The study, led by Raphaël Waziry of Columbia University Irving Medical Center together with colleagues including Sara Hägg of the Karolinska Institutet, tackled a question that has long hovered over the booming field of geroscience: do the different biomarkers of aging, measured in different tissues, actually tell the same story about the same person? Aging researchers currently rely on a patchwork of measures. Telomere length tracks the protective caps on chromosomes, which erode with each cell division. Epigenetic clocks read chemical tags called methyl groups scattered across the genome, whose patterns shift predictably as we age. Physiology-based measures combine routine blood chemistry values, such as albumin, creatinine, glucose, and C-reactive protein, into a composite estimate of biological age. Whether these domains of aging biology converge within a single individual, and whether a measurement in one tissue reflects measurements in another, had remained largely unresolved.

To answer it, the team turned to the Health and Retirement Study, a nationally representative longitudinal survey of more than 37,000 Americans aged 50 and older, coordinated by the University of Michigan’s Institute for Social Research. Participants who provided saliva samples during the 2008 interview wave had their average telomere length measured using quantitative PCR, a technique that compares the copy number of telomere sequences to a single-copy gene to produce a ratio proportional to mean telomere length. Eight years later, in 2016, many of the same participants contributed venous blood samples, which were processed within 24 to 48 hours and analyzed at a CLIA-certified laboratory at the University of Minnesota. In total, 2,406 individuals had complete data on both salivary telomere length and blood-based physiology measures, and 1,029 also had DNA methylation data, allowing direct within-person comparisons across tissues.

The DNA methylation analysis was unusually thorough. Rather than relying on a single epigenetic clock, the researchers computed eight of them, spanning the field’s history from first-generation chronological age predictors to second-generation mortality-focused algorithms. These included the original Horvath pan-tissue clock built on 353 CpG sites across 51 healthy tissues and cell types; the Hannum clock, comprising 71 CpG sites developed in whole blood; PhenoAge, based on 513 CpG sites and trained on phenotypic age; GrimAge, which incorporates epigenetic surrogates for a dozen plasma proteins along with smoking pack-years and is considered among the strongest predictors of time to death, coronary heart disease, and cancer; the skin and blood clock built on 391 CpGs; and the more compact Lin, Weidner, and VidalBralo clocks, which use 99, three, and eight CpG sites respectively. On the physiology side, the team used phenotypic age, an algorithm derived from Gompertz proportional hazard models that integrates nine blood biomarkers: albumin, creatinine, glucose, log C-reactive protein, lymphocyte percent, mean cell volume, red blood cell distribution width, alkaline phosphatase, and white blood cell count.

The headline result was strikingly consistent. Across every blood-based measure examined, longer salivary telomeres were associated with slower biological aging, an inverse relationship that held after adjusting for sex, the interaction between sex and telomere length, and the eight-year gap between the saliva and blood collections. When participants were grouped into tertiles of telomere length, those in the highest tertile showed biological ages roughly one to four years younger than those in the lowest tertile, depending on the clock. The largest effects appeared for the compact Weidner clock, with a beta coefficient of minus 3.97 years, followed by Lin at minus 3.45 years and GrimAge at minus 3.33 years, the latter highly statistically significant. The pattern persisted even in sensitivity analyses that excluded telomere measurements with a T/S ratio above 2.0, values more likely to be artifacts in salivary samples, and when the sex interaction term was dropped from the models.

Notably, the cross-tissue correlations, while statistically significant, were modest in magnitude, with coefficients such as minus 0.067 for the Lin clock and minus 0.064 for Weidner. This stands in sharp contrast to the intra-tissue comparisons within blood, where DNA methylation clocks correlated strongly with the physiology-based phenotypic age. GrimAge led the pack with a correlation coefficient of 0.75, followed by Hannum at 0.68 and the skin and blood clock at 0.66, all highly significant. The interpretation is nuanced: within a single tissue, the different domains of aging biology move together robustly, but the signal weakens when it must travel across tissues, time, and measurement technologies. An eight-year interval between samples, which the authors adjusted for but could not eliminate, likely contributes to this attenuation, as does the well-documented fact that saliva and blood differ substantially in the concentration and abundance of their molecular analytes.

One of the most intriguing findings emerged from the finer-grained analysis of individual physiology biomarkers. When each of the nine blood analytes was correlated separately with the DNA methylation clocks, the strength of association varied widely. Creatinine, a marker of kidney function, and lymphocyte percent, a marker of immune status, showed the strongest correlations with epigenetic aging measures, in some cases exceeding their correlations with chronological age itself. Alkaline phosphatase, by contrast, showed little relationship with any of the clocks. The authors suggest this points to potentially prominent roles for renal function and immunity in accelerating or decelerating biological aging at the epigenetic level, a hypothesis that aligns with longstanding observations that immune cell composition shifts and kidney decline are hallmarks of the aging process.

The study also revealed sex-based variation. In models stratified by sex, the effect estimates linking each tertile increase in telomere length to blood-based aging measures differed between males and females across the various clocks. This was anticipated by the researchers, who included a sex-by-telomere interaction term in their primary models based on prior evidence that telomere dynamics differ between the sexes, with women generally maintaining longer telomeres than men of the same chronological age. The precise biological underpinnings of these differences, whether hormonal, inflammatory, or related to differential exposures across the life course, remain an open question for future work.

The practical implications could prove significant. Saliva is among the most accessible human tissues: collection requires no needles, no trained phlebotomist, and minimal equipment, and samples are cheaper and easier to ship and store than blood. It has already proven its worth in therapeutic drug monitoring for conditions such as epilepsy, diabetes, and multiple sclerosis, and in clinically difficult populations where venous access is not feasible. The new findings suggest that a salivary telomere measurement, even taken years before a blood draw, carries real information about a person’s epigenetic and physiological aging status. For vulnerable populations, including frail elderly patients, young children, and people in low-resource settings, this opens the possibility of integrating aging biology monitoring into routine care using whichever tissue is accessible. The authors caution, however, that saliva composition is influenced by circadian rhythms, medications, sympathetic and parasympathetic nervous activity, and conditions ranging from diabetes to infectious disease, and that immune cell contamination can confound salivary DNA methylation measurements.

The researchers are candid about the limitations. The biomarkers were measured at two different time points, introducing potential imprecision despite statistical adjustment. Only telomere length was available in saliva, so no epigenetic clocks could be computed from that tissue. Each individual was measured only twice, precluding analysis of within-person trajectories over long periods. And because participants were aged roughly 59 to 72 at saliva collection, the results cannot be extrapolated to younger adults. Still, the study’s scale, national representativeness, and unusually comprehensive battery of aging measures make it one of the clearest demonstrations to date that the body’s aging clocks are connected across tissues. As epigenetic testing becomes cheaper and repeated measurement becomes more feasible, the convergence of telomere, methylation, and physiology-based measures, whether sampled from blood or saliva, may form the foundation for the surrogate endpoints that geroscience trials have long sought, and for a future in which a routine swab of the cheek helps clinicians track how quickly, or slowly, their patients are growing old.

Subject of Research: Cross-tissue correspondence between salivary telomere length and blood-based DNA methylation and physiology biomarkers of biological aging in older adults

Article Title: Connections between cross-tissue and intra-tissue biomarkers of aging biology in older adults

Article References: Waziry, R., Gu, Y., Williams, O., & Hägg, S. (2023). Connections between cross-tissue and intra-tissue biomarkers of aging biology in older adults. Epigenetics Communications, 3(1), Article 7. https://doi.org/10.1186/s43682-023-00022-4

Image Credits: AI Generated

DOI: 10.1186/s43682-023-00022-4

Keywords: biological aging, telomeres, DNA methylation, epigenetic clocks, saliva biomarkers, blood biomarkers, GrimAge, phenotypic age, Health and Retirement Study, geroscience, biomarkers, older adults

The research drew on the Longitudinal Aging Study in India Diagnostic Assessment of Dementia, known as LASI-DAD, an in-depth sub-study of the nationally representative Longitudinal Aging Study in India. The investigators analyzed data from 1,929 adults aged 60 and older who had measurements of one-carbon metabolism biomarkers and the clinical biomarkers needed to estimate Phenotypic Age at the first visit, conducted between 2017 and 2019. A subset of 922 participants also had DNA methylation data, enabling the team to compute epigenetic aging measures. For longitudinal analyses, 886 participants had repeated Phenotypic Age measurements and 583 had repeated epigenetic measurements at a second visit between 2022 and 2024.

Biological aging was assessed in two complementary ways. The first was Phenotypic Age, a validated clinical measure derived from nine blood biomarkers and chronological age that reflects metabolic, inflammatory, renal, hepatic, hematologic, and immune function; acceleration was defined as the deviation of Phenotypic Age from what would be expected for a person’s chronological age. The second approach used DNA methylation profiles generated from peripheral blood on Illumina EPIC v2.0 arrays, covering roughly 928,000 CpG sites after quality control. From these data the researchers computed four second- and third-generation epigenetic measures: DNAm PhenoAge, GrimAge2, DunedinPACE, and SystemsAge. DunedinPACE estimates the current pace of aging relative to chronological time, while SystemsAge quantifies aging across eleven physiological systems.

The baseline picture underscored why India is an important setting for this work. Mean age in the cohort was 69.4 years, most participants lived in rural areas, and hyperhomocysteinemia—defined as homocysteine above 16.2 micromoles per liter—affected nearly two thirds of participants, rising to almost 75 percent among men. Vitamin B12 deficiency below 200 picograms per milliliter was present in 34.5 percent of participants and folate deficiency below 3.0 nanograms per milliliter in 13.6 percent, both more common in men. Current smokers and participants with chronic kidney disease showed higher rates of folate deficiency and elevated homocysteine, and PhenoAge acceleration was greater among smokers and those with kidney disease.

In cross-sectional analyses adjusted for age, sex, education, body mass index, smoking, alcohol use, residence, region, kidney function, and comorbidity burden, each doubling of homocysteine was associated with 3.38 additional years of Phenotypic Age acceleration, while each doubling of folate was associated with 1.05 fewer years and each doubling of vitamin B12 with 0.72 fewer years. The longitudinal results told a similar story: within-person increases in homocysteine over the follow-up period corresponded to roughly 3.17 years greater acceleration in Phenotypic Age, whereas increases in folate corresponded to about 1.10 years slower aging. Changes in vitamin B12 were not significantly associated with changes in Phenotypic Age acceleration, and additional adjustment for C-reactive protein did not materially alter the findings, suggesting that systemic inflammation explains only part of the relationship.

The epigenetic analyses revealed a more differentiated pattern. Higher folate was consistently linked to slower molecular aging, with each doubling associated with lower DunedinPACE and lower SystemsAge acceleration both cross-sectionally and longitudinally, with the longitudinal estimates reaching 0.16 units lower DunedinPACE and 0.19 standard deviations lower SystemsAge acceleration per doubling of folate. Associations for vitamin B12 and homocysteine with the epigenetic clocks were smaller and less consistent, and vitamin B12 was paradoxically associated with slightly higher DNAm PhenoAge acceleration in cross-sectional analysis. The authors interpret this heterogeneity as reflecting the distinct biological dimensions captured by different clocks rather than inconsistent biology: Phenotypic Age reflects multisystem physiological dysregulation, while DunedinPACE and SystemsAge are more responsive to dynamic aging processes.

Perhaps the most striking results came from analyses of biomarker transitions. Participants who moved from normal folate status to folate deficiency between visits showed 2.99 years greater increase in Phenotypic Age acceleration than those who remained normal, along with faster DunedinPACE, greater DNAm PhenoAge acceleration, and greater SystemsAge acceleration—the latter reaching 0.71 standard deviations and remaining significant after correction for multiple testing. Participants with persistently elevated homocysteine showed 2.47 additional years of Phenotypic Age acceleration compared with those who stayed in the normal range. These within-person transition analyses, the researchers note, provide the first longitudinal evidence that shifts in one-carbon metabolism status are accompanied by shifts in biological aging measured both clinically and epigenetically.

Sex differences emerged for the molecular measures. The inverse association between folate change and DunedinPACE was stronger in men than in women, and the positive association between homocysteine change and SystemsAge acceleration was also stronger among men, with interaction terms remaining significant after false discovery rate correction. The authors caution that these exploratory interaction analyses involved many comparisons, but they suggest that differences in nutritional status, smoking patterns, renal function, hormonal regulation, and environmental exposures may modify how one-carbon metabolism relates to molecular aging in men and women.

Mechanistically, the findings align with the geroscience hypothesis, which holds that targeting fundamental mechanisms of aging may delay multiple age-related diseases. One-carbon metabolism supplies methyl groups for DNA synthesis and methylation, processes central to genomic stability and epigenetic regulation, and it is connected to glutathione metabolism, redox homeostasis, mitochondrial function, and inflammatory signaling. Elevated homocysteine has been linked to oxidative stress, endothelial dysfunction, and neurodegeneration, while adequate folate supports cellular repair. Consistent with this, prior studies in high-income populations—including analyses of the National Health and Nutrition Examination Survey and trials such as VITACOG, CALERIE, COSMOS, and DO-HEALTH—have reported links between methyl donor nutrients, homocysteine, and epigenetic aging measures, and some interventions have produced modest shifts in DNA methylation clocks.

The study has limitations that the authors acknowledge. As an observational analysis, it cannot establish causation, and residual confounding by diet, supplementation, medication use, and other unmeasured factors cannot be excluded. Circulating biomarkers reflect integrated physiology rather than isolated dietary mechanisms, and vitamin B6, an essential cofactor in the pathway, was not measured in LASI-DAD, leaving the full one-carbon pathway incompletely characterized. Even so, the researchers argue that one-carbon metabolism emerges as a potentially modifiable correlate of biological aging, particularly relevant in low- and middle-income settings where micronutrient deficiencies are common. With India projected to have more than 340 million older adults by 2050, they conclude that future mechanistic and interventional studies should test whether optimizing folate status and lowering homocysteine can slow biological aging and extend health span.

Subject of Research: Longitudinal associations between one-carbon metabolism biomarkers and biological and epigenetic aging in older adults in India

Article Title: Longitudinal associations of one-carbon metabolism biomarkers with biological and epigenetic aging in older adults in India

Article References: Wang, J., Crimmins, E., Kim, J. K., Chang, N.-S., Dey, S., Lee, J., Thyagarajan, B., & Vivek, S. (2026). Longitudinal associations of one-carbon metabolism biomarkers with biological and epigenetic aging in older adults in India. GeroScience. https://doi.org/10.1007/s11357-026-02505-3

Image Credits: AI Generated

DOI: 10.1007/s11357-026-02505-3

Keywords: one-carbon metabolism, folate, homocysteine, vitamin B12, biological aging, epigenetic clocks, Phenotypic Age, DunedinPACE, SystemsAge, LASI-DAD, older adults, India