Globalization or Regionalization? How the structural organization of neuronal networks shapes their activity patterns and vice versa
Photo by Samora Okujeni
Interactions between neurite growth and synaptic activity constitute the basis for the self-organization of neural networks. Neurons that receive little attention increase their neurites, arborized processes that extend from their cell bodies, to form more synapses with other neurons, and in doing so increase their input. Conversely, neurons with excessive input withdraw their neurites. In early development, when neurons are still isolated, they can also move towards other neurons to increase the probability to form a connection. But what are the mechanics behind this behavior of individual neurons and how does this translate into connectivity and activity patterns on a global scale?
Video on Media Portal University of Freiburg
“These questions are difficult to address in intact brains, where all architecture has already been established and neurons are entangled in a dense mesh of neurites. In order to address these questions, we investigated developing networks grown in cell culture. This approach enables us to access the molecular processes that regulate neurite growth and to manipulate them,” Okujeni explains. “Our key to manipulate network growth in cell culture is the enzyme Protein Kinase C. This protein regulates the stability and turnover of the cellular skeleton as a function of synaptic activity. A flexible cytoskeleton allows neurons to move and extend as well as to withdraw their neurites. A stable cytoskeleton with reduced turnover tends to grow continuously, instead. It is too stiff, so to speak.”
There seems to be an optimal level of activity that each neuron strives to achieve by adjusting its level of connectivity. This results in a homeostatic loop in which neurons adjust their growth and migration according to their input. The scientists were then able to address how changing the translation between activity and growth in neurons influences the architecture of the network and the activity patterns as a whole.
The researchers found that shifting the balance towards flexibility promoted aggregation of neurons into regular clusters and leads to short neurites. More stability, on the other hand, led to pronounced long-range connections between homogeneously distributed neurons. Network stability allowed longer neurites with more connections, but impaired withdrawal and movement. Simulations of networks showed that a mix of connections to nearby and distant neurons as well as their organization of neurons in clusters help to keep activity going. These types of structured networks are also believed to have additional advantages, e.g. for information processing.
But what does this mean for activity? The clustered networks showed very high activity in many places that often failed to recruit distant neurons. In homogeneous networks, on the other hand, overall activity was low, yet the occasional massive bursts of activity starting in a few distinct areas fired up the whole network. “However, if the neurons were left without manipulation,” Okujeni explains further, “they would settle in the middle. They formed networks that emphasized local interactions while maintaining global connectivity. Such networks generated more diverse patterns of activity propagation, recruited most of the network and fired at the highest levels.”
Interestingly, networks with the highest connectivity were not the ones that generated the highest activity levels. The degree of clustering seems to be a crucial factor for activity generation. Apparently, self-organization balances growth and activity to form inhomogeneous networks with the richest possible set of activity patterns.
So what would be the advantage in having this structure? Theoretical analyses show that this type of network architecture allows the largest number of different states, which increases space representation, memory capacity and information transmission. The next item on the agenda of the investigators is to test if these predictions hold up in real neuronal networks.
Image Caption
Neurons with a flexible cytoskeleton form very regular clusters with dense connections within clusters but with few connections between clusters. In this image the neurons were stained for cytoskeletal proteins MAP2 (green) and neurofilament (red), expressed in dendrites and axons, respectively.