Population signals from neuronal ensembles in cortex during behavior are commonly measured with EEG, local field potential (LFP), and voltage-sensitive dyes. A genetically encoded voltage indicator would be useful for detection of such signals in specific cell types. Here we describe how this goal can be achieved with Butterfly, a voltage-sensitive fluorescent protein (VSFP) with a subthreshold detection range and enhancements designed for voltage imaging from single neurons to brain in vivo. VSFP-Butterfly showed reliable membrane targeting, maximum response gain around standard neuronal resting membrane potential, fast kinetics for single-cell synaptic responses, and a high signal-to-noise ratio. Butterfly reports excitatory postsynaptic potentials (EPSPs) in cortical neurons, whisker-evoked responses in barrel cortex, 25-Hz gamma oscillations in hippocampal slices, and 2- to 12-Hz slow waves during brain state modulation in vivo. Our findings demonstrate that cell class-specific voltage imaging is practical with VSFP-Butterfly, and expand the genetic toolbox for the detection of neuronal population dynamics.

The deep cerebellar nuclei (DCN) integrate inputs from the brain stem, the inferior olive, and the spinal cord with Purkinje cell output from cerebellar cortex and provide the major output of the cerebellum. Despite their crucial function in motor control and learning, the various populations of neurons in the DCN are poorly defined and characterized. Importantly, differences in electrophysiological properties between glutamatergic and GABAergic cells of the DCN have been largely elusive. Here, we used glutamate decarboxylase (GAD) 67-green fluorescent protein (GFP) knock-in mice to unambiguously identify GABAergic (GAD-positive) and non-GABAergic (GAD-negative, most likely glutamatergic) neurons of the DCN. Morphological analysis of DCN neurons patch-clamped with biocytin-containing electrodes revealed a significant overlap in the distributions of the soma sizes of GAD-positive and GAD-negative cells. Compared with GAD-negative DCN neurons, GAD-positive DCN neurons fire broader action potentials, display stronger frequency accommodation, and do not reach as high firing frequencies during depolarizing current injections. Furthermore, GAD-positive cells display slower spontaneous firing rates and have a more shallow frequency-to-current relationship than the GAD-negative cells but exhibit a longer-lasting rebound depolarization and associated spiking after a transient hyperpolarization. In contrast to the rather homogeneous population of GAD-positive cells, the GAD-negative cells were found to consist of two distinct populations as defined by cell size and electrophysiological features. We conclude that GABAergic DCN neurons are specialized to convey phasic spike rate information, whereas tonic spike rate is more faithfully relayed by the large non-GABAergic cells.