14 April 2010. Besides making long-range connections between different brain regions, neurons also form complex local connections—or microcircuits—that represent a basic unit of neuronal communication. Several studies have begun to tease out these local connections (see ARF related news story), but for the most part, the functional relationships between microcircuit activity and learning remain unknown.
In a study presented online in Nature on April 7, a team led by Karel Svoboda at the Janelia Farm Research Campus in Ashburn, Virginia, used two-photon calcium imaging to look more directly at real-time microcircuit dynamics in awake, active mice. They saw that coordinated activity between functionally related neurons increased as learning progressed, and suggest that this increased coordination may represent a way in which neurons code for learned behavior. Two-photon imaging allowed the researchers to visualize the activity of groups of neurons at single-neuron resolution within a region of 250 times 250 micrometers. This represents a technical advance over other approaches that cannot distinguish individual activity patterns, or correlations between the individual activities of multiple cells in such a focused area.
“Imaging cellular activity while a learning behavior and paradigm is being executed opens up a whole new world of scientific questions and scientific analysis that can be done,” said Grace (Beth) Stutzmann, Rosalind Franklin University in North Chicago, Illinois, who was contacted for comment on this study.
Previous two-photon and multi-photon studies in transgenic mice have been used to perform similar high-resolution observations of cellular dynamics, including changes in calcium levels and caspase activation in AD models (see Kuchibhotla et al., 2008; Kuchibhotla et al., 2009; de Calignon et al., 2010; and ARF related news story). However, these studies have not looked specifically at neuronal activity during a learning task.
In this study, first author Takaki Komiyama and colleagues trained thirsty mice to lick for a water reward in response to one odor (hit) but not lick in response to a second odor (correct rejection). The mice were immobilized by their heads under a two-photon microscope to permit imaging during the licking trials. It turns out that the mice rarely missed a hit, so the scientists assessed behavioral performance by calculating the fraction of correct rejections. Mice learned rapidly, reaching a learning criterion of 60 percent correct rejections within a single training session.
Using microstimulation or retrograde trans-synaptic tracing, Komiyama and his colleagues identified two distinct motor cortex areas involved in the control of licking behavior. The researchers observed ensemble activity of layer 2/3 neurons in these regions by loading one region or the other with a calcium-sensitive dye called OGB-1 AM. The fluorescence signals from the calcium dye evoked during the licking task provided a measure of instantaneous neuronal activity events.
In both motor cortex regions, the authors observed several different neuronal responses. They also observed dynamic activity in individual neurons within a training session. They identified three main types of response: task neurons that were active during both correct lick and correct rejection trials, hit neurons that were active only during correct lick trials, and correct rejection neurons that were active only during correct rejection trials. The time at which the activity of a neuron began to diverge between hit and correct rejection trials (time until divergence) also varied between neurons.
Next, the authors analyzed the functional relationships between the imaged neurons. Neurons with different response types were randomly distributed in space, so neurons with the same response type were not clustered together. However, the timing of neuronal activity was more closely matched between neurons with the same response type than between neurons with different response types. The timing of activity was even more closely matched if neurons shared the same response type and had the same time until divergence. These temporal correlations in activity decreased with distance at lengths similar to the range of a typical dendritic or axonal arbor (~154 micrometers). Therefore, the authors suggest that these correlations likely reflect local synaptic connections between correlated neurons.
Finally, Komiyama and colleagues observed that temporal correlations between neuron pairs with the same response type increased significantly with learning, both within a single training session and across training sessions. They suggest that this strengthened coupling between functionally related neurons may reflect learning-related reorganization of microcircuit connections.
In an interview with ARF, Komiyama noted that these experiments “reveal a striking degree of plasticity of cortical microcircuits within a learning session. It would be interesting to ask what the microcircuit plasticity looks like in models of psychiatric and neurodegenerative diseases like AD, which are well known to affect the ability to form new memories.”
AD researchers contacted for comment on this study note that some behavioral tasks, like the Morris water maze, would be poorly suited to two-photon imaging of head-fixed mice. However, training in other tasks such as object recognition, for example, should be easier to adapt. More portable two-photon imaging units have also been developed for use in rats (see Helmchen et al., 2001), suggesting that future technological advances might expand potential imaging applications.
For her part, Stutzmann noted that the potential to look for learning-related deficits in the connections and correlations between individual neuronal populations “could give us new insight into how memory is encoded differently or how that encoding might go awry in AD.”—Elizabeth Eyler.
Elizabeth Eyler is a freelance writer in Baltimore, Maryland.
Komiyama T, Sato TR, O’Connor DH, Zhang Y-X, Huber D, Hooks BM, Gabitto M, Svoboda K. Learning-related fine-scale specificity imaged in motor cortex circuits of behaving mice. Nature. 2010 Apr 7. Abstract