Functional and Structural Properties of Highly Responsive Somatosensory Neurons in Mouse Barrel Cortex

Abstract

Sparse population activity is a well-known feature of supragranular sensory neurons in neocortex. The mechanisms underlying sparseness are not well understood because a direct link between the neurons activated in vivo, and their cellular properties investigated in vitro has been missing. We used two-photon calcium imaging to identify a subset of neurons in layer L2/3 (L2/3) of mouse primary somatosensory cortex that are highly active following principal whisker vibrotactile stimulation. These high responders (HRs) were then tagged using photoconvertible green fluorescent protein for subsequent targeting in the brain slice using intracellular patch-clamp recordings and biocytin staining. This approach allowed us to investigate the structural and functional properties of HRs that distinguish them from less active control cells. Compared to less responsive L2/3 neurons, HRs displayed increased levels of stimulus-evoked and spontaneous activity, elevated noise and spontaneous pairwise correlations, and stronger coupling to the population response. Intrinsic excitability was reduced in HRs, while we found no evidence for differences in other electrophysiological and morphological parameters. Thus, the choice of which neurons participate in stimulus encoding may be determined largely by network connectivity rather than by cellular structure and function.

Keywords: barrel cortex; patch clamp; somatosensory; sparse coding; two-photon imaging.

Figure 1

Combined in vivo and in vitro approach for the functional and structural characterization of neurons with known stimulus–response properties. The method comprised four main experimental steps (A-D): (A) Intrinsic imaging for localizing whisker-evoked response and injection site, followed by injections of two viruses and a chronic window implant. The viruses comprised a red calcium indicator (jRGECO1a or RCaMP2) and a photo-activatable green fluorescent protein (H2B-PAGFP). (B) Two-photon imaging of whisker-evoked responses and neuronal tagging by light-activation of H2B-PAGFP-postive neurons. (C) Intracellular recordings of neurons tagged in vivo and simultaneous dye injection (biocytin). (D) Neuronal staining and reconstruction of neuronal morphology.

Figure 2

In vivo whisker stimulation, two-photon imaging and neuronal tagging. (A) Intrinsic imaging was performed prior to each two-photon imaging session to identify barrel location (white circle). Vasculature under green light illumination (left) and change in reflected red light RD/R0 during single whisker stimulation (right). (B) Mice (n = 35) were kept under light isoflurane anesthesia during two-photon imaging in barrel cortex L2/3 (green square; left) and piezo-electric stimulation of single whiskers at 90 Hz (right). (C) Co-expression of red calcium indicator for two-photon imaging (here jRGECO1a) and photo-activatable green fluorescent protein (H2B-PAGFP) for photo-tagging cells. Overlay of green and red channels before (top) and after photo-activation of H2B-PAGFP (bottom). (D) Single-neuron calcium transients Δf/f0 (black) in response to whisker stimulation (blue). HRs #1–7 showed large average responses (red curves), while LRs #8–14 showed small average responses (blue curves). (E-G) Criteria for defining HRs (red) and LRs (blue): Peak calcium signals (Δf/f0) averaged across all trials (E), fraction of trials with significant activation (F), and response latency (G). In (G), only neurons for which the response latency could be calculated were included (100% of all HRs, 99% of MR and 65% of LRs). Response latencies take discrete values from the sampling time of calcium imaging at our sampling rate (11.4 Hz). Neurons that did not match these criteria were categorized as medium responders.

Figure 3

A gradient of neuronal activity in response to whisker stimulation. We recorded from 193 HRs, 608 MRs and 1299 LRs in 21 FOVs from 7 mice. (A) Neuronal calcium responses during trials, ranked according to their average z-scores. Top row identifies neurons as HR (red), MR (gray) or LR (blue). (B) Response amplitudes of HRs (left) and LRs (right). (C) Fraction of active trials as a function of the stimulus-evoked peak amplitude (left) and the response latency (right). Only neurons for which the response latency could be calculated were included (right panel; 100% of all HRs, 99% of MR and 65% of LRs). Each dot represents one neuron. (D) Discrete events were extracted from continuous two-photon recordings using a deconvolution algorithm. Shown are z-scored events during spontaneous (top) and stimulus-evoked activity (bottom) for a representative set of 100 neurons recorded in one FOV. Each arrow indicates stimulus onset. (E) Left-to-right: Average distribution of HRs (red), MRs (gray) and LRs (blue) in the population (n = 21 FOVs), probability of calcium events during stimulation (69 ± 10 trials per FOV, mean ± SD) and during spontaneous activity (262 ± 199 s per FOV, mean ± SD)). Statistical comparisons were performed using Friedman test with Dunn–Sidàk’s correction. (F) Average event rates across cell classes during stimulus-evoked and spontaneous activity (n = 21 FOVs, Friedman test with Dunn–Sidàk’s correction). (G) Relationship between the frequency of stimulus-induced and spontaneous calcium events across all neurons (Spearman rank correlation). Trend line (black) based on least squares fit. P-values: ^* < 0.05; ^** < 0.01; ^*** < 0.001.

Figure 4

Network-dependent response profiles during stimulation and spontaneous activity. Data comprised 193 HRs, 608 MRs and 1299 LRs measured in 21 FOVs from 7 mice. Neuronal activity was analyzed with respect to changes in the fraction of active neurons (“ensemble size”, A-C), correlations among cell classes (D-J), and the degree of coupling to the population activity (K-P). (A) Probability of ensemble activation (% of neural population activated by stimulus) under shuffled (dashed line) and experimental conditions (continuous line). For each neuron, activity was shuffled across trials while keeping the total activity constant. Data (n = 1454 trials) were pooled across FOVs. Error shades represent shuffled data from 1^st to 99^th percentile. (B) Stimulus-induced changes in ensemble composition (top) and sub-ensemble recruitment (bottom) as a function of responding ensemble size. (C) Changes in response amplitude of different cell classes as a function of the size of the responding ensemble. The response amplitude was defined as the area under the curve of the z-scored calcium transient. The ensemble size was normalized by the maximum ensemble size per FOV across trials. Inset: Slope of the linear regression of the amplitude on the responding ensemble size (Friedman test with Dunn–Sidàk’s correction). (D-F) Pairwise correlations during stimulation. Correlations were quantified using Pearson r, which was averaged across 69 ± 10 trials (mean ± SD) separately for each FOV (n = 21 FOVs). Shown are the average r between pairs of neurons during stimulation across FOVs (D), the corresponding boxplots with the median (midline), interquartile ranges and 2.7 SD (E) (Friedman test with Dunn–Sidàk’s correction), and the cumulative probability distributions of single pairwise correlations with n = 1135 HRs/HRs, n = 9498 MRs/MRs, n = 41 200 LRs/LRs (F). (G-J) Pairwise correlations during spontaneous activity. Pearson r was calculated based on 258 ± 199 s (mean ± SD) of spontaneous activity. Displayed are the average r between pairs of neurons during baseline across FOVs (G), the corresponding boxplots (H), and the cumulative probability distributions of single pairwise correlations (I). Average spontaneous correlations for neuronal pairs (n = 891 HRs/HRs, n = 891 MRs/MRs, n = 891 LRs/LRs) with matched geometric mean firing rates (GMFR > 0.05 Hz) are shown in J (Kruskall–Wallis test with Dunn–Sidàk’s correction). P-values: ^* < 0.05; ^** < 0.01; ^*** < 0.001. (K) Population coupling as a function of response probability during stimulation. Across all neurons (n = 2100), population coupling and response probability were positively correlated (Spearman rank correlation). (L) A model based on population coupling predicts pairwise correlations similar to those measured. Shown are the pairwise correlations for all pairs of neurons in an example FOV, with a comparison of the predicted (upper left triangle) and the experimentally observed pairwise correlations (lower right triangle). Neurons are sorted by increasing population coupling. (M) Cumulative distribution of residual correlations during spontaneous activity in HRs, MRs and LRs with matching GMFR above 0.05 Hz. (n = 623 HRs/HRs, n = 623 MRs/MRs, n = 623 LRs/LRs, (P < 0.001 HRs versus LRs, Mann–Whitney U test). (N) Average residual correlations during stimulation (top) and spontaneous activity (bottom) for all neuronal pairs (n = 1135 HRs/HRs, n = 9498 MRs/MRs, n = 41 200 LRs/LRs, Kruskall–Wallis test with Dunn–Sidàk’s correction). Error bars represent 95% confidence intervals (C, J, N).

Figure 5

Electrophysiological and morphological characterization of an example high responder (HR) and an example low responder (LR) in barrel cortex L2/3. (A) Example of a patched HR. Thalamocortical slice and pipette position under infrared-differential interference contrast (IR-DIC; A1) to visualize barrels (white), and under fluorescent illumination to visualize H2B-PAGFP-labeled cells (A2). HRs previously tagged in vivo appear brighter than non-tagged neurons (A2, inset). Fluorescent micrograph (A3) of patched neuron with H2B-PAGFP-labeled nucleus (green circle) and the biocytin-filled patch pipette (white line). (B) Example of a patched LR. Thalamocortical slice (IR-DIC, B1) with brightly tagged LRs (H2B-PAGFP; B2, B3). Same conventions as in A. (C,D) Minimum current required to elicit spiking in HR (160 pA; C) and LR (100 pA; D). (E,F) Sustained firing pattern in HR (330 pA; E) and LR (205 pA; F). Insets display the first three spikes from the trace (gray shading). (G,H) Number of spikes fired as a function of current injected in HR (G) and LR (H). Data points are indicated by crosses and the fitting curve by a gray line. (I,J) Histological processing and morphological reconstruction of biocytin-filled HR (I) and LR (J). Left panel: Micrograph of neuron (top) with enlarged view of apical spines (AS; bottom left) and basal spines (BS; bottom right; scale bar: 10 μm). Right panel: 3D neuronal reconstruction with dendrites (HR: red, LR: blue), axons (HR: orange, LR: cyan), layer borders in gray and barrels in black.

Figure 6

Electrophysiological parameters of neuronal excitability. HR neurons (n = 16 cells, 12 mice) and LR neurons (n = 12 cells, 9 mice) were compared in terms of firing frequency per 100 pA injected, input resistance (R_input), resting membrane potential (V_rest), rheobase current, spike threshold, and first inter-interspike interval (ISI₁ in a 10-spike train). Statistical comparisons were performed using a Mann–Whitney U test. Box plots show the median (central horizontal bar), mean (+), as well as the first and third quartiles (lower and upper limits).

Figure 7

Spontaneous excitatory postsynaptic currents (sEPSCs) in HRs and LRs. (A) Left panel: HR example raw trace with sEPSCs (market by asterisks). Right panel: Enlarged view of example sEPSC (marked by gray asterisk in raw trace) with the 20–80% rise time indicated by green crosses and the peak indicated by a gray bar. (B) LR example raw trace (left) and enlarged sEPSC (right). Same conventions as in A. (C) Scattergrams of sEPSC amplitudes as a function of inter-event intervals (left) and rise times (right) with respective marginal histograms (50 bins). (D) Comparison of amplitude (left), inter-event interval (middle) and 20–80% rise time (right) between HR sEPSCs (n = 1200; 8 cells in 5 mice) and LR sEPSCs (n = 1200; 8 cells in 6 mice) using Mann–Whitney U tests. 150 randomly selected events were included from each neuron and pooled for each group. Same conventions for box plots as in Figure 6.

Figure 8

Dendritic morphology of excitatory HRs and LRs. Pyramidal HR neurons (red; n = 19 cells, 15 mice) and LR neurons (blue; n = 10 cells, 7 mice) could be subdivided into three categories based on the structure of the apical dendrite: horizontal, thin-tufted and broad-tufted. (A) Pyramidal neurons with a horizontally oriented apical dendrite. Layer borders are indicated by gray dashed lines. (B) Pyramidal neurons with a thin-tufted apical dendrite. (C) Pyramidal neurons with a thick-tufted apical dendrite. There is no difference of this sub-type proportion between the two groups (P = 0.43, Fisher exact test). (D) Subpial depth of reconstructed cells. Left-to-right: Infrared-DIC micrograph of layers 1–3, fluorescent image of tagged cells in layers 1–3, and depth distribution of cells included in morphological analysis (H: horizontal, B: broad-tufted, T: thin-tufted apical dendrites).