360 Review TRENDS in Neurosciences Vol.26 No.7 July 2003 Structure –stability –function relationships of dendritic spines Haruo Kasai1, Masanori Matsuzaki1, Jun Noguchi1, Nobuaki Yasumatsu1 and Hiroyuki Nakahara2 1 Department of Cell Physiology, National Institute for Physiological Sciences and The Graduate University for Advanced Studies (SOKENDAI), Okazaki 444-8585, Japan 2 Laboratory for Mathematical Neuroscience, RIKEN Brain Science Institute, Wako, Saitama 351-0198, Japan Dendritic spines, which receive most of the excitatory synaptic input in the cerebral cortex, are heterogeneous with regard to their structure, stability and function. Spines with large heads are stable, express large numbers of AMPA-type glutamate receptors, and contribute to strong synaptic connections. By contrast, spines with small heads are motile and unstable and contribute to weak or silent synaptic connections. Their structure –stability –function relationships suggest that large and small spines are ‘memory spines’ and ‘learning spines’, respectively. Given that turnover of glutamate receptors is rapid, spine structure and the underlying organization of the actin cytoskeleton are likely to be major determinants of fast synaptic transmission and, therefore, are likely to provide a physical basis for memory in cortical neuronal networks. Characterization of supramolecular complexes responsible for synaptic memory and learning is key to the understanding of brain function and disease. A prominent feature of neurons is the huge number of contacts that each forms with other neurons via axons and dendrites. Such connections, known as synapses, have long been considered the site of memory storage in the brain [1,2]. Excitatory synapses in particular often form at small appendages of dendrites known as spines. The substantial structural diversity of dendritic spines has attracted the attention of neuroscientists ever since these structures were discovered in the 19th century [3]. It is only recently, however, that modern biophysical techniques have begun to reveal the structure – function relationships of dendritic spines. This review summarizes the most recent progress in our understanding of such relationships for pyramidal neurons of the cerebral cortex, especially with regard to the insight obtained by twophoton excitation-induced uncaging of a caged-glutamate compound. In addition, it addresses current working hypotheses and missing links in this important field of neuroscience. Spine structure and stability Many ultrastructural investigations have described morphological changes in neurons that accompany neuronal Corresponding author: Haruo Kasai ([email protected]). activity. In particular, increases in the number [4] or volume [5] of dendritic spines or changes in spine shape [6] have been observed. It is not possible, however, to study the stability and function of such plasticity by electron microscopy. Recent progress in two-photon excitation imaging has, thus, been largely responsible for the demonstration that the structure of central synapses is plastic both during neuronal activity and at rest [7]. Dendritic spines can take various shapes and have been classified on the basis of their structure as filopodial, thin, stubby, fenestrated or mushroom-like [8]. As a first approximation, in this review spines will be divided into two groups, small spines (filopodial and thin spines) and large spines (stubby, fenestrated, and mushroom spines), representing spines with small and large heads, respectively. Small spines change their form rapidly [9– 11], either disappearing or growing into large spines. New spines are also generated and eliminated during the intense neuronal activity that accompanies the induction of LTP [12,13]. These observations support the idea that structural alterations of spines underlie adaptive and learning processes. By contrast, large spines are relatively stable in vitro [9] and survive for more than a month [10] or even for a year [11] in the mouse neocortex in vivo. Close examination reveals that even large spines show some motility [14,15]. For example, such spines often extend small spinules from the spine head [16]. However, large spines remain within the category of large spines for long periods of time even in vivo [9 – 11], suggesting that memory is maintained in a structural form for extended periods in the brain. An important limitation of these studies is that they do not directly address the correlation between spine structure and function. The most important function of dendritic spines is to sense glutamate released from presynaptic nerve terminals (Fig. 1a). The structure – function relationships of spines were first addressed with the use of immunogold labeling of AMPA-sensitive glutamate receptors, which mediate the fast component of glutamate-mediated synaptic transmission. This approach revealed that spines with large postsynaptic densities (PSDs) tended to exhibit a higher level of AMPA-receptor immunoreactivity than did those with smaller PSDs in both the hippocampus [17,18] and http://tins.trends.com 0166-2236/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0166-2236(03)00162-0 Review 361 TRENDS in Neurosciences Vol.26 No.7 July 2003 (a) (c) (b) MNI-glutamate (f) 10 pA 1 ms 10 pA 1 ms (e) 30 FWHM= 0.45 µm 0.29 µm 20 10 0 –5 –0.4 0.0 0.4 x distance (µm) Current (pA) Fluorescence (A.U.) Current (pA) Fluorescence (A.U.) (d) 1 µm (g) 30 FWHM= 1.10 µm 0.89 µm 20 10 0 –5 –3 –2 –1 0 1 2 3 z distance (µm) 0 5 10 20 40 pA TRENDS in Neurosciences Fig. 1. Two-photon excitation of methoxy-nitroindolino-glutamate (MNI-glutamate) in hippocampal neurons. (a,b) Schematic representations of glutamate release either from a synaptic terminal (a) or by two-photon excitation of a caged-glutamate compound (b) at a single spine on a dendrite. The current traces (dotted lines) were recorded from the postsynaptic neuron; (a) shows an average of ten miniature excitatory postsynaptic currents (mEPSCs) for a cell in which such currents were rapid and (b) shows an average of 28 current traces obtained from the same cell in response to photolysis of MNI-glutamate (for 50 ms with 7 mW of light at a wavelength of 720 nm; indicated by red shading). The solid lines were constructed as described [22]. (c) Recording configuration for mapping of glutamate sensitivity along a dendrite using two-photon uncaging of the caged-glutamate compound. (d,e) Amplitudes of two-photon excitation-induced currents (green dots) along horizontal (d) or vertical (e) lines crossing the center of an AMPA-receptor cluster, and fluorescence profiles of a 0.1-mm-diameter bead (black dots) along the horizontal (d) or vertical (e) axes. The black lines (fluorescence) represent Gaussian fitting of the data, and the green lines (current) are predicted from a theory described previously [22]. The full-width-at-half-maximal diameters (FWHM) represent the spatial resolutions of the optical system (black) and of AMPA-receptor activation (green). Abbreviation: A.U., arbitrary units. (f,g) Fluorescence image of a CA1 pyramidal neuron in a rat hippocampal slice preparation (f) and the corresponding map of glutamate sensitivity at the dendritic surface (g). Numbers in (f) indicate the spines whose glutamate sensitivities are shown in Fig. 2m. Amplitudes of glutamate-induced currents are represented in (g) by the pseudo-color code shown at the bottom of this panel; the map was smoothed by linear interpolation. Panels (d– g) are modified, with permission, from Ref. [22], q (2001) Nature Publishing Group (http://www.nature.com/). primary sensory cortex [19]. Given that the size of PSDs is correlated with that of spine heads [20], these studies suggested that large spines express a greater number of glutamate receptors than do small spines. The relationship between AMPA-receptor immunoreactivity and spine function, however, remained to be clarified. Glutamate application with a femtosecond-pulse laser Investigation of the function of individual spines requires the systematic application of glutamate to identifiable spines along a dendrite. Such studies have been impossible with methodologies that rely only on microelectrodes because brain tissue is too compact to allow the arbitrary movement of an electrode with a resolution at the micrometer level (Fig. 1c). The light-induced release of glutamate from caged compounds was, therefore, applied as a solution to this problem. The spatial resolution of this approach with a conventional light source was, however, limited (axial resolution ,20 mm) by activation of the caged compound outside of the focal plane [21]. A femtosecond-pulse laser was therefore applied to achieve http://tins.trends.com the photolysis of caged-glutamate compounds by twophoton excitation within a small focal point of the objective lens (Fig. 1b) by Matsuzaki, Ellis-Davies, Nemoto, Miyashita, Iino and Kasai [22]. The major obstacle to this approach was the lack of a caged-glutamate compound with sufficient two-photon absorption. To address this problem, Matsuzaki et al. synthesized a new compound, methoxy-nitroindolino-glutamate (MNI-glutamate), that exhibits a cross-section of two-photon action sufficiently large for application to brain tissue; that is, MNIglutamate releases the neurotransmitter at a laser power that does not damage neurons. The same compound was independently synthesized for one-photon application by Papageorgiou and Corrie [23] and later shown to undergo photolysis within 1 ms [24]. MNI-glutamate is now commercially available. The currents induced by focal two-photon uncaging of MNI-glutamate (Fig. 1b) were almost identical to miniature excitatory postsynaptic currents (mEPSCs) (Fig. 1a). The laser-induced glutamate release also occurred within a space restricted to 0.29 mm laterally (Fig. 1d) and TRENDS in Neurosciences Vol.26 No.7 July 2003 0.89 mm axially (Fig. 1e), resulting in the focal activation of AMPA receptors with a resolution of 0.45 mm laterally (Fig. 1d) and 1.10 mm axially (Fig. 1e) (see following discussion for further explanation). MNI-glutamate does not manifest any antagonistic or agonistic effects at glutamate receptors, even at concentrations as high as 10 mM [22]. Thus, two-photon uncaging of glutamate provides a means to achieve rapid and fine 3D spatial control of glutamate concentration, to mimic glutamate release from presynaptic terminals. Spine structure and function Matsuzaki et al. attempted to measure glutamate sensitivity at the surface of dendrites of CA1 pyramidal neurons in fresh slice preparations obtained from the rat hippocampus [22]. The two-photon uncaging of MNI-glutamate was induced at 1064 points in a cubic region encompassing a small portion of a dendrite (Fig. 1f). Sampling points were pseudo-randomized to minimize receptor desensitization. The glutamate-induced currents were recorded at the soma, and their peak amplitudes were displayed by color-coding (Fig. 1g; for simplicity, only those data obtained from photolysis at the surface of the dendrite are shown). Glutamate sensitivity was found to vary markedly among individual spines: whereas some spines exhibited pronounced glutamate sensitivity (spines 3 and 8 in Fig. 1f), the sensitivity of others was relatively low (spines 1, 2 and 6). This was apparently the first experiment in which synaptic function was systematically quantified along the dendrites of central neurons. Glutamate sensitivity was found to be highly correlated with spine structure [22]. Data obtained from a total of nine dendrites thus revealed that glutamate sensitivity was highest at spines with the largest heads (Fig. 2), with thin spines and filopodia exhibiting only a low sensitivity (Figs 1g, 2k). Plots of glutamate sensitivity versus spine-head volume yielded correlation coefficients of between 0.7 and 0.9 (mean ^ SD , 0.8 ^ 0.07; n ¼ 9 dendrites). The actual correlation coefficient must be . 0.8, however, given the stochastic variability in the number of open channels. This relationship between glutamate sensitivity and spinehead volume was recently confirmed by another laboratory [25], in a study in which currents were recorded from the dendritic shaft rather than at the soma. These results were also consistent with those of previous anatomical studies [17,18], suggesting that immunoreactive AMPA receptors in PSDs are functional. The structure – function relationship suggests that, if LTP is long-lasting at the level of the individual synapse, then the affected spines should also be stable over the same period. Long-term stability of large spines has recently been supported by experimental evidence [10,11]. The structure – function relationship also predicts that, if postsynaptic glutamate sensitivity is increased by LTP, this effect should be accompanied by spine enlargement, at least in the long run. The latter prediction is likely to be directly addressed by two-photon uncaging experiments that induce synaptic plasticity. Synapses on small spines and dendritic shafts [26] might represent ‘silent synapses’ that express NMDA-sensitive http://tins.trends.com (a) (b) (c) (d) (e) (k) (f) (g) (h) (i) (j) (l) 0 5 10 20 40 pA (m) 3 1 µm (n) 20 4 10 2 0 0.00 8 5 9 7 6 Autocorrelation coefficient Review Current (pA) 362 1 10 1 0.5 0 -0.5 0.04 0.08 Spine-head volume (µm3) 0 2 4 6 8 10 Spine separation 0.0 1.70 3.30 4.80 6.40 8.0 Mean distance (µm) TRENDS in Neurosciences Fig. 2. Spine structure and expression of functional AMPA receptors. (a –l) Fluorescence images (upper panels) and glutamate-sensitivity maps (lower panels) for various dendritic spines of CA1 pyramidal neurons in rat hippocampal slices. The former were obtained from stacked images containing the respective spine and the latter were obtained from one x–y section (in which the maximal glutamate sensitivity of the spine was detected) and were smoothed by linear interpolation. White lines indicate the contours of the dendritic structures. Representative data are from thin spines (a–e), large spines (f– j) and a filopodial spine (k) in slices prepared from 15–22-day-old animals and from a large spine in a 9-day-old animal (l). Amplitudes of glutamate-induced currents are represented by the pseudo-color code shown. (m) Correlation between spine-head volume and glutamate sensitivity for the dendrite shown in Fig. 1f; numbered points correspond to the numbered spines. (n) Spatial autocorrelation of glutamate sensitivity (blue) and of spine-head volume (red) for the spines shown in Fig. 1f. The autocorrelaton coefficient is obtained from the equation X ðaðiÞ 2 aÞðaði þ jÞ 2 aÞ i ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi RðjÞ ¼ rX X ðaðiÞ 2 aÞ2 ðaði þ jÞ 2 aÞ2 i i where aðiÞ represents either the maximal amplitude of the glutamate-induced current or the spine-head volume of the ith spine, a is the mean value of aðiÞ for all spines and j is the spine separation. The mean distances of two spines with the spine separation of j are also shown. The correlations were obtained from nine dendrites. Modified, with permission, from Ref. [22], q (2001) Nature Publishing Group (http://www.nature.com/). glutamate receptors rather than AMPA-sensitive ones and that are preferential sites for the induction of LTP [27]. The proportion or even the existence of silent synapses in the developing synaptic network, however, remains controversial [28]. This issue is also likely to be reinvestigated with two-photon uncaging experiments that simultaneously reveal the expression of AMPA receptors and NMDA receptors. The results of Matsuzaki et al. [22] also suggest that the decrease in the proportion of large spines apparent in Review TRENDS in Neurosciences Vol.26 No.7 July 2003 mentally retarded subjects [29] is likely to reflect a decreased proportion of substantial and stable synaptic connections. Large spines receive synaptic input from presynaptic terminals with a larger diameter and greater number of synaptic vesicles [20,30] compared with the presynaptic terminals associated with small spines. These observations suggest that the structure and function of synapses are regulated coordinately in presynaptic and postsynaptic components. The spatial distributions of different types of synapse warrant detailed investigation in future anatomical studies of the brain [29,31]. Memory density Given that glutamate-mediated synaptic transmission depends on a diffusible molecule, the length of glutamate action is a crucial determinant of the practical density of synaptic connections. The length of glutamate action is, in turn, dependent on the gating properties of postsynaptic AMPA receptors. If the activation of AMPA receptors is slow, then glutamate released from a presynaptic terminal will reach neighboring synapses before the corresponding postsynaptic receptors are activated. Such a spillover effect of glutamate on AMPA receptors has been mainly addressed theoretically on the basis of kinetic values obtained from macroscopic analysis [32]. With the use twophoton uncaging of glutamate, Matsuzaki et al. were able to narrow the application of transmitter to a lateral region as small as 0.29 mm (Fig. 1d) and to examine the spread of AMPA-receptor activation. Such analysis yielded a value of 0.45 mm for the lateral diameter of the region of AMPA-receptor activation [22], which is consistent with the theoretical prediction [32] and confirms that the area of activation is crucially governed by the gating of AMPA receptors. Given that spine density is , 1 – 3 mm21 in the cerebral cortex [20,31,32], the gating of AMPA receptors is marginally sufficient to confer independent sensing by spines of the input from glutamatergic afferents. It has been suggested that complex cognitive functions require an increased spine density in the cortex [31]. With the use of statistical analysis of the glutamate sensitivities of neighboring spines, Matsuzaki et al. next investigated whether the expression of AMPA receptors is controlled independently in such spines (Fig. 2n) [22]. They found that the statistical correlation between the glutamate sensitivities of neighboring spines was weak or nonexistent. This observation is, thus, consistent with a tenet of spine function: that the narrow neck of spines isolates their metabolism and allows independent control of their functions [3,33]. LTP was shown to spread . 70 mm [34] after stimulation of a bundle of presynaptic fibers, with an overall diameter of 30 mm. It will thus be of interest to determine the spread of LTP after stimulation of individual spines. Finally, Matsuzaki et al. determined how many AMPA receptors were expressed in single spines. Two-photon excitation allowed them to apply constant amounts of glutamate repetitively to a postsynaptic membrane in slice preparations and to determine the number of AMPA receptors on the basis of non-stationary fluctuation analysis [22]. Such analysis revealed that individual spines in CA1 pyramidal neurons express at most 150 http://tins.trends.com 363 AMPA receptors on the proximal dendrite [22]. Using the same technique, Smith et al. reported that the actual number of receptors depended on spine location on the dendrite, being larger for distal spines (mean ¼ 171) than for proximal spines (mean ¼ 91) [25], an observation that might be related to the distance scaling of synaptic input to CA1 pyramidal neurons. Given that AMPA-receptormediated currents can only summate within a spine, any number of AMPA receptors between 0 and 150 is able to code at most 7.2 (log2 150) bits of information. Neuronal networks with all-or-nothing synaptic connections might be nearly as efficient as those with analog connections [35,36]. How the brain utilizes the memory capacity of individual spines will be an important topic for future investigation. Currently available memory devices have memory elements with a spatial dimension in the sub-micrometer range, as is the case with cortical neuronal networks. Photographic film, for example, stores memory in the form of silver aggregates with a minimum spacing of ,1 mm21. Large-capacity hard disks of computers store memory in magnetic form at 10 bits mm21. And holographic memory is able to store information with a resolution of 0.3 mm in a cube, corresponding to a memory capacity of 27 Tbyte cm23. Holographic memory demonstrates that data can be stored in a highly distributed manner in memory devices as it can in the brain. Although algorithms of storage and readout for functional sets of memories differ markedly, memory capacity should ultimately depend on the density of memory elements both in memory devices and in the brain. The high density of synaptic memory, akin to that of currently available memory devices, and the absence of other similarly dense brain structures support the classical notion that the synapse is the major memory element in the cerebral cortex. Implications of structure –stability –function relationships of spines In memory devices, memory storage and readout occur in two different operational modes. New memory thus cannot be stored in the readout mode. By contrast, storage and readout of memory occur by inseparable processes in the brain. This feature has been a major theme for the operation of realistic neuronal networks and it has been ascribed mainly to the highly distributed nature of memory in neuronal networks and in the superposition of memories, distributed over many synapses or ‘weights’, as in holographic memory. One important question then arises: how are both the robustness of readout and the rapid acquisition of new memory achieved? The structure – stability –function relationships of dendritic spines provide new insight into this important issue. For simplicity, we consider only two categories of spines, large and small. First, large spines are stable, or ‘writeprotected’, and maintain preexisting strong connections with little interference from new memory formation. By contrast, small spines are unstable, or ‘write-enabled’, and are mostly responsible for the acquisition of memory, which is possibly mediated by their transformation into large spines. In this manner, the robustness of readout and the rapid acquisition of new memory can be achieved 364 Review TRENDS in Neurosciences Vol.26 No.7 July 2003 together. Second, large spines remain stable even during rest or sleep, whereas small spines can be eliminated. Thus, the lifetime of synaptic memory can vary widely depending on spine structure, with large spines being the structural basis for long-term memory. Finally, small spines are generated during activity-dependent processes or even at rest, providing an inexhaustible source of new synapses and memory. Exploration of neuronal network models that take into account these structure– stability – function relationships of synapses should provide insight into the operation of realistic neuronal networks that experience various behavioral states in the long term. In addition, the structure –stability – function relationships of synapses should be further verified and specified experimentally and their molecular basis elucidated. Molecular basis of spine structure –stability– function relationships The trafficking and turnover of AMPA receptors are rapid [24,37 – 40] and have been proposed to account for the changes in synaptic strength during LTP [41] and LTD [42]. By contrast, given that AMPA-receptor expression is dynamically regulated in spines, the maintenance of memory must depend on stable factors that regulate AMPA-receptor expression; otherwise, the strength of synaptic connections could not be maintained in the long term. Such factors might include spine shape and the underlying cytoskeletal organization, given that functional AMPA-receptor expression correlates with spinehead volume (Fig. 2m) and that large spines are stable. Synaptic structure, including spine-head volume, is thus a good candidate for storage of synaptic memory. It is still possible that certain molecules might particularly contribute to memory storage. Phosphorylated Ca2þ/calmodulin-dependent protein-kinase II (CaMKII) is a candidate to be such a molecule [43], although kinase inhibitors have been found not to block the maintenance phase of LTP [43,44]. Critical assessment of these possibilities should be feasible after the supramolecular organization of synaptic memory has been clarified further. On the basis of the notion that synaptic structure determines function and stability, the molecular control of synaptic strength is likely to encompass the following four aspects: (1) the genesis, motility, enlargement and elimination of small spines, which underlie effective acquisition of new memory; (2) the stability of large spines, which is responsible for the maintenance of longterm memory; (3) the expression of AMPA receptors, which determines the amplitude of rapid synaptic transmission; and (4) the expression of NMDA receptors, which allow the passage of Ca2þ and contribute to synaptic reorganization. Spine instability Small spines are motile, readily generated and eliminated, and undergo enlargement into large spines. Various molecular mechanisms might be responsible for such instability. As predicted by Crick [45], cytoskeletal elements are abundant in dendritic spines [46,47]. Indeed, actin is more concentrated in spines than in other neuronal structures [48]. Regulation of the actin cytoskeleton contributes to many cellular functions [49] and is http://tins.trends.com fundamental to synaptogenesis and spine motility [19,50 – 53]. Spontaneous spine motility requires actin and is blocked by volatile anesthetics [54], suggesting that it might play an important role in the organization of neuronal networks. Actin organization is regulated by small GTPases of the Rho family (Rho, Rac and Cdc42), and these proteins also contribute to the determination of spine morphology [55] (Fig. 3a). The protein encoded by the gene mutated in individuals with fragile-X mental retardation (FMRP) is thought to regulate Rac1 function, and impairment of such regulation might be responsible for the long, thin and tortuous dendritic spines characteristic of this disease [56]. In addition, LIMK-1, an inhibitor of cofilin/ADF [a factor that induces depolymerization of filamentous actin (F-actin)] is implicated in Williams syndrome, which is characterized by profound cognitive deficits [57]. Abnormalities of dendritic spines have also been described in many other mental disorders [29], although it is difficult to rule out the possibility that they are the result, rather than the cause, of pathological neuronal activity. Ca2þ signaling is thought to play an important role in the activity-dependent genesis and motility of small spines [12,58], as well as in the induction of LTP. The potential targets of Ca2þ in such signaling include calmodulin, CaMKII, adenylate cyclase [39], Ras and Rap [42,59], and mitogen-activated protein kinase (MAPK) pathways [42,60] (Fig. 3a). The signaling mechanisms that underlie the conversion of small spines to larger ones have not yet been elucidated. It is possible that the level of Ca2þ signaling is greater in small spines than in large spines, as a result of the smaller head volume of the former, resulting in greater instability. In addition, the small head volume mechanically facilitates the instability of small spines. Spine stability Large spines remain within the category of large spines for long periods of time [9– 11], even though all the molecules that contribute to spine stability turn over within days or weeks [61 – 63]. Several possible mechanisms might underlie the stability of large spines. First, existing clusters of F-actin might exert a positive feedback effect on actin polymerization, cross-linking and stability (Fig. 3b,A). Indeed, F-actin anchors many protein complexes of PSDs, including the CaMKIIa– densin-180 – a-actinin complex [64], the PSD95 – GKAP (a guanylatekinase-associated protein)–shank–cortactin complex [65], and CaMKIIb [66]. Moreover, certain PSD proteins possess enzymatic activities that affect F-actin organization, including SPAR (a spine-associated GTPase-activating protein for Rap) [67], kalirin (a guanine-nucleotideexchange factor for Rac1) [68] and spinophilin (which is also called neurabinII and has actin-bundling activity) [69]. Overexpression of SPAR [67], PSD95 [70] or shank [65] in cultured neurons has also been shown to increase spine volume. Thus, positive feedback mechanisms appear to maintain a large spine size despite the potential rapid turnover of actin [63]. Some actin filaments in F-actin clusters are also capped and colocalize with other structural proteins [51,71] such as drebrin [72]. The small surface-to-volume ratio of large F-actin clusters Review 365 TRENDS in Neurosciences Vol.26 No.7 July 2003 (a) Spine instability (b) Spine stability B CaMK A NR CaM SPRCs F-actin Actin organizers Scaffold proteins Ca/CaM-binding proteins NMDAR-binding scaffold proteins Adhesion molecules GR AMPARs NR NMDARs (c) AMPAR expression GR GR C CaMK Ribosome Actin-capping proteins PSD (d) NMDAR expression NR NR GR TRENDS in Neurosciences Fig. 3. Spine structure– stability– function relationships. (a) Spine instability. The cytoskeleton, various actin-organizing enzymes and Ca2þ/calmodulin (Ca/CaM)-binding proteins contribute to spine genesis, motility, enlargement and elimination. (b) Spine stability. Hypothetical mechanisms of spine stability include the positive feedback effects via filamentous actin (F-actin) organization within the spine (A), via spine interaction with the presynaptic terminal (B), and via protein synthesis inside or close to a spine (C). (c) AMPA-receptor (AMPAR or GR) expression proportional to spine-head volume. The actin cytoskeleton and scaffold proteins might be major molecular determinants of AMPA-receptor expression. (d) NMDA-receptor (NMDAR or NR) expression is independent of F-actin. Although many scaffold proteins bind tightly to NMDA receptors, NMDA-receptor expression is not proportional to postsynaptic density (PSD) size. White circles represent presynaptic vesicles of glutamate. Abbreviations: CaMK, Ca2þ/calmodulin-dependent-kinase; SPRCs, synaptic polyribosome complexes. might render them less sensitive to environmental changes and prolong their life span. Second, postsynaptic spines communicate with presynaptic terminals via adhesion molecules, possibly resulting in mutual stabilization (Fig. 3b,B). Indeed, spines that are attached to presynaptic terminals have been observed to be more stable than those that are not [14]. The adhesion molecules in spines, including cadherins [73] and neuoligin (neurexin) [74], bind to PSDs via S-SCAM (a synaptic scaffolding molecule) [75] or via PSD95 [76]. The presynaptic ligand ephrinB was shown to induce spine enlargement by interaction with complexes of EphB–syndecan-2 [77], intersectin – Cdc42 – N-WASP (neural Wiskott-Aldrich syndrom protein)– actin [78], or kalirin – Rac1– PAK1 (p21-activated kinase 1)– actin [68]. Third, large spine heads might facilitate the incorporation of the cellular machinery for protein synthesis, and proteins newly synthesized in situ could maintain spine http://tins.trends.com size (Fig. 3b,C). Indeed, polyribosomes are preferentially translocated to large spines during synaptic plasticity [79], and synaptic polyribosome complexes [79,80] often exhibit a rough-endoplasmic-reticulum-like configuration (Fig. 3b). These complexes might allow the local synthesis of CaMKII [80,81], FMRP [56], the cytoskeleton-associatedprotein Arc, and even transmembrane proteins [80]. This possible mechanism of spine stability is consistent with the fact that protein synthesis is required for the induction of memory [82] and for late LTP [80,83]. Active protein synthesis in spines might thus function as a synaptic tag for transcription-dependent memory formation [84,85]. However, it remains to be demonstrated that protein synthesis actually occurs within spines. Organelles in spines might also help to maintain spine size independently of protein synthesis. Finally, activity-dependent changes in large spines are likely suppressed by attenuation of Ca2þ signaling caused 366 Review TRENDS in Neurosciences Vol.26 No.7 July 2003 by the large volume of the spine head, contributing to the stability of these spines. AMPA-receptor expression Consistent with the observation that AMPA-receptor expression is proportional to spine-head volume (Fig. 2m), several lines of evidence support the notion that the expression of AMPA receptors is determined by F-actin and PSD proteins (Fig. 3c), the distributions of which are expected to correlate with spine-head volume. Actin-depolymerizing agents have, thus, been shown to induce redistribution of AMPA-receptor immunoreactivity in cultured hippocampal neurons [86]. Such agents also trigger a reduction in the number of functional AMPA receptors [87] and prevent LTP [88,89]. Given that AMPA recptors do not bind F-actin directly and are localized to PSDs, their expression also requires certain PSD proteins that link them to F-actin. Such interactions are mediated, for example, by complexes of AMPA-receptor– SAP97– protein 4.1 –F-actin [90] and AMPAR– stargazin– PSD95 – GKAP–shank–cortactin–F-actin [91]. It will be important in the future to determine how AMPA-receptor distribution is regulated during structural plasticity of spines. NMDA-receptor expression The expression of NMDA receptors differs from that of AMPA receptors. The distribution of NMDA-receptor immunoreactivity is thus only weakly correlated with PSD size in hippocampus [17,18]; the expression of AMPA receptors, but not that of NMDA receptors, is modulated during LTP [42]. Furthermore, in contrast to AMPA receptors: (1) the expression of NMDA receptors does not depend on F-actin [86,88]; (2) NMDA receptors bind tightly to many PSD proteins, including a-actinin, phosphorylated CaMKII, PSD95 and S-SCAM (Fig. 3d) [39]. The expression of these NMDA-receptor-binding proteins is not necessarily proportional to PSD size; and (3) NMDAreceptor expression is modulated by protein kinases A [92] and C [93,94] (Fig. 3d). The presence of NMDA receptors in some small spines greatly promotes their activitydependent plasticity, whereas the lack of these receptors in some large spines provides a basis for their structural and functional stability and underlies the robustness of memory readout. The regulation of NMDA-receptor expression is, thus, important in the organization of neuronal networks. Compared with AMPA-receptor expression, however, little is known of the regulatory mechanisms of NMDA-receptor expression at the level of the individual spine. Proteomic analysis has been performed on many of the scaffold proteins discussed in this section [95], the results of which are now available (http://www.anc.ed.ac.uk/mscs/PPID/). Recapitulation Recent progress in biophysical techniques and molecular biology has provided insight into the structure– function relationships of dendritic spines in the cerebral cortex, as well as support for the century-old hypothesis that spine structure is the basis for memory in the brain. The structure – stability –function relationships of spines have further suggested that small and large spines play distinct http://tins.trends.com roles in learning and memory, enabling rapid acquisition of new memory and robust readout, respectively. Clarification of such relationships will help to link molecular events to higher-order brain functions, and to shed light on and develop new treatments for mental disorders. Achievement of this goal will require a more complete characterization of the functional alterations associated with spine structural plasticity, which will be facilitated by the application of two-photon uncaging methodology, the only approach available that allows stimulation of individual central synapses independently and systematically. Progress in this field of research is rapid, and the time could soon be ripe for construction of a material theory of synaptic memory in brain tissue. Acknowledgements We thank Y. Hata for critical reading of the manuscript. Our work was supported by Grants-in-Aid from the Japanese Ministry of Education, Culture, Sports, Science and Technology and from the Japan Society for the Promotion of Science, and by a research grant from the Human Frontier Science Program Organization. References 1 Squire, L.R. (1987) Memory and Brain, 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 Shepherd, G.M. (1996) The dendritic spine: a multifunctional integrative unit. J. Neurophysiol. 75, 2197 – 2210 4 Moser, M.B. et al. (1994) An increase in dendritic spine density on hippocampal CA1 pyramidal cells following spatial learning in adult rats suggests the formation of new synapses. Proc. Natl. Acad. Sci. U. S. A. 91, 12673 – 12675 5 Fifkova, E. (1985) A possible mechanism of morphometric changes in dendritic spines induced by stimulation. Cell. Mol. Neurobiol. 5, 47 – 63 6 Toni, N. et al. (1999) LTP promotes formation of multiple spine synapses between a single axon terminal and a dendrite. Nature 402, 421 – 425 7 Yuste, R. and Bonhoeffer, T. (2001) Morphological changes in dendritic spines associated with long-term synaptic plasticity. Annu. Rev. Neurosci. 24, 1071 – 1089 8 Hering, H. and Sheng, M. (2001) Dendritic spines: structure, dynamics and regulation. Nat. Rev. Neurosci. 2, 880 – 888 9 Parnass, Z. et al. (2000) Analysis of spine morphological plasticity in developing hippocampal pyramidal neurons. Hippocampus 10, 561 – 568 10 Trachtenberg, J.T. et al. (2002) Long-term in vivo imaging of experience-dependent synaptic plasticity in adult cortex. Nature 420, 788 – 794 11 Grutzendler, J. et al. (2002) Long-term dendritic spine stability in the adult cortex. Nature 420, 812 – 816 12 Maletic-Savatic, M. et al. (1999) Rapid dendritic morphogenesis in CA1 hippocampal dendrites induced by synaptic activity. Science 283, 1923– 1927 13 Engert, F. and Bonhoeffer, T. (1999) Dendritic spine changes associated with hippocampal long-term synaptic plasticity. Nature 399, 66 – 70 14 Korkotian, E. and Segal, M. (2001) Regulation of dendritic spine motility in cultured hippocampal neurons. J. Neurosci. 21, 6115 – 6124 15 Dunaevsky, A. and Mason, C.A. (2003) Spine motility: a means towards an end? Trends Neurosci. 26, 155– 160 16 Fischer, M. et al. (1998) Rapid actin-based plasticity in dendritic spines. Neuron 20, 847 – 854 17 Nusser, Z. et al. (1998) Cell type and pathway dependence of synaptic AMPA receptor number and variability in the hippocampus. Neuron 21, 545 – 559 18 Takumi, Y. et al. (1999) Different modes of expression of AMPA and NMDA receptors in hippocampal synapses. Nat. Neurosci. 2, 618 – 624 19 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 Review TRENDS in Neurosciences Vol.26 No.7 July 2003 20 Harris, K.M. and Stevens, J.K. (1989) Dendritic spines of CA1 pyramidal cells in the rat hippocampus: serial electron microscopy with reference to their biophysical characteristics. J. Neurosci. 9, 2982 – 2997 21 Pettit, D.L. et al. (1997) Chemical two-photon uncaging: a novel approach to mapping glutamate receptors. Neuron 19, 465– 471 22 Matsuzaki, M. et al. (2001) Dendritic spine geometry is critical for AMPA receptor expression in hippocampal CA1 pyramidal neurons. Nat. Neurosci. 4, 1086– 1092 23 Papageorgiou, G. and Corrie, J.E.T. (2000) Effects of aromatic substituents on the photocleavage of 1-acyl-7-nitroindolines. Tetrahedron 56, 8197 – 8205 24 Morrison, J. et al. (2002) Mechanisms of photorelease of carboxylic acids from 1-acyl-7-nitroindolines in solutions of varying water content. Photochem. Photobiol. Sci. 1, 960 – 969 25 Smith, M. et al. (2003) Mechanism of the distance-dependent scaling of Schaffer collateral synapse in CA1 pyramidal neurons. J. Physiol. 548, 245 – 258 26 Durand, G.M. et al. (1996) Long-term potentiation and functional synapse induction in developing hippocampus. Nature 381, 71 – 75 27 Malenka, R.C. and Nicoll, R.A. (1999) Long-term potentiation – a decade of progress? Science 285, 1870– 1874 28 Groc, L. et al. (2002) Spontaneous unitary synaptic activity in CA1 pyramidal neurons during early postnatal development: constant contribution of AMPA and NMDA receptors. J. Neurosci. 22, 5552 – 5562 29 Fiala, J.C. et al. (2002) Dendritic spine pathology: cause or consequence of neurological disorders? Brain Res. Rev. 39, 29 – 54 30 Schikorski, T. and Stevens, C.F. (1997) Quantitative ultrastructural analysis of hippocampal excitatory synapses. J. Neurosci. 17, 5858 – 5867 31 Elston, G.N. et al. (2001) The pyramidal cell in cognition: a comparative study in human and monkey. J. Neurosci. 21, RC163 32 Rusakov, D.A. and Kullmann, D.M. (1998) Extrasynaptic glutamate diffusion in the hippocampus: ultrastructural constraints, uptake, and receptor activation. J. Neurosci. 18, 3158 – 3170 33 Svoboda, K. et al. (1996) Direct measurement of coupling between dendritic spines and shafts. Science 272, 716 – 719 34 Engert, F. and Bonhoeffer, T. (1997) Synapse specificity of long-term potentiation breaks down at short distances. Nature 388, 279 – 284 35 Petersen, C.C. et al. (1998) All-or-none potentiation at CA3– CA1 synapses. Proc. Natl. Acad. Sci. U. S. A. 95, 4732 – 4737 36 van Hemmen, J.L. and Kuhn, R. (1986) Nonlinear neural networks. Phys. Rev. Lett. 57, 913 – 916 37 Sheng, M. and Lee, S.H. (2001) AMPA receptor trafficking and the control of synaptic transmission. Cell 105, 825 – 828 38 Malinow, R. and Malenka, R.C. (2002) AMPA receptor trafficking and synaptic plasticity. Annu. Rev. Neurosci. 25, 103– 126 39 Sheng, M. and Kim, M.J. (2002) Postsynaptic signaling and plasticity mechanisms. Science 298, 776– 780 40 Song, I. and Huganir, R.L. (2002) Regulation of AMPA receptors during synaptic plasticity. Trends Neurosci. 25, 578 – 588 41 Shi, S.H. et al. (1999) Rapid spine delivery and redistribution of AMPA receptors after synaptic NMDA receptor activation. Science 284, 1811 – 1816 42 Zhu, J.J. et al. (2002) Ras and Rap control AMPA receptor trafficking during synaptic plasticity. Cell 110, 443– 455 43 Lisman, J. et al. (2002) The molecular basis of CaMKII function in synaptic and behavioural memory. Nat. Rev. Neurosci. 3, 175 – 190 44 Malinow, R. et al. (1989) Inhibition of postsynaptic PKC or CaMKII blocks induction but not expression of LTP. Science 245, 862 – 866 45 Crick, F. (1982) Do dendritic spines twitch? Trends Neurosci. 5, 44 – 46 46 Matus, A. et al. (1982) High actin concentrations in brain dendritic spines and postsynaptic densities. Proc. Natl. Acad. Sci. U. S. A. 79, 7590 – 7594 47 Fifkova, E. and Delay, R.J. (1982) Cytoplasmic actin in neuronal processes as a possible mediator of synaptic plasticity. J. Cell Biol. 95, 345 – 350 48 Fifkova, E. (1985) Actin in the nervous system. Brain Res. 356, 187 – 215 49 Pollard, T.D. and Borisy, G.G. (2003) Cellular motility driven by assembly and disassembly of actin filaments. Cell 112, 453 – 465 http://tins.trends.com 367 50 Matus, A. (2000) Actin-based plasticity in dendritic spines. Science 290, 754 – 758 51 Smart, F.M. and Halpain, S. (2000) Regulation of dendritic spine stability. Hippocampus 10, 542– 554 52 Segal, M. et al. (2000) Dendritic spine formation and pruning: common cellular mechanisms? Trends Neurosci. 23, 53 – 57 53 Rao, A. and Craig, A.M. (2000) Signaling between the actin cytoskeleton and the postsynaptic density of dendritic spines. Hippocampus 10, 527 – 541 54 Kaech, S. et al. (1999) Volatile anesthetics block actin-based motility in dendritic spines. Proc. Natl. Acad. Sci. U. S. A. 96, 10433 – 10437 55 Tashiro, A. et al. (2000) Regulation of dendritic spine morphology by the rho family of small GTPases: antagonistic roles of Rac and Rho. Cereb. Cortex 10, 927– 938 56 Bardoni, B. and Mandel, J.L. (2002) Advances in understanding of fragile X pathogenesis and FMRP function, and in identification of X linked mental retardation genes. Curr. Opin. Genet. Dev. 12, 284 – 293 57 Meng, Y. et al. (2002) Abnormal spine morphology and enhanced LTP in LIMK-1 knockout mice. Neuron 35, 121 – 133 58 Furuyashiki, T. et al. (2002) Multiple spatiotemporal modes of actin reorganization by NMDA receptors and voltage-gated Ca2þ channels. Proc. Natl. Acad. Sci. U. S. A. 99, 14458 – 14463 59 Manabe, T. et al. (2000) Regulation of long-term potentiation by H-Ras through NMDA receptor phosphorylation. J. Neurosci. 20, 2504 – 2511 60 Wu, G.Y. et al. (2001) Spaced stimuli stabilize MAPK pathway activation and its effects on dendritic morphology. Nat. Neurosci. 4, 151– 158 61 Crick, F. (1984) Memory and molecular turnover. Nature 312, 101 62 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 63 Star, E.N. et al. (2002) Rapid turnover of actin in dendritic spines and its regulation by activity. Nat. Neurosci. 5, 239– 246 64 Walikonis, R.S. et al. (2001) Densin-180 forms a ternary complex with the a-subunit of Ca2þ/calmodulin-dependent protein kinase II and a-actinin. J. Neurosci. 21, 423 – 433 65 Naisbitt, S. et al. (1999) Shank, a novel family of postsynaptic density proteins that binds to the NMDA receptor/PSD-95/GKAP complex and cortactin. Neuron 23, 569 – 582 66 Shen, K. et al. (1998) CaMKIIb functions as an F-actin targeting module that localizes CaMKIIa/b heterooligomers to dendritic spines. Neuron 21, 593 – 606 67 Pak, D.T. et al. (2001) Regulation of dendritic spine morphology by SPAR, a PSD-95-associated RapGAP. Neuron 31, 289– 303 68 Penzes, P. et al. (2003) Rapid induction of dendritic spine morphogenesis by trans-synaptic EphrinB – EphB receptor activation of the Rho-GEF Kalirin. Neuron 37, 263– 274 69 Hsieh-Wilson, L.C. et al. (2003) Phosphorylation of spinophilin modulates its interaction with actin filaments. J. Biol. Chem. 278, 1186– 1194 70 El-Husseini, A.E. et al. (2000) PSD-95 involvement in maturation of excitatory synapses. Science 290, 1364 – 1368 71 Rao, A. et al. (2000) Neuroligation: building synapses around the neurexin-neuroligin link. Nat. Neurosci. 3, 747 – 749 72 Shirao, T. et al. (1992) Cloning of drebrin A and induction of neuritelike processes in drebrin-transfected cells. NeuroReport 3, 109– 112 73 Togashi, H. et al. (2002) Cadherin regulates dendritic spine morphogenesis. Neuron 35, 77 – 89 74 Scheiffele, P. et al. (2000) Neuroligin expressed in nonneuronal cells triggers presynaptic development in contacting axons. Cell 101, 657– 669 75 Nishimura, W. et al. (2002) Interaction of synaptic scaffolding molecule and b-catenin. J. Neurosci. 22, 757 – 765 76 Irie, M. et al. (1997) Binding of neuroligins to PSD-95. Science 277, 1511 – 1515 77 Ethell, I.M. et al. (2001) EphB/syndecan-2 signaling in dendritic spine morphogenesis. Neuron 31, 1001– 1013 78 Irie, F. and Yamaguchi, Y. (2002) EphB receptors regulate dendritic spine development via intersectin, Cdc42 and N-WASP. Nat. Neurosci. 5, 1117 – 1118 79 Ostroff, L.E. et al. (2002) Polyribosomes redistribute from dendritic shafts into spines with enlarged synapses during LTP in developing rat hippocampal slices. Neuron 35, 535– 545 Review 368 TRENDS in Neurosciences Vol.26 No.7 July 2003 80 Steward, O. and Schuman, E.M. (2001) Protein synthesis at synaptic sites on dendrites. Annu. Rev. Neurosci. 24, 299 – 325 81 Mori, Y. et al. (2000) Two cis-acting elements in the 30 untranslated region of a-CaMKII regulate its dendritic targeting. Nat. Neurosci. 3, 1079 – 1084 82 Davis, H.P. and Squire, L.R. (1984) Protein synthesis and memory: a review. Psychol. Bull. 96, 518 – 559 83 Miller, S. et al. (2002) Disruption of dendritic translation of CaMKIIa impairs stabilization of synaptic plasticity and memory consolidation. Neuron 36, 507 – 519 84 Frey, U. and Morris, R.G. (1997) Synaptic tagging and long-term potentiation. Nature 385, 533– 536 85 Tokuyama, W. et al. (2000) BDNF upregulation during declarative memory formation in monkey inferior temporal cortex. Nat. Neurosci. 3, 1134 – 1142 86 Allison, D.W. et al. (1998) Role of actin in anchoring postsynaptic receptors in cultured hippocampal neurons: differential attachment of NMDA versus AMPA receptors. J. Neurosci. 18, 2423– 2436 87 Zhou, Q. et al. (2001) Contribution of cytoskeleton to the internalization of AMPA receptors. Proc. Natl. Acad. Sci. U. S. A. 98, 1261 – 1266 88 Kim, C.H. and Lisman, J.E. (1999) A role of actin filament in synaptic transmission and long-term potentiation. J. Neurosci. 19, 4314– 4324 89 Krucker, T. et al. (2000) Dynamic actin filaments are required for stable long-term potentiation (LTP) in area CA1 of the hippocampus. Proc. Natl. Acad. Sci. U. S. A. 97, 6856– 6861 90 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 91 Chen, L. et al. (2000) Stargazin regulates synaptic targeting of AMPA receptors by two distinct mechanisms. Nature 408, 936 – 943 92 Crump, F.T. et al. (2001) cAMP-dependent protein kinase mediates activity-regulated synaptic targeting of NMDA receptors. J. Neurosci. 21, 5079 – 5088 93 Lan, J.Y. et al. (2001) Protein kinase C modulates NMDA receptor trafficking and gating. Nat. Neurosci. 4, 382 – 390 94 Fong, D.K. et al. (2002) Rapid synaptic remodeling by protein kinase C: reciprocal translocation of NMDA receptors and calcium/calmodulindependent kinase II. J. Neurosci. 22, 2153 – 2164 95 Grant, S.G. and Husi, H. (2001) Proteomics of multiprotein complexes: answering fundamental questions in neuroscience. Trends Biotechnol. 19, S49– S54 Endeavour the quarterly magazine for the history and philosophy of science Online access to Endeavour is FREE to BioMedNet subscribers, providing you with a collection of beautifully illustrated articles on the history of science, book reviews and editorial comment. featuring The pathway to the cell and its organelles: one hundred years of the Golgi apparatus by M. Bentivoglio and P. Mazzarello Joseph Fourier, the ‘greenhouse effect’ and the quest for a universal theory of terrestrial temperatures by J.R. Fleming The hunt for red elixir: an early collaboration between fellows of the Royal Society by D.R. Dickson Art as science: scientific illustration 1490–1670 in drawing, woodcut and copper plate by C.M. Pyle The history of reductionism versus holistic approaches to scientific research by H. Andersen Reading and writing the Book of Nature: Jan Swammerdam (1637–1680) by M. Cobb Coming to terms with ambiguity in science: wave–particle duality by B.K. Stepansky The role of museums in history of science, technology and medicine by L. Taub The ‘internal clocks’ of circadian and interval timing by S. Hinton and W.H. Meck The troubled past and uncertain future of group selectionism by T. Shanahan A botanist for a continent: Ferdinand Von Mueller (1825–1896) by R.W. Home Rudolf Virchow and the scientific approach to medicine by L. Benaroyo Darwinism and atheism: different sides of the same coin? by M. Ruse Alfred Russel Wallace and the flat earth controversy by C. Garwood John Dalton: the world’s first stereochemist by Dennis H. Rouvray Forensic chemistry in 19th-century Britain by N.G. Coley Owen and Huxley: unfinished business by C.U.M. Smith Characteristics of scientific revolutions by H. Andersen and much, much more . . . Locate Endeavour in the BioMedNet Reviews collection. Log on to http://reviews.bmn.com, hit the ‘Browse Journals’ tab and scroll down to Endeavour http://tins.trends.com
© Copyright 2026 Paperzz