Synthesis of a comprehensive population code for contextual features in the awake sensory cortex

  1. Evan H Lyall
  2. Daniel P Mossing
  3. Scott R Pluta
  4. Yun Wen Chu
  5. Amir Dudai
  6. Hillel Adesnik  Is a corresponding author
  1. Biophysics Graduate Group, United States
  2. Department of Molecular and Cell Biology, United States
  3. The Edmond and Lily Safra Center for Brain Sciences and The Life Sciences Institute, The Hebrew University of Jerusalem, Israel
  4. The Helen Wills Neuroscience Institute, University of California, Berkeley, United States
5 figures, 2 videos and 1 additional file

Figures

Figure 1 with 1 supplement
Probing cortical population codes of higher-order features during active touch.

(A) Schematic of the experimental preparation showing a locomoting mouse under a two-photon microscope actively contacting five individual pistons with five individual whiskers (the rest are trimmed …

Figure 1—figure supplement 1
Whisker tracking during an example trial.

(A) Frames from a high-speed video of the whiskers whisking during a single trial. By properly trimming the whiskers and positioning the pistons, each whisker can only come into contact with a …

Figure 2 with 4 supplements
Diverse and selective representations of higher-order tactile stimuli in mouse barrel cortex.

(A) Example map of principal whisker preference in L2/3 of one mouse. (B) Top: calcium responses from an example neuron to each of the 31 possible combinations of the five pistons (mean and 95% …

Figure 2—figure supplement 1
Maps of single whisker tuning in awake, whisking mice.

(A) Three maps of principal whisker preference for L4 neurons from three Scnn1-Cre mice injected with AAV-syn-flexed-GCaMP6s. (B). Same as a but for three mice recorded in L2/3 using Camk2-tTa;tetO-G…

Figure 2—figure supplement 2
The transform from spiking to fluorescence in GCaMP6s-expressing neurons in Camk2-tTa;tetO-GCaMP6s mice.

(A) Top: trial-averaged instantaneous spike rate of a neuron for trials that contain specific numbers of spikes (mean with s.e.m.). The neuron is recorded in vivo via two-photon targeted loose patch …

Figure 2—figure supplement 3
Whisker kinematics across the stimulus set.

(A) Scatter plots comparing the mean of each indicated kinematic variable during multi-whisker stimuli (y-axis) and the mean across the component single whisker stimuli for each corresponding …

Figure 2—figure supplement 4
Electrophysiological measurement of supralinear summation during multi-whisker contact.

(A, B) Trial-averaged firing rates (composed of many touches per trial) in two example neurons for three different single whisker stimuli and their multi-whisker combination. Firing rates for the …

Figure 3 with 4 supplements
Sparse, high-dimensional population codes of higher-order tactile features generated through stimulus-specific supra-linear summation.

(A) 31 example tuning curves from one field of view in one mouse showing neurons that exhibit a peak response for each of the 31 presented stimuli. Shaded regions are 95% confidence intervals (B) …

Figure 3—figure supplement 1
The preferred multi-whisker stimulus for each neuron nearly always includes its principal whisker (PW).

(A) Breakdown of C1 column neurons by whether they have C1 in their best stimulus or not. (B) Breakdown of C1 column neurons by whether C1 is their PW. (C) Normalized tuning curves for all …

Figure 3—figure supplement 2
Differences between tactile representations in L4 and L2/3.

(A) Normalized, cross-validated tuning curves for recorded L4 neurons. (B) Same as A but for L2/3 neurons. (C) Ring graphs showing the percentage of driven neurons that were single-whisker …

Figure 3—figure supplement 3
Comparison between L4 and L2/3 of barrel cortex.

(A) Normalized tuning curves of 50% of the data for all sensory-driven L4 neurons from awake mice ordered on tuning curves calculated on the remaining 50% of data. (B) Normalized linear difference …

Figure 3—figure supplement 4
Altered population codes in anesthetized, passively stimulated mice.

(A) Normalized, cross-validated tuning curves for all sensory-driven L2/3 neurons from awake mice. The black diagonal line depicts the expected population diagonal if the neural population uniformly …

Figure 4 with 7 supplements
Stimulus-specific supra-linear summation in the primary visual cortex.

(A) Schematic of the experimental preparation. (B) Example imaged field of view in layer 2/3 where each identified neuron is color coded according to its preferred single stimulus patch. (C) Five …

Figure 4—figure supplement 1
Comparison between L4 and L2/3 of visual cortex.

(A) Normalized tuning curves computed from 50% of the data for all sensory-driven V1 L4 neurons from awake mice ordered on tuning curves calculated on the remaining 50% of data. N=5 fields of view …

Figure 4—figure supplement 2
Analogous population code for 15° grating patches.

(A) Cross-validated tuning curves and (B) linear difference plots of neurons in layer 4 and layer 2/3 of V1 for 15 visual degree patches (right). N=10 fields of view, 6 animals.

Figure 4—figure supplement 3
Effect of deconvolution on tuning curve calculation.

(A) 2-D histogram of tuning curve elements computed using ∆F/F as compared to deconvolved event rate. Each plot is from one imaging session (either S1 L2/3 or S1 L4). Only trials following catch …

Figure 4—figure supplement 4
Non-cross validated tuning curves and linear difference plots of S1 and V1 neurons.

(A) Non-cross validated tuning curves and (B) linear difference plots for S1 layer 4 and layer 2/3 experiments, sorted on and displaying an average of all trials, rather than sorted based on held …

Figure 4—figure supplement 5
Alternative cross-validated tuning curves and linear difference plots of S1 and V1 neurons.

(A) Cross-validated tuning curves and (B) linear difference plots for S1 layer 4 and layer 2/3 experiments, sorted based on two-thirds of trials, and displaying the average response across the final …

Figure 4—figure supplement 6
Supra-linear difference plots reflect true deviation from linearity in S1.

(A) Linear difference plots, computed based on actual data, reproduced from Figure 3D. (B) Linear difference plots, computed based on a surrogate dataset, in which the underlying tuning curve was …

Figure 4—figure supplement 7
Eye movements are not strongly synchronized to stimulus presentation.

(A) Representative example traces of x and y positions of pupil center, in visual degrees, relative to their means. (B) Trial average of x-position of pupil center. Dashed lines represent onset (0 …

Figure 5 with 2 supplements
Selective supra-linearity and global sub-linearity of spatial summation across cortical areas and layers.

(A) Example V1 layer 2/3 tuning curves of the four multi-patch preferring neurons reproduced from Figure 4C, sorted by descending response magnitude. (B) Linear difference of the same neurons’ …

Figure 5—figure supplement 1
Supra-linearity of response to preferred stimulus reflects true deviation from linearity.

(A) Scatter plot of linear differences for preferred (top-ranked) stimulus versus an average of all non-preferred (non-top-ranked stimuli), for individual neurons (small dots), and for averages …

Figure 5—figure supplement 2
Input-specific supra-linear integration is well captured by a simple linear-nonlinear transformation.

(A1) Example tuning curve of an S1 L4 neuron, plotted as in Figures 2 and 4. (A2) Best-fit generalized linear model parameters for the tuning curve in (A1), fit based on half of trials. Color …

Videos

Video 1
Video of tactile stimulation with the novel, multi-piston tactile stimulator for head-fixed, locomoting mice.
Video 2
High speed tracking of whiskers and contacts during tactile stimulation.

Additional files

Download links