Feedback stabilizes propagation of synchronous spiking in cortical neural networks

Precisely timed action potentials related to stimuli and behavior have been observed in the cerebral cortex. However, information carried by the precise spike timing has to propagate through many cortical areas, and noise could disrupt millisecond precision during the transmission. Previous studies have demonstrated that only strong stimuli that evoke a large number of spikes with small dispersion of spike times can propagate through multilayer networks without degrading the temporal precision. Here we show that feedback projections can increase the number of spikes in spike volleys without degrading their temporal precision. Feedback also increased the range of spike volleys that can propagate through multilayer networks. Our work suggests that feedback projections could be responsible for the reliable propagation of information encoded in spike times through cortex, and thus could serve as an attentional mechanism to regulate the flow of information in the cortex. Feedback projections may also participate in generating spike synchronization that is engaged in cognitive behaviors by the same mechanisms described here for spike propagation.The firing rates of cortical neurons carry information about sensory inputs and motor actions (1). Precisely timed action potentials related to stimuli and behavior have also been observed in the cerebral cortex (2–4), and the mechanisms underlying precisely timed spike initiation have been studied in cortical neurons in vitro (5). However, the information carried by the precise timing of spikes would have to propagate through a hierarchy of cortical areas (6) and noise could disrupt millisecond precision during the transmission.Previous modeling studies have demonstrated that synchronized volleys of spikes are essential for reliably driving the cortex by sparse thalamic inputs (7) and can indeed propagate through the layers of a feedforward network without compromising the temporal precision (8, 9). Furthermore, the temporal precision of a spike volley sharpens as it propagates through the network (10, 11). However, the results of these modeling studies suggest that only sufficiently strong stimuli that evoke spike volleys with a large number of spikes and a small dispersion of spike times would successfully propagate through the feedforward networks without degrading the temporal precision, whereas neural activities that are too weak or too dispersed will die out. This is a critical limitation on the propagation of synchronous spiking compared with the propagation of firing rates; stimuli evoking even low firing rate activity can successfully propagate through multilayer neural networks (12–14).Here we show that when feedback connections are added to a multilayer feedforward model the propagation of synchronous spiking through the network layers is significantly enhanced without compromising temporal precision.In our model with feedback projections, the state space of the model was divided into two areas: propagation and nonpropagation. In the propagation area all trajectories converged into an attractor state representing successful spike volley propagation; any spike volley starting anywhere inside this area successfully propagated through the network and reached the propagation attractor state with millisecond precision. Spike volleys starting outside the propagation area decayed after a few steps of transmission. The feedback changed the initial state of the spike volleys by moving them into the basin of the attractor for successful spike propagation. In addition, the feedback changed the position of the boundary separating propagation and nonpropagation areas, increasing the size of the basin of the propagation attractor.Feedback projections are ubiquitous in the brain (15, 16), but little is known about what they contribute to information processing (17). The results presented here provide testable hypotheses for the functional role of the feedback projections in the brain.Our model suggests that feedback projections could be responsible for allowing information encoded as spike times to propagate through cortical hierarchies, and therefore feedback projections could serve as an attentional mechanism to regulate the flow of information in the cortex. Feedback connections may also participate in generating spike-time synchronization among populations of neurons that are engaged in cognitive behaviors (18–20) by the same mechanisms described here for propagation of synchronized spikes through cortical areas.     

An arithmetic rule for spatial summation of excitatory and inhibitory inputs in pyramidal neurons

Dendritic integration of excitatory and inhibitory inputs is critical for neuronal computation, but the underlying rules remain to be elucidated. Based on realistic modeling and experiments in rat hippocampal slices, we derived a simple arithmetic rule for spatial summation of concurrent excitatory glutamatergic inputs (E) and inhibitory GABAergic inputs (I). The somatic response can be well approximated as the sum of the excitatory postsynaptic potential (EPSP), the inhibitory postsynaptic potential (IPSP), and a nonlinear term proportional to their product (k*EPSP*IPSP), where the coefficient k reflects the strength of shunting effect. The k value shows a pronounced asymmetry in its dependence on E and I locations. For I on the dendritic trunk, k decays rapidly with E–I distance for proximal Es, but remains largely constant for distal Es, indicating a uniformly high shunting efficacy for all distal Es. For I on an oblique branch, the shunting effect is restricted mainly within the branch, with the same proximal/distal asymmetry. This asymmetry can be largely attributed to cable properties of the dendrite. Further modeling studies showed that this rule also applies to the integration of multiple coincident Es and Is. Thus, this arithmetic rule offers a simple analytical tool for studying E–I integration in pyramidal neurons that incorporates the location specificity of GABAergic shunting inhibition.     

Nonlinear convergence boosts information coding in circuits with parallel outputs

Neural circuits are structured with layers of converging and diverging connectivity and selectivity-inducing nonlinearities at neurons and synapses. These components have the potential to hamper an accurate encoding of the circuit inputs. Past computational studies have optimized the nonlinearities of single neurons, or connection weights in networks, to maximize encoded information, but have not grappled with the simultaneous impact of convergent circuit structure and nonlinear response functions for efficient coding. Our approach is to compare model circuits with different combinations of convergence, divergence, and nonlinear neurons to discover how interactions between these components affect coding efficiency. We find that a convergent circuit with divergent parallel pathways can encode more information with nonlinear subunits than with linear subunits, despite the compressive loss induced by the convergence and the nonlinearities when considered separately.

Sensory systems, by necessity, compress a wealth of information gathered by receptors into the smaller amount of information needed to guide behavior. In many systems, this compression occurs via common circuit motifs—namely, convergence of multiple inputs to a single neuron and divergence of inputs to multiple parallel pathways (1). Selective nonlinear circuit elements transform inputs, selecting some parts of the signal while discarding others. Here, we investigate how these motifs work together to determine how much information is retained in compressive neural circuits.These issues are highly relevant to signaling in the retina, because the bottleneck produced by the optic nerve ensures that considerable feed-forward convergence occurs prior to the transmission of signals to central targets. This convergence reduces the dimension of signals as they traverse the retina. In total, signals from ∼100 million photoreceptors modulate the output of ∼1 million ganglion cells (2, 3). If the dynamic range of the ganglion cell is not sufficiently expanded beyond that of the photoreceptors and bipolar cells, this convergent circuit architecture could lead to a compression of input signals in which some information or stimulus resolution is lost—resulting in ambiguously encoded stimuli. It is estimated that the population of ganglion cells collectively transmits approximately 106 bits of information (3–5) and that this is much less than the amount of information available to the photoreceptors (2). However, not much is known about how neuron properties interact with a convergent circuit structure to drive or mitigate a loss of information.Receptive field subunits are a key feature of the retina’s convergent circuitry. Multiple bipolar cells converge onto a single ganglion cell—forming functional subunits within the receptive field of the ganglion cell (6, 7). Ganglion-cell responses can often be modeled as a linear sum of a population of nonlinear subunits. These subunit models have been used to investigate center-surround interactions (8–12) and to explain the nonlinear integration of signals across space (7, 10, 13–15).While it is clear that subunits have the potential to compress inputs, it is not known whether this architecture subserves an efficient code where inputs are encoded with minimal ambiguity. For decades, information theory (16, 17) has been used to quantify the amount of information that neurons encode (3, 5, 18–27). The efficient-coding hypothesis proposes that the distribution of neural responses should be one that is maximally informative about the inputs (21, 22, 28). Take the example of a stimulus variable, such as luminance, where the brightness level is encoded by the number of spikes in the response. An input/output mapping in which most of the possible luminance levels are encoded by the same response (i.e., the same number of spikes or firing rate) makes many bright and dim inputs ambiguous and provides very little information.Information can be maximized at the level of a single neuron by distributing the responses such that they optimally disambiguate inputs (23). A nonlinear response function optimized for the distribution of inputs can make the most of the neuron’s dynamic range. Adaptive rescaling of the response nonlinearity to changes in the input statistics can maintain maximal information in the output (29–31). Alternatively, information can be maximized by optimizing connection weights in the circuit, perhaps in combination with optimizing the nonlinearities (19, 32, 33). These past works, however, have not made explicit how the set of motifs found in most neural circuits, and in the retina in particular, combine to collectively influence coding efficiency.Our contribution here is to dissect a canonical neural circuit in silico and to investigate how much each of its components contribute to or detract from the information encoded by the circuit about stimuli. These circuit components, considered separately, have the potential to discard information. We begin with the simplest motif of converging inputs to single neurons and analyze the role of rectifying nonlinear subunits applied to each of these multiple inputs. We then add a diverging motif which splits the response into two opposing pathways. We find that rectifying nonlinear subunits mitigates the loss of information from convergence when compared to circuits with linear subunits. This is despite the fact that the rectifying nonlinear subunits, considered in isolation, lead to a loss of information. Moreover, this ability of nonlinear subunits to retain information stems from a reformatting of the inputs to encode distinct stimulus features compared with their linear counterparts. Our study contributes to a better understanding of how biologically inspired circuit structures and neuron properties combine to impact coding efficiency in neural circuits.     

Reversal of long-term dendritic spine alterations in Alzheimer disease models

Synapse loss is strongly correlated with cognitive impairment in Alzheimer''s disease (AD). We have previously reported the loss of dendritic spines and the presence of dystrophic neurites in both the hippocampi of transgenic mice overexpressing amyloid precursor protein (APP) and in the human brain affected with AD. In the studies reported here we have asked whether the acute alterations in dendritic spines induced by Aβ, as well as the chronic loss of spine density seen in hAPP transgenic mice, are reversible by treatments that restore the cAMP/PKA/CREB signaling pathway or proteasome function to control levels. The results show that both rolipram and TAT-HA-Uch-L1 restore spine density to near control conditions, even in elderly mice. The results suggest that changes in dendritic structure and function that occur after Aβ elevation are reversible even after long periods of time, and that one could envision therapeutic approaches to AD based on this restoration that could work independently of therapies aimed at lowering Aβ levels in the brain.     

Origin of information-limiting noise correlations

The ability to discriminate between similar sensory stimuli relies on the amount of information encoded in sensory neuronal populations. Such information can be substantially reduced by correlated trial-to-trial variability. Noise correlations have been measured across a wide range of areas in the brain, but their origin is still far from clear. Here we show analytically and with simulations that optimal computation on inputs with limited information creates patterns of noise correlations that account for a broad range of experimental observations while at same time causing information to saturate in large neural populations. With the example of a network of V1 neurons extracting orientation from a noisy image, we illustrate to our knowledge the first generative model of noise correlations that is consistent both with neurophysiology and with behavioral thresholds, without invoking suboptimal encoding or decoding or internal sources of variability such as stochastic network dynamics or cortical state fluctuations. We further show that when information is limited at the input, both suboptimal connectivity and internal fluctuations could similarly reduce the asymptotic information, but they have qualitatively different effects on correlations leading to specific experimental predictions. Our study indicates that noise at the sensory periphery could have a major effect on cortical representations in widely studied discrimination tasks. It also provides an analytical framework to understand the functional relevance of different sources of experimentally measured correlations.The response of cortical neurons to an identical stimulus varies from trial to trial. Moreover, this variability tends to be correlated among pairs of nearby neurons. These correlations, known as noise correlations, have been the subject of numerous experimental as well as theoretical studies because they can have a profound impact on behavioral performance (1–7). Indeed, behavioral performance in discrimination tasks is inversely proportional to the Fisher information available in the neural responses, which itself is strongly dependent on the pattern of correlations. In particular, correlations can strongly limit information in the sense that some patterns of correlations can lead information to saturate to a finite value in large populations, in sharp contrast to the case of independent neurons for which information grows proportionally to the number of neurons. However, the saturation is observed for only one type of correlations known as differential correlations. If the correlation pattern slightly deviates from differential correlations, information typically scales with the number of neurons, just like it does for independent neurons (7). These previous results clarify how correlations impact information and consequently behavioral performance but fail to address another fundamental question, namely, Where do noise correlations, and in particular information-limiting differential correlation, come from? Understanding the origin of information-limiting correlation is a key step toward understanding how neural circuits can increase information transfer, thereby improving behavioral performance, via either perceptual learning or attentional selection.Several groups have started to investigate sources of noise correlations such as shared connectivity (2), feedback signals (8), internal dynamics (9–11), or global fluctuations in the excitability of cortical circuits (12–16). Global fluctuations have received a lot of attention recently as they appear to account for a large fraction of the measured correlations in the primary visual cortex. Correlations induced by global fluctuations, however, do not limit information in most discrimination tasks (with the possible exception of contrast discrimination for visual stimuli). Therefore, if cortex indeed operates at information saturation, the source of information-limiting correlations is still very much unclear.In this paper, we focus on correlations induced by feedforward processing of stimuli whose information content is small compared with the information capacity of neural circuits. Using orientation selectivity as a case study, we find that feedforward processing induces correlations that share many properties of the correlations observed in vivo. Moreover, we also show feedforward processing leads to information-limiting correlations as a direct consequence of the data processing inequality. Interestingly, these information-limiting correlations represent only a small fraction of the overall correlations induced by feedforward processing, making them difficult to detect through direct measurements of correlations. Finally, we demonstrate that correlations induced by global fluctuations cannot limit information on their own, but can reduce the level at which information saturates in the presence of information-limiting correlations. Despite our focus on orientation selectivity, our results can be generalized to other modalities, stimuli, and brain areas.In summary, this work identifies a major source of noise correlations and, importantly, a source of information-limiting noise correlations, while clarifying the interactions between information-limiting correlations and correlations induced by global fluctuations.     

Internal body state influences topographical plasticity of sensory representations in the rat gustatory cortex

Primary sensory cortices are remarkably organized in spatial maps according to specific sensory features of the stimuli. These cortical maps can undergo plastic rearrangements after changes in afferent ("bottom-up") sensory inputs such as peripheral lesions or passive sensory experience. However, much less is known about the influence of "top-down" factors on cortical plasticity. Here, we studied the effect of a visceral malaise on taste representations in the gustatory cortex (GC). Using in vivo optical imaging, we showed that inducing conditioned taste aversion (CTA) to a sweet and pleasant stimulus induced plastic rearrangement of its cortical representation, becoming more similar to a bitter and unpleasant taste representation. Using a behavior task, we showed that changes in hedonic perception are directly related to the maps plasticity in the GC. Indeed imaging the animals after CTA extinction indicated that sweet and bitter representations were dissimilar. In conclusion, we showed that an internal state of malaise induces plastic reshaping in the GC associated to behavioral shift of the stimulus hedonic value. We propose that the GC not only encodes taste modality, intensity, and memory but extends its integrative properties to process also the stimulus hedonic value.     

The Rab5 guanylate exchange factor Rin1 regulates endocytosis of the EphA4 receptor in mature excitatory neurons

Ephrin signaling through Eph receptor tyrosine kinases regulates important morphogenetic events during development and synaptic plasticity in the adult brain. Although Eph-ephrin endocytosis is required for repulsive axon guidance, its role in postnatal brain and synaptic plasticity is unknown. Here, we show that Rin1, a postnatal brain-specific Rab5-GEF, is coexpressed with EphA4 in excitatory neurons and interacts with EphA4 in synaptosomal fractions. The interaction of Rin1 and EphA4 requires Rin1's SH2 domain, consistent with the view that Rin1 targets tyrosine phosphorylated receptors to Rab5 compartments. We find that Rin1 mediates EphA4 endocytosis in postnatal amygdala neurons after engagement of EphA4 with its cognate ligand ephrinB3. Rin1 was shown to suppress synaptic plasticity in the amygdala, a forebrain structure important for fear learning, possibly by internalizing synaptic receptors. We find that the EphA4 receptor is required for synaptic plasticity in the amygdala, raising the possibility that an underlying mechanism of Rin1 function in amygdala is to down-regulate EphA4 signaling by promoting its endocytosis.     

Serotonin stimulation of cAMP-dependent plasticity in Aplysia sensory neurons is mediated by calmodulin-sensitive adenylyl cyclase

Calmodulin (CaM)-sensitive adenylyl cyclase (AC) in sensory neurons (SNs) in Aplysia has been proposed as a molecular coincidence detector during conditioning. We identified four putative ACs in Aplysia CNS. CaM binds to a sequence in the C1b region of AC-AplA that resembles the CaM-binding sequence in the C1b region of AC1 in mammals. Recombinant AC-AplA was stimulated by Ca²⁺/CaM. AC-AplC is most similar to the Ca²⁺-inhibited AC5 and AC6 in mammals. Recombinant AC-AplC was directly inhibited by Ca²⁺, independent of CaM. AC-AplA and AC-AplC are expressed in SNs, whereas AC-AplB and AC-AplD are not. Knockdown of AC-AplA demonstrated that serotonin stimulation of cAMP-dependent plasticity in SNs is predominantly mediated by this CaM-sensitive AC. We propose that the coexpression of a Ca²⁺-inhibited AC in SNs, together with a Ca²⁺/CaM-stimulated AC, would enhance the associative requirement for coincident Ca²⁺ influx and serotonin for effective stimulation of cAMP levels and initiation of plasticity mediated by AC-AplA.     

GABAB receptor deficiency causes failure of neuronal homeostasis in hippocampal networks

Stabilization of neuronal activity by homeostatic control systems is fundamental for proper functioning of neural circuits. Failure in neuronal homeostasis has been hypothesized to underlie common pathophysiological mechanisms in a variety of brain disorders. However, the key molecules regulating homeostasis in central mammalian neural circuits remain obscure. Here, we show that selective inactivation of GABA_B, but not GABA_A, receptors impairs firing rate homeostasis by disrupting synaptic homeostatic plasticity in hippocampal networks. Pharmacological GABA_B receptor (GABA_BR) blockade or genetic deletion of the GB_1a receptor subunit disrupts homeostatic regulation of synaptic vesicle release. GABA_BRs mediate adaptive presynaptic enhancement to neuronal inactivity by two principle mechanisms: First, neuronal silencing promotes syntaxin-1 switch from a closed to an open conformation to accelerate soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) complex assembly, and second, it boosts spike-evoked presynaptic calcium flux. In both cases, neuronal inactivity removes tonic block imposed by the presynaptic, GB_1a-containing receptors on syntaxin-1 opening and calcium entry to enhance probability of vesicle fusion. We identified the GB_1a intracellular domain essential for the presynaptic homeostatic response by tuning intermolecular interactions among the receptor, syntaxin-1, and the Ca_V2.2 channel. The presynaptic adaptations were accompanied by scaling of excitatory quantal amplitude via the postsynaptic, GB_1b-containing receptors. Thus, GABA_BRs sense chronic perturbations in GABA levels and transduce it to homeostatic changes in synaptic strength. Our results reveal a novel role for GABA_BR as a key regulator of population firing stability and propose that disruption of homeostatic synaptic plasticity may underlie seizure''s persistence in the absence of functional GABA_BRs.Neural circuits achieve an ongoing balance between plasticity and stability to enable adaptations to constantly changing environments while maintaining neuronal activity within a stable regime. Hebbian-like plasticity, reflected by persistent changes in synaptic and intrinsic properties, is crucial for refinement of neural circuits and information storage; however, alone it is unlikely to account for the stable functioning of neural networks (1). In the last 2 decades, major progress has been made toward understanding the homeostatic negative feedback systems underlying restoration of a baseline neuronal function after prolonged activity perturbations (2–4). Homeostatic processes may counteract the instability by adjusting intrinsic neuronal excitability, inhibition-to-excitation balance, and synaptic strength via postsynaptic or presynaptic modifications (5, 6) through a profound molecular reorganization of synaptic proteins (7, 8). These stabilizing mechanisms have been collectively termed homeostatic plasticity. Homeostatic mechanisms enable invariant firing rates and patterns of neural networks composed from intrinsically unstable activity patterns of individual neurons (9).However, nervous systems are not always capable of maintaining constant output. Although some mutations, genetic knockouts, or pharmacologic perturbations induce a compensatory response that restores network firing properties around a predefined “set point” (10), the others remain uncompensated, or their compensation leads to pathological function (11). The inability of neural networks to compensate for a perturbation may result in epilepsy and various types of psychiatric disorders (12). Therefore, determining under which conditions activity-dependent regulation fails to compensate for a perturbation and identifying the key regulatory molecules of neuronal homeostasis is critical for understanding the function and malfunction of central neural circuits.In this work, we explored the mechanisms underlying the failure in stabilizing hippocampal network activity by combining long-term extracellular spike recordings by multielectrode arrays (MEAs), intracellular patch-clamp recordings of synaptic responses, imaging of synaptic vesicle exocytosis, and calcium dynamics, together with FRET-based analysis of intermolecular interactions at individual synapses. Our results demonstrate that metabotropic, G protein-coupled receptors for GABA, GABA_BRs, are essential for firing rate homeostasis in hippocampal networks. We explored the mechanisms by which GABA_BRs gate homeostatic synaptic plasticity. Our study raises the possibility that persistence of epileptic seizures in GABA_BR-deficient mice (13–15) is directly linked to impairments in a homeostatic control system.