3 April 2012. Trillions of synapses in the human brain connect neurons near and far, with information traversing different paths within this maze of connectivity. But don’t be fooled: this elaborate network reflects some pretty basic principles, according to two new studies.
One, a modeling study led by Edward Bullmore of University of Cambridge, United Kingdom, published online March 30 in the Proceedings of the National Academy of Sciences, proposes that the brain’s wiring diagram is generated by two competing processes that, when “detuned,” result in patterns found in schizophrenia. Another study, published March 30 in Science and spearheaded by Van Wedeen of Massachusetts General Hospital in Charlestown, distills the weblike connections visualized by diffusion magnetic resonance imaging (MRI) into an orderly, three-dimensional grid. Together, the studies suggest that grasping the logic behind the tangle of neural connections may make it easier to detect areas of pathology in brain disorders of all kinds, including schizophrenia.
It’s a small world after all
Abnormal brain neurodevelopment has long been suspected in schizophrenia, and genetic risk factors such as DISC1 are implicated in various processes of brain-making, including neural proliferation, migration, and synapse formation (see SRF related news story). But these findings have not yet congealed into an understanding of the core differences in the wiring diagram of individuals with schizophrenia. Brain imaging offers a more top-down approach, and several different aspects of brain organization to measure.
One option is to study the default-mode network, a collection of brain areas that are active when a person is awake but at rest. Functional MRI picks up this resting-state activity, and correlations in its ups and downs between two regions point to functional connections between them, thus capturing the inherent organization of at least some of the brain’s pathways. Aberrations in the default-mode network have been found in schizophrenia, including hyperactive resting-state activity that correlated with worse performance on a working memory task (see SRF related news story). Similarly, a recent study identified differences in resting-state activity that distinguished schizophrenia from bipolar disorder—differences that were shared with unaffected first-degree relatives (Meda et al., 2012). This suggests that focusing on resting-state activity could help find the genetic factors contributing to altered networks in disease.
The study by Bullmore and colleagues used resting-state activity, but went beyond cataloging differences in functional connectivity between particular regions to get at the shape, or “topology,” of the brain’s overall connectivity patterns. This approach focuses on the number of connections emanating from any one region, or “node,” to identify hubs that contact many nodes, and clusters of nodes that communicate largely with each other; the result looks much like an airline’s flight map across the world. Researchers have drawn from graph theory to analyze the network properties of a range of brain data, including gene transcription (see SRF related news story), neuroanatomical structure, and functional connectivity (Bullmore and Sporns, 2009).
Seen this way, the brain looks like a “small world” network in that the path between any two nodes is short compared to a random network. (Social networks are similar, with small degrees of separation between people within it, prompting exclamations of “Small world!” upon discovering a mutual acquaintance.) This small world structure is disrupted in schizophrenia (Liu et al., 2008), with studies finding an increased connection distance and fewer hubs in fragmented networks (Bassett et al., 2008; Lynall et al 2010).
To address why this might be, first author Petra Vértes and colleagues set out to understand the forces that shape these networks in the first place. The researchers developed a computer model to mimic the network shape that emerged from resting-state fMRI data from 140 cortical areas in a group of 20 healthy study participants. One prevailing idea is that networks organize themselves “economically” so that the high-maintenance long-distance connections are few; however, the researchers found that simulations based solely on this distance penalty—in which the probability of a connection between two regions dropped with the distance between them—did not reproduce the network’s properties. When they added another “clustering” rule by which the probability to form a connection increased between two regions if they were already connected to a common neighbor nearby (akin to a mutual acquaintance), this reproduced the network in terms of clustering, efficiency (related to the path lengths within the network), and modularity (the community structure of nodes densely connected to one another).
They found that this two-rule “economical clustering” process could also account for network properties on which the computer model was not trained, like the distribution of distances between any two nodes in the network. Also, the particular values for the distance penalty and the clustering rule that best fit the first dataset also predicted the network features obtained from a second fMRI dataset from 12 healthy participants. Finally, the researchers found that the model could account for the altered network configurations found in 19 individuals with childhood-onset schizophrenia, which display less modularity and clustering—but only when the model parameters were tweaked away from the values that worked for healthy participants. This suggests that in schizophrenia, the distance penalty rule is still in effect but less strictly enforced—something that could fragment a network by allowing more long-distance connections.
Though this exercise does not prove that the brain actually uses these rules—which have biologically plausible correlates—it does suggest that the complexity of these networks may stem from a small number of factors.
In the Science paper, Wedeen and colleagues used diffusion spectrum MRI (DSI) to visualize the actual trajectories these connections take within the brain. A variant of diffusion-weighted imaging techniques that infers white matter structure from the direction of movement of water molecules, DSI can detect different diffusion patterns within a single voxel, making it sensitive enough to resolve fiber tracts that crisscross each other (Wedeen et al., 2008).
DSI of whole-brain samples from four different nonhuman primate species and of six live human subjects visualized the spaghetti of white matter, and ensuing analysis reconstructed its paths taken throughout the brain. To get at the spatial relations between these different pathways, the researchers pinpointed the main path of a small region, then delineated the “path neighborhood” for it, which consisted of all other paths that shared at least one voxel with the main path. This revealed a grid-like sheet, with pathways running either parallel to or crossing perpendicularly to the main path—much like the fibers within a woven cloth. In a three-dimensional view, these gridlike sheets were stacked on top of each other, and diagonal paths were not observed. For example, analysis of a fiber tract known to split into two seemingly diagonal branches in the frontal lobe of the rhesus monkey revealed that this split actually consisted of one branch turning 90 degrees into a different plane.
Click on the image to launch the video.
The human brain's connections turn out to be an orderly 3D grid structure with no diagonals. 2D sheets of parallel fibers cross at right angles, "like the warp and weft of a fabric." Image credit: Van Wedeen
Though these surfaces were curved and folded to fit within the brain, their basic structure was preserved across different parts of the brain and across species. The researchers hypothesize that this simple structure reflects the three axes of axon-guiding chemical gradients operating during brain development. This new coordinate system may make it easier to pick up on connectivity pathologies in brain disorders like schizophrenia, which has been characterized by white matter abnormalities (Whitford et al., 2011).
With the rise of imaging genetics, research will soon explore the extent to which a particular variant controls these global views of brain organization. Indeed, another study in the same issue of Science finds the fingerprints of the genetic program upon cortical organization (Chen et al., 2012). Using structural MRI to delineate the cortical surface area of 406 twin pairs, the researchers compared monozygotic to dizygotic twin data to identify 12 genetically based subdivisions similar to recognized regions within the brain. This “genetically parceled” brain atlas, and the connectivity grids and topologies observed in the other studies, represent refinements of brain phenotype which may better illuminate the influences of genetic risk factors in disorders like schizophrenia.—Michele Solis.
Vértes PE, Alexander-Bloch AF, Gogtay N, Giedd JN, Rapoport JL, Bullmore ET. Simple models of human brain functional networks. Proc Natl Acad. Sci USA. 2012 Mar 30. Abstract
Wedeen VJ, Rosene DL, Wang R, Dai G, Mortazavi F, Hagmann P, Kaas JH, Tseng WY. The geometric structure of the brain fiber pathways. Science. 2012 Mar 30;335: 1628-1634. Abstract