Acquired liver fat is a key determinant of serum lipid alterations in healthy monozygotic twins

Authors

  • S.M. Kaye,

    Corresponding author
    • Obesity Research Unit, Department of Medicine, Division of Endocrinology, Helsinki University Central Hospital and University of Helsinki, Helsinki, Finland
    Search for more papers by this author
  • M. Maranghi,

    1. Department of Medicine, Helsinki University Central Hospital, Helsinki, Finland
    2. Department of Internal Medicine and Medical Specialties, Sapienza University, Rome, Italy
    Search for more papers by this author
  • L.H. Bogl,

    1. The Finnish Twin Cohort Study, Department of Public Health, University of Helsinki, Helsinki, Finland
    Search for more papers by this author
  • J. Kaprio,

    1. The Finnish Twin Cohort Study, Department of Public Health, University of Helsinki, Helsinki, Finland
    2. Institute for Molecular Medicine FIMM, University of Helsinki, Helsinki, Finland
    3. National Institute for Health and Welfare, Helsinki, Finland
    Search for more papers by this author
  • A. Hakkarainen,

    1. Department of Radiology, University of Helsinki and HUS Medical Imaging Center, Helsinki, Finland
    Search for more papers by this author
  • J. Lundbom,

    1. Department of Radiology, University of Helsinki and HUS Medical Imaging Center, Helsinki, Finland
    Search for more papers by this author
  • N. Lundbom,

    1. Department of Radiology, University of Helsinki and HUS Medical Imaging Center, Helsinki, Finland
    Search for more papers by this author
  • A. Rissanen,

    1. Obesity Research Unit, Department of Psychiatry, Helsinki University Central Hospital, Helsinki, Finland
    Search for more papers by this author
  • M.-R. Taskinen,

    1. Department of Medicine, Helsinki University Central Hospital, Helsinki, Finland
    Search for more papers by this author
  • K.H. Pietiläinen

    1. Obesity Research Unit, Department of Medicine, Division of Endocrinology, Helsinki University Central Hospital and University of Helsinki, Helsinki, Finland
    2. Institute for Molecular Medicine FIMM, University of Helsinki, Helsinki, Finland
    Search for more papers by this author

  • Conflict of Interest: The authors declared no conflict of interests.

  • Funding agencies: This study was supported by research grants from Helsinki University Central Hospital, Academy of Finland (grants numbers 118555 and 141054 to J.K.), Paavo ja Eila Salonen's donation to Central Finland Health Care District and the following Foundations: Novo Nordisk, Diabetes Research, Jalmari and Rauha Ahokas, Biomedicum Helsinki Foundations, Finnish Foundation for Cardiovascular Research, Research Foundation of the University of Helsinki and Doctoral Programs in Public Health, and Academy of Finland Center of Excellence in Complex Disease Genetics (grant numbers: 213506, 129680)

Correspondence: S.M. Kaye (sanna.kaye@hus.fi)

Abstract

Objective: The effects of acquired obesity on lipid profile and lipoprotein composition in rare BMI-discordant monozygotic (MZ) twin pairs were studied.

Design and Methods: Abdominal fat distribution, liver fat (magnetic resonance imaging and spectroscopy), fasting serum lipid profile (ultracentrifugation, gradient gel-electrophoresis, and colorimetric enzymatic methods), and lifestyle factors (questionnaires and diaries) were assessed in 15 BMI-discordant (within-pair difference [Δ] in BMI >3 kg/m2) and nin concordant (ΔBMI <3 kg/m2) MZ twin pairs, identified from two nationwide cohorts of Finnish twins.

Results: Despite a strong similarity of MZ twins in lipid parameters (intra-class correlations 0.42-0.90, P < 0.05), concentrations of apolipoprotein B (ApoB), intermediate-density lipoprotein cholesterol, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein 3a% (HDL3a%), and HDL3c% were higher (P < 0.05) and those of HDL cholesterol, HDL2-C, and HDL2b% were lower (P < 0.01) in the heavier co-twins of BMI-discordant pairs. The composition of lipoprotein particles was similar in the co-twins. When BMI-discordant pairs were further divided into liver fat-discordant and concordant (based on median for Δliver fat, 2.6%), the adverse lipid profile was only seen in those heavy co-twins who also had high liver fat. Conversely, BMI-discordant pairs concordant for liver fat did not differ significantly in lipid parameters. In multivariate analyses controlling for Δsubcutaneous, Δintra-abdominal fat, sex, Δsmoking and Δphysical activity, Δliver fat was the only independent variable explaining the variation in ΔApoB, Δtotal cholesterol, and ΔLDL-C concentration.

Conclusions: Several pro-atherogenic changes in the amounts of lipids but not in the composition of lipoprotein particles were observed in acquired obesity. In particular, accumulation of liver fat was associated with lipid disturbances, independent of genetic effects.

Introduction

Obesity-associated derangements of plasma lipid profile are a widely documented risk factor for cardiovascular disease [1]. Body fat depots differ in their contribution to lipid metabolism. Intra-abdominal (ia) fat rather than subcutaneous (sc) or gynoid body fat is associated with an atherogenic lipid profile and increased cardiovascular risk [2, 3]. Ectopic fat in liver contributes significantly both to serum triglyceride (TG) and high-density lipoprotein cholesterol (HDL-C) levels [4]. The derangement of lipoprotein composition due to liver steatosis is already seen in children presenting higher amounts of large very low-density lipoprotein (VLDL), small dense low-density lipoprotein (LDL), and smaller HDL particles [5]. Characterizing the individuals at highest cardiovascular risk is a challenge. Not all obese individuals present the same harmful metabolic sequelae probably due to a different predisposition to accumulation of ectopic fat [6, 7]. A subgroup of obese individuals remains insulin sensitive and presents with normal TG and HDL-C concentrations [8]. Detailed lipid concentrations and compositions and their relationships to body fat distribution in obesity are poorly characterized.

Derangements in the lipid profile are determined by complex combinations of genetic and lifestyle factors [9]. In our previous studies, we provided evidence for large genetic influences on variations in LDL-C (79%), HDL-C (73%), and TGs (64%), as has been previously reported elsewhere [10]. In large genome-wide association studies, several single nucleotide polymorphisms associate with lipid levels in humans [11]. Moderate heritability estimates were observed for LDL peak particle size (49%) and HDL mean particle size (56%) and HDL subspecies (46-63%) [10], confirming that substantial environmental influences also exist. The sources of those environmental influences are variable and may be difficult to distinguish from genetic and environmental influences. Obesity and regional fat distribution have a substantial heritability [12] and many genes underlying obesity have been identified [13]. BMI-discordant monozygotic (MZ) twins offer a unique model where the influence of acquired obesity on lipid profile can be disentangled from genetic factors, as in this design the lean and heavy subjects are completely matched for any variations in the genetic background (DNA sequence) [14].

We aimed to determine the associations between acquired obesity and serum lipid profile with a special emphasis on fat distribution, sc, ia, and liver fat in rare MZ twin pairs discordant for BMI. In addition, we tested the similarity of lipoprotein particle quantity and quality in MZ twins, to evaluate the genetic similarity within the twin pairs. We also aimed to carefully adjust the confounding effects of lifestyle factors on lipid profile in our analyses.

Methods

Subjects

The twins included in this study were recruited from two population-based, longitudinal studies, FinnTwin16 (FT16) and FinnTwin12 (FT12), each consisting of five consecutive birth cohorts of Finnish twins [15]. In FT16, twins born between 1975 and 1979 were followed up by questionnaires at 16, 17, 18.5, and 22-27 years of age (response rates 83–97%, n = 5,601 at baseline). In FT12, twins born between 1983 and 1987 were sent questionnaires at 11-12, 14, and 17.5 years (response rates 74-92%, n = 5,184 at baseline). This study included an intensive subsample from the last follow-ups of both FT12 and FT16 studies including 15 pairs (6 male and 9 female pairs) discordant for BMI (within-pair difference in BMI ≥3 kg/m2 and 10-30 kg in weight) and 9 pairs (5 male and 4 female pairs) concordant for BMI (within-pair difference in BMI <3 kg/m2). Within-pair difference of ≥3 kg/m2 represents the top 5% most discordant pairs in the cohorts. All pairs were of Caucasian origin with a mean age of 28.1 years (range 22.8-33.1 years). Except for one obese co-twin who had recently developed type 2 diabetes and was on insulin therapy and another obese co-twin having an inactive ulcerative colitis and being on mesalazine and azathioprine treatment, the subjects were healthy and did not take any medications. The twins' weights had been stable for at least 3 months prior to the study. None of the female subjects were pregnant or lactating. One female subject used oral contraceptives. Monozygosity was confirmed by genotyping of 10 informative genetic markers [16]. The study protocols were approved the ethical committee the Hospital District of Helsinki and Uusimaa, Finland. Written informed consent was obtained from all participants.

Diet, smoking, and physical activity

Energy and macronutrient intakes were assessed from 3-day food records and analyzed by Diet32, which is based on a national Finnish database for food composition [17]. Habitual alcohol intake was assessed by a structured questionnaire. The non-smokers comprised never smokers, former and occasional smokers, and the current smokers comprised those who were daily smokers (n = 6 individuals). Physical activity was assessed using the Baecke physical activity questionnaire, which derives an index for physical activity in total (total index) and separate indices for physical activity at work (work index), sports activities during leisure time (sport index), and physical activity during leisure time excluding sports (leisure time index) [18].

Body composition

Weight and height were measured after 12-h overnight fast to calculate BMI. Body composition was measured using whole body dual X-ray absorptiometry (DXA) scans (software version 8.8, Lunar Prodigy, Madison, WI,USA). A standardized procedure at least 4 h after a light meal with empty bladder was used to avoid differences in the hydration status. Whole body fat percentage was calculated as fat mass/(fat mass + lean mass + bone mineral content).

Magnetic resonance experiments

The magnetic resonance (MR) measurements were performed on a clinical 1.5 T imager (Avanto, Siemens, Germany, Erlangen). To allow measurement of abdominal fat distribution, a T1-weighted axial image stack of 16 slices with thickness of 10 mm and gap of 0 mm was centered at L4/L5 intervertebral disk. Selective fat excitation was used to obtain images with a standard body coil. MR images were analyzed using SliceOmatic v4.3 segmentation software (Tomovision, Montreal, Canada). The areas of sc and ia fat tissue were measured for each slice using a region-growing routine. The results were expressed as total volumes of sc fat and ia fat.

For liver-fat analyses, point resolved spectroscopy (PRESS) localization technique with repetition time (TR)/echo time (TE) of 3,000/30 ms and 16 acquisitions was used to obtain non-suppressed liver spectra. Orthogonal three plane images were used for localization of the cubic 8-27 cm3 voxel of interest within the right lobe of the liver avoiding signal contamination from vascular structures, gallbladder, and adipose tissue. The MRS data were collected using a flex surface coil in combination with spine coils. The liver spectra were analyzed with jMRUI v3.0 software [19] using the AMARES algorithm [20]. Areas of water signal at 4.7 parts per million (ppm) and methylene signal from intracellular TGs at 1.3 ppm were determined using a line-fitting procedure. Spectroscopic intracellular TG content was expressed as methylene/(water + methylene) signal area × 100 and the values were further converted to mass fractions as described earlier [21].

Biochemical analyses

Venous blood samples from the study subjects were drawn after an overnight fast and serum and EDTA plasma were separated by centrifugation and stored at -80°C until analysis for lipid profile. Plasma concentrations of apolipoprotein A1 (ApoA1) and apolipoprotein B (ApoB) were measured by immunoturbidometric methods (for ApoA1, Wako Chemicals GmbH and for ApoB, Orion Diagnostica, Espoo, Finland). Serum apolipoprotein C3 (ApoC3) concentration was quantitated by ELISA [22]. Serum total cholesterol (TC) and TGs were determined using an automated Konelab 60i analyzer (Thermo Fisher Scientific Oy) by enzymatic methods (Refs. 981812 and 981301).

Fasting serum lipoproteins (VLDL, intermediate-density lipoprotein [IDL], LDL, HDL2, and HDL3) were separated by sequential flotation ultracentrifugation using a modification of the method of Havel et al. [23]. Lipoprotein compositions were analyzed using enzymatic methods by Konelab 60i. HDL was separated and isolated by ultracentrifugation from 0.5 ml plasma [24]. Distribution of HDL2b, 2a, 3a, 3b, and 3c subspecies and HDL mean particle size were determined by native gradient gel electrophoresis as previously described with minor modifications [24, 25]. The molecular size intervals for HDL subspecies 2b, 2a, 3a, 3b, and 3c were used, and for each subspecies, the relative area under the densitometric scan was reported. Mean HDL particle size was calculated by multiplying the mean size of each HDL subclass by its relative area under the densitometric scan. LDL peak particle diameters determined using 1 mm 2-10% linear non-denaturating polyacrylamide gradient gels [26].

Statistical methods

Statistical analyses were performed with Stata statistical software (release 11.0; Stata Corporation, College Station, TX, USA). Results are expressed as mean ± SE unless otherwise specified. Twin similarity was assessed using sex-adjusted intra-class correlations (ICCs). Mean values of leaner and heavier co-twins are shown unadjusted. Wilcoxon matched-pairs signed-ranks test was performed to test whether heavier and leaner co-twins differed significantly from each other. Within-pair differences (Δ) were calculated by subtracting the leaner co-twin's value from the heavier co-twin's value and using all twin pairs in analyses. Mann–Whitney U-test was used to test whether Δlipids differed in discordant vs. concordant groups. Pearson partial correlations adjusted for sex were used to calculate correlations between Δs of different adiposity measures. The effects of acquired adiposity on serum lipid profiles were calculated using Pearson partial correlations (Δeach adiposity measure vs. Δeach lipid value) adjusted for sex, Δphysical activity (total index), and Δsmoking status. Similarly, the effects of Δphysical activity on Δlipids were tested by Pearson correlation adjusted for sex and Δsmoking. Multiple regression analyses included Δeach lipid value against Δsc fat, Δia fat and Δliver fat, sex, Δphysical activity, and Δsmoking in the same model. Because further adjustment for Δpercentage of energy from macronutrients or alcohol did not substantially change these results, they were not included in the final models.

Results

Physical characteristics

Physical characteristics and the lipid profiles of MZ twin pairs are summarized in Table 1. Briefly, within-pair differences (Δ) in BMI ranged from 3.1 to 9.4 kg/m2 in discordant pairs and from 0.1 to 2.3 kg/m2 in concordant pairs. The heavier co-twins of BMI-discordant pairs had on average 1.6 times the amount of sc fat, 2.0 times that of ia fat, and 5.1 times the amount of liver fat than their leaner counterparts. These parameters did not differ between the heavier and leaner co-twins in BMI-concordant twin pairs. To further analyze the influence of liver fat we divided BMI-discordant twin pairs into two subgroups based on median for Δliver fat (2.6%) in this study sample: seven BMI-discordant pairs (Δweight 16 kg) where both co-twins had low liver fat percentages (from 0.1 to 1.6%) had no differences in liver fat (Δliver fat 0-0.2%, P = 0.40). The remaining eight BMI-discordant pairs (Δweight 17 kg) differed significantly for liver fat: the heavier co-twins' liver fat% ranged from 3.8% to 9.4% (Δliver fat 2.6-9.0%, P = 0.012).

Table 1. Physical characteristics and lipid profiles of monozygotic twin pairs
 BMI-discordant n = 15 pairsBMI-concordant n = 9 pairs
ΔBMI >3 kg/m2ΔBMI <3 kg/m2
LeanerHeavierLeanerHeavier
  1. Data are mean ± SE, non-normally distributed data are median (25–75% inter-quartile range). Wilcoxon signed ranks test leaner vs. heavier co-twin, *P < 0.05, **P < 0.01, ***P < 0.001. ApoA1, apolipoprotein A1; ApoB, apolipoprotein B; ApoC3, apolipoprotein C3; HDL-C, high density lipoprotein cholesterol; ia, intra-abdominal; IDL-C, intermediate density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; sc, subcutaneous; TC, total cholesterol; TG, triglycerides; VLDL-C, very low density lipoprotein cholesterol.
  2. aThe within-pair difference significantly different (P < 0.05) in discordant vs. in concordant pairs (Mann-Whitney U-test).
Adiposity measures
BMI (kg/m2)24.8 ± 1.030.3 ± 1.0***27.6 ± 1.129.3 ± 1.1**
Body fat %32.2 ± 2.440.6 ± 2.0***31.6 ± 3.133.0 ± 2.6
Sc fat (dm3)3.4 ± 0.45.4 ± 0.1***3.6 ± 1.03.9 ± 0.5
Ia fat (dm3)0.6 (0.3–0.8)1.1 (0.7–1.5)***1.1 (0.6–1.5)1.1 (0.6–1.4)
Liver fat (%)0.6 (0.4–1.1)3.8 (0.6–6.1)**1.3 (0.5–1.6)1.0 (0.5–1.7)
Apolipoproteins
ApoA1 (mg/dl)128.9 (119.3–172.5)129.9 (113.0–144.8)126.4 (112.6–130.5)127.6 (118.8–142.6)
ApoB (mg/dl)69.8 ± 3.381.1 ± 4.0**81.0 ± 9.681.1 ± 7.6
ApoC3 (mg/dl)9.5 ± 0.79.4 ± 0.79.4 ± 2.09.7 ± 1.3
Serum TGs and cholesterol
TG (mg/dl)93.8 (74.3–115.1)112.4 (87.6–130.1)83.2 (63.7–201.8)98.2 (74.3–104.4)
TC (mg/dl)174.6 ± 5.5183.7 ± 6.2*173.4 ± 14.5176.6 ± 13.4
VLDL-C (mg/dl)9.4 ± 1.012.4 ± 1.814.1 ± 4.911.3 ± 3.2
LDL-C (mg/dl)98.4 ± 5.8110.4 ± 5.6*104.1 ± 11.1108.4 ± 11.5
IDL-C (mg/dl)a3.8 (2.9–4.3)5.0 (3.1–6.7)**3.7(2.5–5.5)2.6 (2.3–5.4)
HDL-C (mg/dl)a62.7 ± 3.751.2(46.0–61.8)**50.9 ± 3.652.8 ± 3.8
HDL2-C (mg/dl)31.9 ± 3.521.3 (15.4–25.0)**21.1 ± 3.915.3 (14.8–21.5)
HDL3-C (mg/dl)30.8 ± 1.231.4 ± 1.229.7 ± 1.832.4 ± 1.8
Lipoprotein particle mass and size
VLDL mass (mg/dl)95.2 ± 8.8121.6 ± 14.0143.2 ± 44.3123.8 ± 33.1
LDL mass (mg/dl)243.6 ± 11.4254.3 (228.0–316.8)*267.7 ± 24.6276.7 ± 26.7
IDL mass (mg/dl)25.3 ± 2.329.0 ± 2.9*25.3 ± 3.724.8 ± 3.1
HDL mass (mg/dl)393.4 ± 23.1362.5 ± 22.1334.5 ± 15.4351.8 ± 25.6
LDL size (nm)26.6 ± 0.226.3 ± 0.226.1 ± 0.425.9 ± 0.3
HDL size (nm)9.6 ± 0.19.4 ± 0.1**9.2 ± 0.29.1 ± 0.1
HDL subclasses
HDL2a%27.0 ± 1.027.2 ± 0.824.4 ± 1.925.8 ± 1.4
HDL2b%36.0 ± 2.928.7 ± 3.0**23.7 ± 5.020.3 ± 3.1
HDL3a%24.3 ± 1.627.5 ± 2.130.5 ± 2.133.7 ± 1.9
HDL3b%9.2 ± 1.111.2 ± 1.115.7 ± 3.115.0 ± 1.9
HDL3c%3.5 ± 0.65.4 ± 1.1**5.7 ± 1.55.2 ± 0.8

Within-pair similarity of lipid profile

At first, sex-adjusted ICC was determined from all twins to assess familial, probable genetic control over lipoprotein composition. In pairs concordant for BMI, ICC coefficients ranged from 0.51-0.93 (all P < 0.05) for concentrations of lipids, phospholipids, and proteins in VLDL, LDL, IDL, and HDL particles as well as lipoprotein particle masses. Equally, ICC analysis revealed high within-pair similarity in apolipoproteins ApoB and ApoC3 (0.70 and 0.81 respectively, both P < 0.05), TC concentration (0.85, P = 0.0004), all HDL subclasses (0.60-0.77, all P < 0.05), HDL mean particle size (0.80, P = 0.01), and marginal significance in ApoA1 (0.56, P = 0.06).

Within-pair similarity was significant even in BMI-discordant twins for concentrations of lipids, phospholipids, and proteins in IDL and HDL2 particles (0.48-0.64, P < 0.05). However, we identified some notable exceptions. In VLDL, within-pair similarity in particle composition weakened with weight-discordance as ICC coefficient range 0.11-0.35 was much lower and statistically non-significant compared with the marked resemblance in concordant pairs (0.70-0.85, P < 0.01). There was very little within-pair resemblance for ApoB, VLDL-C, LDL-C, HDL2a%, total TG concentration, and VLDL and LDL masses in BMI-discordant twin pairs (0.12-0.38, P = 0.07-0.29). There was no within-pair similarity in LDL peak particle size and only modest similarity in HDL3 particle in this data.

Differences in heavier vs. leaner co-twins in lipid profile and lipoprotein composition

The BMI-concordant MZ co-twins did not differ for their lipid profile (Table 1). In BMI-discordant pairs, the heavier co-twins had significantly higher levels of ApoB (P = 0.01), TC (P = 0.05), LDL-C (P = 0.05), IDL-C (P = 0.01), and HDL3c% (P = 0.003), as well as higher IDL and LDL masses (P = 0.02 and 0.05 respectively). Levels of HDL-C, HDL2-C, and HDL2b% were lower (P = 0.01 for all) and HDL mean particle size was smaller (P = 0.01) in the heavier co-twins. There were no differences in concentrations of ApoA1, ApoC3, VLDL-C and HDL3-C, LDL peak particle size, VLDL, or HDL masses between the co-twins. Neither total TG (Table 1) nor VLDL-TG (54.4 ± 5.7 mg/dl vs. 69.2 ± 7.7, P = 0.15) concentration differed between leaner and heavier co-twins.

Acquired obesity associated with only minor changes in the lipoprotein particle composition when assessing qualitative composition of each lipoprotein particle (data not shown). The heavier co-twins of the discordant pairs had significantly higher percentages of IDL free cholesterol (7.7 vs. 7.0%, P = 0.05) and cholesterol esters (10.6 vs. 9.0%, P = 0.05) in relation to other components within IDL particle when compared with their leaner counterparts. HDL2 particles were enriched with TG's in the heavier co-twins (6.1 vs. 5.1%, P = 0.02).

Correlations between adiposity, physical activity, and lipid profile within twin pairs

In analyses controlling for genetic effects within all MZ twin pairs, Pearson partial correlation coefficients between Δmarkers of adiposity, and Δlipid profile adjusted for sex, Δphysical activity, and Δsmoking are presented in Table 2. ΔBMI, Δia, Δsc, and Δliver fat were strong correlates of atherogenic lipid parameters independent of genetic effects. After further adjusting for Δliver fat, correlations became weaker for many variables.

Table 2. Pearson partial correlations between within-pair differences (Δ) in adiposity and Δlipid parameters, adjusted for sex, Δsmoking, and Δphysical activity in monozygotic twin pairs (n = 24 pairs)
 ΔLiver fat%ΔBMIΔBMI adjusted for Δliver fatΔSc fatΔSc fat adjusted for Δliver fatΔIa fatΔIa fat adjusted for Δliver fat
  1. (*)P < 0.08, *P < 0.05,
  2. **P < 0.01.
  3. Sc fat, subcutaneous fat in dm3; Ia fat, intra-abdominal fat in dm3.
Apolipoproteins
ΔApoA10.06−0.09−0.12−0.18−0.20−0.10−0.14
ΔApoB0.56**0.55*0.45*0.44*0.390.49*0.32
ΔApoC30.090.00−0.03−0.15−0.180.060.02
Serum TGs and cholesterol
ΔTG0.47*0.52*0.43(*)0.380.320.57**0.46*
ΔTC0.53*0.310.170.300.210.230.00
ΔVLDL-C0.310.56*0.500.44*0.390.60**0.54*
ΔLDL-C0.40(*)0.210.080.240.170.12−0.06
ΔIDL-C0.40(*)0.53*0.46*0.44*0.390.45*0.33
ΔHDL-C−0.12−0.38−0.37−0.37−0.36−0.39(*)−0.38
ΔHDL2-C−0.32*−0.37−0.29−0.32−0.27−0.41(*)−0.32
ΔHDL3-C0.42(*)−0.14−0.33−0.22−0.36−0.07−0.31
Lipoprotein mass and size
ΔVLDLmass0.43(*)0.52*0.44(*)0.380.320.56**0.45*
ΔIDLmass0.110.210.180.170.150.130.09
ΔLDLmass0.44*0.350.230.360.290.270.09
ΔHDLmass0.07−0.10−0.13−0.16−0.19−0.08−0.13
ΔLDL size−0.08−0.32−0.31−0.28−0.27−0.43(*)−0.44(*)
ΔHDL size−0.30−0.48*−0.42(*)−0.47*−0.43(*)−0.41(*)−0.33
HDL subclasses
ΔHDL2a%−0.06−0.05−0.03−0.09−0.07−0.020.06
ΔHDL2b%−0.30−0.46*−0.40(*)−0.45*−0.41(*)−0.42(*)−0.34
ΔHDL3a%0.210.310.260.320.280.240.17
ΔHDL3b%0.310.44*0.370.43(*)0.380.300.19
ΔHDL3c%0.090.180.160.180.170.270.25

In contrast, Δphysical activity adjusted for sex, Δsmoking, and ΔBMI associated strongly with a favorable lipid profile: ApoB: r = −0.53, LDL-C: r = −0.54, HDL-C: r = 0.47, LDL mass: r = −0.48, HDL mass: r = 0.46, HDL3b: r = −0.45 (P < 0.05 for all).

Most adiposity measures were significantly correlated with each other. Sex-adjusted partial correlations were as follows: ΔBMI and Δsc fat: r = 0.93, P < 0.001, ΔBMI and Δia fat: r = 0.72, P < 0.001, ΔBMI and Δliver fat: r = 0.35, P = 0.09, Δia fat and Δsc fat: r = 0.62, P = 0.002, Δia fat and Δliver fat: r = 0.52, P = 0.01.

Because of the high correlation between the measures of adiposity, multivariate regression analyses were performed to assess which component of body fat distribution independently explained pro-atherogenic changes in lipid profile. We entered within-pair differences in body fat distribution (Δsc fat, Δia fat, and Δliver fat), sex, Δsmoking status, and Δphysical activity as independent variables in the models. Out of all body fat depots, Δliver fat was the only one independently explaining the variation in ΔApoB (β = 2.2 ± 1.0, P = 0.05; whole model adjusted R2 = 0.45, P = 0.01), ΔTC (β = 3.6 ± 1.4, P = 0.02; R2 = 0.37, P = 0.04), ΔLDL-C (β = 2.9 ± 1.3, P = 0.04; R2 = 0.56, P = 0.003), ΔHDL3-C (β = 0.7 ± 0.3, P = 0.04; R2 = 0.08, P = 0.32), and also marginally ΔLDL mass (β = 5.7 ± 2.9, P = 0.07; R2 = 0.48, P = 0.01). ΔPhysical activity remained significant in models explaining ΔLDL-C (β = −10.4 ± 2.8, P = 0.002; R2 = 0.56, P = 0.003), ΔLDL mass (β = −16.7 ± 6.2, P = 0.016; R2 = 0.48, P = 0.01), and marginally ΔHDL3b (β = -1.5 ± 0.7, P = 0.06; R2 = 0.28, P = 0.088).

Influence of liver fat on lipid profile in twins from BMI-discordant pairs

Division of the BMI-discordant pairs based on their within-pair differences for liver fat revealed an interesting splitting of the results. Despite the marked difference in weight in both groups, only those obese co-twins who also had high liver fat, showed a pro-atherogenic lipid profile (Figure 1). In this group, the heavier co-twins had higher levels of ApoB, TC, LDL-C, IDL-C, LDL mass, HDL3c and lower levels of HDL-C, smaller HDL mean particle size (P < 0.05), and marginally higher TG concentrations (P = 0.07) than their leaner counterparts. ApoA1, ApoC3, VLDL-C, VLDL, IDL and HDL masses, LDL peak particle size, or other HDL subclasses did not differ between the co-twins neither in liver fat discordant nor in liver fat concordant pairs (data not shown). In BMI-discordant but liver fat concordant pairs, the lipid profiles were very similar between the co-twins: the only difference noted was a slightly smaller HDL peak particle size (9.7 vs. 9.5 nm, P = 0.05) in the heavier co-twins.

Figure 1.

Lipid profiles in BMI-discordant twins divided into two groups based on the median of within-pair difference (Δ) in liver fat 2.6%: [1] concordant (Δlfat <2.6%) or [2] discordant (Δlfat ≥2.6%) for liver fat. Both groups were equally discordant for overall obesity (mean Δbody weight 16 and 17 kg). Data are mean ± SE. P values: leaner vs. heavier co-twins in Wilcoxon signed ranks tests and liver fat discordant vs. concordant pairs in Mann–Whitney U-test. Liver fat concordant group, n = 7 pairs. Liver fat discordant group, n = 8 pairs. Lfat, liver fat; ApoB, apolipoprotein B; LDL-C, low density lipoprotein cholesterol; IDL-C, intermediate density lipoprotein cholesterol; HDL-C, high density lipoprotein cholesterol.

Discussion

In this study, we were able to demonstrate that acquired obesity is associated with several pro-atherogenic derangements in the lipoprotein metabolism especially when obesity is accompanied with a fatty liver. In our study, we used a unique sample of MZ twin pairs discordant for BMI, which allowed us to estimate the effects of acquired adiposity in heavier and leaner individuals matched for genetic background. A further advantage of this study was that half of these rare MZ pairs were highly discordant for liver fat, whereas the other half was composed of co-twins which in spite of large differences in body weight had equally low liver fat values in both twin pair members. Using this unique experiment of nature, we demonstrated that the heavier co-twins with high liver fat had higher concentrations of pro-atherogenic lipoprotein particles and less atheroprotective qualities in their lipid profile. The heavier co-twins with low liver fat resembled their lean co-twins more; however, a non-significant trend for worsening of the lipid profile with overall fatness was observed in them as well. Our data further showed that the serum lipid profile is very similar in MZ twins concordant for body weight and remains highly correlated between co-twins in case of TC, IDL-C, HDL-C, and HDL subclasses even in the BMI-discordant pairs. However, TG, ApoB, LDL-C, HDL2a, VLDL, and LDL masses reacted to obesity as the within-pair similarity was lost in BMI-discordant pairs. This suggests that the lipoprotein metabolism has a strong genetic background but there is an interaction with acquired obesity in a complex manner, depending on the distribution of body fat.

The heterogeneity in the lipid panel in our BMI-discordant twin pairs supports previous data on the hypothesis that there might be a distinctive group of metabolically healthy obese people [27]. Uncomplicated obesity is characterized by low amount of visceral adipose tissue [28], low inflammatory state [29], and a lower degree of ectopic fat deposition, in particular in muscle and liver [27]. Because liver fat and ia fat are highly correlated, their independent roles on the lipid derangements have been difficult to distinguish. This study suggests that liver fat is the main culprit of the lipid disturbances as it was the only body fat depot that remained significant in explaining the lipid parameters in the multivariate models. This finding is in line with the close biological role of the liver as the factory of lipoprotein particles [30].

Fatty liver leads to overproduction of VLDL particles. In addition to overproduction of these TG-rich lipoproteins, hypertriglyceridemia is due to impaired clearance of VLDL particles [7]. Overproduction of VLDL has effects both on HDL and LDL via actions of cholesterol ester transfer protein (CETP) secreted in liver. CETP mediates the bidirectional transfer of lipoprotein core lipids between different particles. When VLDL concentration is high, HDL cholesterol esters are preferentially transferred by CETP to larger VLDL particles that become cholesterol rich and thus potentially more atherogenic [31]. In addition, CETP is involved with the formation of small LDL particles. This most atherogenic subclass of LDL develops when TGs in LDL are gradually hydrolyzed by hepatic lipase resulting in the formation of small LDL particles [32]. Again, our findings showing a pro-atherogenic lipid pattern mainly when the liver is fatty, corresponds well with this biology.

Our data with detailed measures of lipoprotein particles, their size and composition and apolipoproteins have revealed several new findings beyond the traditional measures of basic lipid values. Both LDL-C and LDL particle mass were higher in the heavier co-twins suggesting that the heavier co-twins have an increased number of LDL particles as reflected by the increase of ApoB concentration. In general, HDL is divided into HDL2 and HDL3 subspecies based on their size. While the larger HDL2 has better cholesterol efflux capacity, the smaller and denser HDL3 has been demonstrated to protect LDL from oxidation [33]. In this study, ApoA1, the predominant protein carried on HDL particles, did not differ between heavy and lean co-twins. As there was no difference in HDL mass either, it suggests that HDL particle concentration remains unchanged in obesity. Low HDL-C in heavier co-twins was due to a reduction of cholesterol in large HDL2 particles, HDL2b in particular, the subclass that is considered the most cardioprotective [34]. Subsequently, HDL particle size was smaller in the heavier co-twins of the BMI-discordant pairs in our data.

Accumulation of fat in the liver has proven to be in part genetically controlled: familial clustering of fatty liver has been demonstrated [35]. Approximately 60% of the variability of liver enzymes is heritable [36], and several novel gene candidates and single nucleotide polymorphisms in DNA have been described for explaining either fatty liver [35, 37] or liver enzymes [38]. It is therefore probable that the propensity to accumulate fat in the liver in obesity is also partly genetic. Notably, in our study, the liver was fatty after development of obesity but not in lean twins, which points to the importance of obesity as the trigger of the genetic predisposition. However, despite the fact that PNPLA3 and LYPLAL1 gene variants are significantly associated with liver fat, they do not affect serum lipids [37]. This suggests that other underlying mechanisms are involved in the lipoprotein derangements. Whether the same sets of genes explain liver fat and serum lipids remains to be studied.

It is also possible that environmental factors, such as physical activity and diet modify the function of relevant genes and explain the acquired differences in liver fat content and the lipid profile. Exercise training has been proven to reduce intra-hepatic fat content in adults [39]. The associations with exercise and lipid levels are less consistent. In Sullivan's trial, no improvement in VLDL-TG or ApoB100 secretion rates was seen [39]. In a meta-analysis of randomized controlled-trials in adults improvement was seen only when exercise was combined with diet [40]. It is therefore important that we utilized the information from 3-day food diaries and physical activity questionnaires to examine the roles of lifestyle factors on the lipid profile. Low physical activity, together with high liver fat remained as an independent predictor of high LDL-C. In this study, within-pair differences in 3-day food intake or alcohol did not explain the observed lipid changes.

The limitations of our study include the cross-sectional design and the small sample size. It must be acknowledged that this may have limited our power to detect significant differences in the liver fat concordant twin pairs. Furthermore, the possibility of residual confounding by unmeasured covariates cannot be excluded.

In summary, we demonstrate the diversity of serum lipid profile in a unique collection of MZ twins discordant for BMI. Among the BMI-discordant twin pairs, the occurrence of both liver fat discordant and liver fat concordant pairs offered an extraordinary informative design to study the role of body fat distribution and ectopic fat on serum lipids. We were able to show that acquired obesity was associated with a pro-atherogenic lipid profile, in particular when obesity was accompanied by high liver fat. Of note was that despite the probable strong impact of genes on the liver fat and lipid profile, lean individuals remained to a large extent protected from both the fattening of the liver and from the harmful lipid patterns. Physical activity together with low liver fat was the most important cardioprotective factors in this study.

Acknowledgments

We thank the participants for their invaluable contributions to the study and Juha Rantanen, Laura Suojanen, Tiina Vikstedt, Pentti Pölönen and Helinä Perttunen-Nio for their excellent assistance.

Ancillary