Genetic Factors Shape [18F]FDG PET Metabolic Patterns in Lewy Body Dementia

A genetic twist may help explain why dementia with Lewy bodies can look so different from one patient to the next—and why brain scans sometimes mistake it for Alzheimer’s disease. In a study of 43 people with dementia with Lewy bodies (DLB), researchers found that two well-known genetic factors, the GBA1 mutation and the APOE ε4 allele, were associated with distinct patterns of glucose metabolism on brain PET scans. The results suggest that a patient’s inherited biology may influence not only the symptoms of DLB but also the way the disease appears on one of its most important imaging tests. The findings, published in the European Journal of Nuclear Medicine and Molecular Imaging, point toward a future in which genetic information and quantitative brain imaging are combined to make dementia diagnosis more precise.

DLB is the second most common neurodegenerative dementia after Alzheimer’s disease, but it is often much harder to identify confidently. Patients may experience fluctuating cognition, visual hallucinations, Parkinson-like movement problems and rapid-eye-movement sleep behavior disorder, in which they physically act out dreams. The disorder is driven primarily by the accumulation of abnormal alpha-synuclein, the protein associated with Lewy bodies, yet many patients also have Alzheimer’s-related amyloid and tau pathology. Autopsy studies suggest that Alzheimer’s co-pathology is present in as many as three-quarters of people with DLB. This biological mixture can blur the boundary between diseases, producing symptoms and biomarkers that do not fit neatly into a single diagnostic category.

The new study focused on two genetic influences that may help shape this biological diversity. GBA1 provides instructions for an enzyme involved in lysosomal function, the cellular recycling system that breaks down waste materials. Certain GBA1 variants impair this process and are associated with the accumulation and spread of alpha-synuclein in Parkinson’s disease and DLB. The mutations are particularly common among Ashkenazi Jewish populations, the group examined in this study. APOE ε4, by contrast, is the best-established common genetic risk factor for Alzheimer’s disease. It is linked to amyloid deposition, tau pathology and more aggressive cognitive decline, and is also frequent among people with DLB. Three participants carried both genetic risk factors, allowing the researchers to examine how they might contribute to overlapping but different metabolic signatures.

To see the brain’s energy patterns, the team used [18F]fluorodeoxyglucose positron-emission tomography, or FDG-PET. The radioactive glucose analogue is taken up by metabolically active brain cells, enabling PET to reveal regions in which glucose use is reduced. Rather than relying only on visual inspection, the researchers analyzed the scans in several complementary ways. They measured activity in anatomically defined brain regions, calculated the “cingulate island sign,” compared each scan with established Alzheimer’s-, Parkinson’s- and DLB-related metabolic networks, and applied a machine-learning model trained to distinguish healthy aging from Alzheimer’s disease, Parkinson’s disease and DLB. The study included 18 healthy controls and 17 first-degree relatives of patients, with scans collected between August 2019 and December 2023.

The clearest effect of APOE ε4 appeared in the cingulate island sign, a characteristic feature of DLB imaging. In typical DLB, the occipital cortex at the back of the brain shows reduced glucose metabolism, while the posterior cingulate cortex is relatively preserved compared with nearby regions such as the precuneus and cuneus. This “island” of preserved activity can help distinguish DLB from Alzheimer’s disease, where posterior cingulate metabolism is often more broadly impaired. In the study, DLB patients without APOE ε4 had a substantially stronger cingulate island sign than carriers: the median standardized score was 1.35 compared with 0.23. The difference remained statistically significant after accounting for age and disease duration, suggesting that APOE ε4 may selectively weaken one of the metabolic clues clinicians use to recognize DLB.

That weakened signal could have important consequences for diagnosis. If APOE ε4 increases Alzheimer’s-related amyloid and tau pathology in a person who also has Lewy body disease, the posterior cingulate cortex may become more metabolically vulnerable. The resulting PET scan could lose the distinctive contrast between a relatively preserved posterior cingulate and a metabolically reduced occipital cortex. In practical terms, an APOE ε4 carrier with DLB might produce a scan that appears more Alzheimer-like, particularly in the early stages when abnormalities are subtle. The study did not directly measure amyloid or tau in these participants, so the proposed mechanism remains an interpretation rather than a proven causal chain. Still, it fits earlier work connecting tau accumulation in the posterior cingulate with cognitive impairment and loss of the cingulate island sign in Lewy body disorders.

GBA1 status influenced the scans in a different way. The 14 GBA1 carriers, who represented 32.6 percent of the DLB group, tended to show a more consistent, Parkinson’s- and DLB-like whole-brain metabolic pattern. Non-carriers displayed greater variability and significantly higher expression of an Alzheimer’s disease-related metabolic pattern. Their median Alzheimer’s-related pattern score was 3.9, compared with 2.7 among carriers, a difference that remained significant in the study’s analysis. The non-carrier group also showed more variation in several disease-related networks, suggesting that their DLB diagnosis encompassed a wider range of underlying biological processes. The researchers caution that these were not separate metabolic categories; instead, the patients occupied positions along a continuous spectrum extending from typical DLB and Parkinson’s disease toward Alzheimer’s-like metabolism.

The machine-learning analysis reinforced that interpretation. The model, known as generalized matrix learning vector quantization, was trained using an independent cohort from the University Medical Center Groningen and then used to project the Tel Aviv participants into a multidimensional space based on glucose uptake in 48 brain regions. GBA1 carriers clustered more tightly around the model’s DLB prototype and showed a modest tendency to lie closer to the Parkinson’s disease prototype. Non-carriers were more dispersed, with some positioned toward the Alzheimer’s disease region and others deep within the DLB space, where they appeared to have a stronger overall metabolic expression of disease. The difference in within-group clustering was statistically significant. However, the model correctly classified only 60 percent of DLB patients overall, emphasizing that the approach currently describes biological similarity rather than providing a stand-alone diagnostic test.

The genetic effects also aligned with clinical findings, although the study was not large enough to establish individual-level predictions. Higher expression of Alzheimer’s-, Parkinson’s- and DLB-related metabolic patterns was associated with lower scores on the Mini-Mental State Examination and the Montreal Cognitive Assessment, indicating greater cognitive impairment. APOE ε4 carriers had lower average Mini-Mental State Examination scores than non-carriers, a difference that could itself have contributed to their reduced cingulate island sign. Motor severity, measured with the motor section of the Unified Parkinson’s Disease Rating Scale, correlated with disease duration but not with the PET network scores. Healthy controls and relatives were generally recognized as healthy by the machine-learning model, and the two healthy relatives who carried GBA1 mutations did not show a scan pattern suggesting elevated DLB risk.

The researchers stress that the findings need replication in larger and more diverse populations. The study involved only 43 DLB patients, and the sample was drawn from an Ashkenazi Jewish cohort in which GBA1 variants are unusually frequent. The small number of participants also prevented reliable comparisons between individual mutations, which can have different biological effects. The APOE ε4 carrier group had more cognitive impairment, making it difficult to separate genetic effects from disease severity. In addition, the machine-learning analysis included a limited overlap between participants used for scanner harmonization and those evaluated in the test cohort. No metabolic signature accurately identified a patient’s genetic status on an individual basis. Even so, the convergence of regional measurements, disease-network analysis and machine learning suggests that genetic background is not merely a risk factor operating before disease onset. It may continue to shape how DLB unfolds in the living brain.

The broader message is that DLB may be less like a single disease than a family of related biological trajectories. GBA1 mutations appear to favor a more homogeneous alpha-synuclein-associated pattern resembling Parkinson’s disease and classic DLB, while APOE ε4 may make a specific brain region, the posterior cingulate cortex, more susceptible to Alzheimer’s-like dysfunction. Combining genetic testing with advanced FDG-PET analysis could therefore help clinicians recognize patients whose scans fall outside the textbook pattern. It might also allow researchers to design more informative clinical trials by separating participants with relatively pure synuclein-driven disease from those with substantial Alzheimer’s co-pathology. For now, the results do not change clinical diagnosis or treatment, but they offer a compelling explanation for why the same label—dementia with Lewy bodies—can conceal markedly different metabolic stories inside the brain.