Fig 1.
Biophysically detailed multi-scale model of whisker deflection evoked responses in cortical pyramidal tract neurons.
A: Sensory-evoked signal flow: stimuli of single whiskers (which are arranged in ‘arcs’ and ‘rows’ on the animal’s snout) are transmitted to the brainstem (BS), from there to the VPM thalamus, and from there to the primary sensory cortex of the vibrissal system (vS1). This pathway is somatotopically organized, with barreloids in VPM and barrels in vS1 corresponding to the respective whiskers. B: In vivo labeled L5PT dendrite morphologies used in this study. C: Corresponding receptive fields to passive single whisker touch, measured in vivo (upper panels). Average receptive field across 9 in vivo recorded L5PTs (lower panel). Error bars are std. D: Network model of rat vS1 and VPM provides anatomically realistic estimates of which neurons are connected to a L5PT embedded into the network. In this study, the simulated neurons are located in the C2 column of vS1, thus we refer to the somatotopically aligned C2 whisker as the ‘principal’ whisker, and the adjacent whiskers as surround whiskers. Red and blue markers denote soma locations of presynaptic excitatory and inhibitory somata, respectively. E: Synapse distribution originating from the neuron shown in Panel B. F: Spatiotemporal input pattern to L5PT: combining the anatomical constraints with empirical measurements of the activity of different presynaptic populations (S1 Fig) provides spatiotemporal input patterns that the L5PT can receive after sensory stimulation. G: Trial-to-trial activity of example synapses matching the soma distance from panel F for a principal whisker (C2) and surround whisker (D2) stimulus. H: Biophysically detailed multi-compartmental L5PT models reproduce the cell type’s characteristic electrophysiology (left panel), i.e. back propagation of APs (upper left), dendritic Ca-APs and somatic burst firing (upper right), as well as regular firing properties (lower row). Right panel: biophysically detailed neuron morphologies at the moment of a dendritic Ca AP. I: Simulated response to principal whisker touch. J: Simulated receptive fields across morphologically and biophysically diverse L5PT multi-compartmental models across 81 network embeddings capture broad and heterogeneous receptive fields.
Fig 2.
Input-output computation of L5PTs upon single whisker deflections.
A: Exemplary responses of the multi-compartmental model with respect to the prediction time point (the time point for which the occurrence of an AP is to be predicted) for the three relevant response categories ‘AP’, ‘no AP’, and ‘recent AP’ (AP was elicited shortly before the prediction time point). B: Spatiotemporal input filter that best separates AP and no AP trials assigns strong weight to proximal synapses (top) active in a short time window before the prediction time point (bottom). C: Nonlinear relationship between WNI and AP probability. WNI represents the ‘drive’ a neuron receives; the higher the WNI the higher the probability an AP will be generated. D: Weighted net input–the input filtered by the spatiotemporal filter–separates AP and no AP trials, but not ‘recent AP’ trials, which can be distinguished based on a second measure, ‘time to previous AP’.E: Reduced model structure. APs are generated stochastically based on the AP probability (output of the nonlinearity). If an AP is generated, subsequent APs become less likely due to the post AP penalty, which is subtracted from the WNI. This reduced model directly relates AP output to synaptic input and previously generated APs in the simulated in vivo condition. F: The reduced model’s responses match the biophysically detailed model across many trials (close PSTH match) and on the single trial level (high AUROC score across all time points). Without the post AP penalty, the AUROC score drops during the sensory-evoked response.
Fig 3.
Input-output computation is robust to morphological and biophysical diversity.
A: Reduced models inferred on the different multi-compartmental models are qualitatively similar, with similar temporal and spatial filters, nonlinearity and post AP penalty. B: All models have high AUROC scores, specifically during the sensory-evoked (peak) response. C: We quantify latency, spontaneous AP rate (before the stimulus) and response probability for each pair of multi-compartmental and reduced model. D: Comparing response properties between multi-compartmental and corresponding reduced model shows close match.
Fig 4.
Reduced models predict origins of receptive field variability.
A: Comparison of responses of 7 different multi-compartmental models and their corresponding reduced models to 9 different whisker stimuli (PW and 8 SW) in 81 different network embedding locations. Response probability is the probability that one or more APs are generated 0-25ms after the sensory stimulus. B: Comparison of exemplary receptive field shapes shows close match between biophysically detailed and reduced model. C: Quantification of receptive field similarity for all cell positions and biophysically detailed models. D: Comparison between in vivo and reduced model responses to 9 different whisker stimuli (PW and 8 SW). Green dots represent the mean response probability. E: Exemplary receptive fields of reduced models. F: Influence of biophysics, morphology and cell position on receptive field shape, quantified by computing the correlation coefficient between receptive fields if one of these properties is changed.
Fig 5.
Reduced models predict contribution of input pathways to sensory responses.
A: Absolute contribution of presynaptic populations to WNI following a PW stimulus to a model located at the center of the C2 column: pyramidal neurons in L2/3 (L2PY, L3PY), spiny neurons in L4 (L4SP), intratelencephalic neurons in L5 (L5IT), L5PT, corticocortical neurons at the L5/6 border (L6CC), and relay cells in the ventral posterior medial nucleus of thalamus (VPM). Despite the lack of sensory-evoked responses by L4PY, L6CT and L6INV [14], their contributions were considered in the overall WNI calculations (Figs 2 and 3). B: Sensory-evoked contribution (i.e. absolute contribution minus baseline for each input pathway) of presynaptic populations to WNI following a PW stimulus. C: Sensory-evoked contribution of presynaptic populations to WNI following a SW stimulus. D: Contribution of the main input pathways–VPM and L6CC–depending on the soma location of the L5PT model in a 9x9 grid across the C2 column for a PW stimulus. The black circle denotes the C2 column border. E: Comparison between model responses to 9 different whisker stimuli (PW and 8 SW) under control conditions and when removing sensory-evoked input from L6CC. Removing evoked L6CC activity attenuates responses, in particular to surround whiskers.