Populations of looming responsive VPNs synapse onto single ipsilateral DNs.

(A) Schematic of the Drosophila optic lobe and central brain illustrating two VPNs from different populations that project to their respective glomeruli in the central brain where they synapse onto a DN. (B) Looming sensitive VPN connectivity onto five DNs based on the FAFB EM connectome. The numbers of neurons within a VPN population are given within the circles, and the numbers of synaptic connections made by the whole population onto each DN are given next to the connections. (C) EM reconstructed LPLC1 population from flywire.ai (Dorkenwald et al., 2022). (D) Single cell reconstructions of representative VPNs (LC4, LC6, LC22, LPLC1, LPLC2, and LPLC4). Scale bars 10 µm. Me, Medulla; Lp, Lobula plate; Lo, Lobula.

VPN inputs spatially cluster on DN partners but span multiple sub-branches.

(A) Schematic of example VPNs from two populations (LC4 and LPLC2) synapsing onto a downstream partner (DNp01). (B-F) Left: Synapse locations of VPN populations across the dendritic arbors of the 3D EM reconstructions of five looming responsive DNs (DNp01, DNp02, DNp03, DNp04, and DNp06). VPN populations project their synapses to non-overlapping regions of the DN dendrites. Right: Dendrograms of the same five DNs with VPN synapse locations. VPN populations occupy overlapping sub-branches. Scale bars: 15 µm.

Anatomical receptive field estimations for looming responsive VPN populations tile the whole visual space.

(A) Left: EM reconstructions of the LC4 population (gray) with a quadratic plane (orange) fitted to all LC4 dendrites to estimate the lobula curvature. Middle: 2D lobula projections of LC4 dendrites (gray) and centroids (black). Right: 2D estimation of dendritic receptive fields of the LC4 population projected onto the fly’s visual field. Translucent gray polygons represent the area of the visual field sampled by the dendrites of an individual neuron. (B-F) Same as in (A), but for LC6, LC22, LPLC1, LPLC2, and LPLC4. All Panels: Two example neurons are highlighted in red and blue. Two previously reconstructed Tm5 cells and their centroids are used as landmarks for the center of the eye and a more dorsal position of the central meridian (magenta). The central meridian is denoted by the cyan line. An analysis of how VPN centroids are organized within the lobula space is shown in Figure 3—figure supplement 1.

VPN centroids tile the lobula with biases across the two main axes.

(A) Example LC4 centroid (dorsal: red, ventral: blue) and their dendrite (gray) projections onto the lobula. The black line splits the lobula in half along the dorso-ventral axis. (B) Number of VPN centroids in the dorsal and ventral lobula hemisphere. Some VPN populations show biases. (C, D) Same as in (A, B) but for the posterior (cyan) and anterior (purple) lobula hemispheres.

VPN synapses are not retinotopically organized.

(A,B) Representative examples illustrating no retinotopic organization of two VPN subsets selected from the extreme positions on each lobula axis. (A) Left: Synapse locations of the 10 most dorsal and ventral LC4 neurons on the EM surface reconstruction of DNp01. Right: Synapse locations of the 10 most anterior and posterior LC4s. (B) Same as (A) but for LPLC4-DNp03. Scale bars: 15 µm. (C) Left: Schematic illustrating the calculation of Euclidean distance in the lobula space for a selected pair of neurons. Right: Schematic illustrating the calculation of the distance along the dendrites (µm) for a pair of synapses. For each unique pair of neurons, the average synapse distance is calculated as the average of the distances between all possible unique synapse pairs between the two neurons. (D,E) Representative plots show a lack of retinotopic synapse organization. A positive linear trendline would indicate retinotopically organized synapses. (D) The distance between the centroids of all unique pairs of LC4 presynaptic to DNp01 is not correlated with the average synapse distance for that given pair of LC4 neurons. Linear regression in red. (E) Same as (D) but for LPLC4-DNp03. Analyses as illustrated in (C) for all other VPN-DN pairs are shown in Figure 4—figure supplement 1.

Lack of retinotopic synapse organization in all VPN-DN pairs. Linear regression line in red.

(A-L) (A) The distance between the centroids of all unique pairs of LC4 presynaptic to DNp02 is not correlated with the average synapse distance for that given pair of LC4 neurons. (B) Same as (A) but for LC4-DNp03. (C) Same as (A) but for LC4-DNp04 (D) Same as (A) but for LC4-DNp06 (E) Same as (A) but for LC6-DNp06 (F) Same as (A) but for LC22-DNp03 (G) Same as (A) but for LPLC1-DNp03 (H) Same as (A) but for LPLC1-DNp06 (I) Same as (A) but for LPLC2-DNp01 (J) Same as (A) but for LPLC2-DNp03 (K) Same as (A) but for LPLC2-DNp04 (L) Same as (A) but for LPLC2-DNp06

VPNs project their synapses to neighboring locations.

(A) Schematic representation of classifying nearest neighboring (NN) synapse pairs along the DN dendrites in the original dataset and after shuffling synapse identities. (B) Percentage of unique NN synapse pairs belonging to the same neuron for all VPN-DN pairs for the original and shuffled datasets, Mann–Whitney U test, ****p < 0.0001 (C) Percentage of unique NN synapse pairs belonging to the same neuron for each VPN-DN pair. 1-sample T-test ****p < 0.0001.

Overview of the modeling pipeline.

Fitting passive properties of the models.

(A) Simulated current injection into the soma of a multicompartment model of DNp01, fitted to averaged experimental current injections (−60 pA) from three animals, 20 sweeps per animal. (B) Same as (A) but for DNp03 (current injection: −2 pA).

The SIZ is located downstream of the tether in DNp03.

(A) Maximum intensity projection of RFP labeled DNp03 and GFP labeled para (genotype: para-GFSTF, DNp03-split-GAL4, UAS-RFP). Scale bar: 50 µm. (B-E) Zoomed-in view of (A) showing, (B) the individual para, and (C) DNp03 channels, (D) their overlay, and (E) para colocalization on DNp03. Scale bar: 5 µm. (F) FAFB EM mesh reconstruction of DNp03 (ID: 720575940627645514). Scale bar: 50 µm. (G) Zoomed-in view of the DNp03 mesh reconstruction from (F) showing the tether branchpoint region. Scale bar: 7.5µm. SIZ locations for the other DNs are shown in Figure 8—figure supplement 1-3.

The SIZ is located downstream of the tether in DNp01.

(A) Maximum intensity projection of RFP labeled DNp01 and GFP labeled para (genotype: para-GFSTF, DNp01-split-GAL4, UAS-RFP). Scale bar: 50 µm. Crosses were set at 22°C. (B-E) Zoomed-in view of (A) showing (B) the individual para and (C) DNp01 channels, (D) their overlay, and (E) para colocalization on DNp01 Scale bar: 5 µm.

The SIZ is located downstream of the tether in DNp02 and DNp06.

(A) Maximum intensity projection of RFP labeled DNp06 and DNp02 and GFP labeled para (genotype: para-GFSTF, DNp06-split-GAL4, UAS-RFP). Scale bar: 50 µm. Crosses were set at 18°C. (B-E) Zoomed-in view of top white box in (A) showing (B) the individual para and (C) DNp06/DNp02 channels, (D) their overlay, and (E) para colocalization on DNp06. Scale bar: 5 µm. (F-I) Zoomed-in view of bottom white box in (A) showing (F) the individual para and (G) DNp06/DNp02 channels, (H) their overlay, and (I) para colocalization on DNp02. Scale bar: 5 µm.

The SIZ is located downstream of the tether in DNp02.

(A) Maximum intensity projection of RFP labeled DNp02 and GFP labeled para (genotype: para-GFSTF, DNp02-split-GAL4, UAS-RFP). Scale bar: 50 µm. Crosses were set at 18°C. (B-E) Zoomed-in view of (A) showing (B) the individual para and (C) DNp02 channels, (D) their overlay, and (E) para colocalization on DNp02 Scale bar: 5 µm.

VPN synaptic input is passively normalized at the SIZ.

(A) Left: Schematic of the DNp01 model with the SIZ location in black and randomly selected example synapses from each VPN population color coded. Right: Membrane voltage traces from activations of individual VPN synapses. Example synapse EPSPs are color coded accordingly. All remaining VPN synapse EPSPs are shown in gray. Top: Membrane voltage at the synapse location. Bottom: Membrane voltage of the same simulations at the SIZ. Inset: EPSP amplitude at the SIZ of individual VPN synapses over their distance to the SIZ. (B) Same as (A), but for DNp03 and its VPN inputs.

Linear encoding of VPN synapse numbers.

(A) Depolarization amplitude at the SIZ of DNp01 in response to the activation of multiple synapses. Color-coded circles: Activation of all synapses of an individual VPN. Black circles: Activation of randomly selected synapse sets. (B) Same as (A) but for DNp03 and its VPN inputs.

Shunting of VPN inputs with increasing synapse activation.

(A) Peak depolarization at the SIZ of DNp01 in response to the activation of an increasing number of synapses from a given VPN population. (B) Same as (A) but for DNp03 and its VPN inputs.

Individual VPNs distribute synapses across DN dendrites to avoid shunting.

(A) Example synapse locations across the DNp01 skeleton model for single sets of 15 close and random synapses and for an example VPN with 15 synapses. (B) Same as (A) but for DNp03. Scale bars: 15 µm. (C, D) Composite EPSP amplitudes for the simultaneous activation of the three groups of synapse sets. Composite EPSP amplitudes cover a wider range for the Close group, than for the Random or Single VPN groups. (E, F) Population analysis of data from (C) and (D) shows VPNs distribute their synapses in a random fashion to maximize composite EPSP amplitudes (Kruskal-Wallis tests p < 0.05, followed by a Dunn’s post-hoc test with a Bonferroni adjustment of p-values *p <0.05, **p < 0.01, ***p <0.001, ****p < 0.0001)(G, H): Composite EPSP amplitude depends on the average synapse distance to the SIZ. The Close group covers a broader range of distances to the SIZ and therefore also a larger range of EPSP amplitudes than the other two groups. Analysis for a range of synapse numbers is shown in Figure 12—figure supplement 1.

VPNs distribute their synapses to achieve efficient composite EPSP amplitudes at the SIZ.

(A, B) Composite EPSP amplitudes at the SIZ in response to simultaneous activation of varying numbers of synapses in DNp01 (A) and DNp03 (B). (C,D) Average synapse spread of randomly distributed synapses, closely clustered and single VPNs at varying synapse counts in DNp01 (A) and DNp03 (B).

Model Morphological Parameters