Preliminary longitudinal epigenetic clock analyses of patients with mild cognitive impairment through nutritional intervention

Abstract

Background

Mild cognitive impairment (MCI) is an intermediate stage between normal aging and dementia. Not a small number of patients with MCI progress to Alzheimer's disease within 1 year. Epigenetic changes in DNA methylation, such as those observed in the epigenetic clock, are receiving increasing attention in the field of aging research. It is crucial to demonstrate the effectiveness of interventions for patients with MCI from a biological perspective to promote the prevention of dementia.

Objective

We conducted epigenetic clock analysis, mainly longitudinal evaluation, on patients with MCI who received intervention with vitamin D and/or marine omega-3 fatty acid supplements.

Methods

We conducted epigenetic clock analyses with public DNA methylation datasets from the Gene Expression Omnibus database. This dataset contains two timepoints’ longitudinal DNA methylation data of 25 patients with MCI and 20 controls. They received two-year intervention with vitamin D and/or marine omega-3 fatty acid supplements. We conducted comparative analyses with Wilcoxon signed-rank test, and regression analyses of three models.

Results

In patients with MCI, PhenoAge and GrimAge significantly decelerated after the intervention in comparative analyses. These clocks nominally significantly decelerated also in most of regression analyses. Moreover, natural killer cells showed significant differences before and after the intervention.

Conclusions

We found deceleration of some epigenetic clocks in patients with MCI with interventions. We believe that it is crucial to conduct additional epigenetic clock analyses in cohorts with larger sample sizes or in cohorts with interventions such as occupational therapy and pharmacotherapy are actively conducted.

Keywords

aging Alzheimer's disease dementia DNA methylation epigenetic clock mild cognitive impairment

Introduction

Mild cognitive impairment (MCI) is an intermediate stage between normal aging and dementia. Although no severe functional disorders that obstruct daily activities are noted in MCI, critical changes in cognitive function occur.^1,2 According to a meta-analysis, the incidence rate of MCI per 1000 person-years is estimated to be 22.5 for those aged 75–79, 40.9 for those aged 80–84, and 60.1 for those aged 85 and over.³ Furthermore, approximately 5–10% patients with MCI progress to Alzheimer's disease (AD) within 1 year.³ As a result of the world's aging population, the number of people with dementia is expected to increase to 152.8 million by 2050, approximately three times the current number, causing a significant increase in the burden on medical systems, patients, and caregivers⁴ Consequently, various efforts are being made to intervene at the MCI stage,^2,5 however, only a few objective indicators can be widely and easily used in clinical practice to evaluate these effects. Furthermore, patients do not seek medical attention because of fear of stigma in many cases.⁶

Recently, epigenetic changes in DNA methylation (DNAm) at the cytosine-phosphate-guanine (CpG) sites have attracted attention. Changes in DNAm affect interactions between chromatin, transcription factors, and DNA, leading to changes in gene expression.⁷ Thus, changes in DNAm patterns can lead to functional disorders in biological processes and various diseases by maintaining chromatin structure adaptations.⁸ Moreover, DNAm affects biological aging^9,10; therefore, multiple measurement methods of biological aging have been established based on the genome-wide DNAm status as “epigenetic clocks”.^11–16 The estimated epigenetic age acceleration is associated with psychiatric disorders,^17–23 neurodegenerative disorders,^24,25 and all-cause mortality.²⁶ Epigenetic clocks were developed using elastic net regression in which hundreds to thousands of important CpG sites were selected from approximately 900,000 sites. Biological aging was then calculated from methylation rates and individual coefficients.

The epigenetic clock that Horvath firstly proposed in 2013 is now called HorvathAge, which uses 353 CpG sites based on comprehensive DNAm status in multiple tissues (such as peripheral blood, buccal mucosa, large intestine, heart, kidney, liver, and brain).¹¹ Subsequently, epigenetic clocks, such as HannumAge, SkinBloodAge, PhenoAge, GrimAge, and GrimAge2, have been developed. While HorvathAge, HannumAge, and SkinBloodAge were designed using the actual age itself as the objective variable, PhenoAge, GrimAge, and GrimAge2 were designed using the time to death or the time to the onset of lethal physical disorders as the objective variables in the training data. PhenoAge uses the DNAm status of 513 CpG sites, whereas GrimAge and GrimAge2 use that of 1030 CpG sites.^14–16 GrimAge and GrimAge2 have a two-stage algorithm that predict the biological age of an individual based on the predicted values of approximately 10 aging-related protein levels and cumulative smoking history, which are predicted from the DNAm status. The predicted values for these aging-related proteins are called GrimAge components and GrimAge2 components.^15,16 GrimAge components include adrenomedullin (ADM), beta-2 microglobulin (B2 M), cystatin C, growth differentiation factor-15 (GDF15), leptin, plasminogen activation inhibitor-1 (PAI1), and tissue inhibitor of metalloproteinases-1 (TIMP1).¹⁵ GrimAge2 components additionally include hemoglobin A1c and C-reactive protein.¹⁶ Although these proteins are not used as biomarkers in psychiatric clinical practice, they are strongly associated with the nervous system.^27–30 The predicted cumulative smoking history based on the DNAm status is called DNAmPACKYRS.

Moreover, DNAm-based telomere length (DNAmTL) was developed with 140 CpG sites and predicts telomere lengths.^31,32 DunedinPACE was developed using the DNAm rates of 173 CpG sites based on longitudinal datasets. Comprehensive DNAm status also enables the prediction of multiple cell compositions (CD8+ T cells, CD4+ T cells, natural killer [NK] cells, Bcells, monocytes, granulocytes, and plasmablasts). Changes in the cell composition are also associated with age.³³

These epigenetic clocks use different formulas with different numbers of CpG sites, because the tissues and indicators used for development differ. Each clock captures specific aspects of biological aging. Consequently, DNAm and epigenetic clocks are biological and objective indicators that have been successfully applied in the study of various psychiatric and neurological disorders.³⁴ AD is widely known to be a disease related to aging and is associated with DNAm and epigenetic clocks.^35,36 Additionally, Vyas et al. demonstrated a significant association between cognitive function based on neuropsychiatric evaluation and GrimAge in a cohort of patients with MCI and healthy controls.³⁷ Conversely, few studies longitudinally verified the changes in epigenetic clocks, GrimAge components, and cell composition through intervention in patients with MCI based on DNAm analyses. Orr et al. reported that nicotinamide riboside slowed the PhenoAge and GrimAge.³⁸ While Holmes et al. found that vitamin B group, which reduced total plasma homocysteine levels, normalized epigenetic aging in patients with MCI.³⁹

We believe it is necessary to demonstrate the effectiveness of various interventions in patients with MCI from a biological perspective to promote dementia prevention worldwide. We aimed to investigate whether interventions with vitamin D and marine omega-3 fatty acids in patients with MCI affect epigenetic age acceleration as these nutritional elements are implicated in the relationship between aging and psychiatric disorders.^40,41

Methods

Participants

In this study, we used the public DNAm dataset GSE190540 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE190540, last accessed November 9, 2024) downloaded from the Gene Expression Omnibus database. Okerete and Vyes developed this dataset, and Vyes et al. examined the correlation between cognitive function and epigenetic clocks in patients with MCI and healthy individuals.³⁷ The demographic data for this dataset are listed in Table 1. This dataset contains comprehensive DNA methylation data of 20 healthy individuals and 25 patients with MCI from Massachusetts General Hospital in the United States at two time points (before and 2 years after the intervention) obtained with Illumina Infinium methylation EPIC technology.³⁷ Participants received nutritional interventions involving vitamin D and/or marine omega-3 fatty acid supplementation.^40,41 The diagnosis of MCI was based on the same consensus diagnostic method as the Uniform Dataset protocol employed in the Alzheimer's Disease Research Center program.⁴² All participants provided written informed consent for the study, and the study protocol was approved by the Institutional Review Board of Mass General Brigham.

Table 1.

Demographic data of the dataset of this study.

		Controls	Patients with MCI	p
Total	N	20	25
	Age, mean ± SD	70.05 ± 5.83	69.64 ± 5.30	0.811^a
	Female, N (%)	9 (45)	13 (52)	0.767^b
Females	Age, mean ± SD	71.74 ± 6.43	70.32 ± 6.49	0.593^a
Males	Age, mean ± SD	68.67 ± 5.18	68.90 ± 3.78	0.880^a

MCI: mild cognitive impairment; N: number; SD: standard deviation

The p values were calculated with Mann-Whitney U test.

The p value was calculated with Fisher's exact test.

Evaluation of multiple epigenetic age acceleration, GrimAge components, and cell compositions

Blood samples were subjected to DNA extraction and DNAm rates were calculated via bisulfite transformation. We calculated multiple epigenetic clocks (HorvathAge, HannumAge, SkinBloodAge, PhenoAge, GrimAge, GrimAge2, and DNAmTL), GrimAge components (ADM, B2 M, cystatin C, GDF15, leptin, PAI1, TIMP1, and DNAmPACKYRS), and predicted values of cell compositions (CD8+ Tcell, CD4+ Tcell, NK cell, Bcell, Monocyte, Granulocyte, and PlasmaBlast)³³ using online the DNAm age calculator (https://dnamage.clockfoundation.org/ last accessed October 19, 2024) with DNAm statuses.¹¹ We evaluated epigenetic age acceleration: AgeAccelHorvath, AgeAccelHannum, AgeAccelSkinBlood, AgeAccelPheno, AgeAccelGrim, and AgeAccelGrim2 defined as the residual from regressing HorvathAge, HannumAge, SkinBloodAge, PhenoAge, GrimAge, and GrimAge2, respectively, on the chronological (actual) age. Positive and negative values indicate whether the epigenetic age is higher or lower than the expected age (based on the actual age). The age-adjusted estimate of DNAmTL (DNAmTLadjAge) was defined as the residual from regressing DNAmTL on the actual age. Positive values indicated that the DNAmTL was longer than expected. We calculated DunedinPACE using the R package.⁴³ Regarding GrimAge components, to be consistent with the notation in the online calculator, these were denoted as DNAmADM, DNAmB2 M, DNAmCystatin C, DNAmGDF15, DNAmLeptin, DNAmPAI1, DNAmTIMP1, and DNAmPACKYRS.

Statistical analysis

Statistical analyses were performed using R version 4.3.1 (R Development Core Team, Vienna, Austria). We used Fisher's exact test for sex difference and Mann-Whitney U test for age difference between groups. We used Pearson's correlation analyses to examine the correlation between each epigenetic clock and the actual age of the 20 controls and 25 patients with MCI at two points to confirm the validity of the epigenetic clocks.

First, we conducted regression analyses for the baseline data of 20 controls and 25 patients with MCI, with phenotype (controls or patients with MCI) as the explanatory variable and each epigenetic age acceleration, GrimAge component, and predicted value of cell composition as the objective variables. In the first regression model, we did not include any confounding factors. In the second regression model, we included age and sex as confounding factors. In the third regression model, we included age, sex, and seven DNAm-based cell compositions as confounding factors.

Second, we conducted comparative analyses using the Wilcoxon signed-rank test and regression analyses for the data of the two timepoints (epigenetic age acceleration, GrimAge component, and cell composition) for 25 patients with MCI. In the regression analyses, we used the timepoint of blood sampling (before or 2 years after the intervention) as the explanatory variable, and each age acceleration, GrimAge component, and predicted value of cell composition as the objective variables. In the first regression model, we included the individual as a random variable as a confounding factor. In the second regression model, we added age and sex as confounding factors. In the third regression model, we added seven DNAm-based cell compositions to confounding factors.

Third, as this GSE190540 dataset also contains the data of controls both before and after nutritional interventions, we conducted comparative analyses using the Wilcoxon signed-rank test and regression analyses for each epigenetic age acceleration of the two timepoints of 20 controls. In the regression analyses, we used the timepoint of blood sampling (before or 2 years after the intervention) as the explanatory variable and each age acceleration as the objective variable. In the first regression model, we included the individual as a random variable as a confounding factor. In the second regression model, we added age and sex as confounding factors. In the third regression model, we added seven DNAm-based cell compositions to confounding factors.

Fourth, we conducted regression analyses for the values of change in epigenetic age acceleration, GrimAge component, and cell composition from before to 2 years after the intervention in 20 controls and 25 patients with MCI. In the regression analyses, we used the phenotype (controls or patients with MCI) as the explanatory variable and the value of change in each age acceleration, GrimAge component, and cell composition as the objective variables. Confounding factors were not included in these regression analyses.

We used the following dummy variables for the phenotype: control = 0 and MCI = 1, for the sex: male = 0 and female = 1, for the timepoints: before intervention = 0 and after intervention = 1. Regarding multiple testing, statistical significance was defined as a two-tailed q < 0.05. We calculated q values by adjusting p values with Benjamini Hochberg method. We used the R packages lmertest and lme4 for regression analyses using random variables.

Results

In the groups of patients with MCI and controls, there are no significant differences in sex or age (Table 1). We found significant correlations between actual age and each epigenetic clock in the dataset (HorvathAge: correlation coefficient = 0.473, q = 2.48 × 10⁻⁶; HannumAge: correlation coefficient =0.631, q = 4.43 × 10⁻¹¹; SkinBloodAge: correlation coefficient = 0.663, q = 2.54× 10⁻¹²; PhenoAge: correlation coefficient = 0.617, q = 1.27 × 10⁻¹⁰; GrimAge: correlation coefficient = 0.819, q = 4.38× 10⁻²²; GrimAge2: correlation coefficient = 0.792, q = 5.46× 10⁻²⁰; DNAmTL: correlation coefficient = −0.516, q = 2.22× 10⁻⁷) confirming the reliability of the epigenetic clocks (Figure 1).

Figure 1.

The correlation plot for each epigenetic clock. Controls are displayed as dark circular dots at baseline and pale circular dots at the time 2 years after the intervention. Patients with MCI are displayed as dark square dots at baseline and pale square dots at 2 years after the intervention. “Rho” indicates the Pearson's correlation coefficients. The q values were calculated by adjusting p values of the Pearson's correlation test. The gray lines represent the regression lines. This Figure was created using the R package ggplot2. MCI, mild cognitive impairment.

The results of the regression analyses for phenotypes (controls or patients with MCI) are listed in Supplemental Table 1. Violin plots are shown in Supplemental Figure 1 for each epigenetic age acceleration, in Supplemental Figure 2 for each GrimAge component, and in Supplemental Figure 3 for each cell composition. No epigenetic age acceleration, GrimAge component, or cell composition showed significant associations with the phenotype in either the first, second or third regression models.

The results of the comparative and regression analyses for the time points (before or 2 years after intervention) for patients with MCI are listed in Table 2, and violin plots are shown in Supplemental Figure 4 for each epigenetic age acceleration, in Supplemental Figure 5 for each GrimAge component, and in Supplemental Figure 6 for each cell composition. In the comparative analyses, in age acceleration, AgeAccelPheno (q = 0.046) and AgeAccelGrim (q = 0.046) showed significant differences before and after the intervention. No GrimAge components showed significant differences. In cell compositions, NK cells (q = 0.023) showed significant difference. In the first regression model, no age acceleration or GrimAge components showed significant associations with timepoints. In cell compositions, NK cells (effect size = −0.033, standard error [SE] = 0.011, q = 0.034) showed significant association with timepoints. In the second regression model, no age acceleration or GrimAge components showed significant associations with timepoints. In cell compositions, NK cells (effect size = −0.035, SE = 0.011, q = 0.024) showed significant association with timepoints. In the third regression model, in age acceleration, AgeAccelHannum (effect size = 1.446, SE = 0.469, and q = 0.025) and AgeAccelSkinBlood (effect size = 1.367, SE = 0.467, and q = 0.025) showed significant associations with timepoints. No GrimAge components showed significant associations. The violin plots of AgeAccelPheno, AgeAccelGrim, and AgeAccelGrim2 are shown in Figure 2.

Figure 2.

Violin plot showing ageAccelPheno, AgeAccelGrim, and AgeAccelGrim2 of patients with MCI before and after intervention. The same individuals are connected by lines. The plot was created using the R package ggplot2. MCI, mild cognitive impairment. The q values were calculated by adjusting p values of the Wilcoxon signed-rank test.

Table 2.

The results of comparative and regression analyses for timepoints (before or two years after intervention) in only patients with MCI.

Comparative analyses^a
	Before intervention, mean ± SD	After intervention, mean ± SD	p	q^b
Epigenetic age acceleration
AgeAccelHorvath	0.673 ± 6.206	−0.526 ± 6.282	0.0802	0.160
AgeAccelHannum	0.969 ± 4.948	0.366 ± 5.038	0.339	0.387
AgeAccelSkinBlood	0.117 ± 3.562	0.385 ± 3.883	0.615	0.615
AgeAccelPheno	0.975 ± 5.153	−0.636 ± 4.857	0 . 0115	0.0458
AgeAccelGrim	0.429 ± 2.937	−0.517 ± 3.159	0.00807	0.0458
AgeAccelGrim2	0.396 ± 3.339	−0.495 ± 3.455	0.0255	0.0680
DNAmTLAdjAge	−0.031 ± 0.239	0.002 ± 0.233	0.107	0.172
DunedinPACE	0.999 ± 0.131	1.016 ± 0.108	0.182	0.242
GrimAge components
DNAmADM	365.90 ± 21.95	365.14 ± 23.42	0.672	0.874
DNAmB2M	1730146 ± 95206	1765530 ± 100355	0.0588	0.181
DNAmCystatinC	661589.1 ± 19698.2	663060.5 ± 26347.1	0.731	0.874
DNAmGDF15	1063.90 ± 131.68	1008.91 ± 155.17	0.0802	0.181
DNAmLeptin	5787.42 ± 4331.38	6464.29 ± 4722.96	0.0903	0.181
DNAmPAI1	15059.6 ± 2503.4	14991.5 ± 1972.6	0.833	0.874
DNAmTIMP1	35355.9 ± 878.2	35548.2 ± 892.4	0.0626	0.181
DNAmPACKYRS	20.33 ± 6.85	20.09 ± 7.05	0.874	0.874
Cell compositions
CD8T	0.081 ± 0.058	0.074 ± 0.047	0.578	0.762
CD4T	0.130 ± 0.062	0.159 ± 0.074	0.0275	0.0962
NKcell	0.049 ± 0.057	0.016 ± 0.022	0.00334	0.0233
Bcell	0.019 ± 0.023	0.023 ± 0.03	0.637	0.762
Monocyte	0.074 ± 0.037	0.082 ± 0.023	0.191	0.445
Granulocyte	0.566 ± 0.093	0.566 ± 0.068	0.771	0.771
PlasmaBlast	1.725 ± 0.146	1.718 ± 0.155	0.653	0.762

Regression analyses
	Effect size	Standard error	t value	p	q^b
The first regression model ^c
Epigenetic age acceleration
AgeAccelHorvath	−1.200	0.840	−1.429	0.166	0.332
AgeAccelHannum	−0.603	0.657	−0.917	0.368	0.440
AgeAccelSkinBlood	0.268	0.394	0.680	0.503	0.503
AgeAccelPheno	−1.611	0.575	−2.799	0.00995	0.0796
AgeAccelGrim	−0.946	0.425	−2.227	0.0356	0.142
AgeAccelGrim2	−0.891	0.495	−1.799	0.0847	0.226
DNAmTLAdjAge	0.032	0.026	1.261	0.219	0.351
DunedinPACE	0.017	0.019	0.884	0.385	0.440
GrimAge components
DNAmADM	−0.765	2.744	−0.279	0.783	0.855
DNAmB2M	35383.10	14841.79	2.384	0.0254	0.187
DNAmCystatinC	1471.408	2633.887	0.559	0.582	0.855
DNAmGDF15	−54.993	29.979	−1.834	0.0790	0.211
DNAmLeptin	676.871	561.141	1.206	0.239	0.479
DNAmPAI1	−68.093	369.909	−0.184	0.855	0.855
DNAmTIMP1	192.315	91.690	2.097	0.0467	0.187
DNAmPACKYRS	−0.238	0.645	−0.369	0.716	0.855
Cell compositions
CD8T	−0.00669	0.00631	−1.060	0.300	0.434
CD4T	0.02971	0.01259	2.360	0.0267	0.0935
NKcell	−0.03314	0.01067	−3.107	0.00481	0.0337
Bcell	0.00439	0.00423	1.037	0.310	0.434
Monocyte	0.00859	0.00761	1.129	0.270	0.434
Granulocyte	0.00045	0.02033	0.022	0.983	0.983
PlasmaBlast	−0.00690	0.03245	−0.213	0.833	0.972
The second regression model ^d
Epigenetic age acceleration
AgeAccelHorvath	−1.273	0.960	−1.327	0.193	0.318
AgeAccelHannum	−0.681	0.746	−0.912	0.368	0.490
AgeAccelSkinBlood	0.349	0.487	0.718	0.477	0.545
AgeAccelPheno	−1.829	0.690	−2.649	0.0114	0.0912
AgeAccelGrim	−1.071	0.473	−2.262	0.0301	0.120
AgeAccelGrim2	−1.103	0.548	−2.014	0.0520	0.139
DNAmTLAdjAge	0.040	0.030	1.307	0.199	0.318
DunedinPACE	−1.84 × 10⁻⁴	0.021	−0.009	0.993	0.993
GrimAge components
DNAmADM	−3.427	2.894	−1.184	0.246	0.462
DNAmB2M	10254.170	15636.25	0.656	0.517	0.517
DNAmCystatinC	−4663.068	2904.612	−1.605	0.118	0.389
DNAmGDF15	−82.699	31.026	−2.665	0.0127	0.102
DNAmLeptin	560.622	576.718	0.972	0.340	0.462
DNAmPAI1	−331.469	392.125	−0.845	0.405	0.462
DNAmTIMP1	−83.614	98.744	−0.847	0.404	0.462
DNAmPACKYRS	−1.199	0.809	−1.481	0.146	0.389
Cell compositions
CD8T	−0.00274	0.00743	−0.368	0.715	0.806
CD4T	0.03192	0.01339	2.384	0.0236	0.0827
NKcell	−0.03521	0.01100	−3.201	0.00349	0.0244
Bcell	0.00627	0.00456	1.376	0.179	0.417
Monocyte	0.00707	0.00786	0.899	0.377	0.658
Granulocyte	−0.00518	0.02092	−0.247	0.806	0.806
PlasmaBlast	−0.02476	0.03382	−0.732	0.470	0.658
The third regression model ^e
Epigenetic age acceleration
AgeAccelHorvath	0.792	0.841	0.942	0.354	0.564
AgeAccelHannum	1.446	0.469	3.084	0.00403	0.0255
AgeAccelSkinBlood	1.367	0.467	2.931	0.00637	0.0255
AgeAccelPheno	−1.700	0.782	−2.173	0.0378	0.101
AgeAccelGrim	−0.168	0.457	−0.366	0.717	0.819
AgeAccelGrim2	−0.000632	0.501	−0.00126	0.999	0.999
DNAmTLAdjAge	−0.0235	0.0289	−0.812	0.423	0.564
DunedinPACE	0.0334	0.0194	1.724	0.0957	0.191
GrimAge components
DNAmADM	0.325	2.901	0.112	0.912	0.964
DNAmB2M	32547.44	16369.15	1.988	0.0567	0.207
DNAmCystatinC	−135.708	2972.200	−0.046	0.964	0.964
DNAmGDF15	−79.323	36.392	−2.180	0.0378	0.207
DNAmLeptin	944.009	673.027	1.403	0.172	0.343
DNAmPAI1	−828.499	452.897	−1.829	0.0778	0.207
DNAmTIMP1	40.117	104.334	0.385	0.703	0.938
DNAmPACKYRS	0.561	0.745	0.753	0.456	0.730

The p values were calculated with the Wilcoxon signed-rank test.

The q values were calculated by adjusting p values with Benjamini Hochberg method.

In the first regression model, individual (random variable) was included as confounding factor.

In the second regression model, individual (random variable), age, and sex were included as confounding factors.

In the third regression model, individual (random variable), age, sex, and cell compositions were included as confounding factors.

In the regression analyses, the explanatory variable is timepoint (before or two years after intervention), and the objective variable is each epigenetic age acceleration, GrimAge component, and cell composition. Data of only patients with MCI were used for these regression analyses.

ADM: adrenomedullin; B2M: beta 2 microglobulin; CD4T: CD4 + Tcell; CD8T: CD8 + Tcell; DNAm: DNA methylation; DNAmPACKYRS: DNA methylation-based smoking pack-years; DNAmTLAdjAge: age-adjusted estimate of DNA methylation-based telomere length; GDF15: growth differentiation factor-15; MCI: mild cognitive impairment; NKcell: natural killer cell; PAI1: plasminogen activation inhibitor-1; SD: standard deviation; TIMP1: tissue inhibitor of metalloproteinases-1.

p < 0.05 and q < 0.05 are shown in bold.

The results of the comparative and regression analyses for the timepoints (before or 2 years after intervention) of the controls are listed in Supplemental Table 2, and violin plots are shown in Supplemental Figure 7 for each epigenetic age acceleration. In the comparative analyses, AgeAccelHannum (q < 0.001), AgeAccelPheno (q = 0.019), and DNAmTLAdjAge (q = 0.019) showed significant differences before and after the intervention. In the first regression model, AgeAccelHannum (effect size = −1.96, SE = 0.378, and q < 0.001), AgeAccelPheno (effect size = −2.32, SE = 0.752, and q = 0.016), and DNAmTLAdjAge (effect size = 0.0533, SE = 0.0165, and q = 0.016) showed significant associations with timepoints. In the second regression model, AgeAccelHannum (effect size = −2.14, SE = 0.441, and q < 0.001), AgeAccelPheno (effect size = −2.42, SE = 0.841, and q = 0.028), and DNAmTLAdjAge (effect size = 0.0559, SE = 0.0207, and q = 0.028) showed significant associations with timepoints. In the third regression model, AgeAccelHannum (effect size = −1.62, SE = 0.472, and q = 0.015) showed significant association with timepoints.

The results of the regression analyses for the values of change are listed in Supplemental Table 3 and violin plots are displayed in Supplemental Figure 8 for each age acceleration, in Supplemental Figure 9 for each GrimAge component, and in Supplemental Figure 10 for each cell composition. Any changes of age acceleration, GrimAge components and cell compositions were not significantly associated with the phenotype.

Discussion

In this study, we investigated whether interventions with vitamin D and marine omega-3 fatty acids in patients with MCI affect epigenetic age acceleration. Notably, our results of comparative analyses showed that the intervention for MCI significantly slowed down some epigenetic clocks, particularly those that indicate life expectancy, suggesting they may be indicators of the intervention effects. However, in the regression analysis, the significance was only nominal at best, so we need to be careful in interpreting the results. Moreover, when including cell compositions as confounding factors, the effect size of timepoints on age acceleration shifted to the direction of acceleration overall. Some other epigenetic clocks which were developed using actual age itself showed significant acceleration. However, we should consider cell compositions also change in relation to aging or life expectancy.^33,44 We could also hypothesize that the biological aging speed was decelerated through changes of cell compositions. The results of the analyses on the values of change in epigenetic age acceleration before and after the intervention showed no significant differences between controls and patients with MCI. This intervention may have equivalent anti-biological aging effects in patients with MCI and healthy individuals. Our results are consistent with previous studies which have reported various interventions that delay the biological aging process.^35,45,46 Although our findings were not specific to patients with MCI, we believe that our findings provide a basis for promoting the health of patients with MCI, given that we were able to identify decelerated biological aging through nutritional interventions, even when limited to patients with MCI.

Although the intervention in this study was mainly nutritional, with vitamin D and marine omega-3 fatty acids, it nominally significantly increased the predicted value of CD4+ T cells and significantly decreased the predicted value of NK cells. Furthermore, changes in NK cells (decrease) and monocytes (increase) were greater in patients with MCI than those in controls (nominal significant). An increase in vitamin D levels regulates the number of CD4+ T cells through vitamin D receptors and promotes the production of virus-specific antibodies by B cells through the activation of CD4+ T cells. Furthermore, vitamin D suppresses the onset of cytokine storms by reducing the release of inflammatory cytokines by CD4+ T cells.^47–50 Whereas, omega-3 fatty acids reduce NK cell numbers.^51,52 We previously reported that abnormalities in the number of NK cells are associated with suicide and major depressive disorder.^20,53 Furthermore, the number of reports indicating that NK cells play an important role in the progression of age-related diseases, including AD, is increasing.⁵⁴ These changes in immune cell compositions may not only reflect the psychological stress of patients with AD but also relate to the pathophysiology involving the accumulation of amyloid-β and tau proteins through cytokines.⁵⁵ Huang et al. examined the differences in white blood cell counts between patients with AD, patients with MCI, and healthy controls and found that patients with AD had non-significantly lower monocyte counts.⁵⁶ Nominally significantly greater increasing in monocytes after intervention in patients with MCI may be due to changes in the normal direction.

Nonetheless, this study has some limitations. First, the sample size was small, and the findings are still preliminary. We used Benjamini Hochberg method for adjustment of multiple testing, but this method is not very strict. Second, we were unable to obtain data that could have affected the DNAm, such as information on the individual's growth history, eating habits, and actual smoking history.^57–59 Third, we were unable to obtain detailed information on the interventions (vitamin D, marine omega-3 fatty acids, or both) received by each participant. Fourth, the intervention administered to the participants in this study was mainly a nutritional intervention, and we may obtain greater and different effects if interventions such as occupational therapy, rehabilitation, or drug therapy are conducted.

We examined associations between multiple indices of biological aging in patients with MCI and the interventions with vitamin D and marine omega-3 fatty acids, showing the intervention significantly slowed acceleration of some epigenetic clocks that are good at indicate life expectancy. We believe it is crucial to conduct additional verification in cohorts with larger sample sizes in which interventions such as occupational therapy and pharmacological therapy are actively conducted in the future.

Supplemental Material

sj-docx-1-alz-10.1177_13872877251360230 - Supplemental material for Preliminary longitudinal epigenetic clock analyses of patients with mild cognitive impairment through nutritional intervention

Supplemental material, sj-docx-1-alz-10.1177_13872877251360230 for Preliminary longitudinal epigenetic clock analyses of patients with mild cognitive impairment through nutritional intervention by Kiriko Minami, Toshiyuki Shirai, Satoshi Okazaki, Shohei Okada, Masao Miyachi, Ikuo Otsuka and Akitoyo Hishimoto in Journal of Alzheimer's Disease

Footnotes

Acknowledgments

The authors have no acknowledgments to report.

Author contributions

Kiriko Minami: Conceptualization; Data curation; Formal analysis; Investigation; Writing—original draft.

Toshiyuki Shirai: Data curation; Formal analysis; Investigation; Methodology; Visualization; Writing—original draft.

Satoshi Okazaki: Conceptualization; Data curation; Methodology.

Shohei Okada: Data curation; Formal analysis; Investigation.

Masao Miyachi: Data curation; Formal analysis.

Ikuo Otsuka: Data curation; Formal analysis.

Akitoyo Hishimoto: Conceptualization; Project administration; Writing—review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was partially supported by JST (Moonshot R&D Program) grant number [JPMJMS239F] (Akitoyo Hishimoto and Ikuo Otsuka), JSPS KAKENHI (Grant Numbers JP21K07520, JP24K10710 to Satoshi Okazaki; JP24K10732 to Ikuo Otsuka; JP21H02852, JP24K02383 to Akitoyo Hishimoto), SENSHIN Medical Research Foundation (to Ikuo Otsuka), and the Smoking Research Foundation (to Akitoyo Hishimoto).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

The information of DNA methylation data, sex and age from the GSE190540 dataset are publicly available in the Gene Expression Omnibus database (); this was created by Vyas et al. and has been used in their published paper.³⁷

Supplemental material

Supplemental material for this article is available online.

References

Petersen

Smith

Waring

, et al. Mild cognitive impairment: clinical characterization and outcome. Arch Neurol 1999; 56: 303–308.

Campbell

Lavoie

Farraia

, et al. Quality of life in mild cognitive impairment and mild dementia associated with Alzheimer's disease: a systematic review. Neurol Ther 2025; 14: 7–26.

Gillis

Mirzaei

Potashman

, et al. The incidence of mild cognitive impairment: a systematic review and data synthesis. Alzheimers Dement (Amst) 2019; 11: 248–256.

GBD 2019 Dementia Forecasting Collaborators. Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the Global Burden of Disease Study 2019. Lancet Public Health 2022; 7: e105–e125.

Wang

, et al. Frontiers and hotspots evolution in mild cognitive impairment: a bibliometric analysis of from 2013 to 2023. Front Neurosci 2024; 18: 1352129.

Siette

Meka

Antoniades

. Breaking the barriers: overcoming dementia-related stigma in minority communities. Front Psychiatry 2023; 14: 1278944.

Razin

Cedar

. DNA Methylation and gene expression. Microbiol Rev 1991; 55: 451–458.

Bird

. Perceptions of epigenetics. Nature 2007; 447: 396–398.

López-Otín

Blasco

Partridge

, et al. The hallmarks of aging. Cell 2013; 153: 1194–1217.

10.

Mkrtchyan

Abdelmohsen

Andreux

, et al. ARDD 2020: from aging mechanisms to interventions. Aging (Albany NY) 2020; 12: 24484–24503.

11.

Horvath

. DNA Methylation age of human tissues and cell types. Genome Biol 2013; 14: R115.

12.

Hannum

Guinney

Zhao

, et al. Genome-wide methylation profiles reveal quantitative views of human aging rates. Mol Cell 2013; 49: 359–367.

13.

Horvath

Oshima

Martin

, et al. Epigenetic clock for skin and blood cells applied to Hutchinson Gilford Progeria Syndrome and ex vivo studies. Aging (Albany NY) 2018; 10: 1758–1775.

14.

Levine

Quach

, et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging (Albany NY) 2018; 10: 573–591.

15.

Quach

Wilson

, et al. DNA Methylation GrimAge strongly predicts lifespan and healthspan. Aging (Albany NY) 2019; 11: 303–327.

16.

Binder

Zhang

, et al. DNA Methylation GrimAge version 2. Aging (Albany NY) 2022; 14: 9484–9549.

17.

Fries

Bauer

Scaini

, et al. Accelerated epigenetic aging and mitochondrial DNA copy number in bipolar disorder. Transl Psychiatry 2017; 7: 1283.

18.

Han

LKM

Aghajani

Clark

, et al. Epigenetic aging in major depressive disorder. Am J Psychiatry 2018; 175: 774–782.

19.

Okazaki

Numata

Otsuka

, et al. Decelerated epigenetic aging associated with mood stabilizers in the blood of patients with bipolar disorder. Transl Psychiatry 2020; 10: 129.

20.

Okazaki

Otsuka

Horai

, et al. Accelerated extrinsic epigenetic aging and increased natural killer cells in blood of suicide completers. Prog Neuropsychopharmacol Biol Psychiatry 2020; 98: 109805.

21.

Okazaki

Otsuka

Numata

, et al. Epigenetic clock analysis of blood samples from Japanese schizophrenia patients. NPJ Schizophr 2019; 5: 4.

22.

Tanifuji

Okazaki

Otsuka

, et al. Epigenetic clock analysis reveals increased plasma cystatin C levels based on DNA methylation in major depressive disorder. Psychiatry Res 2023; 322: 115103.

23.

Higgins-Chen

Boks

Vinkers

, et al. Schizophrenia and epigenetic aging biomarkers: increased mortality, reduced cancer risk, and unique clozapine effects. Biol Psychiatry 2020; 88: 224–235.

24.

Horvath

Ritz

. Increased epigenetic age and granulocyte counts in the blood of Parkinson's disease patients. Aging (Albany NY) 2015; 7: 1130–1142.

25.

Levine

Bennett

, et al. Epigenetic age of the pre-frontal cortex is associated with neuritic plaques, amyloid load, and Alzheimer's disease related cognitive functioning. Aging (Albany NY) 2015; 7: 1198–1211.

26.

Marioni

Shah

McRae

, et al. DNA Methylation age of blood predicts all-cause mortality in later life. Genome Biol 2015; 16: 25.

27.

Savaş

Herken

Yürekli

, et al. Possible role of nitric oxide and adrenomedullin in bipolar affective disorder. Neuropsychobiology 2002; 45: 57–61.

28.

Glubb

McHugh

Deng

, et al. Association of a functional polymorphism in the adrenomedullin gene (ADM) with response to paroxetine. Pharmacogenomics J 2010; 10: 126–133.

29.

Sun

Chen

. Associations of serum cystatin C with depressive symptoms and suicidal ideation in major depressive disorder. BMC Psychiatry 2021; 21: 576.

30.

Barlati

Minelli

Ceraso

, et al. Social cognition in a research domain criteria perspective: a bridge between schizophrenia and autism spectra disorders. Front Psychiatry 2020; 11: 806.

31.

Seeboth

Tsai

, et al. DNA methylation-based estimator of telomere length. Aging (Albany NY) 2019; 11: 5895–5923.

32.

Lee

Sun

Ori

APS

, et al. Epigenome-wide association study of leukocyte telomere length. Aging (Albany NY) 2019; 11: 5876–5894.

33.

Houseman

Accomando

Koestler

, et al. DNA Methylation arrays as surrogate measures of cell mixture distribution. BMC Bioinformatics 2012; 13: 86.

34.

Ryan

. “Epigenetic clocks”: theory and applications in human biology. Am J Hum Biol 2021; 33: e23488.

35.

Campisi

Cannella

Celik

, et al. Mitigating cellular aging and enhancing cognitive functionality: visual arts-mediated cognitive activation therapy in neurocognitive disorders. Front Aging Neurosci 2024; 16: 1354025.

36.

Altuna

Urdánoz-Casado

Sánchez-Ruiz de Gordoa

, et al. DNA Methylation signature of human hippocampus in Alzheimer's disease is linked to neurogenesis. Clin Epigenetics 2019; 11: 91.

37.

Vyas

Sadreyev

Gatchel

, et al. Pilot study of second-generation DNA methylation epigenetic markers in relation to cognitive and neuropsychiatric symptoms in older adults. J Alzheimers Dis 2023; 93: 1563–1575.

38.

Orr

Kotkowski

Ramirez

, et al. A randomized placebo-controlled trial of nicotinamide riboside in older adults with mild cognitive impairment. Geroscience 2024; 46: 665–682.

39.

Holmes

Valentin

Jernerén

, et al. Elevated homocysteine is associated with increased rates of epigenetic aging in a population with mild cognitive impairment. Aging Cell 2024; 23: e14255.

40.

Okereke

Reynolds

3rd Mischoulon

, et al. The VITamin D and OmegA-3 TriaL-Depression Endpoint Prevention (VITAL-DEP): rationale and design of a large-scale ancillary study evaluating vitamin D and marine omega-3 fatty acid supplements for prevention of late-life depression. Contemp Clin Trials 2018; 68: 133–145.

41.

Manson

Bassuk

Lee

, et al. The VITamin D and OmegA-3 TriaL (VITAL): rationale and design of a large randomized controlled trial of vitamin D and marine omega-3 fatty acid supplements for the primary prevention of cancer and cardiovascular disease. Contemp Clin Trials 2012; 33: 159–171.

42.

Weintraub

Salmon

Mercaldo

, et al. The Alzheimer's disease centers’ uniform data set (UDS): the neuropsychologic test battery. Alzheimer Dis Assoc Disord 2009; 23: 91–101.

43.

Belsky

Caspi

Corcoran

, et al. DunedinPACE, a DNA methylation biomarker of the pace of aging. Elife 2022; 11: e73420.

44.

Chen

Marioni

Colicino

, et al. DNA methylation-based measures of biological age: meta-analysis predicting time to death. Aging (Albany NY) 2016; 8: 1844–1865.

45.

Buttet

Bagheri

Ugbolue

, et al. Effect of a lifestyle intervention on telomere length: a systematic review and meta-analysis. Mech Ageing Dev 2022; 206: 111694.

46.

Shirai

Okazaki

Otsuka

, et al. Accelerated epigenetic aging in alcohol dependence. J Psychiatr Res 2024; 173: 175–182.

47.

Siopis

Porter

. Contribution of biological age-predictive biomarkers to nutrition research: a systematic review of the current evidence and implications for future research and clinical practice. Adv Nutr 2022; 13: 1930–1946.

48.

Poles

Karhu

McGill

, et al. The effects of twenty-four nutrients and phytonutrients on immune system function and inflammation: a narrative review. J Clin Transl Res 2021; 7: 333–376.

49.

Peng

Liu

Zheng

, et al. Immunological aspects of SARS-CoV-2 infection and the putative beneficial role of vitamin-D. Int J Mol Sci 2021; 22: 5251.

50.

Haroon

Fitzgerald

. Vitamin D and its emerging role in immunopathology. Clin Rheumatol 2012; 31: 199–202.

51.

Mukaro

Costabile

Murphy

, et al. Leukocyte numbers and function in subjects eating n-3 enriched foods: selective depression of natural killer cell levels. Arthritis Res Ther 2008; 10: R57.

52.

Peng

, et al. The ω-3 polyunsaturated fatty acid docosahexaenoic acid enhances NK-cell antitumor effector functions. Cancer Immunol Res 2024; 12: 744–758.

53.

Shindo

Tanifuji

Okazaki

, et al. Accelerated epigenetic aging and decreased natural killer cells based on DNA methylation in patients with untreated major depressive disorder. NPJ Aging 2023; 9: 19.

54.

Liu

. Natural killer cells in aging and age-related diseases. Neurobiol Dis 2023; 183: 106156.

55.

Baj

Seth

. Role of curcumin in regulation of TNF-α mediated brain inflammatory responses. Recent Pat Inflamm Allergy Drug Discov 2018; 12: 69–77.

56.

Huang

Zhang

Wang

, et al. Association of peripheral blood cell profile with Alzheimer's disease: a meta-analysis. Front Aging Neurosci 2022; 14: 888946.

57.

Merrill

Moore

Gladish

, et al. Paternal adverse childhood experiences: associations with infant DNA methylation. Dev Psychobiol 2021; 63: e22174.

58.

Maugeri

Barchitta

. How dietary factors affect DNA methylation: lesson from epidemiological studies. Medicina (Kaunas) 2020; 56: 374.

59.

Jain

Tamayo

, et al. The association of cigarette smoking with DNA methylation and gene expression in human tissue samples. Am J Hum Genet 2024; 111: 636–653.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

20.85 MB