Physiology of electrosensory lateral line lobe neurons in Gnathonemus petersii

Continual Learning in a Multi-Layer Network of an Electric Fish

Distributing learning across multiple layers has proven extremely powerful in artificial neural networks. However, little is known about how multi-layer learning is implemented in the brain. Here, we provide an account of learning across multiple processing layers in the electrosensory lobe (ELL) of mormyrid fish and report how it solves problems well known from machine learning. Because the ELL operates and learns continuously, it must reconcile learning and signaling functions without switching its mode of operation. We show that this is accomplished through a functional compartmentalization within intermediate layer neurons in which inputs driving learning differentially affect dendritic and axonal spikes. We also find that connectivity based on learning rather than sensory response selectivity assures that plasticity at synapses onto intermediate-layer neurons is matched to the requirements of output neurons. The mechanisms we uncover have relevance to learning in the cerebellum, hippocampus, and cerebral cortex, as well as in artificial systems.

Internally Generated Predictions Enhance Neural and Behavioral Detection of Sensory Stimuli in an Electric Fish

Studies of cerebellum-like circuits in fish have demonstrated that synaptic plasticity shapes the motor corollary discharge responses of granule cells into highly-specific predictions of self-generated sensory input. However, the functional significance of such predictions, known as negative images, has not been directly tested. Here we provide evidence for improvements in neural coding and behavioral detection of prey-like stimuli due to negative images. In addition, we find that manipulating synaptic plasticity leads to specific changes in circuit output that disrupt neural coding and detection of prey-like stimuli. These results link synaptic plasticity, neural coding, and behavior and also provide a circuit-level account of how combining external sensory input with internally generated predictions enhances sensory processing.

A quest for excitation: Theoretical arguments and immunohistochemical evidence of excitatory granular cells in the ELL of Gnathonemus petersii

The Electrosensory Lateral Line lobe (ELL) is the first central target where the electrosensory information encoded in the spatiotemporal pattern electroreceptor afferent discharges is processed. These afferents encode the minute amplitude changes of the basal electric field through both a change in latency and discharge rate. In the ELL the time and rate-coded input pattern of the sensory periphery goes through the granular cell layer before reaching the main efferent cells of the network: large fusiform (LF) and large ganglion (LG) cells. The evidence until now shows that granular cells are inhibitory. Given that large fusiform cells are excited by the sensory input, it remains a mystery how the afferent input produce excitation through a layer composed by only inhibitory cells. We addressed this problem by modeling how the known circuitry of the ELL could produce excitation in LF cells with only inhibitory granular cells. Alternatively we show that a network composed of a mix of excitatory and inhibitory granular cell not only performs better, as expected, carrying excitation to LF cells but it does so robustly and at higher sensitivity by enhancing the contrast of the electric image between the periphery and the ELLs output. We then show with refined histological methods that a subpopulation of the granular cells indeed are excitatory, providing the necessary input for this contrast enhancing mechanism.

Identifying self- and nonself-generated signals: lessons from electrosensory systems

This chapter provides a short review of the mechanisms used by electroreceptive fish to discriminate self- from nonself-generated signals. Electroreception is used by animals to detect objects of electric impedance different from the water, to detect natural electrogenic sources and to communicate signals between conspecifics. Electroreceptive animals may generate electric fields either with the purpose of electrically illuminating the neighborhood or as an epiphenomenon of other functions. In addition, the presence of the fish body as a conductive object in a scene funnels the current flow and, consequently, animal movements also generate signals by changing the body shape or the spatial relationship of the body with the surrounding objects. Therefore, mechanisms for discrimination between self and externally generated signals are very important for constructing a coherent representation of the environment. Some mechanisms facilitate and stream the flow of signals carried by the self-generated electric field. Others are designed to reject unwanted interference coming from self-generated movements or even the self-generated electric field. Finally, more complex operations involving sensory motor integration are used for discriminating between self- and conspecific- generated communication signals. Despite the evolutionary distance between animals endowed with electric sense, mechanisms for self-identification reappear with few differences between species. This suggests that many of the possible strategies are present in vertebrates may be found in these fish. Therefore, we have much to learn about self recognition from the study of electroreception.

Dendritic backpropagation and synaptic plasticity in the mormyrid electrosensory lobe

This study is concerned with the origin of backpropagating action potentials in GABAergic, medium ganglionic layer neurones (MG-cells) of the mormyrid electrosensory lobe (ELL). The characteristically broad action potentials of these neurones are required for the expression of spike timing dependent plasticity (STDP) at afferent parallel fibre synapses. It has been suggested that this involves active conductances in MG-cell apical dendrites, which constitute a major component of the ELL molecular layer. Immunohistochemistry showed dense labelling of voltage gated sodium channels (VGSC) throughout the molecular layer, as well as in the ganglionic layer containing MG somata, and in the plexiform and upper granule cell layers of ELL. Potassium channel labelling was sparse, being most abundant in the deep fibre layer and the nucleus of the electrosensory lobe. Intracellular recordings from MG-cells in vitro, made in conjunction with voltage sensitive dye measurements, confirmed that dendritic backpropagation is active over at least the inner half of the molecular layer. Focal TTX applications demonstrated that in most case the origin of the backpropagating action potentials is in the proximal dendrites, whereas the small narrow spikes also seen in these neurones most likely originate in the axon. It had been speculated that the slow time course of membrane repolarisation following the broad action potentials was due to a poor expression of potassium channels in the dendritic compartments, or to their voltage- or calcium-sensitive inactivation. However application of TEA and 4AP confirmed that both A-type and delayed rectifying potassium channels normally contribute to membrane repolarisation following dendritic and axonal spikes. An alternative explanation for the shape of MG action potentials is that they represent the summation of active events occurring more or less synchronously in distal dendrites. Coincidence of backpropagating action potentials with parallel fibre input produces a strong local depolarisation that could be sufficient to cause local secretion of GABA, which might then cause plastic change through an action on presynaptic GABA(B) receptors. However, STP depression remained robust in the presence of GABAB receptor antagonists.

Dendritic spike back propagation in the electrosensory lobe of Gnathonemus petersii

Spike timing-dependent plasticity that follows anti-Hebbian rules has been demonstrated at synapses between parallel fibers and inhibitory interneurons known as medium ganglionic layer (MG) neurons in the cerebellum-like electrosensory lobe of mormyrid fish. This plasticity is expressed when presynaptic activation is associated with a characteristically broad, postsynaptic action potential, lasting 7-15 ms, occurring within a window of up to 60-80 ms following synaptic activation. Since the site of plastic change is presumably in the apical dendrites, it is important to know where, when and how this broad spike is generated and the manner in which such events propagate within the intrinsic network of the electrosensory lobe. The electrosensory lobe has a strict layered organization that makes the preparation suitable for one dimension current source density analysis. Using this technique in an 'in vitro' interface slice preparation, we found that following either parallel fiber stimulation or an orthogonal field stimulus, a sink appeared in the ganglionic layer and propagated into the molecular layer. Intracellular records from MG somata showed these stimuli evoked broad action potentials whose timing corresponds to this sink. TTX application in the deep fiber layer blocked the synaptically evoked ganglionic layer field potential and the 'N3' wave of the outer molecular layer field potential simultaneously, while the molecular layer 'N1' and 'N2' waves corresponding to synaptic activation of the apical dendrites remained intact. These results confirm the hypothesis that the broad spikes of MG cells originate in the soma and propagate through the molecular layer in the apical dendritic tree, and suggest the possibility that this backpropagation may contribute to 'boosting' of the synaptic response in distal apical dendrites in certain circumstances.

Dye coupling without gap junctions suggests excitatory connections of gamma-aminobutyric acidergic neurons

After injections of the low-molecular-weight tracer neurobiotin into the preeminential nucleus of the brain of the mormyrid fish Gnathonemus petersii, we observed that retrogradely labeled, large fusiform projection neurons (LFd cells) of the deep granular layer of the electrosensory lobe (ELL) were surrounded by 30-50 labeled satellite granular cells. More superficially located projection cells, including large fusiform cells in the superficial granular layer (LFs) and large ganglionic (LG) cells in the ganglionic layer, were never surrounded by labeled satellites. LFd-satellite cells have a small soma (diameter 5-8 microm), a few small dendrites, and an apical axon that terminates in the plexiform and ganglionic layers of the ELL. They contact LFd projection neurons with dendrodendritic, dendrosomatic, and somatodendritic puncta adhaerentia-like appositions, designated here as "neurapses." In the electron microscope, these contacts resemble synapses without presynaptic vesicles. Because no gap junctions were found between LFd and satellite granule cells, we suggest that the neurapses allow the passage of neurobiotin, though not biocytin or biotinylated dextran amine. These contacts may provide the intermediate substrate for the postulated, but so far unknown, excitatory connection between primary afferent input and LFd projection neurons, via gamma-aminobutyric acid (GABA)-ergic granular cells. We suggest that certain puncta adhaerentia-like contacts might not be only adhesive structures and that LFd-satellite granular cells might both excite LFd projection cells via neuraptic contacts of their dendrites and cell bodies and inhibit more superficial LF and LG cells via their GABAergic axonal synapses. Our results suggest that puncta adhaerentia-like contacts could be responsible in some cases for the electrical coupling found electrophysiologically in local inhibitory circuits.

Oscillatory and burst discharge in the apteronotid electrosensory lateral line lobe

Oscillatory and burst discharge is recognized as a key element of signal processing from the level of receptor to cortical output cells in most sensory systems. The relevance of this activity for electrosensory processing has become increasingly apparent for cells in the electrosensory lateral line lobe (ELL) of gymnotiform weakly electric fish. Burst discharge by ELL pyramidal cells can be recorded in vivo and has been directly associated with feature extraction of electrosensory input. In vivo recordings have also shown that pyramidal cells are differentially tuned to the frequency of amplitude modulations across three ELL topographic maps of electroreceptor distribution. Pyramidal cell recordings in vitro reveal two forms of oscillatory discharge with properties consistent with pyramidal cell frequency tuning in vivo. One is a slow oscillation of spike discharge arising from local circuit interactions that exhibits marked changes in several properties across the sensory maps. The second is a fast, intrinsic form of burst discharge that incorporates a newly recognized interaction between somatic and dendritic membranes. These findings suggest that a differential regulation of oscillatory discharge properties across sensory maps may underlie frequency tuning in the ELL and influence feature extraction in vivo.