Raymond Dunn is a postdoctoral researcher in the Department of Neurology at UC San Francisco, specializing in systems neuroscience and closed-loop neural circuit interrogation. His research combines whole-brain calcium imaging, targeted optogenetics, and custom software engineering to investigate how neural circuits implement cognitive functions such as short-term memory in the nematode C. elegans. Raymond’s work bridges neuroscience and technology, with a focus on real-time control of biological neural networks using the Mightex Polygon1000 DMD.

Like previous winners, Raymond presented his research to the Mightex team — please see the video below.

Crucially, the Mightex DMD offered several advantages for our research. First, it allowed us to stimulate neurons that cannot be easily targeted through genetics alone. The majority of C. elegans neurons cannot be genetically isolated by a single cis-regulatory promoter, and the intersectional genetics is complex and cumbersome. The Mightex Polygon1000 DMD made this project possible by allowing us to selectively illuminate individual neurons or subcellular compartments with optogenetic stimulation light while continuing to image calcium activity across the entire brain with spinning disk confocal microscopy. For example, we can selectively activate ASIL neuron using the osm-10 promoter, also found in ASHL. However we can deliver activation with spatial precision (Fig. 1B), which is impossible with traditional whole-field optogenetics. This approach allowed us to isolate single neuron contributions to neural dynamics.

Integration with Our Closed-Loop Experimental Framework: We integrated the Mightex DMD into a custom microscopy system running our Closed-Loop Experimental Framework (CLEF), a Python-based automation platform for real-time experimental control. Using Mightex’s software kit (SDK), development we programmed rapid updates to stimulation ongoing neural patterns based on activity and behavior. The system analyzes calcium imaging data in real time, identifies target neurons, and sends updated illumination patterns to the DMD within the timescale of neural dynamics (tens to hundreds of milliseconds) (Fig. 1C). Parts of CLEF were incorporated into Pycro-Manager (Pinkard et al. 2021). The DMD enabled fast pattern updates that adapt to the animal’s ongoing behavior measured with CLEF. We stimulated different neurons depending on whether the animal was moving forward or backward, and targeted specific neurons only when they were in particular activity states (Fig. 2A). This capability let us control for within-animal variability by ensuring each animal experiences all experimental conditions.

Virtual Reality for Studying Sensory Memory: Using closed-loop control of the DMD, we created a virtual reality system where optogenetic stimulation of olfactory neuron AWA is timed to the animal’s instantaneous body posture. When the animal bent its head dorsally during forward crawling, we stimulated AWA to mimic encountering higher odor concentration on the dorsal side. After a period of this posture-contingent stimulation, animals would reverse and then preferentially turn in the direction that was previously associated with the virtual odor source. This behavior requires short-term memory because the animal must remember the direction of the sensory cue through an intervening behavior lasting up to tens of seconds. Using whole-brain calcium imaging, we found that this memory is encoded in the relative phase between two oscillatory neural populations with distributed but coordinated signals: one that drives discrete behavioral command states (forward, reverse, turn) and another that drives head swings during crawling. The spatial targeting capability of the Mightex DMD was essential for discovering unexpected functional relationships between neurons. We found that stimulation of several neurons correlated with headswing produced the same response, strongly suggesting functional redundancy (Fig. 2B-D). We expected that stimulating these neurons would bias subsequent turn choice. Instead, we found that stimulating any of these neurons immediately terminated the reversal state, forcing the animal into a turn, however the turn direction was dictated by the spontaneous distributed network state. This was true even when stimulating specific subcellular compartments like the dorsal or ventral neurites of RIA in the nerve ring, which have been previously shown to exhibit compartmentalized calcium dynamics (Hendricks et al. 2012). All together, we found that five different neuron classes/subcompartments within the headswing complex can terminate reversals yet do not bias turn direction, providing evidence that the encoding of turn direction intent in the interaction of distributed dynamical complexes represents a behaviorally causal memory.

Future Directions: We are currently preparing CLEF for publication as an open-source platform for closed-loop neuroscience experiments. The software package will include demonstration protocols for controlling the Mightex DMD, making it easier for other laboratories to implement similar closed-loop experimental designs. The combination of whole-brain imaging with DMD-based targeted optogenetics has opened new possibilities for studying how neural circuits implement cognitive functions. The Mightex technology has been instrumental in enabling this research, and we expect it will continue to drive discoveries about how distributed neural activity gives rise to memory, decision-making, and flexible behavior.