DNA methylation signatures reveal that distinct combinations of transcription factors specify human immune cell epigenetic identity.

Epigenetic reprogramming underlies specification of immune cell lineages, but patterns that uniquely define immune cell types and the mechanisms by which they are established remain unclear. Here, we identified lineage-specific DNA methylation signatures of six immune cell types from human peripheral blood and determined their relationship to other epigenetic and transcriptomic patterns. Sites of lineage-specific hypomethylation were associated with distinct combinations of transcription factors in each cell type. By contrast, sites of lineage-specific hypermethylation were restricted mostly to adaptive immune cells. PU.1 binding sites were associated with lineage-specific hypo- and hypermethylation in different cell types, suggesting that it regulates DNA methylation in a context-dependent manner. These observations indicate that innate and adaptive immune lineages are specified by distinct epigenetic mechanisms via combinatorial and context-dependent use of key transcription factors. The cell-specific epigenomics and transcriptional patterns identified serve as a foundation for future studies on immune dysregulation in diseases and aging.

Sleep Duration Polygenic Risk and Phenotype: Associations with Biomarkers of Accelerated Aging in the Baltimore Longitudinal Study of Aging.

We sought to explore whether genetic risk for, and self-reported, short sleep are associated with biological aging and whether age and sex moderate these associations. Participants were a subset of individuals from the Baltimore Longitudinal Study of Aging who had complete data on self-reported sleep (n = 567) or genotype (n = 367). Outcomes included: Intrinsic Horvath age, Hannum age, PhenoAge, GrimAge, and DNAm-based estimates of plasminogen activator inhibitor-1 (PAI-1) and granulocyte count. Results demonstrated that polygenic risk for short sleep was positively associated with granulocyte count; compared to those reporting <6 hr sleep, those reporting >7 hr demonstrated faster PhenoAge and GrimAge acceleration and higher estimated PAI-1. Polygenic risk for short sleep and self-reported sleep duration interacted with age and sex in their associations with some of the outcomes. Findings highlight that polygenic risk for short sleep and self-reported long sleep is associated with variation in the epigenetic landscape and subsequently aging.

Mitochondrial genetic variation is enriched in G-quadruplex regions that stall DNA synthesis in vitro.

As the powerhouses of the eukaryotic cell, mitochondria must maintain their genomes which encode proteins essential for energy production. Mitochondria are characterized by guanine-rich DNA sequences that spontaneously form unusual three-dimensional structures known as G-quadruplexes (G4). G4 structures can be problematic for the essential processes of DNA replication and transcription because they deter normal progression of the enzymatic-driven processes. In this study, we addressed the hypothesis that mitochondrial G4 is a source of mutagenesis leading to base-pair substitutions. Our computational analysis of 2757 individual genomes from two Italian population cohorts (SardiNIA and InCHIANTI) revealed a statistically significant enrichment of mitochondrial mutations within sequences corresponding to stable G4 DNA structures. Guided by the computational analysis results, we designed biochemical reconstitution experiments and demonstrated that DNA synthesis by two known mitochondrial DNA polymerases (Pol γ, PrimPol) in vitro was strongly blocked by representative stable G4 mitochondrial DNA structures, which could be overcome in a specific manner by the ATP-dependent G4-resolving helicase Pif1. However, error-prone DNA synthesis by PrimPol using the G4 template sequence persisted even in the presence of Pif1. Altogether, our results suggest that genetic variation is enriched in G-quadruplex regions that impede mitochondrial DNA replication.

Prior psychosocial profile and perceived impact of the COVID-19 pandemic: insights from the Baltimore Longitudinal Study of Aging.

Stress, social isolation, and changes in health behaviors during the COVID-19 pandemic period may have a lasting influence on health. Here, the correlation between current or prior demographic, social and health related characteristics, including psychosocial factors with perceived impact of the COVID-19 pandemic assessed by questionnaire during the early pandemic period is evaluated among 770 participants of the Baltimore Longitudinal Study of Aging. In multinomial logistic regression models participants with higher pre-pandemic personal mastery, a construct related to self-efficacy, were more likely to report "both positive and negative" impact of the pandemic than a solely "negative" impact (OR: 2.17, 95% CI: 1.29-3.65). Higher perceived stress and frequent contact with family prior to the pandemic were also associated with pandemic impact. These observations highlight the relevance of psychosocial factors in the COVID-19 pandemic experience and identify characteristics that may inform interventions in future public health crises.

GeMes, clusters of DNA methylation under genetic control, can inform genetic and epigenetic analysis of disease.

Epigenetic marks such as DNA methylation have generated great interest in the study of human disease. However, studies of DNA methylation have not established population-epigenetics principles to guide design, efficient statistics, or interpretation. Here, we show that the clustering of correlated DNA methylation at CpGs was similar to that of linkage-disequilibrium (LD) correlation in genetic SNP variation but for much shorter distances. Some clustering of methylated CpGs appeared to be genetically driven. Further, a set of correlated methylated CpGs related to a single SNP-based LD block was not always physically contiguous-segments of uncorrelated methylation as long as 300 kb could be interspersed in the cluster. Thus, we denoted these sets of correlated CpGs as GeMes, defined as potentially noncontiguous methylation clusters under the control of one or more methylation quantitative trait loci. This type of correlated methylation structure has implications for both biological functions of DNA methylation and for the design, analysis, and interpretation of epigenome-wide association studies.

Correlates of life course physical activity in participants of the Baltimore longitudinal study of aging.

Physical activity is consistently associated with better health and longer life spans. However, the extent to which length and intensity of exercise across the life course impact health outcomes relative to current activity is undefined. Participants of the Baltimore Longitudinal Study of Aging were asked to categorize their level of physical activity in each decade of life from adolescence to the current decade. In linear mixed effects models, self-reported past levels of physical activity were significantly associated with activity assessed at study visits in the corresponding decade of life either by questionnaire or accelerometry. A pattern of life course physical activity (LCPA) derived by ranking participants on reported activity intensity across multiple decades was consistent with the trajectories of activity estimated from standard physical activity questionnaires assessed at prior study visits. In multivariable linear regression models LCPA was associated with clinical characteristics, measures of body composition and indicators of physical performance independent of current physical activity. After adjustment for minutes of high intensity exercise, LCPA remained significantly associated with peak VO2, fasting glucose, thigh muscle area and density, abdominal subcutaneous fat, usual gait speed, lower extremity performance, and multimorbidity (all p < 0.01) at the index visit. The observed associations suggest that an estimate of physical activity across decades provides complementary information to information on current activity and reemphasizes the importance of consistently engaging in physical activity over the life course.

Cross-sectional analysis of healthy individuals across decades: Aging signatures across multiple physiological compartments.

The study of age-related biomarkers from different biofluids and tissues within the same individual might provide a more comprehensive understanding of age-related changes within and between compartments as these changes are likely highly interconnected. Understanding age-related differences by compartments may shed light on the mechanism of their reciprocal interactions, which may contribute to the phenotypic manifestations of aging. To study such possible interactions, we carried out a targeted metabolomic analysis of plasma, skeletal muscle, and urine collected from healthy participants, age 22-92 years, and identified 92, 34, and 35 age-associated metabolites, respectively. The metabolic pathways that were identified across compartments included inflammation and cellular senescence, microbial metabolism, mitochondrial health, sphingolipid metabolism, lysosomal membrane permeabilization, vascular aging, and kidney function.

Deep conditional generative model for longitudinal single-slice abdominal computed tomography harmonization.

Purpose: Two-dimensional single-slice abdominal computed tomography (CT) provides a detailed tissue map with high resolution allowing quantitative characterization of relationships between health conditions and aging. However, longitudinal analysis of body composition changes using these scans is difficult due to positional variation between slices acquired in different years, which leads to different organs/tissues being captured. Approach: To address this issue, we propose C-SliceGen, which takes an arbitrary axial slice in the abdominal region as a condition and generates a pre-defined vertebral level slice by estimating structural changes in the latent space. Results: Our experiments on 2608 volumetric CT data from two in-house datasets and 50 subjects from the 2015 Multi-Atlas Abdomen Labeling Challenge Beyond the Cranial Vault (BTCV) dataset demonstrate that our model can generate high-quality images that are realistic and similar. We further evaluate our method's capability to harmonize longitudinal positional variation on 1033 subjects from the Baltimore longitudinal study of aging dataset, which contains longitudinal single abdominal slices, and confirmed that our method can harmonize the slice positional variance in terms of visceral fat area. Conclusion: This approach provides a promising direction for mapping slices from different vertebral levels to a target slice and reducing positional variance for single-slice longitudinal analysis. The source code is available at: https://github.com/MASILab/C-SliceGen.

Longitudinal Variability Analysis on Low-dose Abdominal CT with Deep Learning-based Segmentation.

Metabolic health is increasingly implicated as a risk factor across conditions from cardiology to neurology, and efficiency assessment of body composition is critical to quantitatively characterizing these relationships. 2D low dose single slice computed tomography (CT) provides a high resolution, quantitative tissue map, albeit with a limited field of view. Although numerous potential analyses have been proposed in quantifying image context, there has been no comprehensive study for low-dose single slice CT longitudinal variability with automated segmentation. We studied a total of 1816 slices from 1469 subjects of Baltimore Longitudinal Study on Aging (BLSA) abdominal dataset using supervised deep learning-based segmentation and unsupervised clustering method. 300 out of 1469 subjects that have two year gap in their first two scans were pick out to evaluate longitudinal variability with measurements including intraclass correlation coefficient (ICC) and coefficient of variation (CV) in terms of tissues/organs size and mean intensity. We showed that our segmentation methods are stable in longitudinal settings with Dice ranged from 0.821 to 0.962 for thirteen target abdominal tissues structures. We observed high variability in most organ with ICC<0.5, low variability in the area of muscle, abdominal wall, fat and body mask with average ICC≥0.8. We found that the variability in organ is highly related to the cross-sectional position of the 2D slice. Our efforts pave quantitative exploration and quality control to reduce uncertainties in longitudinal analysis.

Plant Protein but Not Animal Protein Consumption Is Associated with Frailty through Plasma Metabolites.

There is evidence that the association of protein intake and frailty may depend on the source of dietary protein. The mechanism underlying this association is not clear. In this study, we explore circulating metabolites as mediators of the relationship between dietary protein and of frailty in participants of the Baltimore Longitudinal Study of Aging (BLSA). Cross-sectional analyses in 735 BLSA participants of associations between plant and animal protein intake and frailty. Usual protein intake from plant and animal sources were estimated with a Food Frequency Questionnaire (FFQ) and frailty was assessed with a 44-item Frailty Index (FI). Compared with the lowest quartile, higher quartiles of plant, but not animal, protein were associated with lower FI. Twenty-five plasma metabolites were associated with plant protein intake; of these, fifteen, including phosphatidylcholines, cholesterol esters, sphingomyelins, and indole metabolites, mediated the association between plant protein intake and FI. The protective association between plant protein consumption and FI is mediated by lower abundance of lipid metabolites and higher abundance of tryptophan-related metabolites.

Single slice thigh CT muscle group segmentation with domain adaptation and self-training.

Purpose: Thigh muscle group segmentation is important for assessing muscle anatomy, metabolic disease, and aging. Many efforts have been put into quantifying muscle tissues with magnetic resonance (MR) imaging, including manual annotation of individual muscles. However, leveraging publicly available annotations in MR images to achieve muscle group segmentation on single-slice computed tomography (CT) thigh images is challenging. Approach: We propose an unsupervised domain adaptation pipeline with self-training to transfer labels from three-dimensional MR to single CT slices. First, we transform the image appearance from MR to CT with CycleGAN and feed the synthesized CT images to a segmenter simultaneously. Single CT slices are divided into hard and easy cohorts based on the entropy of pseudo-labels predicted by the segmenter. After refining easy cohort pseudo-labels based on anatomical assumption, self-training with easy and hard splits is applied to fine-tune the segmenter. Results: On 152 withheld single CT thigh images, the proposed pipeline achieved a mean Dice of 0.888 (0.041) across all muscle groups, including gracilis, hamstrings, quadriceps femoris, and sartorius muscle. Conclusions: To our best knowledge, this is the first pipeline to achieve domain adaptation from MR to CT for thigh images. The proposed pipeline effectively and robustly extracts muscle groups on two-dimensional single-slice CT thigh images. The container is available for public use in GitHub repository available at: https://github.com/MASILab/DA_CT_muscle_seg.

DPP4 inhibition impairs senohemostasis to improve plaque stability in atherosclerotic mice.

Senescent vascular smooth muscle cells (VSMCs) accumulate in the vasculature with age and tissue damage and secrete factors that promote atherosclerotic plaque vulnerability and disease. Here, we report increased levels and activity of dipeptidyl peptidase 4 (DPP4), a serine protease, in senescent VSMCs. Analysis of the conditioned media from senescent VSMCs revealed a unique senescence-associated secretory phenotype (SASP) signature comprising many complement and coagulation factors; silencing or inhibiting DPP4 reduced these factors and increased cell death. Serum samples from persons with high risk for cardiovascular disease contained high levels of DPP4-regulated complement and coagulation factors. Importantly, DPP4 inhibition reduced senescent cell burden and coagulation and improved plaque stability, while single-cell resolution of senescent VSMCs reflected the senomorphic and senolytic effects of DPP4 inhibition in murine atherosclerosis. We propose that DPP4-regulated factors could be exploited therapeutically to reduce senescent cell function, reverse senohemostasis, and improve vascular disease.

Accelerating 2D Abdominal Organ Segmentation with Active Learning.

Abdominal computed tomography CT imaging enables assessment of body habitus and organ health. Quantification of these health factors necessitates semantic segmentation of key structures. Deep learning efforts have shown remarkable success in automating segmentation of abdominal CT, but these methods largely rely on 3D volumes. Current approaches are not applicable when single slice imaging is used to minimize radiation dose. For 2D abdominal organ segmentation, lack of 3D context and variety in acquired image levels are major challenges. Deep learning approaches for 2D abdominal organ segmentation benefit by adding more images with manual annotation, but annotation is resource intensive to acquire given the large quantity and the requirement of expertise. Herein, we designed a gradient based active learning annotation framework by meta-parameterizing and optimizing the exemplars to dynamically select the 'hard cases' to achieve better results with fewer annotated slices to reduce the annotation effort. With the Baltimore Longitudinal Study on Aging (BLSA) cohort, we evaluated the performance with starting from 286 subjects and added 50 more subjects iteratively to 586 subjects in total. We compared the amount of data required to add to achieve the same Dice score between using our proposed method and the random selection in terms of Dice. When achieving 0.97 of the maximum Dice, the random selection needed 4.4 times more data compared with our active learning framework. The proposed framework maximizes the efficacy of manual efforts and accelerates learning.

Label efficient segmentation of single slice thigh CT with two-stage pseudo labels.

Purpose: Muscle, bone, and fat segmentation from thigh images is essential for quantifying body composition. Voxelwise image segmentation enables quantification of tissue properties including area, intensity, and texture. Deep learning approaches have had substantial success in medical image segmentation, but they typically require a significant amount of data. Due to the high cost of manual annotation, training deep learning models with limited human label data is desirable, but it is a challenging problem. Approach: Inspired by transfer learning, we proposed a two-stage deep learning pipeline to address the thigh and lower leg segmentation issue. We studied three datasets, 3022 thigh slices and 8939 lower leg slices from the BLSA dataset and 121 thigh slices from the GESTALT study. First, we generated pseudo labels for thigh based on approximate handcrafted approaches using CT intensity and anatomical morphology. Then, those pseudo labels were fed into deep neural networks to train models from scratch. Finally, the first stage model was loaded as the initialization and fine-tuned with a more limited set of expert human labels of the thigh. Results: We evaluated the performance of this framework on 73 thigh CT images and obtained an average Dice similarity coefficient (DSC) of 0.927 across muscle, internal bone, cortical bone, subcutaneous fat, and intermuscular fat. To test the generalizability of the proposed framework, we applied the model on lower leg images and obtained an average DSC of 0.823. Conclusions: Approximated handcrafted pseudo labels can build a good initialization for deep neural networks, which can help to reduce the need for, and make full use of, human expert labeled data.

Quantification of muscle, bones, and fat on single slice thigh CT.

Muscle, bone, and fat segmentation of CT thigh slice is essential for body composition research. Voxel-wise image segmentation enables quantification of tissue properties including area, intensity and texture. Deep learning approaches have had substantial success in medical image segmentation, but they typically require substantial data. Due to high cost of manual annotation, training deep learning models with limited human labelled data is desirable but also a challenging problem. Inspired by transfer learning, we proposed a two-stage deep learning pipeline to address this issue in thigh segmentation. We study 2836 slices from Baltimore Longitudinal Study of Aging (BLSA) and 121 slices from Genetic and Epigenetic Signatures of Translational Aging Laboratory Testing (GESTALT). First, we generated pseudo-labels based on approximate hand-crafted approaches using CT intensity and anatomical morphology. Then, those pseudo labels are fed into deep neural networks to train models from scratch. Finally, the first stage model is loaded as initialization and fine-tuned with a more limited set of expert human labels. We evaluate the performance of this framework on 56 thigh CT scans and obtained average Dice of 0.979,0.969,0.953,0.980 and 0.800 for five tissues: muscle, cortical bone, internal bone, subcutaneous fat and intermuscular fat respectively. We evaluated generalizability by manually reviewing external 3504 BLSA single thighs from 1752 thigh slices. The result is consistent and passed human review with 150 failed thigh images, which demonstrates that the proposed method has strong generalizability.

Longitudinal phenotypic aging metrics in the Baltimore Longitudinal Study of Aging.

To define metrics of phenotypic aging, it is essential to identify biological and environmental factors that influence the pace of aging. Previous attempts to develop aging metrics were hampered by cross-sectional designs and/or focused on younger populations. In the Baltimore Longitudinal Study of Aging (BLSA), we collected longitudinally across the adult age range a comprehensive list of phenotypes within four domains (body composition, energetics, homeostatic mechanisms and neurodegeneration/neuroplasticity) and functional outcomes. We integrated individual deviations from population trajectories into a global longitudinal phenotypic metric of aging and demonstrate that accelerated longitudinal phenotypic aging is associated with faster physical and cognitive decline, faster accumulation of multimorbidity and shorter survival. These associations are more robust compared with the use of phenotypic and epigenetic measurements at a single time point. Estimation of these metrics required repeated measures of multiple phenotypes over time but may uniquely facilitate the identification of mechanisms driving phenotypic aging and subsequent age-related functional decline.

Personality traits are consistently associated with blood mitochondrial DNA copy number estimated from genome sequences in two genetic cohort studies.

Background: Mitochondrial DNA copy number (mtDNAcn) in tissues and blood can be altered in conditions like diabetes and major depression and may play a role in aging and longevity. However, little is known about the association between mtDNAcn and personality traits linked to emotional states, metabolic health, and longevity. This study tests the hypothesis that blood mtDNAcn is related to personality traits and mediates the association between personality and mortality. Methods: We assessed the big five personality domains and facets using the Revised NEO Personality Inventory (NEO-PI-R), assessed depressive symptoms with the Center for Epidemiologic Studies Depression Scale (CES-D), estimated mtDNAcn levels from whole-genome sequencing, and tracked mortality in participants from the Baltimore Longitudinal Study of Aging. Results were replicated in the SardiNIA Project. Results: We found that mtDNAcn was negatively associated with the Neuroticism domain and its facets and positively associated with facets from the other four domains. The direction and size of the effects were replicated in the SardiNIA cohort and were robust to adjustment for potential confounders in both samples. Consistent with the Neuroticism finding, higher depressive symptoms were associated with lower mtDNAcn. Finally, mtDNAcn mediated the association between personality and mortality risk. Conclusions: To our knowledge, this is the first study to show a replicable association between mtDNAcn and personality. Furthermore, the results support our hypothesis that mtDNAcn is a biomarker of the biological process that explains part of the association between personality and mortality. Funding: Support for this work was provided by the Intramural Research Program of the National Institute on Aging (Z01-AG000693, Z01-AG000970, and Z01-AG000949) and the National Institute of Neurological Disorders and Stroke of the National Institutes of Health. AT was also supported by the National Institute on Aging of the National Institutes of Health Grant R01AG068093.

Cells are powered by internal structures called mitochondria which have their own DNA molecules. How many copies of mitochondrial DNA blood cells contain is one aspect of mitochondrial health and is considered to provide a good indication of an individual's ability to convert glucose into energy. Consequently, changes in the amount of mitochondrial DNA in the blood are linked to conditions like diabetes and cancer, and have also been associated with aging and mortality. A set of well-classified personality traits known as 'the Big Five' have also been shown to affect energy levels and the longevity of individuals. However, it remained unclear if there is a relationship between these characteristics and the number of copies of mitochondrial DNA in the blood. To investigate, Oppong et al. used a specialized test to assess the personality traits of participants from two separate cohorts: Baltimore Longitudinal Study of Ageing and the SardiNIA Project. The genomic sequence of each person was then analyzed to calculate the amount of mitochondrial DNA in their blood, and their mortality was recorded based on whether they were alive or dead multiple years later. Oppong et al. found that low levels of mitochondrial DNA were linked with high scores in neuroticism (a trait typically associated with anxiety, depression, and self-doubt). Further statistical tests revealed that mitochondrial DNA levels mediate the relationship between a person's personality and their risk of death. These findings suggest that personality traits impact the number of mitochondrial DNA molecules in a person's blood, which, in turn, influences how long they are likely to live. However, further work is needed to find out what causes this effect.

A computational solution for bolstering reliability of epigenetic clocks: Implications for clinical trials and longitudinal tracking.

Epigenetic clocks are widely used aging biomarkers calculated from DNA methylation data, but this data can be surprisingly unreliable. Here we show technical noise produces deviations up to 9 years between replicates for six prominent epigenetic clocks, limiting their utility. We present a computational solution to bolster reliability, calculating principal components from CpG-level data as input for biological age prediction. Our retrained principal-component versions of six clocks show agreement between most replicates within 1.5 years, improved detection of clock associations and intervention effects, and reliable longitudinal trajectories in vivo and in vitro. This method entails only one additional step compared to traditional clocks, requires no replicates or prior knowledge of CpG reliabilities for training, and can be applied to any existing or future epigenetic biomarker. The high reliability of principal component-based clocks is critical for applications to personalized medicine, longitudinal tracking, in vitro studies, and clinical trials of aging interventions.

Epigenetic Age Acceleration and Hearing: Observations From the Baltimore Longitudinal Study of Aging.

Objectives: Age-related hearing loss (ARHL) is highly prevalent among older adults, but the potential mechanisms and predictive markers for ARHL are lacking. Epigenetic age acceleration has been shown to be predictive of many age-associated diseases and mortality. However, the association between epigenetic age acceleration and hearing remains unknown. Our study aims to investigate the relationship between epigenetic age acceleration and audiometric hearing in the Baltimore Longitudinal Study of Aging (BLSA). Methods: Participants with both DNA methylation and audiometric hearing measurements were included. The main independent variables are epigenetic age acceleration measures, including intrinsic epigenetic age acceleration-"IEAA," Hannum age acceleration-"AgeAccelerationResidualHannum," PhenoAge acceleration-"AgeAccelPheno," GrimAge acceleration-"AgeAccelGrim," and methylation-based pace of aging estimation-"DunedinPoAm." The main dependent variable is speech-frequency pure tone average. Linear regression was used to assess the association between epigenetic age acceleration and hearing. Results: Among the 236 participants (52.5% female), after adjusting for age, sex, race, time difference between measurements, cardiovascular factors, and smoking history, the effect sizes were 0.11 995% CI: (-0.00, 0.23), p = 0.054] for Hannum's clock, 0.08 [95% CI: (-0.03, 0.19), p = 0.143] for Horvath's clock, 0.10 [95% CI: (-0.01, 0.21), p = 0.089] for PhenoAge, 0.20 [95% CI: (0.06, 0.33), p = 0.004] for GrimAge, and 0.21 [95% CI: (0.09, 0.33), p = 0.001] for DunedinPoAm. Discussion: The present study suggests that some epigenetic age acceleration measurements are associated with hearing. Future research is needed to study the potential subclinical cardiovascular causes of hearing and to investigate the longitudinal relationship between DNA methylation and hearing.