Whither new neurons?

The adult brain has long been seen as a stable circuit, often compared with computer hardware in that the components (neurons) are connected in an elaborate architecture that is not amenable to structural change. It was thought that learning occurred by altering the way the neurons communicated with each other, without any need to modify the composition of the system. As I discussed in an earlier post, this dogma was overturned just about a decade ago; it is now accepted that throughout life, new neurons are continually added to two areas of the brain: the olfactory bulb (see earlier post), and the hippocampus, which is essential for learning and memory and the focus of this post. Even more recently, it was demonstrated that these new neurons actually participate in preexisting networks.

Although the number of new neurons added per day (thousands!) is a miniscule fraction of the total number of neurons in the brain (about 100 billion), it is certainly high enough to significantly affect the functioning of the brain. However, due to technical limitations (reflecting the nascency of the field), it is unclear what these new neurons actually contribute to mature circuits, and how important they really are.

Speculation, of course, is rampant. One of the interesting things about young neurons is that, until a few months after they're born, they are functionally different from older, more mature neurons. New neurons are, in comparison with their elders, more easily excited, and more likely to enhance their connections with other neurons (aka LTP, a phenomenon believed to underlie learning and memory). These features indicate that new neurons may have a special role in the brain, beyond simply replacing old, dying neurons; specifically, the flexibility of their connections suggests that they may participate in learning and memory.

A recent paper, published in Nature Neuroscience, explored how these neurons are recruited into memory circuits when animals are learning new information. The researchers, led by Paul Frankland at the University of Toronto, focused on the hippocampus's role in the formation of spatial memories, which, as you may have guessed, are memories that are concerned with spatial locations (the Neurophilosopher recently had a fantastic post on different mammalian spatial memory systems).

The group used a cognitive test called a Morris Water Maze, which is the most common paradigm for assessing spatial memory in rodents. Basically, a rodent is placed in a circular pool of opaque water (about 4-5 ft in diameter, typically clouded with milk powder or white paint) that contains a platform hidden about 1 cm below the surface. As rodents are highly averse to swimming, they desperately swim around in search of an exit until they find the platform and can "escape." A series of static visual cues are placed around the edge of the pool, which the rodent uses to determine and, after repeated trials, learn the spatial location of the platform. During the course of training, rodents should require progressively less time to find the platform; once learned, the spatial memory should endure after the training has been completed. This ability to remember the location of the platform depends on the hippocampus; if the hippocampus is damaged, the animals never learn the task. Further, if the hippocampus is damaged after the location is learned, the animal will be unable to retrieve the memory.

In this study, the group looked at the brains of mice that were training on the Morris Water Maze, and were thus actively storing the relevant spatial information. To see which neurons were actively participating in this newly forming spatial memory network, they analyzed the expression patterns of genes called immediate early genes (IEGs), which are turned on when a neuron is activated. They combined this technique with a common strategy for determining when a cell has last divided, BrdU, and were thus able to determine the age at which these cells were born and incorporated into the network.

They found that new neurons were, indeed, incorporated into these spatial memory circuits. In fact, new neurons between the ages of 2 and 8 weeks were 2-3 times more likely to be activated during the task than their elders, which is a striking display of preference for these particular cells.

What about when neither the cells nor the memory are new? This was the most tantalizing part of the story. A month and a half after training (during which the mice did not perform the task), the researchers put the same mice back into the pool. They found that the cells born during the training period were still preferentially activated, indicating that they had become stable components of a network encoding the spatial location of the platform, presumably to endure for the lifetime of the animal.

The key here is that spatial memory formation and retrieval preferentially activate new neurons, which implies that such cells make a unique contribution to memory processing. Although the results do not answer the questions "Why new neurons? What are they good for?", it certainly fuels more informed speculation. Perhaps mature neurons are more sensitive to features they learned when they are young; thus, we may need to add new neurons to the network in order to form distinct spatial memories, e.g. when adapting to a new environment.