INTEGRATING TOP DOWN AND BOTTOM UP PROCESSES IN THE MARMOSET AND MACAQUE CORTEX – CORNET

We combine the complementary skills of both teams to improve our understanding of cortical processes by developing studies on the marmoset brain to integrate large scale structural and functional models of the cortex. We use retrograde tracers to derive a weighted and directed matrix of marmoset cortex connectivity. Tracers are injected into the marmoset areas scattered throughout the cortex. The visual cortical areas are the focus of attention in order to establish an extremely precise mapping of the connectivity of this region. High-density electrocorticography (hdECog, 200 electrodes / cm2) is used to simultaneously record the visual areas of the occipital, temporal, parietal and frontal regions, emphasizing the role of local and inter-local rhythmic synchronization. Some experiments with retrograde tracers are conducted on animals for which we have electropysiological data in order to confirm and consolidate electrophysiological and anatomical mapping. These data are used to construct models of dynamic causality in order to reveal the mechanics of visual predictive coding to a degree of precision never reached. The work on the marmoset is extended by developing a Bayesian framework to complete the structural cortical network. We improve the existing algorithms to complement the data and quantify the uncertainty by incorporating distance and weight in our imputation procedures. Imputation is of paramount importance to complete the network in the marmoset, macaque and mouse. We are undertaking further study to improve the existing macaque database from an atlas of 91 to 131 areas. Conversion of our existing data into weighted areas to a finer degree allows us to optimize the correlation with electrophysiology and more flexible segmentation.

We have successfully adapted the functional recording techniques developed in the large macque brain to the much smaller marmoset brain. The Fries lab pioneered the development of ECoG recording techniques which makes possible the recording of high-resolution low field potentials across multiple areas. Adapting this technique to the much smaller marmoset brain is extremely challenging but will allow exploiting the gene editing techniques which are being developed in this species. An important addition has been to combine ECoG recording with tracer injections so that we can derive unprecedented insight into structure function relations in the cortex.
Much work on sensory, motor and cognitive processes is currently being undertaken in the mouse cortex. How relevant is the work in mouse cortex to understanding the biology of primate cortex including human ? To investigate this we have used identical tract tracing in mouse and macaque. Surprisingly this showed in our publication in PloS Biology last year that a similar distance rule applies across brains of very different sizes and to primate and mouse brain. However, these experiments also showed that the density of connections between cortical areas falls off sharply with increasing brain size. These findings suggest that there are important limitations in the usefulness of work carried out in the mouse for understanding the human brain.

We have been able to make an important step in better understanding hierarchical processing in the cortex. In a first step we established a structural hierarchy in the macaque brain, we then examined how this was related to a functional hierarchy in the same brain. By establishing the functional signature of hierarchy in the macaque brain we were then able to examine the similar electrophysiological recordings in the human cortex. This enabled us to establish the functional hierarchy of the human cerebral cortex and these findings were published in the international journal « Neuron ». This work opens up an entirely new area of research. Because hierarchy plays such an important role in the development of modern theories of brain function based on predictive coding, whole brain imaging in the human cortex can now be precisely located in a hierarchical framework. Given our findings on the limitations of the mouse model (see above), we predict that the functional exploration of the human brain will become increasingly important

Donahue CJ, Sotiropoulos S, Jbabdi S, Herandez-Ferandez M, Behrens T, Kennedy H, Knoblauch K, Coalson T, Glasser M, Van Essen D (2016) Using diffusion tractography to predict cortical connection strength and distance: a quantitative comparison with tracers in the monkey. J Neurosci 36:6758-6770.
This reports the altogether unexpected findings that dMRI tractography only modestly captures the strength of interareal connections.

Horvat S, Gamanut R, Ercsey-Ravasz M, Magrou L, Gamanut B, Van Essen DC, Burkhalter A, Knoblauch K, Toroczkai Z, Kennedy H (2016) Spatial Embedding and Wiring Cost Constrain the Functional Layout of the Cortical Network of Rodents and Primates. PLoS Biol 14:e1002512.
This reports that the mouse cortex obeys a similar distance rule to that found previously in the macaque cortex, suggesting that this so called exponential distance rule constitutes a universal feature in cerebral cortex.

Richter CG, Thompson WH, Bosman CA, Fries P (2017) Top-down beta enhances bottom-up gamma. J Neurosci. 3771-16
This reports how feedforward and feedback pathways interact.

Ni J, Wunderle T, Lewis CM, Desimone R, Diester I, Fries P (2016) Gamma-Rhythmic Gain Modulation. Neuron 92:240-251.
This reports the influence of feedforward pathways on their target areas.

The Kennedy-Knoblauch lab has established an extensive data base concerning the inter-areal pathways in the macaque cortex. The weight-distance relationships in this database allowed the development of a predictive EDR large-scale model of the cortex that reveals numerous features that can be observed in the data including hierarchical organization (Markov et al., 2013b; Song et al., 2014). The Fries team is internationally known for its development of cutting edge technology for large-scale electrophysiological recording in cortex, leading to important findings showing that cortical computation unfolds in the interplay between neuronal dynamics and structural neuronal connectivity (Fries, 2009). More recently the Fries team’s interest in predictive coding has extended to developing dynamic causal models for looking at asymmetries in feedforward and feedback pathways in the cortical network (Bastos et al., 2015a; Bastos et al., 2012). The teams of Fries and Kennedy-Knoblauch have collaborated in order to examine the correlation of structural and functional hierarchies in the macaque, visual cortex (Bastos et al., 2015b). This work showed that directed influences constrain a functional hierarchy sharing critical features with the structural hierarchy, but importantly exhibiting task dependent dynamics. In the present project the two teams will extend and deepen this collaboration. This involves combining the complementary skills of both teams in order to further our understanding of cortical processing by developing the smooth brained marmoset as an experimental animal for integrating large-scale structural and functional models of the cortex. In the marmoset we shall use identical retrograde tracer technology as in our earlier work in the macaque in order to derive a weighted and directed matrix of the connectivity of the marmoset cortex. Tracers will be injected in marmoset areas widely distributed across the cortex. A particular focus will be made in visual cortical areas where we shall make fine grain mapping of the connectivity of the region that will be explored with physiological recoding. We shall use high high-density electrocorticography (hdECog, 200 electrodes/cm2) to simultaneously record from visual areas. A particular emphasis will be on the role of local and inter-areal rhythmic synchronization. Retrograde tracer experiment will be carried out in animals that have undergone electrophysiological recording so as to co-register electrophysiological and anatomical maps. Those data will be used to construct dynamic causal models in order to reveal the mechanics of visual predictive coding at an unprecedented level of detail. Work on the marmoset will be further supported by developing a Bayesian framework for structural cortical network completion. We shall improve on existing algorithms for data completion and uncertainty quantification by incorporating distance and weight in our imputation procedures. Imputation will be of critical importance for network completion in marmoset, macaque and mouse. We will undertake further studies to refine and extend the existing macaque data base. This will extend our existing cortical matrix based on 91 areas to an atlas of 131 areas. Converting our existing area weights to a more fine grain level thereby allowing more flexible parcellation and optimising correlation to electrophysiology and imaging data. The present proposal will investigate the large-scale structural models in macaque, marmoset and mouse providing insight as to how the EDR model scales to brain size thereby enabling us to deepen our understanding of the embedded features of the cortex. The weight-distance relationships in these three species with different brain sizes will allow us to determine the scaling features of cortex and will have important consequences to extrapolating our present findings using invasive techniques to the large human brain.