A "sample-and-hold" pulse-counting integrator as a mechanism for graded memory underlying sensorimotor adaptation

Spatiotemporal discrimination in attractor networks with short-term synaptic plasticity

We demonstrate that a randomly connected attractor network with dynamic synapses can discriminate between similar sequences containing multiple stimuli suggesting such networks provide a general basis for neural computations in the brain. The network contains units representing assemblies of pools of neurons, with preferentially strong recurrent excitatory connections rendering each unit bi-stable. Weak interactions between units leads to a multiplicity of attractor states, within which information can persist beyond stimulus offset. When a new stimulus arrives, the prior state of the network impacts the encoding of the incoming information, with short-term synaptic depression ensuring an itinerancy between sets of active units. We assess the ability of such a network to encode the identity of sequences of stimuli, so as to provide a template for sequence recall, or decisions based on accumulation of evidence. Across a range of parameters, such networks produce the primacy (better final encoding of the earliest stimuli) and recency (better final encoding of the latest stimuli) observed in human recall data and can retain the information needed to make a binary choice based on total number of presentations of a specific stimulus. Similarities and differences in the final states of the network produced by different sequences lead to predictions of specific errors that could arise when an animal or human subject generalizes from training data, when the training data comprises a subset of the entire stimulus repertoire. We suggest that such networks can provide the general purpose computational engines needed for us to solve many cognitive tasks.

The central nervous system transcriptome of the weakly electric brown ghost knifefish (Apteronotus leptorhynchus): de novo assembly, annotation, and proteomics validation

Background

The brown ghost knifefish (Apteronotus leptorhynchus) is a weakly electric teleost fish of particular interest as a versatile model system for a variety of research areas in neuroscience and biology. The comprehensive information available on the neurophysiology and neuroanatomy of this organism has enabled significant advances in such areas as the study of the neural basis of behavior, the development of adult-born neurons in the central nervous system and their involvement in the regeneration of nervous tissue, as well as brain aging and senescence. Despite substantial scientific interest in this species, no genomic resources are currently available.

Results

Here, we report the de novo assembly and annotation of the A. leptorhynchus transcriptome. After evaluating several trimming and transcript reconstruction strategies, de novo assembly using Trinity uncovered 42,459 unique contigs containing at least a partial protein-coding sequence based on alignment to a reference set of known Actinopterygii sequences. As many as 11,847 of these contigs contained full or near-full length protein sequences, providing broad coverage of the proteome. A variety of non-coding RNA sequences were also identified and annotated, including conserved long intergenic non-coding RNA and other long non-coding RNA observed previously to be expressed in adult zebrafish (Danio rerio) brain, as well as a variety of miRNA, snRNA, and snoRNA. Shotgun proteomics confirmed translation of open reading frames from over 2,000 transcripts, including alternative splice variants. Assignment of tandem mass spectra was greatly improved by use of the assembly compared to databases of sequences from closely related organisms. The assembly and raw reads have been deposited at DDBJ/EMBL/GenBank under the accession number GBKR00000000. Tandem mass spectrometry data is available via ProteomeXchange with identifier PXD001285.

Conclusions

Presented here is the first release of an annotated de novo transcriptome assembly from Apteronotus leptorhynchus, providing a broad overview of RNA expressed in central nervous system tissue. The assembly, which includes substantial coverage of a wide variety of both protein coding and non-coding transcripts, will allow the development of better tools to understand the mechanisms underlying unique characteristics of the knifefish model system, such as their tremendous regenerative capacity and negligible brain senescence.

Electronic supplementary material

The online version of this article (doi:10.1186/s12864-015-1354-2) contains supplementary material, which is available to authorized users.

A theory of rate coding control by intrinsic plasticity effects

Intrinsic plasticity (IP) is a ubiquitous activity-dependent process regulating neuronal excitability and a cellular correlate of behavioral learning and neuronal homeostasis. Because IP is induced rapidly and maintained long-term, it likely represents a major determinant of adaptive collective neuronal dynamics. However, assessing the exact impact of IP has remained elusive. Indeed, it is extremely difficult disentangling the complex non-linear interaction between IP effects, by which conductance changes alter neuronal activity, and IP rules, whereby activity modifies conductance via signaling pathways. Moreover, the two major IP effects on firing rate, threshold and gain modulation, remain unknown in their very mechanisms. Here, using extensive simulations and sensitivity analysis of Hodgkin-Huxley models, we show that threshold and gain modulation are accounted for by maximal conductance plasticity of conductance that situate in two separate domains of the parameter space corresponding to sub- and supra-threshold conductance (i.e. activating below or above the spike onset threshold potential). Analyzing equivalent integrate-and-fire models, we provide formal expressions of sensitivities relating to conductance parameters, unraveling unprecedented mechanisms governing IP effects. Our results generalize to the IP of other conductance parameters and allow strong inference for calcium-gated conductance, yielding a general picture that accounts for a large repertoire of experimental observations. The expressions we provide can be combined with IP rules in rate or spiking models, offering a general framework to systematically assess the computational consequences of IP of pharmacologically identified conductance with both fine grain description and mathematical tractability. We provide an example of such IP loop model addressing the important issue of the homeostatic regulation of spontaneous discharge. Because we do not formulate any assumptions on modification rules, the present theory is also relevant to other neural processes involving excitability changes, such as neuromodulation, development, aging and neural disorders.

Over the past decades, experimental and theoretical studies of the cellular basis of learning and memory have mainly focused on synaptic plasticity, the experience-dependent modification of synapses. However, behavioral learning has also been correlated with experience-dependent changes of non-synaptic voltage-dependent ion channels. This intrinsic plasticity changes the neuron's propensity to fire action potentials in response to synaptic inputs. Thus a fundamental problem is to relate changes of the neuron input-output function with voltage-gated conductance modifications. Using a sensitivity analysis in biophysically realistic models, we depict a generic dichotomy between two classes of voltage-dependent ion channels. These two classes modify the threshold and the slope of the neuron input-output relation, allowing neurons to regulate the range of inputs they respond to and the gain of that response, respectively. We further provide analytical descriptions that enlighten the dynamical mechanisms underlying these effects and propose a concise and realistic framework for assessing the computational impact of intrinsic plasticity in neuron network models. Our results account for a large repertoire of empirical observations and may enlighten functional changes that characterize development, aging and several neural diseases, which also involve changes in voltage-dependent ion channels.

Spatial gradients and multidimensional dynamics in a neural integrator circuit

In a neural integrator, the variability and topographical organization of neuronal firing rate persistence can provide information about the circuit’s functional architecture. Here we use optical recording to measure the time constant of decay of persistent firing (“persistence time”) across a population of neurons comprising the larval zebrafish oculomotor velocity-to-position neural integrator. We find extensive persistence time variation (10-fold; coefficients of variation 0.58–1.20) across cells within individual larvae. We also find that the similarity in firing between two neurons decreased as the distance between them increased and that a gradient in persistence time was mapped along the rostrocaudal and dorsoventral axes. This topography is consistent with the emergence of persistence time heterogeneity from a circuit architecture in which nearby neurons are more strongly interconnected than distant ones. Collectively, our results can be accounted for by integrator circuit models characterized by multiple dimensions of slow firing rate dynamics.

Low-frequency summation of synaptically activated transient receptor potential channel-mediated depolarizations

Neurons sum their input by spatial and temporal integration. Temporally, presynaptic firing rates are converted to dendritic membrane depolarizations by postsynaptic receptors and ion channels. In several regions of the brain, including higher association areas, the majority of firing rates are low. For rates below 20 Hz, the ionotropic receptors α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor and N-methyl-d-aspartate (NMDA) receptor will not produce effective temporal summation. We hypothesized that depolarization mediated by transient receptor potential (TRP) channels activated by metabotropic glutamate receptors would be more effective, owing to their slow kinetics. On the basis of voltage-clamp and current-clamp recordings from a rat slice preparation, we constructed a computational model of the TRP channel and its intracellular activation pathway, including the metabotropic glutamate receptor. We show that synaptic input frequencies down to 3-4 Hz and inputs consisting of as few as three to five pulses can be effectively summed. We further show that the time constant of integration increases with increasing stimulation frequency and duration. We suggest that the temporal summation characteristics of TRP channels may be important at distal dendritic arbors, where spatial summation is limited by the number of concurrently active synapses. It may be particularly important in regions characterized by low and irregular rates.

Calcium-dependent phosphorylation regulates neuronal stability and plasticity in a highly precise pacemaker nucleus

Specific types of neurons show stable, predictable excitability properties, while other neurons show transient adaptive plasticity of their excitability. However, little attention has been paid to how the cellular pathways underlying adaptive plasticity interact with those that maintain neuronal stability. We addressed this question in the pacemaker neurons from a weakly electric fish because these neurons show a highly stable spontaneous firing rate as well as an N-methyl-D-aspartate (NMDA) receptor-dependent form of plasticity. We found that basal firing rates were regulated by a serial interaction of conventional and atypical PKC isoforms and that this interaction establishes individual differences within the species. We observed that NMDA receptor-dependent plasticity is achieved by further activation of these kinases. Importantly, the PKC pathway is maintained in an unsaturated baseline state to allow further Ca(2+)-dependent activation during plasticity. On the other hand, the Ca(2+)/calmodulin-dependent phosphatase calcineurin does not regulate baseline firing but is recruited to control the duration of the NMDA receptor-dependent plasticity and return the pacemaker firing rate back to baseline. This work illustrates how neuronal plasticity can be realized by biasing ongoing mechanisms of stability (e.g., PKC) and terminated by recruiting alternative mechanisms (e.g., calcineurin) that constrain excitability. We propose this as a general model for regulating activity-dependent change in neuronal excitability.

Timescale-Invariant Pattern Recognition by Feedforward Inhibition and Parallel Signal Processing

The timescale-invariant recognition of temporal stimulus sequences is vital for many species and poses a challenge for their sensory systems. Here we present a simple mechanistic model to address this computational task, based on recent observations in insects that use rhythmic acoustic communication signals for mate finding. In the model framework, feedforward inhibition leads to burst-like response patterns in one neuron of the circuit. Integrating these responses over a fixed time window by a readout neuron creates a timescale-invariant stimulus representation. Only two additional processing channels, each with a feature detector and a readout neuron, plus one final coincidence detector for all three parallel signal streams, are needed to account for the behavioral data. In contrast to previous solutions to the general time-warp problem, no time delay lines or sophisticated neural architectures are required. Our results suggest a new computational role for feedforward inhibition and underscore the power of parallel signal processing.

Calcium dynamics encode the magnitude of a graded memory underlying sensorimotor adaptation

The role of Ca(2+) in the induction of neural correlates of memory has frequently been described in binary terms despite the fact that many forms of memory are graded in their strength and/or persistence. We find that Ca(2+) dynamics encode the magnitude of sensorimotor adaptation of the electromotor output in a weakly electric fish. The neural correlate of this memory is a synaptically induced Ca(2+)-dependent enhancement of intrinsic excitability of neurons responsible for setting the electromotor output. Changes in Ca(2+) during induction accurately predict the magnitude of this graded memory over a wide range of stimuli. Thus despite operating over a range from seconds to tens of minutes, the encoding of graded memory can be mediated by a relatively simple cellular mechanism.

Evolution of time-coding systems in weakly electric fishes

Weakly electric fishes emit electric organ discharges (EODs) from their tail electric organs and sense feedback signals from their EODs by electroreceptors in the skin. The electric sense is utilized for various behaviors, including electrolocation, electrocommunication, and the Jamming avoidance response (JAR). For each behavior, various types of sensory Information are embedded in the transient electrical signals produced by the fish. These temporal signals are sampled, encoded, and further processed by peripheral and central neurons specialized for time coding. There are time codes for the sex or species Identities of other fish or the resistance and capacitance of objects. In the central nervous system, specialized neural elements exist for decoding time codes for different behavioral functions. Comparative studies allow phylogenetic comparison of time-coding neural systems among weakly electric fishes.

Spike integration and cellular memory in a rhythmic network from Na+/K+ pump current dynamics

The output of a neural circuit results from an interaction between the intrinsic properties of neurons within the circuit and the features of the synaptic connections between them. The plasticity of intrinsic properties has been primarily attributed to modification of ion channel function and/or number. In this study, we demonstrate a mechanism for intrinsic plasticity in rhythmically active Drosophila neurons that is not conductance-based. Larval motor neurons show a long lasting sodium-dependent afterhyperpolarization (AHP) following bursts of action potentials that is mediated by the electrogenic activity of Na⁺/K⁺ ATPase. This AHP persists for multiple seconds following volleys of action potentials and is able to function as a pattern-insensitive integrator of spike number that is independent of external calcium. This current also interacts with endogenous Shal K⁺ conductances to modulate spike timing for multiple seconds following rhythmic activity, providing a cellular memory of network activity on a behaviorally relevant time scale.

Timescale-invariant representation of acoustic communication signals by a bursting neuron

Acoustic communication often involves complex sound motifs in which the relative durations of individual elements, but not their absolute durations, convey meaning. Decoding such signals requires an explicit or implicit calculation of the ratios between time intervals. Using grasshopper communication as a model, we demonstrate how this seemingly difficult computation can be solved in real time by a small set of auditory neurons. One of these cells, an ascending interneuron, generates bursts of action potentials in response to the rhythmic syllable-pause structure of grasshopper calls. Our data show that these bursts are preferentially triggered at syllable onset; the number of spikes within the burst is linearly correlated with the duration of the preceding pause. Integrating the number of spikes over a fixed time window therefore leads to a total spike count that reflects the characteristic syllable-to-pause ratio of the species while being invariant to playing back the call faster or slower. Such a timescale-invariant recognition is essential under natural conditions, because grasshoppers do not thermoregulate; the call of a sender sitting in the shade will be slower than that of a grasshopper in the sun. Our results show that timescale-invariant stimulus recognition can be implemented at the single-cell level without directly calculating the ratio between pulse and interpulse durations.

In vitro studies of closed-loop feedback and electrosensory processing in Apteronotus leptorhynchus

Electrosensory systems comprise extensive feedback pathways. It is also well known that these pathways exhibit synaptic plasticity on a wide-range of time scales. Recent in vitro brain slice studies have characterized synaptic plasticity in the two main feedback pathways to the electrosensory lateral line lobe (ELL), a primary electrosensory nucleus in Apteronotus leptorhynchus. Currently-used slice preparations, involving networks in open-loop conditions, allow feedback inputs to be studied in isolation, a critical step in determining their synaptic properties. However, to fully understand electrosensory processing, we must understand how dynamic feedback modulates neuronal responses under closed-loop conditions. To bridge the gap between current in vitro approaches and more complex in vivo work, we present two new in vitro approaches for studying the roles of closed-loop feedback in electrosensory processing. The first involves a hybrid-network approach using dynamic clamp, and the second involves a new slice preparation that preserves one of the feedback pathways to ELL in a closed-loop condition.

mGluR-dependent persistent firing in entorhinal cortex layer III neurons

Persistent firing is believed to be a crucial mechanism for memory function including working memory. Recent in vivo and in vitro findings suggest an involvement of metabotropic glutamate receptors (mGluRs) in persistent firing. Using whole-cell patch-recording techniques in a rat entorhinal cortex (EC) slice preparation, we tested whether EC layer III neurons display persistent firing due to mGluR activation, independently of cholinergic activation. Stimulation of the angular bundle drove persistent firing in 90% of the cells in the absence of a cholinergic agonist. The persistent firing was typically stable for > 4.5 min at which point persistent firing was terminated by the experimenter. The average frequency of the persistent firing was 2.1 Hz, ranging from 0.4 to 5.5 Hz. This persistent firing was observed even in the presence of atropine (2 microM), suggesting that the persistent firing can occur independent of cholinergic activation. Furthermore, ionotropic glutamate and GABAergic synaptic blockers (2 mM kynurenic acid, 100 microM picrotoxin and 1 microM CGP55845) did not block the persistent firing. On the other hand, blockers of group I mGluRs (100 microM LY367385 and 20 microM MPEP) completely blocked or suppressed the persistent firing. An agonist of group I mGluRs (20 microM DHPG) greatly enhanced the persistent firing induced by current injection. These results indicate that persistent firing can be driven through group I mGluRs in entorhinal layer III neurons, suggesting that glutamatergic synaptic input alone could enable postsynaptic neurons to hold input signals in the form of persistent firing.

Amplitude changes in response to target displacements during human eye-head movements

Sensorimotor adaptation, the ability to adjust motor output in response to persistent changes in sensory input, is a key function of the central nervous system. Although a great deal is known about vestibulo-ocular reflex and saccadic adaptation, relatively little is known about the behavior and neural mechanisms underlying gaze adaptation when the head is free to move. In an attempt to understand the mechanisms of gaze adaptation, and constrain hypotheses concerning the locus at which changes in gaze control may be implemented, we altered the size of large, head-unrestrained gaze shifts made to visual targets by surrepetitiously moving the visual target forward (30 degrees -->60 degrees ) or backwards (60 degrees -->30 degrees ) during gaze shifts. In our 10 human subjects, after a few hundred back-step trials, gaze amplitudes were reduced by between 6 degrees and 27 degrees. Similarly, after a few hundred forward adaptation trials, our subjects increased gaze amplitude by between 0 degrees and 26 degrees. Changes in the amplitude of primary gaze shifts occurred regardless of the particular combinations of eye and head movements that made up the amplitude-altered gaze shifts. When gaze shifts were initiated with the eyes in systematically different positions relative to the head, the resulting changes in gaze, eye and head movement amplitudes were consistent with the hypothesis that gaze adaptation occurs at the level of a gaze shift command and not by altering separately the signals that produce eye and head movements.

Long-term potentiation in rat hippocampal neurons is accompanied by spatially widespread changes in intrinsic oscillatory dynamics and excitability

Oscillations in neural activity are a prominent feature of many brain states. Individual hippocampal neurons exhibit intrinsic membrane potential oscillations and intrinsic resonance in the theta frequency range. We found that the subthreshold resonance frequency of CA1 pyramidal neurons was location dependent, varying more than 3-fold between the soma and the distal dendrites. Furthermore, activity- and NMDA-receptor-dependent long-term plasticity increased this resonance frequency through changes in h channel properties. The increase in resonance frequency and an associated reduction in excitability were nearly identical in the soma and the first 300 mum of the apical dendrites. These spatially widespread changes accompanying long-term synaptic potentiation also reduced the neuron's ability to elicit spikes evoked through a nonpotentiated synaptic pathway. Our results suggest that the frequency response of these neurons depends on the dendritic location of their inputs and that activity can regulate their response dynamics within an oscillating neural network.

A new principle for information storage in an enzymatic pathway model

Strong experimental evidence indicates that protein kinase and phosphatase (KP) cycles are critical to both the induction and maintenance of activity-dependent modifications in neurons. However, their contribution to information storage remains controversial, despite impressive modeling efforts. For instance, plasticity models based on KP cycles do not account for the maintenance of plastic modifications. Moreover, bistable KP cycle models that display memory fail to capture essential features of information storage: rapid onset, bidirectional control, graded amplitude, and finite lifetimes. Here, we show in a biophysical model that upstream activation of KP cycles, a ubiquitous mechanism, is sufficient to provide information storage with realistic induction and maintenance properties: plastic modifications are rapid, bidirectional, and graded, with finite lifetimes that are compatible with animal and human memory. The maintenance of plastic modifications relies on negligible reaction rates in basal conditions and thus depends on enzyme nonlinearity and activation properties of the activity-dependent KP cycle. Moreover, we show that information coding and memory maintenance are robust to stochastic fluctuations inherent to the molecular nature of activity-dependent KP cycle operation. This model provides a new principle for information storage where plasticity and memory emerge from a single dynamic process whose rate is controlled by neuronal activity. This principle strongly departs from the long-standing view that memory reflects stable steady states in biological systems, and offers a new perspective on memory in animals and humans.

A Synchronization-Desynchronization Code for Natural Communication Signals

Synchronous spiking of neural populations is hypothesized to play important computational roles in forming neural assemblies and solving the binding problem. Although the opposite phenomenon of desynchronization is well known from EEG studies, it is largely neglected on the neuronal level. We here provide an example of in vivo recordings from weakly electric fish demonstrating that, depending on the social context, different types of natural communication signals elicit transient desynchronization as well as synchronization of the electroreceptor population without changing the mean firing rate. We conclude that, in general, both positive and negative changes in the degree of synchrony can be the relevant signals for neural information processing.