A dendritic substrate for temporal diversity of cortical inhibition

Abstract

In the mammalian neocortex, GABAergic interneurons (INs) inhibit cortical networks in profoundly different ways. The extent to which this depends on how different INs process excitatory signals along their dendrites is poorly understood. Here, we reveal that the functional specialization of two major populations of cortical INs is determined by the unique association of different dendritic integration modes with distinct synaptic organization motifs. We found that somatostatin (SST)-INs exhibit NMDAR-dependent dendritic integration and uniform synapse density along the dendritic tree. In contrast, dendrites of parvalbumin (PV)-INs exhibit passive synaptic integration coupled with proximally enriched synaptic distributions. Theoretical analysis shows that these two dendritic configurations result in different strategies to optimize synaptic efficacy in thin dendritic structures. Yet, the two configurations lead to distinct temporal engagement of each IN during network activity. We confirmed these predictions with in vivo recordings of IN activity in the visual cortex of awake mice, revealing a rapid and linear recruitment of PV-INs as opposed to a long-lasting integrative activation of SST-INs. Our work reveals the existence of distinct dendritic strategies that confer distinct temporal representations for the two major classes of neocortical INs and thus dynamics of inhibition.

Figure 1:. PV- and SST-INs display different dendritic integration properties.

A) (left) Two-photon laser scanning microscopy (2PLSM) image (maximum-intensity projection, MIP) of a PV-IN patch loaded with Alexa-594 (10 μM). The inset shows a selected dendrite and uncaging sites for 10 synapses. (right) Example of photolysis-evoked EPSPs in a dendrite with increasing number of stimulated synapses (from locations in A, activated at 0.12 ms intervals). The traces compare the calculated linear sum of individual responses to the recorded response from quasi-synchronous synapse activation. The input-output curve shows recorded somatic EPSP amplitudes versus the expected linear sum. B) Summary plot of input-output curves for dendrites of PV-INs. Single grey lines represent individual experiments (n=17) and in red the binned average. C-D) Same as A-B, but for SST-INs. E) Same as A-B, but for pyramidal neurons. F) Schematic representation of experimental conditions used to stimulate local excitatory connections and record glutamatergic synaptic activity in PV- and SST-INs in layer 2/3 of primary somatosensory cortex. G) Representative traces (left) and summary plot (right) of EPSCs recorded at −70 mV and +40 mV in PV- and SST-INs (n=8; p<0.001, Mann-Whitney). H) Percentage of non-linearity obtained in PYR (green, n=25), SST-INs (orange, n=23), PV-INs (red, n=17), SST INs in the presence of NMDA blockers (light green, n=14) and SST-INs from GluN1-KO mice (khaki green, n=7). The data are presented as average ± SEM; **p<0.001,***p<0.0001, ****p<0.00001, Kruskal-Wallis test with Dunn’s correction. I) Properties of the simplified model of PV-IN integration. (left) Schematic of the morphology with a somatic compartment and symmetric-branching dendritic tree. (right) Input resistance and transfer resistance profiles along the dendritic tree relative to soma distance. J) Input-output curve in the simplified model. (top inset) Depolarization values from quasi-synchronous input of increasing number of synapses (1 to 6 synapses) at various dendritic locations (color-coded as in I). Dashed and solid lines show linear prediction (sum) and observed (real) responses. (bottom) Relationship between expected (sum) and observed (real) peak EPSP values at different dendritic locations (color-coded), normalized to single event EPSP values. K) Subthreshold input-output relationship of pEPSPs in proximal (<40 μm from soma) and distal (100 μm from soma) dendrites using quantal adjusted pEPSP amplitude. Dashed line indicates slope of 1. L) Summary of non-linearity percentage in proximal (cardinal red) and distal (brick red) PV-INs and distal SST-INs dendrites (orange) using quantal adjusted EPSP amplitude. Data are average ± SEM, *p<0.01,***p<0.0001, Kruskal-Wallis test with Dunn’s correction

Figure 2:. Multivesicular release summates sublinearly in PV-INs dendrites and predominantly linearly in SST-INs dendrites.

A) (Top) Methodology to study neurotransmitter release at synaptic contacts between PNs and PV- or SST-INs using iGluSNFR-A184V expressed in PV- or SST-INs. Glutamate release was evoked from L2/3 PNs via current injection. (Bottom) Two-photon images show a patch-loaded PN axon (red) and SST-INs expressing iGluSNFR-A184V (green). B) Averaged line scan (10 images) of iGluSNFR-A184V fluorescence (green) at contact points between L2/3 PN boutons and PV/SST-IN dendrites after AP (7 APs @ 10 Hz) initiation. Time series traces (10) show mean fluorescence over 1 um (white line), with the black trace as the average (ΔF/F). C) Histogram of ΔF/F amplitude distributions for PV (red) and SST-INs (orange) iGluSNFR events. D) Summary of failure rates and coefficient of variation for L2/3 PN synaptic contacts in SST-(n=8) and PV-INs (n=9). Data include release probability (Pves) and maximum vesicles released per AP (N max Ves), (*p<0.01, Mann-Whitney test). E) Two-photon image of an SST-IN loaded with Alexa 594, with glutamate uncaging at three dendritic locations. Somatically recorded EPSCs from photolysis at two pulse durations. F) Similar to E for PV-INs. G) Summary of paired-pulse ratios of photolysis-evoked EPSCs for PV and SST-INs. H) Paired-pulse ratios binned by distance from soma for PV and SST-INs, (**p<0.001, p<0.00001, Kruskal Wallis test with Dunn’s correction). I, J) Similar analysis in the presence of NMDA receptor antagonists, alone (I) or with NMDA and VGCC blockers (J), (**p<0.001, p<0.00001, Kruskal Wallis test with Dunn’s correction). K, L) Similar analysis for PV-INs with TEA (5 mM), (**p<0.001, p<0.00001, Kruskal Wallis test with Dunn’s correction). Data are mean ± SEM.

Figure 3:. Distinct synaptic distributions in the dendritic trees of Basket and Martinotti INs.

A) Illustration of experimental approach used to quantify glutamate synapse distribution along dendrites of SST and PV-INs. PSD-95 proteins were selectively labelled in PV- and SST-INs and quantified at two different distances along individual dendritic branches. B) (left) Overview image of PV-INs with proximal dendrites labelled with tdTomato (green) and PSD95venus (red). (Middle) STED image of dendritic location defined on the left. (right) Merged between tdTomato obtained in confocal mode (green) and PSD95venus (red) labelling obtained in STED. D, E) Same as B but for SST-INs. Note the absence of PSD-95 puncta close to soma but clear labelling in distal dendrites. C, F) Summary plot of density of PSD-95 puncta in proximal (less than 40 μm) and distal (approximately 100 μm) dendritic location in PV- and SST-IN dendrites (PV_proximal, n=33; PV_distal, n=12, p<0.0001, Mann-Witney test; SST_proximal, n=21; SST_distal, n=14, p>0.01, Mann-Witney test). G) Example of basket and Martinotti INs reconstructions from the millimeter-scale volumetric electron microscopy MICrONS dataset (48). Blue dots indicate the location of identified synapses along the dendrites of a basket and a Martinotti Cell. H) Illustration of synapse distributions in individual dendritic branches from reconstructed Basket (left) and SST-INs (right) displayed in G. Images display synapses along full individual dendritic branches as well at proximal and distal dendritic segments. I) Linear synapse density distributions (left, see Methods) for all basket (n=59) and Martinotti (n=41) cells as a function of the distance from soma. (right) Violin plots of distribution peaks (Basket vs Martinotti, p=1e-11, Mann-Whitney test) and distribution slopes as a variation per 100μm (Basket vs Martinotti, p=6e-14, Mann-Whitney test) for the plots on the left. J) Synaptic count (left, see Methods) for all basket (n=59) and Martinotti (n=41) cells as a function of the distance from soma. (right) Violin plots of distribution peaks (basket vs Martinotti, p=2e-13, Mann-Whitney test) and distribution skewness (basket vs Martinotti, p=1e-8, Mann-Whitney test) for the plots on the left.

Figure 4:. The PV- and SST-IN dendritic programs improve distal signal transmission and optimize synaptic efficacy through different mechanisms.

A) Morphology of the EM reconstructed basket cell (i) and Martinotti cell (ii) used as PV- and SST-IN models, respectively. In the insets, we show the real distributions (colored bars) of synapses on a single representative branch together with the uniform surrogate distribution (plain gray bars). The inset table shows the sparse subset of synapses in the distal (green) and proximal (pink) segments. B) Example numerical simulations of V_m response following the quasi-synchronous stimulation of the synapses in the distal (green, left) and proximal (pink, right) segments in the PV-IN (i) and SST-IN (ii) models. We show the observed response (plain line) and the expected response (dashed line) from the individual event responses. We compare each model to its control situation (i, grey, uniform distribution), (ii, purple, no-NMDA). C) Suppression of EPSP amplitude between observed response and linear predictions (see annotations in B) in the PV-IN (i, red) and SST-IN (ii, orange) models with their own control in the distal (green) and proximal (pink) segments. D) Ratio of suppression between the distal and proximal segments in the PV-IN model (i, red, with its “uniform distribution” control in grey, n=6 branches, p=3e-2, Wilcoxon test) and SST-IN model (ii, orange, with its “no NMDA” control in purple, n=6 branches, p=3e-2, Wilcoxon test). E) V_m dynamics at the soma following background and stimulus-evoked activity in the dendritic branch shown in A. Synaptic stimulation consists in the quasi-synchronous activation of increasing number of synapses $n_{syn}$ (see annotations). We show the PV-IN model (red, with its “uniform distribution” control in grey) and SST-IN model (orange, with its “no NMDA” control in purple). F) Summary plot of the spike probability (in a 50ms post-stim. window) as a function of the number of synapses $n_{syn}$ in the stimulus. G) Onset response level $n_{syn}^{onset}$ in the PV-IN model (red, with its “uniform distribution” control in grey, p=3e-2, n=6 branches, Wilcoxon test) and SST-IN model (orange, with its “no NMDA (AMPA+)” control in purple, p=3e-2, n=6 branches, Wilcoxon test). Onset level $n_{syn}^{onset}$ is defined as the input level where the “Erf” fit goes above spike-proba=0.05. Results are shown as mean ± s.e.m over n=6 branches in panels C,D,F,G.

Figure 5:. The PV- and SST-IN dendritic programs shape two temporally-distinct inhibitory dynamics in cortical networks.

A) Response to packets of excitatory and inhibitory activity controlled by current steps in the PV-IN model (red) and in the SST-IN model (orange). We show the time-varying rate (top, grey) controlling the generation of excitatory and inhibitory events (green and red respectively), a single trial V_m example (middle) and the time-varying output rate bult from multiple trials (n=24). B) Same than A but for a time-varying input rate controlled by a temporally-correlated stochastic process (see Methods). C) Cross-correlation function between input rate and PV-IN (red) and SST-IN (orange) rate. We also show the autocorrelation function of the input rate (grey). D) Positive half-width of the cross-correlation functions (see Methods, shown as mean ± s.e.m. over N=4 input seeds x n= 6 branches; input vs PV, p=6e-3; input vs SST, p=3e-9; input vs SST-noNMDA, p=3e-3; PV vs SST:, p=3e-9; SST vs SST-noNMDA, p=3e-9; Mann-Whitney test). E) Schematic of the Neuropixels dataset from (Siegle et al., 2021). F) Single session examples in PV-cre and SST-cre (ii) mice. We show the photo-tagging trials (top-left) and the summary analysis to split positive and negative units (see Methods). We show the spiking events of positive (colored) and negative (grey) units around stimulus onset (right). G) Cross-correlation function between negative unit rates and PV-positive rate (red) or SST-positive rate (orange) for the natural movie #1 shown in all sessions. H) Half-width of the cross-correlation function in the positive and negative units. PV-cre mice: N=9 sessions x 2 movies, PV+ units vs PV- units, p=7e-2, Mann-Whitney test. SST-cre mice: N=12 sessions x 2 movies, SST+ units vs SST- units, p=2e-5, Mann-Whitney test. PV- units vs SST- units: p=5e-1, Mann-Whitney test. PV+ units vs SST+ units: p=3e-3, Mann-Whitney test. Results are shown as mean+/−sem over sessions and movies. I) (left) Illustration of the genetic mouse model approach to selectively remove NMDARs from SST-INs. (Middle) Two-photon image of SST-INs in V1 expressing GCaMP6s. (Right) Representative fluorescence traces shown as ΔF/F0 (green) together with their deconvolution traces. J) Percentage of SST-INs exhibiting a statistically significant positive response to visual stimuli at both full and half contrast in wild-type subjects and in animals lacking GluN1 subunits selectively in SST-INs. K) Deconvolved responses following stimulus presentation (average over all orientations) at half and full contrast. In the inset, we show the value of a Gaussian curve decay in the fitting window highlighted in grey. L) Half-width of the evoked response decay (evaluated by the width parameter of a Gaussian fit in the window highlighted in grey). SST:WT vs SST:GluN1-KO, p=4e-2, Mann-Whitney test.