Short-term responsiveness of DNA methylation–based aging biomarkers to a multimodal intervention comprising exercise and dietary guidance involving daily consumption of yogurt containing Bifidobacterium longum BB536: an exploratory randomized controlled trial

Introduction

Aging is the predominant risk factor for chronic diseases such as cardiovascular disorders, diabetes, and neurodegeneration, positioning it as a primary target for preventive strategies [1]. The geroscience hypothesis suggests that interventions addressing fundamental aging mechanisms could simultaneously delay multiple age-related disorders and extend health span [2–4]. Reliable biomarkers are indispensable for assessing whether such interventions can modulate biological aging.

Epigenetic clocks have emerged as promising biomarkers of biological aging, capturing variation in morbidity, mortality, and functional decline that is not explained by chronological age alone [5, 6]. These algorithms are constructed by modeling DNA methylation patterns at specific CpG sites to generate composite indices that estimate biological age, or, in some cases, the pace of aging [6]. First-generation clocks, such as the Horvath and Hannum models, were trained to predict chronological age [7, 8], whereas second-generation clocks, including PhenoAge, GrimAge and GrimAge version 2 (GrimAge v2), were explicitly optimized against physiological biomarkers and mortality risk [9–11]. More recently, DunedinPACE was developed to quantify the pace of aging, and accumulating evidence suggests that it may be particularly sensitive to short-term biological responses to lifestyle modification compared with traditional age-prediction clocks [12–14].

Contemporary geroscience frameworks conceptualize aging as the cumulative consequence of multiple “hallmarks of aging”, encompassing chronic inflammation and deregulated nutrient-sensing, among others [3, 15]. These hallmarks progress in parallel and reinforce each other, making single-pathway interventions unlikely to produce substantial improvements. Indeed, obesity exemplifies a clinical state in which several hallmarks of aging—chronic low-grade inflammation and metabolic dysregulation—are concurrently perturbed, and higher BMI is consistently associated with accelerated epigenetic aging across several DNA methylation clocks [16–18]. In this context, lifestyle components have garnered attention because they have partially shared effects across diet and exercise. Dietary modifications primarily influence nutrient-sensing and mitochondrial integrity [19], whereas physical activity enhances mitochondrial efficiency while attenuating chronic inflammation [20]. Importantly, inflammation and nutrient-sensing can respond within weeks to coordinated changes in diet and physical activity. Accordingly, multicomponent lifestyle approaches, by engaging multiple hallmarks concurrently, are hypothesized to produce integrated biological effects detectable over relatively short timeframes, and to yield measurable deceleration in aging-related biomarkers compared with single-modality approaches. Using a biomarker that captures the current pace of aging, such as DunedinPACE, may increase the likelihood of detecting these early signals in the context of short-term lifestyle intervention.

Although interest in lifestyle interventions to promote healthy aging is increasing, randomized controlled trials and post hoc analyses evaluating their impact on pace of aging quantified by DunedinPACE remain limited [21]. In the CALERIE trial, a two-year caloric restriction intervention in non-obese healthy adults significantly slowed the pace of aging as measured by DunedinPACE, yet failed to alter biological age estimated by PhenoAge or GrimAge [13]. Conversely, in the DO-HEALTH trial of community-dwelling older Europeans receiving vitamin D, omega-3 fatty acids, and a home exercise program for three years, omega-3 supplementation alone induced small but statistically significant reductions in PhenoAge, GrimAge v2, and DunedinPACE, whereas vitamin D or exercise alone showed no clear effects on any clock, although an additive benefit was observed only for PhenoAge when all three interventions were combined [22]. On the other hand, comparatively short- or mid-term trials were limited. Small pilot studies with multiple lifestyle programs, including diet, exercise, stress management, and supplementation with Lactobacillus plantarum 299v, were associated with a short-term reduction in salivary Horvath clock, although DunedinPACE was not measured [16, 17]. Conceptually, probiotics are often discussed as multi-component interventions because they comprise diverse microbial-derived factors, such as cell wall components and metabolites, which may collectively interact with host pathways related to multiple hallmarks of aging [23]. From a theoretical perspective, such characteristics could be hypothesized to support the detection of integrated biological responses in short-term intervention studies, although this remains to be empirically validated. This clock- and intervention-dependent variability underscores the complexity of aging processes and highlights the necessity of integrating multiple epigenetic biomarkers and studies to capture robust effects. Nevertheless, short- to mid-term randomized evidence on responsiveness of epigenetic aging measures remains limited [18].

Indeed, a recent cross-sectional analysis reported that adherence to a healthy dietary pattern was associated with slower biological aging, as estimated by multiple DNA methylation clocks, in older adults, although the evidence remains observational [24]. Therefore, we conducted an exploratory randomized controlled trial in overweight men to examine whether a pragmatic multicomponent lifestyle intervention could induce detectable short-term changes in DNA methylation–based measures of biological aging. The intervention combined individualized physical activity and dietary counseling with daily intake of yogurt containing Bifidobacterium longum BB536 (BB536). Based on prior reports suggesting that BB536 has multiple components relevant to aging-related pathways (e.g., inflammatory and oxidative stress signaling) [25, 26], it was included as one element from food components of a multicomponent lifestyle intervention intended to target multiple aging-related pathways concurrently.

Results

Male participants aged 50-74 years were recruited and randomized to the intervention group (INT, n = 24) or control group (CON, n = 24) using predefined allocation factors (chronological age and BMI) as shown in Figure 1. Per-protocol analyses were conducted in INT (n = 21) and CON (n = 24). Baseline characteristics of each group are summarized in Table 1, and participant-level baseline values are provided in Supplementary Table 4. No significant differences were observed between the two groups in chronological age, body weight, BMI, or smoking habits. During intervention, all participants in the INT group complied with exercise guidance (Supplementary Table 1). As the intervention incorporated diet and exercise counseling, an effect on body weight was anticipated. Indeed, following the intervention, body weight and BMI were significantly reduced in the INT group (Table 2).

Figure 1. CONSORT flow diagram of the study participants. The flow diagram illustrates the progress of participants through the phases of the trial. Forty-eight participants were randomized (INT, n = 24; CON, n = 24); in the INT group, one participant withdrew for personal reasons and two were excluded because unintended lifestyle changes introduced after randomization, resulting in per-protocol samples (PPS) of 21 in the INT group and 24 in the CON group, respectively.

To explore how the intervention might influence DNA methylation–based aging biomarkers across multiple dimensions, we examined changes in the pace of aging using DunedinPACE, alongside changes in current estimation of biological age (PC-HannumClock, PC-HorvathClock, PC-PhenoAge, GrimAge, and GrimAge v2) and GrimAge-derived surrogate markers. For DunedinPACE the adjusted mean change was −0.022 (95%CI −0.039 to −0.004) in the INT group and +0.001 (95%CI −0.015 to +0.017) in the CON group, resulting in a modest but statistically significant between-group difference of −0.023 (95%CI −0.045 to −0.001; p = 0.045; Figure 2A). Given that DunedinPACE is a pace of aging measure with 1.0 indicating approximately one biological year per chronological year, this corresponds to an approximately 2.2% slowing in the pace of aging within the INT group and a 2.3% slower pace of aging relative to controls over 12 weeks. Within the biological age family, no clock remained statistically significant after false discovery rate (FDR) correction. Detailed results for all biological age and raw values are shown in Supplementary Tables 1, 3, respectively. Prior to FDR adjustment, GrimAge showed a nominal between-group difference (p = 0.049; Figure 2B), but this did not withstand correction for multiple testing. No significant associations were observed between changes in DNA methylation–based aging measures and body weight reduction (Figure 3).

Figure 2. Effect of the multimodal lifestyle intervention on DNA methylation-based clocks. Data are presented as adjusted mean changes from baseline with 95% confidence intervals (CI). Open and black circles represent an adjusted mean change in the CON and INT group, respectively. Scale bars show 95% CI. The dotted line indicates zero. Asterisks (*) indicate significant between-group differences by ANCOVA (p < 0.05). Panels show; (A) DunedinPACE; (B) GrimAge; (C) GrimAge v2; (D) PC-Horvath clock; (E) PC-Hannum clock; (F) PC-PhenoAge.

Figure 3. Correlation between changes in body weight and DNAm-based aging measures. Scatter plots illustrating associations between changes in body mass index (ΔBMI, kg/m²) and changes in DNA methylation-derived age acceleration (week 12 minus baseline) in the INT group (n = 21). Each point represents an individual participant; solid lines depict simple linear regression fits, shown for visualization only. Pearson correlation coefficients (p) and corresponding p-values are shown in each panel. Panels show; (A) ΔDunedinPACE vs. ΔBMI; (B) ΔGrimAge vs. ΔBMI; (C) ΔGrimAge v2 vs. ΔBMI; (D) ΔPC-Horvath clock vs.ΔBMI; (E) ΔPC-Hannum clock vs. ΔBMI; (F) ΔPC-PhenoAge vs. ΔBMI.

In exploratory analyses of GrimAge-derived surrogate markers, including DNAm-based estimators of plasma proteins (e.g., DNAmCystatinC, DNAmADM, DNAmPAI-1, DNAmGDF15), DNAmCystatinC showed a significant between-group difference and remained significant after FDR correction (p = 0.001; q < 0.01; Figure 4A). Detailed results for all GrimAge-derived surrogate markers and raw values are provided in Supplementary Tables 2, 3, respectively.

Figure 4. Exploratory analyses of DNAm-based surrogate biomarkers of GrimAge. Data are presented as adjusted mean value in DNAm-based surrogate biomarkers of GrimAge with 95% confidence intervals (CI) at post-intervention. Open and black circles represent an adjusted mean value in the CON and INT group, respectively. Scale bars show 95% CI. The dotted line indicates zero. Asterisks (*) indicate significant between-group differences by ANCOVA. Panels show; (A) DNAmCystatinC; (B) DNAmADM; (C) DNAmB2M; (D) DNAmGDF15; (E) DNAmLeptin; (F) DNAmPAI1; (G) DNAmTIMP1.

Discussion

This exploratory randomized controlled trial showed that a multicomponent lifestyle intervention combining dietary guidance, physical activity counseling, and daily intake of yogurt containing BB536 was associated with short-term change in DNA methylation-based aging biomarkers in overweight, middle-aged to older men. Specifically, DunedinPACE, a measure of the current pace of aging that has been prospectively associated with risk of mortality and cardiovascular disease, showed a modest but nominally statistically significant between-group difference over 12 weeks. In contrast, no clock within the biological-age family remained significant after FDR correction. Exploratory analyses of GrimAge-derived surrogate markers identified reduction in DNAmCystatinC as a supportive finding.

In the control group, DunedinPACE changed minimally (+0.001 units, ≈+0.1%), consistent with expected natural drift and the high test–retest reliability of this measure [12]. By contrast, the ≈2.2% slowing of DunedinPACE in the INT group over 12 weeks is modest but within the same order of magnitude as the ≈2–3% slowing reported over 2 years of 25% caloric restriction in the CALERIE trial and the ≈3 months of protection observed over 3 years of omega-3 supplementation in DO-HEALTH across multiple clocks, including DunedinPACE [13, 22]. These comparisons suggest that, if sustained, the magnitude of short-term changes observed here is within the range previously associated with longer-term lifestyle interventions. However, their clinical relevance requires confirmation in adequately powered, long-term trials.

Our findings align with prior evidence linking adverse cardiometabolic profiles to accelerated epigenetic aging, particularly as captured by DunedinPACE. Higher adiposity, metabolic syndrome, and pro-inflammatory dietary patterns have been associated with DunedinPACE acceleration, often exceeding associations observed for other DNA methylation clocks [23], whereas adherence to anti-inflammatory dietary patterns and favorable cardiovascular risk profiles has been associated with slower DunedinPACE [27, 28]. Within this context, the DunedinPACE deceleration observed here may, at least in part, reflect the combined effects of (i) optimized metabolic regulation and reduced lipotoxicity, (ii) enhanced insulin sensitivity, vascular function, and muscle maintenance via regular exercise, and (iii) improved gut homeostasis and lower systemic inflammation. Importantly, DunedinPACE deceleration in the INT group was not statistically correlated with the BMI reduction (Figure 3). In exploratory analysis, the weekly frequency of exercise sessions was not associated with DunedinPACE (Supplementary Figure 1). However, because energy intake, diet quality, or expenditure were not quantified using validated instruments, and exercise intensity/volume (e.g., METs-minutes) was not measured, mediation analyses and interpretation of diet- or activity-linked effects are limited. Taken together, these observations suggest that neither weight loss alone nor exercise habits alone fully account for the observed slowing of biological aging. We hypothesize that additive or integrated modulation of systemic inflammation through microbiota-derived anti-inflammatory metabolites such as tryptophan-derived metabolites [29] and SCFAs [30], alongside diet [31] and exercise [32], may contribute, even if individual main effects are not detectable.

Among the DNAm-based surrogate protein components constituting GrimAge, the renal function-related marker DNAmCystatinC showed a significant between-group difference, with a clear decrease in the intervention group relative to controls, and remained significant after FDR correction within the surrogate-marker family (p = 0.001; q < 0.01, Figure 4A and Supplementary Table 3). Directionally similar but nonsignificant change was also observed in DNAmB2M (Supplementary Table 3), which is incorporated into the GrimAge algorithm as a surrogate for glomerular filtration [10]. GrimAge showed only a nominal between-group difference, but this did not remain significant after FDR correction. These findings may suggest that the observed signals were more localized to renal-related components within the GrimAge framework than broadly reflected across composite epigenetic age measures. Accordingly, the decrease in DNAmCystatinC, together with the directionally similar change in DNAmB2M, may indicate reduced physiological renal stress.

Taken together, the observed decrease in DNAmCystatinC may be interpreted as a supportive signal within a renal- and metabolic-related physiological domain [33, 34]. One biological hypothesis for future investigation is that microbiota-related alterations in indole metabolism may contribute to this domain, as suggested by the potential roles of indoxyl sulfate and indole-3-lactic acid (ILA). Indole, produced by gut microbiota from tryptophan, is converted in the liver to indoxyl sulfate—a compound that promotes chronic kidney disease through oxidative stress, inflammation, and fibrosis—whereas Bifidobacterium species, including BB536, can convert indole into ILA, a beneficial metabolite with anti-inflammatory and potentially renoprotective properties [26, 35–38]. As this study assessed only DNA methylation-based surrogate markers, the underlying biological mechanisms remain unresolved. Future investigations employing transcriptomic and microbiota analyses will be critical to clarify these mechanisms.

From a translational standpoint, it is noteworthy that the dietary component of our intervention emphasized moderating overeating based on daily food records rather than imposing stringent caloric restriction. Adherence to the prescribed exercise program was excellent, with participants completing an average 5.2 sessions per week and thereby achieving approximately 156 min per week, which meets the lower threshold of WHO 2020 recommendation for adults to maintain their health [39] (Supplementary Table 5). Further, psychological stress appears to be linked to epigenetic age acceleration in prior studies [16, 17, 40], raising concern that intensive lifestyle programs could paradoxically attenuate benefits via stress-related pathways. In our trial, however, Profile of Mood States 2 (POMS 2) total mood disturbance scores did not change within or between groups (Table 2), suggesting minimal psychological burden. Notably, although participants did not habitually consume yogurt at breakfast before the trial, adherence to daily BB536-containing yogurt intake was high (≥99%), indicating that this single dietary change was acceptable and sustainable over the intervention period. Taken together, these observations suggest that modest adjustments in diet and physical activity, combined with a realistic and easily adoptable habit of daily BB536-containing yogurt consumption, may constitute a feasible lifestyle strategy for inducing short-term changes in biomarkers related to the pace of biological aging, a hypothesis that warrants confirmation in larger, long-term trials.

Several limitations should be acknowledged. First, because the intervention combined dietary counseling, exercise, and BB536-containing yogurt, the present study cannot attribute the observed epigenetic changes to any single component or their interaction. Accordingly, the findings should be interpreted at the level of the intervention package. Second, this trial was exploratory in nature and was not powered or pre-registered to test a single primary epigenetic endpoint. Accordingly, observed patterns across multiple clocks and DNAm surrogates should be regarded as hypothesis-generating rather than definitive. Third, the relatively small sample size (n = 45) and short intervention duration (12 weeks) limit statistical power and preclude conclusions regarding long-term sustainability of epigenetic changes or clinical outcomes. Although baseline differences in DNA methylation-based measures were adjusted for in the ANCOVA analyses, the small sample size means that some residual imbalance may persist despite randomization. Furthermore, the durability of the observed short-term changes awaits confirmation in future studies, which will be important for clarifying the long-term impact of such interventions. Fourth, the study population consisted exclusively of overweight Japanese men aged 50–74 years, restricting generalizability to women, younger individuals, or other ethnic groups. Previous research in a Western population has suggested that similar multimodal lifestyle interventions could reduce epigenetic aging in both men and women [16, 17], although these trials used different epigenetic clocks and study designs. Future studies should therefore explicitly include female participants to confirm broad applicability. Fifth, given the open-label design, blinding was not feasible, and participants’ awareness of group assignment may have led to additional unmeasured health behavior changes in the INT group. Finally, although the data suggest several hypotheses, mechanistic pathways—such as the role of Bifidobacterium-derived metabolites in modulating systemic inflammation and epigenetic aging—remain speculative and cannot be evaluated within the present design, as noted above. Future research may also benefit from complementing DNA methylation-based measures with RNA-based transcriptomic aging measures, which may be more responsive to rapid biological changes in short-term trials. Large-scale, long-duration randomized trials will then be required to assess the durability of epigenetic changes, active components, and their clinical relevance.

Materials and Methods

Supplementary Materials