People with Alzheimer’s disease share a similar profile. Most begin to show symptoms, notably cognitive decline, late in life. Their brains contain plaques made up of beta amyloid protein and tangles composed of tau protein. They undergo a prolonged decline until, ultimately, they die.
But what if the universal hallmarks of the disease actually represent multiple subtypes at the molecular level?
“Alzheimer’s disease has become an umbrella diagnosis for dementia in a way, even though the only way to definitively diagnose it is post-mortem,” says Nikhil Milind, a rising senior at North Carolina State University. “But more and more people are suspecting that there are different paths that converge at the same end point. We are looking at genetics data, asking a fundamental question about what elements are driving the differences in Alzheimer’s patients.”
Milind may still be an undergraduate, but he’s also the first author on a paper, “Transcriptomic stratification of late-onset Alzheimer’s cases reveals novel genetic modifiers of disease pathology,” just published in PLoS Genetics. Working in the laboratory of Jackson Laboratory (JAX) Gregory Carter, Ph.D.Develops computational strategies using genetic data to understand complex genetic systems involving multiple genes and environmental factors.Associate Professor Greg Carter, Ph.D. , initially as a summer student then, the following summer, as an intern, Milind carried out a computational project analyzing Alzheimer’s patient gene expression data. Working with Associate Research Scientist Christoph Preuss, Ph.D., he found in the data indicates that there is indeed more than one path on the way to Alzheimer’s disease.
Modules and subtypes
Carter, the paper’s senior author, has been looking at gene expression data of Alzheimer’s disease patients and non-affected control subjects. Using machine learning, members of his lab defined modules within the data—groups of genes that all change expression together—and began to look at what they were and what biological processes they represent. Then Milind took it on as his summer student project with Preuss as a mentor for disease genetics. With contributions from Sarah Shapley, another JAX summer student from Baldwin Wallace College, they were able to take it a step further.
“Greg let me play with the data a lot,” Milind says. “What are the complexities, how do you dissect it? We used a method known as iterative WGCNA (weighted gene co-expression network analysis) to take the modules that had been previously generated and refine them into smaller sets. We were able to hone in on specific cells and processes that are relevant to Alzheimer’s disease.”
Alzheimer’s disease, which involves many genetic pathways, plus environmental and behavioral impacts that can span decades, makes it particularly difficult to find patterns in the data. Not only are the patient data sets highly variable from person to person, the control data sets are as well. To begin to get their arms around the problem, Carter and his team found inspiration from research into another highly complex disease: cancer.
“The first paper showing specific subtypes of breast cancer was published more than a decade ago,” says Carter. “When gene expression experiments were first done on tumors, researchers immediately realized there were four distinct types. Now, everyone carries out breast cancer analyses with this context. While it’s easier with cancer because strong selection drives clear genetic signatures in a tumor, working with the gene expression modules allowed us to stratify the Alzheimer’s disease patient population as well.”
Inflammation or no?
As presented in the paper, the team found two distinct subtypes within the Alzheimer’s patient data. One was strongly associated with inflammatory signatures, the other not. Of note is that the two subtypes were found in about equal numbers in the patient data. The finding is intriguing on multiple levels. For example, it could help explain why behavioral factors, such as regular exercise, may lower risk across a broad population, but some regular exercisers still face an Alzheimer’s disease diagnosis. Could it be that physical activity reduces susceptibility to chronic inflammation, greatly reducing risk of the inflammatory subtype but not affecting the incidence of the non-inflammatory one? More research remains to be done, but stratifying patients has the potential to refine our understanding of the mechanisms underlying environments and behaviors that increase or decrease risk.
In the near term, the insight will help refine the engineering of animal models for Alzheimer’s disease, such as those being developed by the IU/JAX MODEL-AD Center, for which Carter is a Principal Investigator. Because the inflammatory subtype is the dominant signature in the patient population as a whole, the genes implicated in it were the initial focus of the mouse model development effort. By focusing on the genes and pathways that help drive the non-inflammatory subtype versus those involved with inflammation, however, it should be possible to model both subtypes in different mouse models.
Moving forward, there are many possibilities for further investigation.
“We’ve analyzed the transcriptome (all of the messenger RNAs in a cell) pretty comprehensively,” says Milind. “But what about the other components, other ‘omes? The epigenome, proteome, metabolome? If we can integrate all of those data, we will be able to describe the subtypes and their molecular signatures much more completely.”
With that kind of knowledge, there is hope that it could point clinical research in useful directions as well.
“The inflammation is mostly driven by microglia (brain-specific immune cells) in the brain,” says Carter, “but we might detect an echo of the signaling in the general circulation. Right now, we don’t quite know. We might also identify inflammation biomarkers that can be detected with brain imaging. That would be a very clean way to distinguish between the two subtypes in living people recently diagnosed with Alzheimer’s disease. In the big picture, if we can redefine Alzheimer’s disease through the molecular data, it will open up all sorts of possibilities and offer resolutions that are not currently possible now.”
A transformational summer
Unlike some of his summer student peers, Milind didn’t have a science career in sharp focus when he first went to college. In fact, when he went to see Carter speak at a seminar at North Carolina State, he admits it was “complicated,” and he struggled to keep up. But that initial contact and the concept of an immersive research experience in a place where everyone around him was doing the same thing made the JAX program sound very appealing.
“It gave back way more than I’d expected,” he says. “It was transformative. I was particularly inspired by the human element, and how JAX got involved with things like the Longest Day, when all the Alzheimer’s research labs joined together to raise funds for the Alzheimer’s Association.”
Now, with a first-author paper already on his CV, Milind is looking to continue his education in computational biomedical genetics.
“Understanding how the research ends is very important to me. My time at JAX showed me that even if you do basic research, you can have a big effect. You can also stay connected with the people you’re working to benefit, like the Alzheimer’s disease community.”