LETTER. A directional tuning map of Drosophila elementary motion detectors

Transcription

1 doi:1.138/nature1232 A diretional tuning map of Drosophila elementary motion detetors Matthew S. Maisak 1 *, Juergen Haag 1 *, Georg Ammer 1, Etienne Sere 1, Matthias Meier 1, Aljosha Leonhardt 1, Taea Shilling 1, Armin Bahl 1, Gerald M. Ruin 2, Aljosha Nern 2, Barry J. Dikson 3, Dierk F. Reiff 1 {, Elisaeth Hopp 1 & Alexander Borst 1 The extration of diretional motion information from hanging retinal images is one of the earliest and most important proessing steps in any visual system. In the fly opti loe, two parallel proessing streams have een anatomially desried, leading from two first-order interneurons, L1 and L2, via T4 and T5 ells onto large, wide-field motion-sensitive interneurons of the loula plate 1.Therefore, T4 and T5 ells are thought to have a pivotal role in motion proessing; however, owing to their small size, it is diffiult to otain eletrial reordings of T4 and T5 ells, leaving their visual response properties largely unknown. We irumvent this prolem y means of optial reording from these ells in Drosophila, using the genetially enoded alium indiator GCaMP5 (ref. 2). Here we find that speifi supopulations of T4 and T5 ells are diretionally tuned to one of the four ardinal diretions; that is, front-to-ak, ak-to-front, upwards and downwards. Depending on their preferred diretion, T4 and T5 ells terminate in speifi sulayers of the loula plate. T4 and T5 funtionally segregate with respet to ontrast polarity: whereas T4 ells seletively respond to moving rightness inrements (ON edges), T5 ells only respond to moving rightness derements (OFF edges). When the output from T4 or T5 ells is loked, the responses of postsynapti loula plate neurons to moving ON () or OFF edges () are seletively ompromised. The same effets are seen in turning responses of tethered walking flies. Thus, starting with L1 and L2, the visual input is split into separate ON and OFF pathways, and motion along all four ardinal diretions is omputed separately within eah pathway. The output of these eight different motion detetors is then sorted suh that ON (T4) and OFF (T5) motion detetors with the same diretional tuning onverge in the same layer of the loula plate, jointly providing the input to downstream iruits and motion-driven ehaviours. Most of the neurons in the fly rain are dediated to image proessing. The respetive part of the head ganglion, alled the opti loe, onsists of several layers of neuropile alled lamina, medulla, loula and loula plate, all uilt from repetitive olumns arranged in a retinotopi way (Fig. 1a). Eah olumn houses a set of identified neurons that, on the asis of Golgi staining, have een desried anatomially in great detail 3 5.Owingto their small size, however, most of these olumnar neurons have never een reorded from eletrophysiologially. Therefore, their speifi funtional role in visual proessing is still largely unknown. This fat is ontrasted y rather detailed funtional models aout visual proessing inferred from ehavioural studies and reordings from the large, eletrophysiologially aessile output neurons of the fly loula plate (tangential ells). As the most prominent example of suh models, the Reihardt detetor derives diretional motion information from primary sensory signals y multiplying the output from adjaent photoreeptors after asymmetri temporal filtering 6. This model makes a numer of rather ounter-intuitive preditions all of whih have een onfirmed experimentally (for review, see ref. 7). Yet, the neurons orresponding to most a e plate Inner hiasm T5 T4 plate Medulla Lamina Retina Medulla Distal Max ΔF/F d f Min Medulla Layers plate T4 T5 Distal plate Figure 1 Diretional tuning and layer-speifi projetion of T4 and T5 ells. a, Shemati diagram of the fly opti loe. In the loula plate, motionsensitive tangential ells extend their large dendrites over many hundreds of olumns. Shown are the reonstrutions of the three ells of the horizontal system 22., Anatomy of T4 and T5 ells, as drawn from Golgi-impregnated material (from ref. 5)., Confoal image of the Gal4-driver line R42F6, shown in a horizontal ross-setion (from ref. 1). Neurons are marked in green (Kir2.1 EGFP laelled), whereas the neuropile is stained in purple y an antiody against the postsynapti protein Dlg. Sale ar, 2 mm. d, Two-photon image of the loula plate of a fly expressing GCaMP5 under the ontrol of the same driver line R42F6. Sale ar, 5 mm. The size and orientation of the image approximately orresponds to the yellow square in. e, Relative fluoresene hanges (DF/F) otained during 4-s grating motion along the four ardinal diretions, overlaid on the greysale image. Eah motion diretion leads to ativity in a different layer. Minimum and maximum DF/F values were.3 and 1. (horizontal motion), and.15 and.6 (vertial motion). f, Compound representation of the results otained from the same set of experiments. Sale ar, 5 mm. Results in e and f represent the data otained from a single fly averaged over four stimulus repetitions. Similar results were otained from six other flies. 1 Max Plank Institute of Neuroiology, Martinsried, Germany. 2 Janelia Farm Researh Campus, Ashurn, Virginia 2147, USA. 3 Institute of Moleular Pathology, 13 Vienna, Austria. {Present address: Institute Biology 1, Alert-Ludwigs University, 7985 Freiurg, Germany. *These authors ontriuted equally to this work. 212 NATURE VOL 5 8 AUGUST Mamillan Pulishers Limited. All rights reserved

2 RESEARCH of the iruit elements of the Reihardt detetor have not een identified so far. Here, we fous on a set of neurons alled T4 and T5 ells (Fig. 1) whih, on the asis of irumstantial evidene, have long een speulated to e involved in motion detetion 1,8 1. However, it is unlear to what extent T4 and T5 ells are diretionally seletive or whether diretion seletivity is omputed or enhaned within the dendrites of the tangential ells. Another important question onerns the funtional separation etween T4 and T5 ells; that is, whether they arry equivalent signals, maye one eing exitatory and the other inhiitory on the tangential ells, or whether they segregate into diretional- and non-diretional pathways 11 or into separate ONand OFF-motion hannels 12,13. To answer these questions, we omined Gal4-driver lines speifi for T4 and T5 ells 14 with GCaMP5 (ref. 2) and optially reorded the visual response properties using two-photon fluoresene mirosopy 15. In a first series of experiments, we used a driver line laelling oth T4 and T5 ells. A onfoal image (Fig. 1, modified from ref. 1) revealed lear laelling (in green) in the medulla (T4 ell dendrites), in the loula (T5 ell dendrites), as well as in four distint layers of the loula plate, representing the terminal arorizations of the four supopulations of oth T4 and T5 ells. These four layers of the loula plate an also e seen in the two-photon mirosope when the alium indiator GCaMP5 is expressed (Fig. 1d). After stimulation of the fly with grating motion along four ardinal diretions (front-to-ak, ak-to-front, upwards and downwards), ativity is onfined to mostly one of the four layers, depending on the diretion in whih the grating is moving (Fig. 1e). The outome of all four stimulus onditions an e omined into a single image y assigning a partiular olour to eah pixel depending on the stimulus diretion to whih it responded most strongly (Fig. 1f). From these experiments it is lear that the four supopulations of T4 and T5 ells produe seletive alium signals depending on the stimulus diretion, in agreement with previous deoxygluose laelling 8. Sudden hanges of the overall luminane evokes no responses in any of the layers (field fliker; n 5 4 experiments, data not shown). However, gratings flikering in ounter-phase lead to layer-speifi responses, depending on the orientation of the grating (Supplementary Fig. 1). The retinotopi arrangement of this input to the loula plate is demonstrated y experiments where a dark edge was moved within a small area of the visual field only. Depending on the position of this area, ativity of T4 and T5 ells is onfined to different positions within the loula plate (Fig. 2a). Consequently, when moving a right vertial edge horizontally from ak to front, ativity of T4 and T5 ells is eliited sequentially in layer 2 of the loula plate (Fig. 2). These two experiments also demonstrate that T4 and T5 ells indeed signal motion loally. We next investigated the question of where diretion seletivity of T4 and T5 ells arises; that is, whether it is already present in the dendrite, or whether it is generated y synapti interations within the loula plate. This question is hard to answer, as the dendrites of oth T4 and T5 ells form a dense mesh within the proximal layer of the medulla (T4) and the loula (T5), respetively. However, signals within the inner hiasm where individual proesses of T4 and T5 ells an e resolved in some preparations show a lear seletivity for motion in one over the other diretions (Fig. 2). Suh signals are as diretionally seletive as the ones measured within the loula plate, demonstrating that the signals delivered from the dendrites of T4 and T5 ells are already diretionally seletive. To assess the partiular ontriution of T4 and T5 ells to the signals oserved in the aove experiments, we used driver lines speifi for T4 and T5 ells, respetively. Applying the same stimulus protool and data evaluation as in Fig. 1, idential results were otained as efore for oth the T4- as well as the T5-speifi driver line (Fig. 3a, ). We onlude that T4 and T5 ells eah provide diretionally seletive signals to the loula plate, in ontrast to previous reports 11. Thus, oth T4 and T5 ells an e grouped, aording to their preferred diretion, into four sulasses overing all four ardinal diretions, reminisent of ON OFF ganglion ells of the rait retina 16. a Inner hiasm plate Inner hiasm plate Distal plate Figure 2 Loal signals of T4 and T5 ells. a, Retinotopi arrangement of T4 and T5 ells. A dark edge was moving repeatedly from front-to-ak within a 15u wide area at different azimuthal positions (left). This leads to relative fluoresene hanges at different positions along the proximal distal axis within layer 1 of the loula plate (right). Sale ar, 5 mm. Similar results have een otained in four other flies., Sequential ativation of T4 and T5 ells. A right edge was moving from ak-to-front at 15u s 21. Sale ar, 5 mm. Similar results have een otained in six other flies., Signals reorded from individual fires within the inner hiasm (left) reveal a high degree of diretion seletivity (right). Sale ar, 5 mm. Similar results were otained from four other flies, inluding oth lines speifi for T4 and T5 ells. Response traes in and are derived from the region of interest enirled in the image with the same olour. We next addressed whether T4 ells respond differently to T5 ells. To answer this question, we used, instead of gratings, moving edges with either positive (ON edge, rightness inrement) or negative (OFF edge, rightness derement) ontrast polarity as visual stimuli. We found that T4 ells strongly responded to moving ON edges, ut showed little or no response to moving OFF edges (Fig. 3). This is true for T4 ells terminating in eah of the four layers. We found the opposite for T5 ells. T5 ells seletively responded to moving OFF edges and mostly failed to respond to moving ON edges (Fig. 3d). Again, we found this for T5 ells in eah of the four layers. We next addressed whether there are any other differenes in the response properties etween T4 and T5 ells y testing the veloity tuning of oth ell populations y means of stimulating flies with grating motion along the horizontal axis from the front to the ak at various veloities overing two orders of magnitude. T4 ells revealed a maximum response at a stimulus veloity of 3u s 21, orresponding to a temporal frequeny of 1 Hz (Fig. 3e). T5 ell responses showed a similar dependeny on stimulus veloity, again with a peak at a temporal frequeny of 8 AUGUST 213 VOL 5 NATURE Mamillan Pulishers Limited. All rights reserved

3 RESEARCH LETTER a e g 1. T4 ells Temporal frequeny (Hz) 1 Hz (Fig. 3f). Thus, there is no ovious differene in the veloity tuning etween T4 and T5 ells. As another possiility, T4 ells might funtionally differ from T5 ells with respet to their diretional tuning width. To test this, we stimulated flies with gratings moving into 12 different diretions and evaluated the relative hange of fluoresene in all four layers of the loula plate. Using the T4-speifi driver line, we found an approximate half width of 6 9u of the tuning urve, with the peak responses in eah layer shifted y 9u (Fig. 3g). No derease of alium was detetale for grating motion opposite to the preferred diretion of the respetive layer. When we repeated the experiments using the T5-speifi driver line, we found a similar dependene of the relative hange of fluoresene on the stimulus diretion (Fig. 3h). We onlude that T4 ells have the same veloity and orientation tuning as T5 ells. The only funtional differene we were ale to detet remains their seletivity for ontrast polarity. Our finding aout the different preferene of T4 and T5 ells for the polarity of a moving ontrast makes the strong predition that seletive f h 1. T5 ells d 1.5 Layer 1 Layer 2 Layer 3 Layer Layer 1 Layer 2 Layer 3 Layer Layer 1 Layer 2 Layer 3 Layer Temporal frequeny (Hz) Layer 1 Layer 2 6 Layer 3 Layer 4 3 Figure 3 Comparison of visual response properties etween T4 and T5 ells. a,, Relative fluoresene hanges (DF/F) of the loula plate terminals of T4 (a) and T5 () ells otained during grating motion along the four ardinal diretions. Results represent the data otained from a single fly eah, averaged over two stimulus repetitions. Sale ars, 5 mm. Similar results have een otained in ten other flies., d, Responses of T4 () and T5 (d) ells to ON and OFF edges moving along all four ardinal diretions. ON (white) and OFF (lak) responses within eah layer are signifiantly different from eah other, with P,.5 exept for layers 3 and 4 in T5 ells, where P,.5. e, f, Responses of T4 (e) and T5 (f) ells to gratings moving horizontally at different temporal frequenies. Relative fluoresene hanges were evaluated from layer 1 of the loula plate and normalized to the maximum response efore averaging. g, h, Responses of T4 (g) and T5 (h) ells to gratings moving in 12 different diretions. Relative fluoresene hanges were evaluated from all four layers of the loula plate normalized to the maximum response efore averaging. Data represent the mean 6 s.e.m. of the results otained in n 5 8 (), n 5 7(d), n 5 6(e), n 5 7(f), n 5 6(g) and n 5 5(h) different flies. Signifianes indiated are ased on two-sample t-test lokade of T4 or T5 ells should seletively ompromise the responses of downstream loula plate tangential ells to either ON or OFF edges. To test this predition, we loked the output of either T4 or T5 ells via expression of the light hain of tetanus toxin 17 and reorded the responses of tangential ells via somati whole-ell path to moving ON and OFF edges. In response to moving ON edges, strong and reliale diretional responses were oserved in all ontrol flies (Fig. 4a). However, T4-lok flies showed a strongly redued response to ON edges, whereas the responses of T5-lok flies were at the level of ontrol flies (Fig. 4, ). When we used moving OFF edges, ontrol flies again responded with a large amplitude (Fig. 4d). However, the responses of T4-lok flies were at the level of ontrol flies, whereas the responses of T5-lok flies were strongly redued (Fig. 4e, f). These findings are reminisent on the phenotypes otained from loking lamina ells L1 and L2 (ref. 13) and demonstrate that T4 and T5 ells are indeed the motion-oding intermediaries for these ontrast polarities on their way to the tangential ells of the loula plate. Whether the residual responses to ON edges in T4-lok flies and to OFF edges in T5-lok flies are due to an inomplete signal separation etween the two pathways or due to an inomplete geneti lok in oth fly lines is urrently unlear. To address the question of whether T4 and T5 ells are the only motion detetors of the fly visual system, or whether they represent one ell lass, in parallel to other motion-sensitive elements, we used tethered flies walking on an air-suspended sphere 18 and stimulated them y ON and OFF edges moving in opposite diretions 19.Asin the previous experiments, we loked T4 and T5 ells speifially y seletive expression of the light hain of tetanus toxin. During alaned motion, ontrol flies did not show signifiant turning responses to either side (Fig. 4g). T4-lok flies, however, strongly followed the diretion of the moving OFF edges, whereas T5-lok flies followed the diretion of the moving ON edges (Fig. 4h, i). In summary, the seletive preferene of T4-lok flies for OFF edges and of T5-lok flies for ON edges not only orroorates our findings aout the seletive preferene of T4 and T5 ells for different ontrast polarities, ut also demonstrates that the signals of T4 and T5 ells are indeed the major, if not exlusive, inputs to downstream iruits and motiondriven ehaviours. Almost a hundred years after T4 and T5 ells have een anatomially desried 3, this study reports their funtional properties in a systemati way. Using alium as a proxy for memrane voltage 2,we found that oth T4 and T5 ells respond to visual motion in a diretionally seletive manner and provide these signals to eah of the four layers of the loula plate, depending on their preferred diretion. Both ell types show idential veloity and orientation tuning whih mathes the one of the tangential ells 21,22. The strong diretion seletivity of oth T4 and T5 ells is unexpeted, as previous studies had onluded that the high degree of diretion seletivity of tangential ells is due to a push pull onfiguration of weakly diretional input with opposite preferred diretion 23,24. Furthermore, as the preferred diretion of T4 and T5 ells mathes the preferred diretion of the tangential ells ranhing within orresponding layers, it is urrently unlear whih neurons are responsile for the null-diretion response of the tangential ells. As for the funtional separation etween T4 and T5 ells, we found that T4 ells seletively respond to rightness inrements, whereas T5 ells exlusively respond to moving rightness derements. Interestingly, parallel ON and OFF motion pathways had een previously postulated on the asis of seletive silening of lamina neurons L1 and L2 (ref. 13). Studies using apparent motion stimuli to proe the underlying omputational struture arrived at ontroversial onlusions: whereas some studies onluded that there was a separate handling of ON and OFF events y motion detetors 12,25,26, others did not favour suh a strit separation 19,27. The present study diretly demonstrates the existene of separate ON and OFF motion detetors, as represented y T4 and T5 ells, respetively. Furthermore, our results anatomially onfine the essential proessing steps of elementary 214 NATURE VOL 5 8 AUGUST Mamillan Pulishers Limited. All rights reserved

4 RESEARCH a 2 1 Control flies 2 1 Blok flies 15 1 PD ND response (mv) d e f Turning response (degrees per seond) g 6 ON-edge 3 3 OFF-edge Figure 4 Voltage responses of loula plate tangential ells and turning responses of walking flies to moving ON and OFF edges. a, d, Average time ourse of the memrane potential in response to preferred diretion motion minus the response to null diretion motion (PD 2 ND response) as reorded in three types of ontrol flies (stimulation period indiated y shaded area)., e, Sameasina, d, ut reorded in T4-lok flies (green) and T5-lok flies (red). The stimulus pattern, shown to the left, onsisted of multiple ON- (a) or OFF-edges (d)., f, Mean voltage responses (PD 2 ND) of tangential ells in the five groups of flies. Reordings were done from ells of the vertial 21 and the horizontal 22 system. Beause no differene was deteted etween them, data were pooled. Data omprise reordings from n 5 2 (TNT ontrol), n 5 12 (T4 ontrol), n 5 16 (T5 ontrol), n 5 17 () and n 5 18 () ells. In oth T4 and T5-lok flies, ON and OFF responses are signifiantly different h i from eah other with P,.1. In T4-lok flies, ON responses are signifiantly redued ompared to all three types of ontrol flies, whereas in T5- lok flies, OFF responses are signifiantly redued, oth with P,.1. g, Average time ourse of the turning response of three types of ontrol flies to ON and OFF edges moving simultaneously to opposite diretions (stimulation period indiated y shaded area). h, Same as in g, ut reorded from T4-lok flies (green) and T5-lok flies (red). i, Mean turning tendeny (6s.e.m.) during the last seond of the stimulation period averaged aross all flies within eah group. Data omprise average values otained in n 5 12 (TNT ontrols), n 5 11 (T4 ontrols), n 5 11 (T5 ontrols), n 5 13 () and n 5 12 (T5 lok) flies. Values of T4 and T5-lok flies are highly signifiantly different from zero with P,.1. Signifianes indiated are ased on two-sample t-test. motion detetion that is, asymmetri temporal filtering and nonlinear interation to the neuropile etween the axon terminals of lamina neurons L1 and L2 (ref. 28) and the dendrites of diretionally seletive T4 and T5 ells (Supplementary Fig. 2). The dendrites of T4 and T5 ells might well e the plae where signals from neighouring olumns interat in a nonlinear way, similar to the dendrites of starurst amarine ells of the verterate retina 29. METHODS SUMMARY Flies. Flies used in alium imaging experiments (Figs 1 3) had the following genotypes: T4/T5 line (w 2 ; 1/1; UAS-GCaMP5,R42F6-GAL4/UAS-GCaMP5, R42F6-GAL4), T4 line (w 2 ; 1/1; UAS-GCaMP5,R54A3-GAL4/UAS-GCaMP5, R54A3-GAL4), T5 line (w 2 ; 1/1; UAS-GCaMP5,R42H7-GAL4/UAS-GCaMP5, R42H7-GAL4). Flies used in eletrophysiologial and ehavioural experiments (Fig. 4) had idential genotypes of the following kind: TNT ontrol flies (w 1 /w 1 ; UAS-TNT-E/UAS-TNT-E; 1/1), T4 ontrol flies (w 1 /w 2 ; 1 /1; VT37588-GAL4/ 1), T5 ontrol flies (w 1 /w 2 ; 1/1; R42H7-GAL4/1), T4-lok flies (w 1 /w 2 ; UAS-TNT-E/1; VT37588-GAL4/1), T5-lok flies (w 1 /w 2 ; UAS-TNT-E/1; R42H7-GAL4/1). Two-photon mirosopy. We used a ustom-uilt two-photon laser sanning mirosope 29 equipped with a 34 water immersion ojetive and a mode loked Ti:sapphire laser. To shield the photomultipliers from the stimulus light, two separate arriers were used: the first was plaed diretly over the LEDs, the seond extended from the fly holder over the arena. Images were aquired at a resolution of pixels and a frame rate of 1.87 Hz, exept where indiated, using SanImage software 3. Eletrophysiology. Reordings were estalished under visual ontrol using a Zeiss Mirosope and a 34 water immersion ojetive. Behavioural analysis. The loomotion reorder was ustom-designed aording to ref. 18. It onsisted of an air-suspended sphere floating in a owl-shaped sphere holder. Motion of the sphere was reorded y two optial traking sensors. Visual stimulation. For alium imaging and eletrophysiologial experiments, we used a ustom-uilt LED arena overing 18u and 9u of the visual field along the horizontal and the vertial axis, respetively, at 1.5u resolution. For the ehavioural experiments, three 12-Hz LCD sreens formed a U-shaped visual arena with the fly in the entre, overing 27u and 114u of the visual field along the horizontal and the vertial axes, respetively, at.1u resolution. Data evaluation. Data were evaluated off-line using ustom-written software (Matla and IDL). Full Methods and any assoiated referenes are availale in the online version of the paper. Reeived 16 April; aepted 2 May Bausenwein, B., Dittrih, A. P. M. & Fishah, K. F. The opti loe of Drosophila melanogaster II. Sorting of retinotopi pathways in the medulla. Cell Tissue Res. 267, (1992). 2. Akeroom, J. et al. Optimization of a GCaMP alium indiator for neural ativity imaging. J. Neurosi. 32, (212). 3. Cajal, S. R. & Sanhez, D. Contriuion alonoimiento delos entros nerviosos de los insetos (Imprenta de Hijos de Niholas Moja, 1915). 4. Strausfeld, N. J. Atlas of an Inset Brain (Springer, 1976). 5. Fishah, K. F. & Dittrih, A. P. M. The opti loe of Drosophila melanogaster. I.A Golgi analysis of wild-type struture. Cell Tissue Res. 258, (1989). 8 AUGUST 213 VOL 5 NATURE Mamillan Pulishers Limited. All rights reserved

5 RESEARCH LETTER 6. Reihardt, W. Autoorrelation, a priniple for the evaluation of sensory information y the entral nervous system. In Sensory Communiation (ed. Rosenlith, W. A.) (MIT Press and John Wiley & Sons, 1961). 7. Borst, A., Haag, J. & Reiff, D. F. Fly motion vision. Annu. Rev. Neurosi. 33, 49 7 (21). 8. Buhner, E., Buhner. S. & Buelthoff, I. Deoxygluose mapping of nervous ativity indued indrosophila rainy visual movement. 1. Wildtype. J. Comp. Physiol. 155, (1984). 9. Strausfeld, N. J. & Lee, J. K. Neuronal asis for parallel visual proessing in the fly. Vis. Neurosi. 7, (1991). 1. Shnell, B., Raghu, V. S., Nern, A. & Borst, A. Columnar ells neessary for motion responses of wide-field visual interneurons in Drosophila. J. Comp. Physiol. A 198, (212). 11. Douglass, J. K. & Strausfeld, N. J. Visual motion-detetion iruits in flies: Parallel diretion- and non-diretion-sensitive pathways etween the medulla and loula plate. J. Neurosi. 16, (1996). 12. Franeshini, N., Riehle, A. & Le Nestour, A. Diretionally seletive motion detetion y inset neurons. In Faets of Vision (ed. Stavenga, H.) (Springer, 1989). 13. Joesh, M., Shnell, B., Raghu, S. V., Reiff, D. F. & Borst, A. ON and OFF pathways in Drosophila motion vision. Nature 468, 3 34 (21). 14. Pfeiffer, B. D. et al. Tools for neuroanatomy and neurogenetis in Drosophila. Pro. Natl Aad. Si. USA 15, (28). 15. Denk, W., Strikler, J. H. & We, W. W. Two-photon laser sanning fluoresene mirosopy. Siene 248, (199). 16. Oyster, C. W. & Barlow, H. B. Diretion-seletive units inrait retina: distriutionof preferred diretions. Siene 155, (1967). 17. Sweeney, S. T., Broadie, K., Keane, J., Niemann, H. & O Kane, C. J. Targeted expression of tetanus toxin light hain in Drosophila speifially eliminates synapti transmission and auses ehavioral defets. Neuron 14, (1995). 18. Seelig, J. D. et al. Two-photon alium imaging from head-fixed Drosophila during optomotor walking ehavior. Nature Methods 7, (21). 19. Clark, D. A., Bursztyn, L., Horowitz, M. A., Shnitzer, M. J. & Clandinin, T. R. Defining the omputational struture of the motion detetor in Drosophila. Neuron 7, (211). 2. Egelhaaf, M. & Borst, A. Calium aumulation in visual interneurons of the fly: Stimulus dependene and relationship to memrane potential. J. Neurophysiol. 73, (1995). 21. Joesh, M., Plett, J., Borst, A. & Reiff, D. F. Response properties of motion-sensitive visual interneurons in the loula plate of Drosophila melanogaster. Curr. Biol. 18, (28). 22. Shnell, B. et al. Proessing of horizontal opti flow in three visual interneurons of the Drosophila rain. J. Neurophysiol. 13, (21). 23. Borst, A. & Egelhaaf, M. Diretion seletivity of fly motion-sensitive neurons is omputedina two-stageproess. Pro. Natl Aad. Si. USA 87, (199). 24. Single, S., Haag, J. & Borst, A. Dendriti omputation of diretion seletivity and gain ontrol in visual interneurons. J. Neurosi. 17, (1997). 25. Eihner, H., Joesh, M., Shnell, B., Reiff, D. F. & Borst, A. Internal struture of the fly elementary motion detetor. Neuron 7, (211). 26. Joesh, M., Weer, F., Eihner, H. & Borst, A. Funtional speialization of parallel motion detetion iruits in the fly. J. Neurosi. 33, (213). 27. Egelhaaf, M. & Borst, A. Are there separate ON and OFF hannels in fly motion vision? Vis. Neurosi. 8, (1992). 28. Takemura, S. Y., Lu, Z. & Meinertzhagen, I. A. Synapti iruits of the Drosophila opti loe: the input terminals to the medulla. J. Comp. Neurol. 59, (28). 29. Euler, T., Detwiler, P. B. & Denk, W. Diretionally seletive alium signals in dendrites of starurst amarine ells. Nature 418, (22). 3. Pologruto, T. A., Saatini, B. L. & Svooda, K. SanImage: Flexile software for operating laser sanning mirosopes. Biomed. Eng. Online 2, 13 (23). Supplementary Information is availale in the online version of the paper. Aknowledgements We thank L. Looger, J. Simpson, V. Jayaraman and the Janelia GECI team for making and providing us with the GCaMP5 flies efore puliation; J. Plett for designing and engineering the LED arena; C. Theile, W. Essauer and M. Sauter for fly work; and A. Mauss, F. Gaiani and T. Bonhoeffer for ritially reading the manusript. This work was in part supported y the Deutshe Forshungsgemeinshaft (SFB 87). M.S.M., G.A., E.S., M.M., A.L., A.Ba and A.Bo are memers of the Graduate Shool of Systemi Neurosienes. Author Contriutions M.S.M. and J.H. jointly performed and, together with A.Bo., evaluated all alium imaging experiments. G.A., E.S. and M.M. reorded from tangential ells. A.L., T.S. and A.Ba. performed the ehavioural experiments. G.R., B.D. and A.N. generated the driver lines and haraterized their expression pattern. D.F.R. performed preliminary imaging experiments. E.H. helped with programming and developed the PMT shielding for the two-photon mirosope. A.Bo. designed the study and wrote the manusript with the help of all authors. Author Information Reprints and permissions information is availale at The authors delare no ompeting finanial interests. Readers are welome to omment on the online version of the paper. Correspondene and requests for materials should e addressed to A.Bo. (orst@neuro.mpg.de). 216 NATURE VOL 5 8 AUGUST Mamillan Pulishers Limited. All rights reserved

6 RESEARCH METHODS Flies. Flies were raised on standard ornmeal-agar medium at 25 uc and 6% humidity throughout development on a 12 h light/12 h dark yle. For alium imaging, we used the genetially enoded single-wavelength indiator GCaMP5, variant G, with the following mutations: T32L, R33P and D38Y (ref. 2). Expression of GCaMP5 was direted y three different Gal4 lines, all from the Janelia Farm olletion 14. Flies used in alium imaging experiments (Figs 1 3) had the following genotypes: T4/T5 line (w 2 ; 1/1; UAS-GCaMP5,R42F6-GAL4/ UAS-GCaMP5,R42F6-GAL4), T4 line (w 2 ; 1/1; UAS-GCaMP5,R54A3-GAL4/ UAS-GCaMP5,R54A3-GAL4), T5 line (w 2 ; 1/1; UAS-GCaMP5,R42H7- GAL4/UAS-GCaMP5,R42H7-GAL4). All driver lines were generated y the methods desried in ref. 14 and were identified y sreening a dataase of imaged lines, followed y reimaging of seleted lines 31. As homozygous for oth the Gal4- driver and the UAS-GCaMP5 genes, T4 flies also showed some residual expression in T5 ells, and T5 flies also in T4 ells. This unspeifi expression, however, was in general less than 25% of the expression in the speifi ells. Flies used in eletrophysiologial and ehavioural experiments (Fig. 4) had idential genotypes of the following kind: TNT ontrol flies (w 1 /w 1 ; UAS-TNT-E/UAS-TNT-E; 1/1), T4 ontrol flies (w 1 /w 2 ; 1 /1; VT37588-GAL4/1), T5 ontrol flies (w 1 /w 2 ; 1/1; R42H7-GAL4/1), T4-lok flies (w 1 /w 2 ; UAS-TNT-E/1; VT37588-GAL4/1), T5-lok flies (w 1 /w 2 ; UAS-TNT-E/1; R42H7-GAL4/1). UAS-TNT-E flies were derived from the Bloomington Stok Center (stok no ) and VT Gal4 flies were derived from the VDRC (stok no ). Before eletrophysiologial experiments, flies were anaesthetized on ie and waxed on a Plexiglas holder using ees wax. The dissetion of the fly utile and exposure of the loula plate were performed as desried previously (for imaging experiments, see ref. 32; for eletrophysiology, see ref. 21). Flies used in ehavioural experiments were taken from 18 uc just efore the experiment and immediately old-anaesthetized. The head, the thorax and the wings were glued to a needle using near-ultraviolet onding glue (Sinfony Opaque Dentin) and strong lue LED light (44 nm, dental uring-light, New Woodpeker). Two-photon mirosopy. We used a ustom-uilt two-photon laser sanning mirosope 33 equipped with a 34 water immersion ojetive (.8 NA, IR- Ahroplan; Zeiss). Fluoresene was exited y a mode loked Ti:sapphire laser (,1 fs, 8 MHz, 7 1,2 nm; pumped y a 1 W CW laser; oth Mai Tai; Spetraphysis) with a DeepSee aessory module attahed for dispersion ompensation ontrol resulting in etter pulse ompression and fluoresene at the target sample. Laser power was adjusted to 1 2 mw at the sample, and an exitation wavelength of 91 nm was used. The photomultiplier tue (H177PB-4, Hamamatsu) was equipped with a dihroi and-pass mirror (52/35, Brightline). Images were aquired at a resolution of pixels and a frame rate of 1.87 Hz, exept in Fig. 2 (7.5 Hz), using the SanImage software 3. Eletrophysiology. Reordings were estalished under visual ontrol using a 34 water immersion ojetive (LumplanF, Olympus), a Zeiss mirosope (Axioteh vario 1, Zeiss), and illumination (1 W fluoresene lamp, hot mirror, neutral density filter OD.3; all from Zeiss). To enhane tissue ontrast, we used two polarization filters, one loated as an exitation filter and the other as an emission filter, with slight deviation on their polarization plane. For eye protetion, we additionally used a 42-nm LP filter on the light path. Behavioural analysis. The loomotion reorder was ustom-designed aording to ref. 18. Briefly, it onsists of an air-suspended sphere floating in a owl-shaped sphere holder. A high-power infrared LED (8 nm, JET series, 9 mw, Roithner Eletronis) is loated in the ak to illuminate the fly and the sphere surfae. Two optial traking sensors are equipped with lens and aperture systems to fous on the sphere ehind the fly. The traking data are proessed at 4 khz internally, read out via a USB interfae and proessed y a omputer at <2 Hz. This allows realtime alulation of the instantaneous rotation axis of the sphere. A third amera (GRAS-2S4M-C, Point Grey Researh) is loated in the ak whih is essential for proper positioning of the fly and allows real-time oservation and video reording of the fly during experiments. Visual stimulation. For alium imaging and eletrophysiologial experiments, we used a ustom-uilt LED arena that allowed refresh rates of up to 55 Hz and 16 intensity levels. It overed 18u (1.5u resolution) and 9u (1.5u resolution) of the visual field along the horizontal and the vertial axis, respetively. The LED arena was engineered and modified ased upon ref. 34. The LED array onsists of individual TA8-81GWA dot-matrix displays (Kingright), eah harouring individual green (568 nm) LEDs. Eah dot-matrix display is ontrolled y an ATmega168 miroontroller (Atmel) omined with a ULN284 line driver (Toshia Ameria) ating as a urrent sink. All panels are in turn ontrolled via an I2C interfae y an ATmega128 (Atmel)-ased main ontroller oard, whih reads in pattern information from a ompat flash (CF) memory ard. Matla was used for programming and generation of the patterns as well as for sending the serial ommand sequenes via RS-232 to the main ontroller oard. The luminane range of the stimuli was.5 33 d m 22. For the alium imaging experiments, two separate arriers were used to shield the photomultipliers from the stimulus light oming from the LED arena. The first was a spetral filter with transpareny to wavelengths.54 nm plaed diretly over the LEDs (ASF SFG 1, Mirohemials). The seond was a layer of lak PVC extending from the fly holder over the arena. Square wave gratings had a spatial wavelength of 3u of visual angle and a ontrast of 88%. Unless otherwise stated, they were moving at 3u s 21. Edges had the same ontrast and were also moving at 3u s 21. For the experiments shown in Figs 1, 2 and 3, eah grating or edge motion was shown twie within a single sweep, resulting in a total of eight stimulation periods. Eah stimulus period lasted 4 s, and susequent stimuli were preeded y a 3-s pause. In the experiment shown in Fig. 2a, a dark edge of 88% ontrast was moved for 1 s at 15u s 21 from the front to the ak at three different positions (22u,44u,66u, from frontal to lateral). At eah position, edge motion was repeated 15 times. For the experiment shown in Fig. 2, a right edge of 88% ontrast was moving at 15u s 21 from the ak to the front, and images were aquired at a frame rate of 7.5 Hz. For the experiments shown in Figs 3e, f, all six stimulus veloities were presented one within one sweep, with the stimulus lasting 4 s, and different stimuli eing separated y 2 s. In the experiments shown in Figs 3g, h, a single sweep ontained all 12 grating orientations with the same stimulus and pause length as aove. For the eletrophysiology experiments (Fig. 4a f), multiple edges were used as stimuli moving simultaneously at 5u s 21. To stimulate ells of horizontal system (HS ells), a vertial, stationary square-wave grating with 45u spatial wavelength was presented. For ON-edge motion, the right (preferred diretion, PD) or the left edge (null diretion, ND) of eah light ar started moving until it merged with the neighouring ar. For OFF-edge motion, the right or the left edge of eah dark ar was moving. To stimulate ells of the vertial system (VS ells), the pattern was rotated y 9u lokwise. For the ehavioural experiments (Fig. 4g i), three 12-Hz LCD sreens (Samsung 2233 RZ) were vertially arranged to form a U-shaped visual arena (w 5 31 m 3 d 5 31 m 3 h 5 47 m) with the fly in the entre. The luminane ranged from to 131 d m 22 and overed large parts of the flies visual field (horizontal, 6135u; vertial, 657u; resolution,,.1u). The three LCD sreens were ontrolled via NVIDIA 3D Vision Surround Tehnology on Windows 7 64-it allowing a synhronized update of the sreens at 12 frames per seond. Visual stimuli were reated using Panda3D, an opensoure gaming engine, and Python 2.7, whih simultaneously ontrolled the frame rendering in Panda3D, read out the traking data and temperature and streamed data to the hard disk. The alaned motion stimulus onsisted of a square-wave grating with 45u spatial wavelength and a ontrast of 63%. Upon stimulation onset, dark and right edges moved into opposite diretions at 1u s 21 for 2.25 s. This stimulation was performed for oth possile edge diretions and two initial grating positions shifted y half a wavelength, yielding a total of four stimulus onditions. Data evaluation. Data were evaluated off-line using ustom-written software (Matla and IDL). For the images shown in Figs 1e, f, 2a and 3a,, the raw image series was onverted into four images representing the relative fluoresene hange during eah diretion of grating motion: (DF/F) stim 5 (F stim 2 F ref )/F ref. The image representing the stimulus fluoresene (F stim ) was otained y averaging all images during stimulation; the image representing the referene fluoresene (F ref ) was otained y averaging three images efore stimulation. Both images were smoothed using a Gaussian filter of 1 pixel half-width. For the images shown in Figs 1f and 3a,, DF/F images were normalized y their maximum value. Then, a partiular olour was assigned to eah pixel aording to the stimulus diretion during whih it reahed maximum value, provided it passed a threshold of 25%. Otherwise, it was assigned to akground. The response strength of eah pixel was oded as the saturation of that partiular olour. For the data shown in Figs 2, and 3 h, the raw image series was first onverted into a DF/F series y using the first three images as referene. Then, a region was defined within a raw image, and average DF/F values were determined within that region for eah image, resulting in a DF/F signal over time. Responses were defined as the maximum DF/F value reahed during eah stimulus presentation minus the average DF/F value during the two images preeding the stimulus. For the ar graphs shown in Fig. 4, f, the average voltage responses during edge motion (.45 s) along the ell s preferred (PD) and null diretion (ND) were alulated. For eah reorded tangential ell, the differene etween the PD and the ND response was determined, and these values were averaged aross all reorded ells. The data shown in Fig. 4g, h were otained from the four stimulus onditions y averaging the turning responses for the two starting positions of the grating and alulating the mean differene etween the turning responses for the two edge diretions. For the ar graph shown in Fig. 4i, the average turning response of eah fly during the last seond of alaned motion stimulation was alulated. These values were averaged aross all reorded flies within eah genotype. 213 Mamillan Pulishers Limited. All rights reserved

7 RESEARCH LETTER 31. Jenett, A. etal. A Gal4-driver line resoure for Drosophila neuroiology. CellRep. 2, (212). 32. Reiff, D. F., Plett, J., Mank, M., Griesek, O. & Borst, A. Visualizing retinotopi halfwave retified input to the motion detetion iruitry of Drosophila. Nature Neurosi. 13, (21). 33. Euler, T. et al. Eyeup sope optial reording of light stimulus-evoked fluoresene signals in the retina. Pfluger Arh. 457, (29). 34. Reiser, M. B. & Dikinson, M. H. A modular display system for inset ehavioral neurosiene. J. Neurosi. Methods 167, (28). 213 Mamillan Pulishers Limited. All rights reserved