Portal de Periódicos da CAPES

The Millennium of the Dendrite?

2000; Cell Press; Volume: 27; Issue: 3 Linguagem: Inglês

10.1016/s0896-6273(00)00054-4

1097-4199

Andrew Matus, Gordon M. Shepherd,

Fractal and DNA sequence analysis

Despite the fact that they constitute as much as 80% of the surface area in most types of neurons, dendrites have not always loomed large in the consciousness of neurobiologists. A recent international conference, the first to be devoted exclusively to dendrites (Workshop on Dendrites, supported by Columbia University and the Instituto Juan March de Estudios e Investigaciones, June 4–7, 2000), suggests that awareness of their importance is increasing rapidly. Many of the studies reported have been made possible by a range of new methodologies that enable the properties of dendrites to be analyzed much more directly than heretofore. As the fruits of these new studies were shared, excitement over the new insights was palpable. The conference was anchored by three keynote lectures. Opening the meeting, Gordon Shepherd plotted the ups and downs in the fortunes of dendrites throughout the history of neuroscience. Following their dominance as the highly visible indicators of neuronal diversity during the early phase of anatomical analysis, when the Golgi stain reigned supreme, dendrites seemed almost to vanish from the landscape of neuroscience during the middle decades of the last century. This was the period when prevailing ideas about brain circuits had them reduced to "neural networks" in which cell body "nodes" were connected to one another by axonal "wires." This imbalance began to be redressed by the work of Wilfrid Rall using compartmental models of dendritic integration, and the key roles of dendrites as integrators of neuronal signals and sites of synaptic plasticity have now been recognized and are being rapidly explored. This historical perspective was further elaborated in lectures by Rodolfo Llinás (NYU) and Eric Kandel (Columbia)—the first examined the gradual growth of ideas about active electrical properties of dendrites, while the second traced the development of knowledge about mechanisms involved in signaling from synapse to dendrite and nucleus that underlie circuit plasticity in the nervous system. In between these contributions, the explosive progress in current research on dendrites was emphatically demonstrated in talks and posters reporting results over a wide range of topics. The sessions were organized around several main areas of investigation: active properties, calcium and plasticity, receptor localization, proteins and RNA targeting, and development. The new generation of practitioners of patch recordings from dendrites was well represented. These studies are converging on a new set of principles of active properties of dendrites. Voltage-gated channels are nearly ubiquitous in dendrites, but at varying densities, making the generation of active responses at any site highly conditional on activity elsewhere in the axon initial segment, soma, and other dendritic branches. The nonuniform densities for different channels also reflect different functions in different parts of the dendritic tree (Daniel Johnston, Baylor). Depolarizing conductances are characteristically controlled by shunting K conductances, as in the extent of back propagating action potentials from the soma; conversely, EPSPs can control voltage-dependent K conductances, influencing the induction of LTP. Dendritic active conductances are highly regulated by some of the usual suspects, such as cyclic AMP. Extending earlier work in motor neurons, distal synapses appear to have larger conductances than proximal synapses (Jeffrey Magee, Louisiana State). The neuron can redistribute its sensitivity to different input patterns (Matthew Larkum, Max Planck, Heidelberg). The site of action potential initiation can shift between the classical site in the axonal initial segment and distal dendrites, depending on the balance of distal EPSPs and proximal IPSPs, with variable linkage between dendrites and axon (Nelson Spruston, Northwestern; Wei Chen, Yale). When added to the complex passive and active dendritic properties originally elucidated by Rall, all of these factors acting together mean that synaptic integration is a dynamic process engaging different parts of the dendritic tree depending on the modulator state of the system. This theme of the dynamic properties was extended by studies of dendritic plasticity using new methods for imaging morphological and molecular events in living neurons. The power of the techniques that are becoming available was illustrated by a demonstration of two-photon confocal imaging in living brain using a miniature scanning head operating through a window in the skull in the awake behaving animal (Fritjof Helmchen, Bell Laboratories). While not yet applied to imaging neuronal morphology, the potential value of this approach for answering questions about how dendritic structure responds to sensory inputs is obvious. At present, such studies are mainly conducted in vitro where time-lapse video recordings of brain tissue slices made using the two-photon method have shown the emergence of new spines following NMDA receptor–dependent induction of LTP (Tobias Bonhoeffer, Max Planck, Muenchen-Martinsried). In another confocal microscopy study, inducing widespread LTP in hippocampal slices using a cocktail of chemicals appeared not to influence spine number but did produce a small increase in the length of a subset of spines (Tim Bliss, Mill Hill, London). Such differences in experimental results reflect the need for further studies employing standard experimental conditions of the kind that are used in electrophysiological studies of LTP. While these studies address the question of how new dendritic spines are produced in response to afferent stimulation, it is clear that spines retain a capacity for morphological plasticity even after synapses have been formed. This appears to be related to their high content of actin that produces spontaneous motility in cultured cells and brain tissue slices (Andrew Matus, Friedrich Meischer Institute, Basel, and Carol Mason, Columbia). Direct visualization of the actin cytoskeleton in spines using actin tagged with GFP shows that actin-based spine motility is blocked by activation of either AMPA- or NMDA-glutamate receptors through a mechanism that involves Ca2+ entry into the spine cytoplasm (Andrew Matus). These studies together suggest that formation and stabilization of excitatory synapses are distinct processes, both of which are regulated by synaptic transmission acting through glutamate receptors. Further back in development lie questions about how the fundamental architecture of dendrites and synaptic contact patterns are established. The size of dendritic arbors in neurons of the frog brain depends on sensory stimuli acting through NMDA receptors (Hollis Cline, Cold Spring Harbor). Among the genes that may be involved in mediating this effect is CPG15, which is switched on in the frog optic tectum by visual activity. When CPG15 protein is expressed ectopically in optical tectal neurons using a viral vector, their dendritic arbors become larger as they continue to extend rapidly without showing the normal transition to a second phase of slower growth. One very promising approach to analyzing molecules that determine dendritic phenotypes is to exploit the capacity for genetic manipulation and the now-complete genomic information of Drosophila. Developing dendrites in Drosophila embryos are highly dynamic but nevertheless form branching patterns that are fairly invariant for the same neuron from embryo to embryo (Fen-Biao Gao, University of California, San Francisco). These patterns are altered in a variety of mutations bearing the usual profusion of exotic names including flamingo, kakapo (a New Zealand flightless parrot in case you were wondering), sequoia, and tumbleweed. Varying from zinc finger proteins (sequoia) to seven-transmembrane proteins with multiple cadherin-type repeats (flamingo), their sheer heterogeneity shows how much we still have to learn. Interestingly, though, in several of these mutations, dendrites fail to respect the midline boundary, suggesting that they may contribute to a mechanism for limiting growth and guidance of dendrite branches. One of the key factors determining the extent of dendrite growth is synaptogenesis, and time-lapse video recording is being used to determine how synaptic partners form their initial contacts (Stephen J. Smith, Stanford). Imaging pairs of dye-labeled Mauthner axons and their target postsynaptic neurons in the spinal cord of living zebra fish embryos shows that motile axonal growth cones and dendritic filopodia both participate in the initial contact with the axon growth cone, sometimes "pausing" and seeming to palpate the dendrites of the postsynaptic cell. In other cases, the axon passes by without slowing and dendritic filopodia from the target cell later grow out to contact it. Using cultured rat hippocampal neurons, the Smith group have followed the formation of synaptic contacts using the synaptic vesicle protein VAMP, tagged with GFP, as marker for nascent presynaptic sites. VAMP–GFP is transported along dendrites together with other presynaptic protein components in preformed "packets" that appear to stabilize at cell–cell contact sites where they become capable of activity-evoked recycling within an hour after initial contact formation. In this case too, contacts may be initiated by filopodial protrusions from dendrites. Morphological plasticity is only one of several aspects of dendrite function that depend on activity-regulated cytoplasmic calcium levels. Optical recording with Ca2+-sensitive dyes has shown that stimulus-induced Ca2+ signals in spines and dendrites are reduced by around 50% when AMPA receptors are blocked, whereas blocking NMDA receptor entirely eliminates the stimulus-evoked Ca2+ signal (Arthur Konnerth, TU Muenchen). Terminating calcium fluxes in the spine head requires clearing Ca2+ from the spine cytoplasm for which there are two major routes, diffusion of Ca2+ through the spine neck into the dendrite shaft and uptake into intracellular membrane-bound stores via an ATP-dependent Ca2+ pump. These two mechanisms appear to operate at different levels in individual spines, suggesting that spines can be divided into two basic categories, "diffusers," which are correlated with short-necked spines, and "pumpers," which presumably represent those with a well-developed internal membrane compartment or spine apparatus (Rafael Yuste, Columbia). Spike evoked Ca2+ entry can act synergistically with mGluR activation to produce regenerative Ca2+ waves in apical dendrites and soma (William Ross, New York Medical College). Novel constraints on compartmentalization of calmodulin as a key transducer of dendritic intracellular calcium are being revealed (Sally Kim, University of Texas, Houston). As this brief summary suggests, Ca2+ appeared repeatedly throughout the meeting as a key regulator in various aspects of dendrite function while itself being subject to several modes of regulation. Another layer of control is provided by neurotrophins, especially BDNF, which can excite neurons in several major brain areas at nanomolar concentrations. Focal application of BDNF produces Na+ signals within individual dendrite segments and Ca2+ transients that may be limited to single dendritic spines. These highly localized responses may provide a postsynaptic mechanism for BDNF-mediated LTP (Konnerth). A link between BDNF and LTP is further suggested by experiments with BDNF knockout mice where LTP, especially in its late phase, is severely compromised. Evidence that this effect results from the lack of BDNF itself and not some side effect is provided by experiments showing that LTP in hippocampal tissue slices from these mutant mice can be rescued by reintroducing the BDNF gene locally into cells using an adenovirus vector (Bonhoeffer). BDNF also emerges as an important regulator of dendritic morphology from experiments in which BDNF and GFP (as a marker of dendrite morphology) were simultaneously introduced into cells in tissue slices of ferret cortex using the biolistic device popularly known as the "gene gun" (Lawrence Katz, Duke). BDNF released from these cells has a powerful autocrine effect on morphological plasticity so that within 24 hr they develop roughly 3-fold more dendrites than control cells expressing only GFP. At the same time, dendritic spines are destabilized so that almost none remain. New experiments using green and red fluorescent proteins to examine neighboring cells show that the radius of action of BDNF is less than 10 μm, indicating that its effects are remarkably local. In addition to these external views of dendritic and spine morphology, the new microscopy provides views of what's going on inside dendrites. On this topic, another session was devoted to the important issue of how macromolecules are targeted to their appropriate destinations within dendrites, a subject that affects both proteins and mRNAs. Targeting of proteins to different neuronal domains appears to involve their segregation into distinct populations of pleiomorphic carrier vesicles. Their sorting is being analyzed by follow the trafficking of representative dendritic or axonal proteins fused to GFP in living neurons (Gary Banker, Oregon Health Sciences). Polarized delivery of the transferrin receptor, a dendritic protein, can be accounted for by selective microtubule-based transport. However, carrier vesicles containing NgCAM, an axonal protein, are transported into both axons and dendrites indicating that mechanisms downstream of transport are necessary to account for its selective delivery to the axonal plasma membrane. It has been known for many years that there are polyribosomes in dendrites, with small clusters adjacent to spines implying local protein synthesis in relation to synaptic function and plasticity. How does dendritic mRNA get to the appropriate sites, and what are the links between soma and dendritic mRNA? Experiments on the immediate early gene Arc offer a promising experimental approach since the Arc mRNA is rapidly synthesized and transported into dendrites in response to synaptic stimulation. When different pathways to the same set of dendrites are stimulated, Arc mRNA accumulates selectively at sites postsynaptic to the activated axonal pathway in a mechanism that depends on NMDA receptors, suggesting the existence of a synapse-oriented targeting mechanism (Oswald Steward, University of California, Irvine). Cytoplasmic polyadenlylation may be a link between NMDA receptor activation and mRNA translation in dendrites (Justin Fallon, Brown University). In addition to their other roles, neurotrophins may also be involved in mRNA targeting since targeting of mRNA for the alpha subunit of calcium/calmodulin kinase II (αCaMKII) to synaptic sites in dendrites is promoted by BDNF and NT-3 (Gary Bassell, Albert Einstein College of Medicine). What moves mRNA within dendrites? High-resolution in situ hybridization studies have shown a variety of mRNAs including αCamKII associated with granules inside dendrites. Staufen, an RNA-binding protein initially identified in Drosophila oocytes, is associated with large RNA-containing granules in dendrites, and new studies using staufen–GFP suggest that these granules move on microtubules, implicating the microtubule cytoskeleton in targeting dendritic mRNAs (Martin Koehrmann, EMBL, Heidelberg). These studies remind one of the dramatic impact in the middle 80s of the first microscopic views of particle movement through axons. Now those same questions can be addressed in dendrites, where communication between nucleus and synaptic sites is likely to be much more complex than between nucleus and axonal terminals. A key question is whether individual dendritic compartments and even individual synapses are under differential mRNA control, either from the nucleus or by local protein synthesis. Among dendritic proteins whose levels may vary, neurotransmitter receptors are likely to have one of the major impacts on neuronal function. A session devoted to their localization and regulation provided much new data. The complexity of receptor distribution on relation to innervation patterns is strikingly demonstrated by GABAergic interneurons. In hippocampal area CA1, there are three distinct classes of these cells innervating the soma and initial axon segment of pyramidal cells and at least ten different types innervating their dendritic fields (Peter Somogyi, MRC, Oxford). Postsynaptically, hippocampal pyramidal cells express at least 12 different GABAA receptor subunits, providing an enormous potential range of functional GABA receptors with affinities for GABA varying by as much as 10-fold. Inhibitory GABA receptors and their anchoring molecule gephyrin are segregated on the dendrite surface from excitatory AMPA- and NMDA-type glutamate receptors and their associated scaffold proteins such as PSD-95/SAP90 (Ann Marie Craig, Washington University). NMDA receptors form clusters independently of neuronal activity, but total receptor numbers vary, increasing when activity is suppressed and decreasing when receptors are blocked. Altogether, these results suggest the existence of a homeostatic mechanism that maintains threshold sensitivity against a changing background of activity. What, then, happens to levels of synaptic receptors under conditions of synaptic potentiation such as LTP? Many GluR1-containing AMPA receptors appear to be "stored" in the dendritic shaft and are delivered to the synapse via a stimulus-dependent mechanism that involves interaction with a PDZ domain scaffold protein (Roberto Malinow, Cold Spring Harbor). Delivery of GluR1 or GluR4, which is present earlier in development, can lead to a long-lasting increase in transmission despite the fact that individual AMPA receptor molecules turn over rapidly. This suggests the existence of "slot" proteins that may be delivered along with GluR1 and GluR4 and act as placeholders for AMPA receptor replacement. One of the features that is certain to be significant in determining the relationship between dendrite and synapse is the relationship between the postsynaptic junctional structure—the PSD—and the dendrite cytoskeleton. Over recent years, a complex nexus of molecular interactions has been discovered that links components of the PSD to the dendrite cytoskeleton. Adapter proteins such as PSD-95/SAP90 and GRIP/ABP connect ion channels and neurotransmitter receptors to the actin cytoskeleton. Now that three-dimensional structures for interaction domains are available for several of these adapter proteins, attempts are beginning to fit them into an integrated model of the PSD (Mary Kennedy, Caltech). These same proteins are likely to play an important role in regulating the function of their receptor and ion channel binding partners. For example, the GluR2 AMPA receptor subunit binds to the postsynaptic adapter proteins PICK1 and GRIP. The interaction with GRIP is regulated by phosphorylation, suggesting that several levels of receptor-to-scaffold interactions await discovery (Richard Huganir, Johns Hopkins). Somehow, activity detected by receptors has to be signaled back to the nucleus to regulate gene transcription. This function is still mainly in the black box stage, but one aspect of its operation emerges from evidence that the R1 subunit of the metabotropic GABAB receptor binds via a coiled-coil domain in its C terminus to the rat homolog of the CREB2 transcription factor (Jeremy Henley, MRC Bristol). GABABR1 and CREB2 are colocalized in patches on dendrites and CREB2 becomes redistributed when GABA receptors are activated, suggesting a novel signaling pathway from dendritic receptor to nuclear transcription unit. The complexity of the diverse interactions discussed at the meeting was encapsulated in a contribution describing computer models that provide insight into the enhanced information processing capacity of active dendrites (Bartlett Mel, USC). It showed how much we do not yet understand and indicated how necessary it will be to develop, in parallel with experimental analysis, a theoretical basis for understanding dendrites as complex systems. Summing up, Steven Siegelbaum (Columbia) drew the themes of the conference together and captured the feeling of the participants that the fallow period, when many neuroscientists regarded dendrites as irrelevant to neuronal function, is well and truly over. Judging by the enthusiasm and wealth of new data presented at the meeting, dendrites are set to receive a lot more attention in the new millennium.‡To whom correspondence should be addressed (e-mail: [email protected] [A. M.], [email protected] [G. M. S.]).