Генетические данные о оси печени и мозга: риск липидного метаболизма и нейродегенеративного заболевания.

Introduction

Neurodegenerative diseases (NDs) of the central nervous system, including Alzheimer’s disease (AD), Parkinson’s disease (PD), multiple sclerosis (MS), and amyotrophic lateral sclerosis (ALS), affect human health and lifespan, with an increasing incidence rate worldwide. With the gradual loss of neurons with age, ND can lead to deterioration including cognitive performance, motor control, and other essential neurological functions [ 1 ]. Despite extensive research, the complete molecular and cellular pathways underlying neurodegenerative disorders remain incompletely elucidated. While advancing age represents a primary contributor to these diseases, multiple physiological elements and external environmental influences also contribute to disease manifestation [ 2 , 3 ]. Recent studies have concluded that metabolic risk factors contribute to the occurrence and development of ND, suggesting promising strategies for prevention and treatment [ 4 , 5 ].

Recent research has established the liver-brain axis as a crucial pathway in human health [ 6 ]. Studies have uncovered multiple molecular mechanisms underlying the liver-brain axis in neurodegeneration. For example, liver endogenic hormone FGF21, exerts neuroprotective effects by maintaining mitochondrial function via AMPKα/AKT signaling, reducing neuroinflammation through NF-κB pathway inhibition, and suppressing oxidative stress [ 7 ]. Also, liver-produced bile acids can cross the blood–brain barrier directly and activate nuclear receptors in neurons and glial cells, or through an indirect FXR-FGF15/19 and TGR5-GLP-1 signaling pathway to affect neuroinflammation and neurodegeneration [ 8 ]. Additionally, hepatic dysfunction of branched-chain amino acids metabolism impacts neurotransmitter synthesis and neuronal function, leading to neurotoxicity and cognitive decline [ 9 ].

Lipid metabolism in the liver, is related to the maintenance of body homeostasis, including ND. Neuroinflammation is considered a common pathophysiological mechanism of ND. Microglia in the brain are crucial for maintaining homeostasis and alleviating neuroinflammation, and lipid and lipoprotein metabolism are important metabolic processes that meet the bioenergetic needs of microglia to perform protective functions [ 3 , 10 ]. Hence, as an important fuel for neurons in the brain, lipid metabolism has been associated with various NDs [ 11 ]. Previous placebo-crossover trials have shown that medium-chain triglycerides can significantly improve cognitive function and lipid metabolomics in patients with early to intermediate stage AD [ 12 ]. Both circulating and brain cholesterol levels are inversely linked with motor dysfunction in PD patients [ 13 , 14 ]. However, studies also suggest an elevated PD risk in patients with high cholesterol [ 15 , 16 ], and evidence from clinical trials indicates that omega-3 unsaturated fatty acids supplementation fails to improve AD patient outcomes, including cognitive, functional, and depressive symptom [ 17 ].

Given the conflicting research findings, the mechanistic links between specific lipid metabolism parameters and ND disease development remains controversial and warrants further investigation. To address this knowledge gap, this study employed Mendelian randomization to explore potential genetically defined causal relationships between lipid metabolism parameters and ND within the context of the liver‒brain axis. This approach allowed us to overcome some limitations of observational studies and provided more robust evidence for causal associations [ 18 , 19 ]. Understanding these connections could have significant clinical implications for disease prevention and treatment strategies.

Methods

Study design

A two-sample Mendelian randomization (MR) was used to examine potential causal relationships between lipid metabolism and NDs. A schematic representation of the study and the analysis pipeline is shown in Fig.  1 . Following the three fundamental MR assumptions, our study was structured as follows:

Study overview. Schematic representation of the Mendelian randomization (MR) study design and analysis pipeline

Genetic Instrument Selection (Assumption 1: Relevance) Six well-established, clinically relevant lipid metabolites were selected (apolipoprotein A1 (ApoA), apolipoprotein B (ApoB), total cholesterol (CHOL), high-density lipoprotein cholesterol (HDLC), low-density lipoprotein cholesterol (LDLC), and triglycerides (TG)) as exposures based on their potential roles in NDs, as these metabolites are routinely measured in clinical settings and have been previously implicated in neurological disorders. Genetic variants from Genome-Wide Association Studies (GWAS) databases served as instrumental variables (IVs), those with strong associations to lipid metabolism were selected, rather than direct measurements from blood. The selection process applied strict significance thresholds ( P  < 5 × 10⁻⁸) and linkage disequilibrium (LD) clumping (r 2  < 0.001, clumping window = 10,000 kb) to ensure independent signals. Single-nucleotide polymorphisms (SNPs) with F-statistic values > 10 were extracted from the GWAS datasets, followed by allele harmonization. The selection of instrumental variables (IVs) was performed independently for each unique exposure-outcome combination, ensuring that each pair had its own distinct set of genetic instruments.

The upper part illustrates the MR assumptions and potential relationships between SNPs, exposures (lipid metabolism measures), and outcomes (neurodegenerative diseases), with confounders (obesity and inflammation). The solid arrows indicate assumed causal relationships, whereas the dashed lines with red X symbols represent associations that should not exist to satisfy MR assumptions. The lower part outlines the step-by-step analysis process, from the GWAS dataset to multivariable MR analysis. Key elements include instrument selection criteria, primary and sensitivity MR analyses, and the final multivariable analysis adjusting for potential confounders. Specific lipid biomarkers, neurodegenerative diseases, and confounding factors are listed. MR: Mendelian randomization; SNPs: single-nucleotide polymorphisms; GWAS: genome-wide association studies; LD: linkage disequilibrium.

Confounder Check (Assumption 2: No Horizontal Pleiotropy) To address potential pleiotropy, our analysis incorporated several key confounding factors: body mass index (BMI), blood glucose, and inflammatory markers (C-reactive protein [CRP] and IL-17). Horizontal pleiotropy was evaluated via the MR‒Egger intercept test, and heterogeneity across causal effects was examined via the Cochran’s Q test and the Rucker Q’ test. To address any potential residual confounding from incomplete covariate adjustment, sensitivity analyses using MR-Egger regression could detect and account for potential pleiotropy arising from unmeasured confounders. The analysis also included albumin and gamma-glutamyl transpeptidase (GGT) as markers of liver function. All the data were sourced from publicly available GWAS databases.

Outcome Validation (Assumption 3: Exclusion Restriction) By focusing on these key lipid parameters as exposures, this study aimed to provide a targeted analysis of their potential causal relationships with four NDs as outcomes: Alzheimer's disease (AD), amyotrophic lateral sclerosis (ALS), Parkinson's disease (PD), and multiple sclerosis (MS). The inverse variance weighted (IVW) method assessed causal effects between lipid parameters and NDs. All source GWAS datasets incorporated key adjustments to minimize potential confounding: the UK Biobank lipid parameters data were adjusted for age, sex, and the first 20 genetic principal components; the ND outcome datasets included similar demographic and genetic adjustments (age, sex, and principal components for ALS and PD; age and principal components for AD; sex and principal components for MS). Supporting analyses via MR Egger, weighted median, simple mode, and weighted mode were conducted. Sensitivity analyses incorporated the MR multidirectional residual sum with outliers (MR-PRESSO) to correct for outliers. Reverse MR was conducted by setting ND as the exposure and lipid metabolism parameters as the outcomes. A validation analysis was performed using a separate ND dataset. To ensure the use of the most up-to-date GWAS data, the recently published large-scale metabolite GWAS covering European and Asian population was incorporated as a validation dataset [ 20 ]. The associations between the lipid metabolites and NDs were reanalyzed using these updated summary statistics. To elucidate the potential influences of mediating factors on causal inference, a multifactorial MR analysis was conducted, considering four potential confounders: BMI, blood glucose, CRP, and IL-17. BMI and blood glucose were selected because of their likely associations with lipid metabolism abnormalities, whereas CRP and IL-17 were chosen because of their potential relevance to ND occurrence.

Data sources

European ancestry GWAS databases were utilized, prioritizing the most recent and comprehensive datasets. Lipid metabolism parameters were obtained from the UK Biobank. Data for AD, PD, and MS were sourced from various consortia and databases, including the International Genomics of Alzheimer's Project, the International Parkinson's Disease Genomics Consortium, the International Multiple Sclerosis Genetics Consortium, the FinnGen consortium, and the UK Biobank. ALS data were obtained through a comprehensive meta-analysis of GWAS published in recent literature. Table 1 presents details of the validation and mediator datasets. For AD, clinically diagnosed cases were included based on established diagnostic criteria. ALS cases were classified according to El Escorial Criteria. MS cases were clinically confirmed through standard diagnostic procedures. The PD dataset included both idiopathic and genetic forms of the disease, as this distinction was not made in the original GWAS data. Detailed demographic information for the study populations of AD, PD, MS, and ALS can be found in Supplementary file 1.

Table 1 Data sources for exposures, neurodegenerative diseases, and confounders of interest Trait GWAS accession number N subjects N SNPs Population PMID Exposure  ApoA ebi-a-GCST90092808 115,082 11,590,399 European 35,213,538 ebi-a-GCST90301945 136,016 13,389,637 European, Asian 38,448,586  ApoB ebi-a-GCST90092809 115,082 11,590,399 European 35,213,538 ebi-a-GCST90301946 136,016 13,389,637 European, Asian 38,448,586  CHOL ebi-a-GCST90092985 115,082 11,590,399 European 35,213,538 ebi-a-GCST90302075 136,016 13,389,637 European, Asian 38,448,586  HDLC ebi-a-GCST90092822 115,082 11,590,399 European 35,213,538 ebi-a-GCST90302133 136,016 13,389,637 European, Asian 38,448,586  LDLC ebi-a-GCST90092883 115,082 11,590,399 European 35,213,538 ebi-a-GCST90302014 136,016 13,389,637 European, Asian 38,448,586  TG ebi-a-GCST90092992 115,082 11,590,399 European 35,213,538 ebi-a-GCST90302076 136,016 13,389,637 European, Asian 38,448,586 Outcome  AD ieu-b-5067 488,285 12,321,875 European 35,565,667 ukb-b-14699 423,738 9,851,867 European -  ALS ebi-a-GCST90027163 138,086 10,426,600 European 34,873,335 ieu-a-1085 36,052 7,740,345 European 27,455,348  PD ieu-b-7 482,730 17,891,936 European 31,701,892 ukb-b-6548 422,464 9,851,867 European -  MS ieu-b-18 115,803 6,304,359 European 31,604,244 ukb-b-17670 462,933 9,851,867 European - Confounder  BMI ebi-a-GCST90029007 532,396 11,973,091 European 29,892,013  Blood glucose ebi-a-GCST90092819 114,870 11,590,399 European 35,213,538  CRP ebi-a-GCST90029070 575,531 10,713,245 European 35,459,240  IL-17 ebi-a-GCST90019425 10,708 15,568,168 European 33,328,453  GGT ebi-a-GCST90014004 389,672 10,783,692 European 34,017,140  Albumin ebi-a-GCST90018945 315,268 19,053,186 European 34,594,039 For the two datasets corresponding to the same exposure/outcome variable, the first dataset was used for the main analysis, while the second was used for validation analysis GWAS Genome-Wide Association Study, SNPs Single Nucleotide Polymorphisms, ApoA Apolipoprotein A1, ApoB Apolipoprotein B, CHOL Total Cholesterol, HDLC High-Density Lipoprotein Cholesterol, LDLC Low-Density Lipoprotein Cholesterol, TG Triglycerides, AD Alzheimer's Disease, ALS Amyotrophic Lateral Sclerosis, PD Parkinson's Disease, MS Multiple Sclerosis, BMI Body Mass Index, CRP C-Reactive Protein, IL-17 Interleukin-17, GGT Gamma-glutamyl transferase, PMID PubMed ID

Data sources for exposures, neurodegenerative diseases, and confounders of interest

For the two datasets corresponding to the same exposure/outcome variable, the first dataset was used for the main analysis, while the second was used for validation analysis

GWAS Genome-Wide Association Study, SNPs Single Nucleotide Polymorphisms, ApoA Apolipoprotein A1, ApoB Apolipoprotein B, CHOL Total Cholesterol, HDLC High-Density Lipoprotein Cholesterol, LDLC Low-Density Lipoprotein Cholesterol, TG Triglycerides, AD Alzheimer's Disease, ALS Amyotrophic Lateral Sclerosis, PD Parkinson's Disease, MS Multiple Sclerosis, BMI Body Mass Index, CRP C-Reactive Protein, IL-17 Interleukin-17, GGT Gamma-glutamyl transferase, PMID PubMed ID

Statistical analysis

All analyses used R version 4.2.2 (RStudio, USA). The two-sample MR package for Mendelian randomization and the MVMR package for multifactorial MR were employed. To account for multiple testing across six lipid parameters and ND, Bonferroni correction ( P  < 0.0083 [0.05/6]) was applied. This stringent threshold aims to minimize false-positives. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines, ensuring transparent and comprehensive reporting, methodological rigor, and facilitating result interpretation.

Results

Associations between lipid parameters and NDs

Firstly the potential association between hepatic function and NDs was investigated by analyzing serum albumin and GGT. No significant associations were observed (Supplementary Table 1). Then the relationships between six lipid metabolism parameters were illustrated, namely, ApoA, ApoB, CHOL, HDLC, LDLC, and TG, and the occurrence of ND in Fig.  2 . IVW analysis revealed a significant association between high CHOL levels and ALS incidence (OR = 1.10, 95% CI: 1.03–1.18, P  = 4.23 × 10⁻ 3 ). This association was consistently validated via two additional MR methods: weighted median (OR = 1.11, P  = 4.03 × 10⁻ 2 ) and weighted mode (OR = 1.11, P  = 4.78 × 10⁻ 2 ) (Supplementary Table 2). Additionally, heightened LDLC levels were significantly correlated with the incidence of ALS (OR = 1.10, 95% CI: 1.03–1.17, P  = 3.28 × 10⁻ 3 ), with consistent validation across the weighted median (OR = 1.12, P  = 9.78 × 10⁻ 3 ), weighted mode (OR = 1.12, P  = 1.23 × 10⁻ 2 ), and MR Egger methods (OR = 1.12, P  = 2.27 × 10⁻ 2 ) (Supplementary Table 2). Furthermore, a potential correlation was observed between ApoB and increased ALS incidence (OR = 1.08, 95% CI: 1.02–1.15, P  = 0.016), although the significance diminished after Bonferroni correction. Conversely, a potential inverse association between ApoB and AD incidence was noted (OR = 0.91, 95% CI: 0.85–0.99, P  = 0.024), yet this association also did not survive Bonferroni correction. The number of SNPs selected at each step of the IV selection process (e.g., initial SNPs, after LD clumping, after harmonization, after confounder checks) was shown in Supplementary Table 3.

Forest plot of lipid markers and neurodegenerative disease associations

The plot displays odds ratios (ORs) with 95% confidence intervals (CIs) for the effects of ApoA, ApoB, CHOL, HDLC, LDLC, and TG on Alzheimer's disease (AD), amyotrophic lateral sclerosis (ALS), multiple sclerosis (MS), and Parkinson's disease (PD). Each row represents a specific lipid‒disease association, with the number of SNPs used as instruments (Nsnp), OR, 95% CI, and P value. Significant associations with P  < 0.05 and P  < 0.0083 (0.05/6) are highlighted with single and double asterisks, respectively. The forest plot on the right visually represents the ORs and their confidence intervals. The orange squares highlight associations with P  < 0.05, and the red squares highlight associations with P  < 0.0083.

In the context of PD and MS, no discernible associations were identified with lipid metabolism markers. Despite the use of alternative MR analysis methods, including MR Egger, weighted median, simple mode, and weighted mode, and even without multiple comparison corrections, no statistically significant results emerged (Supplementary Table 2).

Importantly, in the assessment of heterogeneity via the Egger and IVW methods, as well as the detection of horizontal pleiotropy via the Egger method, many of the aforementioned MR results displayed heterogeneity and/or pleiotropy (Supplementary Table 4). For example, the relationship between ApoB and AD exhibited significant heterogeneity (Egger Q = 88.3, P  = 3.83 × 10⁻ 5 ). Importantly, in the evaluation of the associations between CHOL and LDLC with ALS, no significant heterogeneity was detected (CHOL for ALS: heterogeneity Egger P = 0.203; LDLC for ALS: heterogeneity Egger P = 0.355), with no pleiotropy detected (CHOL for ALS: pleiotropy Egger P = 0.252; LDLC for ALS: pleiotropy Egger P = 0.610), supporting the reliability of the aforementioned results.

Sensitivity analysis

MR-PRESSO were applied to identify and exclude outliers with heterogeneity or pleiotropy, reanalyzing the data to yield the outlier-corrected results presented in Fig.  3 A. The results revealed that the association between CHOL and ALS remained significant (OR = 1.10, 95% CI: 1.03–1.18, P  = 6.14 × 10⁻ 3 ), and similarly, the relationship between LDLC and ALS remained significant (OR = 1.10, 95% CI: 1.03–1.17, P  = 5.56 × 10⁻ 3 ).

Mendelian randomization results with outlier correction and secondary dataset validation

(A) Forest plot showing outlier-corrected OR results from MR-PRESSO analysis for associations between lipid metabolism markers and neurodegenerative diseases. (B) Forest plot displaying validation dataset-OR results when an independent dataset was used to validate associations between lipid metabolism markers and neurodegenerative diseases. Both plots present odds ratios (ORs) with 95% confidence intervals (CIs) for the effects of ApoA, ApoB, CHOL, HDLC, LDLC, and TG on AD, ALS, MS, and PD. Each row represents a specific lipid‒disease association, including the OR, 95% CI, and P value. Significant associations with P  < 0.05 and P  < 0.0083 (0.05/6) are highlighted with single and double asterisks, respectively. The forest plots visually depict the ORs and their confidence intervals, allowing easy comparison of effect sizes and significance across different lipid‒disease relationships in both the outlier-corrected and validation analyses. The orange squares highlight associations with P  < 0.05, and the red squares highlight associations with P  < 0.0083.

Subsequently, by employing a reverse MR approach to investigate potential biases induced by reverse causation, ALS was considered as the exposure factor, exploring its impact on CHOL and LDLC. The results indicated a lack of significant effects of ALS on CHOL and LDLC (CHOL: OR = 0.93, 95% CI: 0.79–1.10, P  = 0.411; LDLC: OR = 0.94, 95% CI: 0.82–1.08, P  = 0.392, IVW approach), with no observed heterogeneity or pleiotropy (Supplementary Table 5).

Finally, to validate these findings, the latest independent metabolite GWAS dataset was used and the associations between CHOL and LDLC with ALS remained significant in this validation analysis (Fig.  3 B), reinforcing result reliability. The overall results closely mirrored the primary analysis, with the association between CHOL and ALS remaining significant (OR = 1.11, 95% CI: 1.03–1.20, P  = 7.28 × 10⁻ 3 ), and similarly, the relationship between LDLC and ALS maintaining significance (OR = 1.08, 95% CI: 1.01–1.16, P  = 3.67 × 10⁻ 2 ). These sensitivity analyses enhanced the robustness of the primary results, providing additional support for the observed associations between lipid metabolism markers and ALS.

Multivariable MR

Multivariable MR-IVW analysis assessed CHOL and LDLC effects on ALS while adjusting for potential confounders. Four mediator factors were included, including BMI and glucose, both of which are associated with metabolism, and CRP and IL-17, which are linked to inflammatory responses. Throughout the adjustment for these factors, the direct effects of CHOL and LDLC on ALS remained unaltered (Fig.  4 ). This MVMR-IVW analysis strengthened the findings by accounting for confounders and confirming independent lipid-ALS associations.

Multifactorial Mendelian randomization results

(A) Forest plot showing multivariable Mendelian randomization results for the association of CHOL with ALS risk, adjusting for potential confounders. (B) Forest plot displaying multivariable Mendelian randomization results for the association of LDLC with ALS risk, adjusted for potential confounders. Both plots present odds ratios (ORs) with 95% confidence intervals (CIs) for CHOL or LDLC and each confounder (BMI, blood glucose, CRP, and IL-17) in separate models. Each row shows the number of SNPs used, OR, 95% CI, and P value for the lipid marker and the confounder. Statistically significant associations ( P  < 0.0083) are indicated by double asterisks and highlighted in red. The forest plots visually represent the ORs and their confidence intervals, demonstrating that both CHOL and LDLC maintain significant positive associations with ALS risk even after adjusting for these potential confounders, whereas the confounders themselves are not significantly associated with ALS risk.

Discussion

This two-sample MR study found associations between CHOL and LDLC levels and ALS risk. The causal associations of higher CHOL and LDLC levels with ALS were confirmed, suggesting potential genetically defined predictors of ALS.

The liver‒brain axis has emerged as a crucial pathway in the pathogenesis of various diseases, including ND [ 21 ]. Recent research has provided substantial evidence for the role of the liver‒brain axis in ND. The liver can regulate brain Aβ levels [ 22 ], while hepatic epoxide hydrolase activity affects brain Aβ and cognition [ 23 ]. The liver‒brain axis also influences body metabolism and behavior, especially by linking chronic liver inflammation to alterations in neural transmission and behavioral changes [ 21 , 24 ]. Mounting evidence suggests a significant role for the liver in ND, with lipid metabolism serving as a key mediator. Lipid parameters have been implicated in the pathophysiology of several neurodegenerative conditions. In AD, altered cholesterol metabolism and oxidized lipids correlate with disease severity [ 25 ]. For ALS, lipid dysregulation has been linked to neuroinflammation and motor neuron degeneration [ 26 ]. In PD, specific lipid species alterations have been identified in both serum and CSF, suggesting potential biomarkers and therapeutic targets [ 27 ]. Research on MS has revealed associations between lipid peroxidation products and disease activity, as well as changes in cholesterol metabolites reflecting myelin damage [ 28 ]. Additionally, Liver X Receptors function as oxysterol sensors in the liver, receiving oxysterol signaling to regulate cholesterol transport and lipid synthesis for metabolic homeostasis. These oxysterols have emerged as critical mediators in neurodegenerative pathogenesis, mediating processes including amyloid-beta accumulation in AD [ 29 ], metabolic alterations in PD development [ 30 ], neuroinflammation in MS [ 31 ], and oxidative stress in ALS [ 32 ]. The accumulating evidence across these ND underscores the importance of further investigating the causal relationships between lipid metabolism parameters and neurodegeneration, potentially opening new avenues for diagnosis, prognosis, and therapeutic interventions.

Results show that elevated levels of CHOL and LDLC are both associated with ALS. The pathogenesis of ALS involves multiple biological processes, including protein misfolding, neuroinflammation, oxidative stress, and metabolic imbalance [ 33 – 35 ]. ALS patients typically exhibit a hypermetabolic state, with high metabolism associated with poorer prognosis [ 36 , 37 ]. Lipid metabolism, as a crucial energy source for neuronal function, is likely an important risk factor for ALS. Research has identified several ALS-related genes and protein targets associated with cholesterol metabolism. For example, RNA/DNA-binding protein fused in sarcoma (FUS) and ATPase valosin-containing protein (VCP)/p97 regulate the rate-limiting enzyme HMGCR in cholesterol metabolism [ 38 , 39 ], whereas ATXN2 and TDP-43 directly affect cholesterol levels [ 40 , 41 ]. However, there have been controversial conclusions. Some studies have shown an increase in TC levels in ALS patients; however, this conclusion may have been biased due to the lack of a control group or the small sample size [ 42 – 44 ]. Even in studies with control groups and large sample sizes, scholars have reached varying conclusions regarding changes in CHOL levels in ALS patients [ 45 – 48 ]. Also, previous studies report conflicting LDLC-ALS associations: small sample meta-analyses found no significant relationship [ 49 ], while MR studies suggested LDLC mediates polyunsaturated fatty acids (PUFAs)-related traits to ALS risk [ 50 ]. These discrepancies may result from lacking control groups, small samples, and varying methods in collection, diagnosis, and analysis. Additionally, confounding factors and reverse causality can bias observational studies. While large-scale randomized clinical trials (RCTs) could address these issues, their development is often limited by practical constraints. By employing genetic variables as substitutes, MR revealed the associations between CHOL and LDLC levels and ALS and avoid the influence of bias on the conclusions. Consistent with the conclusions of this study, recent MR analyses of ALS risk loci also revealed a causal relationship between ALS and high cholesterol levels [ 51 ], which further strengthen the interpretation of the result.

While this study primarily highlighted the significant associations between CHOL/LDLC and ALS, a marginally significant association between ApoB and ALS was also observed. Although this association did not survive Bonferroni correction, it merits attention given ApoB's established role as the principal protein component of LDL particles [ 52 , 53 ]. This finding aligns with the stronger associations between LDLC and ALS, as ApoB serves as a key indicator of LDL particle number and metabolism [ 54 ]. The marginally significant nature of the ApoB association might reflect the complex relationship between different components of lipid metabolism in neurodegeneration. Recent studies have suggested that alterations in lipid metabolism, including changes in ApoB-containing lipoproteins, may contribute to ALS pathogenesis through mechanisms involving cholesterol transport and cellular lipid homeostasis [ 55 ]. However, given the marginal statistical significance, this relationship requires further investigation in larger cohorts and through mechanistic studies to fully understand its biological relevance in ALS pathogenesis.

Findings of this study complement and extend recent Mendelian randomization studies examining metabolite-ND relationships. Notably, Lord et al. [ 56 ] and Huang et al. [ 57 ] conducted comprehensive metabolome-wide MR studies focusing on blood circulating metabolites and AD. Huang et al. investigated 123 circulating metabolites, while Lord et al. examined 19 metabolites previously linked to midlife cognition, both revealing significant associations between various metabolites and AD risk. While this study specifically focused on lipid metabolism parameters and uniquely incorporated multivariable MR analysis to account for confounding factors including BMI, blood glucose, and inflammatory markers. Through this targeted approach, combined with validation using independent datasets and consideration of the liver-brain axis, specific causal relationships were identified between lipid parameters and NDs, especially covering four NDs: AD, PD, ALS, and MS. Metabolome-wide MR studies also identified distinct metabolic profiles in MS, with serine, lysine, and ketone bodies (acetone and acetoacetate) showing risk-increasing effects, while the relationship with lipids appears more complex – total cholesterol and phospholipids showed opposing effects depending on their lipoprotein carriers [ 58 ]. This complexity in lipid metabolism's role across different NDs is further highlighted by the finding of no significant associations between lipid parameters and MS, suggesting disease-specific metabolic pathways. These collective findings underscore the importance of disease-specific approaches when investigating metabolic factors in neurodegeneration, as causal metabolic pathways may vary substantially across different neurodegenerative conditions.

In this study, among the six lipid metabolism parameters, only ApoB showed a near-significant association with AD risk. This finding aligns with the Mendelian randomization study conducted on 329,896 UK Biobank participants, which reported that only ApoB was significantly associated with increased dementia risk [ 59 ]. Interestingly, results revealed potentially opposing relationships for apolipoproteins: elevated ApoA levels and decreased ApoB levels might be linked to AD occurrence. This presents a paradox, as ApoA is typically considered the primary lipoprotein of HDLC with protective effects against atherosclerosis and coronary heart disease [ 52 , 60 ], whereas ApoB, which is predominantly associated with LDLC, is generally viewed as a risk factor for these conditions [ 52 , 53 , 61 ]. Consistent with this study, a recent report from the European Alzheimer's & Dementia Biobank Mendelian Randomization Collaboration identified an association between high HDLC concentrations and increased AD risk [ 62 ]. The proposed mechanism involves high plasma HDLC disrupting homeostasis between plasma and cerebrospinal fluid apoE/apoA1 HDL particles. These conflicting results underscore the current controversy surrounding the relationship between HDLC, its associated apolipoprotein ApoA1, and AD risk, highlighting the need for further clinical data analysis. Additionally, this study revealed contrasting findings for ApoB in AD and ALS patients. Previous research on early-onset AD patients revealed no significant difference in plasma LDLC levels between carriers and noncarriers of APOB coding variants, whereas a strong association between high LDLC levels and early-onset AD was found in noncarriers [ 63 ], suggesting that factors beyond ApoB contribute to plasma LDLC differences in AD. Consequently, the association between ApoB and ND may be influenced by multiple factors, and more definitive evidence is needed before ApoB can be considered a reliable predictor of ND risk.

Findings of specific associations between CHOL/LDLC levels and ALS suggest that therapeutic strategies targeting lipid metabolism may be particularly relevant for ALS. However, this specificity also implies that such approaches may need to be tailored differently for other NDs. The observed lack of genetic associations with PD and MS does not rule out the involvement of lipid metabolism in these conditions but suggests that different metabolic pathways or mechanisms may be involved. This highlights the importance of developing disease-specific therapeutic strategies rather than applying a one-size-fits-all approach to NDs.

In this study, no significant associations between lipid metabolism parameters and MS or PD were observed. MS, which is characterized primarily by neuroinflammation and demyelination, has been linked to alterations in lipid metabolism. As demyelination occurs, cholesterol from neuronal myelin may induce changes in circulating lipids, with clinical cohort evidence suggesting associations between MS progression and elevated levels of LDL, total cholesterol, and triglycerides [ 28 , 64 – 66 ]. However, despite indications that lipid metabolism levels in MS patients may aid in disease assessment, robust evidence supporting a causal relationship between lipid metabolism abnormalities and MS pathogenesis remains elusive. Similarly, while PD research has demonstrated connections to lipid metabolism, with changes in lipid composition potentially inducing protein alterations, such as α-synuclein accumulation, and some PD-associated genes (e.g., GBA and LRRK2) are thought to be involved in lipid metabolic processes [ 67 – 71 ], direct evidence for a causal relationship between lipid metabolism and PD onset is lacking. Results are consistent with other Mendelian randomization studies, which suggest that alterations in lipid metabolism do not play a prominent role in PD etiology [ 72 , 73 ]. These results underscore the complexity of the liver‒brain axis in ND and highlight the need for further research to elucidate the precise mechanisms linking lipid metabolism to ND.

Results also revealed an intriguing pattern: while specific lipid metabolism parameters showed significant associations with NDs, general liver function markers (albumin and GGT) demonstrated no significant correlations. This discrepancy might be explained by the highly specific nature of liver-brain axis interactions in neurodegeneration. The brain, as one of the most lipid-rich organs, requires specific lipid metabolic pathways for myelin maintenance and synaptic function [ 74 , 75 ], rather than depending on general liver synthetic capacity. This specificity is further supported by evidence that neurodegenerative pathologies often involve disruption of specific metabolic pathways rather than global liver dysfunction. For instance, cholesterol metabolism specifically influences synaptic plasticity and neurotransmitter release [ 76 ], and specific lipid-mediated signaling pathways, rather than general liver function, play crucial roles in neuroinflammation and neurodegeneration [ 77 , 78 ]. This pathway specificity explains why alterations in particular lipid parameters, rather than general liver function markers, may serve as more relevant indicators of ND risk and progression.

Strengths and limitation

This study employed multiple MR methods (IVW, MR Egger, weighted median, and weighted mode) to enhance the reliability of findings and assess potential biases such as pleiotropy and heterogeneity. Comprehensive sensitivity analyses were performed, including MR-PRESSO for outlier detection and correction, and reverse MR analysis, which strengthened the robustness of the results. The multivariable MR approach accounted for potential confounding factors including BMI, blood glucose, CRP, and IL-17, minimizing possible bias in causal inference. Large-scale GWAS data from European populations were utilized, providing substantial statistical power for detecting causal relationships. The study adhered strictly to the STROBE guidelines, ensuring transparent and comprehensive reporting of methods and results.

Several limitations of this study should be acknowledged. Although genetic variants from GWAS data were used as instruments, the findings may still be affected by weak instrument bias, requiring careful balance between statistical power and potential pleiotropy in future studies. Second, while our datasets included Asian participants in both primary (2,407 ALS cases and 11,775 controls) and validation analyses, our study population remained predominantly of European genetic background, potentially limiting result generalizability. Genetic variants associated with lipid metabolism and neurodegenerative diseases can vary significantly across different populations due to distinct genetic architectures and environmental factors. Future studies incorporating multi-ethnic GWAS data are needed to validate these findings across diverse populations and establish whether these causal relationships are consistent across different ethnic backgrounds. Further research may be needed to validate these results in diverse populations. Third, the cross-sectional nature of the severity data for ND may have limited the ability to identify causal links between lipid-related traits and disease progression. Longitudinal studies may help better elucidate lipid metabolism's impact on disease progression and validate cross-sectional severity measures. Fourth, the potential lack of harmonization in analytical methods used for measuring lipid parameters across different GWAS datasets may introduce methodological heterogeneity. As these data were collected from various sources and potentially analyzed using different approaches, systematic variations in the measurements of cholesterol, triglycerides, LDL, HDL, ApoA and ApoB levels might exist. This methodological issue should be considered when interpreting the results. Additionally, the idiopathic and genetic forms of PD is unable to distinguish in this analysis, as this differentiation was not available in the original GWAS datasets. This combined analysis of different PD subtypes may have masked potential subtype-specific associations between lipid metabolism parameters and PD risk, suggesting that future studies should aim to analyze these forms separately when such data becomes available. Although the source GWAS datasets included adjustments for key demographic and genetic confounders, the degree of adjustment varied across studies. While the MR framework helps protect against confounding through the use of genetic instruments, residual bias from differential adjustment of covariates cannot be completely ruled out. However, sensitivity analyses using MR-Egger and multivariable approaches suggest that the main findings are robust to potential unmeasured confounding. While the MR analysis supports a causal role for elevated cholesterol in ALS, MR cannot determine the exact biological pathways driving the observed relationships. Further research using complementary designs can help confirm these findings and elucidate the underlying pathways.

Conclusions

In this two-sample MR study, higher CHOL and LDLC levels were associated with increased ALS risk, suggesting cholesterol as a genetically-linked risk factor for ALS and providing rationale for lipid profile monitoring and lipid-modifying interventions in ALS management. Additionally, potential associations between ApoB and both ALS and AD risk were observed, but more exploration is required to clarify the biological pathways involved before clinical translation.

Supplementary Information

Supplementary Material 1.

Supplementary Material 2.

Acknowledgements

Not applicable.

Abbreviations

Neurodegenerative diseases

Alzheimer's disease

Parkinson's disease

Multiple sclerosis

Amyotrophic lateral sclerosis

Mendelian randomization

Instrumental variables

Apolipoprotein A1

Apolipoprotein B

Total cholesterol

High-density lipoprotein cholesterol

Low-density lipoprotein cholesterol

Triglycerides

Gamma-glutamyl transferase

Genome-Wide Association Studies

Single-nucleotide polymorphisms

Linkage Disequilibrium

Inverse variance weighted

Strengthening the Reporting of Observational Studies in Epidemiology

MR multidirectional residual sum with outliers

Odds ratios

Confidence intervals

Fused in sarcoma

ATPase valosin-containing protein

3-Hydroxy-3-methylglutaryl-coenzyme A reductase

Polyunsaturated fatty acids

Randomized clinical trials

Body mass index

C-reactive protein

Interleukin-17

Authors’ contributions

ZW and ZY designed the project, analyzed most of the data, and wrote the manuscript. GS assisted in project design and manuscript revision. JZ and DZ designed the project, supervised the study, and revised the manuscript. All the authors read and approved the final manuscript.

Authors’ information (optional)

Not applicable.

Funding

Not applicable.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Zeyu Wang and Zixiao Yin share co-first authorship.

Contributor Information

Dong Zhang, Email: zhangd2010@hotmail.com.

Jianguo Zhang, Email: zjguo73@126.com.

References

Associated Data

Supplementary Materials

Supplementary Material 1.

Supplementary Material 2.

Data Availability Statement

No datasets were generated or analysed during the current study.