Abstract Background Biological aging is a critical risk factor of age-related diseases, but its impact on diabetic individuals remains unclear. This study aimed to examine the associations of biological aging with the incident cardiovascular diseases (CVDs) and life expectancy loss in diabetic individuals. Methods We included 12,828 diabetic individuals in UK biobank. Biological aging was calculated by Klemera-Doubal method Biological Age (KDMAge) and phenotypic age (PhenoAge). Cox proportional hazard models were fitted to investigate the associations of biological aging with incident coronary heart disease (CHD), atrial fibrillation (AF), heart failure (HF), stroke, and degenerative valvular heart disease (VHD) in diabetic individuals. We also evaluated life expectancy loss in accelerated aging individuals, the interactions between biological aging and clonal hematopoiesis of indeterminate potential (CHIP), and performed causal mediation analysis. Results During a median follow-up of 13.1 years, we documented 3794 incident CVDs in diabetic individuals. PhenoAge accelerated aging was significantly associated with all CVD subtypes, with hazard ratios ranging from 1.23 to 1.62, and KDMAge showed even stronger associations. Accelerated biological aging was also associated with over 2 years of life expectancy loss. CHIP and PhenoAge accelerated aging had a significant synergistic effect on CHD, HF, and VHD. Inflammatory activation contributed significantly to accelerated aging-associated CHD and HF. Conclusions Biological aging significantly increases CVD risk and reduces life expectancy in diabetic population, with effects modified by CHIP status. Targeting biological aging mechanisms may help prevent CVDs and premature mortality in diabetic population. Graphical abstract AF, atrial fibrillation; CHD, coronary heart disease; CHIP, clonal hematopoiesis of indeterminate potential; CI, confidence interval; CVD, cardiovascular disease; HF, heart failure; HR, hazard ratio; KDMAge, Klemera-doubal method biological age; PAR, partial population attributable risk; PhenoAge, phenotypic age; VHD, valvular heart disease.[38] graphic file with name 12933_2025_2855_Figa_HTML.jpg Supplementary Information The online version contains supplementary material available at 10.1186/s12933-025-02855-w. Keywords: Biological aging, PhenoAge, KDMAge, Diabetes, Cardiovascular diseases Research insights What is currently known about this topic? * Diabetes increases cardiovascular disease (CVD) risk, which remains the leading cause of mortality in this population. * Biological aging metrics predict mortality better than chronological age alone. * Clonal hematopoiesis of indeterminate potential (CHIP) is associated with an increased CVD risk in diabetic individuals. What is the key research question? * Is accelerated biological aging associated with adverse outcomes in diabetic individuals? What is new? * Accelerated biological aging was associated with CVDs development and loss of life expectancy in diabetic individuals. * Synergistic effect between accelerated biological aging and CHIP mutations substantially increased CVD risk. * Inflammation activation, especially neutrophil degranulation pathway, may play an important role in the association between accelerated biological aging and CVDs. How might this study influence clinical practice? * Early identification and management of biological aging may help reduce cardiovascular burden of diabetes. Background Diabetes is a global health crisis with profound healthcare and economic consequences, and the number of cases is projected to surpass 783 million by 2045 [[39]1]. Individuals with diabetes confront a substantially higher risk of cardiovascular disease (CVD), which persist as the predominant cause of mortality in this population [[40]2]. Consequently, elucidating the risk factors that drive CVD development in this vulnerable population remains imperative for mitigating the escalating burden of diabetes. Diabetes frequently accelerates biological aging and induces multi-organ dysregulation, which advances independently of chronological age and parallels disease progression [[41]3]. Diverse measurements reflecting the aging landscape within intrinsic biological systems have been proposed, including telomere length and sophisticated algorithms integrating epigenetic, proteomic, metabolomic, inflammatory, and organ function parameters [[42]4]. These metrics have demonstrated superior capacity in predicting mortality and age-related diseases compared to chronological age alone [[43]5–[44]7]. Among these, the Klemera-Doubal method Biological Age (KDMAge) and phenotypic age (PhenoAge) have emerged as widely adopted surrogates for biological aging, calculated from readily accessible clinical indicators and blood chemistry parameters [[45]8, [46]9]. Longitudinal investigation has established the prognostic significance of PhenoAge in diabetes development, progression to diabetic complications, and mortality in the general population [[47]10]. Nevertheless, studies examining the relationship between KDMAge/PhenoAge and CVD outcomes, along with their underlying mechanistic pathways, remain absent in the diabetic population. Moreover, the impact of accelerated biological aging on life expectancy in this population remains inadequately characterized. Biological aging is substantially influenced by genetic determinants [[48]11]. Recent investigations have identified clonal hematopoiesis of indeterminate potential (CHIP)—defined by the presence of cancer-associated somatic mutations in hematopoietic cells without evidence of hematologic malignancy—as a crucial acquired genetic factor intimately associated with biological aging [[49]12, [50]13]. Of note, CHIP and accelerated biological aging demonstrate a pronounced synergistic effect on adverse clinical outcomes [[51]13]. Although the association between CHIP and incident CVD risk in diabetic individuals has been established, it remains uncertain whether these relationships are modulated by biological aging [[52]14]. Given these knowledge gaps, we leveraged comprehensive prospective cohort data from the UK Biobank to evaluate within the diabetic population: (1) the association between biological aging and CVD development while characterizing aging-associated mediator pathways; (2) potential synergistic interactions between accelerated biological aging and CHIP in the pathogenesis of CVDs; and (3) the quantitative impact of accelerated biological aging on life expectancy loss. Our findings may inform targeted preventive strategies and personalized risk assessment in this high-risk population. Methods Study participants The UK Biobank is a large-scale, prospective cohort study comprising over 0.5 million participants aged 40–69 years recruited between 2006 and 2010 across the United Kingdom. The study collected comprehensive baseline data encompassing physical examinations, sociodemographic characteristics, lifestyle factors, detailed medical history, and biological samples [[53]15]. Plasma proteomic profiling was subsequently performed on 52,704 participants using the OLINK platform [[54]16]. For the present analysis, we identified 26,310 participants with diabetes at baseline (Supplementary Table [55]S1). We included 12,828 participants after excluding those with incomplete data for biological age calculation (n = 9,834), or with pre-existing cardiovascular conditions (n = 3,648), including coronary heart disease (CHD), atrial fibrillation (AF), heart failure (HF), stroke, and any valvular heart disease (VHD). In pathway mediator analysis, we finally enrolled 3,553 diabetic participants who had available proteomic data (Supplementary Figure [56]S1). This study was conducted in accordance with the Declaration of Helsinki and was approved by the North West Multi-centre Research Ethics Committee (REC reference: 11/NW/0382). Assessment of biological aging We evaluated biological aging using two well-validated algorithms—PhenoAge and KDMAge [[57]8, [58]9]. Both algorithms were calculated based on clinical biomarkers collected at baseline assessment. KDMAge was computed using forced expiratory volume in 1 s, systolic blood pressure, concentrations of albumin, alkaline phosphatase, blood urea nitrogen, creatinine, C-reactive protein, glycated hemoglobin, and total cholesterol. The calculations of KDMAge were performed separately for males and females [[59]8]. PhenoAge was derived from chronological age and nine biomarkers: concentrations of albumin, alkaline phosphatase, creatinine, glucose, C-reactive protein, lymphocyte percentage, mean cell volume, red blood cell distribution width, and white blood cell count (Supplementary Methods 1, Table [60]S2) [[61]9]. All calculations were performed using the BioAge R package ([62]https://github.com/dayoonkwon/BioAge) [[63]17]. Age acceleration was defined as the residual values derived from regressing biological age measures against chronological age. Participants with age acceleration > 0 were classified as having accelerated aging. Clonal hematopoiesis of indeterminate potential CHIP was assessed using whole-exome sequencing data available for all study participants. Somatic variants associated with CHIP were identified using the Mutect2 following the Genome Analysis Toolkit best practices pipeline. The analysis focused on a curated panel of 58 CHIP-driver genes (Supplementary Table [64]S3). To minimize the number of false-positive CHIP calls, we further excluded variants with a total sequencing depth of < 20×, a depth for the reference or alternate allele of < 5×, a variant allele fraction of < 2.0%, or without any variant support in either forward or reverse sequencing reads. Sequencing artefacts and germline variants were removed as described previously [[65]18]. Covariates assessment All covariates were collected at baseline through touch-screen questionnaires, physical measurements, and verbal interviews. Socio-demographic characteristics included age, sex (male or female), ethnicity (White, Black, Asian, Mixed, or Other), Townsend deprivation index (TDI), education level (college/university degree or others), and household income (≥£31,000 or not). Lifestyle factors were comprised of smoking status (never, former, or current), alcohol intake, physical activity, and dietary score. Alcohol intake (g/day) was estimated based on the frequency of alcohol intake and the standardized amounts of various beverage types [[66]19]. Physical activity was classified as ideal if participants reported ≥ 150 min/week of moderate activity, ≥ 75 min/week of vigorous activity, or 150 min/week of combined activity [[67]20]. The diet score was evaluated by the consumption of fruits, vegetables, fish, red meat, and processed meat [[68]19]. Information on medications (lipid-lowering, antihypertensive, and anti-diabetic medication) was collected by verbal interview at baseline. We also assessed clinical comorbidities and medical history, including hypertension, hyperlipidemia, and cancer, which were systematically identified using standardized diagnostic codes from the International Classification of Diseases, 10th revision (ICD-10). Blood pressure was collected by physical measurements. Diabetes type was classified as insulin-dependent, non-insulin dependent, and others based on the ICD-10 diagnosis. Diabetes duration (year) was calculated by recruitment and diabetes diagnosis date. Detailed information for covariates was provided in Supplementary Methods [69]2 and Tables [70]S4–[71]S6. Ascertainment of outcomes The primary outcomes were incident CVD events, including CHD, AF, HF, any stroke, and degenerative VHD. Outcomes were identified through multiple sources, including death registers, primary care records, hospital inpatient records, and self-reported history. Disease events were defined using ICD-10 codes: I20-I25 for CHD, I48 for AF, I50 for HF, and I60-I64 and I69 for stroke (I60-I62 for hemorrhagic stroke, I63 for ischemic stroke) [[72]14]. Degenerative VHD cases, a composite of aortic valve stenosis (AVS), aortic valve regurgitation (AVR), and mitral valve regurgitation (MVR) were identified using both ICD-10 and the Office of Population Censuses and Surveys Classification of Interventions and Procedures version 4 (OPCS-4) codes, following previously established criteria [[73]19]. Participants were followed up until the first occurrence of any CVD event, death, or the censoring date (December 12, 2022), whichever occurred first (Supplementary Table [74]S4). Statistical analysis Missing categorical variables were coded as missing, while continuous variables were imputed with median values [[75]21]. Continuous variables were tested for normality using the Shapiro-Wilk test and presented as median (interquartile range, [IQR]) or mean (standard deviation, SD) accordingly. Categorical variables were presented as numbers (percentages). Between-group comparisons were performed by Student’s t-test or Wilcoxon rank-sum test for continuous variables and Pearson chi-squared or Fisher exact test for categorical variables, as appropriate. To facilitate direct comparison of effect sizes between PhenoAge and KDMAge acceleration measures, we standardized both measures as Z-scores with mean value of 0 and standard deviation of 1. For dose-response analyses, we categorized both measures into quartiles based on their original distributions. Cox proportional hazards models were used to estimate the hazard ratios (HRs) and 95% confidence intervals (CIs) in the associations of standardized continuous values, quartiles (Q1 as reference), and binary classification (non-accelerated aging as reference) with incident CVDs. All models were fitted with age, sex, ethnicity, TDI, education, household income, BMI, medication use (lipid-lowering, antihypertensive, and anti-diabetic medication), comorbidities (hypertension, hyperlipidemia, and cancer), types of diabetes, diabetes duration, smoking status, ideal physical activity, alcohol intake, and diet score. The proportional hazards assumption was tested based on Schoenfeld residuals (Supplementary Table [76]S7). Variance inflation factor and tolerance values were computed to detect potential multicollinearity among all independent variables (Supplementary Table [77]S8) [[78]22]. Partial population attributable risk (PAR) percentages of accelerated aging for each incident CVD were calculated using par macro in SAS [[79]23]. Restricted cubic splines (RCS) models were further constructed with three knots placed at the 10th, 50th, and 90th percentiles with the same covariates of Cox models. Subgroup analyses were performed by stratifying age (< 65 vs. ≥ 65 years), sex (female vs. male), BMI categories (normal: ≤25 kg/m², overweight: >25 to < 30 kg/m², obese: ≥30 kg/m²), household income (<£31,000 vs. ≥£31,000), education level (university/college degree vs. others), diabetes type (insulin-dependent, non-insulin dependent, or others) and CHIP status (no vs. yes). To evaluate the joint effects of CHIP and accelerated aging, we conducted additive interaction analyses where participants were categorized into four groups based on their CHIP and biological aging status, with those having neither CHIP nor accelerated biological aging serving as the reference. The relative excess risk due to interaction (RERI) and attributable proportion (AP) were calculated using the interaction R, with 95% CIs estimated using the delta method. Multi-state models were further used to analyze transitions from baseline status to CVD onset and death related to accelerated aging status [[80]10]. We then conducted flexible parametric proportional hazards analyses using the stpm2 command in Stata to evaluate the impact of accelerated aging on life expectancy following the previous studies [[81]24, [82]25]. Given that chronic disease prevalence increases substantially, and mortality data becomes more reliable after age 45 years, we focused our life expectancy analyses on individuals aged 45 years and older. For proteomic analysis, GLIPR1 (99.7% missing), NPM1 (74.0% missing), and PCOLCE (63.6% missing) were first excluded due to excessive missing values. The remaining missing values were imputed using the K-nearest neighbors’ algorithm [[83]26]. To identify differentially expressed proteins (DEPs) between individuals with and without accelerated aging, we implemented linear regression models using limma R package, adjusting for the same covariates used in Cox models along with technical batch effects. Statistical significance was determined using the Benjamini-Hochberg (BH) procedure to control the false discovery rate (FDR), with proteins showing FDR < 0.05 considered as significant DEPs. We then conducted pathway enrichment analysis on the upregulated DEPs using Reactome pathways. To quantitatively represent each enriched pathway, we performed principal component analysis (PCA) on all constituent proteins within each pathway [[84]27, [85]28]. PC1, which captured the maximum variance in pathway-specific protein expression, was extracted as a composite score for the corresponding pathway. Subsequently, we conducted causal mediation analyses using Regmedint R package to investigate how these pathways potentially mediate the relationship between accelerated aging and incident CVDs [[86]29]. Multiple testing correction was performed using the BH method, accounting for the number of tested pathways (57 for PhenoAge and 75 for KDMAge) across five CVDs. For the mediation analysis, pure natural direct effect (PNDE), total natural indirect effect (TNIE) indirect effects, total effect (TE), and proportion of effect mediated were estimated with corresponding 95% CIs. To evaluate the robustness of our primary findings, we conducted several sensitivity analyses, including: (1) multiple imputation of missing data using the MICE R package with subsequent replication of main analyses (Supplementary Methods 3); (2) exclusion of participants who experienced CVDs within the first year of follow-up; and (3) implementation of Fine-Gray competing risk models to account for death as a competing event. Statistical analyses were performed with R version 4.2.1, Stata 18MP, and SAS version 9.4. Two-sided P < 0.05 was considered statistically significant. Results Baseline participant characteristics Among 12,828 participants with diabetes, the median age was 61.2 years (IQR: 54.8–65.4). 5311 (41.4%) and 6,233 (48.6%) participants exhibited accelerated aging based on PhenoAge and KDMAge, respectively. Individuals with accelerated biological aging were predominantly male with higher BMI and greater hypertension burden. Additionally, they demonstrated lower socioeconomic status and adverse lifestyle patterns, including reduced physical activity, poorer dietary quality, and lower alcohol consumption. Besides, they were more likely to be diagnosed with insulin-dependent diabetes and had a longer duration of diabetes. Notably, individuals with PhenoAge accelerated aging tended to harbor more CHIP mutations (Table [87]1). Table 1. Baseline characteristics of participants Overall (N = 12,828) PhenoAge KDMAge Non-accelerated aging (N = 7,517) Accelerated aging (N = 5,311) P value Non-accelerated aging (N = 6,595) Accelerated aging (N = 6,233) P value Age, years, median [IQR] 61.2 [54.8, 65.4] 61.2 [54.8, 65.4] 61.2 [54.8, 65.6] 0.284 61.1 [54.6, 65.3] 61.3 [54.9, 65.6] 0.007 Male, n (%) 7770 (60.6) 4469 (59.5) 3301 (62.2) 0.002 4283 (64.9) 3487 (55.9) < 0.001 Ethnicity, n (%) 0.002 0.018 White 11247 (87.7) 6532 (86.9) 4715 (88.8) 5727 (86.8) 5520 (88.6) Black 174 (1.4) 116 (1.5) 58 (1.1) 103 (1.6) 71 (1.1) Asian 487 (3.8) 319 (4.2) 168 (3.2) 265 (4.0) 222 (3.6) Mixed 592 (4.6) 346 (4.6) 246 (4.6) 309 (4.7) 283 (4.5) Others 228 (1.8) 148 (2.0) 80 (1.5) 135 (2.0) 93 (1.5) Missing 100 (0.8) 56 (0.7) 44 (0.8) 56 (0.8) 44 (0.7) TDI, median [IQR] −1.45 [−3.26, 1.75] −1.55 [−3.30, 1.61] −1.28 [−3.22, 1.95] < 0.001 −1.57 [−3.32, 1.68] −1.30 [−3.20, 1.85] 0.003 BMI, median [IQR] 30.28 [27.06, 34.27] 29.76 [26.67, 33.27] 31.20 [27.66, 35.59] < 0.001 29.22 [26.18, 32.70] 31.57 [28.23, 35.71] < 0.001 Education, n (%) 0.161 < 0.001 University or college degree 3233 (25.2) 1940 (25.8) 1293 (24.3) 1844 (28.0) 1389 (22.3) Others 9346 (72.9) 5435 (72.3) 3911 (73.6) 4629 (70.2) 4717 (75.7) Missing 249 (1.9) 142 (1.9) 107 (2.0) 122 (1.8) 127 (2.0) Household income, n (%) 0.012 < 0.001 ≥£31,000 4114 (32.1) 2488 (33.1) 1626 (30.6) 2320 (35.2) 1794 (28.8) <£31,000 6534 (50.9) 3767 (50.1) 2767 (52.1) 3237 (49.1) 3297 (52.9) Missing 2180 (17.0) 1262 (16.8) 918 (17.3) 1038 (15.7) 1142 (18.3) Smoking status, n (%) < 0.001 0.627 Never 6159 (48.0) 3671 (48.8) 2488 (46.8) 3173 (48.1) 2986 (47.9) Former 5232 (40.8) 3101 (41.3) 2131 (40.1) 2687 (40.7) 2545 (40.8) Current 1321 (10.3) 681 (9.1) 640 (12.1) 669 (10.1) 652 (10.5) Missing 116 (0.9) 64 (0.9) 52 (1.0) 66 (1.0) 50 (0.8) Alcohol intake, grams/day, median [IQR] 5.81 [5.81, 22.45] 6.97 [5.81, 22.89] 5.81 [5.81, 21.53] 0.008 7.89 [5.81, 23.71] 5.81 [5.81, 21.03] < 0.001 Ideal physical activity, n (%)^* 8100 (63.1) 4887 (65.0) 3213 (60.5) < 0.001 4210 (63.8) 3890 (62.4) 0.098 Diet score, n (%) < 0.001 < 0.001 0 435 (3.4) 231 (3.1) 204 (3.8) 200 (3.0) 235 (3.8) 1 6857 (53.5) 3901 (51.9) 2956 (55.7) 3451 (52.3) 3406 (54.6) 2 4571 (35.6) 2743 (36.5) 1828 (34.4) 2397 (36.3) 2174 (34.9) 3 965 (7.5) 642 (8.5) 323 (6.1) 547 (8.3) 418 (6.7) Comorbidities Hypertension, n (%) 8154 (63.6) 4688 (62.4) 3466 (65.3) 0.001 3840 (58.2) 4314 (69.2) < 0.001 Hyperlipidemia, n (%) 5304 (41.3) 3165 (42.1) 2139 (40.3) 0.04 2737 (41.5) 2567 (41.2) 0.729 Cancer, n (%) 1573 (12.3) 905 (12.0) 668 (12.6) 0.374 732 (11.1) 841 (13.5) < 0.001 Diabetes type < 0.001 < 0.001 Insulin dependent 1087 (8.5) 447 (5.9) 640 (12.1) 452 (6.9) 635 (10.2) Non− insulin dependent 5349 (41.7) 3066 (40.8) 2283 (43.0) 2754 (41.8) 2595 (41.6) Others 6392 (49.8) 4004 (53.3) 2388 (45.0) 3389 (51.4) 3003 (48.2) Diabetes duration 5.40 [2.36, 10.25] 4.49 [1.96, 8.86] 6.91 [3.31, 12.24] < 0.001 5.30 [2.35, 9.94] 5.51 [2.38, 10.60] 0.021 CHIP, n (%) 499 (3.9) 266 (3.5) 233 (4.4) 0.016 250 (3.8) 249 (4.0) 0.581 Medications Antihypertensive medication, n (%) 7504 (58.5) 4221 (56.2) 3283 (61.8) < 0.001 3618 (54.9) 3886 (62.3) < 0.001 Lipid-lowering medication, n (%) 9098 (70.9) 5280 (70.2) 3818 (71.9) 0.045 4896 (74.2) 4202 (67.4) < 0.001 Anti-diabetic medication, n (%) 9374 (73.1) 4950 (65.9) 4424 (83.3) < 0.001 4790 (72.6) 4584 (73.5) 0.252 Biological ages KDMAge, years, median [IQR] 67.83 [62.48, 72.65] 66.13 [60.87, 70.82] 70.35 [65.21, 75.06] < 0.001 63.75 [58.78, 67.70] 72.54 [68.38, 76.17] < 0.001 KDMAge acceleratiom, median [IQR] −0.16 [−3.65, 3.34] −1.59 [−5.01, 1.39] 2.05 [−1.17, 5.92] < 0.001 −3.54 [−6.67, −1.59] 3.47 [1.54, 6.40] < 0.001 KDMAge accelerated aging, n (%) 6233 (48.6) 2706 (36.0) 3527 (66.4) < 0.001 0 (0.0) 6233 (100.0) < 0.001 PhenoAge, years, median [IQR] 63.14 [56.47, 69.80] 58.67 [52.74, 63.32] 71.31 [65.87, 76.75] < 0.001 60.54 [54.17, 66.39] 66.11 [59.35, 73.22] < 0.001 PhenoAge acceleration, median [IQR] − 1.70 [− 6.14, 4.40] − 5.41 [− 8.34, − 2.83] 5.94 [2.55, 11.72] < 0.001 − 4.07 [− 7.77, 0.45] 1.28 [− 3.72, 8.03] < 0.001 PhenoAge accelerated, aging, n (%) 5311 (41.4) 0 (0.0) 5311 (100.0) < 0.001 1784 (27.1) 3527 (56.6) < 0.001 Components of biological ages FEV1, L, median [IQR] 2.59 [2.08, 3.16] 2.62 [2.11, 3.20] 2.55 [2.04, 3.11] < 0.001 2.71 [2.17, 3.27] 2.48 [1.99, 3.03] < 0.001 SBP, mm Hg, median [IQR] 143.00 [131.00, 155.00] 142.00 [131.00, 155.00] 143.00 [132.00, 156.00] 0.013 137.00 [126.00, 147.00] 150.00 [138.00, 162.00] < 0.001 Total Cholesterol, mg/dL, median [IQR] 172.89 [149.96, 200.58] 174.21 [150.77, 202.21] 170.94 [148.80, 198.37] < 0.001 165.05 [144.28, 188.84] 182.60 [157.64, 211.89] < 0.001 HbA1c, %, median [IQR] 4.98 [4.31, 5.87] 4.64 [4.09, 5.29] 5.64 [4.86, 6.69] < 0.001 4.74 [4.14, 5.42] 5.36 [4.57, 6.45] < 0.001 BUN, mg/dL, median [IQR] 15.46 [13.03, 18.29] 15.27 [12.89, 17.87] 15.80 [13.28, 18.96] < 0.001 14.65 [12.44, 17.17] 16.39 [13.84, 19.47] < 0.001 Lymphocyte, %, median [IQR] 27.95 [23.20, 33.19] 29.61 [24.96, 34.68] 25.60 [21.15, 30.70] < 0.001 28.40 [23.80, 33.70] 27.40 [22.51, 32.70] < 0.001 MCV, fL, median [IQR] 90.40 [87.50, 93.29] 90.41 [87.70, 93.13] 90.35 [87.15, 93.43] 0.285 90.82 [87.90, 93.61] 89.92 [87.10, 92.90] < 0.001 Serum glucose, mmol/L, median [IQR] 6.47 [5.26, 8.84] 5.62 [4.91, 6.61] 9.31 [7.08, 12.07] < 0.001 6.09 [5.12, 7.84] 7.06 [5.51, 10.03] < 0.001 RDW, %, median [IQR] 13.43 [12.97, 14.03] 13.28 [12.87, 13.74] 13.77 [13.20, 14.50] < 0.001 13.40 [12.92, 13.95] 13.50 [13.00, 14.10] < 0.001 WBC, 1000 cells/ uL, median [IQR] 7.35 [6.20, 8.70] 7.10 [6.05, 8.37] 7.74 [6.48, 9.14] < 0.001 7.10 [6.01, 8.35] 7.65 [6.46, 9.03] < 0.001 Albumin, g/dL, median [IQR] 4.51 [4.33, 4.70] 4.58 [4.40, 4.76] 4.42 [4.24, 4.60] < 0.001 4.56 [4.38, 4.75] 4.46 [4.28, 4.65] < 0.001 Creatinine, mg/dL, median [IQR] 0.80 [0.69, 0.93] 0.79 [0.69, 0.91] 0.82 [0.70, 0.97] < 0.001 0.79 [0.68, 0.90] 0.82 [0.70, 0.96] < 0.001 CRP, mg/dL, median [IQR] 0.17 [0.08, 0.36] 0.14 [0.07, 0.27] 0.25 [0.12, 0.51] < 0.001 0.10 [0.06, 0.19] 0.30 [0.17, 0.55] < 0.001 ALP, U/L, median [IQR] 84.20 [70.10, 101.00] 80.90 [68.00, 96.10] 89.20 [74.30, 107.85] < 0.001 78.00 [65.90, 92.00] 91.30 [76.70, 109.90] < 0.001 [88]Open in a new tab Values are presented as median (IQR, interquartile range) or number (percentage). BMI, body mass index; BUN, blood urea nitrogen; CRP, C-reactive protein; FEV1, forced expiratory volume in one second; HbA1c, glycated hemoglobin; MCV, mean cell volume; RDW, red cell distribution; width; SBP, systolic blood pressure; TDI, Townsend deprivation index; WBC, white blood cell *Ideal physical activity: ≥150 min/week moderate or ≥ 75 min/week vigorous or 150 min/week mixed (moderate + vigorous) activity Associations between biological accelerated aging and incident CVDs During a median follow-up of 13.1 years (IQR: 9.94-14.0), we documented 2,383 incident cases of CHD, followed by 1,487 AF, 1,049 HF, 668 any stroke, and 642 VHD. Individuals with any incident CVD demonstrated greater age acceleration (Supplementary Figures [89]S2 and [90]S3). For continuous measures, each 1-SD increment in PhenoAge acceleration was associated with significantly higher risks across all cardiovascular endpoints: CHD (HR: 1.19, 95% CI 1.14–1.23), AF (HR: 1.17, 95% CI 1.11–1.24), HF (HR: 1.40, 95% CI 1.32–1.48), stroke (HR: 1.28, 95% CI 1.19–1.37), and VHD (HR: 1.21, 95% CI 1.12–1.31). KDMAge acceleration demonstrated comparable or stronger associations, with HRs ranging from 1.20 (95% CI 1.15–1.25) for AF to 1.31 (95% CI 1.27–1.36) for HF. When comparing extreme quartiles, participants in the highest PhenoAge quartile (Q4) exhibited substantially elevated risks compared to those in the lowest quartile (Q1): CHD (HR: 1.43, 95% CI 1.27–1.62), AF (HR: 1.43, 95% CI 1.22–1.67), HF (HR: 2.20, 95% CI 1.80–2.69), stroke (HR: 1.92, 95% CI 1.51–2.44), and VHD (HR: 1.46, 95% CI 1.14–1.85). KDMAge demonstrated even more pronounced effects (Table [91]2). Both measurements showed significant associations primarily with ischemic stroke and MVR (Supplementary Tables [92]S9 and [93]S10). Table 2. Associations between biological aging and risk of incident cardiovascular diseases in UK biobank Outcomes PhenoAge KDMAge Cases per 1000 py Cases/N HR (95% CI) P value PAR Cases per 1000 py Cases/N HR (95% CI) P value PAR CHD Continuous per 1SD 15.49 2383/12,828 1.19 (1.14–1.23) < 0.001 15.49 2383/12,828 1.23 (1.19–1.26) < 0.001 Q1 11.47 458/3207 Ref 12.95 509/3207 Ref Q2 14.18 553/3207 1.09 (0.96–1.24) 0.174 12.56 496/3207 1.08 (0.95–1.23) 0.235 Q3 16.31 624/3207 1.19 (1.05–1.34) 0.007 15 576/3207 1.24 (1.09–1.4) 0.001 Q4 20.42 748/3207 1.43 (1.27–1.62) < 0.001 21.88 802/3207 1.62 (1.44–1.82) < 0.001 Non-accelerated aging 13.26 1221/7517 Ref 12.78 1035/6595 Ref Accelerated aging 18.83 1162/5311 1.24 (1.14–1.35) < 0.001 4.6% (2.9–6.3%) 18.5 1348/6233 1.37 (1.26–1.49) < 0.001 9.5% (7.1–11.9%) AF Continuous per 1SD 9.31 1487/12,828 1.17 (1.11–1.24) < 0.001 9.31 1487/12,828 1.2 (1.15–1.25) < 0.001 Q1 6.41 264/3207 Ref 7.82 317/3207 Ref Q2 8.69 351/3207 1.09 (0.93–1.28) 0.31 6.86 281/3207 0.93 (0.79–1.1) 0.388 Q3 10.31 409/3207 1.17 (1–1.37) 0.058 9.64 384/3207 1.2 (1.03–1.41) 0.022 Q4 12.07 463/3207 1.43 (1.22–1.67) < 0.001 13.17 505/3207 1.44 (1.24–1.67) < 0.001 Non-accelerated aging 7.88 751/7517 Ref 7.28 610/6595 Ref Accelerated aging 11.44 736/5311 1.23 (1.10–1.37) < 0.001 4.2% (2.1–6.2%) 11.57 877/6233 1.4 (1.26–1.56) < 0.001 10.8% (7.5–14.1%) HF Continuous per 1SD 6.43 1049/12,828 1.4 (1.32–1.48) < 0.001 6.43 1049/12,828 1.31 (1.27–1.36) < 0.001 Q1 3.33 140/3207 Ref 4.09 170/3207 Ref Q2 5.12 212/3207 1.24 (1–1.53) 0.053 4.16 173/3207 1.03 (0.83–1.28) 0.802 Q3 7.2 292/3207 1.53 (1.25–1.88) < 0.001 6.45 263/3207 1.44 (1.18–1.77) < 0.001 Q4 10.38 405/3207 2.2 (1.8–2.69) < 0.001 11.32 443/3207 2.26 (1.88–2.72) < 0.001 Non-accelerated aging 4.56 444/7517 Ref 4.2 359/6595 Ref Accelerated aging 9.22 605/5311 1.62 (1.43–1.84) < 0.001 20.7% (15.8–25.6%) 8.9 690/6233 1.80 (1.57–2.05) < 0.001 29.1% (23.5–34.5%) Any stroke Continuous per 1SD 4.08 668/12,828 1.28 (1.19–1.37) < 0.001 4.08 668/12,828 1.22 (1.15–1.29) < 0.001 Q1 2.62 110/3207 Ref 3.69 153/3207 Ref Q2 3.79 157/3207 1.3 (1.01–1.66) 0.038 2.73 114/3207 0.77 (0.6–0.99) 0.045 Q3 4.18 171/3207 1.36 (1.06–1.74) 0.014 3.98 163/3207 1.06 (0.84–1.34) 0.618 Q4 5.85 230/3207 1.92 (1.51–2.44) < 0.001 6.01 238/3207 1.52 (1.22–1.88) < 0.001 Non-accelerated aging 3.42 334/7517 Ref 3.17 271/6595 Ref Accelerated aging 5.05 334/5311 1.33 (1.13–1.55) 0.001 7.6% (3.5–11.7%) 5.07 397/6233 1.51 (1.29–1.78) < 0.001 15.7% (10.0–21.2%) VHD Continuous per 1SD 3.7 642/12,828 1.21 (1.12–1.31) < 0.001 3.7 642/12,828 1.28 (1.21–1.34) < 0.001 Q1 2.57 112/3207 Ref 3.16 137/3207 Ref Q2 3.29 143/3207 1.07 (0.84–1.37) 0.588 2.41 105/3207 0.78 (0.6–1.02) 0.069 Q3 4.04 175/3207 1.19 (0.94–1.52) 0.150 3.54 154/3207 1.09 (0.85–1.38) 0.511 Q4 4.9 212/3207 1.46 (1.14–1.85) 0.002 5.72 246/3207 1.64 (1.31–2.04) < 0.001 Non-accelerated aging 3.03 309/7517 Ref 2.77 248/6595 Ref Accelerated aging 4.65 333/5311 1.30 (1.11–1.53) 0.001 6.7% (2.7–10.6%) 4.69 394/6233 1.56 (1.32–1.84) < 0.001 17.9% (11.8–23.9%) [94]Open in a new tab All models were adjusted for age, sex, ethnicity, TDI, education, household income, BMI, medication use (lipid-lowering, antihypertensive, and anti-diabetic medication), comorbidities (hypertension, hyperlipidemia, and cancer), types of diabetes, diabetes duration, smoking status, ideal physical activity, alcohol intake, and diet score. AF, atrial fibrillation; BMI, body mass index; CHD, coronary heart disease; HF, heart failure; PAR, population attributable risk; py, person-year; SD, standard deviation; TDI, Townsend deprivation index; VHD, valvular heart disease The magnitude of these associations was reflected in substantial PAR, ranging from 4.6% (95% CI 2.9–6.3%) for CHD to 20.7% (95% CI 15.8–25.6%) for HF with PhenoAge, and higher for KDMAge, spanning from 9.5% (95% CI 7.1–11.9%) to 29.1% (95% CI 23.5–34.5%). RCS revealed that KDMAge exhibited notable non-linear relationships with all cardiovascular outcomes (P for non-linearity < 0.001), while PhenoAge demonstrated predominantly linear associations (Supplementary Figure [95]S4). KDMAge showed a stronger association with HF in insulin-dependent individuals (P for interaction < 0.001). The association between PhenoAge and CHD was modified by CHIP, where individuals harboring CHIP had significantly higher risk of incident CHD in the presence of biological aging (P for interaction = 0.011; Supplementary Table [96]S11). Associations remained robust in multiple sensitivity analyses (Supplementary Tables [97]S12–[98]S14). Joint effect of CHIP and accelerated aging Significant positive additive interactions between CHIP and PhenoAge accelerated aging was observed, but not with KDMAge (Fig. [99]1, Supplementary Figures [100]S5 and [101]S6). Compared to participants with neither condition, individuals with both accelerated PhenoAge accelerated aging and CHIP mutations demonstrated substantially elevated risks of CHD (HR: 1.80, 95% CI 1.43–2.28), HF (HR: 2.65, 95% CI 1.94–3.64), and VHD (HR: 2.42, 95% CI 1.66–3.53). Significant synergistic effects were confirmed by RERI (CHD: 0.682, 95% CI 0.188–1.176; HF: 0.956, 95% CI 0.022–1.89; VHD: 1.047, 95% CI 0.006–2.089) and AP (CHD: 37.8%, 95% CI 16.7–58.9%; HF: 36.0%, 95% CI 9.1–63.0%; VHD: 43.3%, 95% CI 11.4–75.1%). Fig. 1. [102]Fig. 1 [103]Open in a new tab The joint association between CHIP and PhenoAge accelerated aging in the risk of incident cardiovascular diseases. All models were adjusted for age, sex, ethnicity, TDI, education, household income, BMI, medication use (lipid-lowering, antihypertensive, and anti-diabetic medication), comorbidities (hypertension, hyperlipidemia, and cancer), types of diabetes, diabetes duration, smoking status, ideal physical activity, alcohol intake, and diet score. AF, atrial fibrillation; AP, Attributable proportion; BMI, body mass index; CHD, coronary heart disease; CHIP, clonal hematopoiesis of indeterminate potential; HF, heart failure; RERI, Relative excess risk due to interaction; py, person-year; TDI, Townsend deprivation index; VHD, valvular heart disease Dynamic progression of diabetic individuals with accelerated aging Among 12,828 participants, 3,614 (28.2%) developed incident non-fatal CVDs, and 2065 experienced deaths, comprising 1077 deaths without prior CVD (8.4% of total cohort) and 988 deaths following CVD diagnosis (27.3% of those with CVD) (Fig. [104]2A). Both PhenoAge and KDMAge accelerated aging demonstrated significant associations with initial CVD development (PhenoAge HR: 1.27, 95% CI 1.19–1.35; KDMAge HR: 1.31, 95% CI 1.24–1.38), progression from CVD to death (PhenoAge HR: 1.58, 95% CI 1.39–1.79; KDMAge HR: 1.36, 95% CI 1.21–1.53), and direct death without CVD (PhenoAge HR: 1.43, 95% CI 1.28–1.62; KDMAge HR: 1.35, 95% CI 1.22–1.51) (Fig. [105]2B). Fig. 2. [106]Fig. 2 [107]Open in a new tab Multi-state trajectory and life expectancy loss in individuals with accelerated aging. A Numbers (percentages) of participants and median follow-up period in each status and three transitions from baseline to cardiovascular events, and then to death. B Hazard ratios (95% CIs) for transitions of three status based on PhenoAge and KDMAge acceleration. C Years of life loss in participants with PhenoAge/KDMAge accelerated aging compared with those without accelerated aging. Years of loss in upper three quantile by PhenoAge D and KDMAge E acceleration compared with the lowest quartile among people with diabetes. All models were adjusted for age, sex, ethnicity, TDI, education, household income, BMI, medication use (lipid-lowering, antihypertensive, and anti-diabetic medication), comorbidities (hypertension, hyperlipidemia, and cancer), types of diabetes, diabetes duration, smoking status, ideal physical activity, alcohol intake, and diet score. AF, atrial fibrillation; BMI, body mass index; CHD, coronary heart disease; CVD, cardiovascular disease; HF, heart failure; TDI, Townsend deprivation index; VHD, valvular heart disease. Life expectancy analysis quantified the substantial longevity impact associated with accelerated aging (Fig. [108]2C). At age 45 years, individuals with accelerated aging exhibited reduced life expectancy of 2.58 years (95% CI 0.19–4.97) and 2.17 years (95% CI 0.53–3.81) based on PhenoAge and KDMAge, respectively, compared to non-accelerated counterparts. This gap persisted across age groups but diminished with advancing age. A dose-response relationship was observed, with the highest acceleration quartile experiencing the greatest life expectancy reduction (Fig. [109]2D and E, Supplementary Table [110]S15). These findings remained robust in multiple sensitivity analyses (Supplementary Figures [111]S7 and [112]S8; Tables [113]S16 and [114]S17). Potential pathway mediators in the association between accelerated aging and incident CVDs To investigate potential molecular mechanisms underlying accelerated biological aging and cardiovascular outcomes, we performed comprehensive proteomic analyses. We first identified 711 and 1,163 DEPs associated with PhenoAge (664 upregulated, 47 downregulated) and KDMAge accelerated aging (1,156 upregulated, 7 downregulated), respectively (Supplementary Figure [115]S9). Pathway enrichment analysis of upregulated proteins revealed 57 and 75 significantly enriched pathways for PhenoAge and KDMAge accelerated aging, respectively. Top enriched pathways included neutrophil degranulation, immunoregulatory interactions between lymphoid and non-lymphoid cells, and extracellular matrix organization (Fig. [116]3A and B). Fig. 3. [117]Fig. 3 [118]Open in a new tab Reactome pathway enrichment and causal mediation analysis. Bar plot showing the top 10 enriched pathways in up-regulated differentially expressed proteins associated with PhenoAge A and KDMAge B accelerated aging. C Summary of causal mediation of neutrophil degranulation in the association of PhenoAge and KDMAge accelerated aging. Models were adjusted for age, sex, ethnicity, TDI, education, household income, BMI, medication use (lipid-lowering, antihypertensive, and anti-diabetic medication), comorbidities (hypertension, hyperlipidemia, and cancer), types of diabetes, diabetes duration, smoking status, ideal physical activity, alcohol intake, and diet score. AF, atrial fibrillation; BMI, body mass index; CHD, coronary heart disease; HF, heart failure; PNDE, pure natural direct effect; TDI, Townsend deprivation index; TE, total effect; TNIE, total natural indirect effect; VHD, valvular heart disease. Causal mediation analyses revealed that for CHD, the neutrophil degranulation pathway mediated 53.3% (95% CI 16.2–90.4%, P = 0.005) of PhenoAge accelerated aging effects and 31.5% (95% CI 11.5–51.5%, P = 0.002) of KDMAge. For HF, this pathway accounted for 52.2% (95% CI 26.1–78.3%, P < 0.001) and 39.3% (95% CI 20.2–58.4%, P < 0.001) of the total effects for PhenoAge and KDMAge accelerated aging, respectively (Fig. [119]3C; Supplementary Tables [120]S18 and [121]S19). Discussion In this large-scale prospective study, we provided novel evidence that accelerated biological aging was significantly associated with incident CVDs in diabetic participants, resulting in a significant reduction of life expectancy. Notably, we identified a synergistic effect, whereby individuals with both accelerated biological aging and CHIP mutations exhibited substantially higher risks of CVDs, particularly CHD, HF, and VHD. Through comprehensive pathway and causal mediation analyses, we demonstrated that inflammatory pathways—especially neutrophil degranulation—substantially mediate the associations between accelerated biological aging and adverse cardiovascular outcomes. These findings suggest that screening for accelerated biological aging could enable earlier implementation of targeted CVD prevention strategies, potentially reducing the substantial burden of adverse outcomes in the growing diabetic population. Aging is a major risk factor for CVDs; however, individuals often age at different pace, giving rise to the concept of biological aging [[122]5]. Previous research has demonstrated that accelerated biological aging increased the risk of cardiometabolic multimorbidity development and associated mortality in the general population [[123]30, [124]31]. However, the relationship between biological aging and adverse cardiovascular events in diabetic populations remains unclear [[125]10]. Pan et al. utilized a multi-state model to illustrate the significant association between accelerated aging and the risk progression from baseline to T2D onset, macrovascular or microvascular complications, and mortality [[126]10]. CVDs remain the leading cause of death in diabetic individuals [[127]2]. To the best of our knowledge, our study represents the first comprehensive investigation demonstrating that accelerated biological aging is associated with increased risks of CHD, AF, HF, stroke, and VHD in diabetic individuals. Notably, HF showed the largest effect size among all outcomes examined, suggesting that aging may have particularly detrimental effects on cardiac structure and function [[128]32]. We observed that the risk effects of accelerated biological aging on CVDs in diabetic populations exceeded those reported in general populations [[129]30]. It was hypothesized that in relatively unhealthy populations, accelerated aging may function as a “second hit,” amplifying its detrimental impact on physiological systems already compromised by diabetes-related pathophysiology. Telomere attrition and epigenetic alterations, such as DNA methylation and histone modifications, represent other biological aging metrics [[130]33]. However, these indicators often underestimate the complexity of aging and cannot accurately assess biological aging systematically [[131]5, [132]33]. PhenoAge and KDMAge, as composite measures of multiple blood biomarkers, have significant associations with CVDs independent of chronological age, demonstrating their ability to capture physiologically relevant aging processes and making them useful tools for cardiovascular risk stratification in diabetic populations [[133]8, [134]9]. Our study further provided novel insights into the comprehensive impact of biological aging on the entire life course of diabetic participants. Moreover, biological aging confers substantially greater mortality risk following CVD events compared to its effect on initial CVD development, further supporting the “second hit” hypothesis we discussed earlier, which was consistent with the previous findings [[135]10]. Quantitively, individuals with accelerated biological aging face a loss of life expectancy exceeding 2 years at age 45. Previous studies have also shown that accelerated biological aging increases mortality risk in both the general population and patients with rheumatoid arthritis [[136]9, [137]25]. Additionally, other biological age-related indicators also support the direct impact of biological aging on lifespan. For instance, telomere shortening has been associated with mortality and loss of life expectancy in the general population [[138]34], while biological age estimation through DNA methylation has been linked to long-term mortality in patients with type 2 diabetes [[139]35]. Unlike non-modifiable aging markers such as telomere length, PhenoAge and KDMAge include clinically actionable components like inflammatory status, blood pressure, and cholesterol levels. Interventions targeting these components can improve biological age scores, offering diabetic individuals tangible evidence of treatment benefits and potentially enhancing therapeutic engagement. Most importantly, our study provides refined risk stratification for diabetic participants based on biological aging and CHIP, offering a novel screening tool for identifying high-risk individuals. CHIP represents somatic mutations in hematopoietic stem and progenitor cells that contribute to various CVDs through sustained inflammatory responses [[140]14, [141]36–[142]38]. We noted that the joint effect between biological aging and CHIP was observed only with PhenoAge but not KDMAge. Genome-wide association study and the composition of biological aging metrics suggest that PhenoAge and KDMAge are associated with inflammatory and metabolic levels, respectively [[143]13]. Aging not only enhances inflammatory responses, contributing to hematopoietic cell inflamm-aging and promoting CHIP development, but also compromises intestinal barrier integrity, facilitating the release of microbial metabolites such as ADP-heptose, which subsequently accelerates CHIP progression [[144]39, [145]40]. Thus, the dual inflammatory hits from both PhenoAge and CHIP may partially explain the additive effects on long-term cardiovascular event risk in diabetic individuals. Intervention targeting the ADP-heptose-ALPK1 axis may represent a promising therapeutic target to mitigate the aging-induced exacerbation of CHIP detrimental effects [[146]40]. Interestingly, CHIP was not associated with long-term CVD in individuals without PhenoAge biological aging, which extends previous research conclusions regarding CHIP in diabetic populations [[147]14]. Multiple mechanisms mediate the relationship between aging and CVD development, including oxidative stress, persistent inflammation, and mitochondrial dysfunction [[148]41]. Our findings suggested that individuals experiencing accelerated biological aging had significantly enhanced immune responses and extracellular matrix production pathways, particularly neutrophil degranulation. With advancing age, neutrophils exhibit diminished phagocytic capacity while potentially demonstrating enhanced degranulation responses, leading to excessive inflammation and tissue damage. Additionally, reduced formation of neutrophil extracellular traps (NETs) coupled with impaired degradation results in NETs accumulation, perpetuating chronic inflammation [[149]42]. The causal mediation analysis indicated that neutrophil degranulation played a crucial mediating role in both CHD and HF, consistent with findings in the general population [[150]28, [151]43]. Myeloperoxidase inhibitors (primarily targeting an enzyme released by neutrophils) may show promise in alleviating aging-associated CVD risks [[152]44]. However, proteomic data did not identify mediating pathways associated with AF, stroke, or VHD. This may be partly attributable to the UK Biobank’s restricted protein panel of only 2,923 proteins, constraining our exploration of additional pathways. Furthermore, the microenvironmental aging phenotypes specific to atrial tissue, valve structures, and cerebral tissue may play more critical roles in the pathogenesis of these conditions [[153]45–[154]47]. Beyond molecular pathways, lack of physical activity or frailty status in aging populations may also serve as important mediators in CVD development [[155]48, [156]49]. SGLT2 inhibitors have demonstrated favorable effects in frail elderly patients with diabetes [[157]50]. The anti-inflammatory properties of SGLT2 inhibitors may confer potential beneficial effects in populations with biological aging, warranting further clinical investigation [[158]51]. Several limitations should be acknowledged in our study. First, the observational nature of our study design hinders the establishment of causal relationships. Second, potential unmeasured confounders might not have been included in our analyses. Third, as the UK Biobank collected only baseline characteristics and lacks longitudinal, repeated-measure data, it is difficult for us to account for changes in biological aging over time. This limitation prevents us from assessing individual trajectories of biological age changes and their associations with long-term prognosis. Finally, the generalizability of our findings is limited by the predominantly European ancestry of the UK Biobank population, making it challenging to extrapolate these results to other ethnic groups. In addition, as a volunteer cohort, participants in the UK Biobank tend to be healthier and wealthier [[159]52]. This may introduce selection bias, potentially leading to conservative estimates of the association between biological aging and adverse outcomes in diabetic individuals. Conclusions In conclusion, our study demonstrates significant associations between accelerated biological aging and increased risks of CVDs and shortened life expectancy in diabetic individuals, with neutrophil degranulation playing a crucial mediating role in this pathophysiological process. In the context of the increasing global prevalence of diabetes and population aging, prioritizing the prevention of cardiovascular events among diabetic individuals represents an urgent public health imperative. Developing targeted interventions to slow biological aging processes, particularly those affecting inflammatory pathways, may represent a promising strategy for reducing the substantial cardiovascular burden associated with diabetes. Electronic supplementary material Below is the link to the electronic supplementary material. [160]Supplementary Material 1^ (19.7MB, docx) Acknowledgements