Assessing Local and Branch-specific Activity in Dendrites

Abstract

Dendrites are elaborate neural processes which integrate inputs from various sources in space and time. While decades of work have suggested an independent role for dendrites in driving nonlinear computations for the cell, only recently have technological advances enabled us to capture the variety of activity in dendrites and their coupling dynamics with the soma. Under certain circumstances, activity generated in a given dendritic branch remains isolated, such that the soma or even sister dendrites are not privy to these localized signals. Such branch-specific activity could radically increase the capacity and flexibility of coding for the cell as a whole. Here, we discuss these forms of localized and branch-specific activity, their functional relevance in plasticity and behavior, and their supporting biophysical and circuit-level mechanisms. We conclude by showcasing electrical and optical approaches in hippocampal area CA3, using original experimental data to discuss experimental and analytical methodology and key considerations to take when investigating the functional relevance of independent dendritic activity.

Keywords: GABAergic circuitry; blind dendritic patch clamp electrophysiology; branch specific activity; dendritic spikes; feature tuning; two-photon imaging.

Figure 1.. Probing dendritic activity

A-B. Sample reconstructed morphology (left), corresponding individual (middle), and average (right) post-synaptic voltage responses to stratum radiatum (SR) input stimulation obtained with patch clamp recordings from CA3 pyramidal neurons in acute slices. Shaded area represents SEM. A. Somatic recording showing subthreshold post-synaptic potentials (PSP, black) and action potentials (purple). B. Dendritic recording showing subthreshold PSPs (black), simple (green), and complex (red) spikes. Note that only a partial reconstruction was possible from the dendritic fill. C-E. Sample field of view and calcium events recorded from a GCaMP-expressing CA3 pyramidal neuron in vivo in head-fixed behaving mice. C. Field of view showing the imaged neuron and selected regions of interest (ROI). D. Top, calcium ΔF/F traces from selected ROIs showing global (left) and localized (right) activity. E. Expanded views of the time windows demarcated above.

Figure 2.. Hypothetical branch-specific activity patterns

A. Schematic representation of branch-specific activity. In this hypothetical scenario, different dendritic branches are tuned to different parts of the feature space, with one branch (d2) sharing tuning with the soma. B. Graphs showing artificial readings of activity in the different compartments noted in A.

Figure 3.. Propagation of dendritic depolarizations

A-E. Schematics of dendritic depolarization propagation in different scenarios (PSP = post-synaptic potential; exc = excitatory input; inh = inhibitory input). A. Passive attenuation. B. Inhibitory gating. C. GABAergic disinhibition. D. Nonlinear summation. E. Back-propagation.

Figure 4.. Key elements of the dendritic blind patch clamp recording procedure

A. Photograph of the agar mold used to hold hippocampi or cortico-hippocampal complexes for transverse slicing; edges are outlined in red. B. Photograph of the tissue inside the agar mold, blue and green outlines show cortex and hippocampus respectively, dotted lines denote anatomical axes. C. Photograph of the tissue and agar block inside the vibratome chamber. D. Bright field images of serial sections along the septo-temporal axis, the purple outline indicates transverse slices suitable for recordings. E. Graphs showing artificial readings of micromanipulator motion, pipette pressure, voltage command, current, and resistance. F. Graphs showing artificial readings of the current response to a voltage pulse (top) and schematics of the pipette approaching the dendrites (bottom) during the different phases of dendritic patch clamping denoted in E (a, approach; b, contact; c, seal; d, break in).

Figure 5.. Dendritic responses to laminar input stimulation.

A. Images of an acute hippocampal slice captured during dendritic patch clamp experiments. Top, bright field DIC image with stimulation (SLM, green; SR, red; SL, blue) and recording (black) pipette locations and hippocampal strata (purple) outlined. Middle, epifluorescence image showing the recording pipette loaded with Alexa 594. Bottom, higher magnification epifluorescence image corresponding to the inset demarcated above. Due to of the deep location within the tissue, note that the fill of the apical dendrite and soma is faint but discernable (yellow arrows). B. Schematic representation of layered inputs targeting CA3 pyramidal neuron and location of the dendritic recording. C. Top, sample trace of action potentials evoked by current step injection. Bottom, higher magnification of the time window demarcated above showing a doublet spike. D. Sample traces of SLM (green), SR (red) and SL (blue) input stimulation-evoked PSPs. Note the compound PSP evoked by SR stimulation which displays an initial depolarization followed by a hyperpolarization. E. Summary graph of PSP amplitude recorded in CA3 pyramidal neuron dendrites in response to SL, SR and SLM stimulation (individual data shown as thin lines, population averages shown as thick lines, error bars represent SEM, n = 11; Friedman ANOVA, p = 0.011; Wilcoxon signed-rank tests: SL vs SR, p = 0.005; SR vs SLM, p = 0.007; SL vs SLM, p = 0.831).

Figure 6.. Example fields of view with variable expression densities

A-D. Sample in vivo fields of view in CA3 of mice expressing “sparse” (A), “light” (B), “medium” (C) and “dense” (D) levels of GCaMP. Each image is a maximum projection image of a 10 minute long recording session. Images have been rotated and/or flipped such that the apical dendrites are pointed upwards and to the right (red arrows). Table. Titers of virus used in mixtures to achieve the displayed results. Columns A and B used a mixture of AAV2.1-CaMKII-Cre and AAV2.1-Syn-FLEX_GCaMP6f. Column C used a mixture of AAV2.1-CaMKII-Cre and AAV2.1-Syn-FLEX-jGCaMP7b. Column D used AAV2.1-CaMKII0GCaMP6f. The first two rows indicate the titer used for each individual virus. AAV2.1-CaMKII-Cre was diluted from stock titer of 2.71x10¹³ with ACSF. The volume ratio of the two viruses is indicated in the third row. The final row indicates the titer of the viruses if they were mixed in a 1:1 volume ratio.

Figure 7.. Correcting for Z motion with a static structural label

A. Left, maximum projection image of the green channel in a mouse co-expressing GCaMP7b and tdTomato (Ai14 homozygous mouse injected with virus mixture from column C of Figure 6). Two regions of interest from neighboring sections of a dendrite are indicated. Middle, extracted fluorescence traces from ROI 1 and 2. Right, if taken at face value these adjacent segments are uncorrelated despite their close proximity. B. Same as A but for the red channel. Notice that the red signals, which should be static, fluctuate by large amounts and are anti-correlated with each other, indicative of motion in the Z plane. C. Left, merge of the green and red channels. Middle, extracted green/red fluorescence ratio for ROI 1 and 2. Note the relatively constant baseline in comparison to the traces in A. Right, when the dendritic signals are estimated as the green/red ratio, they exhibit a strong positive correlation.

Figure 8.. ROI labeling and cross-talk

A. Illustration of ROI labeling process to enable connecting dendrites to parent soma in dense data sets. Top row, an experimenter goes through the recorded video frame by frame to find instances of clear activation of an isolated cell body and connected dendrites. Middle row, the boundaries defining that soma and its connected dendrites are manually annotated, to define regions of interest (ROIs). Bottom row, boundaries identified from individual frames are plotted over the maximum intensity projection to identify neurons still to be labeled. This process continues until no more isolated neurons can be identified. B. Top, highlighted neurons with overlapping dendritic ROIs (different mouse than in A). Most of the pixels of dendrites D1b and D2a are overlapping, complicating the estimation of their temporal activity. Bottom, estimated ΔF/F traces from the above-labeled ROIs. In this estimate, dendrite D1b has 3 calcium events (asterisks) that are not visible in the parent soma S1 or sister dendrite D1a, which could erroneously be interpreted as branch-specific activity. Attributing these events to D1b eliminates them from the estimated activity of D2a (open circles).