A diagnosis of Alzheimer’s disease does not necessarily explain everything happening in a person’s brain. Amyloid plaques accumulate between neurons, while tau tangles form inside them. But these changes can coexist with vascular damage and other abnormal protein deposits. When multiple disease processes contribute to a person’s dementia, the condition is known as mixed dementia.
People with similar symptoms may have different combinations of underlying diseases, potentially influencing how they respond to a drug. “Typically, more than just Alzheimer’s disease is contributing to their dementia,” said Sterling Johnson, a dementia researcher at the University of Wisconsin–Madison.
Johnson and Sarah Biber, an Associate Professor of Neurology at Washington University in St. Louis, are tackling that complexity through a new proteomics initiative co-led with Indiana University’s Tatiana Foroud, Professor in the Department of Medical and Molecular Genetics.. The project, which is part of the Consortium for Clarity in ADRD Research Through Imaging (CLARiTI), plans to analyze approximately 21,000 plasma samples collected over time from 10,000 participants in the US Alzheimer’s Disease Research Centers program. The researchers hope to identify patterns of blood proteins that help distinguish the different diseases contributing to cognitive impairment.
DDN spoke with Biber and Johnson at the Alzheimer’s Association International Conference in London about the biology of mixed dementia, how proteomics could improve clinical trials, and whether longitudinal biomarkers could help predict whether and when cognitive symptoms will develop.
What is the central scientific hypothesis of this new initiative, and what are you hoping to achieve?
Johnson: We are very interested in detecting the multiple pathologies that ultimately contribute to cognitive impairment in someone presumed to have Alzheimer’s disease. This study is about identifying proteins in blood that may help us understand the different neurodegenerative diseases a person may be simultaneously experiencing and how these combinations can have different presentations, prognoses, and relevance for treatment.
This is something we have been thinking about for a long time, but the tools were not previously available. Now we have NULISAseq Neuro 220, a panel containing assays that are highly relevant not only to Alzheimer’s disease, but to other neurodegenerative pathologies as well. The plasma samples we will use provide a particularly strong way to road-test the panel because they represent the heterogeneity we see in Alzheimer’s disease and related disorders. We have to disentangle that heterogeneity and see whether we can bring some clarity to it.
Biber: What is especially valuable about this cohort is that it includes thousands of participants from across the National Institute of Aging’s Alzheimer’s Disease Research Centers (ADRC) program who have been followed for years. We have blood samples collected over time for these individuals, which is what we’ll use for proteomics profiling, but we also have a wide variety of types of data on these people, all linked at the participant level over time. This gives us a much more holistic view of each participant’s clinical picture and how it changes over time. For example, annual detailed clinical assessments are available on each of them and, in many cases, standardized longitudinal imaging as well. About 60 percent have also agreed to brain donation which will allow us to integrate critical neuropathological data.
Our ultimate goal is to understand and disentangle the mixture of different etiologies contributing to an individual’s dementia risk and symptoms, and neuropathology provides the “ground truth” we need to piece together how data collected during life might be used to detect a broad range of pure and mixed etiologies.
All of this together will allow us to build multimodal AI models that ask: What combinations of clinical, imaging, proteomic, genetic, and other data indicate the presence of these different underlying etiologies? If that works, we’d have new biomarker signatures that allow us to detect these conditions earlier and get a better sense of how someone is likely to progress.
What makes this cohort different from other large biomarker cohorts and how could it help future clinical trials?
Biber: We have the full spectrum of both relatively pure and mixed etiologies. The cohort is also unusual in the breadth of the biological heterogeneity it represents, the depth of its longitudinal multimodal data, and the proportion of participants who have agreed to brain donation and neuropathological examination.
Because these individuals are followed for many years through the ADRC program, we may be able to observe them before the onset of dementia and then over the course of disease development. That gives us critical insight into the changes associated with the transition from normal cognition to mild cognitive impairment and dementia. But really, it is the combination of all these things that makes the resource so valuable: longitudinal blood, detailed clinical phenotyping, imaging, genetics, and eventually brain neuropathology.
The implications for clinical trials could be substantial. One factor that complicates trials is that we often do not know which participants have relatively pure Alzheimer’s disease and which have Alzheimer’s together with other neurodegenerative diseases. If this project is successful, it could lead to assays that help trial sponsors prescreen participants and identify the people whose underlying pathologies make them the best fit for a particular trial. It could also help researchers understand how the presence of additional etiologies affects response to a drug or other treatment. In the future, this is how we get to personalized medicine where treatments are tailored to the particular mix of conditions each patient brings to the clinic.
You are also planning an open data challenge. What kinds of questions would you put to outside teams?
Biber: We are in the process of bringing together partners for a large, open data challenge. We hope that could include Sage Bionetworks and the DREAM Challenge platform, as well as other organizations.
We want to make the data available and give teams concrete prediction problems. For example, we might ask things like: Based on these multimodal data, can you predict who will transition from normal cognition to mild cognitive impairment during the next three years? Can you estimate the relative contributions of different etiologies in an individual?
We would invite teams from around the world to develop and compare different methods. We already expect considerable interest in the dataset because it is such an unusual resource, but a challenge could bring even more researchers to it and help generate insights more rapidly.
One of the markers you will measure is eMTBR-tau243. Why did you choose this marker and what might it add beyond p-tau217?
Johnson: What makes eMTBR-tau243 particularly interesting is that it appears to give us different information from p-tau217. Plasma p-tau217 is extremely useful for identifying whether Alzheimer’s disease biology is present, particularly amyloid pathology. But eMTBR-tau243 appears to be much more directly related to the burden of fibrillar tau tangles and therefore may tell us where someone is biologically within the course of the disease.
This thinking is based on my own early experience with this panel and this assay in people who are well characterized with amyloid and tau PET, which is consistent with recent work from others showing a clear relationship between eMTBR-tau243 and tau PET, including an ability to identify people with more extensive fibrillar tau burden. That is why the combination is so interesting to us. One hypothesis we want to study is whether people with elevated p-tau217 but normal or relatively low eMTBR-tau243 represent a particularly attractive stage for amyloid-lowering treatment — Alzheimer’s biology is present, but extensive fibrillar tau has not yet developed. We do not yet know whether that predicts treatment response, so that remains a hypothesis to test. But it illustrates why biological staging may become just as important as determining whether someone is biomarker positive.
You have argued that tau-directed trials may need a “Goldilocks” population — people with detectable but still mild-to-moderate tau — rather than no measurable tau or advanced disease. What is the biological rationale for that therapeutic window, and how could it be defined reproducibly across PET tracers, brain regions, and staging systems?
Johnson: The biological rationale is that you want enough tau pathology to know that the therapeutic target is present, but not so much that the disease has progressed to extensive, potentially irreversible neurodegeneration. If there is very little tau, there may be too little target or too little dynamic range to demonstrate an effect; if tau burden is already advanced, intervening on tau may simply be too late.
Exactly where that therapeutic window lies depends on the question and context. This is still an active area of research. Defining it reproducibly will require harmonizing measurements across PET tracers and brain regions and anchoring those measurements to biological stage and, ultimately, treatment response. We do not yet know where those thresholds should be.
Biber: This project may help us learn something important about it because, although a large proportion of the cohort is currently cognitively unimpaired, that doesn’t necessarily mean they have no Alzheimer’s pathology in their brains. The ADRCs are actively recruiting people who are cognitively unimpaired. Some will already have pathology, some will develop it later, and others may never develop it. Having all those groups represented should help us understand the biological transitions and the stages at which an intervention may have the greatest effect.
At AAIC this year, where do you see unresolved debate or disagreement in the biomarker field?
Johnson: One major focus now is using biomarkers to predict when a person will become symptomatic. The question is shifting from whether symptoms will occur to when they will occur. But how confident can we really be when that prediction is based on a single biomarker? I think the field is now asking what the right panel or combination of markers is for predicting the timing of symptoms.
Biber: There is clearly continuing controversy over biomarkers and how much they can tell us. There is also continuing debate over whether blood biomarkers alone will be sufficient. Do we still need PET? Do we need other modalities, or can blood biomarkers eventually substitute for some of them?
I hope we reach a point where blood biomarkers are sufficient for many of these purposes, and I think they may be, but that debate is not yet settled.
