Lipid metabolism in long-lived families: the Leiden Longevity Study

Introduction

Over the last century, life expectancy has dramatically increased in western societies (Oeppen and Vaupel 2002); nevertheless, inter-individual differences in life expectancy remain. For example, family members of nonagenarian siblings and centenarians have a 30% survival benefit for three generations compared with their birth cohort (Perls et al. 2007; Schoenmaker et al. 2006). Offspring of nonagenarians siblings and centenarians are characterized by a lower prevalence of several major diseases, including diabetes, hypertension, coronary artery disease (CAD) (Adams et al. 2008; Westendorp et al. 2009), and strokes (Terry et al. 2004), implying that these differences in life expectancy are associated with a familial, possible genetic, factor that influences the risk of metabolic diseases.

The mechanisms underlying the variation in human life expectancy are still largely unknown but various studies suggest involvement of lipid metabolism. Specifically, low plasma apoE levels in elderly apolipoprotein E gene (APOE) ε3/ε3 carriers are associated with a decreased mortality risk (Mooijaart et al. 2006) and in comparison with same-aged controls, offspring of centenarians or nonagenarian siblings had a tendency for higher plasma levels of high-density lipoprotein cholesterol (HDL-C), lower levels of low-density lipoprotein cholesterol (LDL-C) (Barzilai et al. 2001), and larger lipoprotein particle sizes (Barzilai et al. 2003; Heijmans et al. 2006). Studies into longevity performed so far, often analyzed a subset of blood lipid parameters only and were relatively small. Their findings thus require confirmation in larger studies, particularly because they could not account for the correlation between the various lipid parameters. The importance of the latter is illustrated by the disappearance of the association of HDL particle size with coronary artery disease and familial longevity after adjustment for triglyceride and/or apolipoprotein B levels (Barzilai et al. 2003; El Harchaoui et al. 2009; Heijmans et al. 2006).

To improve the understanding of the association between lipid metabolism and human longevity, we performed multiple analyses of classical blood lipid parameters, lipoprotein particle sizes and apoE levels in long-lived families from the Leiden Longevity Study. We compared 1,664 offspring of nonagenarian sibships, who were shown to have a life-long survival advantage (Schoenmaker et al. 2006) and a decreased incidence of metabolic and cardiovascular disease (Westendorp et al. 2009), with 711 similarly aged partners as controls. Since men and women differ in lipid metabolism and life expectancy (Freedman et al. 2004), we also stratified for gender in our analysis.

Materials and methods

Results

To test the association of blood lipid parameters with human longevity, we studied individuals for whom complete measurements were available (total cholesterol, HDL-C, LDL-C, triglycerides levels, ApoE, mean HDL, and LDL particle sizes). Compared with controls (n = 711), offspring of long-lived siblings (n = 1,664) had lower triglyceride levels (1.67 vs. 1.75 mmol/L; p = 0.001), larger mean HDL particle sizes (9.05 vs. 9.03 nm; p = 0.01), larger mean LDL particle sizes (21.32 vs. 21.22 nm; p < 10⁻³), and a smaller total cholesterol/HDL-C ratio (4.08 vs. 4.20; p = 0.004). These observations were independent of age, sex, and the interaction between them (Table 1). Further adjustment for self-reported body mass index (BMI), which was available in 619 controls and 1,399 offspring, did not affect these associations (triglyceride levels, p = 0.01; HDL particle size, p = 0.03; LDL particle size, p < 10⁻³; total cholesterol/HDL-C ratio, p = 0.03). As shown in Table 1, no differences were observed for apoE levels for the complete group, even when the analysis was restricted to APOE ε3/ε3 carriers only. We observed that the genotype distribution for the APOE polymorphism differs between offspring and controls (ε2/ε2 = 0.4, ε2/ε3 = 10.2, ε2/ε3 = 3.2, ε3/ε3 = 62.3, ε3/ε4 = 21.3 and ε4/ε4 = 2.5; ε2/ε2 = 1.0, ε2/ε3 = 13.3, ε2/ε3 = 2.3, ε3/ε3 = 60.8, ε3/ε4 = 21.2, and ε4/ε4 = 1.2, respectively, p = 0.019).

Sex-specific analysis revealed that female offspring had lower triglyceride levels (1.51 mmol/L vs. 1.62 mmol/L, p = 0.001), a lower cholesterol/HDL-C ratio (3.75 vs. 3.88, p = 0.018), larger LDL particle sizes (21.60 nm vs. 21.49 nm, p = 0.01), and larger HDL particle sizes (9.25 nm vs. 9.21 nm, p = 0.031) compared with controls. After adjusting for BMI, HDL particle size and the cholesterol/HDL-C ratio were no longer significant different between offspring and controls (data not shown). In males however, offspring only had larger LDL sizes (20.99 nm vs. 20.86, p = 0.023) compared to offspring and for the cholesterol/HDL-C ratio a trend was observed were offspring had a lower ratio compared to controls (4.47 vs. 4.64, p = 0.080). Adjustment for BMI did not affect these results (data not shown).

These univariate analyses do not account for the strong correlations that exist between the various lipid parameters. For example, the Pearson's correlation coefficient between LDL particle size and triglyceride levels is −0.487, and the correlation between HDL-C and LDL particle size is 0.661 (Table 2). To assess the association between longevity and the lipid parameters in more detail and independent of each other, we used multiple logistic regression where the starting model included the all the available lipid parameters, except for total cholesterol and the ratio LDL-C/HDL-C which were not included because they are highly correlated with LDL-C and HDL-C, respectively, which introduces multicollinearity. In this initial model (Table 3), the lipid parameters LDL size (increase in log OR per unit = 0.21; p = 0.017) and triglyceride levels (decrease in log OR per unit = 0.26; p = 0.026) were significantly different between offspring and controls independent from the other lipid variables. After backwards regression, were age, sex, and the interaction between age and sex were not removed to account for the study design, the final model included LDL particle size (increase in log OR per unit = 0.16; p = 0.014) and triglyceride levels (decrease in log OR per unit = 0.21; p = 0.05).

Sex-specific multiple regression indicated differences between men and women. In the initial multivariate logistic regression model, none of the lipid parameters was significantly different in male offspring compared with controls, but for LDL size a trend was observed (p = 0.055) (Table 3). After backwards regression, male offspring had a larger LDL particle size than controls (increase in log OR per unit = 0.21; p = 0.023), whereas no differences in triglyceride levels were observed. For female offspring, in the initial starting model, only triglyceride levels, were lower among female offspring (decrease in log OR per unit = 0.42; p = 0.011) (Table 3). After backwards regression, no difference in LDL size were observed in female offspring as compared with controls, whereas triglyceride levels were significantly lower among female offspring (decrease in log OR per unit = 0.22; p = 0.01).

The valine allele of rs5882 in the cholesterol ester transfer protein gene was previously found to be associated with larger LDL and HDL particle sizes (Barzilai et al. 2003). The association of the valine allele with larger LDL size was replicated in a linear regression analysis in controls and offspring combined (p = 0.008; adjusted for age, sex, and triglyceride levels), but this allele was not associated with HDL size (p = 0.200). The polymorphism, however, did not explain the difference in LDL size between controls and offspring (increase in log OR per unit = 0.16; p = 0.009 after accounting for rs5882, as compared with increase in log OR per unit = 0.17; p = 0.014 without rs5882). Nor was the valine allele associated with the offspring group (p = 0.131).Similar results were obtained for man and women separately.

Discussion

We investigated the association of classical serum lipid levels, apoE levels and lipoprotein particle sizes with human longevity, in a study that was sufficiently large to account for correlations among lipid parameters and to test for differences between sexes. Our main finding is that both LDL particle size and triglyceride levels are associated with human longevity independent of other lipid parameters. Further analysis indicated that LDL particle size is particularly associated with male, whereas triglyceride levels are associated with female longevity.

Our study (N = 2,375) confirms the association of LDL particle size previously observed in a smaller study (N = 474) (Barzilai et al. 2003). The differences in the final end model were comparable with the findings from Barzilai for males (0.21 nm increase), but not for females (in the final model only triglyceride levels were significantly different between offspring and controls). Remarkable, an association of triglyceride levels with longevity was not reported before. In one study, only a trend was observed (Heijmans et al. 2006) and in another study triglycerides were not mentioned in relation to longevity (Barzilai et al. 2003). In a very small study where fasting samples were used, no association between longevity and triglyceride levels were found (Barzilai et al. 2001). Since we used non-fasting samples and had a large study population, this could explain why we did found an association with triglyceride levels and longevity. Our results did not confirm the previously reported association of LDL and HDL cholesterol levels (Barzilai et al. 2001) and HDL particle size (Barzilai et al. 2003). For HDL particle size we did observe an association in a univariate analysis, but a multiple analysis that accounted for the correlation between lipid parameters showed that this result depended on other blood lipid parameters in particular triglyceride levels (Heijmans et al. 2006).

Because the APOE genotype was previously found to confound the association between ApoE level and all-cause mortality in elderly subjects (Mooijaart et al. 2006), it was necessary to stratify the data according to APOE genotype. Still we did not find an association of apoE level with longevity. However, we did find that the beneficial APOEε2 allele was more common in the offspring compared with controls, which might explain why offspring have a more beneficial lipid profile compared with controls (Mooijaart et al. 2006).

Since offspring of nonagenarian siblings from the Leiden Longevity Study have a lower prevalence of late-onset metabolic diseases (Westendorp et al. 2009), it could be hypothesized that low triglyceride levels and a larger mean LDL particle size are protective against diseases like diabetes type II and CAD. High triglyceride levels are a component of the metabolic syndrome, which is associated with a higher risk for cardiovascular diseases (CVD) and type II diabetes (Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults 2001). Triglyceride levels itself are also independently associated with a higher risk for CVD (Sarwar et al. 2007). Elevated triglyceride levels are associated with increased levels of remnants lipoproteins, which might initiate and promote the development of atherosclerotic plaques (Nordestgaard et al. 2007). Mean LDL particle size is determined by summing the size-weighted relative contribution made by each subspecies (i.e., small, medium and large LDL particles) to the total concentration of LDL particles (Otvos et al. 1992). Thus, people with a small mean LDL particle size have relatively more small LDL particles as compared with people with a large mean LDL particle sizes. Small LDL particle sizes are associated with type II diabetes (Festa et al. 2005), metabolic syndrome (Gentile et al. 2008; Kathiresan et al. 2006), insulin resistance (Garvey et al. 2003; Goff et al. 2005), incident CVD (Mora et al. 2009), and future CAD (El Harchaoui et al. 2007). Possible atherosclerosis could be induced by those small LDL particles, as indicated in a study of two knockout mice strains with similar cholesterol levels, were the Ldr^−/−Apob^100/100 mice had a larger number of smaller lipoprotein proteins and more atherosclerosis than the Apoe^−/−Apob^100/100 mice (Veniant et al. 2008). Thus from literature it is clear that LDL particle size and triglyceride levels have been contributing to cardiovascular diseases.

LDL particle size seems to be particularly associated with male longevity whereas triglyceride levels are associated with female longevity, indicating that the role of lipid metabolism in longevity may be different in men and women. In women it was found that LDL particle size associated with incident CVD, however in that study LDL particle size was comparable but not superior to standard lipid levels in predicting CVD (Mora et al. 2009). This could support our finding that triglyceride levels eliminated the association between LDL size and longevity in women. Moreover, in a meta-analysis of population based prospective studies it was concluded that high triglyceride levels are a larger risk factor for cardiovascular disease in women than in men (Hokanson and Austin 1996). But in a more recent meta-analysis, it was found that the effect of triglyceride levels on cardiovascular disease did not differ between men and women (Sarwar et al. 2007). On the other hand, men are characterized by smaller and denser LDL particles and higher triglyceride levels than women (Freedman et al. 2004; Lemieux et al. 2002), which is also the case in our study. In addition, men have higher hepatic lipase activity compared with women, resulting in a smaller mean LDL particle size (Carr et al. 2001; Magkos et al. 2009). Thus different plasma concentrations of enzymes involved in intravascular lipid remodeling between men and women (Magkos et al. 2009) may explain our observed sex differences regarding LDL particle size and familial longevity. But further research on lipid enzymes difference between men and women in relation to longevity is needed to confirm this hypothesis.

This study has several strong points. Firstly, due to the large sample size it was possible to adjust for the correlations among the lipid parameters and to stratify for gender. Secondly, since controls are the partners of the offspring of long-lived siblings, they did not differ on any major indicators of lifestyle, including current smoking, self-reported BMI, and level of education (Westendorp et al. 2009). A weak point is that the NMR spectroscopy-based lipid parameters were not determined in a single measurement which introduced batch effects, for which it was necessary to adjust during analysis. We choose not to correct for the treatment of high cholesterol levels with lipid lowering medication, which mostly consist of 3-hydroxy-3-methylglutaryl-coenzyme-A reductase inhibitor (statin) treatment (93.8%), because per individual it is unknown if statin treatment reduces moderate to severe hypercholesterolemia to above or below population average lipid levels. Since the offspring group is using lesser lipid lowering medication than the controls, this might bias our results in such a way that we underestimate the effect of lipid metabolism on longevity. Also, the use of non-fasting samples introduced noise in our data-set, which may lead to an underestimation of the effects found our study.

Our study confirms the link between lipid metabolism and human longevity and identifies LDL particle size and triglyceride levels as the key lipid parameters. To gain more insight into why there are differences between males and females, it is necessary to include information on lipid enzymes, genetic and expression data of genes known to be involved in lipid metabolism in future studies investigating the relation between lipid metabolism and longevity in humans.