In the broadest terms, neurodegeneration is caused by aggregates of misfolded proteins, and traditionally, specific proteins have been associated with individual diseases, such as amyloid beta and tau with Alzheimer’s disease and alpha-synuclein with Parkinson’s disease.

However, in practice, aggregates of multiple different proteins are found both in individual patients, regardless of their diagnosis, and in healthy elderly individuals. Disease categories such as tauopathies and synucleinopathies also demonstrate that the same misfolded protein can manifest in multiple different ways depending on the context.

Now, investigators have developed a new approach to classifying neurodegenerative disorders that used the overall patterns of protein aggregation, rather than specific proteins, to define six clusters of patients that crossed traditional diagnostic categories.

“We find that perhaps the way that clinicians have been diagnosing these disorders… is not necessarily the way these disorders work,” Danielle Bassett told BioWorld. “The way we’ve been trying to carve nature at joints is not the way that nature has joints. The joints are elsewhere.”

Scientifically, she said, the work shows that “there are biological mechanisms that are behind some of these observations [of clinical heterogeneity] that we’ve had for a long time.”

Bassett, who is a physicist by training and describes her interest in the broadest sense as “problems at the intersection of basic science, engineering and clinical medicine that can be tackled using systems-level approaches,” holds a joint appointment from no fewer than five departments at the University of Pennsylvania. She is J. Peter Skirkanich Professor of bioengineering, neurology, psychiatry, electrical and systems engineering, and physics and astronomy. She is the senior author of the paper describing the new classification approach, which appeared in the Aug. 3, 2020, online issue of Nature Biomedical Engineering.

In their work, the researchers used both unsupervised and supervised machine learning to look at the overall patterns of seven different pathologies across 15 brain areas in nearly 900 postmortem brains.

Those pathologies were tau tangles, alpha-synuclein and TDP-43 inclusion bodies, neuritic plaques (whose main component is beta-amyloid), neuronal loss, angiopathy and gliosis.

Not six of one, half dozen of the other

The approach segregated patients into six different clusters that had some overlap with traditional neurodegenerative diagnostic schemes – one cluster, for example, contained many patients that were diagnosed with synucleinopathies under the traditional scheme.

However, patients with high levels of both alpha-synuclein and what the authors called Alzheimer’s disease neuropathologic change – that is, both amyloid beta and tau pathology – formed a distinct cluster, demonstrating that the focus on one or two proteins could lump together patients with very different overall patterns.

Bassett said that she and her colleagues were surprised by just how common co-pathologies that straddled diagnostic categories were in the brains they analyzed.

“We had certainly predicted large amounts of co-pathology,” she said, “but we hadn’t predicted the extent of it.”

The team also showed that the clusters could explain some of the clinical heterogeneity in clinical measures, including cognitive performance.

The particular pathologies they looked at, Bassett explained, are “ones that have consistently been acquired” in the databases the team used for its analysis, enabling them to analyze a large group of patients.

But the team also showed that the same clusters could potentially be generated from data available from living patients – a combination of the Mini-Mental Status Exam cognitive test, protein levels in cerebrospinal fluid (CSF), and patients’ genotypes at the APOE and MAPT loci. (Even though CSF measuring of protein levels cannot give direct information on where in the brain those proteins come from, elevated protein levels in different brain regions lead to different protein levels in the CSF.)

Bassett said markers that are being used for diagnoses today are “quite different” from those in the databases. Ultimately, she and her colleagues hope to use modern diagnostic methods such as neuroimaging to develop less invasive identification methods for the same clusters.

An earlier paper on which Bassett is a co-author described using neural activity to measure the connection strength between multiple brain regions and developing a similar transdiagnostic approach to neuropsychiatric disorders.

And a blood test, analogous to one presented at the 2020 Alzheimer's Association International Conference (AAIC), could also be a less invasive route to the new diagnostic categories, though Bassett and her colleagues have not looked at possible blood biomarkers for their clusters yet.

Asked whether the work could eventually form the basis of new diagnostic categories for neurodegenerative disorders, she said that “this is certainly a step in that direction,” though at this point the findings need to be validated and clinically useful strategies developed to capture the categories she and her team have discovered.

“I’m optimistic,” she said, “but there is a lot of work that has to happen.”

No Comments