Overeating Saturated Fat Promotes Fatty Liver and Ceramides Compared With Polyunsaturated Fat: A Randomized Trial

Overweight men and women were recruited by local advertising. Inclusion criteria included participants ages 20 to 55 years and body mass index (BMI) of 25 to 32 kg/m²; exclusion criteria were the following: diabetes (fasting glucose >7 mM at two occasions) or liver disease; pregnancy; lactation; alcohol abuse; claustrophobia; abnormal clinical chemistry test results; use of drugs influencing energy metabolism; use of omega-3 supplements or extreme diets; regular heavy exercise (>3 h/wk); intolerance to gluten, egg, or milk protein; and implanted metals. Subjects were instructed to maintain their habitual diet and physical activity level throughout the study. Subjects were fasted overnight (10 to 12 hours) and were discouraged from physical exercise and alcohol intake 48 hours before measurements.

The LIPOGAIN-2 study was a 12-week, double-blind, parallel-group, randomized trial in free-living subjects, carried out from August 2014 through June 2015 at the Uppsala University Hospital, Uppsala, Sweden. Subjects were randomized by computer-generated lists (generated by a statistician not otherwise involved in the project), stratified by sex, age, and BMI. Double-blinding was ensured by labeling, and the code was concealed from all investigators until study completion and statistical analyses of primary outcomes. During the first 8 weeks, subjects were overfed (weight gain), followed by 4 weeks of caloric restriction (weight loss). The trial was registered at www.clinicaltrials.gov as NCT02211612↗.

Sixty-one participants were randomized to eat muffins containing either sunflower oil (high in PUFA, linoleate, 18:2n-6) or palm oil (high in SFA, mainly palmitate 16:0) during 8 weeks. Both oils were refined; fatty acid composition has been described (3). Body weight was measured, and muffins were provided to subjects weekly at the clinic. Muffins were baked in large batches under standardized conditions in a metabolic kitchen at Uppsala University. Muffins were added to the habitual diet (to be consumed anytime during the day) and individually adjusted weekly (altered by ±1 muffin/d depending on rate of weight gain) to achieve a 3% weight gain (on average, 2.9 ± 0.5 muffins were added, equaling ∼40 g of oil/d). Except for fat type, muffins were identical in composition (51 energy% fat, 44 energy% carbohydrates, and 5 energy% protein). Following the overfeeding period, subjects switched to a 4-week, low-calorie diet, consisting of ∼800 kcal/d (∼52 energy% carbohydrate, ∼26 energy% protein, and ∼18 energy% fat; Modifast; Nutrition & Santé).

Liver and pancreas fat content, liver volume, and volume of total body and visceral fat were assessed by MRI using a 1.5T Achieva clinical scanner (Philips Healthcare), modified to allow arbitrary table speed. Collection and analyses of the MRI data were performed at one center under blinded conditions. A single operator, trained by an experienced radiologist, performed all measurements. The absolute amount of liver fat in liters was calculated by the multiplication of liver volume by liver fat content (%). Total body and visceral fat were quantified using a previously described automated image analysis algorithm (16). Total body fat was also measured using whole-body air displacement plethysmography (ADP; Bod Pod; COSMED®), according to the manufacturer’s instructions. Total body water content was measured by bioelectrical impedance analysis (Tanita BC-558; Tanita Corporation). Data from the ADP were corrected for total body water content by using a three-compartment model (17).

In a subgroup (n = 5 in each group), palmitate uptake in liver, pancreas, heart, and skeletal muscle was measured on a 3T positron emission tomography (PET)-magnetic resonance (MR) scanner (Signa PET/MR; GE Healthcare). A single bed-position dynamic PET scan, covering the splanchnic bed, was performed during 60 minutes, starting simultaneously with a bolus injection of [¹¹C]palmitate (361.35 ± 11.9 MBq). Regions of interest for liver, pancreas, and skeletal muscle tissue, as well as the blood pool in ascending aorta and portal vein, were outlined manually based on the anatomical information in a T1-weighted MR image acquired simultaneously with the PET scan. Myocardium and left-ventricular cavity were segmented automatically on PET images using the cardiac module in Carimas. Quantification of [¹¹C]palmitate uptake in liver, pancreas, and skeletal muscle was performed using an image-derived input function based on average radioactivity concentrations in an ascending aorta and portal vein, corrected for radioactive metabolites, using the measured parent fraction remaining in plasma at 10, 20, 30, 40, and 50 minutes postinjection, as well as plasma to whole-blood radioactivity concentration at 5, 10, 20, 30, 40, and 50 minutes postinjection. Four different parameters were calculated based on the [¹¹C]palmitate data:The accumulation rate constant [influx constant (K_i)] was estimated using Patlak analysis for the 5- to 60-minute postinjection interval.Standardized uptake values (SUVs) were estimated, adjusting radioactivity concentrations for intersubject differences in body weight and administered amount of [¹¹C]palmitate [SUV = radioactivity concentration (becquerel/mL)/body weight (gram) × administered tracer (becquerel)] for the uptake windows of 30 to 35 minutes and 55 to 60 minutes postinjection.Myocardial fatty acid uptake was assessed by the multiplication of the plasma nonesterified fatty acid (NEFA) concentration by the fractional uptake rate (determined by the division of the myocardial concentration peak value by the integral of the metabolite-corrected plasma curve).Myocardial fatty acid oxidation was estimated by the fitting of a biexponential function to the myocardial clearance of radioactivity (18).

Ceramides from serum and adipose tissue were extracted using the butanol–methanol methods (19, 20). Adipose tissue ceramides were then further purified using straight-phase HPLC, coupled to a fraction collector (21). Ceramides from both serum and adipose tissue were detected and quantified using ultraperformance liquid chromatography/tandem mass spectrometry, as previously described (22).

Dietary intake was assessed by 4-day weighed food records (at baseline and week 8) and processed with DietistNET dietary assessment software. During this 4-day periods, subjects wore accelerometers (Philips Respironics Actical) on their right ankle to assess 24-hour physical activity. Fatty acid composition was measured in the intervention oils used in the isocaloric food items, as well as in plasma cholesterol esters (CEs), phospholipids (PLs), and adipose tissue triglycerides by gas chromatography, as previously described (23). Subcutaneous adipose tissue biopsies were taken as previously described (3).

SCD activity was estimated as the 16:1n-7/16:0 ratio. Delta-5 desaturase activity was estimated as the 20:4n-6/20:3n-6 ratio and delta-6 desaturase as the 18:3n-6/18:2n-6 ratio.

Participants completed an oral glucose tolerance test by consuming 75 g glucose dissolved in 200 mL water (blood sampling at 0, 30, 60, and 120 minutes). Glucose, total cholesterol, high-density lipoprotein (HDL)-cholesterol, low-density lipoprotein (LDL)-cholesterol (directly measured), triglycerides, C-reactive protein, apolipoprotein (apo)B, and apoA1 were analyzed in plasma and insulin in serum by standard laboratory methods at Uppsala University Hospital.

Vascular cell adhesion molecule-1, intercellular adhesion molecule-1, endostatin, and TNF receptor 1 and 2 (TNF-R1 and TNF-R2) were analyzed in plasma by ELISA (R&D Systems). d-3-Hydroxybutyrate was analyzed using a kinetic enzymatic method using the Ranbut reagent (RB1008; Randox Laboratories).

Serum fibroblast growth factor (FGF)21, proprotein convertase subtilisin/kexin type 9 (PCSK9), and FGF19 levels were determined by ELISA (Catalog nos. DF2100, DPC900, and DF1900, respectively; R&D Systems), according to the manufacturer’s instructions. Serum NEFA levels were determined using an enzymatic colorimetric assay [HR Series NEFA-HR(2); Wako Diagnostics]. Levels of serum BAs and 7a-hydroxy-4-cholesten-3-one (C4), a BA synthesis marker, were determined by liquid chromatography–tandem mass-spectrometry using deuterium-labeled standards for C4 and BA (24). Unesterified lathosterol, a serum marker of total cholesterol synthesis, was determined by gas chromatography–mass spectrometry (25). C4 and lathosterol levels were corrected for total serum and total cholesterol (C4c and lathosterol/c), respectively (26, 27).

Differences in mean values between groups postintervention were analyzed with ANCOVA, adjusted for baseline values. Differences in mean values between subgroups (e.g., for PET-MR) were analyzed with Mann-Whitney test. Data are given as means (SD) or medians [interquartile range (IQR)]. Correlations between variables are given as Pearson’s r or Spearman’s rho. A P value <0.05 was considered statistically significant. JMP version 13.1.0 was used; heatmaps were created using R. Primary outcome measure for the trial was change in lean tissue mass (reported in a separate manuscript); secondary outcome measures included changes in liver and pancreas fat, VAT, total body fat, palmitate uptake, and blood lipids. Sample size was determined using Lehr formula, based on our previous trial, using similar intervention and outcome measures (3). For lean tissue, n = 22 subjects per group was needed to detect a 0.55-L difference between groups, with α = 0.05 and β = 0.20. For liver fat, n = 23 subjects per group was needed to detect a 0.52% units difference between groups with α = 0.05. For the exploratory assessment of hepatic palmitate uptake, we had no previous data to use for sample-size determination. With the current results in hand, n = 13 subjects per group would have been needed to detect a 20% difference in net uptake rate between groups with α = 0.05.

This study was conducted in accordance with the Declaration of Helsinki. All subjects provided written, informed consent before inclusion, and the study was approved by the Regional Ethical Review Board in Uppsala (Dnr 2014/186). All authors reviewed and approved the final manuscript.

Changes in fatty acid composition in both plasma and adipose tissue reflected the assigned interventions, indicating high compliance (28). Linoleate, the major fatty acid in the PUFA muffins, increased in plasma CE, PL, and subcutaneous adipose tissue in the PUFA group (P < 0.0001 for comparison between groups), whereas palmitate, the major fatty acid in the SFA muffins, increased in both plasma PL and adipose tissue in the SFA group during the intervention (P < 0.0001 for difference between groups). In addition, based on checklists/self-report, subjects in the SFA group consumed 96.8%, and subjects in the PUFA group consumed 97.1% of the provided muffins, and the 4-day weighed dietary records indicated no changes other than those desired/expected in fat type. Finally, cholesterol levels changed in accordance to what would be expected based on increased intakes of SFA and PUFA, respectively (vide infra). Thus, both key dietary biomarkers, as well as self-reported data, suggested excellent compliance to both diets.

SFA overfeeding deteriorated the blood lipid profile (e.g., increasing LDL-cholesterol and apoB), whereas PUFA overfeeding reduced atherogenic blood lipids, despite weight gain (Fig. 1) (28). Compared with SFA, PUFA caused a reduction (P = 0.004) in circulating lathosterol, a marker of cholesterol synthesis; this was, however, attenuated when corrected for total cholesterol, indicating no change in cholesterol synthesis, in line with unchanged PCSK9. Neither fasting serum total BAs nor their composition or other markers of cholesterol and BA synthesis (e.g., FGF19 and C4) differed between diets (28). Serum total BAs increased postprandially in both groups, but areas under the curve did not differ between groups.

Liver fat increased by 53% (1.54 ± 2.0% points; 30 ± 40 mL) in the SFA group, whereas in the PUFA group, there was a 2% decrease (−0.09 ± 1.55% points; −1 ± 30 mL; P = 0.001 for between-group difference), despite similar body-weight gain (Fig. 2A). The different liver fat accumulation was reflected by plasma alanine aminotransferase (ALT) levels, which increased by 18% (Δ0.08 ± 0.18 µkat/L; from 0.45 ± 0.27 µkat/L to 0.53 ± 0.29 µkat/L) in the SFA group and remained unchanged (Δ−0.01 ± 0.14 µkat/L; from 0.48 ± 0.32 µkat/L to 0.47 ± 0.27 µkat/L) in the PUFA group (P = 0.035 for between-group difference). The change in liver fat was inversely correlated with change in circulating linoleate in both plasma (r = −0.48, P = 0.0001 for CE and r = −0.38, P = 0.003 for PL) and adipose tissue (r = −0.32, P = 0.04) but directly associated with change in the circulating palmitate (r = 0.30, P = 0.02 and r = 0.52, P = 0.0001 for CE and PL, respectively) and SCD activity index in CE (r = 0.49, P = 0.0001). The changes in liver fat did not translate to significant changes in glucose tolerance or insulin sensitivity between groups (28).

Pancreas fat accumulation was similar (P = 0.52) in SFA (0.49 ± 2.29%) and PUFA (0.46 ± 1.69%) groups (Fig. 2B).

The accumulation of VAT was similar (P = 0.17) in SFA (0.37 ± 0.29 L) and PUFA (0.26 ± 0.30 L) groups, as was the accumulation of total body fat (2.22 ± 1.57 L vs 1.77 ± 1.63 L, respectively, P = 0.28; Fig. 2C and 2D).

In the exploratory subgroup study using PET-MRI, the change in hepatic palmitate uptake did not differ between groups, neither when expressed as SUVs at 30 or 60 minutes nor when assessing net influx rate K_i during different time intervals (28). Change in liver fat in this subgroup reflected that seen in the full sample. (Liver fat quantification failed at one occasion; hence, this comparison is for n = 4 vs n = 5.) In both groups combined (n = 10), none of the measurements of palmitate uptake changed during the intervention (data not shown). The change in hepatic palmitate uptake (net uptake rate K_i 5 to 60 minutes) tended to be inversely associated with change in liver fat, but there was no association between hepatic palmitate net uptake rate and liver fat at baseline (r = 0.18, P = 0.61; data not shown).

SFA- and PUFA-enriched diets had markedly different effects on serum ceramides (Cer), dihydroceramides (DiCer), glucosylceramides (GluCer), and lactosylceramides (LacCer). Overall, SFA increased, whereas PUFA decreased ceramide levels—changes that were already evident at 4 weeks (Fig. 3). Change in liver fat was directly associated with change in species from all ceramide groups (28). Interestingly, the differential effects on liver fat by SFA- and PUFA-rich diets were diminished (P = 0.08) when adjusted for changes in C16-containing ceramides (Cer16:0, DiCer16:0, and GluCer16:0; data not shown). Changes in circulating 16:1n-7 and the SCD index were positively correlated with changes in multiple ceramide species, whereas changes in circulating linoleate showed consistent, inverse associations (28).

Although changes in adipose tissue ceramides were also generally different between SFA and PUFA groups, changes were less pronounced and with few exceptions, not statistically different between groups.

SFA and PUFA overfeeding had differential effects, also on circulating sphingomyelin species: 16:0, 16:1, 18:0, and 18:1 species increased significantly more by SFA (changes evident already at 4 weeks), and changes in 16:0 and 16:1 species were directly associated with changes in both liver and VAT fat levels (r ∼ 0.3, P < 0.03; data not shown).

Changes in palmitate uptake in pancreas, skeletal muscle, or heart did not differ between groups (28). In both groups combined, none of the measurements changed as a result of the intervention (data not shown). Myocardial palmitate uptake or oxidation did not differ between groups (28), but uptake increased by 40% during the intervention (P = 0.037) in both groups combined (data not shown).

Changes in serum Cer, DiCer, GluCer, and LacCer were strongly, positively associated (r = 0.63 to 0.95), with changes in pancreatic (but not hepatic) palmitate uptake assessed as SUV at 30 and 60 minutes (Fig. 4A–4D). Similar but attenuated associations were observed when palmitate uptake was assessed as net uptake rate K_i. Furthermore, similar results were observed for the individual C16 and C18 species (P = 0.002 to 0.03 and P = 0.001 to 0.07, respectively; data not shown).

We did not observe any differences in markers of inflammation (C-reactive protein, TNF-R1, TNF-R2), endothelial function (vascular cell adhesion molecule-1, intercellular adhesion molecule-1, endostatin), or free radical-induced lipid peroxidation (isoprostane 8-iso-prostaglandin F2α) as a result of PUFA compared with SFA overfeeding (28).

Fifty-one subjects (n = 26/25 from SFA and PUFA groups, respectively) completed the 4-week hypocaloric period (∼800 kcal/d), which caused similar (P = 0.92) weight loss (−4.55 ± 2.39 kg and −4.48 ± 2.63 kg, respectively). Furthermore, both groups decreased similarly (P = 0.82) in waist circumference (−4.62 ± 5.68 vs −4.26 ± 5.29 cm, respectively) and total body fat (kilograms, ADP; −3.70 ± 2.61 vs −3.21 ± 2.57 kg, respectively; P = 0.94). All differences observed during weight gain between groups in blood lipids were abolished by weight loss. Likewise, no differences between groups were observed for liver enzymes or ceramides after weight loss (data not shown). In both groups combined, changes in ceramides and sphingolipids showed diverse effects in response to weight loss. Total serum BAs (P = 0.012) and FGF19 (P = 0.042; a circulating inhibitor of BA synthesis) were higher in the SFA vs PUFA group after weight loss.