Genetic Analysis of TOR Complex Gene Variation With Human Longevity: A Nested Case–Control Study of American Men of Japanese Ancestry

Abstract

The mechanistic target of rapamycin (mTOR) pathway is crucial for life span determination in model organisms. The aim of the present study was to test tagging single-nucleotide polymorphisms that captured most of the genetic variation across key TOR complex 1 (TORC1) and TOR complex 2 (TORC2) genes MTOR, RPTOR, and RICTOR and the important downstream effector gene RPS6KA1 for association with human longevity (defined as attainment of at least 95 years of age) as well as health span phenotypes. Subjects comprised a homogeneous population of American men of Japanese ancestry, well characterized for aging phenotypes and who have been followed for 48 years. The study used a nested case–control design involving 440 subjects aged 95 years and older and 374 controls. It found no association of 6 tagging single-nucleotide polymorphisms for MTOR, 61 for RPTOR, 7 for RICTOR, or 5 for RPS6KA1 with longevity. Of 40 aging-related phenotypes, no significant association with genotype was seen. Thus common genetic variation (minor allele frequency ≥10%) in MTOR, RPTOR, RICTOR, and RPS6KA1 is not associated with extreme old age or aging phenotypes in this population. Further research is needed to assess the potential genetic contribution of other mTOR pathway genes to human longevity, gene expression, upstream and downstream targets, and clinically relevant aging phenotypes.

Age is the most powerful independent risk factor for major causes of death in developed countries. The rate of aging is influenced by environmental and genetic factors. Identification of the molecular mechanisms responsible for this complex polygenic phenotype represents a major current challenge. Because the genetic component becomes paramount for the attainment of extreme old age (1–3) and caloric restriction is the most effective means of extending life span in diverse species (4), studies of polymorphisms in genes belonging to intracellular pathways that are modulated by caloric restriction have been conducted in nonagenarian and centenarian populations (4,5). These pathways affect insulin/insulin-like growth factor-1 signaling and growth responses. The hyperfunction theory of aging postulates that in later life, the once-beneficial processes contributing to growth become deleterious by causing hypertrophic and hyperplastic pathologies (6).

The mechanistic (formerly “mammalian”) target of rapamycin (mTOR) is positioned at a central hub of nutrient-sensing pathways (7,8) and is pivotal to life span extension in response to caloric restriction (9,10). Despite good evidence showing that insulin/insulin-like growth factor-1 and mTOR pathways are critical for life span regulation of model organisms (6,11,12), to date, attainment of extreme old age in humans has been consistently replicated only for variation in one gene in these pathways, the forkhead box O transcription factor 3 gene (FOXO3) (5,6).

mTOR is an evolutionarily conserved serine/threonine phosphoinositide 3-kinase-related kinase with crucial roles in cell growth and metabolism in response to nutrients, growth factors, cellular energy, and stress (6,8,13,14). Rapamycin forms a stoichiometric complex with mTOR to inhibit its kinase activity (15,16) and can thus extend mean and maximum life span of middle-aged male and female mice, with no adverse health consequences, only benefits (17,18).

mTOR is a component of two structurally and functionally distinct multiprotein complexes—TOR complex 1 (TORC1) and TOR complex 2 (TORC2). In mammals, two accessory proteins—regulatory-associated protein of mTOR (raptor or RPTOR) and rapamycin-insensitive companion of mTOR (rictor or RICTOR)—distinguish TORC1 from TORC2, respectively (19,20). mTORC1 exerts its effects on temporal control of cell growth by regulating several cellular processes, including translation, transcription, ribosome biogenesis, nutrient transport, and autophagy (14). The two TOR complexes constitute an evolutionarily conserved ancestral signaling network that responds to energy levels and sources of energy in order to control the fundamental process of cell growth and maintenance. In addition to its key role in aging, mTOR has been implicated in cancer, cardiovascular disease, obesity, and diabetes (21).

Rapamycin inhibits TORC1 by forming a complex with FK506-binding protein (FKBP12), which then binds to and inhibits TORC1 in its complex with raptor (16,22). In contrast, rapamycin does not bind to TORC2. Binding of raptor to eukaryotic initiation factor 4E-binding protein-1 stimulates ribosomal protein S6 kinase (23) and ribosomal protein S6 kinase, 90kDa, polypeptide 1 (RPS6KA1). RPS6KA1 is a downstream effector of nutrient-responsive mTOR signaling. It regulates multiple transcription factors, increases cell growth—in part by increasing mTOR signaling—and links nutrient availability to aging in organisms as diverse as yeast and mice (24). Rapamycin reduces mTOR-mediated phosphorylation of RPS6KA1 (16), so increasing life span and resistance to age-related pathologies, as well as producing gene expression patterns similar to those seen in caloric restriction or pharmacological activation of adenosine monophosphate (AMP)-activated protein kinase (AMPK), a conserved regulator of the metabolic response to caloric restriction (25).

Here we report the results of a genetic association study of tagging single-nucleotide polymorphisms (tagSNPs) that provided virtually complete coverage of key mTOR complex genes MTOR, RPTOR, and RICTOR, as well as RPS6KA1, in long-lived Americans of Japanese ancestry. The study tested for association of these genetic variants with living to more than 95 years of age and with multiple aging-related parameters in these subjects.

Methods

Study Population

We used American men of Japanese ancestry drawn from the Honolulu Heart Program (HHP) and Honolulu-Asia Aging Study (HAAS; HHP and HAAS) population. Recruitment, design, and procedures have been outlined in detail elsewhere (26,27). Briefly, recruitment of subjects aged 45–68 years (mean age: 54 years) began on Oahu in 1965–1968. These men have been followed with 12 regular examinations and blood work until the present, or death. A nested case–control design was utilized for the present study. “Cases” (longevity phenotype) were defined as HHP and HAAS participants who had survived to beyond the upper 1% of the 1910U.S. birth cohort–specific survival (minimum 95 years of age; n = 440 as of June 2011, 296 having died [age of death = 97.0±2.2 SD; range: 95–106 years]) (28) and 144 still being alive (mean age = 96.3±1.6 S D; range: 95–106 years). Controls consisted of 374 individuals from the HHP and HAAS cohort who had died near the mean for the 1910U.S. birth cohort–specific survival for middle-aged men (77 years of age). To achieve a case:control ratio of approximately 1:1, the HHP and HAAS study population was sampled for controls who died up to the age of 81 years (mean age at death = 78.1±1.8 S D years [range: 73–81 years]), consistent with the 3.5 years longer life expectancy of Japanese American men in Hawaii (29). Analyses showed no evidence of population stratification in the data set (5). All participants were drawn from records of study participants updated up to June 2011. Archived phenotypic data and blood samples from Examination 4 of the HHP (1991–1993), which coincided with the commencement of the HAAS, were used as the baseline exam for this nested case–control study. We had phenotype data for only 439 cases because one subject did not attend Exam 4. Participants included 3,734 men aged 71–93 years at Exam 4 (mean age: 77.9±4.7 years), representing approximately 80% of the survivors from the original HHP (30).

Procedures performed were in accordance with institutional guidelines and were approved by the Institutional Review Board of Kuakini Medical Center. Written informed consent was obtained from all study participants or from family representatives, if participants could not provide consent.

Genotyping

Total cellular DNA was isolated from leukocytes using the PureGene system (Gentra Systems) and quantified using PicoGreen staining (Molecular Probes, Eugene, OR). SNPs genotyped were from within the gene and from a 5-kb region of DNA flanking either end of the genes MTOR, RPTOR, RICTOR, and RPS6KA1. TagSNPs were selected using Haploview, a program that defines a haplotype based on high correlation between the first and last markers and all intermediate markers. Supplementary Figures S1–S4 show the linkage disequilibrium blocks captured by the tagSNPs used. We restricted our selection to SNPs having a minor allele frequency of ≥10% in the Japanese population (as indicated by the HapMap database: http://hapmap.ncbi.nlm.nih.gov/; HapMap release 27/phase II + III, Feb 2009 on NCBI B36 assembly, dbSNP b126 – http://www.ncbi.nlm.nih.gov/projects/SNP/). The minimal coefficient of determination (r ²) value at which all alleles are to be captured was set to a threshold of .8 for the identification of all tagSNPs. Genotype data were managed through an integrated database system sample management-data processing system that has been proven to be accurate in other successful studies (5). All positive controls on each genotyping plate were evaluated for consistency. Call rates for markers exceeded 98%.

SNPs for MTOR, RPTOR, and RICTOR were genotyped at the University of Hawaii Cancer Center on the Illumina GoldenGate platform (high-throughput SNP genotyping on universal bead arrays (31)). SNPs for RPS6KA1 were detected by the TaqMan OpenArray System (Life Technologies). We implemented a series of quality control checks based on the Illumina metrics. For inclusion of data for a SNP, its call rate had to exceed .95 and the Hardy–Weinberg equilibrium p needed to be >.01. Of the 81 tagSNPs genotyped, all met these criteria and were included. In addition, more than 96% concordance was observed for all SNPs for HapMap samples that were assayed for quality assurance.

Statistical Analyses

Genotype frequencies of SNPs were evaluated for deviation from Hardy–Weinberg equilibrium. The Pearson chi-square test was used to compare cases and controls for difference in genotype frequencies. The 95% confidence intervals for allele frequencies were calculated as described (32). For analyses of phenotypes, linear regression models were used for continuous variables (with log transformation of skewed variables) and logistic regression models were used for categorical variables, to test for associations with genotypes, adjusting for case–control status and age at baseline. The analyses used the SAS statistical package (33).

Results

Characteristics of Participants

Table 1 shows the age-adjusted baseline characteristics of the HHP and HAAS study population at the 1991–1993 examination when blood was drawn for genotyping in the present study (Exam 4 of the cohort). The more limited set of parameters recorded at Exam 1 (1965–1968) are also shown. Unadjusted data are shown in Supplementary Table S1. Presented are biological characteristics, general health status, disease prevalence, and functional status of long-lived “cases” (longevity phenotype) and average-lived “controls” (average phenotype). The long-lived cases were older, leaner (based on lower waist:hip ratio), had lower plasma glucose level, lower plasma insulin level, better self-rated health, and lower prevalence of cardiovascular disease (hypertension, coronary heart disease, and stroke) and cancer. The cases appeared better able to walk 0.8 km than controls, but grip strength was similar. Cognitive score was slightly higher in cases. When adjusting for age, the differences between long-lived cases became even starker, with cases demonstrating significantly lower blood pressure and clinical phenotypes in midlife and better grip strength and cognitive score in late life.

Table 1.

Age-Adjusted Baseline Characteristics of Case and Control Participants (mean ± SD)

Parameter	Controls	Cases	p
n	374	439	—
Biological and physiological phenotypes
Height (cm) at Exam 1	163.2±5.4	162.8±5.4	.38
Height (cm) at Exam 4	161.2±5.6	161.3±5.5	.96
Weight (kg) at Exam 1	64.1±9.4	62.0±7.9	.009
Weight (kg) at Exam 4	59.6±9.6	61.6±8.0	.020
BMI (kg/m2) at Exam 1	24.2±3.0	23.5±2.7	.016
BMI (kg/m2) at Exam 4	23.0±3.1	23.6±2.9	.040
BMI at age 25	24.0±3.0	23.4±2.7	.016
Waist:hip ratio	0.95±0.06	0.94±0.05	.015
Systolic BP (mmHg) at Exam 1	135.1±18.8	128.7±18.3	.000
Systolic BP (mmHg) at Exam 4	151.0±26.3	150.4±22.4	.79
Diastolic BP (mmHg) at Exam 1	83.1±11.8	81.2±10.5	.076
Diastolic BP (mmHg) at Exam 4	79.0±12.7	81.6±11.6	.022
EHT (%) >140/90 /meds at Exam 4	74±43	79±41	.24
EHT (%) 160/95 /meds at Exam 4	58±50	57±50	.79
EHT (%) 140/90 /meds at Exam 1	43±49	32±47	.018
EHT (%) 160/95 /meds at Exam 1	26±42	14±37	.0028
ISH (%) SBP>140 at Exam 1	36±47	22±43	.0016
ISH (%) SBP>160 at Exam 1	12±29	3.6±24	.0006
ISH (%) SBP >140 at Exam 4	62±49	66±47	.34
ISH (%) SBP >160 at Exam 4	35±45	31±48	.44
ABI <0.9 (%)	25±38	3.7±28	<.0001
Plasma glucose (mg/dl)	118±37	109±22	.0005
Insulin (mIU/dl)	17.4±15.6	13.8±9.5	.0036
Log insulin (mIU/dl)	2.6±0.6	2.5±0.5	.022
HOMA	5.5±7.1	3.8±3.0	.0008
Total cholesterol (mg/dl)	183±34.2	195±32.6	.0003
HDL cholesterol (mg/dl)	51.3±14.4	51.3±13.1	.97
LDL cholesterol (mg/dl)	105±32.7	114±30.6	.0034
Triglycerides (mg/dl)	146±119	152±79	.54
Other clinical phenotypes
Poor or fair self-rated health (%)	44±49	25±44	.0001
Coronary heart disease (%)	25±44	13±31	.0014
Stroke history (%)	9.3±27	1.2±15	.0001
Cancer (%)	23±40	7.9±31	.000009
Diabetes (%)	37±48	20±41	.0001
On diabetes medication (%)	18±38	6.9±26	.0003
COPD (emphysema; %)	4.8±21	1.4±12	.038
Physical and cognitive functional phenotypes
Grip strength (kg)	26.9±7.1	31.7±5.2	1×10–15
Difficulty walking 0.8 km (%)	40±46	6.6±32	1×10–17
Cognitive (CASI) score	75.6±19.8	87.1±10.6	2×10–14

Notes: The Honolulu Heart Program (HHP) has had 12 full exam cycles since study onset in 1965. Shown are phenotypic variables for long-lived cases and average-lived controls at two exam cycles: HHP Exam 4 (1991–1993), the baseline exam for the present study (ie, when blood was collected for genetic studies and participants were elderly (mean age 76 years; range: 71–93) and Exam 1 (1965–68), when study participants were middle-aged (mean age 54 years; range 45–68). Cases consisted of all Kuakini HHP and Honolulu-Asia Aging Study (HAAS) participants (both living and dead) for which DNA had been collected at Exam 4 and who achieved longevity (age ≥95 years) by June 2011, the follow-up period for this nested case–control study. Data are shown for Exam 4 unless otherwise indicated. ABI = ankle brachial index (ratio of BP in the lower legs to the BP in the arms); BMI = body mass index; BP = blood pressure; CASI = cognitive abilities screening instrument (score range: 1–100); CHD = coronary heart disease; COPD = chronic obstructive pulmonary disease (emphysema); EHT = essential hypertension (moderate: systolic/diastolic BP 140–160 mmHg/90–95 mmHg or on antihypertensive medication; severe: systolic/diastolic BP >160/>95 mmHg or on antihypertensive medication); HDL = high-density lipoprotein; HOMA = homeostasis model assessment; ISH = isolated systolic hypertension (moderate: systolic BP 140–160 mmHg; severe: systolic BP >160 mmHg); LDL = low-density lipoprotein.

Genetic Association Findings

The minor allele frequencies of each SNP tested in MTOR, RPTOR, RICTOR, and RPS6KA1 in cases and controls are presented in Tables 2–5, together with results of chi-square analyses. Differences that appeared statistically significant for two SNPs in RPTOR (rs7225574 and rs2589143) and one SNP in RICTOR (rs1239256) were lost after correction for multiple comparisons.

Table 2.

Frequency of Minor Allele of Each Tagging SNP of MTOR on Chromosome 1p36 in Long-Lived Case vs Control Participants

SNP ID	Position	Total Number of Cases and Controls	Minor Allele Frequency of Cases and Controls, % (95% CI)	χ2	p Value
rs1057079	11205058	440/374	12.6 and 12.8 (–0.035 to 0.030)	1.42	.49
rs11581010	11212458	440/374	10.7 and 11.1 (–0.035 to 0.026)	0.36	.84
rs3806317	11248216	440/374	13.8 and 15.2 (–0.049 to 0.019)	0.81	.67
rs11121703	11293792	439/374	13.9 and 15.4 (–0.049 to 0.020)	1.65	.44
rs4845858	11304720	439/372	19.0 and 18.3 (–0.031 to 0.045)	0.63	.73
rs3765904	11321864	440/374	7.3 and 7.4 (–0.026 to 0.025)	1.22	.54

Notes: We captured 57 of 57 alleles at r ² ≥ .8, ie, we captured 100% of alleles, and mean of maximum r ² was .97 for the six tagSNPs. CI = confidence interval; SNP = single-nucleotide polymorphism.

Table 5.

Frequency of Minor Allele of Each Tagging SNP of RPS6KA1 on Chromosome 1p36.11 in Long-Lived Case vs Control Participants

SNP ID	Position	Total Number of Cases and Controls	Minor Allele Frequency of Cases and Controls, % (95% CI)	χ2	p Value
rs12094989	26720973	184/358	27.2 and 30.0 (–0.084 to 0.029)	1.72	.42
rs17162190	26745419	177/346	27.4 and 31.4 (–0.097 to 0.019)	1.70	.42
rs2278978	26745832	183/355	23.0 and 24.9 (–0.073 to 0.034)	3.10	.22
rs4970510	26783341	180/346	13.6 and 14.6 (–0.053 to 0.035)	0.21	.90

Notes: The cases and controls used for the RPS6KA1 study were a subset described previously (5). We captured 24 of 24 alleles at r ² ≥ .8, ie, we captured 100% of SNPs, and mean of maximum r ² was .94 for the four tagSNPs. CI = confidence interval; SNP = single-nucleotide polymorphism.

Table 3.

Frequency of Minor Allele of Each Tagging SNP of RPTOR on Chromosome 17p25.3 in Long-Lived Case vs Control Participants

SNP ID	Position	Total Number of Cases and Controls	Minor Allele Frequency of Cases and Controls, % (95% CI)	χ2	p Value
rs4889856	78513632	436/366	33.3 and 34.8 (–0.062 to 0.031)	3.70	.16
rs8064502	78562434	440/374	36.4 and 31.7 (0.001 to 0.093)	4.01	.14
rs4890052	78564245	438/371	34.0 and 36.4 (–0.070 to 0.023)	2.68	.26
rs7220588	78598231	437/371	34.3 and 37.1 (–0.074 to 0.019)	2.91	.23
rs901065	78599655	425/360	35.3 and 31.9 (–0.013 to 0.080)	2.89	.24
rs8066867	78620072	430/360	27.3 and 29.0 (–0.062 to 0.027)	1.61	.45
rs4889782	78640510	440/374	30.1 and 32.1 (–0.065 to 0.025)	1.73	.42
rs4889875	78651500	440/374	34.8 and 37.0 (–0.069 to 0.024)	1.29	.53
rs8065598	78678243	438/371	31.4 and 32.6 (–0.058 to 0.033)	0.30	.86
rs4969256	78745643	440/374	52.3 and 49.1 (–0.017 to 0.081)	2.68	.26
rs9898212	78751489	440/374	28.6 and 29.1 (–0.049 to 0.039)	0.89	.64
rs11150745	78757626	440/374	34.9 and 31.1 (–0.008 to 0.083)	2.72	.26
rs11150746	78761389	440/374	28.6 and 29.1 (–0.049 to 0.039)	0.08	.96
rs4969266	78761546	440/374	47.0 and 47.2 (–0.050 to 0.047)	0.51	.77
rs11150747	78761607	439/373	19.0 and 19.5 (–0.043 to 0.034)	0.23	.89
rs4969426	78762808	440/374	20.6 and 23.8 (–0.073 to 0.008)	4.34	.11
rs8069962	78766464	440/374	25.6 and 29.0 (–0.078 to 0.009)	2.44	.30
rs12601434	78767275	438/372	31.4 and 28.2 (–0.013 to 0.076)	1.93	.38
rs11654508	78771947	410/353	35.5 and 32.1 (–0.014 to 0.081)	2.38	.30
rs7225574	78777941	440/374	13.1 and 16.0 (–0.064 to 0.005)	5.13	.03
rs2672886	78781651	426/362	49.8 and 49.7 (–0.049 to 0.050)	0.58	.75
rs734338	78782340	440/374	46.4 and 46.5 (–0.050 to 0.046)	1.86	.51
rs7212127	78787662	440/374	11.5 and 11.4 (–0.030 to 0.032)	0.01	.99
rs2589134	78793593	438/374	34.6 and 34.0 (–0.040 to 0.053)	0.45	.80
rs2672901	78796666	440/373	29.3 and 32.0 (–0.072 to 0.018)	4.97	.08
rs2589143	78810687	437/371	49.7 and 44.7 (–0.000 to 0.098)	6.69	.04
rs2589157	78820023	440/374	30.2 and 28.5 (–0.027 to 0.062)	0.71	.70
rs3751945	78820329	440/374	41.3 and 43.3 (–0.069 to 0.027)	2.41	.30
rs2333988	78821447	439/374	18.9 and 16.2 (–0.011 to 0.063)	2.90	.23
rs2672893	78827863	440/374	39.0 and 36.7 (–0.025 to 0.069)	0.34	.84
rs4969282	78828266	440/377	28.4 and 26.2 (–0.022 to 0.065)	1.00	.61
rs9895380	78829484	440/374	10.2 and 10.4 (–0.032 to 0.028)	1.49	.47
rs2672890	78834207	440/374	36.6 and 39.2 (–0.073 to 0.021)	1.29	.52
rs2589118	78836553	440/374	51.5 and 49.6 (–0.030 to 0.067)	0.66	.72
rs2063785	78836846	440/374	8.3 and 8.8 (–0.033 to 0.022)	1.59	.45
rs2589141	78843139	440/374	42.0 and 44.3 (–0.070 to 0.026)	0.83	.66
rs2289762	78859110	440/374	38.1 and 35.2 (–0.018 to 0.076)	2.21	.33
rs11656061	78864749	440/374	44.7 and 40.8 (–0.009 to 0.087)	3.08	.21
rs2289763	78865491	440/374	16.4 and 17.2 (–0.045 to 0.028)	3.33	.31
rs7217623	78873014	440/374	22.4 and 22.1 (–0.037 to 0.044)	4.03	.13
rs2280146	78875669	440/374	22.8 and 23.5 (–0.048 to 0.034)	2.75	.25
rs868432	78877735	440/374	48.3 and 48.5 (–0.051 to 0.046)	0.52	.77
rs4969301	78885904	440/374	25.5 and 27.5 (–0.064 to 0.022)	1.23	.54
rs12938300	78894022	439/374	49.4 and 45.3 (–0.008 to 0.090)	0.80	.67
rs908236	78898600	440/374	34.1 and 33.7 (–0.042 to 0.050)	0.29	.86
rs6565494	78898778	417/357	24.3 and 23.9 (–0.039 to 0.047)	0.20	.90
rs2271608	78899458	440/374	9.8 and 10.2 (–0.033 to 0.025)	0.50	.78
rs4969227	78900598	440/374	15.3 and 16.7 (–0.050 to 0.022)	0.60	.74
rs4969311	78906360	440/374	25.1 and 26.9 (–0.060 to 0.025)	1.12	.57
rs7210742	78912010	440/374	39.8 and 40.9 (–0.059 to 0.036)	0.47	.79
rs1468032	78915387	424/360	49.1 and 48.4 (–0.042 to 0.057)	0.15	.93
rs2292639	78915955	440/374	51.0 and 49.2 (–0.030 to 0.067)	0.64	.73
rs6565500	78916499	440/374	37.7 and 39.4 (–0.065 to 0.030)	1.26	.53
rs4969314	78923007	379/342	13.5 and 13.9 (–0.040 to 0.031)	1.73	.42
rs1468035	78923953	440/374	38.1 and 39.6 (–0.063 to 0.032)	0.47	.79
rs12951778	78930344	440/374	25.1 and 26.3 (–0.055 to 0.030)	0.02	.99
rs3751934	78938498	440/374	44.5 and 43.0 (–0.033 to 0.063)	0.72	.70
rs3751932	78939414	440/374	12.8 and 13.5 (–0.040 to 0.026)	1.86	.39
rs4969322	78944104	440/373	53.3 and 49.5 (–0.010 to 0.087)	2.30	.32
rs7502124	78945137	434/372	24.2 and 26.6 (–0.067 to 0.018)	1.28	.53
rs11653897	78951177	440/374	22.3 and 23.1 (–0.049 to 0.032)	0.48	.79

Notes: There were 424 SNPs with minor allele frequency ≥0.10 in the region studied. Eight SNP assays failed, leaving 61 out of 69 tagSNPs. We captured 406 of 424 alleles at r ² ≥ .8, ie, we captured 96% of alleles, and mean of maximum r ² was .92 for the 61 tagSNPs. CI = confidence interval; SNP = single-nucleotide polymorphism.

Phenotype–Genotype Analyses

Phenotype–genotype comparisons for each SNP with 40 different aging-related phenotypes in cases showed weak associations for 5 of 6 SNPs in MTOR, 38 of 61 in RPTOR, and 5 of 7 in RICTOR prior to correction for multiple testing (Supplementary Table S2). In all, 83 genotypic associations were observed with blood pressure (BP) and related clinical phenotypes in the case of 23 of the RPTOR SNPs. To complement the case–control analysis, multivariate tests for association of quantitative phenotypes with SNP allele counts (0, 1, and 2) were conducted for phenotypes determined at Exam 1 and Exam 4 using canonical correlation. For BP, the mean of Exam 1 and Exam 2 data (substituting Exam 3 BP values if Exam 2 values were missing) were used. Variables were transformed when necessary to reduce skewness of the residuals from the regression of each phenotype on age and case–control status. Wilk’s lambda test for association between the Exam 1 variables—log values of systolic and diastolic BP (log(SBP) and log(DBP)), serum cholesterol, body mass index, and log of subscapular skinfold thickness, adjusted for age at Exam 1—and case–control status was not significant (p = .29). Similarly, for the Exam 4 variables—grip strength, waist:hip ratio, body mass index, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, fasting serum glucose, log fasting insulin, log triglyceride level, and log of (101 − CASI [cognitive abilities screening instrument] score), adjusted for age at Exam 4 and case–control status—Wilk’s test was not significant (p = .22). Although we cannot rule out associations between particular age-related phenotypes and individual SNPs, the findings suggest that, overall, the actual number of associations observed did not exceed the expected numbers of associations that might be observed by chance alone.

Discussion

The present nested case–control study found no association of genetic variation in mTOR pathway genes MTOR, RPTOR, RICTOR, and RPS6KA1 with longevity (survival to age 95 years or older) in American men of Japanese ancestry. SNPs with at least moderate effects on longevity will create an excess of high chi-square test values. Instead, we noted a deficiency in such values (Tables 2–4). There were 81 ${\chi }_{\text{2}df}^2$ tests for association between genetic variants and case–control status. The quantiles of ${\chi }_{\text{2}df}^2$ for the upper 10, 5, and 1 percentiles are 4.61, 5.99, and 9.21, respectively; the percent of chi-square values that fell beyond these limits were actually less than expected: 8.6%, 1.2%, and 0%. For minor allele frequencies of 0.1–0.4 and a nominal significance level of 0.05, the study had 90% power to detect odds ratios ranging from 1.58–1.95 and 99% power to detect odds ratios of 1.83–2.36. Therefore, although almost all SNPs with odds ratios of 2.0 and greater are expected to be significant at the .05 level in our study, there were actually slightly fewer significant results than expected if genetic variants had no effect.

Table 4.

Frequency of Minor Allele of Each Tagging SNP of RICTOR on Chromosome 5p13.1 in Long-Lived Case vs Control Participants

SNP ID	Position	Total Number of Cases and Controls	Minor Allele Frequency of Cases and Controls, % (95% CI)	χ2	p Value
rs10941413	38938664	438/373	23.3 and 22.5 (–0.033 to 0.049)	0.27	.87
rs10461998	38984797	440/374	46.0 and 45.6 (–0.044 to 0.053)	0.40	.81
rs1239256	38987371	440/374	34.7 and 35.6 (–0.056 to 0.037)	5.84	.05
rs6878291	39030707	438/374	38.6 and 39.7 (–0.059 to 0.036)	4.64	.10
rs7703002	39045342	440/374	13.5 and 14.0 (–0.039 to 0.028)	0.34	.84
rs12233987	39053547	435/369	48.4 and 48.9 (–0.054 to 0.044)	0.93	.63
rs17364997	39054742	440/374	22.7 and 23.0 (–0.044 to 0.038)	2.16	.34

Notes: We captured 43 of 43 alleles at r ² ≥ .8, ie, we captured 100% of alleles and mean of maximum r ² was .91 for the seven tagSNPs. CI = confidence interval; SNP = single-nucleotide polymorphism.

Because the tagging SNPs we selected provided maximal coverage of genetic variability across each gene plus 5kb of 5′ and 5kb of 3′ DNA, taking advantage of linkage disequilibrium, our data appear to exclude a major effect of common genetic variation in these genes in the etiology of achieving extreme old age, at least in the specific population tested. The population used for the present study has proven capable of detecting genetic association with extreme old age in the case of FOXO3 (5), thereby demonstrating it has the capacity to detect an association should one exist, at least for common variants (minor allele frequency ≥10%) as used in the present study. Nevertheless, the possibility remains for the existence of one or more polymorphisms not in linkage disequilibrium with the SNPs tested or of population-specific variants that could be associated with longevity not present in the current population.

To date, only limited association findings have been validated for genetic variants and human longevity (34–36). The two genes that have received widespread replication are APOE and FOXO3, each having been associated with human longevity in more than 10 independent populations and under different study conditions (37). Several other genes have had several replications, and a limited number of others have had one to two replications (37). This paucity of data reflects, in part, the difficulty of obtaining robust populations of suitable study participants who have attained extreme old age and the complexity of human longevity as a phenotype. Therefore, any data, whether positive or negative, is instructive in deciphering the molecular genetic etiology of human longevity.

The only other comprehensive genetic study of MTOR, RPTOR, RICTOR, and RPS6KA1 in human longevity involved a population of 417 elderly subjects in the Netherlands (38). Genotype data for SNPs in these genes failed to find any association with longevity. However, this study cannot be directly compared with ours for the following reasons: (i) The tagSNPs selected did not provide complete coverage of every gene. (ii) Both the cases (aged >89 years) and controls (aged 55–60 years) were much younger. The relative “youth” of their elderly cohort may explain why no association of variation in FOXO3 was found with longevity in this cohort (38,39), whereas in our cohort, with participants aged older than 95 years (5), a highly significant association was found. The latter finding has been replicated consistently in geographically diverse populations of long-lived cases (generally nonagenarian and centenarians) worldwide (37). The inability to detect an association may be due to the fact that some longevity-associated alleles become markedly more enriched as the age of longevity cases becomes greater. Although the difference between age 89 years and age 95 years may seem small, only 32% of typical 89-year-old people achieve the age of 95 years and thus the latter age is a much more extreme phenotype (40). This has important consequences for studies of human longevity. For example, the minor allele frequency of the longevity-associated allele of FOXO3 increases by 42% from the octogenarian years to nonagenarian years (5). Allele frequency is a major determinant of statistical power in genetic studies. Thus, the Dutch cohort likely has significantly less power to detect genetic variants responsible for extreme old age than the current study. (iii) Cases and controls were from different cities, although the authors claim population substructure differences were not detected. (iv) Participants were genotyped at different times at different centers using different Illumina platforms (38,39,41), which could introduce artifacts into the findings. (v) The Dutch group compared expression of 40 genes in the mTOR pathway between their long-lived subjects and controls (38). Although there was no difference in expression of the longevity-associated gene FOXO3, nor of MTOR, RICTOR, and RPS6KA1, they did find significantly lower expression of RPTOR in both long-lived cases and their offspring. Six other genes also exhibited either higher or lower expression in the long-lived group of subjects. Although it might be possible that some of the expression differences could be caused by polymorphisms in, say, the promoter of these seven genes, they could instead be downstream consequences of genetic or other influences on transcription or posttranscriptional mechanisms of the seven genes.

Although we did not find variation in genotype to be associated with human longevity, we hypothesized that there may be association with aging-related phenotypes. After correction for multiple testing, no significant association was, however, found between the various genotypes and 40 aging-related biologic, clinical, and functional phenotypes.

Certain features of our case and control groups merit comment. Because we only studied men, it remains to be seen whether genotypes of any of the SNPs tested could be associated with longevity or other aging-related phenotypes in women. In addition, baseline examination findings suggested that the longevity case participants were healthier than controls despite the fact that cases were on average 10.8 years older. The cases possessed significantly better biological and physiological risk factor profiles, less age-related disease, and better physical and cognitive function (Table 1). This lends additional credence to the concept of a “healthy aging” phenotype, whereby certain individuals are somehow able to delay or avoid major clinical disease and disability until late in life.

Our long-lived cases also had metabolic profiles that suggested higher insulin sensitivity at younger ages, namely lower waist:hip ratio, lower glucose levels, lower insulin levels, and lower homeostasis model assessment values. This profile clearly has at least some genetic basis because protective variants in FOXO3 were associated with phenotypes that reflect insulin sensitivity in previous work (5,42). Although there was bias in the selection of controls because only their deaths qualified them for control status, this was unavoidable.

An important advantage of the present study is the nested case–control study design, which selects cases and controls from an ongoing cohort study, with longitudinally collected data. This has advantages relative to a regular case–control study in that several phenotypes of interest, such as biological and physiological phenotypes, disease prevalence, and functional status, were obtained by direct clinical examination when the participants were younger. This eliminated recall bias, namely, the differential accuracy of recall between cases and controls, because the information was collected from our subjects before case–control status was determined. In contrast, findings from studies of exceptional survivors, such as centenarians, in which evidence for phenotypes suggestive of slower aging have been reported (43–45), could have suffered from recall bias; older participants may not have recalled their past medical history precisely, nor their functional status when they were younger, or may have learnt of their condition before self-reporting their exposure history to researchers. In the present study, major diseases were, moreover, adjudicated by a morbidity and mortality committee that used performance-based measures of physical and cognitive function to supplement self-reported information.

There are several other strengths to our study. First, the candidate genes selected for analysis were chosen a priori based on hypothesis-driven criteria. That is, studies of model organisms of aging using various methods, such as knockouts, have shown that the mTOR pathway is important for aging and longevity (46). The mTOR pathway and many of its functions appear to be evolutionarily conserved. Second, not only did the study involve a nested case–control design, but it had a high event rate (deaths) during a long period of follow-up, which enhances statistical power. Third, the HHP cohort is a highly homogeneous group of individuals and no population stratification was detected in our study participants (5).

A possible weakness was that cases and controls had an average age difference of 10.8 years, so we cannot completely exclude the possibility of confounding by birth cohort effects. There are several reasons why this is not likely: (i) There was only a 19-year difference in birth year from oldest to youngest participants; (ii) Subgroup analyses showed no differences in socioeconomic status (education and occupation) between cases and controls; and (iii) Contrary to typical cohort studies in which younger cohorts have longer survival, older participants at baseline were the most likely to have lived 95 and more years. Because our study was conducted in only one population, it should be replicated in other populations. Further research should also be directed at effects of the genetic variants with aging of specific organs.

Although our present study does not support an association between the SNPs in the mTOR pathway genes examined and longevity, it remains possible that variants having a minor allele frequency less than 0.10 or variants in other genes in this pathway could have significant effect(s) on longevity or the health or rate of aging of specific organs or both. This is particularly relevant to a genetic study of the rare phenotype of attainment of extreme old age (>95 years of age in our study). As a result it is possible that at least some of the genetic variants that contribute to this phenotype are themselves rare or were below the minor allele frequency of 10% threshold we set for selection of SNPs for testing. Further research is needed to examine the association of uncommon and rare variants of mTOR pathway genes with longevity. Indeed, disruption of mTOR signaling has been associated with sarcopenia (47), and there is evidence that rapamycin treatment can reverse age-related heart dysfunction (48). Furthermore, gene expression analysis in calorically restricted mice indicates a potential neuroprotective role for the mTOR signaling pathway (49). Moreover, as mentioned earlier, there are significant differences in the expression of mTOR pathway genes in older ages (38).

In the current era of genome-wide association studies, genome-wide expression studies, and next-generation sequencing, one might question the relevance of candidate gene and pathway analyses such as that adopted in the present study. Although some headway has been made using data from published genome-wide association studies for 125 diseases and risk factors relevant to mortality in order to obtain a genetic score for time to death (50), finding actual genetic variants would require very large numbers of subjects of extreme old age in order for such studies to be sufficiently powered to generate a result that is significant at the genome-wide level. Although that particular study confirmed that longevity is a complex and highly polygenic trait, it also suggested that it is not easily explainable by common genetic variants related to diseases and physiological traits. In the largest genome-wide linkage analysis for human longevity, involving 2,118 white nonagenarian sibling pairs in 15 study centers of 11 European countries, 4 regions with significant logarithm of the odds of linkage (Lod) scores were identified (2). Fine mapping and a fixed-effect meta-analysis approach showed that APOEε4 and APOEε2 alleles at the TOMM40, APOE, and APOC1 locus explained the linkage at 19q13.3–q13.32, confirming the role of APOE as a longevity gene. The authors concluded that linkage results for the other three loci were not explained by common variants, suggesting that rare variants play an important role in longevity. Thus, although case–control studies, such as ours involving much older (>95 years) subjects than these, would seem to have value because longevity genes would be more concentrated in the extremely old, a major limitation is the difficulty of studies such as ours in obtaining sufficiently high numbers of subjects older than 95 years of age to test polymorphisms other than common variants, such as those we tested.

In conclusion, the present study has found no association of common genetic variation in MTOR, RPTOR, RICTOR, and RPS6KA1 with living to age 95 or older in American men of Japanese ancestry. Further studies of these and other TOR pathway genes in both sexes and in other populations may be warranted.

Supplementary Material

Supplementary material can be found at: http://biomedgerontology.oxfordjournals.org/

Funding

This research was supported by the National Heart, Lung, and Blood Institute (contract NO1-HC-05102), the National Institute on Aging (contract NO1-AG-4-2149; and grants U01-AG-019349, R01-AG-038707, and R01-AG-027060), and the Hawaii Community Foundation (grant 2004-0463).

Conflict of Interest

The authors declared no conflict of interest.