Noise-induced properties of active dendrites

Dendrites play an essential role in the integration of highly fluctuating input in vivo into neurons across all nervous systems. Yet, they are often studied under conditions where inputs to dendrites are sparse. The dynamic properties of active dendrites facing in vivo–like fluctuating input thus remain elusive. In this paper, we uncover dynamics in a canonical model of a dendritic compartment with active calcium channels, receiving in vivo–like fluctuating input. In a single-compartment model of the active dendrite with fast calcium activation, we show noise-induced nonmonotonic behavior in the relationship of the membrane potential output, and mean input emerges. In contrast, noise can induce bistability in the input–output relation in the system with slowly activating calcium channels. Both phenomena are absent in a noiseless condition. Furthermore, we show that timescales of the emerging stochastic bistable dynamics extend far beyond a deterministic system due to stochastic switching between the solutions. A numerical simulation of a multicompartment model neuron shows that in the presence of in vivo–like synaptic input, the bistability uncovered in our analysis persists. Our results reveal that realistic synaptic input contributes to sustained dendritic nonlinearities, and synaptic noise is a significant component of dendritic input integration.

Neuronal interactions are mediated via synapses primarily located on large, tree-like dendritic structures. These dendritic trees integrate synaptic inputs and determine the extent of the neuron’s spiking output (1). Dendrites contribute substantially to neuronal plasticity and function (2–4). Their active calcium dynamics can induce nonlinear regenerative events, such as dendritic spikes, both in vitro (5) and in vivo (6). Dendritic spikes also serve a functional role in in vivo cortical visual processing, where a typical pyramidal neuron receives massive amounts of intensely fluctuating synaptic input from excitatory and inhibitory presynaptic neurons in the circuit (7). Although dendrites’ importance in integrating relevant signals in vivo is evident (3), how a dendritic tree transfers its input into output in noisy conditions is essentially unknown. Synaptic noise is commonly assumed to be a nuisance, and nervous systems develop strategies to filter it. However, noise can also induce new organized behaviors in systems that lack in deterministic conditions (8). Examples of noise leading to spontaneous order in biological and physical systems include stochastic resonance (9), noise-induced phase transitions (10), and noise-induced bistability (11). To investigate whether in vivo–like fluctuating input induces novel dynamic states in active dendrites, we analytically study a canonical model of a dendritic compartment with voltage-gated calcium channels.Various types of voltage-gated calcium channels in the dendrite have been studied (12, 13). In spinal motoneurons, where input is sparse, dendritic calcium channels can mediate a persistent depolarizing drive underlying their bistable dendritic properties (14). Despite the diversity of dendritic calcium channels in central nervous system neurons, no in vitro experiments have reported calcium channels that directly induce bistability. However, the experimental evidence suggests long-lasting dendritic elevated responses to a brief input in vivo, known as dendritic plateau potentials (15–17). Dendritic plateau potentials play an essential role in enhancing localized learning and information storage capacity at a specific dendritic branch (15). In this paper, we study whether the interaction of nonlinear calcium dynamics and in vivo–like fluctuation contributes to dendritic plateau potentials in the central nervous system.We study a canonical single-compartment model of an active dendrite to understand how the realistic in vivo input interacts with dendritic calcium channels. Our results indicate that stochastic phenomena emerge in dendritic dynamics with in vivo–like fluctuating input that is entirely absent in the deterministic condition. We observe that in vivo–like noise induces nonmonotonic or bistable dynamics in the input–output relation of a single-compartment dendritic model, depending on the timescales of calcium-gating variables. To investigate the relevance of these results for spatially extended neurons, we further numerically study a multicompartmental model neuron receiving realistic synaptic inputs across the dendritic branches. Our paper reveals that the interaction between in vivo–like input fluctuations and dendritic voltage-gated calcium channels could lead to dendritic plateau potentials.     

Dendrites are dispensable for basic motoneuron function but essential for fine tuning of behavior

Dendrites are highly complex 3D structures that define neuronal morphology and connectivity and are the predominant sites for synaptic input. Defects in dendritic structure are highly consistent correlates of brain diseases. However, the precise consequences of dendritic structure defects for neuronal function and behavioral performance remain unknown. Here we probe dendritic function by using genetic tools to selectively abolish dendrites in identified Drosophila wing motoneurons without affecting other neuronal properties. We find that these motoneuron dendrites are unexpectedly dispensable for synaptic targeting, qualitatively normal neuronal activity patterns during behavior, and basic behavioral performance. However, significant performance deficits in sophisticated motor behaviors, such as flight altitude control and switching between discrete courtship song elements, scale with the degree of dendritic defect. To our knowledge, our observations provide the first direct evidence that complex dendrite architecture is critically required for fine-tuning and adaptability within robust, evolutionarily constrained behavioral programs that are vital for mating success and survival. We speculate that the observed scaling of performance deficits with the degree of structural defect is consistent with gradual increases in intellectual disability during continuously advancing structural deficiencies in progressive neurological disorders.Dendrites are structural ramifications of a neuron specialized for receiving and processing synaptic input (1). The estimated 100 billion neurons in the human brain (2) form approximately 100 trillion synapses onto a total of approximately 100,000 miles of dendritic cable. The functions of dendrites are proposed to range from simply providing enough surface for synaptic input (3) to highly compartmentalized units of molecular signaling and information processing (4–6). In addition to functioning as passive receivers, dendrites may be equipped with output synapses (7) and active membrane currents (8), which add tremendous computational power to a single neuron (6, 9, 10).Accordingly, dendritic abnormalities are highly consistent anatomical correlates of numerous brain disorders (11, 12), including autism spectrum disorders, Alzheimer’s disease, schizophrenia, Down syndrome, Fragile X syndrome, Rett syndrome, anxiety, and depression. However, in many cases it remains unclear whether dendritic defects are the cause or the consequence of impaired brain function. Trying to understand dendrite function poses major technical challenges because it requires selective manipulation of dendritic structure without disturbing other properties of the affected neuron, followed by quantitative analysis of neuronal function and the resulting behavioral consequences.This study uses the Drosophila genetic model system to selectively abolish dendrites from a subset of identified wing muscle motoneurons that have well-described and stereotyped dendritic morphologies (13) and firing patterns during flight (14) and courtship song (15, 16). Surprisingly, we find that motoneurons that lack 90% of their dendrites are still contacted by appropriate synaptic partners and produce qualitatively normal firing patterns and wing movements during flight and courtship song. However, normal dendritic architecture is essential for particularly challenging tasks, such as the integration of optomotor input for adequate control of flight power output, or the temporal accuracy of switching between different song elements during courtship to ensure mating success. Our data demonstrate that the vast majority of basic motor functions can be satisfactorily accomplished with motoneurons that have significant dendritic defects but normal axonal structure and membrane currents. However, a complex 3D dendritic architecture is mandatory for intricate regulation of behavioral output, which in turn imposes a positive selection pressure on the maintenance of such complex dendritic trees through evolution.     

Territories of heterologous inputs onto Purkinje cell dendrites are segregated by mGluR1-dependent parallel fiber synapse elimination

In Purkinje cells (PCs) of the cerebellum, a single “winner” climbing fiber (CF) monopolizes proximal dendrites, whereas hundreds of thousands of parallel fibers (PFs) innervate distal dendrites, and both CF and PF inputs innervate a narrow intermediate domain. It is unclear how this segregated CF and PF innervation is established on PC dendrites. Through reconstruction of dendritic innervation by serial electron microscopy, we show that from postnatal day 9–15 in mice, both CF and PF innervation territories vigorously expand because of an enlargement of the region of overlapping innervation. From postnatal day 15 onwards, segregation of these territories occurs with robust shortening of the overlapping proximal region. Thus, innervation territories by the heterologous inputs are refined during the early postnatal period. Intriguingly, this transition is arrested in mutant mice lacking the type 1 metabotropic glutamate receptor (mGluR1) or protein kinase Cγ (PKCγ), resulting in the persistence of an abnormally expanded overlapping region. This arrested territory refinement is rescued by lentivirus-mediated expression of mGluR1α into mGluR1-deficient PCs. At the proximal dendrite of rescued PCs, PF synapses are eliminated and free spines emerge instead, whereas the number and density of CF synapses are unchanged. Because the mGluR1-PKCγ signaling pathway is also essential for the late-phase of CF synapse elimination, this signaling pathway promotes the two key features of excitatory synaptic wiring in PCs, namely CF monoinnervation by eliminating redundant CF synapses from the soma, and segregated territories of CF and PF innervation by eliminating competing PF synapses from proximal dendrites.Monoinnervation of cerebellar Purkinje cells (PCs) by single climbing fibers (CFs) is established in the early postnatal period (1–3). The soma of a neonatal PC is innervated by more than five CFs with similar synaptic strengths, from which a single CF is functionally strengthened (4, 5). The strengthened (“winner”) CF starts dendritic translocation, whereas the other weaker (“loser”) CFs remaining on the soma are eliminated (6–8). In this process, P/Q-type voltage-dependent Ca²⁺ channels (VDCCs) promote functional differentiation and dendritic translocation of winner CFs, and the early phase of CF synapse elimination (9–11), whereas the late phase of CF synapse elimination is critically dependent on the formation of parallel fiber (PF) synapses and activation of the type 1 metabotropic glutamate receptor (mGluR1)-protein kinase Cγ (PKCγ) pathway (12–17).Segregated dendritic innervation by CFs and PFs is another distinguished feature of the PC synaptic wiring. Although hundreds of thousands of PFs innervate the distal dendritic domain, a single CF monopolizes the proximal dendritic domain, and both innervate a narrow intermediate domain (18). Given that both dendritic translocation of winner CFs and formation of PF synapses proceed upwards from the base of the dendritic tree (6, 19), CFs and PFs must compete with each other to establish their segregated territories. However, the developmental route and the underlying mechanisms of this process are unknown.Our findings indicate that CF and PF territories on PC dendrites are dynamically refined during the early postnatal period, and that the mGluR1-PKCγ signaling pathway regulates segregation by promoting PF synapse elimination. Thus, this signaling cascade plays key roles in sculpting the excitatory synaptic wiring in PCs by eliminating both redundant CF synapses from the soma (3, 20) and competing PF synapses from the proximal dendrites.     

NMDA receptor–BK channel coupling regulates synaptic plasticity in the barrel cortex

Postsynaptic N-methyl-D-aspartate receptors (NMDARs) are crucial mediators of synaptic plasticity due to their ability to act as coincidence detectors of presynaptic and postsynaptic neuronal activity. However, NMDARs exist within the molecular context of a variety of postsynaptic signaling proteins, which can fine-tune their function. Here, we describe a form of NMDAR suppression by large-conductance Ca²⁺- and voltage-gated K⁺ (BK) channels in the basal dendrites of a subset of barrel cortex layer 5 pyramidal neurons. We show that NMDAR activation increases intracellular Ca²⁺ in the vicinity of BK channels, thus activating K⁺ efflux and strong negative feedback inhibition. We further show that neurons exhibiting such NMDAR–BK coupling serve as high-pass filters for incoming synaptic inputs, precluding the induction of spike timing–dependent plasticity. Together, these data suggest that NMDAR-localized BK channels regulate synaptic integration and provide input-specific synaptic diversity to a thalamocortical circuit.

Glutamate is the primary excitatory chemical transmitter in the mammalian central nervous system (CNS), where it is essential for neuronal viability, network function, and behavioral responses (1). Glutamate activates a variety of pre- and postsynaptic receptors, including ionotropic receptors (iGluRs) that form ligand-gated cation-permeable ion channels. The iGluR superfamily includes α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs), kainate receptors, and N-methyl-D-aspartate receptors (NMDARs), all of which form tetrameric assemblies that are expressed throughout the CNS (2).NMDARs exhibit high sensitivity to glutamate (apparent half maximal effective concentration in the micromolar range) and a voltage-dependent block by Mg²⁺ (3, 4), slow gating kinetics (5), and high permeability to Ca²⁺ (6, 7) (for a review, see ref. 8). Together, these characteristics confer postsynaptic NMDARs with the ability to detect and decode coincidental activity of pre- and postsynaptic neurons: presynaptic glutamate release brings about the occupation of the agonist-binding site and AMPAR-driven postsynaptic depolarization, removing the voltage-dependent Mg²⁺ block. The coincidence of these two events leads to NMDAR activation and a Ca²⁺ influx through the channel (8, 9), which initiates several forms of synaptic plasticity (10, 11).Large-conductance Ca²⁺- and voltage-gated K⁺ (BK) channels are opened by a combination of membrane depolarization and relatively high levels of intracellular Ca²⁺ (12, 13). In CNS neurons, such micromolar Ca²⁺ increases are usually restricted to the immediate vicinity of Ca²⁺ sources, including voltage-gated Ca²⁺ channels (VGCCs) (14–16) and ryanodine receptors (RyRs) (17, 18). In addition, Ca²⁺ influx through nonselective cation-permeable channels, including NMDARs, has also been shown to activate BK channels in granule cells from the olfactory bulb and dentate gyrus (19–21). In these neurons, Ca²⁺ entry through NMDARs opens BK channels in somatic and perisomatic regions, causing the repolarization of the surrounding plasma membrane and subsequent closure of NMDARs. Because BK channel activation blunts NMDAR-mediated excitatory responses, it provides a negative feedback mechanism that modulates the excitability of these neurons (19, 20). Thus, the same characteristics that make NMDARs key components in excitatory synaptic transmission and plasticity can paradoxically give rise to an inhibitory response when NMDARs are located in the proximity of BK channels. However, it is unclear whether functional NMDAR–BK coupling is relevant at dendrites and dendritic spines.The barrel field area in the primary somatosensory cortex, also known as the barrel cortex (BC), processes information from peripheral sensory receptors for onward transmission to cortical and subcortical brain regions (22, 23). Sensory information is received in the BC from different nuclei of the thalamus. Among these nuclei, the ventral posterior medial nucleus, ventrobasal nucleus, and posterior medial nucleus are known to directly innervate layer 5 pyramidal neurons (BC-L5PNs) (24–27). In basal dendrites of BC-L5PN, the coactivation of neighboring dendritic inputs can initiate NMDAR-mediated dendritically restricted spikes characterized by large Ca²⁺ transients and long-lasting depolarizations (28–30), providing the appropriate environment for BK activation.To determine whether functional NMDAR–BK coupling plays a role in synaptic transmission, and potentially synaptic plasticity, we investigated the thalamocortical synapses at basal dendrites of BC-L5PNs. We found that the suppression of NMDAR activity by BK channels occurs in the basal dendrites of about 40% of BC-L5PNs, where NMDAR activation triggers strong negative feedback inhibition by delivering Ca²⁺ to nearby BK channels. This inhibition regulates the amplitude of postsynaptic responses and increases the threshold for the induction of synaptic plasticity. Our findings thus unveil a calibration mechanism that can decode the amount and frequency of afferent synaptic inputs by selectively attenuating synaptic plasticity and providing input-specific synaptic diversity to a thalamocortical circuit.     

Activity-dependent dendritic spine neck changes are correlated with synaptic strength

Most excitatory inputs in the mammalian brain are made on dendritic spines, rather than on dendritic shafts. Spines compartmentalize calcium, and this biochemical isolation can underlie input-specific synaptic plasticity, providing a raison d’etre for spines. However, recent results indicate that the spine can experience a membrane potential different from that in the parent dendrite, as though the spine neck electrically isolated the spine. Here we use two-photon calcium imaging of mouse neocortical pyramidal neurons to analyze the correlation between the morphologies of spines activated under minimal synaptic stimulation and the excitatory postsynaptic potentials they generate. We find that excitatory postsynaptic potential amplitudes are inversely correlated with spine neck lengths. Furthermore, a spike timing-dependent plasticity protocol, in which two-photon glutamate uncaging over a spine is paired with postsynaptic spikes, produces rapid shrinkage of the spine neck and concomitant increases in the amplitude of the evoked spine potentials. Using numerical simulations, we explore the parameter regimes for the spine neck resistance and synaptic conductance changes necessary to explain our observations. Our data, directly correlating synaptic and morphological plasticity, imply that long-necked spines have small or negligible somatic voltage contributions, but that, upon synaptic stimulation paired with postsynaptic activity, they can shorten their necks and increase synaptic efficacy, thus changing the input/output gain of pyramidal neurons.Dendritic spines are found in neurons throughout the central nervous system (1), and in pyramidal neurons receive the majority of excitatory inputs, whereas dendritic shafts are normally devoid of glutamatergic synapses (2–7). These facts suggest that spines are likely to play an essential role in neural circuits (1), although it is still unclear exactly what this role is (8, 9). Because of their peculiar morphology, hypotheses regarding the specific function of spines have focused on their role in biochemical compartmentalization, whereby a small spine head, where the excitatory synapse is located, is separated from the parent dendrite by a thin neck, isolating the spine cytoplasm from the dendrite (10). Indeed, spines are diffusionally restricted from dendrites (11–13) and compartmentalize calcium after synaptic stimulation (14–16). This local biochemistry and the high calcium accumulations observed following temporal pairing of neuronal input and output (14, 17, 18) are thought to be responsible for input-specific synaptic plasticity (19–21). However, besides this biochemical role, spines have also been hypothesized to play an electrical role, altering excitatory postsynaptic potentials (EPSPs) (22–30). Consistent with this idea, activating spines with two-photon uncaging of glutamate generates potentials whose amplitudes are inversely proportional to the length of the spine neck (31), and these responses are much larger in spines than in adjacent dendritic shafts (32). Also, spine conductances can be activated independently of dendritic ones (33–36). These data suggest that spines could serve as electrical compartments but, at the same time, raise the issue of the functional significance of the thousands of long-necked spines that cover the dendrites of pyramidal neurons, which would therefore have negligible somatic voltage contributions.In this study we first undertook a series of experiments to discern the potential effect that the spine neck length has on the synaptic potentials generated by minimal synaptic stimulation at identified spines. We find that EPSP amplitudes are inversely correlated with spine neck lengths and that, as also seen in glutamate uncaging experiments (31), long-necked spines do not appear to generate any significant somatic depolarizations. In a separate set of experiments, we used a spike timing-dependent long-term potentiation (STD-LTP) induction protocol to trigger rapid shortening of the stimulated spine neck, which was accompanied by increases in the amplitude of the evoked potentials. In essence, we thus found a way to rapidly increase the voltage contribution of long-necked spines. To dissect the plausible mechanisms of the effect, we conducted biophysical simulations in the software NEURON. Our models show that the observed phenomenon could be accounted for by rapid regulation of synaptic conductance or, alternatively, stem from electrical attenuation effects due to the changes in spine neck resistance associated with changes in neck length. The spine neck resistance values necessary to entirely account for such attenuation are at odds with reported estimates (13, 32), so one would be inclined to assume that a rapid increase in synaptic conductance leads to the observed changes in somatic EPSP size. However, because spine neck resistance values have so far been inferred only indirectly, one cannot rule out the possibility that a combination of (synaptic) conductance and (neck) resistance changes could contribute to the observed activity-dependent changes in somatic EPSP size.     

Sleep promotes the formation of dendritic filopodia and spines near learning-inactive existing spines

Changes in synaptic connections are believed to underlie long-term memory storage. Previous studies have suggested that sleep is important for synapse formation after learning, but how sleep is involved in the process of synapse formation remains unclear. To address this question, we used transcranial two-photon microscopy to investigate the effect of postlearning sleep on the location of newly formed dendritic filopodia and spines of layer 5 pyramidal neurons in the primary motor cortex of adolescent mice. We found that newly formed filopodia and spines were partially clustered with existing spines along individual dendritic segments 24 h after motor training. Notably, posttraining sleep was critical for promoting the formation of dendritic filopodia and spines clustered with existing spines within 8 h. A fraction of these filopodia was converted into new spines and contributed to clustered spine formation 24 h after motor training. This sleep-dependent spine formation via filopodia was different from retraining-induced new spine formation, which emerged from dendritic shafts without prior presence of filopodia. Furthermore, sleep-dependent new filopodia and spines tended to be formed away from existing spines that were active at the time of motor training. Taken together, these findings reveal a role of postlearning sleep in regulating the number and location of new synapses via promoting filopodial formation.

Learning and memory consolidation are associated with the rewiring of neuronal network connectivity (1–3). Previous studies have shown that motor training leads to the formation and elimination of postsynaptic dendritic spines of pyramidal neurons in the primary motor cortex (M1) (4–8). Learning-induced new spines stabilize and persist over long periods of time (4). The extent of spine remodeling correlates with behavioral improvement after learning (4, 9), and the disruption of spine remodeling impairs learned motor behavior (10–12). These studies suggest that learning-induced new synapses contribute to changes in neuronal circuits that are likely important for the retention of learned behaviors (13, 14).Cumulative evidence suggests that sleep affects synaptic structural plasticity in many brain regions (15–17). For example, sleep has been shown to promote spine formation and elimination in developing somatosensory and visual cortices (18, 19). In the motor cortex, sleep promotes branch-specific formation of new dendritic spines following motor learning and selectively stabilizes learning-induced new synaptic connections (11, 12). Sleep has also been shown to regulate dendritic spine numbers in hippocampal CA1 area (20–22). In addition, many lines of evidence have revealed the function of sleep in increasing, decreasing, or stabilizing synaptic strength and neuronal firing in various brain regions (23–31). Together, these studies strongly suggest that sleep has an important role in promoting synaptic structural plasticity in neuronal circuits during development and after learning.While sleep promotes the formation of new spines after learning (12), it remains unknown how postlearning sleep regulates new synapse formation along dendritic branches. Synapse formation is a prolonged process often involving the generation of dendritic filopodia, thin and long protrusions without bulbous heads (32–35). These highly dynamic filopodia have been shown to initiate the contact with presynaptic axonal terminals and transform into new spines (36, 37). It is not known whether sleep promotes new spine formation via filopodia formation and subsequent transformation. Furthermore, it is also unclear whether sleep-dependent formation of new dendritic protrusions (filopodia and spines) is distributed on dendritic branches in a random or nonrandom manner. On the one hand, new synapses may be formed in clusters with synapses of similar functions to allow nonlinear summation of inputs important for increasing memory storage capacity (9, 38–43). On the other hand, new connections may be formed preferentially near less active/strong synapses to avoid competition for limited synaptic resources (44–47).In this study, we found that dendritic filopodia and spines formed after motor training were partially clustered with existing spines on apical tuft dendrites of layer 5 (L5) pyramidal neurons in the mouse primary motor cortex. Posttraining sleep was critical for the clustered formation of new filopodia, some of which were transformed into new spines. In addition, the clustered new filopodia and spines tended to be formed near existing spines that were inactive at the time of motor training. These findings reveal a role for sleep in neuronal circuit plasticity by promoting clustered spine formation via dendritic filopodia near learning-inactive existing spines.     

Control of cerebellar granule cell output by sensory-evoked Golgi cell inhibition

Classical feed-forward inhibition involves an excitation–inhibition sequence that enhances the temporal precision of neuronal responses by narrowing the window for synaptic integration. In the input layer of the cerebellum, feed-forward inhibition is thought to preserve the temporal fidelity of granule cell spikes during mossy fiber stimulation. Although this classical feed-forward inhibitory circuit has been demonstrated in vitro, the extent to which inhibition shapes granule cell sensory responses in vivo remains unresolved. Here we combined whole-cell patch-clamp recordings in vivo and dynamic clamp recordings in vitro to directly assess the impact of Golgi cell inhibition on sensory information transmission in the granule cell layer of the cerebellum. We show that the majority of granule cells in Crus II of the cerebrocerebellum receive sensory-evoked phasic and spillover inhibition prior to mossy fiber excitation. This preceding inhibition reduces granule cell excitability and sensory-evoked spike precision, but enhances sensory response reproducibility across the granule cell population. Our findings suggest that neighboring granule cells and Golgi cells can receive segregated and functionally distinct mossy fiber inputs, enabling Golgi cells to regulate the size and reproducibility of sensory responses.Classical feed-forward inhibition (FFI) involves a sequence of excitation rapidly terminated by inhibition. This temporal sequence narrows the time window for synaptic integration and enforces precise spike timing (1–7). FFI is thought to be important for regulating the temporal fidelity of spike responses in many neural systems, including the motor system, where rapid and adaptable changes in muscle activity are essential for coordinated motor control (8–10). The cerebellum plays a central role in fine sculpting of movements, and damage to the cerebellum produces severe motor deficits, most notably enhanced temporal variability of voluntary movements (11, 12). These findings suggest that cerebellar circuits have the ability to preserve precise timing information during behavior (5, 6, 13), and in vitro studies have shown that feed-forward inhibitory networks in the input layer of the cerebellum provide a mechanism for maintaining the temporal fidelity of information transmission (6, 14, 15).Synaptic inhibition in the granule cell layer is generated by Golgi cells, GABAergic interneurons that provide direct inhibitory input to granule cells (6, 15–17). The prevailing view is that, when mossy fibers are activated, granule cells receive both monosynaptic excitation and disynaptic FFI from Golgi cells, providing temporally precise inhibitory input that narrows the window for the temporal summation of discrete mossy fiber inputs (6, 14, 18). This classical excitation–inhibition sequence forms the basis of a variety of contemporary cerebellar models (7, 9, 18, 19). However, the exact temporal relationship between sensory-evoked excitation and inhibition in granule cells has never been determined in vivo. Here, we combined in vivo whole-cell voltage-clamp recordings from granule cells and in vitro dynamic clamp experiments to investigate both the temporal dynamics of Golgi-cell–mediated inhibition and its importance for shaping sensory responses in the input layer of the cerebellum.     

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.     

Cortical control of adaptation and sensory relay mode in the thalamus

A major synaptic input to the thalamus originates from neurons in cortical layer 6 (L6); however, the function of this cortico–thalamic pathway during sensory processing is not well understood. In the mouse whisker system, we found that optogenetic stimulation of L6 in vivo results in a mixture of hyperpolarization and depolarization in the thalamic target neurons. The hyperpolarization was transient, and for longer L6 activation (>200 ms), thalamic neurons reached a depolarized resting membrane potential which affected key features of thalamic sensory processing. Most importantly, L6 stimulation reduced the adaptation of thalamic responses to repetitive whisker stimulation, thereby allowing thalamic neurons to relay higher frequencies of sensory input. Furthermore, L6 controlled the thalamic response mode by shifting thalamo–cortical transmission from bursting to single spiking. Analysis of intracellular sensory responses suggests that L6 impacts these thalamic properties by controlling the resting membrane potential and the availability of the transient calcium current I_T, a hallmark of thalamic excitability. In summary, L6 input to the thalamus can shape both the overall gain and the temporal dynamics of sensory responses that reach the cortex.Sensory signals en route to the cortex undergo profound signal transformations in the thalamus. One important thalamic transformation is sensory adaptation. Adaptation is a common characteristic of sensory systems in which neural output adjusts to the statistics and dynamics of past stimuli, thereby better encoding small stimulus changes across a wide range of scales despite the limited range of possible neural outputs (1–3). Thalamic sensory adaptation is characterized by a steep decrease in action potential (AP) activity during sustained sensory stimulation (4–7), decreasing the efficacy at which subsequent sensory stimuli are transmitted to the cortex.The widely reported duality of thalamic response mode is another key property of thalamic information processing which further affects how sensory input reaches the cortex. In burst mode, sensory inputs are relayed as short, rapid clusters of APs; in contrast, in tonic mode the same inputs are translated into single APs. Both tonic and burst modes have been described during anesthesia/sleep and wakefulness/behavior, with a pronounced shift toward the tonic mode during alertness (8–12).Although the exact information content of thalamic bursts is not yet clear, it has been suggested that bursting may signal novel stimuli to the cortex, whereas the tonic mode enables linear encoding of fine stimulus details, e.g., when an object is examined (13, 14). One issue hampering the interpretation of burst/tonic responses is that currently it is unknown if the cortex itself is involved in the rapid changes in firing modes seen in the awake and anesthetized animal (15, 16) and which mechanisms initiate these shifts in vivo.On the biophysical level, the response mode depends on the resting membrane potential (RMP), which controls the availability of the transient low-threshold calcium current (I_T) (17). Depolarization decreases the size of the I_T-mediated low-threshold calcium spike (LTS), and fewer burst spikes are fired (18). Similarly, RMP influences adaptation in that depolarization reduces the voltage distance to the AP threshold, thereby increasing the probability that smaller, depressed inputs will trigger APs (6). Thus, the dynamics of the RMP may govern several key properties of signal transformation in the thalamus, thereby providing a common mechanism for controlling thalamic adaptation and response mode.Although subcortical inputs have been shown to influence thalamic firing modes (7, 9), we investigated the impact of cortical activity on thalamic sensory processing. Cortico–thalamic projections from cortical layer 6 (L6) are a likely candidate for regulating thalamic sensory processing with high spatial and temporal precision, because these projections provide a major input to the thalamus and, as shown by McCormick et al. (19), depolarize and modulate firing of thalamic cells in vitro.However, because of the inability to study sensory signals in brain slices, the role of L6 on thalamic input/output properties during sensory processing is not clear. Here, in the ventro posteromedial nucleus (VPM) of the mouse whisker thalamus, we investigate how L6 impacts the transmission of whisker inputs to the cortex. Recent advances in cell-type–specific approaches to dissect specific circuits in vivo (20–22) allowed us to activate the L6–thalamic pathway specifically and determine its impact on thalamic sensory processing.We found that cortical L6 can change key properties of thalamic sensory processing by controlling the interaction of intrinsic membrane properties and sensory inputs. This mechanism enables the cortex to control the frequency-dependent adaptation and the gain of its own input.     

Rapid homeostasis by disinhibition during whisker map plasticity

How homeostatic processes contribute to map plasticity and stability in sensory cortex is not well-understood. Classically, sensory deprivation first drives rapid Hebbian weakening of spiking responses to deprived inputs, which is followed days later by a slow homeostatic increase in spiking responses mediated by excitatory synaptic scaling. Recently, more rapid homeostasis by inhibitory circuit plasticity has been discovered in visual cortex, but whether this process occurs in other brain areas is not known. We tested for rapid homeostasis in layer 2/3 (L2/3) of rodent somatosensory cortex, where D-row whisker deprivation drives Hebbian weakening of whisker-evoked spiking responses after an unexplained initial delay, but no homeostasis of deprived whisker responses is known. We hypothesized that the delay reflects rapid homeostasis through disinhibition, which masks the onset of Hebbian weakening of L2/3 excitatory input. We found that deprivation (3 d) transiently increased whisker-evoked spiking responses in L2/3 single units before classical Hebbian weakening (≥5 d), whereas whisker-evoked synaptic input was reduced during both periods. This finding suggests a transient homeostatic increase in L2/3 excitability. In whole-cell recordings from L2/3 neurons in vivo, brief deprivation decreased whisker-evoked inhibition more than excitation and increased the excitation–inhibition ratio. In contrast, synaptic scaling and increased intrinsic excitability were absent. Thus, disinhibition is a rapid homeostatic plasticity mechanism in rodent somatosensory cortex that transiently maintains whisker-evoked spiking in L2/3, despite the onset of Hebbian weakening of excitatory input.During deprivation-induced sensory map plasticity in cerebral cortex, changes in sensory input trigger both homeostatic plasticity mechanisms that maintain stable cortical firing rates and Hebbian mechanisms, in which inactive inputs lose (and active inputs gain) representation in sensory maps (1). Diverse mechanisms for homeostasis exist, including synaptic scaling (2–4), plasticity of intrinsic excitability (5, 6), and changes in sensory-evoked inhibition and excitation–inhibition (E-I) ratio (7–11). How homeostatic and Hebbian mechanisms interact to control map stability and plasticity remains unclear.One key unknown is the relative dynamics of homeostatic and Hebbian plasticity. Homeostasis mediated by synaptic scaling is slow, occurring over hours in vitro and days in vivo. This process is evident in visual cortex, where eyelid closure during the critical period classically drives Hebbian weakening of closed eye spiking responses (after 2 d of deprivation) followed several days later by a slower homeostatic increase in visual responses (12, 13) mediated by excitatory synaptic scaling (3, 14, 15). We investigated whether more rapid forms of homeostasis also exist that shape the earliest stages of cortical plasticity. Recent results in visual cortex show that eyelid closure rapidly weakens inhibitory circuits (within 1 d), and this process increases network excitability and, therefore, is an initial homeostatic response to deprivation (10, 16). This disinhibition correlates with rapid structural plasticity in inhibitory axons and dendrites (17) and is mediated by a reduction in excitatory drive to parvalbumin-positive interneurons (10). Whether rapid homeostasis by disinhibition or other mechanisms is a general feature of cortical plasticity outside the visual cortex is unknown. Theoretical work shows that rapid homeostasis by inhibitory and/or intrinsic plasticity can guide development of realistic sensory tuning and sparse sensory coding in cortical networks, suggesting broad relevance (18).We tested for rapid homeostasis during the onset of whisker map plasticity in the rodent primary somatosensory (S1) cortex, a major model of cortical plasticity. Each cortical column in the S1 whisker map corresponds to one facial whisker, termed its principal whisker (PW). Trimming or plucking a subset of whiskers in young adults weakens spiking responses to deprived PWs in layer 2/3 (L2/3) of deprived columns (19, 20). This process is mediated by Hebbian synaptic weakening at L4–L2/3 and L2/3–L2/3 excitatory synapses (21–23). No homeostatic restoration or strengthening of deprived whisker responses is known. However, PW response weakening is often preceded by an unexplained initial delay of ∼7 d, in which deprived whisker-evoked spiking responses remain stable (24, 25). We hypothesized that this initial delay reflects not a lack of plasticity but rapid homeostasis that (i) masks initial Hebbian weakening of L2/3 excitatory input and (ii) is mediated by loss of inhibition and/or increased intrinsic excitability in L2/3 neurons. Such rapid homeostasis would be a unique component of whisker map plasticity.     

A stable proportion of Purkinje cell inputs from parallel fibers are silent during cerebellar maturation

Cerebellar Purkinje neurons integrate information transmitted at excitatory synapses formed by granule cells. Although these synapses are considered essential sites for learning, most of them appear not to transmit any detectable electrical information and have been defined as silent. It has been proposed that silent synapses are required to maximize information storage capacity and ensure its reliability, and hence to optimize cerebellar operation. Such optimization is expected to occur once the cerebellar circuitry is in place, during its maturation and the natural and steady improvement of animal agility. We therefore investigated whether the proportion of silent synapses varies over this period, from the third to the sixth postnatal week in mice. Selective expression of a calcium indicator in granule cells enabled quantitative mapping of presynaptic activity, while postsynaptic responses were recorded by patch clamp in acute slices. Through this approach and the assessment of two anatomical features (the distance that separates adjacent planar Purkinje dendritic trees and the synapse density), we determined the average excitatory postsynaptic potential per synapse. Its value was four to eight times smaller than responses from paired recorded detectable connections, consistent with over 70% of synapses being silent. These figures remained remarkably stable across maturation stages. According to the proposed role for silent synapses, our results suggest that information storage capacity and reliability are optimized early during cerebellar maturation. Alternatively, silent synapses may have roles other than adjusting the information storage capacity and reliability.

Typical central excitatory synapses are formed onto dendritic spines, the distinctive morphology of which enables their unambiguous identification (1–3). It has generally been assumed that the presence of a dendritic spine equates to the existence of a functional excitatory transmission (4). Based on this assumption, the observation of spine motility in several brain areas (motor cortex and somatosensory cortex) has been considered to reflect synaptic plasticity (5). Indeed, a number of studies have established a correlation between learning and spine formation (6–8) or pruning (9, 10). However, in some conditions, morphological and synaptic plasticity have been shown to be dissociated (11), in line with the view that morphology does not provide all the information necessary to infer synaptic function.The cerebellum contains the majority of brain neurons (12, 13) and the predominant excitatory synapses found in this structure connect granule cells (GC) to Purkinje cells (PC). These synapses are formed on typical spines borne by PC dendrites (14), the majestic shape of which is likely related to the huge amount of independent inputs they receive. The GC-to-PC synapse is generally acknowledged to be an essential site for plasticity (15–19). However, in sharp contrast to other parts of the brain such as motor and somatosensory cortices, PC spines appear to be constitutive (20, 21), i.e., they appear to be an inherent property of PCs, independent of external factors. Indeed, pruning of these synapses has not been reported. Novel spine formation has been reported, but likely as a result of dendritic tree expansion (22, 23). The high density of spines along PC dendritic branchlets (5 to 17 per linear micrometer in rat; refs. 14 and 24–27) and their regular ordering in a helical pattern (28) support the idea that they optimize space occupancy with little room for spine addition, in accordance with their constitutive nature.The apparent morphological homogeneity of PC spines is in sharp contrast with the spectacular heterogeneity observed in the strength of GC-to-PC synapses. An in vivo study has reported that the receptive field of a PC was much smaller than that of the GCs putatively connecting to it (29), suggesting that most GC-to-PC synapses are electrically silent. This has been confirmed by an in vitro report (30) showing that synaptic transmission between paired-recorded GC and PC was detected nearly 10 times less frequently than expected from the occurrence of morphologically defined synaptic connections predicted by anatomical data (14, 31–33). Taken together, these two studies conducted in adult rats suggest that most (85% according to ref. 30) morphologically and molecularly defined GC-to-PC synapses are silent, i.e., they do not transmit any detectable electrical signal.If silent synapses do not transmit information, what is their role? Are they a reserve for additional information storage? Or do they result from information storage optimization (34)? According to this latter proposal (17, 34), since the requirement for optimized information storage is more and more critical as the amount of learned information increases, one might expect that the proportion of silent synapses increases with the amount learning. As previously suggested (34), this hypothesis could be tested by comparing the proportion of silent synapses in young versus adult animals. Indeed, the mouse cerebellar circuitry is not fully in place until the third postnatal week. Then, for at least 3 wk, the mouse acquires basic skills (eating, walking, and social interactions), adapts to changes in muscle strengths and sensitivity to stimuli, and improves its agility (35). Although the amount of cerebellar learning occurring over this maturation period is unclear, it can be reasonably assumed that cerebellar operation continuously optimizes. Here, we investigate how the proportion of silent synapses changes over this period of maturation.We determine the proportion of silent GC-to-PC synapses by a method based on the determination of the average postsynaptic response per activated synapse (average synaptic weight, w¯) in superfused acute slices. Thanks to the geometrical and repetitive architecture of the cerebellar cortex, calcium imaging is used to quantitatively map GC inputs. This mapping, combined with postsynaptic recording of transmission, and the determination of two cerebellar anatomical features (the average synapse density and spacing between PC planar dendritic trees) enables the determination of w¯. By comparison with the properties of synapses that produce an electrical postsynaptic response (investigated by paired recording and quantal analysis), we show that the proportion of silent synapses is higher than 70% and stable between the postnatal stages of interest. This suggests that cerebellar maturation has insignificant impact on the proportion of silent synapses.     

Adhesion GPCR Latrophilin 3 regulates synaptic function of cone photoreceptors in a trans-synaptic manner

Cone photoreceptors mediate daylight vision in vertebrates. Changes in neurotransmitter release at cone synapses encode visual information and is subject to precise control by negative feedback from enigmatic horizontal cells. However, the mechanisms that orchestrate this modulation are poorly understood due to a virtually unknown landscape of molecular players. Here, we report a molecular player operating selectively at cone synapses that modulates effects of horizontal cells on synaptic release. Using an unbiased proteomic screen, we identified an adhesion GPCR Latrophilin3 (LPHN3) in horizontal cell dendrites that engages in transsynaptic control of cones. We detected and characterized a prominent splice isoform of LPHN3 that excludes a element with inhibitory influence on transsynaptic interactions. A gain-of-function mouse model specifically routing LPHN3 splicing to this isoform but not knockout of LPHN3 diminished Ca_V1.4 calcium channel activity profoundly disrupted synaptic release by cones and resulted in synaptic transmission deficits. These findings offer molecular insight into horizontal cell modulation on cone synaptic function and more broadly demonstrate the importance of alternative splicing in adhesion GPCRs for their physiological function.

Vision is a key sensory modality essential for the survival of most living organisms. In mammals, it is enabled by the retina: a neural structure composed of more than 60 distinct neurons each uniquely wired into the circuitry and with particular roles in image processing (1, 2). Vision begins with the detection of light by rod and cone photoreceptors. Rod photoreceptor cells are exquisitely sensitive to light and mediate vision at low light levels (3, 4). However, most vertebrates including humans rely on cone cells for daytime vision (5). Accordingly, cones have an extremely broad range of light sensitivity spanning 6 to 7 orders of magnitude (6), quickly adapting to changes in luminance and providing high spatial and temporal visual acuity (5, 7, 8). The molecular, cellular, and circuit mechanisms that allow cones to perform their tasks has been a subject of intense interest, providing groundbreaking discoveries that illuminate fundamental organizational principles that govern signal processing by neural circuits in general.The capture and processing of photons by the phototransduction cascade of cones generates graded changes in membrane potential: hyperpolarizing to light and depolarizing with darkness (7). These voltage signals alter the ongoing rate of neurotransmitter glutamate release at the cone synapse to relay information about light and dark to the retinal circuitry (9). The molecular entity that mediates this transformation is the L-type voltage-gated Ca²⁺ channel, Ca_V1.4 (10–12). It is located at specialized active zones containing synaptic ribbons and couples light-driven changes in voltage to changes in local Ca²⁺ levels thereby regulating the vesicular fusion machinery (13, 14). The Ca_V1.4 channel forms a macromolecular complex with a number of synaptic molecules and thus plays a pivotal role in both the structural and functional organization of the presynaptic active zone of photoreceptors (15). Accordingly, changes in Ca_V1.4 function imposed by binding partners or environment have a tremendous impact on the synaptic communication of cone photoreceptors and vision (16–18).Cones form synaptic contacts with three types of neurons. They synapse with postsynaptic ON- and OFF-type bipolar cells (BC) to relay visual information to the downstream neuronal circuitry (19, 20). Cones also contact lateral inhibitory neurons known as horizontal cells (HCs) that connect adjacent to BC dendrites, forming a tripartite synaptic triad (20). This elaborate synaptic arrangement of cones is a site of major influence on how visual information is processed contributing to unique cone physiology and adaptive capacity for daylight detection (21, 22).The function of HCs and their physiological mechanisms are particularly intriguing. HCs powerfully modulate synaptic transmission at cone synapses (23). Light-evoked hyperpolarization of HCs counteracts light-induced suppression of glutamate release from cone terminals, thereby providing strong negative feedback (23, 24). Because each HC contacts multiple cones, this negative influence on surrounding cones is a major mechanism for producing lateral inhibition, a classical feature of signal processing in the retina that enhances contrast and spatial resolution of vision (25). In addition, feedback from individual HC dendrites to specific cone terminals and ribbons can also act locally, fine-tuning synaptic output to local illumination gradients (26–29).While the role of HCs from a circuit perspective is well understood, the mechanisms that they use to provide negative feedback are subject to debate and controversy. At least three different explanations have been provided: direct ephaptic effects (30, 31), changes in synaptic pH (28, 32, 33), and modulation by GABA released from HCs (34, 35). These models are not necessarily mutually exclusive and unifying theories have been proposed (35, 36). Importantly, one of the central effects invariably observed in response to HC feedback is modulation of the Ca_V1.4 function at the active zones of cone terminals (32, 37). However, there is a significant void in our understanding of molecular mechanisms by which HCs modulate transmission of cone signals, mostly due to a paucity of players known to operate at this synapse. Identification and functional characterization of molecular elements involved in coordinating HC influence in cone synapses can transform our understanding of this enigmatic area of visual neuroscience.Here, we performed an unbiased proteomic profiling of proteins selectively enriched in cone synapses. This led to identification of an adhesion G protein–coupled receptor (aGPCR), latrophilin3 (LPHN3), whose role in retina physiology, photoreceptor synaptic development, and function was previously unexplored. We show that alternative splicing of LPHN3 in the retina generates unique isoforms with distinct properties. Using mouse models, we demonstrate that changes in LPHN3 splicing regulate cone synaptic transmission transsynaptically by affecting Ca_V1.4 function. These findings reveal a molecular player with a pivotal role in regulating synaptic function of cone photoreceptors.     

Dynamics of dendritic spines in the mouse auditory cortex during memory formation and memory recall

Long-lasting changes in synaptic connections induced by relevant experiences are believed to represent the physical correlate of memories. Here, we combined chronic in vivo two-photon imaging of dendritic spines with auditory-cued classical conditioning to test if the formation of a fear memory is associated with structural changes of synapses in the mouse auditory cortex. We find that paired conditioning and unpaired conditioning induce a transient increase in spine formation or spine elimination, respectively. A fraction of spines formed during paired conditioning persists and leaves a long-lasting trace in the network. Memory recall triggered by the reexposure of mice to the sound cue did not lead to changes in spine dynamics. Our findings provide a synaptic mechanism for plasticity in sound responses of auditory cortex neurons induced by auditory-cued fear conditioning; they also show that retrieval of an auditory fear memory does not lead to a recapitulation of structural plasticity in the auditory cortex as observed during initial memory consolidation.Mammalian brains are characterized by a tremendous level of plasticity. This plasticity is believed to underlie the ability to extract and store information about past experiences and is crucial for animals and humans to interact adaptively in a changing environment. Therefore, detection and localization of a physical representation of a memory has been an intriguing aspect for a long time (1). Plastic changes in synapses are believed to be substrates of memory (2). The development of imaging techniques that allow chronic monitoring of dendritic spines, the morphological correlates of excitatory synapses on pyramidal neurons, in the living animal has provided valuable insights in the dynamics of neuronal circuits (3–5). It has recently been shown that not only chronic perturbations of sensory inputs (6, 7), but also temporally restricted learning experiences, impact the turnover of synaptic structures in the motor cortex and frontal association cortex of the mouse (8–10) and the high vocal center in zebra finches (11).Auditory-cued fear conditioning (ACFC) is an associative learning paradigm that has been widely used to analyze mechanisms of learning in the auditory modality (12). During a conditioning session, subjects quickly learn to associate a previously neutral sound cue [the conditional stimulus (CS)] with an aversive stimulus like a mild foot shock [unconditional stimulus (US)]. It is well established that memory traces after initial formation undergo several processes at different time scales that lead to their consolidation and render them to a stable state that is, e.g., resistant to trauma introduced by an electroconvulsive shock (13). Interestingly, similar molecular cascades are triggered not only during memory formation, but also when a memory trace is retrieved (14, 15). Furthermore, memory traces that were recently retrieved become sensitive again to manipulations like electroconvulsive shock (16), blockade of NMDA receptors (17), or blockade of protein synthesis (18). These similarities have been suggested to reflect remodeling of memory traces following recall (14, 19). However, data on the dynamics of synaptic structures during memory recall is lacking up to date.A number of brain structures have been identified mediating the formation of a memory induced by ACFC (12). Whereas inputs via the auditory cortex (ACx) to the amygdala, an essential brain structure for this learning paradigm, appear to be always sufficient to support fear conditioning (20), their necessity can depend on the spectrotemporal properties of the auditory CS (21, 22). The ACx as the primary sensory cortical area for the auditory modality has been extensively analyzed in the past during classical conditioning to sound stimuli (23–25) or pairing of sounds with artificial stimulation of the cholinergic system, which can substitute for aspects of the US (26–28). These paradigms lead to changes in the receptive fields of ACx neurons that are specific to the conditioned sound. There is evidence based on local pharmacological or optogenetic manipulations that plasticity of the ACx itself is necessary for experience-induced alterations in sound responses and does not simply reflect plasticity elsewhere in the auditory pathway (29, 30). Indeed, there is evidence based on electrophysiological recordings that synaptic plasticity at intracortical synapses can be induced by pairing of a sound with stimulation of the cholinergic system in vivo (28). Structural plasticity following ACFC has been observed in the frontal association cortex (10). However, it remains elusive if plasticity in sound responses in the ACx induced by ACFC (23–25) also has a structural correlate at the synaptic level.In this study, we asked two major questions: Does ACFC in behaving mice induce structural plasticity in synaptic circuits of the ACx? To what extent do memory formation and memory recall share similarities at the level of synaptic structures? We addressed these questions by combining sound-cued fear conditioning and memory testing with chronic in vivo imaging of dendritic spines in the ACx.     

Robustness of sensory-evoked excitation is increased by inhibitory inputs to distal apical tuft dendrites

Cortical inhibitory interneurons (INs) are subdivided into a variety of morphologically and functionally specialized cell types. How the respective specific properties translate into mechanisms that regulate sensory-evoked responses of pyramidal neurons (PNs) remains unknown. Here, we investigated how INs located in cortical layer 1 (L1) of rat barrel cortex affect whisker-evoked responses of L2 PNs. To do so we combined in vivo electrophysiology and morphological reconstructions with computational modeling. We show that whisker-evoked membrane depolarization in L2 PNs arises from highly specialized spatiotemporal synaptic input patterns. Temporally L1 INs and L2–5 PNs provide near synchronous synaptic input. Spatially synaptic contacts from L1 INs target distal apical tuft dendrites, whereas PNs primarily innervate basal and proximal apical dendrites. Simulations of such constrained synaptic input patterns predicted that inactivation of L1 INs increases trial-to-trial variability of whisker-evoked responses in L2 PNs. The in silico predictions were confirmed in vivo by L1-specific pharmacological manipulations. We present a mechanism—consistent with the theory of distal dendritic shunting—that can regulate the robustness of sensory-evoked responses in PNs without affecting response amplitude or latency.Mechanistic understanding of the principles that underlie sensory-evoked neuronal responses remains a key challenge in neuroscience research. Although electrophysiological and optical imaging techniques provide access to activity patterns of individual and/or populations of neurons in live animals, information about the organization of the underlying synaptic input patterns that drive neuronal activity remains scarce. Here, we investigate the mechanisms underlying whisker deflection-evoked responses in pyramidal neurons (PNs) in the vibrissal part of rat primary somatosensory cortex (vS1, i.e., barrel cortex) (1). Specifically, we wanted to know whether and how L1 interneurons (INs) shape responses of L2 PNs. L1 is densely populated by apical tuft dendrites from multiple types of excitatory PNs and a sparse population of GABAergic INs (2). Recent studies in acute parasagittal (3) and coronal (4) brain slices in vitro have shown that L1 INs have axonal projections largely confined to L1, where they form synaptic connections with the dendrites from PNs located in L2/3 (5) and L5 (4). These connections place L1 INs in a perfect position to manipulate the activity of PNs, for example, by feed-forward inhibition and/or more indirect mechanisms such as disinhibition (4, 6). However, the influence of L1 INs on the sensory-evoked responses of PNs remains unclear.To address this, we performed whole-cell patch-clamp recordings in vivo and reconstructed the 3D morphologies of the recorded L1 INs. These data, acquired under the same experimental conditions as previously used to determine whisker-evoked spiking and 3D morphologies for PN cell types (7), were used to inform and constrain simulation experiments. Specifically, we converted the 3D soma/dendrite morphology of an in vivo-labeled L2 PN into a biophysically detailed full-compartmental model (8) and integrated the neuron model into a recently reported detailed reconstruction of the excitatory circuitry in vS1 (9). This integration enabled us to statistically measure the number and subcellular distribution of cell type-specific synaptic contacts impinging onto the exemplary L2 PN from L1 INs and L2–5 PNs, respectively. These spatially constrained synaptic input patterns were further constrained temporally by using the measured cell type-specific spiking probabilities and latencies (7, 10). Finally, we made in silico experiments (i.e., numerical simulations) and investigated how the interplay between biophysical properties of the dendrites and well-constrained spatiotemporal synaptic input patterns give rise to the whisker-evoked responses measured in vivo (11).     

PirB regulates a structural substrate for cortical plasticity

Experience-driven circuit changes underlie learning and memory. Monocular deprivation (MD) engages synaptic mechanisms of ocular dominance (OD) plasticity and generates robust increases in dendritic spine density on L5 pyramidal neurons. Here we show that the paired immunoglobulin-like receptor B (PirB) negatively regulates spine density, as well as the threshold for adult OD plasticity. In PirB^−/− mice, spine density and stability are significantly greater than WT, associated with higher-frequency miniature synaptic currents, larger long-term potentiation, and deficient long-term depression. Although MD generates the expected increase in spine density in WT, in PirB^−/− this increase is occluded. In adult PirB^−/−, OD plasticity is larger and more rapid than in WT, consistent with the maintenance of elevated spine density. Thus, PirB normally regulates spine and excitatory synapse density and consequently the threshold for new learning throughout life.Experience generates both functional and structural changes in neural circuits. The learning process is robust at younger ages during developmental critical periods and continues, albeit at a lower level, into adulthood and old age (1–3). For example, young barn owls exposed to horizontally shifting prismatic spectacles can adapt readily to altered visual input, but adult owls cannot. The experience in the young owls results in a rearranged audiovisual map in tectum that is accompanied by ectopic axonal projections (1). Experience-dependent structural changes have also been observed in the mammalian cerebral cortex. Enriched sensory experience or motor learning are both associated with an increase in dendritic spine density, and a morphological shift from immature thin spines to mushroom spines which harbor larger postsynaptic densities (PSDs) and stronger synapses (4–7). On the flip side, bilateral sensory deprivation induces spine loss (8, 9). Abnormal sensory experience also results in structural modification of inhibitory synapses and circuitry that is temporally and spatially coordinated with changes in excitatory synapses on dendritic spines (10–13).These experience-driven spine changes are thought to involve synaptic mechanisms of long-term potentiation (LTP) and long-term depression (LTD). In hippocampal slices, induction of LTP causes new spines to emerge, as well as spine head enlargement on existing spines (14–16); induction of LTD results in rapid spine regression (14, 17). Importantly, the emergence or regression of spines starts soon after the induction of LTP or LTD, suggesting that these structural changes underlie the persistent expression of long-term plasticity (14, 17).Little is known about molecular mechanisms that restrict experience-dependent plasticity at circuit and synaptic levels and connect it to spine stability. Paired Ig-like receptor B (PirB), a receptor expressed in cortical pyramidal neurons, is known to limit ocular dominance (OD) plasticity both during the critical period and in adulthood (18). PirB binds major histocompatibility class I (MHCI) ligands, whose expression is regulated by visual experience and neural activity (19–21) and thus could act as a key link connecting functional to structural plasticity. If so, mice lacking PirB might be expected to have altered synaptic plasticity rules on the one hand and changes in the density and stability of dendritic spines on the other.     

Peripheral elevation of TNF-α leads to early synaptic abnormalities in the mouse somatosensory cortex in experimental autoimmune encephalomyelitis

Sensory abnormalities such as numbness and paresthesias are often the earliest symptoms in neuroinflammatory diseases including multiple sclerosis. The increased production of various cytokines occurs in the early stages of neuroinflammation and could have detrimental effects on the central nervous system, thereby contributing to sensory and cognitive deficits. However, it remains unknown whether and when elevation of cytokines causes changes in brain structure and function under inflammatory conditions. To address this question, we used a mouse model for experimental autoimmune encephalomyelitis (EAE) to examine the effect of inflammation and cytokine elevation on synaptic connections in the primary somatosensory cortex. Using in vivo two-photon microscopy, we found that the elimination and formation rates of dendritic spines and axonal boutons increased within 7 d of EAE induction—several days before the onset of paralysis—and continued to rise during the course of the disease. This synaptic instability occurred before T-cell infiltration and microglial activation in the central nervous system and was in conjunction with peripheral, but not central, production of TNF-α. Peripheral administration of a soluble TNF inhibitor prevented abnormal turnover of dendritic spines and axonal boutons in presymptomatic EAE mice. These findings indicate that peripheral production of TNF-α is a key mediator of synaptic instability in the primary somatosensory cortex and may contribute to sensory and cognitive deficits seen in autoimmune diseases.Sensory, motor, and cognitive dysfunctions are common at the early stages of autoimmune diseases, including in more than half of multiple sclerosis (MS) patients (1–5). The increased production of proinflammatory cytokines and infiltration of peripheral immune cells into the central nervous system (CNS) are closely associated with neuronal damage in the spinal cord and many brain regions (6–10). Elevation of several cytokines such as TNF-α and IFN-γ precedes infiltration of peripheral immune cells and could have a significant impact on neuronal function (11–19), potentially contributing to early sensory and cognitive impairments. However, the link between the cytokine elevation and CNS deficits in autoimmune diseases remains unclear.Experimental autoimmune encephalomyelitis (EAE) is a commonly used animal model to study the pathogenesis and treatment of MS (20). Previous studies have shown early behavioral changes, including decreased exploratory behavior and increased startle response before the onset of paralysis in EAE mice (21). The mechanisms underlying these early behavioral changes remain to be addressed. In EAE, activated antigen-specific T cells migrate through the blood brain barrier (BBB) and subsequently recruit additional myeloid and lymphoid cells to the spinal cord and susceptible brain regions (22–24). The resulting inflammatory cascade eventually leads to axonal loss and neuronal death. However, inflammatory infiltrates usually accompany the onset of clinical EAE and are therefore unlikely to contribute to sensory and behavioral deficits in presymptomatic mice. On the other hand, elevation of proinflammatory cytokines occurs during the early phase of EAE induction (25). Cytokines such as TNF-α are produced by activated cells within peripheral lymphoid tissues as well as by CNS-resident microglia in response to a number of inflammatory stimuli. Peripherally produced TNF-α can enter the circulation and cross the BBB through active transport (26–28) or passively after pathologic BBB disruption (27, 29). TNF-α has been shown to affect synaptic transmission and synaptic scaling (30–32), as well as regulate the production of other cytokines and chemokines to impact neuronal function. The induction and progression of EAE can be inhibited by either i.p. or intracranial injection of anti-TNF antibodies (33). Therefore, TNF-α elevation under pathological conditions could cause significant deficits in the CNS and contribute to behavioral abnormalities in EAE.TNF-α is known to affect neuronal functions by acting at two receptor types, TNF receptor (TNFR)1 and TNFR2, each having divergent effects in the CNS (16, 34). TNFR2 responds mainly to transmembrane TNF-α, which is thought to be a prosurvival signal. In contrast, TNFR1 responds mainly to circulating soluble TNF-α (solTNF) and plays a role in inflammation, proliferation, and neural migration. Activation of TNFR1 has been found to affect synaptic scaling by recruiting AMPA receptors to the postsynaptic cell membrane (30–32). TNFR1 activation also initiates the JNK pathway thought to be involved in dendritic cytoskeleton stability via regulation of microtubules (16). A dominant negative TNF inhibitor, XPro1595, blocks the effects of solTNF to selectively decrease the activation of TNFR1 (35, 36). Recently, decreased activation of TNFR1 via XPro1595 has been shown to promote axonal integrity and improve clinical outcome of mice with EAE (10).In this study, we examined the changes in synaptic connections in the primary somatosensory cortex in response to inflammation in an EAE mouse model. Using transcranial two-photon microscopy (37, 38), we followed postsynaptic dendritic spines and presynaptic axonal boutons over time in mice immunized with myelin oligodendrocyte glycoprotein peptide (MOG_35–55). We found that, in conjunction with an elevation of peripheral TNF-α production, the turnover of dendritic spines and axonal boutons was significantly increased in the cortex within 7 d of MOG immunization, several days before the onset of paralysis. The turnover of dendritic spines and axonal boutons continued to increase as the disease progressed. This early instability of cortical synaptic connections in EAE mice is independent of T-cell infiltration and microglia activation, but depends on the elevation of peripheral TNF-α in EAE mice.     

A Kalirin missense mutation enhances dendritic RhoA signaling and leads to regression of cortical dendritic arbors across development

Normally, dendritic size is established prior to adolescence and then remains relatively constant into adulthood due to a homeostatic balance between growth and retraction pathways. However, schizophrenia is characterized by accelerated reductions of cerebral cortex gray matter volume and onset of clinical symptoms during adolescence, with reductions in layer 3 pyramidal neuron dendritic length, complexity, and spine density identified in multiple cortical regions postmortem. Nogo receptor 1 (NGR1) activation of the GTPase RhoA is a major pathway restricting dendritic growth in the cerebral cortex. We show that the NGR1 pathway is stimulated by OMGp and requires the Rho guanine nucleotide exchange factor Kalirin-9 (KAL9). Using a genetically encoded RhoA sensor, we demonstrate that a naturally occurring missense mutation in Kalrn, KAL-PT, that was identified in a schizophrenia cohort, confers enhanced RhoA activitation in neuronal dendrites compared to wild-type KAL. In mice containing this missense mutation at the endogenous locus, there is an adolescent-onset reduction in dendritic length and complexity of layer 3 pyramidal neurons in the primary auditory cortex. Spine density per unit length of dendrite is unaffected. Early adult mice with these structural deficits exhibited impaired detection of short gap durations. These findings provide a neuropsychiatric model of disease capturing how a mild genetic vulnerability may interact with normal developmental processes such that pathology only emerges around adolescence. This interplay between genetic susceptibility and normal adolescent development, both of which possess inherent individual variability, may contribute to heterogeneity seen in phenotypes in human neuropsychiatric disease.

Schizophrenia (SZ) is a debilitating disease that affects ∼1% of the population (1). Clinical symptoms, such as auditory hallucinations and delusions, emerge during the second or third decade of life. Recent studies, however, have emphasized that it is impairments in cognitive and sensory processes underlying the clinical symptoms that are the greatest contributors to functional impairment and long-term disability (2–4). Specifically, auditory processing deficits have been consistently described at the neurophysiological level and include increased threshold for detecting differences between successive auditory stimuli and decreased amplitude of the mismatch negativity response to silent gaps, processes which require an intact of the auditory cortex (5, 6). Current therapeutics have limited efficacy for cognitive and sensory function impairments and confer substantial morbidity owing to undesired side effects, underscoring the need for therapeutics targeting underlying molecular mechanisms.Pyramidal cells (PCs) represent the most abundant neuronal type in the cerebral cortex (7), and their integrity is essential to the cognitive and sensory processes that are disrupted in SZ (5,8). Among the most consistent and highly replicated findings from human postmortem studies of SZ are reductions in dendrite length, branching, and spine density in layer 3 PCs (9). Interestingly, these deficits appear to be more reliably defined in layer 3, as layer 5 PCs have not been consistently shown to be impaired in postmortem studies of SZ (10). Because dendritic spines are the site of most excitatory synapses, much research to date has aimed to determine mechanisms of their reduction in SZ. However, total spine number is a function of both total dendrite length and spine density. Moreover, dendritic length and branching determine a PC’s receptive field (11, 12), help to segment computational compartments (13), and contribute substantially to how the received signals are integrated and transmitted to the cell body (14, 15). Thus, there is a compelling need to investigate dendritic length and branching alterations in SZ.One of the driving physiological requirements for dendritic morphogenesis is the flexibility for adjustment in development and in response to experience (16). Initial rapid growth establishes a nearly full-sized dendritic arbor prior to adolescence (17). While dendrites retain the physiological flexibility to undergo modest changes in branch points or angles to refine circuitry, the overall net arbor size remains relatively constant across adolescence and into adulthood due to a homeostatic balance between growth and retraction pathways (17, 18).Although rapid dendritic growth is complete prior to adolescence, this developmental epoch is a particularly active period of structural changes in the brain, leading to loss of cortical gray matter volume (19–22). In SZ, this reduction in gray matter volume is accelerated (23, 24), coincident with the onset of clinical symptoms (25, 26). The predominant component of cortical gray matter is neuropil, which comprises dendrites and axonal processes (27). It is possible that accelerated gray matter reductions during adolescence in SZ may be, in part, due to regression of dendritic architecture beginning during that time. It stands to reason, then, that a genetic susceptibility in a pathway involved in dendritic morphogenesis may be further exacerbated during the adolescent transition and lead to the onset of clinical symptoms of SZ.Among the genes found to influence dendritic morphogenesis (16, 28) is Kalrn. Of the multiple isoforms generated from the Kalrn gene through alternative splicing, the longer isoforms (KAL9 and KAL12) possess two guanine nucleotide exchange factor (GEF) domains, the second of which activates the GTPase RhoA. A missense mutation (rs143835330) coding for a proline to threonine amino acid change in the Kalrn gene (Kalrn-PT) was first identified in a resequencing analysis in individuals with SZ (29). The P→T change in Kalrn-PT is adjacent to the RhoA GEF domain in the KAL9 and KAL12 isoforms (29) and was shown to act as a modest gain of function for RhoA activity in a heterologous overexpression system (30). The activity of the first GEF domain, which activates Rac1, was shown to be unaltered by the Kalrn-PT mutation (30).Importantly, RhoA regulation of dendritic morphogenesis requires molecular precision. Although constitutively active RhoA reduces dendritic morphogenesis in rodent PCs (31, 32), expression of dominant negative (DN) RhoA fails to affect dendritic outgrowth (33). Thus, targeting specific pathways upstream of RhoA activation is necessary to rescue structural impairments. For example, p75 is a neurotrophin receptor which directly binds to Nogo receptor (NGR) and, in response to ligand binding, subsequently activates RhoA (34, 35). In cerebellar granule neurons, the KAL9 isoform of the Kalrn gene has been shown to directly bind to p75 and provide the GEF domain required for RhoA activation (34). The NGR1/p75/KAL9 pathway is known to restrict neurite outgrowth (34). Specifically disrupting NGR1-mediated signaling via Nogo neutralizing antibodies promotes neurite outgrowth and extension in vitro (34, 36), and in vivo knockdown of neuronal-specific Nogo-A leads to increases in both branching and total length in Layer 2/Layer 3 (L2/3) dendrites (37). Interestingly, increased levels of Nogo messenger RNA as well as elevated levels of KAL9 protein have been described in SZ (38, 39), suggesting enhanced activity of this pathway may contribute to the impairments in dendritic morphogenesis in disease. Although to date Nogo is the most well studied, numerous myelin-associated inhibitors (MAIs) have been identified as additional NGR1 ligands and similarly serve to limit neurite outgrowth (40). Interestingly, a highly potent MAI, oligodendrocyte-myelin glycoprotein (OMGp), increases in expression across adolescence (41).Thus, we hypothesized that Kalrn-PT would act as a RhoA gain of function in neurons and lead to adolescent-onset reductions in dendritic length and complexity.