Karel Svoboda's lab is searching for the substrates of experience-dependent plasticity in the developing and adult neocortex. The functional properties of the brain must change in response to salient sensory experiences, but the nature of these changes at the level of synapses, neurons, and their networks (also known as the engram) is unknown.
Most of the cognitive functions in mammals, ranging from perception to memory formation, are performed in the neocortex, a massive network of neurons. Neurons are linked into circuits by synapses, which pass information between neurons. How do the circuits and synaptic mechanisms underlying this awesome neural network produce our perception of the world? How plastic are these neural circuits, that is, how do the physical properties of the neural network change in response to experience? The answers to these questions will profoundly change our understanding of the function and diseases of the brain.
Cortical tissue is dauntingly complex. Each module of cortical tissue (~1 mm3 of gray matter) contains nearly a million neurons, each of which connects to thousands of other neurons. To observe neurons and synapses in the intact brain, we build and use sensitive tools. Two-photon laser scanning microscopy (2PLSM) allows us to image single synapses in intact tissues and to track changes in intracellular calcium and signal transduction events. Excitation of neuronal elements by focal uncaging of neurotransmitters allows us to probe the structure of cortical circuits efficiently. We combine these optical methods with electrophysiological measurements of synaptic currents and potentials and molecular manipulations of neurons.
Svoboda Research Abstract Slideshow 2
Figure 1: Stability of PSD-95 clusters in the intact mouse neocortex. A sparse subpopulation of layer 2/3 neurons expressed a red cytoplasmic protein (mCherry1) and PSD-95 fused to a green fluorescent protein (GFP). Left, overview of a dendritic branch. Right, time-lapse imaging of a region of interest (white box in the left image). The relatively stable sizes of the green spots indicate stable relative synapse sizes.
Noah Gray, Robby Weimer, and Karel Svoboda
Figure 2: Two-photon fluorescence lifetime microscopy of Ras activation in single dendritic spines in brain slices. A: The imaged neuron expresses a genetically encoded Ras sensor based on GFP (green fluorescent protein) and mRFP1 (monomeric red fluorescent protein). The fluorescence lifetime of the GFP decreases with Ras activation. Color indicates fluorescence lifetime. At t = 0, glutamate was uncaged next to the spine marked by the arrowhead. In response, Ras activity increased locally and the spine grew in volume. B: Ras activity increased over 10 mm of dendritic length. C: Changes in spine volume in selected regions of interest were specific to a synapse.
Ryohei Yasuda and Karel Svoboda
Figure 3: Synaptic input map for a layer 2/3 neuron in the barrel cortex (colored background). Overlaid is the morpholoy of the neuron (blue) and electrophysiological traces from which the synaptic input map was constructed (black). A schematic circuit diagram is also shown (white).
From the Svoboda lab. For details, see Shepherd, G.M., Pologruto, T.A., and Svoboda, K. 2003. Neuron 38:277–289.
Figure 4: Experience-dependent synaptic plasticity in the adult neocortex in vivo. Left, images collected 24 hours apart. Note stable (yellow arrow) and transient (blue and red arrows) spines. Scale bar, 5 mm.
Right, electron micrograph of new spine, indicating that it participates in synaptic transmission. Scale bar, 0.125 mm.
From Trachtenberg, J.T., Chen, B.E., Knott, G.W., Feng, G., Sanes, J.R., Welker, E., and Svoboda, K. 2002. Nature 420:788–794. © 2002 Nature Publishing Group.
Figure 5: Experience-dependent spine plasticity in barrel cortex in vivo. Three images collected 45 minutes apart were color coded and overlaid. White regions indicate stable dendritic segments. Green (red) spines retracted (sprouted) during the observation period.
Photo: Karel Svoboda
Cortical Circuits and Their Plasticity
The neocortex consists of about a dozen types of neurons. How are these neurons wired into circuits? The wiring diagram of the brain is fundamental to understanding cortical function and plasticity. However, little quantitative information about circuitry is available. What are the sources of input to a neuronal subtype and what are their relative strengths? Which synaptic pathways change with novel sensory experience?
We are using laser scanning photostimulation (LSPS) to measure the structure of cortical circuits. LSPS allows us to rapidly map the spatial distributions of synaptic input to individual neurons. We have applied LSPS to the rodent barrel cortex. Similar to other sensory areas in the mammalian cortex, the barrel cortex is arranged in maps. Each whisker is represented by a small cortical region (barrel). Whisker maps are shaped by experience during development and reshaped in the adult.
We have discovered that the barrel cortex contains interdigitated columnar circuits—barrel and septal columns—that process input from distinct thalamic nuclei. In developmental studies we found that cortical columns develop with remarkable specificity, without detectable "overgrowth" and "pruning." By comparing circuits in normal and deprived animals, we performed an unbiased search for the synaptic pathways that undergo experience-dependent plasticity during development. We found that the strengths of excitatory layer 4 → layer 2/3 connections changed with opposing signs in barrel and septal columns. These measurements provide an explanation of previous in vivo measurements of plasticity at the level of synaptic pathways. We applied LSPS to analyze the development and plasticity of barrel cortex circuits in mice lacking the FMR protein, an animal model for fragile X mental retardation, and detected highly specific developmental defects in particular synaptic pathways.
Experience-Dependent Structural Synaptic Plasticity
Understanding plasticity at the level of circuits is just one piece of the puzzle. What changes at the level of neurons and synapses in particular synaptic pathways in response to novel sensory experience? Answers to this question are fundamental to the mechanisms of plasticity and the memory capacity of the brain. Our approach has been to image synapses in transgenic mice in vivo over long times as plasticity happens.
We find that the large-scale arborization of axons and dendrites is stable, but that neurons display a rich repertoire of micrometer-level structural plasticity of dendritic spines, axonal terminals, and axonal branch tips. A subpopulation of dendritic spines appear and disappear. Retrospective electron microscopy has revealed that spine growth and retraction are associated with synapse formation and elimination, respectively, and underlie aspects of experience-dependent changes in circuits.
We have recently turned our attention to synapse stability. Many synapses persist for months, perhaps the entire life of the animal, and also maintain their size. This is remarkable because synapses are tiny structures composed of just a handful of proteins of a given type. Protein lifetimes are on the order of days. To study the underpinnings of synapse stability, we have developed methods to measure the trafficking of synaptic proteins at the level of individual synapses in vivo. Remarkably, we find that the major scaffolding protein PSD-95 unbinds from synapses over tens of minutes and exchanges with proteins in neighboring synapses. Larger synapses scavenge diffusing PSD-95 more efficiently and also hold on to PSD-95 longer. The interactions between synapses and their proteins are thus tuned to maintain the synaptic status quo against dissipation by diffusion.
The Function and Plasticity of Single Synapses
Cortical plasticity and memory are thought to manifest themselves physiologically at synapses. A cortical synapse consists of a presynaptic terminal and a postsynaptic spine, tiny compartments (~0.1 μm3) that contain only a few signal-transducing molecules, such as channels and receptors. To understand signal and noise in synaptic transmission, it is therefore necessary to study single synaptic contacts. We have developed methods based on calcium imaging to detect the activation of N-methyl-D-aspartate receptors (NMDA-Rs) in individual spines in response to the release of single quanta of neurotransmitter. We measured the fractional activation of synaptic receptors after neurotransmitter release and found that synaptic receptors are far from saturation. In fact, often only a single receptor opened in response to neurotransmitter release. Synapses are therefore noisy, but capable of linearly transmitting presynaptic release. We have also dissected the mechanisms of calcium influx and handling to gain an understanding of the life cycle of calcium ions in dendrites and spines.
Can we learn about calcium-dependent signaling pathways in spines? One approach uses FRET (fluorescence resonance energy transfer) to image protein-protein interactions, but such measurements are difficult in intact tissue, particularly in microcompartments. To overcome these problems, we have built a sensitive microscope that combines two-photon excitation with fluorescence lifetime imaging microscopy (2pFLIM). We have used 2pFLIM to dissect the role of the small GTPase Ras in synaptic plasticity at the level of single synapses.
Genetics of Neural Circuits
To decipher complex computations and behaviors at the level of cortical circuits we will need to be able to activate and inactivate specific populations of neurons rapidly and reversibly in vivo. To address the problems of inactivation, we have developed MIST (molecules for the inducible inactivation of synaptic transmission). We set out to harness the molecular machinery underlying neurotransmitter secretion to interfere with synaptic transmission, triggered by administration of small-molecule drugs. We use drugs that induce dimerization of small protein domains. These domains were attached to a variety of proteins important in the synaptic vesicle cycle. Candidate systems were transfected into cultured neurons and screened for potent, inducible, and reversible inactivation of the synaptic vesicle cycle. We developed two distinct MISTs, based on the synaptic vesicle proteins synaptophysin and VAMP (vesicle-associated membrane protein). These systems have been tested in vitro and in vivo. For example, we have used the L7 promoter to introduce the VAMP-based system into Purkinje neurons. In L7-MIST mice, injection of dimerizer abolishes learning of the rotarod task. We are introducing MISTs into specific subsets of cortical neurons, including thalamus-projecting L6 cells, various subtypes of GABAergic interneurons, and subregions of the hippocampus and neocortex.
Aspects of this work were supported by grants from the National Institutes of Health and the McKnight Foundation.