N-Cadherin Orchestrates Self-Organization of Neurons within a Columnar Unit in the Drosophila Medulla

Abstract

Columnar structure is a basic unit of the brain, but the mechanism underlying its development remains largely unknown. The medulla, the largest ganglion of the Drosophila melanogaster visual center, provides a unique opportunity to reveal the mechanisms of 3D organization of the columns. In this study, using N-cadherin (Ncad) as a marker, we reveal the donut-like columnar structures along the 2D layer in the larval medulla that evolves to form three distinct layers in pupal development. Column formation is initiated by three core neurons, R8, R7, and Mi1, which establish distinct concentric domains within a column. We demonstrate that Ncad-dependent relative adhesiveness of the core columnar neurons regulates their relative location within a column along a 2D layer in the larval medulla according to the differential adhesion hypothesis. We also propose the presence of mutual interactions among the three layers during formation of the 3D structures of the medulla columns.SIGNIFICANCE STATEMENT The columnar structure is a basic unit of the brain, but its developmental mechanism remains unknown. The medulla, the largest ganglion of the fly visual center, provides a unique opportunity to reveal the mechanisms of 3D organization of the columns. We reveal that column formation is initiated by three core neurons that establish distinct concentric domains within a column. We demonstrate the in vivo evidence of N-cadherin-dependent differential adhesion among the core columnar neurons within a column along a 2D layer in the larval medulla. The 2D larval columns evolve to form three distinct layers in the pupal medulla. We propose the presence of mutual interactions among the three layers during formation of the 3D structures of the medulla columns.

Keywords: Drosophila; N-cadherin; columnar unit; differential adhesion; optic lobe; visual system.

Figure 1.

N-cadherin and phalloidin visualize developing columns. A, Schematics of the developing larval and pupal visual systems. In the larval eye disc, R8 and R7 are sequentially differentiated behind the morphogenetic furrow. The NBs, located on the surface of the larval brain, produce Mi1 medulla neurons. In the pupal visual system, the retina, lamina, medulla, and lobula are topographically connected. Columns are identifiable along the planes indicated by dotted lines in the larval and pupal brains along the three layers (top, M1–M2; middle, M4–M5; bottom, M9–M10). B–D, The donut-like columnar patterns are identifiable in L3 (B1), 24 h APF (C1–C3), and 48 h APF stages (D1–D3) by Ncad (green), phalloidin (blue), and 24B10 (magenta). B1, Single layer of columnar pattern in L3 larval stage. C1–C3, D1–D3, Three layers of columnar patterns in 24 h APF (C1–C3) and 48 h APF (D1–D3) pupal stages. Layers M1–M2 (C1,D1), M4–M5 (C2,D2), and M9-M10 (C3,D3) are shown. Areas indicated by white rectangles are shown at right (B1,C1–C2,D1–D2). In contrast to the clear and regular patterns in layers M1–M2 and M4–M5 (C1,C2), the columnar pattern is irregular in layers M9-M10 at 24 h APF (C3), which become organized by 48 h APF (D3). Distal to the top, dorsal to the right. The following images are oriented in the same manner. C4, D4, Horizontal sections of 24 and 48 h APF brains to visualize the layers M1–M2, M4–M5, and M9-M10. Scale bar: B1, 5 μm.

Figure 2.

Projection patterns of essential neurons within the columns. Projection patterns of R8, R7, Mi1, L1–L5, Tm3, and T4-T5 are visualized by UAS-myrGFP (green or white) and compared with Ncad pattern (magenta). A1–A3, sensF2-Gal4 visualizes R8 axon terminals showing either donut-like (arrow) or horseshoe-like shape (arrowhead) that overlap with the donut-like Ncad pattern in L3 larval stage (A1). In 48 h APF pupal stage, obscured donut-like (arrows) and faint dot-like patterns are identifiable in layers M1–M2 (A2) and M3 (A3), respectively. B1–B3, PM181-Gal4 visualizes dot-like R7 axon terminals inside the donut-like Ncad pattern in L3 larval (B1) and 48 h APF pupal stages in layers M4–M5 (B3). Faint dot-like patterns of R7 are also identifiable in layers M1–M2 (B2). C1–C3, bshM-Gal4 visualizes grid-like Mi1 axon terminals outside the donut-like Ncad pattern in L3 larval (C1) and 48 h APF pupal stages in layers M1–M2 (C2). Mi1 penetrates inside the donut-like Ncad pattern in layers M4–M5 (C3). In layers M9-M10, Mi1 terminals overlap with the donut-like Ncad pattern (C4). D1–D3, R69B02-Gal4 visualizes dot-like L1–L5 axon terminals inside the donut-like Ncad pattern in L3 larval stage (D1). L1–L5 terminals then overlap with the donut-like Ncad pattern in layers M1–M2 (D2) and M4–M5 (D3) in 48 h APF pupal stage. E1–E3, R13E12-Gal4 visualizes grid-like Tm3 axon terminals outside the donut-like Ncad pattern in L3 larval (E1) and 48 h APF pupal stages in layers M1–M2 (E2), M4–M5 (E3), and M9-M10 (E4). F1, R42F06-Gal4 visualizes T4-T5 axon terminals that overlap with the donut-like Ncad pattern in layers M9-M10 in 48 h APF pupal stage. Scale bar: A1, 5 μm. G, Schematics of the developing columnar units in L3 larval and 48 h APF pupal stages.

Figure 3.

Sequential projections of the columnar neurons. A–E, Sequential projections of R8 (A), R7 (B), L1–L5 (C), and Mi1 (D) are visualized by UAS-myrGFP (white) in L3 larval stages. Magenta represents 24B10. Green represents Ncad. A, R8 terminals are visualized by sensF2-Gal4. R8 terminals are not identifiable before 28 h L3 stage (A1, arrows), start to innervate the medulla neuropil by 32 h, and form the donut-like pattern by 38 h L3 stage (A2,A3, arrowheads). B, R7 terminals are visualized by PM181-Gal4. R7 terminals are not identifiable before 34 h L3 stage (B1, arrows), start to innervate the medulla neuropil after 36 h, and form the dot-like pattern by 42 h L3 stage (B1,B2, arrowheads). C, L1–L5 terminals are visualized by R69B02-Gal4. L1–L5 terminals are not identifiable before 44 h L3 stage (C1, arrows), start to innervate the medulla neuropil, and form the dot-like pattern by 48 h L3 stage (C2, arrowheads). Cell bodies of lamina neurons (asterisks). D, Mi1 terminals are visualized by bshM-Gal4. Mi1 terminals are not identifiable before 24 h L3 stage (D1, arrows), start to innervate the medulla neuropil by 28 h, and form the grid-like pattern by 36 h L3 stage (D2–D4, arrowheads). E, Columnar neurons project to the medulla neuropil in the order of Mi1, R8, R7, and L1–L5. F, Terminals of medulla neurons visualized by drf-Gal4 are not identifiable in the medulla neuropil by 32 h L3 stage. G–K, Projections of individual lamina neurons are visualized by clones expressing UAS-myrGFP (green) in 48 h APF pupal stage in layers M1–M2 (G1,H1,J1,K1), M3 (I1), and M4–M5 (G2,J2,K2). Magenta represents Ncad. L1, L2, L3, L4, and L5 are visualized by clones expressing GFP under the control of c202-Gal4, R16H03-Gal4, R20A03-Gal4, R31C06-Gal4, and bshL-Gal4, respectively. L1 and L2 contribute to layers M1–M2, and L1 and L5 contribute to layers M4–M5. L, Individual Mi1 neurons visualized by clones expressing UAS-myrGFP (green) are projecting to the peripheral region of the columns in L3 stage. Scale bars: A1–G1, 5 μm.

Figure 4.

R8, R7, and Mi1 are the core columnar neurons. Columnar Ncad pattern (green), and projection patterns of R axons (24B10; magenta) and columnar neurons (UAS-myrGFP; white) in WT (A) and in the absence of R8 (B), R7 (C), and Mi1 (D) in L3 larval (A1–D1) and 48 h APF pupal stages (A2–D2, layers M1–M2; A3–D3, layers M4–M5; A4–D4,A5–C5, layers M9-M10). A, In WT, columnar patterns of Ncad are regularly arranged. B, In the presence of sens mutant clones, R axon terminals of 24B10 are lost, and columnar patterns of Ncad are disorganized (arrows) in larval medulla (B1) and in pupal layers M1–M2 (B2), M4–M5 (B3), and M9-M10 (B4). B5, Projection pattern of Mi1 is disorganized in layers M9-M10 due to the loss of R8 in layers M1-M3. B6, sens mutant clone in retina is visualized by the absence of GFP (arrows). C, In sev mutant brains, columnar patterns of Ncad are disorganized (arrows) in larval medulla (C1) and in pupal layers M4–M5 (C3) and M9-M10 (C4). D, In bsh¹/bsh² brains, columnar patterns of Ncad are disorganized or fused with each other (arrows) in larval medulla (D1) and in pupal layers M1–M2 (D2), M4–M5 (D3), and M9-M10 (D4). D4, Neuropil structures visualized by Ncad are missing (arrows). A5–C5, Regularly arranged donut-like Mi1 terminals visualized by bshM-Gal4 in layers M9-M10 are partially lost in the presence of sens mutant clones (B5, arrows) and in sev mutant brains (C5, arrows). Scale bar: A, 5 μm.

Figure 5.

R8 and R7 are essential for medula column formation. A, B, Columnar Ncad pattern (green), and projection patterns of R axons (24B10; magenta) and columnar neurons (UAS-myrGFP; white). A1–A4, R8 is partially eliminated by expressing rpr under the control of sensF2-Gal4 in larval (A1) and 48 h APF pupal stage (A2, layers M1–M2; A3, layers M4–M5; A4, layers M9-M10). Columnar structures are lost when R8 terminals are missing (A1–A4, arrows), whereas R7 terminals are present (A3, arrows). B1–B4, R7 is partially eliminated by expressing sevRNAi under the control of PM181-Gal4 in larval (B1) and 48 h APF pupal stage (B2, layers M1–M2; B3, layers M4–M5; B4, layers M9-M10). R axons and columnar structures are disorganized (B1,B3,B4, arrows). C, Columnar Ncad pattern (green) and projection pattern of L1–L5 neurons (UAS-myrGFP; white). L1–L5 are partially eliminated by expressing rpr under the control of R69B02-Gal4 in larval stage. Columns are not disorganized in the region where L1–L5 are eliminated (arrows). Scale bar: A1, 5 μm.

Figure 6.

Differential expression of N-cadherin in the columnar neurons. Ncad protein localized at the terminals of columnar neurons are visualized by combining Ncad-FsF-GFP (green or white) and FLPase strains that are specific to R7 (A), R8 (B), Mi1 (C), and L1–L5 (D) in L3 larval medulla. Magenta represents Ncad. A, Ncad-GFP is visualized in R7 under the control of R20C11-FLPG5. Ncad-GFP exists in the central hole area inside the columns (arrows). B, Ncad-GFP is visualized in R8 under the control of sens-FLPG5. C, Ncad-GFP is visualized in Mi1 under the control of hth-FLP. Mi1 was identified by Bsh expression. D, Ncad-GFP is visualized in L1–L5 under the control of R27G05-FLPG5 (arrows). Higher-magnification images are separately visualized in the top right corner of each panel. Scale bar: A, 5 μm. E, Densities of photons that derive from Ncad-GFP at the terminals of columnar neurons are compared in L3 larval brains. Histogram of photon densities at the terminals of R8, R7, Mi1, and L1–L5. Percentages of pixel areas that contain 1, 2, 3, and 3–5 photons are compared for each neuron type. Number of axon terminals examined: R8 (n = 7), R7 (n = 8), Mi1(n = 6), L1–L5 (n = 7). *p < 0.05 (t test). **p < 0.01 (t test).

Figure 7.

Roles of N-cadherin in columnar organizations. Columnar Ncad pattern (green), and projection patterns of R axons (24B10; magenta) and columnar neurons (UAS-myrGFP; white) in WT and Ncad knockdown backgrounds in L3 larval medulla. A–D, R7 terminals visualized by PM181-Gal4. A, Control. B, Ncad RNAi in R7. R7 terminals are expanded and the columnar Ncad pattern is disorganized (arrows). C, Ncad RNAi in a subset of R7 under the control of R20C11-FLPG5; GMR>Gal80>Gal4. R7 terminals are expanded and redirected to the column periphery (arrows) as quantified in M. D, Ncad mutant MARCM clones visualizing R7 terminals that are expanded and redirected to the periphery of the column (arrows). E–H, R8 terminals visualized by sensF2-Gal4. E, Control. F, Ncad RNAi in R8. R8 terminals are expanded and the columnar Ncad pattern is disorganized (arrows). G, Ncad RNAi in a subset of R8 under the control of sens-FLPG5; GMR>Gal80>Gal4. R8 terminals are expanded and redirected to the column periphery (arrows) as quantified in N. H, Ncad mutant MARCM clones visualizing R8 terinals that are expanded to the column periphery (arrows). Ncad staining is abolished in and around the mutant clones (arrows). I, J, Mi1 terminals visualized by bshM-Gal4. I, Control. J, Ncad RNAi in Mi1. Mi1 terminals and columnar Ncad pattern are not significantly affected. Scale bar: A, 5 μm. K, L, The effect of UAS-Ncad RNAi (BDSC27503) on Ncad protein level was tested by using optix-Gal4 UAS-CD8GFP (green) in larval medulla. Ncad (white). K, Control. L, Ncad signal is significantly downregulated in GFP-positive area indicated by white dotted lines. M, N, Expansion and redirection ratios of R7 (M) and R8 (N) are quantified according to the definition in O. **p < 0.01 (t test). ***p < 0.001 (t test). O, o and i indicate the total GFP amount within the outer and inner rings of Ncad or phalloidin staining, respectively. w indicates the total GFP amount within the entire growth cone of a single R8 axon. Relative GFP density (R7 and Mi1) or amount (R8) between distinct domains was used to calculate the expansion and redirection ratios. Details are described in Materials and Methods.

Figure 8.

Ncad knockdown reveals differential adhesion between columnar neurons. Influences of Ncad RNAi expression in columnar neurons on other neurons in L3 larval medulla. A–D, R7 terminals are visualized by PM181-Gal4 UAS-myrGFP (green). A, B, Mi1 terminals are visualized by bshM-LexA LexAop-myrTomato (red). A, Control. R7 and Mi1 terminals do not overlap each other (arrows). B, Ncad RNAi in R7. R7 terminals are expanded and Mi1 terminals are disorganized. R7 and Mi1 overlap each other as evident from yellow signals (arrows) as quantified in I. C, D, R8 terminals are visualized by sensF2-LexA LexAop-myrTomato (red). C, Control. R7 terminals are enwrapped by R8 terminals (arrows). D, Ncad RNAi in R7. R7 and R8 overlap each other as evident from yellow signals (arrows) as quantified in J. E–H, R8 terminals are visualized by sensF2-Gal4 UAS-myrGFP (green). E, F, Mi1 terminals are visualized by bshM-LexA LexAop-myrTomato (red). E, Control. R8 and Mi1 terminals are adjacent with each other. F, Ncad RNAi in R8. R8 terminals are expanded and Mi1 terminals are disorganized (arrows). R8 and Mi1 overlap each other as quantified in K. G, H, R7 terminals are visualized by sevEnS-LexA LexAop-myrTomato (red). G, Control. R7 terminals are enwrapped by R8 terminals. H, Ncad overexpression in R8. R7 terminals are expanded toward the column center and overlap with R8 terminals as quantified in L. Rectangles indicated by white dots are separately shown at right (A–H). Scale bars: A, 5 μm. I–L, Expansion of Mi1 (I), R8 (J), and R7 terminals (L), and overlaps between R8 and Mi1 (K) are quantified and statistically tested. *p < 0.05 (t test). **p < 0.01 (t test). ***p < 0.001 (t test).

Figure 9.

Ncad overexpression in the core columnar neurons. Columnar Ncad pattern (green), projection patterns of R axons (24B10; magenta), and columnar neurons (UAS-myrGFP; white) in WT and Ncad overexpression backgrounds in L3 larval medulla. A, B, Mi1 terminals visualized by bshM-Gal4. A, Control. B, Ncad overexpression in Mi1. Mi1 terminals are expanded toward the column center (arrows) as quantified in I. C, D, MARCM clones visualizing Mi1 by bshM-Gal4 UAS-myrGFP. C, Control. D, Ncad overexpression in a subset of Mi1 neurons visualized by GFP. Expansion of Mi1 terminals toward the center of the columns is indicated by arrows. E, F, R8 terminals visualized by sensF2-Gal4. E, Control. F, Ncad overexpression in R8. R8 terminals are redirected toward the column center (arrows) and the central hole area. G, H, R7 terminals visualized by PM181-Gal4. A, Control. B, Ncad overexpression in R7. R7 terminals and columnar Ncad pattern are not significantly affected. Scale bar: A, 5 μm. K, L, The effect of UAS-Ncad on Ncad protein level was tested by using optix-Gal4 UAS-CD8GFP (green) in larval medulla. Ncad (white). K, Control. L, Ncad signal is augmented in GFP-positive medulla neurons. I, J, Expansion and redirection ratios of Mi1 (I) and R8 (J) are quantified according to the definition in Figure 7O and Materials and Methods. ***p < 0.001 (t test).

Figure 10.

The continuous mathematical model of column formation. A, Schematic of the configuration of a column in the control. The variables u, v, and w are the densities of the terminals of R7 (green), R8 (red), and Mi1 (blue), respectively. B, Numerical results of sequential projections of R8 (B1; t = 0) and R7 (B2; t = 100). Identical initial distributions of R8 and R7 produce clear concentric patterns of R7, R8, and Mi1 (B3; t = 1500). C, Segregated distribution patterns of PM181-Gal4 UAS-myrGFP (green) and 24B10 (red) just below the lamina before the arrival at the medulla, where R8 partially engulfs R7. D–H, Results of numerical simulations based on the initial distribution pattern shown in D1. D1, D2, Control (a₁₁ = 2.4, a₂₂ = 1.8, a₃₃ = 0.8). R7, R8, and Mi1 are clearly segregated (D2; t = 1000). E, R7 knockdown (a₁₁ = 1.7). R7 is excluded from the ring of R8 and partially enwraps R8 (t = 1000). F, R8 knockdown (a₂₂ = 0.6). R8 is expanded and fuses with the neighboring columns (t = 1000). G, Mi1 overexpression (a₃₃ = 1.8). Mi1 penetrates the ring of R8 (t = 1000). H, R8 overexpression (a₂₂ = 2.4). R8 and R7 are expanded toward the column center and column periphery, respectively (t = 1000). B3, D1, D2, E–H, Right panels, 3D views of the results, in which z axis indicates the density of neurites. I–K, The adhesiveness of R7, R8, and Mi1 (a₁₁, a₂₂, and a₃₃) is changed between 0.4 and 3.2 to examine the range of parameter settings by which the normal column pattern can be reproduced. I, a₁₁ is fixed to 2.4. J, a₂₂ is fixed to 1.8. K, a₃₃ is fixed to 0.8. Normal concentric columnar patterns are obtained if the order of adhesiveness is conserved (yellow lines; a₁₁ > a₂₂ > a₃₃).