Buffering mechanisms in aging: a systems approach toward uncovering the genetic component of aging

Abstract

An unrealized potential to understand the genetic basis of aging in humans, is to consider the immense survival advantage of the rare individuals who live 100 years or more. The Longevity Gene Study was initiated in 1998 at the Albert Einstein College of Medicine to investigate longevity genes in a selected population: the "oldest old" Ashkenazi Jews, 95 years of age and older, and their children. The study proved the principle that some of these subjects are endowed with longevity-promoting genotypes. Here we reason that some of the favorable genotypes act as mechanisms that buffer the deleterious effect of age-related disease genes. As a result, the frequency of deleterious genotypes may increase among individuals with extreme lifespan because their protective genotype allows disease-related genes to accumulate. Thus, studies of genotypic frequencies among different age groups can elucidate the genetic determinants and pathways responsible for longevity. Borrowing from evolutionary theory, we present arguments regarding the differential survival via buffering mechanisms and their target age-related disease genes in searching for aging and longevity genes. Using more than 1,200 subjects between the sixth and eleventh decades of life (at least 140 subjects in each group), we corroborate our hypotheses experimentally. We study 66 common allelic site polymorphism in 36 candidate genes on the basis of their phenotype. Among them we have identified a candidate-buffering mechanism and its candidate age-related disease gene target. Previously, the beneficial effect of an advantageous cholesteryl ester transfer protein (CETP-VV) genotype on lipoprotein particle size in association with decreased metabolic and cardiovascular diseases, as well as with better cognitive function, have been demonstrated. We report an additional advantageous effect of the CETP-VV (favorable) genotype in neutralizing the deleterious effects of the lipoprotein(a) (LPA) gene. Finally, using literature-based interaction discovery methods, we use the set of longevity genes, buffering genes, and their age-related target disease genes to construct the underlying subnetwork of interacting genes that is expected to be responsible for longevity. Genome wide, high-throughput hypothesis-free analyses are currently being utilized to elucidate unknown genetic pathways in many model organisms, linking observed phenotypes to their underlying genetic mechanisms. The longevity phenotype and its genetic mechanisms, such as our buffering hypothesis, are similar; thus, the experimental corroboration of our hypothesis provides a proof of concept for the utility of high-throughput methods for elucidating such mechanisms. It also provides a framework for developing strategies to prevent some age-related diseases by intervention at the appropriate level.

Figure 1. Genotypic Frequency Comparison between Control and Proband

Genotypic frequencies of SNPs associated with some of the genes implicated in CVD. Comparison is between control individuals (∼70 years old), and probands (∼100 years old). Offspring of centenarians are excluded from this analysis. Significant change, p < 0.0066 (after applying Bonferroni correction), was found for two genes, CETP and APOC-3. Most genotypes exhibit no change in frequency between the two groups though they may still be factors contributing to lifespan.

Figure 2. Frequency Trend of Two Buffering Genes

Longevity genes are expected to exhibit monotonic increase in their favorable genotype when sampled in progressively older age groups. The graph shows a highly significant (p < 0.0006) monotonic increased frequency across ages for favorable genotypes in APOC-3 CC, and a significant (p < 0.047) for CETP-VV, fulfilling our definition for candidate longevity genes.

Figure 3. Observed U-Shape Trend of Age-Related Buffered Disease Genes

Longevity genes are hypothesized to buffer the phenotypic effect of certain deleterious age-related disease genotypes, thus allowing the accumulation of the latter in a population endowed with longevity genotypes. Presented here are the frequency trends across ages of deleterious genotypes in KLOTHO and LPA. Frequencies decline until the population age is ∼80 years old, close to the current average lifespan, and then increases to nearly the frequency in younger age, fulfilling our definition of a buffered disease gene. We used a binomial model with identity link function with both linear and quadratic terms for age, and tested for the significance of the quadratic component. We found a statistically significant quadratic component at the level of p < 0.035 for both KLOTHO and LPA.

Figure 4. Interaction between CETP and the Buffered LPA Gene

An interaction between the longevity gene and its target buffered disease gene is revealed by population subdivision. A subpopulation endowed with the favorable longevity genotype will show either no change, or an increase, in the frequency of its target deleterious genotype. In a population lacking the favorable longevity genotype, a monotonic decline (similar to the decline in nonbuffered disease gene) will be observed. Presented here are the frequency trends across ages of deleterious genotype in LPA in the subpopulation having favorable longevity genotype CETP-VV versus the subpopulation lacking it, i.e., CETP-IV and CETP-II. The two trends show a significant difference (p < 0.037, see Text S1 for statistical considerations).

Figure 5. Interaction between CETP and the Buffered KLOTHO gene

Comparable to Figure 4, we show here the frequency trends across ages of deleterious genotype in KLOTHO in the subpopulation having favorable longevity genotype CETP-VV versus subpopulation lacking it, i.e., CETP-IV and CETP-II. Because of the lack of offspring data, we included centenarians (see text). The two frequency trends follow a similar U-shape with age and show no significant interaction term (p < 0.38). Nonsignificant results have also been observed for the interaction term between APOC-3 and KLOTHO (unpublished data).

Figure 6. Protein–Protein Interaction Network

Protein–Protein interaction networks of subsets of proteins relevant to the buffering hypothesis analysis. Connecting lines between gene symbols indicate interactions; different types of interactions are denoted by symbols on the lines. Green square, regulation; blue square, expression; light green triangle, transport; + in grey circle, positive effect; − in grey circle, negative effect. CETP interacts with LPA through LPAL2, an LPA-like 2 protein. This interaction is also mediated by LPAL2 through APOA1, and apolipoprotein A-I. Also, though our analysis shows no interaction between LPA and APOC-3, pathway analysis shows a direct interaction, suggesting additional and non-overlapping functionality between CETP and APOC-3 (see text).