Prolonged exposure to leisure screen time notably accelerates biological aging: Evidence from observational studies and genetic associations

Abstract

LST is steadily increasing and is associated with various health issues. However, its impact on aging remains unclear. A total of 7212 participants from NHANES 1999–2002 were included. LTL, ALM, and FI were selected as aging phenotypes. Observational association between LST and aging traits was analyzed using linear regression models. MR analyses based on 112 genetic variants were performed to test the causal estimates from LST on aging. TWAS and PPI analyses were conducted to investigate underlying biological mechanisms. After adjusting for physical activity, per 1 ​h increase in LST, participants had a shorter LTL (β ​= ​−1.39, 95 ​% CI: −2.47 to −0.30), a lower ALM (β ​= ​−1.09, 95 ​% CI: −1.39 to −0.70), and an increased FI (β ​= ​8.22, 95 ​% CI: 4.29 to 12.30). Likewise, TSMR analyses indicated that genetically increased LST was significantly associated with shorter LTL (β ​= ​−2.63, 95 ​% CI: −4.86 to −0.35), lower ALM (β ​= ​−6.56, 95 ​% CI: −9.43 to −3.60), and increased FI (β ​= ​20.16, 95 ​% CI: 15.73 to 24.77). The trend remained robust after tests for pleiotropy and heterogeneity, consistent with the results of MVMR. 4 hub genes and 15 co-localized genes are identified, respectively, from PPI networks and TWAS. Pathways related to immune reactions, oxidative stress, and protein metabolism were significantly enriched. This study revealed that increased LST is significantly associated with adverse aging phenotypes. Reducing LST may help alleviate the burden of aging.

Introduction

With the progress of global aging, the aging epidemic has increasingly been a socio-economic burden. Aging always occurs with various diseases, such as cancers, cardiovascular diseases, and frailty syndrome [1]. To better understand the phenotype and process of aging, Otín et al. illustrated a nine-feature model, mainly including genomic instability, telomere attrition, loss of proteostasis, mitochondrial dysfunction, and altered intercellular communication [2]. Telomeres, acting as the protective caps of chromosomes, shorten with age, leading to deoxyribonucleic acid (DNA) damage accumulation and resulting in senescent cell phenotypes and apoptosis [3]. Therefore, short leukocyte telomere length (LTL) has been established as a sign of systemic aging [4]. According to these features and recent literature, we adopted several aging-related traits, including LTL (an indicator for biological aging), sarcopenia phenotypes (appendicular lean mass, ALM), and frailty index (FI) in recent biobank databases [5].

Over the years, digital technology has been more accessible to people worldwide, directly leading to increased leisure screen time (LST) and long sedentary behaviors [6]. LST is defined as time spent in screen behaviors unrelated to school or work [7], such as watching TV or videos, or using a computer. While numerous researchers have recognized the benefits of physical activity (PA) for healthy aging [8], growing evidence has proved that screen time-based sedentary behavior is associated with various adverse aging-related outcomes, independent of physical activity [9], including the increased risk of cardiovascular disease mortality [10], and increased incidence of various cancers [11]. However, few researchers have investigated the association between aging and LST, and controversial results exist. Some reported that higher sedentary time is associated with shorter LTL [1,[12], [13], [14]], while others revealed no significant association between them [15].

Considering the weak strength of evidence in observational studies and conflicting results of existing research, we employed Mendelian randomization (MR), a genetically epidemiologic method, which is not disturbed by confounding effects and reverse causality by using genetic variants as instrumental variables (IVs), to comprehensively assess the genetic relationship and potential causal association between LST and aging. Besides, two-sample MR (TSMR) and multivariable MR (MVMR) methods, as well as transcriptome-wide association studies (TWAS), were also conducted to explore shared genes and biological pathways.

Methods

Database and study population

The National Health and Nutrition Examination Survey (NHANES) is a health program conducted by the National Center for Health Statistics (NCHS) and is a part of the Centers for Disease Control and Prevention (CDC) [16]. The whole program began in the early 1960s with intermittent surveys of specific population groups or health topics. NHANES releases a continuous, comprehensive, cross-sectional survey every two years that focuses on measuring various health and nutritional statuses of Americans. Participants will be interviewed at home and examined at a Mobile Examination Center (MEC) to collect anthropometric and physiological findings and blood samples.

This study used publicly available data from the NHANES between 1999 and 2002. The total population was 21,004. Afterward, we excluded participants younger than 18 (n ​= ​9563) or with incomplete interviews or examinations (n ​= ​4229). Finally, there were a total of 7212 participants (3533 females and 3679 males) in our study. All participants provided informed consent, and the NCHS Ethics Review Board approved the program.

Leisure screen time

LST was evaluated through a questionnaire asking as follows: I will now gather information regarding your television (TV) watching and computer usage habits. In the last 30 days, how much time did you generally spend daily sitting and engaging in activities such as watching television or videos, or using a computer for non-work purposes? Response options included none, less than 1 ​h, 2 ​h, 3 ​h, 4 ​h, and 5 ​h or more. In this study, we begin by combining none, less than 1 ​h, and 1 ​h into “≤ 1” and renaming 5 ​h or more to “≥ 5” for a more concise representation. After that, to investigate the linear relationship between LST and other factors, we stipulated and redefined the samples of none, less than 1 ​h, and 1 ​h as 1 ​h, the samples of 2 ​h as 2 ​h, the samples of 3 ​h as 3 ​h, the samples of 4 ​h as 4 ​h, and the samples of 5 ​h or more as 5 ​h.

Leukocyte telomere length

LTL was presented as the mean T/S ratio, which is measured as telomere length relative to standard reference DNA. After blood samples were obtained from MECs, a small portion of each sample would be stored to conduct later analyses on DNA samples. The telomere length surveying was performed in the laboratory of Dr. Elizabeth Blackburn at the University of California, using the quantitative polymerase chain reaction (PCR) method to measure the T/S ratio [17]. Each sample would be tested three times in three days. Samples were assayed in duplicate wells, resulting in 6 data points. Control DNA values are used to normalize differences between runs. If more than four control DNA values were 2.5 standard deviations from the mean of all assay runs, they were excluded from further analysis (<6 ​% of runs). Potential outliers were identified and excluded from the calculation for each sample (<2 ​% of samples). LTL was not normally distributed; thus, we made it naturally log-transformed in the study.

Arm lean mass

ALM is a valuable tool for evaluating muscle health, particularly in older adults and those at risk for sarcopenia, a condition characterized by age-related loss of muscle mass and function. As an objective measure of muscle mass, ALM is an important indicator of overall body composition and physical health. Participants went to the MECs and received dual-energy X-ray absorptiometry (DXA), and each survey would have five records since missing or invalid data had been multiplied. The DXA scans provide bone and soft tissue measurements for the total body, for both arms and legs, the trunk, and the head. As for this study, we need arm lean mass (ALM), which contains “Right Leg Lean excl Bone Mineral Content (grams)”, “Right Arm Lean excl Bone Mineral Content (grams)”, “Left Leg Lean excl Bone Mineral Content (grams)”, and “Left Arm Lean excl Bone Mineral Content (grams)”, each item took the mean and added up. It is not meaningful to calculate limb muscle mass without considering the impact of individual fitness levels on the results. Thus, in order to eliminate individual differences, we took the ALM in kilograms and then adjusted it with body mass index (BMI) as described in previous studies [18,19]. For consistency with LTL, we naturally logarithmize ALM.

Fragility index

A total of 49 deficits were accumulated for calculating FI, including one mark on cognition, 16 marks on dependence, seven depressive symptoms, 13 comorbidities, five marks on hospital utilization and access to care, one mark on physical performance and anthropometry, and six laboratory values. FI was also naturally log-transformed.

Covariate

We chose covariates based on existing studies, including age, gender, ethnicity, BMI, poverty, marital status, hypertension, diet, education, smoking, and physical activity, which were collected through questionnaires, feedback, and interviews. A number of options are available with regard to ethnicity: “Non-Hispanic White”, “Mexican American”, “Non-Hispanic Black”, “Other Hispanic”, and “Other Race - Including Multi-Racial”. BMI was calculated by dividing weight in kilograms by the square of height in meters. Poverty was defined as a ratio of family income to the poverty threshold. Marital status included married, widowed, divorced, separated, never married, and living with a partner. Education was assessed by “What is the highest grade or level of school completed or the highest degree have you received?” and the results could be “Less Than 9th Grade”, “9-11th Grade (Includes 12th grade with no diploma)”, “High School Grad/GED or Equivalent”, “Some College or AA degree”, and “College Graduate or above”, the variables were then rewritten in the form of three new variables: “Below High School”, “High School Grad/GED or Equivalent”, and “After High School”. The Dietary Inflammatory Index (DII) was employed to evaluate the dietary patterns of the population. A value of either +1, −1, or 0 was assigned based on the effect of the food parameter on inflammation. A value of +1 was assigned if the effect was pro-inflammatory, indicating a significant increase in interleukin-1 beta (IL-1β), interleukin-6 (IL-6), tumor necrosis factor-α (TNF-α), or C-reactive protein (CRP), or a significant decrease in interleukin-4 (IL-4) or interleukin-10 (IL-10). Conversely, a value of −1 was assigned if the effect was anti-inflammatory, indicating a significant decrease in IL-1β, IL-6, TNF-α, or CRP, or a significant increase in IL-4 or IL-10. A value of zero was assigned in instances where the food parameter did not result in a notable alteration in inflammatory markers. We combined several items to determine smoking status as follows: participants who smoked less than 100 cigarettes in life or smoked more than 100 cigarettes in life but smoke not at all now would be categorized as not smoking; otherwise, people who smoked more than 100 cigarettes in life and smoked some days or every day would be defined as smoking. Hypertension was defined as having been diagnosed with hypertension, or taking anti-hypertensive medication, or if not, having a systolic blood pressure level ≥130 ​mmHg or a diastolic blood pressure level ≥80 ​mmHg. In order to measure the physical activity intensity of participants, we separately collected the average time that each interviewee spent walking or bicycling, on any physical activities specifically designed to strengthen muscles, and on any tasks in or around the home or yard for at least 10 ​min that required moderate or more tremendous physical effort, then added them up. In order to facilitate comparison, the various activity times were quantified using the metabolic equivalent (MET) metric as described in previous research [20]. The corresponding MET was derived from the oxygen consumption/(kg-min) during other activities. The oxygen consumption of a healthy adult in a quiet seated position was established as 1 MET, which is equivalent to 3.5 ml/(kg-min). We also made all covariates naturally log-transformed.

TSMR and MVMR

The genome-wide association studies data for LST and PA was obtained from the GWAS Catalog database (GCST90104339 and GCST90104341) published by Wang Z et al. [21] in 2022, which includes 526,725 European ancestry individuals. The GWAS data for LTL was obtained from the Integrative Epidemiology Unit (IEU) GWAS project (ieu-b-4879) published by Veryan Codd et al. [22] in 2021, which includes 472,174 samples from European populations. The GWAS data for ALM was obtained from the IEU GWAS project (ebi-a-GCST90000025) published by Yu-Fang Pei et al. [23] in 2020, which includes 450,243 samples from European populations. The GWAS data for the FI was obtained from the IEU GWAS project (ebi-a-GCST90020053) published by Janice L Atkins et al. [24] in 2021, which includes 175,226 samples from European populations.

We established a p-value threshold of 5 ​× ​10⁻⁸ for identifying single-nucleotide polymorphisms (SNPs) strongly associated with the exposure, which were designated as robust instrumental variants. Additional selection and the F-statistics for each SNP were calculated as described previously [25]. In TSMR analysis, we utilized three different methodologies: random-effects inverse-variance weighted (IVW), weighted median, and MR-Egger, as mentioned previously [25,26]. We conducted MR-Egger intercept analysis and leave-one-out sensitivity analysis to investigate horizontal pleiotropy [27], while Cochran's Q test was employed to evaluate heterogeneity [28]. MVMR enables the evaluation of the effects of multiple exposures on specified outcomes [29]. In this approach, we incorporated physical activity into MVMR to ascertain the most robust causal relationship between LST and aging.

TWAS and protein-protein interaction (PPI) analyses

In our transcriptome-wide association studies (TWAS) [30], we employed the FUSION methodology to transform GWAS data into TWAS, utilizing RNA-seq data from European whole blood samples (N ​= ​558) derived from the Genotype-Tissue Expression version 8 (GTEx v8) [31], as previously described [26].

As previously outlined [32], we employed the Search Tool for Interacting Genes (STRING) database (accessible at https://string-db.org) to construct Protein-Protein Interaction (PPI) networks for predicting the functional interactions among proteins.

Statistical analysis

Data are presented as mean ​± ​standard deviation (SD) for continuous variables, and for categorical variables, it is frequencies or percentages. Later, when analyzing the baseline characteristics, continuous variables were expressed as means with standard errors, and categorical variables were expressed as percentages with standard errors. We developed separate linear regression models to investigate the relationship between LST and other factors, including LTL, ALM, and FI. We used both unadjusted and multivariate-adjusted models. In this study, model 1 was unadjusted; model 2 was adjusted for age, gender, ethnicity, BMI, poverty, education, and hypertension; model 3 was adjusted for covariates in model 2 and marital status, smoking status, dietary pattern, and physical activity intensity. Afterward, we delineated three age groups on the basis of 40 and 60 and verified the existence of the linear relationship described above in different age groups and sexes. All analyses were performed using the R statistical packages (TwoSampleMR [33], ggplot2, cluster profile [26], enrich plot [27], and DOSE [28]) alongside the FUSION software [23], and a two-sided p-value of less than 0.05 was considered statistically significant.

Results

Baseline characteristics

The flowchart is presented in Fig. 1. Table S1 displays the weighted distribution characteristics of individuals in this study. In summary, most participants were female (50.53 ​%), with a mean age of 46.34 and a mean BMI of 28.09. Only 15.57 ​% of participants reported never being married. Most people were non-smokers (75.29 ​%). 21.36 ​% of the participants had less than a high school education. As for the aging traits, the mean LTL was 1.06, the mean ALM was 7.68, and the mean FI was 0.12. Less than one-third of the sample had a daily LST of no more than 1 ​h.

Fig. 1.

The flowchart of this study. The figure was built by the Biorender.

Linear association

The results of the linear regression analyses exploring the association between LST and aging are presented in Table 1. After adjustment for relevant covariates, there is a statistically significant association between LST status (≥5 ​h) and LTL (β ​= ​−2.37, P ​< ​0.05). Meanwhile, a linear trend existed (P_trend ​= ​1.50 ​× ​10⁻²) for a reduced LTL as LST increases (β ​= ​−1.39, 95 ​% CI: −2.47 to −0.30). Compared with participants with a daily LST ≤1 ​h, those with a daily LST of 2 ​h, 3 ​h, 4 ​h, 5 ​h, and above hours all had a significantly lower ALM (p ​< ​0.05). For a 1 ​h increase in LST, participants had a lower ALM (β ​= ​−1.09, 95 ​% CI: −1.39 to −0.70). Participants with a daily LST of 5 ​h or more exhibited an increased FI, and a linear growth was significant (P_trend ​< ​0.001) per 1 ​h increase in LST (β ​= ​8.22, 95 ​% CI: 4.29 to 12.30). In general, after adjusted for all covariates, per 1 ​h increase in LST, participants had a shorter LTL (β ​= ​−1.39, 95 ​% CI: −2.47 to −0.30), a lower ALM (β ​= ​−1.09, 95 ​% CI: −1.39 to −0.70), and an increased FI (β ​= ​8.22, 95 ​% CI: 4.29 to 12.30). Meanwhile, a comparison of the average treatment difference between the LST intervention and the control group at the one-year primary endpoint demonstrated that the impact of LST was equivalent to 0.06 years of LTL-related aging change, 0.10 years of ALM-related aging change, and 0.14 years of FI-related aging change.

Table 1.

Percent differences (95 ​%CI) in aging traits by LST in NHANES.

	Model 1a	Model 2b	Model 3c
Leukocyte telomere length
LST≤1	0 (ref)	0 (ref)	0 (ref)
LST ​= ​2	−1.49 (-3.15,0.20)	−0.20 (-1.98,1.51)	−0.30 (-2.47,1.92)
LST ​= ​3	−4.40 (-6.95,-1.78)∗∗	−2.57 (-5.07, 0.10)	−2.57 (-5.44, 0.30)
LST ​= ​4	−5.35 (-7.96,-2.66)∗∗∗	−2.57 (-4.97,-0.20)∗	−2.18 (-4.88, 0.60)
LST≥5	−4.02 (-6.57,-1.39)∗∗	−2.57 (-4.59,-0.40)∗	−2.37 (-4.78, 0.000)∗
P For trend	<0.001	7.00E-03	1.50E-02
Per 1 ​h increase	−2.66 (-4.02,-1.39)∗∗∗	−1.59 (-2.57,-0.50)∗∗	−1.39 (-2.47,-0.30)∗
Appendicular lean mass
LST≤1	0 (ref)	0 (ref)	0 (ref)
LST ​= ​2	2.33 (0.90,3.87)∗∗	−0.80 (-1.29,-0.30)∗∗	−0.80 (-1.39,-0.20)∗
LST ​= ​3	1.41 (-0.50,3.36)	−1.49 (-2.23,-0.70)∗∗∗	−1.29 (-2.18,-0.50)∗∗
LST ​= ​4	5.34 (3.05,7.79)∗∗∗	−1.49 (-2.47,-0.60)∗∗	−1.29 (-2.27,-0.20)∗
LST≥5	2.12 (0.10,4.08)∗	−2.86 (-3.44,-2.27)∗∗∗	−2.66 (-3.44,-1.98)∗∗∗
P For trend	0.003	<0.001	<0.001
Per 1 ​h increase	1.61 (0.70,2.43)∗	−1.19 (-1.49,-0.80)∗∗∗	−1.09 (-1.39,-0.70)∗∗∗
Frailty index
LST≤1	0 (ref)	0 (ref)	0 (ref)
LST ​= ​2	2.63 (-3.44,9.09)	−0.70 (-6.95, 5.97)	−0.60 (-8.06, 7.57)
LST ​= ​3	12.98 (4.81,21.65)∗∗	4.81 (-3.05, 13.31)	3.67 (-4.88, 13.09)
LST ​= ​4	20.90 (11.07,31.39)∗∗∗	4.19 (-3.92, 12.98)	4.50 (-5.35, 15.37)
LST≥5	45.64 (37.16,54.81)∗∗∗	28.15 (19.96, 36.75)∗∗∗	26.62 (17.82, 36.07)∗∗∗
P For trend	<0.001	<0.001	<0.001
Per 1 ​h increase	16.53 (12.86,20.44)∗∗∗	8.65 (4.81, 12.52)∗∗∗	8.22 (4.29, 12.30)∗∗∗

∗P ​< ​0.05; ∗∗P ​< ​0.01; ∗∗∗P ​< ​0.001.

^aUnadjusted model.

^bAdjusted for age, gender, ethnicity, BMI, poverty, hypertension, education.

^cAdjusted for age, gender, ethnicity, BMI, poverty, hypertension, education, marital status, dietary pattern, smoking status, and physical activity.

In the fully adjusted model, we further conducted subgroup analyses categorized by gender and age (Fig. 2). Compared with women, men were more likely to have shorter LTL under the influence of LST (β ​= ​−2.86, 95 ​% CI: −4.62 to −1.10). Participants more than 50 years old were more likely to have shorter LTL under the influence of LST than participants under 50 years old (p ​< ​0.05). Moreover, compared to other races, non-Hispanic white participants had shorter LTL, lower ALM and longer FI under the influence of the LST. However, the differences that existed between these subgroups were not statistically significant (P_interaction ​> ​0.05).

Fig. 2.

The fully adjusted model subgroupedby gender (A), age (B), and ethnicity (c) in the NHANES.

MR analysis

A total of 112 independent SNPs were utilized as IVs for LST (Table S2). The F-statistic for all included IVs exceeded the threshold of 10, indicating the robustness of genetic instruments in estimating potential causal effects. From the TSMR analyses (Fig. 3D), we observed that genetically increased LST was significantly associated with shorter LTL (β ​= ​−2.63, 95 ​% CI: −4.86 to −0.35), lower ALM (β ​= ​−6.56, 95 ​% CI: −9.43 to −3.60), and increased FI (β ​= ​20.16, 95 ​% CI: 15.73 to 24.77) using the IVW analysis. Such estimates were further demonstrated by the methods of MR-Egger and Weighted median (Table S3). We found a significant causal relationship between TV time and FI, ALM, and LTL. In addition, the potential for bidirectional relationships was also analyzed. We then validated the robustness of our results and found no pleiotropy or heterogeneity in our study (Fig. 3A–C, Tables S4–5). Sensitivity analysis using the leave-one-out method showed that the elimination of a single SNP did not bias the IVW results, enhancing the credibility of the results (Supplementary Figs. S1–3). The causal relationship between LST and aging remains significant after adjusting for physical activity in the MVMR analysis. Totally, our analysis suggests a causal relationship between genetically increased LST and aging.

Fig. 3.

MR analyses the causal association between LST and aging. The scatter plots from genetically predicted LST on LTL (A), ALM (B), and FI (C). The results of TSMR and MVMR analyses (D).

Gene-based analysis

To further understand the functional consequences of genetic variation and underlying molecular biology mechanisms of LST on aging, we performed TWAS on LST and aging traits. As LST was found to correlate positively with the FI, 90 shared genes were identified as potential target genes associated with the LST and FI in the same direction (Table S8). LST was also found to correlate negatively with LTL and ALM. 354 and 890 shared genes in the opposite direction were identified as potential targets for LST in LTL and ALM (Tables S6–7). Four common genes (HLA-DQA1, CYB561D2, GFRA2, GGT7) acting in all aging traits were extracted from the Venn diagram (Fig. 4A). Based on these four core genes, we constructed a PPI network to explore the association between these genes (Fig. 4B). Twenty associated genes circled around the four hub genes, giving insights into underlying biological mechanisms. We summarized the top 5 common genes shared between LST and aging traits according to FCP (Table S12), some of which were located at the same cytoband, including 20q13.33 (LIME1, ARFRP1, STMN3, DNAJC5, ZGPAT) and 17q23.3 (CCDC47, TACO1, SMARCD2) and 12q24.31 (SETD8, EIF2B1).

Fig. 4.

Identification of genes commonly linked between LST and aging by the Venn diagram (A) and protein-protein interaction (PPI) analysis (B).

Pathway-based functional enrichment analysis

Finally, we conducted pathway-based functional enrichment analyses for the shared genes above to identify the potential mechanisms (Fig. S4, Tables S9–11). For LST and LTL, several pathways were enriched, including response to topologically incorrect protein (p ​= ​2.30 ​× ​10⁻⁴), proteasomal protein catabolic process (p ​= ​4.22 ​× ​10⁻⁵), regulation of Rac protein signal transduction (p ​= ​2.92 ​× ​10⁻⁴), immune reactions (MHC class II protein complex assembly) (p ​= ​3.54 ​× ​10⁻⁶), and immunoglobulin production involved in immunoglobulin-mediated immune response (p ​= ​1.50 ​× ​10⁻⁴). For LST and ALM, pathways related to oxidative stress were significantly overrepresented (p ​= ​9.69 ​× ​10⁻⁶). For LST and FI, pathways related to DNA replication-independent chromatin assembly (p ​= ​6.33 ​× ​10⁻³), tricarboxylic acid cycle (p ​= ​5.60 ​× ​10⁻³), and response to metal ion (p ​= ​5.07 ​× ​10⁻³) were enriched.

Discussion

To our knowledge, this is the first MR analysis investigating the association between LST and aging by using large-scale GWAS summary-level data. Previous studies have already been carried out using cohort or cross-sectional designs, consistent with them [1,12,13], our study verified a significant association between LST and aging after sensitivity analyses. We firstly used data from NHANES and found that with per 1 ​h increase in LST, participants had a shorter LTL (β ​= ​−1.39, 95 ​% CI: −2.47 to −0.30), a lower ALM (β ​= ​−1.09, 95 ​% CI: −1.39 to −0.70), and an increased FI (β ​= ​8.22, 95 ​% CI: 4.29 to 12.30). In subgroup analyses, men, participants older than 50 years of age, and non-Hispanic white participants were more likely to have shorter LTL under the influence of LST. To eliminate the potential bias from observational studies, we further conducted TSMR and MVMR. With or without adjustment for physical activity, the same result as the aforementioned was presented, strengthening the causal power of the results. Moreover, we conducted TWAS and pathway-based functional enrichment analyses to investigate the underlying biological mechanisms behind LST and aging.

Andersen et al. found that age and sex could influence LST [34]. According to some research based on Children's cohorts [35], men reported a longer LST than women. Moreover, specific screen time also altered as growing [36]. To figure out whether these two factors also work in adults and affect aging traits significantly, we conducted subgroup analyses. We found that age and gender do cause discrepancies in adults. Compared with women, men were more likely to be influenced by LST and had shorter LTL and an increased FI. One possible speculation pointed to the higher estrogen in women, which lengthens the telomere, leading to the increased telomerase activity and antioxidation effect [37].

The role physical activity plays in aging improvement and LTL shortening has been discussed for a long time. According to a recent meta-analysis [38], there has yet to be a consensus. Some found the result was not statistically related; some insisted on the positive correlation between physical activity and LTL shortening, while others reported an inverted U-shaped relationship, which meant moderate levels of physical activity might lead to longer LTL. Our study also sought to determine if the relationship between LST and aging was independent of the strong-related confounding factor of physical activity [39]. Therefore, we adjusted for the physical activity to rule out the potential bias it caused. Whether adjusted for physical activity or not, the study revealed that genetic predisposition to LST-based sedentary behavior is significantly associated with adverse aging traits independent of physical activity. The reason might offset the positive effect of reduced oxidative stress and inflammation and the negative effect of reactive oxygen species (ROS) produced by exercise [15].

The result that LST was related to aging traits independent of physical activity raises questions about the mechanism by which LST might affect the aging process. One possible explanation is that exposure to LST could induce modulation of the metabolic process [40], which will lower the activity of lipoprotein lipase in skeletal muscle and indirectly inhibit the local lipid metabolism, resulting in atrophy of muscle and lower ALM. Another explanation is that the time spent on LST, like watching television, is associated with poor dietary habits, such as increased consumption of soft drinks and sweets [41]. Since oxidative stress and inflammation have been identified as significant determinants of telomeres [42], such diet patterns might produce more ROS, accelerating genomic instability and telomere loss [43]. The profile of protein expression and protein equilibrium might also be influenced, mainly expressed in the accumulation of incorrect protein and overactivated proteolysis featured by increased ubiquitin proteins [44]. LST-based sedentary behaviors also brought about fatness, which could shorten LTL as well [45]. With the growing burden of senescent cells, they produce more inflammatory factors, further worsening the abovementioned process, which might deteriorate aging.

Pathways related to immune response, oxidate stress, inflammation reaction, protein metabolism, chromosome organization, and cellular response to metal ions were obviously enriched in our results. Among these biological processes, the first three played a vital role and interacted with other overrepresented pathways. Sedentary behaviors were associated with higher cytokine levels, which could regulate cellular processes and inflammatory responses [46]. Some studies demonstrated that sedentary behaviors were associated with immunosuppressive effects [47]. Concurrently, our study identified an enrichment of immune function-related pathways; however, the direction of their regulation of immune function remains to be ascertained. Pathways related to immune responses were significantly overrepresented between LST and LTL/ALM. Since immune response, oxidative stress, and inflammation reactions constitute a complete and complex correlative network [48], any variation might imbalance the inner homeostasis and accelerate aging. Parsons et al. found SB was positively associated with higher levels of inflammatory markers, such as IL-6, TNF-α, and C-reactive protein, which was independent of physical activities [49]. The release of such inflammatory factors could foster autoimmunity, promote immune cells to produce ROS, activate the inflammatory response, and induce apoptosis, which might impair partial intracellular processes [50]. ROS also influenced the topology of the final protein by targeting specific molecules, during which more ROS were produced, during which more ROS were produced, perpetuating a deteriorative cycle. Subsequently, systemic inflammation and cellular senescence are exacerbated due to increased apoptosis, cell senescence, and oxidative stress induced by the inflammatory factors [51]. Since cellular oxidative stress and inflammation balance are key determinants of the rate of LTL shortening [52], such procedures would further shorten LTL. Another possible explanation might be the constant DNA damage response (DDR) induced by LST. When DNA is impaired, DDR is activated to block the cell cycle and promote DNA repair to maintain the integrity of the genome [53]. This reaction was under multiple post-translational modifications, such as ubiquitination [54], which was enriched in our pathway enrichment result. Once LST induced inflammation reaction and oxidative stress, ubiquitination would be promoted by the combination of ubiquitination FBW7 and telomere protective proteins1 TPP1, which accelerated protein degradation, uncapped telomere, and continued DDR signaling [55]. Continuous DNA damage and DDR signaling could form a certain amount of senescent cells. These cells could up-regulate genes that secrete inflammatory factors, inactivate telomerase, and dysfunctional the telomere, resulting in lower LTL. In addition, senescent cells showed potential in paracrine-related cytokines, leading to aging and senescence-associated secretory phenotype, which activated the immune systems [56] and initiated senescence arrest [57]. Besides, the pathway concerned with Rac protein signal transduction was notably overrepresented. Rac signal pathway showed competence in regulating apoptosis, inflammation reaction, and immune-cell-mediated phagocytosis [58]. Furthermore, these inflammatory factors could lower ALM by accelerating muscle proteolysis directly and indirectly [59]. Significant evidence showed that higher ubiquitin ligases CHIP and MURF1 levels existed in aging muscle cells, which could promote ubiquitination and degradation of misfolded proteins [60].

As for the association between FI and LST, pathways concerning cellular response to metal ions and mitochondria-related energy metabolism were enriched significantly. Mitochondria offered a place for the tricarboxylic acid cycle, during which abundant adenosine triphosphate (ATP) and a certain number of ROS were produced [61]. These substances hold oxidizing solid properties and will target mitochondrial DNA for damage, leading to the dysfunction of the mitochondria and, eventually, inducing apoptosis by insufficient ATP supply and amplifying oxidative stress. The uncontrollable oxidative stress would lead to the failure of chromosome end replication, causing telomere shortening in a ROS-dependent way [3]. Evidence indicated that the participation of metal ions is associated with protein aggregation and oxidative stress and may act as a channel between the two pathological processes [62]. Metal ions could combine structural proteins in a particular way, resulting in changes in the topology of proteins and abnormality of protein folding [63], which were closely related to oxidative stress, as we mentioned above. Besides, they could mediate redox reactions by uniting specific peptides and catalyzing the oxidation of proteins [64]. Moreover, metal ions showed potential in regulating immune functions during aging, including activating the immune system, stimulating antigen presentation, and promoting cytokine production of cytokines. Take zinc, for example; not only could it help protect against oxidative stress by activating NF-kB and AP-167, but it also prevents apoptosis in aging by activation of endonuclease or by blocking the combination between glucocorticoid receptor and ligand [65].

We explored the common genes shared between LST and LTL/ALM/FI. Venn diagram exhibited four genes in the intersection (Fig. 4), including HLA-DQA1, CYB561D2, GFRA2, and GGT7. A large-scale genome-wide association meta-analysis found a new lifetime-related gene segment-HLA-DQA1, attached to a major histocompatibility complex (MHC), which could code the units of antigen presentation and was associated with multiple autoimmunity disease and relevant features [66]. For instance, HLA-DQA seemed to increase intrathecal synthesis of immunoglobulins in multiple sclerosis patients by regulating the ratio of intrathecal B cells to plasma blasts [67]. CYB561D2, a member of the cytochrome b561 family, acts as a powerful antioxidant. ROS could activate it and induce the expression of immunosuppressive genes by the STAT1 pathway in tumor cells [68]. GFRA2, a member of the family of receptors of neurotrophic factors, is expressed in both congenital and adaptive human peripheral blood mononuclear cells and seems involved in the inflammatory response [69]. GGT7, a new member of the family of γ-Glutamyltranspeptidase, could induce the production of endogenous ROS and lead to continuous oxidative stress, resulting in abnormal methylation [70]. GGT7 also shows a close relationship with the infiltration level of immune cells in a cancer state. According to the respective functions of these four genes summarized above, we hypothesize that the immune response is implicated in counteracting the inflammatory response and oxidative stress, and LST might be involved in the occurrence of the latter.

In addition, we conducted TWAS analyses and showed the top 5 genes overlapping LST and LTL, LST and ALM, and LST and FI, respectively (Table S12). Most co-expressed genes were located on 20q13.13, 17q23.3, and 12q24.31. Eun et al. [71] found LCK-interacting membrane protein1 (LIME1), a raft-associated transmembrane protein mainly expressed in immune cells, participated in the transmission of inflammatory signals and subsequent activation and migration of immune cells, which then is similar to the process discussed above to influence the result of LST on aging traits. Another gene overlapping between LST and LTL located at 20q13.33 was DNAJC5, which showed the ability to maintain proteostasis [72]. The mutation of DNAJC5 increased the mal-folded protein and decreased the proteolytic process by ubiquitination, which would accumulate lipofuscin over time. Lipofuscin was not only seen as a sign of senescence, it also blocked the metabolism of senescent mitochondria, thus producing more ROS, which shorten LTL by deteriorating oxidative stress [52].

ACSS2, located on the near sub-band of 20q13.33, overlapped LST and ALM with great significance in our result. Typically, serine-responsive SAM-containing metabolic enzyme complex interacts with the histone acetyltransferase SAS protein complex, which then promotes histone H4K16 acetylation (H4K16ac) enrichment and the occupancy of protein Bdf1 at sub-telomeric regions [73]. This interaction maintains telomere silencing by antagonizing the spreading of Sir2 along the telomeres. However, with the interaction of ACSS2 with H4K16 acetyltransferase hMOF, this telomere silencing is reduced and induces telomere silencing and cell senescence [73]. However, more studies are still needed to investigate the specific mechanism between LST and ALM. A newly released GWAS reported a strong gene-aggregation between LST and CCDC92. CCDC92 is a coiled-coil domain protein that is associated with higher insulin, higher triglyceride, and lower HDL-cholesterol levels. Since HDL levels are genetically positively correlated with longevity [66], it might act as a mediator between LST and ALM. The specific mechanism of CCDC47 on LST and ALM is still unknown. However, we should consider the possibility of a similar mechanism to CCDC47 since they are coiled helix domain (CCDC) family proteins. As for the other genes on chromosome 17q23.3, there still lacks evidence specifically testifying to the direct effect of overlapped genes on LST and ALM. Tara et al. [74] found that loss of TACO1 might lead to dysfunction of mitochondria and a decline in motor function in mice. The insufficiency of ATP and disuse of muscular atrophy may cause decreased ALM. Another overlapped gene segment is SMARCD2, the mutation of which was reported to induce neutrophil developmental defects and specific particle deficiency. However, more studies are needed to elucidate further the inner relationship between reduced immune and ALM. Several studies supported the idea that SETD8, a lysine monomethyltransferase, is closely related to aging. SETD8 acted as a barrier against cellular senescence, the anti-aging effect of which is achieved by maintaining the silencing marker H20K1me21 at the p4 site of the aging switch gene [75]. Meanwhile, SETD8 could prevent cell senescence by inhibiting nucleolar and mitochondrial activity through histone H4K20 monomethylation. Moreover, ubiquitin-specific proteases (USPs) also help SETD8 in preventing cell senescence. USP17 prevents cell senescence by stabilizing methyltransferase and transcriptional inhibition of p21. USP29 could regulate DNA damage by directly affecting the amount of SETD8 through deubiquitination, which might play a part in the DDR induced by LST. In a word, LST may affect aging traits through multiple mechanisms. On the one hand, LST may induce modulation of metabolic processes, leading to muscle atrophy and lower ALM. It may also be associated with poor dietary habits, resulting in more ROS and accelerating genomic instability and telomere loss. On the other hand, LST-related sedentary behaviors can bring about obesity, influence protein expression and balance, and activate inflammatory responses, thereby accelerating aging. In addition, LST may affect aging through continuous DNA damage responses and influencing the expression of common genes.

The main strength of this study is that we utilized both linear regression analyses and the Mendelian randomization method to investigate the causal relationship and variation trend between LST and aging and to eliminate the potential bias caused by cross-sectional studies. A second strength is the inclusion of a wide range of covariates that might influence LST or aging traits, which helps better understand the direct effect of LST on aging. Since poor body conditions (like low ALM and high FI) might induce participants' long sedentary behavior, the possibility of reverse causality could not be neglected. Moreover, the included GWAS data came from the entire population, including Non−Hispanic White, Non−Hispanic Black, Mexican American, Other Hispanic, etc., which means a certain degree of universality of our results. Briefly, our study is subject to certain limitations. First, the data on LST and covariates like physical activity were collected by questionnaires and interviews, which might be deviated by retrospective bias. However, the current LST measurements in databases such as NHANES are based on self-reported data and do not include objective assessments [[76], [77], [78]]. Objective methods that can accurately measure LST are required in future research to better handle the deficiency. Meanwhile, we categorized television watching, video listening, and computer use into a general LST. Nevertheless, different SB domains and contexts might determine the direction of the ultimate association. For example, compared with sitting while using a computer, sitting while watching TV is associated with lower muscle activity and energy expenditure levels. Moreover, consistent with previous studies [79], our research identified relatively small effect sizes for certain associations, particularly those related to LST. To quantify the years of aging change associated with a 1-h increase in LST for three aging metrics (LTL, ALM, and FI), we applied the ratio of beta 1 (LST-aging characteristics) to beta 2 (age-aging characteristics) [80]. Our findings indicated that the effect of LST corresponded to an aging change of 0.06 years for LTL, 0.10 years for ALM, and 0.14 years for FI. These analyses may facilitate a clearer understanding of the broader impacts of leisure screen time on biological aging, even in the context of small effect sizes. To testify to the potential influence, we newly added and analyzed the relationship between TV and computer time and aging in the Mendelian randomization section (Table S3). We found a significant causal relationship between TV time and FI, ALM, and LTL, which indicates a valuable subdivision of LST. However, the distribution of LST might also influence the result by some mediations. For instance, those who spent LST at night might have a lower LTL due to the influence of sleeping quality. Additionally, the resulting effect of the activity replaced by LST-based sedentary behaviors deserves further exploration and explanation. Some activities (e.g., sleep, healthy meal preparation, hobbies, social interaction) might be causally related to the aging outcomes to a certain extent. Therefore, objective methods for accurately measuring LST and the resulting effects of confound factors are needed, and deep investigation into other populations and different LST domains is required in further studies.

Conclusion

In conclusion, this study provides evidence that the genetic predisposition to higher LST was associated with shorter LTL, lower ALM, and increased FI, which indicates LST might be a risk factor for aging. For per 1 ​h increase in LST, participants had a shorter LTL (β ​= ​−1.39, 95 ​% CI: −2.47–−0.30), a lower ALM (β ​= ​−1.09, 95 ​% CI: −1.39–−0.70), and an increased FI (β ​= ​8.22, 95 ​% CI: 4.29–12.30). This is the first MR analysis study to investigate the causal associations between LST and aging using genetic approaches. Additionally, the study conducted TWAS and pathway-based functional enrichment analysis to explore underlying biological mechanisms in both traits, offering genetic insights into understanding the latent interactions between LST and aging. Since reducing LST may be conducive to improving aging traits, guidelines should be refined to reduce LST according to different age groups and genders. Adults, especially men, those over 50, and non−Hispanic white participants should self-restrict daily LST depending on personal circumstances.

Ethics approval and consent to participate

All data used by this study were publicly available from participant studies with the approval ​of the ethical standards committee related to human experimentation. No additional ethical approval was required in this study.

Consent for publication

All authors consented to the submission and publication of this study.

Author contributions

Jingwei Zhang, Hongwei Liu, and Quan Cheng funded, conceived and designed the research. Jie Wen, Yuchen Wang, and Xueyi Mao wrote the first draft of the manuscript. Jie Wen, Yuchen Wang, Xueyi Mao, Ruoyan Lei, Jinglin Zhou, Jingwei Zhang, Hongwei Liu, and Quan Cheng contributed to data acquisition, analysis, and interpretation. Jie Wen, Yuchen Wang, Xueyi Mao, Ruoyan Lei, Jinglin Zhou, Jingwei Zhang, Hongwei Liu, and Quan Cheng contributed to the revision of the paper. All authors contributed to the article and approved the final manuscript.

Availability of data and material

All data used in this study were available in the original research. Data generated in this study were included in the main text and supplementary files.

Funding

This research was funded by the National Natural Science Foundation of China (NO. 82401594), the Natural Science Foundation of Hunan Province (NO. 2022JJ30943 and NO. 2023JJ30971), and the Huxiang Youth Talent Program (Hejian Category) under the Lotus Talent Plan of Hunan Province (NO. 2023RC3074).

Declaration of competing interest

All authors have no interest in conflict to be declared.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Fig. S1

The leave-out one plot from genetically predicted LST on LTL.

Fig. S2

The leave-out one plot from genetically predicted LST on ALM.

Fig. S3

The leave-out one plot from genetically predicted LST on FI.

Fig. S4

Pathway-based functional enrichment analysis between LST and LTL (A), ALM (B), and FI (C).