key: cord-0879160-0qn2a8jp authors: Gao, Xu; Huang, Ninghao; Guo, Xinbiao; Huang, Tao title: Role of sleep quality in the acceleration of biological aging and its potential for preventive interaction on air pollution insults: Findings from the UK Biobank cohort date: 2022-04-14 journal: Aging Cell DOI: 10.1111/acel.13610 sha: 2eb6bccf0e0614ebdc0f4fec3396af16c826529e doc_id: 879160 cord_uid: 0qn2a8jp Sleep has been associated with aging and relevant health outcomes, but the causal relationship remains inconclusive. In this study, we investigated the associations of sleep behaviors with biological ages (BAs) among 363,886 middle and elderly adults from UK Biobank. Sleep index (0 [worst]–6 [best]) of each participant was retrieved from the following six sleep behaviors: snoring, chronotype, daytime sleepiness, sleep duration, insomnia, and difficulties in getting up. Two BAs, the KDM‐biological age and PhenoAge, were estimated by corresponding algorithms based on clinical traits, and their residual discrepancies with chronological age were defined as the age accelerations (AAs). We first observed negative associations between the sleep index and the two AAs, and demonstrated that the change of AAs could be the consequence of sleep quality using Mendelian randomization with genetic risk scores of sleep index and BAs. Particularly, a one‐unit increase in sleep index was associated with 0.104‐ and 0.119‐year decreases in KDM‐biological AA and PhenoAge acceleration, respectively. Air pollution is another key driver of aging. We further observed significant independent and joint effects of sleep and air pollution (PM(2.5) and NO(2)) on AAs. Sleep quality also showed a modifying effect on the associations of elevated PM(2.5) and NO(2) levels with accelerated AAs. For instance, an interquartile range increase in PM(2.5) level was associated with 0.009‐, 0.044‐, and 0.074‐year increase in PhenoAge acceleration among people with high (5–6), medium (3–4), and low (0–2) sleep index, respectively. Our findings elucidate that better sleep quality could lessen accelerated biological aging resulting from air pollution. physiological alterations in normal aging, sleep behaviors change with aging independent of many factors including medical comorbidity and medications . Some researchers thus suggest that sleep disorders are the consequence of aging-related changes in neuroendocrine function . Meanwhile, sleep is also considered a restorative process that not only allows for energy renewal but also for cellular restoration (Carroll & Prather, 2021) . Hence, the other hypothesis is that the declines in sleep quality may lead to accelerated aging by triggering DNA damage and chronic inflammation to influence the compensatory/resiliency systems of the human body (Carroll & Prather, 2021) . The lack of an established accurate measurement of aging, however, hinders the scientists from elucidating the directionality and causal relationship between aging and sleep. Aging is the sum of changes that occur at hierarchically organized levels in the human body , which makes it hard to be captured by single age-related biomarkers employed in previous relevant studies. All individuals age chronologically at the same rate, but there is marked variation in their biological ages as we observe in real life that the people with the same chronological age may not share the same aging-related symptoms . Some could experience age-related decline faster than others. Moreover, as biological aging is a complex biological process in multiple organ systems, a single aging-related biomarker, such as telomere length or oxidative stress biomarkers, may not be able to completely depict the whole landscape of the aging process of individuals due to the heterogeneity of cells (Lopez-Otin et al., 2013) . Therefore, the identification of "biological age (BA)" has been proposed and has been explored over the last 10 years. To date, several forms of BAs which could be estimated based on the functions of cardiovascular, metabolic, renal, immune, and pulmonary systems (e.g., KDM-biological age (Klemera & Doubal, 2006; Levine, 2013) and PhenoAge ), or based on aging-related DNA methylation profiles (known as "epigenetic clocks" (Horvath & Raj, 2018) ), have been developed. Their discrepancies with chronological age, that is, the age accelerations (AAs), have been highly associated with aging-related health outcomes and mortality (Horvath & Raj, 2018) . However, previous studies on the associations of sleep with BAs or aging-related symptoms usually employed only one or two sleep behaviors or were conducted in a specific population with limited participants (Carroll et al., 2017 Carskadon et al., 2019; Han et al., 2018; Sun et al., 2020) . Therefore, there is a dearth of studies exploring the associations between sleep and aging in larger populations with more detailed sleep information, and causal inference approaches are needed to uncover their causal nature explicitly. Additionally, air pollution, especially the fine particulate matter [PM < 2.5 μm (PM 2.5 )], is a critical environmental exposure that could advance aging (Peters et al., 2021) and affect sleep quality (Liu et al., 2020) . Previous studies have linked aberrant accelerated epigenetic clocks with the increased exposure to various air pollutants in different populations (Peters et al., 2021) . Plenty of evidence has also documented increased risks of sleep disorders (Liu et al., 2020) and sleep-related neurological impairments, for example, dementia and cognitive decline Schikowski & Altug, 2020) , associated with elevated air pollution levels. Nevertheless, since the direction of the sleep-aging relationship is still undetermined, no studies yet evaluated whether the sleep quality or BAs could modify or mediate the adverse effects of air pollution on biological aging or sleep quality, which is critical for developing interventions to mitigate the adverse effects of air pollution. Therefore, we examined the causal associations of sleep (as reflected by six sleep behaviors) with KDM-biological age and PhenoAge based on the measures of clinical traits of multiple organs in the UK Biobank, a national-wide population-based cohort study in the UK. We subsequently explored the associations of five major air pollutants (PM 2.5, PM with an aerodynamic diameter between 2.5 and 10 µm [PM coarse ], PM with an aerodynamic diameter of less than with AAs and sleep, and explored whether AAs could mediate or modify the associations of air pollution with sleep, or whether the AAs could reflect the predisposition of participants regarding the impact of air pollution on their sleep quality. Table 1 presents the baseline characteristics of 363,886 study participants by sleep index category. Participants' age (mean ± standard deviation) was 56.5 ± 8.1 years and most of them were white. About 35% and 55% participants were former and never smokers, respectively. The majority of them had healthy physical activity and >10 years of education. Nearly half were with a healthy daily intake of alcohol. Only about 24%, 5%, and 5.5% participants were with hypertension, diabetes, and CHD diagnosed by doctors, respectively. Insomnia complaint is the most frequent (~75%) sleep-related behavior that the participants had and most participants (~82%) could get up easily in the mornings. About 30% have a high sleep quality (sleep index = 5-6) and 15% have a low sleep quality (sleep index = 0-2). The high sleep quality group has lower BAs and higher AAs than the low sleep group, and the medium sleep quality group (sleep index = 3-4) has the BAs and AAs at the intermediate level in between. Both BAs were highly correlated with the other and with the chronological age ( Figure S1 ). The average concentrations of air pol- We first examined the associations of both forms of AAs with each of the six sleep behaviors (Table 2) . After controlling for all potential hypertension, diabetes, and coronary heart disease. The examination center was controlled for as a random effect to account for the potential residual bias from examinations. Bolded values that were below the significance threshold, which was 0.05/(8*2) = 0.0031, were considered as statistically significant. observed. Sensitivity analysis in psychiatric illness-free participants showed slightly attenuated but still robust associations (Table S2) , Genetic-predicted effects 0.099 (0.068) 0.15 0.008 (0.011) 0.48 a Effects were estimated by one SD change in the sleep index or biological age accelerations. Models were adjusted for age, sex, race, BMI, smoking status, healthy alcohol intake, healthy physical activity, years of education (<10 years or ≥10 years), hypertension, diabetes, and coronary heart disease. The examination center was controlled for as a random effect. The genetic risk scores for sleep index and biological age accelerations were unweighted. Bolded values that were below the significance threshold, which was 0.05/(2*2*2) = 0.00625, were considered as statistically significant. which suggests that the influence of underlying psychiatric illness on the primary findings might be minor. Furthermore, to investigate the causal associations between sleep and AAs, we conducted two MR analyses in two directions. Constructed genetic risk scores (GRSs) have the acceptable variance explained ranging from ~2.3% (for sleep duration) to ~51% (for KDMbiological AA) for the MR analyses (Table S3 ). As demonstrated in Table 3 with unweighted GRSs for sleep index and AAs, we observed significant negative associations between genetic-predicted sleep index and both AAs. However, the genetic-predicted KDMbiological AA was positively associated with sleep index, and the negative association between genetic-predicted PhenoAge acceleration and sleep index was not statistically significant. Similar trends were also observed for each of the six sleep behaviors (Table S4) and another sensitivity analysis with weighted GRSs of sleep additionally demonstrated similar trends for each scenario (Table S5 ). Altogether, these results indicate that sleep quality was more likely to be the determinant of the change in AAs rather than a consequence. Given it was more plausible that accelerated AAs were the consequences of impaired sleep quality, along with the known association between air pollution and biological aging, we validated the associations of the five air pollutants with each AA and explored the hypothesis that whether sleep index could mediate the effects of air pollutants on AAs. As shown in Table S6 (Table S7 ). According to the robust joint effects of air pollution and sleep on AAs, we additionally explored the potential modifying effect of sleep quality on the associations between air pollution and AAs since sleep is a modifiable factor that could be intervened by human behaviors relatively easily. As shown in Figure 3 and Table S8, we observed a robust modifying effect of sleep quality on the air pollution-AA associations (p-values of interaction terms <0.001). Particularly, an IQR increase in PM 2.5 concentration was associated with 0.009-, 0.044-, and 0.074-year increase in PhenoAge acceleration among people with high, medium, and low sleep quality, respec- AAs could be distinguished under relatively lower PM 2.5 (~10 μg/m 3 , Figure 4 ) and NO 2 levels (~30 μg/m 3 , Figure 5 ). In this large cohort of middle-and elderly-aged adults, we dem- 2020), such as the risks of breathing problems, insomnia, sleep efficiency, and overall sleep quality, as well as on aging in different populations (Peters et al., 2021) . With the previously determined causal association between sleep and AAs in the first step, we expected to find a robust mediation effect of sleep on the air pollution-AA relationship. However, we observed very limited changes in the effects of different air pollutants on AAs in models mutually adjusting for sleep index, which suggests that sleep does not play a major role in mediating the effects of air pollution on the two AAs. Since sleep is predominantly related to cognitive health and brain aging , especially the structural and physiological changes that occur in the brain, such limited mediation effects could be explained by the minimal capacity of the two BAs in measuring aging-related neurological changes. The clinical traits implemented in the constructions of the two BAs were not specifically associated with such abnormal alterations related to the central neuron system. This is in line with the findings of our sensitivity analysis in psychiatric illness-free participants, suggesting that the two BAs may not be closely related to the aging-related damage of the neurological systems. Instead, we observed prominent modifying effects of sleep on the air pollution-BA relationship, which implies that a healthy and adequate sleep may help attenuate the adverse aging effects of air pollution on non-neurological systems ( Figure S2 ). Also, we noted response curves with different features of the PM 2.5 and NO 2 with both AAs. Because KDM-biological age is more related to the capacity and function of systems and organs (Hägg et al., 2019) , and PhenoAge is skewed to predict the mortality risk of humans . suggest that sleep may help lessen the detrimental impacts of PM 2.5 and NO 2 on body functions at a lower exposure level and could also attenuate the lethal effects of the two pollutants on mortality when they reached a higher level. Regarding the underlying biological mechanisms under the interesting modifying effects of sleep, one of the potential explanations is the reduction of oxidative stress and inflammation during healthy and adequate sleep. First, air pollutants can induce oxidative stress, the ability to respond to which has been identified as a key determinant of biological aging (Peters et al., 2021) . Sleep may help with the anti-oxidative mechanism by removing reactive oxygen species resulting from air pollution insults (Atrooz & Salim, 2020) . But given our BAs were constructed based on nonoxidative biomarkers, this hypothesis could be validated in future biological studies. Beyond this, sleep may also help enhance immune defenses and stabilize the dysregulation of inflammatory responses The major strengths of this study include the large sample size and relatively sufficient phenotype and biochemistry data for the biological age estimation. Several limitations are notable when interpreting the results. First, UK Biobank is a volunteer cohort, and participants are likely healthier than the general population, which may limit the effect of sleep on BAs in our analysis as their AAs are expected to be lower than the general population theoretically. Furthermore, the measurement bias of air pollution must be noted. The air pollution data we used were mostly a single measurement of the annual average outdoor air pollution level in 2010 since the home addresses of the participants were unavailable during follow-up. Because the initial assessment visit of UK Biobank was from 2006 to 2010, we were unable to determine the lag or short-term ( . Also, as a study with a cross-sectional nature, even though the MR could make causal inference to somewhat extent, caution must still be taken in the causal interpretation on sleep and aging. And, given the genomewide association studies (GWAS) we used to build the GRS for BAs is the only available one but was conducted in UK Biobank, these SNPs we selected may be biased and not objective adequately. Therefore, we conducted a sensitivity analysis using 11 SNPs reported in a GWAS of DNA methylation age to create the GRS for both BAs/AAs (Gibson et al., 2019) as the DNA methylation age has been highly associated with the two BAs we investigated in this study . The findings were in line with our primary findings by showing that sleep quality was more likely to be the determinant of the change in AAs rather than a consequence (Table S10) . Last, participants in this study were mostly of European descent, which limits the generalization of our findings to other races. In conclusion, our study is the first identifying the accelerating effect of poor sleep quality on biological aging. With this premise, we further found that sleep and air pollution were independently associated with biological aging, and sleep quality may modify the aging effects of air pollution. These findings not only provide exploratory evidence supporting sleep as an aging contributor but also underscore the importance of high-quality sleep as an intervention approach to mitigate the negative impact of air pollution on human aging. Nevertheless, aging is associated with a myriad of changes in psychological, social, spiritual, financial, and lifestyle across the Study design and methods of UK Biobank have been reported in detail previously (Sudlow et al., 2015) . In brief, UK Biobank is a large- Due to the lack of specialized questionnaires of sleep, such as PSQI or FIRST at the baseline survey of UK Biobank, we instead used an algorithm based on self-reported sleep quality information that was first introduced in 2020 for UK Biobank cohort (Fan et al., 2020) . This algorithm has been used widely since 2020 and was used to make an index for sleep with 5 sleep-related items (Fan et al., 2020; Li, Zheng, et al., 2021; Li, Xue, et al., 2021) or 4 items (Sambou et al., 2021) , and showed the capacity as an alternative approach to reflect the sleep patterns of the participants of UK Biobank. Therefore, to optimize the sleep quality assessment in the UK Biobank, we used six self-reported sleep behaviors in this study: snoring, chronotype, daytime sleepiness, sleep duration, and insomnia that were used in 5-item sleep score (Fan et al., 2020) and the "difficulties in getting up in the morning" pattern that was found to be related to the risk of terminated health-span (Sambou et al., 2021) to enhance the measurement of sleep quality and the capacity of the below-mentioned sleep index (Table S11 ). Detailed assessment of sleep behaviors can be found in supplementary methods. According to the six sleep behaviors, we generated a sleep index. The low-risk categories of each component were no self-reported snoring, early chronotype ("morning" or "morning than evening"), no frequent daytime sleepiness ("never/rarely" or "sometimes"), normal sleep duration (7-8 h/day), reported never or rarely having insomnia symptoms, and getting up easy in morning ("fairly easy" or "very easy"). For each sleep behavior, the participant received a score of 1 if he or she was classified as the low-risk group or 0 if otherwise as the highrisk group. All component scores were summed to obtain a continuous sleep index ranging from 0 (worst) to 6 (best), with a higher index indicating a general better sleep quality. We further defined a sleep index category as high (5-6), medium (3-4), and low (0-2) based on the continuous sleep index. We computed the BAs derived from a total of 12 blood chemistry traits, systolic blood pressure, and lung function data (Table S11) with two commonly accepted algorithms, the Klemera-Doubal method (i.e., in data from the National Health and Nutrition Examination Survey (NHANES) following the method originally described by Klemera et al. (Klemera & Doubal, 2006; Levine, 2013) and Levine et al. (2018) with two sets of nine clinical traits (Table 1 ). The two BAs were constructed with different purposes. KDM-biological age was computed from an algorithm derived from a series of regressions of nine individual biomarkers on chronological age in the reference population to quantify the decline of system integrity, and PhenoAge was computed from an algorithm derived from multivariate analysis of mortality hazards to estimate the risk of death (Hastings et al., 2019; Parker et al., 2020) . The two BAs were developed in the white population (Hastings et al., 2019) and were stable in other populations and cohorts (Parker et al., 2020) . The selected traits, algorithms, and corresponding R code can be found in the R package 'BioAge' at: https://github.com/dayoo nkwon/ BioAge and corresponding publications Kwon & Belsky, 2021) . Missing values of each trait consisted of <10% of all traits and were, therefore, imputed by the median value of the corresponding trait. The residual differences between the estimated BAs and chronological age were considered as AAs since this approach could minimize the heterogeneities between the measurement platforms of each component of BAs (Hägg et al., 2019; Horvath & Raj, 2018) . The residuals were calculated by a linear regression procedure in which one of the BAs was the outcome and chronological age was the independent variable. AAs were the targeted outcomes/modifiers in our primary analyses. As previously described , the annual average con- Questionnaire (IPAQ) was adopted to assess physical activity. The history of hypertension and diabetes was based on self-reported information and medical records. Detailed information on genotyping, imputation, and quality control in the UK Biobank study has been described previously (Sudlow et al., 2015) . We created GRSs for each sleep behavior, sleep index, We first examined the associations of each sleep behavior and the sleep index with both forms of AAs using mixed-effect linear regression models in which the AAs were the outcomes. We adjusted for covariates described in the previous section including chronic diseases and additionally controlled for the examination center in the model as a random effect to account for the potential residual bias from examinations. Since psychiatric disorders may affect sleep quality, we additionally conducted a sensitivity analysis in 285,054 participants that were free of dementia, depression, and anxiety. These disorders were ascertained using hospital inpatient records and Patient Health Questionnaire (PHQ)-4 questionnaire as previously described (Milaneschi et al., 2021; Zhang et al., 2021) . Then, we performed one-stage MR analyses to explore the causal relationships between sleep and BAs in two scenarios using mixed-effect linear regression models. The GRS of sleep index was used to predict sleep index genetically and to study the effect of sleep on BAs. The GRS of BA was used to predict AAs genetically and to study the effects of the two AAs on sleep quality. The statistically significant genetic-predicted effect in either scenario that was in the same direction as the observed effect in the primary model would indicate the plausibility of a causal effect. Corresponding dose-response curves between sleep index and the two BAs were further assessed by restricted cubic spline regression (Desquilbet & Mariotti, 2010) . Models were adjusted for the previously described covariates and the sleep index = 1, 3, and 5 were selected as knots. Based on the results of MR analyses, the next part of our study would test whether (a) sleep quality could mediate or modify the association between elevated air pollution levels and accelerated AAs Start-up Grant (BMU2021YJ044). We thank Dr. Chen Chen for the language assistance. None reported. Xu Gao conceptualized the study, conducted data clean, estimated biological ages, and draft and reviewed the manuscript; Ninghao Huang coordinated the data collection, reviewed and revised the manuscript; Xinbiao Guo critically reviewed and revised the manuscript; Tao Huang coordinated and supervised the data collection, and reviewed and revised the manuscript. Sleep deprivation, oxidative stress and inflammation Quantification of the pace of biological aging in humans through a blood test, the DunedinPoAm DNA methylation algorithm. eLife, 9 Epigenetic aging and immune senescence in women with insomnia symptoms: findings from the Women's Health Initiative Study Sleep and biological aging: A short review. Current Opinion in Endocrine and Metabolic Research Postpartum sleep loss and accelerated epigenetic aging A pilot prospective study of sleep patterns and DNA methylation-characterized epigenetic aging in young adults The effect of sleep deprivation and disruption on DNA damage and health of doctors Sleep duration and cardiovascular disease risk: Epidemiologic and experimental evidence Cellular aging and restorative processes: Subjective sleep quality and duration moderate the association between age and telomere length in a sample of middle-aged and older adults Dose-response analyses using restricted cubic spline functions in public health research Sleep patterns, genetic susceptibility, and incident cardiovascular disease: A prospective study of 385 292 UK biobank participants Time and the metrics of aging Short-term air pollution, cognitive performance and nonsteroidal anti-inflammatory drug use in the Veterans Affairs Normative Aging Study A meta-analysis of genome-wide association studies of epigenetic age acceleration Developments in molecular epidemiology of aging Is sleep duration associated with biological age (BA) Comparability of biological aging measures in the National Health and Nutrition Examination Study DNA methylation-based biomarkers and the epigenetic clock theory of ageing Genome-wide analysis of insomnia in 1,331,010 individuals identifies new risk loci and functional pathways A new approach to the concept and computation of biological age Genetic associations for two biological age measures point to distinct aging phenotypes A toolkit for quantification of biological age from blood-chemistry and organ-function-test data: BioAge. medRxiv Modeling the rate of senescence: Can estimated biological age predict mortality more accurately than chronological age? An epigenetic biomarker of aging for lifespan and healthspan Sleep in normal aging Poor sleep behavior burden and risk of COVID-19 mortality and hospitalization Adherence to a healthy sleep pattern and incident heart failure: A prospective study of 408 802 UK Biobank Participants Associations between rotating night shifts, sleep duration, and telomere length in women Air pollution exposure and adverse sleep health across the life course: A systematic review The hallmarks of aging Effect of sleep loss on C-reactive protein, an inflammatory marker of cardiovascular risk Association of inflammation with depression and anxiety: Evidence for symptom-specificity and potential causality from UK Biobank and NESDA cohorts Association of blood chemistry quantifications of biological aging with disability and mortality in older adults Hallmarks of environmental insults Investigation of the relationships between sleep behaviors and risk of healthspan termination: A prospective cohort study based on 323,373 UK-Biobank participants The role of air pollution in cognitive impairment and decline UK biobank: An open access resource for identifying the causes of a wide range of complex diseases of middle and old age Associations of sleep quality and sleep duration with frailty and pre-frailty in an elderly population Rugao longevity and ageing study Joint exposure to various ambient air pollutants and incident heart failure: A prospective analysis in UK Biobank WHO global air quality guidelines: Particulate matter (PM2. 5 and PM10), ozone, nitrogen dioxide, sulfur dioxide and carbon monoxide Inflammatory health effects of indoor and outdoor particulate matter Consumption of coffee and tea and risk of developing stroke, dementia, and poststroke dementia: A cohort study in the UK Biobank The role of DNA methylation in cardiovascular risk and disease: methodological aspects, study design, and data analysis for epidemiological studies Additional supporting information may be found in the online version of the article at the publisher's website.How to cite this article: Gao, X., Huang, N., Guo, X., & Huang, T. (2022) . Role of sleep quality in the acceleration of biological aging and its potential for preventive interaction on air pollution insults: Findings from the UK Biobank cohort.Aging Cell, 21, e13610. https://doi.org/10.1111/acel.13610