3-Hydroxybutyrate, a metabolite in sustaining neuronal cell vitality: a mendelian randomization and in vitro experimentation

Abstract

Background

Recent research has implicated mitochondrial DNA copy number (mtDNA-CN) and Tau protein levels in the blood as potential biomarkers for early Alzheimer’s disease (AD) risk assessment, correlating with metabolite profiles. However, intermediary metabolites mediating these associations remain elusive.

Methods

Employing a two-sample and a mediation Mendelian randomization (MR) analysis of the IEU OpenGWAS database, involving 383,476 participants from a genome-wide association study (GWAS) and an exome-wide association study (ExWAS), we identified intermediary metabolites linking mtDNA-CN and Tau.Meanwhile, the effects of mediating metabolites on HT22 cell viability and its mitochondrial morphology were also assessed in conjunction with in vitro experiments.

Results

Our study revealed an association of mtDNA-CN on Tau (OR = 3.102, 95% CI: 1.016–9.472, P = 0.047), as well as on other 31 metabolites such as 3-Hydroxybutyrate (3HB), Docosahexaenoic acid (DHA), Acetate, Albumin, Apolipoprotein A-I (APOA1), and so on. Notably, 3HB was further implicated in a relationship with Tau (OR = 6.030, 95% CI: 1.054–34.491, P = 0.043), acting as a mediator between mtDNA-CN and Tau. In vitro experiments demonstrated that 3HB positively sustained HT22 cell viability by MTT assay and mitigated mitochondrial swelling under low glucose conditions, as observed via HIS-SIM. In Western blot (WB) and quantitative real-time PCR (qPCR) assays, phosphorylation levels of Tau at serine 262 (p-Tau262) and serine 396 (p-Tau396) were tended to decline following 3HB intervention. Additionally, a positive correlation was identified between 3HB concentration and mtDNA-CN.

Conclusions

These findings underscore the potential of 3HB as a biomarker and mediator in early AD risk assessment. Moreover, 3HB’s ability to enhance cell viability, maintain mitochondrial morphology, decrease phosphorylated Tau protein expression and increase mtDNA-CN under stressful conditions, suggesting its therapeutic potential in improving the imbalance of energy metabolism in the AD brain.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12986-025-00960-x.

Introduction

With the increasing trend of global ageing, Alzheimer’s disease (AD), a neurodegenerative disease with a high prevalence in the elderly population, has become an urgent public health challenge [1]. Despite significant advances in etiological exploration, pathomechanisms, and drug development over more than a century, the exact pathogenesis of AD remains incompletely understood, which constrains the development of effective intervention strategies [2–4]. This dilemma highlights the importance of early diagnosis and preventive treatment of AD. In recent years, the scientific community has gradually turned its attention to the early pathogenetic features of AD, in particular disorders of brain energy metabolism [5, 6] and mitochondrial dysfunction [7, 8], which are considered new directions in research on early diagnosis and treatment of the disease. In clinical practice, the risk of early onset of AD has been estimated by detecting mitochondrial DNA copy number (mtDNA-CN) [9] and Tau protein levels [10, 11] in the blood, providing a new perspective on biomarkers for early diagnosis.

Mitochondria, as the core of cellular energy metabolism, convert glucoses, amino acids, lipids, and their derivatives into adenosine triphosphate (ATP) to safeguard the energy needs of the brain through the processes of oxidative phosphorylation (OXPHOS), the tricarboxylic acid (TCA) cycle, and the electron transport chain (ETC) [12]. During the pathological process of AD, the metabolic pattern of neuronal cells undergoes reprogramming, which manifests itself as a decrease in glucose utilisation, activation of alternative energy pathways, and the use of alternative energy substrates (e.g., ketone bodies, fatty acids, and amino acids, etc.) to maintain cell viability and function [13, 14]. However, these metabolic adjustments may disrupt the dynamic balance between mitochondrial fission and fusion, triggering structural abnormalities in mitochondria, exacerbating dysfunction, and creating a vicious cycle [15]. Notably, mitochondrial DNA (mt-DNA), as the genetic material of mitochondria, encodes key proteins directly involved in energy metabolism, and its copy number variations reflect mitochondrial function and cellular metabolic status [16]. Studies claimed that a decrease in mtDNA-CN is an early warning signal for an increased risk of AD [17], and this metabolic association appears to be predominantly influenced by Tau pathology, as opposed to Aβ deposition [18]. Subsequent studies have further identified the association of mtDNA-CN on P-tau181 regarded as the blood biomarker of AD [9], which supports the present study to analyse the association of mtDNA-CN on the AD blood biomarker-Tau.

Variations in metabolite levels, arising as products or substrates of biochemical reactions within the human body, not only signify disruptions in the metabolic pathways associated with AD [19] but also point to the feasibility of utilising dietary and lifestyle interventions to prevent AD [20, 21]. One study identified 14 metabolites (e.g., phenylalanylserine, androsterone sulfate, succinylcarnitine, etc.) out of 422 metabolites associated with AD risk, revealing the important role of the peptide pathway, lipid pathway, and carbohydrate metabolism involved in these metabolites in the pathogenesis of AD [22]. Another study identified four AD-related metabolites, linking low Glutamine levels to impaired cognition, while suggesting that fish oil intake enhances blood levels of 22:6, docosahexaenoic acid, which is beneficial for cognitive function [23].

Although extensive research on metabolite-AD prevalence associations exist, the link between metabolites and AD blood biomarkers (mtDNA-CN, Tau) remains elusive, necessitating deeper investigation. Hence, we employed a mediation Mendelian randomization (MR) approach, leveraging large-scale GWAS summary statistics, to identify genetic instruments and screen 807 blood metabolites for those associated with these biomarkers. Our goal was to uncover potential early diagnostic indicators for AD interventions. Additionally, we assessed the impact of these mediator metabolites on neuronal cell viability and mitochondrial morphology through in vitro experiments, aiming to inform potential therapeutic strategies of the prevention for AD.

Materials and methods

Study design and data sources

A two-sample and a mediation MR approach were carried out to identify the mediating metabolites among 807 metabolites that exhibit associations with mtDNA-CN and Tau. Moreover, the study also established glucose deprivation (GD) and low glucose supply (LG) HT22 cell models to observe the effects of the mediating metabolites on neuronal cells and their mitochondria. The flowchart of the study is shown in Fig. 1. In MR analysis, genetic instruments for the mtDNA-CN, 807 metabolites, and Tau were obtained from the IEU OpenGWAS database (https://gwas.mrcieu.ac.uk/) [24] (Supplementary Table S1). The summary data of mtDNA-CN were obtained from a GWAS and ExWAS study involving 383,476 participants, which developed an automatic mitochondrial copy (AutoMitoC) method to ascertain 71 loci of mtDNA-CN. Subsequently, the functions of the mitochondria involved in these loci were analysed either [25]. The data of 807 metabolites were collected from human blood encompassing glucose, fatty acids, pyruvate, amino acids, and so on [26, 27]. The data of Tau were derived from a GWAS study of Tau protein levels measured in 1000 human plasma samples [28]. The original research that generated the data obtained ethical approval and participant consent, so using summary-level data in our analysis does not require ethical approval.

Fig. 1.

The flowchart of the study. mtDNA-CN, mitochondrial DNA copy numbe; MR, Mendelian Randomization; IVW, inverse variance weighted; MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; GD, glucose deprivation; LG, low glucose; 3HB, 3-Hydroxybutyrate

Mendelian randomization

Statistical analyses were performed using the ‘TwoSampleMR’ package in R version 4.3.2 to screen for significant (P < 5 × 10^− 8) and independently uncorrelated (r² < 0.001, kb = 10000) Single Nucleotide Polymorphisms (SNPs) as a genetic instrument [29, 30]. Inverse variance weighted (IVW) was used as the primary MR analysis approach, with weighted median and other methods as supplements [31]. We removed palindromic single nucleotide variants (SNVs) with intermediate allele frequencies and then used the PhenoScanner (http://www.phenoscanner.medschl.cam.ac.uk/) [32] and LDlink (https://ldlink.nih.gov/) [33] to validate that the selected SNPs were not associated with confounders. Thus, the assessment of the causal correlation between exposure and outcome factors would be more reliable. Specifically, we evaluated the interrelationships among mtDNA-CN, 807 metabolites, and Tau through two-sample MR. The results of MR analysis were expressed as P-value, Odds Ratio (OR), and 95% Confidence Interval (CI). P-value < 0.05 is recognised as statistically significant, which means there was suggestive evidence of a potential causal relationship between exposure and outcome [34]. To further validate the reliability of the statistical analysis results, heterogeneity testing and pleiotropy testing were used, in which the pleiotropy testing was mainly performed by the MR-Egger method [35] – [36]. After preliminary two-sample MR analysis, potential associations between mtDNA-CN, metabolites, and Tau were identified, and the metabolites in which there were associations with mtDNA-CN and Tau were recognised as mediators. To reveal the metabolites mediated between mtDNA-CN and Tau, we employed mediation analysis to calculate their direct, indirect, total, and mediated effect values [37].

Cell culture

The standard growth medium for mouse hippocampal neuronal (HT22) cells consisted of basal DMEM culture medium (C11995500BT, containing 4.5 g/L D-glucose, Gibico), 10% fetal bovine serum (FSP500, Excell) and 1% penicillin-streptomycin (15140122, Gibco), and cells were stored at 90% humidity, 37.0 °C temperature, and 5% CO2 concentration.

Glucose deprivation and 3HB supply

To assess the effect of 3HB on neuronal cells, a GD HT22 cell model was established in the research [38, 39]. HT22 cells were cultured in 96-well plates (3599, CORNING) at an initial density of 4000 cells/well for 24 h. After 24 h, the standard growth medium (containing 4.5 g/L D-glucose) was replaced with glucose-free DMEM culture medium (11966025, without glucose, Gibico), at the same time, 3HB (HY-113378, MedChemExpress) with different concentrations (0, 5, 10, 20, 40, 80, 160 µmol/L) was also used for treating the cells for different durations (6, 12, 18, 24, 30 h). Secondly, this study was further designed to investigate the effects of 3HB on a LG HT22 cell model. HT22 cells were cultured in 96-well plates at an initial density of 4000 cells/well for 24 h. After 24 h of incubation, the standard growth medium (containing 4.5 g/L D-glucose) was replaced with the DMEM culture medium containing 6.25 mmol/L concentration of glucose (composed of glucose-containing DMEM and glucose-free DMEM formulated in a 1:3 ratio) and simultaneously treated with 3HB (80, 160, 320 µmol/L) for different durations (12, 24, 48 h). The control group was constituted by DMEM culture solution (C11965500BT, containing 4.5 g/L glucose, Gibico) that was administered with an equivalent volume containing a glucose concentration of 25 mmol/L.

MTT assay

HT22 cell viability was evaluated with the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) reagent [40]. Upon completion of the interventions, cells were treated with MTT (5 mg/ml in PBS) for 4 h using established protocols. Later, MTT was discarded, DMSO was added, and the absorbance at 570 nm was measured by an microplate reader (N19290-01, Thermo Fisher) after shaking and mixing. Data were analysed and plotted using GraphPad Prism 8.

Microscopic observation and Immunofluorescence

To observe the effect of 3HB on the morphology and structure of HT22 cells, we utilised a CKX53 Inverted Phase Contrast Microscope (IE41345, OLYMPUS, Japan) equipped with a 20x objective lens. Subsequently, we utilised a High Intelligent and Sensitive Microscope (HIS-SIM) (Guangzhou Computational Super-resolution Biotech Co., Ltd.) to observe the mitochondria of cells by immunofluorescent staining. The LG HT22 cell models with 3HB intervention for 24 h were incubated with MitoTracker Green (C1048, Beyotime, Shanghai, China) for 30 min in a confocal petri dish (D29-20-0-N, Cellvis), and then incubated with Hoechst 33,258 (C1018 Beyotime, Shanghai, China) for 10 min. Both incubations were washed with PBS three times. All images were processed and analysed using Image J and Photoshop 21.0.0.

Western blotting

We conducted Western Blotting (WB) experiments. After cell collection, the cells were lysed on ice for 30 min by adding lysis buffer (RIPA lysis buffer: protease inhibitor: phosphatase inhibitor = 100:1:1). Following centrifugation at 12,000 rpm at 4 °C for 15 min, the supernatant was collected, and the total protein concentration was measured using a Bicinchoninic Acid (BCA) kit (P0011, Beyotime). All samples were standardized using Loading Buffer and ddH2O. Samples and protein Marker (26617, Thermo Science) were separated by SDS–PAGE and then transferred to PVDF (ISEQ00010, Merck) membranes. The membranes carrying the sample proteins were blocked with a protein-free rapid blocking solution (PS108P, Epizyme) at room temperature for 15 min and then incubated with primary antibodies overnight at 4 °C. After washing three times with TBST buffer for 5 min each, the membranes were incubated with secondary antibodies at room temperature for 1 h. Subsequently, following washing by TBST buffer 3 times for 5 min, the blots were detected with BeyoECL substrate (P0018M, Beyotime) and a Molecular Imager ChemiDoc XRS System (Universal Hood IIS, BIO-RAD).

DNA isolation and qPCR

We conducted Quantitative real-time polymerase chain reaction (qPCR) experiments. Total DNA was extracted using the SevenEasy DNA Extraction Kit (SM260-01, Seven). The concentration of double-stranded DNA (dsDNA) was determined using a microvolume spectrophotometer (NanoPhotometer-NP80, implen, Germany). DNA templates with well purity were used as standards, and a standard curve was generated through gradient dilution to calculate mitochondrial DNA copy numbers. qPCR experiments were performed using TB Green^® Premix Ex Taq™ (RR420A, TaKaRa) in the CFX Connect Real-Time System (CFX Connect, BIO-RAD).

Results

Results of MR

SNPs for being palindromic and ambiguous alleles were deleted. Furthermore, Pleiotropy-exhibiting SNPs were discarded based on the results of MR-Egger. As a result, a total of 1898 SNPs were screened for the causal associations of mtDNA-CN on metabolites, and metabolites on Tau, of which 67 were associated with mtDNA-CN, 1685 were associated with metabolites, and 146 were associated with Tau (Supplementary Table S2). The mtDNA-CN was associated with 31 metabolites, which included 3-Hydroxybutyrate (3HB), Docosahexaenoic acid (DHA), Acetate, Albumin, Apolipoprotein A-I (APOA1), Citrulline, Concentration of large HDL particles (CLHDLP), Concentration of medium HDL particles (CMHDLP), and so on. Among these 31 metabolites, 18 metabolite data were able to extract validated instrumental variables for assessing the causal associations of metabolites with Tau proteins, and the P-values of the heterogeneity testing and pleiotropy testing were generally higher than 0.05, suggesting that the heterogeneity was not significant and horizontal pleiotropy did not exist (Supplementary Table S3; Supplementary Table S4). Notably, one of the metabolites that were causally associated with both mtDNA-CN and Tau was 3HB (Supplementary Fig. S1), as shown in the forest plot (Fig. 2). The results of the exploration of the mediator showed that there was a causal association between mtDNA-CN and Tau protein (OR = 3.102, 95% CI: 1.016–9.472, P = 0.047), in which 3HB played a potential mediating role, with a mediating effect of 1.599 (Supplementary Table S5). 3HB had an indirect effect of 0.160 on Tau, and mtDNA-CN had a direct effect of 1.132 on Tau. Consequently, the total effect on Tau via 3HB was 1.292. 3HB accounted for 12.4% of the total effect, and mtDNA accounted for 87.6% of the total effect on Tau. mtDNA had a significant effect on Tau, and 3HB acted as a partial mediator between mtDNA-CN and Tau.In the association analyses, there was no heterogeneity (Heterogeneity > 0.05) and no horizontal pleiotropy (MR-Egger intercept > 0.05), indicating stable causal associations between mtDNA-CN, 3HB, and Tau.

Fig. 2.

Forest plot of the MR estimates used IVW of mtDNA-CN on metabolites and metabolites on Tau. (A) Causal association of mtDNA-CN on metabolites; (B) Causal association of metabolites on Tau. Pval < 0.05 was considered significant. (C) Scater plot of SNP effect of mtDNA-CN on 3HB. (D) Scater plot of SNP effect of 3HB on Tau. (E) MR effect size for mtDNA-CN on 3HB. (F) MR effect size for 3HB on Tau. mtDNA-CN, mitochondrial DNA copy number; DHA, 22:6, docosahexaenoic acid; 3HB, 3-Hydroxybutyrate; APOA1, Apolipoprotein A-I; CLHDLP, Concentration of large HDL particles; CMHDLP, Concentration of medium HDL particles; FCMHDL, Free cholesterol in medium HDL; PMHDL, Phospholipids in medium HDL; TCHDL, Total cholesterol in HDL; TCMHDL, Total cholesterol in medium HDL; TLMHDL, Total lipids in medium HDL; TLSHDL, Total lipids in small HDL; TrIDL, Triglycerides in IDL

Results of cell viability assay

As shown in Fig. 3, The viability of HT22 cells in GD models exhibited a gradual decline over time, with a marked reduction observed at 6 h, which persisted at a low level after 12 h. Based on glucose deprivation, cell viability showed differences when the cells were intervened with 80, 160 µmol/L concentrations of 3HB at the time of 24 h, and 30 h. Conversely, no significant differences were observed when 3HB was introduced at concentrations of 5, 10, 20, or 40 µmol/L. Furthermore, the viability of HT22 cells in LG models exhibited differences at 12 h, 24 h, and 48 h compared with the control group. When 3HB was given at the same time with a low glucose concentration, the viability of HT22 cells was maintained at a stable level at 12 h, 24 h, and 48 h, but such effect did not increase with the increase of concentration of 3HB (Supplementary Fig. S2).

Fig. 3.

3HB increased the cell viability of HT22 cell GD models and LG models. (A) The cell viability of HT22 cell GD models after induction for different time (0, 6, 12, 18, 24, and 30 h) was detected by MTT assay; (B, C) The cell viability of HT22 cell GD models intervened with different concentrations (0, 5, 10, 20, 40, 80, 160 µmol/L) of 3HB for different time (24 h and 30 h) was measured by MTT assay; (D, E, F) The cell viability of HT22 cell LG models intervened with different concentrations (80, 160, 320 µmol/L) of 3HB for different time (12, 24, and 48 h) was measured by MTT assay. The experiment was carried out in 3 independent replicates, each with n ≥ 3 wells/replicate samples; +++P < 0.0001 versus Control; *P < 0.05, **P < 0.001, ***P < 0.0001 versus GD; #P < 0.05, ##P < 0.001, ###P < 0.0001 versus G6.25. GD, glucose deprivation; LG, low glucose supply; G6.25, glucose 6.25 mmol/L; 3HB, 3-Hydroxybutyrate; MTT, 3-(4,5-dimethylthiazol-2-yl) -2,5-diphenyltetrazolium bromide

Morphological changes

As shown in Fig. 4, Under the microscope, HT22 cells grew adherently to the wall and showed typical neurone-like morphological features, with polygonal cell morphology, well-defined borders, nuclei located in the centre of the cell or to the side, round or oval, and small neurospheres seen around the cells. The LG HT22 cell model was slow-growing and even appeared to be wrinkled and coalesced in cytoplasm after 24 h. However, the morphology of cells exposed to 3HB (80, 160, 320 µmol/L) for 24 h in the presence of a low glucose supply remained close to normal. At 48 h, the cells in each group showed crumpling, membrane rupture, and nuclear consolidation (Supplementary Fig. S3).

Fig. 4.

(A) Light microscopy of HT22 cell LG models intervened by 3HB for different time (12, 24, and 48 h). Scale bar = 20 μm. (B) The mitochondrial morphology of HT22 cell LG models intervened with different concentrations (80, 160, and 320 µmol/L) of 3HB for 24 h was imaged by HIS-SIM. Hoechst, MitoTracker, Merge: Scale bar = 5 μm; Zoom: Scale bar = 1 μm. (C, D, E, F, G) Quantitative analysis of mitochondrial structures based on HIS-SIM imaging in Part B, involving the mitochondria of mean area, mean perimeter, form factor, aspect ratio, and mean branch length. (H, I, J, K) Representative Western blot and quantitative data showing the effects of 3HB on the expression of Tau5, phospho-Tau (Ser262) (p-Tau262), phospho-Tau (S396) (p-Tau396). (L) The mtDNA copy number of HT22 cell LG models intervened with different concentrations (80, 160, and 320 µmol/L) of 3HB for 24 h was analyzed by qPCR. #P < 0.05, ##P < 0.001, ###P < 0.0001 versus G25; *P < 0.05, **P < 0.001, ***P < 0.0001 versus G6.25; HIS-SIM, High Sensitivity Structured Illumination Microscope; qPCR, Quantitative real-time polymerase chain reaction; LG, low glucose supply; 3HB, 3-Hydroxybutyrate; G25, 25 mmol/L glucose; G6.25, 6.25 mmol/L glucose; H80, 3HB 80 µmol/L; H160, 3HB 160 µmol/L; H320, 3HB 320 µmol/L

Mitochondrial changes of cells

As shown in Fig. 4, Observed through a High Sensitivity Structured Illumination Microscope (HIS-SIM), the mitochondria of HT22 cells cultured in DMEM containing glucose at a concentration of 25 mmol/L for 24 h were of regular morphology, showing a tubular network or punctate shape. When the glucose concentration was 6.25 mmol/L, the mitochondrial morphology of the cells changed, exhibiting swollen or large spherical structures. The mitochondrial swelling of the cells was not obvious after the simultaneous provision of 6.25 mmol/L glucose and 80 µmol/L 3HB. When 6.25 mmol/L glucose and 160 µmol/L 3HB were provided, some mitochondria were observed as having large spherical structures. When 6.25 mmol/L glucose and 320 µmol/L 3HB were supplied, some of the mitochondria showed a prolonged state.

Western blotting

As shown in Fig. 4 (Supplementary Fig. S4), In HT22 cells intervened with glucose and 3HB, no significant alteration was observed in the expression of total Tau protein (Tau5). However, in the low - glucose HT22 model, phosphorylated Tau increased, with a notable increase in the expression of p-Tau396. Following intervention with 3HB, both p-Tau262 and p-Tau396 expression tended to decline.

Results of qPCR

As shown in Fig. 4, The qPCR results (Supplementary Table S6) illustrated the effects of low sugar and 3HB on mitochondrial DNA copy number in HT22 cells. Intervention with 6.25 mmol/L glucose and higher concentrations of 3HB significantly altered the mitochondrial DNA copy number, suggesting that the concentration of 3HB was positively correlated with mitochondrial DNA copy number.

Discussion

Studies have shown that increased blood levels of Tau proteins are positively correlated with the severity of AD and continue to worsen with disease progression [41, 42]. Moreover, the level of p-tau181 has emerged as a biomarker for predicting pathological changes in tau and Aβ in the brain [43–45]. Emerging studies on the early pathological mechanisms of AD have focused on mitochondrial dysfunction and the imbalance of brain energy metabolism [6, 46], which point in the direction of research on early diagnosis and prevention strategies for AD. The function, quantity, and quality of mitochondria, the central hub of intracellular energy production, can be assessed by mtDNA-CN measurements [18, 47]. It is noteworthy that the reduction of mtDNA-CN is strongly associated with a tendency to exacerbate the pathological features of AD [48, 49]. Therefore, mtDNA-CN and Tau protein levels in the blood have received attention as potential biomarkers for assessing the early risk of AD [9–11].

Recent investigations have revealed early energy metabolism abnormalities in AD patients, manifested by decreased efficiency of glucose metabolism and altered patterns of fatty acid metabolism [50, 51]. These metabolic disturbances are associated with mtDNA-CN and Tau protein levels [52, 53]. To further investigate the causal association of metabolites on AD blood biomarkers. In this study, MR analysis was used to find the mediating metabolites between mtDNA-CN and Tau. The results suggest that 3HB has a potential mediating role in the association between mtDNA-CN and Tau. 3HB, also known as 3 Hydroxybutyric acid, or beta Hydroxybutyric acid, is a ketone body distributed in the human brain, liver, and muscle, which is mainly generated from fatty acids by β-oxidation in the mitochondria of the liver [54], and enters the brain via the blood circulation through the blood-brain barrier. Besides, 3HB can also be generated from the reduction of acetoacetic acid in the brain [55]. When glucose supply or uptake is insufficient, 3HB not only serves as an energy substrate in the brain, but also has been implicated as a signalling molecule in the regulation of oxidative stress and neuroprotection [55, 56]. In earlier studies, it was noted that 3HB is significantly elevated during fasting, ketogenic diet, exercise, or when the organism develops insulin resistance [57, 58]. Mice on a ketogenic diet have reduced levels of hyperphosphorylated Tau in the hippocampus, amygdala, and cortex [59]. With subsequent studies, scholars have observed a clear role for 3HB in degenerative brain diseases with mitochondrial dysfunction and abnormal energy metabolism [60, 61]. In the ketosis state, 3HB increases mitochondrial membrane potential by modulating the AMPK signalling pathway, regulates the NAD+/NADH ratio, enhances the nuclear levels of FOXO1, FOXO3a, and PGC1α in neurons in a SIRT2-dependent manner, and is involved in the regulation of mitochondrial mitophagy and quality control [62]. In conclusion, the results of MR revealed a correlation between the mtDNA-CN, 3HB, and Tau, which not only provides ideas for early screening of relevant biomarkers for AD but also provides a basis for preventive treatment of AD. On the one hand, monitoring alterations in the blood ratios of them could enhance early and accurate AD screening. On the other hand, modulating 3HB levels through dietary modifications and lifestyle adjustments could emerge as a promising intervention for AD prevention at an early stage.

In order to further verify the effect of 3HB on neuronal cells, the present study observed the morphological changes of HT22 cells and their mitochondria under no-glucose and low-glucose environments by in vitro experiments with 3HB. The results showed that glucose is an important and indispensable energy source for neuronal cells, and neuronal cells without an exogenous glucose supply for 6 h instantly lost more than half of their cell viability. Some studies have confirmed that neuronal cells can rely on direct glucose uptake and glycolysis to maintain normal function, and brain glucose insufficiency or impaired metabolism may directly lead to neuronal dysfunction and degeneration [63]. In AD patients, the activity of GLUT4, an insulin-dependent glucose transporter protein distributed on neuronal cells, is reduced, which in turn leads to impaired energy metabolism due to reduced neuronal glucose uptake, while ketone bodies can serve as an alternative glucose energy donor [5]. In addition, 3HB can act as a surrogate energy substrate for neuronal cells during glucose deficiency or insufficiency to maintain cell viability and potentially ameliorate mitochondrial morphology changes due to impaired energy metabolism. 3HB has potential therapeutic benefits in AD and can compensate for impaired brain glucose metabolism [64]. Exercise, calorie restriction, ketogenic diets, exogenous 3HB administration, 3HB derivatives, or agonist administration can be beneficial to elevate the levels of 3HB [65], thereby potentially aiding in the prevention and treatment of AD, which has been implicated in mitochondrial energy metabolism dysfunction [66, 67]. However, we observed that the maintenance effect of 3HB on neuronal cell viability did not increase significantly with the elevation of its concentration. Therefore, when supplementing neuronal cells with 3HB via exogenous sources, the optimal concentration level should be maintained. More importantly, how to maintain the optimal level of 3HB in the brain to improve AD deserves to be studied and explored through more clinical practice.

Ketone bodies (KB) hold significant therapeutic potential in the management of Alzheimer’s disease (AD). Emerging evidence suggests that KB enhance mitochondrial efficiency and compensate for the brain’s normal glucose dependency, thereby improving cerebral energy metabolism [68]. This is particularly critical for AD patients, as their brain energy metabolism is frequently impaired, and KB can bypass glucose dependency to directly supply energy to the brain [69]. Furthermore, KB exhibit the capacity to clear misfolded proteins, a process that plays a pivotal role in AD pathogenesis [70]. Through strategies such as ketogenic diets, exogenous KB supplementation, and metabolic modulation, KB may emerge as a viable therapeutic option for AD. However, to ensure both safety and efficacy, large-scale controlled clinical trials are imperative. Future research should further explore the therapeutic potential of KB in AD and optimize relevant intervention strategies to unlock their full clinical value.

It is worth mentioning that the ketogenic diet has long been debated as a viable treatment for AD. The eating pattern is distinguished by a high fat content and a low carbohydrate intake. It is intended to improve brain energy metabolism and alleviate metabolic stress on brain cells in AD patients by boosting ketone body formation as an alternate energy source to glucose [61, 66]. However, research has indicated that ketogenic diets’ benefits are not universally applicable and may potentially have negative consequences on some people [71]. Our findings indicate that differing glucose and ketone body ratios have distinct impacts on neuronal survival, as well as mitochondrial morphology and structure. Although 3HB can sustain some neuronal survival in no-sugar settings, its effect is limited, indicating that the applicability of ketogenic diets should be carefully considered. Furthermore, in a low-glucose environment, neuronal cells showed negative changes such as morphological atrophy, decreased viability, and mitochondrial swelling, whereas moderate 3HB supplementation effectively stabilised neuronal structure, maintained cell viability, and reduced mitochondrial swelling. Some research have also proven that glucose and 3HB concentration influence the morphological structure of cellular mitochondria [72, 73], which can directly affect the activity and efficiency of OXPHPS in regulating cellular energy metabolism [74]. This emphasises the importance of the ratio of glucose and ketone body in sustaining neuronal health. As a result, future research should focus on exploring how to design personalised ketogenic dietary regimens to combat AD by precisely modulating the ratio of glucose and ketone body.

Conclusions

Our study, leveraging Mendelian randomization analysis in conjunction with in vitro experiments, unravels the potential linkage of 3HB as a mediating metabolite between mtDNA-CN and Tau, offering novel insights into the correlation between biomarkers for early AD risk assessment. Notably, 3HB enhances neuronal cell viability and preserves mitochondrial morphology, suggesting its potential role as a protective factor in regulating energy metabolism imbalance in AD brains.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

3HB

3-Hydroxybutyrate

ATP

Adenosine triphosphate

AD

Alzheimer’s disease

APOA1

Apolipoprotein A-I

AutoMitoC

Automatic mitochondrial copy

CLHDLP

Concentration of large HDL particles

CMHDLP

Concentration of medium HDL particles

CI

Confidence Interval

HDAC6

Deacetylase

DHA

Docosahexaenoic acid

ETC

Electron transport chain

ExWAS

Exome-wide association study

GWAS

Genome-wide association study

GD

Glucose deprivation

HIS-SIM

High Intelligent and Sensitive Microscope

IVW

Inverse variance weighted

LG

Low glucose

MR

Mendelian randomization

mt-DNA

Mitochondrial DNA

mtDNA-CN

Mitochondrial DNA copy number

MFN1

Mitofusin 1

HT22

Mouse hippocampal neuronal

OR

Odds Ratio

OXPHOS

Oxidative phosphorylation

SNPs

Single Nucleotide Polymorphisms

SNVs

Single nucleotide variants

TCA

Tricarboxylic acid

Author contributions

Xiaoling Hu was responsible for the conception, data analysis, visualization, and initial drafting of the manuscript. Yu Lin was responsible for the study design, data curation and figures process. Kaiwen Huang and HuiLin Xu prepared the experimental materials and guided the experiment. Changmei Huang Fu, Jiayin Ou and Xiude Fan verified the charts and data. Zhe Li prepared the experimental materials.Jiansong Fang and Shuhuan Fang contributed to the study design and approved the final version of the manuscript. Xiaoling Hu is the first author.Jiansong Fang and Shuhuan Fang are the corresponding authors. All authors have read and agree to the published version of the manuscript.All authors reviewed the manuscript.

Funding

This work was supported by the Natural Science Foundation of Guangdong Province, China (Grant No. 2022A1515011645) and Open Bidding for Selecting the Best Candidates in Guangzhou University of Chinese Medicine (Grant No. A1-2601-23-414-453Z66).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

The original research that generated the data obtained ethical approval and participant consent, so using summary-level data in our analysis does not require ethical approval.

Consent for publication

All authors approved the submitted version.

Competing interests

The authors declare no competing interests.