Alzheimer’s disease (AD) is a progressive neurodegenerative disorder and the most common cause of cognitive decline and dementia in the elderly. It is characterized by brain atrophy, extracellular amyloid-β (Aβ) plaques, and intracellular tau neurofibrillary tangles. These pathological changes begin before clinical symptoms appear, making early detection of biomarkers essential for improving diagnosis and treatment.

This study aimed to identify DNA methylation signatures in blood samples from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) to explore their association with clinical features. We analyzed genome-wide methylation data from 225 cognitively normal (CN) individuals, 387 with Mild Cognitive Impairment (MCI), and 37 AD patients to evaluate the role of epigenetic drift in disease susceptibility and progression.

Stochastic Epigenetic Mutations (SEMs) were estimated using the method by Gentilini et al. (PMID: 26342808) as a measure of epigenetic drift. Epigenetic age acceleration (EAA) was assessed using several DNA methylation-based clocks.

No significant association was found between EAA and AD risk. However, SEM burden in both hypo- and hypermethylated states showed a progressive increase from CN to MCI to AD, suggesting a link with disease severity. The epigenetic load was lowest in CN, intermediate in MCI, and highest in AD, supporting its potential role as a biomarker for early detection.

These results highlight the promise of SEMs as early, non-invasive biomarkers for Alzheimer’s disease. They offer insights into the contribution of epigenetic drift to AD pathogenesis and support the utility of methylation-based signatures in distinguishing disease stages.

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by β-amyloid plaques and t neurofibrillary tangles. It includes a rare Mendelian form and a more common late-onset sporadic form (LOAD), influenced by genetic and non-genetic factors. While the APOE-ε4 allele is the main genetic risk factor for LOAD, disease expression varies widely. Epigenetic mechanisms such as DNA methylation and chromatin remodeling are increasingly recognized as contributors to AD susceptibility and progression.

This study aimed to identify epigenetic markers associated with LOAD and its clinical features. Genome-wide DNA methylation profiles from 451 individuals, 151 AD patients, 163 centenarians, and 137 cognitively healthy controls, were analyzed using the Illumina MethylationEPIC v2.0 BeadChip. Two complementary approaches were used: epigenetic clocks to assess biological age acceleration and quantification of Stochastic Epigenetic Mutations (SEMs) to measure epigenetic drift.

No significant association was found between epigenetic and chronological age. However, AD patients showed reduced proportions of CD8+ and CD4+ T cells, NK cells, B cells, and monocytes, with a slight increase in neutrophils. Differential methylation analysis revealed 8 CpG sites, including APBA2, SYT14, and LINC00293. DMR analysis identified regions overlapping these genes, with a key DMR in LINC00293 also significant in centenarians. SEM burden in both hypo- and hypermethylated states was associated with increased AD risk.

This is among the first studies to integrate EPICv2 data and SEM analysis in sporadic AD. The involvement of LINC00293 and SEM patterns suggests novel epigenetic biomarkers and pathways relevant to AD pathogenesis and potential targets for intervention.

Aging is the major risk factor for neurodegenerative diseases, including Alzheimer disease (AD). Many of these neurodegenerative diseases are associated with the deposition of specific peptide and proteins in the brain. Our potential to prevent and/or treat these diseases depends heavily on our ability to diagnose them early. The accumulating knowledge of the pathobiology of AD are indicating that the pre-clinical manifestation of the disease with the earliest established biomarkers (loss of amyloid b peptide in the cerebrospinal fluid and accumulation of amyloid plaques under positron emission tomography), is preceded by a phase of neuronal hyperexcitability with excessive glutamate and soluble amyloid b peptide forms in ther extracellular space of the brain.

Under Age-It we have therefore set up models of neuronal hyperexcitability using both neuroblastoma cells and primary neurons, under conditions in which glutamate and Ab are present at a “sub-threshold” concentrations, that do not cause evident cell toxicity. We will show that under such conditions of treatment, extrasynaptic ionotropic glutamate receptors are activated to a low and local extent so that the overall intracellular Ca²⁺ concentration is not increased. These events allow the cells to control all the cellular dysfunctional outcomes typically associated with Ab₄₂/Glu, but they result in a slow Ca²⁺-dependent progressive increase of reactive oxygen species (ROS), that are mainly produced by NADPH oxidases (NOXs) in neuroblastoma cells and mostly produced by the same enzyme class in primary neurons. This profile is consistent with the upregulation of proteins involved in the Ca²⁺-dependent regulation of NOX enzymes found in the most recent large proteomic studies involving relatives of familial AD cases fifteen-to-five years before clinical onset and will provide opportunities to investigate the long-term biological, molecular, and metabolic signalling consequences of NOX-mediated ROS production in the context of neuronal hyperactivation preceding AD.

Short abstract

Alzheimer’s Disease (AD) is under several genetic and environmental influences, which are seldom integrated in longitudinal population studies.

We estimated incident AD risk in 18,355 dementia-free participants with clinical, pollution and genetic data available from a prospective Italian population cohort (2005-2010; ≥35 years; 51.9% women). 235 incident AD cases were identified and validated up to 31/12/2022, through the interrogation of the main regional dementia centers’ records.

Yearly exposures to nitrogen dioxide (NO₂) and particulate matter <10 μm (PM₁₀) were estimated from land boxes of the regional environmental agency and polygenic susceptibility to AD was computed through a Polygenic Risk Score (AD-PRS). These were tested for association with incident AD risk using stepwise Cox Proportional Hazard regressions, along with age, sex, education, lifestyles, quality of life (SF-36), adiposity and prevalent health conditions of participants.

Eight features were retained in the final model: age, PM₁₀, AD-PRS, NO₂, physical and mental wellbeing, prevalent cerebrovascular disease and sex. We observed significant risk associations for PM₁₀ and NO₂ levels, AD-PRS and age, and a protective association of mental wellbeing. AD-PRS and air pollutants explained 0.7% and 5.1% of incident AD risk, over a total R² of 14.5% for all the selected features. Time-dependent Area Under the Curve (AUC) was excellent despite a slight decrease over follow-up time (0.97 to 0.92).

These findings reveal a notable influence of environmental - rather than genetic - exposures on AD risk and suggest the integration of internal and external exposome layers as an effective strategy for AD risk prediction.

Extended abstract

Background: Alzheimer’s Disease (AD), the most common type of dementia, is a multifactorial disease under several genetic and environmental influences. However, longitudinal population studies jointly investigating these exposures are scarce.

Methods: To clarify these influences, we estimated incident AD risk in 18,355 participants apparently free from prevalent dementia and with clinical, pollution and genetic data available from the Moli-sani study, a prospective Italian population cohort (2005-2010; ≥35 years; 51.9% women). 235 incident AD cases were identified up to Dec 31, 2022 (median (IQR) follow-up: 15.4 (1.8) years) through the interrogation of the main regional dementia centers’ records, and validated by qualified neurologists.

Yearly exposures to nitrogen dioxide (NO₂) and particulate matter <10 μm (PM₁₀) were estimated from land boxes of the regional environmental agency, interpolated via Kriging algorithms and linked to each participant based on its residence address, then averaged over the follow-up period. Polygenic susceptibility to AD based on common variants was estimated through a Polygenic Risk Score (AD-PRS), trained on the largest AD Genome-Wide Association Scan available. These exposures were tested for association with incident AD risk using stepwise Cox Proportional Hazard regressions based on Akaike Information Criterion minimization, along with other predictors like age, sex, education, lifestyles, quality of life (SF-36 health survey), adiposity and prevalent health conditions (cardiovascular disease, diabetes and cancer). Missing data were imputed through a k-nearest neighbor algorithm (k = 10), when necessary.

Results: Eight features were retained in the final Cox model: age, PM₁₀, AD-PRS, NO₂, physical and mental wellbeing, prevalent cerebrovascular disease and sex. We observed risk associations for PM₁₀ and NO₂ levels (HR [95% CI]: 6.68 [4.91-9.08] and 1.86 [1.39-2.49] for participants above vs below median air pollutant exposure levels), AD-PRS (2.55 [1.90-3.41] per SD increase) and age (1.11 [1.09-1.12] per year), and a protective association for mental wellbeing (0.97 [0.96-0.99] per unit increase). AD-PRS and air pollutants explained 0.7% and 5.1% of the variance in incident AD risk, over a total Nagelkerke R² of 14.5% for all the selected features. Time-dependent Area Under the Curve (AUC) was excellent despite a slight decrease over follow-up time (from 0.97 to 0.92).

Conclusion: These findings reveal a notable influence of environmental - rather than genetic - exposures on AD risk, identifying air pollutants as potential public health targets to lower dementia risk. Moreover, they suggest that the integration of internal and external exposome layers may be the key to effective risk prediction algorithms in the general population. Further validation of these observations in independent cohorts is warranted to substantiate these findings.

Background: Plasma biomarkers are the preferable tool in diagnostic field for their characteristics of economy and easily repeatability. Our previous results showed that Neurofilament Light Chain (NfL), Glial Fibrillary Acidic Protein (GFAP) and Phosphorylated-tau-181 (pTau181), detected in plasma, were able to distinguish different stages of Alzheimer’s Disease (AD): Subjective cognitive decline (SCD), Mild cognitive impairment (MCI) and dementia-AD (d-AD)¹. Moreover, recently we reported that plasma pTau217 levels reflect severity of cognitive impairment and discriminate between patients positive and negative for cerebrospinal fluid (CSF) and neuroimaging biomarkers². Nether less, we identified cut-offs for plasma NfL, pTau181 and ptau217 that have accuracy and predictive value in cognitive decline^1,2. The aim of this study was to extend and confirm our previous findings and establish a plasma biomarker panel useful for the diagnosis of AD and its preclinical stages.

Materials and Methods: Study included 200 patients affected by different stages of cognitive decline (SCD, MCI and d-AD). All patients, recruited at the Neurology Unit of Careggi Hospital in Florence, underwent a comprehensive clinical and neurological assessment, CSF and blood sampling. Plasma was isolated from blood and analyzed at SiMoA SR-X platform (Quanterix Corp) to detect NfL, GFAP and pTau181, and at Lumipulse G-600 (Fujirebio) for pTau217 measurement. All patients were genotyped for Apolipoprotein E (APOE). Statistical analyses were performed using SPSS software version 28 (IBM SPSS Statistics), p < .05 was set as significant.

Results: A statistically linear relationship emerged between diagnosis and mean concentration of plasma biomarkers (NfL, GFAP, pTau181 and pTau217): SCD showed the lowest concentration, instead AD patients had highest levels. A statistically significant association was found between plasma biomarkers and AD biomarkers measured in CSF. A linear regression analysis showed a positive correlation between plasma biomarkers and age at disease onset. Moreover, a positive correlation emerged with Epsilon4 (e4) allele of APOE gene, too. APOE ε4 carriers had increased plasma biomarkers levels compared to non-carriers.

Discussion and conclusions: Study results showed that plasma NfL, GFAP, pTau181 and pTau217 were able to discriminate between different stages of dementia. Preclinical stages (SCD and MCI) had the lowest plasma concentration, AD the highest concentration of NfL, GFAP, pTau181 and pTau217. Plasma biomarkers change as the progression of cognitive decline. Plasma and CSF AD biomarkers were directly correlated. So, plasma levels reflect the underling AD pathology and Amyloid Beta (Aβ) accumulation. These findings confirmed our previous results, that plasma NfL, GFAP, pTau181 and pTau217 can predict cognitive decline^1,2. In addition, patients carrying APOE-e4, had plasma biomarkers increased than non-carriers. Studies speculated that APOE-ε4 affect Aβ clearance³. In conclusion, taking together, plasma NfL, GFAP, pTau181 and pTau217 provide a plasma biomarker panel useful for clinical assessment of dementia.

References

1 Ingannato A, Bagnoli S, Mazzeo S, et al. Plasma GFAP, NfL and pTau 181 detect preclinical stages of dementia. Front Endocrinol (Lausanne). 2024;15:1375302. Published 2024 Apr 9.

2 Giacomucci G, Crucitti C, Ingannato A, et al. The two cut-offs approach for plasma p-tau217 in detecting Alzheimer's disease in subjective cognitive decline and mild cognitive impairment. Alzheimers Dement (Amst). 2025;17(2):e70116. Published 2025 May 11.

3 Castellano JM, Kim J, Stewart FR, et al. Human apoE isoforms differentially regulate brain amyloid-β peptide clearance. Sci Transl Med. 2011;3(89):89ra57.