Dscam2 mediates axonal tiling in the Drosophila visual system

Abstract

Sensory processing centres in both the vertebrate and the invertebrate brain are often organized into reiterated columns, thus facilitating an internal topographic representation of the external world. Cells within each column are arranged in a stereotyped fashion and form precise patterns of synaptic connections within discrete layers. These connections are largely confined to a single column, thereby preserving the spatial information from the periphery. Other neurons integrate this information by connecting to multiple columns. Restricting axons to columns is conceptually similar to tiling. Axons and dendrites of neighbouring neurons of the same class use tiling to form complete, yet non-overlapping, receptive fields. It is thought that, at the molecular level, cell-surface proteins mediate tiling through contact-dependent repulsive interactions, but proteins serving this function have not yet been identified. Here we show that the immunoglobulin superfamily member Dscam2 restricts the connections formed by L1 lamina neurons to columns in the Drosophila visual system. Our data support a model in which Dscam2 homophilic interactions mediate repulsion between neurites of L1 cells in neighbouring columns. We propose that Dscam2 is a tiling receptor for L1 neurons.

Figure 1. Dscam2 is required for visual system development

a, Drosophila Dscam family members. The percentage identity between the extracellular domains is shown at the left, and the number of amino acid residues in the protein at the right. Dscam isoforms differ within three immunoglobulin domains (coloured horseshoes). Dscam2 has two isoforms differing at immunoglobulin domain 7 (red horseshoe). Immunoglobulin domains, horseshoes; FN domains, black boxes; transmembrane domains, blue bars. b, Homologous recombination (HR) scheme to knock out Dscam2 (see Methods). kb, kilobases; w+ indicates the white gene which is used as a marker to detect recombinants. c, Molecular verification of the targeting event by polymerase chain reaction. WT, wild type. d, e, Dscam2 mutants are protein-null. The images show wild-type (d) and Dscam2 mutant (e) pupal brains stained with a Dscam2 antibody 40 h after puparium formation (APF). f, g, R7 and R8 projections in the medulla stained with monoclonal antibody 24B10 (red) are disorganized in adult Dscam2 mutant brains. The projections of other neuronal classes were also disrupted (see Supplementary Fig. 1).

Figure 2. Dscam2 restricts L1 arbors to columns

a, Schematic of lamina neuron and R-cell projections in the medulla. Each cell targets to a specific layer (m1–m6, left) and is restricted to a single column (right). b, c, R7 and R8 do not require Dscam2. The terminals of mutant R7 (b) and mutant R8 (c) (yellow, bottom) in adult brains are indistinguishable from wild-type R7 and R8 (yellow, top). All R7 and R8 axons (red) are stained with monoclonal antibody 24B10 in this figure. d, MARCM scheme. A lamina-specific enhancer was used to drive FLP in lamina precursor cells. Homozygous mutant cells lacking the Gal80 repressor were labelled with actin-Gal4 and UAS-CD8GFP (green). e, f, L2 cells do not require Dscam2. Wild-type (e) and mutant (f) L2 terminals (green) were indistinguishable. g–i, L1 cells require Dscam2 for columnar restriction. Wild-type L1 axons (g) arborized in the m1 and m5 layers of the medulla and were restricted to a single column (dashed lines). Mutant L1 cells (h, i) targeted to the correct layers, but their arbors were not restricted to a single column. Animals were analysed at about 70% APF in e–i.

Figure 3. Dscam2 binds homophilically and its expression is correlated with L1 arbor retraction during development

a, b, Cell aggregation assay (see Methods). a, Control. Cells expressing Dscam2A (marked by co-expression of red fluorescent protein) mixed with cells expressing Dscam2A (marked by co-expression of green fluorescent protein). b, Cells expressing Dscam2A (red) and Dscam2B (green) segregate from one another, showing that homophilic interactions are isoform-specific. c, d, Pull-down assay. Dscam, Dscam2A, Dscam2B and Dscam3 ectodomain–Fc fusion proteins bound their cognate Flag-tagged full-length protein in extracts of transfected S2 cells (c; see Methods). d, Dscam2A or Dscam2B ectodomain–Fc fusion proteins bind to themselves but not other Dscam proteins. Right, inputs. The flag symbols in c and d indicate the Flag epitope. e–g, Wild-type L1 arbor development. At 30 h APF (e), wild-type L1 cells consist of a terminal growth cone and nascent m1 arbors. At about 40 h APF (f), L1 arbors in adjacent columns contact each other. The third column from the left does not contain a labelled L1 cell and this permits the detection of invading neurites from columns 2 and 4. At about 70 h APF (g), L1 processes are restricted to a single column. h, i, Dscam2 protein expression in the medulla during pupal development. Dscam2 expression peaks during the retraction phase of L1 development (40 h APF; h), and is then downregulated (70 h APF; i). j, Dscam2 distribution (green) is non-uniform at 40 h APF. Left, image also stained with monoclonal antibody 24B10 (red). Middle, ×2.5 magnification of the boxed region. Right, at this stage, L1 arbors reside immediately above R7 and immediately below R8 in layers with strong Dscam2 staining.

Figure 4. Dscam2 homophilic interactions are required for axonal tiling

a, Reverse MARCM scheme. The Dscam2 mutation is on the Gal80-containing chromosome so that the wild-type, but not mutant, cells are labelled. b, A wild-type L1 neuron (green) in a wild-type background generated by MARCM (control). c–e, Non-autonomous tiling phenotypes in wild-type cells using the reverse MARCM technique. Note the unidirectional nature of the phenotype. R cells (red) are labelled with monoclonal antibody 24B10. f, Left, Possible outcomes of reverse MARCM. Non-autonomous phenotypes could arise from interactions with a Dscam2 mutant cell in the same or an adjacent column. A requirement in the same column would generate a bidirectional phenotype, whereas a requirement in an adjacent column would generate a unidirectional phenotype. Right, Observed result and interpretation. The reverse MARCM phenotype is exclusively unidirectional (see also Supplementary Fig. 2), indicating that Dscam2 homophilic interactions mediate repulsion. We propose this is due to a mutant ‘unlabelled’ L1 neuron (red) in the adjacent column. g, Model for columnar restriction of L1 arbors. Neurites from L1 cells in adjacent columns interact through Dscam2 homophilic contacts. This generates a repulsive signal resulting in the retraction of neurites to their column of origin. It is important to note that if Dscam2 expression is not restricted to L1 neurons in the layer, then isoform-specific or co-receptor-specific mechanisms may restrict Dscam2 activity to L1 neurons within these layers.