Chi-Hon Lee, MD, PhD, Head, Unit on Neuronal Connectivity
Shuying Gao, PhD, Postdoctoral Fellow
Chun-Yuan Ting, PhD, Postdoctoral Fellow
Lei Xu, PhD, Postdoctoral Fellow
Shinichi Yonekura, PhD, Postdoctoral Fellow
Phoung Chung, BA, Biological Laboratory Technician

Using the Drosophila color-vision circuitry as a model, we are studying how neurons form complex connections or circuits during development and how the assembled neuronal circuits function to guide animal behaviors. The fly retina contains three types of photoreceptors, R1-6, R7, and R8, each responding to a specific spectrum of light and connecting to a specific layer in the brain in the retinotopic fashion. We are focusing on the connections made by the UV-sensitive R7 neurons. Using wavelength-selection behaviors, we isolated a number of mutants in which various aspects of R7-brain connectivity are affected. We are currently cloning and analyzing these mutants. One mutant, pex (premature extension), which affects R7 retinotopic map formation, is a novel allele of the activin receptor baboon. We further revealed that auto/paracrine activin signals through dSmad2 transcription factor and importin-alpha3 in R7 growth cones to restrict the formation of synapses within their retinotopic visual columns. To study the function of color-vision circuitry, we are focusing on identifying the medulla target neurons that serve as synaptic targets for photoreceptor neurons. We determined their connection patterns and neurotransmitter usage. From this information, we plan to derive the functional architecture of the color-vision circuits.

Molecular mechanisms regulating synaptic target selection of R7 photoreceptor axons

Ting, Yonekura, Xu, Gao; in collaboration with Chiba, Herman, Hsu

In the fly visual system, the UV-responsive R7 photoreceptor neurons connect to the M6 layer in the medulla neuropil. In addition, the connections form a retinotopic map: each R7 axon innervates a single medulla column while preserving its relationship to its neighbors in the eye. Layer-specific connectivity and the retinotopic map are the characteristic features of all complex visual systems. Using a forward genetic approach in Drosophila, we are studying how layer-specific connections and the retinotopic map are established during development.

Our developmental analyses revealed that the layer-specific connections mediated by R7s develop in two distinct stages. During the late larval and early pupal stages (the first target-selection stage), R7 neurons sequentially differentiate and project axons into the R7 temporary layer, where they remain for one to two days. During the late pupal stage (the second target-selection stage), all R7 growth cones regain motility and synchronously project into the destined layer, the M6 layer. Cell-ablation analysis revealed that, during the first stage, layer-specific connectivity is dictated by interactions between R7 afferents and the dendrites of medulla neurons. Characterization of the development of R7-brain connections provides a framework for studying the isolated mutants.

In a forward genetic screen based on wavelength-selection behavior, we isolated N-cadherin, LAR, and a novel mutant ovs (overshoot), which all affect the development of R7 layer-specific connections in a cell-autonomous fashion. N-cadherin encodes a homophilic calcium-dependent adhesion molecule, and LAR encodes a receptor tyrosine phosphatase. N-cadherin or LAR mutant R7 axons retract to the superficial R8-recipient layer during development. Mosaic analyses showed that N-cadherin is required in both R7 afferents and medulla neurons to direct R7-specific connections, suggesting that N-cadherin mediates homophilic interactions between R7 growth cones and dendrites of medulla neurons. Structure-function analyses further revealed that the cytoplasmic domain, and hence catenin-binding activity, of N-cadherin is largely dispensable for R7 layer-specific targeting. The results indicate that the adhesive, but not the signaling, activity of N-cadherin is essential for the development of R7 layer-specific connections. In contrast to N-cadherin, ovs mutant R7 axons overshoot the M6 layer and terminate in a deeper layer. We hypothesize that ovs provides an opposing (or balancing) effect on the R7 growth cones, presumably by counteracting N-cadherin or LAR function. We are currently cloning ovs in the hope that its molecular identity will provide insight into the regulatory mechanisms of R7 target selection.

As regards layer-specific connectivity, the retinotopic map is formed in two separate stages. In the larval stage, a coarse topographic map is established by two partially redundant mechanisms: one is mediated by the secreted protein DWnt4 and its receptor Dfrizzled2 and the other by afferent-afferent and afferent-target interactions mediated by various adhesion molecules. In the past year, we studied how the R7 retinotopic map is maintained and further refined at the pupal stages. In a genetic screen based on wavelength-selection behavior, we identified pex (premature extension), which affects refinement of the R7 retinotopic map. Pex is a novel allele of baboon, which encodes a type I activin receptor. Baboon mutant R7 axons target to the appropriate retinotopic medulla column in the correct layer, but they extend collaterals to innervate neighboring medulla columns, indicating that the map formation, but not the layer-specific targeting, is disrupted.

We found that the canonical activin signaling components are conserved in R7s. Mutations disrupting dSmad2, which encodes the transcription factor downstream of baboon, result in baboon-like R7 phenotypes. In collaboration with Tory Herman, we found that importi-alpha3 mutants exhibit R7 retinotopic map defects essentially identical to those in baboon/dSmad2 mutants. Importin-alpha3 and dSmad2 are present in the R7 growth cones and form physical complexes. Removing importin-alpha3 in R7s disrupts normal nuclear accumulation of dSmad2, suggesting that importin-alpha3 is required for nuclear import of dSmad2. While activin is expressed in both R7s and subsets of medulla neurons, disruption of activin function only in R7s by using RNAi or dominant negative construct produces baboon-like R7 phenotypes. Together, these data suggest that auto/paracrine activin refines the R7 retinotopic map through baboon and dSmad2/importin-α3 complexes.

Mapping color-vision circuits

Gao, Ting; in collaboration with Meinertzhagen, Wang

To understand how color information is processed in the fly brain, we set out to identify the second-/third-order interneurons in the medulla ganglion, which synapse with R7 (UV channel), R8 (green/blue channel), or lamina neurons (green channel). To this end, we exploited the fact that the photoreceptor neurons (but not other neurons in the visual system) use histamine as their neurotransmitter; therefore, their synaptic target neurons must express histamine-gated chloride channels (encoded by the ort gene). We constructed a transgenic fly line that expresses the Gal4 transcription factor under the control of the ort promoter. Using the ort-Gal4 line to drive a membrane-tethered GFP marker, we identified subsets of medulla and lamina neurons that likely serve as synaptic targets for R7/8 and R1-6, respectively. Electron-microscopic studies confirmed that these ort (+) medulla neurons indeed form synaptic connections with R7 and R8 axonal termini. Furthermore, expressing the ort gene under the control of ort-Gal4 completely rescues the electrophysiology and behavioral defects of the ort mutants, indicating that our ort-Gal4 construct faithfully recapitulates the endogenous ort expression pattern.

To determine how information flows in the visual system, we identified (1) the connection patterns of these ort (+) interneurons in the medulla, (2) their dendritic and axonal compartments, and (3) their expression of neurotransmitters and receptors. Using a number of enhancer-trap lines, we found that ort (+) neurons can be further categorized into three subtypes according to the neurotransmitters they use (cholinergic, glutaminergic, and GABAnergic). To determine the connection patterns of the R7/R8 target neurons, we performed single-cell mosaic analyses by using a flipase-based genetic system. We identified seven types of first-order interneurons, including one medulla-intrinsic cell type, which communicates between the external and internal medulla neuropils, and six types of transmedullary neurons (projection neurons). The projection neurons extend dendritic arbors into various medulla layers and project axons to the lobula neuropil to form a topographic map. Two types of projection neurons appear to receive input from two color channels, suggesting that they might function as color-opponent or -summation neurons. The Tm5 type neurons extend dendritic arbors into the M2 (L2) and M3 (R8) layers and convey both green and blue color information to the lobula, the higher visual center. Similarly, the Tm20-type neurons connect both R7 (UV channel) and R8 (blue) inputs to a specific layer in the lobula. We propose that the Tm5 and Tm20 neurons are color-opposing neurons, which calculate the intensity differences between spectra. Our study suggests that the fly color-vision circuit exhibits a design similar to that of primates. Furthermore, the results point to the lobula ganglion as the higher visual center for color vision.

COLLABORATORS

Akira Chiba, PhD, University of Illinois, Urbana, IL
Tory Herman, PhD, University of Oregon, Eugene, OR
Shu-ning Hsu, BA, University of Illinois, Urbana, IL
Ian Meinertzhagen, PhD, DSc, Dalhousie University, Halifax, Canada
Jing Wang, PhD, University of California San Diego, La Jolla, CA

For further information, contact leechih@mail.nih.gov.