We worked first on the design of highly predictable and repetitive spatiotemporal patterns of stimulation. A thorough study of the system was first done using rich spatiotemporal stimulus dynamics. We made a detailed characterization of the responses of a large population of single units during multi-whisker stimulations. We obtained the population filters which describe the stimuli eliciting the most neuronal activity in the barrel cortex. These filters were used as the deflection profiles of individual whiskers during the construction of the stimulation patterns used to reveal the expectancy signals.
For the construction of the stimulation protocol, we tested different stimulus trains. We built a long protocol consisting on three stages: a pre-test, a training and a post-test. Test phases contain truncated patterns of stimulation in which the expectancy signals can be revealed only after the training. We also looked for short term effects, by including randomly interleaved truncated tests in 10% of the training stimulus train. Converging spatiotemporal patterns of stimulation were provided in the rostro-caudal axis, which has been shown to elicit the bigger responses on the barrel cortex. The optimal training length was of one hour.
In order to provide the controlled multi-whisker deflections, we calibrated and programmed a dedicated tool, called the Matrix (Jacob et al. 2010). This device allows a micrometrically controlled displacement of the 24 caudal whiskers in the snout of an isoflurane anesthetized rat. We measured simultaneously neuronal activity in the ´barrel cortex´ using multielectrode silicone probes. According to the ‘predictive coding’ theory, different populations of neurons may be encoding predictive signals and error signals in different layers of the cortex. In view of this, we selected a linear configuration containing 64 channels that allow the identification of the layers.
Experiments were done and the data was analyzed separating into multi-units and single unit activity and a more global scale of analysis called ´local field potential´. We found different responses during the training phase and in the test phase. Also, three different timescales were shown to exist. The first one, related with a predictive signal appears in an anticipated manner relative to the missing stimulus. The other two signals are delayed, and they denote either a surprise, or a delayed mismatch, as it is suggested by the early and late timing of the neuronal activity.
These results provided a neuronal premise on which we were able to design a model of the activity of the population of the cells in the targeted barrel of the barrel cortex. By means of simulations of the average neuronal activity of columns of the cortex related specifically with this and neighbor barrels, we aimed at reproducing the measured results. This modeling is on progress and we expect to provide further understanding of the system by means of experimental predictions for future experiments.
Regarding the dissemination of results:
A refereed article was published in Nature Communications 9(1) 2018
Five poster presentations were given in International Conferences:
1. COSYNE. Lisbon, Portugal. Mar 2019
2. Neuralnet GDR. Paris, France. Dec 2018
3. Sfn. San Diego, US. Nov 2018
4. Barrels XXXI. Riverside, US. Nov 2018
5. FENS. Berlin, Germany. Jul 2018
6. Sfn. Washington, US. Nov 2017
Results were also presented in three lectures:
1. Argentina Neuroscience Society Meeting. Córdoba, Argentina. Oct 2018
2. TENSS Summer School. Transylvania, Romania. Jun 2018
3. Barrels XXX. Baltimore, US. Nov 2017
Finally, five invited seminar talks were given:
1.Neuroscience Institute. UC Berkeley. Berkeley, US. Nov 2018
2-5.Institutes IFIByNE, IbioBA, Institute Leloir and IFIBIO. Buenos Aires, Argentina. Oct 2018
Publication of the related results on the framework of this project are planned to appear in the form of a manuscript during the year 2019.
