SOMATOLEARNING · The cortical circuits of associative learning

The cortical circuits of associative learning

Research ought to aim at elucidating the mechanisms of learning. One important quality of humans and other evolved mammals is their capacity to cope with ever changing environments. They owe this trait to a formidable evolutionary development of the neocortex that increases their capacity for analysing and adapting to the environment. Although associative learning occurs in species without neocortex (e.g. the aplysia), investigating how cortex impact learning is crucial: accumulating evidences support our hypothesis that primary sensory areas do more than simply integrating sensory inputs. The urge of studying the mechanisms of learning is also driven by questions of public health: if one understands the mechanisms of learning, one can also design better strategies to treat its dysfunctions. At the extreme end of the spectrum of inherited learning impairments, fragile X syndrome is a major cause of intellectual disability (IQ < 70) and autism that affects one in 3 000 males (Crawford et al., 2002). Strikingly, as many as one in 350 women carries the premutation form and will potentially transmit the fully expressed syndrome to her progeny (Crawford et al., 2001).The USD 40 billion of indirect costs caused by mental retardation handicap which prevents individuals to enter the economically active life (Mortality and Morbidity Weekly Report, January 2004) calls for more efforts for developing treatment enabling their integration. The severity of fragile X impairments might represent an unfathomable challenge. But milder forms of learning deficit exist and they are also more common: one estimate at 5 % the ratio of school-aged children suffering from learning disabilities which cannot be explained by mental retardation, autism, or schizophrenia (source UCLA / Wallis Foundation). Thus, studying the anatomy of learning is both a fundamental quest of neuroscience and a societal priority. Associative learning tags a neutral sensory stimulus with the emotional valence of an aversive or appetitive event. One well described consequence is the enlarged representation of the conditioned stimulus (CS) in the sensory map. However, virtually nothing is known about the cortical circuitry underlying this phenomenon. Neuronal circuits in primary sensory cortices are patterned in columns, one for each sensory stimulus. Together, the cortical columns form a topographic map of our sensory body, the homunculus. This cortical map is plastic and can be modified by sensory experience and by learning. In rodents, the deflection of a whisker previously paired with a shock enlarges the representation of the conditioned whisker in the barrel cortex. What are the circuits propagating the CS inputs across cortical columns, throughout cortex? What aspects of the behavioural fear response do they control? Our objectives are to: (a) identify cortical circuits underlying the transformation of the cortical whisker map upon learning; (b) investigate the cellular mechanisms of plasticity; (c) probe the relationships between the circuit changes and the behavioural response. Experimental work and main results A first study was published early 2011 (Rosselet et al. Frontiers in Neural Circuit). We trained mice in a whisker-discriminative paradigm to associate the deflection of one whisker row with a mild tail shock and probed the changes in the intracortical circuits of barrel cortex using glutamate uncaging with laser scanning photostimulation (LSPS). This technique combines optics and electrophysiology to find in a slice all presynaptic cells to one neuron and examine the pattern of their connections. This technique was previously used to track the development of cortical networks. We used it for the first time to describe their plasticity linked to associative learning. We found multiple circuit alterations in cortical columns that were adjacent to the column of CS and, surprisingly, no change in the column of CS itself. Our claim is that these alterations serve the recall of memorised events by propagating the CS inputs across cortical columns, throughout barrel cortex. This effect should amplify the CS signal before it is sent to other key players of associative learning. In support of our model, we show that these circuit changes are linked to the learning performance of animals as they did not occur in mice that were trained the exact same way but did not show a behavioural fear response, and the strength of a particular projection correlated with the ability of the animal to discriminate the CS amongst other sensory stimuli. We are currently working on another study to address our objectives. To do so, we use conditional mutants which are models of fragile X syndrome. Fragile X mental retardation (FMRP), the missing protein in fragile X, modulates synaptic plasticity, mostly acting as a break. Our strategy is to use the conditional mutant mouse to alter this break in a specific layer of barrel cortex. This trick will allow us to assess the consequences of an improper function of barrel cortex in our training paradigm and more generally to evaluate the role of primary sensory cortices in associative learning and the consequences of their impairment. A first step to validate our model was to verify that FMRP was indeed modulating the plasticity of cortical maps in adult mice. To do so, we used a simpler paradigm to enlarge the representation of whiskers by clipping all but two rows of whiskers. We then mapped with LSPS the neuronal circuits in the columns corresponding to the spared whiskers and found that plasticity was enhanced in the conditional fragile X mutant. Our ongoing study of the cellular mechanisms suggests that FMRP is acting on neurotransmitter release to 'rein' the strengthening of circuits induced by partial sensory deprivation. These results are exciting as they unveil yet another role of FMRP and give us the thumb-up to test these mice in our conditioning protocols.

Data: CORDIS, © European Union