Identification of new genetic variants of HLA-DQB1 associated with human longevity and lipid homeostasis—a cross-sectional study in a Chinese population

Baseline information in longevity and control subjects

There were significant differences in HDL-c (P=0.003), LDL-c (P=0.010), TG (P=3.052×10^-4) and TC (P=0.012) between centenarians and controls. We detected that HDL-c (P=1.461×10^-7), LDL-c (P=2.128×10^-4) and TC (P=8.181×10^-6) were significantly different between nonagenarians and controls. A comparison of longevity (centenarians and nonagenarians) in samples and controls, revealed that significant differences between HDL-c (P=3.438×10^-9), LDL-c (P=1.269×10^-5), TG (P=0.035) and TC (P=6.522×10^-7) had been found. (Supplementary Table 1)

Identifying the new longevity-associated genetic variations

Relationship of longevity-associated variants and plasma lipid homeostasis

For rs1049107, Supplementary Table 4 shows the proportion of participants with different plasma lipid levels, normal or abnormal, and LDL-c/HDL-c ratios ≤2 and those >2, according to the polymorphism genotypes. The rs1049107 T allele was associated with a lower LDL-c/HDL-c ratio (P=0.047, OR: 1.901) in the longevity group.

In HDL-c subgroups of longevity samples, the rs1049107 TT genotype samples and T allele carriers showed a better HDL-c level than did the rs1049107 CC genotype samples (P=0.007, OR: 10.636; P=0.006, OR: 11.277). The normal HDL-c subgroup contained a higher rs1049107 T allele proportion (P=1.458×10^-4, OR: 10.689), but this subgroup also had a lower rs3891176 C allele ratio than the major allele (P=0.013, OR: 0.360).

The combined rs1049107 genotype (TT+CT) in the control group, assuming a dominant effect of the T allele, was associated with a higher LDL-c level (P=1.243×10^-7, OR: 0.098). In longevity group rs1049107 CT and rs1049100 CT, the genotype was associated with a higher TG level (P=9.095×10^-7, OR: 0.025; P=1.799×10^-6, 0.028), but it was opposite in the controls (P=0.004, OR: 11.667; P=0.004, OR: 11.667) compared with major homozygotes.

Overall, the dominant effect of the rs1049107 T allele was associated with a normal HDL-c level (P=0.001, OR: 72.757,95%CI: 4.378-984.360) and a higher TG level (P=0.035, OR: 0.312, 95%CI: 0.105-0.924) in the longevity group and with a higher LDL-c level in controls (P=0.050, OR: 0.383, 95%CI: 0.147-0.999) by a logistic regression analysis.

For the rs3891176 CA genotype, the dominant effect of the variant C allele and rs3891176 C allele were associated with a lower LDL-c/HDL-c ratio (Supplementary Table 3) by a logistic regression analysis, and the dominant effect of the rs3891176 C allele was associated with a lower LDL-c/HDL-c ratio (P=0.034, OR: 2.402, 95%CI: 1.068-5.402). Furthermore, in the longevity group, we detected that major homozygotes of rs3891176 had significantly higher HDL-c levels compared to minor allele carriers (P=0.044) (Table 2).

Analysis of plasma-lipid homeostasis and longevity-associated haplotypes

To analyse the effect of haplotypes on plasma lipids, we divided samples between plasma lipid levels and age. We found that rs41542812, rs1049107 and rs1049100 would form a block in HLA-DQB1 and that the CCC haplotype, which was formed by major alleles, seemed to be a reference. In the centenarian group, we obtained none of the significant haplotypes for four plasma lipid indices. In nonagenarians, the block contained three haplotypes, and the CTT haplotype was associated with a higher LDL-c/HDL-c ratio (P=0.037, OR: 0.301, 95%CI: 0.093-0.969). In the longevity group, the GTT and CTT haplotypes were associated with higher HDL-c levels (P=0.027, OR:4.750, 95%CI: 1.057-21.337; P=0.045, OR:6.667, 95%CI: 0.857-51.883). The CTT haplotype was associated with a higher TG level (P=1.096×10^-4=, OR: 0.178, 95%CI: 0.070-0.451), and the CTT haplotype was associated with LDL-c/HDL-c <=2 (P=0.027, OR: 3.403, 95%CI: 1.098-10.544; Supplementary Table 5).

Interaction between longevity-associated variants and internal milieu lipid levels

An interaction analysis between longevity associated variants and internal milieu lipid levels on age detected significant results in a corrected model (R²=0.375, R²_Adjusted=0.332). Significant values of rs14917 & rs1491399 (P=0.001), rs1491 & LDL-c/HDL-c (P=0.016), rs927299 & LDL-c/HDL-c (P=0.028), and rs1491 & rs927299 & LDL-c/HDL-c (P=0.031) were detected in this model.

Longevity-associated variants and haplotypes identified

Because of the low prevalence of delayed age-related disease, the longevity population was an ideal population to be considered a “real negative control” in human successful healthy ageing and age-related disease. Genetic research about longevity had identify some longevity-associated factors, such as FOXO3A [15,16], SIRT1 [17], APOE [18,19], IL6 and ANKRD20A9P [20]. There had been multiple genetic studies of Chinese longevity, and FOXO3 [21], IGFBP-3 [22], CETP [23], and SIRT1 [24]. Studies in Sardinian centenarians showed the association between HLA-DQB1 and longevity [25,26]. The association between longevity and HLA-DQB1*05 or HLA-DQB1*03 was identified in a Japanese population [7,8]. However, there was no association study about SNPs in HLA-DQB1 and longevity. In this study, we identified four new SNPs in HLA-DQB1, rs41542812 (DQB1*05), rs1049107 (DQB1*03), rs1049100 (DQB1*03) and rs3891176 (DQB1*02), associated with longevity in our cohort (Table 1). Further, we identified that rs41542812, rs1049107, and rs1049100 were closely linked to disequilibrium in one block; that is, the haplotypes GTT and CTT, which significantly increased the chance of longevity (P=3.996×10⁹, OR: 4.367, 95%CI: 2.608-7.313; P=1.812×10⁵, OR: 3.677, 95%CI: 1.970-6.865) to our knowledge have also never been reported before.

In these new longevity-associated variants, we noted through the use of stratified analysis that only rs1049107 was a centenarian-associated variant in long-lived individuals, but the other three variants were associated with both centenarians and nonagenarians (Table 1). This result suggested that the longevity-associated genetic variance seems to increase with ageing, so we observed the variance only in centenarians but not nonagenarians.

Longevity variants associated with lipid homeostasis in LLIs

The main cholesterol-carrying lipoproteins are LDL-c and HDL-c. LDL-c is always considered the bad lipoprotein and can increase CVD and metabolic syndrome risk. In contrast, HDL-c is considered the good lipoprotein and can decrease metabolic-related disease and CVD risk. The importance of the balance of plasma lipids for maintaining health is obvious. The LDL-c/HDL-c ratio is increasingly being considered a valuable tool in CVD-related clinical research. Many variants have been identified as potentially affecting human lipids, as have some longevity-associated variants. The changes in lipid levels in our cohort were consistent with those in previous studies [27–29]. However, evidence for the balance of plasma lipids being controlled by longevity-associated loci is still lacking.

Furthermore, we identified HLA-DQB1 rs1049107, T-carriers (P_HDL=0.006, OR: 11.277; P_TG=9.095×10^-7, OR: 0.025; P_LDL/HDL=0.047, OR: 1.901) and HLA-DQB1 rs1049100, T-carriers (P_TG=1.799×10^-6, OR: 0.028) associated with lipid homeostasis in LLIs.

The interaction analysis of longevity variants and lipid levels

The interaction analysis between lipids and genotype results show that HLA-DQB1 longevity variant alleles or haplotype carriers have a trend of better blood lipid levels, thus benefiting healthy longevity in LLIs (Table 3; Supplementary Figure 1). The interaction analysis results suggest that with longevity-associated variants enriched in the longevity population, the plasma lipid levels were better controlled than by adjusting age. They also suggested that it is possible for longevity variants to possess functions to regulate the homeostasis of plasma lipids in the internal environment of LLIs.

Functional analysis of the new longevity-associated variants in HLA-DQB1

HLA-DQB1 contained 5 exons. Exon 1 encodes the leader peptide, exons 2 and 3 encode the two extracellular protein domains, exon 4 encodes the transmembrane domain, and exon 5 encodes the cytoplasmic tail. All four novel variants in HLA-DQB1 are functional missense mutations. I These four variants, rs41542812 (Gln158His), rs1049107 (Gly157Ser) and rs1049100 (Val148Ile), lays in β chain of HLA-DQB1. Variety in β chain may change the binding site for special antigen, which plays a key role in maintaining internal homeostasis in vivo, such as lipid homeostasis. In addition, rs3891176 (Ala29Ser), located in exon 1, may change the function of the leader peptide and transform the signal that HLA-DQB1 received (Figure 3). Therefore, we first identified the association between longevity and these four variants. oxLDL (oxidized low-density lipoprotein) or LPS bands on HLA-DQB1 of T cell, and stimulate T cell producing cytokines. Cytokines effect on liver cell, and increase or decrease fatty acid metabolism. OxLDL could also bands on LDLR (low-density lipoprotein receptor) or HLA-DQB1 of liver cell. The signal could also change fatty acid metabolism and HDL-c producing of liver cell. With this possible process, plasma lipid homeostasis could affect by HLA-DQB1 variants (Figure 3).

We speculated that the HLA-DQB1 variants decreased the expression and/or antigen binding function of the HLA Class II antigen DQ protein β chain on the Antigen Presenting Cell (i.e., the macrophage), reduces the cytokines released by T-lymphocytes, thus reducing liver cell synthesis and release of cholesterol, to maintain the balance of cholesterol metabolism in vivo. This theory remains to be studied more in depth to obtain the corresponding evidence.

Findings in this research suggest a possible gene effecting model. HLA-DQB1 affects the longevity through a certain biological pathway, and these pathways could influence the balance of plasma lipids through unknown biological process. Longevity and the balance of plasma lipids could affect each other. A longevity gene could help keep plasma lipids balanced, and a better balance of plasma lipids could promote homeostasis and healthy longevity. However, the mechanisms of longevity and plasma lipid levels are still not clear. Furthermore, a gene associated with longevity and plasma lipid levels could be identified in multiple large longevity populations, and rare or low-frequency variants must receive more attention. The mechanism could be studied in cell or animal models by using new technology, such as Crispr/Cas 9. It would be useful for us to understand the mechanism of longevity and help the wider community achieve healthy ageing.

Subjects

The current study was based on the Longevity and Health of Aging Population (LHAP) study conducted in Bama County, Guangxi, China, in 2008 and the Chinese Longitudinal Healthy Longevity Survey (CLHLS) during 1997-2015 (Supplementary Table 6).

To investigate the association between genotype and plasma lipid levels, there were 100 long-lived individuals (aged 90-107 years, 32 males, 68 females), and 172 controls (aged 40-65 years, 75 males, 97 females). The longevity group contained 26 centenarians (5 males, 21 females) and 74 nonagenarians (27 males, 47 females). All control subjects lacked longevity history (no lineal family members aged above 85 for three generations). The Ethics Committee of Beijing Hospital, Ministry of Health, approved the study protocol. All participants were informed and provided informed consent in writing. All clinical investigations were conducted according to the principles of the Declarations of Helsinki. Laboratory parameters were recorded, including total cholesterol (TC), triglyceride (TG), high density lipoprotein-cholesterol (HDL-c) and low-density lipoprotein-cholesterol (LDL-c).

Exome target sequencing

We sequenced the segment of exomes using a targeted NGS approach to analyse SNVs (single nucleotide variations) and small insertions and/or deletions (Indels). The design parameters were as follows: 1) the bait length was 120 bp; 2) the bait tiling frequency was 2×; and 3) the human reference genome (GRch37) was used to search the baits. Genomic DNA (3 μg) was used to construct DNA libraries that contained index sequences for identifying DNA samples. The targeted genomic segment was captured with the Agilent Capture kit 39M (Agilent Technologies, Santa Clara, CA, USA) by NGS on the Illumina HiSeq 2500 platform to identify mutations in 100 Bama longevity subjects using 100-bp paired-end reads. Base sequences were aligned to the reference genome (GRch37), and variations SNVs and Indels) were called using the Samtools software and annotated by comparing them with the 1000 Genomes database. The average coverage of each base in the targeted region was 50×, which resulted in ≥90% of the targeted bases being sufficiently covered for variant calling (≥10×). The average coverage was ≥95% at the NGS stage. We compared the data from our 100 LLIs to publicly available genotype data from a CHS (Chinese Han South) population (n = 97) from the 1000 Genome project phase I. We found 115327 SNVs, in total. Of the 115327 SNVs, 84914 were consistent with those in the public SNP database, and 30413 were previously unknown.

Selection of variations

We selected candidate variation from the target sequence dataset using four continuous steps.

1) To minimize error rates and the impact of random sampling variance, at least three reads had to be obtained from both the forward and reverse DNA strands (double-stranded coverage) for all variants that were kept for association tests.

2) Variants located at repeat sequences, including STR (short tandem repeat) or single nucleotide repeat expansions, may be falsely called; thus, we first excluded these possible artificial variants.

3) The Longevity Gene database was downloaded, and all significant genes were screened. All variants of these significant genes were called and compared with the CHS population in the 1000 Genome project to obtain candidate variants with MAF (minor allele frequency) >10% in the 100 Bama longevity subjects compared to those in one public database. Seventeen genes that included 26 variants were selected as candidate longevity-associated genes. Through genotyping based on longevity and local younger populations, four genetic variants in HLA-DQB1 were identified as longevity-associated gene variants (Supplementary Table 1).

Genotyping and quality control for genotyping

The Sanger Sequencing method was used in the genotyping case-control study. To perform genotyping quality control, all associated allele carriers, and 10% of cases and 10% of controls who carried non-risk alleles, were re-genotyped by Sanger sequencing. Sequencing primers are shown in Supplementary Table 7.

Statistical analysis

Genotypes were evaluated for departure from the Hardy–Weinberg equilibrium (HWE) in cases and controls using the chi-squared goodness test. Variants with P < 0.05 were considered to deviate from the HWE. The MAFs of the variants were used as high-risk allele frequencies, and the defined type I error rate was 0.05. The genotype frequencies of the CHS population from the 1000 Genome database were used as the references for selecting candidate SNVs. A t-test was used to analyse the mean difference between the groups. A chi-squared test was used to establish the differences in the distribution of genotype and allele frequencies between the cases and controls. The odds ratio (OR) was used to estimate the strength of association between variables, with the OR and 95% confidence intervals (95%CI). The ORs and 95% CIs were calculated using the SPSS V17.0 software (SPSS; SAS Institute, Cary, NC, USA). A two-sided P value < 0.05 was considered statistically significant.