Figures and data
Spatial characterization of Brp::rGFP clusters at single-AZ level.
(A) Schematic of the split-GFP tagging strategy. The GFP11 fragment is inserted by CRISPR/Cas9 just prior to the stop codon of brp, while GFP1-10 is expressed in specific cell types via GAL4-UAS system. (B) Brp::rGFP signals in DPM. The entire MB lobe structure is shown. The black box in the left panel indicates the zoom-in area (right panel). GFP1-10 and CD4::tdTomato (magenta) were co-expressed using VT64246-GAL4. Scale bar, 20 μm (left); 5 μm (right). (C) Schematic showing brain regions and cell types analyzed in Figure 1C-F and H. (D) Brp::rGFP (green) and anti-Brp immunostaining signals (blue) in the mushroom body calyx. GFP1-10 was expressed using pan-neuronal driver R57C10-GAL4. Scale bar, 1 μm. (E) Brp::rGFP (green) and Cac::tdTomato (orange) in PAM-γ5 dopamine neuron terminals. GFP1-10 and Cac::tdTomato were co-expressed using R15A04-GAL4. Scale bar, 1 μm. (F) Brp::rGFP (green) in R8 photoreceptors upon constant dark or light conditions. GFP1-10 and CD4::TdTomato (magenta) were co-expressed using Rh6-GAL4. Brp::rGFP cluster number of each R8 photoreceptor terminal in the medulla were quantified. n = 30 (3 brain samples, 10 terminals from each brain). Welch’s t test. Scale bar, 5 μm. (G) Image processing pipeline. Raw images were processed with image deconvolution to improve the signal quality. 3D maxima (green pixels) are detected for each Brp::rGFP cluster. 3D ROIs are indicated by yellow circles. Nearest neighbor distance (NND), r is defined as 3 × mean NND. Scale bar, 1 μm. (H) Number of Brp::rGFP clusters detected in DPM and APL. The dashed lines indicate the number of electron dense projections identified in the hemibrain connectome for DPM (17530) and APL (10789) on the MB lobes and peduncle (Scheffer et al., 2020). Schematics illustrate the innervation patterns of DPM and APL neurons. For both DPM and APL, n = 5. Lower panels show the representative 3D reconstructions of Brp::rGFP clusters in DPM and APL. Scale bars, 20 μm. Data were presented as mean ± s.e.m.
Compartmentalized AZ structures of KCs.
(A) Histograms of the signal intensity of individual Brp::rGFP clusters from three KC subtypes. Values are normalized by the mean intensity in each data set (each sample). Each line represents the histogram of one independent sample. γ KCs (MB009B-GAL4), n = 6; α/β KCs (MB370B-GAL4), n = 5; α’/β’ KCs (MB008B-GAL4), n = 5. (B) 3D reconstructions of Brp::rGFP clusters in three KC subtypes, colored by the Brp::rGFP intensity. The approximate locations of compartments are as indicated. Schematics illustrate the innervation patterns of different KC subtypes. Min. and Max. were set to represent the lowest and highest 5% of Brp::rGFP intensity value in the dataset respectively. Scale bars, 20 μm. (C) Signal intensity of Brp:rGFP clusters in each compartment. Brp::rGFP clusters were quantified compartmentally. Medians in different compartments are showed as the ratio against the average of five compartments in the corresponding KC subtype. Each line represents one independent sample.
The variability of Brp::rGFP concentrations depends on cell type.
(A) Brp::rGFP clusters with distinct intensities but similar size in DPM. The left panel shows a cropped image of Brp::rGFP in DPM. White boxes indicate areas zoomed-in in panels 1 and 2. Transparent white lines in panels 1 and 2 show the lines on which the intensity profiles were plotted. Intensity profiles were plotted for AZ1 (magenta) and AZ2 (blue) respectively. (B-D) Scatter plots showing the correlation between the Brp::rGFP intensity and cluster volume in DPM, APL and α/β KCs. Data of three representative samples were shown. Pearson correlation coefficient R were calculated for each sample. (E) Correlations between the intensity and volume of Brp::rGFP clusters in different cell types. DPM (n = 6), APL (n = 5), γ KCs (n = 6, MB009B-GAL4), α’/β’ KCs (n = 5, MB370B-GAL4), α/β KCs (n = 5, MB008B-GAL4). DPM vs. α’/β’ KCs: P = 0.0253; DPM vs. α/β KCs: P = 0.0009; APL vs. α/β KCs: P = 0.0035; γ KCs vs. α/β KCs: P = 0.0418. Values marked with different lowercase letters represent significant difference (P < 0.05); Data were presented as box plots showing center (median), whiskers (Min. to Max.).
Stereotyped AZ distribution of PPL1-α3 DAN.
(A) Brp::rGFP in PPL1 DANs. UAS-GFP1-10 was expressed using MB504B-GAL4. Dashed line marks the rough boundary between α3 and α2α’2 compartments. Schematic illustrates the innervation patterns of PPL1 DANs within the MB. Scale bar, 20 μm. (B) Brp::rGFP intensity profiles of α3 and α2α’2 compartments. Left panels show the max-projection images of α3 and α2α’2 optical coronal sections. Transparent gray stripes indicate areas where intensity profiles are plotted in the right panels. Scale bars, 10 μm. (C) 3D reconstruction colored by the AZ density in PPL-1 DANs. Color scale: Min. = 0, Max. = 40. Dashed line indicates the rough boundary between α3 and α2α’2 compartments. Scale bar, 20 μm. (D) Stereotyped AZ distribution pattern in PPL1-α3 across individuals. 3D reconstructions show the AZ density across different brain samples. Scale bar, 20 μm.
Cell-type-specific stereotypy of AZ spatial distributions.
(A) 3D reconstructions of Brp::rGFP clusters in DPMs, colored by AZ density. Color scale: Min. = 0, Max. = 35. Black arrows indicate consistently high AZ density regions across brain samples. Scale bars, 20 μm. (B) 3D reconstructions of Brp::rGFP clusters in APLs, colored by AZ density. Color scale: Min. = 0, Max. = 30. Black dashed square indicates the area zoomed in. APL reconstructions are arranged from left to right according to the overall AZ density in α3.
Local intensity analysis revealed sub-compartmental AZ structures.
(A) Brp::rGFP correlation analysis between nearest neighbors (NN). The intensity of a cluster is plotted on the x-axis and the intensity of its nearest neighbor is plotted on the y-axis. High correlation indicates nearest neighbors have similar Brp::rGFP intensity. (B) Scatter plots showing the Brp::rGFP intensity correlation between nearest neighbors in a representative sample of γ KCs. Values were normalized to the mean. Pearson correlation coefficient R is shown. (C) Correlation of Brp::rGFP intensities between nearest neighbor AZs in different cell types. Brp::rGFP intensities were log transformed. DPM vs. γ KCs: P = 0.0066; DPM vs. α’/β’ KCs: P = 0.0027; DPM vs. α/β KCs: P = 0.0066; Scale bar: 1 μm. (D) 3D reconstructions of Brp::rGFP clusters in a γ KC sample, colored by local intensity and AZ density. Black arrows indicate areas with both high local intensities and high AZ densities. Min. and Max. were set to represent the lowest and highest 5% in local intensity value respectively. Color scale of AZ density: Min. = 0 and Max = 35. (E) 3D reconstructions of Brp::rGFP clusters in a DPM neuron, colored by local intensity and AZ density. Black triangle arrows indicate areas with high AZ densities but low local intensities. Min. and Max. are set to represent the lowest and highest 5% in local intensity value respectively. Color scale of AZ density: Min. = 0, Max. = 35. (F) Correlation analysis between local intensity and AZ density. The local intensity of a cluster is plotted on the x-axis and its AZ density is plotted on the y-axis. (G) Scatter plots showing the correlation between Brp::rGFP local intensity and AZ density in a representative sample of γ KCs. Values were normalized to the mean. Pearson correlation coefficient R is shown. (H) Correlations between AZ density and local intensity in different cell types. DPM (n = 6), APL (n = 5), γ KCs (n = 6, MB009B-GAL4), α’/β’ KCs (n = 5, MB370B-GAL4), α/β KCs (n = 5, MB008B-GAL4). DPM vs. γ KCs: P = 0.0171; DPM vs. α’/β’ KCs: P = 0.0171; APL vs. γ KCs: P = 0.0014; APL vs. α’/β’ KCs: P = 0.0014; APL vs. α/β KCs: P = 0.0171.
Associative conditioning re-organizes sub-compartmental active zone clusters.
(A) Brp::rGFP (green) and CD4::tdTomato cytoplasm membrane marker (magenta) in KCs, visualized by using R13F02-GAL4. (B) Single odor conditioning induces long-lasting memory. Left panel, the experimental design of the aversive single odor conditioning. The paired group receives a concurrent presentation of 2% 4-MCH and 90 V electric shock. The unpaired group first receives the electric shock and then 4-MCH 1 min later. Right panel, preference of Canton S between 2% 4-MCH and paraffin oil at 3, 10, 30, 90, 270 minutes and 1 day after single odor conditioning. n = 8 for all groups. Error bars show S.E.M. (C) Correlation coefficient (AZ density vs. local intensity of individual Brp::rGFP clusters) of each compartment at 90 min after conditioning. Pair (n = 11) vs. Unpair (n = 12), γ2 (P = 0.0261), β2 (P = 0.0060), α1 (P = 0.0026), α’3 (P = 0.0026). (D) Heatmap showing the difference of correlation coefficient (AZ density vs. local intensity of Brp::rGFP clusters) between the pair and unpair group at different time points after conditioning. The color indicates the difference and the asterisks in the compartments indicate the significant difference. For 3 min, Pair (n = 12), Unpair (n = 11); For 20 min, Pair (n = 12), Unpair (n = 11); For 90 min, Pair (n = 11) vs. Unpair (n = 12), P = 0.0247; For 270 min, Pair (n = 11), Unpair (n = 11); For 1 D, Pair (n = 12), Unpair (n = 9); For all results involved statistical comparison, only significant results are shown. Scale bars, 20 μm. Mann-Whitney test with original False Discovery Rate method of Benjamini and Hochberg correction. *P<0.05. Data were presented as box plots showing centre (median), whiskers (Min. to Max.). (E) Schematics showing learning induced local AZ remodeling. Single odor aversive training transiently dismisses high Brp level – high AZ density hot spots in specific compartments of KC terminals.
Split-GFP tagging does not affect Brp neuronal expression specificity and expression pattern in the brain.
(A) Fold change of brp mRNA level against reference genes, measured by qRT-PCR. mRNA of Ubiquitin-5E and αTubulin84B were used as the reference. n = 3 for all groups. Error bars show S.E.M. ns = non-significant. ***p ≤ 0.001. Mann-Whitney test. (B) Anti-Brp (nc82) immunostaining signal in the MB of flies with or without GFP11 insertion. Both have expressions of GFP1-10 by R57C10-GAL4. n = 11 for both groups. Error bars show S.E.M. (C) Anti-Brp immunostaining of brains of flies with or without pan-neuronal Brp::rGFP tagging using R57C10-GAL4. Different planes of the image stack were shown. Scale bars, 100 μm.
Neuron-specific assembly of Brp::rGFP.
Brp::rGFP and CD4::tdTomato were visualized in astrocyte-like glia using R86E01-GAL4, and in neuropeptide releasing neurons using Amon-GAL4, respectively. Images were acquired using comparable settings. Scale bars, 50 μm.
Rab3 knock-down in KCs increases Brp::rGFP intensity.
Rab3 was knocked-down in KCs using RNAi by R13F02-GAL4. The median of Brp::rGFP intensity of individual clusters was calculated for each γ compartment. n = 6 for both groups. Scale bars, 50 μm in the overview, 2 μm in the insets.
Effect of iterative Richardson-Lucy image deconvolution.
(A) Brp::rGFP clusters of a DPM neuron in raw image, and in images deconvolved for 5∼40 iterations. Max projections of the XY and XZ planes were shown. White transparent lines indicate pixels where the intensity profiles were plotted in B. Numbers denote different Brp::rGFP clusters. Scale bars, 1 μm. (B) Upper panels show intensity profiles of Brp::rGFP clusters undergone different iterations of deconvolution. Lower panels show the full width at half maximum (FWHM) value of intensity profiles. (C) 3D ROIs were generated using deconvolved images and then were used to analyze the volume and intensity of individual Brp::rGFP clusters in raw images. Scale bars, 1 μm.
Heatmaps of F-scores showing the performance of the pipeline in detecting individual Brp::rGFP clusters in different cell types.
An intensity threshold (threshold) is applied to reduce background noise when detecting 3D maxima. Noise tolerance adjusts the sensitivity of 3D maxima detection. Color represents F-score value.
Gross anatomy of KC terminals is not affected by Brp::rGFP tagging.
Morphology of KC terminals in flies with or without GFP11 insertion. Both groups express GFP1-10 and CD4::tdTomato, driven by R13F02-GAL4. Scale bars, 20 μm.
Brp::rGFP tagging in KCs does not affect short-term memory performance.
10% 4-MCH or 10% 3-OCT was paired with 90 V electric shocks. Odor preference was tested immediately after conditioning. GFP reconstitution was induced in KCs using R13F02-GAL4. R13F02>Brp::rGFP, n = 10; Canton S, n = 8. ns = non-significant. Welch’s t test.
