Periodic Reporting for period 4 - BrainInBrain (Neural circuits underlying complex brain function across animals - from conserved core concepts to specializations defining a species’ identity)
Período documentado: 2021-07-01 hasta 2022-08-31
The core function of brains is to compute the current state of the world, compare it to a desired state of the world and select behavior to minimize any mismatch. The neural circuits underlying these functions are unknown. Yet they are a key to understand brains, one of the major scientific challenges of our time. Understanding how brains process information, on the level of individual neurons and circuits, will be essential to develop novel ways of computing, as traditional, transistor based technology has reached its limits. In this project I used insects, whose tiny brains solve the same basic problems as our brains, but with 100,000 times fewer cells. I focused on the central complex (CX), a single brain region consisting of only a few thousand neurons, which is crucial for sensory integration, encoding of behavioral goals and motor control, essentially a ‘brain in the brain’. To simplify the problem further I focused on navigation behavior, where the desired and actual states of the world become the desired and current headings of the animal, and mismatches result in compensatory steering. Following a comparative approach, I established a framework in which to unravel how head direction is combined with goal directions to drive behavior. We have identified pathways that deliver sensory information to the CX to encode a heading signal, circuits suited to encode navigational goals, and have identified the mathematical algorithms of how current heading and goal directions are compared to initiate steering. Overall, we have significantly contributed to elevating the insect CX as a key model system for context dependent action selection.
To generate an initial framework for this project we first developed a biologically constrained computational model of the CX (Stone et al., 2017). This circuit uses identified CX neurons to perform path integration, a navigation strategy used by many animals, including mammals. Importantly, whereas designed for path integration, the underlying computations performed by this circuit can drive any navigational decision. To illuminate the underlying biological circuits in detail, we developed a novel approach to extract connectomics information from large insects. This work first yielded the projectome of the bumblebee CX (Sayre et al., 2021) - to date the most detailed published account of the CX neural composition of any insect besides the fly Drosophila. While all core cell types were highly conserved, we found key differences between bees and flies, which, due to tight structure function links, support different computations (Pisokas et al, 2020). We also discovered neurons in the bee CX without counterpart in the fly, forming circuits suited to encode navigational goals. Synaptic level connectome analysis has since highlighting the conserved 400 million year old core of the CX decision making circuit and highlighted species-specific differences responsible to specific behavioral abilities (2 papers in prep.). To attach relevance to these anatomical data, we used electrophysiology to reveal the sensory pathways relaying optic flow signals to the CX path integrator (Honkanen et al, under review). We identified a neural network that reshapes simple motion signals from the visual periphery into responses encoding the bee’s movement through space. We also revealed that the CX in the migratory Bogong moth encodes the orientation of the Milky Way (Adden et al, in prep.) - distinctly different information than that encoded by corresponding bee neurons. This showed that despite the overall conservation of the CX, sensory inputs depend on a species’ ecology, a finding supported by our connectomics data.
The second part of our work aimed at illuminating goal encoding during path integration. Using a new behavioral paradigm, we demonstrated that bumblebees use path integration while walking over distances of less than two meters instead of hundreds of meters during flight and that their vector memories (goals) were highly resilient to disturbances (Patel et al., 2022). Importantly, the method can be expanded to study more complex forms of vector navigation, including potential cognitive maps. To make these behaviors and vector memories accessible to electrophysiology, we translated the paradigm from freely walking bees into a virtual reality setup, enabling multi-electrode recordings during behavior.
Third, on the output side of the CX, while connectomics data revealed that the computations underlying the initiation of steering commands are likely adapted to the movement strategies of a species (Sayre et al 2021), a model of the premotor command center downstream of the CX predicting a single steering pathway shared between many behaviors (Adden et al, 2022). Combining results from all stages of the project, we have constructed an overarching computational CX model that covers all processing steps from sensory input to motor output (Goulard et al., in prep.). This model provides the detailed computational framework that was the main goal of the project. Finally, besides generating new methods and results, we also created a transparent, open science tool to deposit, manage and share all data resulting from this project: the InsectBrainDatabase (www.insectbraindb.org; Heinze et al, 2021).
The second part of our work aimed at illuminating goal encoding during path integration. Using a new behavioral paradigm, we demonstrated that bumblebees use path integration while walking over distances of less than two meters instead of hundreds of meters during flight and that their vector memories (goals) were highly resilient to disturbances (Patel et al., 2022). Importantly, the method can be expanded to study more complex forms of vector navigation, including potential cognitive maps. To make these behaviors and vector memories accessible to electrophysiology, we translated the paradigm from freely walking bees into a virtual reality setup, enabling multi-electrode recordings during behavior.
Third, on the output side of the CX, while connectomics data revealed that the computations underlying the initiation of steering commands are likely adapted to the movement strategies of a species (Sayre et al 2021), a model of the premotor command center downstream of the CX predicting a single steering pathway shared between many behaviors (Adden et al, 2022). Combining results from all stages of the project, we have constructed an overarching computational CX model that covers all processing steps from sensory input to motor output (Goulard et al., in prep.). This model provides the detailed computational framework that was the main goal of the project. Finally, besides generating new methods and results, we also created a transparent, open science tool to deposit, manage and share all data resulting from this project: the InsectBrainDatabase (www.insectbraindb.org; Heinze et al, 2021).
Besides its scientific results, the project has pushed beyond the state of the art by developing new methods. First, intracellular electrophysiology in insects without genetically labelled neurons had been restricted to restrained preparations. To allow asking questions of how motivational state regulates behavior, we have established a method combining walking behavior on a treadmill with sharp electrode recordings, not requiring genetically labelled neurons for visual control. Second, the investigation of complex vector navigation behavior observed in bees had been restricted to field studies, as these behaviors are carried out over hundreds of meters, rendering their neural basis inaccessible. Our new bumblebee walking paradigm in a controlled, indoor arena, overcame this issue, allowing us to study properties of vector memory, leading the way to study complex forms of vector navigation. Translating our paradigm onto a treadmill in a virtual reality arena now enables us to observe the build-up and usage of vector memories via electrophysiology. Importantly, we have established a new multi-resolution approach to block-face serial section electron microscopy, combining medium resolution overview scans with synaptic resolution scans of individual CX modules to extrapolate an overall connectome at a fraction of the effort compared to other current approaches, pushing our group to the forefront of comparative connectomics research in insects.
Finally, the data generated by this project have sparked an unexpected collaboration between nanotechnology, computer science and insect neurobiology. This collaboration has grown into an EIC funded consortium across five European countries, aiming at developing a novel form of neuromorphic computers, inspired by our anatomically constrained CX models (Winge et al., 2020). Due to their very small physical footprint and an energy consumption on the level of biological brains, these devices are highly relevant for the development of more sustainable computer technology.
Finally, the data generated by this project have sparked an unexpected collaboration between nanotechnology, computer science and insect neurobiology. This collaboration has grown into an EIC funded consortium across five European countries, aiming at developing a novel form of neuromorphic computers, inspired by our anatomically constrained CX models (Winge et al., 2020). Due to their very small physical footprint and an energy consumption on the level of biological brains, these devices are highly relevant for the development of more sustainable computer technology.