Circuits of neuronal network that compute information in real-time span across the brain tissue in 3D. To understand the computational mechanisms of the brain, we need methods that read out neural activity in large cell populations simultaneously, on both the somatic and dendritic scales. Following signal propagation and integration across a single neuron and recording the activity of hundreds of neurons pose challenges that not many imaging systems has been designed to tackle at once. To overcome this problem, we developed a high-resolution, acousto-optic (AO) two-photon microscope with continuous 3D trajectory and random-access scanning modes, that reaches a near cubic millimeter scan range, and can be adapted to imaging different spatial scales. With our technology, neuronal activity can be followed even in behaving animals in large scanning volumes, with a measurement speed and signal-to noise ratio (SNR) increased by several orders of magnitude. This feat, in combination with the exceptional penetration depth characteristic of two-photon imaging, makes our methodology very convenient for in vivo measurements of neuronal populations.

Katona et al. 2012 Nature Methods, Szalay et al. 2016 Neuron





The activity of neuronal populations has long been studied in the visual cortex, but fast 3D volumetric random-access scanning of hundreds of in hundreds of cells in the visual cortex in vivo has been made possible by using a high-resolution, acousto-optic two-photon microscope. We conduct in vivo two-photon imaging of cell assemblies in the V1 area in behaving mice by using multicell bolus loading of a calcium indicator dye. Neuronal network responses are recorded during different types of visual stimulation (moving bar, moving grating or visual discrimination protocol). In addition, active cells are selected based on the previously recorded somatic activity and their dendritic responses are followed along with the network activity in 3D by using whole-cell patch clamp techniques.

Katona et al. 2012 Nature Methods





Two-photon uncaging of neurotransmitter molecules is the method of choice to mimic and study the subtleties of neuronal communication either in the intact brain or in slice preparations. It has the advantage of high spatial and temporal resolution of two-photon excitation to study dendritic integration. Two-photon uncaging can also be used to map receptor densities (e.g. for GABA receptors) even in 3D. Used in combination with two-photon imaging, two-photon uncaging provides an opportunity to study the long-term structural and functional consequences of stimulation of structures such as dendritic spikes and dendritic spines. As currently available caged materials have several drawbacks, we used quantum chemical modelling to show the mechanisms of hydrolysis and two-photon activation and synthesized more effective caged compounds. We have also developed a new enzymatic elimination method removing neurotransmitters inadvertently escaping from their compound during experiment. This method, usable both in one- and two-photon experiments, allows for the use of materials with an increased rate of photochemical release. The efficiency of the new compound and the enzymatic method and of the new compound are demonstrated in neurophysiological experiments.

Pálfi et al. 2018 Organic & Biomolecular Chemistry





Sharp-wave ripples (SPW-Rs) are rapid bursts of synchronized neuronal activity elicited by the hippocampus. They are widely thought to play a critical role in the consolidation of episodic memory and are associated with the reactivation of neuronal ensembles within specific circuits during memory formation. Fast-spiking, parvalbumin-expressing interneurons (FS-PV INs) are thought to provide fast integration in these oscillatory circuits by suppressing regenerative activity in their dendrites. We used fast 3D AO two-photon imaging and a caged glutamate to challenge this classical view of the functional role of interneurons, by demonstrating a network activity-dependent dynamic switch in dendritic integration mode: that FS-PV IN dendrites can generate propagating Ca2+ spikes during SPW-Rs. The spikes originated from dendritic hot spots and were mediated dominantly by L-type Ca2+ channels. Ca2+ spikes were associated with intrinsically generated membrane potential oscillations. These oscillations required the activation of voltage-gated Na+ channels, had the same frequency as the field potential oscillations associated with SPW-Rs, and controlled the phase of action potentials. We also demonstrated that the smallest functional unit that can generate ripple-frequency oscillations is a segment of a dendrite. The main difficulty in imaging hot spot activity in complex dendritic arbors is the inadequate temporal and spatial resolution of currently available imaging technologies: our fast 3D scanning methods overcome these limitations, by providing a high temporal resolution up to tens of microseconds, which allows simultaneous measurement of even the fastest regenerative events in multiple dendritic segments of the thin distal dendritic arborization, with a high spatial discretization on the size scale of dendritic hot spots during SPW-Rs.

Chiovini et al. 2014 Neuron





Continuing our work on the 3D AO scanning technique, that can simultaneously read out neural activity on the somatic and dendritic scales, we have developed a novel method, 3D DRIFT AO microscopy. By this, we can drift the excitation spot quickly in any direction in 3D, while continuously recording fluorescence data. Therefore, we can extend the pre-selected individual scanning points to small 3D lines, surfaces, or volume elements to cover ROIs along with neighboring background areas or volume elements. In this way, we can preserve fluorescent information for motion artefact elimination, and, hence, to record spines, dendrites and networks form the moving brain of behaving animals. 3D DRIFT AO microscopy makes it possible to functionally image and stimulate large volumes of neural tissue: entire cortical columns, multiple separate cortical areas, or larger areas. Neuronal networks at multiple spatial scales, from small structures to the level of large neuronal assemblies (over 5000 neurons) can be now simultaneously imaged and activated in the V1 region and in higher order visual regions with high spatial (<500nm, in the center) and temporal (>50kHz/ROI) resolution in several cubic millimeter volumes in 3D. This method can increase measurement speed and signal collection efficiency by several orders of magnitude (up to >10,000,000-fold; Szalay et al. 2016 Neuron, Chiovini et al. 2014 Neuron). 





Canonical neural circuit motifs are assemblies of connected cell types that are repeated across areas. Recent studies identified a cortical circuit motif controlled by VIP interneurons, which preferentially inhibit other interneurons and thereby disinhibit principal neurons. As a previous work showed that auditory cortex VIP interneurons respond not only to sensory stimuli but also to reinforcement feedback, we hypothesized that VIP interneurons may transduce global reinforcement signals for local computation. By using 3D random-access two-photon imaging of VIP neurons across the dorsal cortex while mice performed an auditory discrimination task, we found that most VIP neurons across the cortex were robustly activated by reward and punishment and tended to be activated together. Our results reveal that VIP interneurons have multiple response modes with both local and global contributions. We propose that by signaling reinforcement events, VIP interneurons form a new information channel for orchestrating cortex-wide learning mechanisms.





By understanding how functional neuronal assemblies relate to subjective perceptions and behavior, and eventually finding and reactivate these assemblies in a precise, biologically relevant manner, an “artificial sense”, a sensory (visual) prosthetic could be created. In this, a similar cortical activation would be elicited as during a classical visual stimulation. Functional cortical connectivity can be mapped precisely and in a large volume by two-photon AO microscopy. In this project, that has received fund from the European Research Council (ERC), we will scan neuronal activity with high speed and simultaneously photoactivate neurons with high efficiency and subcellular precision in the entire V1 region of the cortex. Using our microscope in combination with novel caged neurotransmitters and optogenetic tools, we want to map cell assemblies and to understand how they form larger clusters and how they are associated with visual features. We aim to restore visual perception by 3D artificial photostimulation in behaving mice.