Download:
-
China is rapidly experiencing population aging, but studies on aging are limited by the traditional aging measure: chronological age (CA), particularly in the field of genomics. Several novel aging measures have been proposed, but they lack comparative evaluation. Therefore, this study aimed to develop and select a more accurate measure for aging and identify novel aging-associated genetic variants. We developed three aging measures with potential for promotion based on biomarkers screened by Lasso regression in 7,584 participants from the Changzhou cohort. We assessed their performance and chose PhenoAge as the best aging measure by area under the receiver operating characteristic (ROC) curve (AUC). To identify aging-associated loci, we conducted a genome-wide association study (GWAS) using PhenoAge acceleration (PhenoAgeAccel) as the aging phenotype among 1,215 genotyped participants through a linear regression model. A total of 24 aging-associated variants were identified, with 3 previously recognized as aging loci and 21 suggesting novel contributions. This study then further explored the functions of these loci on a multi-omics scale. These findings may provide new insights into aging and facilitate the development of a practical screening program based on PhenoAge, which aims to promote healthy aging in China.
Aging measures are important strategies for constructing and quantifying aging rate (1). Chronological age (CA) is the principal aging measure, offering advantages in simplicity and widespread acceptance. However, it introduces heterogeneity across studies (2). To overcome CA’s limitations, several promising aging measures have been proposed (3). Among these, biological age (BA) measures based on composite biomarkers are cost-effective and demonstrate improved performance in replicating reported aging loci and suggesting genetic correlations with age-related phenotypes (4). Given the numerous BA types, evaluating the most suitable BA for the Chinese population is crucial.
The Changzhou cohort is an independent prospective cohort from Wujin District, Changzhou City, Jiangsu Province, China (5). A total of 7,584 individuals were included in this study. The study design workflow is illustrated in
Supplementary Figure S1 . Lasso regression was used to screen for candidate biomarkers. Three BA measures — KDM-BA, PhenoAge, and HD (6) — are more appropriate for application in China. These measures were developed based on candidate biomarkers, and the measure with the highest AUC value was selected. Aging acceleration was estimated as the residual of these aging measures after adjusting for CA using a linear regression model. Details regarding the subjects and BA measures are provided in theSupplementary Material .We performed quality control for 1,244 genotyped participants and included 1,215 participants with approximately 8.15 million variants. PLINK2.0 was used to perform linear regression analysis on the aging phenotype (7), adjusting for the first 10 principal components (PCs), age, sex, and genotyping batch. In the identification of aging-associated loci, we set the significance level at P<10-5 and considered leading SNPs with a threshold of r2≥0.6. ANNOVAR, Reactome, GWAS Catalog, IEUgwas, and GTEx were used for functional mapping and annotation of identified loci. To further explore the functions of these loci on the proteome and phenome, we analyzed protein-aging association and performed a phenome-wide association analysis (PheWAS) in 430,000 UK Biobank (UKB) participants. In the protein-aging association analysis, we adjusted for sex, age, lifestyle, and other appropriate factors. The details of the above analyses are provided in the
Supplementary Material .A total of 7,584 participants were followed up from 2012 to 2019, with a mean follow-up time of 6.37 ± 0.53 years. There were 3 BA measures constructed based on sex and 18 biomarkers screened by Lasso regression. Based on the AUC, PhenoAge (AUC=0.79) was selected as the optimal aging measure for this study (Figure 1). The formula and distribution of PhenoAge and PhenoAgeAccel are shown in
Supplementary Figure S2 .
Figure 1.Screening of variables based on Lasso regression and evaluation of aging measures based on AUC. (A) The variation characteristics of the coefficient of variables in Lasso regression; (B) The selection process of the optimum value of the parameter lambda in the Lasso regression model by cross-validation method; (C) AUC of three aging measures in the Changzhou Cohort; (D) AUC of three aging measures in the Changzhou genotyped group.
Abbreviation: AUC=area under the receiver operating characteristic curve; HD=homeostatic dysregulation; ROC=receiver operating characteristic curve.We identified 24 aging-associated loci tagged by 24 lead SNPs for PhenoAgeAccel (Table 1), 3 of which were previously reported. The genes mapped by these loci were primarily involved in programmed cell death, cell cycle, and immune system pathways. The effects of several identified variants on aging were significantly heterogeneous among age and sex subgroups. Notably, lead SNPs rs9376269, rs76772550, and rs67548191 were identified as eQTLs of IFNGR1, SLC24A4, and GGPS1, respectively, in both the IEUgwas and GTEx databases. IFNGR1 protein was also positively correlated with PhenoAgeAccel P<0.001 in the UKB. Therefore, rs9376269 may affect aging through its influence on the transcription and translation of IFNGR1.
SNP CHR POS RA EA MAF BETA SE P Related genes Region rs78629466 1 156087706 G A 0.098 1.539 0.344 7.49×10−6 LMNA Intronic rs34186915 1 206806424 C A 0.034 2.547 0.557 4.87×10−6 EIF2D, DYRK3 Intergenic rs67548191 1 235324984 C T 0.017 4.721 1.065 9.30×10−6 RBM34 Upstream rs6712152 2 155748871 G C 0.044 2.552 0.492 2.11×10−7 KCNJ3, LINC01876 Intergenic rs75159275 2 156053319 T A 0.036 2.929 0.544 7.15×10−8 KCNJ3, LINC01876 Intergenic rs17266628 3 122127926 G A 0.279 −1.096 0.225 1.09×10−6 FAM162A Intronic rs534553017 4 186795694 C T 0.015 −5.479 1.145 1.72×10−6 SORBS2 Intronic rs143045396 4 190324195 G T 0.054 2.077 0.45 4.02×10−6 LINC02508, LINC01262 Intergenic rs199764613 5 120002898 G A 0.026 −4.132 0.891 3.56×10−6 PRR16 Intronic rs761656272 5 159746937 TAATA T 0.018 −3.620 0.819 9.99×10−6 CCNJL Intergenic rs9376269 6 137539505 G C 0.476 −1.082 0.205 1.25×10−7 IFNGR1 Intronic rs10237037 7 7848087 C T 0.036 2.655 0.546 1.18×10−6 UMAD1 Intronic rs78007164 7 75998408 C T 0.028 3.006 0.619 1.18×10−6 YWHAG, SSC4D Intergenic rs137923974 7 148380093 G A 0.029 −2.961 0.607 1.10×10−6 C7orf33, CUL1 Intergenic rs1446270 9 12046808 C T 0.476 −0.937 0.199 2.57×10−6 PTPRD-AS2, TYRP1 Intergenic rs546542 9 77332549 C T 0.278 −1.028 0.228 6.49×10−6 RORB, TRPM6 Intergenic rs79594032 14 20724036 C T 0.022 4.566 0.959 1.95×10−6 OR11H4, TTC5 Intergenic rs2060609 14 43338730 G A 0.162 1.251 0.277 6.21×10−6 LRFN5, FSCB Intergenic rs76772550 14 92970259 C A 0.184 −1.277 0.266 1.56×10−6 SLC24A4, RIN3 Intergenic rs4778079 15 93593528 C T 0.207 1.157 0.251 3.90×10−6 RGMA Intronic rs138259742 16 73965374 T A 0.016 4.998 1.123 8.57×10−6 LINC01568, LOC101928035 Intergenic rs1432070 18 70162291 C T 0.231 −1.073 0.233 4.01×10−6 LINC01899, CBLN2 Intergenic rs11881034 19 6281607 C T 0.292 0.985 0.22 7.87×10−6 MLLT1, ACER1 Intergenic rs285200 20 42366834 C T 0.013 6.257 1.241 4.63×10−7 GTSF1L, LINC01728 Intergenic Abbreviation: BETA=beta coefficient; CHR=chromosome; EA=effect allele; MAF=minor allele frequency; POS=position; RA=reference allele; SE=standard error; SNP=single nucleotide polymorphism. Table 1. Results for 24 identified aging-associated risk loci.
In the PheWAS, we found 212 significant associations between 24 variants and 75 outcomes (Figure 2). Half of these variants were associated with age-related phenotypes such as diabetes, hypertension, and cardiovascular diseases. Additionally, other variant-phenotype associations involved metabolic diseases, infectious diseases, blood diseases, and certain tumors. These results suggested that these aging-associated variants were associated with aging-related phenotypes, and that metabolic, infectious, and cardiovascular diseases contributed most to aging in the Changzhou population.
Figure 2.Phenotypes associated with aging-associated variants in UKB (n=437,464). (A) Manhattan plot for phenome-wide associations of rs9376269, rs6712152. (B) Heatmap for phenotypes associated with aging-associated variants.
Note: The red horizontal line and “*” denote phenome-wide significance (P<2.70×10−5). -
This study investigated appropriate aging measures and aging-associated variants in the Changzhou cohort. Among BA measures, PhenoAge was identified as the most applicable for this population through comparative evaluation. Twenty-four aging-associated loci were identified based on PhenoAgeAccel.
Previous aging studies, primarily based on CA, identified numerous aging-associated loci. Of these, only Apolipoprotein E (APOE) gene has been widely recognized as an aging-associated gene (2). A limitation of these studies is the difficulty in eliminating heterogeneity introduced by varying longevity thresholds (2). To address this, alternative aging measures, particularly BA measures, have been developed. However, the optimal measure for quantifying aging remains uncertain. Our research compared aging measures from similar previous studies and identified PhenoAge as optimal for the current population.
Compared with previous similar studies, our study employed lasso regression to further screen biomarkers, improving the reliability of aging measures. Additionally, we performed a GWAS of PhenoAgeAccel and identified 24 aging-associated loci. Of these, rs78629466, rs78007164, and rs1446270 were located within previously reported age-associated loci. These identified lead SNPs are mainly enriched in programmed cell death, the cell cycle, and the immune system. Cell death and the cell cycle are fundamental to aging, and chronic activation of the immune system, resulting in low-grade inflammation, is one of the hallmarks of aging. Furthermore, a prior GWAS also found a strong association between PhenoAgeAccel and chronic inflammatory and autoimmune diseases (4). In GTEx analysis, rs76772550 and rs67548191 were identified as eQTLs of SLC24A4 and GGPS1, respectively. These two genes have been reported to be associated with brain and ovary aging. Notably, rs9376269, one of the most significant lead SNPs in our study, was an eQTL of IFNGR1, which was also associated with PhenoAgeAccel. IFNGR1 has been reported to play a crucial role in antimicrobial, antiviral, and antitumor responses (8). In a previous mouse experiment, interferon-γ was also found to be involved in the aging process (9). Combining these results, we hypothesized that rs9376269 might affect the immune system by regulating the expression of IFNGR1. Additionally, we further found that the identified aging-associated variants are not only implicated in the immune system but are also significantly associated with diabetes and other metabolic phenotypes. These findings support the view of a previous longevity study that defensive mechanisms (such as immunity) and metabolism play a key role in aging and longevity in the Chinese population (10). In this study, the parents of participants may have experienced the adversities of famine periods. According to the Developmental Origins of Health and Disease theory, such conditions might be associated with a higher incidence of cardiovascular and metabolic diseases in their offspring. Thus, the effect of aging genes mainly related to inflammation and metabolism is more pronounced in this study.
This study has several limitations. The participants were recruited from Changzhou in Jiangsu Province, a relatively developed city in China with a predominantly Han population, potentially limiting the generalizability of our findings to the broader Chinese population. Additionally, the small sample size and stringent quality control standards may have limited our ability to replicate previously reported aging-associated variants, particularly those related to APOE.
In summary, our study established PhenoAge as an aging measure that may demonstrate greater applicability in the Chinese population. Furthermore, we identified several novel aging-associated variants. Our findings showed significant potential for developing a practical, PhenoAge-based aging screening method to promote healthy aging in China.
-
The well-dedicated participants and staff who participated in the Changzhou cohort study.
HTML
| Citation: |
|