Finding reverses long-held beliefs and has implications for designing therapies
PRINCETON, N.J. — In a finding that eventually could lead to new methods for treating brain diseases and injuries, Princeton scientists have shown that new neurons are continually added to the cerebral cortex of adult monkeys. The discovery reverses a dogma nearly a century old and suggests entirely new ways of explaining how the mind accomplishes its basic functions, from problem solving to learning and memory.
Elizabeth Gould and Charles Gross report in the Oct. 15 issue of Science that the formation of new neurons or nerve cells — neurogenesis — takes place in several regions of the cerebral cortex that are crucial for cognitive and perceptual functions. The cerebral cortex is the most complex region of the brain and is responsible for highest-level decision making and for recognizing and learning about the world. The results strongly imply that the same process occurs in humans, because monkeys and humans have fundamentally similar brain structures.
The meeting will take place at CaixaForum, the new social and cultural center of Fundació “la Caixa” in Barcelona. The building is an early 20th-century factory that has been completely rebuilt and fully appointed with state-of-the-art technical equipment. Originally designed by the Catalan architect Josep Puig i Cadafalch in 1911, it is one of the jewels of modernista industrial architecture in Spain and has been listed as a Historical Monument of National Interest
Our fundamental interest is to understand the mechanisms used by the nervous system to process and analyze information about the sensory environment. Information processing in the nervous system operates on a number of fundamental levels: the dynamic electrical signalling of individual neurons (regulated by ion channels and their distribution within the cell membrane), the communcation between neurons that takes place largely at synapses, and the collective behavior of sets of neurons within neural circuits that are excited (or inhibited) according to external stimuli and internal conditions. Neural information processing is neither static nor uniform; rather at all levels, from molecules to networks, it depends on preceeding states, including developmental interactions with the environment, lifelong learning and training experience, and the immediate cellular and behavioral context. The sensory systems (hearing, vision, touch, smell, taste, and balance) have proven to be excellent experimental systems to study neural information processing. Sensory information processing is often easily traced from one level to the next, the external stimulus can be well controlled, and the constraints of the system and requirements for specific types of processing are often readily discernable. Previous and ongoing work in the laboratory has focussed on electrical signalling and ion channels in individual neurons, and synaptic communication between cells. A new direction that we are taking is to examine neural circuit organization and synaptic plasticity in cortex.
Experiments are complemented with computational models. It is very difficult to accurately predict the outcomes of experimental manipulations when working on a complex non-linear dynamic system like a spiking neuron with voltage-dependent conductances. Our goal is not to generate models for their own sake, although high quality computational representations of real neurons may be useful in several ways, but to use the models to provide a formal and quantitative test situation to evaluate hypotheses and determine where our understanding is weak. The comparison of these models to equivalent experiments allows us to make further predictions and develop new hypotheses; the failures of the models drive us to seek new information from experiments and lead to refinements of the models. The strength of this approach is that we can combine experimental and theoretical work in a single laboratory, which both shortens development time and allows the incorporation of many specific experimental measurements that may not necessarily be available from the literature. This approach results in fairly highly constrained models, closely tied to biophysics, and with known strengths and weaknesses with respect to the data set they represent.
Models are generated in NEURON, MATLAB, or C.