 |
Reverse Engineering the Cortical Microcircuitry
Summary: Rafael Yuste's laboratory studies the structure and function of cortical circuits and the biophysical properties of dendritic spines. Using a bottom-up approach based on imaging the spontaneous and evoked activity of networks of cortical neurons in thalamocortical slices, he is attempting to decipher the cortical microcircuitry of mouse primary neocortex and reverse engineer its fundamental principles of operation. A second line of work is focused on understanding the biophysical properties of one of the more basic structural elements present in those circuits, the dendritic spines. This work could help to generate a general theory of cortical function and a better understanding of the pathophysiology of epilepsy.
Understanding how the neocortex functions is the main challenge of modern science and medicine. The cortex, which constitutes the largest part of the brain in mammals, is the primary site of mental functions and mental disease. No unitary theory of how the cortex works exists. Nevertheless, the basic cortical microcircuitry, defined as the intra- and interlaminar projections in a local region of neocortex (a column of a cross section of <1 mm2), develops in a stereotyped fashion, is similar (at least superficially) in different parts of the cortex and in different mammals, and does not appear to have changed much in evolution since its relatively recent appearance. Because of this, a "canonical" cortical microcircuit may implement a basic computational algorithm. Moreover, the widely different types of computations performed by the cortex would suggest that this algorithm, as the common denominator of all these different computations, should be simple. Thus, nature might have found a general strategy of computing, and the neocortex might be a universal computer, somewhat analogous to a "Turing machine." Our goal is to identify the functional modules that could implement this basic algorithm, as a step toward understanding the basic computational nature of cortical function.
Rather than search for the general principles of cortical function at the behavioral level, we consider the cortical circuit at its most basic nature, as a system of interconnected elements. Thus, one could attempt to infer the basic principles implemented in this circuit from the rules of connectivity and the spontaneous or evoked dynamics generated by isolated cortical preparations. This is reverse engineering: we take apart the black box, examine the wires and transistors to identify the logic that has been implemented in the circuit, and make educated guesses about its function. While we do not ignore behavior, we consider that the purpose of the cortex may be not to generate behavior directly but to generate mental states, internal representations of the world that are mentally manipulated by the animal and can result in behavior, or not. If the cortex is indeed internally driven and has an intrinsic function—one related to, but not strictly determined by, the outside world—it could be useful to study its microcircuit in isolation. Based on my work and that of some of my colleagues, I think this question can be successfully addressed in brain slices.
Anatomical and physiological studies have suggested that the connectivity of the cortical microcircuitry is complex, but not random. For example, it is clear that inhibitory neurons target their connections very specifically. Less is known about the pyramidal-pyramidal connections that constitute the "skeleton" of the cortex. It is conceivable that these connections are also precise and that the neocortex, like the retina, is composed of dozens or hundreds of classes of neurons with specialized circuit functions. Because of this, it seems to me that a fundamental question in cortical research is to identify the elements of neocortical circuits and study how they are connected and interact to generate different activity patterns. The assumption is that cortical activity, out of all the possible spatiotemporal patterns, will occupy a restricted dynamical space, and this should help in the design and testing of functional models of cortical circuits.
A major limitation in past work was the difficulty of revealing functional connections in large numbers. Over the past 15 years, we have developed optical methods using calcium indicators that enable the imaging of the activity of populations of hundreds or even thousands of neurons simultaneously. This technique relies on the faithful generation of somatic calcium transients by action potentials. We are applying these optical techniques to study the cortical microcircuit in brain slices of mouse primary visual cortex (V1B-binocular region) because of (1) its relatively small size (~200,000 neurons), which enables a significant sampling of different neuronal classes, (2) availability of genetically modified animals, (3) some understanding of the receptor field transformation that these neurons carry in vivo, and (4) developmental plasticity paradigms during the critical period for monocular deprivation. For comparison and evaluation of "canonical" elements, we are also using mouse primary somatosensory and frontal cortex.
Our long-term study of these circuits focuses on several complementary projects, using thalamocortical brain slices to study the cortical circuitry at the cellular and multicellular level. The techniques used are electrophysiology and a variety of optical methods, including infrared-DIC, voltage- and ion-sensitive dye imaging with cooled CCD cameras, and two-photon microscopy. We also use cell cultures, biolistics transfection, electron microscopy, and numerical simulations and modeling. We are also developing novel optical methods, such as second harmonic generation, to measure membrane potentials and are making extensive use of transgenic and knockout mice.
We are currently focusing on two questions: (1) What is the function of dendritic spines? Spines are the recipients of most excitatory inputs in the cortex, so they are an essential element in cortical circuits, although they are still poorly understood. Two-photon microscopy has enabled functional studies of dendritic spines and has shown that they compartmentalize calcium because of their morphological features and local calcium influx and efflux mechanisms. Spines have recently been shown to exhibit rapid morphological plasticity. This has suggested that the function of the spine, or the synapse, is equally dynamic. Besides a calcium function, we suspect that spines may also play an electrical role in the circuit, and we are testing this possibility. In our recent work, we have used second harmonic generation to image membrane potential in spines and have discovered that the spine neck can electrically isolate inputs. Therefore, spines can protect the dendrite from changes in input resistance during synaptic transmission and thus help to linearize input summation. Spines could be the anatomical signature of a linear circuit.
(2) What are the multicellular patterns of activity under spontaneous or evoked activation of the circuit? Cortical neurons do not respond individually, but are activated in multicellular units of activation. These ensembles may represent a functional state of the circuit, such as an attractor. We recently showed that these ensembles can be engaged by the stimulation of thalamic inputs, so the role of sensory inputs could be to reawaken these internal states. Our results agree with the theories that propose that the function of the cortex is internally driven, resembling central pattern generators found in evolutionary simpler circuits. We want to use imaging of the spontaneous and evoked activity populations of cells to understand better the basic biology of these neuronal ensembles and relate them to the structure of the cortical circuit.
My laboratory is currently supported by the National Eye Institute; the Kavli Institute for Brain Science; the New York State Office of Science, Technology and Academic Research (NYSTAR) program; and the Pew Charitable Trusts.
Last updated: November 12, 2007
|
|
|
|
