A panel of 48 CSF proteins—Aβ or tau not among them—distinguished people with Alzheimer's from controls with 94 percent accuracy, according to a study published September 6 in Science Translational Medicine. Adding classic AD biomarkers to this curated panel bumped up accuracy even more. The work was led by Thomas Wingo, Nicholas Seyfried, and Allan Levey at Emory University in Atlanta.
- Panel of 48 CSF proteins performs at least as well as canonical biomarkers in detecting Alzheimer's.
- Combining both predicted decline the best.
- Curiously, CSF Aβ42 and amyloid-PET correlated with different proteomic changes.
The new proteomic panel includes proteins from myriad biological processes involved in AD pathogenesis. It tracked with changes in brain metabolism and volume, and even predicted future cognitive decline. Strikingly, the study illuminated mechanistic differences underlying the standard AD biomarkers. For example, low CSF Aβ42 correlated with changes in synaptic proteins, whereas amyloid-PET scans aligned with proteins in aggregation and metabolism.
“The results are impressive: Not only can the CSF panel of 48 proteins match existing canonical AD biomarkers, but its ability to predict cognitive decline and dementia severity slightly exceeds them,” wrote Gerold Schmitt-Ulms of the University of Toronto.
“This paper continues a series of homeruns by this group,” wrote Russell Swerdlow of Kansas University Medical Center in Kansas City. “Papers like this one illustrate the power of using biomarkers as biologic clues, as opposed simply to answers of what causes AD.”
The data are the latest in a long line of proteomics studies by these Emory scientists. To glean biological insight into disease and discover new biomarkers, they had previously plumbed the depths of brain and CSF proteomes from people across the spectrum of AD (Higginbotham et al., 2020; Feb 2022 news). From these broad analyses, they recently curated a panel of 59 proteins to measure in the CSF of carriers of autosomal-dominant AD mutations (Aug 2023 news on Johnson et al., 2023). Lo and behold, some of these CSF proteins became more or less abundant as early as 30 years before symptoms were expected to start, perhaps even a tad before classic AD biomarkers such as CSF Aβ42 or p-tau217 change. Those findings pointed to different biological processes at play along the path of AD pathogenesis, and nominated new biomarkers and therapeutic targets.
For their current study, first author Rafi Haque and colleagues designed yet another panel, and put it to the test. Would it detect AD? Would it predict cognitive decline? The proteins in this panel were mostly the same as those in the previous study with ADAD mutation carriers, and came from a broad range of synaptic, metabolic, and inflammatory processes that had previously come up for different aspects of AD pathogenesis.
Notably, unlike more cumbersome techniques required for the scientists' larger, unbiased proteomic analyses, this pared-down panel was measured with selective reaction monitoring. This high-throughput technique churns out results in a few minutes, Levey told Alzforum.
The scientists used CSF samples from 706 participants in the AD Neuroimaging Initiative (ADNI), including 220 who were cognitively normal, 376 with a clinical diagnosis of mild cognitive impairment, and 110 with AD dementia. Participants were also grouped into biomarker categories based on measures of CSF Aβ42 and p-tau181. Of the 48 proteins, 18 were significantly elevated, and four reduced, in AD relative to control samples.
The most abundant proteins in AD relative to control CSF were YWHAZ and YWHAB; they by themselves detected AD with 78 percent accuracy. As members of the 14-3-3 family, these proteins have previously been tied to AD and other neurodegenerative diseases. They play a role in many aspects of neuronal function. SMOC1, one of the Aβ-plaque-associated, extracellular matrix proteins collectively known as the matrisome, came in a close second, followed by a host of metabolic proteins. The four proteins that were scarcer in AD samples—VGF, SCG2, NPTXR, and NPTX2—play roles in different aspects of synaptic function.
Collectively, the panel of 48 proteins distinguished AD from control samples with an accuracy of 94 percent, while three standard AD biomarkers—CSF Aβ42, p-tau181, and total tau—together clocked in at 91 percent. Putting both panels together raised the accuracy to 96 percent.
The researchers took stock of numerous associations between the CSF panel and other AD measurements. For example, they found that drops in FDG-PET uptake and hippocampal volume were linked to changes in similar sets of proteins. Oddly, despite the notion that FDG-PET is a measure of brain glucose metabolism, metabolic proteins were not among those tied to this type of scan. Instead, reduction of synaptic proteins and Aβ42, and increased abundance of p-tau181, total tau, and the two 14-3-3 proteins, were most strongly tied to both FDG-PET and hippocampal shrinkage.
An even bigger surprise came when the scientists investigated which proteins on the CSF panel associated with CSF Aβ42 versus amyloid-PET. Considered by most researchers to reflect amyloid deposition in the brain, the two markers are usually 90 percent concordant across studies. However, they were remarkably discordant in their associations with proteins on the 48-protein panel. CSF Aβ42 and florbetapir-PET associated with 21 and 25 protein changes, respectively, but only 11 of those overlapped. Low CSF Aβ42 tied most closely to reduction in synaptic proteins, while florbetapir-PET uptake linked with abundance of SMOC1, 14-3-3, and metabolic proteins.
To the authors, this suggest that although these two measures track closely within a given individual, they may reflect distinct underlying mechanisms. While CSF Aβ42 tracks with changes in synaptic function, amyloid-PET is associated with Aβ aggregation. “Thus, we advocate against using these two biomarkers interchangeably,” the authors wrote.
The situation with tau was the opposite. CSF p-tau181 and total tau are thought to reflect different aspects of AD pathogenesis; even so, both tracked with roughly the same proteomic changes.
To Swerdlow, these findings underscore the importance of exploring the biology of the markers used to define and categorize disease. “For example, data presented here challenge the assumption that plaque absorption of CSF Aβ automatically accounts for reduced AD CSF Aβ levels,” he wrote. “This is consistent with mechanistic studies that show neuronal health alters APP intracellular targeting and Aβ secretion.”
Who Will Get Worse? The abundance of 48 CSF proteins predicted future changes in dementia status, cognitive decline, and hippocampal volume in a research cohort. [Courtesy of Haque et al., Science Translational Medicine, 2023.]
The CSF 48 panel shone the brightest in its association with cognitive impairment and dementia. It outdid a combination of standard AD CSF biomarkers in predicting baseline scores in the Montreal Cognitive Assessment (MoCA) and Clinical Dementia Rating Scale Sum of Boxes (CDR-SB). More importantly to Levey’s mind, it strongly predicted future decline and neurodegeneration. Specifically, the panel outperformed canonical CSF biomarkers in “knowing” who would slip on MoCA and CDR-SB scores, as well as whose hippocampus would shrink among participants who had at least three follow-up assessments (image below). Combining the CSF 48 panel with CSF Aβ42, p-tau181, and total tau further sharpened the predictive value.
Mechanisms Behind Markers. Florbetapir-PET (AV45) and CSF Aβ42 associated with mostly different protein changes. In contrast, CSF p-tau181 and total tau were tied to similar proteomic shifts. [Courtesy of Haque et al., Science Translational Medicine, 2023.]
Predicting cognitive decline is where the panel will likely be most useful, Seyfried told Alzforum. In clinical trials, the CSF 48 panel could help pick out participants who are on the precipice of decline, and/or for tracking biological responses to treatment. Standard CSF biomarkers are great for diagnosis, he said, but not for predicting decline. This is because dozens of other complex processes in the brain contribute to vulnerability versus resilience to disease progression, even between people with the same level of Aβ or tau pathology. “With this panel, we can tap into these other processes that track better with cognitive impairment,” Seyfried said.
James Doecke of CSIRO Brisbane and Colin Masters of the University of Melbourne emphasized the need for better biomarkers. “The challenge has never been greater: advances in therapeutic targeting of Aβ demand better theranostics which will identify responders/nonresponders for aducanumab, lecanemab, and donanemab; and markers of efficacy to provide information on when to stop treatments,” they wrote.
In addition, the 48-protein panel could give memory clinic patients a better idea of what to expect following a diagnosis with standard AD biomarkers, said Jasmeer Chhatwal of Massachusetts General Hospital in Boston. “This is the question my patients most want answered: How fast am I going to decline?” Chhatwal said. He is eager to see it validated in a memory clinic population. Following such validation, this methodology can be scaled up for broader use outside of research cohorts, Levey and Seyfried believe.
Chhatwal previously reported that vascular risk factors are one scourge that colludes with Aβ to hasten cognitive decline. He noted that future studies could assess the extent to which vascular dysfunction, and other modifiable risk factors, contribute to proteomic changes on this panel (May 2018 news).—Jessica Shugart
- Proteomics Highlight Alzheimer’s Changes in Matrisome, MAPK Signaling
- Proteins in Biofluids Foreshadow Dementia by 30 Years
- Bad Synergy—Together, Vascular Problems and Aβ Hasten Memory Slippage
- Higginbotham L, Ping L, Dammer EB, Duong DM, Zhou M, Gearing M, Hurst C, Glass JD, Factor SA, Johnson EC, Hajjar I, Lah JJ, Levey AI, Seyfried NT. Integrated proteomics reveals brain-based cerebrospinal fluid biomarkers in asymptomatic and symptomatic Alzheimer's disease. Sci Adv. 2020 Oct;6(43) Print 2020 Oct PubMed.
- Johnson EC, Bian S, Haque RU, Carter EK, Watson CM, Gordon BA, Ping L, Duong DM, Epstein MP, McDade E, Barthélemy NR, Karch CM, Xiong C, Cruchaga C, Perrin RJ, Wingo AP, Wingo TS, Chhatwal JP, Day GS, Noble JM, Berman SB, Martins R, Graff-Radford NR, Schofield PR, Ikeuchi T, Mori H, Levin J, Farlow M, Lah JJ, Haass C, Jucker M, Morris JC, Benzinger TL, Roberts BR, Bateman RJ, Fagan AM, Seyfried NT, Levey AI, Dominantly Inherited Alzheimer Network. Cerebrospinal fluid proteomics define the natural history of autosomal dominant Alzheimer's disease. Nat Med. 2023 Aug;29(8):1979-1988. Epub 2023 Aug 7 PubMed.
No Available Further Reading
- Haque R, Watson CM, Liu J, Carter EK, Duong DM, Lah JJ, Wingo AP, Roberts BR, Johnson EC, Saykin AJ, Shaw LM, Seyfried NT, Wingo TS, Levey AI. A protein panel in cerebrospinal fluid for diagnostic and predictive assessment of Alzheimer's disease. Sci Transl Med. 2023 Sep 6;15(712):eadg4122. PubMed.