The Zimmer Lab

We developed a nuclear localized calcium indicator protein that together with fast volumetric fluorescence microscopy enabled us to make true a long-standing dream: to measure brain wide activity in real time and at single cell resolution (Movie, Figure 1A) (Schrödel et al., 2013, Nature Methods). A milestone discovery obtained from such recordings was that most of the active neurons in the awake brain coordinate amongst each other to form collective global activity patterns, even in the absence of any sensory stimulus. Brain states with characteristic neuronal activity pattern are repeated in a cyclical fashion. This can be visualized with computational methods revealing features characteristic of dynamical systems (Figure 1A-B) (Kato et al., 2015, Cell). A thorough investigation of these neuronal population dynamics showed that they correspond to action commands and their assembly into a behavioral command sequence (forward-crawling, backward crawling, turning; Figure 1B) i.e., worms engage in fictive behavioral patterns under such conditions (Kato et al., 2015, Cell).

These basic locomotor states, however, are not the only behavioral representations that can be found in our datasets: investigating the activities of motor neurons we discovered a central pattern generator circuit that operates on a faster time-scale, and which controls explorative movements of the head and nose, termed head casts. Intrigued by these findings, we investigated these behaviors and thereby discovered a hierarchy of movement patterns that is reflected by a hierarchy of neuronal dynamics in the brain (Kaplan et al., 2020, Neuron) (Figure 2). 

Our findings impact the current understanding of the principles of neural network function. First, neural networks are not simple input-output devices but exhibit rich intrinsic dynamical properties that interact on par with external inputs. Second, the high-level and long-timescale organization of behavior i.e., the assembly of action sequences, is represented globally at the network level. Activities of smaller groups of neurons that control faster time-scale movement patterns are hierarchically nested within these global network states. This uncovered for the first time a neuronal correlate for a behavioral hierarchy, a neuro-ethological principle postulated decades ago by the ethologists Tinbergen and Dawkins.

The wiring architecture of neuronal networks is assumed to be a strong determinant of their dynamical computations. An ongoing effort in neuroscience is therefore to generate comprehensive synapse-resolution connectomes alongside brain-wide activity maps. However, the structure-function relationship, i.e. how the anatomical connectome and neuronal dynamics relate to each other on a global scale remains unsolved. To address this, we systematically compared graph features in the C. elegans connectome with correlations in nervous system wide neuronal dynamics obtained from whole brain Ca²⁺-imaging (Figure 3, upper panels). We found that direct connectivity is a weak predictor of neuronal interactions. However, non-local connectivity features such as specific triplet motifs or similarity measures in network neighborhoods (e.g., input similarities) correlate well with functional interactions between neurons (Figure 3, lower panels). Surprisingly, quantities such as connection strength or the amount of common inputs do not improve these predictions, suggesting that the network’s topology is sufficient. We demonstrate that highly interconnected hub neurons (termed rich club neurons) in the connectome are key to these relevant graph features. Consistently, inhibition of multiple hub neurons specifically disrupts brain-wide activity dynamics. Thus, we propose that the rich club hub-architecture and non-local connectivity features provide an anatomical substrate for coordinated global brain dynamics and that neuronal communication within circuits cannot be fully understood when studied in isolation but within the full context of the connectome (Uzel et al., 2022, Current Biology).

What are the functions of brain wide behavior representations? An intriguing feature of them is that they encompass not only motor circuits but also neurons previously thought to exclusively process sensory information. Importantly, such observations were recently made across species, like fruit flies and mice (reviewed in Kaplan et al., 2020, Current Opinion in Neurobiology). One of the major current research goals in the lab is to elucidate the functions of this newly emerging neurobiological principle. In one of our pioneer studies in these directions, we made the surprising observation that even primary sensory neurons retrieve information about the current behavioral state: the BAG carbon dioxide and oxygen sensory neurons are temporally inhibited during short backward crawling episodes. This so termed ‘corollary discharge’ signal is transmitted from the RIM reversal interneuron via extra synaptic tyramine signaling to a cognate ligand gated chloride channel (LGC-55) expressed in BAG (Figure 4). Interestingly, worms can smell their own self-produced carbon dioxide plume when crawling backward into their previous path (Figure 4A). We found that transient inhibition serves the function to reduce perception of this self-induced stimulus (reafference cancelation), which improves perception of external carbon dioxide in the environment (Riedl et al., 2022, Current Biology). We thus propose that reafference cancellation is one of the perhaps multiple functions of brain wide motor representations.

We developed a new paradigm to effectively switch between wakefulness and sleep during whole brain recordings. Animals are alerted and aroused when exposed to atmospheric oxygen levels as they prefer intermediate oxygen concentrations found in their natural habitats. During developmentally timed sleep-prone episodes termed lethargus, they exhibit sleep behavior at intermediate oxygen levels but can be effectively kept awake by atmospheric oxygen (Figure 1A-B). We found that sleep is a state of brain wide quiescence exempting a set of sleep active neurons. While behaviors in awake animals are represented by dynamic network activity as described above, sleep appears as a stationary fixed point in neuronal state space (Figure 1B). When worms are left unperturbed this sleep state spontaneously emerges, but can be rapidly interrupted by the arousing oxygen stimulus. Our work demonstrates for the first time a global activity signature of sleep in an invertebrate, and it is the first brain-wide single cell resolution sleep study in any animal. Our results support the hypothesis stating that sleep arises via a self-organizing network mechanism; however, the brain exerts control over its global state via sleep active neurons and opposing arousal circuits (Nichols et al., 2017, Science).

In order to understand brain functions we aim for a comprehensive quantitative description of its behavioral outputs. We developed high throughput behavioral recording techniques to measure the crawling movements of worms in terms of detailed posture time series (Figure 5). Using a computational method based on principal components analysis, we showed that behavior can be composed out of two elementary motor patterns: the undulation mode contributes to the stereotyped sinusoidal crawling pattern of the animals, while the turning mode contributes to steering and reorientation movements (Figure 5). We identified peptidergic interneurons (termed AVK and DVA) that can gradually control both motor patterns to achieve a spectrum of different behaviors. This work suggests that animals can flexibly generate behavioral diversity out of simple elementary movement modes (Hums et al., 2016, eLife).