Ceramides may mediate saturated fat–induced insulin resistance, but there are no data comparing ceramide concentrations between human tissues. We therefore performed lipidomic analysis of human subcutaneous (SCfat) and intra-abdominal (IAfat) adipose tissue, the liver, and serum in eight subjects. The liver contained (nmol/mg tissue) significantly more ceramides (1.5–3-fold), sphingomyelins (7–8-fold), phosphatidylethanolamines (10–11-fold), lysophosphatidylcholines (7–12-fold), less ether-linked phosphatidylcholines (2–2.5-fold) but similar amounts of diacylglycerols as compared to SCfat and IAfat. The amounts of ceramides and their synthetic precursors, such as palmitic (16:0) free fatty acids and sphingomyelins, differed considerably between the tissues. The liver contained proportionally more palmitic, stearic (18:0), and long polyunsaturated fatty acids than adipose tissues. Stearoyl-CoA desaturase 1 (SCD1) activity reflected by serum, estimated from the 16:1/16:0-ratio, was closely related to that in the liver (r = 0.86, P = 0.024) but not adipose tissues. This was also true for estimated elongase (18:1/16:1, r = 0.89, P = 0.01), and Δ5 (20:4/20:3, r = 0.89, P = 0.012) and Δ6 (18:3[n-6]/18:2, r = 1.0, P < 0.001) desaturase activities. We conclude that the human liver contains higher concentrations of ceramides and saturated free fatty acids than either SCfat or IAfat.
Ceramides are sphingolipids with a variety of metabolic functions, such as mediation of saturated fat–induced insulin resistance (1,2,3), inflammation, and regulation of cellular stress and apoptosis (4,5). In humans, ceramide concentrations in skeletal muscle have shown to be increased in obese insulin-resistant subjects as compared to lean insulin-sensitive subjects (6). Serum concentrations of sphingosine, an immediate metabolite of ceramide (7), are higher in type 2 diabetic patients as compared to healthy controls (8). We have recently shown that ceramide concentrations in SCfat are higher in subjects with high liver fat content as compared to age-, gender-, and BMI-matched subjects with low liver fat content (9).
Ceramides can be synthesized via de novo synthesis from saturated fatty acids, such as palmitate (16:0), or from sphingomyelin via sphingomyelinase (4). In insulin-resistant subjects with a fatty liver, stable isotope tracer studies have suggested that the production of saturated fatty acids via de novo lipogenesis is 5-fold higher as compared to healthy subjects (10,11). In ob/ob mice, hepatic ceramide concentrations are directly related to the degree of steatosis (12). In addition, increased concentrations of saturated fatty acids in serum characterize subjects with metabolic abnormalities, such as the metabolic syndrome, type 2 diabetes, and cardiovascular disease (13,14,15,16). In vitro, formation of ceramides from saturated fatty acids inhibits insulin signaling (17). Thus, oversupply of saturated fat may contribute to ceramide accumulation and insulin resistance. The relative content of ceramides in different human tissues has not yet been compared.
The major lipid species (triacylglycerols, phospholipids, cholesterol, and cholesterol esters) of serum (18), the liver (19,20,21), and adipose tissue (18,22) in humans have been characterized in several studies. Today's novel lipidomic strategies enable, however, identification and accurate quantification of a wide range of individual lipid species in biological samples (23), such as sphingomyelins, ether-linked lipids, lysophosphatidylcholines in addition to ceramides. In the present study, we applied a lipidomics approach with two analytical platforms to characterize lipid and fatty acid composition of SCfat and IAfat, and the liver in humans.
Methods and Procedures
Clinical characteristics of the study subjects are shown in Table 1. Eight subjects were recruited from patients undergoing a laparoscopic gastric bypass operation based on the following inclusion criteria: (i) age 18–60 years; (ii) no known acute or chronic disease except for obesity or type 2 diabetes based on history, physical examination, standard laboratory tests (blood counts, serum creatinine, thyrotropin, and electrolyte concentrations), and electrocardiogram; (iii) alcohol consumption <20 g per day. Two subjects had type 2 diabetes. The patients had not lost weight prior to surgery. The nature and potential risks of the study were explained to all subjects before obtaining their written informed consent. The study protocol was approved by the ethics committee of the Helsinki University Central Hospital.
Table 1. Clinical characteristics of the study subjects
On the morning prior to surgery, blood samples were taken after an overnight fast for measurement of plasma glucose, serum insulin, C-peptide, liver enzymes, serum triglyceride, and low-density lipoprotein- and high-density lipoprotein-cholesterol concentrations, and for performing metabolomic analyses. Body weight was recorded using a calibrated weighing scale. Wedge biopsies of the liver, and IAfat and SCfat (10–400 mg) were taken at surgery. Approximately one-half of the liver sample was sent to the pathologist for routine histopathological assessment, whereas the rest was immediately frozen and stored in liquid nitrogen. Fat content of the liver biopsy specimens (% of hepatocytes with macrovesicular steatosis in light microscope) was determined by an experienced liver pathologist (J.A.) in a blinded fashion. There were no histological signs of inflammatory changes in the hepatic tissue of the study subjects.
For analysis of serum lipids, 10 µl aliquots of serum were spiked (0.2–0.6 µg/sample) with an internal standard mixture containing nine lipid compounds: PC (17:0/0:0), MG(17:0/0:0/0:0) [rac], PG(17:0/17:0) [rac], Cer(d18:1/17:0), PC(17:0/17:0), PA(17:0/17:0), PE(17:0/17:0), DG(17:0/17:0/0:0) [rac], and TG(17:0/17:0/17:0). Standard stock solutions (5–20 mg lipid/5 ml chloroform:methanol 2:1) were first prepared separately, and internal standard mixtures for serum, for other tissues, and a mixture of labeled standards were prepared by mixing 40–300 µl of the stock solutions and diluting with chloroform:methanol 2:1. The volume of the standard solutions added to the samples was 20 µl for serum and 10 µl for other tissues. Serum samples were mixed with 10 µl of 0.9% NaCl, and the lipids were extracted with 100 µl of chloroform:methanol (2:1). The samples were vortexed for 2 min, and after 30 min standing, they were centrifuged at 10,000 rpm for 3 min. The lower lipid extracts were separated, and another standard mixture containing three labeled lipid compounds (PCo(16:0/0:0-D3), PC(16:0/16:0-D6), and TG(16:0/16:0/16:0-13C3)) was added (0.1–0.2 µg/sample).
Prior to analyzing lipid composition of the liver and adipose tissues, frozen samples (5 mg) were mixed with an internal standard mixture containing 0.5–1 µg/sample of PC(17:0/0:0), Cer(d18:1/17:0), PC(17:0/17:0), PE(17:0/17:0), and TG(17:0/17:0/17:0), and 200 µl chloroform:methanol (2:1). The tissues were homogenized with grinding balls in a mixer mill at 25 Hz for 2 min, and 50 µl of 0.9% NaCl was added. The samples were vortexed for 2 min, and after 30 min standing, centrifuged at 10,000 rpm for 3 min. The labeled lipid standard mixture was added into the separated lipid extracts (1 µg/sample) before UPLC-MS analysis.
Lipid extracts were analyzed on a Waters Q-Tof Premier mass spectrometer (Waters, Milford, MA) combined with an Acquity Ultra Performance LC (UPLC; Waters). The column (at 50 °C) was an Acquity UPLC BEH C18 1 × 50 mm with 1.7 µm particles. The solvent system included A. water (1% 1 mol/l NH4Ac, 0.1% HCOOH) and B. acetonitrile/isopropanol (5:2, 1% 1 mol/l NH4Ac, 0.1% HCOOH). The gradient started from 65% A/35% B, reached 100% B in 6 min and remained there for the next 7 min. There was a 5-min re-equilibration step before next run. The flow rate was 0.200 ml/min and the injected amount was 1.0 µl. Reserpine was used as the lock spray reference compound. The lipid profiling was carried out using ESI+ mode, and the data were collected at mass range of m/z 300–1,200 with a scan duration of 0.2 s. The data were processed by using MZmine software (version 6.0; ref. 24), and the lipid identification was based on an internal spectral library (12).
Fatty acid analysis
Another set of serum aliquots (50 µl) was extracted with chloroform:methanol (2:1; 200 µl) in Eppendorf tubes after addition of internal standards (heptadecatrienoate 10 µg; heptadecanoic acid 10 µg). The samples were vortexed for 1 min, and after 30 min of extraction, centrifuged at 10,000 rpm for 5 min. The lower layer was taken into a glass tube and evaporated into dryness under nitrogen.
Frozen liver and fat tissues (5 mg) were spiked with 40–60 µg of standards and homogenized in capped 2 ml microtubes (Sarstedt, Nümbrecht, Germany) at −20 °C with chloroform:methanol (2:1; 400–600 µl) by using zirconium oxide balls in a mixer mill (25 Hz for 5 min). A volume of 100–150 µl NaCl (0.9%) was added by vortexing, and after 1 h extraction time, the lower layer was separated by centrifuging at 10,000 rpm for 5 min; 100–200 µl aliquots from the extracts were evaporated into dryness and used for fatty acid analyses.
The evaporation residues were redissolved into 700 µl of petroleum ether (boiling point 40–60 °C) by vortexing. Sodium methoxide (250 µl; 0.5 mol/l NaOMe in MeOH) and a couple of boiling stones were added, and the mixture was boiled at 45 °C for 5 min. The samples were acidified with 500 µl of 15% NaHSO4, 100 µl of petroleum ether was added, and the samples were centrifuged in 2 ml microtubes at 10,000 rpm for 5 min. The petroleum ether layers were separated into GC vials, evaporated and redissolved into 100 µl of hexane.
Gas chromatographic analysis of fatty acids was performed using Agilent 5890 series II GC (Agilent Technologies, Santa Clara, CA), equipped with a 25 m FFAP column (0.32 mm ID). The injection volume was 2 µl and split ratio 1:23. Helium was used as carrier gas, and the oven temperature program was from 70 °C to 240 °C at 7 °C/min. The injector and detector (FID) temperatures were 260 °C and 300 °C, respectively.
Measurements of IL-6 and TNFα protein expression in adipose tissue
A frozen sample of SCfat (100–250 mg) was homogenized in lysis buffer. The homogenate was centrifuged for 30 min (+4 °C, 14,000 rpm), and the supernatant was stored at −80 °C until measurement of the interleukin-6 (IL-6) and tumor necrosis factor-α (TNFα) concentrations with the human IL-6 or TNFα immunoassay kit (Quantikine; R&D Systems, Minneapolis, MN). Total protein was measured with the BC Assay protein quantitation kit (Optima Interchim, Montlucon, France).
Fat cell size
Immediately after taking the adipose tissue biopsies, 2–10 mg of adipose tissue pieces were mixed with collagenase and incubated at +37 °C for 1 h. The cells were then removed to a Bürcker's chamber, and the diameter of randomly chosen 200 adipocytes was measured at a magnification of ×25 using a light microscope (Leitz, Wetzlar, Germany). Average cell size was used as fat cell size.
Analytical procedures, measurements, and calculations
Plasma glucose concentrations were measured in duplicate with the glucose oxidase method using Beckman Glucose Analyzer II (Beckman Instruments, Fullerton, CA) (25). Serum-free insulin concentrations were measured with the Auto-DELFIA kit (Wallac, Turku, Finland), and C-peptide concentrations by radioimmunoassay (26). Serum high-density lipoprotein-cholesterol and triglyceride concentrations were measured with the enzymatic kits from Roche Diagnostics using an autoanalyzer (Roche Diagnostics Hitachi, Hitachi, Tokyo, Japan). The concentration of low-density lipoprotein-cholesterol was calculated using the Friedewald formula (27). Serum aspartate aminotransferase, alanine aminotransferase, and γ-glutamyl transpeptidase activities were determined as recommended by the European Committee for Clinical Laboratory Standards. Serum IL-6 and TNFα concentrations were measured with ELISA kits (Quantikine).
Desaturase and elongase activities were estimated from product:precursor ratios of the percentages of individual fatty acids according to the following equations: SCD1 = (16:1/16:0) or (18:1/18:0), Δ6 desaturase = (18:3[n-6]/18:2), Δ5 desaturase = (20:4/20:3), elongase = (18:0/16:0), or (18:1/16:1) (ref. 28).
The data are shown as means ± s.e.m. Correlation analyses were performed using Spearman's nonparametric rank correlation coefficient. The unpaired Student's t-test was used to compare SCfat and IAfat. One-way ANOVA was used to compare the adipose tissues and the liver, or adipose tissues, the liver, and serum. The least significant difference test was used for the post hoc analyses. Calculations were made using GraphPad Prism version 4.00 for Windows (GraphPad Software, San Diego, CA), Systat Statistical Package (Systat version 10; Systat, Evanston, IL), and SPSS 14.0 for Windows (SPSS, Chicago, IL). A P value of <0.05 was considered statistically significant.
Lipid composition of SCfat and IAfat
Because phospholipids are mainly present on cell membranes, we normalized lipid concentrations of adipose tissues to cell size. There were no differences in fat cell size between SCfat (108 ± 2 µm) and IAfat (107 ± 2 µm, not significant). IAfat contained more ceramides (9.6 ± 0.8 vs. 6.0 ± 0.4 nmol/mg tissue/µm, IAfat vs. SCfat, P = 0.003) and less PC(e) (2.7 ± 0.2 vs. 3.6 ± 0.02 nmol/mg tissue/µm, P = 0.009) than SCfat after normalization to fat cell size. There were no differences in any other lipid classes normalized to fat cell size (data not shown). The results remained unchanged when the concentrations or percentages of lipids without normalization to fat cell size were compared (Figure 1, Table 2). IL-6 protein concentrations were higher in SCfat (0.20 ± 0.04 pg/mg protein) than in IAfat (0.09 ± 0.02 pg/mg protein, P = 0.023), whereas TNFα protein concentrations were comparable between adipose tissue depots (1.46 ± 0.34 vs. 1.31 ± 0.31 pg/mg protein, SCfat vs. IAfat, not significant).
Table 2. The absolute (nmol/mg tissue) lipid composition of subcutaneous and intra-abdominal adipose tissue, and the liver
Lipid composition of the liver as compared to adipose tissue depots
In both SCfat and IAfat, triacylglycerols represented the majority (99.2% in both intra-abdominal and subcutaneous fat) of lipids. The rest (0.8%) was phospholipids. In the liver, the respective percentages were 75.5 and 24.5%. The absolute amounts of ceramides, SM, PC, PE, PE(e), Lyso(tot), and LPC were higher in the liver as compared to adipose tissues (Table 2). The amounts of PC(e), LPE, LPE(e), and triacylglycerol were lower in the liver than in either adipose tissue depots. The concentrations of DAG were similar in the liver as compared to adipose tissues (Table 2). The % of ceramides, SM, GPA, PC, PC(e), PE, PE(e), Lyso(tot), LPC, and DAG were significantly higher in the liver than in either of the adipose tissue depots (Figure 1 and Supplementary Table S1 online).
Fatty acid composition and desaturase activities of the liver, and subcutaneous and intra-abdominal fat
The composition of esterified and free fatty acids was similar between SCfat and IAfat (Table 3, Figure 2). The liver and adipose tissues contained similar relative amounts of the essential linoleic (18:2n-6) and alpha-linolenic (18:3n-3) acids, and of free linoleic acid (Table 3). The liver contained 1.3–1.4-fold more saturated palmitic (16:0) and 2.6–2.7-fold more stearic (18:0) fatty acids than either adipose tissues. The liver also contained greater proportions of polyunsaturated gamma-linolenic (18:3n-6), dihomo-γ-linolenic (20:3n-6), arachidonic (20:4n-6), eicosapentaenoic (20:5n-3), and docosahexaenoic (22:6n-3) acids than adipose tissues. The liver contained less myristatic (14:0), palmitoleic (16:1n-9), and oleic (18:1n-9) acids than either adipose tissue.
Table 3. The proportional fatty acid composition of subcutaneous and intra-abdominal adipose tissue, the liver, and serum
The estimated activity of Stearoyl-CoA desaturase (SCD1), determined both from 16:1/16:0- and 18:1/18:0-ratios, was significantly lower in the liver as compared to either adipose tissue (Figure 3). In contrast, the estimated activities of Δ5 and Δ6 desaturase were higher in liver than in either adipose tissue depot (Figure 3). Elongase activity estimated from the 18:0/16:0-ratio but not from the 18:1/16:1-ratio was higher in the liver than in adipose tissues (P < 0.0001 for ANOVA, Figure 3). The estimated SCD1 and elongase activities were inversely related to relative amounts of long polyunsaturated fatty acids in the liver but not in adipose tissues (Table 4).
Table 4. Relationships between palmitic free fatty acids, estimated SCD1 and elongase activities, and long polyunsaturated fatty acids, and SCfat, IAfat, and the liver
Lipid composition of serum in relation to the liver, and subcutaneous and intra-abdominal fat
The lipid (triacylglycerol and phospholipid) composition of serum was more closely related to that of the liver than of adipose tissues (Figure 1 and Supplementary Table S1 online). The relative amounts of PC, ceramides, and DAG were similar in serum and the liver. PC(e), PE(e), SM, Lyso, LPC, LPE, and LPE(e) were more abundant in serum as compared to the liver, whereas GPA, PE, and triacylglycerol were less abundant in serum than in the liver. The proportions of palmitic (16:0), palmitoleic (16:1n-7), stearic (18:0), linolenic (18:3n-3), and docosahexaenoic (22:6n-3) acids, and the activities of SCD1, Δ6 desaturase, and elongases were similar in the liver and serum (Figure 3).
The estimated activities of SCD1 (r = 0.86, P = 0.024 [16:1/16:0]) and elongase (r = 0.89, P = 0.01 [18:1/16:1]) reflected by serum were positively correlated to those in the liver but not to those in adipose tissue. The estimated Δ5 (r = 0.89, P = 0.012) and Δ6 (r = 1.0, P < 0.001) desaturase activities reflected by serum were closely related to those in the liver but not adipose tissues.
Ceramides, sphingomyelins, and palmitic free fatty acids in the liver and adipose tissues
Palmitic free fatty acid concentrations were significantly lower in the liver as compared to adipose tissues (Figure 4). Sphingomyelin and ceramide concentrations were higher in the liver as compared to intra-abdominal and subcutaneous tissues. The relative amounts of palmitic free fatty acid in all three tissues were not related to the relative amounts of ceramides (data not shown). The % ceramides was tightly related to the % SM in the liver (r = 0.91, P = 0.002) and serum (r = 0.90, P = 0.037) but not in SCfat (r = 0.41, not significant) or IAfat (r = 0.52, not significant).
To the best of our knowledge, this is the first study in which lipidome compositions of the liver, SCfat and IAfat, and serum have been compared in humans. Because of ethical limitations to sample human liver and IAfat, the tissues analyzed were obtained from obese subjects undergoing laparoscopic gastric bypass operation. The liver contained more ceramides, sphingomyelins, phosphatidylethanolamines, lysophosphatidylcholines, less ether-linked phosphatidylcholines but similar amounts of diacylglycerols as compared to SCfat and IAfat. Serum lipid and fatty acid composition as well as the estimated activities of desaturases and elongases reflected by serum more closely resembled that in the liver than adipose tissue. The estimated activities of desaturases and elongases in the liver but not in the adipose tissues were directly related to the estimated activities of these enzymes reflected by serum.
The absolute and relative amounts of ceramides were 1.5–3-fold and 6–7-fold higher in the liver as compared to SCfat and IAfat. Ceramides can be generated either via de novo biosynthesis from saturated fatty acids or from sphingomyelins via sphingomyelinase (7). Sphingomyelins were 7–8-fold higher in the liver than in the adipose tissues, and their proportions were tightly related to those of ceramides, suggesting that the activity of sphingomyelin synthase (7) in the liver may be linked to substrate availability. The absolute amounts of palmitic free fatty acids in the liver were significantly lower than those in the adipose tissue and not correlated with ceramides. This does not exclude the possibility that de novo biosynthesis from palmitate contributes significantly to the generation of ceramides in the liver. Amounts of diacylglycerols, another class of bioactive lipids associated with hepatic steatosis and insulin resistance both in animal models (29) and humans (19), did not differ between the liver and adipose tissue depots.
The lipid composition of the liver and serum differed significantly from that of both SCfat and IAfat. The two adipose tissues were nearly identical when lipid data were expressed per cell size or given as absolute concentrations. However, IAfat contained more ceramides than subcutaneous fat, which could contribute to greater insulin resistance of antilipolysis in this tissue (7,30). Protein expressions of neither TNFα nor IL-6 in human SCfat and IAfat depots have been previously compared. We found protein concentrations of TNFα to be comparable and those of IL-6 higher in subcutaneous as compared to IAfat. These data imply that neither TNFα nor IL-6 was the cause or the consequence of higher ceramide concentrations in intra-abdominal as compared to SCfat.
In SCD1-deficient mice, the total ceramide content in skeletal muscle is ∼40% lower as compared to wild-type mice (31), implying that the SCD1 activity may be related to ceramide metabolism. These mice are resistant to development of obesity, insulin resistance, and hepatic steatosis (32,33). However, in the present study, we did not observe any relationship between ceramide content and SCD1 activity in the liver, adipose tissues, or serum (data not shown). Hepatic expression of genes involved in fatty acid β-oxidation in SCD1-deficient mice is increased, and those involved in lipid synthesis decreased (32). The present study in humans shows that estimated SCD1 activity in the liver but not in adipose tissue is inversely related to proportional amounts of long-chain polyunsaturated fatty acids, including docosahexaenoic acid (22:6n-3), a marker of peroxisomal β-oxidation (34,35). This is consistent with in vitro studies showing that polyunsaturated fatty acids suppress SCD1 gene expression (36) and SREBP-1c (37,38), which activates the transcription of multiple rate-limiting genes responsible for fatty acid synthesis (39,40,41). In the present study, we show for the first time in the human liver that polyunsaturated fatty acids, especially the essential linoleic acid (18:2n-6), is inversely correlated with saturated fatty acids (16:0). These data would suggest that the decrease of polyunsaturated fatty acids in the liver may contribute to the pathogenesis of hepatic steatosis (42,43), and suggest the dietary recommendations of increasing intake of essential fatty acids in the treatment of nonalcoholic fatty liver disease (43).
A recent 20-year prospective study indicated that the activities of SCD1, and Δ5 and Δ6 desaturases predict the metabolic syndrome (44). Moreover, serum fatty acid composition associates with insulin resistance in cross-sectional studies (14,45,46,47), and proportional changes in the composition of fatty acids in serum are related to the changes in insulin sensitivity independent of body weight (48,49,50). In the present study, we show that the estimated activities of desaturases in the liver are similar and tightly related to those reflected by serum. Two subjects in the present study had type 2 diabetes, which could have affected their fatty acid composition. This did not seem to be the case but cannot be determined with certainty because of the low number of subjects and thus remains a weakness of the present study. These findings imply that insulin resistance reflected by estimated serum desaturase activities (44,48) primarily reflects changes in the hepatic lipid metabolism.
The sources of fatty acids of the liver include dietary fatty acids, which can enter the liver either by uptake of intestinally derived chylomicron remnants, or through spillover into the plasma fatty acid pool; de novo lipogenesis; and peripheral lipolysis (10,51). The relative amounts of essential fatty acids (18:2n-6 and 18:3n-3) in the present study were similar in the liver and adipose tissues, suggesting that distribution of dietary fatty acids, both essential and nonessential, do not differ between the tissues. The liver contained proportionally more palmitic (16:0) and stearic (18:0) acids than adipose tissues. This is consistent with data showing that de novo lipogenesis occurs in the liver (52,53) and may thus contribute to differences in fatty acid composition between the tissues.
This study was supported by research grants from the Academy of Finland, the Sigrid Juselius Foundation, the Finnish Diabetes Research Foundation, Biomedicum Helsinki Foundation, and Novo Nordisk Foundation. This work is part of the project “Hepatic and adipose tissue and functions in the metabolic syndrome” (http:www.hepadip.org), which is supported by the European Commission as an integrated project under the Sixth Framework Programme (Contract LSHM-CT-2005-018734). We gratefully acknowledge Mia Urjansson, Katja Sohlo, and Laxman Yetukuri for excellent technical assistance, and the volunteers for their help.