Heritability and responses to high fat diet of plasma lipidomics in a twin study

Abstract

Lipidomics have a great potential as clinical tool for monitoring metabolic changes in health and disease. Nevertheless hardly anything is known about the heritability of lipids. Therefore, it is necessary to clarify how and how much we can affect these progresses in individuals. In our interventional twin study (46 healthy, non-obese twin pairs) we investigated the lipid profile in plasma samples after switching from a low fat diet to an isocaloric high fat diet (HFD) to characterize the metabolic adaptation. Additionally we used the ACE model for Additive genetics, Common and unique Environment as well as linear mixed modelling to analyse the heritability of lipids. The heritability of lipids varied between 0-62% and applied to lipid species rather than to lipid classes. Phospholipids showed the highest inheritance. In addition, sex, body mass index (BMI) and age were important modifiers. The lipid profile changed already after one week of HFD and diverged further after 5 weeks of additional HFD. Basal concentrations of specific lipids within phospholipids are strongly inherited and are likely to be associated with heritable disease risks. BMI, sex and age were major modifiers. Nutrition strongly alters specific lipid classes, and has to be controlled in clinical association studies.

Figure 1

Changes in plasma concentration of the different lipid classes. After LF the classes reacted differently with a monotonous reaction, a time dependent increase or decrease or a counter regulation. The stable lipid classes (TAG and LPC, Supplementary Figure S1) are not shown here. (rep.M. ANOVA + Bonferroni adjustment; *p < 0.05, **p < 0.01, ***p < 0.001, mean ± SEM).

Figure 2

Effects of BMI and age on plasma concentrations of lipid classes. The mesh plots show the interaction of BMI and age based on the fixed factors of the linear mixed models with the plasma concentration of the lipid class on the z axis. The axes ranged from 18 to 80 years (age) and 17.5 to 30 kg/m² (BMI). Each mesh covers 4.13 years and 0.83 kg/m². Not shown here are classes, which were not influenced by at least one of the factors.

Figure 3

Effects of BMI, age and sex on plasma concentrations of lipid species. The 3D bubble plot shows the interaction of BMI, age and sex based on the fixed factors of the linear mixed models with young women with low BMI being the reference. The size of a bubble represents the median concentration at LF. The numbering of the species has been restricted to those with a deviation from the principal diagonal. The colours also represent the different lipid species. The deviation of the bubbles from the principal diagonal represents the effect of the respective species on the parameter where the bubble is deviated to. The species were sorted alphabetically with rising chain length and the plot is shown here in three perspectives. (See also Media S1).

Figure 4

Percentage of variance contribution determined by a linear mixed model based on the repeated measured ACE model (LF, HF1, HF6) for lipid classes (A) and lipid species (B). The parameters were sorted/ranked by the rounded values for heritability (colour red), common environment (colour orange), individual environment (colour blue) and unknown effects (colour white). (See also Tables S1 and S2).

Figure 5

Study design. The participants performed two isocaloric diet periods. During the first six weeks they received a low-fat, carbohydrate-rich diet as standardization for their eating behaviours. After a first clinical investigation day (LF) the participants switched to a high-fat, low-carbohydrate diet and underwent two more CIDs after one (HF1) and additional five weeks (HF6) on the high-fat diet. During the week before a clinical investigation day the participants received detailed meal plans for each day to ensure a standardized dietary pattern (STP) and at least every two weeks nutritional counselling (NC) took place.