Skip to main content

Neural circuits underlying complex brain function across animals - from conserved core concepts to specializations defining a species’ identity

Periodic Reporting for period 3 - BrainInBrain (Neural circuits underlying complex brain function across animals - from conserved core concepts to specializations defining a species’ identity)

Reporting period: 2020-01-01 to 2021-06-30

The core function of all brains is to compute the current state of the world, compare it to a desired state of the world and select motor programs that drive behavior to minimize any mismatch. The neural circuits underlying these functions are unknown. Yet they are the key to understand brains in general, i.e. one of the major scientific challenges of our time. Beyond being one of the central enigmas of modern and ancient science, understanding how brains process information, on the level of individual neurons and circuits, will be essential to develop novel ways of computing, especially as traditional, transistor based technology has reached its limits. Several problems have hindered progress: 1) Our brain is extremely complex, consisting of billions of neurons, interconnected by more synapses than there are stars in our galaxy. 2) The circuits underlying complex behaviors in all vertebrates are widely distributed, containing millions of neurons. 3) Generally, an animal’s desired state of the world is rarely known, making the identification of neural correlates challenging. 4) Even in simpler model species, most studies have focused on sensory driven, reflex-like processes, ignoring self-initiated behavior, therefore missing the key processes. With this project I have begun to overcome these problems using insects, whose tiny brains solve the same basic problems as our brains, but with 100,000 times fewer cells. Moreover, I focus on the central complex, a single conserved brain region consisting of only a few thousand neurons, which is crucial for sensory integration, motor control and state-dependent modulation, essentially a ‘brain in the brain’. To simplify the problem further I focus on navigation behavior. Here, the desired and actual states of the world are equal to the desired and current headings of the animal, with mismatches resulting in compensatory steering. It has recently become known how the central complex encodes the animal’s current heading. Using multiple methods including behavioral training, connectomics, electrophysiology, classic neuroanatomy and computational modeling, I pursue a highly comparative approach to unravel how head direction is combined with goal directions to drive behavior. To reveal which of the circuits involved comprise the conserved core circuitry that exists across insects I compare species with distinct lifestyles. This also aims at revealing how these circuits have evolved to match each species’ unique ecology. To directly link intentional behavior to neural activity I use behavioral training to generate animals with highly defined goal directions, and correlate neural activity with the animal’s ‘intentions’ and actions - at the level of individual neurons. Across species, my work on the central complex aims at uncovering central principles of sensory integration, premotor control, and state-dependent modulation of sensory motor transformations. Therefore, the overall aims of my work is to establish a coherent framework to study key concepts of these fundamental brain functions in unprecedented detail - using a single, conserved, but flexible neural circuit.
The first half of the project was dominated by two parallel developments: First, establishing the conceptional framework and baseline computational model, in which to embed the detailed physiological, behavioral and anatomical experiments, and second, developing the detailed methods and recruiting the team needed to pursue the goals outlined above.
We first developed and published a model of the insect central complex that is completely constrained by neuroanatomical and functional data (in collaboration with B. Webb, University of Edinburgh, UK) (Stone et al., 2017). Similar to the theoretical considerations above, the resulting neural circuit compares the current heading of an insect with a desired heading. In our model, the desired heading is computed by integrating the speed and directions of a bee during a foraging trip and represents the shortest path back to the nest (home vector). This circuit thus uses the anatomical substrate of the insect central complex to carry out path integration, a navigation strategy that is used by many animals, including mammals. This first anatomically constrained model for a path integration circuit maps all functions needed during this complex navigation task to individual cell types of the central complex. Not only does this model use all major neurons known to be conserved across insect species, is requires these cells to have precisely the (sometimes odd) morphological features they have. Whereas we developed the model to carry out path integration, using neurons recorded and described in the bee brain, we realized that the computations performed by the circuit are of a more general nature. How the involved elements can be used, in principle, to carry out any navigation behavior, was proposed in detail in two review papers written by my team (Heinze, 2017; Honkanen et al, 2019). This model has also sparked investigations on Drosophila in several laboratories (V. Jayaraman, G. Maimon, M. Dickinson, personal communications) and has had major influence in developing recent models of desert ant navigation (A. Wystrach, M. Mangan, personal communications).
While establishing the conceptual framework in which to tackle the three main questions of the project, we have formed a highly trained team that has developed the methods needed to perform key experiments. Three PhD students, two postdocs and several students have so far worked on this project and have made substantial progress in pushing physiological, behavioral, and anatomical methods beyond the state of the art in our field (see below). In short, the established novel methods include intracellular recordings from behaving bees, in situ hybridization to detect expression of immediate early genes in navigating bumblebees, behavioral paradigms to probe navigational memory, novel virtual reality display systems for tethered behavior and electrophysiology, and finally, multi-resolution serial block-face electron microscopy (SBF-EM) for establishing local connectomics data across multiple insect species with limited available man-hours. Currently we are in the process of setting up one more method, namely extracellular multiunit recordings from behaving moths and bees, enabling physiological studies on animals during naturalistic behavioral states.
The data generated as proof of concept during methods development is already substantial. We have reconstructed a full dataset of all large columnar neurons of the bumblebee CX from SBF-EM data, which currently constitutes the most detailed account of central-complex neural composition in any insect (publication in preparation). We have used the honeybee as a test subject to establish a workflow using novel imaging techniques (see below) and have thereby generated the first detailed description of the neural composition of the central complex of this important model species. This workflow can now be optimally used for the species of interest of this project, for which specimen have already been collected and prepared.
During electrophysiological studies we have obtained the largest ever obtained dataset (more than 600 recordings) of visually responsive neurons in the central brain of a bee. These data are currently being analyzed and will allow to generate a functional map of the uncharted parts of the bee brain, localizing circuits involved in processing visual information, in particular optic flow. Whereas no clear pathway has emerged for how optic-flow information reaches the central complex, a much more complex picture has begun to emerge, involving many brain regions and dozens of cell types linking the central complex to more peripheral visual processing centers in parallel (two publications in preparation).
Behavioral data accumulated while optimizing our behavioral assay has yielded a complete dataset consistent with the idea that path integration in this species indeed involves optic flow based distance measurement. Many GB of movies are currently being analyzed to quantify the observed effects and to rule out alternative solutions (publication in preparation).
Finally, the first years of the project were also used to create a transparent, open science tool to deposit, manage and share all data resulting from this project. This has yielded the InsectBrainDatabase, a free platform to share anatomical and physiological data from insect brains (www.insectbraindb.org). Beyond making data from this project freely accessible to the community, this website has developed into a platform used by multiple research groups to exchange data and enable efficient cross-species comparisons of insect neuroscience data.
The data that provided the foundation of our computational model of the central complex resulted from several lines of work in bees that we extended beyond the state of the art during the first two years of the project.
Firstly, physiological recordings from bees had led to the discovery of optic-flow based speed neurons (encoding the bee’s speed), whose detailed physiological properties are the basis for optimal performance of the path integration model. These recordings have provided the framework for developing new LED-based virtual reality arenas to be used in conjunction with intracellular and extracellular recordings in behaving bees. These new recordings are now aimed at directly observing the integration of speed and directional cues in the central complex of bees. The dependencies of this process on the behavioral and motivational state of the bee require recordings while the bee carries out directed movements with a clear behavioral goal. Substantial development work has been carried out to enable these experiments. We have so far succeeded in performing intracellular recordings from walking bees and also developed a setup to combine walking bees with extracellular recordings. The latter will now enable to record multiple neurons from the central-complex circuit at the same time and correlate their activity to each other and to the ongoing behavior, offering mechanistic insight into the circuitry. This technique will not only allow us to explore path integration memory, but will enable to correlate steering behavior with neural activity in the central complex, one of the main goals of the project. We will pursue steering initiation in walking bumblebees as well as in flying Bogong moths, the second main model of the study. Finally, this setup will also allow to reveal how sensory representations sent to the central complex are affected by behavioral relevance of a stimulus.
Second, an initial block-face electron microscopical dataset was used to reveal the most likely cellular substrates for path integration memory. The initial strategy for obtaining these data only covered parts of the central complex at high resolution, or the entire structure at low resolution. As a proof of concept we have used these data to fully reconstruct all major columnar neurons of the bumblebee central complex, generating the most detailed anatomical dataset of any insect central complex to date. Using novel techniques, in collaboration with the center for microscopy and microanalysis (CMM) at the Queensland Brain Institute, Australia, we now are able to obtain multi-resolution image stacks, in which we combine medium resolution overview image stacks with synaptic resolution data from several individual modules of the central complex, all from the same individual bee. Within the next two years this newly established strategy will enable us to obtain comprehensive connectivity data on the central complex of all species involved in this study (and beyond), for the first time in any insect larger than the fruit fly, and with more than twice the resolution of the bumblebee dataset. These modular connectomics data will enable us to delineate possible mechanisms for formation of the working-memory underlying path integration, as well as to verify the overall computations predicted by our computational model.
Third, a completely different approach towards identifying navigational memory was developed as well. We have identified and cloned all major ‘immediate early genes’ (IEGs) in the bumblebee and established in situ hybridization to detect their expression patterns. As the expression of these genes correlates with ongoing neural activity, and the memory accumulated during path integration is likely encoded by ongoing neural activity in specific neurons of the central complex, we have started to use mapping the expression of IEGs to localize a footprint of the path integration memory in bees with a highly defined home vector. Complementary to electrophysiology, this technique will allow us to directly locate activity patterns generated by path integration behavior, even in case our computational model is partially incorrect.
Finally, we have established a paradigm to behaviorally verify that bumblebees use optic flow patterns to obtain distance information during homing. While this was known for honeybees and partially for desert ants, in bumblebees behavioral data was lacking, even though the existence of optic-flow based speed neurons in the central complex suggested that this feature is used. Our paradigm has so far yielded data highly consistent with optic flow based distance measurement and final control experiments are currently underway. The developed assay now allows to generate bees with defined home vectors to use in the above IEG approach, as well as to probe the limits of the path integration memory (longevity, resistance to cooling, etc.) behaviorally. This opens the possibility to test effects on any kind of manipulation (pharmacologically, etc) on the homing ability of bees.
Overall, we have developed physiological, anatomical and behavioral methods that extend beyond the state of the art, in line with, as well as beyond what was originally planned for this project. The newly established workflows and strategies will now enable to directly tackle the main questions of the project and fill the conceptual and computational framework developed during the first two years with experimental data. Already these developments have sparked collaborations with physics (hardware implementation of the central-complex model using nano-technology) and have provided ground-truth data for expanding our connectomics approach to many more species. To enable the community to maximally benefit from the accumulating data, we have also developed an open platform to deposit, manage and share anatomical and functional data on insect brains, the ‘InsectBrainDatabase’ (www.insectbraindb.org). This open science platform not only allows to view all data generated by this project, but has started to serve as a community tool to share and compare data, providing a central hub for research data from non-classical model organisms.