Dendritic coincidence detection is thought fundamental to neuronal processing yet remains largely unexplored in awake animals. Specifically, the underlying dendritic voltage-calcium relationship has not been directly addressed. Here, using simultaneous voltage and calcium two-photon imaging of Purkinje neuron spiny dendrites, we show how coincident synaptic inputs and resulting dendritic spikes modulate dendritic calcium signaling during sensory stimulation in awake mice. Sensory stimulation increased the rate of post-synaptic potentials and dendritic calcium spikes evoked by climbing fiber and parallel fiber synaptic input. These inputs are integrated in a time-dependent and non-linear fashion to enhance the sensory evoked dendritic calcium signal. Intrinsic supralinear dendritic mechanisms, including voltage-gated calcium channels and metabotropic glutamate receptors, are recruited cooperatively to expand the dynamic range of sensory evoked dendritic calcium signals. This establishes how dendrites can use multiple interplaying mechanisms to perform coincidence detection, as a fundamental and ongoing feature of dendritic integration in behaving animals.

Approximately 20 hr after ANNINE-6plus labeling, simultaneous imaging from PN dendrites was performed in linescan mode at 2 kHz sampling rate. Labeled PNs were clearly visible in both red and green channels indicating successful labeling with ANNINE-6plus and GCaMP6f. During recordings, mice were alert and headfixed sitting on a rotating treadmill. Mice were allowed to sit awake on the treadmill for at least 1 hr before beginning the experiment. Bidirectional linescans, 512 pixels in width, lasting 10.5 s were performed at a line rate of 2 kHz. To limit photo-damage during linescans and to improve signal-to-noise ratio, the objective collar was rotated to elongate the excitation volume predominantly in the z-direction to ~5 μm. During the 10.5 s, no bleaching was observed. Fine corrections in linescan orientation (with respect to PN dendrites) were done prior to the experiment using the rotating stage, on which the mouse treadmill and micro-manipulators were placed. The linescans measuring 256 μm in width were carefully positioned as superficial as possible as to include the full dendritic width of the most distal dendritic spiny PN branches, thus maximizing the total membrane area covered by the linescan, typically less than 50 μm below the pia mater.

ANNINE-6plus is purely electrochromic, showing linear responses across the full physiological voltage range and is well suited for recording neuronal membrane potential, with a temporal resolution limited only by the fluorescence lifetime (Fromherz et al., 2008). The femtosecond-pulsed Ti:sapphire laser was used to excite fluorescence at 1020 nm, near the red spectral edge of absorption. To confirm optimal ANNINE-6plus sensitivity near the red spectral edge of absorption and the mechanism of voltage sensitivity (Kuhn et al., 2004), different excitation wavelengths were tested (Kuhn and Roome, 2019; Roome and Kuhn, 2018; Roome and Kuhn, 2019). Excitation near the red spectral edge of absorption to optimize voltage sensitivity allows for long-term simultaneous voltage and calcium dendritic recordings at least 500 s per recording session at different dendritic depths. As ANNINE-6plus is relatively hydrophobic compared to other voltage-sensitive dyes for intracellular application, the labeling lasts for at least 2 weeks (Roome and Kuhn, 2018; Roome and Kuhn, 2019; Roome and Kuhn, 2020). Due to an extended excitation point spread function (~1 × 1 × 5 µm³) used to increase the signal-to-noise ratio, the voltage signal is the average membrane potential in this volume encompassing spines and dendritic shafts.

Linescan TIFF images were initially cropped in width using ImageJ (US National Institute of Health) to contain only the dendrite from the labeled PN using the red channel as a guide, and to eliminate green signals originating from neighboring PNs. Full linescan traces were imported into Matlab and interpolated (from 2 to 10 kHz). All subsequent data analysis was performed using custom-written programs that we wrote previously with Matlab (MathWorks). Minor movement corrections were made to linescan images in the spatial dimension, to minimize movement artifacts that occur during the sensory stimulus. This was achieved by maximizing the cross correlation between the average pre-stimulus linescan spatial profile and the spatial profile of individual lines collected during the stimulus. It should be stated that because of the nature of linescan imaging, perfect movement correction cannot be achieved.

Single unit electrophysiological recordings were analyzed using Spike2 spike detection software (CED) to create a binary trace for simple spikes (SS) (and to eliminate spikes originating from neighboring PNs). All complex spikes (occurring at ~1 Hz) were easily identified by eye, which could be confirmed by comparing simultaneous electrophysiology, voltage, and calcium traces. Complex spike binary traces were created using the initial sodium spike of the somatic complex spike to mark its onset. All binary traces, raw images, and electrophysiology were then imported into Matlab for analysis.

To calculate ΔF/F of full linescans, the green channel image was first scaled (using Matlab ‘regress’ function) and fit to the red channel and then subtracted from the red channel. This removed crosstalk from the green channel. Average baseline red channel fluorescence was then subtracted from the time-varying red channel fluorescence and the result was divided by the average baseline of red channel fluorescence to give ΔF/F for the red channel. The ΔF/F calculation for the green channel was made in the same way as for the red channel, but it was not necessary to first subtract red channel fluorescence from the green channel as the voltage signal was neglectable compared to GCaMP6f. Relative fluorescence changes imaged with an excitation wavelength of 1020 nm were converted with a factor of 2.1 mV/% to estimate voltage changes.

The example TIFF images (cell63_1 to cell63_10) show raw linescan data as a stack of two images, with time along horizontal axis, space along vertical axis. Image 1 of each TIFF stack is the dendritic voltage linescan recording, image 2 is the simultaneous dendritic calcium linescan recording.

Two photon linescan imaging (2 kHz) recorded dendritic voltage and calcium signals simultaneously from the spiny dendrites of labeled PNs. Throughout the experiment, fully awake mice (i.e., awake for at least 1 hr after isoflurane anesthesia) were seated on a rotating treadmill and head-fixed under a two-photon microscope.

The raw linescan recordings (10.5 s in length) include a 100 ms sensory stimulus beginning 5.5 s after recording onset, delivering an air puff directed at the ipsilateral eye.