Anatomical Distribution and Behavioral Contributions of 13A and 13B Hemilineages

(A) Schematic showing segmental distribution of 13A (green) and 13B (cyan) neurons across pro-, meta-, and meso-thoracic segments (T1, T2, T3) of VNC. (B) Confocal image: Six GABAergic 13A neurons (green arrowheads) and six 13B neurons (cyan arrowheads) in each VNC hemisegment, labeled with GFP (green) driven by R35G04-GAL4-DBD, GAD-GAL4-AD. Neuropil in magenta (nc82). Panel B’ provides a zoomed-in view of T1 region. (C) EM reconstructions: 62 13A neurons (green) and 64 13B neurons (cyan) in right T1. Ventral side up. (D) Continuous activation of 13A and 13B neurons labeled by R35G04-GAL4-DBD, GAD-GAL4-AD in dusted flies reduces front leg rubbing and head sweeps, and induces unusual leg extensions. Control: AD-DBD EMPTY SPLIT >UAS CsChrimson (gray). Experiment: R35G04-GAL4-DBD, GAD-GAL4-AD > UAS CsChrimson (red). Box plots indicate the percentage of time dusted fly engaged in a given behavior over a 4-minute assay (n = 7). The solid blue line marks the mean, dark shading the 95% confidence interval, red dashed line the median, and light shading ±1 standard deviation. ***P ≤ 0.001, *P ≤ 0.05. (E-F) Continuous activation of 13A and 13B neuron subsets induces front leg extension in headless flies. (E, E′) Representative video frames showing headless flies (dusted and undusted) with extended front legs (orange arrowhead) following continuous optogenetic activation of neurons labeled with R35G04-GAL4-DBD, GAD-GAL4-AD > UAS-CsChrimson. Dashed box in E highlights the front legs; schematic illustrates the extended posture. (F) Quantification of leg extension phenotypes in dusted and undusted headless flies. Bar plots show the percentage of flies displaying leg extension (red) or a normal posture (gray). Percentages are calculated as the number of flies showing each posture divided by the total number of flies per condition. Dusted: n = 9; undusted: n = 5. (G–H) Silencing 13A and 13B neuron subsets locks front legs in flexion in headless flies. (G, G′) Representative video frames showing dusted and undusted headless flies with sustained front leg flexion following silencing of neurons labeled with R35G04-GAL4-DBD, GAD-GAL4-AD > UAS-TNTe. Blue arrowheads indicate the flexed posture. (H) Quantification of leg flexion phenotypes in dusted and undusted headless flies. Bar plots show the percentage of flies displaying sustained flexion (red). All flies (100%) in both dusted (n = 13) and undusted (n = 9) conditions showed the phenotype. Also see Figure 1—Video 1.

Spatial Map of Premotor 13A Neurons Correlates With Their Connections to Motor Neurons (MNs)

(A) Hierarchical Clustering of 13A Hemilineage. Clustering of 13A neuron types in the right T1 segment was performed using NBLAST, resulting in identification of 10 morphological groups or Clusters. EM reconstructions of distinct 13A clusters are shown. Neurons are named based on morphological clustering. For example, all neurons in the 13A-3 cluster have similar morphology, with 10 neurons labeled as 13A-3 (a-j) (olive). Images of each 13A neuron are shown in Figure 2—Figure Supplement 1. Also see Figure 2—Video 2. A= anterior, L= Lateral. Ventral side is up. (B) Cosine similarity graph showing pairwise similarity between 13A neurons based on their MN connectivity patterns. 13A neurons are organized based on anatomical clusters obtained with NBLAST as described above. It depicts a correlation between anatomy of 13A neurons and their connections to MNs. For example, 13A-1a, -1b, -1c, -1d (cluster 1) connect to same set of MNs, therefore have high cosine similarity with each other (as seen across the diagonal). Graph also gives insights into groups of 13As that control similar muscles. For example, cluster 1 neurons have high cosine similarity with cluster 3 13A neurons (while, 3g neuron is an exception).

Inhibitory Circuitry for Antagonistic Muscle Control

(A) Schematic of inhibitory circuit motifs (A1) Feedforward inhibition by 13A/B neurons. (A2) Flexor inhibition and extensor disinhibition: 13As-i inhibit flexor MNs and disinhibit extensor MNs by inhibiting 13As-ii. (A3) Extensor inhibition and flexor disinhibition: 13As-ii inhibit extensor MNs and disinhibit flexor MNs by inhibiting 13As-i. (A4) 13B mediated disinhibition: 13Bs disinhibit MNs by targeting premotor 13As, while some also directly inhibit antagonistic MNs. (A5) Reciprocal inhibition among 13A groups that inhibit antagonistic MNs may induce flexor-extensor alternation. (B) Connectivity matrix: Inhibitory connections regulating antagonistic MNs of the medial joint. Leg schematic shows tibia extensor (orange) and flexor (blue) muscles, innervated by respective MNs. Flexor-inhibiting 13A neurons (13As-i) in blue, and extensor inhibiting 13As (13As-ii) in orange. The thickness of edges between nodes is determined by number of synapses. Node colors were assigned based on the type of neurons, with specific colors denoting different subtypes of 13A/B neurons and MNs. Feedforward inhibition: Primary neurons (13A-10f-α, 9d-γ, and 10e-δ) and 13A-10c (13As-i) connect to tibia flexor MNs (blue edges), making a total of 85, 219, 155, 157 synapses, respectively. Twelve secondary 13As ii inhibit tibia extensor MNs (orange edges), with strong connections from 13A-9f, -9e, and -10d totaling 188, 275, 155 synapses, respectively. Reciprocal inhibition: Three neurons from 13As-i inhibit six from 13As ii, with 13A-10e-δ connecting to 13A-9f (19 synapses), -9e (31), and -10d (14). 13A-10c connects to 13A-8a (6), -8b (12), and -8e (5). 13A-9d-γ connects to 13A-8e (8). Conversely, three from 13As-ii inhibit two neurons from 13As i, with 13A-9f connecting to 13A-10f-α (25) and -9d-γ (6), and 13A-10d connecting to 13A-10f-α (8), -9d-γ (7), and -10e-δ (15). 13A-9e connects to 13A-10f-α (21) and -10c (47) (black edges). Disinhibition by 13B neurons: 13B connects to 13As-i (13A-10f-α and -9d-γ) (totaling 78 and 50 synapses) (green edges), disinhibiting flexor MNs. 13B-2g and -2i also directly inhibit tibia extensor MNs. (C) Reciprocal inhibition for multi-joint coordination: Primary 13As (10e-δ and 10f-α) target a combination of proximal (sternotrochanter, tergotrochanter, trochanter extensor, tergoplural promotor), medial (tibia flexor), and distal (tarsus depressor) MNs, while secondary 13As (9e and 9f) target antagonist MNs including sternal posterior rotator, pleural remotor abductor, and tibia extensor. Reciprocal connections between them indicate that generalist 13As coordinate multi-joint muscle synergies through inhibition of antagonistic motor groups. Leg schematic shows the muscles innervated by the corresponding MNs in various leg segments (Th = thorax, C = coxa, Tr = trochanter, Fe = femur, Ti = tibia, Ta = tarsus). (Data for 13A-MN connections are shown in Figure 2—figure supplement 1 I9, I6, I7, H9, H4, and H5; 13A-13A connections shown in Figure 3—figure supplement 1C). (D) Proprioceptive feedback: Sensory feedback from proprioceptors onto reciprocally connected 13As could turn off corresponding MNs and activate antagonistic MNs. Flexion-sensing proprioceptors target extensor MNs and 13As-i that inhibit tibia flexor MNs. Extension-sensing proprioceptors target tibia flexor MNs and two 13As (13As-ii) that inhibit extensor MNs. Claw extension neurons also connect to 13A-δ. One 13B that disinhibits flexor MNs also receives connection from extension-sensing proprioceptors. Also see Figure 3—figure supplement 3.

13A and 13B Neurons Are Required for Leg Coordination During Grooming

(A-A”) Intra-joint coordination and muscle synergies. (A) Angular velocities of proximal (P, blue), and medial (M, cyan) joints predominantly move synchronously, while distal (D, purple) can move in or out of phase during leg rubbing. The schematic (right) indicates the corresponding joint angles. (A’-A”) The proximal and medial joint movements within a leg occur effectively in phase, with a mean lag of ∼0.8 frames (8 ms) during leg rubbing (A′) and during head grooming sweeps (A″). Bar plots show the lag; each dot indicates one animal. Frame = 10ms. (B) Neuronal labeling of 13A and 13B neurons. Top: Confocal image of six Dbx positive 13A neurons per hemisegment labeled by GFP using R35G04-GAL4-DBD, Dbx-GAL4-AD in VNC. Neuroglian (magenta) labels axon bundles. Bottom: Confocal image of three 13B neurons per hemisegment labeled by GFP using R11B07-GAL4-DBD, GAD-GAL4-AD. Nc82 (magenta) labels neuropil. (C-I) Effects of neuronal activity manipulation in dusted flies. Silencing and activation of 13A neurons in dusted flies using R35G04-GAL4-DBD, Dbx-GAL4-AD with UAS Kir or UAS CsChrimson, respectively (n=12 silencing, n=19 activation). Control: AD-GAL4-DBD EMPTY SPLIT with UAS Kir or UAS CsChrimson. For 13B neurons, R11B07-GAL4-DBD, GAD-GAL4-AD with UAS GtACR1, or UAS CsChrimson respectively (n=7 silencing, n=9 activation); control: AD-GAL4-DBD EMPTY SPLIT with UAS GtACR1 or UAS CsChrimson. Each panel compares control (blue) and experimental (orange) groups. Each dot represents the mean feature value for a single fly. Bars indicate the group mean, and whiskers represent the 95% confidence interval of the group mean. P-values (raw and false discovery rate [FDR]–corrected) are shown above each panel. (C-D) Proximal inter-leg distance: Silencing of 13A (C) or 13B (D) neurons during head grooming reduces the distance between the femur-tibia joints of the left and right front legs.(E-I) Frequency modulation: Silencing 13A or 13B neurons reduces mean frequency of proximal joint oscillations in dusted flies. (F, G). Activation of 13A neurons reduced frequency, although this change did not survive FDR correction. However, continuous activation of 13A and 13B neurons increased variability in frequency. (H, I). Mean of the per-animal standard deviation (STD) that reflects variability or spread of data is shown.

Pulsed Activation of 13A Neurons Triggers Rhythmic Actions in Clean Undusted Flies

(A) Schematic showing proximal joint angles of left (PL) and right (PR) legs (B) Left-right coordination and muscle synergies during anterior grooming. Dusted flies perform alternating leg rubs and head sweeps. Proximal joint angular velocities are shown. PL (blue) and PR (red) joints move anti-phase during leg rubs and in-phase during head sweeps (highlighted yellow box). Positive values indicate extension, and negative indicates flexion. (B’) Each flexion and extension cycle lasts ∼140 ms, with each phase around 70 ms. (B’’) Mean lags between proximal joints of the left and the right legs during leg rubbing and head sweeps. High lag during leg rubbing (left pannel) indicates out of phase movement and low lag during heads sweeps (right) indicates in phase movement. Bar plots show the lag; each dot indicates one animal. Frame = 10ms. (C-F) Effect of optogenetic activation using 70ms on and 70ms off pulses in specific 13A neurons (R35G04-DBD, Dbx-GAL4-AD >UAS CsChrimson) in undusted flies. (C) Angular velocity of PL and PR leg joints shows anti-phase leg rubs and sustained in-phase head sweeps, with light pulses active from time=0. (D) Behavioral ethogram showing various grooming actions (head, front leg, abdomen, back leg, wing, thorax) and walking triggered by 70ms on and 70ms off pulsed activation of 13A neurons in undusted flies, with light pulses on from time=0. (E) Maximum angular velocity of proximal joints during head sweeps upon pulsed 13A activation in undusted flies is comparable to that observed in dusted flies. (F) The frequency of proximal joint movements during leg rubbing (left) and head sweeps (right) induced by 13A pulsed activation is also similar between dusted and undusted flies. Control flies: AD-DBD EMPTY SPLIT > UAS-CsChrimson, dusted; Experimental flies: R35G04-DBD, Dbx-GAL4-AD > UAS-CsChrimson, undusted. Light pulses were delivered at 70 ms on / 70 ms off. Each panel compares control (blue) and experimental (orange) groups. Each dot represents the mean feature value for a single fly. Bars indicate the group mean, and whiskers represent the 95% confidence interval of the group mean. P-values (raw and false discovery rate [FDR]–corrected) are shown above each panel.

Modeling the 13A Circuits.

(A) Circuit diagram showing inhibitory circuits and synaptic weights based on connectome. (B) Adjacency matrices from the empirically estimated weights, indicated in the simplified circuit diagram in (A). The 13B neurons in this model do not connect to each other, receive excitatory input from the black box, and only project to the 13As (inhibitory). Their weight matrix, with only two values, is not shown. Excitatory and inhibitory connections are shown in red and blue, respectively. (C) Adjacency matrices of the model circuits same as in (B) but after fine-tuning. The three “joint” angles of the left leg (left) and the right leg (right) as they change over the time of one episode (500 frames). Colors indicate “joints” as follows: distal (cyan), medial (pink), and proximal (blue) for the left leg. Right leg: distal (purple), medial (orange), and proximal (red). (D) Same as (D) but zoomed-in to between 300-400 frames. (E) Same as (E) but showing angular velocities [°/frame]. (F) Firing rates (activity levels) of the two 13A neurons (red and blue) over one episode (500 frames), for both legs (left, right). (G) Same as (G) but zoomed-in to between 300-400 frames. (H) Video frames from the beginning, middle, and end of a video of one episode. Left leg is represented by three “joints”: distal (cyan), medial (pink), and proximal (blue). Right leg: distal (purple), medial (orange), and proximal (red). The legs originate from the “base” (yellow). As legs move over the “body” (the environment – dust is represented as the green Gaussian distribution), the dust (green) is getting removed (black background). The bottom of each movie frame shows the activity of the two left 13A nodes and six left MNs (blue). The right leg nodes are shown in red, on the right side. Brightness of the nodes indicates the activity level. See Figure 6-Video 1. (I) The dynamics of angular velocities of the left leg’s “joints”, and left 13A activation levels, over 100 episodes (500 frames each), when no stimulus is given (indicated by empty matrix on the top). Each row of each matrix is one episode. (The simulation started running for 50 frames before Time=0 but it starts with very high peaks which were not plotted here for better visualization.) (J) Same as in J, but stimulation with pulses of varying durations is given. Top row of each matrix: pulse duration=2 frames; bottom (100th) row of each matrix: pulse duration=100 frames. The pulse stimulation is indicated in the top matrix.

The Modeled 13A Circuits Can Produce Rhythmic Behavior and Activity Without Rhythmic External Input

(A) The three “joint” angles of the left leg (left) and the right leg (right) as they change over the time of one episode (500 frames). Colors indicate “joints” in the as follows: distal (cyan), medial (pink), and proximal (blue). Right leg: distal (purple), medial (orange), and proximal (red). (B) Same as (A) but showing angular velocities [°/frame]. (C) Firing rates (activity levels) of the two 13A neurons (red and blue) over one episode (500 frames), for both legs (left, right). (D) Same as (C) but zoomed-in to between 300–400 frames.