Strategies for assembling columns and layers in the Drosophila visual system

Abstract

A striking feature of neural circuit structure is the arrangement of neurons into regularly spaced ensembles (i.e. columns) and neural connections into parallel layers. These patterns of organization are thought to underlie precise synaptic connectivity and provide a basis for the parallel processing of information. In this article we discuss in detail specific findings that contribute to a framework for understanding how columns and layers are assembled in the Drosophila visual system, and discuss their broader implications.

Keywords: Columnar restriction; Columns; Drosophila visual system; Layer specificity; Layers; Synaptic specificity.

Fig. 1

The Drosophila visual system. (a) Anatomy of the Drosophila visual system (Adapted from Fischbach and Diettrich 1989). (b) Diagram illustrating the modular organization of the Drosophila visual system. Four topographically matched modules from the retina and each region of the optic lobe are shown. Ommatidia (retina), cartridge (lamina), column (medulla), lobula complex modules (lobula and lobula plate). (c) Illustration of a cross section through a lamina cartridge. The axons of R1-R6 photoreceptors synapse onto the dendrites of L1-L3 lamina neurons. The R cell axons form a ring around the dendrites, establishing a cylindrical structure that may optimize wiring efficiency. (d) R cell axons form tetrad synapses. At each R cell synapse, input is provided to four postsynaptic elements. L1 and L2 are present at every R cell synapse, but the other two components are variable and can include L3, amacrine (Am) or glial (not shown) processes

Fig. 2

Alternative splicing of Dscam1 and Dscam2 regulates synaptic exclusion. (a) Properties of Dscam1 and Dscam2 alternative splicing are very different, but allow both to exclude processes from the same cell at tetrad synapses. (b) A schematic of a tetrad synapse (variable components not shown). A random array of Dscam1 isoforms are expressed in L1 and L2. Since these isoforms are not identical between the two cells, homophilic repulsion does not occur. L1 and L2 express distinct isoforms of Dscam2. This allows for self-repulsion, but not repulsion between the two different cells. Through this indirect mechanism of excluding inappropriate partners at synapses, postsynaptic specificity is achieved

Fig. 3

Multiple mechanisms for restricting processes to single columns. Columnar restriction can be achieved through repulsion between neighboring cells of the same type, adhesion to cells within the same column and autocrine signaling that limits growth cone movement. The end result is that connections are made within the column rather than with correct target cells that reside in neighboring columns

Fig. 4

Targeting to the outer or inner medulla. A diagram of medulla development at an early pupal stage (~12 hours after puparium formation [h APF]). Lamina growth cones expressing CadN and Sema-1a are prevented from innervating the inner medulla through repulsive interactions with PlexA expressing medulla tangential cells (MeT), and interactions with other CadN expressing processes in the outer medulla. Mi1 = a medulla instrinsic 1 neuron. The asterisk indicates the youngest lamina neuron axons within the medulla neuropil

Fig. 5

Outer layers develop in a stepwise manner from broad domains. h APF = hours after puparium formation (a) A representation of the adult morphologies of lamina neuron axons L1-L5. The arborizations of lamina neuron axons help define specific outer medulla layers. (b) A drawing of lamina neuron growth cones L1-L5 in early pupal development. Prior to arborizing in discrete layers lamina growth cones terminate in distal or proximal domains within the outer medulla. (c) An illustration of M2 development. A CadN-dependent interaction between the axons of lamina neurons L2 and L5 mediates the branching of L5 axons into the M2 layer. (d) A diagram of M3 development. The M3 layer develops in part through the sequential innervation of L3 and R8 axons. DFezf cell autonomously promotes the targeting of L3 growth cones to the proximal domain of the outer medulla. L3 growth cones then segregate into the developing M3 layer in part through repulsion from medulla tangential fibers (MeT). DFezf activates the expression of Netrin which is secreted from L3 growth cones, and serves as an M3-specific cue for R8 growth cones. (Arrows in the second panel from the left indicate the retraction of the leading edge of an L3 growth cone, and extension of filopodia laterally across the column within the developing M3 layer. The arrow in the third and fourth panels from the left show the secretion of Netrin from L3 growth cones, which becomes concentrated within the developing M3 layer)

Fig. 6

A dynamic model of layer assembly in the medulla. Outer medulla layers are established in a stepwise manner during development through a precise sequence of interactions between specific cell types. To illustrate this, the figure concentrates on the stepwise targeting of L3 lamina neuron axons within the medulla during pupal development. (a) L3 axons (green) are prevented from innervating the serpentine layer and inner medulla by adhesive (CadN-dependent) and repulsive (Sema-1a/PlexA) interactions, that serve as a barrier to further extension. MeT = medulla tangential neurons. The gray neuron represents a potential CadN expressing target of L3 axons. (b) Prior to innervating the target layer, L3 axons (light green) terminate in a proximal domain of the outer medulla shared by the growth cone of another lamina neuron (dark green). Specificity for the proximal domain is regulated by dFezf (not shown). An additional lamina neuron subclass (blue-green) terminates in a distal domain of the outer medulla. (c) (left panel) L3 growth cones undergo a stereotyped structural rearrangement that segregates them into the developing target layer. Another lamina neuron (dark green) forms an additional arborization in the distal outer medulla. These events contribute to the emergence of discrete layers. (middle and right panels) DFezf activates Netrin expression in L3 neurons, and Netrin (purple) is secreted from L3 growth cones (green) providing an M3-specific cue for R8 photoreceptor growth cones (red). The sequential targeting of L3 and R8 growth cones contributes to M3 development. (d) Within the target layer, L3 axons (green) may distinguish between appropriate (dark orange) and inappropriate (light orange) synaptic targets through specific cell recognition molecules such as Dpr and Dip proteins