Search

SummaryWhen stimulated, neural populations in the visual cortex exhibit fast rhythmic activity with frequencies in the gamma band (30-80 Hz). The gamma rhythm manifests as a broad resonance peak in the powerspectrum of recorded local field potentials, which exhibits various stimulus dependencies. In particular, in macaque primary visual cortex (V1), the gamma peak frequency increases with increasing stimulus contrast. Moreover, this contrast dependence is local: when contrast varies smoothly over visual space, the gamma peak frequency in each cortical column is controlled by the local contrast in that column’s receptive field. No parsimonious mechanistic explanation for these contrast dependencies of V1 gamma oscillations has been proposed. The stabilized supralinear network (SSN) is a mechanistic model of cortical circuits that has accounted for a range of visual cortical response nonlinearities and contextual modulations, as well as their contrast dependence. Here, we begin by showing that a reduced SSN model without retinotopy robustly captures the contrast dependence of gamma peak frequency, and provides a mechanistic explanation for this effect based on the observed non-saturating and supralinear input-output function of V1 neurons. Given this result, the local dependence on contrast can trivially be captured in a retinotopic SSN which however lacks horizontal synaptic connections between its cortical columns. However, long-range horizontal connections in V1 are in fact strong, and underlie contextual modulation effects such as surround suppression. We thus explored whether a retinotopically organized SSN model of V1 with strong excitatory horizontal connections can exhibit both surround suppression and the local contrast dependence of gamma peak frequency. We found that retinotopic SSNs can account for both effects, but only when the horizontal excitatory projections are composed of two components with different patterns of spatial fall-off with distance: a short-range component that only targets the source column, combined with a long-range component that targets columns neighboring the source column. We thus make a specific qualitative prediction for the spatial structure of horizontal connections in macaque V1, consistent with the columnar structure of cortex.

Neural responses to a localized visual stimulus are modulated by the content of its surrounding. This phenomenon manifests in several forms of contextual modulation, including three interrelated properties of the visual cortex: surround suppression, inverse response and surround facilitation. We devise a unified biologically realistic circuit model accounting for all these phenomena and show that i) surround suppression in L2/3 is only partially due to the recruitment of lateral inhibition; ii) long-range feedback projections are necessary for inverse response and iii) the width of the response profile in the feedback layer determines inverse size tuning. The model predicts the modulations induced by silencing somatostatin-expressing cells or higher visual areas or changing the stimulus contrast. These predictions are consistent with the experimental observations when available and can be tested in existing setups otherwise. We then show the robustness of the identified mechanisms in a model with three interneuron subclasses, built to fit the classical responses and able to predict inverse size-tuning curves.HighlightsOne model explains three different types of contextual modulation: (classical) surround suppression, (inverse) response to ‘holes’ in full field drifting gratings and cross orientation surround facilitation.Feedback, feedforward and lateral inhibitory inputs contribute to classical surround suppression in L2/3 of mouse V1 in different amounts.Observed responses to ‘holes’ in full field drifting gratings require long-range feedback projections.Surround modulation to the response to ‘holes’ in full field drifting gratings requires an increase in the characteristic length scale of the spatial pattern of activity in higher visual areas.The mechanisms uncovered by an analytically tractable model are also at work in a cell-type specific model that predicts response to a ‘hole’ stimulus.

SummaryWhen monkeys learn to group visual stimuli into arbitrary categories, lateral intraparietal area (LIP) neurons become category-selective. Surprisingly, the representations of learned categories are overwhelmingly biased: nearly all LIP neurons in a given animal prefer the same category over other behaviorally equivalent categories. We propose a model where such biased representations develop through the interplay between Hebbian plasticity and the recurrent connectivity of LIP. In this model, two separable processes of positive feedback unfold in parallel: in one, category selectivity emerges from competition between prefrontal inputs; in the other, bias develops due to lateral interactions among LIP neurons. This model reproduces the levels of category selectivity and bias observed under a variety of conditions, as well as the redevelopment of bias after monkeys learn redefined categories. It predicts that LIP receptive fields would spatially cluster by preferred category, which we experimentally confirm. In summary, our model reveals a mechanism by which LIP learns abstract representations and assigns meaning to sensory inputs.

The cortical microcircuit can dynamically adjust to dramatic changes in the strength, scale, and complexity of its input. In the primary visual cortex (V1), pyramidal cells (PCs) integrate widely across space when signals are weak, but narrowly when signals are strong, a phenomenon known as contrast-dependent surround suppression. Theoretical work has proposed that local interneurons could mediate a shift from cooperation to competition of PCs across cortical space, underlying this computation. We combined calcium imaging and electrophysiology to constrain a stabilized supralinear network model that explains how the four principal cell types in layer 2/3 (L2/3) of mouse V1– somatostatin (SST), parvalbumin (PV), and vasoactive intestinal peptide (VIP) interneurons, and PCs– transform inputs from layer 4 (L4) PCs to encode drifting gratings of varying size and contrast. Using bidirectional optogenetic perturbations, we confirmed key predictions of the model. Our data and modeling showed that recurrent amplification drives a transition from a positive PC→VIP⊣SST⊣PC feedback loop at small size and low contrast to a negative PC→SST⊣PC feedback loop at large size and high contrast to contribute to this flexible computation. This may represent a widespread mechanism for gating competition across cortical space to optimally meet task demands.

Electrophysiological recording in the primary visual cortex (V1) of mammals have revealed a number of complex interactions between the center and surround. Understanding the underlying circuit mechanisms is crucial to understanding fundamental brain computations. In this paper we address the following phenomena that have been observed in V1 of animals with orientation maps: 1) surround suppression that is accompanied by a decrease in the excitatory and inhibitory currents that the cell receives as the stimulus size increases beyond the cell’s summation field; 2) surround tuning to the center orientation, in which the strongest suppression arises when the surround orientation matches that of the center stimulus; and 3) feature-specific suppression, in which a surround stimulus of a given orientation specifically suppresses that orientation’s component of the response to a center plaid stimulus. We show that a stabilized supralinear network that has biologically plausible connectivity and synaptic efficacies that depend on cortical distance and orientation difference between neurons can consistently reproduce all the above phenomena. We explain the mechanism behind each result, and argue that feature-specific suppression and surround tuning to the center orientation are independent phenomena. Specifically, if we remove some aspects of the connectivity from the model it will still produce feature-specific suppression but not surround tuning to the center orientation. We also show that in the model the activity decay time constant is similar to the cortical activity decay time constant reported in mouse V1. Our model indicates that if the surround activates neurons that fall within the reach of the horizontal projections in V1, the above mentioned phenomena can be generated by V1 alone without the need of cortico-cortical feedback. Finally, we show that these results hold both in networks with rate-based units and with conductance-based spiking units. This demonstrates that the stabilized supra-linear network mechanism can be achieved in the more biological context of spiking networks.

Predicting the response of the cortical microcircuit to perturbations is a prerequisite to determine the mechanisms that mediate its response to stimulus; yet, an encompassing perspective that describes the full ensemble of the network’s response in models that accurately recapitulate recorded data is still lacking. Here we develop a class of mathematically tractable models that exactly describe the modulation of the distribution of cell-type-specific calcium-imaging activity with the contrast of a visual stimulus. The inferred parameters recover signatures of the connectivity structure found in mouse visual cortex. Analysis of this structure subsequently reveals parameter-independent relations between the responses of different cell types to perturbations and each interneuron’s role in circuit-stabilization. Leveraging recent theoretical approaches, we derive explicit expressions for the distribution of responses to partial perturbations which reveal a novel, counter intuitive effect in the sign of response functions. Finally applying the theory to inferring feedback to V1 during locomotion, we find that it is predominantly mediated by both SOM and VIP modulation.

An inhibition-stabilized network (ISN) is a network of excitatory and inhibitory cells at a stable fixed point of firing rates for a given input, for which the excitatory subnetwork would be unstable if inhibitory rates were frozen at their fixed point values. It has been shown that in a low-dimensional model (one unit per neuronal subtype) of an ISN with a single excitatory and single inhibitory cell type, the inhibitory unit shows a “paradoxical” response, lowering (raising) its steady-state firing rate in response to addition to it of excitatory (inhibitory) input. This has been generalized to an ISN with multiple inhibitory cell types: if input is given only to inhibitory cells, the steady-state inhibition received by excitatory cells changes paradoxically, that is, it decreases (increases) if the steady-state excitatory firing rates decrease (increase).We generalize these analyses of paradoxical effects to low-dimensional networks with multiple cell types of both excitatory and inhibitory neurons. The analysis depends on the connectivity matrix of the network linearized about a given fixed point, and its eigenvectors or “modes”. We show the following: (1) A given cell type shows a paradoxical change in steady-state rate in response to input it receives, if and only if the network with that cell type omitted has an odd number of unstable modes. Excitatory neurons can show paradoxical responses when there are two or more inhibitory subtypes. (2) More generally, if the cell types are divided into two nonoverlapping subsets A and B, then subset B has an odd (even) number of modes that show paradoxical response if and only if subset A has an odd (even) number of unstable modes. (3) The net steady-state inhibition received by any unstable mode of the excitatory subnetwork will change paradoxically, i.e. in the same direction as the change in amplitude of that mode. In particular, this means that a sufficient condition to determine that a network is an ISN is if, in response to an input only to inhibitory cells, the firing rates of and inhibition received by all excitatory cell types all change in the same direction. This in turn will be true if all E cells and all inhibitory cell types that connect to E cells change their firing rates in the same direction.

SO_SCPLOWUMMARYC_SCPLOWContext strongly guides perception by influencing the saliency of sensory stimuli. Accordingly, sensory neurons respond differently to the same stimulus depending on the context in which the stimulus is embedded. In the visual system, the relationship between the features of the stimulus and those of the surround determine how excitatory neurons are modulated. Their responses are suppressed when center and surround share the same features but not when they are different. The mechanisms that relieve surround suppression and thereby generate contextual modulation remain unclear. Here we show that the disinhibitory circuit consisting of somatostatin-expressing (SOM) and vasoactive-intestinal-peptide-expressing (VIP) inhibitory neurons contextually modulates excitatory neurons in mouse primary visual cortex. When the stimulus and the surround share the same features, VIP neurons are inactive and SOM neurons can suppress excitatory neurons. However, when the features of the stimulus differ from those of the surround, VIP neurons are activated, thereby inhibiting SOM neurons and relieving excitatory neurons from suppression. Optogenetic silencing of VIP neurons, via disinhibition of SOM neurons, reduces the contextual modulation of excitatory neurons. We have identified a canonical cortical disinhibitory circuit which contributes to contextual modulation of excitatory neurons, and may thereby regulate perceptual saliency.

Selective visual attention modulates neural activity in the visual system in complex ways and leads to enhanced performance on difficult visual tasks. Here, we show that a simple circuit model, the stabilized supralinear network, gives a unified account of a wide variety of effects of attention on neural responses. We replicate results from studies of both feature and spatial attention, addressing findings in a variety of experimental paradigms on changes both in firing rates and in correlated neural variability. Finally, we expand this circuit model into an architecture that can perform visual tasks—a convolutional neural network—in order to show that these neural effects can enhance detection performance. This work provides the first unified mechanistic account of the effects of attention on neural and behavioral responses.

One of the most fundamental organizational principles of the brain is the separation of excitatory (E) and inhibitory (I) neurons. In addition to their opposing effects on post-synaptic neurons, E and I cells tend to differ in their selectivity and connectivity. Although many such differences have been characterized experimentally, it is not clear why they exist in the first place. We studied this question in an artificial neural network equipped with multiple E and I cell types. We found that a deep convolutional recurrent network trained to perform an object classification task was able to capture salient distinctions between E and I neurons. We explored the necessary conditions for the network to develop distinct selectivity and connectivity across cell types. We found that neurons that project to higher-order areas will have greater stimulus selectivity, regardless of whether they are excitatory or not. Sparser connectivity is required for higher selectivity, but only when the recurrent connections are excitatory. These findings demonstrate that the differences observed across E and I neurons are not independent, and can be explained using a smaller number of factors.

SummaryVariability and correlations in cortical activity are ubiquitously modulated by stimuli. Correlated variability is quenched following stimulus onset across multiple cortical areas, suppressing low-frequency components of the LFP and of Vm-LFP coherence. Modulation of Fano factors and correlations in area MT is tuned for stimulus direction. What circuit mechanisms underly these behaviors? We show that a simple model circuit, the stochastic Stabilized Supralinear Network (SSN), robustly explains these results. Stimuli modulate variability by modifying two forms of effective connectivity between activity patterns that characterize excitatory-inhibitory (E/I) circuits. Increases in the strength with which activity patterns inhibit themselves reduce correlated variability, while increases in feedforward connections between patterns (transforming E/I imbalance into balanced fluctuations) increase variability. These results suggest an operating regime of cortical dynamics that involves fast fluctuations and fast responses to stimulus changes, unlike previous models of variability suppression through suppression of chaos or networks with multiple attractors.