7 December 2009. Learning and memory depend on the brain’s capacity for synaptic reorganization, which itself relies on the dynamic nature of synapse-bearing dendritic spines. These tiny connective protrusions grow and shrink, form and disappear with time (e.g., see ARF related news story and ARF related news story), and are subject to loss in Alzheimer disease. Now, with the use of in vivo imaging, two groups have caught dendrites in the act of rewiring cortical circuits in live mice experiencing motor or sensory learning. Both studies come to a similar conclusion: Small changes go a long way toward encoding memories, and like memories, once dendrites are formed they can last a lifetime. The results are reported in two papers published back to back in Nature online November 29.
One study comes from Yi Zuo and colleagues at University of California Santa Cruz, who find that when mice learn to reach past a barrier to pick up a seed, new spines form rapidly in their motor cortex. New spine formation is followed by elimination of some existing spines, resulting in a rewiring of cortical circuits that persists over weeks to months. In the second study, Wen-Biao Gan and coworkers at New York University report similar results, and go further to look at overall dendritic spine dynamics before and after motor training or in response to sensory experience. They find that a relatively small number of long-lasting, training-induced spines persist in the midst of a large number of permanent connections. Their findings suggest that small adjustments in a fabric of existing connections results in the rewiring that allows new learning and long-term memory to coexist, at least in the motor and sensory cortex.
Both groups used a similar approach of transcranial imaging of mice that expressed green fluorescent protein in layer 5 cortical neurons. Gan has been a leader in the imaging of live spines (see ARF related news story on Grutzendler et al., 2002), and in the new work, coauthors Guang Yang and Feng Pan serially imaged mice before and after sessions that taught the animals to run at increasing speed on a rotating rod. In untrained mice, new spines accounted for about 7 percent of total spines in primary motor cortex after two days; that number doubled in the trained mice. New spine formation returned to normal levels after two days without training, but continued to be elevated if the mice continued to train.
The effect is not limited to the motor cortex. The researchers report a similar sprouting of new spines in the whisker barrel sensory cortex of mice placed in a cage accessorized with hanging beads that provide a novel sensory experience by tickling their whiskers.
What is the significance of the new spines? Repeated imaging of the same animals revealed that most new spines were eliminated in the two weeks after a brief training or sensory experience, but a few percent remained. Further training or exposure increased the proportion of spines that persisted. The endurance of spines formed in motor cortex in the initial days of training, but not after, correlated with performance on the rotating rod a week later. These results suggest that the spines formed in response to experience are important for memory retention.
Motor learning and novel sensory experience triggered not just new spine formation, but also removal of existing connections. In contrast to the rapid onset of spine formation, mice required longer training (7-14 days) before the researchers began to detect spine elimination. Motor performance at seven days correlated with removal of old spines, suggesting that both new synapse formation and existing synapse removal are required for rewiring of cortical circuits during learning.
In the work from the Zuo lab, first authors Tonghui Xu and Xinzhu Yu show similar results in mice trained on a different fine motor task, reaching for a seed through a hole in a plastic barrier. One hour after a single training session, they could already see a doubling of new spine formation in the corresponding motor cortex. The extent of new spines correlated with the animals’ success at picking up the seed. They saw no early loss of dendrites, but similar to Gan’s results, they found that pruning occurred over the next week that returned total spine number to normal levels. As in the other paper, most of the new spines were unstable, but a fraction persisted over time.
The Gan lab took the analysis further to ask how, with all the making and breaking of connections, the brain stores long-term memories. They approached this question by looking at what happened to spines throughout the life of the mice. The researchers measured the persistence of individual spines, starting with 30-day-old mice. Whether spines originated early in life, or appeared because of learning, the half-life was about the same, in the range of 70-90 months. This indicates that some of the structures likely last a lifetime (laboratory mice typically only live about 36 months), and so could support lifelong memory. Gan and coworkers also found that learning-induced spines accounted for only about 0.04 percent of total spines in the adult mice, while spines formed early in life made up half or more of the population. “Our data suggest that learning-induced remodeling is pretty small, but that’s sufficient. After you learn things, you change a very small fraction, but you keep most of the existing connections,” Gan said.
In support of this idea, both groups show that adult mice who learned the task as youngsters could perform it months later without triggering an increase in new spine formation. However, the adults did undergo remodeling when they learned a different task, and that additional remodeling did not affect the spines from previous training.
This leads to a model where spines formed early in life provide a scaffold of basic cortical function and lifelong memory storage. On top of that, Gan writes in the paper, “Learning and daily sensory experience leave minute but permanent marks on cortical connections, and suggest that lifelong memories are stored in largely stably connected synaptic networks.”
Gan says he is interested in looking at remodeling in aging mice, and possibly in Alzheimer mouse models, too. “One could start to look at age-related memory decline by looking at whether learning-induced remodeling is reduced in aging brain. We see a very good correlation between remodeling and performance, so it will be useful to understand the process, and to ask whether we actually could find some way to enhance the remodeling.” Gan has looked at the effects of amyloid plaques on dendrites in vivo (see ARF related news story on Tsai et al., 2004), and says they are now thinking of imaging dendrite remodeling in an AD mouse model to see if and at what stage learning-induced remodeling might be compromised.—Pat McCaffrey.
Xu T, Yu X, Perlik AJ, Tobin WF, Zweig JA, Tennant K, Jones T, Zuo Y. Rapid formation and selective stabilization of synapses for enduring motor memories. Nature. 2009 Nov 29. Abstract
Yang G, Pan F, Gan WB. Stably maintained dendritic spines are associated with lifelong memory. Nature. 2009 Nov 29. Abstract