The cohort of this study included 484 individuals aged 65–107 years (mean age 83.68 ± 9.56 years), consisting of 312 females (mean age 84.07 ± 9.45) and 172 males (mean age 82.98 ± 9.77). The samples were obtained from elderly individuals residing in nursing homes across the Calabria region in southern Italy. This was part of a broader investigation into the quality of aging in the entire region. Subjects were eligible to participate in the study if they were of Calabrian ancestry. Socio-demographic information and health details, including medical history and medication use, were gathered through a standardized questionnaire, medical visits, and clinical exams. Fasting venous blood was collected from each participant for clinical testing, laboratory analysis, and DNA extraction.

The study conforms to the Declaration of Helsinki regarding research involving human subjects and the protocol was approved by the local Ethical Committee (Comitato Etico Regione Calabria-Sezione Area Nord) on 2017-10-31 (code n. 25/2017). Each subject signed an informed consent for the permission to collect blood samples and usage of register-based information for research purposes.

Trained nurses measured participants’ waist and hip circumferences, from which the waist-to-hip ratio (WHR) was calculated. Height and weight were recorded with participants wearing light clothing and no shoes, and body mass index (BMI) was calculated as weight (kg) divided by height squared (m²). Blood pressure (BP), including systolic (SBP) and diastolic values (DBP), was measured three times on the right arm using a mercury sphygmomanometer after a rest period of at least 5 min. The average of the three readings was used as the final BP value. Routine hematological and biochemical markers, reported in Table 1, were analysed at the Italian National Research Center on Aging (Cosenza) using standardized methods. The triglyceride-glucose (TyG) index was calculated as Ln [fasting triglycerides (mg/dL) × fasting plasma glucose (mg/dL)/2], following the method by Simental-Mendía et al. (2008). The glomerular filtration rate (GFR) was estimated using the creatinine-based Berlin Initiative Study 1 (BIS1) equation, specifically developed for older adults (Schaeffner et al., 2012). The formula used was: eGFR BIS1 = 3,736 × creatinine^−0.87 × age^−0.95 × 0.82 (if female).

Hypertension was defined as SBP/DBP ≥140/90 mmHg or use of antihypertensive medication), diabetes as fasting plasma glucose >125 mg/dL or use of antidiabetic therapy, and the presence of cardiovascular disease (CVD) based on a comprehensive review of medical history, clinical symptoms, imaging results, and physical or laboratory examinations, evaluated by a board-certified cardiologist in accordance with international clinical guidelines. Chronic kidney disease (CKD) was defined as an estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m² using the BIS1 equation.

Multimorbidity was assessed using the modified Cumulative Illness Rating Scale (CIRS), which evaluates the burden of chronic conditions across 14 physiological systems (cardiac, vascular, respiratory, gastrointestinal, hepatic, renal, genitourinary, musculoskeletal, dermatologic, neurologic, endocrine, metabolic, breast, and psychiatric). Each system was scored from 0 (no issue) to 4 (severe impairment), with total scores ranging from 0 to 56. Higher scores indicate greater overall morbidity.

Twelve SNPs encompassing the CHGA gene and its 5′and 3′flanking regions were genotyped in all subjects included in the study. The SNPs were selected from previous studies in which significant genetic associations with cardio-metabolic diseases and/or associated endophenotypes were reported.

Schematic representation of the human chromogranin A gene structure and its derived peptides, as well as the SNPs that were genotyped, are shown in Supplementary Figure S1.

DNA was extracted from whole blood according to standardized procedures. Multiplex SNP genotyping was carried out using PCR, followed by primer extension and MALDI-TOF mass spectrometry with iPLEX Gold technology from Sequenom (Sequenom Inc., San Diego, CA, United States). PCR and single base extension primers were designed using Sequenom MassARRAY Assay Designer software (version 3). PCR amplification was performed using standard methods, and unincorporated nucleotides were removed using shrimp alkaline phosphatase (SAP). A primer extension reaction was then conducted with a mass extension primer and terminator. The extension products were desalted on resin and spotted onto the 384-element SpectroCHIP (Sequenom) for MALDI-TOF analysis, utilizing SpectroACQUIRE v3.3.1.3 software (Sequenom). Data were analyzed with MassARRAY Typer v3.4 Software (Sequenom).

To ensure quality control, approximately 10% of the samples were re-genotyped to verify the accuracy of the genotype identification process, with duplicate concordance exceeding 99.8% for all SNPs. Additional quality control steps included excluding SNPs that were not in Hardy–Weinberg equilibrium in controls (p-value <0.05) or had low genotyping success (<90%).

CHGA gene polymorphisms significantly associated with the phenotypes measured in our study were a subject for comprehensive bioinformatics analysis. In particular, bioinformatics tools of the quantitative trait loci (QTL) databases such as the GTEx portal (https://gtexportal.org) (The GTEx Consortium, 2020), and QTLbase2 (http://mulinlab.tmu.edu.cn/qtlbase) (Huang et al., 2023), were used to evaluate whether the CHGA gene polymorphisms represent significant QTLs correlating with a variation of molecular traits such as mRNA expression (eQTL), methylation (mQTL), histone modification (hQTL) and splicing events (sQTL).

The GTEx portal serves as an extensive public resource for researching gene expression and regulation specific to different tissues. QTLbase2 is a database that gathers and compiles genome-wide QTL summary data for various human molecular traits across more than 70 different tissues and cell types. The functionality of these SNPs was further explored by FuncPred (https://snpinfo.niehs.nih.gov/snpinfo/snpfunc.html) (Xu and Taylor, 2009) and Regulome DB (http://regulomedb.org/) (Boyle et al., 2012), which annotate and prioritize potential regulatory variants in the human genome using a collection of functional information gathered from a variety of tools and resources.

Analyses were performed by dividing the sample into older adults (269 subjects, 65–85 years; median 76.77 ± 6.3) and long-lived (215 subjects, 86–107 years; 92.33 ± 4.6). Continuous variables are presented as means with standard deviations. To assess data distribution normality, the Kolmogorov-Smirnov test was employed. Where necessary, variables were log-transformed to meet normality assumptions before analysis. The independent-samples t-test or the Mann-Whitney U test, depending on the distribution characteristics, were used to compare differences between groups for continuous variables. Categorical variables are expressed as percentages and were compared using the chi-squared (χ2) test. Hardy-Weinberg equilibrium for each SNP was evaluated in controls using the χ2 test.

Logistic regression models were used to estimate the impact of genotypes on the chance to reach a very advanced age. Genetic data were coded with respect to a dominant, a recessive, and an additive model of inheritance. Then, for each SNP, the most likely genetic model was estimated based on minimum level of statistical significance (Wald test p-value). For SNPs with minor allele homozygotes <3%, either in the group older adults or long-lived group, only the dominant model was considered. In such models, sex was used as a covariate.

Multivariate linear and logistic regression models were used to calculate the association of genetic variants with continuous and dichotomized traits, respectively, adjusting for age, sex, and BMI. SNP × sex interaction terms were included in logistic regression models, adjusting for age. The false discovery rate (FDR) was calculated to correct for multiple comparisons of genotypes, and FDR <0.05 was considered to indicate a statistical significance for association (Benjamini and Hochberg, 1995). Complementary validation analyses included (i) the use of the TSI population from the 1000 Genomes Project (www.1000genomes.org) (n = 114; 59 males and 55 females), limited to allele frequencies and age-association given the absence of phenotypic data, and (ii) a stratified bootstrap approach (Bias-Corrected and Accelerated method, 5,000 resamples, stratified by sex) to internally assess the robustness of the observed associations.

Linkage disequilibrium (LD) and haplotype analysis were conducted using the SHEsis software (Shi and He, 2005).

Statistical analyses were performed using SPSS software (version 29.0, SPSS Inc., Chicago, IL, United States).

A total of 269 older adults (ages 65-85) and 215 oldest-old individuals (ages 86-107) participated in this study. Summary statistics for the anthropometric, biochemical, and clinical variables for both age groups are presented in Table 1. In brief, no significant differences were observed in sex distribution, with a sex ratio imbalance favouring females in both groups. As shown in Table 1, significant differences between the groups were observed in some anthropometric and quantitative metabolic variables, as well as in the presence of morbidities.

Twelve SNPs were selected for this study, and details of their gene position, functional annotation and minor allele frequency (MAF) observed in our samples and in the 1,000 Genomes TSI population (Tuscany, Italy) can be found in Supplementary Table S1. Two of them (rs9658630 and rs9658667) were excluded from the analysis because of low genotyping success rates (<90%), and one (rs9658664) because it deviated from HWE (P < 0.05). The remaining nine SNPs successfully passed the quality control procedures. The genotypic frequencies of these SNPs in each of the two age groups are shown in Figure 1.

Table 2 shows the results of the logistic regression analyses adjusting for sex on the probability of reaching advanced age. Five out of the nine polymorphisms (namely, rs9658628, rs9658631, rs9658634, rs7159323 and rs7610) exhibited a statistically significant association under a dominant genetic model. It emerged that the rs9658628-A allele was associated with a diminished probability of belonging to the oldest age group, with an adjusted odds ratio (OR) of 0.51 (95% CI: 0.35–0.75; p = 0.0005), signifying that the presence of the minor A allele is associated with diminished survival at advanced age. Also, individuals carrying the minor allele at rs9658631-(G), rs9658634-(A), rs7159323-(A), and rs7610-(T) had a higher chance of being in the oldest age group. The adjusted ORs were 1.86 (1.29–2.68), 1.79 (1.24–2.58), 1.88 (1.30–2.71), and 1.80 (1.24–2.62), with p-values of 0.0008, 0.0017, 0.0007, and 0.002, respectively. All the associations remained significant after adjusting for multiple testing with false discovery rate (FDR) q < 0.05. To strengthen the robustness of our results, we performed an additional sensitivity analysis using a Bias-Corrected and Accelerated (BCa) bootstrap analysis with 5,000 resamples, stratified by sex, which confirmed the stability of the observed SNP associations with longevity (results reported in Table 2). In a separate analysis, we tested for sex × SNP interactions (Table 2) and found no significant interactions, indicating that the associations were consistent across sexes.

In further support of these findings, the same associations were replicated using a cohort of adult Italian volunteers (Tuscans; TSI; 114 individuals, 59 males and 55 females), whose genotypes for the relevant SNPs were obtained from the 1000 Genomes Project and used as controls for comparison with long-lived individuals. The ORs were 0.58 (0.36–0.93), 1.92 (1.19–3.10), 1.86 (1.15–3.00), 1.85 (1.14–2.98), 2.10 (1.30–3.40), respectively for rs9658628, rs9658631, rs9658634, rs7159323 and rs7610. No association was observed when comparing our sample of older people with the group of Tuscans, suggesting that these SNPs may not differ significantly in the general older population.

For identifying independent variants in CHGA associated with advanced ages, the LD patterns (r2) of five significant SNPs were examined and depicted in Figure 2. A strong LD was detected among the variants rs9658631, rs9658634, and rs7159323 (r² > 0.9), suggesting that the associations observed at the individual SNPs are not independent and likely reflect a single association signal influencing longevity. The haplotype analysis for these polymorphic sites revealed that haplotype composed of the minor alleles of the three variants (G-A-A) was associated with a significantly increased chance of reaching old age (OR 1.67, CI: 1.24–2.25, p = 0.0006). Conversely, the haplotype containing the major alleles (T-G-C) appeared to have an opposite effect, with an odds ratio of 0.59 (CI: 0.44–0.80, p = 0.0006) (Table 3). Thus, haplotype analysis confirmed association results with the single markers.

Next, we aimed to investigate whether the five variants associated with advanced age were also associated with clinical parameters and health conditions that are established risk factors for age-related diseases that contribute to significant morbidity and mortality in the elderly. We focused on the age group 64–85 years and assessed the variability of each SNPs in relation to the parameters listed in Table 1. The significant results (p < 0.05), adjusted for age, sex, and BMI, are presented in Table 4. To assess the robustness of these findings, we also performed BCa bootstrap analyses and tested for sex × SNP interactions, which showed that the associations remained consistent and that no significant interactions were present (Table 4).

The table shows that the presence of the minor rs9658628-A allele, which was negatively correlated with survival after 85 years of age, was also associated with an increased triglyceride glucose (TyG) index (β = 0.104; p = 0.019), a well-established marker of insulin resistance. Additionally, individuals carrying the rs9658628-A allele had a significantly higher likelihood of having diabetes compared to non-diabetics (OR = 1.93, 95% CI: 1.02–3.66; p = 0.038). Notably, carriers of this allele also scored significantly lower on the Mini-Mental State Examination (MMSE) than non-carriers (β = −2.30; p = 0.006), indicating poorer cognitive performance. Conversely, the rs9658628-A allele appeared to confer a protective effect against cardiovascular disease (CVD), with carriers showing a significantly lower risk compared to non-carriers (OR = 0.46, 95% CI: 0.24–0.87; p = 0.017). The analysis also showed that carriers of the T minor allele of rs7610, who were more likely to live to an older age, had significantly higher diastolic blood pressure (DBP) compared with non-carriers (beta = 2.92; p = 0.019). There was also a marginal, but not significant (p = 0.057) trend towards higher systolic blood pressure (SBP).

Regarding the three SNPs in strong LD (rs9658631, rs9658634, rs7159323), no significant associations were observed with the clinical parameters tested. A marginal effect (p = 0.06) was observed towards an association with higher LDL-C levels.

However, the associations that reached nominal significance did not remain significant after FDR correction for multiple testing.

We conducted a comprehensive functional annotation of the associated SNPs using integrative multi-omics databases. Supplementary Table S2 provides a synthesis of annotations specifically related to the impact of these SNPs on the CHGA gene, while Supplementary Table S3 presents QTLbase-derived annotations referring also to effects on other genes.

Regulatory information was obtained through RegulomeDB and FuncPred. All variants were predicted to have regulatory functions and affect putative transcription factor-binding sites.

To assess transcriptional effects, we examined the expression quantitative trait loci (eQTLs) using the GTEx portal and QTLbase. According to GTEx data, the rs9658628-G allele significantly reduced CHGA mRNA levels in the cerebellum, while the rs7610-T allele was associated with increased CHGA expression in the pituitary. The rs7610-T allele also correlated with reduced ITPK1 expression in the aorta and tibial arteries, and lower LINCO2287 levels in the cerebellum. In contrast, no significant effect on CHGA expression across tissues was observed for the linked variants rs9658631, rs9658634, and rs7159323. However, their minor alleles were associated with increased RIN3 (rs9658631, rs9658634, rs7159323) and ITPK1 (rs7159323) expression in the aorta.

QTLbase further supported and expanded these findings (see Supplementary Table S2, 3). For rs9658628, additional effects were noted in brain tissues, including altered expression of CHGA in the prefrontal cortex and ITPK1 and FBLN5 in other brain regions. This variant also influenced the expression of several other genes across multiple tissues. Similarly, the rs7610 minor allele was linked to reduced CHGA expression in the hippocampus, lower FBLN5 levels in the nucleus accumbens, and altered ITPK1 and TMEM251 expression in vascular and adipose tissues, respectively. Linked variants (rs9658631, rs9658634, rs7159323) were reported as eQTLs for CHGA in the hippocampus, with minor alleles associated with lower expression.

In addition, methylation QTL (mQTL) analysis from QTLbase identified several methylation associations for CHGA in blood (rs9658268 and rs7610: 6 mQTLs; rs9658234: 23 mQTLs; rs7159323: 18 mQTLs) and brain tissue (rs965863: 2 mQTLs; rs9658234: 4 mQTLs; rs7159323: 2 mQTLs). These SNPs also acted as mQTLs for other genes, particularly ITPK1, in both blood and brain.