A screen for constituents of motor control and decision making in Drosophila reveals visual distance-estimation neurons

Abstract

Climbing over chasms larger than step size is vital to fruit flies, since foraging and mating are achieved while walking. Flies avoid futile climbing attempts by processing parallax-motion vision to estimate gap width. To identify neuronal substrates of climbing control, we screened a large collection of fly lines with temporarily inactivated neuronal populations in a novel high-throughput assay described here. The observed climbing phenotypes were classified; lines in each group are reported. Selected lines were further analysed by high-resolution video cinematography. One striking class of flies attempts to climb chasms of unsurmountable width; expression analysis guided us to C2 optic-lobe interneurons. Inactivation of C2 or the closely related C3 neurons with highly specific intersectional driver lines consistently reproduced hyperactive climbing whereas strong or weak artificial depolarization of C2/C3 neurons strongly or mildly decreased climbing frequency. Contrast-manipulation experiments support our conclusion that C2/C3 neurons are part of the distance-evaluation system.

Figure 1. Diagram of the ring-gap climbing assay and performance of 2,415 GAL4 lines driving UAS-shibire^ts to inactivate sets of neurons at 34 °C.

(a) Mattfinished Perspex climbing disk with five concentric ring-gaps of 2.0 to 4.0 mm widths (in steps of 0.5 mm) and a bowl-like centre recess milled into it. Gaps are filled half-high with water plus detergent. Fifteen anesthetized flies of a given line are placed into the recess and a coated glass plate is lowered onto the disk surface. The cover glass is lifted by 3.0 mm when all flies are awake. Flies can distribute for 10 min; an overhead camera records their whereabouts once a 1s; the red backlight improves visibility. (b) Each diamond of the scatter plot represents the median outcome of 3 to 32 replications per line in sets of 15 flies (16,189 experiments; Supplementary Table 1. The median maximum of the mean distance from centre of the 15 flies reached any time during the 10-min recording is represented on the x-axis (in relative units of the radius of the disk [0, centre to 1, outer rim]) whereas the y-axis reflects the median percentage of dead flies after 10 min. In terms of the median maximum mean distance from centre the lowest 10% of all R-lines are classified as under-performers and the uppermost 10% as over-performers. We further classified the R-lines into groups with either normal or significantly increased rate of dead flies. The boundary was chosen at 10% of the R-lines with the most losses. These definitions divide the R-lines into 6 classes. The centre of the green cross represents, in both dimensions, the medians of 1,412 control experiments, each comprising 15 flies of pBDPGAL4U driving UAS-shi^ts at 34 °C. Example lines of Figs 1c and 2 are marked by red rings. The thick lines represent the 25%-/75%-quartiles, and the thin lines the 10%-/90%-quantiles. (c) Examples of the development of the mean distance from centre of 15 flies per experiment over time (relative units; 0, centre and 1, rim of the disk). Maximum mean distances are indicated by circles.

Figure 2. High-speed video analysis data of the example lines shown in Fig. 1.

(a1) Shows a sketch of the setup and (a2) centre details. A catwalk with ramps on both sides and the gap in the middle is situated on a water-surrounded platform. Side and top views of climbing flies are taken at 200 fps against a LED-backlight and under LED-downlight, respectively. (b1–g1) Overhead camera shots at t = 10 min, the end of the experiments (the drilled pipe for filling the gaps can be seen). Arrowheads and numbers indicate dead flies in the grooves. (b2–g2) Fraction of climbing trials with percentage of falling, and (b3–g3) fraction of successful crossings, respectively, both per 10 approaches to the gap per fly at 34 °C. N  ≥ 10 flies per line and per two widths. Lines were tested on surmountable (2.5 to 4.0 mm) and insurmountable widths (5.0 and 6.0 mm) to test climbing and width-estimation capabilities. Boxes indicate 25%-/75%-quartiles, thick lines medians. Whiskers indicate 10%-/90%-quantiles. Insets show typical side views; successful crossing of a demanding gap width (b3) by a pBDPGAL4U > shi^ts control fly, (c3) failure to initiate crossing at an easy gap width by a R61H03 > shi^ts fly, (d3) failure in surmounting an easy gap by a R24D10 > shi^ts fly, (e3) malfunction of the tarsi (arrow) by a R23D06 > shi^ts fly, (f3) fruitless attempt by a R22F08 > shi^ts fly at a gap of insurmountable width, and (g3) successful crossing of a gap of just surmountable width by a R68A11 > shi^ts fly. The latter flies perform better than controls; median leg lengths (h) n = 14; femur + tibia + tarsi) proof that they are not bigger than pBDPGAL4U > shi^ts flies. Wilcoxon rank sum tests are performed per gap width against respective pBDPGAL4U > shi^ts data (repeated from b in light green); *p < 0.05; **p < 0.01; ***p < 0.001. A detailed account of statistics is given in Supplementary Table 4.

Figure 3. Overeager phenotype replicated by blocking just C2 and/or C3 feedback neurons.

(a1) R22F08 expression pattern comprises scattered expression in C2 feedback neurons; scale bar, 100 μm. The split-GAL4 system is used to specifically target all C2 (b), all C3 neurons (c), or all C2 and C3 neurons (d). For visualization purposes, individual cells were labelled by the Multicolor FlpOut technique (b1,c1). Scale bar in b1, 20 μm, is valid for (b1–d1). (d1) Golgi reconstructions of C2 and C3 neurons in an outline of the optic lobes for comparison (modified after Fischbach and Dittrich, 1989). (a2–d2) fraction of climbing trials, and (a3–d3) fraction of successful crossings, respectively, both per 10 approaches to the gap per fly driving UAS-shi^ts at 34 °C. N = 10 or more flies per line and per two widths. Line-number combinations indicate AD-split GAL4 and DBD-split GAL4. Expression of shi^ts in all C2, all C3, or all C2 and C3 neurons leads to increased numbers of attempts at insurmountable gap widths. Wilcoxon rank sum tests against pBDPGAL4U > shi^ts control data (light green, taken from Fig. 3b) of the same gap width at 34 °C; *p < 0.05; **p < 0.01; ***p < 0.001. Panels a2 and a3 are reproduced from Fig. 2f2.

Figure 4. The overeager phenotype turns into an overcautious phenotype when activating instead of inactivating all C2, C3 or C2 and C3 feedback neurons.

(a) Control line driving UAS-dTrpA1 at 29 °C. (b–d) Lines and conventions as for Fig. 3 but lines are driving UAS-dTrpA1 at 29 °C. N ≥ 10 flies or more per line and per two widths. Wilcoxon rank sum tests are performed per gap width against pBDPGAL4U > dTrpA1 data obtained at 29 °C (light green, repeated from A); *p < 0.05; **p < 0.01; ***p < 0.001.

Figure 5. Overcautious or overeager behaviour can be elicited by visual properties of the gap’s opposite sidewall.

Catwalks were either made of dark plastic material (as for all other experiments in this paper), clear Perspex, or dark plastic material and black and white vertical stripes at the opposite sidewall. Gap width was invariably 4.0 mm. Trials are most significantly reduced at the clear gap and highly significantly increased at the decorated gap. Neither blocking (shi^ts at 34 °C) nor activation (dTrpA1 at 29 °C) of C2 neurons using (R20C11 AD-split GAL4 – R25B02 DBD-split GAL4) are effective when visibility of the opposite sidewall is low as in the case of the clear catwalk. If visibility is given C2-neuron blocking significantly increases the fraction of climbing trials as compared to C2-neuron activation.