Identification and characterization of two functional variants in the human longevity gene FOXO3

Abstract

FOXO3 is consistently annotated as a human longevity gene. However, functional variants and underlying mechanisms for the association remain unknown. Here, we perform resequencing of the FOXO3 locus and single-nucleotide variant (SNV) genotyping in three European populations. We find two FOXO3 SNVs, rs12206094 and rs4946935, to be most significantly associated with longevity and further characterize them functionally. We experimentally validate the in silico predicted allele-dependent binding of transcription factors (CTCF, SRF) to the SNVs. Specifically, in luciferase reporter assays, the longevity alleles of both variants show considerable enhancer activities that are reversed by IGF-1 treatment. An eQTL database search reveals that the alleles are also associated with higher FOXO3 mRNA expression in various human tissues, which is in line with observations in long-lived model organisms. In summary, we present experimental evidence for a functional link between common intronic variants in FOXO3 and human longevity.

Fig. 1

Association plot shows all 205 SNVs tested in the FOXO3 gene region on chromosome 6 (hg18). The two SNVs with the best P values are indicated (red circles). The four exons are shown in the gene model. Number of individuals: 594 German long-lived individuals (≥100 years) and 918 controls (60–75 years). P_CCA, allelic P values, determined in a case–control association study by chi-squared test with one degree of freedom; −log₁₀P_CCA, negative decadic logarithm of allelic P values

Fig. 2

SRF binds to the longevity allele A but not to the major allele G of rs4946935. Nuclear extracts from Panc1 a and Jurkat b cells were submitted to EMSA with the indicated oligonucleotides. Supershift experiments were performed with antibodies as listed. The position of the supershifted complex is indicated by an arrow. One biological replicate of four is shown. Gt goat, IgG immunoglobulin G, NC nonspecific control (NC I: anti-STAT5A, rabbit; NC II: anti-PDX1, rabbit; NC III, anti-PDX1, goat); PDX1 pancreatic and duodenal homeobox 1, rb rabbit, SRF serum response factor

Fig. 3

FOXO3 SNVs rs12206094 and rs4946935 influence luciferase promoter activity, and knockdown of SRF links this transcription factor to rs4946935. a, b Luciferase promoter assays were perfomed in Panc1 cells transfected with allele-specific constructs for rs12206094 (a) and rs4946935 (b) in both the presence and absence of FCS. For both SNVs, the promoter activity in cells transfected with constructs containing the respective major allele was set = 1 (black dot). Each white dot represents one independent experiment (n = 5). The tables below the figures show the median as well as the minimum and maximum values for the ratio of the longevity allele to the respective alternative allele, taking into account all experiments. c, d Knockdown of transcription factors was done using specific siRNAs. The luciferase activity in cells treated with control siRNA (con siRNA) instead of the transcription factorspecific siRNA was set = 1 (black dot) for each allele of each SNV. Each white dot represents one independent experiment (n = 5). The tables below the figures show the median as well as the minimum and maximum values for the ratios of the activity in presence of con siRNA and transcription factor-specific siRNA for each allele, taking into account all experiments. a–d For determination of specific luciferase activity, activity of the firefly luciferase was normalized to the activity of the renilla luciferase. P ≤ 0.05 was considered statistically significant (two-sided Wilcoxon signed-rank test for paired data). A.U., arbitrary luminescence units

Fig. 4

IGF-1 treatment reduces luciferase promoter activity in the presence of the longevity alleles of rs12206094 and rs4946935, respectively. a, b Luciferase promoter assays were perfomed in Panc1 cells transfected with allele-specific constructs for rs12206094 (a) and rs4946935 (b). The activity of cells treated with DMSO as control instead of IGF-1 was set = 1 (black dot) for each allele of each SNV. Each white dot represents one independent experiment (n = 6). Final DMSO concentrations did not exceed 0.1%. The tables below the figures show the median as well as the minimum and maximum values for the ratios of the activity in presence of DMSO and 100 ng/mL IGF-1 for each allele, taking into account all experiments. For determination of specific luciferase activity, activity of the firefly luciferase was normalized to the activity of the renilla luciferase. P ≤ 0.05 was considered statistically significant (two-sided Wilcoxon signed-rank test for paired data). A.U., arbitrary luminescence units