Adapted from an original article by Jennifer Altman which appeared in Alzheimer Actualités, a newsletter published by the Ipsen Foundation.
Although so much has been learned in the past 25 years about the pathological processes underlying Alzheimer disease (AD), there is still little known about how these processes lead to memory loss. One striking early change in the brains of patients with AD is the loss of synapses, and hence of connectivity. In recent years, however, the cellular and molecular mechanisms involved in memory storage have been intensively studied, with particular emphasis on changes in the synaptic connections between neurons during learning. At the Colloquium on Alzheimer’s Disease held in Paris on 16 April 2007, speakers explored how this modern understanding of basic synaptic function and synaptic plasticity can contribute to unraveling the relationship between neurodegeneration and the development of behavioral deficits. The meeting was organized by Dennis Selkoe, Harvard Medical School, and Yves Christen from the Fondation IPSEN, Paris.
Synapses, the tiny junctions through which information is transmitted from one neuron to another, have two parts: the presynaptic axon terminal, or bouton, which releases neurotransmitter, and the post-synaptic membrane on the dendrite, which contains the receptors and channels that respond to the neurotransmitter. The post-synaptic component is often sited on small, needle-like projections termed dendritic spines and is characterized by the post-synaptic density, a complex scaffold containing many different proteins, below the membrane (Morgan Sheng, MIT).
When neurotransmitter molecules attach to the post-synaptic receptors, ions flow through the associated channels, resulting in a change in the electrical charge across the dendritic membrane. This signal spreads through the neuron, integrating with signals from other synapses, until the neuron membrane becomes sufficiently positive to generate a signal along its axon that will result in transmitter release on to the next neuron in the chain.
Until recently synapses were thought to be stable and static structures: once formed, they endured unless subjected to pathological attack. In contrast, research in the past 20 years into the molecular mechanisms of memory storage has demonstrated both a continual turnover of molecules within a synapse and that old synapses disappear and new ones are formed. The speakers at this meeting demonstrated a dazzling range of highly sophisticated and demanding techniques, ranging from tracking individual molecules, in vivo observation and later reconstruction of single synapses, and detailed brain imaging to tracking localized electrical and chemical changes in the brain in active animals. Combined with the powerful tools provided by genetic manipulation of both cells in culture and of mice, these virtuoso approaches are yielding a detailed dissection of the molecules and mechanisms involved in the maintenance and turnover of synapses, as well as some of the pathological processes that affect their well-being.
Modern high-resolution imaging and electrophysiological recording techniques are establishing just how dynamic synapses are. On the structural level, dendritic spines in the cortex of active mice can now be imaged in vivo over several days (Antony Holtmaat, Cold Spring Harbor, Laboratory, New York). In rodents, sensory neurons in the facial whiskers connect with a specialized area of the somato-sensory cortex in which the neurons are arranged in barrel-like clusters. The main branching of these neurons, the axon terminals and the dendritic trees, are very stable. However, on the dendrites, two types of spines are seen, one large and persistent, the other thinner and transient. When all but one whisker are shaved away, the barrel field representing the remaining whisker expands and new synapses form. Painstaking reconstructions from electron micrographs reveal that two-thirds are new contacts within existing boutons and the rest are in new boutons. The behavior of the newly forming synapses indicates that they “sample and select” where they will make contacts (see Holtmaat et al., 2005; Holtmaat et al., 2006; and ARF related news story).
The movements of individual receptor molecules can now be traced using a type of fluorescent label known as a quantum dot. This and similar techniques are showing that the post-synaptic membrane is always in movement: receptor and channel molecules, scaffold proteins in the post-synaptic density, and even membrane lipids are in a continual dance (Antoine Triller, Inserm-École Normale Supérieure, Paris, France) (see Bouzigues et al., 2007). Receptors diffuse out from the post-synaptic membrane into the surrounding extra-synaptic membrane and back and may even move into a neighboring synapse. Although the movement is largely determined by diffusion, the rate at which molecules diffuse is affected by several factors, such as synaptic activity and calcium concentration in the post-synaptic zone. Interactions between receptors, the underlying scaffold proteins, and the cytoskeleton mean that receptors become trapped at certain points, slowing down their progress. Some receptors seem to stay relatively long in the synaptic membrane, whereas others move rapidly in and out—these may represent two populations, one involved in maintaining the synapse, the other in plasticity. The emerging view of the synapse is of a field composed of many molecules that is constantly renewing its components (see ARF related news story).
Receptors not only move around in the membrane but also cycle between the membrane and the underlying cytoplasm. Some excitatory synapses use glutamate as a transmitter, which binds to post-synaptic AMPA receptors. As a synapse becomes more active, the number of AMPA receptors increases, so that transmission across the synapse strengthens—this is the basis of long-term potentiation (LTP), an experimental model of memory storage (see below). Conversely, in underused synapses, AMPA receptor numbers decline and transmission weakens, a phenomenon known as long-term depression (LTD).
The AMPA receptors are transported in and out of the membrane by a complex sequence of chemical reactions, involving a family of GTPases known as Rab proteins (José Esteban, University of Michigan Medical School, Ann Arbor). Assisted by Rab 11, AMPA receptors stored in an intracellular pool are moved to a pool in the head of the spine. From there they are inserted into the synaptic membrane by a process known as exocytosis, which involves Rab 8 working with a complex of proteins in the post-synaptic density termed the exocyst (see, e.g., Gerges et al., 2006). From the membrane, AMPA receptors are returned to the intracellular pool by endocytosis, which requires the help of first Rab 5 and then Rab 4. All these molecules are, of course, liable to mutations that affect the way synapses work, and may in time provide therapeutic targets for supporting the health of synapses.
The classic model of the molecular mechanisms thought to be responsible for establishing memories is derived from the experimental generation of LTP and LTD. Put simply, LTP is a strengthening of transmission across a synapse in response to coincident excitation from different sources; LTD is the converse: transmission weakens when the inputs are not excited coincidentally. Establishing LTP requires that glutamate released by the presynaptic terminal first activates AMPA receptors, which results in the electrical potential across the post-synaptic membrane becoming more positive; this brings a second type of glutamate receptor, the NMDA receptor, into play. Calcium ions entering through the NMDA receptor-associated channel trigger the biochemical processes that result in more AMPA receptors appearing in the post-synaptic membrane, so strengthening transmission across that synapse.
The molecular circuit involved in LTD, which results in the weakening and sometimes even disappearance of a synapse, seems more complex. One cause of neuron death, for example, after a stroke, is massive overstimulation of NMDA receptors resulting in a rapid increase in calcium-ion concentration in the post-synaptic dendrites. This in turn activates the enzymes of the programmed cell-death pathway, or apoptosis. Now, pharmacological and genetic evidence is pointing to a low-level activation of some components of the apoptotic pathway under conditions that result in LTD (Sheng). NMDA receptors seem also to be necessary for maintaining the health of dendritic spines, possibly through their physical presence in the post-synaptic density (Bernardo Sabatini, Harvard Medical School). (See recent review by Alvarez and Sabatini, 2007.)
Amyloid-β and Synaptic Health
Conventional pathology has long shown a loss of synapses in areas close to deposits of amyloid-β (Aβ) in brains of patients with Alzheimer disease. The Aβ peptides that form fibrils in these deposits derive from the amyloid precursor protein (APP), and the biological functions of APP and its derivatives are key questions. This is especially important in view of attempts to produce drugs and vaccines to reduce the production of Aβ. Here, attention moves to the presynaptic terminals, which release Aβ when stimulated. However, the Aβ then depresses further synaptic activation—an effect seen only at excitatory synapses (Roberto Malinow, Cold Spring Harbor, Laboratory, New York State, USA; Edward Koo, University of California, San Diego, USA). (See ARF related news story; Hsieh et al., 2006.)
The release of Aβ has been measured in the cortex of active mice using a microdialysis technique that detects chemical changes in the fluid surrounding neurons (John Cirrito, Washington University, St Louis, Missouri). In these experiments, synaptic activity, whether natural or evoked by electrical stimulation, rapidly and dynamically modulated the level of Aβ around the terminals (see ARF related news story). Work with in vitro brain slices has established that Aβ is not present in the synaptic vesicles that contain the neurotransmitter. Its release is, however, driven by the exocytosis of the vesicles and also depends on uptake of APP from the extra-neuronal space.
The LTP/LTD model is being used to examine how Aβ affects synaptic transmission. Both the form of Aβ commonly found as fibrils in plaques, Aβ42, and a mutated form of the normally more benign Aβ40 mimic the mechanism of LTD and rapidly inhibit the maintenance of LTP. The basal transmission across the synapses is not affected (Michael Rowan, Trinity College, Dublin, Ireland). In slices of hippocampus tested in vitro, increasing Aβ production by applying APP increases the endocytosis of AMPA receptors from the post-synaptic membrane, just as happens in LTD (Malinow). The reduction of functional AMPA receptors leads to a concomitant decrease in NMDA receptor function, leading to loss of synapses and then of spines. The increased internalization of AMPA receptors and the inhibition of LTP are probably both caused by Aβ directly reducing the influx of calcium ions through NMDA receptors (Sabatini).
Aβ—Big or Small?
The loss of dendritic spines is one of the earliest changes in AD. Brain imaging combined with three-dimensional computer reconstruction is revealing details of the course of this degeneration in the brains of mice engineered to produce human forms of Aβ (Floyd Bloom, The Scripps Research Institute, La Jolla, California). By 40 days, the volume of the hippocampus is substantially less than in control mice; the loss is greatest in the dentate gyrus, where neurons that receive their inputs from the entorhinal cortex already have fewer dendritic spines. (Both the dentate gyrus and the entorhinal cortex are the earliest areas to be affected in humans developing Alzheimer disease.) However, the amyloid deposits do not appear in the brains of these mice until about 7 months. By 12 months, substantial deposits of compact amyloid are present, forming sheets along synaptic planes; after this, most additional amyloid is diffuse rather than compact (see ARF related news story).
Although clearly not related to Aβ deposits, the early loss of spines in the dentate gyrus could be caused by recently discovered, soluble forms of oligomeric Aβ. This form of Aβ seems to have direct and deleterious effects on spine health and synaptic function (Sabatini; Selkoe; William Klein, Northwestern University, Evanston, Illinois; Rowan; Koo). Aβ oligomers have been found in the brains of mice carrying human APP genes and in brains of Alzheimer patients (Selkoe; Klein).
Oligomers composed of two or three Aβ molecules can be collected from cultured cells containing the APP gene. When used to bathe hippocampal slice preparations, these oligomers result in a lower density of dendritic spines (Sabatini) and inhibit the maintenance of LTP (Selkoe; Rowan); injecting them into the brains of rats temporarily compromises learning ability (Selkoe). Trimers are more potent than dimers, while monomers have no effect. Dimers and trimers seem to be more stable at 12 degrees C, whereas at body temperature aggregates containing 12 molecules, known as ADDLs (Aβ-derived diffusible ligands), are more prevalent (Klein). These too inhibit the later phase of LTP, enhance LTD, cause aberrations in spine number and shape, and have a deleterious effect on memory (see ARF related news story and ARF news story).
How the oligomers disrupt spines and synapses is slowly being teased out. ADDLs bind to the dendrites of certain cortical neurons, mostly on their spines and close to the NMDA receptors (Klein). However, the molecule that binds any form of Aβ to the membrane has still not been identified. Once attached, both large and small oligomers suppress the action of various molecules in the post-synaptic density, particularly those involved in the opening of the NMDA receptor-associated channels (Sabatini; Selkoe; Klein). The small oligomers also seem to stimulate several proinflammatory molecules, including COX2, and TNFα, that have long been implicated in the pathology of AD (Rowan). Particularly interesting is recent evidence that Aβ oligomers interfere with the receptors for the hormone insulin, which regulates the supply of glucose to cells throughout the body (Selkoe, Klein). Loss of spines and inhibition of LTP can be reversed by chemicals that interact with many of the molecules involved, as well as by antibodies that specifically inactivate different forms of Aβ; surprisingly, the suppression of the insulin receptor can be reversed by increasing the level of insulin. All of these provide possible leads for developing treatments for AD.
Yet another possible deleterious action of Aβ may result from its ability to bind to APP itself (Koo). The part of the APP molecule that lies inside the membrane, known as the C-terminal domain, contains a site that can be cleaved by the apoptotic enzyme caspase—once again linking the biology of APP to the cell-death pathway (see above). The fragment produced by this cleavage can be released from neurons and is toxic to cells in vitro. When Aβ binds to APP, it seems to enhance the likelihood of this caspase cleavage and the production of this toxic fragment (see Saganich et al., 2006 and ARF related news story).
The mechanisms by which synapses may be disrupted in AD could reflect what happens during the initial generation of synapses early in life. Many developmental behavioral problems, such as those seen in children with autistic spectrum disorders (ASD) and related conditions, such as Rett’s and Fragile X syndrome, are being tracked back to a disturbance in the normal pattern of synapse formation during brain development (Thomas Bourgeron, Institut Pastuer, Paris, France). Genetic mutations have been identified in children affected by ASD that alter proteins in the post-synaptic density involved in the generation and stabilization of synapses. These proteins may well also be implicated in loss of synapses in AD. Another genetic mutation related to ASD causes the loss of an enzyme needed for the generation of melatonin, a hormone that helps regulate the sleep-wake cycle. As sleep disturbances are common in AD, the activity of this enzyme may well be worth investigating (see Durand et al., 2007 and related ARF live discussion on Fragile X syndrome).
Deepening knowledge of synaptic structure and function is clearly extending the understanding of the subtle interplay of physiology and pathology at the root of neurodegenerative diseases. Many lines of research are now converging to provide a—sometimes literally—more three-dimensional view of how synapses are maintained and destroyed. A detailed picture is emerging that is beginning to provide the sound biological basis needed for the development of therapeutic tools for ameliorating devastating neurological conditions, both those in which established synapses are destroyed and those in which the initial development of connections in the brain is disturbed. The next few years should prove very interesting!