A new day for an old emotion: studying fear learning using awake mouse functional magnetic resonance imaging (Commentary on Harris et al.)

Determining the relationship between functional brain networks and learning-induced changes in neuronal activity is a daunting task. Apart from the sheer anatomical complexity of the mammalian brain, one reason why this is such a challenge is that there is a high likelihood that information is processed and transferred along different neural structures over time. Therefore, establishing a causal relationship between specific neuronal pathways and learned emotional behavior is not straightforward and requires improvements in the currently available techniques used for measuring and analyzing neural activity. An ideal technique should be non-invasive, non-destructive to brain tissue, provide large-scale spatial as well as refined temporal data and, importantly, it should provide data obtained in awake experimental animals under minimized stress conditions. In the current issue, Harris et al. (2015) move the field a step forward towards such a technically challenging objective. In their work, they sought to determine the neural circuitry of fear learning in awake mice using functional MRI (Harris et al., 2015). They applied a traditional behavioral neuroscience procedure to train mice to associate a conditioned stimulus (CS; flashing light-emitting diodes) to an unconditioned fear-eliciting stimulus (electric shock of the paws). Twenty-four hours after initial training, mice were scanned in a 7-Tesla MRI scanner for their blood oxygenation level-dependent (BOLD) response to the CS. Results in the CS-paired group were compared with a CS-unpaired group. To get to this point, however, the research team trained mice under two different protocols of restraint (5 vs. 12 days) and MRI noise presentation (attenuated vs. full sound delivery). They determined that more days of habituation training is better, as is a slower build-up of the MRI noise, which led to mice optimally tolerating the restraint and sounds after 12 days. The 12-day paradigm resulted in reduced blood levels of stress hormone (e.g., corticosterone in rodents) and body movements during restrained imaging. The principal finding of the study is that the amygdala, specifically in the right hemisphere, encodes the fear-eliciting CS. The finding that the right amygdala shows high responsiveness to a paired CS laden with strong negative emotional value is not new. For years the amygdala has been shown under a variety of conditions to be a critical site involved in fear learning and the processing of emotional memory. However, the nucleus accumbens of awake mice in the non-paired group showed a heightened response to the CS, which in their case signaled that the electric shock would not be presented (thus, it constituted a ‘safety’ signal). Such a safety signal was not present in mice in the CS paired group. This novel finding is a direct result of the technical ingenuity applied to the fMRI technique in this study. Harris et al. (2015) provide an outstanding validation of the awake mouse fMRI technique using this highly characterized rodent fear conditioning model (Harris et al., 2015). To further validate the use of fMRI in awake mice to understand the neurobiology of fear-learning, it will be important for future experiments to conduct experimental manipulations that establish a direct link between fMRI signal changes and the expression levels of the synaptic protein machinery involved in plasticity during learning and memory retrieval stages. It will also be important to determine whether there are time-varying brain regional changes in CS-induced activation at time points beyond a 24-h period (can the contributions to long-term emotional memory storage of other regions outside the amygdala be measurable by using awake mouse fMRI?)

The application of fMRI in awake rodents is a powerful approach, more so when combined with a variety of novel experimental manipulations that can uncover new and important mechanisms (Ferris et al., 2011). However, as with other neuroscience techniques, the current state of this approach is not without its share of shortcomings (Febo, 2011). Among the most urgent concerns expressed by peers in the field is the role of stress in affecting the final results. A further concern is whether the amount of movement during data acquisition is tolerable and does not influence the quality and reproducibility of the data. Harris and colleagues directly tackled these two issues and their results lend support to the notion that stress and body movements can be minimized through improvements of the procedures used to habituate mice to restraint. This is consistent with prior results in rats (King et al., 2005).

There are more pressing questions regarding the measurements in fMRI in awake rodents. Does fMRI of the awake mouse provide better data representing underlying neuronal encoding properties of large-scale brain networks? By using the awake mouse fMRI technique, will we be able to capture biologically meaningful functional signatures embedded in malleable synaptic circuits? In their paper, Harris and colleagues hint that the answer to these questions is ‘yes’. As suggested by this research group, manipulation of mouse genetics in specific neural circuits to affect specific protein functions will be a way forward. Another limitation in the field of animal imaging is the lack of behavior expression during the acquisition of the BOLD fMRI data. As the animals are restrained (and in many cases anesthetized), this sets an upper limit to the amount of applications that can be done at the moment (Febo, 2011). Virtual reality technology such as that employed in combination with two-photon imaging of cortical calcium signal dynamics in awake mice might be the answer to addressing this limitation (Dombeck et al., 2010). The solution to the above issues, however, cannot be to merely conduct fMRI studies in anesthetized animals, which would only provide small incremental steps (if any) in our knowledge of the brain and its networks (and be fraught with controversy regarding which anesthetic type and amount is the best; Febo, 2011). The solution will also not be to not apply animal imaging at all and instead only use a reductionist approach ignoring the complexity of the CNS by focusing on localized manipulations of single neuronal pathways. The more innovative solutions favored by the present work by Harris and colleagues needs to produce advancements in habituation to restraint, establishment of an ‘ideal’ or improved mouse model for awake imaging, cause–effect genetic manipulations, raising statistical power and rigor, and the use of safe contrast-based mechanisms that can bring us closer to directly listening to neurons and synapses instead of indirectly listening behind the walls of the cerebrovasculature.

Disclosure statement

The author does not have any personal or financial interest to disclose.