Afferent Growth Cone Interactions Control Synaptic Specificity in the Drosophila Visual System  Thomas R Clandinin, S.Lawrence Zipursky  Neuron  Volume 28, Issue 2, Pages 427-436 (November 2000) DOI: 10.1016/S0896-6273(00)00122-7

Figure 1 R1–R6 Target Selection in Wild Type (A) Schematic side views of three adjacent ommatidia (blue ellipses). Individual R1–R6 cells within a single ommatidium are displaced away from the central axis of the ommatidium by a constant angle (denoted X°). This is equal to the angle of displacement between the central axes of two adjacent ommatidia (denoted C1 and C2). This correlation causes R cells, in this case R6 and R5, in neighboring ommatidia, to look at the same point in space (red lines). In this example, R6 is from the left ommatidium; R5 is from the middle ommatidium. (B) Each ommatidium contains six outer R cells, R1–R6, that are responsive to visible light, and one ultraviolet light–sensitive cell, R7. The R8 cell body lies below R7 and is not displayed in these schematics. Six outer R cells in adjacent ommatidia look at the same point in space (red circles). Input from each of these groups of six R cells is superimposed by connecting each of the six R cells to the same lamina cartridge (data not shown). Within a single ommatidium, each R cell samples a different point in visual space (with each point in space represented by a different colored circle) and connects to a different target (see [C]). Two additional R cells, R7 and R8, have different visual pigments and project through the lamina to a different ganglion, the medulla. Lower panels compare the relative arrangement of R cells in the adult retina and the midpupal eye. R cell identity and position in subsequent figures was assessed in pupae. (C) Schematic diagram of the connectivity pattern of one ommatidium, color coded as in (B). Upper layer, retina; lower layer, lamina. The axon bundle twists (arrow) such that the position of the terminus is displaced 180° with respect to the cell bodies (Vigier 1909; Braitenburg 1967). This twist cannot be directly visualized in dye-injected ommatidia (D) but can be inferred from the resulting orientation of the projection pattern. (D) Side view of DiI-labeled R cell projections from one ommatidium. Colored arrowheads mark R1–R6 termini in the lamina. Small arrows denote the axons of R7 and R8. Chevron marks two labeled axons from a partially filled neighboring ommatidium. Scale bar, 5 μM. (E and F) R cell projections from one ommatidum, looking from the retina onto the surface of the lamina, along the long axis of the R cell axon bundle. The extension of the axons across the surface of the lamina is visible in this orientation but their projections into lamina cartridges are not. Scale bar, 5 μm. (F) Schematic view of the lamina, looking down onto the surface from the retina. Hatched circles represent lamina cartridges (not visible in [E]). Neuron 2000 28, 427-436DOI: (10.1016/S0896-6273(00)00122-7)

Figure 6 Axon Fascicle Composition Is Reorganized in Lamina Cartridges In the upper right corner, a single ommatidial bundle in the dorsal hemisphere of the left eye is associated with a column of lamina neurons, designated L1–L5, shown at ∼50% pupal development. Immediately below the cell bodies of the lamina nuclei, R cell axons branch out across the lamina and select cartridges arranged in a characteristic pattern (lower part of figure). The lamina axons from the column associated with an R cell bundle from one ommatidium project through the cartridge directly below the column. In the upper left portion of the figure, two cross-sectional views are presented, from two different levels of a single lamina cartridge (yellow arrows). Two different patterns of axon fasciculation are visible in the two sections; in the upper view (immediately below the cell bodies of the lamina neurons), R cell axons from the same ommatidium and L1–L4 from the same cartridge form a single fascicle, with L5 separate. In the lower view, within a lamina cartridge, R1–R6 axons from six different adjacent ommatidial fascicles surround the axons of L1 and L2; L3–L5 occupy distinct, characteristic positions in the fascicle. The equator of the lamina is to the left; anterior is toward the background. The 180° twist that occurs in the axon bundle between the retina and the lamina is not visible and would be located immediately distal to the plane of this schematic. Neuron 2000 28, 427-436DOI: (10.1016/S0896-6273(00)00122-7)

Figure 2 Development of R1–R6 Cell Projection Specificity in the Lamina (A–D) Profiles of adjacent R cell termini at various times during pupal development. (A) Side view of R cell termini from multiple ommatidia. The projections of R7 and R8, extending through the lamina into the medulla are not visible in this plane. (C–D) Top views of two adjacent ommatidia. Hours denote time after puparium formation; eclosion to adult occurs at 100 hr. Individual growth cones at the periphery of the projection are color coded as in Figure 1C. Magnification as in Figure 1C. (E–G) Structure of the lamina target. (E) is a schematic side view. Each R cell axon fascicle from a single ommatidium (red) is associated with a column of five lamina nuclei of two distinct types (indicated as purple and blue spheres). Each column is surrounded by glial cytoplasmic processes (indicated in green). The lamina plexus lies immediately below each column of nuclei and is shown at 24% pupal development, just before the complex interweaving of axons between cartridges takes place within the layer. The marked plane of section is displayed in (F). (F) The lamina target as viewed from the retina. R cell fascicles (red), glial cytoplasmic processes (green), and the nuclei of single lamina target neurons in each cartridge (blue) form a regular lattice-like array. Individual axons projecting away from each fascicle are not visible in this focal plane. Scale bar, 10 μm. (G) Schematic top view, illustrating the relationship of R cell projections (in red) to the array of lamina columns (blue) and glial cytoplasm (green), at two different developmental stages. Neuron 2000 28, 427-436DOI: (10.1016/S0896-6273(00)00122-7)

Figure 3 Interactions between R Cells Are Required for R1–R6 Targeting (A and E) Wild type, (B and F) phyllopod, (C and G) lozengespr, (D and H) seven-up. (A–D) Schematic summaries of each result. Pairs of R cells are assigned unique colors. phyllopod causes R1, R6, and R7 to be transformed into nonneuronal cone cells (denoted c, in yellow). In lozengesprite and seven-up, different subsets of R1–R6 cells are transformed into R7 cells (black). (E–H) Arrowheads mark individual R cell termini as viewed looking from the retina onto the surface of the lamina. The variations in axonal morphology, including the number and extent of filopodia, seen in these panels reflects small differences in developmental stage (and not the particular transformation mutant used). For instance, the sample shown in (H) is from a slightly older animal. As the projection of R2 across the lamina surface is shorter than that of the other R cells (including R5), its axon is frequently obscured by the fibers lying above it (e.g., E and Figure 4B and Figure 4D). Only ommatidia in which all R cells were DiI labeled (as assessed in the retina) were scored. Axon identities were assigned based on projection length and relative position in wild type, phyllopod, and seven-up. The highly irregular arrangement of projections seen in lozengespr precludes discrimination between R1, R2, R5, and R6 axons and their orientation. In seven-up (H), the projections correspond to the R2/R5 pair. As their orientation is altered, it is not possible to assign a unique identity to either axon. These results demonstrate that defasciculation of R cell axons from the ommatidial bundle does not require a normal R cell complement, while interactions between R cells are required to select the correct pattern of targets. Neuron 2000 28, 427-436DOI: (10.1016/S0896-6273(00)00122-7)

Figure 4 Ommatidial Polarity and a Target-Derived Cue Control R Cell Projection Orientation (A and B) Wild type, (C and D) frizzled, (E and F) nemo. (A, C, and E) Schematic summaries. “Eq” denotes the position of the equator, the dorsoventral midline in both the retina and the lamina. (B, D, and F) Representative R cell projection patterns in the lamina; inset panels display the corresponding ommatidia in the retina. White arrowheads, R3 growth cones; the position of the ommatidial fascicle is marked with an “x”. White arrows denote the relative orientation of the equator. The vector that bisects R7 and R3, pointing toward R3, defines ommatidial orientation. The vector from the ommatidial fascicle along the R3 axon defines projection orientation. In wild type, the orientation of the projection was 180° rotated with respect to the orientation of the corresponding ommatidium, consistent with the 180° twist known to occur in the ommatidial axon bundle (Figure 1B). In frizzled, ommatidia that are correctly oriented (lower image and inset in [D]) project their axons in the correct direction, rotated 180° toward the equator. Conversely, ommatidia that are incorrectly oriented (upper image and inset in [D]) project their axons incorrectly, 180° rotated away from the equator. Therefore, ommatidial orientation determines the orientation of the projection along the dorsoventral axis of the target. In nemo, the orientation of the projection is no longer 180° rotated with respect to the ommatidium, demonstrating that there must be a cue in the target that can reorient the projection to the correct axis (see text). R cell identities were assigned in the retina by morphological criteria using serially reconstructed ommatidia. At this stage of development, R7 is an elongated cell lying between R1 and R6 that contacts all other outer R cells except R3; contact between R2 and R4 prevents R7 from contacting R3. The basal R8 cell lies between R1 and R2. Other R cells also have characteristic morphologies at this stage. Neuron 2000 28, 427-436DOI: (10.1016/S0896-6273(00)00122-7)

Figure 5 Vector Plots of Ommatidial Orientation versus Projection Orientation (A) Wild type, (B and C) frizzled, (D and E) spiny legs, (F) nemo. Individual orientation vectors were determined for each ommatidium in the retina (i.e., R3; red arrow) and its corresponding R3 projection in the lamina (black arrow). The two vectors for each were then compared and plotted with the R3 axon projection orientation set toward either the equator or the pole, as appropriate. Results for frizzled and spiny legs were separated into two groups based on ommatidial orientation: (B) frizzled and (D) spiny legs correspond to ommatidia that were correctly oriented in the retina; (C) frizzled and (E) spiny legs correspond to ommatidia that were flipped 180°. The relative angles between the projection vector and the ommatidial vector in wild-type cluster around 180°, as described in Figure 1C, is consistent with a very precise rotation of the ommatidial bundle. In frizzled and spiny legs, the angle between the two vectors is almost always clustered around 180°, demonstrating that the orientation of the ommatidium determines whether the projection is directed toward the equator or the pole. This observation also implies that the 180° rotation of the axon bundle occurs regardless of whether the ommatidium is normally oriented or flipped. We also observed three exceptional cases (thin black arrows) in which ommatidia that were misoriented projected their axons correctly, suggesting that there may be a weak cue in the target that can reorient projections with respect to the equator. In nemo, the angle between the ommatidial vector and the projection vector is reduced by an amount consistent with the known effect of nemo on ommatidial orientation. That is, ommatidial orientation is up to 45° displaced and the angle between the projection vector and the ommatidial vector is correspondingly reduced (Figure 5F). Ommatidial and R cell projection orientation were determined as described in the legend to Figure 4. Neuron 2000 28, 427-436DOI: (10.1016/S0896-6273(00)00122-7)