Synaptic interactions in neocortical local circuits: dual intracellular recordings in vitro

Coherent oscillations and short-term plasticity in corticothalamic networks

The neocortex and thalamus are a unified oscillatory machine. Different types of brain rhythms, which characterize various behavioral states, are combined within complex wave-sequences. During the stage of sleep that is associated with low-frequency and high-amplitude brain rhythms, the excitatory component of a cortically generated slow oscillation is effective in triggering thalamically generated rhythms and in increasing their spatiotemporal coherence over widespread territories. Thus, the study of coherent oscillations, as they appear naturally during states of vigilance in animals and humans, requires intact-brain preparations in which the neocortex and thalamus engage in a permanent dialog. Sleep oscillations are associated with rhythmic spike-bursts or spike-trains in thalamic and cortical neurons, which lead to persistent excitability changes consisting of increased depolarizing responses and decreased inhibitory responses. These short-term plasticity processes could be used to consolidate memory traces acquired during wakefulness, but can also lead to paroxysmal (hypersynchronous) episodes, similar to those observed in some epileptic seizures.

The effect of etomidate on sensory transmission in the dorsal column pathway in the urethane-anaesthetized rat

The effect of Etomidate, a general anaesthetic, on sensory afferent transmission was measured in the dorsal column pathway in urethane-anaesthetized rats. Extracellular recordings were made of peripherally evoked responses by single cells in the cuneate nucleus, ventroposterolateral nucleus of the thalamus and laminae IV-VI of the primary somatosensory cortex. Cortical mass responses were also recorded. In further experiments, cortical mass responses were evoked antidromically by stimulation in the pyramidal tract. The effect of incremental administration of Etomidate on evoked responses was recorded. These results are compared with the previously reported effects of urethane, a 'conventional' anaesthetic. Etomidate did not alter cuneate or ventroposterolateral thalamic cell responses but it caused a dose-dependent reduction in cortical cell responsiveness. It failed to alter antidromically evoked cortical mass responses. Etomidate differs from the majority of anaesthetics, which act in the thalamus, and appears to cause perturbation at the cortical level.

Intrinsic electrophysiology of neurons in thalamorecipient layers of developing rat auditory cortex

During early postnatal life, several critical events contribute to the functional development of rat sensory neocortex. Thalamocortical innervation of sensory cortex is completed during the first postnatal week and extrathalamic innervation develops over the first several weeks. In auditory cortex, acoustic-evoked potentials first occur in week 2 and develop most rapidly over weeks 2-3. Thus, rapid functional maturation of cortical circuits in sensory cortex occurs during the second and third postnatal weeks. The electrophysiological properties of cortical neurons that receive afferent inputs during this time may play an important role in development and function. In this study we examined the intrinsic electrophysiology, including spiking patterns, of neurons in layers II/III and IV of auditory cortex during postnatal weeks 2 and 3. Many neurons displayed characteristics consistent with previous descriptions of response classes (regular spiking, fast spiking, intrinsic bursting). In addition, we identified two groups, Rectifying and On-spiking neurons, that were characterized by (i) brief spike trains in response to maintained intracellular depolarizations, and (ii) striking outward rectification upon depolarization. Unusually brief spike trains (1-2 spikes) and short spike latencies (<10 ms) further distinguished On-spiking from Rectifying cells. Biocytin labeling demonstrated that On-spiking and Rectifying cells could be either pyramidal or nonpyramidal neurons. The intrinsic physiology of these cell groups may play an important role in auditory cortex function.

Long-range inhibition within the zebra finch song nucleus RA can coordinate the firing of multiple projection neurons

The zebra finch forebrain song control nucleus RA (robust nucleus of the archistriatum) generates a phasic and temporally precise neural signal that drives vocal and respiratory motoneurons during singing. RA's output during singing predicts individual notes, even though afferent drive to RA from the song nucleus HVc is more tonic, and predicts song syllables, independent of the particular notes that comprise the syllable. Therefore RA's intrinsic circuitry transforms neural activity from HVc into a highly precise premotor output. To understand how RA's intrinsic circuitry effects this transformation, we characterized RA interneurons and projection neurons using intracellular recordings in brain slices. RA interneurons fired fast action potentials with steep current-frequency relationships and had small somata with thin aspinous processes that extended throughout large portions of the nucleus; the similarity of their fine processes to those labeled with a glutamic acid decarboxylase (GAD) antibody strongly suggests that these interneurons are GABAergic. Electrical stimulation revealed that RA interneurons receive excitatory inputs from RA's afferents, the lateral magnocellular nucleus of the anterior neostriatum (LMAN) and HVc, and from local axon collaterals of RA projection neurons. To map the functional connections that RA interneurons make onto RA projection neurons, we focally uncaged glutamate, revealing long-range inhibitory connections in RA. Thus these interneurons provide fast feed-forward and feedback inhibition to RA projection neurons and could help create the phasic pattern of bursts and pauses that characterizes RA output during singing. Furthermore, selectively activating the inhibitory network phase locks the firing of otherwise unconnected pairs of projection neurons, suggesting that local inhibition could coordinate RA output during singing.

The Ornstein-Uhlenbeck process does not reproduce spiking statistics of neurons in prefrontal cortex

Cortical neurons of behaving animals generate irregular spike sequences. Recently, there has been a heated discussion about the origin of this irregularity. Softky and Koch (1993) pointed out the inability of standard single-neuron models to reproduce the irregularity of the observed spike sequences when the model parameters are chosen within a certain range that they consider to be plausible. Shadlen and Newsome (1994), on the other hand, demonstrated that a standard leaky integrate-and-fire model can reproduce the irregularity if the inhibition is balanced with the excitation. Motivated by this discussion, we attempted to determine whether the Ornstein-Uhlenbeck process, which is naturally derived from the leaky integration assumption, can in fact reproduce higher-order statistics of biological data. For this purpose, we consider actual neuronal spike sequences recorded from the monkey prefrontal cortex to calculate the higher-order statistics of the interspike intervals. Consistency of the data with the model is examined on the basis of the coefficient of variation and the skewness coefficient, which are, respectively, a measure of the spiking irregularity and a measure of the asymmetry of the interval distribution. It is found that the biological data are not consistent with the model if the model time constant assumes a value within a certain range believed to cover all reasonable values. This fact suggests that the leaky integrate-and-fire model with the assumption of uncorrelated inputs is not adequate to account for the spiking in at least some cortical neurons.

Differential subcellular localization of forward and feedback interareal inputs to parvalbumin expressing GABAergic neurons in rat visual cortex

In rat visual cortex, forward and feedback interareal pathways innervate both pyramidal and gamma-aminobutyric acid (GABA)ergic (Johnson and Burkhalter [1996] J. Comp. Neurol. 368:383-398). GABAergic neurons consist of different cell types of which the largest group expresses parvalbumin (PV; Gonchar and Burkhalter [1997] Cereb. Cortex 4:347-358). Here, we report that PV neurons in layers 2/3 are synaptic targets of forward and feedback projections between area 17 and the lateromedial area (LM) of rat visual cortex. In both forward and feedback pathways, approximately 90% of axon terminals in layer 2/3 labeled by tracing with biotinylated dextran amine formed synapses with PV-negative profiles. In both pathways, most of these profiles resembled dendritic spines. Although there were no differences in the innervation of PV-negative targets, the two pathways differed in the innervation of PV-positive neurons. In each pathway, approximately 10% of terminals formed synapses with PV-positive profiles. However, in the forward pathway, the size of the contacted PV-positive profiles was larger than in the feedback pathway. Moreover, in the forward pathway, axon terminals on PV-positive profiles were larger, contained more mitochondria and docked synaptic vesicles than feedback synapses on PV neurons. Our results show that PV neurons provide a major target for area 17 <-> LM forward and feedback pathways terminating in upper layers. In each pathway, the proportion of axons contacting PV neurons is similar. However, both pathways differ in the subcellular localization and morphology of synapses on PV neurons. These asymmetries may contribute to the inequality in the strength of disynaptic inhibition evoked by forward and feedback inputs (Shao and Burkhalter [1996] J. Neurosci. 16:7353-7365).

Isodirectional tuning of adjacent interneurons and pyramidal cells during working memory: evidence for microcolumnar organization in PFC

Studies on the cellular mechanisms of working memory demonstrated that neurons in dorsolateral prefrontal cortex (dPFC) exhibit directionally tuned activity during an oculomotor delayed response. To determine the particular contributions of pyramidal cells and interneurons to spatial tuning in dPFC, we examined both individually and in pairs the tuning properties of regular-spiking (RS) and fast-spiking (FS) units that represent putative pyramidal cells and interneurons, respectively. Our main finding is that FS units possess spatially tuned sensory, motor, and delay activity (i. e., "memory fields") similar to those found in RS units. Furthermore, when recorded simultaneously at the same site, the majority of neighboring neurons, whether FS or RS, displayed isodirectional tuning, i.e., they shared very similar tuning angles for the sensory and delay phases of the task. As the trial entered the response phase of the task, many FS units shifted their direction of tuning and became cross-directional to adjacent RS units by the end of the trial. These results establish that a large part of inhibition in prefrontal cortex is spatially oriented rather than being untuned and simply regulating the threshold response of pyramidal cell output. Moreover, the isodirectional tuning between adjacent neurons supports a functional microcolumnar organization in dPFC for spatial memory fields similar to that found in other areas of cortex for sensory receptive fields.

Impact of network activity on the integrative properties of neocortical pyramidal neurons in vivo

During wakefulness, neocortical neurons are subjected to an intense synaptic bombardment. To assess the consequences of this background activity for the integrative properties of pyramidal neurons, we constrained biophysical models with in vivo intracellular data obtained in anesthetized cats during periods of intense network activity similar to that observed in the waking state. In pyramidal cells of the parietal cortex (area 5-7), synaptic activity was responsible for an approximately fivefold decrease in input resistance (Rin), a more depolarized membrane potential (Vm), and a marked increase in the amplitude of Vm fluctuations, as determined by comparing the same cells before and after microperfusion of tetrodotoxin (TTX). The model was constrained by measurements of Rin, by the average value and standard deviation of the Vm measured from epochs of intense synaptic activity recorded with KAc or KCl-filled pipettes as well as the values measured in the same cells after TTX. To reproduce all experimental results, the simulated synaptic activity had to be of relatively high frequency (1-5 Hz) at excitatory and inhibitory synapses. In addition, synaptic inputs had to be significantly correlated (correlation coefficient approximately 0.1) to reproduce the amplitude of Vm fluctuations recorded experimentally. The presence of voltage-dependent K+ currents, estimated from current-voltage relations after TTX, affected these parameters by <10%. The model predicts that the conductance due to synaptic activity is 7-30 times larger than the somatic leak conductance to be consistent with the approximately fivefold change in Rin. The impact of this massive increase in conductance on dendritic attenuation was investigated for passive neurons and neurons with voltage-dependent Na+/K+ currents in soma and dendrites. In passive neurons, correlated synaptic bombardment had a major influence on dendritic attenuation. The electrotonic attenuation of simulated synaptic inputs was enhanced greatly in the presence of synaptic bombardment, with distal synapses having minimal effects at the soma. Similarly, in the presence of dendritic voltage-dependent currents, the convergence of hundreds of synaptic inputs was required to evoke action potentials reliably. In this case, however, dendritic voltage-dependent currents minimized the variability due to input location, with distal apical synapses being as effective as synapses on basal dendrites. In conclusion, this combination of intracellular and computational data suggests that, during low-amplitude fast electroencephalographic activity, neocortical neurons are bombarded continuously by correlated synaptic inputs at high frequency, which significantly affect their integrative properties. A series of predictions are suggested to test this model.

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 (approximately 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.

Emergent oscillations in a realistic network: the role of inhibition and the effect of the spatiotemporal distribution of the input

We have simulated a network of 10,000 two-compartment cells, spatially distributed on a two-dimensional sheet; 15% of the cells were inhibitory. The input to the network was spatially delimited. Global oscillations frequently were achieved with a simple set of connectivity rules. The inhibitory neurons paced the network, whereas the excitatory neurons amplified the input, permitting oscillations at low-input intensities. Inhibitory neurons were active over a greater area than excitatory ones, forming a ring of inhibition. The oscillation frequency was modulated to some extent by the input intensity, as has been shown experimentally in the striate cortex, but predominantly by the properties of the inhibitory neurons and their connections: the membrane and synaptic time constants and the distribution of delays. In networks that showed oscillations and in those that did not, widely distributed inputs could lead to the specific recruitment of the inhibitory neurons and to near zero activity of the excitatory cells. Hence the spatial distribution of excitatory inputs could provide a means of selectively exciting or inhibiting a target network. Finally, neither the presence of oscillations nor the global spike activity provided any reliable indication of the level of excitatory output from the network.

The functional influence of burst and tonic firing mode on synaptic interactions in the thalamus

Thalamocortical and perigeniculate (PGN) neurons can generate action potentials either as Ca2+ spike-mediated high-frequency bursts or as tonic trains. Using dual intracellular recordings in vitro in monosynaptically connected pairs of PGN and dorsal lateral geniculate nucleus (LGNd) neurons, we found that the functional effect of synaptic transmission between these cell types was strongly influenced by the membrane potential and hence the firing mode of both the pre- and postsynaptic neurons. Activation of single action potentials or low-frequency spike trains in PGN or thalamocortical neurons resulted in the generation of PSPs that were 0.5-2.0 mV in amplitude. In contrast, the generation of Ca2+ spike-mediated bursts of action potentials in the presynaptic cell increased these PSPs to an average of 4.4 mV for the IPSP and 3.0 mV for the EPSP barrage, because of temporal summation and/or facilitation. If the postsynaptic neuron was at a resting membrane potential (e.g., -65 mV), these PSP barrages could result in the activation of a low-threshold Ca2+ spike and burst of action potentials. These results demonstrate that the burst firing mode of action potential generation is a particularly effective means by which perigeniculate and thalamocortical neurons may influence one another. We propose that the activation of burst discharges in these cell types is essential for the generation of some forms of synchronized rhythmic oscillations of sleep and of epileptic seizures.

Complex microstructures of sensory cortical connections

The neocortex has a distinctive laminar and modular organization. Although important questions remain regarding structure and function at this level of organization, recent studies are addressing a finer scale of synaptic and network microstructure. New findings concerning network properties are rapidly emerging from approaches in which dual or triple intracellular recordings in vitro are combined with analyses of cell and synaptic morphology, as well as from experiments designed to label multiple cell populations.

Target-cell-specific facilitation and depression in neocortical circuits

In neocortical circuits, repetitively active neurons evoke unitary postsynaptic potentials (PSPs) whose peak amplitudes either increase (facilitate) or decrease (depress) progressively. To examine the basis for these different synaptic responses, we made simultaneous recordings from three classes of neurons in cortical layer 2/3. We induced repetitive action potentials in pyramidal cells and recorded the evoked unitary excitatory (E)PSPs in two classes of GABAergic neurons. We observed facilitation of EPSPs in bitufted GABAergic interneurons, many of which expressed somatostatin immunoreactivity. EPSPs recorded from multipolar interneurons, however, showed depression. Some of these neurons were immunopositive for parvalbumin. Unitary inhibitory (I)PSPs evoked by repetitive stimulation of a bitufted neuron also showed a less pronounced but significant difference between the two target neurons. Facilitation and depression involve presynaptic mechanisms, and because a single neuron can express both behaviors simultaneously, we infer that local differences in the molecular structure of presynaptic nerve terminals are induced by retrograde signals from different classes of target neurons. Because bitufted and multipolar neurons both formed reciprocal inhibitory connections with pyramidal cells, the results imply that the balance of activation between two recurrent inhibitory pathways in the neocortex depends on the frequency of action potentials in pyramidal cells.

Reliability and state dependence of pyramidal cell-interneuron synapses in the hippocampus: an ensemble approach in the behaving rat

Spike transmission probability between pyramidal cells and interneurons in the CA1 pyramidal layer was investigated in the behaving rat by the simultaneous recording of neuronal ensembles. Population synchrony was strongest during sharp wave (SPW) bursts. However, the increase was three times larger for pyramidal cells than for interneurons. The contribution of single pyramidal cells to the discharge of interneurons was often large (up to 0.6 probability), as assessed by the presence of significant (<3 ms) peaks in the cross-correlogram. Complex-spike bursts were more effective than single spikes. Single cell contribution was higher between SPW bursts than during SPWs or theta activity. Hence, single pyramidal cells can reliably discharge interneurons, and the probability of spike transmission is behavior dependent.

Postsynaptic pyramidal target selection by descending layer III pyramidal axons: dual intracellular recordings and biocytin filling in slices of rat neocortex

Paired intracellular recordings in slices of adult rat neocortex with biocytin filling of synaptically connected neurons were used to investigate the pyramidal targets, in layer V, of layer III pyramidal axons. The time-course and sensitivity of excitatory postsynaptic potentials to current injected at the soma, and locations of close appositions between presynaptic axons and postsynaptic dendrites, indicated that the majority of contributory synapses were located in layer V. Within a "column" of tissue, radius < or = 250 microm, the probability that a randomly selected layer III pyramid innervated a layer V pyramid was 1 in 4 if the target cell was a burst firing pyramid with an apical dendritic tuft in layers II/I. If, however, the potential target was a regular spiking pyramid, the probability of connectivity was only 1 in 40, and none of the 13 anatomically identified postsynaptic layer V targets had a slender apical dendrite terminating in layers IV/III. Morphological reconstructions indicated that layer III pyramids select target layer V cells whose apical dendrites pass within 50-100 microm of the soma of the presynaptic pyramid in layer III and which have overlapping apical dendritic tufts in the superficial layers. The probability that a layer V cell would innervate a layer III pyramid lying within 250 microm of its apical dendrite was much lower (one in 58). Both presynaptic layer III pyramids and their large postsynaptic layer V targets could therefore access similar inputs in layers I/II, while small layer V pyramids could not. One prediction from the present data would be that neither descending layer V inputs to the striatum or thalamus, nor transcallosal connections would be readily activated by longer distance cortico-cortical "feedback" connections that terminated in layers I/II. These could, however, activate corticofugal pathways to the superior colliculus or pons, both directly and via layer III.

Salient features of synaptic organisation in the cerebral cortex

The neuronal and synaptic organisation of the cerebral cortex appears exceedingly complex, and the definition of a basic cortical circuit in terms of defined classes of cells and connections is necessary to facilitate progress of its analysis. During the last two decades quantitative studies of the synaptic connectivity of identified cortical neurones and their molecular dissection revealed a number of general rules that apply to all areas of cortex. In this review, first the precise location of postsynaptic GABA and glutamate receptors is examined at cortical synapses, in order to define the site of synaptic interactions. It is argued that, due to the exclusion of G protein-coupled receptors from the postsynaptic density, the presence of extrasynaptic receptors and the molecular compartmentalisation of the postsynaptic membrane, the synapse should include membrane areas beyond the membrane specialisation. Subsequently, the following organisational principles are examined: 1. The cerebral cortex consists of: (i) a large population of principal neurones reciprocally connected to the thalamus and to each other via axon collaterals releasing excitatory amino acids, and, (ii) a smaller population of mainly local circuit GABAergic neurones. 2. Differential reciprocal connections are also formed amongst GABAergic neurones. 3. All extrinsic and intracortical glutamatergic pathways terminate on both the principal and the GABAergic neurones, differentially weighted according to the pathway. 4. Synapses of multiple sets of glutamatergic and GABAergic afferents subdivide the surface of cortical neurones and are often co-aligned on the dendritic domain. 5. A unique feature of the cortex is the GABAergic axo-axonic cell, influencing principal cells through GABAA receptors at synapses located exclusively on the axon initial segment. The analysis of these salient features of connectivity has revealed a remarkably selective array of connections, yet a highly adaptable design of the basic circuit emerges when comparisons are made between cortical areas or layers. The basic circuit is most obvious in the hippocampus where a relatively homogeneous set of spatially aligned principal cells allows an easy visualization of the organisational rules. Those principles which have been examined in the isocortex proved to be identical or very similar. In the isocortex, the basic circuit, scaled to specific requirements, is repeated in each layer. As multiple sets of output neurones evolved, requiring subtly different needs for their inputs, the basic circuit may be superimposed several times in the same layer. Tangential intralaminar connections in both the hippocampus and isocortex also connect output neurones with similar properties, as best seen in the patchy connections in the isocortex. The additional radial superposition of several laminae of distinct sets of output neurones, each representing and supported by its basic circuit, requires a co-ordination of their activity that is mediated by highly selective interlaminar connections, involving both the GABAergic and the excitatory amino acid releasing neurones. The remarkable specificity in the geometry of cells and the selectivity in placement of neurotransmitter receptors and synapses on their surface, strongly suggest a predominant role for time in the coding of information, but this does not exclude an important role also for the rate of action potential discharge in cortical representation of information.

CA1 pyramidal to basket and bistratified cell EPSPs: dual intracellular recordings in rat hippocampal slices

1. Dual intracellular recordings in the CA1 region of adult rat hippocampal slices and biocytin filling of synaptically connected cells were used to study the excitatory postsynaptic potentials (EPSPs) elicited in basket (n = 7) and bistratified interneurones (n = 7) by action potentials activated in simultaneously recorded pyramidal cells. 2. Interneurones could be subdivided according to their electrophysiological properties into classical fast spiking, burst firing, regular spiking and fast spiking cells with a rounded spike after-hyperpolarization. These physiological classes did not, however, correlate with morphological type. EPSPs were not recorded in regular spiking cells. 3. Average EPSP amplitudes were larger in bistratified cells (range, 0.5-9 mV) than in basket cells (range, 0. 15-3.6 mV) and the probability of obtaining a pyramidal cell-interneurone EPSP was also higher for the bistratified cells (1:7) than for the basket cells (1:22). EPSP 10-90 % rise times in bistratified cells (0.7-2 ms) and their widths at half-amplitude (3. 9-11.2 ms) were slightly longer than in basket cells (rise times, 0.4-1.6 ms; half-widths, 2.2-9.7 ms). 4. The majority of these EPSPs (6 of 8 tested) increased in amplitude and duration with postsynaptic depolarization, although in two (of 4) basket cells the voltage relation was conventional. 5. All EPSPs tested in both basket (n = 7) and bistratified cells (n = 5) decreased in amplitude with repetitive presynaptic firing. The average amplitudes of second EPSPs elicited within 15 ms of the first were between 34 and 94 % of the average amplitude of the first EPSP. Third and fourth EPSPs in brief trains were further depressed. This depression was associated with an increase in the incidence of apparent failures of transmission indicating a presynaptic locus.

Quantal analysis of synaptic processes in the neocortex

The application of fluctuation analysis to studies of synaptic function in the neocortex is discussed. Analysis of failures of transmission has been valuable in indicating whether a presynaptic or a postsynaptic site is responsible for a change in synaptic efficacy. When combined with detailed ultrastructural verification of all synapses involved in an individual cell to cell connection, a reasonable estimate of quantal size and release probability under conditions of low frequency activity can be obtained. However, both the number of available release sites in functional terms and the probability that an action potential (AP) will release transmitter from any given site can vary from AP to AP at higher frequencies. A variety of presynaptic mechanisms that modulate release are now apparent. For example, one mechanism dominates release patterns at one class of connection which is insensitive to absolute firing frequency, but responsive to changes in frequency. At another class of connection, a different mechanism dominates, resulting in high sensitivity to frequency.