Research Article

Topological and modality-specific representation of somatosensory information in the fly brain

See allHide authors and affiliations

Science  03 Nov 2017:
Vol. 358, Issue 6363, pp. 615-623
DOI: 10.1126/science.aan4428

A somatosensory map in the fly brain

The organization of sense organs and sensory brain centers shows conserved principles for many sensory modalities, even between insects and mammals. Tsubouchi et al. systematically mapped the somatosensory circuits in fruit flies. The findings revealed topological and modality-specific mechanosensory representations in the insect ventral nerve cord and brain. The authors dissected preferential responses to wing and leg movement and contributions to the control of upwind behavior that occur only when the flies are on the ground.

Science, this issue p. 615


Insects and mammals share similarities of neural organization underlying the perception of odors, taste, vision, sound, and gravity. We observed that insect somatosensation also corresponds to that of mammals. In Drosophila, the projections of all the somatosensory neuron types to the insect’s equivalent of the spinal cord segregated into modality-specific layers comparable to those in mammals. Some sensory neurons innervate the ventral brain directly to form modality-specific and topological somatosensory maps. Ascending interneurons with dendrites in matching layers of the nerve cord send axons that converge to respective brain regions. Pathways arising from leg somatosensory neurons encode distinct qualities of leg movement information and play different roles in ground detection. Establishment of the ground pattern and genetic tools for neuronal manipulation should provide the basis for elucidating the mechanisms underlying somatosensation.

Responding to external stimuli is a crucial function, even for the simplest unicellular eukaryotes, and these organisms share specific molecules that are used for sensory detection in higher animals (1). In metazoans, the organization of sense organs and associated sensory centers in the brain often shows conserved organizational principles for many sensory modalities, even between the species belonging to distant evolutionary clades such as insects and mammals (25). Is there such correspondence in the somatosensory system? Mammalian and insect somatosensory systems initially appear rather different. In mammals, somatosensory neurons derive from the neural crest, and they reside deep in the body in the dorsal root ganglia. They send their axons to specific laminae in the dorsal spinal cord, from where information is sent to different parts of the thalamus, depending on somatosensory modality (6, 7). In insects, somatosensory neurons derive from the epidermis and reside directly beneath the exoskeletal body surface (Fig. 1A) (8). Terminals of different types of insect somatosensory neurons tend to form a layered organization in the ventral nerve cord (VNC), the ganglia that are equivalent to the vertebrate’s dorsal spinal cord (9, 10). However, despite extensive studies on the structure and functional properties of primary neurons and associated local secondary neurons in the VNC (1114), knowledge about their patterning and connections to the higher brain have remained limited (15, 16). Here we provide the first systematic mapping of insect somatosensory neural circuitry using Drosophila melanogaster as a model.

Fig. 1 Layered somatosensory axon terminals in the ventral nerve cord.

(A) Schematics of mechanosensory organs. Cross-sectional view. The planar view is also shown for the rightmost panel. Red and blue indicate neurons and support cells, respectively. Scale bars, 10 μm. (B to F) Schematics of the entire adult fly (B), somatosensory neurons in the leg (C), wing (D), haltere (E), and abdominal body wall of the fourth and fifth segments (F). See figs. S1 and S2 for detail. (G) Entire neuropil of the ventral nerve cord (VNC). Three-dimensional (3D) reconstruction image of the nc82 antibody labeling. Dotted rectangles indicate the areas visualized in the following panels. (H1 to H6) 3D reconstruction of the layered axon terminals in the leg neuropil. Fluorescent signals were segmented using Amira software. The anterior oblique view is shown. Lgs1N [(H1), magenta], Les1N [(H2), green], Lsr1N [(H3), pink], Lco1N [(H4), yellow], Lcs1N [(H5), blue], and merged image (H6). See fig. S3, A to G, and movie S1 for detail. (I1 to I5) 3D reconstruction of the layered axon terminals in the wing neuropil. Tes1N [(I1), green], Wgs1N [(I2), purple], Wes1N [(I3), yellow], Wcs1N [(I4), blue], and merged image (I5). See fig. S3, J to M, and movie S2 for detail. (J1 to J3) 3D reconstruction of the layered axon terminals in the abdominal ganglia. Amd1N [(J1), magenta], Aes1N [(J2), green], and merged image (J3). See fig. S3O for detail. (H to J) Directions with arrows: D (dorsal), V (ventral), L (lateral), M (medial), and A (anterior). Scale bars, 50 μm. (K) Schematics of the arrangement of presynaptic terminals for each type of primary somatosensory neurons in the frontal cross section.

Modality-specific layered organization of the somatosensory axon terminals in the VNC

To understand the global architecture of the somatosensory pathways in the Drosophila central nervous system (CNS), we first analyzed the organization of somatosensory axon terminals in the VNC. There are a number of GAL4 expression driver strains to visualize somatosensory neurons (8, 14). We screened more than 10,000 promoter-fusion and enhancer-trap GAL4/LexA driver lines (1721) and identified an array of 22 strains. Each strain labels specific types of somatosensory neurons in the legs, wings, halteres, and abdominal body surface (figs. S1 and S2). We selected a set of strains so that they together label essentially all of the somatosensory cells. Using these specific driver lines, we then mapped the distribution of axon terminals in the respective neuropils of the VNC (Fig. 1 and fig. S3).

Drosophila has six major types of somatosensory neurons (Fig. 1A). Gustatory sensilla (gs) neurons detect food stimuli through the pores of external taste hairs (22, 23). External sensilla (es) neurons detect movement of the bristles caused by contact, wind, etc. (24). Stretch receptor (sr) neurons contact neighboring leg joints to detect stretch and tension between them (25). The chordotonal organs (co) have thin elongated structures that span across different parts of the exoskeleton to detect tension or vibration (26, 27). Campaniform sensilla (cs) feature dome-shaped thin cuticles to detect deformation of the exoskeleton (28, 29). Finally, multidendritic (md) neurons extend complex dendrites spanning under the epidermis to detect diverse stimuli such as touch, pain, heat, and coldness (30).

Among these different types of somatosensory neurons, the legs feature all but the last neuron type (i.e., gs, es, sr, co, and cs) (Fig. 1, B and C, and fig. S1) (8, 24). Most of their axons terminate in the bulbous leg neuropils of the VNC (Fig. 1G). Previous studies suggested a layered organization of some neuron types (e.g., es lying ventrally and cs and co dorsally) (10, 13, 24); we found that axon terminals of all five types of leg somatosensory neurons have clearly segregated layered projections (Fig. 1, H and K, and movie S1). The leg gs neurons are distributed in the distalmost leg segments (fig. S1, A and B). They express 27 gustatory receptor genes and were categorized into seven types according to their projection patterns (23). One of the sugar-detecting neuron types (VT, expressing Gr5a) arborizes only in the thoracic VNC, which we here call Lgs1NT (where “L” stands for the leg, “1N” for primary neuron, and “T” for thorax). All other gs neurons project both to the VNC and the brain, which we collectively call Lgs1NB (“B” for brain). In the VNC, axons of both Lgs1NT and Lgs1NB terminate in the most ventral part of the leg neuropil (Fig. 1H1 and fig. S3, A and B).

The leg es neurons (Les1N) are studded along the entire leg surface (fig. S1C) (24). Their axon terminals in the VNC lie slightly more dorsally than that of the leg gs neurons, in the shape of a flat disc that horizontally spreads across the ventral leg neuropil (Fig. 1H2 and fig. S3C). The disc features concentric fine-scale somatotopy. We surgically cut off the legs at different joints and observed the projection patterns 2 weeks later after axons arising from the excised parts are degenerated. Axons of the es neurons in the distal leg segments terminate in the central parts of the disc, whereas those deriving from the more proximal segments terminate in more peripheral rings (fig. S3H).

Leg sr neurons (Lsr1N) are scattered in several leg joints (fig. S1D) (25); in the VNC they arborize at the level just above the disc of the es neuron terminals (Fig. 1H3 and fig. S3D). Most of the leg co neurons express the inactive gene (Lco1N1) (fig. S1E) (31), whereas another line labels a different small subset (Lco1N2) (fig. S1F). Lcs1N labels a subset of leg cs neurons (fig. S1G). The axons of co and cs neurons terminate in the most dorsal layer, just in the middle of the leg neuropil (Fig. 1, H4 and H5, and fig. S3, E to G). Their projections are segregated within a layer: The axon terminals of the co neurons form several thick bundles (fig. S3, F2 and G2), and cs neurons terminate in the space between these bundles (fig. S3E2). Cross-sectional views of the registered images show that the presynaptic sites of different neuron types hardly overlap (fig. S3I). Thus, the leg neuropils have a modality-specific layered organization in the order of gs, es, sr, co, and cs axon terminals from ventral to dorsal (Fig. 1, H6 and K, and movie S1).

Somatosensory neurons of the thoracic body surface (es) and wings (gs, es, and cs) send their axons to the wing neuropil (3234) (Fig. 1, D and G). The thoracic es neurons scattered on the body surface (Tes1N) (32) project broadly to the most ventral layer of the wing neuropil (Fig. 1I1 and fig. S3J). The wing gs neurons (Wgs1N) and wing es neurons (Wes1N) (33, 35) are located along the anterior wing margin (fig. S2, A and B). They terminate in the wing neuropil in the form of thin bundles: Axons of Wgs1N (Fig. 1I2 and fig. S3K) terminate just above the terminals of Tes1N (Fig. 1I1 and fig. S3J), whereas Wes1N terminate slightly more dorsally (Fig. 1I3 and fig. S3L). Finally, wing cs neurons (Wcs1N) (34) are located on the dorsal and ventral surfaces of the wing base, as well as on the L3 wing vein (fig. S2, C and E). Their axon terminals are located further dorsally in the middle and dorsolateral parts of the wing neuropil (Fig. 1I4 and fig. S3M). Thus, the wing neuropil has four modality-specific layers in the order of thoracic es and wing gs, es, and cs (Fig. 1, I5 and K, and movie S2).

Fig. 5 Schematic comparison of somatosensory pathways in flies and mammals.

(A) Pathways in Drosophila identified in this study. (B) Known pathways in mammals. (C) Possible correspondence between different types of somatosensory neurons, showing inverse dorsoventral order. Note that some fish and amphibians also have putative gustatory neurons on the body, but their projection targets are not well known.

Somatosensory neurons in the haltere terminate in the haltere neuropil (Fig. 1, E and G). Unlike wings, halteres have no gs neurons and only a few es neurons. Haltere cs neurons (Hcs1N) (34) are located on the dorsal and ventral sides of its basal part (fig. S2, D and F); their axons terminate in a comparable dorsal level (fig. S3N) as that of the wing cs neurons in the wing neuropil (fig. S3M).

The abdominal body surface features three types of mechanosensory neurons (md, es, and co) (Fig. 1F). The abdominal md neurons (Amd1N) extend dendrites under the dorsal and lateral body walls (fig. S2G). Abdominal es neurons (Aes1N) (36) are located on the dorsal and ventralmost abdomen (fig. S2H), and a few co neurons (Aco1N) (37) lie along the ventral midline (fig. S2I). Among them, the abdominal co neurons do not arborize in the abdominal ganglia; instead, they project directly to the prothoracic ganglion to converge with the terminals of the foreleg co neurons. The md and es neurons terminate in the abdominal ganglia to form segregated layers; axons of the md neurons project to the most ventral layer, whereas es neurons arborize in a more dorsal layer (Fig. 1J and fig. S3O). Thus, the abdominal VNC neuropils again have segregated layers that consist of md and es axon terminals, respectively (Fig. 1, J and K).

Our analysis revealed that although the repertoires of somatosensory neuron types that terminate in the leg, wing, haltere, and abdominal VNC neuropils are different, all neuropils have modality-specific layered organization of somatosensory axon terminals.

Modality-specific and somatotopic organization of somatosensory axon terminals in the brain

We next investigated what parts of the brain receive somatosensory information, bearing in mind two likely arrangements: direct pathways by the axons of the peripheral somatosensory neurons and indirect pathways by the secondary interneurons ascending from the VNC. We first analyzed the former (Fig. 2A).

Fig. 2 Projection of primary and secondary neurons in the brain.

(A and B) Schematics of the projection patterns of the identified mechanosensory neurons (A) and ascending interneurons (B), dorsal view. Cell bodies and arborizations are schematized with open and painted circles, respectively. Brain regions that contain terminal arborizations are painted in pale colors. (C) Anterior view of the entire neuropil of the brain. 3D reconstruction images of the nc82 antibody labeling. Dotted rectangles indicate the areas visualized in the following panels. (D to G) Neuronal fibers (magenta) and presynaptic terminals (green-white) of the primary somatosensory neurons, anterior view. See fig. S4, A to D, for lateral view. Axons deriving from Wcs1N (Wl and Wm in yellow arrows) and Hcs1N (Hl and Hm in white arrows) (D), Lgs1NB (E), Lco1N1B (F), and Lco1N2B (G) are shown. Compared with the axons of Lco1N1B, those of Lco1N2B in the AVLP neuropil are longer and have branched terminals at their distal ends (blue arrow). (H to M) Neuronal fibers (magenta) and presynaptic terminals (green-white) of the secondary ascending interneurons in the brain, anterior view. See fig. S5, G to O, for lateral view. Wcs2N1 (H), Hcs2N1 (I), Lco2N1 (J), Lco2N2 (K), Les2N1 (L), and Les2N2 (M). (N to P) Overlay of the axons of primary and secondary neurons. (N) Wcs1N (green), Hcs1N (blue), Wcs2N1 (magenta) and Hcs2N1 (yellow). (O) Lco1N1B (yellow), Lco1N2B (blue), Lco2N1 (green), and Lco2N2 (magenta). (P) Lgs1NB (blue), Les2N1 (yellow), and Les2N2 (magenta). Regions of the brain (GNG, SLP, PVLP, AVLP, and WED) are indicated with white doted lines. (See movie S3 for detail.) Scale bars in (C) to (P), 50 μm. (Q) Schematic somatosensory pathways. Note that although neurons appear to terminate in diverse brain regions, their terminals mostly occupy a small volume of the brain that span across neighboring regions.

Most of the wing and haltere cs neurons form two axon branches that terminate in the VNC and project to the brain, respectively. The latter form a single common bundle, and upon entering the brain they spread into four sub-bundles that fan out in the posterior gnathal ganglia [GNG (Fig. 2D and fig. S4A); see definitions by the Insect Brain Name Consortium (38)]. The lateral two sub-bundles extend dorsolaterally to arborize in the region of the GNG just posterior to zones C and E of the antennal mechanosensory and motor center, which receives information about antennal deflection (5, 39), whereas the medial two sub-bundles terminate more proximally in the posterior medio-ventral GNG.

Because the axons arising from wing and haltere cs neurons project via a common bundle from the VNC, their projection targets are not immediately apparent. To analyze topological organization of their terminals, we surgically excised the wing or haltere of one side and observed the projection patterns after axonal degeneration. Only the lateral two sub-bundles remained when a haltere was cut off (fig. S4E); they are therefore specific to the wing cs neurons (named Wl and Wm for wing-lateral and wing-medial sub-bundles, respectively). Conversely, the two medial sub-bundles remained after wing excision (fig. S4F), showing their origin from the haltere cs neurons (named Hl and Hm sub-bundles). Thus, wing and haltere cs neurons form a distinct somatotopic sensory map in the form of parallel sub-bundles in the posterior GNG.

Whereas a type of sweetness-responsive gs neurons (Lgs1NT) terminate within the VNC, all other gs neurons detecting sweetness, bitterness, contact pheromones, etc. (collectively called Lgs1NB) project their axons directly to the middle part of the GNG (23) (Fig. 2E and fig. S4B). They terminate in the brain region just posterior to the terminals of the bitterness-responsive gustatory sensory neurons of the proboscis (20).

Most of the leg co neurons terminate within the VNC, but some project toward the brain. We distinguish those subgroups as T and B for the thorax and brain, respectively. The axons of both Lco1N1B and Lco1N2B subpopulations form a single bundle that runs anterior-dorsally in the brain along the lateral side of the GNG (Fig. 2, F and G, and fig. S4, C and D) and wedge (WED) neuropils to terminate in the ventralmost part of the anterior ventrolateral protocerebrum (AVLP). Do co neurons arising from different legs make a somatotopic map in this axon bundle? To address this question, we surgically cut off two of the three legs on one side of the animal and observed projections after axonal degeneration. The remaining axons terminate at different positions along the trajectory; axons from the foreleg and midleg project toward the distal end and a slightly proximal part of the bundle, respectively (fig. S4, G and H), whereas those from the hindleg terminate much more proximally (fig. S4I). Thus, the leg co neurons also make a clear somatotopic map in the form of longitudinal projection (fig. S4J).

Modality-specific convergence of primary neurons and secondary interneurons in the brain

Because many somatosensory neurons terminate within the VNC, information should also be conveyed to the brain via secondary ascending interneurons. If we can identify interneurons whose dendritic arbors arise in a specific somatosensory layer of the VNC neuropils, they would be likely candidates to convey modality-specific information to the brain. We screened GAL4 driver lines and identified six types of such interneurons (Fig. 2, B and H to M; fig. S5, G to L; and movie S3).

One type of wing cs interneurons (Wcs2N1, where “2N” stands for putative secondary interneuron) and one type of haltere cs interneurons (Hcs2N1) have postsynaptic sites in matching layers of the primary cs neuron axon terminals of the wing and haltere in the respective neuropils (fig. S5, A and B). In the brain, they terminate in the distal and proximal regions of the posterior GNG (Fig. 2, H and I, and fig. S5, G and H), where they converge with that of the primary neurons Wcs1N1 and Hcs1N1, respectively (Fig. 2N and fig. S5M).

A comparable trajectory was also observed in the leg co neurons. The postsynaptic sites of the two types of putative leg co interneurons (Lco2N1 and Lco2N2) overlap with the terminal branches of the primary neurons Lco1N1 in the leg neuropil (fig. S5, C and D). In the brain, Lco2N1 axons project via the same trajectory as that of the primary co neurons (Fig. 2J and fig. S5I), whereas Lco2N2 axons run more medially through the middle of the GNG (Fig. 2K and fig. S5J). Nevertheless, they eventually converge in the same region of the ventralmost AVLP neuropil, which overlaps with the distalmost projection of the primary co neurons Lco1N1B and Lco1N2B (Fig. 2O and fig. S5N).

Because none of the es neurons project directly to the brain, their information should be transmitted via interneurons. Do they supply the same brain region as the mechanosensory neurons of other modalities? We found two types of interneurons (Les2N1 and Les2N2) that have postsynaptic sites in the ventralmost leg neuropil layers that receive terminals of Les1N1 as well as Lgs1N1 (fig. S5, E and F). Les2N1 and Les 2N2 have three [dorsal (D), middle (M), and ventral (V)] and two (D and V) axon projections in the brain, respectively (Fig. 2B), some of which are reminiscent of the previously reported taste-associated interneurons TPN1 and TPN 2 (40) (fig. S5, E and F). The ventral projections (Les2N1V and Les2N2V) terminate in the middle part of the GNG that is adjacent to the terminals of the leg gs neurons. Other projections pass through the GNG to project further dorsally: Les2N1D and Les2N1M terminate in the dorsal and middle parts of the posterior ventrolateral protocerebrum (PVLP), respectively, whereas Les2N2D terminates in the ventral part of the superior lateral protocerebrum (SLP) (Fig. 2, B, L, M, and P, and fig. S5, K, L, and O). The targets of those indirect pathways arising from the leg es and gs neurons do not overlap with, but lie slightly above and medial to, the direct and indirect targets of the leg co neurons.

These results suggest a rather simple architectural concept of the putative somatosensory centers in the fly CNS (Fig. 2Q). First, axons arising from different types of primary somatosensory neurons terminate in different layers of the VNC and different subregions of the brain. Second, direct and indirect pathways of the same sensory modality tend to converge into overlapping subregions of the brain. These characteristics infer that specific subregions of the brain receive distinct types of somatosensory information, forming somatotopic and modality-specific sensory representation.

Difference of mechanosensory information conveyed to the brain by different types of primary and secondary neurons

So far we found that different somatosensory neurons of the same modality send overlapping information to certain subregions of the brain. To reveal differences of their functional properties, we next examined the activity of the primary and secondary neurons using cell-specific calcium imaging. We placed the heads of the flies beneath a thin metal plate, so that the tethered flies could move their wings and legs freely on a trackball, and observed activity of the axon terminals of the neurons expressing GCaMP6s (41) using a two-photon microscope (Fig. 3A and movie S4).

Fig. 3 Calcium imaging of specific primary sensory neurons and secondary interneurons.

(A) Schematics of imaging setup (left), photograph of a fly with a tungsten rod stimulation (yellow arrow) to the wing (middle), and schematics showing areas of recording in the ventral brain (right). Neuronal activity was measured at the terminal parts of axons. See movie S4 and figs. S6 to S8 for examples of actual imaging. (B and C) Responses of wing cs neurons (B1 to B3) and haltere cs neurons (C1 to C3). Dotted circles indicate the areas that show activity differences. + and – symbols indicate the relative intensity of the interested signals. Scale bars, 10 μm. (D) Responses of wing and haltere cs neurons during artificial deflection of wing or haltere and during tethered flight (n = 3 to 4 animals, 3 to 8 data sets in total). Box plot of the average intensity for 2 s in the region of interest [shown in (B) and (C)], measured from 3 s before and after the onset of wing or haltere deflection and flight (where F is the average intensity before the onset of deflection or flight). **P < 0.01. The data represent responses of Wm and Hm sub-bundles; Wl and Hl sub-bundles also showed similar responses. (E to H) Responses of co neurons from single legs. Other legs were excised shortly before assay. Activity in the area of AVLP and WED neuropils (for each leg) and GNG (for hindleg) was measured. Dotted circles indicate the areas that show activity differences. + and – symbols indicate the relative intensity of responses. Scale bars, 10 μm. (I) Responses of leg co neurons deriving from a single leg (n = 3 animals, 9 to 19 data sets). Intensity for 2 s just before the flies start another behavior was averaged (where F is the value for the longest “stop” period during an imaging session). **P < 0.01; dash indicates not significant. (J to M) Responses of primary and secondary neurons during different classes of leg movement behavior (n = 3 to 9 animals/3 to 83 data sets). Observed behaviors: walk, groom (groom legs and body parts), stop on the trackball, touch (touch other objects briefly while moving legs in the air), flail (move legs in the air without contact to other objects), and stay (stay silent in the air). Intensity for a 2-s period just before the flies start another behavior was averaged (where F is the value for the longest “stop” period during an imaging session). **P < 0.01; *P < 0.05; dash indicates not significant. P values were determined by Wilcoxon-Mann–Whitney U-tests for pairwise comparisons and Kruskal-Wallis tests for multiple comparisons.

Because cs neurons respond to the deformation of the cuticular epidermis (29), we measured the activity of Wcs1N and Hcs1N in the brain when we mechanically moved the wing or haltere with a tungsten rod. Only the lateral (Wl and Wm) or medial (Hl and Hm) axon sub-bundles showed activity when we specifically stimulated the wing or haltere, respectively (Fig. 3, B to D, and fig. S6, A and B). Thus, the somatosensory representation we anatomically revealed earlier (fig. S4, E and F) is confirmed physiologically as the selective activity patterns of axon sub-bundles. Consistent with the reported response property of the leg cs (14), activity persisted during the period of deflection (fig. S6, A and B). Neurons also showed strong responses upon large and rapid deflections during flight behavior (Fig. 3, B to D, and fig. S6C).

To analyze the physiological somatosensory representation of the leg co neurons, we surgically removed two of the three legs on one side of the body and recorded the axons of the remaining leg co neurons (Fig. 3, E to I, and fig. S7A). Flies with only forelegs and midlegs showed specific responses during walking behavior at the distal and slightly proximal portion of the axon bundle in the AVLP and WED regions, respectively (Fig. 3, E, F, and I), whereas those with only hindlegs showed responses not in the AVLP and WED regions but only in the proximal part of the axon bundle in the posterior GNG (Fig. 3, G, H, and I). This finding nicely correlates with the anatomy of axon projections from the co neurons of each leg that we examined (fig. S4, G to J).

Primary neurons and secondary interneurons of the leg co converge in the ventral AVLP. Do different neurons convey different aspects of somatosensory signals to the brain? We compared video-recorded leg movement of the flies with the responses of leg co neurons. We classified the flies’ leg movement into seven classes depending on whether the legs move or not and how they contact with other objects. When the flies are on the trackball, they either walk, move their legs to groom the body, or stop. When we lower the trackball and keep the flies in the air, they either groom their legs or body, move their legs with brief contact to other objects such as the metal plate of the imaging setup (touch), move their legs freely in the air (flail), or stay without moving (stay) (movie S4).

Both types of primary co neurons respond more strongly when the flies move their legs than when they do not (walk and/or groom versus stop on the trackball; groom, touch, and/or flail versus stay in the air) (Fig. 3, J and K, and fig. S7, B and C). They show weakest activity when the legs have no movement and ground contact (stay in the air). Response levels in other behavioral classes were more varied in Lco1N1B than in Lco1N2B.

Secondary interneurons whose dendrites overlap with the axon terminals of the leg somatosensory neurons responded to leg movement, suggesting the sensory input. Whereas the primary co neurons showed stronger responses when the legs move, a coassociated secondary interneuron type Lco2N1 showed stronger responses when the legs contact objects continuously. Leg movement with ground contact (walk, groom, and/or stop) or persistent contact with other objects (groom in the air) showed higher activity level than during the leg movement with brief or no contact (touch, flail, and/or stay) (Fig. 3L, and fig. S8, A and B). Likewise, the response when the flies stop on the ball (stop) was stronger than when they stay in the air (stay). Secondary interneurons Les2N1D, which are associated with the leg es and gs, showed the inverse pattern: Responses are weaker when the legs contact objects continuously (walk, groom, stop on the ball, and groom in the air) than during leg movement with brief contact (touch) (Fig. 3M and fig. S8, C and D).

Importance of contact-responsive sensory neurons and interneurons for the walking response against wind

Our physiological assay revealed varying patterns of leg neuron responses depending on cell types, leg movement, and ground contact. To further understand the roles of those neurons in behavior control, we disturbed their functions to analyze behavioral effects. Flies show a rapid arrest of walking when they encounter strong airflow (Fig. 4A), called wind-induced suppression of locomotion (WISL), and a specific subset of Johnston’s organ neurons are responsible for wind detection (39). However, the same speed of airflow stimulus does not cause arrest when the flies are in the air. Thus, to induce WISL the flies should integrate wind stimuli with signals from other body parts to determine whether they are on the ground or in the air.

Fig. 4 Behavioral analysis of the responses against wind.

(A) Schematics of the WISL assay. A group of ~25 flies was placed in a thin transparent chamber at one time, illuminated from below and video recorded for 240 s from above. Airflow (1.35 m/s) was provided from one side of the chamber for 120 s. Movement of the flies was tracked to calculate walking velocity between neighboring video frames (1/30 s). Twelve trials (total: 300 flies) were repeated, and movement of 68 ± 17 flies (mean ± SD) was tracked successfully for each video frame. See movie S5 for examples of actual video recording. (B to J) WISL analysis. Thin vertical lines represent mean ± SD of velocity for each video frame. (B) Genetic control (carrying UAS-Kir2.1 but no GAL4 to activate it) shows robust WISL. Black, mean; gray, SD. (C to J) Lines represent velocity of genetic control flies (GAL4 only, blue, mean; light blue, SD) and flies with silenced neuronal activity (GAL4 + Kir2.1, red, mean; pink, SD). We were not able to measure the specific contribution of Lco1N1, because all of the driver lines we found for this neuron type induce expression also in the wind-detecting Johnston’s organ neurons. (K) WISL index. The ratio between the mean velocity for all frames of wind-on and wind-off states was calculated for each trial, and mean ± SEM of the 12 trials was compared. **P < 0.01. P values were determined by ANOVA (analysis of variance) with post-hoc Tukey HSD tests.

To identify the neurons involved in ground detection, we induced expression of an inward rectifier potassium channel, Kir2.1 (42), to suppress the functions of specific neurons in the legs (movie S5). Control flies that carry only the Kir2.1 transgene or only the GAL4 construct showed normal WISL (Fig. 4B and blue lines in Fig. 4, C to J). The strain we used for labeling Lco1N1 also activates expression in the wind-detecting co neurons of the antennal Johnston’s organ. Accordingly, suppression of these cells severely disturbed WISL (Fig. 4C). When we suppressed the leg es neurons, the basal walking activity was decreased even without wind (Fig. 4D), indicating that the function of the leg es neurons is crucial for normal walking behavior. Nevertheless, the WISL index, which compares the decrease of average walking velocity between wind-off and wind-on states, showed a considerable decrease when the leg es neurons are blocked (Fig. 4K), indicating that the flies with no signals from these cells cannot respond properly against wind stimuli. WISL was also disturbed when we silenced leg sr neurons and distinct subsets of leg co neurons (Fig. 4, E, F, and K), suggesting their roles for ground detection. On the contrary, silencing leg gs and cs neurons did not disturb WISL (Fig. 4, G, H, and K).

Thus, only es, sr, and co neurons are important for WISL that occurs only on the ground. The es neurons detect contacts between legs and ground, and the sr neurons are known to be important for proprioception and load sensing (25). The role of Lco1N2 for ground detection matches well with its physiological property, which showed stronger responses when flies stop on the ground versus in the air (Fig. 3K). Suppression of es- and gs-associated and co-associated interneurons Les2N1 and Lco2N1, respectively, also disturbed WISL (Fig. 4, I, J, and K). This is reasonable considering that these interneurons showed different activity levels, depending on ground contact (Fig. 3, L and M).

Suppression of any of the three leg primary neuron types (es, sr, and co) and certain interneurons alone was sufficient to disturb WISL, but none of them induced complete suppression. These suggest that integration of the information from all those neuron types would be required when the flies determine whether they are on the ground or in the air.


We performed systematic identification of the primary and secondary somatosensory neurons of Drosophila and provided a systematic overview of the somatosensory neuronal circuitry of insects from periphery to brain (Fig. 5A). A set of expression driver strains will open up a new research field for detailed functional analysis and selective manipulation of the somatosensory system. Identification of clearly layered terminals in all the sensory neuropils of the VNC should provide the basis to identify and analyze further secondary interneurons arising from each modality-specific layer.

Architecture of the Drosophila somatosensory system

Only three distinct types of sensory information are transmitted directly to the brain by primary neurons (leg gs, co, and wing and haltere cs). Such connection has also been reported in other insects (15, 16), suggesting that this might be a general feature across insecta. Whereas only a small portion of leg co neurons project directly to the brain, most wing and haltere cs neurons innervate the brain; these cs neurons are known to detect various aspects of wing-beat force during flight to provide feedback control (43). Direct projections to the brain would be important for these neurons to enable fast transmission of information about rapidly changing sensory parameters during flight.

We found that ground detection for WISL, which would require slower temporal resolution than flight control, is mediated by both direct and indirect pathways (Fig. 4). Primary neurons and secondary interneurons of the same sensory modality tend to converge in specific subregions of the brain, forming modality-specific somatosensory representation. In spite of the similar axon trajectory in the brain (Fig. 2O), these neurons convey information about leg movement in different ways (Fig. 3, J to L).

Interneurons associated with the leg co and es terminate in neighboring but different regions of the lateral brain (Fig. 2, O and P), yet some of them have shared roles in WISL control. Because their signals are transmitted to distinct parts of the brain, yet-unidentified higher-order neurons in the brain should converge those signals to the motor control circuitry.

In this respect, it is important to note that most ascending secondary interneurons we identified have presynaptic output sites, not only in the brain but also in the VNC (fig. S5, A2 to F2 and A4 to F4). Local circuitry in the leg neuropil is important for controlling leg movement (11, 14, 44). Those local neurons are likely candidates that receive output from the ascending interneurons, because axon terminals of sensory neurons hardly have postsynaptic sites. Similar local output has also been found in other sensory modalities; many olfactory and visual projection interneurons have collateral output synapses in the antennal lobe and optic lobes (45, 46).

There are three pairs of leg neuropils. Among them, the foreleg neuropil has specialized arborization of the gs neurons that exist only in the foreleg (23). Other than this, we did not find substantial differences of arborization patterns between the fore-, mid-, and hindleg neuropils.

Homology between insect and mammalian somatosensory systems

The present results provide data for a systematic comparison of the insect somatosensory system with its mammalian counterparts. Insects and mammals share similarities of neural organization underlying the perception of odors, taste, vision, sound, and gravity (25), and our data also reveal marked similarity for the mechanosensory sysem (Fig. 5, A and B). In insects, some primary neurons project directly to distinct parts of the ventral and lateral brain, whereas others terminate within the VNC. Likewise, in mammals, some neurons project directly to the ventral brain at the medulla oblongata, whereas others terminate within the spinal cord (6, 47, 48). Modality-specific pathways tend to converge in different subregions of the medulla, as well as in the thalamus of the mammalian brain (6, 49). Similarly, direct and indirect pathways tend to converge in common subregions of the insect brain, and neurons conveying information about different somatosensory modalities tend to terminate in different subregions. As in mammals, these subregions often lie adjacent to each other in certain parts of the brain; for example, the entire terminal arborizations of the leg co and es secondary interneurons are confined in a 40-μm-wide, 150-μm-tall cylindrical volume in the lateral brain (Fig. 2, O and P, and fig. S5, N and O).

Somatosensory signals are sent predominantly to the ipsilateral brain side in insects and contralateral in mammals. Considering that descending neurons tend to project ipsilaterally in insects but contralaterally in mammals (5052), however, somatosensory signals and motor control computation are processed primarily in the same side of the brain in both cases.

Layers of sensory axon terminals in the insect VNC and mammalian spinal cord (9, 10, 13, 53) are also organized in a similar order (Fig. 5C). Insect multidendritic neurons and mammalian free nerve endings share various characteristics in common: Their dendrites both have free endings without forming particular sense organs to detect pain, temperature, and other submodalities (54, 55). The md neurons project to the most ventral layer of the VNC, whereas free nerve endings innervate the most dorsal layer of the spinal cord. Axons from the insect external sensilla and mammalian hair receptors, both of which detect haptic contact to the tips of the bristles and hairs, terminate in the second-ventral and second-dorsal layers, respectively. Insect chordotonal organ and mammalian muscle spindle, as well as insect campaniform sensilla and mammalian Golgi tendon organ, also show similarity with respect to their functions in motor control (12). These receptor systems supply afferents to the most dorsal and most ventral layers in insects and mammals, respectively. A fly’s stretch receptors and mammalian Merkel cell neurites—as well as Meissner, Ruffini, and Pacinian corpuscles—terminate in the third-ventral and third-dorsal layer, respectively. Although correspondence between them is less obvious, they similarly detect deformation of the exoskeleton and skin (1, 6, 48). Thus, functionally comparable somatosensory terminals are layered in reverse order between the two systems (Fig. 5C). Considering that the dorsoventral axis of the mammalian body is developmentally upside down compared with the insect one (56, 57), the corresponding order of sensory arrangements is actually conserved exactly between the two systems.

Do corresponding somatosensory cell types express common genes? Modality-dependent molecular specialization is not apparent even within insects or mammals, because the same genes are often expressed in multiple cell types and only a few genes share expression in the corresponding cell types across taxa (table S1). This might be a rather general feature; receptor molecules as well as developmental origins of the sensory organs are not identical between insects and mammals also in olfactory and auditory systems, yet sensory centers in the brain share architectural similarities (4, 5).

With our somatosensory analysis, transphyletic correspondence of neuronal circuitry has been found in all of the sensory modalities. Corresponding organization has been suggested also for associative centers and motor systems (25, 9, 5860). The fact that essentially all important components of the brain system share conserved features across the two evolutionary clades, which have been separated since at least the end of the Ediacaran period more than 550 million years ago, would suggest that basic development programs for the orderly and secrete segregation of those circuits may have evolved before deuterostome-protostome or deuterostomia-ecdysozoa divergence.

Supplementary Materials

Materials and Methods

Figs. S1 to S8

Table S1

References (61121)

Movies S1 to S5

References and Notes

  1. Acknowledgments: We are grateful to W. A. Johnson for ppk-GAL4, K. Emoto for ppk-LexA, D. van Meyel for tutl-GAL4, C. Montell for inactive-GAL4, J. Carlson and T. Tanimura for Gr5a-GAL4 and Gr89a-GAL4, B. Pfeiffer for LexAop2-mCD8::GFP, J. Urban and G. Technau for MZ-series enhancer trap strains, the members of the Nippon GAL4 Consortium and D. Yamamoto for the NP-series strains, G. Rubin and Y. Aso for GMR-GAL4 driver strains, Bloomington Stock Centre for other transgenic strains, and the Developmental Studies Hybridoma Bank for the nc82 antibody. We thank A. Büschges, A. Kamikouchi, S. Kohatsu, S. Namiki, T. Shimogori, D. Yamada, D. Yamamoto, and M. Yoshihara for discussion and sharing technical information; M. Ito and Janelia FlyLight project for sharing neuron image data; T. Kawase for visualizing 3D data; K. Branson and J. Bender for behavior analysis; and S. Shuto, K. Yamashita, H. Hirose, Y. Ishida, and R. Tatsumi for technical assistance. 3D data of the identified neurons are available via the Virtual Fly Brain database (, and the raw movie data of calcium imaging and behavioral assay are archived on our laboratory data server ( This work was supported by grants from the Core Research for Evolutional Science and Technology (Japan Science and Technology Agency and Japan Agency for Medical Research and Development) program and from the Strategic Research Program for Brain Sciences to K.I.; a Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology of Japan to K.I. and A.T.; and a Japan Society for the Promotion of Science fellowship to T.Y. and T.K.Y.
View Abstract

Navigate This Article