Skip to main content

The origins of dendritic computation within mammalian neural circuits

Periodic Reporting for period 3 - DENDRITECIRCUITS (The origins of dendritic computation within mammalian neural circuits)

Reporting period: 2019-07-01 to 2020-12-31

This project aims to address a fundamental question in neuroscience: how are dendritic computations engaged by neural circuits during behaviour? The vast majority of the synaptic input to neurons is made onto their dendrites. The active properties of dendrites can transform these inputs, enabling them to act as computational devices. However, it is unclear whether these features improve behaviourally relevant computations carried out by neural circuits, which depends crucially on the spatiotemporal input patterns delivered by the circuit. Tackling this problem will require a multi-level approach, combining experiment and theory. First, we are determining how the interactions between synaptic inputs can yield computational ‘elements’ in dendrites. Second, we are measuring the input patterns that activate these ‘elements’ in vivo. Third, we are identifying the neurons delivering these patterns, and how the spatiotemporal pattern of their activation can trigger the dendritic nonlinearities. Finally, we are manipulating these nonlinearities directly in the behaving animal to identify their contribution to task performance, and use operant conditioning to reveal behavioural engagement of local dendritic plasticity mechanisms. To achieve these aims we are exploiting a range of novel imaging and recording technologies, applied in parallel in vitro and in vivo, including the use of dual-colour pre- and postsynaptic imaging, combined with ‘all-optical’ simultaneous imaging and optogenetic interrogation, that are allowing us to image dendritic activity, identify the neurons of input, and manipulate both of them during behaviour. These experiments will ultimately provide us with fundamental insights into how single neurons act as computing devices, and how the computations that drive behaviour are implemented on the level of single cells and neural circuits. The possibility that dendritic computations may improve behavioural performance represents a new frontier in neuroscience that can reveal how single neurons contribute to behaviour.
We have made good progress in the reporting period, meeting or exceeding the milestones in the original proposal. Specifically, we have:

1. Shown that active dendrites constitute a key cellular mechanism for driving robust and precise grid cell firing and therefore for ensuring reliable spatial navigation (Schmidt-Hieber et al. Nature Neuroscience 2017).

2. Helped to develop and test Neuropixels, a revolutionary new silicon probe for high-density recording of neural activity (Jun et al., Nature 2017), which we are now exploiting for high-resolution recordings of dendritic activity in behaving animals.

3. Developed a new approach for all-optical closed-loop interrogation of neural circuits in the intact brain, allowing enabling the manipulation of neural circuit activity 'on the fly' during behavior (Zhang et al. 2018).

4. Demonstrated that dendritic signals triggered by climbing fiber input to Purkinje cells in a novel behavioural task signal reward expectation, delivery and omission (Kostadinov et al. 2019).

These technological and conceptual advances can help provide important new insights into the computations being carried out by dendrites of single neurons in behaviour.
In numerous areas we have made progress beyond the state of the art, as reflected in our publications in leading international journals.

The experiments planned until the end of the project will define the rules for how dendrites behave as fundamental computational units in the nervous system. They will reveal crucial new insights about integration and plasticity rules in dendrites, and define how the circuit elements harness these rules, and ultimately show their engagement in vivo during behavior. What will be the wider impact of this knowledge? First, we will develop a quantitative understanding of dendritic computation and of the underlying mechanisms, allowing us to make simplified models of single neurons which capture the essence of the dendritic computation. These models can in turn be used to build biologically more realistic larger-scale network models. Second, we will reveal how different elements of the local neural circuit map onto the dendritic tree, and in turn engage dendritic nonlinearities, which will provide important new information about how neural circuits are organized, and also provide important constraints for morphological plasticity and provide clues towards developmental mechanisms. Finally, a causal demonstration of how dendritic mechanisms are useful for behavioural performance will provide a transformational advance in neuroscience and in our understanding of how single neurons contribute to computations in the brain. Identification of the underlying biophysical and circuit mechanisms, particularly with respect to specific circuit motifs involved in regulation of dendritic excitability, will provide clear targets for manipulation of neuronal processing that may ultimately have translational relevance.
All-optical interrogation of individual neurons