Methods

Two photon calcium imaging

We use two-photon calcium imaging as the principal method in our group. This approach permits the recording of neuronal activity in defined neuronal populations in living animals across days. Two-photon calcium imaging enables us to obverse learning-related neuronal activity in behaving mice and to study the stability of neuronal activity patterns.

Two-photon targeted cell-attached recordings

Cell-attached recordings is a variant of patch-clamp recordings that we employ to listen to neuronal activity of individual and visually identified neurons. We use this approach to calibrate calcium imaging recordings with simultaneously recorded electrophysiology. These technically challenging calibration experiments enable us to better interpret signals obtained with calcium imaging.

Light-sheet microscopy

We take advantage of the ZMB at the University of Zurich to clear entire mouse brains and  to image the distribution of cells and their processes and projections throughout the brain using mesoSPIM light-sheet microscopes.

Optogenetics

In addition to the recording of neural activity, we also use several methods to perturb brain circuits, using pharmacology or optogenetics. For example, we have previously used, in collaboration with the Bohacek Lab, optogenetic stimulation of the locus coeruleus together with two-photon imaging of the hippocampus.

Mouse behavior

We use spontaneous and task-guided mouse behavior in a head-fixed setting to study the neural processes underlying the respective behavior. We typically spend a lot of time with few individual mice in order to make them learn what we want them to do. The behavior of mice is variable and often predictable, and we embrace the challenge of understanding the often unexpected diversity of the spontaneous behaviors and the behavioral strategies that we observe.

Data analysis and deep-learning

As a rule of thumb, we spend more time with the analysis of the experiments compared with the amount of time for their planning and execution. Therefore, every member of the Rupprecht group should be or should become highly proficient in Python and/or Matlab. Among other analysis methods that include correlation functions and modeling, we also use deep networks to analyze our data. For example, we have developed deep learning methods for spike inference and have used deep networks for 3D image analysis or denoising.