ARTICLE IN BRIEF
Two studies support the hypothesis that patterns of pathology in dementing illnesses strongly correlate with the patterns of connectivity in the healthy brain.
Neurologists have long been intrigued by the differing patterns of brain atrophy in neurodegenerative diseases. Recent findings from autopsy studies and cell culture models have strongly suggested that disease pathology spreads across synapses, and that this may underlie the observed patterns of degeneration. Now, two papers in the March 22 Neuron provide yet more support for that hypothesis — they show that patterns of pathology in dementing illnesses strongly correlate with the patterns of connectivity in the healthy brain.
The results, according to Reisa Sperling, MD, associate professor of neurology at Harvard Medical School in Boston, who was not involved in either study, “really cement the concept” that “the intrinsic network organization of the brain is actually very relevant to the study of neurodegenerative disease. Things that fire together wire together, and, sadly, die together.”
Investigators, led by Helen Juan Zhou, PhD, and William Seeley, MD, associate professor of neurology at the University of California, San Francisco, compared the pattern of brain functional connectivity, derived from task-free fMRI of healthy individuals, with atrophy patterns in five different neurodegenerative diseases: Alzheimer disease, behavioral variant frontotemporal dementia, semantic dementia, progressive nonfluent aphasia, and corticobasal syndrome.
“We wanted to know whether healthy brain network architecture can actually inform us about how the brain will degenerate in the context of disease,” said Dr. Seeley. That idea has been around for several decades, he noted, but his intent was to use graph theory to quantitatively characterize brain networks, and then to rigorously test several competing hypotheses related to connectivity-vulnerability relationships.
One such hypothesis was “trophic failure.” In this model, the most vulnerable brain regions are those with the fewest connections to other regions, since they receive fewer trophic inputs than other, more connected regions.
Another hypothesis is “nodal stress,” in which regions experiencing the greatest flow of information are at greatest risk of failure, whether by excitotoxicity, wear and tear, or some other use-related process.
A third hypothesis is “transneuronal spread,” in which the disease spreads outward from “epicenters” of degeneration, regions in which the disease may actually begin, through synaptic contacts. In this model, the most vulnerable regions are those connectionally closest to the epicenters.
Graph theoretical analysis is ideally suited to analyze these competing hypotheses, Dr. Seeley said, since in its mathematical form, it assigns each brain region a set of parameters, including connectedness or clustering, flow, and functional distance from other regions.
The atrophy patterns for each disease were used as templates to examine activity in healthy brains within these same regions. Using fMRI from the healthy brain, each of over 2500 regions within the affected areas were assigned values for connectedness, flow, and distance from epicenters. Epicenters were identified by statistically testing each region to identify those regions whose connectivity patterns in the healthy brain most closely mirrored the disease vulnerability pattern, Dr. Seeley said. “We ended up with over 3 million node pair connections,” for the five atrophy patterns, “reflecting all these different regions and their patterns of connectivity.”
The analysis indicated that the shorter the path to an epicenter in health, the greater the atrophy in disease, a finding that held true across the five disease patterns. “This supports the role of transneuronal spread as an explanation for disease progression,” Dr. Seeley concluded. There was a weaker effect from total flow in all five patterns, supporting a contribution from nodal stress factors, but no consistent effect from functional clustering, arguing against the trophic failure hypothesis.
A fourth model, referred to as “shared vulnerability,” could not be easily tested in this analysis. In that model, cell-intrinsic factors, such as gene expression patterns, determine a neuron's vulnerability. Dr. Seeley conceded that such cell-intrinsic factors could still contribute if they correlate with epicenter proximity, but for these factors to give the appearance of the network dependence seen in the study would require several unlikely coincidences.
There are really two major questions for every neurodegenerative disease, he said: how and where it starts, and how it spreads. “Our data do not address the question of how the disease starts. There has to be something special about the epicenters that makes them vulnerable to disease.” But, he said, “I think the evidence is starting to pile up to support [the transneuronal] mechanism” for spread.
An intriguing unanswered question is whether each disease requires disease-specific epicenters. “The question I think we could now address is, what if you started the disease in another region that has no particular importance in Alzheimer's disease? Would you find the same thing? Once the misfolded protein is there, does it spread obeying its own rules and kinetics, or does it have to be situated in a particular circuit milieu to propagate and spread?”
In the second paper, Ashish Raj, PhD, assistant professor of computer science in radiology at Weill Cornell Medical College in New York, used structural, rather than functional, connectivity in the healthy brain as the basis for building a model of spatial diffusion of disease within the brain. In his model, disease activity acts like a diffusing gas, with the diffusion rate proportional to concentration of the pathogenic substance, but instead of spreading in all directions equally, the disease moved along the most persistent connectivity subnetworks.
Within the mathematical formula of Dr. Raj's model, these stable subnetworks are represented as “eigenmodes,” or persistent oscillation patterns. In the model, an eigenmode acts as an attractor or trap for the pathological agent. “Once the agent enters this subnetwork, it remains confined within it, and escapes from it only very slowly,” Dr. Raj said.
By comparing the most persistent subnetworks within the model of the healthy brain to the atrophy patterns of the common dementing illness, Dr. Raj showed that one mapped onto the other; specifically, the differing spatial patterns of atrophy in Alzheimer's disease and frontotemporal dementia matched different subnetworks within the healthy brain.
Dr. Raj noted that his work is in some sense complementary to that of Dr. Seeley. His study was based on structural connections and used a model of spread to derive the classical patterns of dementias, while Dr. Seeley's was based on functional connections, and used dementia patterns to ask about the most logical mechanism of spread.
Dr. Seeley noted: “The fact that we came to a very similar set of conclusions without either side knowing about the other's plans is a really pretty convincing thing.”
In an interview with Neurology Today, Dr. Sperling praised both studies, but noted that there is still a gulf between the demonstration that disease spreads through networks and the conclusion that misfolded proteins are the toxic agent. The two ideas “are consonant,” she said, “but we somehow need to fill in the black box between these two levels.”
Using the tools from these studies could help in the design of preventive clinical trials, Dr. Sperling noted, allowing a more fine-grained subgrouping of patients than is currently possible, based on atrophy pattern and likely progression. Further, she said, “I believe that eventually you could develop an algorithmic approach that would be helpful on an individual subject level,” to better understand prognosis. “The ultimate next step is: how can we use these networks to monitor response to therapy, and predict progression clinically? That's the name of the game.”