Featured Review Cellscience Reviews Vol 2 No.2 ISSN 1742-8130

Morphological correlates of long-term potentiation and depression

Scott M. Thompson^1,2,3,4 , Hayley A. Mattison^1,2, Michael W. Nestor^1,2,3

Introduction

Tremendous progress has been made over the last decades in understanding the biological basis of one of the quintessential advancements of human evolution: our ability to learn and to form memories that can last with remarkable precision for virtually our entire lifespan. Somewhat unusually for biology, the field benefited from a strong early theoretical framework. Particularly important was experimental psychologist Donald Hebb’s prescient postulation that use-dependent changes in synaptic strength are the elemental units of learning (Hebb, 1949). When Bliss and Lømo (1973), together with Per Andersen and Tony Gardner-Medwin, observed that a brief period of synaptic activity caused a long-lasting increase in the strength of the activated synapses, the phenomenon now called long-term potentiation or LTP (Figure 1), they already suspected that they had discovered the cellular basis of learning and memory. Intensive investigation of LTP in the intervening decades has cemented this relationship. Perhaps the most persuasive evidence comes from studies in which either pharmacological treatments or genetic manipulations that block LTP have been found to also block the ability of animals to learn certain behaviors (Morris et al., 2003).

Figure 1. Long-term potentiation Typical results from an experiment in which long-term potentiation (LTP) is induced in an ex vivo hippocampal brain slice. Excitatory postsynaptic potentials (EPSPs) are recorded intracellularly from a CA1 pyramidal cell in response to electrical stimulation of Schaffer collateral inputs. After a period of stable baseline recording, a high frequency stimulation protocol (typically 100 Hz, 1 sec; HFS) is delivered. The amplitude of the EPSP is doubled immediately the tetanus, followed by a slow decline to the final potentiated level, which can be maintained for many hours. Early LTP refers to the first 30-60 minutes after induction and does not require protein synthesis, whereas late LTP (>60 min after induction) is prevented by protein synthesis inhibitors.

We now know a great deal about the biological processes underlying the induction and expression of LTP (for reviews see: Malenka and Nicoll, 1999; Malenka and Bear, 2004; Malinow, 2003). Most investigators now agree that the most common form of LTP is induced when repetitive synaptic activation causes sufficient depolarization of the postsynaptic cell to relieve the constitutive block of the N-methyl-D-aspartate receptor (NMDAR)-gated ion channel by Mg²⁺ ions present in the extracellular fluid. NMDAR channels then allow Ca²⁺ to enter the postsynaptic cell where it activates a variety of second messenger pathways, particularly the Ca²⁺- and calmodulin-dependent protein kinase, CaMK. How these pathways then cause synaptic strength to increase remains controversial, but the strongest evidence favors postsynaptic mechanisms: increased insertion of new glutamate receptors into the postsynaptic membrane and/or a direct phosphorylation of pre-existing receptors that increases their ability to generate current (Malinow et al., 2000; Malenka and Bear, 2004). These processes are believed to account for potentiation lasting up to 2 hours after induction. Other, less well understood second messenger pathways are thought to lead ultimately to a change in the expression of unknown genes, encoding proteins that, in some undetermined way, underlie maintenance of the potentiation for many hours or days (Martin et al., 2000).

It has long been speculated that memories that last for time periods longer than hours or days must be encoded in the form of quasi-permanent morphological changes in the activated synapses. Compared to the remarkable progress in unraveling the physiology and molecular biology of learning and memory, the evidence that lasting changes in synaptic structure can be induced by the same types and patterns of synaptic activity that induce LTP is much less compelling, but has recently begun to yield to new experimental approaches.

The goal of this review is to discuss critically the most recent evidence for morphological synaptic plasticity and its relationship to electrophysiological plasticity, focusing exclusively on the hippocampus because that is where most experiments have been performed. We hope that by pointing out contradictory data, differences in experimental techniques, and unanswered questions, we will encourage further experiments and resolution of outstanding issues.

Dendritic spines and the needle-in-the haystack

The search for morphological plasticity has focused on dendritic spines. Dendritic spines form the postsynaptic half of excitatory synapses onto principal cells in the limbic cortex and neocortices, where learning is likely to take place and where many types of memories are stored. Each spine consists of a bulbous ‘head’ attached to the parent dendrite via a thin ‘neck’ of cytoplasm (for reviews see: Calverley and Jones, 1990; Edwards, 1995a; Harris, 1999; Nimchinsky et al., 2002; Yuste and Bonhoeffer, 2004). Each dendritic spine receives input from only one presynaptic nerve terminal (Fiala et al., 1998), although a given terminal may form synapses with multiple spines (see below). NMDARs and non-NMDARs are expressed in the plasma membrane of the spine head, which is opposed to the active zone of the presynaptic terminal. The receptors are anchored by a complex network of intracellular scaffolding proteins, apparent as a thickening known as the postsynaptic density (PSD), which can be seen readily in the electron microscope (Harris et al., 1992). The PSD links the receptors with other proteins whose functions are critical for the induction and expression of LTP (see: Kim and Sheng, 2004). Spines also contain intracellular organelles for regulating the intracellular Ca²⁺ concentration (Cooney et al., 2002).

The large heterogeneity in spine head shapes and sizes and in the length and thicknesses of the spine necks (e.g. Fiala et al., 1998) has engendered the search for the correlations amongst morphology and synaptic function and learning. That is, maybe some spines look different because they took part in the storage of some piece of information (an engram). For more than a century, dendritic spines could only be studied with the Golgi staining method in fixed tissue using either light or electron microscopy. The basic strategy in these studies was to maximize an animal’s ‘learning’ either by overtraining on specific behaviors or by enriching the environment, and then sacrificing the animal and examining synaptic profiles in Golgi stained tissue sections. Analogous experiments were also performed in which LTP was induced in ex vivo brain slices, followed by ultrastructural analysis of the fixed tissue. While some of these studies revealed provocative suggestions of learning associated changes, the overall variability of the spine population always left some room for doubt about the conclusions. Like finding the needle-in-the-haystack, there was no possibility in such studies of unambiguously demonstrating that those spines that displayed atypical morphologies were in fact the spines that had been involved in the learning or the LTP, despite efforts to potentiate as large a fraction of the synapses as possible. Nevertheless, these efforts did reveal some of the candidate mechanisms that might underlie structural plasticity.

What might an engram look like?

There are several conceivable ways in which a change in the strength of the synaptic connection formed by a pair of cells might be encoded morphologically: 1) an increase in the number of sites of contact between the two cells, 2) an increase in the size of the spine head or postsynaptic density to accommodate more receptors, or 3) a change in the spine neck diameter so that depolarization generated in the spine head might more readily depolarize the parent dendrite. There are at least some data from fixed tissue to support all three possibilities after induction of LTP or in whole animal learning paradigms. Increases in the number of sites should be apparent as an increase in spine density, and some (Moser et al., 1994; Geinisman, 2000 and O'Malley et al., 2000), but not all (Sorra and Harris, 1998), studies have supported this possibility. Enlargement of spine heads and increases in the area of the PSD have also been reported in some studies (Fifkova and Van Harreveld, 1977; Desmond and Levy, 1983; 1986a; 1986b; Chang and Greenough, 1984; Calverley and Jones, 1990), but not all (Lee et al., 1980; Andersen and Soleng, 1998; Sorra and Harris, 1998).

In 1998, a seminal paper from the laboratory of Andrew Matus (Fischer et al., 1998) changed significantly the way in which we think about and study dendritic spines. Hippocampal cell cultures were transfected with a green fluorescent protein (GFP)-tagged actin construct. The molecules became incorporated into the cytoskeleton of the cells and were particularly concentrated within the actin-rich PSD, thus allowing the spines of living cells to be visualized with unprecedented resolution. To the surprise of many, time lapse image sequences of labeled spines revealed that they were undergoing continuous ongoing changes in shape. The spine head, in particular, was observed to change its dimensions, or ‘morph’, by as much as 30% within as little as two minutes in a manner that required changes in the polymerization of actin. There was, however, no evidence that any new processes were formed or that existing processes were retracted over the time course of these experiments. While not directly related to the issue of functional synaptic plasticity, these observations did provide a dramatic demonstration that spines were considerably more motile than most people had imagined from looking at fixed tissue (Crick, 1982). Indeed, subsequent studies have replicated these rapid shape changes in a variety of neural tissues under a variety of experimental conditions, including in situ in intact animals (Majewska and Sur, 2003; Holtmaat et al., 2005). Unlike previously described manipulations that affect spine size or number (Woolley and McEwen, 1994; McKinney et al., 1999), which take days to be detectable, the observations of Fischer et al. (1998) provided the first evidence that spines could change at a time scale comparable to the changes in synaptic efficacy underlying LTP. Furthermore, this key paper suggested that imaging of fluorescently labeled dendritic spines might allow for simultaneous induction of functional and structural plasticity.

More is better

Because over 90% of the excitatory axodendritic synapses occur on the heads of dendritic spines in the adult hippocampus (Harris and Kater, 1994), any change that results in an increase in the number of spines is likely to represent an increase in synapse number. The advent of two photon laser scanning microscopy (TPLSM) made it possible to obtain high resolution images of fluorescently labeled dendritic spines in living ex vivo brain slices over the short time scales corresponding to the induction and expression of LTP. Two groups have found compelling evidence of an increase in the number of dendritic protrusions following LTP induction protocols in the CA1 region using organotypic slice cultures of the rat hippocampus.

The first evidence came from Roberto Malinow’s laboratory (Maletic-Savatic et al., 1999). Induction of LTP with tetanic stimulation delivered via a glass electrode positioned 3-10 µm away from the target dendrite resulted in an increase in the number of filopodia-like protrusions (>2.75 µm in length) extending from the dendritic shaft (Figure 2A). Growth was significant within 20 minutes and lasted for about one hour. The induction of filopodial outgrowth was blocked reversibly by NMDAR antagonists. Interestingly, these filopodial structures persisted in the absence of further stimulation and in almost one-third of the cases, these structures developed a bulbous head within one hour after stimulation. This is consistent with the idea that filopodia may mature directly into dendritic spines once they are stabilized by synaptic contact (Ziv and Smith, 1996).

Figure 2. Models of morphological plasticity accompanying LTP Drawings indicate the changes in synaptic contacts before, 5-30 min after (early), and 60-120 min after (late) LTP induction. In models A-D, the strengthening of preexisting connections is mediated by an increase in the number of synaptic contacts, whereas potentiation is accompanied by an increase in synaptic size in model E. A. In the study of Maletic-Savetic et al. (1999), LTP induction was accompanied by the appearance of filopodia, some of which later matured into new dendritic spines. B. Experiments by Engert and Bonhoeffer (1999) demonstrated that LTP induction resulted in the appearance of new dendritic spines directly from dendritic shafts. Whether the new spines in these studies contacted presynaptic terminals of the same axons opposed to the original spine remains unknown. C. Toni et al. (1999) demonstrated that LTP induction led first to the appearance of perforated postsynaptic densities and the formation of spinules in the spine head that penetrated into the nerve terminal. Later, presynaptic nerve terminals formed synapses with multiple spines from the same parent dendritic shaft. D. Richards et al. (2004) have demonstrated that dendritic processes can grow towards a source of glutamate. This suggests that new dendritic filopodia may establish multiple-spine, single dendrite synapses by growing towards the presynaptic terminal active during LTP induction and then maturing into a spine. In all of these models, there may also be an increase in the number or conductance of postsynaptic AMPARs in the original spine. E. In the study of Matsuzaki et al. (2004), LTP is accompanied by a large transient swelling of the spine head, followed by a smaller maintained increase in spine head volume.

Shortly thereafter, Engert and Bonhoeffer (1999) reported that LTP induction in rat hippocampal slice cultures caused the direct outgrowth of dendritic spines (Figure 2B), rather than filopodia. These authors blocked neurotransmitter release in the culture by removing extracellular Ca²⁺ and blocking voltage-gated Ca²⁺ channels with Cd²⁺. They then locally superfused control Ca²⁺ containing saline within an isolated, 30 µm-in-diameter region of a dendrite. This allowed the investigators to confine the potentiated synapses and monitor changes in spine number within only a small ‘haystack.’ In their experiments, LTP was induced not with tetanization, but by conjunctive pairing of low frequency evoked EPSPs and brief depolarizing current pulses. Within 30 minutes of the induction of LTP, between two and nine short, stubby spines appeared de novo from the dendritic shafts within the superfused region. The LTP-induced emergence of spines was blocked by NMDAR antagonists, as was the emergence of filopodia in the experiments of Maletic-Savetic et al. Importantly, spine protrusion was not observed when the pairing protocol failed to induce LTP, suggesting common triggering mechanisms.

While LTP-like stimulation protocols induced NMDAR-dependent dendritic process outgrowth in both sets of experiments, the nature of the processes differed. Engert and Bonhoeffer observed an outgrowth in what appeared to be true dendritic spines, whereas Maletic-Savatic et al., observed outgrowth of filopodia. During synaptogenesis, pyramidal cell dendrites in the hippocampus extend numerous filopodia. These filopodia are highly motile and can extend up to 10 µm, compared to a spine which is generally < 2 µm in length (Ziv & Smith, 1996; Fiala et al., 1998). It has been suggested that a filopodium initiates contact with a growing axon nearby. This contact results in the accumulation of active zone exocytotic proteins in the axon at the site of contact, followed by the accumulation of PSD proteins in the filopodium itself or in the the shaft to which the filopodium retracts (Ziv and Smith, 1996; Friedman et al., 2000; Marrs et al. 2001; Washbourne et al., 2002; Okabe et al., 2001). It is therefore thought that the contact between a filopodium and an axon is the initial step in the formation of a mature axospinous synapse. The filopodium then evolves directly into a spine or facilitates the formation of a shaft synapse from which a spine arises at later stages of synaptogenesis (Ziv and Smith, 1996; Fiala et al., 1998; Yuste and Bonhoeffer, 2004).

The most plausible explanation for the discrepancy in the morphological changes observed in the Engert and Bonhoffer and Maletic-Savatic et al. studies is the developmental age of the hippocampal slice cultures at the time of the experiments (Jontes and Smith, 2000). The hippocampal slices had been maintained in vitro for 2-9 days in the experiments of Maletic-Savatic et al., compared to 14-28 days in the study of Engert and Bonhoeffer. Given the age of the pups at the time the cultures were prepared and assuming that synaptogenesis and maturation proceed at the same pace in the cultures as in situ (De Simoni et al., 2003), then the experiments of Maletic-Savatic et al. correspond to postnatal (PN) days 9-16, an age range in which synaptogenesis in intact hippocampus is still in progress and dendritic protrusions remain highly motile and morphologically unstable (Dailey and Smith, 1996; Ziv and Smith, 1996; Fiala et al., 1998). The tissue used by Engert and Bonhoeffer, in contrast, corresponds to PN 17 - 35, an age range over which mature synapses are formed onto stable spines in intact tissue. While these differences are small, they could be significant. As illustrated in Figure 3, the transition from filopodia to spines occurs rapidly in these cultures. Thus, induction of LTP in younger tissue might lead first to the protrusion of filopodia that only later mature into spines, whereas LTP induction might cause spine outgrowth directly in older tissue.

Figure 3. Development of dendritic filopodia and spines in rat hippocampal slice cultures Photomicrographs of CA1 pyramidal cell dendrites, transfected with GFP, at 7, 9, and 12 days in vitro. Note that within this short time period, filopodia (open arrowheads) become rare and are replaced by mature-looking dendritic spines (closed arrowheads). Scale bar = 2 µm.

The Ca²⁺- and calmodulin-dependent protein kinase CaMK has long been recognized as a primary sensor for NMDAR-mediated Ca²⁺ influx in the induction of LTP. Through phosphorylation of glutamate receptors, CaMK increases the single channel conductance of pre-existing AMPARs in the postsynaptic plasma membrane (Derkach et al., 1999) and/or triggers the insertion of new AMPARs from intracellular stores in the spine head (Hayashi et al., 2000). Through these mechanisms, CaMK may be responsible for multiple aspects of LTP expression. A recent paper from Dominique Muller’s laboratory has shown that CaMK plays a key role in spine morphogenesis. Acute injection of either the CaMK activator calmodulin or a constitutively active, autophosphorylated form of CaMK into pyramidal cells in rat hippocampal slice cultures (PN 17 - 24) resulted in the outgrowth of new dendritic processes (Jourdain et al., 2003). Both filopodia and spines were found to emerge where few had been present before, with the emergence of filopodia preceding the appearance of spines. This effect was prevented by pharmacological inhibitors of CaMK. In addition, they replicated the results of Maletic-Savetic et al. by demonstrating that induction of LTP was accompanied by the emergence of filopodia followed by spines in a CaMK-dependent manner

These studies clearly establish an increase in spine number as a potential anatomical correlate of LTP. Presumably, a persistent increase in the ability of the presynaptic axons to excite the target neuron, in accordance with Hebb’s rule, would be accounted for by an increase in the number of synaptic contacts made by dendritic spines. For this to be true, however, the newly developed spines would have to form synapses selectively with the same presynaptic terminals that were active during the induction of the LTP. Unfortunately, both studies left the identity of the presynaptic cell unknown and therefore the issue remains unresolved.

Can more become less?

The strength of excitatory synaptic transmission is bidirectional, capable of displaying long-term depression (LTD, Figure 4A) as well as LTP, and it is reversible, capable of displaying depotentiation and de-depression (Fujii et al., 1991; Dudek and Bear, 1993; Lee et al., 2000; Nägerl et al., 2004). Although the morphological aspects of these forms of synaptic plasticity have received less attention than those accompanying LTP induction, new evidence indicates that induction of LTD and depotentiation may also result in morphological changes that are seemingly the converse of those observed upon LTP induction. These morphological changes thus mirror, in some respects, the electrophysiological data.

Figure 4. Models of morphological plasticity accompanying LTD A. Typical results from an experiment in which long-term depression (LTD) is induced in an ex vivo hippocampal brain slice. EPSPs are recorded intracellularly from a CA1 pyramidal cell in response to electrical stimulation of Schaffer collateral inputs. After a period of stable baseline recording, a low frequency stimulation protocol (typically 3 Hz, 15 sec; LFS) is delivered. The amplitude of the EPSP is decreased in amplitude immediately thereafter, followed by a slow recovery to the final depressed level, which can be maintained for many hours. B. In the study of Nägerl et al. (2004), LTD induction is accompanied by a loss of pre-existing dendritic spines. C. Zhou et al. (2004) demonstrated that LTD induction is accompanied by a decrease in the volume of the spine head.

In another study from Tobias Bonhoeffer’s lab, this time using organotypic hippocampal slice cultures from transgenic mice expressing GFP, Nägerl et al. (2004) reported that induction of LTP with a theta burst stimulation protocol triggered the formation of spines de novo, consistent with results of Engert and Bonhoeffer (1999). Fewer spines were found to emerge in this study (1 spine per 100 µm) compared to their previous report (ca. 3 spines per 100 µm). This difference might be attributed to the age of the cultures (PN 16-35 rat vs. PN 14-25 mouse), the use of theta burst stimulation rather than high frequency stimulation induction protocols, or the use of mice rather than rats, which may differ in terms of the developmental time course of synaptogenesis. Importantly, however, the number of new spines was roughly equal to the number of spines lost during the same time period (ca. 1 spine per 100 µm dendrite every 4 hours), so that, on average, there was no net change in spine density after LTP induction. In this important sense, therefore, the results do not agree with Maletic-Savetic et al. and Engert and Bonhoeffer, who reported that there was a net increase in spine number after LTP induction. Taken at face value, then, this study would imply that without LTP and learning there would be a steady decline in synapse number due to ongoing spine loss.

Nägerl et al. next induced LTD with a low frequency stimulus train in naïve slices and observed that spines fully retracted. Maximal spine loss occurred within 3 hours of low frequency stimulation and amounted to the loss of 2-3 spines per 100 µm of dendrite; significantly higher than the rate of ongoing spine loss in baseline conditions. Individual spines were observed to retract over the course of 10 minutes to one hour. This low frequency stimulation-induced spine retraction was blocked with the NMDAR antagonist AP5, like electrophysiological LTD. The loss of spines that occurred in the presence of AP5 was similar to the number of spines lost in baseline conditions in which no stimulus is given (~0.25/100 µm dendrite/hr), further indicating that NMDAR activation is required for LFS-induced spine loss to occur. Finally, Nägerl et al. examined the age-dependence of the LTD-associated spine loss and found that spines become progressively more stable with increasing age, with the greatest susceptibility occurring between PN 12 – PN 20, in accordance with the ease with which electrophysiological LTD can be induced (Wagner and Alger, 1995). This finding lends support to the idea that there is a developmentally defined decline in morphological plasticity with increasing age (Dailey and Smith, 1996; Kirov et al., 2004).

In summary, these studies have provided clear evidence that changes in spine number accompany and parallel changes in synaptic strength. Left unclear from these studies, however, is the identity or fate the presynaptic partners of the newly formed or recently retracted spines.

Spine splitting or ‘come hither’ signaling?

One conceptual problem with the observations of increased spine density after LTP is the problem of specificity. If, as Hebb postulated, learning of a particular bit of information depends on the correlated action potential discharge of two cells, then how does a global, or even a local increase in new spines result in a strengthening of the connection between the two cells? Is the formation of new presynaptic terminals also induced? Do these new spines form synapses with presynaptic partners whose activity was not correlated with the postsynaptic cell? A resolution to this problem may be apparent in another study from Dominique Muller’s laboratory.

In order to get around the needle-in-the-haystack problem in ultrastructural tissue samples, Toni et al. (1999) used a clever histochemical trick to label those synapses that had been tetanized during the LTP induction. Potassium chromium–trisoxalate was applied to the tissue during osmium fixation so as to create an electron dense precipitate with any Ca²⁺ ions present within the tissue (Buchs et al., 1994). They then examined the distribution of this precipitate in ex vivo hippocampal brain slices using electron microscopy. In control slices, only about 5% of all dendritic spine profiles contained precipitate. Delivery of an LTP-inducing tetanic stimulus train resulted in a doubling of the number of labeled spine profiles in a manner that was entirely dependent upon NMDAR activation. The authors thus assumed that the majority of the labeled spines represent spines that have been activated by the tetanus and have therefore contributed to the potentiation of excitatory synaptic transmission. The precipitate thus labels the needles in the haystack!

Using this trick, Toni et al. (1999) found that labeled spines were twice as likely to have a perforated PSD 30 min after induction of LTP as compared to spines in non-tetanized slices, in which only 20% of PSDs were perforated. These perforated synapses displayed small (ca. 200 nm) protrusions of the postsynaptic membrane into the presynaptic terminal, called spinules (Figure 2C). The appearance of perforated synapses and spinules was delayed from the tetanus and transient: no increase was seen at 5 or 15 min post-tetanus or at 45 - 120 min post-tetanus.

Following the appearance of perforated PSDs, there was an increase in the percentage of presynaptic boutons that formed synapses with not only the precipitate-labeled spine, but also another postsynaptic spine. The increase in these multiple spine boutons was only apparent 60 min after the induction of LTP (Figure 2C). The fraction of all boutons that had multiple postsynaptic partners increased from 5% in non-tetanized slices to 15% in tetanized slices one hour after the tetanus. These observations suggest that LTP is accompanied by the appearance of new spines that preferentially synapse with pre-existing nerve terminals. Like LTP, the increases in both perforated synapses and multiple spine boutons were prevented by antagonists of NMDARs and inhibitors of CaMK.

The resolution to the specificity issue came from serial reconstructions of multiple spine boutons. When all of the spines contacting a multiple spine bouton were traced back to their parent dendrites, it was found that two thirds of all spines forming synapses with multiple spine boutons arose from the same parent dendrite when the samples were taken from tissue fixed 60 min after LTP induction. In non-tetanized tissue or in tissue fixed 5 min after LTP induction, in contrast, 90% of the spines forming synapses with multiple spine boutons arose from different dendrites. If a second period of tetanic stimulation was delivered one hour after the induction of LTP, then both spines contacting the multiple spine bouton were observed to contain precipitate, indicating that the newly formed synapses were already functional. These results are thus interpreted to indicate that LTP induction results in an NMDAR- and CaMK-dependent formation of new dendritic spines and that these spines tend overwhelmingly to form synapses with the same presynaptic terminals that formed synapses with the cell’s original spines.

Mechanistically, the appearance of spinules and perforated synapses 30 min after LTP induction, followed by the appearance of the multiple spine boutons 60 min after LTP induction, may indicate that LTP induces the formation of the new spines by causing a splitting of pre-existing spines. In this model (Figure 2C), the protrusion of the spinule from the spine head and the perforation of the postsynaptic density are the visible signs of a spine splitting itself and its opposed presynaptic terminal.

A significant challenge to this model was raised by Fiala et al. (2002). They used electron microscopy and serial reconstruction to examine seven multiple spine boutons in which the spines arose from the same dendrite in hippocampal slices (PN 15) in which LTP had been induced two hours earlier. They found that 1-7 mature axons passed through the gap between the dendrite, the spines, and their presynaptic bouton. This observation is clearly inconsistent with the simple notion of spine splitting and suggests, instead, that the newly formed spine might have emerged de novo from the dendrite, as suggested by Engert and Bonhoeffer (1999) and Maletic-Savetic et al. (1999). The newly formed spines would then preferentially contact the presynaptic terminal whose activity caused the new spine to form. This model implies a preferential growth of the new process towards a particular nerve terminal, but how might that be accomplished? During development, it is believed that the protrusion of filopodia induces the aggregation of presynaptic exocytotic machinery at the site of contact with the presynaptic axon (Washbourne et al., 2002). Is it possible that, in more mature tissue, the nerve terminal can attract an outgrowing process to grow towards it?

Richards et al. (2005) have provided compelling evidence that glutamate itself may act as a ‘come hither’ presynaptic signal that attracts growing postsynaptic dendritic processes. Furthermore, their results indicate that growing dendritic processes grow preferentially towards a source of glutamate. Using live cell imaging of green fluorescent protein-labeled CA1 pyramidal cell dendrites in hippocampal slice cultures (PN 27 - 48), Richards et al. noted that blocking action potential-evoked transmitter release with tetrodotoxin (TTX) resulted in an increase in the formation of spinule-like projections from pre-existing dendritic spine heads. With electron microscopy, they showed that these spinule-like projections had a PSD and were opposed to vesicle-containing nerve terminals within two hours of TTX application, and that the original PSD remained after protrusion of the projection. These spinule-like projections thus formed functional synapses rapidly. The authors also showed that a brief (ca. 1 min) application of glutamate pulses from an iontophoretic electrode induced the outgrowth of spinule-like processes from spine heads. The glutamate induced processes were about twice as long (ca. 1.5 µm) as those arising spontaneously in the presence of TTX. Most remarkably, these glutamate-induced processes grew directly towards the iontophoretic electrode. These results clearly establish that exogenous glutamate can promote dendritic process outgrowth, much like tetanic stimulation, and that these processes grow towards a source of glutamate, given the right conditions. There is some ambiguity in the pharmacological identity of the postsynaptic receptors mediating these responses, but there is at least some evidence for a role of NMDARs. In addition, it was not shown that synaptically released glutamate could produce a similar outgrowth. Nevertheless, we can speculate that the formation of same-dendrite, multiple-spine boutons, observed by Toni et al. (1999), may arise from directed process outgrowth from near the base of the original spine towards the strongly active presynaptic terminal that took part in the induction of LTP in the postsynaptic cell. In this model, one spine would be the original spine whereas the other would have grown out from the dendritic shaft (Figure 2D).

Bigger is better

A change in synapse number may not be the only correlate of synaptic plasticity. Equally compelling evidence is starting to emerge that changes in the size of dendritic spines accompany both LTP and LTD. Dendritic spines come in a variety of sizes, both in the volume of the spine head and in the thickness and length of the spine neck (Harris and Stevens, 1989; Harris et al., 1992). Changes in spine size occur during development and synaptogenesis (Knafo et al., 2005), resulting in a decrease in the range of these parameters as the brain matures (Harris et al., 1992). It has long been presumed that these changes reflect, in part, the responses of spines to the large amounts of learning that inevitably occur in the central nervous system after birth (Newey et al., 2005). Indeed, there is classic evidence that the amount of intellectual stimulation provided by an animal’s environment results in dramatic differences in the richness of dendritic spines (Greenough et al., 1978).

What is the relationship between spine size and synaptic function? A direct answer to this question has recently been provided by Matsuzaki et al. (2001). Using TPSLM to photolyze a caged glutamate compound in a small volume corresponding roughly to the size of a single spine head, they were able to stimulate α-amino-3-hydroxy-5-methyl-4-isoxazolepropionate-preferring glutamate receptors (AMPARs) on single dendritic spines on CA1 pyramidal cells in hippocampal slice cultures (PN 14 – 18) and measure the amplitude of the resulting postsynaptic currents with simultaneous whole-cell voltage-clamp recordings. They observed a strong correlation between spine head volume and the amplitude of the excitatory postsynaptic current, suggesting that larger spine heads contained more AMPARs. This finding is consistent with the positive correlation between spine head size and postsynaptic density area. We have confirmed this observation using single photon glutamate photolysis in PN 21 – 28 rat hippocampal slice cultures (Bagal et al., 2005). These observations again raise the possibility that the largest spines are those at which LTP has previously been induced through the animal’s prior experience. Furthermore, if LTP involves the insertion of AMPARs into the postsynaptic membrane, then must LTP by accompanied by an increase in spine head size?

Matsuzaki et al. (2004) went on to use the same techniques to address this question. CA1 pyramidal cells were transfected biolistically with green fluorescent protein to allow spines to be visualized. LTP was induced in an extracellular saline containing a low concentration of Mg²⁺. High frequency trains of UV light pulses were delivered to release caged glutamate and the resulting LTP was recorded electrophysiologically. Using this stimulation protocol, they observed a >50% increase in spine head volume in 90% of the spines in which they induced LTP, as well as potentiation of AMPAR-mediated currents recorded from the stimulated neurons. The greatest spine head enlargement was approximately 200% over baseline and was maximal within 1-5 minutes after stimulation. This was followed by a partial recovery in spine head volume at 10 min after tetanization. Nevertheless, spine head enlargement remained at an average of 50% over baseline for up to 60 min. Only 9% of spines outside the uncaging area were said to become enlarged during this time, suggesting that the uncaging protocol resulted in the stimulation of only one spine and that the volume change was synapse specific. The degree of spine head enlargement depended strongly on the initial, naive spine size. Although almost all spines showed a transient enlargement, 55% of small spines, but almost no large spines, became persistently increased in volume after stimulation with caged glutamate. High frequency Schaffer collateral stimulation produced similar increases in spine head volume. Like conventional LTP, both the transient and sustained increase in spine head size were prevented by the NMDAR antagonist APV. Interestingly, the CaMK inhibitor KN62 and the inhibitor of actin polymerization, latrunculin A, inhibited the long lasting increase in spine head volume, but not the transient increase.

These findings suggest that spine enlargement induced by glutamate release concurrent with postsynaptic potentiation is due to LTP induction (Figure 2E). Consistent with observations that smaller spine heads contain smaller PSDs (Chicurel and Harris, 1992) and fewer AMPARs (Matsuzaki et al., 2001), the amount of potentiation of exogenous glutamate responses was greater at small spines than at large spines (Matsuzaki et al., 2004), as was the increase in spine head volume. In their experiments, 44% of small spines showed a significant increase in the amplitude of AMPAR-mediated currents, whereas large spines displayed a 25% decrease in AMPAR current. Furthermore, all spines that displayed potentiation of AMPAR currents also displayed an increase in volume, although not all spines that displayed an increase in volume also displayed potentiation.

One interpretation of these results is that small spines may represent the "silent" synapses described in electrophysiological studies (Isaac et al., 1995; Liao et al., 1995), or at least have lower AMPAR to NMDAR ratios. This also suggests that the insertion of more AMPARs occurs pari passu with the spine head enlargement. That is, the spine head must become enlarged in order to accommodate the new AMPARs. Furthermore, there may be an upper limit to spine size and space to insert new AMPARs, thus accounting for saturation of LTP.

One interesting aspect of this model is that it can seemingly account for the maturation of spines during development. During the third to sixth postnatal weeks in the rat, spine heads generally become larger and more uniform and small headed spines are no longer apparent (Harris et al., 1992). Spine maturation might therefore reflect the LTP that occurs during ongoing learning and memory formation. Even more speculatively, the achievement of uniformly large spine heads, incapable of expressing LTP, may herald the end of the critical period and its associated plasticity in cortical regions.

Spine head growth may not be an obligatory concomitant of LTP, however, according to results from our laboratory (Bagal et al., 2005). Using comparable cultured hippocampal slices, but single photon techniques for uncaging glutamate and stimulating individual dendritic spines, we also observed potentiation of the responses to the uncaged glutamate, but two significant differences. First, we did not detect any difference in the amount of potentiation expressed by small and large spines. Second, we did not detect any difference in spine head volume after the induction of the potentiation.

The differences in the results of the two studies may reflect methodological differences. Particularly relevant may be the difference in stimulation protocols. Matsuzaki et al. used 1 - 2 Hz trains of glutamate pulses for 1 min to induce potentiation whereas Bagal et al. used a single glutamate pulse paired with one brief depolarizing voltage step (200 ms to 10 mV). One possible explanation is therefore that the increase in spine head size reflects disturbances in ionic gradients as the result of intense stimulation. While the Matsuzaki et al. stimulation protocol is the same as that used in many conventional synaptic LTP experiments, it is important to note that conventional synaptic stimulation causes glutamate release at a given spine in response to only some fraction of stimuli (Murthy et al., 1997) whereas photostimulation always releases glutamate. It is at least possible therefore that the swelling observed by Matsuzaki et al. reflects excessive stimulation. Arguing against this possibility, however, is the finding that the persistent spine head swelling is prevented by a CaMK antagonist.

Another key methodological difference is the use of wide field illumination versus TPLSM to image the spines. Because the depth of field is wider with the former approach, and light is collected from fluorophores above and below the focal plane, the former technique is less sensitive to small differences in focus. Indeed, there is evidence of focus problems in the data published by Matsuzaki et al., in our opinion.

On the other hand, Matsuzaki et al. used perforated patch-clamp recording techniques, in which there is little of no dialysis of the cell cytosol whereas Bagal et al. used whole-cell recording, in which intracellular dialysis is substantial. We have observed that ongoing, constitutive spine head motility (e.g. Fischer et al., 1998) is considerably less in cells recorded with whole-cell electrodes than in undisturbed cells transfected with GFP (unpublished observations). Nevertheless, LTP is also very sensitive to intracellular dialysis (e.g. Malinow and Tsien, 1990) but was of a comparable magnitude in both studies.

At the very least, the ability of Bagal et al. to detect potentiation of glutamate responses without spine head enlargement suggests that changes in spine head size are not necessary for the expression of LTP. Indeed, Lang et al. (2004) used TPLSM as well as conventional synaptic LTP-inducing protocols and found that, regardless of spine shape, LTP-inducing stimuli did not cause an appreciable change in spine morphology, although they did find a transient expansion of spine head size after LTP-inducing stimulation. Furthermore, spine morphology remains relatively stable for periods of months in vivo (Grutzendler et al., 2002) and even after induction of epileptiform seizures in vivo (Mizrahi et al., 2004), arguing that spine shape changes do not necessarily result from intense synaptic activity.

Honey, I shrunk the spine!

LTD is the systemic weakening of synaptic efficacy in response to either high frequency stimulation (in the cerebellum) or low frequency stimulation (in the hippocampus, Figure 4). The depression of synaptic responses can last for hours or weeks and has been implicated in learning (Kemp and Manahan-Vaughan, 2004). In many respects, LTD is the inverse of LTP and, indeed, LTD-inducing stimulation protocols reverse LTP in electrophysiological experiments and visa versa (Dudek and Bear, 1993). But what about morphological changes?

Mu-Ming Poo and colleagues (Zhou et al., 2004) examined whether stimulation patterns that induce LTD lead to morphological changes in dendritic spines. They loaded CA1 pyramidal neurons in ex vivo slices taken from PN 14 - 18 rats with the fluorescent dye calcein, placed stimulating electrodes about 20-30 µm away from the neuron of interest, and used TPLSM to visualize spines before and after stimulation. A standard LTD-inducing low frequency stimulation protocol (LFS, 1 Hz, 15 minutes) triggered a decrease in the diameter of spine heads (Figure 4C). The reduction in spine head diameter averaged about 70% and persisted for >30 min, thus paralleling the physiologically assessed decrease in excitatory synaptic transmission. Another 6% of spines retracted and disappeared after a single episode of LTD-inducing LFS. LFS thus seems to produce opposite morphological changes seen with LTP: decreases in spine size and number.

To determine whether functional and morphological changes accompanying LTP were reversible, Zhou et al. first delivered a high-frequency stimulation (HFS) protocol (900 pulses at 100 Hz), which induced LTP and caused a small increase in spine head size (16%), thus confirming the results of Matsuzaki et al. (2004). When LFS was delivered 30 minutes after HFS, they observed a reduction in the size of the spine heads that apparently reversed the growth observed due to HFS, although it is unclear whether the same spines were assayed in these experiments. Similarly, LFS induced decreases in spine head size were reversed by HFS. The results suggest that LTP and LTD are associated not only with opposite morphological effects, but that these morphological effects are reversible. It should be noted, however, that this does not mean that they share a common signaling pathway.

What signaling pathways are important for mediating LTD-associated changes in spine size? As with electrophysiological depression, Zhou et al. observed no change in spine size after LFS delivered in the presence of AP5. Calcineurin (PP2B) and its downstream effector PP1 have been implicated in electrophysiological LTD (Mulkey et al., 1994; Gerges et al., 2005; Morishita et al., 2005; Xia and Storm, 2005). Consistent with these data, the calcineurin inhibitor FK506 prevented spine shrinkage in response to LFS. Interestingly, however, okadaic acid and calyculin A, inhibitors of PP1 and PP2A, prevented LTD but not spine shrinkage. It thus appears that electrophysiological and morphological responses to LFS share common initial induction mechanisms, but that the pathways then diverge downstream from the activation of NMDARs.

The actin cytoskeleton is important in regulating the morphological changes that occur in dendritic spines (Smart and Halpain, 2000) and is sensitive to NMDAR-mediated Ca²⁺ influx (Rosenmund and Westbrook, 1993). One protein that is involved in the regulation of actin polymerization and affects the morphology of dendritic spines is cofilin. Cofilin is a protein that increases the number of pointed ends on actin and increases the rate of depolymerization of gelsolin-capped actin filaments (Fass et al., 2004). The activity of cofilin is inhibited by LIM kinase-mediated phosphorylation and stimulated by dephosphorylation (Wang et al., 2005). To determine if cofilin is involved in promoting the morphological changes resulting from LTD induction, Zhou et al. designed two peptides: each corresponding to the N-terminal sequence of the cofilin protein, one mimicking a phosphorylated ser-3 site and one without. Phosphorylated cofilin (P-cofilin) competes with the endogenous cofilin as a substrate for phosphatases thus reducing dephosphorylation of endogenous cofilin. This should, in turn, prevent the shrinkage of spines. Indeed, LFS failed to induce spine shrinkage in neurons loaded with P-cofilin via the whole-cell pipette. Loading cells with the non-phosphorylated cofilin peptide led to a small increase in spine shrinkage in the absence of LFS. When LFS was administered for only 10 minutes (as opposed to 15 minutes in original experiments), LFS-induced spine shrinkage in cells loaded with non-phosphorylated cofilin was increased about 15% over LFS-alone when measured 45 minutes after LFS. These results thus implicate changes in the phosphorylation state of cofilin as a critical regulator of activity-dependent morphological plasticity. Consistent with this conclusion, cofilin dephosphorylation, like LFS-induced spine shrinkage, is blocked by FK506 but not okadaic acid (Wang et al., 2005). Cofilin does not seem to be a substrate for CaMK, however, but is phosphorylated at ser-3 by other kinases whose activity are regulated by Rho-family GTPases (Arber et al., 1998).

Can these results be reconciled with those of Nägerl et al. (2004), in which LTD is accompanied by spine retraction? Zhou et al. did see a small amount of spine loss, albeit less than that reported by Nägerl et al. One potential explanation is that Zhou et al. followed the changes in spine head diameter for only one hour after LFS, whereas Nägerl et al. monitored spine loss over the course of six hours. Individual spines varied in the amount of time taken to retract so that maximal spine loss was not observed until about 3 hours of LFS (Nägerl et al., 2004). Had Nägerl et al. measured spine head diameter in addition to spine loss, perhaps they would have observed spine head shrinkage prior to spine retraction. In addition, perhaps Zhou et al. would have observed significant spine retraction if they had monitored the LTD-induced changes in morphology for three hours instead of one. Finally, Zhou et al. used whole-cell recording in acute slices whereas Nägerl et al. imaged un-impaled cells in organotypic hippocampal slice cultures, and this could also be a potential explanation for their discrepant observations.

Taken together, these findings suggest a model in which induction of LTP causes an increase in spine size concomitant with an increase in AMPAR number. Large spine heads, in this model, would be resistant to further LTP and are thus the physical manifestation of stored memory traces. Induction of LTD, in contrast, leads to shrinkage of spines in parallel to the internalization and decrease in AMPAR number (e.g. Carroll et al., 1999) and the subsequent destabilization of the synapse. These synapses, with their decreased efficacy may be candidates for synapse pruning.

Conclusions and Implications

Morphological plasticity in the hippocampus may thus result in either changes in spine size or spine number or both. The morphological endpoint may depend upon the age of the tissue. For example, in the mature nervous system, the actin cytoskeleton may be less ‘plastic,’ rendering full spine retraction in response to LTD-inducing activity patterns more unlikely. Similarly, because it is a fundamental mechanism in synaptogenesis, only the less mature nervous system may have the capacity to generate new spines de novo in response to LTP-inducing activity patterns. It is unfortunate that most of these studies have been performed in tissue at an age that is on the cusp of the maturation of the CNS (ca. 1-3 weeks after birth). More information from the mature adult brain would clearly be useful.

Deformed dendritic spines and deficient spine density are a hallmark of many neurological conditions, notably in virtually every disease in which cognitive performance is impaired. Alzheimer’s disease is perhaps the best characterized neurological disease with significant learning and memory dysfunction, and substantial decreases in dendritic spine density in pyramidal cells of the neocortex and hippocampus are observed in human tissue from Alzheimer’s patients (e.g. Anderton et al., 1998). Dendritic spine loss is reported in other non-Alzheimer’s type dementias (e.g. Mehraein et al., 1975), and may represent a pathological acceleration of the normal decrease in dendritic spine density observed in senescence (Scheibel, 1979; de Ruiter and Uylings, 1987). Furthermore, pyramidal cells in several different forms of mental retardation have a lower than normal density of spines, including Down’s syndrome (Takashima et al., 1994) and fragile X syndrome (Comery et al., 1997). Decreases in spine density and structural synaptic abnormalities are also common in human tissue from psychotic schizophrenic patients (Kung et al., 1998), and in hippocampi from patients suffering from uncontrolled epileptic seizures (Scheibel et al., 1974). It is our sincere hope that a better understanding of the role that dendritic spines play in determining synaptic efficacy and in the expression of synaptic plasticity will provide considerable insights into the relationship between these cognitive deficits and the morphological changes in spines in these disease states.

References

Andersen, P., and Soleng, A.F. (1998). Long-term potentiation and spatial training are both associated with the generation of new excitatory synapses. Brain Res Rev 26, 353–359.

Anderton, B. H., Callahan, L., Coleman, P., Davies. P., Flood, D., Jicha, G. A., Ohm, T., and Weaver, C. (1998). Dendritic changes in Alzheimer’s disease and factors that may underlie these changes. Prog Neurobiol 55, 595-609.

Arber, S., Barbayannis, F.A., Hanser, H., Schneider, C., Stanyon, C.A., Bernard, O., and Caroni, P. (1998). Regulation of actin dynamics through phosphorylation of cofilin by LIM-kinase. Nature 393, 805-809.

Bagal, A.A., Kao, J.P.Y., Tang, C-M., and Thompson, S.M. (2005). Long-term potentiation of exogenous glutamate responses at single dendritic spines. Proc Natl Acad Sci USA 102,14434-4439.

Bliss, T.V., and Lømo, T. (1973). Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. J Physiol Lond 232, 331–356.

Buchs, P.A., Stoppini, L., Parducz, A., Siklos, L., and Muller, D. (1994). A new cytochemical method for the ultrastructural localization of calcium in the central nervous system. J Neurosci Methods 54, 83-93.

Calverley, R.K.S., and Jones, D.G. (1990). Contributions of dendritic spines and perforated synapses to synaptic plasticity. Brain Res Brain Res Rev 15, 215–249.

Carroll, R.C., Lissin, D.V., von Zastrow, M., Nicoll, R.A., and Malenka, R.C. (1999). Rapid redistribution of glutamate receptors contributes to long-term depression in hippocampal cultures. Nat Neurosci 2, 454-60.

Chang, F.L. and Greenough, W.T. (1984). Transient and enduring morphological correlates of synaptic activity and efficacy change in the rat hippocampal slice. Brain Res 309, 35–46.

Chicurel, M.E., and Harris, K.M. (1992). Three-dimensional analysis of the structure and composition of CA3 branched dendritic spines and their synaptic relationships with mossy fiber boutons in the rat hippocampus. J Comp Neurol 325, 169-82.

Comery, T. A., Harris, J. B., Willems, P. J., Oostra, B. A., Irwin, S. A., Weiler, I. J., and Greenough, W. T. (1997). Abnormal dendritic spines in fragile X knockout mice: maturation and pruning deficits. Proc Natl Acad Sci USA 94, 5401-5404.

Cooney, J.R., Hurlburt, J.L., Selig, D.K., Harris, K.M., and Fiala, J.C. (2002). Endosomal compartments serve multiple hippocampal dendritic spines from a widespread rather than a local store of recycling membrane. J Neurosci 22, 2215-24.

Crick, F. (1982). Do dendritic spines twitch? Trends Neurosci 5, 44–46.

Dailey, M. E., and Smith, S. J. (1996). The dynamics of dendritic structure in developing hippocampal slices. J Neurosci 16, 2983–2994.

Derkach, V., Barria, A., and Soderling, T.R. (1999). Ca²⁺/calmodulin-kinase II enhances channel conductance of alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionate type glutamate receptors. Proc Natl Acad Sci U S A 96, 3269-74.

de Ruiter, J. P., and Uylings, H. B. (1987). Morphometric and dendritic analysis of fascia dentate granule cells in human aging and senile dementia. Brain Res 402, 217-229.

De Simoni, A., Griesinger, C.B., and Edwards, F.A. (2003). Development of rat CA1 neurones in acute versus organotypic slices: role of experience in synaptic morphology and activity. J Physiol 550, 135 - 147.

Desmond, N.L. and Levy, W.B. (1986a). Changes in the numerical density of synaptic contacts with long-term potentiation in the hippocampal dentate gyrus. J Comp Neurol 253, 466–475.

Desmond, N.L. and Levy, W.B. (1986b). Changes in the postsynaptic density with long-term potentiation in the dentate gyrus. J Comp Neurol 253, 476-82.

Dudek, S.M., and Bear, M.F. (1993). Bidirectional long-term modification of synaptic effectiveness in the adult and immature hippocampus. J Neurosci 13, 2910-8.

Edwards, F.A. (1995). Anatomy and electrophysiology of fast central synapses lead to a structural model for long-term potentiation. Physiol Rev 75, 759-87.

Engert, F. and Bonhoeffer, T. (1999). Dendritic spine changes associated with hippocampal long-term synaptic plasticity. Nature 399, 66–70.

Fass, J., Gehler, S., Sarmiere, P., Letourneau, P., and Bamburg, J.R. (2004). Regulating filopodial dynamics through actin-depolymerizing factor/cofilin. Anat Sct Int 79, 173-183.

Fiala, J.C., Allwardt, B., and Harris, K.M. (2002). Dendritic spines do not split during hippocampal LTP or maturation. Nat Neurosci 5, 297-8.

Fiala, J.C., Feinberg, M., Popov, V., and Harris, K.M. (1998). Synaptogenesis via dendritic filopodia in developing hippocampal area CA1. J Neurosci 18, 8900–8911.

Fifkova, E. and Anderson, C.L. (1981). Stimulation-induced changes in dimensions of stalks of dendritic spines in the dentate molecular layer. Exp Neurol 74, 621–627.

Fischer, M., Kaech, S., Wagner, U., Brinkhaus, H. and Matus, A. (2000). Glutamate receptors regulate actin-based plasticity in dendritic spines. Nat Neurosci 3, 887–894.

Fischer, M., Kaech, S., Knutti, D. and Matus, A. (1998). Rapid actin based plasticity in dendritic spines. Neuron 20, 847–854.

Fujii, S., Saito, K., Miyakawa, H., Ito, K., and Kato, H. (1991). Reversal of long-term potentiation (depotentiation) induced by tetanus stimulation of the input to CA1 neurons of guinea pig hippocampal slices. Brain Res 555, 112-22.

Geinisman, Y. (2000). Structural synaptic modifications associated with hippocampal LTP and behavioral learning. Cereb Cortex 10, 952–962.

Gerges, N.Z., Alzoubi, K.H., and Alkadhi, K.A. (2005). Role of phosphorylated CaMKII and calcineurin in the differential effect of hypothyroidism on LTP of CA1 and dentate gyrus. Hippocampus 15, 480-90.

Greenough, W.T., West, R.W., and DeVoogd, T.J. (1978). Subsynaptic plate perforations: changes with age and experience in the rat. Science 202, 1096-8.

Grutzendler, J., Kasthuri, N., and Gan, W.B. (2002). Long-term dendritic spine stability in the adult cortex. Nature 420, 812-816.

Harris, K.M. (1999). Structure, development, and plasticity of dendritic spines. Curr Opin Neurobiol 9, 343–348.

Harris, K. M., Jensen, F. E. and Tsao, B. (1992). Three-dimensional structure of dendritic spines and synapses in rat hippocampus (CA1) at postnatal day 15 and adult ages: implications for the maturation of synaptic physiology and long-term potentiation. J Neurosci 12, 2685–2705.

Harris, K.M., and Kater, S.M. (1994). Dendritic spines: cellular specializations imparting both stability and flexibility to synaptic function. Ann Rev Neurosci 17, 341-371.

Harris, K.M. and Stevens, J.K. (1989). Dendritic spines of CA 1 pyramidal cells in the rat hippocampus: serial electron microscopy with reference to their biophysical characteristics. J Neurosci 9, 2982 - 2997.

Hayashi, Y., Shi, S-H., Esteban, J.A., Piccini, A., Poncer, J-C., and Malinow, R. (2000). Driving AMPA Receptors into Synapses by LTP and CaMKII: Requirement for GluR1 and PDZ Domain Interaction. Science 287, 2262 - 2267.

Hebb, D.O. (1949). The Organization of Behavior. New York: John Wiley Inc.

Holtmaat, A.J., Trachtenberg, J.T., Wilbrecht, L., Shepherd, G.M., Zhang, X., Knott, G.W., and Svoboda, K. Transient and persistent dendritic spines in the neocortex in vivo. Neuron 45, 279-291.

Isaac, J.T., Nicoll, R.A., and Malenka, R.C. (1995). Evidence for silent synapses: implications for the expression of LTP. Neuron 15, 427-434.

Jontes, J. D. and Smith, S. J. (2000). Filopodia, spines, and the generation of synaptic diversity. Neuron 27, 11–14.

Jourdain, P., Fukunaga, K., and Muller, D. (2003). Calcium/calmodulin-dependent protein kinase II contributes to activity-dependent filopodia growth and spine formation. J Neurosci 23, 10645 - 10649.

Kemp, A., and Manahan-Vaughan, D. (2004). Hippocampal long-term depression and long-term potentiation encode different aspects of novelty acquisition. Proc Natl Acad Sci U S A 101, 8192-7.

Kim, E., and Sheng, M. (2004). PDZ domain proteins of synapses. Nat Rev Neurosci 5, 71-81.

Kirov, S.A., Goddard, C.A., and Harris, K.M. (2004). Age-dependence in the homeostatic upregulation of hippocampal dendritic spine number during blocked synaptic transmission. Neuropharmacology 47, 640-648.

Knafo, S., Libersat, F., and Barkai, E. (2005). Olfactory learning-induced morphological modifications in single dendritic spines of young rats. Eur J Neurosci 21, 2217-2226.

Kung, L., Conley, R., Chute, D. J., Smialek, J., and Roberts, R. C. (1998) Synaptic changes in the striatum of schizophrenic cases: a controlled postmortem ultrastructural study. Synapse 28, 125-139.

Lang, C., Barco, A., Zablow, L., Kandel, E.R., Siegelbaum, S.A., and Zakharenko, S.S. (2004). Transient expansion of synaptically connected dendritic spines upon induction of hippocampal long-term potentiation. Proc Natl Acad Sci U S A 101, 16665-16670.

Lee, K. S., Schottler, F., Oliver, M. and Lynch, G. (1980). Brief bursts of high-frequency stimulation produce two types of structural change in rat hippocampus. J Neurophysiol 44, 247-258.

Lee, H.K., Barbarosie, M., Kameyama, K., Bear, M.F. and Huganir, R.L. (2000). Regulation of distinct AMPA receptor phosphorylation sites during bidirectional synaptic plasticity. Nature 405, 955-959.

Liao, D., Hessler, N.A., Malinow, R. (1995). Activation of postsynaptically silent synapses during pairing-induced LTP in CA1 region of hippocampal slice. Nature 375, 400-404.

Majewska, A., Sur, M. (2003). Motility of dendritic spines in visual cortex in vivo: changes during the critical period and effects of visual deprivation. Proc Natl Acad Sci U S A 100, 16024-16029.

Malenka, R.C., and Bear, M.F. (2004). LTP and LTD: an embarrassment of riches. Neuron 44, 5-21.

Malenka, R.C. and Nicoll, R.A. (1999) Long-Term Potentiation--A Decade of Progress? Science 285, 1870-1874.

Maletic-Savatic, M., Malinow, R. and Svoboda, K. (1999). Rapid dendritic morphogenesis in CA1 hippocampal dendrites induced by synaptic activity. Science 283, 1923–1927.

Malinow R. (2003). AMPA receptor trafficking and long-term potentiation. Philos Trans R Soc Lond B Biol Sci 358, 707-714.

Malinow, R., Mainen, Z.F. and Hayashi, Y. (2000). LTP mechanisms: from silence to four-lane traffic. Curr Opin Neurobiol 10, 352-357.

Malinow, R. and Tsien, R.W. (1990). Presynaptic enhancement shown by whole-cell recordings of long-term potentiation in hippocampal slices. Nature 346, 177-80.

Marrs, G.S., Green, S.H., and Dailey, M.E. (2001). Rapid formation and remodeling of postsynaptic densities in developing dendrites. Nat Neurosci 4, 1006-1013.

Martin, K.C., Barad, M., and Kandel, E.R.(2000). Local protein synthesis and its role in synapse-specific plasticity. Curr Opin Neurobiol. 10, 587-592.

Matsuzaki, M., Ellis-Davies, G.C., Nemoto, T., Miyashita, Y., Iino, M., and Kasai, H. (2001). Dendritic spine geometry is critical for AMPA receptor expression in hippocampal CA1 pyramidal neurons. Nat Neurosci 4, 1086-92.

Matsuzaki, M., Honkura, N., Ellis-Davies, G.C. and Kasai, H. (2004). Structural basis of long-term potentiation in single dendritic spines. Nature 429, 761–766.

McKinney, R.A., Capogna, M., Dürr, R., Gähwiler, B.H., and Thompson, S.M. (1999). Miniature synaptic events maintain dendritic spines via AMPA receptor activation. Nat Neurosci 2, 44-49.

Mehraein, P., Yamada, M., and Tarnowska-Dziduszko, E. (1975). Quantitative Study on dendrites and dendritic spines in Alzheimer’s disease and senile dementia. Adv Neurol 12, 453-458.

Mizrahi, A., Crowley, J.C., Shtoyerman, E., and Katz, L.C. (2004). High-resolution in vivo imaging of hippocampal dendrites and spines. J Neurosci 24, 3147-3151.

Morishita, W., Marie, H., and Malenka, R.C. (2005). Distinct triggering and expression mechanisms underlie LTD of AMPA and NMDA synaptic responses. Nat Neurosci 8, 1043-1050.

Morris, R.G., Moser, E. I., Riedel, G., Martin, S. J., Sandin, J., Day, M. and O'Carroll, C. (2003) Philos Trans R Soc Lond B Biol Sci 358, 773-786.

Moser, M.B., Trommald, M. and Andersen, P. (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.

Mulkey, R.M., Endo, S., Shenolikar, S., and Malenka, R.C. (1994). Involvement of a calcineurin/inhibitor-1 phosphatase cascade in hippocampal long-term depression. Nature 369, 486-488.

Murthy, V.N., Sejnowski, T.J., and Stevens, C.F. (1997). Heterogeneous release properties of visualized individual hippocampal synapses. Neuron 18, 599-612.

Nägerl, U.V., Eberhorn, N., Cambridge, S.B., and Bonhoeffer, T. (2004). Bidirectional activity-dependent morphological plasticity in hippocampal neurons. Neuron 44, 759-767.

Newey, S.E, Velamoor, V., Govek, E.E., and Van Aelst, L. (2005). Rho GTPases, dendritic structure, and mental retardation. J Neurobiol 64, 58-74.

Nimchinsky, E.A., Sabatini, B.L. and Svoboda, K. (2002). Structure and function of dendritic spines. Annu Rev Physiol 64, 313–353.

Okabe, S., Miwa, A., and Okado, H. (2001). Spine formation and correlated assembly of presynaptic and postsynaptic molecules. J Neurosci. 21, 6105-6114.

Richards, D.A., Mateos, J.M., Hugel, S., de Paola, V., Caroni, P., Gähwiler, B.H, and McKinney, R.A. (2005). Glutamate induces the rapid formation of spine head protrusions in hippocampal slice cultures. Proc Natl Acad Sci U S A 102, 6166-6171.

Rosenmund, C., and Westbrook, G.L. (1993). Calcium-induced actin polymerization reduces NMDA channel activity. Neuron 10, 805-814

Scheibel, A. B. (1979) The hippocampus: organizational patterns in health and senescence. Mech Ageing Dev 9, 89-102.

Scheibel, A. B., Crandall, P. H., and Scheibel, A. B. (1974). The hippocampal-dentate complex in temporal lobe epilepsy. Epilepsia 15, 55-80.

Smart, F.M., and Halpain, S. (2000). Regulation of dendritic spine stability. Hippocampus 10, 542-554.

Sorra, K.E. and Harris, K.M. (1998). Stability in synapse number and size at 2 hr after long-term potentiation in hippocampal area CA1. J Neurosci 18, 658–671.

Takashima, S., Iida, K., Mito, T., and Arima, M. (1994) Dendritic and histochemical development and aging in patients with Down’s syndrome. J Intellect Disabil Res 38, 265-273.

Toni, N., Buchs, P.A, Nikonenko, I., Bron, C.R., and Muller, D. (1999). LTP promotes formation of multiple spine synapses between a single axon terminal and a dendrite. Nature 402, 421-425.

Wagner, J.J., and Alger, B.E. (1995). GABAergic and developmental influences on homosynaptic LTD and depotentiation in rat hippocampus. J Neurosci 15, 1577-1586.

Wang, Y., Shibasaki, F., and Mizuno, K. (2005). Calcium signal-induced cofilin dephosphorylation is mediated by Slingshot via calcineurin. J Biol Chem 280, 12683-12689.

Washbourne, P., Bennett, J.E., and McAllister, A.K. (2002). Rapid recruitment of NMDA receptor transport packets to nascent synapses. Nat Neurosci 5, 751-759.

Woolley, C.S., and McEwen, B.S. (1994). Estradiol regulates hippocampal dendritic spine density via an N-methyl-D-aspartate receptor-dependent mechanism. J Neurosci 14, 7680-7687.

Xia, Z., and Storm, D.R. (2005). The role of calmodulin as a signal integrator for synaptic plasticity. Nat Rev Neurosci 6, 267-276.

Yuste, R. and Bonhoeffer, T. (2004). Genesis of dendritic spines: Insights from ultrastructural and imaging studies. Nat Rev Neurosci 5, 24-34.

Zhou, Q., Homma, K.J., and Poo, M.M. (2004). Shrinkage of dendritic spines associated with long-term depression of hippocampal synapses. Neuron 44, 749-57.

Ziv, N., and Smith, S. (1996). Evidence for a role of dendritic filopodia in synaptogenesis and spine formation. Neuron 17, 91–102.