Local Axonal Conduction Shapes the Spatiotemporal Properties of Neural Sequences

Exploiting Signal Propagation Delays to Match Task Memory Requirements in Reservoir Computing

Recurrent neural networks (RNNs) transmit information over time through recurrent connections. In contrast, biological neural networks use many other temporal processing mechanisms. One of these mechanisms is the inter-neuron delays caused by varying axon properties. Recently, this feature was implemented in echo state networks (ESNs), a type of RNN, by assigning spatial locations to neurons and introducing distance-dependent inter-neuron delays. These delays were shown to significantly improve ESN task performance. However, thus far, it is still unclear why distance-based delay networks (DDNs) perform better than ESNs. In this paper, we show that by optimizing inter-node delays, the memory capacity of the network matches the memory requirements of the task. As such, networks concentrate their memory capabilities to the points in the past which contain the most information for the task at hand. Moreover, we show that DDNs have a greater total linear memory capacity, with the same amount of non-linear processing power.

Unsupervised restoration of a complex learned behavior after large-scale neuronal perturbation

Reliable execution of precise behaviors requires that brain circuits are resilient to variations in neuronal dynamics. Genetic perturbation of the majority of excitatory neurons in HVC, a brain region involved in song production, in adult songbirds with stereotypical songs triggered severe degradation of the song. The song fully recovered within 2 weeks, and substantial improvement occurred even when animals were prevented from singing during the recovery period, indicating that offline mechanisms enable recovery in an unsupervised manner. Song restoration was accompanied by increased excitatory synaptic input to neighboring, unmanipulated neurons in the same brain region. A model inspired by the behavioral and electrophysiological findings suggests that unsupervised single-cell and population-level homeostatic plasticity rules can support the functional restoration after large-scale disruption of networks that implement sequential dynamics. These observations suggest the existence of cellular and systems-level restorative mechanisms that ensure behavioral resilience.

Bursts from the past: Intrinsic properties link a network model to zebra finch song

Neuronal intrinsic excitability is a mechanism implicated in learning and memory that is distinct from synaptic plasticity. Prior work in songbirds established that intrinsic properties (IPs) of premotor basal-ganglia-projecting neurons (HVC X ) relate to learned song. Here we find that temporal song structure is related to specific HVC X IPs: HVC X from birds who sang longer songs including longer invariant vocalizations (harmonic stacks) had IPs that reflected increased post-inhibitory rebound. This suggests a rebound excitation mechanism underlying the ability of HVC X neurons to integrate over long periods of time and represent sequence information. To explore this, we constructed a network model of realistic neurons showing how in-vivo HVC bursting properties link rebound excitation to network structure and behavior. These results demonstrate an explicit link between neuronal IPs and learned behavior. We propose that sequential behaviors exhibiting temporal regularity require IPs to be included in realistic network-level descriptions.

Breaking reflection symmetry: evolving long dynamical cycles in Boolean systems

In interacting dynamical systems, specific local interaction rules for system components give rise to diverse and complex global dynamics. Long dynamical cycles are a key feature of many natural interacting systems, especially in biology. Examples of dynamical cycles range from circadian rhythms regulating sleep to cell cycles regulating reproductive behavior. Despite the crucial role of cycles in nature, the properties of network structure that give rise to cycles still need to be better understood. Here, we use a Boolean interaction network model to study the relationships between network structure and cyclic dynamics. We identify particular structural motifs that support cycles, and other motifs that suppress them. More generally, we show that the presence of dynamical reflection symmetry in the interaction network enhances cyclic behavior. In simulating an artificial evolutionary process, we find that motifs that break reflection symmetry are discarded. We further show that dynamical reflection symmetries are over-represented in Boolean models of natural biological systems. Altogether, our results demonstrate a link between symmetry and functionality for interacting dynamical systems, and they provide evidence for symmetry's causal role in evolving dynamical functionality.

Neural Sequences and the Encoding of Time

Converging experimental and computational evidence indicate that on the scale of seconds the brain encodes time through changing patterns of neural activity. Experimentally, two general forms of neural dynamic regimes that can encode time have been observed: neural population clocks and ramping activity. Neural population clocks provide a high-dimensional code to generate complex spatiotemporal output patterns, in which each neuron exhibits a nonlinear temporal profile. A prototypical example of neural population clocks are neural sequences, which have been observed across species, brain areas, and behavioral paradigms. Additionally, neural sequences emerge in artificial neural networks trained to solve time-dependent tasks. Here, we examine the role of neural sequences in the encoding of time, and how they may emerge in a biologically plausible manner. We conclude that neural sequences may represent a canonical computational regime to perform temporal computations.

Uncoordinated sleep replay across hemispheres in the zebra finch

Bilaterally organized brain regions are often simultaneously active in both humans1,2,3 and animal models,4,5,6,7,8,9 but the extent to which the temporal progression of internally generated dynamics is coordinated across hemispheres and how this coordination changes with brain state remain poorly understood. To address these issues, we investigated the zebra finch courtship song (duration: 0.5-1.0 s), a highly stereotyped complex behavior10,11 produced by a set of bilaterally organized nuclei.12,13,14 Unilateral lesions to these structures can eliminate or degrade singing,13,15,16,17 indicating that both hemispheres are required for song production.18 Additionally, previous work demonstrated broadly coherent and symmetric bilateral premotor signals during song.9 To precisely track the temporal evolution of activity in each hemisphere, we recorded bilaterally in the song production pathway. We targeted the robust nucleus of the arcopallium (RA) in the zebra finch, where population activity reflects the moment-to-moment progression of the courtship song during awake vocalizations19,20,21,22,23,24 and sleep, where song-related network dynamics reemerge in "replay" events.24,25 We found that activity in the left and right RA is synchronized within a fraction of a millisecond throughout song. In stark contrast, the two hemispheres displayed largely independent replay activity during sleep, despite shared interhemispheric arousal levels. These findings demonstrate that the degree of bilateral coordination in the zebra finch song system is dynamically modulated by behavioral state.

Using temperature to analyze the neural basis of a time-based decision

The basal ganglia are thought to contribute to decision-making and motor control. These functions are critically dependent on timing information, which can be extracted from the evolving state of neural populations in their main input structure, the striatum. However, it is debated whether striatal activity underlies latent, dynamic decision processes or kinematics of overt movement. Here, we measured the impact of temperature on striatal population activity and the behavior of rats, and compared the observed effects with neural activity and behavior collected in multiple versions of a temporal categorization task. Cooling caused dilation, and warming contraction, of both neural activity and patterns of judgment in time, mimicking endogenous decision-related variability in striatal activity. However, temperature did not similarly affect movement kinematics. These data provide compelling evidence that the timecourse of evolving striatal activity dictates the speed of a latent process that is used to guide choices, but not continuous motor control. More broadly, they establish temporal scaling of population activity as a likely neural basis for variability in timing behavior.

Direct laser writing of 3D electrodes on flexible substrates

This report describes a 3D microelectrode array integrated on a thin-film flexible cable for neural recording in small animals. The fabrication process combines traditional silicon thin-film processing techniques and direct laser writing of 3D structures at micron resolution via two-photon lithography. Direct laser-writing of 3D-printed electrodes has been described before, but this report is the first to provide a method for producing high-aspect-ratio structures. One prototype, a 16-channel array with 300 µm pitch, demonstrates successful electrophysiological signal capture from bird and mouse brains. Additional devices include 90 µm pitch arrays, biomimetic mosquito needles that penetrate through the dura of birds, and porous electrodes with enhanced surface area. The rapid 3D printing and wafer-scale methods described here will enable efficient device fabrication and new studies examining the relationship between electrode geometry and electrode performance. Applications include small animal models, nerve interfaces, retinal implants, and other devices requiring compact, high-density 3D electrodes.

Optogenetic manipulation of inhibitory interneurons can be used to validate a model of spatiotemporal sequence learning

The brain uses temporal information to link discrete events into memory structures supporting recognition, prediction, and a wide variety of complex behaviors. It is still an open question how experience-dependent synaptic plasticity creates memories including temporal and ordinal information. Various models have been proposed to explain how this could work, but these are often difficult to validate in a living brain. A recent model developed to explain sequence learning in the visual cortex encodes intervals in recurrent excitatory synapses and uses a learned offset between excitation and inhibition to generate precisely timed "messenger" cells that signal the end of an instance of time. This mechanism suggests that the recall of stored temporal intervals should be particularly sensitive to the activity of inhibitory interneurons that can be easily targeted in vivo with standard optogenetic tools. In this work we examined how simulated optogenetic manipulations of inhibitory cells modifies temporal learning and recall based on these mechanisms. We show that disinhibition and excess inhibition during learning or testing cause characteristic errors in recalled timing that could be used to validate the model in vivo using either physiological or behavioral measurements.

Robust encoding of natural stimuli by neuronal response sequences in monkey visual cortex

Parallel multisite recordings in the visual cortex of trained monkeys revealed that the responses of spatially distributed neurons to natural scenes are ordered in sequences. The rank order of these sequences is stimulus-specific and maintained even if the absolute timing of the responses is modified by manipulating stimulus parameters. The stimulus specificity of these sequences was highest when they were evoked by natural stimuli and deteriorated for stimulus versions in which certain statistical regularities were removed. This suggests that the response sequences result from a matching operation between sensory evidence and priors stored in the cortical network. Decoders trained on sequence order performed as well as decoders trained on rate vectors but the former could decode stimulus identity from considerably shorter response intervals than the latter. A simulated recurrent network reproduced similarly structured stimulus-specific response sequences, particularly once it was familiarized with the stimuli through non-supervised Hebbian learning. We propose that recurrent processing transforms signals from stationary visual scenes into sequential responses whose rank order is the result of a Bayesian matching operation. If this temporal code were used by the visual system it would allow for ultrafast processing of visual scenes.

A neural hub for holistic courtship displays

Courtship displays often involve the concerted production of several distinct courtship behaviors. The neural circuits that enable the concerted production of the component behaviors of a courtship display are not well understood. Here, we identify a midbrain cell group (A11) that enables male zebra finches to produce their learned songs in concert with various other behaviors, including female-directed orientation, pursuit, and calling. Anatomical mapping reveals that A11 is at the center of a complex network including the song premotor nucleus HVC as well as brainstem regions crucial to calling and locomotion. Notably, lesioning A11 terminals in HVC blocked female-directed singing but did not interfere with female-directed calling, orientation, or pursuit. In contrast, lesioning A11 cell bodies strongly reduced and often abolished all female-directed courtship behaviors. However, males with either type of lesion still produced songs when in social isolation. Lastly, imaging calcium-related activity in A11 terminals in HVC showed that during courtship, A11 signals HVC about female-directed calls and during female-directed singing, about the transition from simpler introductory notes to the acoustically more complex syllables that depend intimately on HVC for their production. These results show how a brain region important to reproduction in both birds and mammals enables holistic courtship displays in male zebra finches, which include learning songs, calls, and other non-vocal behaviors.

Thalamus drives vocal onsets in the zebra finch courtship song

While motor cortical circuits contain information related to specific movement parameters1, long-range inputs also have a critical role in action execution2,3. Thalamic projections can shape premotor activity2-6 and have been suggested7 to mediate the selection of short, stereotyped actions comprising more complex behaviours8. However, the mechanisms by which thalamus interacts with motor cortical circuits to execute such movement sequences remain unknown. Here we find that thalamic drive engages a specific subpopulation of premotor neurons within the zebra finch song nucleus HVC (proper name) and that these inputs are critical for the progression between vocal motor elements (that is, 'syllables'). In vivo two-photon imaging of thalamic axons in HVC showed robust song-related activity, and online perturbations of thalamic function caused song to be truncated at syllable boundaries. We used thalamic stimulation to identify a sparse set of thalamically driven neurons within HVC, representing ~15% of the premotor neurons within that network. Unexpectedly, this population of putative thalamorecipient neurons is robustly active immediately preceding syllable onset, leading to the possibility that thalamic input can initiate individual song components through selectively targeting these 'starter cells'. Our findings highlight the motor thalamus as a director of cortical dynamics in the context of an ethologically relevant behavioural sequence.

Single spikes drive sequential propagation and routing of activity in a cortical network

Single spikes can trigger repeatable firing sequences in cortical networks. The mechanisms that support reliable propagation of activity from such small events and their functional consequences remain unclear. By constraining a recurrent network model with experimental statistics from turtle cortex, we generate reliable and temporally precise sequences from single spike triggers. We find that rare strong connections support sequence propagation, while dense weak connections modulate propagation reliability. We identify sections of sequences corresponding to divergent branches of strongly connected neurons which can be selectively gated. Applying external inputs to specific neurons in the sparse backbone of strong connections can effectively control propagation and route activity within the network. Finally, we demonstrate that concurrent sequences interact reliably, generating a highly combinatorial space of sequence activations. Our results reveal the impact of individual spikes in cortical circuits, detailing how repeatable sequences of activity can be triggered, sustained, and controlled during cortical computations.

Motor learning drives dynamic patterns of intermittent myelination on learning-activated axons

Myelin plasticity occurs when newly formed and pre-existing oligodendrocytes remodel existing patterns of myelination. Myelin remodeling occurs in response to changes in neuronal activity and is required for learning and memory. However, the link between behavior-induced neuronal activity and circuit-specific changes in myelination remains unclear. Using longitudinal in vivo two-photon imaging and targeted labeling of learning-activated neurons in mice, we explore how the pattern of intermittent myelination is altered on individual cortical axons during learning of a dexterous reach task. We show that behavior-induced myelin plasticity is targeted to learning-activated axons and occurs in a staged response across cortical layers in the mouse primary motor cortex. During learning, myelin sheaths retract, which results in lengthening of nodes of Ranvier. Following motor learning, addition of newly formed myelin sheaths increases the number of continuous stretches of myelination. Computational modeling suggests that motor learning-induced myelin plasticity initially slows and subsequently increases axonal conduction speed. Finally, we show that both the magnitude and timing of nodal and myelin dynamics correlate with improvement of behavioral performance during motor learning. Thus, learning-induced and circuit-specific myelination changes may contribute to information encoding in neural circuits during motor learning.

Excitatory and inhibitory neural dynamics jointly tune motion detection

Neurons integrate excitatory and inhibitory signals to produce their outputs, but the role of input timing in this integration remains poorly understood. Motion detection is a paradigmatic example of this integration, since theories of motion detection rely on different delays in visual signals. These delays allow circuits to compare scenes at different times to calculate the direction and speed of motion. Different motion detection circuits have different velocity sensitivity, but it remains untested how the response dynamics of individual cell types drive this tuning. Here, we sped up or slowed down specific neuron types in Drosophila's motion detection circuit by manipulating ion channel expression. Altering the dynamics of individual neuron types upstream of motion detectors increased their sensitivity to fast or slow visual motion, exposing distinct roles for excitatory and inhibitory dynamics in tuning directional signals, including a role for the amacrine cell CT1. A circuit model constrained by functional data and anatomy qualitatively reproduced the observed tuning changes. Overall, these results reveal how excitatory and inhibitory dynamics together tune a canonical circuit computation.

A feedforward inhibitory premotor circuit for auditory-vocal interactions in zebra finches

Significance During conversations, we frequently alternate between listening and speaking. This involves withholding responses while the other person is vocalizing and rapidly initiating a reply once they stop. Similar exchanges also occur in other animals, such as songbirds, yet little is known about how brain areas responsible for vocal production are influenced by areas dedicated to listening. Here, we combined neural recordings and mathematical modeling of a sensorimotor circuit to show that input-dependent inhibition can both suppress vocal responses and regulate the onset latencies of vocalizations. Our resulting model provides a simple generalizable circuit mechanism by which inhibition precisely times vocal output and integrates auditory input within a premotor nucleus.

Sleep replay reveals premotor circuit structure for a skilled behavior

Neural circuits often exhibit sequences of activity, but the contribution of local networks to their generation remains unclear. In the zebra finch, song-related premotor sequences within HVC may result from some combination of local connectivity and long-range thalamic inputs from nucleus uvaeformis (Uva). Because lesions to either structure abolish song, we examine "sleep replay" using high-density recording methods to reconstruct precise song-related events. Replay activity persists after the upstream nucleus interfacialis of the nidopallium is lesioned and slows when HVC is cooled, demonstrating that HVC provides temporal structure for these events. To further gauge the importance of intra-HVC connectivity for shaping network dynamics, we lesion Uva during sleep and find that residual replay sequences could span syllable boundaries, supporting a model in which HVC can propagate sequences throughout the duration of the song. Our results highlight the power of studying offline activity to investigate behaviorally relevant circuit organization.

Role of NMDAR plasticity in a computational model of synaptic memory

A largely unexplored question in neuronal plasticity is whether synapses are capable of encoding and learning the timing of synaptic inputs. We address this question in a computational model of synaptic input time difference learning (SITDL), where N-methyl-d-aspartate receptor (NMDAR) isoform expression in silent synapses is affected by time differences between glutamate and voltage signals. We suggest that differences between NMDARs' glutamate and voltage gate conductances induce modifications of the synapse's NMDAR isoform population, consequently changing the timing of synaptic response. NMDAR expression at individual synapses can encode the precise time difference between signals. Thus, SITDL enables the learning and reconstruction of signals across multiple synapses of a single neuron. In addition to plausibly predicting the roles of NMDARs in synaptic plasticity, SITDL can be usefully applied in artificial neural network models.

DARPP-32 distinguishes a subset of adult-born neurons in zebra finch HVC

Adult male zebra finches (Taeniopygia guttata) continually incorporate adult-born neurons into HVC, a telencephalic brain region necessary for the production of learned song. These neurons express activity-dependent immediate early genes (e.g., zenk and c-fos) following song production, suggesting that these neurons are active during song production. Half of these adult-born HVC neurons (HVC NNs) can be backfilled from the robust nucleus of the arcopallium (RA) and are a part of the vocal motor pathway underlying learned song production, but the other half do not backfill from RA, and they remain to be characterized. Here, we used cell birth-dating, retrograde tract tracing, and immunofluorescence to demonstrate that half of all HVC NNs express the phosphoprotein DARPP-32, a protein associated with dopamine receptor expression. We also demonstrate that DARPP-32+ HVC NNs are contacted by tyrosine hydroxylase immunoreactive fibers, suggesting that they receive catecholaminergic input, have transiently larger nuclei than DARPP-32-neg HVC NNs, and do not backfill from RA. Taken together, these findings help characterize a group of HVC NNs that have no apparent projections to RA and so far have eluded positive identification other than HVC NN status.

Using focal cooling to link neural dynamics and behavior

Establishing a causal link between neural function and behavioral output has remained a challenging problem. Commonly used perturbation techniques enable unprecedented control over intrinsic activity patterns and can effectively identify crucial circuit elements important for specific behaviors. However, these approaches may severely disrupt activity, precluding an investigation into the behavioral relevance of moment-to-moment neural dynamics within a specified brain region. Here we discuss the application of mild focal cooling to slow down intrinsic neural circuit activity while preserving its overall structure. Using network modeling and examples from multiple species, we highlight the power and versatility of focal cooling for understanding how neural dynamics control behavior and argue for its wider adoption within the systems neuroscience community.

A neuronal circuit that generates the temporal motor sequence for the defensive response in zebrafish larvae

Animals use a precisely timed motor sequence to escape predators. This requires the nervous system to coordinate several motor behaviors and execute them in a temporal and smooth manner. We here describe a neuronal circuit that faithfully generates a defensive motor sequence in zebrafish larvae. The temporally specific defensive motor sequence consists of an initial escape and a subsequent swim behavior and can be initiated by unilateral stimulation of a single Mauthner cell (M-cell). The smooth transition from escape behavior to swim behavior is achieved by activating a neuronal chain circuit, which permits an M-cell to drive descending neurons in bilateral nucleus of medial longitudinal fascicle (nMLF) via activation of an intermediate excitatory circuit formed by interconnected hindbrain cranial relay neurons. The sequential activation of M-cells and neurons in bilateral nMLF via activation of hindbrain cranial relay neurons ensures the smooth execution of escape and swim behaviors in a timely manner. We propose an existence of a serial model that executes a temporal motor sequence involving three different brain regions that initiates the escape behavior and triggers a subsequent swim. This model has general implications regarding the neural control of complex motor sequences.

Neuronal selectivity to complex vocalization features emerges in the superficial layers of primary auditory cortex

Early in auditory processing, neural responses faithfully reflect acoustic input. At higher stages of auditory processing, however, neurons become selective for particular call types, eventually leading to specialized regions of cortex that preferentially process calls at the highest auditory processing stages. We previously proposed that an intermediate step in how nonselective responses are transformed into call-selective responses is the detection of informative call features. But how neural selectivity for informative call features emerges from nonselective inputs, whether feature selectivity gradually emerges over the processing hierarchy, and how stimulus information is represented in nonselective and feature-selective populations remain open question. In this study, using unanesthetized guinea pigs (GPs), a highly vocal and social rodent, as an animal model, we characterized the neural representation of calls in 3 auditory processing stages-the thalamus (ventral medial geniculate body (vMGB)), and thalamorecipient (L4) and superficial layers (L2/3) of primary auditory cortex (A1). We found that neurons in vMGB and A1 L4 did not exhibit call-selective responses and responded throughout the call durations. However, A1 L2/3 neurons showed high call selectivity with about a third of neurons responding to only 1 or 2 call types. These A1 L2/3 neurons only responded to restricted portions of calls suggesting that they were highly selective for call features. Receptive fields of these A1 L2/3 neurons showed complex spectrotemporal structures that could underlie their high call feature selectivity. Information theoretic analysis revealed that in A1 L4, stimulus information was distributed over the population and was spread out over the call durations. In contrast, in A1 L2/3, individual neurons showed brief bursts of high stimulus-specific information and conveyed high levels of information per spike. These data demonstrate that a transformation in the neural representation of calls occurs between A1 L4 and A1 L2/3, leading to the emergence of a feature-based representation of calls in A1 L2/3. Our data thus suggest that observed cortical specializations for call processing emerge in A1 and set the stage for further mechanistic studies.

Addition of new neurons and the emergence of a local neural circuit for precise timing

During development, neurons arrive at local brain areas in an extended period of time, but how they form local neural circuits is unknown. Here we computationally model the emergence of a network for precise timing in the premotor nucleus HVC in songbird. We show that new projection neurons, added to HVC post hatch at early stages of song development, are recruited to the end of a growing feedforward network. High spontaneous activity of the new neurons makes them the prime targets for recruitment in a self-organized process via synaptic plasticity. Once recruited, the new neurons fire readily at precise times, and they become mature. Neurons that are not recruited become silent and replaced by new immature neurons. Our model incorporates realistic HVC features such as interneurons, spatial distributions of neurons, and distributed axonal delays. The model predicts that the birth order of the projection neurons correlates with their burst timing during the song.

Functions of local neural circuits depend on their specific network structures, but how the networks are wired is unknown. We show that such structures can emerge during development through a self-organized process, during which the network is wired by neuron-by-neuron recruitment. This growth is facilitated by steady supply of immature neurons, which are highly excitable and plastic. We suggest that neuron maturation dynamics is an integral part of constructing local neural circuits.

Cellular transcriptomics reveals evolutionary identities of songbird vocal circuits

Birds display advanced behaviors, including vocal learning and problem-solving, yet lack a layered neocortex, a structure associated with complex behavior in mammals. To determine whether these behavioral similarities result from shared or distinct neural circuits, we used single-cell RNA sequencing to characterize the neuronal repertoire of the songbird song motor pathway. Glutamatergic vocal neurons had considerable transcriptional similarity to neocortical projection neurons; however, they displayed regulatory gene expression patterns more closely related to neurons in the ventral pallium. Moreover, while γ-aminobutyric acid-releasing neurons in this pathway appeared homologous to those in mammals and other amniotes, the most abundant avian class is largely absent in the neocortex. These data suggest that songbird vocal circuits and the mammalian neocortex have distinct developmental origins yet contain transcriptionally similar neurons.

Neural Coding: Axonal Delays Make Waves

In songbirds, premotor activity is locked to song with millisecond precision. New results suggest that delayed signal propagation along local axons is critical for smooth sequencing. A cross-species analysis hints that such delays may be important in mammals too.