Astrocytic modulation of population encoding in mouse visual cortex via GABA transporter 3 revealed by multiplexed CRISPR/Cas9 gene editing
- Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, United States
- Picower Institute for Learning and Memory, Massachusetts Institute of Technology, United States
- Department of Biology, Eberly College of Science and Huck Institutes of the Life Sciences, Pennsylvania State University, United States
- F.M. Kirby Neurobiology Center, Boston Children’s Hospital, Harvard Medical School, United States
- Department of Neurosurgery, Boston Children’s Hospital, United States
- Whitehead Institute for Biomedical Research, United States
Figures
Figure 1 with 3 supplements
A Gat3-specific multiplexed CRISPR construct successfully knocks out Gat3.
(A) Gat3 expression across cortical layers in the mouse visual cortex (scale bar = 100 µm). (B) Quantification of Gat3 expression across cortical layers; expression density in white strip is shown at right (n = 3 mice, shaded area = SEM). (C) Schematic diagram illustrating the construct design, which consists of six CRISPR KO sgRNAs targeting the mouse Gat3 gene. These sgRNAs are separated by Csy4 enzyme cleavage sites, allowing their individual release in virus-injected cells. (D) Western blot and its quantification showing efficient knockout of Gat3 in cultured astrocytes co-transfected with Cas9 and Gat3-MRCUTS plasmids compared to astrocytes transfected with Cas9 plasmid alone (n = 3 independent experiments, *p < 0.05, two-tailed unpaired t-test, error bars = SEM). (E) DNA sequencing reads of one gRNA targeted region from mouse brain tissue collected after virus injection shows frequency of deletions at the target site in KO tissue (n.s., pmismatches = 0.723; ***pdeletions < 0.001; n.s., pinsertions = 0.158, two-tailed unpaired t-test, error bars = SEM). (F) Schematic of viral injections and cranial window implant over V1 for two-photon imaging. Viral constructs of the multiplexed gRNAs and red-shifted calcium indicator were co-injected in the left hemisphere of either wild-type mice or Cas9-expressing transgenic mice. (G) Representative immunohistochemistry images from a control and KO animal (scale bar = 100 µm, applies to all images in a row). (H) Comparison of Gat3 fluorescence intensity at the imaging sites and at the non-injected site within individual slices. Baseline intensity was determined by the non-injected right (contralateral) hemisphere to account for variability between slices (ncontrol = 9 slices, 4 mice, n.s., pcontrol = 0.443; nGat3 KO = 10 slices, 4 mice, ***pGat3 KO < 0.001, Mann–Whitney U test, ***pBetween groups < 0.001, two-way ANOVA).
Figure 1—figure supplement 1
Experimental design and timeline.
(A) Transgenic mice expressing Cre-dependent GCaMP6s in excitatory neurons (CaMKII-Cre) had cranial windows implanted over V1 and were imaged 1–2 weeks later. (B) Neuronal responses to visual stimuli were imaged using two-photon microscopy in head-fixed mice. Following baseline imaging, either vehicle (5% DMSO in corn oil) or SNAP-5114 (50 mg/kg) was injected i.p. followed by imaging the same field-of-view (FOV). (C) Timeline for imaging. Imaging after vehicle and SNAP-5114 treatment was performed on the same FOV on two different days. Both were compared to baseline imaging.
Figure 1—figure supplement 2
Systemic administration of SNAP-5114 affects select V1 neuronal response properties.
(A) Average maximum response magnitude of visually responsive neurons before and after vehicle administration (nbaseline = 1798 neurons, nvehicle = 1648 neurons, nbaseline = 20 sessions, nvehicle = 20 sessions, *p < 0.05, LME t-stats). (B) Same as A but before and after SNAP-5114 administration (nbaseline = 1709 neurons, nSNAP-5114 = 2002 neurons, nbaseline = 25 sessions, nSNAP-5114 = 38 sessions, **p < 0.01, LME t-stats). (C) Orientation selectivity index (OSI) distribution of all visually responsive neurons before and after vehicle administration (nbaseline = 1798 neurons, nvehicle = 1648 neurons, nbaseline = 20 sessions, nvehicle = 20 sessions, n.s., p = 0.359, LME t-stats). (D) Same as C but before and after SNAP-5114 treatment (nbaseline = 1709 neurons, nSNAP-5114 = 2002 neurons, nbaseline = 25 sessions, nSNAP-5114 = 38 sessions, n.s., p = 0.622, LME t-stats). (E) Average tuning curves of visually responsive neurons per session (in lighter shade) before and after vehicle administration. Average tuning curve of all sessions is in bold (nbaseline = 1798 neurons, nvehicle = 1648 neurons, nbaseline = 20 sessions, nvehicle = 20 sessions, error bars = SEM). (F) Same as E but before and after SNAP-5114 treatment (nbaseline = 1709 neurons, nSNAP-5114 = 2002 neurons, nbaseline = 25 sessions, nSNAP-5114=38 sessions, error bars = SEM). All sessions were from five mice studied under vehicle and SNAP-5114 treatment conditions (see Figure 1).
Figure 1—figure supplement 3
Gat3 expression changes in control and KO brain slices.
(A) Coronal slices of control animal brains. Left column: hemisphere with no virus injection. Right columns: hemispheres with virus injection from two different animals. Reduced superficial Gat3 expression is observed in control animals with variability between animals. Animal 1 represents the maximal effect of AAV injection on Gat3 expression, and animal 2 represents a more typical effect (scale bar = 100 µm, applies to all images). (B) Same as A but for Gat3 KO animal brains. Compared to the injected hemispheres of control animals, the injected hemispheres of Gat3 KO animals showed a more severe reduction of Gat3 expression throughout the cortical layers (scale bar = 100 µm, applies to all images). White dotted lines indicate pial surface of the cortex.
Figure 2
Genetic knockout of Gat3 in the visual cortex alters inhibitory output onto single pyramidal neurons.
(A) Schematic of ex vivo whole-cell patch clamp electrophysiology setup. Gat3-MRCUTS was co-injected with a tdTomato virus to label the injection site for recordings. (B) Representative traces of spontaneous inhibitory postsynaptic currents (sIPSCs) of L2/3 pyramidal neurons in visual cortex brain slices. (C) Comparison of frequency of sIPSCs between control and Gat3 KO brain slices (ncontrol = 20 cells, nGat3 KO = 23 cells, ***p < 0.001, two-tailed unpaired t-test). (D) Cumulative probability histograms for inter-event intervals (***p < 0.001, Kolmogorov–Smirnov test). (E) Comparison of average amplitude of sIPSCs (n.s., p = 0.9351, two-tailed unpaired t-test).
Figure 3 with 1 supplement
Genetic knockout of Gat3 in the visual cortex alters spontaneous activity of single neurons.
(A) Schematic of control and experimental mice preparation. Both control mice (wild-type) and experimental mice (transgenic mice with cells constitutively expressing Cas9-EGFP under CAG promoter) received co-injection of Gat3-MRCUTS and a neuronal calcium sensor (jRGECO1a) in V1 during stereotactic surgeries. (B) Two-photon imaging setup consisting of a running wheel and a pupil camera to acquire locomotion and pupil dynamics, respectively. (C) Top: Example field-of-view (FOV) images of an imaging session from each group (scale bar = 50 µm). Bottom: Example average Ca2+ traces from a control and a Gat3 KO mouse. The normalized Ca2+ traces of all neurons within the same FOV were averaged. (D) Representative heatmaps of normalized spontaneous calcium activity of neurons in each session from each group. (E) Firing rates of individual neurons in control and Gat3 KO groups. Inset shows the average firing rates of all neurons from each group (ncontrol = 838 neurons, 4 mice, nGat3 KO = 606 neurons, 4 mice, ***p < 0.001, Linear mixed effects model (LME) t-stats, see Methods, error bars = SEM). (F) Distribution of pairwise correlation coefficients of neurons (ncontrol = 31,460 pairs, nGat3 KO = 18,004 pairs, n.s., p = 0.224, LME t-stats).
Figure 3—figure supplement 1
Comparisons of average firing rates of single neurons and neuron-to-neuron pairwise correlation during spontaneous activity.
(A) Empirical cumulative distribution functions plot of firing rates for each animal shown as individual traces (ncontrol = 4 mice, nGat3 KO = 4 mice). (B) Average firing rates of neurons; different sessions per animal indicated by different colors (ncontrol = 13 sessions, 4 mice, nGat3 KO = 11 sessions, 4 mice, *p < 0.05, LME t-stats, error bars = SEM). (C) Average neuron-to-neuron pairwise correlation coefficient of neurons per session by animal (ncontrol = 13 sessions, 4 mice, nGat3 KO = 11 sessions, 4 mice, n.s., p = 0.224, LME t-stats, error bars = SEM).
Figure 4 with 1 supplement
Genetic knockout of Gat3 in the visual cortex alters the visual response properties of neurons.
(A) Example Ca2+ traces of a single neuron from control (top) and Gat3 KO (bottom) during presentation of drifting gratings. The average of all trials is plotted in a dark line overlaid on the lighter individual trial traces (16 trials in total). (B) Average maximum response magnitudes of neurons to their preferred grating orientation. Visually responsive neurons were pooled across animals within each group (ncontrol = 526 neurons, 4 mice, nGat3 KO = 366 neurons, 4 mice, **p < 0.01, LME t-stats). (C) Representative tuning curves of control individual neurons (in lighter shade) and the average tuning curve (in bold) of all neurons in each FOV centered around their preferred orientation (ncontrol = 32 neurons, error bars = SEM). (D) Same as B but for Gat3 KO (nGat3 KO = 38 neurons, error bars = SEM). (E) Comparison of orientation selectivity index (OSI) distribution of visually responsive neurons between the two groups (n.s., p = 0.183, LME t-stats). Insets show the percentage of cells with OSI greater or less than 0.3. (F) Example Ca2+ traces of a single neuron from control (top) and Gat3 KO (bottom) during natural movies where the dotted lines indicate the onset of a movie. The average of all trials is plotted in a dark line overlaid on the lighter individual trial traces (32 trials in total). (G) Example plots showing variability of each trial response (in lighter shade) of a single neuron to a natural video; dotted line indicates the stimulus onset. (H) Reliability indices of neurons to their preferred stimuli in control and Gat3 KO group (ncontrol = 707 neurons, 4 mice, nGat3 KO = 436 neurons, 4 mice, *p < 0.05, LME t-stats). (I) Generalized linear model (GLM)-based single neuron encoding model of visual stimulus information, pupil dynamics, and running speed. Variance explained (R2) is computed to assess the encoding property of neurons. (J) Distribution of R2 of individual neurons from each group (ncontrol = 647 neurons, 4 mice, nGat3 KO = 565 neurons, 4 mice). (K) Comparison of average R2 values of individual neurons between the two groups (*p < 0.05, LME t-stats, error bars = SEM). (L) Proportions of neurons encoding each parameter (visual stimuli, pupil dynamics, and movement) from each imaged population (n.s., pVisual stimuli = 0.116; n.s., pPupil = 0.662; n.s., pMovement = 0.172, LME t-stats).
Figure 4—figure supplement 1
Comparisons of visual responses of single neurons to drifting gratings.
(A) Average tuning curves of neurons by session in each group (ncontrol = 12 sessions, 4 mice, nGat3 KO = 11 sessions, 4 mice, error bars = SEM). (B) Orientation selectivity index (OSI) distribution of neurons for each animal (ncontrol = 4 mice, nGat3 KO = 4 mice, n.s., p = 0.183, LME t-stats, error bars = SEM). (C) Maximum response magnitudes of visually responsive neurons to their preferred gratings compared by animal (ncontrol = 4 mice, nGat3 KO = 4 mice, **p < 0.01, LME t-stats). (D) Average R2 of neurons for single neuron encoding of drifting gratings and behavioral variables by sessions (ncontrol = 12 sessions, 4 mice, nGat3 KO = 11 sessions, 4 mice, **p < 0.01, LME t-stats, error bars = SEM).
Figure 5 with 1 supplement
Genetic knockout of Gat3 alters population-level properties of cortical neurons.
(A) Schematic of a single neuron encoding model of population activity using generalized linear model (GLM). Calcium traces of randomly sampled neurons in a fixed population size were used to train a GLM model for prediction of the target neuron’s activity. (B) Distribution of R2 values of individual neurons (ncontrol = 707 neurons, 4 mice, nGat3 KO = 436 neurons, 4 mice, training population size = 20 neurons). (C) Comparison of average R2 value of all neurons between two groups (*p < 0.05, LME t-stats, error bars = SEM). (D) The maximum value of the predictor weights (b) from each neuron’s GLM fitting was extracted and grouped into ranges of below 0.05, 0.05–0.1, and above 0.1. The difference in proportions of the weights showed the different level of encoding of other neurons between the two groups (*p < 0.05, ***p < 0.001, Mann–Whitney U test). (E) Support Vector Machine (SVM)-based decoding analysis of neuronal population activity induced by drifting gratings in neuronal populations of various sizes. Comparison of decoding accuracy of visual stimulus information (area under the receiver operating characteristic curve) of populations between two groups (ncontrol = 12 sessions, 4 mice, nGat3 KO = 9 sessions, 4 mice, ***p < 0.001, two-way ANOVA). (F) Same as E but for natural movies (ncontrol = 11 sessions, 4 mice, nGat3 KO = 11 sessions, 4 mice, ***p < 0.001, two-way ANOVA). Inset: comparison of average area under the receiver operator characteristic curve (AUROC) between different visual stimuli within each group (***p < 0.001, Mann–Whitney U test, error bars = SEM). (G) A simplified diagram of a visual cortex L2/3 microcircuit consisting of neurons and astrocytes. The microcircuit contains different types of inhibitory neurons that exert inhibitory or disinhibitory effects on pyramidal neurons. Extra-synaptic expression of Gat3 in astrocytic processes allows astrocytes to control extracellular GABA levels that may differentially influence a wide network of cells.
Figure 5—figure supplement 1
Comparisons of neuronal responses to natural movies.
(A) Average reliability indices of neurons per session (ncontrol = 12 sessions, 4 mice, nGat3 KO = 10 sessions, 4 mice, *p < 0.05, LME t-stats, error bars = SEM). (B) Representative noise correlation coefficient matrices. (C) Average signal correlation coefficients between pairs of neurons (ncontrol = 24,929 pairs, nGat3 KO = 10,110 pairs, ncontrol = 12 sessions, 4 mice, nGat3 KO = 10 sessions, 4 mice, n.s., p = 0.741, LME t-stats, error bars = SEM). (D) Average noise correlation coefficients between pairs of neurons (ncontrol = 24,929 pairs, nGat3 KO = 10,110 pairs, ncontrol = 12 sessions, 4 mice, nGat3 KO = 10 sessions, 4 mice, n.s., p = 0.1349, LME t-stats, error bars = SEM). (E) R2 distribution of population activity encoding GLM model performance as a function of different population sizes of neurons used for model training. (F) Average R2 values per session and per animal for each population size (ncontrol = 12 sessions, 4 mice, nGat3 KO = 10 sessions, 4 mice, *p5-20 < 0.05 each, LME t-stats, error bars = SEM).
Additional files
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Astrocytic modulation of population encoding in mouse visual cortex via GABA transporter 3 revealed by multiplexed CRISPR/Cas9 gene editing
eLife 14:RP107298.
https://doi.org/10.7554/eLife.107298.3
