting that it might flip a critical switch within the premotor circuits of the thoracic ganglia. In stick insects, backward walking is brought about by a switch in phase of the two muscles that operate the thorax-coxa joint, the protractor and retractor (19). In forward walking, loading of the leg during stance phase leads to activation of the retractor, pushing the leg back. Unloading activates the protractor, bringing the leg forward in swing phase. In backward walking, these proprioceptive inputs are reversed: Loading activates the protractor and unloading, the retractor. A descending neuron for backward locomotion could potentially induce this switch, directly or indirectly, by providing presynaptic inhibition onto the excitatory outputs of core CPG interneurons (20), thereby reversing the sign of feedback signals from the load sensors on the legs (21).

Is there an antagonistic counterpart to MDN that would trigger forward locomotion, as proposed on theoretical grounds (20) and analogous to the AVB command interneurons in C. elegans (18)? Or might the thoracic locomotor circuits instead operate by default in the forward mode, requiring only a single descending signal to discriminate between backward and forward walking? Although its function is presently less clear, we suggest that MAN stabilizes the switch to backward walking by inhibiting a default forward-walking state. The identification of MDN and MAN provides a first glimpse into the nature and logic of the signals that pass between the brain and local motor circuits, steering the animal toward or away from specific targets.

Supplementary Material

Supplementary Text

Materials and Methods

Figs. S1 to S5

Tables S1 to S8

References (2231)

Movies S1 to S7

References and Notes

  1. Acknowledgments: We thank L. Tirian and the entire VT project team for generating and staining VT GAL4 and split-GAL4 lines; A. Büschges and M. Palfreyman for insightful discussions and comments on the manuscript; M. Columbini and D. Kummerer for constructing the various assay chambers; K. Heinze for help with the laser sheet setup; and H. Grasberger and W. Li for help in developing the tracker. Basic research at the IMP is funded in part by Boehringer Ingelheim GmbH.
View Abstract

Navigate This Article