Introduction

The neuroexposome framework underscores the intricate relationships between environmental factors and neurological health, illuminating how external exposures can critically influence brain function and contribute to the onset and progression of neurodegenerative diseases1,2. As dementia rises to the forefront of global health concerns, its ramifications extend beyond individuals, deeply impacting families, caregivers, and entire communities3. Current projections suggest a staggering increase in dementia prevalence, with estimates indicating that millions more will be affected in the coming decades, posing significant challenges to healthcare systems and societal infrastructure4.

Among the spectrum of dementia types, Alzheimer’s disease (AD) is the most widespread, characterized by the pathological accumulation of amyloid-β plaques and hyperphosphorylated tau proteins5,6. These hallmark features lead to a cascade of neurodegenerative processes that progressively erode cognitive function, ultimately impairing memory, reasoning, and daily living activities5,6. Following AD, vascular dementia (VaD) emerges as the second most common form, primarily resulting from compromised cerebral blood flow due to ischemic lesions7. This disruption compromises neuronal vitality and impacts cognitive domains, causing significant impairment and loss of independence7.

Metabolomic profiling is a revolutionary tool in dementia research, offering profound insights into the biochemical alterations associated with these neurodegenerative disorders8. By analyzing specific metabolites present in biological samples, researchers can detect disruptions in metabolic pathways that may correlate with disease severity and environmental influences9,10. This approach provides a unique biochemical fingerprint, allowing for the identification of potential biomarkers linked to cognitive decline11,12,13.

The main aim of this study is to undertake a detailed exploration of the metabolomic landscape of dementia within the understudied Egyptian population. Through this observational analysis, we aspire to establish a foundational understanding as a launching point for future longitudinal studies focused on the neuroexposome. By advancing our comprehension of the relationship between environmental exposures and the development of dementia, we hope to contribute to more targeted interventions and strategies that could mitigate the risks associated with this debilitating condition.

Results

Cohort structure

The study cohort was stratified into three groups: 61 individuals diagnosed with clinical Alzheimer’s disease (AD), 76 with clinical vascular dementia (VaD) or preclinical cognitive impairment due to a vascular event, and 100 cognitively healthy controls. Details of the cohort structure and composition are provided in Table 1.

Table 1 Cohort Structure
Full size table

Data exploration and analysis

To inspect the data, we applied Principal Component Analysis (PCA) to reduce the dimensionality of the metabolomic data while retaining the most informative variance. We plotted the first principal components (PC1 and PC2) to explore clustering patterns among AD patients, VaD patients, and healthy controls. Each point on the PCA plot represents an individual sample, with different colors indicating the various phenotypic groups. We drew ellipses to capture each group’s 95% confidence intervals, which helps visualize the separation and potential overlap among the groups. The degree of separation between clusters reflects metabolic differences and highlights the distinct biochemical signatures associated with each condition (Fig. 1).

Fig. 1: PCA of Metabolomic Data in AD, VaD, and Control Groups.
figure 1

PCA plot illustrating the separation of patients and healthy controls based on all metabolomic profiles identified. Dots colored green are the control group, pink is the Alzheimer’s disease group, and blue is the vascular dementia group.

Full size image

A generalized regression analysis model, adjusted for age and sex as covariates, was employed to analyze 291 metabolites, identifying significant differences between Alzheimer’s disease (AD) and control groups, vascular dementia (VaD) and control groups, and between VaD and AD groups. Specifically, 59 metabolites were significantly altered in the AD group compared to controls, including 20 environmental and agrochemical contaminants. In the VaD group, 69 metabolites showed significant changes compared to controls, including 31 environmental and agrochemical contaminants. Nine significant metabolites were identified as specific to the VaD group compared to the AD group (Supplementary Tables 1–5). False Discovery Rate (FDR) adjusted p-value at <0.01 (strict), <0.05 (moderate), and <0.1 (exploratory) were used as thresholds for statistical significance.

PCA is used to investigate clustering patterns in AD versus control groups and VaD versus control groups to inspect the significant metabolomic data in each group (Fig. 2). In addition, the heatmap visually represents the clustering of significant metabolites between AD patients and controls and between VaD patients and controls (Fig. 3). By organizing samples and metabolites based on similarity, these heatmaps reveal distinct metabolic profiles associated with each condition. The color variations indicate differences in metabolite abundance, with a clear separation between the patient and control groups. This clustering highlights metabolic dysregulation in AD and VaD, potentially identifying altered pathways linked to the disease process. The applied hierarchical clustering further reinforces group differentiation, demonstrating metabolic alterations in the disease process.

Fig. 2: PCA of significant metabolites in AD and VaD patients compared to controls.
figure 2

A PCA plot illustrating the separation of AD patients and healthy controls based on significant metabolic profiles. B PCA plot illustrating the separation of VaD patients and healthy controls based on significant metabolic profiles.

Full size image
Fig. 3: Heatmaps of Significant Metabolites in AD and VaD Patients compared to Controls.
figure 3

A Heatmap depicting the relative abundance of significant metabolites in AD patients versus controls, with hierarchical clustering highlighting metabolic differences. B Heatmap representing the relative levels of significant metabolites in VaD patients compared to controls, with clustering patterns indicating distinct metabolic alterations.

Full size image

Metabolomic data analysis in Alzheimer’s disease

Based on the generalized regression model comparing the Alzheimer’s disease (AD) group to the control group, 26 metabolites demonstrated high statistical significance (FDR-adjusted p < 0.001), nine showed moderate significance (FDR < 0.05), and four were included for exploratory purposes due to mild significance (FDR < 0.1). Notably, DL-2-Hydroxyvaleric acid, L-Norvaline, and Biotin exhibited the highest increases, with fold changes of 4.5, 4, and 3.6, respectively (FDR p-values: 1.99E-24, 9.19E-19, 5.25E-16). Among the environmental and agrochemical contaminants, 12 showed high significance, six had moderate significance, and two were mildly significant. The most prominent increases were observed in Prometryn (triazine herbicide), Dimethenamide ESA (chloroacetamide herbicide metabolite), and Foramsulfuron (sulfonylurea herbicide), with fold changes of 4.4, 3.4, and 2.6, respectively (FDR p-values: 1.07E-24, 3.02E-13, 7.03E-08). Visualizations through the volcano and box plots further clarify significant metabolite alterations and their distribution, thus enhancing the understanding of metabolic dysregulation in AD patients (Figs. 4, 5).

Fig. 4: Volcano plots of significant metabolites in Alzheimer’s Disease (AD) patients compared to controls.
figure 4

A Volcano plot displaying the distribution of significant metabolites and B Volcano plot displaying the distribution of significant environmental and agrochemical contaminants based on Log2 (fold change) and -Log10 (FDR-adjusted p-value). Metabolites with FDR P < 0.01 are highlighted in red, those with FDR P < 0.05 in green, FDR P < 0.1 (exploratory) in blue, and non-significant metabolites (FDR P > 0.1) in gray. Metabolites with more than 1 log fold-change are labeled.

Full size image
Fig. 5: Box plots of significant metabolites in Alzheimer’s Disease (AD) patients compared to controls.
figure 5

A Box plots representing the relative abundance of significant metabolites. B Box plots representing the relative abundance of significant environmental and agrochemical contaminants. The figure illustrates variations in individual metabolite intensity across groups. Metabolites with FDR P < 0.01 are labeled (***), those with FDR P < 0.05 are labeled (**), and FDR P < 0.1 are labeled (*).

Full size image

Metabolomic data analysis in vascular dementia

Based on the generalized regression model comparing the vascular dementia (VaD) group to the control group, 33 metabolites showed high statistical significance, seven exhibited moderate significance, and two had mild significance. The most significantly increased metabolites were DL-2-Hydroxyvaleric acid, L-Homocarnosine, and L-Norvaline, with fold changes of 5.4, 3.7, and 3.4, respectively (FDR p-values: 2.63E-42, 4.93E-18, 3.09E-16). Among the environmental and agrochemical contaminants, 24 demonstrated high significance, two were moderately significant, and 4 had mild significance. Key increases were noted in Prometryn, Prothioconazole-desthio (triazole fungicide metabolite), and Dimethenamide ESA, reporting fold changes of 5, 4.5, and 3.6, respectively (FDR p-values: 1.01E-36, 4.10E-30, 1.10E-17). The distribution and variability of these significant metabolites were further illustrated through volcano and box plots, emphasizing the differences in abundance between the groups (Figs. 6, 7).

Fig. 6: Volcano plots of significant metabolites in Vascular Dementia (VaD) patients compared to controls.
figure 6

A Volcano plot displaying the distribution of significant metabolites and B Volcano plot displaying the distribution of significant environmental and agrochemical contaminants based on Log2 (fold change) and −Log10 (FDR-adjusted p-value). Metabolites with FDR p < 0.01 are highlighted in red, those with FDR p < 0.05 in green, FDR p < 0.1 (exploratory) in blue, and non-significant metabolites (FDR P > 0.1) in gray. Metabolites with more than 1 log fold-change are labeled. The volcano plots provide an overview of metabolite significance and fold change between VaD patients and controls.

Full size image
Fig. 7: Box plots of significant metabolites in Vascular Dementia (VaD) patients compared to controls.
figure 7

A Box plots representing the relative abundance of significant metabolites. B Box plots representing the relative abundance of significant environmental and agrochemical contaminants. The figure illustrates variations in individual metabolite intensity across groups. Metabolites with FDR p < 0.01 are labeled (***), those with FDR p < 0.05 are labeled (**), and FDR p < 0.1 are labeled (*).

Full size image

Comparative metabolomic analysis between vascular dementia and Alzheimer’s disease

Generalized regression analysis, adjusted for age and sex, was employed to assess associations between vascular dementia (VaD) and Alzheimer’s disease (AD) groups, with statistical significance defined as a false discovery rate (FDR) adjusted p < 0.1. This analysis identified a distinct metabolomic profile associated with the VaD phenotype (Fig. 8). In panel A, nine metabolites showed significant differences in abundance between the two groups with FDR p < 0.1, with metabolites such as 5’-S-methylthioadenosine, cCMP, and Guanosine being markedly elevated in VaD compared to AD patients (FDR p-values: 5.87E-08, 0.00072426, 0.0008751). These highly significant metabolites suggest potential discriminatory biomarkers that need further validation. Conversely, metabolites such as 4-hydroxybenzoic acid, N-carbamoyl-DL-aspartic acid, and Biotin showed lower levels in VaD in comparison to AD patients (FDR p-values: 1.08E-08, 7.34E-06, 5.50E-05). Panel B supports these findings with box plots showing apparent group-wise differences in metabolite intensities. These patterns suggest underlying differences in metabolic pathways, possibly reflecting distinct pathophysiological mechanisms in VaD and AD, and highlight the potential use of serum metabolomics in differentiating between the two dementia subtypes.

Fig. 8: Comparative metabolomic analysis of significant metabolites in VaD compared to AD groups.
figure 8

A Volcano plot displaying the distribution of significant metabolites in VaD compared to AD patients based on Log2 (fold change) and -Log10 (FDR-adjusted p-value). Metabolites with FDR P < 0.01 are highlighted in red, those with FDR P < 0.05 in green, FDR P < 0.1 (exploratory) in blue, and non-significant metabolites (FDR P > 0.1) in gray. B Box plots representing the relative abundance of significant metabolites in AD versus VaD patients, illustrating variations in metabolite intensity across groups. Metabolites with FDR P < 0.01 are labeled (***), those with FDR P-values < 0.05 are labeled (**), and FDR P < 0.1 are labeled (*). The volcano plots provide an overview of metabolite significance and fold change, while the box plots emphasize individual metabolite differences between AD and VaD patients.

Full size image

Pathway-enrichment analysis

Compared to controls, significantly altered metabolites in each disease group were subjected to over-representation analysis (ORA) to identify disproportionately affected metabolic pathways, with statistical significance defined as a p < 0.05 relative to the background metabolome. The pathway enrichment analysis revealed both shared and distinct metabolic alterations in the Alzheimer’s disease (AD) and vascular dementia (VaD) groups (Fig. 9). The citrate cycle (TCA cycle) was significantly enriched in both AD patients (p-value: 0.00282) and VaD patients (p-value: 0.0036), underscoring a common disruption in central energy metabolism associated with neurodegeneration. The AD group showed distinct enrichment in pathways related to amino acid and lipid metabolism, including branched-chain amino acid (BCAAs) biosynthesis (p-value: 0.00567), butanoate metabolism (p-value: 0.0199), and phenylalanine, tyrosine, and tryptophan biosynthesis (p-value: 0.0585). These findings suggest potential impairments in mitochondrial function, lipid-derived energy regulation, and amino acid metabolism in AD patients. In contrast, the VaD group was characterized by unique enrichment in purine metabolism (p-value: 0.0044), vitamin B6 metabolism (p-value: 0.00851), and alanine, aspartate, and glutamate metabolism (p-value: 0.00949), reflecting potential alterations linked to vascular dysfunction and oxidative stress.

Fig. 9: Comparative pathway enrichment analysis in Alzheimer’s disease (AD) and vascular dementia (VaD).
figure 9

Bubble plots display the metabolomic pathway enrichment analysis results in A AD group and B VaD group. Pathways are ranked by statistical significance, shown as –log10(p-value) on the x-axis. Bubble size indicates the enrichment ratio (observed vs expected hits), while color indicates statistical significance (p-value) from red (most significant) to yellow (least significant).

Full size image

Discussion

Dementia is a multifaceted disease influenced by genetic, environmental, age-related, sex-specific, and lifestyle factors, contributing to its phenotypic heterogeneity14. Given that cognitive, behavioral, and functional impairments may take years, there remains a significant gap in understanding perspective metabolite associations that could enhance our understanding of the disease process and improve early detection strategies15.

In this study, we have identified 20 and 31 significantly elevated serum levels of environmental and agrochemical contaminants in AD and VaD patients, reinforcing previous findings that link pesticide exposure to neurodegenerative processes through several mechanisms16,17,18. Pesticides are recognized neurotoxins and chronic exposure is associated with neuronal loss, cognitive decline, and an increased risk of neurodegenerative diseases, including dementia17. We have identified several classes of herbicides and their metabolites. Triazine, Uracil, and pyrrolidinone herbicides exhibit disruption in mitochondrial function and oxidative phosphorylation, leading to increased oxidative stress and neuronal apoptosis19. Chloroacetamide herbicide metabolites interfere with lipid metabolism, potentially affecting neuronal membrane integrity and neurotransmission20. Phenylurea and phenoxy herbicide metabolites contribute to neuroinflammation and axonal degeneration, further exacerbating neurodegenerative disease pathology17. A previous study showed that β-Triketone herbicides disrupt dopamine pathways, which may contribute to Parkinsonian neurodegeneration21. Furthermore, Phenylpyrrole and triazole fungicide metabolites induce oxidative stress and impair synaptic plasticity, mechanisms implicated in cognitive decline22. Strobilurin fungicides impair mitochondrial function, leading to neuronal energy deficits and apoptosis23. Additionally, organophosphate fungicides and insecticides irreversibly inhibit acetylcholinesterase, leading to excessive synaptic acetylcholine accumulation, excitotoxicity, and neuronal death17. The detection of elevated pesticide levels in our AD and VaD patients suggests a potential role of environmental toxins in disease pathogenesis, further emphasizing the need for stringent monitoring of neurotoxic exposures and their long-term impact on cognitive health.

Glycerophospholipids such as phosphatidylinositols, phosphatidylcholines, and phosphatidylethanolamines play a crucial role in biological pathways related to lipid metabolism, lipid peroxidation, and lipid transport24. At the cellular level, glycerophospholipids are integral to signaling pathways involved in apoptosis25, adhesion26, migration27, and proliferation28. Consistently, the significant elevation of glycerophospholipids has previously demonstrated association with neuronal atrophy, contributing to the pathogenesis of dementia29. The findings of this study reveal a marked increase in levels of lipid metabolites among patients. DL-2-Hydroxyvaleric acid exhibited the highest increase in the AD group with 4.5-fold change (FDR p-value: 1.99E-24) and 5.4-fold change in the VaD group (FDR p-value: 2.63E-42), followed by 13-Hydroxyoctadecadienoic acid (13-HODE), Phosphatidylinositols, Phosphatidylcholines, and Phosphatidylethanolamines. It is worth noting that this is the first time DL-2-Hydroxyvaleric acid has been observed to have a marked increase in AD and VaD patients.

Conversely, chronic pesticide exposure has been implicated in disrupting lipid metabolism, contributing to notable metabolic alterations30,31. Previous studies revealed that the lipid β-oxidation pathway and disturbances in glycerophospholipid metabolism are common pathways affected by several pesticide classes31,32,33. Our findings reveal a marked elevation in pesticide serum metabolites in AD and VaD patients, which may suggest a potential association with alterations in the serum lipid metabolome.

It is well established that energy metabolism and mitochondrial dysfunction play a pivotal role in disease pathogenesis, as impaired mitochondrial function disrupts ATP production, increases oxidative stress, and contributes to neuronal degeneration, ultimately exacerbating cognitive decline34. However, direct evidence supporting the alterations in metabolites affecting energy metabolism and mitochondrial dysfunction is unclear. L-carnitine plays a crucial role in lipid metabolism and β-oxidation, facilitating the transport of long-chain fatty acids into the mitochondria for energy production35. Notably, previous studies have reported altered serum carnitine levels in individuals with AD and mild cognitive impairment (MCI), potentially contributing to the disease’s progression through disruptions in energy metabolism and interactions with key risk factors like APOE e436,37,38,39,40,41. In the current study, serum L-carnitine was markedly increased in AD patients (2-fold, FDR p-value: 0.00055) and VaD patients (1.9-fold, FDR p-value: 0.00048). Similarly, L-malic acid and L-succinic acid serve as intermediates in the citric acid cycle within the mitochondria, further contributing to energy production42. A prior study indicated a marginal and non-significant increase in malic acid serum levels among AD patients43, and another study reported lower mean succinate levels in the cerebrospinal fluid of AD patients44. However, our current study reveals a significant increase in both malic acid (3-fold, FDR p-value: 1.26E-11) and succinic acid (1.9 folds, FDR p-value: 0.001) serum levels in AD patients. Similarly, in VaD patients, Malic acid shows a 2.6-fold increase (FDR p-value: 1.05E-08), and succinic acid shows a 2.3-fold increase (FDR p-value: 9.83E-07). Moreover, Tiglyl-glycine (N-acyl glycine), whose presence in body fluids can serve as a diagnostic marker for mitochondrial lipid beta-oxidation disorders (HMDB0000959), was found to be markedly increased by 3-fold in AD (FDR p-value: 6.21E-10) and 1.7-fold in VaD patients (FDR p-value: 0.007), suggesting a potential disruption in lipid metabolism and mitochondrial dysfunction. Furthermore, L-ornithine is critical in metabolic pathways linked to neurodegenerative diseases45. L-ornithine’s central role is in the polyamine synthesis pathway, which is significantly dysregulated in AD with altered ornithine decarboxylase (ODC) enzyme activity, potentially influencing neurodegenerative and restorative processes46. Previous studies have reported altered plasma levels of L-ornithine in AD patients43,45,47,48,49. Consistently, our findings reveal a significant reduction in AD patients (0.40 folds, FDR p-value: 2.79E-07) and VaD patients (0.47 folds, FDR p-value: 2.31E-05)45,48,49. On the other hand, chronic pesticide exposure can perturb pathways related to mitochondrial energy metabolism. Yan Q. et al. revealed that pathways such as the carnitine shuttle, short-chain fatty acid (butanoate) metabolism, glycolysis and gluconeogenesis, and pyruvate metabolism were found to be altered with higher pesticide exposure31. These findings may support the link between the increased pesticide serum metabolite levels and the implicated metabolome alterations observed.

Previous research reported discrepancies in aromatic amino acid levels in AD, depending on several factors, including the sample’s origin (serum, CSF, brain, and different brain regions), the stage of the disease, and the methodology used for detection50. The altered levels of these amino acids can have significant implications for AD pathology, potentially affecting neurotransmitter levels51,52,53. In serum, previous studies reported either a decreased level or an insignificant alteration of phenylalanine, tyrosine, and tryptophan in AD patients compared to healthy controls47,54,55. In addition, several studies reported decreased phenylalanine levels in CSF of AD patients50,56,57. However, in postmortem brain tissue, findings are less consistent51,52,53,58. In the current study, we report a significant reduction in serum phenylalanine in AD (0.6-folds, FDR p-value: 0.03) and VaD patients (0.4-folds, FDR p-value: 0.001), a marked increase in serum tryptophan in AD (1.6-folds, FDR p-value: 0.03) and VaD patients (2.4-folds, FDR p-value: 7.39E-07), and no significant change in tyrosine level in AD and VaD patients. Therefore, the relationship between AD and phenylalanine, tyrosine, and tryptophan levels is complex. Further investigation is required to understand fully the specific changes occurring in different compartments and their contribution to the disease’s progression.

Lower levels of BCAAs, such as isoleucine, leucine, and valine, were reported to be associated with an increased risk of AD59. However, some studies have reported decreased valine levels in individuals with MCI but elevated levels in those with AD59. Another study found no significant differences in blood levels of BCAAs among healthy individuals, MCI patients, and those diagnosed with AD60. Harshfield et al. found that individuals who were dementia-free but showed signs of small vessel disease had lower serum valine levels associated with an increased risk of developing dementia61. However, a study analyzing serum amino acid metabolite levels found no significant differences in the total amino acid concentrations between healthy controls and patients with vascular dementia (VaD)62. The findings of this study reveal elevated serum valine and leucine levels of 1.7-fold (FDR p-value: 0.01) and 2-fold (FDR p-value: 0.0004) in AD patients, while no significant alterations in both valine and leucine were observed in VaD patients. It is worth noting that unlike most amino acids, which the liver significantly metabolizes during the first pass, BCAAs undergo minimal processing there. As a result, blood levels of BCAAs often directly reflect dietary intake63. This might explain the different alterations in blood levels across studies.

Moreover, significantly reduced levels of 5′-Methylthioadenosine (MTA) in dementia patients may reflect disruptions in pathways involving polyamine metabolism, methylation, and purine salvage, which are critical for neuronal function and survival64,65. MTA is also involved in cellular signaling, epigenetic regulation, and oxidative stress response, and its deficiency may implicate neurodegeneration66. Additionally, reduced MTA levels could impair DNA and histone methylation, contributing to epigenetic changes associated with AD67. Furthermore, decreased MTA may suggest impaired purine metabolism, leading to neuron energy deficits, and could indicate a chronic inflammatory state, exacerbating neurotoxicity68. In this study, MTA showed a significant reduction in AD patients (0.3-fold, FDR p-value: 5.02E-11) and VaD patients (0.3-fold, FDR p-value: 5.19E-12). Moreover, the comparative analysis revealed a significant increase in serum MTA levels in VaD compared to AD patients (1.9-fold, FDR p-value: 5.87E-08).

Another purine, Guanosine, has been increasingly recognized for its neuroprotective potential in AD69. Several in vitro and in vivo studies have demonstrated that Guanosine protects against β-amyloid-induced apoptosis, neuroinflammation, and oxidative stress70. Moreover, Guanosine plays an important role in vascular injury, where it is released from astrocytes after ischemic events and is, therefore, involved in the brain’s response to vascular damage. Previous postmortem cerebrospinal fluid analysis from AD patients showed no significant difference in guanosine levels compared to controls71. However, data on serum guanosine levels in AD remains unavailable, indicating an area for future investigation. In the current study, we observed a significant reduction in Guanosine serum level in AD patients compared to controls (0.5-fold, FDR p-value: 0.00084). On the contrary, a 1.8-fold increase in guanosine serum level was observed in VaD patients compared to controls (FDR p-value: 0.00261), and interestingly, a 2-fold increase in guanosine serum level was observed in VaD compared to AD patients (FDR p-value: 0.0008751). Together, these findings underscore the importance of MTA and Guanosine in maintaining neuronal health and suggest their potential to discriminate between the two dementia subtypes. However, this needs further research and validation.

In conclusion, our study highlights significant metabolic alterations in dementia patients, particularly in lipid metabolism, energy metabolism and mitochondrial dysfunction, amino acid metabolism, and purine metabolism. The dysregulation of fatty acids, glycerophospholipids, and energy-related metabolites underscores their potential role in the disease process. The distinct metabolic profiles observed in VaD patients compared to those with AD point to underlying differences in specific metabolic pathways. These differences may reflect unique pathophysiological processes in each dementia subtype and underscore the potential of serum metabolomics to distinguish between AD and VaD. Furthermore, an important aspect of our findings is the influence of environmental and agrochemical contaminants on serum metabolome disturbances and neurotoxicity. The detection of elevated pesticide levels in AD and VaD patients supports the growing body of evidence linking their exposure to neurodegenerative processes. Overall, our findings provide new insights into the metabolic landscape of dementia in an understudied population. A deeper understanding of these metabolic disruptions and their interactions with environmental factors may lead to more effective disease prevention and management strategies.

Despite the insights gained from this study, several limitations must be acknowledged. One primary concern is the use of non-fasted biological samples, which may introduce variability in metabolite levels due to recent dietary intake. Furthermore, the absence of detailed records on participants’ dietary habits restricts our ability to fully understand the influence of nutrition on metabolic profiles. Additionally, data on medications, pesticide exposure, air pollution, and food-related contaminants were not collected in the current study, limiting our capacity to fully assess the environmental factors that could significantly impact the metabolic alterations.

Further research should build upon this study’s findings by incorporating more comprehensive datasets that include detailed dietary assessments, lifestyle factors, and environmental exposures such as agrochemical contaminants, pesticides, and air quality metrics. Longitudinal studies could enhance our understanding of how these variables influence metabolic profiles over time, particularly in diverse populations. Expanding the cohort to include a broader range of neurodegenerative disorders could facilitate a comparative analysis of metabolic disruptions across different conditions. Utilizing advanced technologies in metabolomics and integrating them with genetic and epigenetic data may further elucidate the complex interactions within the neuroexposome framework. Ultimately, these efforts will contribute to more targeted prevention and intervention strategies, improving health outcomes for individuals affected by dementia and other neurodegenerative diseases.

Methods

Participants

We recruited one hundred thirty-seven (n = 137) dementia patients from six different centers across Egypt: Cairo, Mansoura, Beni Suef, Assiut, Aswan, and Sohag, as a part of the Egyptian Dementia Network (EDN) registry. Clinical assessment was performed by neurologists, geriatricians, or geriatric psychiatrists at the university hospitals of each center. In addition, we recruited one hundred (n = 100) healthy volunteers as a reference group.

Diagnosis of subjects

The cohort includes two groups of patients. The first group is diagnosed with clinical Alzheimer’s Disease (AD) (n = 61). The second group is diagnosed with preclinical vascular MCI and clinical vascular dementia (VaD) combined (n = 76). Clinicians diagnosed patients based on the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5)72 and International Statistical Classification of Diseases and Related Health Problems 10th Revision (ICD-10)73. The ICD-10 criteria for dementia require evidence of memory decline (especially for new information), impaired cognitive functions such as judgment and planning, and no clouding of consciousness to confirm the decline. Associated emotional or behavioral changes (e.g., irritability, apathy, social inappropriateness) must be present, and symptoms should persist for at least six months for a definitive diagnosis; additional signs like aphasia, agnosia, or apraxia may also occur. Clinical Alzheimer’s dementia is diagnosed when these criteria are met without evidence of other causes such as stroke, Parkinson’s disease, metabolic disorders, or substance misuse. In contrast, clinical vascular dementia also meets the general dementia criteria but is characterized by uneven cognitive deficits (often with memory more affected than reasoning), clinical signs of focal brain damage (e.g., unilateral weakness, increased reflexes, Babinski sign), and a history or imaging evidence of cerebrovascular disease such as stroke or infarcts. Preclinical vascular MCI represents an earlier stage of vascular dementia, where mild cognitive impairment is present but daily functioning remains largely intact. Clinicians assessed impairment of activities of daily living (ADL)74 and instrumental activities of daily living (IADL)75 to rule out delirium or other psychiatric diseases. A thorough history was taken, and a series of blood tests were conducted—including liver and kidney function tests, complete blood count (CBC), blood sugar curve, lipogram, and thyroid function tests (free T3, free T4, and TSH)—to gain information about co-morbidities and rule out the presence of other systemic conditions, such as hypothyroidism, that could contribute to pseudodementia. They further assessed cognitive ability using psychometric tools such as the Mini-Mental State Examination (MMSE)76 and the Montreal Cognitive Assessment (MoCA)77. Clinicians used neuroimaging techniques, such as MRI or CT, to identify evidence of cerebrovascular disease (e.g., stroke or infarcts), which supports the diagnosis of vascular dementia by confirming the vascular origin of the cognitive decline.

Ethics statement

All participants or legal guardians provided written informed consent following the Declaration of Helsinki. Cognitively impaired individuals provided informed consent after properly assessing each participant’s capacity to consent; if a participant cannot provide informed consent, a legally authorized representative, such as the legal guardian, provides consent on their behalf. Even when a proxy provides consent, researchers seek the participant’s assent to the extent possible and respect any dissent expressed by the participant. All protocols were approved by the Institution Review Board at the American University in Cairo (IRB-AUC# 2024-2025-028).

Serum samples preparation

Non-fasted venous whole blood samples were collected for participants in serum-separator tubes, centrifuged at 3500 rpm for 15 min, and then serum was collected and stored at −80 °C. We applied a one-phase liquid extraction of the serum samples by adding four ice-cold Acetonitrile (ACN) volumes to 1 sample volume and vortexed for 30 s. Extraction mixtures were centrifuged at 15,000 × g for 10 min at 4 °C to separate the precipitated protein debris. The supernatant was filtered through a 0.22 µm Polytetrafluoroethylene (PTFE) filter and used directly for metabolomics LCMS analysis. Pooled patients and control samples were separately prepared similarly and analyzed as quality control runs.

LC-MS analysis

Samples were analyzed using liquid chromatography-electrospray ionization–tandem mass spectrometry (LC-ESI-MS/MS) with an ExionLC AC-SCIEX Triple Quad 5500 + MS/MS system (Sciex) on both positive and negative modes for a broader spectrum of detection. Ascentis® Express 90 Å C18 Column (Sigma Aldrich) was used for separation. Two eluents, five mM ammonium formate pH 3 and acetonitrile (LC grade), were implemented as mobile phases. The gradient was programmed as follows: 5% B at 0–1 min, 5–100% B from 1 to 20 min, 100% B from 20 to 25 min, 5% at 25.01, and 5% from 25.01 to 30 min with the flow rate set at 0.3 ml/min and the injection volume was 5 µl. For MS/MS analysis, positive and negative ionization modes were applied with an EMS-IDA-EPI scan from 100 to 1000 Da for MS1 with the parameters mentioned in Table 2 and from 50 to 1000 Da for MS2 with a de-clustering potential: 80/−80; collision energy: 35/−35 for positive and negative modes respectively.

Table 2 Parameters for positive and negative ionization modes MS/MS analysis
Full size table

LC-MS data acquisition and peak characterization

All data were processed using MS-DIAL 4.92 open-source software. Two libraries (MSMS-Pos-MassBank.msp and MSMS-Neg-MassBank.msp) were employed for identification. Analysis parameters were at 0.2 Da for both MS1 and MS2 tolerance. The minimum peak height was 100,000 amplitude, and the mass slice width was 0.1 Da. The identification score cut-off was 80%, and the retention time tolerance was 100 min. The adduct ion setting was defined for [M + H]+, [M + NH4]+, [M+Na]+, [M + H-H2O]+, and [M-H]-, [M-H2O-H]-, [M+Na-2H]-, [M+Cl]- for positive and negative modes respectively.

Data preprocessing and statistical analysis

Peak areas depicted the abundance of metabolites. Data processed with MS-DIAL were exported into a spreadsheet format and analyzed with R statistical software version 4.4.278. We implemented quality control steps, Probabilistic Quotient Normalization (PQN), and Log and Z-score transformations for data preprocessing. We employed PCA and heatmaps for data visualization. To assess the effect of phenotype on metabolite levels, we employed the generalized regression analysis model adjusted for age and sex as covariates. Multiple test correction was achieved using the Benjamini-Hochberg method with a False Discovery Rate (FDR) adjusted p-value at <0.01 (strict), <0.05 (moderate), and <0.1 (exploratory) set for statistical significance. We performed Metabolite Set Enrichment Analysis (MSEA) using MetaboAnalyst 6.0 (https://www.metaboanalyst.ca/home.xhtml). Over-representation analysis (ORA) was conducted by inputting the HMDB IDs of significantly altered metabolites, with the KEGG human metabolic pathways database used as the reference library.