氧化锌/氧化钽双介质层忆阻器的突触特性分析

人工突触的开发是模拟人脑功能的关键。忆阻器由于具备独特的电阻记忆行为,在人工突触研究领域得到广泛的关注。氧化钽和氧化锌均是优良的忆阻器材料,鉴于有关ZnO/TaOx双介质忆阻器突触特性的研究较少,本文将氧化锌介质层引入Ti/TaOx/ITO忆阻器中拟改善其突触性能。研究发现,器件Ti/ZnO/TaOx/ITO在功耗和电导调制线性度方面皆有改善,并有着电阻渐变的行为,利于器件突触功能的实现及应用。为此对Ti/ZnO/TaOx/ITO双介质层器件进行了电压脉冲训练,并成功模拟了学习饱和、经验学习以及短时程记忆向长时程记忆转变等生物突触行为。     

Locally dynamic synaptic learning rules in pyramidal neuron dendrites

Long-term potentiation (LTP) of synaptic transmission underlies aspects of learning and memory. LTP is input-specific at the level of individual synapses, but neural network models predict interactions between plasticity at nearby synapses. Here we show in mouse hippocampal pyramidal cells that LTP at individual synapses reduces the threshold for potentiation at neighbouring synapses. After input-specific LTP induction by two-photon glutamate uncaging or by synaptic stimulation, subthreshold stimuli, which by themselves were too weak to trigger LTP, caused robust LTP and spine enlargement at neighbouring spines. Furthermore, LTP induction broadened the presynaptic-postsynaptic spike interval for spike-timing-dependent LTP within a dendritic neighbourhood. The reduction in the threshold for LTP induction lasted approximately 10 min and spread over approximately 10 microm of dendrite. These local interactions between neighbouring synapses support clustered plasticity models of memory storage and could allow for the binding of behaviourally linked information on the same dendritic branch.     

Long-term potentiation of electrotonic coupling at mixed synapses

Long-term potentiation of chemical synapses is closely related to memory and learning. Studies of this process have concentrated on chemically mediated excitatory synapses. By contrast, activity-dependent modification of gap junctions, which also widely exist in higher structures such as hippocampus and neocortex, has not been described. Here we report that at mixed synapses between sensory afferents and an identified reticulospinal neuron, the electrotonic coupling potential can be potentiated, as well as the chemically mediated excitatory postsynaptic potential, for a prolonged time period using a stimulation paradigm like that which produces long-term potentiation in hippocampus. The effect on coupling is due to an increase in gap-junctional conductance. Our data indicate that the potentiation of both synaptic components requires an increase in intracellular calcium, involves activation of NMDA (N-methyl-D-aspartate) receptors, and is specific to the tetanized pathway.     

忆阻器类脑神经突触的研究进展

大脑之所以能够控制人和动物的复杂生命活动,使生物体在多变的自然环境得以生存,得益于大规模神经网络中高效、快速、精准的信息传递。神经突触作为神经元之间信息传递的重要机构,保证了神经网络的高效运转,因此构建具有神经突触功能的电子突触是研究仿生系统和类脑神经网络的必经之路。研究人员尝试各种电子元件对神经突触进行模拟,其中忆阻器由于其独特的器件结构和具有“记忆特性”的电学性能,成为构建类脑神经突触的最佳选择。文章全面概述近年来忆阻器模拟神经突触的研究进展,包括忆阻器模拟神经突触的可塑性、再可塑性、非联想学习、联想学习等功能,总结了忆阻器神经突触在人工神经网络中的应用、存在的问题和挑战,并对忆阻器神经突触的研究进行展望。     

Repetitive motor learning induces coordinated formation of clustered dendritic spines in vivo

Many lines of evidence suggest that memory in the mammalian brain is stored with distinct spatiotemporal patterns. Despite recent progresses in identifying neuronal populations involved in memory coding, the synapse-level mechanism is still poorly understood. Computational models and electrophysiological data have shown that functional clustering of synapses along dendritic branches leads to nonlinear summation of synaptic inputs and greatly expands the computing power of a neural network. However, whether neighbouring synapses are involved in encoding similar memory and how task-specific cortical networks develop during learning remain elusive. Using transcranial two-photon microscopy, we followed apical dendrites of layer 5 pyramidal neurons in the motor cortex while mice practised novel forelimb skills. Here we show that a third of new dendritic spines (postsynaptic structures of most excitatory synapses) formed during the acquisition phase of learning emerge in clusters, and that most such clusters are neighbouring spine pairs. These clustered new spines are more likely to persist throughout prolonged learning sessions, and even long after training stops, than non-clustered counterparts. Moreover, formation of new spine clusters requires repetition of the same motor task, and the emergence of succedent new spine(s) accompanies the strengthening of the first new spine in the cluster. We also show that under control conditions new spines appear to avoid existing stable spines, rather than being uniformly added along dendrites. However, succedent new spines in clusters overcome such a spatial constraint and form in close vicinity to neighbouring stable spines. Our findings suggest that clustering of new synapses along dendrites is induced by repetitive activation of the cortical circuitry during learning, providing a structural basis for spatial coding of motor memory in the mammalian brain.     

CaMKII regulates the density of central glutamatergic synapses in vivo

Synaptic connections undergo a dynamic process of stabilization or elimination during development, and this process is thought to be critical in memory and learning and in establishing the specificity of synaptic connections. The type II calcium- and calmodulin-dependent protein kinase (CaMKII) has been proposed to be pivotal in regulating synaptic strength and in maturation of synapses during development. Here we describe how CaMKII regulates the formation of central glutamatergic synapses in Caenorhabditis elegans. During larval development, the density of ventral nerve cord synapses containing the GLR-1 glutamate receptor is held constant despite marked changes in neurite length. The coupling of synapse number to neurite length requires both CaMKII and voltage-gated calcium channels. CaMKII regulates GLR-1 by at least two distinct mechanisms: regulating transport of GLR-1 from cell bodies to neurites; and regulating the addition or maintenance of GLR-1 to postsynaptic elements.     

Synaptic computation

Neurons are often considered to be the computational engines of the brain, with synapses acting solely as conveyers of information. But the diverse types of synaptic plasticity and the range of timescales over which they operate suggest that synapses have a more active role in information processing. Long-term changes in the transmission properties of synapses provide a physiological substrate for learning and memory, whereas short-term changes support a variety of computations. By expressing several forms of synaptic plasticity, a single neuron can convey an array of different signals to the neural circuit in which it operates.     

基于启发式退火拓扑择优机制的稀疏联想记忆实现

借鉴统计物理学中的“退火”概念，针对已有稀疏互联联想记忆模型中只考虑网络连接随机稀疏方式，缺乏面向特定模式存储任务的确定性操作，使用非平衡态统计分析方法，讨论了有限代谢能量资源约束下的网络结构最优稀疏原则，给出了相应的理论推导.在此基础上，研究了面向特定学习任务的网络稀疏结构自适应方法，构建了基于启发式退火拓扑择优机制的稀疏联想记忆模型.实验表明，该模型既具有一定的生物学基础，维持了网络结构广泛稀疏互联的特性，又能在网络资源受限条件下达到最优联想记忆性能，符合神经生物系统本身自组织、自学习的特点.     

Glycine enhances NMDA-receptor mediated synaptic potentials in neocortical slices

One class of excitatory amino-acid receptors, the N-methyl-D-aspartate (NMDA) receptors, mediates transmission at a small, but important, group of synapses in the neocortex. These receptors are implicated in neuronal plasticity during development in young mammals and in memory acquisition in adults. Recently, responses of isolated membrane patches to NMDA were shown to be greatly enhanced by glycine. This, together with the demonstration that the strychnine-insensitive glycine-binding site is distinct from, but linked to, the NMDA receptor has excited intense interest in glycine as a synaptic modulator. Before proposing a physiological function, however, it is important to determine whether glycine could enhance synaptic responses to NMDA receptor activation in intact, adult tissue. An earlier study failed to demonstrate enhancement of NMDA responses when glycine was applied and it was proposed that in intact tissue the high-affinity glycine site was already saturated by endogenous glycine. It remained possible that glycine concentrations can be maintained at low levels close to synaptic receptors. We have examined responses of neurons in slices of adult neocortex to focal applications of excitatory amino acids and glycine and report enhancement by glycine of NMDA receptor-mediated excitatory postsynaptic potentials.     

Recovery of spatial learning deficits after decay of electrically induced synaptic enhancement in the hippocampus

A widespread interest in a long-lasting form of synaptic enhancement in hippocampal circuits has arisen largely because it might reflect the activation of physiological mechanisms that underlie rapid associative learning. As its induction normally requires the 'Hebbian' association of activity on a number of input fibres, we refer to the process as long-term enhancement (LTE) rather than long-term potentiation (LTP), to emphasize its distinction from the ubiquitous, non-associative 'potentiation' phenomena that occur at most synapses, including those exhibiting LTE. Among other evidence that LTE might actually have a role in associative memory is the demonstration that repeated high-frequency stimulation, which saturated the inducible LTE, caused a severe deficit in spatial learning, although it had no effect on well established spatial memory. These results were consistent with a widespread view that information need only temporarily be stored in the hippocampal formation in order for long-term memories to be established in neocortical circuits. In this context, it is important to understand whether the possible underlying synaptic changes are of a permanent character, or are relatively transient. A second question is whether the actual cause of the observed learning deficit is the distruption of the synaptic weight distribution, and/or the limitation of further synaptic change, which presumably results from experimental saturation of the LTE mechanism. Alternatively, the deficit could be a consequence of some unobserved secondary effect of the high-frequency electrical stimulation. Here we demonstrate that learning capacity recovers in about the same time that it takes LTE to decay, which strongly favours the first possibility and supports the idea that LTE-like processes actually underlie associative memory.     

Neurotrophic regulation of synapse development and plasticity

Neurotrophic factors are traditionally thought to be secretory proteins that regulate long-tern survival and differe, ntiation of neurons. Recent studies have revealed a previously unexpected role for these factors in synaptie de velopment ami plasticity in diverse neuronal populations. Here we review experimeuts carried oul in our own laboratory in the last few years.. We have made two important discoveries.First,we were among the first to report that brain-derived. neurotrophie faclor (BDNF) facilitates hippocampal hmg-term potentiation (LTP), a form of synaptic plaslicity believed to be involved in learning and memory. BDNF modulates LTP al CAI synapses by enhaneing synaptic responses to high frequency, tetanic slimulalion. This is achieved primafily by facilitating synaptie vesicle doeking, possibly due to an in crease in the levels of the vesicle prolein synaptobrevin and synaptoplysin in the nerve terminals. Gene knockout study demonstrates thai the effects of BDNF are primarily mediated through presynaptic mechanisms. Second, we demonstrated a form of long-term, neurotrophin-mediated synaptic regulation. We showed that long-term treatment of the neuromuscu lar synapses with neurotrophin-3 (NT3) resulted in an enhancement of both spontaneous and evoked synaptic currcuts, as well as profound changes in thc number of synaptic varicosities and syuaptic vesicle proteins in motoneurons, all of which are indicative of more mature synapses. Our current work addresses the following issues:(i) activity-dependent trafficking of neurotrophin receptors, and its role in synapse-specific modulation; (ii) signal transduction mechanisms medialing the acute enhancement of synaplic transmission by neurotrophins; (iii) acute and long-tenn synaptie actions of the GDNF family; (iv) role of BDNF in late-phase LTP and in the development of hippocampal circuit.     

Kainate receptors are involved in synaptic plasticity

The ability of synapses to modify their synaptic strength in response to activity is a fundamental property of the nervous system and may be an essential component of learning and memory. There are three classes of ionotropic glutamate receptor, namely NMDA (N-methyl-D-aspartate), AMPA (alpha-amino-3-hydroxy-5-methyl-4-isoxazole-4-propionic acid) and kainate receptors; critical roles in synaptic plasticity have been identified for two of these. Thus, at many synapses in the brain, transient activation of NMDA receptors leads to a persistent modification in the strength of synaptic transmission mediated by AMPA receptors. Here, to determine whether kainate receptors are involved in synaptic plasticity, we have used a new antagonist, LY382884 ((3S, 4aR, 6S, 8aR)-6-((4-carboxyphenyl)methyl-1,2,3,4,4a,5,6,7,8,8a-decahydro isoquinoline-3-carboxylic acid), which antagonizes kainate receptors at concentrations that do not affect AMPA or NMDA receptors. We find that LY382884 is a selective antagonist at neuronal kainate receptors containing the GluR5 subunit. It has no effect on long-term potentiation (LTP) that is dependent on NMDA receptors but prevents the induction of mossy fibre LTP, which is independent of NMDA receptors. Thus, kainate receptors can act as the induction trigger for long-term changes in synaptic transmission.     

Dendritic spikes as a mechanism for cooperative long-term potentiation

Strengthening of synaptic connections following coincident pre- and postsynaptic activity was proposed by Hebb as a cellular mechanism for learning. Contemporary models assume that multiple synapses must act cooperatively to induce the postsynaptic activity required for hebbian synaptic plasticity. One mechanism for the implementation of this cooperation is action potential firing, which begins in the axon, but which can influence synaptic potentiation following active backpropagation into dendrites. Backpropagation is limited, however, and action potentials often fail to invade the most distal dendrites. Here we show that long-term potentiation of synapses on the distal dendrites of hippocampal CA1 pyramidal neurons does require cooperative synaptic inputs, but does not require axonal action potential firing and backpropagation. Rather, locally generated and spatially restricted regenerative potentials (dendritic spikes) contribute to the postsynaptic depolarization and calcium entry necessary to trigger potentiation of distal synapses. We find that this mechanism can also function at proximal synapses, suggesting that dendritic spikes participate generally in a form of synaptic potentiation that does not require postsynaptic action potential firing in the axon.     

Long-term potentiation of NMDA receptor-mediated synaptic transmission in the hippocampus.

Neurotransmission at most excitatory synapses in the brain operates through two types of glutamate receptor termed alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionate (AMPA) and N-methyl-D-aspartate (NMDA) receptors; these mediate the fast and slow components of excitatory postsynaptic potentials respectively. Activation of NMDA receptors can also lead to a long-lasting modification in synaptic efficiency at glutamatergic synapses; this is exemplified in the CA1 region of the hippocampus, where NMDA receptors mediate the induction of long-term potentiation (LTP). It is believed that in this region LTP is maintained by a specific increase in the AMPA receptor-mediated component of synaptic transmission. We now report, however, that a pharmacologically isolated NMDA receptor-mediated synaptic response can undergo robust, synapse-specific LTP. This finding has implications for neuropathologies such as epilepsy and neurodegeneration, in which excessive NMDA receptor activation has been implicated. It adds fundamentally to theories of synaptic plasticity because NMDA receptor activation may, in addition to causing increased synaptic efficiency, directly alter the plasticity of synapses.     

模式识别中不同记忆编码的模式和模板的匹配

通过一项全部报告实验和两项抽样汉字识别实验,探讨了模式识别中不同记忆编码的模式和模板如何匹配的问题.实验结果显示:全部报告实验中,待识别汉字模式平均转换到短时记忆的量为3.64个,比进入到感觉记忆中的信息量(8个)少得多,这表明感觉记忆信息向短时记忆的转换受到了约束;抽样汉字识别实验1中,匹配汉字模式识别量与其转换量之间无显著差异(p＞0.5),说明转换到短时记忆的信息被等消耗识别;抽样汉字识别实验2中,匹配汉字模式识别量约为其转换量的2倍,差异显著(p＜0.001),基于受约束转换和等消耗识别现象,由模式识别信息来源的判据的判据可知,模式识别直接利用了感觉记忆编码的模式信息.抽样汉字识别实验2中,与模式进行匹配的模板只具有短时记忆编码,因此,有理由认为:模式和模板能以不同的记忆编码直接匹配.     

LTP promotes formation of multiple spine synapses between a single axon terminal and a dendrite

Structural remodelling of synapses and formation of new synaptic contacts has been postulated as a possible mechanism underlying the late phase of long-term potentiation (LTP), a form of plasticity which is involved in learning and memory. Here we use electron microscopy to analyse the morphology of synapses activated by high-frequency stimulation and identified by accumulated calcium in dendritic spines. LTP induction resulted in a sequence of morphological changes consisting of a transient remodelling of the postsynaptic membrane followed by a marked increase in the proportion of axon terminals contacting two or more dendritic spines. Three-dimensional reconstruction revealed that these spines arose from the same dendrite. As pharmacological blockade of LTP prevented these morphological changes, we conclude that LTP is associated with the formation of new, mature and probably functional synapses contacting the same presynaptic terminal and thereby duplicating activated synapses.     

RIM1alpha is required for presynaptic long-term potentiation.

Two main forms of long-term potentiation (LTP)-a prominent model for the cellular mechanism of learning and memory-have been distinguished in the mammalian brain. One requires activation of postsynaptic NMDA (N-methyl d-aspartate) receptors, whereas the other, called mossy fibre LTP, has a principal presynaptic component. Mossy fibre LTP is expressed in hippocampal mossy fibre synapses, cerebellar parallel fibre synapses and corticothalamic synapses, where it apparently operates by a mechanism that requires activation of protein kinase A. Thus, presynaptic substrates of protein kinase A are probably essential in mediating this form of long-term synaptic plasticity. Studies of knockout mice have shown that the synaptic vesicle protein Rab3A is required for mossy fibre LTP, but the protein kinase A substrates rabphilin, synapsin I and synapsin II are dispensable. Here we report that mossy fibre LTP in the hippocampus and the cerebellum is abolished in mice lacking RIM1alpha, an active zone protein that binds to Rab3A and that is also a protein kinase A substrate. Our results indicate that the long-term increase in neurotransmitter release during mossy fibre LTP may be mediated by a unitary mechanism that involves the GTP-dependent interaction of Rab3A with RIM1alpha at the interface of synaptic vesicles and the active zone.     

Rapid BDNF-induced retrograde synaptic modification in a developing retinotectal system

In cultures of hippocampal neurons, induction of long-term synaptic potentiation or depression by repetitive synaptic activity is accompanied by a retrograde spread of potentiation or depression, respectively, from the site of induction at the axonal outputs to the input synapses on the dendrites of the presynaptic neuron. We report here that rapid retrograde synaptic modification also exists in an intact developing retinotectal system. Local application of brain-derived neurotrophic factor (BDNF) to the Xenopus laevis optic tectum, which induced persistent potentiation of retinotectal synapses, led to a rapid modification of synaptic inputs at the dendrites of retinal ganglion cells (RGCs), as shown by a persistent enhancement of light-evoked excitatory synaptic currents and spiking activity of RGCs. This retrograde effect required TrkB receptor activation, phospholipase Cgamma activity and Ca2+ elevation in RGCs, and was accounted for by a selective increase in the number of postsynaptic AMPA-subtype glutamate receptors at RGC dendrites. Such retrograde information flow in the neuron allows rapid regulation of synaptic inputs at the dendrite in accordance to signals received at axon terminals, a process reminiscent of back-propagation algorithm for learning in neural networks.