Rapid plasticity of dendritic spine: hints to possible functions?

Progress in Neurobiology 63 (2001) 61±70

www.elsevier.com/locate/pneurobio

Rapid plasticity of dendritic spine: hints to possible functions? Menahem Segal* Department of Neurobiology, The Weizmann Institute, Rehovot 76100, Israel

Abstract Contrary to a century-old belief that dendritic spines are stable storage sites of long term memory, the emerging picture from a recent ¯urry of exciting observations using novel high resolution imaging methods of living cells in culture is that of a dynamic structure, which undergoes fast morphological changes over periods of hours and even minutes. Concurrently, the nature of stimuli which cause formation or collapse of dendritic spines has changed from a mysterious Hebbian-governed plasticity producing stimulus to the more trivial activation of the synapse by strong/weak stimulation. The molecular mechanisms underlying spine plasticity are beginning to emerge; the role of presynaptic and/or postsynaptic activity, genetic, central or local factors in the formation and retraction of spines are currently being analyzed. A common mechanism for both, formation/elongation and pruning/retraction of spines, involving changes in intracellular calcium concentration ([Ca2+]i), is emerging. It appears that [Ca2+]i is related to changes in spines in a bell shape form: lack of synaptic activity causes transient outgrowth of ®lopodia but eventual elimination of spines, a moderate rise in [Ca2+]i causes elongation of existing spines and formation of new ones, while a massive increase in [Ca2+]i such as that seen in seizure activity, causes fast shrinkage and eventual collapse of spines. Nuclear signals (e.g. CREB), activated by an increase in [Ca2+]i, are involved in the central regulation of spine formation, while spine shrinkage and elongation are probably triggered by local [Ca2+]i changes. This hypothesis provides a parsimonious explanation for con¯icting reports on activity-dependent changes in dendritic spine morphology. Still, the many di€erences between cultured neurons, with which most of current studies are conducted, and the neuron in the real brain, require a cautious extrapolation of current assumptions on the regulation of spine formation. 7 2000 Elsevier Science Ltd. All rights reserved.

Contents 1.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

2.

Enhancing neuronal activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

3.

Formation and pruning of dendritic spines in the developing brain . . . . . . . . . . . . . . . . . . . . 63

4.

Suppression of synaptic connections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

5.

Time constants of changes in spine dynamics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

6.

In vitro experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

7.

The spine is a unique calcium compartment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

8.

A unifying hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

* Tel.: +972-8934-2553; fax: +972-8934-4140. E-mail address: [email protected] (M. Segal). 0301-0082/00/$ - see front matter 7 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 3 0 1 - 0 0 8 2 ( 0 0 ) 0 0 0 2 1 - 6

62

M. Segal / Progress in Neurobiology 63 (2001) 61±70

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

1. Introduction Ramon y Cajal, who ®rst described dendritic spines in central neurons, was also the ®rst to propose that they constitute the site of long term, stable memory in central neurons. The great variety of their shapes, sizes, density and inhomogeneous distribution on the parent dendrites attracted tremendous attention among neuroscientists of all disciplines. This attraction re¯ects a desire to formulate common elementary rules that govern the formation, maturation, and disappearance of dendritic spines and relate spine changes to memory formation, storage and erasure. Some of the `hot' issues, still not resolved, involve the formation of dendritic spines; do they result from synaptic interaction with a€erent inputs, or are they genetically predetermined much like cell shape and neurotransmitter content? Likewise, is the formation of spines orchestrated centrally via nuclear signals, or peripherally, in response to local demand? What are the roles of speci®c plasticity-producing stimuli, hormones and growth factors in the mature neuron? More speci®cally, what are the main parameters that express long term changes in spine functions; is it dendritic spine density, spine shape, size of synaptic contact, or the presence of multiple synapses on a spine? What are the molecular events that underlie stable long term changes in spines, how long do they last and what maintains their stability? Finally, what constitutes a spine? When should we stop calling a long protrusion branching o€ a dendrite `a ®lopodium' and begin to call it `a spine'? Should the same rules be applied to spines of di€erent neurons, which are obviously di€erent in their basic morphological and subcellular properties, as well as their connectivities? Early studies of the intact brain using Golgi impregnated material, attempted to correlate changes in dendritic spine morphology with development, enriched environment and aging (Purpura, 1974; Globus et al., 1973; Scheibel et al., 1975), with the rationale being that these extreme conditions may exert major changes in brain structure and function that can be seen under the microscope. Indeed, it was found that in the process of aging, cortical neurons lose up to 50% of their dendritic spines (Feldman and Dowd, 1975; Scheibel et al., 1975). Likewise, enriched environment was associated with an increase in the spine density, as well as in other morphological alterations in cortical dendrites. These early studies correlated experience and spine densities over long periods of time, but could not

relate speci®c events to changes in spines in speci®c neurons. Consequently, several di€erent models for assaying activity-related changes in dendritic spine morphology have been developed. Tetanic stimulation in the intact brain as well as in the in vitro hippocampal slice has been shown to produce long term potentiation (LTP) of reactivity to a€erent stimulation, a convenient and popular model of synaptic plasticity (Bliss and Collingridge, 1993). Other useful models include the induction of status epilepticus by kainic acid (Suzuki et al., 1997), pilocarpine (Isokawa, 1998), or kindling stimulation (Spigelman et al., 1998; Bundman and Gall, 1994), which can produce local or generalized seizures. Other studies use more natural stimulations such as enriched environment (Moser et al., 1997; Jones et al., 1997), imprinting (Bock and Braun, 1999b), learning (O'Malley et al., 1998; Moser et al., 1997) and hormonal states (Woolley and McEwen, 1993). A single learning event in the life of a young chick has been shown to produce rapid and marked changes in spine density in selective regions of its brain (Lowndes and Stewart, 1994). A great variety of animal species, age, brain region and cell type studied, type and duration of treatment and methods of observation and analysis contribute to the heterogeneity of observed changes in spine dimensions. The rather labor-intensive studies with dendritic spines, which are at the limit of optical resolution, contribute to the slow turnover rate of results in vivo and make it dicult to reach a universal understanding of the rules that govern spine plasticity. A partial account of reported changes in spine dimensions following speci®c treatments illustrates this diculty:

2. Enhancing neuronal activity Several studies examined the structure of dendritic spines after submission of the tissue to a stimulation protocol that produces LTP either in vivo or in a slice (Lee et al., 1980; Desmond and Levy, 1983; O'Malley et al., 1998; van Harreveld and Fifkova, 1975). The extensive search for molecular mechanisms underlying LTP revealed a sequence of events, starting from an in¯ux of calcium into the dendritic spines, through activation of several protein kinases, which phosphorylate synaptic proteins, to activation of nuclear cascades leading to protein synthesis and eventually, to a stable, long lasting LTP (Andersen et al., 1980; Bliss and Collingridge, 1993; Malenka and Nicoll, 1993; Bear,

M. Segal / Progress in Neurobiology 63 (2001) 61±70

1995). Tetanic stimulation produces a spectrum of e€ects on dendritic spines, ranging from changes in spine dimensions (Lee et al., 1980), synaptic contact area and shape (Desmond and Levy, 1983), spine head bifurcation (Rusakov et al., 1997), to the formation of novel spines (Toni et al., 1999; Trommald et al., 1996) or disappearance of existing ones (Rusakov et al., 1997). None of these studies have actually veri®ed that the altered/novel spines contribute to the enhanced reactivity to the a€erent stimulation, the hallmark of LTP. In some studies, a control condition, e.g. a drug which blocks LTP (the NMDA receptor antagonist 2APV) is used as an evidence that the novel spines are involved in LTP. This is only a correlative evidence, which does not clearly show that the novel spine is part of the enhanced synapse. One of the indicators for the presence of LTP is the ability to reduce the potentiated synapse back to baseline using a depotentiation protocol. Yet, in no study was such a protocol used to show that spines are restored to their previous state once LTP is erased. Furthermore, potentiation is taking place in a selective subset of synapses, and in fact, a potentiated region is likely to be ¯anked by unpotentiated, or even depressed regions (Andersen et al., 1980). Such a spatial segregation of potentiated and depressed synapses has not been shown morphologically. Finally, recent studies with acute slices cast doubt whether LTP by itself produces morphological changes in spines or not. Instead Kirov et al. (1999) suggest that slicing of the tissue, in preparation of the LTP experiment, causes a drastic increase in spine density. Slicing is indeed a traumatic experience for the tissue, far more than a single tetanic stimulation. Right after its formation, the slice undergoes a period of electrical silence, probably due to the slicing trauma, which includes depolarization, release of potassium, and subsequent release of excitatory neurotransmitter substances. Depending on the slicing angle and the proximity to the slice surface, di€erent cells in the slice may experience di€erent amounts of trauma including transection of dendrites and axons and dea€erentation, which will determine their ability to undergo LTPinduced spine plasticity. In this respect, the cultured slice provides a more stable preparation, where the cells do not experience acute stress. On the other hand, the cultured cell is missing major extrinsic excitatory a€erents, and its spine density is lower than that of its in vivo counterpart (Collin et al., 1997). For both cell types, heterogeneity in the observed spine changes following tetanic stimulation can be expected for reasons listed above. Needless to say, a study which will examine all possible parameters is an unful®lled dream of morphologists and physiologists alike. A much stronger synaptic stimulation, causing local or generalized seizure, is intuitively expected to produce more coherent increase in spine density. In fact,

63

seizure produces dramatically opposite e€ects, involving either formation of novel synapses (Isokawa, 1998; Spigelman et al., 1998) and spines where they were not seen before (e.g. somatic spines, Bundman and Gall, 1994), or elimination of existing ones (Drakew et al., 1996). 3. Formation and pruning of dendritic spines in the developing brain Developmental studies may provide an important clue regarding the mechanisms of spine formation. Neurons are born without dendrites or spines, and develop them over weeks as they become innervated by incoming ®bers. Still, it is dicult to discern whether the spines are formed before or after the formation of synapses. A recent in vivo EM study (Fiala et al., 1998) indicates that incoming ®bers make synapses with dendritic ®lopodia, which may then convert into mature spines, suggesting that spine formation follows synapse formation. The importance of a€erent innervation has been stressed in a recent in-vitro study, where spine density has been associated with innervation pattern within the dendritic tree of a single neuron, such that innervated dendrites contain higher spine density than uninnervated ones (Kossel et al., 1997). In contrast, spines in some neurons are formed in the absence of a€erent ®bers such as in reeler or weaver mice. Born without their main excitatory granular cell a€erents, cerebellar Purkinje cells of these mice still bear spines, indicating that they may be pre-programmed to produce dendritic spines, even in the absence of a€erent connectivity (Sotelo, 1990). In this case, the role of a€erent activity is, at best, to modulate the dimensions of existing spines. Spines that are unattached to presynaptic ®bers can also be formed in purkinje cells by exposure to Substance P (Baloyannis et al., 1992). Thus, although it is common to assume that all the spines in vivo are innervated by a€erent ®bers, there may be many exceptions to this rule. One important process in dendritic development involves spine pruning. In early development, cortical spine density increases with age, followed by a reduction to reach mature values (Lund and Holbach, 1991; Rakic et al., 1986). While these observations were made originally in monkeys, similar age- and sexdependent pruning were also found in rodents and birds (Munoz-Cueto et al., 1990; Wallhausser and Scheich, 1987). It is not clear if spine pruning is an active process, associated with an increase in synaptic activity, or a passive process, caused by a lack of a€erent activation of the pruned spine. In the avian forebrain, where spine pruning was ®rst described in association with early learning, it appears to be an active process, which can be blocked by NMDA antagonists (Bock and Braun, 1999a, 1999b). Similar ob-

64

M. Segal / Progress in Neurobiology 63 (2001) 61±70

servations were made in retinal ganglion cells (Lau et al., 1992). The fact that spines of di€erent neuronal types reach their maximal density at di€erent postnatal ages (Lund and Holbach, 1991) indicates that spine pruning may be a cell- and pathway-speci®c phenomenon, and not a global e€ect in the entire developing brain. A recent approach to the study of the regulation of spine formation is the use of transgenic mice. Transgenic mice are used extensively for the analysis of the involvement of a single gene product in brain functions, and a number of transgenic mice phenotype contain varying degrees of learning disabilities and impaired expression of long term potentiation. Most of these mice do not seem to have signi®cant or speci®c changes in their neuronal morphology. There are, however, some exceptions. A large increase in spine density was found in mice lacking the NR3A subunit of the NMDA receptor, which was correlated with an increase in the response to NMDA (Das et al., 1998). Other studies, which manipulated NMDA receptors either by a knockout or overexpression (Okabe et al., 1998a, 1998b; Rampon et al., 2000) did not report changes in dendritic spine density. A di€erent approach to the genetic analysis is presented in a series of studies that examine the role of drebrin in spine morphology. Overexpression of drebrin, a cytoskeletal element, has been found to a€ect the shape of dendritic spines and their density in cultured cortical neurons (Hayashi and Shirao, 1999). Thus, this study provides a clear link between genetic factors and the putative environmental regulators of dendritic spine plasticity. An interesting correlation between dendritic spine morphology and mental retardation was found recently. It has long been suggested that mental retardation is associated with dendritic spine dysgenesis, (Purpura, 1974) but the molecular basis of this has been described only recently. The fragile-X mental retardation protein (FMRP) which is absent in fragile-X mentally retarded children has been isolated and cloned. FMRP-knock out mice have long, dense, ®lopodia-like immature spines distributed along the dendrites of cortical pyramidal neurons (Comery et al., 1997), proposing a role for FMRP in maturation and pruning of dendritic spines. The physiological and molecular signi®cance of this observation is still being investigated. 4. Suppression of synaptic connections One approach to the analysis of the factors which govern maintenance of spines involves the study of e€ects of removal of a€erents to the spiny cell. In the young brain, dea€erentation may cause an initial formation of ®lopodia followed by an eventual reduction

of dendritic spine density (Benshalom, 1989; Cheng et al., 1997; Schauwecker and McNeill, 1996). Interestingly, blockade of action potential discharges with tetrodotoxin does not mimic the severe e€ect of deafferentation (McKinney et al., 1999), indicating that existing presynaptic terminals may release neurotransmitters or some growth factors in an amount that is sucient to maintain the shape and size of the postsynaptic spine without the need for a spike-associated synchronized a€erent activity. Alternately, the di€erence between dea€erentation and blockade of action potentials can re¯ect the possibility that the degenerated terminals activate glial processes which are instrumental in removing remaining spines. The nature of these potential mechanisms has not been explored as yet, but it is obvious that dendritic spines need a€erent activity for their maintenance beyond their initial growth (Okabe et al., 1999). 5. Time constants of changes in spine dynamics One of the clues to possible functions of dendritic spines can be found in the time course of changes in spines following an environmental challenge. If spines are indeed made to store information for long periods of time, one can expect them to react slowly to environmental changes, and once they are formed, to remain unchanged. This approach was taken in early studies, where development, enriched environment and aging process were used for the analysis of spine changes. Over the years it became apparent that this is not necessarily the rule: A reduction in spine density and dendritic morphology in the hibernating ground squirrel can be completely restored within 2 h of arousal from torpor (Popov et al., 1992). Likewise, changes in dendritic spine density vary quite precisely across the estrus cycle, with large variations seen over periods of 24 h (Woolley and McEwen, 1993). Also, tetanic stimulation or even just the cutting of brain slices can produce marked increases in spine density in less than 1 h (Kirov et al., 1999). The newly described short time constants of spine changes in vivo indicate that they may serve more transient roles than originally thought. The apparent rapid changes in spine morphology as well as the ability to label individual neurons with a ¯uorescent tag and follow changes in spines in real time has popularized the use of simple, in vitro preparations for the study of the factors which regulate spine plasticity. 6. In vitro experiments Tissue culture provides a controlable environment, where changes in individual spine morphology along

M. Segal / Progress in Neurobiology 63 (2001) 61±70

with incoming a€erents can be monitored with relative ease (Fig. 1). Cultured neurons are di€erent from their in vivo counterparts in many morphological and pharmacological properties, with the most striking di€erence being a smaller spine density, about 10% of that found in vivo (Papa et al., 1995), but they are likely to share basic rules of cellular interactions. Nonetheless, results with cultured neurons should be veri®ed with in vivo preparations. Following this rationale, we replicated original in vivo observations by Woolley and McEwen (1993), to show that estradiol causes a two fold increase in dendritic spine density in cultured rat hippocampal pyramidal neurons (Murphy and Segal, 1996). Thereafter, we have initiated a series of studies on the molecular cascades leading from the activation of the estrogen receptor to the formation of novel dendritic spines. The e€ects of estradiol appear to be mediated by downregulation of brain derived neurotrophic factor (BDNF) (Murphy et al., 1998b) which regulates expression of glutamate decarboxylase (GAD) (Murphy et al., 1998a), the enzyme responsible for the production of GABA. Reduction of GAD

Fig. 1. A: 3D confocal microscopic reconstruction of a dendrite from a cultured mouse hippocampal neuron, ®xed and stained with DiI, (red), and counter-immunostained with an antibody to synaptophysin (yellow). Most of the dendritic spines are touched by synaptophysinimmunoreactive terminals (from Braun and Segal, unpublished observations). B: Calcein-®lled, 3D reconstructed dendrite from a live rat hippocampal neuron in culture. Left, control condition, several spines are seen to branch o€ the main dendrite. Spontaneous motility of the spines is low in this mature neuron. Middle, after exposure of the dendrite to low concentration of glutamate, the spines expand. Right, after exposure of the dendrite to a high concentration of glutamate, the spines shrink (modi®ed from Korkotian and Segal, 1999b).

65

causes a shift in the excitation/inhibition balance in the culture, towards an increase in excitability of the neurons. Enhanced excitability and subsequent network activity cause accumulation of intracellular calcium in the pyramidal neurons, which stimulates the phosphorylation (p-) of cAMP response element binding protein (CREB). The p-CREB then presumably activates downstream genes responsible for the creation of novel dendritic spines (Murphy and Segal, 1997). The basic assumption underlying this hypothesis is that activity regulates dendritic spines in a dynamic fashion, such that an increase in activity causes formation of new dendritic spines, while a blockade of the activity inhibits this e€ect. In support of this hypothesis is our observation that TTX, which blocks action potential discharges, as well as drugs which amplify inhibitory tone in the culture, e.g. diazepam, block the e€ects of estradiol on spine density (Murphy et al., 1998b). Also, blockade of inhibition using GABA receptor antagonists which causes an increase in excitatory network activity, also causes an increase in dendritic spine density that is dependent on a rise of intracellular calcium concentration (Papa and Segal, 1996). While this is an appealing hypothesis, it does not account for some con¯icting observations made recently in both cultured neurons and in vivo. Similar pharmacological treatments in a cultured slice, involving blockade of inhibition, caused a marked reduction in dendritic spine density (Drakew et al., 1996). Also, direct activation of glutamate receptors caused shrinkage of spines rather than formation of new ones (Segal, 1995c; Halpain et al., 1998). Likewise, in vivo epileptic seizures, which are also produced by blockade of inhibition, can cause elimination of dendritic spines, as well as their formation (see above). One of the unique advantages of the in vitro preparation is that it allows for the labeling of living individual cells and for the tracking down of changes in their morphology over extended periods of time. This circumvents the need to compare large independent populations, and allows a precise timing of the observed changes in dendritic spines. This ability facilitated the detection of rapid changes in spine morphology, in the minute time scale. The changes include spontaneous ¯uctuations around the same mean length, using green-¯uorescent protein (GFP)-tagged actin ®laments (Fischer et al., 1998). Further studies of spine motility indicate that these ¯uctuations appear not to depend on action potential discharges, but are developmentally regulated, as they are larger in the young than in the adult neurons, and involve rapid changes in actin polymerization (Dunaevsky et al., 1999). Another recent study has demonstrated the formation of novel spines in relation to the induction of LTP in pyramidal neurons of the cultured slice (Engert and Bonhoe€er, 1999). Similarly, a rapid growth of

66

M. Segal / Progress in Neurobiology 63 (2001) 61±70

®lopodia in response to local electrical stimulation has been described (Maletic-Savatic et al., 1999). Finally, expansion of existing spines and formation of novel ones in response to the release of calcium from stores (Korkotian and Segal, 1999a, 199b) have been demonstrated in dissociated cultures of rat hippocampal neurons. Once again, there is a marked contrast between these results, and those showing rapid collapse of dendritic spines in response to glutamate (Halpain et al., 1998; Segal, 1995c). These results also contrast with those suggesting that an NMDA antagonist can in fact induce rather than retard formation of novel spines (Rocha and Sur, 1995). These discrepancies will be addressed below. 7. The spine is a unique calcium compartment Using high resolution imaging of calcium sensitive dyes, studies in the past decade have demonstrated that the dendritic spine is a unique calcium compartment. Synaptic and chemical stimulation has been reported to raise [Ca2+]i to higher levels in the spine than in its parent dendrite. In fact, this change can sometimes be restricted to individual spines (Segal, 1995a, 1995b; Yuste and Denk, 1995). The spine possesses voltage gated calcium channels (Segal, 1995b), glutamate-activated channels, primarily of the NMDA type (Kovalchuk et al., 2000) and calcium stores (Korkotian and Segal, 1998). The source of synapticallyinduced elevated [Ca2+]i in individual spines is still debated; the [Ca2+]i rise results from either activation of NMDA channels (Kovalchuk et al., 2000), or from the release of calcium from stores (Emptage et al., 1999). In addition to the conventional fast synaptic current and its associated calcium rise, a unique synaptic calcium transient that may be of metabotropic origin, and store-related, has also been described (Takechi et al., 1998). Variations in the length of the spine neck, which may not have electrotonic relevance, appear to have important consequences for regulation of calcium in the spine and its parent dendrite. Morphological variations in spine dimensions are expressed in the ability of the spine to either transfer synaptically related [Ca2+]i changes into the parent dendrite, or be a€ected by a calcium transient arising in the dendrite (Volfovsky et al., 1999). Thus, the longer the spine neck, the more independent the spine is from the parent dendrite, and the lesser is the e€ect it exerts on the dendrite. These observations were made initially in cultured hippocampal neurons (Korkotian and Segal, 1998; Volfovsky et al., 1999) and were con®rmed recently in neurons in the hippocampal slice (Majewska et al., 2000). Spine independence may result in the ability of the spine to maintain high concen-

tration of calcium for a longer period, and may have biochemical relevance for calcium-dependent processes. The view that the spine is an independent biochemical compartment has been suggested before (Wickens, 1988; Segal, 1995d) and expended in a recent review by Shepherd (1996). The length of the spine is therefore extremely important for the biochemical functions of the spine. Subsequently, the nature of the stimuli that can cause elongation or retraction of spines is of extreme signi®cance for the understanding of the role of spines in synaptic processing. 8. A unifying hypothesis The con¯icting reports of the role of enhanced neuronal activity in spine formation forces a modi®cation of the current hypothesis to suggest that there is a bimodal relationship between spines and intracellular calcium concentration, such that a moderate increase in [Ca2+]i will cause formation of novel spines, and elongation of existing ones, whereas a large and persistent increase in [Ca2+]i will cause shrinkage and eventual elimination of spines. Furthermore, it is suggested that local changes in [Ca2+]i will change the length of local spines, while a central change in [Ca2+]i will lead to CREB phosphorylation and formation of novel spines or their disappearance and shrinkage of their parent dendrites. Thus, both local and central factors play a role in the regulation of spine morphology. This hypothesis is supported by recent observations suggesting that acute exposure to glutamate causes a fast shrinkage of dendritic spines (Segal, 1995c; Halpain et al., 1998). A more moderate increase in [Ca2+]i, which is seen following the release of calcium from internal stores, will cause elongation of existing spines (Korkotian and Segal, 1999a). In fact, a direct support of this hypothesis comes from our most recent experiments showing that a short pulse application of glutamate, initiating a low and transient rise of [Ca2+]i, causes elongation of spines, while a larger pulse of glutamate initiating a longer rise of [Ca2+]i results in shrinkage of the same set of spines (Korkotian and Segal, 1999b) (Fig. 1(b)). A dual role of [Ca2+]i changes in regulation of opposing cellular mechanisms is not novel. A similar dependence on [Ca2+]i variations is also proposed for the growth cone (Kater et al., 1994). Its motility is controlled in a narrow dynamic range which, when exceeded in either direction causes shrinkage of the growth cone. Similarly, LTP requires an increase in postsynaptic [Ca2+]i, but the same is true with long term depression (LTD). The paradoxical calciumdependent ability of a synapse to undergo both LTP and LTD is assumed to depend on the magnitude of the rise of [Ca2+]i, with a smaller rise leading to acti-

M. Segal / Progress in Neurobiology 63 (2001) 61±70

vation of phosphatases and eventual LTD, and a larger one leading to activation of kinases and eventual LTP (Malenka and Nicoll, 1993; Bear, 1995). Interestingly, LTP can be evoked with a tetanic stimulation lasting no more than 1 s, whereas LTD requires a calcium-triggered process that lasts at least 20 min. The similarity of the distinction between LTP and LTD to shrinkage and elongation of spines is striking in this respect as well, since the former process can take place within a minute or so (Halpain et al, 1998), whereas elongation of spines is a much slower process (Korkotian and Segal, 1999a, 1999b). Functionally, it is likely that elongation of spines may be associated with independence of the spine from the parent dendrite and a reduction in synaptic eciency (Volfovsky et al., 1999), whereas shrinkage of spines will be associated with a stronger link between the spine and the parent dendrite and an increase in synaptic eciency. It is premature to claim that LTP is associated with shrinkage of spines, and LTD with their elongation, but the analogy can lead to experiments that test these hypotheses. Likewise, the involvement of protein kinases and phosphatases in these processes can be examined. It is suggested that processes leading to shrinkage and elongation of dendritic spines are controlled locally, while processes leading to formation of novel spines may represent extension of local biochemical events or may be engineered centrally, through messages arriving from the soma. Such might be the case with the e€ect of estradiol, where a global increase in dendritic spine density is seen following a nuclear message (phosphorylation of CREB), and this action is not restricted to a single dendrite. One of the open questions in this analysis is the involvement of presynaptic terminals, and the changes they undergo in the process of formation and shrinkage of spines. Do the boutons expand/shrink/multiply/ disappear together with the postsynaptic change, do they lead or follow the postsynaptic changes, and what are the molecular mechanisms underlying these changes? The main criticism of recent studies, which demonstrate the formation of novel spines in LTP, (Engert and Bonho€er, 1999) is their lack of evidence that the novel spines are innervated and functional. In our earlier studies, we found that estradiol also causes an increase in the number of terminals staining for synaptophysin (Murphy and Segal, 1997), indicating that the presynaptic boutons also undergo an expansion. Whether this e€ect is mediated by a retrograde signal arriving from the activated postsynaptic element or by an independent presynaptic mechanism is still open. The assumption that an increase in [Ca2+]i can produce both, an extension and shrinkage of spines can explain seemingly contradictory results of blocked inhibition or enhanced excitation on spine morphology. It

67

is likely that in some studies the resulting excitatory activity is so intense as to raise [Ca2+]i to very high levels leading to shrinkage of spines and their eventual disappearance (Drakew et al., 1996), whereas in other studies conducted by the same group, a more moderate increase in excitatory activity (McKinney et al, 1999) results in expansion and growth of existing spines. In fact, electrical stimulation which may mimic a€erent activity, causing a moderate increase in [Ca2+]i has been reported recently to facilitate extension of dendritic ®lopodia (Malecic-Savatic et al., 1999). Similarly, tetanic stimulation leading to LTP can cause di€erent morphological e€ects in the focal area on the dendrite where the stimulated ®bers are concentrated and a large change in [Ca2+]i is expected, than in the periphery of the stimulated zone, where a smaller calcium change is expected. Interestingly, a similar observation was made with respect to LTP, it has been suggested that the dendritic area where LTP is produced is ¯anked by regions where LTD is produced as a result of the same stimulation (Teyler et al., 1994; Andersen et al., 1980). Finally, the universality of this hypothesis has to be examined in di€erent cell types and spine types. Di€erent neuron types display di€erent experimentally observable changes in [Ca2+]i depending on their relative abundance of calcium binding proteins and buffers. There might also be di€erences in the genetic machinery for regulating spine formation. For example, Purkinje cells are highly spiny, but di€erent regions of their dendrites produce spines in response to di€erent stimuli (Bravin et al., 1999). By the same token, the cells may di€er in the balance between central triggers for production of spines, mediated by CREB phosphorylation, and local control of these processes. Regardless of these possible sources of heterogeneity, the possibility that elongation and contraction of spines may re¯ect common dependency on [Ca2+]i, may serve as a unifying mechanism for regulation of spine morphology (Segal et al., 2000). The current results obtained primarily with in vitro preparations, paint a very di€erent picture of spine motility and stability than ever before appreciated. The spine appears as a dynamic structure, which can change continuously as a function of current demand. The current reliance on in vitro systems may bias the view of the role of spines in long term processes, after all, most of the studies cited herein deal with the immature brain, and may not re¯ect on processes that take place in the adult one. Thus, rapid motility may be a process by which a spine searches and ®nds its target, but this motility may be lost weeks/months after the connection is stable. Also, the neuron in the brain is surrounded by glia, which may act to restrict motility of the spine. Given these limitations, then, it appears that the demand that changes the spine at

68

M. Segal / Progress in Neurobiology 63 (2001) 61±70

such a fast rate has to do with the need to restrict changes in [Ca2+]i to the area of the activated synapse. This view was proposed before (Segal, 1995d) as a means for protecting the neuron from excessive rise in [Ca2+]i following synaptic activation. This may be one of the mechanisms, in addition to rapid manipulation of glutamate receptor of the AMPA type (Lissin et al., 1999; Shi et al., 1999), which will serve to modulate synaptic current and subsequently, the e€ect of the synapse on neuronal activity.

Acknowledgements Supported by a grant from the US±Israel Binational Science Foundation. I wish to thank Drs. K. Braun and E. Korkotian for the use of ®gures and unpublished observations.

References Andersen, P., Sundberg, S.H., Sveen, O., Swann, J.W., Wigstrom, H., 1980. Possible mechanisms for long-lasting potentiation of synaptic transmission in hippocampal slices from guinea-pigs. J. Physiol. 302, 463±482. Baloyannis, S.J., Costa, V., Deretzi, G., 1992. Intraventricular administration of substance P induces unattached Purkinje cell dendritic spines in rats. Int. J. Neurosci. 62, 251±262. Bear, M., 1995. Mechanism for a sliding synaptic modi®cation threshold. Neuron 15, 1±4. Benshalom, G., 1989. Structural alterations of dendritic spines induced by neural degeneration of their presynaptic a€erents. Synapse 4, 210±222. Bliss, T.V.P., Collingridge, G.L., 1993. A synaptic model of memory: long term potentiation in the hippocampus. Nature 361, 31±39. Bock, J., Braun, K., 1999a. Blockade of N-methyl-D-aspartate receptor activation suppresses learning-induced synaptic elimination. Proc. Natl. Acad. Sci. USA 96, 2485±2490. Bock, J., Braun, K., 1999b. Filial imprinting in domestic chicks is associated with spine pruning in the associative area, dorsocaudal neostriatum. Eur. J. Neurosci. 11, 2566±2570. Bravin, M., Monardo, L., Vercelli, A., Rossi, F., Strata, P., 1999. Control of spine formation by electrical activity in the adult rat cerebellum. Proc. Natl. Acad. Sci. USA 96, 1704±1709. Bundman, M.C., Gall, C.M., 1994. Ultrastructural plasticity of the dentate gyrus granule cells following recurrent limbic seizures. Part II: Alterations in somatic synapses. Hippocampus 4, 611± 622. Cheng, H.W., Rafols, J.A., Goshgarian, H.G., Anavi, Y., Tong, J., McNeill, T.H., 1997. Di€erential spine loss and regrowth of striatal neurons following multiple forms of dea€erentation: a Golgi study. Exptl. Neurol. 147, 287±298. Collin, C., Miaguchi, K., Segal, M., 1997. Dendritic spines in cultured hippocampal slices: correlating structure and function. J. Neurophysiol. 77, 1614±1623. Comery, T.A., Harris, J.B., Willems, P.J., Oostra, B.A., Irwin, S.A., Weiler, I.J., Greenough, W.T., 1997. Abnormal dendritic spines in fragile-X knockout mice: maturation and pruning de®cits. Proc. Natl. Acad. Sci. USA. 94, 5401±5404. Das, S., Sasaki, Y.F., Rothe, T., Premkumar, L.S., Takasu, M., Crandall, J.E., Dikkes, P., Conner, D.A., Rayud, P.V., Cheung,

W., Chen, H.S., Lipton, S.A., Nakanishi, N., 1998. Increased NMDA current and spine density in mice lacking the NMDA receptor subunit NR3A. Nature 393, 377±381. Desmond, N.L., Levy, W.B., 1983. Synaptic correlates of associative potentiation/ depression: an ultrastructural study in the hippocampus. Brain Res. 265, 21±30. Drakew, A., Muller, M., Gahwiler, B.H., Thompson, S.M., Frotscher, M., 1996. Spine loss in experimental epilepsy: quantitative light and electron microscopic analysis of intracellularly stained CA3 pyramidal cells in hippocampal slice cultures. Neuroscience 70, 31±45. Dunaevsky, A., Tashiro, A., Majewska, A., Mason, C., Yuste, R., 1999. Developmental regulation of spine motility in the mammalian central nervous system. Proc. Natl. Acad. Sci. USA 96, 13438±13443. Emptage, N., Bliss, T.V., Fine, A., 1999. Single synaptic events evoke NMDA receptor mediated release of calcium from internal stores in hippocampal dendritic spines. Neuron 22, 115±124. Engert, F., Bonhoe€er, T., 1999. Dendritic spine changes associated with hippocampal long-term synaptic plasticity. Nature 399, 66± 70. Fischer, M., Kaech, S., Knutti, D., Matus, A., 1998. Rapid actin based plasticity in dendritic spines. Neuron 20, 847±854. Feldman, M.L., Dowd, C., 1975. Loss of dendritic spines in aging cerebral cortex. Anat. Embryol. 148, 279±301. Fiala, J.C., Feinberg, M., Popov, V., Harris, K.M., 1998. Synaptogenesis via dendritic ®lopodia in developing hippocampal area CA1. J. Neurosci. 18, 9800±9811. Globus, A., Rosenzweig, M.R., Bennett, E.L., Diamond, M.C., 1973. E€ects of di€erential experience on dendritic spine counts in rat cerebral cortex. J. Comp. Physiol. Psychol. 82, 175±181. Hayashi, K., Shirao, T., 1999. Change in the shape of dendritic spines caused by overexpression of drebrin in cultured cortical neurons. J. Neurosci. 19, 3918±3925. Halpain, S., Hipolito, A., Sa€er, L., 1998. Regulation of F-actin stability in dendritic spines by glutamate receptors and calcineurin. J. Neurosci. 18, 9835±9844. Isokawa, M., 1998. Remodeling dendritic spines in the rat pilocarpine model of temporal lobe epilepsy. Neurosci. Lett. 258, 73±76. Jones, T.A., Klintsova, A.Y., Kilman, V.L., Sirevaag, A.M., Greenough, W.T., 1997. Induction of multiple synapses by experience in the visual cortex of adult rats. Neurobiol. Learning Memory 68, 13±20. Kater, S.B., Davenport, R.W., Guthrie, P.B., 1994. Filopodia as detectors of environmental cues: signal integration through changes in growth cone calcium levels. Prog. Brain Res. 102, 119±160. Kirov, S.A., Sorra, K.E., Harris, K.M., 1999. Slices have more synapses than perfusion-®xed hippocampus from both young and mature rats. J. Neurosci. 19, 2876±2886. Korkotian, E., Segal, M., 1998. Fast confocal imaging of calcium released from stores in dendritic spines. Eur. J. Neurosci. 10, 2076±2084. Korkotian, E., Segal, M., 1999a. Release of calcium from stores alters the morphology of dendritic spines in cultured hippocampal neurons. Proc. Natl. Acad. Sci. USA 96, 12068±12072. Korkotian, E., Segal, M., 1999b. Bidirectional regulation of dendritic spine dimensions by glutamate receptors. NeuroReport 10, 2875± 2877. Kossel, A.H., Williams, C.V., Schweizer, M., Kater, S.B., 1997. A€erent innervation in¯uences the development of dendritic branches and spines via both activity-dependent and non-activitydependent mechanisms. J. Neurosci. 17, 6314±6324. Kovalchuk, Y., Eilers, J., Lisman, J., Konnerth, A., 2000. NMDA receptor-mediated subthreshold Ca2+ signals in spines of hippocampal neurons. J. Neurosci. 20, 1791±1799. Lau, K.C., So, K.F., Tay, D., 1992. APV prevents the elimination of

M. Segal / Progress in Neurobiology 63 (2001) 61±70 transient dendritic spines on a population of retinal ganglion cells. Brain Res. 595, 171±174. Lee, K.S., Schottler, F., Oliver, M., Lynch, G., 1980. Brief bursts of high-frequency stimulation produce two types of structural change in rat hippocampus. J. Neurophysiol. 44, 413±422. Lissin, D.V., Carroll, R.C., Nicoll, R.A., Malenka, R.C., von Zastrow, M., 1999. Rapid, activation-induced redistribution of ionotropic glutamate receptors in cultured hippocampal neurons. J. Neurosci 19, 1263±1272. Lowndes, M., Stewart, M.G., 1994. Dendritic spine density in the lobus parolfactorius of the domestic chick is increased 24 h after one-trial passive avoidance training. Brain Res. 654, 129±136. Lund, J.S., Holbach, S.M., 1991. Postnatal development of thalamic recipient neurons in the monkey striate cortex. Part I: Comparison of spine acquisition and dendritic growth of layer 4C alpha and beta spiny stellate neurons. J. Comp. Neurol. 309, 115±128. Majewska, A., Brown, E., Ross, J., Yuste, R., 2000. Mechanisms of calcium decay kinetics in hippocampal spines: role of spine calcium pumps and calcium di€usion through the spine neck in biochemical compartmentalization. J. Neurosci. 20, 1722±1734. Malenka, R.C., Nicoll, R.A., 1993. NMDA-receptor-dependent synaptic plasticity: multiple forms and mechanisms. Trends Neurosci. 16, 521±527. Maletic-Savatic, M., Malinow, R., Svoboda, K., 1999. Rapid dendritic morphogenesis in CA1 hippocampal dendrites induced by synaptic activity. Science 283, 1923±1927. McKinney, R.A., Capogna, M., Durr, R., Gahwiler, B.H., Thompson, S.M., 1999. Miniature synaptic events maintain dendritic spines via AMPA receptor activation. Nature Neurosci. 2, 44±49. Moser, M.B., Trommald, M., Egeland, T., Andersen, P., 1997. Spatial training in a complex environment and isolation alter the spine distribution di€erently in rat CA1 pyramidal cells. J. Comp. Neurol. 380, 373±381. Munoz-Cueto, J.A., Garcia-Segura, L.M., Ruiz-Marcos, A., 1990. Developmental sex di€erences and e€ect of ovariectomy on the number of cortical pyramidal cell dendritic spines. Brain Res. 515, 64±68. Murphy, D.D., Segal, M., 1996. Regulation of dendritic spine density in cultured rat hippocampal neurons by steroid hormones. J. Neurosci. 16, 4059±4068. Murphy, D.D., Segal, M., 1997. Morphological plasticity of dendritic spines in central neurons is mediated by activation of cAMP response element binding protein. Proc. Natl. Acad. Sci. USA 94, 1482±1487. Murphy, D.D., Cole, N.B., Greenberger, V., Segal, M., 1998a. Estradiol increases dendritic spine density by reducing GABA neurotransmission in hippocampal neurons. J. Neurosci. 18, 2550±2559. Murphy, D.D., Cole, N.B., Segal, M., 1998b. BDNF mediates estradiol- induced dendritic spine formation in hippocampal neurons. Proc. Natl. Acad. Sci. USA 95, 11412±11417. Okabe, S., Vicario-Abejon, C., Segal, M., McKay, R.D.G., 1998a. Survival and synaptogenesis of hippocampal neurons without NMDA receptor function in culture. Eur. J. Neurosci. 10, 2192± 2198. Okabe, S., Collin, C., Auerbach, J.M., Meiri, N., Bengzon, J., Kennedy, M.B., Segal, M., McKay, R.D.G., 1998b. Hippocampal synaptic plasticity in mice overexpressing an embryonic subunit of the NMDA receptor. J. Neurosci. 18, 4177±4188. Okabe, S., Kim, H.D., Miwa, A., Kuriu, T., Okado, H., 1999. Continual remodeling of postsynaptic density and its regulation by synaptic activity. Nature Neurosci. 2, 804±811. O'Malley, A., O'Connell, C., Regan, C.M., 1998. Ultrastructural analysis reveals avoidance conditioning to induce a transient

69

increase in hippocampal dentate spine density in the 6 h posttraining period of consolidation. Neuroscience 87, 607±613. Papa, M., Bundman, M.C., Greenberger, V., Segal, M., 1995. Morphological analysis of the development of dendritic spines in primary cultures of hippocampal neurons. J. Neurosci. 15, 1±11. Papa, M., Segal, M., 1996. Morphological plasticity in dendritic spines of cultured hippocampal neurons. Neuroscience 71, 1005± 1011. Popov, V.I., Bocharova, L.S., Bragin, A.G., 1992. Repeated changes of dendritic morphology in the hippocampus of ground squirrels in the course of hibernation. Neuroscience 48, 45±51. Purpura, D.P., 1974. Dendritic spine dysgenesis and mental retardation. Science 186, 1126±1128. Rakic, P., Bourgeois, J.P., Eckenho€, M.F., Zecevic, N., GoldmanRakic, P.S., 1986. Concurrent overproduction of synapses in diverse regions of the primate cerebral cortex. Science 232, 232± 235. Rampon, C., Tang, Y.P., Goodhouse, J., Shimizu, E., Kyin, M., Tsien, J.Z., 2000. Enrichment induces structural changes and recovery from nonspatial memory de®cits in CA1 NMDAR1knockout mice. Nature Neurosci. 3, 238±244. Rocha, M., Sur, M., 1995. Rapid aquisition of dendritic spines by visual thalamic neurons after blockade of N-methyl D-aspartate receptors. Proc. Natl. Acad. Sci. USA 92, 8026±8030. Rusakov, D.A., Richter-Levin, G., Stewart, M.G., Bliss, T.V., 1997. Reduction in spine density associated with long-term potentiation in the dentate gyrus suggests a spine fusion-and-branching model of potentiation. Hippocampus 7, 489±500. Scheibel, M.E., Lindsay, R.D., Tomiyasu, U., Scheibel, A., 1975. Progressive dendritic changes in aging human cortex. Exptl. Neurol. 47, 392±403. Schauwecker, P.E., McNeill, T.H., 1996. Dendritic remodeling of dentate granule cells following a combined entorhinal cortex/®mbria fornix lesion. Exptl. Neurol. 141, 145±153. Segal, M., 1995a. Imaging of calcium variations in dendritic spines of cultured hippocampal neurons. J. Physiol. 486, 285±296. Segal, M., 1995b. Fast imaging of [Ca]2+ reveals presence of voltage i gated calcium channels in dendritic spines of cultured hippocampal neurons. J. Neurophysiol. 74, 484±488. Segal, M., 1995c. Morphological alternations in dendritic spines of rat hippocampal neurons exposed to NMDA. Neurosci. Lett. 193, 73±75. Segal, M., 1995d. Dendritic spines for Neuroprotection. Trends in Neurosci. 11, 468±471. Segal, M., Korkotian, E., Murphy, D.D., 2000. Dendritic spine induction and pruning- common cellular mechanisms? Trends in Neurosci. 23, 53±57. Shepherd, G.M., 1996. The dendritic spine: A multifunctional integrative unit. J. Neurophysiol. 75, 2197±2200. Shi, S.H., Hayashi, Y., Petralia, R.S., Zaman, S.H., Wenthold, R.J., Svoboda, K., Malinow, R., 1999. Rapid spine delivery and redistribution of AMPA receptors after synaptic NMDA receptor activation. Science 284, 1811±1816. Spigelman, I., Yan, X.X., Obenaus, A., Lee, E.Y., Wasterlain, C.G., Ribak, C.E., 1998. Dentate granule cells form novel basal dendrites in a rat model of temporal lobe epilepsy. Neuroscience 86, 109±120. Sotelo, C., 1990. Cerebellar synaptogenesis: what we can learn from mutant mice. J. Exp. Biol. 153, 225±249. Suzuki, F., Makiura, Y., Guilhem, D., Sorensen, J.C., Onteniente, B., 1997. Correlated axonal sprouting and dendritic spine formation during kainate-induced neuronal morphogenesis in the dentate gyrus of adult mice. Exptl. Neurol. 145, 203±213. Takechi, H., Eilers, J., Konnerth, A., 1998i. A new class of synaptic response involving calcium release in dendritic spines. Nature 396, 757±760. Teyler, T.J., Cavus, I., Coussens, C., Discenna, P., Grover, L., Lee,

70

M. Segal / Progress in Neurobiology 63 (2001) 61±70

Y.P., Little, Z., 1994. Multideterminant role of calcium in hippocampal synaptic plasticity. Hippocampus 4, 623±634. Toni, N., Bucks, P.A., Nikohenko, I., Bron, C.R., Muller, D., 1999. LTP promotes formation of multiple spine synapses between a single axon terminal and a dendrite. Nature 402, 421±425. Trommald, M., Hulleberg, G., Andersen, P., 1996. Long term potentiation is associated with new excitatory spine synapses on rat dentate granule cells. Learning and Memory 3, 218±228. van Harreveld, A., Fifkova, E., 1975. Swelling of dendritic spines in the fascia dentata after stimulation of the perforant ®bers as a mechanism of post-tetanic potentiation. Exptl. Neurol. 49, 736± 749. Volfovsky, N., Parnas, H., Segal, M., Korkotian, E., 1999. Geometry of dendritic spines a€ects calcium dynamics in hippo-

campal neurons: theory and experiments. J. Neurophysiol. 82, 450±462. Wallhausser, E., Scheich, H., 1987. Auditory imprinting leads to di€erential 2-deoxyglucose uptake and dendritic spine loss in the chick rostral forebrain. Brain Res. 428, 29±44. Wickens, J., 1988. Electrically coupled but chemically isolated synapses: dendritic spines and calcium in a rule for synaptic modi®cation. Prog. Neurobiology 31, 507±528. Woolley, C.S., McEwen, B.S., 1993. Roles of estradiol and progesterone in regulation of hippocampal dendritic spine density during the estrous cycle in the rat. J. Comp. Neurol. 336, 293±306. Yuste, R., Denk, W., 1995. Dendritic spines as basic functional units of neuronal integration. Nature 375, 682±684.