• Open Access

Common genetic variants of the β2-adrenergic receptor affect its translational efficiency and are associated with human longevity



Xiao-Li Tian, Department of Human Population Genetics, Institute of Molecular Medicine, Peking University, 5 Yiheyuan Road, Beijing 100871, China. Tel.: +86 10 6275 5397; fax: +86 10 6275 6926; e-mail: tianxiaoli@pku.edu.cn


β-adrenoceptors are the common pharmacological targets for the treatment of cardiovascular diseases and asthma. Genetic modifications of β-adrenergic system in engineered mice affect their lifespan. Here, we tested whether genes encoding for key components of the β-adrenergic signaling pathway are associated with human longevity. We performed a 10-year follow-up study of the Chinese longitudinal healthy longevity survey. The Han Chinese population in this study consisted of 963 long-lived and 1028 geography-matched young individuals. Sixteen SNPs from ADRB1, ADRB2, ADCY5, ADCY6, and MAPK1 were selected and genotyped. Two SNPs, rs1042718 (C/A) and rs1042719 (G/C), of ADRB2 in linkage disequilibrium (D' = 1.0; r2 = 0.67) were found to be associated with enhanced longevity in men in two geographically isolated populations. Bonferroni-corrected P-values in a combined analysis were 0.00053–0.010. Men with haplotype A-C showed an increased probability to become centenarians (the frequency of A-C in long-lived and young individuals are 0.332 and 0.250, respectively, OR = 1.49, CI 95% = 1.17–1.88, = 0.0007), in contrast to those with haplotype C-G (the frequency of C-G in long-lived and young individuals are 0.523 and 0.635, respectively, OR = 0.63, CI 95% = 0.51–0.78, = 0.000018). The permuted P-values were 0.00005 and 0.0009, respectively. ADRB2 encodes the β2-adrenergic receptor; the haplotype A-C markedly reduced its translational efficiency compared with C-G (= 0.002) in transfected HEK293 cells. Thus, our data indicate that enhanced production of β2-adrenergic receptors caused by genetic variants is inversely associated with human lifespan.


The β-adrenergic system is critical to the regulation of vascular tone (Chang et al., 2009), cell growth and apoptosis (Singh et al., 2010), lipid metabolism (Zee et al., 2006), and the immunoresponse (Woszczek et al., 2005). Therefore, it is clinically one of the most common pharmacological targets for treating various cardiovascular diseases and asthma. Because of their protective effects from various types of cardiovascular stresses, beta-receptor blockage was hypothesized to increase longevity (Milne & Hong, 2004). In support, it was recently demonstrated that mice lacking ADCY5, encoding type 5-adenylyl cyclase (AC5), are stress resistant and have experience a 30% increase in median lifespan (Yan et al., 2007). AC5 is a member of adenylyl cyclase (AC) family that is activated by β-adrenergic receptors (βARs) and is responsible for the production of cAMP in vivo (Roth et al., 2002). Null function of AC5 blocks the signaling transduction of βARs. On the other hand, transgenic mice engineered to overexpress a β2AR in cardiac tissue have reduced lifespan, with males displaying a more dramatic phenotype (Gao et al., 2003). These findings establish a role for βAR signaling in mammalian longevity regulation and also suggest that this pathway may underscore sexual dimorphic feature of this complex phenotype. However, a role for βAR signaling in human longevity remains to be established, and it is important to address this question because β-adrenergic agonists and antagonists are widely used in treating various human diseases.

Here, we tested for such associations by evaluating the genetic contribution of the β-adrenergic signaling pathway to human longevity in Chinese populations. Sixteen SNPs were selected from the ADRB1, ADRB2, ADCY5, ADCY6, and MAPK1 genes, together encoding the key components of the pathway including the β1-adrenergic receptor (β1AR), β2-adrenergic receptor (β2AR), type 5 adenylyl cyclase (AC5), type 6 adenylyl cyclase (AC6), and mitogen-activated protein kinase 1 (MAPK1), respectively. The genotypic frequencies of these genes were compared between long-lived and middle-age individuals. We identified two common genetic variants of ADRB2 that suppress its translation and are predominantly associated with longevity in men. These findings support that βAR signaling is linked to aging and establish this pathway as a candidate to explain the sexually dimorphic nature of human aging.


Characterization of populations

Chinese Longitudinal Healthy Longevity Survey (CLHLS) was a follow-up study initiated in 1998, investigating the effects of social–biological factors on healthy conditions of the oldest-old population (aged 80+ and mostly 90+). Of 9093 participants were interviewed and 4120 donated the blood samples in 1998. Of those who donated the blood samples, 1072 were centenarians (259 men and 813 women) from Han Chinese nationality and used in this study (Fig. S1, Supporting information). The reliability of age reports in this population has been assessed and validated on our cohort in previous study (Wang et al., 1998). In this study, population 1 (stage 1) included 384 centenarians and 384 gender- and geographically matched young individuals from northern China, while population 2 (stage 2) consisted of 579 cases with exceptional long life span (age ≥ 95 year for men and age ≥ 100 year for women) and 644 geographically matched young individuals from southern China (Table 1). All participants in this study were from Han Chinese. The matched design eliminated the potential effects of population stratification caused by gender and geographic region. However, as a consequence, the total number of individuals in this study was limited by a smaller number of male centenarians; thus, to include enough men with exceptional long lifespan in the population 2, the cutoff age was ≥ 95 years. The Han Chinese population in this study consisted of 963 long-lived (median age is 101 years) and 1028 geography-matched young individuals (median age is 44 years). The mean ages (mean ± SEM) of long-lived individuals in Population 1 and 2 were 102.1 ± 0.20 and 101.0 ± 0.16. The differences between case and control groups for age, gender, body mass index (BMI), systolic blood pressure (SBP), and diastolic blood pressure (DBP) in populations 1 and 2 are shown in Table 1.

Table 1. Characteristics of populations
ItemsPopulation 1Population 2
Case (n = 384)Control (n = 384) P Case (n = 579)Control (n = 644) P
  1. The data are presented as mean ± SEM (standard error of the mean).

  2. BMI, body mass index; Case, long-lived individuals; Control, young individuals; DBP, diastolic blood pressure; m/f, male/female; P-values are calculated from t-test or chi-square test when comparing case and control within population; SBP, systolic blood pressure.

Age (year)102.1 ± 0.2046.8 ± 0.55< 0.05101.0 ± 0.1645.6 ± 0.50< 0.05
Gender (m/f)96/28896/2881.00242/337280/3640.553
BMI (kg m−2)20.4 ± 0.324.2 ± 0.2< 0.0519.3 ± 0.223.1 ± 0.2< 0.05
SBP (mmHg)146 ± 1.3123 ± 1.1< 0.05147 ± 1.0120 ± 0.6< 0.05
DBP (mmHg)82 ± 0.778 ± 0.6< 0.0584 ± 0.676 ± 0.4< 0.05

Genotyping and single SNP association study

Sixteen SNPs were selected from ADRB1, ADRB2, ADCY5, ADCY6, and MAPK1 from the β-adrenergic signaling pathway (Fig. S2, Supporting information). As a first step, the 16 SNPs were genotyped by directly sequencing the amplified genomic DNA from 96 centenarians and 96 young individuals from population 1 (Table S1, Supporting information) with the listed PCR primers (Table S2, Supporting information). None of the SNPs were found to depart from Hardy–Weinberg equilibrium expectations (Table S3, Supporting information). The distributions of minor allele frequencies (MAFs) between case and control groups were compared by the Pearson chi-square test. Of all SNPs analyzed, four from ADRB2 (rs1042718 and rs1042719), ADCY6 (rs2291727), and MAPK1 (rs2266968) exhibited significant differences (< 0.05; Table S3, Supporting information).

We then genotyped the four SNPs in the remaining individuals of population 1 and found that the MAFs of SNPs rs1042718 and rs1042719 from ADRB2 increased in the centenarian group compared with controls (= 0.014–0.028, Table 2). The SNPs, rs2291727 and rs2266968, from ADCY6 and MAPK1 did not differ in the distributions of their MAFs. Four SNPs, rs1042718, rs1042719, rs2291727, and rs2266968, were also genotyped by TaqMan method to validate the genotyping. The consistencies between two methods were greater than 99.3% (not shown).

Table 2. Allelic association of ADRB2 with longevity
SNPsAllele (major/minor)Population 1Population 2Combined
MAF (case/control)OR (95% CI) P/aP MAF (case/control)OR (95% CI) P/aP MAF (case/control)OR (95% CI)P/aPbP/PW
  1. Combined, population 1 plus population 2; MAF, minor allele frequency; OR (95% CI), odds ratio with 95% confidence interval; P, P-value calculated from chi-square test; PW, power when alpha is set as 0.05; aP, P-value adjusted by gender; bP, P-value corrected by Bonferroni method.

rs1042718C/A0.350/0.2981.27 (1.03–1.58)0.028/0.0280.345/0.2971.25 (1.05–1.48)0.011/0.0130.347/0.2971.25 (1.10–1.43)7.92E-4/0.0011.27E-2/0.74
rs1042719G/C0.451/0.3891.29 (1.05–1.58)0.014/0.0140.465/0.4051.28 (1.09–1.50)0.003/0.0030.460/0.3991.28 (1.13–1.45)1.15E-4/1.27E-41.84E-3/0.85

The association of rs1042718 and rs1042719 from ADRB2 with longevity was replicated in population 2 (stage 2) (= 0.003–0.011, Table 2). When two populations were combined, the P-values and statistical powers for rs1042718 and rs1042719 became (P/power) 0.00079/0.74 and 0.00011/0.85. The Bonferroni-corrected P-values were 0.012 and 0.0018, respectively.

Quantitative trait analysis demonstrated that SNPs rs1042718 and rs1042719 of ADRB2 did not exhibit any correlation with BMI, systolic, or diastolic blood pressures in either the oldest-old or their young-matched subjects (Fig. S3, Supporting information).

Gender effects on MAFs of SNPs

To eliminate gender effects on the distributions of MAFs, we compared the MAFs of SNPs between the oldest-old and control groups within gender using the chi-square test. We found that the MAFs of rs1042718 and rs1042719 from ADRB2 were only elevated in men (the Bonferroni-corrected = 0.0005–0.010; statistical power = 0.92–0.93, Table 3) but not in female centenarians compared with their young counterparts (= 0.091–0.102; Table S4, Supporting information). The interactions between genders and SNPs were then tested; the P-values for interaction were 0.07 and 0.019 for rs1042718 and rs1042719, respectively, as shown in Table S5 (Supporting information).

Table 3. Allelic association of ADRB2 with longevity in males
SNPsAllele (major/minor)Population 1Population 2Combined
MAF (case/control)OR (95% CI) P MAF (case/control)OR (95% CI) P MAF (case/control)OR (95% CI)P/PWbP
  1. Combined, population 1 plus population 2; MAF, minor allele frequency; OR (95% CI), odds ratio with 95% confidence interval; P, P-value calculated from chi-square test; PW, power when alpha is set as 0.05; bP, P-value corrected by Bonferroni method.

rs1042718C/A0.365/0.2631.61 (1.04−2.49)0.0330.335/0.2591.44 (1.10−1.88)0.0080.343/0.2601.49 (1.19−1.87)6.31E−4/0.921.01E-2
rs1042719G/C0.458/0.3321.70 (1.12-2.58)0.0120.469/0.3661.53 (1.19−1.96)0.0010.466/0.3571.57 (1.27−1.94)3.31E−5/0.935.30E-4

Enrichment of MAF in groups with increased death ages

To test whether the MAFs of SNPs, rs1042718 and rs1042719 of ADRB2, were enriched with increased age of mortality, we calculated the MAFs of the two SNPs in four male groups: (i) young group (living subjects; life expectancy was reported as 71.3 years for men in China in 2009); (ii) death age = 95; (iii) death ages from 96 to 100; and (iv) death ages > 100. One-way ANOVA and post hoc multiple comparisons were used to test the differences between groups. It was found that the MAFs of the two SNPs were significantly increased with age of mortality (Fig. 1 and Table S6, Supporting information).

Figure 1.

Enrichment of MAF in groups with increased death ages. MAF, minor allele frequency; group 1, young group (living subjects, the life expectancy was reported as 71.3 year for men in China in 2009, n = 376); group 2 (n = 51), death age = 95; group 3 (n = 121), death ages from 96 to 100, and group 4 (n = 166), death ages > 100. The P-values for group 1 vs. 2, group 1 vs. 3, and group 1 vs. 4 were 0.480, 0.049, and 0.0003 for rs1042718 as well as 0.152, 0.002, and 0.0004 for rs1042719, respectively. *Statistically significant.

Genotypic association analysis

To determine through which inheritance models rs1042718 and rs1042719 of ADRB2 were associated with the longevity trait, a genotypic association test was performed under dominant, recessive, and additive models by logistic regression in the combined analysis. We found that rs1042718 and rs1042719 were strongly associated with male longevity in both dominant (= 0.00001–0.003) and additive models (= 0.00001–0.001; Table 4). Association signals were relatively weak in recessive model (= 0.008–0.017). The marginal P-values for the associations were also observed in women (Table 4).

Table 4. Genotypic association of ADRB2 with longevity in males and females from the combined populations
SNPAllele (major/minor)ModelGenotypingMaleFemale
CaseControlOR (95% CI) P CaseControlOR (95% CI) P
  1. Add, additive model; Dom, dominant; OR (95% CI), odds ratio with 95% confidence interval; P, P-value calculated from chi-square test; Rec, recessive.

rs1042718C/ADomA/A + C/A1931721.56 (1.16–2.09)0.0033703521.24 (0.99–1.54)0.060
C/C145201  255300  
RecA/A39222.08 (1.21–3.59)0.00866631.10 (0.77–1.59)0.595
C/A + C/C299351  559589  
AddA/A39221.50 (1.19–1.90)0.00166631.16 (0.98–1.37)0.094
C/A154150  304289  
 C/C145201  255300  
rs1042719G/CDomC/C + G/C2552262.05 (1.48–2.83)1.37E-54594341.41 (1.11–1.80)0.005
G/G82149  163218  
RecC/C59421.68 (1.10–2.58)0.0171091180.96 (0.72–1.28)0.789
G/C + G/G278333  513534  
AddC/C59421.67 (1.32–2.10)1.37E-51091181.16 (0.98–1.36)0.081
G/C196184  350316  
G/G82149  163218  

Linkage disequilibrium and haplotypic association analysis

Linkage disequilibrium (LD) and haplotype blocks were defined by r2 (> 0.6) and visualized by the solid spine of LD method (lower panel in Fig. 2). The common haplotypes (frequency ≥ 5%) were analyzed for an association with longevity. Three haplotypes, C-G, C-C, and A-C, formed by rs1042718 and rs1042719 of ADRB2 were thus selected and tested in a combined analysis (338 male cases and 460 male controls). The frequency of haplotype C-G was decreased (the frequency of C–G is 0.523 and 0.635 in cases and controls, respectively, OR = 0.63, CI 95% = 0.51–0.78, = 0.000018), whereas that of A-C was elevated (the frequency of A–C is 0.332 and 0.250 in cases and controls, respectively, OR = 1.49, CI 95% = 1.17–1.88, = 0.0007) in oldest-olds relative to their male controls (Table 5, Fig. 2); the permutation corrected P-values were 0.00005 and 0.0009, respectively. None of the three haplotypes showed statistical differences in their frequencies between case and control in women (Data not shown).

Figure 2.

Haplotype structure and cloning strategy of ADRB2-firefly luciferase fusion genes (A) and haplotypic effects on protein production (B). A/G, C/G, G/A, C/A, and G/Cwere the major/minor alleles for SNPs rs1042713, rs1042714, rs1042717, rs1042718, and rs1042719, respectively; CMV, Human cytomegalovirus promoter region; ADRB2, β2-adrenergic receptor cDNA; Flu, firefly luciferase cDNA; TK, Herpes simplex virus thymidine kinase promoter region; Rlu, Renilla luciferase cDNA; a, b, and c represented the fusion genes comprising Flu and ADRB2 cDNA harboring haplotypes (C-G), (C-C), and (A-C), respectively; d and e were the constructs to control the transfection. OR, odds ratio; 95% CI, 95% confidence interval; Perm P, permuted P-values.

Table 5. Haplotypic association of ADRB2 with male longevity in combined population
HaplotypeFrequency (case/control)OR (95% CI) P Perm P
  1. OR (95% CI), odds ratio with 95% confidence interval; P: P-value from chi-square test; Perm P: P-value for permutation correction.

C-G0.523/0.6350.63 (0.51–0.78)1.86E-55.00E-5
A-C0.332/0.2501.49 (1.17–1.88)7.00E-40.0009
C-C0.133/0.1071.29 (0.92–1.80)0.1190.323

Effects of haplotypes on translational efficiency of ADRB2 gene

Five common SNPs are found in the ADRB2 coding region: rs1042713, rs1042714, rs1042717, rs1042718, and rs1042719. Among these, two synonymous SNPs, rs1042718 (Arg175Arg) and rs1042719 (Gly351Gly), were associated with longevity in our study. It is common that an exonic SNP affects protein production by influencing RNA stability. To test whether the two SNPs were able to alter the translational efficiency of ADRB2, we respectively cloned the ADRB2 coding regions that harbored the three common haplotypes (C-G, C-C, and A-C) formed by rs1042718 and rs1042719 and fused each with the firefly luciferase reporter gene (Fig. 2). To exclude interferences from other SNPs on an expression assay, we uniformly used the major allele for rs1042713, rs1042714, and rs1042717 in each construct (Fig. 2); the expression vectors were independently transfected into HEK293A cells. Renilla luciferase was used to assess the transfection efficiency. As the only differences among the three fused genes are caused by haplotypic structures, the total translational activity is solely dependent on the respective haplotypes and measure by the ratio of firefly to Renilla luminescence. Therefore, this system accurately assesses the effects of haplotypes on translational efficiency (Tranque et al., 1998). We found that the haplotype A-C led to a reduction in translation efficiency compared with C-G (= 0.002; Fig. 2). The resulting protein level from the fused gene that carried C-C was closer to that of C-G and marginally different from that of A-C (= 0.059). The correlations between expression levels and ORs were shown in Fig. 2.


Longevity is a complex trait that is affected by both genetic and environmental factors (Herskind et al., 1996; Li et al., 2009; Zeng et al., 2010). A great effort has been made to assess to what extent genetics contributes to the trait over the past decades. For instance, a Danish twin-based study indicated that the heritability of longevity was about 0.33 and 0.44 for men and women, respectively (Herskind et al., 1996). A recent meta-analysis of four GWAS identified 273 SNPs associated with longevity at low P-values (< 0.0001) (Newman et al., 2010). Although these studies may not directly indicate genetic contribution to longevity, these suggest there should be intrinsic relationship between certain genetic determinants and the trait. Interestingly, it was found that genetic factors have increasing effects, particularly after the certain age (60 years) (Von Hjelmborg et al., 2006). It is possible that genetic factors contribute to longevity directly by mediating the signaling pathways associated with aging process or indirectly by buffering the deleterious effects caused by other factors (Herskind et al., 1996; Bergman et al., 2007). It has been hypothesized that instituting beta-receptor blockage pharmacotherapy at an early age will increase longevity by countering the effects of sympathetically mediated stress (Milne & Hong, 2004). This, however, remains to be validated.

Here, we assessed the genetic contribution of βAR signaling system to the human longevity trait in Han Chinese long-lived and younger individuals. The average lifespan in China was 71 and 74 years of ages for men and women, respectively. Thus, those who were at the ages of 95 or above were considered as the ‘long-lived,’ and middle-aged groups were taken as controls. The SNPs selected from ADRB1, ADRB2, ADCY5, ADCY6, and MAPK1 were genotyped and evaluated for their association with the longevity trait. We identified two synonymous SNPs of ADRB2 (rs1042718 and rs1042719) that were significantly associated with longevity in men through dominant and additive models. Despite the fact that no positive signals were observed in women in the allele association study, we noticed that these two SNP exhibited marginal P-values in genotypic association in women. This may suggest that the effect size in women is smaller than in men. The association with male longevity was replicated in two geographically isolated Han Chinese populations. The MAFs of the two SNPs were increased with the death ages, providing additional support for the association. In further studies, we demonstrated that the haplotype formed by these two SNPs, which increased the probability of becoming a centenarian, also led to decreased protein production of β2AR. In contrast, over-production of the β2AR is associated with decreased male lifespan. The other SNPs analyzed were not found to be associated with human longevity in the studied populations. To our knowledge, this is the first indication that ADRB2 is linked to human longevity and that a suppression of β2AR production may have beneficial effects on lifespan.

β1AR and β2AR are two well-studied subtypes of beta adrenoceptors (Xiao et al., 2006). β1AR is predominantly expressed in the heart, cerebral cortex, and kidney, while β2AR appears more abundant in blood vessels, lung, skeletal muscle, and adipose tissue. It has been shown with engineered mouse models that the β1 receptor is critical for the regulation of cardiac inotropy and chronotropy (Rohrer et al., 1996). Both subtypes are important in regulating vascular relaxation, metabolic rate, and diet-induced thermogenesis (Rohrer et al., 1999). In the β-adrenergic signaling system (Fig. S2, Supporting information), the βAR dissociates a Gsα subunit from the Gsα-β-γ complex to activate adenylyl cyclase (AC). AC catalyzes the conversion of ATP into cAMP, activating PKA, and subsequently promoting phosphorylation of downstream target proteins such as Raf-1; which leads to its inactivation (Dhillon & Kolch, 2002). Reduced PKA signaling is known to result in enhanced yeast replicative and chronological lifespan (Lin et al., 2002; Gendron et al., 2003), and it was recently documented that mice lacking AC5 have decreased PKA activity, leading to the activation of the Raf/MEK/ERK signaling pathway and enhanced lifespan in mice (Yan et al., 2007). The prolonged lifespan is presumably due to protecting mice from aging-induced pressure overload and catecholamine-induced stress (Vatner et al., 2009). Despite the fact that no significant associations were found between the ADCY5 or ADCY6 genotype and human longevity in the present study, we demonstrated that β2AR, the upstream activator of AC5 and/or AC6, is strongly associated with longevity in men. Furthermore, we found that the haplotype A-C formed by two SNPs (rs1042718 and rs1042719) reduced the expression of β2AR and was positively associated with longevity, while the haplotype C-G increased the translational efficiency of β2AR and was negatively associated with longevity. Notably, an existing study also demonstrated that the haplotype harboring A-C alleles decreases the translational level, providing strong evidence that A-C alleles are functional (Panebra et al., 2010). Thus, our data indicate that the β2AR-triggered signaling pathway contributes to human lifespan.

Several possibilities exist to explain mechanism(s) linking β2AR signaling to human lifespan. In addition to interacting with GTP-binding proteins (Gs and Gi) that control the activation of AC, β2AR interacts directly with protein kinase and adaptor proteins to regulate a range of other biological functions raising the possibility that the benefits of reduced β2AR signaling may be multifactorial in nature.(Benovic, 2002). Second, β2AR is an important mediator of lipolysis (Lima et al., 2005) and has been genetically linked to obesity (Masuo et al., 2005) and insulin resistance (Park et al., 2008). Given the tight coupling between longevity and metabolism, it is reasonable to speculate that reduced β2AR activity may improve insulin sensitivity with age and that these metabolic changes may promote longevity. It was well documented that β-adrenergic responses decrease with age (Xiao et al., 1998). While the reasons for this decline are not apparent, it is intriguing to speculate that this is a response to events that occur with age and that individuals with a lower response earlier in life may be protected from changes associated with the aging process.

Finally, certain genetic variants of β2AR have been linked to a lower risk of coronary events in elderly population (Heckbert et al., 2003) as well as myocardial infarction (Zee et al., 2005), asthma (Wechsler et al., 2009), hypertension (Ge et al., 2005), and obesity (Masuo et al., 2005). We did not test how genotypes affect disease status in centenarian population. This is due to the fact that the disease status was based on the report, not on medical examination. Interestingly, a recent study has found that a β2AR genetic polymorphism was associated with healthy aging (Kulminski et al., 2010).

One interesting potential molecular link between β2AR activation and aging has emerged from a recent study (Hara et al., 2011), which shows that stimulation of β2ARs in response to chronic stress leads through PKA and β-arrestin to accelerated DNA damage and destabilization of p53 through Akt-mediated Mdm2 activation. Given the support for accumulated DNA damage as a driver of aging, it is tempting to speculate that one benefit of reduced β2AR signaling involves attenuation of this response to chronic stress.

It has been well documented that the longevity trait exhibits sexual dimorphism. Women occupy about 80% of centenarians (Li et al., 2009). In this study, we find that the association of β2AR with human longevity is linked to gender. The two synonymous SNPs of ADRB2 and their haplotypes tend to be associated with the longevity trait preferentially in men. That the MAFs of the two SNPs were enriched with the death ages strongly supports the association. Previously, it has been reported that transgenic mice with cardiac-selective expression of β2AR showed a dramatic reduction in male lifespan, while this phenotype was much less severe in females (Gao et al., 2003; Thireau et al., 2010), supporting the hypothesis that reduced β2AR levels in vivo have a beneficial effect on male lifespan. Perhaps, consistently, it has been reported that male but not female mice lacking the protein kinase A regulatory subunit RIIb have extended lifespan (Enns et al., 2009). The reason why the β-adrenergic system exhibits sexual dimorphism in the control of lifespan is largely unknown. In a transgenic mouse model, it has been shown that orchiectomy rescues this life-threatening phenotype caused by the over-expression of β2AR, suggesting that sex hormones such as testosterone are critical to the underlying mechanism or pathway (Gao et al., 2003; Thireau et al., 2010). On the other hand, the transgenic mice develop heart failure, and therefore, the mechanism by which over-expression of β2AR shortens male lifespan may not reflect the natural process of aging. For this reason, it would be better to compare the longevity of mouse models lacking either β1AR or β2AR because these mice appear normal. The lifespan phenotypes of these mice remain to be reported (Rohrer et al., 1999).

Indeed, the β-adrenergic system exhibits sexual dimorphism in several aspects, including the expression of βARs (Paulose & Kanungo, 1982), memory impairment (Cahill & Van Stegeren, 2003), growth hormone release (Krieg et al., 1986), vascular constriction (Li & Duckles, 1994), and Ca2+ influx in cardiac myocyte (Vizgirda et al., 2002). Notably, epinephrine-induced cAMP accumulation in hepatocytes is about 3-fold higher in females than in males (Yagami et al., 1990). Growth hormone release from pituitary cells induced by β2-adrenergic stimulation is more potent in male than in female rats (Krieg et al., 1986). The rate of lipolysis stimulated by 8-bromo-cAMP and isoprenaline is greater in the subcutaneous than in omental adipocytes, and this is not seen in males (Lundgren et al., 2008). β1 or β2AR blocking treatment has a cardioprotective effect after high-risk vascular surgery in men but not in women (Matyal et al., 2008). Whether these differences are associated with the sexual dimorphism of β2ARs in the regulation of human lifespan requires further investigation.

Despite the fact that our findings were strongly supported by several lines of evidence, the limitations need to be considered in this study. Firstly, only 96 centenarians and 96 controls were selected for primarily screen SNPs for the association. Clearly, this design may be restrictive to the identification of SNPs that were prominently associated with the longevity trait. Secondly, due to a limited number of SNPs genotyped in this study, the independence or structures of population 1 and 2 could not be evaluated by, for example, principal component analysis, which requires a large number of SNPs. Thirdly, the ages for control groups were relatively young. This design can maximize the contrast between cases and controls. On the other hand, it can also bring in other strata, for instance, reproductive age, early onset of diseases, and so on. Lastly, it is well documented that women have longer life expectancy than men (Zeng & Vaupel, 2004). It is therefore very difficult to find a large population of male centenarians. A small sample size will increase statistic errors and bias in conclusions, and thus, whether or not β2AR is associated with longevity trait in other ancestries needs to be further investigated. Nonetheless, such data can be used as one of the tests for meta-analysis of human longevity in the future.

In summary, we show that common haplotypes of β2AR, which affect its translational efficiency, are associated with longevity in men and that the level of β2AR protein is inversely associated with male lifespan. These findings may well have clinical implications for treating patients with βAR agonists or antagonists. On the other hand, it is well known that the evaluation of a long-term effect of a pharmacological intervention in humans is methodologically difficult, as well as being time- and funding-intensive in clinical practice. Our study suggests that the identification of genetic variants that exert similar effects to the pharmacological interventions may provide an alternate ‘shortcut’ route to achieving such goals.

Experimental procedures

Human subjects

We performed a 10-year follow-up study for the ‘Chinese Longitudinal Healthy Longevity Survey (CLHLS)’ from 1998 to 2008 (Zeng & Gu, 2008). In the Survey, 9093 oldest-olds aged 80–116 were recruited in 1998 and interviewed in 1998, 2000, 2002, 2005, and 2008 with a questionnaire and basic physical examinations. In 2011, we recruited additional 21 blood samples from the male oldest-old in south of China. An individual's age reported in the survey was based on the Resident Identity Card of the People's Republic of China and the Household Register, with additional evidence including genealogical records, marriage certificates, demographic statistical data of local administration offices, migration documents, certification for employment or military service, and the witness of neighbors. The Resident Identity Card and Household Register are both issued by the Ministry of Public Security of the People's Republic of China to an individual and a family, respectively. The Resident Identity Card has the holder's name and a number that carries the information on place and date of birth; the Household Register presents the basic information on names, genders, dates of birth, relationship, marriage status, and the professions of all family members.

We selected 384 centenarians and 384 gender-matched young individuals from northern China as population 1. The population 2 consisted of 579 cases with exceptional long lifespan (age ≥ 95-year for men and age ≥ 100-year for women) and 644 young individuals from southern China. A detailed protocol diagram for the inclusion of the participants was provided as Fig. S1 (Supporting information). The study protocol was approved by the Institutional Review Board, Institute of Molecular Medicine at Peking University. The signed informed consent forms were obtained from all participants or their representative family members in cases of the oldest-olds who were incapable of signing. The study conformed to the principles outlined in the Declaration of Helsinki. (The detailed information on sampling and storage was in ‘Sampling and genotyping’ in Data S1, Supporting information).

Genotyping of SNPs and Validation

Human genomic DNA was isolated from a 5-mm-diameter punch-out from each blood spot or from the EDTA-anti-coagulated blood using the proteinase K methods. Based on the HapMap (CHB+JPT), the 16 SNPs from the ADRB1, ADRB2, ADCY5, ADCY6, and MAPK1 genes were selected for genotyping by directly sequencing. The primers used to amplify the SNPs are presented (Table S2, Supporting information). (The detailed information on genotyping and validation was in ‘Sampling and genotyping’ of Data S1, Supporting information).

Constructs and cell transfection

The ADRB2 coding region was generated and cloned by PCR with primers 5′-CGCCCTCGAGATGGGGCAACCCGGGAAC-3′ and 5′-CTGCTAAGCTTTTACCAGCAGTGAGTCAT-3′. We constructed plasmids that carried a CMV promoter, the human full-length coding region of ADRB2, and firefly luciferase fused at its 3′-end. Three common haplotypes (frequencies > 5%) formed by rs1042718 (C/A) and rs1042719 (G/C) were generated by PCR-based mutagenesis and confirmed by sequencing. The major alleles of SNPs rs1042713, rs1042714, and rs1042717 were used to avoid any possible interference. The translational efficiency was measured by firefly luciferase activity. Transfection efficiency was standardized by reference to Renilla luciferase activity resulting from the parallel introduction of this plasmid into the test cells.

The protein expression level of the ADRB2 gene was measured by the ratio of light units of β2AR firefly to Renilla. Data represent mean ± SEM of four independent experiments performed in triplicate (more detailed description was in ‘Functional characterization of SNPs’ in Data S1, Supporting information).

Statistical analysis

Allele and genotype frequencies of SNPs were calculated and tested for departure from Hardy–Weinberg equilibrium using the chi-square test. Differences in allele and genotype distributions between case (long-lived population) and control (young population) were analyzed using logistic regression under various genetic models (Li et al., 2009). A Bonferroni method was used to correct multiple comparison in allele association tests (Holm, 1979).

Linkage disequilibrium (LD) and haplotype blocks were constructed with genotyping data from this study and defined (r2 > 0.60) and visualized by the solid spine of LD method using Haploview 4.0 (http://www.broadinstitute.org/scientific-community/science/programs/medical-and-population-genetics/haploview/haploview). The haplotypes with frequencies ≥ 5% were subjected to an association test with longevity. The permutation P-value was obtained by simulating 100 000 times for multiple comparison correction in haplotype association analysis (Rice et al., 2008).

The two-tailed P-values, odds ratios, and 95% confidence intervals (95% CI) are presented for all association tests. One-way anova with post hoc LSD test was used to compare BMI, systolic blood pressure, and diastolic blood pressure among three genotypes for each SNP, and to compare the ratios of firefly to Renilla luciferase (Ratio of Flu/Rlu) of three haplotypic groups. A P-value of < 0.05 was considered statistically significant.


This study was supported by grants from the National Basic Research Program of the Chinese Ministry of Science and Technology (973 Grant No.: 2007CB512100), Key Program from the National Natural Science Foundation of China NSFC (Grant No.: 30730047 and 70533010) and NIA/NIH (R01 AG023627).

Author contributions

Xiao-Li Tian and Ling Zhao designed the study; Yi Zeng supervised the collection of blood samples and information from the oldest-old subjects; Hanbing Cui, Xiaomin Chen, Zhiming Zhu, Hongbo He, and Xianming Mo collected the blood samples and clinical information from geographically matched controls. Ling Zhao, Fan Yang, Ke Xu, Huiqing Cao, Gu-Yan Zheng, Yang Zhang, and Jianxin Li performed experiments; Xiao-Li Tian, Ling Zhao, Fan Yang, Brian K. Kennedy, and Yousin Suh analyzed and interpreted data; Xiao-Li Tian and Ling Zhao wrote the manuscript; Brian K. Kennedy, Fan Yang, Ke Xu, Yi Zeng, and Yousin Suh were involved in the final approval of the manuscript.