2025 Theses Doctoral

Innate spectral preference and multisensory learning mechanisms in D. Melanogaster

Liu, Po Fung Andrew

Light plays a multifaceted role in shaping behavior, yet the innate value of different wavelengths and their role in learning remain poorly understood. Here, I investigate how Drosophila Melanogaster uses spectral cues to guide both innate phototaxis and learned olfactory associations. Using a high-throughput Y maze assay, I systematically characterize innate preferences across a broad range of wavelengths and intensities.

I show that phototactic behavior can be explained by two principal components—an achromatic and a chromatic axis—that together account for over 90% of behavioral variance. To trace the neural basis of these behaviors, I map photoreceptor contributions to phototaxis from the retina through the lobula. I further test the hypothesis that spectral light carries innate valence signals that can reinforce associative learning. Using monochromatic light as an unconditioned stimulus in a classical conditioning paradigm, I demonstrate wavelength- and intensity-dependent learning: violet light elicits either appetitive or aversive associations depending on intensity, while blue light consistently drives aversive learning. These effects are modulated by PAM and PPL1 dopaminergic circuits, as wellas by Rh7 and CRY photoreceptor pathways.

Finally, I show that spectral cues can also function as contextual signals in multisensory learning, enabling flies to retrieve odor memories in a context-dependent manner. Functional imaging had revealed that UV light robustly activates γd Kenyon cells (KCs), with a majority of responses being UV-selective. Consistent with these responses, I show that flies can use UV as a contextual cue to associate odors with reward. The extent of contextual learning varied by wavelength, with UV vs dark supporting it best, and blue, amber, and red showing limited or no support. In addition, our results suggest extinction-like dynamics, aligning with a reinforcement learning framework. Together, these results reveal that spectral information carries both innate and learned value, and they outline a circuit framework for how light guides complex behaviors in the fly brain.