Active dendrites enable strong but sparse inputs to determine orientation selectivity

The dendrites of neocortical pyramidal neurons are excitable. However, it is unknown how synaptic inputs engage nonlinear dendritic mechanisms during sensory processing in vivo, and how they in turn influence action potential output. Here, we provide a quantitative account of the relationship between synaptic inputs, nonlinear dendritic events, and action potential output. We developed a detailed pyramidal neuron model constrained by in vivo dendritic recordings. We drive this model with realistic input patterns constrained by sensory responses measured in vivo and connectivity measured in vitro. We show mechanistically that under realistic conditions, dendritic Na⁺ and NMDA spikes are the major determinants of neuronal output in vivo. We demonstrate that these dendritic spikes can be triggered by a surprisingly small number of strong synaptic inputs, in some cases even by single synapses. We predict that dendritic excitability allows the 1% strongest synaptic inputs of a neuron to control the tuning of its output. Active dendrites therefore allow smaller subcircuits consisting of only a few strongly connected neurons to achieve selectivity for specific sensory features.

There is longstanding evidence from in vitro experiments that dendrites of mammalian neurons are electrically excitable (1, 2), and theoretical work has demonstrated that these active properties can be exploited for computations so that single neurons can perform functions that could otherwise only be performed by a network (3–7). Recently, technical breakthroughs have enabled dendritic integration to be studied in vivo using both imaging and electrophysiological techniques (8, 9). These experiments have revealed that the integration of synaptic events in vivo can be highly nonlinear and that this process influences the response properties of single neurons and neuronal populations in vivo (10–16). For example, patch-clamp recordings from dendrites in mouse primary visual cortex (V1) have demonstrated that dendritic spikes are triggered by visual input and that they may contribute to the orientation selectivity of somatic membrane potential (17). However, important mechanistic questions are still unanswered. How many synaptic inputs must be locally coactive on a dendrite to recruit dendritic spikes? What is the contribution of individual dendritic spikes to somatic action potential (AP) output and its orientation selectivity? Moreover, we do not understand how the answers to these questions depend on the type of dendritic spike. Finally, how do active dendrites, by supporting dendritic spikes, influence which synaptic inputs control AP output and its tuning?These issues are extremely challenging to address experimentally. We have therefore taken a modeling approach, constrained by in vitro and in vivo experimental data, in order to provide a quantitative understanding of the relationship between synaptic input, dendritic spikes, and AP output during sensory processing in V1. We constructed a detailed active model of a layer (L) 2/3 pyramidal neuron in mouse V1 and combined this with a model of the presynaptic inputs it receives during visual stimulation with drifting gratings in vivo (17–21).Our model reproduces key features of the experimental data on dendritic and somatic responses to visual stimulation as observed in vivo and allows us to identify the synaptic inputs that trigger dendritic Na⁺ spikes and NMDA spikes. We also provide a quantitative explanation for how these dendritic spikes determine neuronal output in vivo. Our results show that dendritic spikes can be triggered by a surprisingly small number of synaptic inputs—in some cases even by single synapses. We also find that during sensory processing, already few dendritic spikes are effective at driving somatic output. Overall, this strategy allows a remarkably small number of strong synaptic inputs to dominate neural output, which may reduce the number of neurons required to represent a given sensory feature.     

Collapse of complexity of brain and body activity due to excessive inhibition and MeCP2 disruption

Complex body movements require complex dynamics and coordination among neurons in motor cortex. Conversely, a long-standing theoretical notion supposes that if many neurons in motor cortex become excessively synchronized, they may lack the necessary complexity for healthy motor coding. However, direct experimental support for this idea is rare and underlying mechanisms are unclear. Here we recorded three-dimensional body movements and spiking activity of many single neurons in motor cortex of rats with enhanced synaptic inhibition and a transgenic rat model of Rett syndrome (RTT). For both cases, we found a collapse of complexity in the motor system. Reduced complexity was apparent in lower-dimensional, stereotyped brain–body interactions, neural synchrony, and simpler behavior. Our results demonstrate how imbalanced inhibition can cause excessive synchrony among movement-related neurons and, consequently, a stereotyped motor code. Excessive inhibition and synchrony may underlie abnormal motor function in RTT.

A diverse and complex repertoire of body movements requires diverse and complex neural activity among cortical neurons. Moreover, interactions between movement-related neurons and the body must be sufficiently high dimensional to carry these movement signals with high fidelity. The complexity of movement-related neural activity and neuron–body interactions can be compromised if synchrony among neurons is excessive. Indeed, it is well understood theoretically that excessive correlations can limit the information capacity of any neural code (1–3)—if all neurons are perfectly synchronized, then different neurons cannot encode different motor signals. Synchrony is also known to play a role in pathophysiology of movement-related disorders, like Parkinson’s disease (4–6). However, synchrony and correlations also contribute to healthy function in the motor system (7–14). For instance, particular groups of synchronized neurons seem to send control signals to particular muscle groups (7, 8) and propagation of correlated firing contributes to motor planning (10). Synchrony can also play a role in motor learning (12–14). These findings suggest that correlated activity among specific subsets of neurons encodes specific motor functions. Thus, it stands to reason that if this synchrony became less selective and more stereotyped across neurons, then the motor code would become less complex and lose specificity, resulting in compromised motor function.Here we explored this possibility in rats, in the caudal part of motor cortex where neurons associated with hindlimb, forelimb, and trunk body movement are located (15–17). We focused on two conditions. First, we studied a transgenic rat model of Rett syndrome (RTT), which has disrupted expression of the MeCP2 gene. Second, we studied normal rats with acutely altered inhibitory neural interactions. Both of these cases are associated with abnormal motor behavior and, possibly, abnormal synchrony. Abnormal synchrony is a possibility, because both of these cases are linked to an imbalance between excitatory (E) and inhibitory (I) neural interactions, which in turn is likely to result in abnormal synchrony. For instance, many computational models suggest that synchrony is strongly dependent on E/I interactions (18–21). Likewise, in experiments, pharmacological manipulation of E/I causes changes in synchrony (19, 22, 23) and the excessive synchrony that occurs during epileptic seizures is often attributed to an E/I imbalance (24, 25). Similarly, the majority of people with RTT suffer from seizures (26) and many previous studies establish E/I imbalance as a common problem in RTT (27). MeCP2 dysfunction, which is known to cause RTT, seems to be particularly important in inhibitory neurons (28). For instance, two studies have shown that disrupting MeCP2 only in specific inhibitory neuron types can recapitulate the effects of brain-wide disruption of MeCP2 (29, 30). However, whether the E/I imbalance favors E or I at the population level seems to vary across different brain regions in RTT. Studies of visual cortex (29) and hippocampus (31) suggest that the balance tips toward too much excitation (perhaps explaining the prevalence of seizures), while studies of somatosensory cortex (32, 33) and a brain-wide study of Fos expression (34) suggest that frontal areas, including motor cortex, are tipped toward excessive inhibition. These facts motivated our choice to study pharmacological disruption of inhibition here. While it is clear that E/I imbalance is important in RTT, it is much less clear how it manifests at the level of dynamics and complexity of neural activity that is responsible for coordinating body movements. Thus, in addition to pursuing the general questions about synchrony and complexity in the motor system discussed above, a second goal of our work was to improve understanding of motor dysfunction due to MeCP2 disruption.Taken together, these facts led us to the following questions: How does MeCP2 disruption impact the complexity of body movements, movement-related neural activity, and motor coding? Are abnormalities in the MeCP2-disrupted motor system consistent with excessive inhibition in motor cortex? We hypothesized that both MeCP2 disruption and excessive inhibition lead to reduced complexity of interactions between cortical neurons and body movements, excessive cortical synchrony, and reduced complexity of body movements. Our findings confirmed this hypothesis and suggest that RTT-related motor dysfunction may be due, in part, to excessive synchrony and inhibition in motor cortex.     

Gating of steering signals through phasic modulation of reticulospinal neurons during locomotion

The neural control of movements in vertebrates is based on a set of modules, like the central pattern generator networks (CPGs) in the spinal cord coordinating locomotion. Sensory feedback is not required for the CPGs to generate the appropriate motor pattern and neither a detailed control from higher brain centers. Reticulospinal neurons in the brainstem activate the locomotor network, and the same neurons also convey signals from higher brain regions, such as turning/steering commands from the optic tectum (superior colliculus). A tonic increase in the background excitatory drive of the reticulospinal neurons would be sufficient to produce coordinated locomotor activity. However, in both vertebrates and invertebrates, descending systems are in addition phasically modulated because of feedback from the ongoing CPG activity. We use the lamprey as a model for investigating the role of this phasic modulation of the reticulospinal activity, because the brainstem–spinal cord networks are known down to the cellular level in this phylogenetically oldest extant vertebrate. We describe how the phasic modulation of reticulospinal activity from the spinal CPG ensures reliable steering/turning commands without the need for a very precise timing of on- or offset, by using a biophysically detailed large-scale (19,600 model neurons and 646,800 synapses) computational model of the lamprey brainstem–spinal cord network. To verify that the simulated neural network can control body movements, including turning, the spinal activity is fed to a mechanical model of lamprey swimming. The simulations also predict that, in contrast to reticulospinal neurons, tectal steering/turning command neurons should have minimal frequency adaptive properties, which has been confirmed experimentally.In many vertebrate and invertebrate motor systems, a phasic modulation occurs in the descending control system determining the level of activity (1–3) during rhythmic movements. The physiological role of this modulation has remained enigmatic, because it has been shown that tonic activity is sufficient to effectively drive motor activity like locomotion. One example is the reticulospinal neurons in the brainstem that serve as the major interface between higher level commands and the networks in the spinal cord in all vertebrates from lamprey to primates (4–6). In this study, we investigate the motor system of the lamprey, belonging to the most ancient group of vertebrates that has been investigated in considerable detail not only at the brainstem–spinal cord level but also with regard to the forebrain systems underlying the control of action (2, 7, 8). A bilateral symmetric activation of reticulospinal neurons will activate the locomotor networks in the spinal cord, resulting in coordinated swimming movements (2, 9–11). The reticulospinal neurons act on the excitatory and inhibitory network interneurons in the spinal cord through NMDA and AMPA receptors (12). Most reticulospinal neurons can be involved in several motor patterns (13). Whereas a bilaterally symmetric activation leads to locomotion, a unilateral addition of excitation to one side will enhance motor activity on this side and result in turning. This is the basis of steering during locomotion (14).The phasic modulation of reticulospinal neurons is most pronounced in the fastest-conducting group involved in steering (15). It results from feedback from the network neurons in the rostral segments of the spinal cord during locomotor movements, so that the reticulospinal neurons become active in phase with those segments, and during the inactive period they are instead inhibited (15–19). This feedback is conveyed to reticulospinal neurons via ascending spinobulbar neurons (15–20) that provide an “efference copy” regarding the cycle-to-cycle activity in the locomotor network. Spinobulbar neurons (21) provide the excitatory and inhibitory drive to reticulospinal neurons, resulting in modulation of their activity in phase with the ipsilateral rostral parts of the spinal cord (21, 22) and this forms a closed spino-reticulo-spinal loop (17).Visuomotor coordination (23) and steering results from activation of tectal output neurons that monosynaptically activate reticulospinal neurons (24, 25), which represent the interface between tectum and the spinal cord networks. Here, we focus on the role of the phasic modulation of the reticulospinal neurons. We show that the phasic modulation of the reticulospinal cells is advantageous in that the steering commands from tectum become gated and thereby arrive in the correct phase of the locomotor cycle. The tectal commands, therefore, need not be timed very precisely in relation to the locomotor cycle. The reticulospinal modulation ascertains that the command signal will be accurately timed provided that the tectal command signal itself remains constant and thus has a limited spike-frequency adaptation, which indeed applies to the tectal output neurons as shown here experimentally (25). We explore the effect of steering signals through a combined simulation-experimental approach. Biologically detailed lamprey spinal cell models (26) are used in large-scale simulations of the locomotor network (7) to replicate electric activity in command centers and the spinal cord. A mechanical model of swimming (27, 28) is used for a quantitative evaluation of the locomotor response.     

Pop-out search instigates beta-gated feature selectivity enhancement across V4 layers

Visual search is a workhorse for investigating how attention interacts with processing of sensory information. Attentional selection has been linked to altered cortical sensory responses and feature preferences (i.e., tuning). However, attentional modulation of feature selectivity during search is largely unexplored. Here we map the spatiotemporal profile of feature selectivity during singleton search. Monkeys performed a search where a pop-out feature determined the target of attention. We recorded laminar neural responses from visual area V4. We first identified “feature columns” which showed preference for individual colors. In the unattended condition, feature columns were significantly more selective in superficial relative to middle and deep layers. Attending a stimulus increased selectivity in all layers but not equally. Feature selectivity increased most in the deep layers, leading to higher selectivity in extragranular layers as compared to the middle layer. This attention-induced enhancement was rhythmically gated in phase with the beta-band local field potential. Beta power dominated both extragranular laminar compartments, but current source density analysis pointed to an origin in superficial layers, specifically. While beta-band power was present regardless of attentional state, feature selectivity was only gated by beta in the attended condition. Neither the beta oscillation nor its gating of feature selectivity varied with microsaccade production. Importantly, beta modulation of neural activity predicted response times, suggesting a direct link between attentional gating and behavioral output. Together, these findings suggest beta-range synaptic activation in V4’s superficial layers rhythmically gates attentional enhancement of feature tuning in a way that affects the speed of attentional selection.

Throughout cortex, sensory information is organized into maps. This phenomenon is readily observable in visual cortex where maps organize information in both the radial (e.g., within cortical columns) and tangential (e.g., across a cortical area) dimensions (1–4). Importantly, sensory information attributed to these maps is malleable. For example, selective attention is linked to profound changes in neural activity organizing sensory information in both space and time (5–34).In visual cortex, cortical columnar microcircuits comprise many neurons that respond to the same location of visual space and similar stimulus features. For example, primary visual cortex (V1) features “orientation columns” consisting of neurons sharing response preference for the same stimulus orientation (35, 36) and “ocular dominance columns” consisting of neurons that preferentially respond to the same eye (37). Similar columnar organization for feature selectivity has been described across many other visual cortical areas, including area V2 (36, 38, 39), area V3 (40), middle temporal area (area MT) (41–43), and inferotemporal cortex (44–46). Midlevel visual cortical area V4, a well-studied area contributing to attentional modulation, follows suit with columnar organization of visual responses and feature preferences (44, 47–53). Yet, we do not know the extent to which attention impacts feature preferences along columns. While canonical microcircuit models of cortex predict laminar differences for attentional modulation [e.g., feedback-recipient extragranular layers modulating before granular layers (54–56)], how this modulation interacts with columnar feature selectivity is largely unknown.We sought to determine the spatiotemporal profile of feature preferences within the V4 laminar microcircuit during attentional selection. To address this question, we performed neurophysiological recordings along V4 layers in monkeys performing an attention-demanding pop-out search task. We identified feature columns demonstrating homogeneous feature preference along cortical depth. When the search array item presented in the column’s receptive field (RF) was unattended, the upper cortical layers were most selective. However, when attended, feature selectivity in the deep layers enhanced the most, resulting in overall strongest feature selectivity in both extragranular compartments. We further found that the enhancement of feature selectivity associated with attention was rhythmically gated in the beta range. While beta activity was measurable across both unattended and attended conditions, rhythmic gating of feature selectivity was only present with attention. Moreover, beta power modulating the neural response was predictive of response time (RT), suggesting a link between attentional gating and behavior. Synaptic currents revealed the beta rhythm originates in superficial cortical layers, which is compatible with top-down influence.     

Local generation of multineuronal spike sequences in the hippocampal CA1 region

Sequential activity of multineuronal spiking can be observed during theta and high-frequency ripple oscillations in the hippocampal CA1 region and is linked to experience, but the mechanisms underlying such sequences are unknown. We compared multineuronal spiking during theta oscillations, spontaneous ripples, and focal optically induced high-frequency oscillations (“synthetic” ripples) in freely moving mice. Firing rates and rate modulations of individual neurons, and multineuronal sequences of pyramidal cell and interneuron spiking, were correlated during theta oscillations, spontaneous ripples, and synthetic ripples. Interneuron spiking was crucial for sequence consistency. These results suggest that participation of single neurons and their sequential order in population events are not strictly determined by extrinsic inputs but also influenced by local-circuit properties, including synapses between local neurons and single-neuron biophysics.A hypothesized hallmark of cognition is self-organized sequential activation of neuronal assemblies (1). Self-organized neuronal sequences have been observed in several cortical structures (2–5) and neuronal models (6–7). In the hippocampus, sequential activity of place cells (8) may be induced by external landmarks perceived by the animal during spatial navigation (9) and conveyed to CA1 by the upstream CA3 region or layer 3 of the entorhinal cortex (10). Internally generated sequences have been also described in CA1 during theta oscillations in memory tasks (4, 11), raising the possibility that a given neuronal substrate is responsible for generating sequences at multiple time scales. The extensive recurrent excitatory collateral system of the CA3 region has been postulated to be critical in this process (4, 7, 12, 13).The sequential activity of place cells is “replayed” during sharp waves (SPW) in a temporally compressed form compared with rate modulation of place cells (14–20) and may arise from the CA3 recurrent excitatory networks during immobility and slow wave sleep. The SPW-related convergent depolarization of CA1 neurons gives rise to a local, fast oscillatory event in the CA1 region (“ripple,” 140–180 Hz; refs. 8 and 21). Selective elimination of ripples during or after learning impairs memory performance (22–24), suggesting that SPW ripple-related replay assists memory consolidation (12, 13). Although the local origin of the ripple oscillations is well demonstrated (25, 26), it has been tacitly assumed that the ripple-associated, sequentially ordered firing of CA1 neurons is synaptically driven by the upstream CA3 cell assemblies (12, 15), largely because excitatory recurrent collaterals in the CA1 region are sparse (27). However, sequential activity may also emerge by local mechanisms, patterned by the different biophysical properties of CA1 pyramidal cells and their interactions with local interneurons, which discharge at different times during a ripple (28–30). A putative function of the rich variety of interneurons is temporal organization of principal cell spiking (29–32). We tested the “local-circuit” hypothesis by comparing the probability of participation and sequential firing of CA1 neurons during theta oscillations, natural spontaneous ripple events, and “synthetic” ripples induced by local optogenetic activation of pyramidal neurons.     

Homeostatic regulation of axonal Kv1.1 channels accounts for both synaptic and intrinsic modifications in the hippocampal CA3 circuit

Homeostatic plasticity of intrinsic excitability goes hand in hand with homeostatic plasticity of synaptic transmission. However, the mechanisms linking the two forms of homeostatic regulation have not been identified so far. Using electrophysiological, imaging, and immunohistochemical techniques, we show here that blockade of excitatory synaptic receptors for 2 to 3 d induces an up-regulation of both synaptic transmission at CA3–CA3 connections and intrinsic excitability of CA3 pyramidal neurons. Intrinsic plasticity was found to be mediated by a reduction of Kv1.1 channel density at the axon initial segment. In activity-deprived circuits, CA3–CA3 synapses were found to express a high release probability, an insensitivity to dendrotoxin, and a lack of depolarization-induced presynaptic facilitation, indicating a reduction in presynaptic Kv1.1 function. Further support for the down-regulation of axonal Kv1.1 channels in activity-deprived neurons was the broadening of action potentials measured in the axon. We conclude that regulation of the axonal Kv1.1 channel constitutes a major mechanism linking intrinsic excitability and synaptic strength that accounts for the functional synergy existing between homeostatic regulation of intrinsic excitability and synaptic transmission.

Chronic modulation of activity regimes in neuronal circuits induces homeostatic plasticity. This implicates a regulation of both intrinsic excitability (homeostatic plasticity of intrinsic excitability) and synaptic transmission (homeostatic plasticity of synaptic transmission) to maintain network activity within physiological bounds (1). In most cases, these two forms of homeostatic plasticity act synergistically but involve different molecular actors. Homeostatic intrinsic plasticity is associated with the regulation of voltage-gated ion channels (2–9), while homeostatic synaptic plasticity involves the regulation of postsynaptic receptors to neurotransmitters (10–17) or the regulation of the readily releasable pool of synaptic vesicles (18–20). However, the function of voltage-gated ion channels is not limited to the control of intrinsic excitability. Several studies point to the role of axonal voltage-gated channels in shaping presynaptic action potential (AP) waveform and subsequently controlling neurotransmitter release and synaptic transmission (21–35). Moreover, some studies describe homeostatic plasticity of the AP waveform via voltage-gated channel regulation (36–38), while other studies report an absence of this phenomenon (39).Kv1.1 channels are responsible for the fast-activating, slow-inactivating D-type current (I_D) in CA3 neurons. This current has been shown to create a delay in the onset of the first AP and to determine intrinsic excitability in various neuronal types, including CA1 and CA3 pyramidal neurons of the hippocampus (3, 40), L5 pyramidal neurons of the cortex (26, 34), and L2/3 fast-spiking interneurons of the somatosensory cortex (41, 42). Furthermore, Kv1.1 channels have been shown to control axonal AP width and subsequently presynaptic calcium entry and neurotransmitter release. In fact, pharmacological blockade of Kv1.1 channels broadens presynaptic APs and increases synaptic transmission at neocortical and hippocampal glutamatergic synapses and at cerebellar GABAergic synapses (21, 22, 26, 30, 32, 41, 43, 44). Moreover, Kv1.1 channels have been shown to be responsible for the phenomenon of depolarization-induced analog digital facilitation of synaptic transmission (d-ADF). In fact, at CA3–CA3 and L5–L5 synapses, a somatic subthreshold depolarization of the presynaptic cell leads to inactivation of axonal Kv1.1 channels, inducing the broadening of the presynaptic AP, an increase in spike-evoked calcium entry, and a facilitation of presynaptic glutamate release (26, 31, 33, 34, 45, 46). Therefore, Kv1.1 channels control both intrinsic excitability and glutamate release in CA3 pyramidal neurons.Kv1.1 channels have been shown to be involved in homeostatic regulation of neuronal excitability. Chronic activity enhancement by kainate application leads to an increase in I_D current and a decrease in excitability in dentate gyrus granule cells (47). Conversely, chronic sensory deprivation leads to Kv1.1 channel down-regulation and enhancement of excitability in the avian cochlear nucleus (48). In this study, we examined whether the increase in synaptic transmission could also be due to Kv1.1 channel down-regulation, which would possibly explain the observed synergy between homeostatic plasticity of excitability and synaptic transmission.We show here that chronic activity deprivation induced with an antagonist of ionotropic glutamate receptors (kynurenate) in hippocampal organotypic cultures provokes both an increase in CA3 pyramidal cells excitability and an enhancement of synaptic transmission at monosynaptically connected CA3 neurons. Deprived cultures display a decrease in Kv1.1 channel staining in the axon initial segment. Bath application of dendrotoxin-K (DTX-K), a selective blocker of Kv1.1 channels, leads to a larger excitability increase in control cultures than in deprived cultures. Focal puffing of DTX-K on the axon increases excitability in control but not in deprived cultures, showing that homeostatic plasticity of excitability in deprived cultures is partly due to the down-regulation of axonal Kv1.1 channels. In addition, we found that axonal Kv1.1 down-regulation in deprived cultures is responsible for a spike broadening in CA3 neurons, leading to elevated release probability at CA3–CA3 synapses. Consistent with these observations, d-ADF, a Kv1.1-dependent form of synaptic facilitation, is present in control cultures but not in deprived cultures. Altogether, these results show that chronic activity blockade of the hippocampal CA3 circuit induces the down-regulation of axonal Kv1.1 channels leading to a homeostatic increase in both excitability and presynaptic release probability.     

Single-neuron firing cascades underlie global spontaneous brain events

The resting brain consumes enormous energy and shows highly organized spontaneous activity. To investigate how this activity is manifest among single neurons, we analyzed spiking discharges of ∼10,000 isolated cells recorded from multiple cortical and subcortical regions of the mouse brain during immobile rest. We found that firing of a significant proportion (∼70%) of neurons conformed to a ubiquitous, temporally sequenced cascade of spiking that was synchronized with global events and elapsed over timescales of 5 to 10 s. Across the brain, two intermixed populations of neurons supported orthogonal cascades. The relative phases of these cascades determined, at each moment, the response magnitude evoked by an external visual stimulus. Furthermore, the spiking of individual neurons embedded in these cascades was time locked to physiological indicators of arousal, including local field potential power, pupil diameter, and hippocampal ripples. These findings demonstrate that the large-scale coordination of low-frequency spontaneous activity, which is commonly observed in brain imaging and linked to arousal, sensory processing, and memory, is underpinned by sequential, large-scale temporal cascades of neuronal spiking across the brain.

The brain at rest exhibits slow (<0.1 Hz) but highly organized spontaneous activity as measured by functional MRI (fMRI) (1, 2). Much research in this area has utilized the temporal coordination of these signals to assess the functional organization a large number of brain networks. In recent years, however, new attention has been directed to a less-studied aspect of this signal, namely the conspicuous and discrete spontaneous events that occur simultaneously across the brain (3–5). These global resting-state fMRI events appear to reflect transient arousal modulations at a timescale of ∼10 s (4, 6) and also to be closely related to activity among clusters of cholinergic projection neurons in the basal forebrain (4, 5).The nature of global brain events is of great interest, as is their spatiotemporal dynamics. Some evidence suggests they take the form of traveling waves, propagating coherently according to the principles of the cortical hierarchy (7, 8), and shaping functional connectivity measures important for assessing the healthy and diseased brain (7, 9, 10). Other work has linked such global activity to phenomena as varied as modulation of the autonomic nervous system (11–14), cleansing circulation of cerebrospinal fluid in the glymphatic system (15–19), and memory consolidation mediated by hippocampal sharp-wave ripples (20, 21). In general, the global activity measured through brain imaging appears coordinated over timescales of seconds with a range of other neural and physiological events (11, 12, 14, 21–23). In a few cases, the relationship between local and global neural events has been studied using simultaneous measurements. For example, brain-wide fMRI fluctuations and local field potential (LFP) power changes are locked to the issuance of hippocampal ripples (21, 24). However, very little is understood about the extent to which single neurons participate in the expression and coordination of global spontaneous events of the seconds timescale. To approach this topic, recent technological advances have made it possible to track and compare the spiking activity of a large number of isolated neurons recorded simultaneously across multiple brain areas.A few recent studies utilizing high-density neuronal recording (25) have accumulated initial evidence suggesting a close relationship between the brain state and neuronal population dynamics (26–28). A large proportion of neurons, regardless of their location, showed strong modulations in their discharging rate that were coordinated in time with physiological arousal measures (28), across thirsty and sated states (26), and during exploratory and nonexploratory behaviors (27). Nevertheless, these studies leave open the question of how neuronal population dynamics are organized at a finer timescale of seconds surrounding spontaneous global events during immobile rest, and whether and how such dynamics are coincident with arousal modulations, hippocampal ripples, and sensory excitability. To investigate this topic, we examine the spiking activity recorded from large neuronal populations of neurons in immobilized mice, focusing on their seconds-scale coordination with global events and with one another. We further studied the impact of this spontaneous spiking on the magnitude of visually evoked responses and its time locking with other physiological signals related to arousal, such as LFP changes, hippocampal ripples, and changes in pupil diameter.     

Optogenetic stimulation of glutamatergic neurons in the cuneiform nucleus controls locomotion in a mouse model of Parkinson’s disease

In Parkinson’s disease (PD), the loss of midbrain dopaminergic cells results in severe locomotor deficits, such as gait freezing and akinesia. Growing evidence indicates that these deficits can be attributed to the decreased activity in the mesencephalic locomotor region (MLR), a brainstem region controlling locomotion. Clinicians are exploring the deep brain stimulation of the MLR as a treatment option to improve locomotor function. The results are variable, from modest to promising. However, within the MLR, clinicians have targeted the pedunculopontine nucleus exclusively, while leaving the cuneiform nucleus unexplored. To our knowledge, the effects of cuneiform nucleus stimulation have never been determined in parkinsonian conditions in any animal model. Here, we addressed this issue in a mouse model of PD, based on the bilateral striatal injection of 6-hydroxydopamine, which damaged the nigrostriatal pathway and decreased locomotor activity. We show that selective optogenetic stimulation of glutamatergic neurons in the cuneiform nucleus in mice expressing channelrhodopsin in a Cre-dependent manner in Vglut2-positive neurons (Vglut2-ChR2-EYFP mice) increased the number of locomotor initiations, increased the time spent in locomotion, and controlled locomotor speed. Using deep learning-based movement analysis, we found that the limb kinematics of optogenetic-evoked locomotion in pathological conditions were largely similar to those recorded in intact animals. Our work identifies the glutamatergic neurons of the cuneiform nucleus as a potentially clinically relevant target to improve locomotor activity in parkinsonian conditions. Our study should open avenues to develop the targeted stimulation of these neurons using deep brain stimulation, pharmacotherapy, or optogenetics.

In Parkinson’s disease (PD), midbrain dopaminergic (DA) cells are lost, resulting in motor dysfunction, including severe locomotor deficits (e.g., gait freezing, akinesia, and falls) (1). Growing evidence indicates that part of these deficits can be attributed to changes in the mesencephalic locomotor region (MLR) (refs. 2–7; for review, see ref. 8). This brainstem region plays a key role in locomotor control by sending projections to reticulospinal neurons that carry the locomotor drive to the spinal cord in vertebrates (lamprey: refs. 9 and 10; salamander: refs. 11 and 12; and mouse: refs. 13 to 19; for review, see ref. 20). The DA neurons of the substantia nigra pars compacta (SNc) indirectly control MLR activity through the basal ganglia (15, 21, 22). In parallel, the MLR receives direct DA projections from the SNc (23–26) and from the zona incerta (27). The DA innervation of the MLR degenerates in a monkey model of PD (28). Therefore, the loss of DA cells in PD has major effects on MLR activity. In PD, locomotor deficits are associated with MLR cell loss, abnormal neural activity, altered connectivity, and metabolic deficits, likely resulting in a loss of amplification of the locomotor commands (for review, see ref. 8). Accordingly, motor arrests and gait freezing are associated with a decrease in MLR activity in PD (ref. 29; for review, see ref. 8).One approach to improve locomotor function in PD would be to increase MLR activity. L-DOPA, the gold standard drug used to improve motor symptoms in PD, increases MLR activity and this likely contributes to the locomotor benefits (30). The MLR has been proposed to contribute to the locomotor benefits of deep brain stimulation (DBS) of the subthalamic nucleus (31–33), which has direct and indirect projections to the MLR (refs. 34, 35; for review, see ref. 20). However, the benefits of L-DOPA and subthalamic DBS on locomotor deficits may wane over time, highlighting the need to find new therapeutic approaches (36, 37). Since 2005, the MLR has been explored as a DBS target (38). The results vary, from modest to promising (39, 40). However, the best target in the MLR in PD conditions is not yet identified. The MLR is a heterogeneous structure, with the cuneiform nucleus (CnF) controlling the largest range of locomotor speeds, and the pedunculopontine nucleus (PPN) controlling slow speeds, posture and in some cases locomotor arrests (refs. 15, 17–19; for review, see ref. 20). Human DBS protocols targeted the PPN, but left the CnF unexplored, despite its major importance in locomotor control in animal research (19, 41, 42). In humans, a recent anatomical analysis of DBS electrode position relative to the pontomesencephalic junction suggests that some of the beneficial effects attributed to the PPN could be due to CnF activation (43).To add further complexity, three main cell types are present in the MLR: glutamatergic, GABAergic, and cholinergic cells. It is still unknown which cell type is the best target to improve locomotor function in PD conditions. Optogenetic studies uncovered that glutamatergic cells in the CnF play a key role in generating the locomotor drive for a wide range of speeds (15, 17–19). Glutamatergic cells of the PPN control slower speeds (17, 18) and, in some cases, evoke locomotor arrests (18, 44, 45). The GABAergic cells in the CnF and PPN stop locomotion likely by inhibiting glutamatergic cells (15, 17). The role of the PPN cholinergic cells is not resolved, as their activation can increase or decrease locomotion (refs. 15, 17, 18; for review, see ref. 20). Clinically, DBS likely stimulates all cells around the electrode, including the GABAergic cells that stop locomotion, and this could contribute to the variability of outcomes.Here, we aimed at identifying a relevant target in the MLR to improve the locomotor function in parkinsonian conditions. We hypothesized that the selective activation of CnF glutamatergic neurons should improve locomotor function in a mouse model of PD. We induced parkinsonian conditions in mice by bilaterally injecting into the striatum the neurotoxin 6-hydroxydopamine (6-OHDA), which is well known to damage the nigrostriatal DA pathway and to induce a dramatic decrease in locomotor activity (e.g., refs. 22 and 46). Using in vivo optogenetics in mice expressing channelrhodopsin in a Cre-dependent manner in Vglut2-positive neurons (Vglut2-ChR2-EYFP mice), we show that the photostimulation of glutamatergic neurons in the CnF robustly initiated locomotion, reduced immobility, increased the time spent in locomotion, and precisely controlled locomotor speed. Our results should help in defining therapeutic strategies aimed at specifically activating CnF glutamatergic neurons to improve locomotor function in PD using optimized DBS protocols, pharmacotherapy, or future optogenetic tools for human use.     

Visual stimuli recruit intrinsically generated cortical ensembles

The cortical microcircuit is built with recurrent excitatory connections, and it has long been suggested that the purpose of this design is to enable intrinsically driven reverberating activity. To understand the dynamics of neocortical intrinsic activity better, we performed two-photon calcium imaging of populations of neurons from the primary visual cortex of awake mice during visual stimulation and spontaneous activity. In both conditions, cortical activity is dominated by coactive groups of neurons, forming ensembles whose activation cannot be explained by the independent firing properties of their contributing neurons, considered in isolation. Moreover, individual neurons flexibly join multiple ensembles, vastly expanding the encoding potential of the circuit. Intriguingly, the same coactive ensembles can repeat spontaneously and in response to visual stimuli, indicating that stimulus-evoked responses arise from activating these intrinsic building blocks. Although the spatial properties of stimulus-driven and spontaneous ensembles are similar, spontaneous ensembles are active at random intervals, whereas visually evoked ensembles are time-locked to stimuli. We conclude that neuronal ensembles, built by the coactivation of flexible groups of neurons, are emergent functional units of cortical activity and propose that visual stimuli recruit intrinsically generated ensembles to represent visual attributes.There is a growing consensus in neuroscience that ensembles of neurons working in concert, as opposed to single neurons, are the underpinnings of cognition and behavior (1–3). At the microcircuit level, the cortex is dominated by recurrent excitatory connections (4, 5). Such densely interconnected excitatory networks are ideal for generating reverberating activity (1, 6, 7) that could link neurons into functional neuronal ensembles. Moreover, most cortical neurons are part of highly distributed synaptic circuits, receiving inputs from and projecting outputs to, thousands of other neurons (8, 9). In fact, the basic excitatory neurons of the cortex, pyramidal cells, appear to be biophysically designed to perform large-scale integration of inputs (10). All of these structural features indicate the possibility that rather than relying on the firing of individual neurons, cortical circuits may generate responses built out of the coordinated activity of groups of neurons. These postulated emergent circuit states could represent the building blocks of mental and behavioral processes (1–3, 11).In the visual cortex, there has been continuing progress in understanding functional properties and receptive fields of single neurons using single-unit electrophysiology and optical imaging (12–14). These single-neuron studies have provided a solid foundation for neuroscience. However, the focus on single neurons may provide an incomplete picture of this highly distributed neural circuit (3, 15). In fact, in recent years, the network activity patterns of the primary visual cortex (V1) in vitro and in vivo have been shown to be highly structured in spatiotemporal properties (13, 16–18). For example, using voltage-sensitive imaging, one can measure large-scale cortical dynamics with high temporal resolution, albeit without single-cell resolution (19–22). At this bird’s-eye view, wave-like spontaneous spatiotemporal patterns of activity appear similar to those patterns measured during visual stimulation (19, 21, 22). These findings imply that groups of neurons are active together in the absence of any visual input and that the same groups of neurons are also active together in response to visual stimulation. However, to test this hypothesis, one must measure the circuit activity with single-cell resolution.With calcium imaging, multineuronal activity can be visualized with single-cell resolution (16, 23), so it has become possible to discern exactly which neurons are activated under spontaneous and visually evoked conditions, cell by cell. Indeed, calcium imaging of brain slices from mouse visual cortex has revealed that groups of neurons become coactive spontaneously (24, 25) and that the same groups of neurons can be triggered by stimulation of thalamic afferents (26, 27). However, the patterns of activity found in slices may differ from the patterns of activity in vivo. Therefore, to determine the relation between spontaneous and evoked cortical activity patterns properly, it is necessary to measure them in vivo.Using two-photon calcium imaging in vivo, we have now mapped the spontaneous reverberating activity patterns in the V1 from awake mice with single-cell resolution and analyzed their relation to the activity patterns evoked by visual stimulation. We find patterns of coactive neurons that we term “ensembles,” defined as “a group of items viewed as a whole rather than individually” (28). Although the mere existence of these coactive neurons does not prove their functional importance, we provide converging lines of evidence that these ensembles are, in fact, functional units of cortical activity. This work provides a step in the progression of defining neuronal ensembles, rather than receptive fields of individual cells, as a building block of cortical microcircuits and suggests that these intrinsic neuronal ensembles are recruited when the cortex performs some of its most basic functions.     

Local dendritic balance enables learning of efficient representations in networks of spiking neurons

How can neural networks learn to efficiently represent complex and high-dimensional inputs via local plasticity mechanisms? Classical models of representation learning assume that feedforward weights are learned via pairwise Hebbian-like plasticity. Here, we show that pairwise Hebbian-like plasticity works only under unrealistic requirements on neural dynamics and input statistics. To overcome these limitations, we derive from first principles a learning scheme based on voltage-dependent synaptic plasticity rules. Here, recurrent connections learn to locally balance feedforward input in individual dendritic compartments and thereby can modulate synaptic plasticity to learn efficient representations. We demonstrate in simulations that this learning scheme works robustly even for complex high-dimensional inputs and with inhibitory transmission delays, where Hebbian-like plasticity fails. Our results draw a direct connection between dendritic excitatory–inhibitory balance and voltage-dependent synaptic plasticity as observed in vivo and suggest that both are crucial for representation learning.

Many neural systems have to encode high-dimensional and complex input signals in their activity. It has long been hypothesized that these encodings are highly efficient; that is, neural activity faithfully represents the input while also obeying energy and information constraints (1–3). For populations of spiking neurons, such an efficient code requires two central features: First, neural activity in the population has to be coordinated, such that no spike is fired superfluously (4); second, individual neural activity should represent reoccurring patterns in the input signal, which reflect the statistics of the sensory stimuli (2, 3). How can this coordination and these efficient representations emerge through local plasticity rules?To coordinate neural spiking, choosing the right recurrent connections between coding neurons is crucial. In particular, recurrent connections have to ensure that neurons do not spike redundantly to encode the same input. An early result was that in unstructured networks the redundancy of spiking is minimized when excitatory and inhibitory currents cancel on average in the network (5–7), which is also termed loose global excitatory–inhibitory (E-I) balance (8). To reach this state, recurrent connections can be chosen randomly with the correct average magnitude, leading to asynchronous and irregular neural activity (5) as observed in vivo (4, 9). More recently, it became clear that recurrent connections can ensure a much more efficient encoding when E-I currents cancel not only on average, but also on fast timescales and in individual neurons (4), which is also termed tight detailed E-I balance (8). Here, recurrent connections have to be finely tuned to ensure that the network response to inputs is precisely distributed over the population. To achieve this intricate recurrent connectivity, different local plasticity rules have been proposed, which robustly converge to a tight balance and thereby ensure that networks spike efficiently in response to input signals (10, 11).To efficiently encode high-dimensional input signals, it is additionally important that neural representations are adapted to the statistics of the input. Often, high-dimensional signals contain redundancies in the form of reoccurring spatiotemporal patterns, and coding neurons can reduce activity by representing these patterns in their activity. For example, in an efficient code of natural images, the activity of neurons should represent the presence of edges, which are ubiquitous in these images (3). Early studies of recurrent networks showed that such efficient representations can be found through Hebbian-like learning of feedforward weights (12, 13). With Hebbian learning the repeated occurrence of patterns in the input is associated with postsynaptic activity, causing neurons to become detectors of these patterns. Recurrent connections indirectly guide this learning process by forcing neurons to fire for distinct patterns in the input. Recent efforts rigorously formalized this idea for models of spiking neurons in balanced networks (11) and spiking neuron sampling from generative models (14–17). The great strength of these approaches is that the learning rules can be derived from first principles and turn out to be similar to spike-timing–dependent plasticity (STDP) curves that have been measured experimentally.However, to enable the learning of efficient representations, these models have strict requirements on network dynamics. Most crucially, recurrent inhibition has to ensure that neural responses are sufficiently decorrelated. In the neural sampling approaches, learning therefore relies on strong winner-take-all dynamics (14–17). In the framework of balanced networks, transmission of inhibition has to be nearly instantaneous to ensure strong decorrelation (18). These requirements are likely not met in realistic situations, where neural activity often shows positive correlations (19–22).We here derive a learning scheme that overcomes these limitations. First, we show that when neural activity is correlated, learning of feedforward connections has to incorporate nonlocal information about the activity of other neurons. Second, we show that recurrent connections can provide this nonlocal information by learning to locally balance specific feedforward inputs on the dendrites. In simulations of spiking neural networks we demonstrate the benefits of learning with dendritic balance over Hebbian-like learning for the efficient encoding of high-dimensional signals. Hence, we extend the idea that tightly balancing inhibition provides information about the population code locally and show that it can be used not only to distribute neural responses over a population, but also for an improved learning of feedforward weights.     

Harvesting entropy and quantifying the transition from noise to chaos in a photon-counting feedback loop

Many physical processes, including the intensity fluctuations of a chaotic laser, the detection of single photons, and the Brownian motion of a microscopic particle in a fluid are unpredictable, at least on long timescales. This unpredictability can be due to a variety of physical mechanisms, but it is quantified by an entropy rate. This rate, which describes how quickly a system produces new and random information, is fundamentally important in statistical mechanics and practically important for random number generation. We experimentally study entropy generation and the emergence of deterministic chaotic dynamics from discrete noise in a system that applies feedback to a weak optical signal at the single-photon level. We show that the dynamics transition from shot noise to chaos as the photon rate increases and that the entropy rate can reflect either the deterministic or noisy aspects of the system depending on the sampling rate and resolution.Continuous variables and dynamical equations are often used to model systems whose time evolution is composed of discrete events occurring at random times. Examples include the flow of ions across cell membranes (1), the dynamics of large populations of neurons (2), the birth and death of individuals in a population (3), traffic flow on roads (4), the trading of securities in financial markets (5, 6), infection and transmission of disease (7), and the emission and detection of photons (8). We can identify two sources of unpredictability in these systems: the noise associated with the underlying random occurrences that comprise these signals, which are often described by a Poisson process, and the macroscopic dynamics of the system, which may be chaotic. When both effects are present, the macroscopic dynamics can alter the statistics of the noise, and the small-scale noise can in turn feed the large-scale dynamics. This can lead to subtle and nontrivial effects including stochastic resonance and coherence resonance (9, 10). Dynamical unpredictability and complexity are quantified by Lyapunov exponents and dimensionality, whereas shot noise is characterized by statistical metrics like average rate, variance, and signal-to-noise ratio. Characterizing the unpredictability of a system with both large-scale dynamics and small-scale shot noise remains an important challenge in many disciplines including statistical mechanics and information security.Many cryptographic applications, including public key encryption (11), use random numbers. Because the unpredictability of these numbers is essential, physical processes are sometimes used as a source of random numbers (12–25). Physical random number generators are usually tested using the National Institute of Standards and Technology (26) and Diehard (27) test suites, which assess their ability to produce bits that are free of bias and correlation. These tests are an excellent assessment of the performance of a physical random number generator in practical situations but leave an important and fundamental problem unaddressed. Deterministic postprocessing procedures, such as hash functions (25), are often used to remove bias and correlation. Because these procedures are algorithmic and reproducible, they cannot in principle increase the entropy rate of a bit stream. Thus, the reliability of a physical random number generator depends on an accurate assessment of the entropy rate of physical process that generated the numbers (28). It remains difficult to assess the unpredictability of a system based on physical principles.Evaluation of entropy rates from an information-theoretic perspective is also centrally important in statistical mechanics (29–36). One might expect that the unpredictability of a system with both small-scale shot noise and large-scale chaotic dynamics would depend on the scale at which it is observed. In many systems, the dependence of the entropy rate on the resolution, ε, and the sampling interval, τ, can reflect the physical origin of unpredictability (37–40). This dependence has been studied experimentally in Brownian motion, RC circuits, and Rayleigh–Bénard convection (34, 35, 37, 41, 42).Here, we present an experimental exploration and numerical model of entropy production in a photon-counting optoelectronic feedback oscillator. Optoelectronic feedback loops that use analog detectors and macroscopic optical signals produce rich dynamics whose timescales and dimensionality are highly tunable (43–47). Our system applies optoelectronic feedback to a weak optical signal that is measured by a photon-counting detector. The dynamic range of this system (eight orders of magnitude in timescale and a factor of 256 in photon rate) allows us to directly observe the transition from shot noise-dominated behavior to a low-dimensional chaotic attractor with increasing optical power—a transition that, to our knowledge, has never been observed experimentally. We show that the entropy rate can reflect either the deterministic or stochastic aspects of the system, depending on the sampling rate and measurement resolution, and describe the importance of this observation for physical random number generation.     

A functional division of Drosophila sweet taste neurons that is value-based and task-specific

Sucrose is an attractive feeding substance and a positive reinforcer for Drosophila. But Drosophila females have been shown to robustly reject a sucrose-containing option for egg-laying when given a choice between a plain and a sucrose-containing option in specific contexts. How the sweet taste system of Drosophila promotes context-dependent devaluation of an egg-laying option that contains sucrose, an otherwise highly appetitive tastant, is unknown. Here, we report that devaluation of sweetness/sucrose for egg-laying is executed by a sensory pathway recruited specifically by the sweet neurons on the legs of Drosophila. First, silencing just the leg sweet neurons caused acceptance of the sucrose option in a sucrose versus plain decision, whereas expressing the channelrhodopsin CsChrimson in them caused rejection of a plain option that was “baited” with light over another that was not. Analogous bidirectional manipulations of other sweet neurons did not produce these effects. Second, circuit tracing revealed that the leg sweet neurons receive different presynaptic neuromodulations compared to some other sweet neurons and were the only ones with postsynaptic partners that projected prominently to the superior lateral protocerebrum (SLP) in the brain. Third, silencing one specific SLP-projecting postsynaptic partner of the leg sweet neurons reduced sucrose rejection, whereas expressing CsChrimson in it promoted rejection of a light-baited option during egg-laying. These results uncover that the Drosophila sweet taste system exhibits a functional division that is value-based and task-specific, challenging the conventional view that the system adheres to a simple labeled-line coding scheme.

The taste systems of many animal species are known to possess a dedicated “channel” for detecting sugars, a class of chemicals that is highly nutritious. For example, mice have been shown to encode gustatory receptors that specifically sense sugars, and the taste neurons that express these sugar receptors on their tongues generally do not express receptors that sense chemicals of another taste modality (e.g., bitterness) (1–3). Furthermore, activation of these sugar-sensing taste neurons by artificial means has been shown to be able to drive appetitive sugar-induced innate responses (e.g., licking) and act as a positive reinforcer for learning (3–5). In some recent studies, these properties of the sweet taste neurons have been found to be present in some of their central nervous system (CNS) targets (e.g., taste-sensitive neurons in the insular cortex), too (6, 7). Thus, one school of thought is that taste coding for sweetness in mice may follow the simple “labeled-line” rule: sweet taste neurons, and potentially some of their central targets, are hardwired to detect sugars specifically and drive sugar-induced reinforcing neural signals and appetitive behaviors (1–7).Drosophila melanogaster also possess sugar-detecting taste neurons. Pioneering early studies have shown that sugar-sensing taste neurons in flies are molecularly, anatomically, and functionally distinct from taste neurons that sense bitterness; sweet-sensing and bitter-sensing taste neurons express different gustatory receptors, project axons to different areas in the brain, and are required to promote different (appetitive versus aversive) behaviors (8–12). Moreover, the activation of sweet neurons by artificial means can drive appetitive behaviors and act as a positive reinforcer for learning (10, 13, 14), while artificial activation of bitter-sensing neurons can induce rejection behaviors and be used as a punishment for learning (10, 13, 15). Interestingly, while these results suggest that Drosophila sweet neurons and their mammalian counterparts have some shared properties, subsequent studies suggest that significant differences exist between them, too. First, the Drosophila genome appears to encode many more sweet receptors than mouse genome does (12, 16–23). Second, Drosophila sweet neurons appear to be able to detect some chemicals that belong to another taste modality [e.g., acetic acid (AA)] (24–27). Third, Drosophila sweet neurons can be found on several body parts (e.g., proboscis and legs) (8, 12, 18, 20, 23, 28–30). Interestingly, sweet neurons on different body parts of Drosophila do not promote identical behavioral outputs (8, 20, 23, 24, 28, 29). For example, labellar sweet neurons and esophageal sweet neurons on the proboscis have been shown to promote proboscis extension reflex (PER) and ingestion, respectively, whereas leg sweet neurons have been shown to promote PER and slowing down of locomotion (8, 12, 28, 29). Collectively, these results suggest that in contrast to the apparent homogeneity of sweet neurons in some mammals, a functional division exists among Drosophila sweet neurons, although the different behavioral responses promoted by different Drosophila sweet neurons generally appear appetitive in nature.In this work, we report yet another striking feature of Drosophila sweet neurons that sets them apart from their mammalian counterparts, namely a functional division that is value-based and task-specific. We discovered this by taking advantage of a context-dependent but highly robust sugar rejection behavior exhibited by egg-laying females (31–34). Previous studies have shown that when selecting for egg-laying site in a small enclosure (dimension ∼16 × 10 × 18 mm), Drosophila readily accept a sucrose-containing agarose for egg-laying when it is the sole option but strongly reject it when a plain option is also available (31, 32). Importantly, silencing their sweet neurons causes the females to no longer reject the sucrose option when choosing between the sucrose versus plain options (31, 32). Thus, in addition to promoting appetitive behaviors and acting as a positive reinforcer, activation of sweet neurons on an egg-laying option can also decrease the value of such an option (thereby causing its rejection over an option that does not activate sweet neurons). These observations not only suggest the existence of an apparent “antiappetitive” role of Drosophila sweet neurons when the task of animals is to select for egg-laying sites but also raise a key question as to whether such counterintuitive, value-decreasing property of sweetness detection during egg-laying may be 1) solely an emergent property of specific neurons in the brain that respond similarly to all peripheral sweet neurons but are sensitive to animals’ behavioral goal and context or 2) carried out by specific sweet neurons at the periphery and then transmitted into the brain via a unique neural pathway activated by these neurons. To disambiguate between these possibilities, we genetically targeted different subsets of sweet neurons to assess their circuit properties as well as their behavioral roles as the animals decided in either a regular or a virtual sweet versus plain decision during egg-laying, taking advantage of a high-throughput closed-loop optogenetic stimulation platform we developed recently. Our collective results support the second scenario and suggest that the value-decreasing property of sweetness/sucrose is conveyed specifically by the sweet neurons on the legs of Drosophila—and not by other sweet neurons—and the unique postsynaptic target(s) of the leg sweet neurons that send long-range projections to the superior lateral protocerebrum (SLP) in the brain. These results reveal a previously unappreciated functional and anatomical division of the Drosophila sweet taste neurons that is both task-specific and value-based, pointing to a level of complexity and sophistication that seems unmatched by their mammalian counterparts so far.     

Emergence of functional subnetworks in layer 2/3 cortex induced by sequential spikes in vivo

During cortical circuit development in the mammalian brain, groups of excitatory neurons that receive similar sensory information form microcircuits. However, cellular mechanisms underlying cortical microcircuit development remain poorly understood. Here we implemented combined two-photon imaging and photolysis in vivo to monitor and manipulate neuronal activities to study the processes underlying activity-dependent circuit changes. We found that repeated triggering of spike trains in a randomly chosen group of layer 2/3 pyramidal neurons in the somatosensory cortex triggered long-term plasticity of circuits (LTPc), resulting in the increased probability that the selected neurons would fire when action potentials of individual neurons in the group were evoked. Significant firing pattern changes were observed more frequently in the selected group of neurons than in neighboring control neurons, and the induction was dependent on the time interval between spikes, N-methyl-D-aspartate (NMDA) receptor activation, and Calcium/calmodulin-dependent protein kinase II (CaMKII) activation. In addition, LTPc was associated with an increase of activity from a portion of neighboring neurons with different probabilities. Thus, our results demonstrate that the formation of functional microcircuits requires broad network changes and that its directionality is nonrandom, which may be a general feature of cortical circuit assembly in the mammalian cortex.Layer 2/3 neurons in the barrel cortex play a central role in integrative cortical processing (1–4). Neurons in layer 2/3 are interconnected with each other, and their axons and dendrites traverse adjacent barrel areas (5, 6). Recent calcium (Ca²⁺) imaging studies in awake animals showed that two very closely localized layer 2/3 pyramidal neurons are independently activated by different whiskers (7). In addition, adjacent layer 2/3 neurons have different receptive field properties; signals from different whiskers may emerge on different spines in the same neurons (8, 9). These findings suggest that the organization of functional subnetworks in somatosensory layer 2/3 is heterogeneous at the single-cell level and that microcircuits are assembled at a very fine scale (10). In vivo whole-cell recording experiments have also shown that most, but not all, layer 2/3 pyramidal neurons receive subthreshold depolarization by single-whisker stimulation with much broader receptive fields than neurons in layer 4 (11, 12). These anatomical and functional data suggest that electric signals relayed to the cortex by whisker activation are greatly intermingled within layer 2/3 neurons, and that studying the mechanisms by which these layer 2/3 neurons make connections may be critical for understanding the cortical network organizing principles underlying somatosensation.A previous modeling study suggested that spike timing-dependent plasticity (STDP) can lead to the formation of functional cortical columns and activity-dependent reorganization of neural circuits (13–16). However, how spikes arising in multiple neurons in vivo influence their connectivity is poorly understood. In this study using two-photon glutamate photolysis, which allowed us to control neuronal activity in a spatially and temporally precise manner, we examined activity-dependent cellular mechanisms during network rearrangement generated by repetitive spike trains in a group of neurons. We found that repetitive spikes on a group of neurons induced the probability of the neurons firing together. This circuit plasticity required spiking at short intervals among neurons and is expressed by N-methyl-D-aspartate (NMDA) receptor- and Calcium/calmodulin-dependent protein kinase II (CaMKII)-dependent long-lasting connectivity changes. The probability of firing was differentially affected by the order of the spike sequence but was not dependent on the physical distance between neurons. Thus, our data show that neuronal connectivity within a functional subnetwork is established in not only a preferred but also a directional manner.     

In vivo imaging of immediate early gene expression reveals layer-specific memory traces in the mammalian brain

The dynamic processes of formatting long-term memory traces in the cortex are poorly understood. The investigation of these processes requires measurements of task-evoked neuronal activities from large numbers of neurons over many days. Here, we present a two-photon imaging-based system to track event–related neuronal activity in thousands of neurons through the quantitative measurement of EGFP proteins expressed under the control of the EGR1 gene promoter. A recognition algorithm was developed to detect GFP-positive neurons in multiple cortical volumes and thereby to allow the reproducible tracking of 4,000 neurons in each volume for 2 mo. The analysis revealed a context-specific response in sparse layer II neurons. The context-evoked response gradually increased during several days of training and was maintained 1 mo later. The formed traces were specifically activated by the training context and were linearly correlated with the behavioral response. Neuronal assemblies that responded to specific contexts were largely separated, indicating the sparse coding of memory-related traces in the layer II cortical circuit.In the mammalian brain, memory traces in cortical areas are poorly understood. In contrast to the medial temporal lobe, particularly the hippocampus, which is involved in the temporary storage of declarative memories (1, 2), the neocortex is believed to store remote memories (3–6). However, remarkably little knowledge regarding the sites and dynamics of remote memory storage has been revealed at the cellular level owing to the complexity of the connections and the large number of neurons within the cortical circuit.In vivo electrophysiological recording of neuronal firing revolutionized neurobiology by linking neuronal activity with animal behavior. The small number of neurons recorded by the electrodes, however, was a limitation, as information coding and decoding may use an army of neurons forming neuronal assemblies (7, 8). Efforts to record the activity of larger populations of neurons in cortical volumes have been actively pursued by either increasing the number of electrode probes (7, 9–11) or using calcium indicator–based imaging (12–15) and immediate early gene (IEG)-based reporters (16–18). The expression of IEGs is correlated with the averaged neuronal activation on external stimuli (19, 20), implying that the marked neurons are involved in behavior (1, 21–25). Studies using in vivo imaging of IEGs have revealed cortical coding in the visual cortex and in other cortical areas, reflecting electrical activation in individual neurons (16, 17). Among IEGs, the expression of early growth response protein 1 (EGR1, also known as zif268) is associated with high-frequency stimulation and the induction of long-term plasticity during learning (26, 27). To measure neuronal activation in cortical circuits during a behavioral task, we used an EGR1 expression reporter mouse line in which the expression of the EGFP protein is under the control of the Egr1 gene promoter. We designed offline recording strategies to monitor task-associated neuronal activity by quantifying changes in cellular EGFP signals in the mouse cortex. Patterns of activated neuronal assemblies during different tasks were visualized in the entire cortical volume. Furthermore, through computer recognition-based reconstruction, we were able to track the activity-related cellular EGFP signals from multiple cortical areas for 2 mo to reveal memory-related changes in the cortical circuit.     

From the Cover: PNAS Plus: Whole-brain calcium imaging with cellular resolution in freely behaving Caenorhabditis elegans

The ability to acquire large-scale recordings of neuronal activity in awake and unrestrained animals is needed to provide new insights into how populations of neurons generate animal behavior. We present an instrument capable of recording intracellular calcium transients from the majority of neurons in the head of a freely behaving Caenorhabditis elegans with cellular resolution while simultaneously recording the animal’s position, posture, and locomotion. This instrument provides whole-brain imaging with cellular resolution in an unrestrained and behaving animal. We use spinning-disk confocal microscopy to capture 3D volumetric fluorescent images of neurons expressing the calcium indicator GCaMP6s at 6 head-volumes/s. A suite of three cameras monitor neuronal fluorescence and the animal’s position and orientation. Custom software tracks the 3D position of the animal’s head in real time and two feedback loops adjust a motorized stage and objective to keep the animal’s head within the field of view as the animal roams freely. We observe calcium transients from up to 77 neurons for over 4 min and correlate this activity with the animal’s behavior. We characterize noise in the system due to animal motion and show that, across worms, multiple neurons show significant correlations with modes of behavior corresponding to forward, backward, and turning locomotion.How do patterns of neural activity generate an animal’s behavior? To answer this question, it is important to develop new methods for recording from large populations of neurons in animals as they move and behave freely. The collective activity of many individual neurons appears to be critical for generating behaviors including arm reach in primates (1), song production in zebrafinch (2), the choice between swimming or crawling in leech (3), and decision-making in mice during navigation (4). New methods for recording from larger populations of neurons in unrestrained animals are needed to better understand neural coding of these behaviors and neural control of behavior more generally.Calcium imaging has emerged as a promising technique for recording dynamics from populations of neurons. Calcium-sensitive proteins are used to visualize changes in intracellular calcium levels in neurons in vivo which serve as a proxy for neural activity (5). To resolve the often weak fluorescent signal of an individual neuron in a dense forest of other labeled cells requires a high magnification objective with a large numerical aperture that, consequently, can image only a small field of view. Calcium imaging has traditionally been performed on animals that are stationary from anesthetization or immobilization to avoid imaging artifacts induced by animal motion. As a result, calcium imaging studies have historically focused on small brain regions in immobile animals that exhibit little or no behavior (6).No previous neurophysiological study has attained whole-brain imaging with cellular resolution in a freely behaving unrestrained animal. Previous whole-brain cellular resolution calcium imaging studies of populations of neurons that involve behavior investigate either fictive locomotion (3, 7), or behaviors that can be performed in restrained animals, such as eye movements (8) or navigation of a virtual environment (9). One exception has been the development of fluorescence endoscopy, which allows recording from rodents during unrestrained behavior, although imaging is restricted to even smaller subbrain regions (10).Investigating neural activity in small transparent organisms allows one to move beyond studying subbrain regions to record dynamics from entire brains with cellular resolution. Whole-brain imaging was performed first in larval zebrafish using two-photon microscopy (7). More recently, whole-brain imaging was performed in Caenorhabditis elegans using two-photon (11) and light-field microscopy (12). Animals in these studies were immobilized, anesthetized, or both and thus exhibited no gross behavior.C. elegans’ compact nervous system of only 302 neurons and small size of only 1 mm make it ideally suited for the development of new whole-brain imaging techniques for studying behavior. There is a long and rich history of studying and quantifying the behavior of freely moving C. elegans dating back to the mid-1970s (13, 14). Many of these works involved quantifying animal body posture as the worm moved, for example as in ref. 15. To facilitate higher-throughput recordings of behavior, a number of tracking microscopes (16–18) or multiworm imagers were developed (19, 20) along with sophisticated behavioral analysis software and analytical tools (21, 22). Motivated in part to understand these behaviors, calcium imaging systems were also developed that could probe neural activity in at first partially immobilized (23) and then freely moving animals, beginning with ref. 24 and and then developing rapidly (17, 18, 25–29). One limitation of these freely moving calcium imaging systems is that they are limited to imaging only a very small subset of neurons and lack the ability to distinguish neurons that lie atop one another in the axial direction of the microscope. Despite this limitation, these studies, combined with laser-ablation experiments, have identified a number of neurons that correlate or affect changes in particular behaviors including the AVB neuron pair and VB-type motor neurons for forward locomotion; the AVA, AIB, and AVE neuron pairs and VA-type motor neurons for backward locomotion; and the RIV, RIB, and SMD neurons and the DD-type motor neurons for turning behaviors (17, 18, 25, 26, 28, 30, 31). To move beyond these largely single-cell studies, we sought to record simultaneously from the entire brain of C. elegans with cellular resolution and record its behavior as it moved about unrestrained.