Voltage imaging to understand connections and functions of neuronal circuits

Abstract

Understanding of the cellular mechanisms underlying brain functions such as cognition and emotions requires monitoring of membrane voltage at the cellular, circuit, and system levels. Seminal voltage-sensitive dye and calcium-sensitive dye imaging studies have demonstrated parallel detection of electrical activity across populations of interconnected neurons in a variety of preparations. A game-changing advance made in recent years has been the conceptualization and development of optogenetic tools, including genetically encoded indicators of voltage (GEVIs) or calcium (GECIs) and genetically encoded light-gated ion channels (actuators, e.g., channelrhodopsin2). Compared with low-molecular-weight calcium and voltage indicators (dyes), the optogenetic imaging approaches are 1) cell type specific, 2) less invasive, 3) able to relate activity and anatomy, and 4) facilitate long-term recordings of individual cells' activities over weeks, thereby allowing direct monitoring of the emergence of learned behaviors and underlying circuit mechanisms. We highlight the potential of novel approaches based on GEVIs and compare those to calcium imaging approaches. We also discuss how novel approaches based on GEVIs (and GECIs) coupled with genetically encoded actuators will promote progress in our knowledge of brain circuits and systems.

Keywords: GECI; GEVI; membrane voltage; neurophysiology; optical imaging.

Fig. 1.

Depolarization envelope. A: neuronal membrane potential oscillates between DOWN and UP states in response to glutamatergic input via microiontophoresis (ex vivo) (Milojkovic et al. 2004) or physiologically in vivo (Steriade et al. 2001). B: the same kind of synaptic input is ineffective in the DOWN state but very effective during the UP state (generation of APs). C: 3 examples of published GEVI imaging illustrating reliable reporting of the slow component (depolarization envelope) of the membrane potential transients during underlying bursts of APs. Examples are taken from an early GEVI, VSFP2.3 (from Akemann et al. 2010 with permission; Lundby et al. 2008), a further evolved GEVI, ArcLight Q175 (from Han et al. 2013 with permission; Jin et al. 2012), and a more recently developed GEVI, Ace2N-2AA-mNeon (from Gong et al. 2015 with permission).

Fig. 2.

Overview of GEVI molecular designs: voltage-sensing domain-based and opsin-based GEVIs. Each of A–G displays a schematic depiction of 1 GEVI type with conformation at resting membrane voltage and after membrane depolarization along with an example optical recording trace obtained with a leading representative. A: FRET-based voltage-sensitive probes of the VSFP1/2 type where a pair of fluorescent proteins (FP,D: FRET donor; FP,A: FRET acceptor) is attached to a voltage-sensor domain, consisting of four segments (S1–S4). Depolarization of membrane voltage induces a rearrangement of the 2 fluorescent proteins with an increase in FRET (black arrow) that is optically reported as a change in the ratio between donor and acceptor fluorescence. Example trace taken from Lam et al. (2012) with permission. B and C: single fluorescent protein and circularly permuted (cp) fluorescent protein probes of the VSFP3 family. Examples are ArcLight and FlicR1 (traces taken from Abdelfattah et al. 2016 with permission). D: in FRET-based voltage-sensitive probes of the VSFP-Butterfly family the voltage-sensor domain is sandwiched between two fluorescent proteins. FRET mechanism as in A; example trace for Nabi2.244 taken from Sung et al. (2015) with permission. E: in ASAPs a single cp fluorescent protein is inserted between the 3rd and 4th transmembrane voltage-sensing domains. Example trace taken from St-Pierre et al. (2014) with permission. F and G: in GEVIs based on opsins a change in membrane potential induces increased absorption of red light by the retinal chromophore. This effect is read out either via retinal fluorescence (F) or via quenching of a FP donor (G). Examples are Quasars (from Hochbaum et al. 2014 with permission) and Ace (from Gong et al. 2015 with permission). Also shown is a list of selected representatives for each group.

Fig. 3.

Voltage imaging of dendritic integration. A: compound photograph of a neuron; “stim.” marks position of the synaptic stimulation electrode. B: low-resolution image of the part of the dendritic tree where optical signals were recorded. Regions of interest (ROIs) are marked by numbers. C: simultaneous electrical (ROI 1) and optical (ROIs 2–13) recordings of dendritic membrane potential changes upon a single synaptic stimulation shock (stim.). D: schematic representation of voltage waveforms along a basal dendrite receiving excitatory synaptic input in distal segment (green) and inhibitory synaptic input near the soma (blue). E₁: basilar dendritic tree of a pyramidal neuron. E₂: schematic representation of the basilar dendritic tree. Dendrites 1 and 2 receive inputs from a “blue” network of neurons. Dendrites 3 and 4 receive inputs from a “red” network of neurons. E₃: dendritic voltage waveforms explain the origin of somatic and axonal electrical responses.

Fig. 4.

Schematic of how GEVIs can interrogate circuits at multiple levels—using a GEVI expressed specifically in L2/3 cortical neurons as an example. A: single-cell voltage signals from genetically identified L2/3 neurons to assess how these can be integrated across the dendrite to generate output signals at the soma and axon. B: GEVI-based single-cell and pathway-specific local network interrogation using optical (e.g., using ChR2) or electrical (electrophysiological) stimulation of inputs. L5 output neurons are not labeled. Synaptically connected L2/3 neurons integrate their subthreshold inputs and provide a spatial map of connections within the locally activated network. C: simplified illustration of how GEVIs can report longer-range activity from, e.g., 2 different thalamocortical inputs (1 and 2) to L2/3 of motor cortex with hypothetical different levels of thalamocortical synaptic input (left and right hemispheres). Input 1 leads to a larger GEVI signal than input 2 to illustrate the concept of how the size of the GEVI signal represents the level of connectivity at the input. Also shown is the concept of spread of subthreshold activity via recruitment of horizontal monosynaptic connections across cortical boundaries. D: simplified systems-level example to explain the concept of how GEVI expressed in L2/3 neurons (right image) can generate an “input map” in vivo (left image). For example, a visually evoked movement could first generate an input map in L2/3 neurons of visual cortex (V) that propagates via unseen output (L5, to other cortical and subcortical structures) to synaptically activate motor cortex (M) input, seen by L2/3 neurons, to then drive L5 motor output.

Fig. 5.

Cortex-wide voltage imaging. A, top: schematic dorsal view of the mouse head. Bottom: transcranial mCitrine fluorescence image of the left cortical hemisphere in a mouse expressing VSFP butterfly (Akemann et al. 2012). The imaged cortical region is indicated as a dotted rectangle (cyan) in the schematic at top; M, S, and V indicate motor, somatosensory, and visual cortical regions. Donor (mCitrine) and acceptor (mKate2) signals were acquired with 2 synchronized CCD cameras at 50 frames/s in 4-s-lasting trials during which either a visual stimulus (5 ms light flash, B) or a single deflection of the contralateral C1 whisker (C) was delivered. B and C: peristimulus, single-trial sequence of ΔR/R₀ (indicating changes in membrane voltage) images spatially filtered with a 2 × 2 Gauss kernel. The first image (grayscale) in B shows the mCitrine baseline image with pixels outside of the optical window masked. Note the spread of neuronal activation from sensory to motor cortices. D, top: time courses of 25 consecutive trials of signals over regions of interest selected within somatosensory and motor areas (black and red circles in A). Time window of the response to whisker deflection is indicated by dotted rectangle. Depolarizing events outside of this time window are spontaneous. Bottom: average of the evoked S1 and M1 response (25 trials) at enlarged timescale. Note similarity between spontaneous and stimulation-evoked events. Partially redrawn from Akemann et al. (2015) with permission.