2020 Theses Doctoral
Characterization of global brain state dynamics in Drosophila melanogaster
Internal states, such as arousal and hunger, elevate the probability of a set of behaviors and persist on longer timescales than the behaviors that they predict. These states are triggered by sensors (e.g. neurotransmitters, biogenic amines) within the animal that detect internal homeostatic conditions and external factors. However, the sustained nature of internal states and the diversity of behaviors associated with a singular state suggest that state is represented not only by hormonal and modulatory signals but also by the coordinated activity of neurons within the central brain. Additionally, recent evidence suggests that internal states are represented throughout cortex in rodents and in many neuropil regions in Drosophila. In this thesis, I suggest how persistent states are represented globally in the brain by observing the activity of neurons, at the single-neuron level, distributed throughout the brain of Drosophila melanogaster and determining on what timescales their neural activity predicts behavior.
To do this, we first establish a strategy to rapidly capture brain-wide activity of an awake, freely behaving Drosophila adult. We employ Swept Confocally Aligned Planar Excitation (SCAPE) microscopy, which has been shown to be an effective tool for volumetric imaging in a wide range of living samples, including zebrafish and Drosophila larvae. SCAPE's volumetric imaging speeds exceed those of point-scanning methods ten- to hundred-fold, and offers additional advantages, such as reduced phototoxicity and high signal-to-noise. The optical geometry of SCAPE consists of a single objective located directly above the sample. Therefore, this single stationary objective lens allows for imaging of intact, behaving animals like adult flies. Here, we characterize the spatial resolution of the system with respect to in vivo imaging of neurons in the adult fly brain. We show that we can achieve single-cell resolution, even in closely-spaced or dense neuronal populations. Additionally, we show that high-speed imaging of calcium activity throughout the whole brain can be performed at 20 fly brain volumes per second. These rates allow us to monitor neural dynamics occurring on the time scale of hundreds of milliseconds, which lets us capture the dynamics of popular calcium indicators like GCaMP. Moreover, we have demonstrated the feasibility of this approach to optically record odor responses of individual neurons in the olfactory circuit, while the animal freely behaves on a spherical treadmill.
Having established a system for whole-brain imaging in Drosophila, we then use this methodology to explore the representation of two internal states: arousal, in flies freely running on a spherical treadmill, and hunger, in food-deprived flies consuming sugar. We define internal state as neural activity that predicts behavior on long timescales. To determine the timescale with which individual neurons best predict behavior, we define a regression model in which the activity of each cell is proportional to behavior filtered with unique time constant (tau_i). In freely running flies, we see that the neural activity exhibits a strikingly large dominant mode - nearly all cells across the brain are correlated with locomotion. While the median timescale is short, the distribution of timescales across all cells is broad, with some neurons correlated with locomotion on a much longer timescale, representing arousal based on our definition. In food-deprived flies fed sugar, no dominant mode exists; the neural activity tracking feeding is relatively subtle at the global scale. However, by applying the regression model to determine the timescales of individual cells, we do identify some ensembles of neurons possessing either a short timescale (tau_i < 10s), likely representing reward, or a long timescale (tau_i > 60s), putatively representing hunger. To investigate the populations that make up these different timescales, we used both genetic labeling and hierarchical clustering to determine the identity of neurons of interest. For example, in the freely running flies, we notice that cells in a dorsomedial region called the pars intercerebralis exhibit consistently large tau_i with respect to locomotion. Similarly, by genetically labeling neurons producing the hormone DH44, we see that in food-deprived flies consuming sugar, these neurons exhibit large tau_i with respect to feeding. Thus, we have identified dimensions of global dynamics, including a broadly distributed behavioral state as well as subspaces supporting putative neural correlates of the internal states of arousal and hunger. These data presented in this thesis, and the techniques we have established, have the potential to significantly impact our understanding of internal states at a global level in Drosophila melanogaster and can be extended to other organisms.
This item is currently under embargo. It will be available starting 2022-03-26.
More About This Work
- Academic Units
- Neurobiology and Behavior
- Thesis Advisors
- Axel, Richard
- Ph.D., Columbia University
- Published Here
- May 12, 2020