Review
Structural dynamics of dendritic spines in memory and cognition Haruo Kasai1, Masahiro Fukuda1, Satoshi Watanabe1, Akiko Hayashi-Takagi2 and Jun Noguchi1 1
Laboratory of Structural Physiology, Center for Disease Biology and Integrative Medicine, Faculty of Medicine, and Center for NanoBio Integration, University of Tokyo, Bunkyo-ku, Tokyo 113-0033, Japan 2 Department of Psychiatry and Behavioral Neurosciences, Johns Hopkins University School of Medicine, Baltimore MD 21287, USA
Recent studies show that dendritic spines are dynamic structures. Their rapid creation, destruction and shapechanging are essential for short- and long-term plasticity at excitatory synapses on pyramidal neurons in the cerebral cortex. The onset of long-term potentiation, spine-volume growth and an increase in receptor trafficking are coincident, enabling a ‘functional readout’ of spine structure that links the age, size, strength and lifetime of a synapse. Spine dynamics are also implicated in long-term memory and cognition: intrinsic fluctuations in volume can explain synapse maintenance over long periods, and rapid, activity-triggered plasticity can relate directly to cognitive processes. Thus, spine dynamics are cellular phenomena with important implications for cognition and memory. Furthermore, impaired spine dynamics can cause psychiatric and neurodevelopmental disorders. Dendritic spines On pyramidal neurons in the cerebral cortex, excitatory synapses terminate at spines, which are short protrusions joined to the main dendrite by a thin neck. Discovered in the 19th century and intensely scrutinized in the 20th century, dendritic spines are found in higher animals [1,2] and some insects [3,4]. Spines exist only on certain types of neurons, including pyramidal neurons in the cortex, medium spiny neurons in the basal ganglia and Purkinje cells in the cerebellum. Spines are more abundant in higher brain regions and highly variable in shape. Moreover, dendritic spines are the most actin-rich structures in the brain [5,6], and their morphology and density are abnormal in several mental disorders [7]. The best-known example of input-specific, activity-dependent synaptic plasticity—Donald Hebb’s canonical basis for learning and memory [8] —is long-term potentiation (LTP) of spine synapses in the hippocampus [9]. The link between LTP and spine structure was suggested by the finding that the size of the postsynaptic density (PSD) is related to the size of the spine head [10] and the number of AMPA-type glutamate receptors within it [11–13]. These ultrastructural studies, however, could not determine the functional state of a spine. This structure–function relationship was first established in 2001 using two-photon uncaging of a caged–glutamate compound [14–20]. Later Corresponding author: Kasai, H. (
[email protected])
reports showed that spine enlargements are associated with LTP in single identified spines (Figure 1a) [21], indicating that Hebb’s learning rule applies even at the level of a single synapse. Many studies have since confirmed that the induction of LTP or long-term depression (LTD), another form of activity-dependent plasticity, induces structural plasticity of spines in stimulated dendritic branches [22–30]. Given the apparent stability in vivo of dendritic and axonal arbors at low magnification [31–33], the properties that govern spine dynamics over the long-term could play a major role in reorganizing cortical circuitry throughout life [34,35]. In fact, spines are frequently generated and eliminated even in the adult neocortex, and these events have been suggested as substrates for stable memory formation [35–38]. Both formation and enlargement of spines are important during synaptic rearrangements in the visual cortex that follow sensory deprivation [39]. It is important to note that spine structural dynamics include broader phenomena than LTP/LTD. Namely, they include the generation and elimination of spines [31,32,40– 42], and long-term, activity-independent fluctuations (described in detail below) (Figure 1b) [42]. In addition, spines become larger in response to the force of actin polymerization [43], which occurs within seconds [21] of LTP induction [44]. Spines seem to display expansive force continuously to maintain their shape and function [43]. These findings suggest that spine synapses are not just electrochemical but also mechanical in nature. The purpose of this article is to present the new findings on spine dynamics that can be extrapolated to a broad spectrum of higher-order brain functions. We summarize the relationships between spine structural dynamics and functional plasticity, explain the long-term maintenance of spine structures, propose an explanation for the impairment of spine dynamics in mental disorders and introduce the possible relationships between rapid spine dynamics and cognitive processes. Activity-dependent structural plasticity of dendritic spines and receptor trafficking At the level of the dendritic spine, structural dynamics and receptor trafficking both contribute to functional plasticity. For example, spine enlargement occurs within a minute (Figure 1a) [21], a time course that matches the rapid
0166-2236/$ – see front matter ß 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.tins.2010.01.001
121
Review
Trends in Neurosciences Vol.33 No.3
We speculate that receptor trafficking does not support the long-term maintenance of plasticity in the absence of structural changes (Figure 1c), based on the turnover rates for individual molecules: the lifetime of a spine is more than a thousand times greater than that of its molecular constituents. Even PSD-95, among the most stable molecules in the spine, redistributes within 3 h in both dissociated cultures and the intact brain in vivo [56,57]. Spine lifetimes, by contrast, can be more than a year in vivo [32,35]. The autophosphorylation of CaMKII, once thought to help maintain long-term changes in synaptic strength, has since been revealed to be rather transient [58]. Thus, maintenance of the molecular status of the spine can require some association with its structure. Given this evidence, it would be interesting to test whether the functional plasticity induced without structural changes to the spine [24,46,59,60] can persist as long as the functional plasticity that accompanies structural alterations. We next address how spine structure can be maintained over days (Figure 1b). Figure 1. Structural dynamics of dendritic spines. Rapid and slow changes in spine shape and size result in rapid and slow synaptic plasticities. (a) Time course of spine enlargement induced by the two-photon uncaging of a caged–glutamate compound. Spine enlargement is specific to the stimulated spine (red circle) and does not spread to the neighboring spine (white diamond). Enlargement occurs within a few seconds and lasts over 2 h. (b) Slow intrinsic fluctuations in spine sizes occur in the presence of inhibitors of NMDA receptors (NMDAR) over periods of days. The spines show relatively small changes in size (quantified in Figure 2). Spine generation (spines 4, 6, 7) and elimination (spines 8 and 10) still occur frequently. (c) Relationship between structural plasticity and receptor trafficking, the two mechanisms of synaptic plasticity. Structural changes induced by plasticity can be maintained for a long time, enabling the functional states of spines to also be maintained for a long time—if there are structure–function relationships. Such relationships can be caused by the allocation of AMPA receptors in proportion to the spine size, resulting in the ‘‘readout of structure.’’ Trafficking of receptors can facilitate and hasten the readout processes. Figures in (a) and (b) are reproduced with permission from the Nature publishing group [21] and the Society for Neuroscience [42], respectively.
induction of LTP. Enlarged spines also explain the longterm maintenance of LTP, given that the number of functional AMPA receptors correlates with spine volume [14– 20] (Figure 1c). And spine enlargement [45,46], like the late phase of LTP and long-term memory itself [47], depends on protein synthesis. Moreover, spine structures are stable for days in cultured hippocampal slices (Figure 1b) [42,48] and for years in the cortex in vivo [31,32,34]. These data support the idea that structural plasticity is the central cellular mechanism that underlies memory formation. Receptor trafficking also affects functional plasticity by facilitating the readout of spine structure (Figure 1c). During LTP induction, AMPA receptors in cytosolic vesicles are inserted into the plasma membrane [49,50] and diffuse laterally to supply new receptors to the PSD [51,52]. These receptors are anchored to PSD proteins in an actin-dependent manner [53–55]. LTD protocols also affect the endocytosis of AMPA receptors [49]. Hence, numerous experiments support the link between LTP/LTD induction, receptor trafficking and the acquisition of long-term memory in behaving animals [49]. It is important to note, however, that the rapid exchange of AMPA receptors between synaptic and extrasynaptic regions [51] means that trafficking mechanisms by themselves cannot account for long-term changes in synaptic function. 122
Long-term intrinsic fluctuations and maintenance of spines Neuronal networks reflect the properties of spine populations, rather than those of a single spine, because the generation of action potentials in a neuron requires the activation of many synapses. We recently identified a key phenomenon that affects the long-term behaviors of spine populations. Through the observation and systematic quantification of spine dynamics over periods of days (Figure 1b) [42], it became apparent that spine volumes grew and shrank spontaneously. These volumetric changes occurred even when NMDA receptors and Na+ channels were completely blocked to prevent activity-dependent plasticity (Figure 2a). Such changes, termed ‘intrinsic fluctuations,’ encompass all phenomena (other than activity-dependent plasticity) that contribute to the structural dynamics of spines [42], and represent at least a part of fluctuations of spine volumes reported in vivo [38,61,62]. These fluctuations reflect an inevitable lack of structural stability in spines whose molecular components turn over with a time constant of less than 3 h [56,57]. Despite this constant change, the average daily volume change was close to zero for all spine sizes (Figure 2b). Thus, spine populations at various sizes are in equilibrium, and individual spines can act as analog memory elements despite the stochastic changes in spine volume (up to 20% per day in young hippocampus) (Figure 3c). At first glance, intrinsic fluctuations would seem incompatible with long-term memory storage. However, two additional findings reveal that this variability does not defeat the maintenance of spine function over time (Figure 1c). First, the volume of an average spine remains largely the same for a certain period of time [42], consistent with the persistence for weeks of LTP in vivo [42,63]. Second, spine lifetimes can be very long (Figure Ia in Box 1), in line with in vivo two-photon imaging [32,34] (for theory and examples, see Box 1). The high proportion of small spines in the volume distribution [64] (Figure 2d) can now be explained by a positive correlation between the size of the spine and the amplitude
Review
Trends in Neurosciences
Vol.33 No.3
Figure 2. Long-term structural dynamics of dendritic spines. (a) Fluctuations in the head volume of an individual spine as determined by the fluorescence intensity of the spine head. There are two mechanisms controlling the long-term dynamics of spine volumes: activity-dependent plasticity (A, blue arrows) and intrinsic fluctuations (I, black arrows). Activity-dependent plasticity is dependent on the activation of NMDA receptors, whereas intrinsic fluctuations exist even in the presence of inhibitors of NMDA receptors and Na+ channels. Activity-dependent plasticity and intrinsic fluctuations represent learning mechanisms and maintenance processes, respectively. The Langevin equations used to model the plasticity are shown. We assume that every spine larger than 0.02 mm3 has a presynaptic partner, although the presynaptic terminal is drawn only for the small spine. (b) Mean values of fluctuations in spine head volume in the presence (black, I) and in the absence (blue, A) of NMDA receptor inhibitors. (c) Standard deviations of fluctuations in spine head volume in the presence (black, I) and in the absence (green, C) of NMDA receptor inhibitors. The activity-dependent fraction is displayed by blue (A). (d) Probability–density distributions predicted from the C and I data given in (b, c). The prediction is well fit by the actual data for spine head volumes in slice cultures incubated for 3 days in the absence (C) or presence (I) of NMDA receptor inhibitors [42].
of its intrinsic fluctuations (Figure 2c, I). Small spines accumulate because of smaller fluctuations, and larger spines become less common because of their greater fluctuation amplitudes (Figure 2c) [42] (for a theoretical basis, see Box 1). Similar spine–volume distributions have been found in various conditions in vivo and in vitro, supporting the interpretation that the mechanisms are the same. In contrast to the magnitude of intrinsic fluctuations, the frequency of activity-dependent structural changes is greater in smaller spines (Figure Id in Box 1; for quantification of LTP, see Box 1). All spines are considered to receive a synaptic contact [20,42], and small spines correspond approximately to the notion of ‘‘silent synapses’’ [65] that express NMDA but not AMPA receptors [14,16,20]. Small spines display greater increases in cytosolic Ca2+ induced by NMDA receptor stimulation [16,66] and are preferential sites for LTP induction [21,65]. The structural changes that accompany LTP include the enlargement of small spines and an increased number of medium-sized spines (Figure 2d, C). Such effects on the volume distribution, however, were relatively weak (Figure 2d, compare I and C) because intrinsic fluctuations occur constantly and appear to be the dominant mechanism in determining spine–volume distributions. Large spines form only gradually, because activity-dependent enlargement acts preferentially on small, rather than medium, spines (Figure 2a) [42]. This simple fact predicts that older spines will tend to be larger (Figure Ib in Box 1) [42] and have longer life expectancies (Figure Ia in Box 1). Thus, the history of a spine is reflected in its volume, unlike a one-bit memory element in a computer, which cannot indicate its own history. Indeed, this history effect explains a feature of memory first described by Ebbinghaus in 1885—that older memories are more persistent [67]. In his seminal work, Ebbinghaus estimated the decay of memories by memorizing a list of nonsense syllables. After a variable period, he memorized the list again and quan-
tified memory as the decrease in memorization time. He found a rapid loss of memory in the first day, followed by a slower decline over the next 31 days. This non-exponential pattern of memory decay suggested that longer-lasting memories were more persistent. A graph of the time savings was fitted by a logarithmic curve (Figure Ic in Box 1, red dashed line) that was later called the ‘‘savings function’’ or ‘‘forgetting curve’’ [68]. Thus, the nature of memory depends on its history, which can be explained in turn by the history effect of the dendritic spine. If one posits that the creation or enlargement of a few small spines within Broca’s area represent memory traces for new syllables, then the idea of intrinsic fluctuations predicts a logarithmic decay in the mean volume of small spines (Figure Ic in Box 1, solid line) that is closely related to the savings function (red dashed line). At the same time, the model explains how a proportion of the memory encoded within small spines can be saved into larger spines as a result of intrinsic fluctuations (Figure 2a). Intrinsic fluctuations are random forces that alternately strengthen and destroy the smallest spines. But the stochastic elimination of existing spines (Figure 1b, spines 8 and 10), in turn, enables the spontaneous generation of new spines (Figure 1b, spines 4, 6, 7) by clearing space and recycling molecular resources. Indeed, the spontaneous creation of new spines continues into adulthood and is often detected [42] in the mature neocortex [35,37,61]. Although new spines start small (Figure 2a), they represent functional synapses that can readily enlarge in an activity-dependent manner (Figure 1a). The creation of new spines accelerates 20–60 min after neuronal activity [40,69], but the sprouting cannot be synapse-specific as there is no synapse before its generation, in principle. Thus, activity-dependent synapse formation seems to reflect considerable chance [48]. Perhaps the random nature of this make-and-test process enables animals to adapt to an unexpected environment. These new spines could be seeds of new memory [35,37,38]. 123
Review
Trends in Neurosciences Vol.33 No.3
Box 1. Population dynamics of dendritic spines To understand the population behaviors of spines, we have introduced a mathematical model in which spine volumes change continuously in a semi-random manner. The properties of the changes or fluctuations depend on spine volume (Figure 2a). Similar approaches have been used in the study of many other biological problems, such as inter-spike intervals [116], population genetics and ecology [117]. The simplest mathematical model for random continuous fluctuations is Brownian motion W(t) [116,117]. In this framework, a continuous random variable or stochastic process V(t) that has an average change or drift of m(V) and a standard deviation of s(V) is described as dV ðtÞ dW ðtÞ ¼ sðV ðtÞÞ þ mðV ðtÞÞ; (1) dt dt where W(t) represents the standard Brownian motion with a variance of 1/day [116,118]. We have applied this Langevin equation to the volume fluctuations of spines V(t) using standard deviation and drift values obtained from experiments (Figure 2b,c). We found that the actual volume distributions of spines in control (C) treatments and in the absence of activity-dependent plasticity (I) are well predicted by the model (Figure 2d) [42]. This model also accounts for spine elimination by defining it as the shrinkage of spines past a certain threshold value (0.02 mm3) (Figure 2a). Thus, we can predict that larger spine will have a longer life (Figure Ia in Box 1). The life expectancy of a spine with a volume of 0.3 mm3 is 57 days in the absence of activity-dependent plasticity, assuming that the intrinsic fluctuations have a coefficient of variation (CV=s/V) of 20%/day [42]. If the CV is reduced by a factor of v, the life expectancy is prolonged by a factor of v2. If CV is 5%/day, we can predict that the lifetime of a spine with a volume of 0.3 mm2 is 2 years, consistent with in vivo measurements in mice [34]. If the CV is 1%/ day, the life expectancy is approximately 64 years [42]. Thus, if spine volumes fluctuate by only 1%/day, they could account for the lifelong persistence of some spines in a human being. This model also predicts that on average, older spines will have a larger mean volume (Figure Ib in Box 1) and a longer life expectancy. The average time course of spine volume for spines with the initial volume of 0.021 mm3 can be obtained using the same calculation (Figure Ic in Box 1), which can be fitted with the same saving function (the dashed red line in Figure Ic in Box 1) that represents the time course of forgetting. We can interpret our data on the activity-dependent plasticity of spines using two opposing processes, the enlargement (LTP) and shrinkage (LTD) of spines. If one assumes that the unitary amplitudes
All the above studies are consistent with a model in which activity-dependent spine–volume changes regulate newmemory acquisition (by enlarging/stabilizing or eliminating the smallest spines) and existing-memory persistence (by changing volumes of spines) [70].Thus, activity-dependent plasticity selects memory content and modifies memory strength, supporting the random-generation-and-test model of new-memory acquisition. In contrast, intrinsic fluctuations in spine volume may change the strength of a memory but seldom affect its content. We will next discuss how spine dynamics can account for abnormal spine profiles in various neurological and psychiatric conditions. Abnormalities in spine dynamics and mental disorders: a working model Many clinical investigators have proposed that synapses are major sites of pathogenesis for mental disorders such as schizophrenia, autism and other conditions that show normal gross anatomy [7,71–73]. Here, we summarize the reports of dendritic spine abnormalities in these disorders and present a hypothesis for how these changes might yield such diverse symptoms. 124
for enlargement and shrinkage are qE and qS, respectively, and the rates of enlargement and shrinkage are lE and lS, respectively, the average change and standard deviation can be expressed as: m ¼ q E lE q S lS ;
s 2 ¼ qE2 lE þ qS2 lS :
(2)
Then, the rates of enlargement and shrinkage are expressed as: lE ¼
s 2 þ qS m ; qE ðq E þ q S Þ
lE ¼
s2 þ qS m : q E ðqE þ q S Þ
(3)
If one applies this equation to the activity-dependent component of fluctuations (Figure 2b,c, A) by assuming qE=qS=0.02, then the frequencies of enlargement and shrinkage are predicted as shown in Figure Id in Box 1.
Figure I. (a) Relationship between spine head volume and mean life expectancy of a spine according to model I. (b) Dependence on spine age of the mean head volume of a spine following model I. (c) Average time course of head volume changes for spines that obey model I and have initial head volumes of 0.021 mm3. The red dashed line represents the savings function, k/[(Log[at])c+k], where k=0.13, a=480 and c=1.2. Figures reproduced, with permission from the Society for Neuroscience, from Ref. [42]. (d) The predicted contributions of LTP and LTD to the activity-dependent fluctuations (A) obtained using Eq. (3) in Box 1.
Dendritic spines in mental retardation In human patients [7,74,75] and most (but not all) animal models of mental retardations [76–79], dendritic spines tend to be abnormally small and immature. Many types of mental retardation have been traced to genetic defects in scaffolding and adhesion molecules thought to maintain synapses [80,81]. Presuming that defects in scaffolding and adhesion proteins would affect intrinsic fluctuations, these abnormally small spines might represent an inability to maintain spine structure. In effect, the lack of these molecules unleashes intrinsic fluctuations (Figure 3) to erode large spines and cause a proliferation of small spines (Figure 2d), thereby masking the effects of activity-dependent plasticity on volume distribution [42]. The inability to preserve large spines created by activity-dependent plasticity prevents the accumulation of proper knowledge in the developmental period, potentially causing a deficit in general intelligence. Dendritic spines are abnormally small in mutant mice that lack a copy of the shank-1 gene, which encodes a major scaffolding protein in the spine [79]. In these mice, interestingly, spatial reference learning was enhanced,
Review
Trends in Neurosciences
Vol.33 No.3
Figure 3. Various forms of spine motility and mental disorders. There are three forms of spine motility: rapid, activity-dependent plasticity; long-term, activity-dependent plasticity; and long-term intrinsic fluctuations. Schizophrenia and mental retardation can arise from selective impairments in activity-dependent plasticity and intrinsic fluctuations, respectively. Intrinsic fluctuations or synaptic stability can also be impaired in autism.
although the memory after 4 weeks was much worse than controls [79]. One interpretation of these results could be that the overproliferation of small spines augments newmemory formation, whereas the lack of large spines indicates poor retention of older memories [80]. The superior performance of the mutant on the spatial learning task might represent a murine equivalent of the exceptional abilities that sometimes manifest in autistic savants. Dendritic spines in schizophrenia Unlike autism or mental retardation, schizophrenia is rarely linked to abnormalities in spine size [82]. This observation suggests that synaptic maintenance, as influenced by intrinsic fluctuations, remains unimpaired among schizophrenics (Figure 3). Instead, several lines of evidence point to links between schizophrenia and abnormal activity-dependent plasticity (Figure 3) [71,83]. The strongest such evidence is referred to as the NMDA hypothesis [84,85]. It stems from the observation that phencyclidine, a NMDA receptor antagonist, produces diverse schizophrenic symptoms that are ameliorated by NMDA receptor agonists such as D-serine. The dopamine dysregulation model, which relates to the classical dopamine hypothesis of schizophrenia, suggests that the dopamine-fueled elevation of intracellular cAMP plays an important role in NMDA-dependent synaptic plasticity [86]. Furthermore, the expression products of many schizophrenia-susceptibility genes are localized near glutamatergic synapses in a statistically significant manner (and can exist in other regions as well) [83,87] (Figure 3). Reduced spine density in the prefrontal cortex of schizophrenics [71,86] indicates a loss of balance between synaptogenesis and elimination [42,88]; both processes are strongly regulated by NMDA receptor transmission [89]. Therefore, the reduced density of dendritic spines in schizophrenic patients must represent a reduced generation relative to elimination of synapses [42]. Adult-onset schizophrenia can also develop in response to an inevitable post-adolescent decline in activity-dependent plasticity [90]. At this age, prodromal individuals who exhibit some schizophrenia-associated problems [91] already demonstrate a reduced capacity for activity-dependent synaptic plasticity. This growing deficit is often obscured, however, by high levels of synaptic remodeling at this age. When the decrease in activity-dependent plasticity is uncovered in adulthood, it is often accom-
panied by evidence of abnormal neural connectivity and the full onset of symptoms (Figure 3). Although positive symptoms such as delusions and hallucinations remain the major criteria for schizophrenia diagnosis, clinicians have given greater attention in recent decades to a range of cognitive deficits and negative symptoms [91]. Schizophrenics experience the loss of attention, working memory, episodic memory, verbal fluency, emotion and volition [91]—nearly every higher-order brain function is impaired except intelligence quotient [92], which can be spared because it is believed to reflect cognitive functions acquired during childhood and adolescence [93]. In the context of spine dynamics, two synergic mechanisms could explain the symptoms of schizophrenia. The first mechanism rests on the idea that normal spine dynamics regulate experience-dependent plasticity within neuronal networks [71,83]. By extension, abnormal spine dynamics can permit the chaotic reorganization of neuronal networks, culminating in schizophrenia. Accordingly, the disordered thoughts and infrequent brainwave synchrony has led to the characterization of schizophrenia as a disease of disconnected brain regions [94,95]. Particularly because spiny glutamatergic neurons and their projections are so prevalent in the affected brain regions, spine dysfunction could be the culprit in disrupting normal connectivity. But a second, complementary and more provocative possibility is that spine dynamics take part directly in cognition (Figure 4). Rapid structural dynamics of spines in cognition: a new hypothesis Within the brain, the coordinated firing of neurons in space and time underlies myriad functions, including the cognitive processes of perception, emotion and volition [8,96,97]. A major challenge in neuroscience is to delineate the physical conditions of the conscious brain [96,98], sometimes referred to as the neuronal correlates of consciousness [99]. Unfortunately, many studies of these correlates reduce every neuron to its action potential. This presents an incomplete picture of cognitive function, and indeed the brain is more than its ions. Recent experiments using optogenetic tools suggest that mechanisms other than spikes can participate in the creation of internal representations [100]. In this section, we attempt to identify connections between the 125
Review
Figure 4. Hypothetical consequences of synchronous activities in a neuronal network. These figures illustrate the positive feedback mechanism in which recursive spike chains trigger the enlargement of specific spines, which in turn facilitate the spike chains. Each neuron displays characteristic neuronal motility in response to a spike chain, and an assembly of neurons exhibits synchronous and complex neural motility. (a) Selective enlargement of a specific set of spines in a neuron (yellow) by distinct spike chains 1 and 2. (b) Characteristic neuronal motilities induced by various spike chains. (c) Possible consequences of spine enlargement on synaptic functions. Spine enlargement enhances AMPA receptor function, applies positive pressure on a presynaptic terminal and can enhance neurotransmitter release.
cognitive and synaptic neurosciences to suggest a new synaptic basis for cognitive function. The first connection is about attention. As a physical event, attention manifests as the synchronous firing of a neuronal population that responds preferentially to attended stimuli [97,101]. Synchronous firing can also provide a solution to the binding problem [102,103], in which the brain recognizes an object with diverse attributes as a unitary percept. Therefore, the question arises: How does the brain recognize synchronous firing among spatially scattered neurons? From synaptic physiology, we know that the synchronous firing of pre- and postsynaptic neurons induces rapid spine enlargement [45,104]. In this way, a single spine can encode information from both neurons, and the thousands of spines (Figure 4b) in a pyramidal neuron have the potential to detect synchronous spikes from thousands of other neurons scattered throughout the brain [45] (Figure 4a). The rapid structural dynamics could even reflect a specific spike chain (Figure 4a,b). Thus, a pyramidal neuron can detect and register the synchronous firing of neuronal assemblies. During cognitive processes, billions of spiny neurons show synchronous motilities in their spines. 126
Trends in Neurosciences Vol.33 No.3
The second connection concerns speed. The speed of thought is not instantaneous under any circumstances, but particularly not for cognitive processes. For example, visual perception takes at least 0.1 s of recursive network computation [96], and tactile perception and free will each require 0.5 s of electrical activity in the human cortex [105]. These delays are on the order of the delay in spine enlargement after repetitive stimulation (0.5–3 s). The slight lag is attributed to the Ca2+ triggering of actin polymerization; once begun, the extension of actin filaments occurs rapidly (0.03–1 mm/s) [43,106]. This rapid enlargement does not seem to rely on the same enzymatic reactions needed for long-term enlargement, which involve CaMKII, Kalirin7, Rac and PAK [43,58,107]. The third connection relates to the ordering of long-term changes. Just as cognition precedes the formation of explicit memory [108], rapid enlargement precedes long-term enlargement of spines (Figure 1a). Thus, if cognitive processes involve spine dynamics, they use the same substrates as memory, and effectively leave their traces for memory. The fourth connection explores unconsciousness. General anesthetics easily and completely dispel cognition. Spine enlargement can also be blocked by anesthesia, because rapid spine motility has never been described in an anesthetized animal under natural stimulation—not even when the neurons continued to fire action potentials [109,110]. In support of this consideration, volatile anesthetics are reported to interfere with actin organization in resting cells [111]. In addition, the cerebellum is not a direct source of consciousness [112], although the cerebellar Purkinje cells have prominent spines. This could be correlated with the fact that these cells lack NMDA receptors and thus would not show rapid activity-dependent enlargement. The fifth connection addresses similarities between the phenomena of cognition and structural dynamics. Cognitive processes are specific—meaning defined or individual or particular—but their ultimate origin and subsequent progression are stochastic [113]. Likewise, spine motilities—which represent connections between specific neurons—are very rich cellular behaviors caused by the complex interplay of chemical and physical forces from the cytoskeleton, plasma membrane and surrounding cells. They, too, are often stochastic. The sixth and last connection recognizes self-modification. Cognitive processes are self-modifying, giving rise to memory, emotion and executive functions [113]. Likewise, spine enlargement can dynamically alter the functions of neuronal networks. Although we know that enlargement is associated with increased functional expression of AMPA receptors [21], we suspect that there are other functional consequences of spine enlargement that await discovery. The massive actin polymerization generates an expansive force [43] that acts on surrounding tissues (Figure 4c), but to what effect? One possibility is that it exerts a mechanical force on neighboring spines to alter their functions as well [43]. Another possibility is that the mechanical force enhances presynaptic function. This hypothesis is based on evidence that physical forces or chemical stimuli applied to the plasma membrane promote
Review exocytosis [114,115]. If such forms of regulation can indeed be exerted in presynaptic terminals, then spines can actively maintain and trigger the firing of a particular neuronal network or spike chain through an increase in the release probability of presynaptic terminals. This type of regulation must be highly complex and stochastic, as described above. Thus, spine–volume changes could alter synaptic connections in various ways and rapidly selfmodify neuronal–network functions. The rapid, responsive movement of synapses shares many features with cognition. Dendritic spines can take part directly in cognitive processes to make them more individual, active and stochastic—unlike a computer, in which memory elements obey simple and deterministic rules. Thus, cognitive processes can be easier to understand when we take account of the spine structural dynamics. Direct imaging of spine motilities in vivo has the ability to substantiate these possibilities. Concluding remarks We have described the close relationship between spine structure and function, and introduced the hypothesis that intrinsic fluctuations in spine volumes account for the longterm maintenance of spines. The biophysical properties of these fluctuations could mirror the psychological properties of complex behaviors such as forgetting. Intrinsic fluctuations also predict the spontaneous generation of abundant new spines, leading to the random-generationand-test model of new-memory acquisition. We propose that abnormalities in the two types of spine structural plasticity—intrinsic fluctuations and activitydependent plasticity—are involved in the pathogenesis of mental retardation and schizophrenia, respectively. We also introduce the idea that rapid spine dynamics underlie cognition. These are testable hypotheses that can be examined quantitatively in the future. Regardless of those future results, the ability to visualize and manipulate spine dynamics in vivo will be useful in our investigations of cognitive processes and the synaptic bases of psychiatric disorders. Such investigations could provide new avenues for studying brain functions and can lead to novel diagnostic and therapeutic approaches for psychiatric disorders. Acknowledgements We thank M. Fukuda, K. Kasai, A. Sawa and K. Toyama for helpful discussions. This work was supported by Grants-in-Aids from the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan (H.K., J.N.).
References 1 Ramon y Cajal, S. (1995) Histology of the Nervous System of Man and Vertebrate, Oxford University Press 2 Greenough, W.T. and Bailey, C.H. (1988) The anatomy of a memory: convergence of results across a diversity of tests. Trends Neurosci. 11, 142–147 3 Coss, R.G. and Perkel, D.H. (1985) The function of dendritic spines: a review of theoretical issues. Behav. Neural Biol. 44, 151–185 4 Leiss, F. et al. (2009) Characterization of dendritic spines in the Drosophila central nervous system. Dev. Neurobiol. 69, 221–234 5 Dunaevsky, A. et al. (1999) Developmental regulation of spine motility in the mammalian central nervous system. Proc. Natl. Acad. Sci. U. S. A. 96, 13438–13443
Trends in Neurosciences
Vol.33 No.3
6 Matus, A. (2000) Actin-based plasticity in dendritic spines. Science 290, 754–758 7 Fiala, J.C. et al. (2002) Dendritic spine pathology: cause or consequence of neurological disorders? Brain Res. Rev. 39, 29–54 8 Hebb, D.O. (1949) The Organization of Behavior, John Wiley and Sons 9 Bliss, T.V. and Lomo, T. (1973) Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. J. Physiol. 232, 331–356 10 Harris, K.M. et al. (1992) Three-dimensional structure of dendritic spines and synapses in rat hippocampus (CA1) at postnatal day 15 and adult ages: implications for the maturation of synaptic physiology and long-term potentiation. J. Neurosci. 12, 2685–2705 11 Nusser, Z. et al. (1998) Cell type and pathway dependence of synaptic AMPA receptor number and variability in the hippocampus. Neuron 21, 545–559 12 Takumi, Y. et al. (1999) Different modes of expression of AMPA and NMDA receptors in hippocampal synapses. Nat. Neurosci. 2, 618–624 13 Kharazia, V.N. and Weinberg, R.J. (1999) Immunogold localization of AMPA and NMDA receptors in somatic sensory cortex of albino rat. J. Comp. Neurol. 412, 292–302 14 Matsuzaki, M. et al. (2001) Dendritic spine geometry is critical for AMPA receptor expression in hippocampal CA1 pyramidal neurons. Nat. Neurosci. 4, 1086–1092 15 Smith, M. et al. (2003) Mechanism of the distance-dependent scaling of Schaffer collateral synapse in CA1 pyramidal neurons. J. Physiol. (Lond.) 548, 245–258 16 Noguchi, J. et al. (2005) Spine-neck geometry determines NMDA receptor-dependent Ca2+ signaling in dendrites. Neuron 46, 609–622 17 Beique, J.C. et al. (2006) Synapse-specific regulation of AMPA receptor function by PSD-95. Proc. Natl. Acad. Sci. U. S. A. 103, 19535–19540 18 Asrican, B. et al. (2007) Synaptic strength of individual spines correlates with bound Ca2+-calmodulin-dependent kinase II. J. Neurosci. 27, 14007–14011 19 Holbro, N. et al. (2009) Differential distribution of endoplasmic reticulum controls metabotropic signaling and plasticity at hippocampal synapses. Proc. Natl. Acad. Sci. U. S. A. 106, 15055– 15060 20 Zito, K. et al. (2009) Rapid functional maturation of nascent dendritic spines. Neuron 61, 247–258 21 Matsuzaki, M. et al. (2004) Structural basis of long-term potentiation in single dendritic spines. Nature 429, 761–766 22 Okamoto, K. et al. (2004) Rapid and persistent modulation of actin dynamics regulates postsynaptic reorganization underlying bidirectional plasticity. Nat. Neurosci. 7, 1104–1112 23 Otmakhov, N. et al. (2004) Persistent accumulation of calcium/ calmodulin-dependent protein kinase II in dendritic spines after induction of NMDA receptor-dependent chemical long-term potentiation. J. Neurosci. 24, 9324–9331 24 Lang, C. et al. (2004) Transient expansion of synaptically connected dendritic spines upon induction of hippocampal long-term potentiation. Proc. Natl. Acad. Sci. U. S. A. 101, 16665–16670 25 Zhou, Q. et al. (2004) Shrinkage of dendritic spines associated with long-term depression of hippocampal synapses. Neuron 44, 749–757 26 Nagerl, U.V. et al. (2004) Bidirectional activity-dependent morphological plasticity in hippocampal neurons. Neuron 44, 759–767 27 Kopec, C.D. et al. (2006) Glutamate receptor exocytosis and spine enlargement during chemically induced long-term potentiation. J. Neurosci. 26, 2000–2009 28 Bastrikova, N. et al. (2008) Synapse elimination accompanies functional plasticity in hippocampal neurons. Proc. Natl. Acad. Sci. U. S. A. 105, 3123–3127 29 Segal, M. (2005) Dendritic spines and long-term plasticity. Nat. Rev. Neurosci. 6, 277–284 30 Okada, D. et al. (2009) Input-specific spine entry of soma-derived Vesl1S protein conforms to synaptic tagging. Science 324, 904–909 31 Trachtenberg, J.T. et al. (2002) Long-term in vivo imaging of experience-dependent synaptic plasticity in adult cortex. Nature 420, 788–794 32 Grutzendler, J. et al. (2002) Long-term dendritic spine stability in the adult cortex. Nature 420, 812–816 33 Stettler, D.D. et al. (2006) Axons and synaptic boutons are highly dynamic in adult visual cortex. Neuron 49, 877–887
127
Review 34 Zuo, Y. et al. (2005) Development of long-term dendritic spine stability in diverse regions of cerebral cortex. Neuron 46, 181–189 35 Yang, G. et al. (2009) Stably maintained dendritic spines are associated with lifelong memories. Nature 462, 920–924 36 Chklovskii, D.B. et al. (2004) Cortical rewiring and information storage. Nature 431, 782–788 37 Holtmaat, A. and Svoboda, K. (2009) Experience-dependent structural synaptic plasticity in the mammalian brain. Nat. Rev. Neurosci. 10, 647–658 38 Xu, T. et al. (2009) Rapid formation and selective stabilization of synapses for enduring motor memories. Nature 462, 915–919 39 Hofer, S.B. et al. (2009) Experience leaves a lasting structural trace in cortical circuits. Nature 457, 313–317 40 Engert, F. and Bonhoeffer, T. (1999) Dendritic spine changes associated with hippocampal long-term synaptic plasticity. Nature 399, 66–70 41 Lendvai, B. et al. (2000) Experience-dependent plasticity of dendritic spines in the developing rat barrel cortex in vivo. Nature 404, 876–881 42 Yasumatsu, N. et al. (2008) Principles of long-term dynamics of dendritic spines. J. Neurosci. 28, 13592–13608 43 Honkura, N. et al. (2008) The subspine organization of actin fibers regulates the structure and plasticity of dendritic spines. Neuron 57, 719–729 44 Gustafsson, B. and Wigstrom, H. (1990) Long-term potentiation in the hippocampal CA1 region: its induction and early temporal development. Prog. Brain Res. 83, 223–232 45 Tanaka, J. et al. (2008) Protein synthesis and neurotrophindependent structural plasticity of single dendritic spines. Science 319, 1683–1687 46 Yang, Y. et al. (2008) Spine expansion and stabilization associated with long-term potentiation. J. Neurosci. 28, 5740–5751 47 Kelleher, R.J., III et al. (2004) Translational regulatory mechanisms in persistent forms of synaptic plasticity. Neuron 44, 59–73 48 De, R.M. et al. (2008) LTP promotes a selective long-term stabilization and clustering of dendritic spines. PLoS Biol. 6, e219 49 Kessels, H.W. and Malinow, R. (2009) Synaptic AMPA receptor plasticity and behavior. Neuron 61, 340–350 50 Newpher, T.M. and Ehlers, M.D. (2009) Spine microdomains for postsynaptic signaling and plasticity. Trends Cell Biol. 19, 218– 227 51 Heine, M. et al. (2008) Surface mobility of postsynaptic AMPARs tunes synaptic transmission. Science 320, 201–205 52 Makino, H. and Malinow, R. (2009) AMPA receptor incorporation into synapses during LTP: the role of lateral movement and exocytosis. Neuron 64, 381–390 53 Lisman, J.E. and Zhabotinsky, A.M. (2001) A model of synaptic memory: a CaMKII/PP1 switch that potentiates transmission by organizing an AMPA receptor anchoring assembly. Neuron 31, 191–201 54 Nakagawa, T. et al. (2004) Quaternary structure, protein dynamics, and synaptic function of SAP97 controlled by L27 domain interactions. Neuron 44, 453–467 55 Kato, A.S. et al. (2008) AMPA receptor subunit-specific regulation by a distinct family of type II TARPs. Neuron 59, 986–996 56 Okabe, S. et al. (2001) Rapid redistribution of the postsynaptic density protein PSD-Zip45 (Homer 1c) and its differential regulation by NMDA receptors and calcium channels. J. Neurosci. 21, 9561–9571 57 Gray, N.W. et al. (2006) Rapid redistribution of synaptic PSD-95 in the neocortex in vivo. PLoS Biol. 4, e370 58 Lee, S.J. et al. (2009) Activation of CaMKII in single dendritic spines during long-term potentiation. Nature 458, 299–304 59 Bagal, A.A. et al. (2005) Long-term potentiation of exogenous glutamate responses at single dendritic spines. Proc. Natl. Acad. Sci. U. S. A. 102, 14434–14439 60 Wang, X.B. et al. (2007) Independent expression of synaptic and morphological plasticity associated with long-term depression. J. Neurosci. 27, 12419–12429 61 Zuo, Y. et al. (2005) Long-term sensory deprivation prevents dendritic spine loss in primary somatosensory cortex. Nature 436, 261–265 62 Holtmaat, A. et al. (2006) Experience-dependent and cell-type-specific spine growth in the neocortex. Nature 441, 979–983 63 Abraham, W.C. (2003) How long will long-term potentiation last? Philos. Trans. R. Soc. Lond. B Biol. Sci. 358, 735–744
128
Trends in Neurosciences Vol.33 No.3 64 Benavides-Piccione, R. et al. (2002) Cortical area and species differences in dendritic spine morphology. J. Neurocytol. 31, 337– 346 65 Malenka, R.C. and Nicoll, R.A. (1999) Long-term potentiation – a decade of progress? Science 285, 1870–1874 66 Sobczyk, A. et al. (2005) NMDA receptor subunit-dependent [Ca2+] signaling in individual hippocampal dendritic spines. J. Neurosci. 25, 6037–6046 67 Ebbinghaus, H. (1885) Uber das Gedachtnis, Dunker & Humbolt 68 Wixted, J.T. and Ebbesen, E.B. (1997) Genuine power curves in forgetting: a quantitative analysis of individual subject forgetting functions. Mem. Cognit. 25, 731–739 69 Maletic-Savatic, M. et al. (1999) Rapid dendritic morphogenesis in CA1 hippocampal dendrites induced by synaptic activity. Science 283, 1923–1927 70 Morita, S. et al. (2009) Generation, elimination and fluctuations of synapses in the cerebral cortex. Commun. Integr. Biol. 2, 1–4 71 Lewis, D.A. and Gonzalez-Burgos, G. (2008) Neuroplasticity of neocortical circuits in schizophrenia. Neuropsychopharmacology 33, 141–165 72 Sudhof, T.C. (2008) Neuroligins and neurexins link synaptic function to cognitive disease. Nature 455, 903–911 73 Harrison, P.J. (1999) The neuropathology of schizophrenia. A critical review of the data and their interpretation. Brain 122, 593–624 74 Purpura, D.P. (1974) Dendritic spine ‘‘dysgenesis’’ and mental retardation. Science 186, 1126–1128 75 Ramakers, G.J. (2002) Rho proteins, mental retardation and the cellular basis of cognition. Trends Neurosci. 25, 191–199 76 Nimchinsky, E.A. et al. (2001) Abnormal development of dendritic spines in FMR1 knock-out mice. J. Neurosci. 21, 5139–5146 77 Boda, B. et al. (2004) The mental retardation protein PAK3 contributes to synapse formation and plasticity in hippocampus. J. Neurosci. 24, 10816–10825 78 Hayashi, M.L. et al. (2004) Altered cortical synaptic morphology and impaired memory consolidation in forebrain-specific dominantnegative PAK transgenic mice. Neuron 42, 773–787 79 Hung, A.Y. et al. (2008) Smaller dendritic spines, weaker synaptic transmission, but enhanced spatial learning in mice lacking Shank1. J. Neurosci. 28, 1697–1708 80 Kasai, H. et al. (2003) Structure-stability-function relationships of dendritic spines. Trends Neurosci. 26, 360–368 81 Kasri, N.N. and Aelst, L.V. (2008) Rho-linked genes and neurological disorders. Pflugers Arch. 455, 787–797 82 Roberts, R.C. et al. (1996) Reduced striatal spine size in schizophrenia: a postmortem ultrastructural study. Neuroreport 7, 1214–1218 83 Hashimoto, R. et al. (2007) Postsynaptic density: a key convergent site for schizophrenia susceptibility factors and possible target for drug development. Drugs Today (Barc.) 43, 645–654 84 Moghaddam, B. (2003) Bringing order to the glutamate chaos in schizophrenia. Neuron 40, 881–884 85 Goff, D. and Coyle, J. (2001) The emerging role of glutamate in the pathophysiology and treatment of schizophrenia. Am. J. Psychiatry 158, 1367–1377 86 Keshavan, M.S. et al. (2008) Schizophrenia, ‘‘just the facts’’: what we know in 2008. Part 3: neurobiology. Schizophr. Res. 106, 89–107 87 Harrison, P.J. and West, V.A. (2006) Six degrees of separation: on the prior probability that schizophrenia susceptibility genes converge on synapses, glutamate and NMDA receptors. Mol. Psychiatry 11, 981– 983 88 Alvarez, V.A. and Sabatini, B.L. (2007) Anatomical and physiological plasticity of dendritic spines. Annu. Rev. Neurosci. 30, 79–97 89 Colonnese, M. and Constantine-Paton, M. (2006) The emerging role of glutamate in the pathophysiology and treatment of schizophrenia. J. Comp. Neurol. 494, 738–751 90 Holtmaat, A.J. et al. (2005) Transient and persistent dendritic spines in the neocortex in vivo. Neuron 45, 279–291 91 Tandon, R. et al. (2009) Schizophrenia, ‘‘just the facts’’ 4. Clinical features and conceptualization. Schizophr. Res. 110, 1–23 92 Woodberry, K.A. et al. (2008) Premorbid IQ in schizophrenia: a metaanalytic review. Am. J. Psychiatry 165, 579–587 93 Plomin, R. and Spinath, F. (2004) Intelligence: genetics, genes, and genomics. J. Pers. Soc. Psychol. 86, 112–129
Review 94 McGlashan, T. and Hoffman, R. (2000) Schizophrenia as a disorder of developmentally reduced synaptic connectivity. Arch. Gen. Psychiatry 57, 637–648 95 Whalley, H. et al. (2009) Connecting the brain and new drug targets for schizophrenia. Curr. Pharm. Des. 15, 2615–2631 96 Lamme, V.A. (2003) Why visual attention and awareness are different. Trends Cogn. Sci. 7, 12–18 97 Sakamoto, K. et al. (2008) Discharge synchrony during the transition of behavioral goal representations encoded by discharge rates of prefrontal neurons. Cereb. Cortex 18, 2036–2045 98 Searle, J. (2004) Mind: A Brief Introduction, Oxford University Press 99 Crick, F. and Koch, C. (1998) Consciousness and neuroscience. Cereb. Cortex 8, 97–107 100 Airan, R.D. et al. (2009) Temporally precise in vivo control of intracellular signalling. Nature 458, 1025–1029 101 Gregoriou, G.G. et al. (2009) High-frequency, long-range coupling between prefrontal and visual cortex during attention. Science 324, 1207–1210 102 Gray, C.M. (1999) The temporal correlation hypothesis of visual feature integration: still alive and well. Neuron 24, 31–47 103 von der Malsburg, C. (1999) The what and why of binding: the modeler’s perspective. Neuron 24, 95–104 104 Harvey, C.D. and Svoboda, K. (2007) Locally dynamic synaptic learning rules in pyramidal neuron dendrites. Nature 450, 1195–1200 105 Libet, B. (2004) Mind Time: The Temporal Factor in Consciousness, Harvard University Press
Trends in Neurosciences
Vol.33 No.3
106 Pollard, T.D. and Borisy, G.G. (2003) Cellular motility driven by assembly and disassembly of actin filaments. Cell 112, 453–465 107 Penzes, P. and Jones, K.A. (2008) Dendritic spine dynamics – a key role for kalirin-7. Trends Neurosci. 31, 419–427 108 Dudai, Y. (2002) Memory from A to Z, Oxford University Press 109 Hubel, D.H. and Wiesel, T.N. (1959) Receptive fields of single neurones in the cat’s striate cortex. J. Physiol. 148, 574–591 110 Metin, C. et al. (1988) The primary visual cortex in the mouse: receptive field properties and functional organization. Exp. Brain Res. 69, 594–612 111 Kaech, S. et al. (1999) Volatile anesthetics block actin-based motility in dendritic spines. Proc. Natl. Acad. Sci. U. S. A. 96, 10433–10437 112 Penfield, W. (1975) The Mystery of Mind, Princeton University Press 113 Block, N. (1996) How can we find the neural correlate of consciousness? Trends Neurosci. 19, 456–459 114 Rosenmund, C. and Stevens, C.F. (1996) Definition of the readily releasable pool of vesicles at hippocampal synapses. Neuron 16, 1197– 1207 115 Kishimoto, T. et al. (2006) Vacuolar sequential exocytosis of large dense-core vesicles in adrenal medulla. EMBO J. 25, 673–682 116 Tuckwell, H.C. (1988) In Introduction to Theoretical Neurobiology (Vol. 2), Cambridge University Press 117 Kloeden, P.E. and Platen, E. (1999) Numerical Solution of Stochastic Differential Equations, Springer-Verlag 118 Risken, H. (1989) The Fokker–Planck Equation, Springer
129