After 14 years in Miami, HAI is on the move again. This January conference started in chilly Boston and moved around the north for some time, meeting in Chicago, Seattle, and Toronto, before settling in a warmer clime in 2011. For 2025, HAI pulled up stakes again and landed in San Juan, Puerto Rico. There, 419 attendees from 16 countries shared data about how to line up blood biomarkers with PET data, how locus coeruleus connectivity might influence tangle spread, and what the field is learning from its growing focus on studying non-white populations.
At Human Amyloid Imaging Conference, Plasma Tau Was the Star
The hunt for blood-based biomarkers is continuing apace. At the Human Amyloid Imaging (HAI) meeting, held January 15-17, researchers focused on new techniques—all based on tau and blood, conference title notwithstanding—for identifying whether a person has tau accumulation in his or her brain. Even as this blood-PET work advances, uncertainty around which tau PET tracers are best is still dogging validation efforts.
The conference, which left Miami after 14 years there and convened in San Juan, Puerto Rico, drew 419 attendees from 16 countries. It featured technical sessions on data harmonization, on ways to deal with the heterogeneity that makes Alzheimer’s and other tauopathies so difficult to image, and on evolving methods of measuring neuropathology, among other topics.
This year’s breakout? Several dozen of the meeting’s 222 abstracts dealt with relating blood-based biomarkers to brain-imaging data in hopes of defining the most informative peripheral indicator of what AD pathology might be afoot in the central nervous system. This, the thinking goes, will lead to cheaper blood tests that can become widely and routinely available, while pricier PET scans can remain useful in clinical trials and specialty applications such as uncertain diagnoses.
Much discussion was sparked by Gemma Salvadó’s talk. Salvadó, Lund University, Sweden, presented a study attempting to determine the stage of a person’s Alzheimer’s disease by analyzing the levels of multiple tau plasma biomarkers simultaneously. Using mass spectrometry, Salvadó and colleagues measured, cross-sectionally, the baseline concentrations of 12 phosphorylated and non-phosphorylated tau species in plasma samples from 549 people in the Swedish BioFINDER cohort. They ranked each marker based on how well its level reflected progression of the person’s amyloid PET, tau PET in the medial temporal lobe and neocortex, cortical thickness, as well as their PACC and MMSE scores taken over the course of four years of study.
The team ultimately settled on three markers that best correlated with disease progression across the population. Two of them, p-tau217r and p-tau205r, represent ratios between phosphorylated and non-phosphorylated species. The former became positive with the onset of amyloid pathology, the latter with MCI symptoms. The third marker, an alternatively spliced isoform called 0N-tau, increased later still, when MCI progressed to dementia.
Blood Groups. Plasma p-tau217r, p-tau205r and 0N-tau scores change as people progressed from cognitive normality into amyloid positivity, MCI, then dementia. A new plasma staging model (below) proposes four phases of disease based on the amount or ratio of each marker. (Courtesy of Lund University.)
From these data, the researchers built a four-stage model (see image above) and validated it in 140 people from the TRIAD cohort based in Montreal. In both datasets, the model revealed four distinct stages of clinical and PET pathology that corresponded with plasma tau levels. People at higher stages had steeper tau accumulation and cognitive decline over the study period. It “shows that you can use changes in blood markers to accurately stage disease,” Salvadó said.
“It’s a great effort,” Bruna Bellaver, University of Pittsburgh, who chaired the session, told Alzforum. Bernard Hanseeuw, Cliniques Universitaires Saint-Luc, Brussels, who presented an alternative approach to tau plasma biomarker testing (see below), told Alzforum that Salvadó’s study convincingly showed that plasma can recapitulate PET stages of AD.
That said, the model was mostly verified in clear-cut stages rather than with the more difficult intermediate cases where PET and plasma results only partially overlap. “The percentage of mismatches makes it difficult to use in clinical practice where individual precision is needed, although it could work to enrich clinical trials,” Hanseeuw wrote.
Salvadó readily acknowledged that this plasma staging system is at the proof-of-concept stage, and far from ready for clinical use. “We need to validate it in a bigger and more diverse population,” she told Alzforum. At this point, it is difficult to even determine how precisely each biomarker reflects tau in the brain, she said. “We do not know whether the tau PET stages can be used as a gold standard yet.”
Although Bellaver considers the data promising, she was concerned about the finding that people in stage 1 had tau tangles in the neocortex. Neocortical tangles usually appear later in disease progression, years after p-tau217.
Salvadó said her group will next refine the model by incorporating additional biomarkers such as MTBR-243, which indicates tangles in the brain and can be measured both in CSF and plasma (Aug 2024 conference news). They also plan to determine how other co-pathologies, such as cerebrovascular disease and α-synuclein deposits, could affect the levels of various tau species in plasma.
An alternative approach, Hanseeuw said, is to avoid the question of biomarkers entirely by measuring tau’s aggregation capacity rather than its phosphorylation. In his talk, Hanseeuw presented early data from VeraBIND, a new assay by the Austin, Texas-based diagnostics company Veravas. The assay detects tau species with hyperphosphorylation at any location along the length of the protein. It involves multiple antibodies linked to beads and combines the tau isolated from a proband’s plasma sample with regular tau protein. The trick? By way of a luminescent readout, the assay reveals whether aggregation can occur—in other words, whether the donor’s blood contains species of tau that are actively able to aggregate “normal tau.”
Hanseeuw told Alzforum that this might be more clinically useful than simply knowing whether blood contains more of a certain form of tau. The approach is also free of known confounds such as chronic kidney disease and other comorbid conditions influencing p-tau217 plasma levels (Bornhorst, et al., 2025). Furthermore, only around 6 percent of people with amyloid but no tangles as measured by PET go on to develop cognitive impairment within the next three to five years (Ossenkoppele et al., 2022). “It’s important for [people] to know whether they have a significant short-term risk, and p-tau217 doesn’t give that information,” Hanseeuw said.
Hanseeuw and colleagues tested the new system in plasma samples from 58 cognitively unimpaired and 33 impaired people, along with plasma p-tau217 levels and MK6420 Tau PET scans from them. Compared with p-tau217 assays, the VeraBIND assay was less specific and less sensitive at predicting a person’s amyloid status, but it outperformed p-tau217 at predicting tau status. The VeraBIND assay was also more accurate than p-tau217 at detecting those cases who had tangles but no amyloid.
Bye-Bye p-Tau? The phosphorylation-agnostic VeraBIND assay predicts tangles in the brain more accurately than does plasma p-tau217, which is closely associated with amyloid status. (Courtesy of Bernard Hanseeuw.)
Bellaver called the VeraBIND data promising, despite the small sample size. “They saw more specificity for tau than for amyloid. That’s what we’re looking for,” she told Alzforum.
Suzanne Schindler, Washington University, St. Louis, agrees there’s a great need to keep improving plasma assays since few research centers—and even fewer clinics—can perform tau PET. “We know people with really high p-tau217 are likely to have high tau PET, but the prediction isn’t great,” she said. If a person’s plasma levels are already high, she said, it may not be worth enrolling them in a trial or starting them on a therapy that works best at earlier stages of pathology.
Right now, the quest for the best plasma biomarkers is hobbled by lingering uncertainty about how best to quantify tangles in the brain, which, after all, is the standard blood tests are supposed to be held against. Several tau PET tracers are competing in this regard, and the differences between them are significant (Jul 2024 news). Few direct comparison studies exist.
At HAI, Pamela Ferreira, University of Pittsburgh, showed results from one. Ferreira compared how well four tau PET tracers correlated with plasma p-tau217 using data from HEAD. In this head-to-head study, people at different disease stages or healthy controls receive PET scans with two to four different tracers. Ferreira and colleagues analyzed MK6240, flortaucipir, PI2620, and RO948 uptake across disease stages. At HAI, she reported that MK6240 most closely tracked with plasma p-tau217, especially during early disease stages, although all four tracers showed statistically significant associations.
6240 Wins This One. In HEAD, plasma p-tau217 levels correlate more closely with reads from MK6240 than three other tau PET tracers, especially during early tangle deposition. (Courtesy of Bruna Bellaver.)
“It looks like MK6420 is better across the board,” Schindler said of the Merck compound, which is still in clinical trials but is used by some research groups. Her group and many others primarily use flortaucipir, since it is the only FDA-approved tau tracer thus far, despite its known off-target binding in the hippocampus. Ferreira’s data, Schindler said, “is telling us that we're not doing the optimal test.”
When Tau Wanders Off, Subcortical Axon Firing Goes Mum
When it comes to tau tangles, cortical deposits receive the lion’s share of scientists’ attention, but it is the tiny subcortical regions where tangles originate that might set the course for Alzheimer's. New data presented at the Human Amyloid Imaging conference, held January 15-17 in San Juan, Puerto Rico, suggests as much. Researchers say that connectivity among the white-matter tracts linking the locus coeruleus to other brain regions determines how well these systems can compensate for one another’s functions and offset the effects of neuronal death in the early stages of Alzheimer's disease.
The locus coeruleus—with its two lobes the size of a grain of rice in our brainstems—serves as a hub that drives brain-wide functions such as the sleep-wake cycle. It’s also considered the birthplace of tau pathology. Autopsy data suggests that tangles accumulate in the LC before they fan out into the medial temporal lobe (Ehrenberg et al., 2017). This spreading process seems to be a one-way street: Neurodegeneration in the LC accelerates tau spread, but high levels of tangles in the medial temporal lobe don’t disrupt the LC (May 2024 news).
Studying how tau in the LC affects pathology and cognitive function in living people, however, has proven extraordinarily challenging. Not only is the LC too small to see on most scanners, but all available tau PET tracers have off-target binding to neuromelanin and ferromagnetic materials throughout this region. “Everything we do in the locus coeruleus is a proxy,” Heidi Jacobs of Massachusetts General Hospital, Boston, told Alzforum.
At HAI, Jacobs and colleagues presented alternative methods for measuring how tau affects subcortical networks, particularly these regions’ ability to help one another out when neurons begin to die during Alzheimer's. In one poster, Yuliya Patsyuk, Maastricht University, Netherlands, used 7T diffusion MRI scanning in 47 cognitively unimpaired people to measure microstructural integrity in the tracts connecting the LC to other brain regions.
Spot the Dot? 7T diffusion MRI scans show that the fiber tract (red) connecting the locus coeruleus (arrow to blue) to the entorhinal cortex (yellow) is among first to break down as plasma tau levels increase. (Courtesy of the Jacobs lab.)
Patsyuk found that people with worse diffusion within the tract connecting the LC to the entorhinal cortex (EC), and less structural integrity, tended to have higher plasma levels of phosphorylated tau species and of the astroglial activation marker GFAP. The result implies a link between white matter breakdown, tau aggregation and neuroinflammation.
The pattern seemed specific to the LC-EC tract. The fibers connecting the LC and prefrontal cortex did not show the same associations between microstructural integrity and pTau blood biomarker levels. This suggests that the LC-EC tract is particularly vulnerable to break down even before other pathologies appear and supports the idea that tau pathology moves via anatomical connections (Feb 2018 news).
Similarly, Elouise Koops, MGH, found worse LC-EC tract degradation and higher levels of entorhinal tangles in women from the COLBOS cohort of people who carry the PSEN1 Paisa mutation. Although none of the participants had yet developed dementia, mutation carriers whose LC-EC tracts contained scant fibers did worse on memory tests compared to those with denser connections. Members of the kindred who don’t carry the mutation showed no link between tract density and performance.
Tau Versus Fiber Tracts. Among asymptomatic PSEN1 E280A mutation carriers (dark blue), those with more tangles in the entorhinal cortex had thinner fibers connecting the LC and EC . This was not true for but not non-carriers (light blue). (Courtesy of the Jacobs lab.)
Patsyuk does not know whether white-matter degradation follows tau accumulation or whether a breakdown in fiber microstructure allows tau to spread more easily. For that, she told Alzforum, the team will need longitudinal data, which they hope to acquire soon.
Because the LC and other subcortical systems, such as the hypothalamus and Raphe nuclei, are so highly interconnected with one another, and with the rest of the brain for control of brain-wide functions, Jacobs believes they can likely compensate for each other to a degree when one begins to lose neurons in the early stages of Alzheimer's. For instance, the LC’s ability to regulate arousal via norepinephrine release might be able to mask neuronal loss in the hypothalamus that would otherwise affect sleep. But once the LC begins losing neurons, too, the functional loss becomes more severe, Jacobs said.
This compensatory ability may affect cognition too. Lukas Heinrich, MGH, used fMRI to record functional connectivity between the LC and the medial temporal lobe in 128 people from the HABS cohort while they were performing a memory test. Those with higher cognitive reserve—calculated as a combination of education level, IQ, and occupation—at baseline outperformed those with lower reserve regardless of whether they had brain amyloid in the neocortex.
People with high connectivity between the LC and medial temporal lobe continued performing well over the course of about seven years even if they accumulated amyloid elsewhere in the brain, suggesting these strong connections could delay the impact of plaques on cognition. Emma Wiklund, MGH, found similar patterns when she studied the LC’s glucose metabolism using FDG-PET. People with more metabolic activity in the LC, she found, performed more consistently on cognitive tests over the course of seven years regardless of their overall brain levels of plaques or tangles.
Saving Grace. Strong functional connectivity between the LC and MTL slowed the rate of cognitive decline in people with high amyloid loads (measured by PiB-PET) and high cognitive reserve scores (CR). (Courtesy of the Jacobs lab.)
Melissa Murray, Mayo Clinic, Jacksonville, Florida, said the results make sense to her. Practicing a skill builds connections between brain regions, hence the idea that education builds a strong white matter “highway” between the locus coeruleus and other brain regions could help explain why it is protective in Alzheimer's disease. Murray was impressed by the methods the Jacobs lab has developed to study the LC, despite the challenges of imaging this area. “A lot of people just kind of move past it because it's a minefield,” she told Alzforum.
Jacobs acknowledged that no one has yet been able to show definitive longitudinal data that would prove whether and how tau spreads from the LC to the entorhinal cortex. If the hypothesis holds true, she said, “maybe we can find ways to halt tau propagation and delay disease progression.”
Lea Grinberg, University of California, San Francisco, told Alzforum she is glad scientists are paying more attention to the locus coeruleus. “People are starting to link the dots,” she said. Evidence from Grinberg’s lab suggests that some neurons and nuclei are more vulnerable to tau and other pathologies than are their neighbors, although why is a mystery. “If we can understand this, we can create neuroprotective therapeutics,” she said. Examples include atomoxetine, an ADHD treatment targeting norepinephrine transporters that has improved brain metabolism in an Alzheimer's repurposing trial. A recent proteomics study suggested atomoxetine might boost norepinephrine in the LC (Sep 2024 news).
Jacobs’ group is exploring a different therapeutic approach. They are testing whether stimulating the LC can slow cognitive decline. In a recently started clinical trial called WALLe, clinicians stimulate the LC and nearby nuclei by attaching electrodes to a participant’s ear, near the spot where the vagus nerve terminates, and apply electrical current to it. This is an application of transcutaneous auricular vagus never stimulation (Kim et al., 2022).
Hear the Shocker? The WALLe trial is testing in cognitively normal volunteers whether stimulating the vagus via an electrode placed on the outer ear will trigger norepinephrine release in the locus coeruleus and, in this way, benefit memory and neuroinflammation. (Courtesy of the Jacobs lab.)
So far, the trial has enrolled 98 cognitively unimpaired older people who receive this stimulation for about 20 minutes per day over the course of two weeks. The researchers are testing whether this approach affects performance on a face-name memory test, a cognitive composite score, and markers of inflammation. They will monitor changes in brain structure via 7T MRI over the course of four months.
An older, more invasive approach, deep-brain stimulation, is effective in Parkinson’s disease (Jan 2013 news). Can tickling the vagus nerve in this way help cognition? Stay tuned.—Sara Reardon
Sara Reardon is a freelance writer in Bozeman, Montana.
Studying Diverse Populations May Require New Biomarkers
Although efforts to identify biomarkers for Alzheimer's disease are advancing rapidly, many still struggle to obtain information from populations other than non-Hispanic white people. At the Human Amyloid Imaging (HAI) conference, held January 15-17 in San Juan, Puerto Rico, researchers presented new data suggesting that clinical staging frameworks and standard biomarker sets might not accurately reflect disease progression in all racial and ethnic groups.
Black and Hispanic people are more likely to develop all-cause dementia than their white counterparts, but have historically made up less than 5 percent of clinical trial participants. In his keynote talk at HAI, Sid O’Bryant, University of North Texas Health Science Center, Fort Worth, said this disparity partly stems from recruitment challenges and unequal access to healthcare facilities. But it is changing, and emerging data indeed points to some biological differences in the way dementia develops in different populations.
One recent analysis from the screening and recruitment data of the ongoing AHEAD lecanemab secondary prevention trial found that black people were less likely to have brain amyloid than non-Hispanic white people with the same amount of cognitive impairment. This means that they could not be included in AHEAD, which, as an amyloid-removal trial, had a minimum amyloid threshold as an inclusion criteria (Nov 2024 news).
Such findings challenge scientists’ general understanding of the etiology and features of Alzheimer's disease and related dementias, O’Bryant believes. For instance, most of the data used to design the amyloid-tau-neurodegeneration classification (ATN) framework came from the Alzheimer's Disease Neuroimaging Study, which included more than 90 percent white people in its earliest iteration, and from the Mayo Clinic Study of Aging, which started in 2004 in Minnesota’s Olmstead county and is 97 percent white (Lim et al., 2023; Schwarz et al., 2025). “How these biomarkers look across a broader range of populations is simply unknown,” O’Bryant said.
More recent cohort extension studies, such as the Health and Aging Brain Study’s Health Disparities (HABS-HD) project that O’Bryant directs, have statistical power to begin examining disease progression in diverse populations (Nov 2023 conference news). In one presentation at HAI, Annie Cohen, University of Pittsburgh School of Medicine, showed one of the first attempts to quantify these differences on a large scale using data from the HABS-HD cohort. Her group collected MRIs, tau, and amyloid PET scans from 1,082 non-Hispanic white, 1,079 Mexican American and 605 non-Hispanic black participants who had varying levels of cognitive impairment.
The researchers modelled how Aβ, tau, and neurodegeneration progressed over time in each of these groups. In essence, the stages of white HABS-HD participants’ disease fit the ATN framework—amyloid PET became positive before tau PET or before MRI showed significant hippocampal neurodegeneration. In contrast, in both black and Hispanic cohorts, all three pathologies became detectable at the same time. They also progressed at around the same rate, whereas in whites the three pathologies progressed at different rates. “As we get into studying the individuals who've been historically excluded from research for a variety of reasons, we have to be more careful in interpreting how well [the ATN] framework will work,” Cohen told Alzforum.
Graded on a Curve. Amyloid accumulation (blue) precedes other pathologies in non-Hispanic whites (left columns). Plaques, tangles (yellow), hippocampal atrophy (pink), and white-matter damage (purple) occur simultaneously in Mexican-American (center) and non-Hispanic black (right) populations. Left set of 6 graphs represents ApoE4 noncarriers; right set, carriers. (Courtesy of Annie Cohen.)
Reisa Sperling, also at Harvard, who co-directs the Harvard Aging Brain Study, finds the Texas HABS-HD data compelling in showing that amyloid can’t be the only target when studying diverse populations with dementia. “We have to go after multiple processes that contribute to cognitive decline,” she told Alzforum. Future trials, she said, should study combinations of factors across diverse populations. An example is the Alzheimer Plasma Extension observational study (APEX) which studies biomarkers, cognition and the social determinants of health in people who didn’t qualify for the AHEAD lecanemab trial. Sperling said that 54 percent of the people in APEX are racial minorities.
Cohen’s group also confirmed previous findings that while carrying the ApoE4 allele significantly increased Alzheimer's risk in non-Hispanic white people, the allele had little effect among black and Hispanic people. “It may be that there are other social determinants of health, other factors that increase risk so much it washes out ApoE4,” she told Alzforum.
Lea Grinberg, University of California, San Francisco, said that while the HABS-HD study doesn’t invalidate ATN staging, it does suggest that broader criteria might be needed. Suzanne Schindler, Washington University, St. Louis, told Alzforum that clinicians might conflate Alzheimer's with non-AD dementias that are more common in non-white populations. “We're applying an Alzheimer's framework to people who don't have Alzheimer's disease, and that's why things aren't lining up,” she said. Emerging data on proteomics may help improve such diagnoses, she said.
Cohen encouraged researchers to look at biomarker patterns within racial minority groups with dementia, rather than only trying to compare diverse cohorts against white cohorts.
In one poster at HAI, for example, Lindsey Kuchenbecker and senior author Minerva Carrasquillo, Mayo Clinic, Jacksonville, Florida, analyzed plasma biomarkers in the Florida Consortium of African American Alzheimer’s Disease Studies (FCA3DS). The purpose, Carrasquillo said, was to find proteins that are not necessarily associated with tau or amyloid aggregation, but could identify disease processes such as immune response or vascular disease that might be particularly relevant in non-white populations. “We have to do a better job at studying the full spectrum of risk factors in these underrepresented populations so that we can have a better understanding of the biology underlying the disease,” she told Alzforum.
Using the SomaScan 7k assay, which measures levels of 7,300 plasma proteins, Kuchenbecker and colleagues compared the blood proteomes of 145 cognitively unimpaired people with 183 who had been clinically diagnosed with Alzheimer's dementia. No PET or fluid testing was used to confirm their diagnoses. A machine learning algorithm bubbled up 32 unique proteins that distinguished the two groups.
Fourteen of these had been previously linked with Alzheimer's pathology in some way. Some were linked to amyloid or tau pathology, including the protein-folding markers SFRP1 and SPON1. Others were associated with neurodegeneration, such as synaptic dysfunction marker NPTXR, the inflammation-related protein IGFBP-2, and a synaptic vesicle protein called CPLX2 that was spotted in plasma seven years prior to symptom onset in autosomal-dominant AD.
Eighteen others represented novel potential markers. They include the cytokine CCL25, which has been associated with poorer performance on cognitive tests among MCI patients (see also Ferguson et al., 2020), the chemokine CCL22, which is made by dendritic cells and macrophages and has been reported in several neuroinflammatory conditions, and TAGLN, which affects cell migration and could be associated with cerebrovascular dysfunction (e.g., Gomez et al., 2025).
Blood Ties. Plasma samples from a black cohort revealed 98 proteins that increased in clinically diagnosed AD and 22 that decreased. Some had been previously associated with neuroinflammation and vascular degeneration, others had not. (Kuchenbecker et al., bioRxiv 2024).
The researchers replicated their findings in the Accelerating Medicines Project’s AD Diverse Cohorts’ postmortem brain proteomics dataset—an extension of the AMP-AD brain collection project. In these proteomes, the 32 proteins Carrasquillo’s group found in the FCA3DS plasma samples distinguished autopsy-confirmed Alzheimer's cases from controls with 91percent accuracy.
Kuchenbecker acknowledged that these disease end-stage brain proteins might not have been present in the plasma since blood samples were not available, but said the predictive power of the FCA3DS’s initial blood dataset’s suggests these proteins can distinguish between AD and controls, not just dementia cases diagnosed as AD. The scientists uploaded a manuscript to bioRxiv last July (Kuchenbecker et al., 2024). They are now looking for these hits in other ADRC cohorts to see whether they occur in non-Hispanic white and other diverse populations
Schindler would like to see plasma p-tau217 analyzed in this cohort. It is not included in the SomaScan 7k assay used in the study. “It's great to look at stuff in addition to p-tau217, but it's so important that leaving it out just doesn't make sense anymore,” she said.
Nevertheless, Schindler is excited to see statistically valid findings emerging from HABS-HD and other diverse cohorts. “For a while I was skeptical that there would be racial and ethnic differences because Alzheimer's is the same no matter the color of your skin,” she said. That quantifiable differences exist, she said, “has an important impact on diagnostics and treatments.”
“These studies highlight the importance of studying representative cohorts that mirror the racial and ethnic diversity of older adults at risk for cognitive impairment. The findings strongly suggest that late-onset dementia may involve distinct biological pathways and upstream drivers, depending on racial and ethnic background,” wrote Gil Rabinovici, University of California, San Francisco. “Better understanding the social determinants that drive these findings is therefore a critical element of our overall strategy to prevent dementia.”—Sara Reardon
Sara Reardon is a freelance writer in Bozeman, Montana.
Comments
No Available Comments
Make a Comment
To make a comment you must login or register.