Unreliable Synapses Research Articles

Fundamental bounds on learning performance in neural circuits

How does the size of a neural circuit influence its learning performance? Larger brains tend to be found in species with higher cognitive function and learning ability. Intuitively, we expect the learning capacity of a neural circuit to grow with the number of neurons and synapses. We show how adding apparently redundant neurons and connections to a network can make a task more learnable. Consequently, large neural circuits can either devote connectivity to generating complex behaviors or exploit this connectivity to achieve faster and more precise learning of simpler behaviors. However, we show that in a biologically relevant setting where synapses introduce an unavoidable amount of noise, there is an optimal size of network for a given task. Above the optimal network size, the addition of neurons and synaptic connections starts to impede learning performance. This suggests that the size of brain circuits may be constrained by the need to learn efficiently with unreliable synapses and provides a hypothesis for why some neurological learning deficits are associated with hyperconnectivity. Our analysis is independent of specific learning rules and uncovers fundamental relationships between learning rate, task performance, network size, and intrinsic noise in neural circuits.

Long-term depression triggers the selective elimination of weakly integrated synapses

Long-term depression (LTD) weakens synaptic transmission in an activity-dependent manner. It is not clear, however, whether individual synapses are able to maintain a depressed state indefinitely, as intracellular recordings rarely exceed 1 h. Here, we combine optogenetic stimulation of identified Schaffer collateral axons with two-photon imaging of postsynaptic calcium signals and follow the fate of individual synapses for 7 d after LTD induction. Optogenetic stimulation of CA3 pyramidal cells at 1 Hz led to strong and reliable depression of postsynaptic calcium transients in CA1. NMDA receptor activation was necessary for successful induction of LTD. We found that, in the days following LTD, many depressed synapses and their "neighbors" were eliminated from the hippocampal circuit. The average lifetime of synapses on nonstimulated dendritic branches of the same neurons remained unaffected. Persistence of individual depressed synapses was highly correlated with reliability of synaptic transmission, but not with spine size or the amplitude of spine calcium transients. Our data suggest that LTD initially leads to homogeneous depression of synaptic function, followed by selective removal of unreliable synapses and recovery of function in the persistent fraction.

Inverse stochastic resonance induced by synaptic background activity with unreliable synapses

Inverse stochastic resonance (ISR) is a recently pronounced phenomenon that is the minimum occurrence in mean firing rate of a rhythmically firing neuron as noise level varies. Here, by using a realistic modeling approach for the noise, we investigate the ISR with concrete biophysical mechanisms. It is shown that mean firing rate of a single neuron subjected to synaptic bombardment exhibits a minimum as the spike transmission probability varies. We also demonstrate that the occurrence of ISR strongly depends on the synaptic input regime, where it is most prominent in the balanced state of excitatory and inhibitory inputs.

Role of synaptic dynamics and heterogeneity in neuronal learning of temporal code

Temporal codes are believed to play important roles in neuronal representation of information. Neuronal ability to classify and learn temporal spiking patterns is thus essential for successful extraction and processing of information. Understanding neuronal learning of temporal code has been complicated, however, by the intrinsic stochasticity of synaptic transmission. Using a computational model of a learning neuron, the tempotron, we studied the effects of synaptic unreliability and short-term dynamics on the neuron's ability to learn spike timing rules. Our results suggest that such a model neuron can learn to classify spike timing patterns even with unreliable synapses, albeit with a significantly reduced success rate. We explored strategies to improve correct spike timing classification and found that firing clustered spike bursts significantly improves learning performance. Furthermore, rapid activity-dependent modulation of synaptic unreliability, implemented with realistic models of dynamic synapses, further improved classification of different burst properties and spike timing modalities. Neuronal models with only facilitating or only depressing inputs exhibited preference for specific types of spike timing rules, but a mixture of facilitating and depressing synapses permitted much improved learning of multiple rules. We tested applicability of these findings to real neurons by considering neuronal learning models with the naturally distributed input release probabilities found in excitatory hippocampal synapses. Our results suggest that spike bursts comprise several encoding modalities that can be learned effectively with stochastic dynamic synapses, and that distributed release probabilities significantly improve learning performance. Synaptic unreliability and dynamics may thus play important roles in the neuron's ability to learn spike timing rules during decoding.

Vibrational resonance in feedforward neuronal network with unreliable synapses

In this paper, we investigate vibrational resonance in feedforward neuronal network coupled in an all-to-all fashion. In contrast to earlier most work, where only reliable synaptic connections are considered, we mainly examine the effects of unreliable synapses on signal propagation in this work. It is shown that the neurotransmitter release probability and excitatory synaptic strength largely influence the signal propagation, and better tuning of these synaptic parameters makes the feedforward neuronal network support stable signal propagation. Furthermore, it is found that high-frequency driving plays important roles in causing the response of the feedforward neuronal network to subthreshold low-frequency signal and strengthening the ability of the signal propagation between layers. In particular, the optimal amplitude of high-frequency driving is largely influenced by the stochastic effect of neurotransmitter release and the coupling strength. Finally, we compare our results with those obtained in corresponding feedforward neuronal networks connected with reliable synapses but in a random coupling fashion. It is demonstrated that unreliable synaptic coupling is more efficient than the random coupling for the transmission of local input signal. Considering that unreliable synapses are inevitable in neuronal communication, the presented results could have important implications for the weak signal detection and information propagation in neural systems.

Stochastic resonance in Hodgkin–Huxley neuron induced by unreliable synaptic transmission

We systematically investigate the stochastic dynamics of a single Hodgkin–Huxley neuron driven by stochastic excitatory and inhibitory input spikes via unreliable synapses in this paper. Based on the mean-filed theory, a novel intrinsic neuronal noise regulation mechanism stemming from unreliable synapses is presented. Our simulation results show that, under certain conditions, the stochastic resonance phenomenon is able to be induced by the unreliable synaptic transmission, which can be well explained by the theoretical prediction. To a certain degree, the results presented here provide insights into the functional roles of unreliable synapses in neural information processing.

Population rate coding in recurrent neuronal networks with unreliable synapses

Neuron transmits spikes to postsynaptic neurons through synapses. Experimental observations indicated that the communication between neurons is unreliable. However most modelling and computational studies considered deterministic synaptic interaction model. In this paper, we investigate the population rate coding in an all-to-all coupled recurrent neuronal network consisting of both excitatory and inhibitory neurons connected with unreliable synapses. We use a stochastic on-off process to model the unreliable synaptic transmission. We find that synapses with suitable successful transmission probability can enhance the encoding performance in the case of weak noise; while in the case of strong noise, the synaptic interactions reduce the encoding performance. We also show that several important synaptic parameters, such as the excitatory synaptic strength, the relative strength of inhibitory and excitatory synapses, as well as the synaptic time constant, have significant effects on the performance of the population rate coding. Further simulations indicate that the encoding dynamics of our considered network cannot be simply determined by the average amount of received neurotransmitter for each neuron in a time instant. Moreover, we compare our results with those obtained in the corresponding random neuronal networks. Our numerical results demonstrate that the network randomness has the similar qualitative effect as the synaptic unreliability but not completely equivalent in quantity.

Synchronization of the small-world neuronal network with unreliable synapses

As is well known, synchronization phenomena are ubiquitous in neuronal systems. Recently a lot of work concerning the synchronization of the neuronal network has been accomplished. In these works, the synapses are usually considered reliable, but experimental results show that, in biological neuronal networks, synapses are usually unreliable. In our previous work, we have studied the synchronization of the neuronal network with unreliable synapses; however, we have not paid attention to the effect of topology on the synchronization of the neuronal network. Several recent studies have found that biological neuronal networks have typical properties of small-world networks, characterized by a short path length and high clustering coefficient. In this work, mainly based on the small-world neuronal network (SWNN) with inhibitory neurons, we study the effect of network topology on the synchronization of the neuronal network with unreliable synapses. Together with the network topology, the effects of the GABAergic reversal potential, time delay and noise are also considered. Interestingly, we found a counter-intuitive phenomenon for the SWNN with specific shortcut adding probability, that is, the less reliable the synapses, the better the synchronization performance of the SWNN. We also consider the effects of both local noise and global noise in this work. It is shown that these two different types of noise have distinct effects on the synchronization: one is negative and the other is positive.

Emergence of reliable spike patterns in models of CA1 cells contacted by unreliable synapses

Excitatory synapses onto CA1 pyramidal cells fail four times out of five on average, yet the firing of CA1 neuron is elicited at specific phases of the EEG theta cycle with a high degree of precision when a rat is traversing a place field. We use a multicompartmental biophysical model of several reconstructed CA1 cells, and a model of a stochastic glutamatergic synapse that includes facilitation and depression to study the conditions and properties for reliable and precise CA1 firing. The model of the synapse is tightly constrained by experimental data obtained with minimal stimulations in vitro. Synapses are presynaptically stimulated with CA3/entorhinal spike trains that have been recorded in vivo in the behaving rat. We report that under those conditions, CA1 pyramidal cells are capable of generating precise spike patterns that are theta-modulated. The precise timing of the pattern depends mainly on either the recruitment of high initial probability synapses, or on the recruitment of weaker perisomatic synapses receiving fast bursts of 1–3 presynaptic action potentials. The patterns generated are robust to noise, and contain a marked theta-frequency component, even though the input spikes are not coherent at any particular frequency. We also report that spike patterns may include gamma-like frequency components in part due to the synaptic dynamics, and to the presence of fast bursts in the presynaptic inputs. We conclude that even though afferent synapses are unreliable, CA1 pyramidal cells are able to generate precisely timed patterns of spiking that mimic those that are reported in vivo. Our model further predicts that about 400 presynaptic cells are involved in the firing of a single CA1 pyramidal cell when the rat traverses a place field.

Multifractal distribution of spike intervals for two oscillators coupled by unreliable pulses

Two neurons coupled by unreliable synapses are modelled by leaky integrate-and-fire neurons and stochastic on–off synapses. The dynamics is mapped to an iterated function system. Numerical calculations yield a multifractal distribution of interspike intervals. The covering, information and correlation dimensions are calculated as a function of synaptic strength and transmission probability.

Beyond spike timing: the role of nonlinear plasticity and unreliable synapses.

Spike-timing-dependent plasticity (STDP) strengthens synapses that are activated immediately before a postsynaptic spike, and weakens those that are activated after a spike. To prevent an uncontrolled growth of the synaptic strengths, weakening must dominate strengthening for uncorrelated spike times. However, this weight-normalization property would preclude Hebbian potentiation when the pre- and postsynaptic neurons are strongly active without specific spike-time correlations. We show that nonlinear STDP as inherent in the data of Markram et al. [(1997) Science 275:213-215] can preserve the benefits of both weight normalization and Hebbian plasticity, and hence can account for learning based on spike-time correlations and on mean firing rates. As examples we consider the moving-threshold property of the Bienenstock-Cooper-Munro rule, the development of direction-selective simple cells by changing short-term synaptic depression, and the joint adaptation of axonal and dendritic delays. Without threshold nonlinearity at low frequencies, the development of direction selectivity does not stabilize in a natural stimulation environment. Without synaptic unreliability there is no causal development of axonal and dendritic delays.

Resonance and selective communication via bursts in neurons having subthreshold oscillations

Revealing the role of bursts of action potentials is an important step toward understanding how the neurons communicate. The dominant point of view is that bursts are needed to increase the reliability of communication between neurons [Trends Neurosci. 20 (1997) 38]. In this paper we present an alternative but complementary hypothesis. We consider the effect of a short burst on a model postsynaptic cell having damped oscillation of its membrane potential. The oscillation frequency (eigenfrequency) plays a crucial role. Due to the subthreshold membrane resonance and frequency preference, the responses (i.e. voltage oscillations) of such a cell are amplified when the intra-burst frequency equals the cell's eigenfrequency. Responses are negligible, however, if the intra-burst frequency is twice the eigenfrequency. Thus, the same burst could be effective for one cell and ineffective for another depending on their eigenfrequencies. This theoretical observation suggests that, in addition to coping with unreliable synapses, bursts of action potentials may provide effective mechanisms for selective communication between neurons.

A model for fast analog computation based on unreliable synapses.

We investigate through theoretical analysis and computer simulations the consequences of unreliable synapses for fast analog computations in networks of spiking neurons, with analog variables encoded by the current firing activities of pools of spiking neurons. Our results suggest a possible functional role for the well-established unreliability of synaptic transmission on the network level. We also investigate computations on time series and Hebbian learning in this context of space-rate coding in networks of spiking neurons with unreliable synapses.

Fast burst firing and short-term synaptic plasticity: A model of neocortical chattering neurons

We present an ionic conductance model of chattering neurons in the neocortex, which fire fast rhythmic bursts in the gamma frequency range (∼40 Hz) in response to stimulation [Gray C. M. and McCormick D. A. (1996) Science 274, 109–113]. The bursting mechanism involves a “ping-pong” interplay between soma-to-dendrite back propagation of action potentials and an afterdepolarization generated by a persistent dendritic Na + current and a somatic Na + window current. The oscillation period is primarily determined by a slowly inactivating K + channel and passive membrane properties. The model behavior is compared quantitatively with the experimental data. It is shown that the cholinergic muscarinic receptor activation can transform the model cell's firing pattern from tonic spiking to rapid bursting, as a possible pathway for acetylcholine to promote 40-Hz oscillations in the visual cortex. To explore possible functions of fast burst firing in the neocortex, a hypothetical neural pair is simulated, where a chattering cell is presynaptic to an inhibitory interneuron via stochastic synapses. For this purpose, we use a synapse model endowed with a low release probability, short-term facilitation and vesicle depletion. This synapse model reproduces the behavior of certain neocortical pyramid-to-interneuron synapses [Thomson A. M. et al. (1993) Neuroscience 54, 347–360]. We showed that the burstiness of cell firing is required for the rhythmicity to be reliably transmitted to the postsynaptic cell via unreliable synapses, and that fast burst firing of chattering neurons can provide an exceptionally powerful drive for recruiting feedback inhibition in cortical circuits. From these results, we propose that the fast rhythmic burst firing of neocortical chattering neurons is generated by a calcium-independent ionic mechanism. Our simulation results on the neural pair highlight the importance of characterizing the short-term plasticity of the synaptic connections made by chattering cells, in order to understand their putative pacemaker role in synchronized gamma oscillations of the visual cortex.