Sensory experience regulates cortical inhibition by inducing IGF1 in VIP neurons

Abstract

Inhibitory neurons regulate the adaptation of neural circuits to sensory experience, but the molecular mechanisms by which experience controls the connectivity between different types of inhibitory neuron to regulate cortical plasticity are largely unknown. Here we show that exposure of dark-housed mice to light induces a gene program in cortical vasoactive intestinal peptide (VIP)-expressing neurons that is markedly distinct from that induced in excitatory neurons and other subtypes of inhibitory neuron. We identify Igf1 as one of several activity-regulated genes that are specific to VIP neurons, and demonstrate that IGF1 functions cell-autonomously in VIP neurons to increase inhibitory synaptic input onto these neurons. Our findings further suggest that in cortical VIP neurons, experience-dependent gene transcription regulates visual acuity by activating the expression of IGF1, thus promoting the inhibition of disinhibitory neurons and affecting inhibition onto cortical pyramidal neurons.

Extended Data Figure 1. Validation of the sensory stimulation protocol and the RiboTag-based cell-type-specific purification of mRNA

a) Quantitative real-time PCR (qPCR) for known experience-regulated genes on RNA isolated from the visual cortex of mice that were dark-housed for varying durations (overnight, 3 days, 7 days or 14 days) and then either sacrificed in the dark or exposed to light for 1, 3, or 7.5 hours, and then sacrificed. Data are normalized to the maximal value in each dataset and represent the mean and standard error of four biological replicates. b) qPCR for known experience-regulated genes on RNA isolated from the visual cortex of mice that were dark-housed for 14 days and then either exposed to light for 1, 3, or 7.5 hours (dark + light, black) or kept in the dark during these hours (dark + dark, red). All mice of a given time point were dissected in very close temporal proximity. Data are normalized to the maximal value in each dataset and represent the mean and standard error of four biological replicates. c) qPCR for known cell-type-specific marker genes on RNA isolated from RiboTag mice expressing Cre in distinct neuronal subtypes. Data are normalized to the maximal value in each dataset and represent the mean and standard error of three biological replicates. d) qPCR for known early-induced transcription factors on RNA isolated from RiboTag mice expressing Cre in distinct neuronal subtypes. Data are normalized to the maximal value in each dataset and represent the mean and standard error of three biological replicates.

Extended Data Figure 2. Validation of the RiboTag-Seq approach

a) Matrix of Spearman correlation coefficients between biological replicates across all samples (Scale of correlation coefficients extends from 0.7 to 1, see color bar) (S = standard housing, 0 = dark-housed only, 1/3/7.5 = 1/3/7.5 hours of light exposure after dark housing). b) Line plots of RNA-Seq data showing the expression values (normalized reads across all exons of a gene) for cell-type-specific marker genes and ubiquitously expressed house-keeping genes in different Cre lines (Emx1-Cre: red; Gad2-Cre: blue, Pv-Cre: brown, Sst-Cre: green, and Vip-Cre: purple) across all time points of the experiment. Data represent the mean and standard error of three biological replicates.

Extended Data Figure 3. Characterization of the experience-induced gene programs in subtypes of cortical neurons

a) Average matrix of Spearman correlation coefficients computed from the expression values of 1000 random sets of 602 genes (including experience-regulated genes, with replacement). b) Matrix of the Spearman correlation coefficient computed from the expression levels of control transcripts that match the expression distribution of experience-regulated genes (n = 602). c) Cladogram resulting from hierarchical clustering of all samples (except samples from Gad2-Cre). Cladograms were computed using the mean expression values (i.e. normalized reads across all exons of a gene) for all expressed transcripts (n = 14,280). d) Cladogram resulting from hierarchical clustering of the mean expression values of a set of control transcripts that match the expression distribution of experience-regulated genes (n = 602). e) Pie charts showing the subdivision of experience-regulated genes on the basis of kinetics in each set of Cre lines (red: rapidly induced; gray: induced with delayed kinetics; orange: induced only after two weeks of dark housing; green: rapidly suppressed; magenta: suppressed with delayed kinetics; blue: suppressed only after two weeks of dark housing). f) Left, matrix of Spearman correlation coefficients between Cre lines computed using the mean expression values (normalized reads across all exons of a gene) of early-induced genes one hour after light exposure. Right, matrix of Spearman correlation coefficients between Cre lines computed using the mean expression values of late-induced genes 7.5 hours after light exposure. For each matrix, the correlations upon permuting the expression values are also shown. (Colorbar at right; scale begins at zero.)

Extended Data Figure 4. Characterization of cell-type-specific and experience-induced gene in subtypes of cortical neurons

a) Scatter plots showing the Log10 expression values for each expressed gene in a given Cre line (x-axis) plotted against the maximum Log10 expression values for that gene found in all other Cre lines (y-axis). Black line denotes unity, and the red line is the 5.5-fold enrichment threshold set to include Vglut1 as a cell-type-specific gene in Emx1-Cre. Data represent the mean values of three biological replicates. b) Scatter plots of all expressed genes, for each Cre line plotting the mean Log2 fold-enrichment in that Cre line (x-axis) against the mean Log2 of the absolute value of the maximum fold-change observed in that Cre line. Data represent the mean values of three biological replicates. Genes that pass both enrichment and induction thresholds in 3 of 3 biological replicates are shown in red. c) Bar graph showing the maximum expression value (in normalized reads) for VIP-neuron-specific experience-regulated genes.

Extended Data Figure 5. Conditional knockout of Igf1 in VIP neurons does not affect cortical morphology or gross morphology of VIP neurons

a) Example image of cortices from Igf1 WT (Vip-Cre/+, LSL-tdTomato/+, Igf1 wt/wt) or cKO (Vip-Cre/+, LSL-tdTomato/+, Igf1 flx/flx) mice. VIP neurons are labeled in white, with DAPI shown in blue (cortical layers are indicated on the left, scale bar = 200 µm). b–d) Bar graphs showing the area of each cortical layer (b), number of VIP neurons per image per layer (c), or soma size of VIP neurons (d) in Igf1 WT (black) or cKO (red) mice. Data represent the mean and standard error of three biological replicates.

Extended Data Figure 6. Igf1 knockdown in VIP neurons affects inhibitory inputs onto VIP neurons but not onto neighboring neurons

a) AAV shRNA constructs. shRNA cassettes against Igf1 or a control gene (Luc) were cloned downstream of the U6 promoter into an AAV vector that drives Cre-dependent expression of EGFP. b) qPCR validation of the efficacy of Igf1 shRNA constructs. Cultured cortical neurons were infected with lentiviral constructs either expressing no shRNA (Vector only), a control shRNA (against Luc) or shRNAs against Igf1. Four days post infection the cultures were quieted overnight with TTX and AP-5 and then harvested either prior to or after being depolarized with 55 mM KCl for 6 hours; RNA was then isolated and qPCR was performed. Data are normalized to the maximal value in each replicate and represent the mean and standard error of three biological replicates. c) Bar graph showing normalized soma size of P21 visual cortex VIP neurons infected with control shRNA or shRNA targeting Igf1 (shRNA Control: n = 103, shRNA Igf1: n = 174, p =0.41, Mann Whitney U-Test). d) Bar graphs showing the levels of IGF-1 in the serum of P20 mice that were injected intracortically with AAVs driving the expression of control shRNA (black) or Igf1 shRNA. Data represent the mean and SEM of the serum IGF-1 levels of 4 mice per group. e–l) Bar graphs and cumulative distribution plots showing mIPSC amplitudes and frequency/interevent interval upon early widespread knockdown of Igf1 in VIP (e,f), SST (g,h), PV (i,j) and excitatory (k,l) neurons after injection of AAVs into P3 cortices of the respective Cre-mice. VIP neurons (identified by EGFP-positive cells in Vip-Cre mice): control shRNA and Igf1 shRNA (Amplitude p = 0.96, Frequency p = 0.04). SST neurons (identified by EGFP-positive cells in Sst-Cre mice): control and Igf1 shRNA (Amplitude p = 0.89, Frequency p = 0.55). PV neurons (identified by EGFP-positive cells in Pv-Cre mice): control and Igf1 shRNA (Amplitude p = 0.084, Frequency p = 0.93). Pyramidal neurons (identified by morphology): control and Igf1 shRNA (Amplitude p = 0.84, Frequency p = 0.15). (All p values are derived from Mann Whitney U-Tests; numbers inside bars indicate the number of cells recorded)

Extended Data Figure 7. Effects of IGF-1 overexpression in excitatory and Sst-positive neurons

a) RiboTag-Seq identifies Igf1.4 as the major Igf1-isoform expressed in VIP neurons. Representative tracks of histograms of the RNA-Seq reads in each Cre line across the Igf1 genomic locus. Data is from the 7.5 hour light exposure RiboTag-Seq datasets. b) AAV constructs for the Cre-dependent expression of HA-/Myc-tagged IGF-1 (Igf1.4) and EGFP (i.e. IGF1-OE, top) or of EGFP alone (i.e. control, bottom). F2A = Furin cleavage site followed by the 2A peptide; black and white triangles represent a Cre-dependent Flex-switch. c) Western blot analysis of IGF1-receptor activation status in lysates of serum starved HEK293T cells that were stimulated with conditioned media (CM) containing epitope-tagged isoforms of IGF-1. CM was produced by transfecting HEK293T cells with the respective construct and collecting the culture media. IGF1-receptor is detected with antibodies against either activated IGF1-receptor (anti Igf1R pTyr^1136/1138) or total IGF-1 receptor (anti Igf1R total). (Molecular weight markers are on the right and the arrow indicates the band of the IGF1-receptor). d,e) Bar graphs and cumulative distribution plots showing mIPSC frequency / interevent interval (d) and amplitude (e) of mIPSCs recorded from EGFP positive neurons in P20 Sst-Cre mice that were intracortically injected with AAVs driving the expression of control (C) or IGF1-OE (OE) constructs (Amplitude, p = 0.16; Frequency, p = 0.01; Mann Whitney U-Test; numbers inside bars indicate the number of cells recorded). f,g) Bar graphs and cumulative distribution plots showing mIPSC frequency / interevent interval (f) and amplitude (g) of mIPSCs recorded from EGFP positive neurons in P20 Emx1-Cre mice that were intracortically injected with AAVs driving the expression of control (black, n = 20) or IGF1-OE AAVs (red, n = 20). (Amplitude, p = 0.99; Frequency, p = 0.01, Mann Whitney U-Test) h–j) Scatter plots of IGF-1-interacting proteins showing the Log10 normalized mean expression values in Vip-Cre neurons versus each of the other Cre lines (h - Emx1, i - Pv, j - Sst). k) Quantification of EPSC paired-pulse recordings from VIP neurons infected with control shRNA (black n = 9) or Igf1 shRNA (red n = 11) expressing AAVs. The ratio of the second EPSC amplitude divided by the first EPSC amplitude is plotted against inter-stimulus interval (p = 0.1, two-way ANOVA).

Extended Data Figure 8. VIP neuron-derived IGF-1 does not disrupt ocular dominance plasticity

a) Widespread infection of VIP neurons by AAV-shRNA-hUbc-Flex-EGFP. High-titer injection of AAVs into the visual cortex of P18-20 Vip-Cre/+ mice leads to infection of the majority VIP neurons (green = EGFP, red = anti-VIP, blue = DAPI, arrowheads = infected VIP neurons; Scale bars 150 µm, 20 µm in the inlet). b–c) Average of VEP traces recorded in the visual cortices of mice that were injected with AAVs expressing control shRNA (black/grey) or shRNA against Igf1 (red/pink) shRNA and that were (grey, pink) or were not (black, red) subjected to monocular deprivation in the eye contralateral to the recording site (MD vs NoMD, respectively). d) Monocular deprivation induces a significant reduction in the VEP amplitude in response to low spatial frequency stimulation in mice that had AAVs expressing control shRNA and Igf1 shRNA injected into their visual cortices (control shRNA NoMD, n = 5 mice; control shRNA MD, n = 7 mice; Igf1 shRNA NoMD, n = 5; Igf1 shRNA MD, n = 8. p* < 0.05, Mann Whitney U-Test). e) Mice that had AAVs expressing control shRNA (black) and Igf1 shRNA (red) injected into their visual cortices display normal ocular dominance plasticity as monocular deprivation (MD) induces a shift to the ispilateral eye in both groups (control shRNA NoMD, n = 5 mice; control shRNA MD, n = 7; Igf1 shRNA NoMD, n = 5 ; Igf1 shRNA MD, n = 8 mice. p** < 0.0001, One-way ANOVA-Tukey post test).

Figure 1. VIP neurons mount a unique transcriptional response to sensory experience

a) Approach for purifying ribosome-bound RNA from Cre-expressing neurons in visual cortex following manipulation of visual experience. b) Representative RNA-Seq tracks of cell-type-specific marker genes (exons shaded). c) Line plots showing the expression levels of selected early-induced genes in each Cre-line at each time point (n=3, bars = SEM). d) Histograms showing the distribution of all observed fold-changes for each Cre-line (mean of three replicates). e) Color-coded matrix showing the −Log10 P-value of pairwise corrected two-tailed t-tests of the fold-change distributions between each Cre-line. f, g) Matrices of Spearman correlation coefficients computed from the expression levels of experience-regulated genes (f) or genes not regulated by experience (g) (mean of three replicates). h, i) Cladograms of (h) all experience-regulated genes created by using the mean values of each gene in each sample or (i) average cladogram created from 1000 random sets of 602 random genes using the mean expression values of each gene in each sample (p=0.011, Monte Carlo test).

Figure 2. IGF-1 is an experience-induced cell-type-specific secreted factor in VIP neurons

a) Heatmap showing the fold-change of all experience-regulated genes across all stimulus intervals. b) Line plots of secreted factors that are experience-regulated and expressed in a cell-type-specific manner (n=3, bars = SEM). c) Fluorescent In situ hybridization for Igf1 and Vip in mouse visual cortex after dark housing and light exposure for 7.5 hours (white box = magnified area; scale bar = 200 µm in main image, 50 µm in magnification). d–f) Quantification of fluorescent in situ hybridization for Igf1, Crh, Prok2, and inhibitory markers in visual cortices of dark-housed/light-exposed mice (n=3, bars represent SEM).

Figure 3. IGF-1 promotes inhibitory inputs to VIP neurons in a cell autonomous manner

a–d) Bar graph and cumulative distribution of the frequency and interevent intervals of mIPSCs or mEPSCs recorded from Igf1 WT or cKO VIP neurons (mIPSC frequency: p = 0.046; amplitude p = 0.3. mEPSC frequency: p = 0.44; amplitude p = 0.9, Mann-Whitney U-Test). e) Example image of sparsely infected VIP neurons upon injection of AAV-shRNA-hUbc-Flex-EGFP into mice expressing tdTomato in all VIP neurons (white box = magnified area; arrows = infected VIP neurons; arrowheads = non-infected VIP neurons; scale bars, 100 µm in main image, 50 µm in magnification) f–i) Bar graph and cumulative distribution of mIPSC/mEPSC frequency, interevent interval and amplitude recorded from VIP neurons sparsely infected with control or Igf1 shRNAs (mIPSC Frequency: shRNA1 p = 0.05, shRNA2 p = 0.042. mIPSC Amplitude: shRNA1 p = 0.004, shRNA2 p = 0.001. mEPSC Frequency: shRNA1 p = 0.13, shRNA2 p = 0.07. mEPSC Amplitude: p = 0.77, shRNA2 p = 0.44. Mann-Whitney U-Test). j) Sholl analyses of VIP neurons infected with control or Igf1 shRNA (p = 0.76, two-way repeated-measures ANOVA). k) Expression of epitope-tagged Igf1.4 in VIP-neurons. Cortices of P3 Vip-Cre/+ mice were injected with AAVs driving Cre-dependent expression of SSHAIgf1.4Myc-F2A-EGFP (top) or F2A-EGFP (bottom) and stained at P20 for HA (red) and EGFP (green) (Scale-bar = 10 µm). l–o) Bar graphs and cumulative distribution plots showing mIPSC/mEPSC amplitude and frequency / interevent interval in VIP neurons infected with control or IGF1-OE AAVs. (mIPSC: Amplitude, p = 0.05; Frequency, p = 0.02. mEPSC: Amplitude, p = 0.55; Frequency, p = 0.86. Mann Whitney U-Test) a–d, f–j, l–o) Numbers inside bars indicate the number of cells recorded

Figure 4. VIP neuron-derived IGF-1 regulates VIP neuron function and regulates visual acuity in an experience-dependent manner

a) Paired-pulse recordings from VIP neurons infected with control or Igf1 shRNA (p = 0.96, two-way ANOVA). b) Left: average traces of light-evoked IPSCs (eIPSC) from paired recordings of VIP neurons infected with control or Igf1 shRNA (green traces) and neighboring pyramidal neurons (black traces). Right: quantification of eIPSC amplitude of the VIP neuron after infection with AAVs expressing control or Igf1 shRNA normalized, to the eIPSC amplitude of the paired pyramidal neuron (p = 0.01, Mann-Whitney U-Test). c) Average firing rate of VIP neurons infected with Igf1 or control shRNA (p = 0.04, Mann-Whitney U-Test). b, c) Numbers inside bars indicate the number of cells recorded d) Schematic of the schedule for monocular deprivation (MD) experiments. e–f) Representative traces of visually evoked potential (VEP) amplitude as a function of spatial frequency (cpd) in the contralateral visual cortex of mice that received bilateral injections of AAVs expressing Igf1 or control shRNA into their visual cortices and were subjected to MD or not. g) VIP neuron-derived IGF-1 restricts visual acuity in an experience-dependent manner. Visual acuity in mice injected with AAVs expressing Igf1 or control shRNA with no MD or MD (P24 – P28; control shRNA NoMD, n = 5; control shRNA MD, n = 5; Igf1 shRNA NoMD, n = 5; Igf1 shRNA MD, n = 7; p* < 0.05, p*** < 0.0001, n.s. – not significant, One way ANOVA-Tukey post test).