Role of hyperpolarization-activated conductances in the lateral superior olive: A modeling study

This modeling study examines the possible functional roles of two hyperpolarization-activated conductances in lateral superior olive (LSO) principal neurons. Inputs of these LSO neurons are transformed into an output, which provides a firing-rate code for a certain interaural sound intensity difference (IID) range. Recent experimental studies have found pharmacological evidence for the presence of both the Gh conductance as well as the inwardly rectifying outward GKIR conductance in the LSO. We addressed the question of how these conductances influence the dynamic range (IID versus firing rate). We used computer simulations of both a point-neuron model and a two-compartmental model to investigate this issue, and to determine the role of these conductances in setting the dynamic range of these neurons. The width of the dynamic regime, the frequency-current (f-I) function, first-spike latency, subthreshold oscillations and the interplay between the two hyperpolarization activated conductances are discussed in detail. The in vivo non-monotonic IID-firing rate function in a subpopulation of LSO neurons is in good correspondence with our simulation predictions. Two compartmental model simulation results suggest segregation of Gh and GKIR conductances on different compartments, as this spatial configuration could explain certain experimental results.     

Robustness of neuronal tuning to binaural sound localization cues against age-related loss of inhibitory synaptic inputs

Sound localization relies on minute differences in the timing and intensity of sound arriving at both ears. Neurons of the lateral superior olive (LSO) in the brainstem process these interaural disparities by precisely detecting excitatory and inhibitory synaptic inputs. Aging generally induces selective loss of inhibitory synaptic transmission along the entire auditory pathways, including the reduction of inhibitory afferents to LSO. Electrophysiological recordings in animals, however, reported only minor functional changes in aged LSO. The perplexing discrepancy between anatomical and physiological observations suggests a role for activity-dependent plasticity that would help neurons retain their binaural tuning function despite loss of inhibitory inputs. To explore this hypothesis, we use a computational model of LSO to investigate mechanisms underlying the observed functional robustness against age-related loss of inhibitory inputs. The LSO model is an integrate-and-fire type enhanced with a small amount of low-voltage activated potassium conductance and driven with (in)homogeneous Poissonian inputs. Without synaptic input loss, model spike rates varied smoothly with interaural time and level differences, replicating empirical tuning properties of LSO. By reducing the number of inhibitory afferents to mimic age-related loss of inhibition, overall spike rates increased, which negatively impacted binaural tuning performance, measured as modulation depth and neuronal discriminability. To simulate a recovery process compensating for the loss of inhibitory fibers, the strength of remaining inhibitory inputs was increased. By this modification, effects of inhibition loss on binaural tuning were considerably weakened, leading to an improvement of functional performance. These neuron-level observations were further confirmed by population modeling, in which binaural tuning properties of multiple LSO neurons were varied according to empirical measurements. These results demonstrate the plausibility that homeostatic plasticity could effectively counteract known age-dependent loss of inhibitory fibers in LSO and suggest that behavioral degradation of sound localization might originate from changes occurring more centrally.     

Neural Computation via Neural Geometry: A Place Code for Inter-whisker Timing in the Barrel Cortex?

The place theory proposed by Jeffress (1948) is still the dominant model of how the brain represents the movement of sensory stimuli between sensory receptors. According to the place theory, delays in signalling between neurons, dependent on the distances between them, compensate for time differences in the stimulation of sensory receptors. Hence the location of neurons, activated by the coincident arrival of multiple signals, reports the stimulus movement velocity. Despite its generality, most evidence for the place theory has been provided by studies of the auditory system of auditory specialists like the barn owl, but in the study of mammalian auditory systems the evidence is inconclusive. We ask to what extent the somatosensory systems of tactile specialists like rats and mice use distance dependent delays between neurons to compute the motion of tactile stimuli between the facial whiskers (or 'vibrissae'). We present a model in which synaptic inputs evoked by whisker deflections arrive at neurons in layer 2/3 (L2/3) somatosensory 'barrel' cortex at different times. The timing of synaptic inputs to each neuron depends on its location relative to sources of input in layer 4 (L4) that represent stimulation of each whisker. Constrained by the geometry and timing of projections from L4 to L2/3, the model can account for a range of experimentally measured responses to two-whisker stimuli. Consistent with that data, responses of model neurons located between the barrels to paired stimulation of two whiskers are greater than the sum of the responses to either whisker input alone. The model predicts that for neurons located closer to either barrel these supralinear responses are tuned for longer inter-whisker stimulation intervals, yielding a topographic map for the inter-whisker deflection interval across the surface of L2/3. This map constitutes a neural place code for the relative timing of sensory stimuli.     

Functional refinement in the projection from ventral cochlear nucleus to lateral superior olive precedes hearing onset in rat

Principal neurons of the lateral superior olive (LSO) compute the interaural intensity differences necessary for localizing high-frequency sounds. To perform this computation, the LSO requires precisely tuned, converging excitatory and inhibitory inputs that are driven by the two ears and that are matched for stimulus frequency. In rodents, the inhibitory inputs, which arise from the medial nucleus of the trapezoid body (MNTB), undergo extensive functional refinement during the first postnatal week. Similar functional refinement of the ascending excitatory pathway, which arises in the anteroventral cochlear nucleus (AVCN), has been assumed but has not been well studied. Using whole-cell voltage clamp in acute brainstem slices of neonatal rats, we examined developmental changes in input strength and pre- and post-synaptic properties of the VCN-LSO pathway. A key question was whether functional refinement in one of the two major input pathways might precede and then guide refinement in the opposite pathway. We find that elimination and strengthening of VCN inputs to the LSO occurs over a similar period to that seen for the ascending inhibitory (MNTB-LSO) pathway. During this period, the fractional contribution provided by NMDA receptors (NMDARs) declines while the contribution from AMPA receptors (AMPARs) increases. In the NMDAR-mediated response, GluN2B-containing NMDARs predominate in the first postnatal week and decline sharply thereafter. Finally, the progressive decrease in paired-pulse depression between birth and hearing onset allows these synapses to follow progressively higher frequencies. Our data are consistent with a model in which the excitatory and inhibitory projections to LSO are functionally refined in parallel during the first postnatal week, and they further suggest that GluN2B-containing NMDARs may mediate early refinement in the VCN-LSO pathway.     

Encoding of spatio-temporal input characteristics by a CA1 pyramidal neuron model

The in vivo activity of CA1 pyramidal neurons alternates between regular spiking and bursting, but how these changes affect information processing remains unclear. Using a detailed CA1 pyramidal neuron model, we investigate how timing and spatial arrangement variations in synaptic inputs to the distal and proximal dendritic layers influence the information content of model responses. We find that the temporal delay between activation of the two layers acts as a switch between excitability modes: short delays induce bursting while long delays decrease firing. For long delays, the average firing frequency of the model response discriminates spatially clustered from diffused inputs to the distal dendritic tree. For short delays, the onset latency and inter-spike-interval succession of model responses can accurately classify input signals as temporally close or distant and spatially clustered or diffused across different stimulation protocols. These findings suggest that a CA1 pyramidal neuron may be capable of encoding and transmitting presynaptic spatiotemporal information about the activity of the entorhinal cortex-hippocampal network to higher brain regions via the selective use of either a temporal or a rate code.     

Gamma oscillations as a mechanism for selective information transmission

In the past decades, many studies have focussed on the relation between the input and output of neurons with the aim to understand information processing by neurons. A particular aspect of neuronal information, which has not received much attention so far, concerns the problem of information transfer when a neuron or a population of neurons receives input from two or more (populations of) neurons, in particular when these (populations of) neurons carry different types of information. The aim of the present study is to investigate the responses of neurons to multiple inputs modulated in the gamma frequency range. By a combination of theoretical approaches and computer simulations, we test the hypothesis that enhanced modulation of synchronized excitatory neuronal activity in the gamma frequency range provides an advantage over a less synchronized input for various types of neurons. The results of this study show that the spike output of various types of neurons [i.e. the leaky integrate and fire neuron, the quadratic integrate and fire neuron and the Hodgkin–Huxley (HH) neuron] and that of excitatory–inhibitory coupled pairs of neurons, like the Pyramidal Interneuronal Network Gamma (PING) model, is highly phase-locked to the larger of two gamma-modulated input signals. This implies that the neuron selectively responds to the input with the larger gamma modulation if the amplitude of the gamma modulation exceeds that of the other signals by a certain amount. In that case, the output of the neuron is entrained by one of multiple inputs and that other inputs are not represented in the output. This mechanism for selective information transmission is enhanced for short membrane time constants of the neuron.     

Glycinergic transmission regulates dendrite size in organotypic culture

We previously demonstrated that inhibitory synaptic transmission influences dendrite development in vivo. We now report an analogous finding in an organotypic culture of a glycinergic projection nucleus, the medial nucleus of the trapezoid body (MNTB), and its postsynaptic target, the lateral superior olive (LSO) of gerbils. Cultures were generated at 6–7 days postnatal and grown in serum containing medium with or without the glycine receptor antagonist, strychnine (SN), at 2 μM. LSO neurons were then labeled with biocytin, and the dendritic arbors were analyzed morphometrically. Compared to neurons from age-matched in vivo tissue, the neurons cultured in control media were somewhat atrophic, including decreases in dendritic branching and length. Incubation in strychnine led to a dramatic increase in dendritic branching and total dendritic length. Control neurons averaged 6.3 branches, compared to 18 branches/neuron in SN-treated cultures. There was a similar increase in primary dendrites and total dendritic length. The physical elimination of MNTB cells did not mimic SN treatment, presumably because glycinergic LSO neurons generated intrinsic connections. In fact, the LSO soma area was significantly greater following MNTB removal, suggesting that these afferents provide a second signal to postsynaptic neurons. These results suggest that spontaneous glycinergic transmission regulates the growth of postsynaptic processes. © 1996 John Wiley & Sons, Inc.     

Anatomical Organization of Multiple Modulatory Inputs in a Rhythmic Motor System

In rhythmic motor systems, descending projection neuron inputs elicit distinct outputs from their target central pattern generator (CPG) circuits. Projection neuron activity is regulated by sensory inputs and inputs from other regions of the nervous system, relaying information about the current status of an organism. To gain insight into the organization of multiple inputs targeting a projection neuron, we used the identified neuron MCN1 in the stomatogastric nervous system of the crab, Cancer borealis. MCN1 originates in the commissural ganglion and projects to the stomatogastric ganglion (STG). MCN1 activity is differentially regulated by multiple inputs including neuroendocrine (POC) and proprioceptive (GPR) neurons, to elicit distinct outputs from CPG circuits in the STG. We asked whether these defined inputs are compact and spatially segregated or dispersed and overlapping relative to their target projection neuron. Immunocytochemical labeling, intracellular dye injection and three-dimensional (3D) confocal microscopy revealed overlap of MCN1 neurites and POC and GPR terminals. The POC neuron terminals form a defined neuroendocrine organ (anterior commissural organ: ACO) that utilizes peptidergic paracrine signaling to act on MCN1. The MCN1 arborization consistently coincided with the ACO structure, despite morphological variation between preparations. Contrary to a previous 2D study, our 3D analysis revealed that GPR axons did not terminate in a compact bundle, but arborized more extensively near MCN1, arguing against sparse connectivity of GPR onto MCN1. Consistent innervation patterns suggest that integration of the sensory GPR and peptidergic POC inputs occur through more distributed and more tightly constrained anatomical interactions with their common modulatory projection neuron target than anticipated.     

Target-dependent morphological segregation of Aplysia sensory outgrowth in vitro.

The adult nervous system is characterized by partial or complete morphological segregation of terminals from different afferent neurons innervating the same postsynaptic target. This segregation is thought to result, in part, from competition between the afferent terminals. To explore the role of the target cell in the spatial distribution of presynaptic inputs, the sensory neurons of Aplysia were cultured either with or without a common target motor neuron. In the presence of a common target, the outgrowth from two different sensory neurons tends to occupy separate postsynaptic regions. When cultured without a target motor neuron, processes from different sensory neurons do not segregate, but rather grow freely along one another. Thus, morphological segregation of sensory outgrowth requires interaction with a target neuron and may reflect competition between presynaptic terminals for a limited number of synaptic sites on the motor neuron, or for a postsynaptic trophic factor.     

Emergence of network structure due to spike-timing-dependent plasticity in recurrent neuronal networks. I. Input selectivity–strengthening correlated input pathways

Spike-timing-dependent plasticity (STDP) determines the evolution of the synaptic weights according to their pre- and post-synaptic activity, which in turn changes the neuronal activity. In this paper, we extend previous studies of input selectivity induced by (STDP) for single neurons to the biologically interesting case of a neuronal network with fixed recurrent connections and plastic connections from external pools of input neurons. We use a theoretical framework based on the Poisson neuron model to analytically describe the network dynamics (firing rates and spike-time correlations) and thus the evolution of the synaptic weights. This framework incorporates the time course of the post-synaptic potentials and synaptic delays. Our analysis focuses on the asymptotic states of a network stimulated by two homogeneous pools of “steady” inputs, namely Poisson spike trains which have fixed firing rates and spike-time correlations. The (STDP) model extends rate-based learning in that it can implement, at the same time, both a stabilization of the individual neuron firing rates and a slower weight specialization depending on the input spike-time correlations. When one input pathway has stronger within-pool correlations, the resulting synaptic dynamics induced by (STDP) are shown to be similar to those arising in the case of a purely feed-forward network: the weights from the more correlated inputs are potentiated at the expense of the remaining input connections.     

Spike propagation synchronized by temporally asymmetric Hebbian learning

 Synchronously spiking neurons have been observed in the cerebral cortex and the hippocampus. In computer models, synchronous spike volleys may be propagated across appropriately connected neuron populations. However, it is unclear how the appropriate synaptic connectivity is set up during development and maintained during adult learning. We performed computer simulations to investigate the influence of temporally asymmetric Hebbian synaptic plasticity on the propagation of spike volleys. In addition to feedforward connections, recurrent connections were included between and within neuron populations and spike transmission delays varied due to axonal, synaptic and dendritic transmission. We found that repeated presentations of input volleys decreased the synaptic conductances of intragroup and feedback connections while synaptic conductances of feedforward connections with short delays became stronger than those of connections with longer delays. These adaptations led to the synchronization of spike volleys as they propagated across neuron populations. The findings suggests that temporally asymmetric Hebbian learning may enhance synchronized spiking within small populations of neurons in cortical and hippocampal areas and familiar stimuli may produce synchronized spike volleys that are rapidly propagated across neural tissue. Received: 28 May 2002 / Accepted: 3 June 2002 RID="*" ID="*" Correspondence to: R. E. Suri Intelligent Optical Systems (IOS), 2520 W 237th St, Torrance, CA 90505-5217, USA (e-mail: rsuri@intopsys.com, Tel.: +1-310-5307130 ext. 108, Fax: +1-210-5307417)     

Exclusive-OR function of single arborized neuron

We designed four arborized neurons which are able to evaluate the exclusive-or (XOR) function from two inputs. The input neurons form exclusively excitatory synapses on a dendritic tree which is a patchwork of passive (ohmic) and active cable segments. The active segments are described by the Hodgkin-Huxley model. The dynamics of the neurons and their output are obtained by numerical integration of the cable equation. In neurons 1 and 2 the XOR function is based on the annihilation of colliding action potentials. In neuron No. 3 the design takes advantage of the refractory period of action potentials. In neuron No. 4 voltage inversion is used as it occurs for inactivated sodium conductance in the Hodgkin-Huxley model. In all cases the XOR function depends critically on an appropriate timing of the input signals and on delays of the voltage transients in different branches of the dendrite.     

Differential expression pattern of chloride transporters<Emphasis Type="Italic"> NCC,NKCC2, KCC1, KCC3, KCC4,</Emphasis> and<Emphasis Type="Italic"> AE3</Emphasis> in the developing rat auditory brainstem

During development of inhibitory synapses, the action of the two neurotransmitters GABA and glycine shifts from depolarizing to hyperpolarizing. The shift is due to an age-dependent regulation of the intracellular free chloride concentration ([Cl(-)](i)) in postsynaptic neurons. A model system to study this maturation process is a glycinergic projection in the mammalian auditory brainstem. It is formed in the superior olivary complex (SOC) by neurons of the medial nucleus of the trapezoid body, whose axons terminate in the lateral superior olive (LSO). LSO neurons of perinatal rats and mice are depolarized upon glycine application, whereas older cells (>postnatal day (P) 8) are hyperpolarized. Here we examined the expression of six secondary active chloride transporter genes ( NCC, NKCC2, KCC1, KCC3, KCC4, and AE3) in the rat SOC to unravel the molecular mechanisms underlying this change. RT-PCR analysis demonstrated brainstem expression of KCC1, KCC3, KCC4, and AE3, but not of NCC and NKCC2. RNA in situ hybridization showed that only AE3 is highly expressed both at P3 (high [Cl(-)](i)) and P12 (low [Cl(-)](i)) in LSO neurons. KCC1 and KCC4 are weakly expressed in LSO neurons at P3 and P12, respectively. This study completes the expression analysis of all known chloride transporters sensitive to loop diuretic drugs in the SOC and demonstrates differences in the maturation between hippocampal and brainstem inhibitory synapses.     

GABA-mediated modulation of the discharge pattern and rate-level function of two simultaneously recorded neurons in the inferior colliculus of the big brown bat, Eptesicus fuscus

Neurons in the central nucleus of the inferior colliculus (IC) receive excitatory and inhibitory inputs from both lower and higher auditory nuclei. Interaction of these two opposing inputs shapes response properties of IC neurons. In this study, we examine the interaction of excitation and inhibition on the responses of two simultaneously recorded IC neurons using a probe and a masker under forward masking paradigm. We specifically study whether a sound that serves as a probe to elicit responses of one neuron might serve as a masker to suppress or facilitate the responses of the other neuron. For each pair of IC neurons, we deliver the probe at the best frequency (BF) of one neuron and the masker at the BF of the other neuron and vice versa. Among 33 pairs of IC neurons recorded, this forward masking produces response suppression in 29 pairs of IC neurons and response facilitation in 4 pairs of IC neurons. The degree of suppression decreases with recording depth, sound level and BF difference between each pair of IC neurons. During bicuculline application, the degree of response suppression decreases in the bicuculline-applied neuron but increases in the paired neuron. Our data indicate that the forward masking of responses of IC neurons observed in this study is mostly mediated through GABAergic inhibition which also shapes the discharge pattern of these neurons. These data suggest that interaction among individual IC neurons improves auditory sensitivity during auditory signal processing.     

Gain adjustment of inhibitory synapses in the auditory system

A group of central auditory neurons residing in the lateral superior olivary nucleus (LSO) responds selectively to interaural level differences and may contribute to sound localization. In this simple circuit, ipsilateral sound increases firing of LSO neurons, whereas contralateral sound inhibits the firing rate via activation of the medial nucleus of the trapezoid body (MNTB). During development, individual MNTB fibers arborize within the LSO, but they undergo a restriction of their boutons that ultimately leads to mature topography. A critical issue is whether a distinct form of inhibitory synaptic plasticity contributes to MNTB synapse elimination within LSO. Whole-cell recording from LSO neurons in brain slices from developing gerbils show robust long-term depression (LTD) of the MNTB-evoked IPSP/Cs when the MNTB was activated at a low frequency (1 Hz). These inhibitory synapses also display mixed GABA/glycinergic transmission during development, as assessed physiologically and immunohistochemically (Kotak et al. 1998). While either glycine or GABA_A receptors could independently display inhibitory LTD, focal delivery of GABA, but not glycine, at the postsynaptic-locus induces depression. Furthermore, the GABA_B receptor antagonist, SCH-50911, prevents GABA or synaptically induced depression. Preliminary evidence also indicated strengthening of inhibitory transmission (LTP) by a distinct pattern of inhibitory activity. These data support the idea that GABA is crucial for the expression inhibitory LTD and that this plasticity may underlie the early refinement of inhibitory synaptic connections in the LSO.     

The Possible Role of Spike Patterns in Cortical Information Processing

When the same visual stimulus is presented across many trials, neurons in the visual cortex receive stimulus-related synaptic inputs that are reproducible across trials (S) and inputs that are not (N). The variability of spike trains recorded in the visual cortex and their apparent lack of spike-to-spike correlations beyond that implied by firing rate fluctuations, has been taken as evidence for a low S/N ratio. A recent re-analysis of in vivo cortical data revealed evidence for spike-to-spike correlations in the form of spike patterns. We examine neural dynamics at a higher S/N in order to determine what possible role spike patterns could play in cortical information processing. In vivo-like spike patterns were obtained in model simulations. Superpositions of multiple sinusoidal driving currents were especially effective in producing stable long-lasting patterns. By applying current pulses that were either short and strong or long and weak, neurons could be made to switch from one pattern to another. Cortical neurons with similar stimulus preferences are located near each other, have similar biophysical properties and receive a large number of common synaptic inputs. Hence, recordings of a single neuron across multiple trials are usually interpreted as the response of an ensemble of these neurons during one trial. In the presence of distinct spike patterns across trials there is ambiguity in what would be the corresponding ensemble, it could consist of the same spike pattern for each neuron or a set of patterns across neurons. We found that the spiking response of a neuron receiving these ensemble inputs was determined by the spike-pattern composition, which, in turn, could be modulated dynamically as a means for cortical information processing.     

猫体感皮层神经元对同时刺激隐神经的A类和C类纤维的反应型式

记录了麻痹猫的体感皮层(SI)神经元的自发和隐神经的A类和C类纤维传入诱发放电(A-ED和C-ED)。用NCCVF分析神经元放电。结果表明,SI区神经元对同时刺激隐神经的A类和C类纤维的反应呈多种型式:(1)A-ED和C-ED共存,包括Ⅰ.A-ED和C-ED始终相互伴随出现;Ⅱ.在刺激之初,只出现A-ED,但是,当阻断A类纤维传入并由C类纤维传入诱发神经元放电后,再同时刺激A类和C类纤维时,A-ED和C-ED便同时出现。(2)A-ED制约C-ED,特点是,只要A-ED存在,C/ED就不出现。只有阻断A类纤维传入后,C-ED才产生。(3)单一A-ED,不管在什么刺激条件下,这类神经元都只有A-ED,而不产生C-ED 结论:根据反应型式的不同,可将SI区的神经元分为Ⅰ.A类和C类纤维传入同时驱动的神经元;Ⅱ.A-ED制约C-ED的神经元;Ⅲ.只由A类纤维传入驱动的神经元。     

Oscillations and oscillatory behavior in small neural circuits

In order to determine the dynamical properties of central pattern generators (CPGs), we have examined the lobster stomatogastric ganglion using the tools of nonlinear dynamics. The lobster pyloric and gastric mill central pattern generators can be analyzed at both the cellular and network levels because they are small, i.e., contain only 25 neurons between them and each neuron and synapse are repeatedly identifiable from animal to animal. We discuss how the biophysical properties of each neuron and synapse in the two circuits act cooperatively to generate two different patterns of sequential activity, how these patterns are altered by neuromodulators and perturbed by noise and sensory inputs. Finally, we show how simplified Hindmarsh–Rose models can be made into analog electronic neurons that mimic the lobster neurons and in addition be incorporated into artificial CPGs with robotic applications.     

Correlation of unit activity of the right and left amygdala in food motivation and emotional stress

Correlation between activities of neurons in the right and left central nuclei of amygdala of rabbits recorded during quiet wakefulness, after 24-h food deprivation, after satiation and during emotional stress (demonstration of a dog) was studied by plotting crosscorrelation histograms. The histogram peaks shifted from zero were observed in 50-67% cases. In hungry animals, in a greater number of cases (52%), the discharge of a neuron in the left amygdala was the first in a pair, and the discharge of the right neuron was delayed (peaks from 10 to 50 and from 130 to 150 ms). The opposite order of discharges was less frequent (36%). When a rabbit saw a dog, the number of common inputs to neurons increased and the leading role of the right amygdalar neurons grew (57%) due to an increase in inhibitory influences from the right to the left amygdala. In most cases, the interaction between amygdalar neurons occurred at the frequencies of the delta range, predominantly, from 2 to 4 Hz.     

Emergence of network structure due to spike-timing-dependent plasticity in recurrent neuronal networks. II. Input selectivity—symmetry breaking

Spike-timing-dependent plasticity (STDP) is believed to structure neuronal networks by slowly changing the strengths (or weights) of the synaptic connections between neurons depending upon their spiking activity, which in turn modifies the neuronal firing dynamics. In this paper, we investigate the change in synaptic weights induced by STDP in a recurrently connected network in which the input weights are plastic but the recurrent weights are fixed. The inputs are divided into two pools with identical constant firing rates and equal within-pool spike-time correlations, but with no between-pool correlations. Our analysis uses the Poisson neuron model in order to predict the evolution of the input synaptic weights and focuses on the asymptotic weight distribution that emerges due to STDP. The learning dynamics induces a symmetry breaking for the individual neurons, namely for sufficiently strong within-pool spike-time correlation each neuron specializes to one of the input pools. We show that the presence of fixed excitatory recurrent connections between neurons induces a group symmetry-breaking effect, in which neurons tend to specialize to the same input pool. Consequently STDP generates a functional structure on the input connections of the network.