Fasting Plasma Insulin Concentrations Are Associated With Changes in Hepatic Fatty Acid Synthesis and Partitioning Prior to Changes in Liver Fat Content in Healthy Adults

Nonalcoholic fatty liver disease (NAFLD), the hepatic manifestation of the metabolic syndrome (1), encompasses a spectrum of conditions from hepatic steatosis through to cirrhosis (2); obesity is a known risk factor. Why intrahepatocellular fat starts to accumulate remains unclear, but it is likely to involve an imbalance between fatty acid (FA) delivery to the liver, FA synthesis and oxidation within the liver, and triglyceride (TG) export from the liver (3). Insulin plays a key role in all of these processes.

Within the liver, insulin has dual action: 1) it stimulates the phosphorylation of the transcription factor Forkhead box protein O1 (FoxO1), which activates gluconeogenesis, and 2) it activates the transcription factor sterol regulatory element-binding protein 1c (SREBP-1c), which enhances the transcription of genes required for FA and TG synthesis (4,5). The induction of FA synthesis (de novo lipogenesis [DNL]) may contribute to insulin resistance (6). In insulin resistance, the FoxO1 pathway becomes resistant to insulin, so gluconeogenesis continues while insulin sensitivity is maintained in the SREBP-1c pathway, leading to accelerated DNL (4). Enhanced hepatic DNL may have significant qualitative implications because the primary FA product is saturated (palmitoyl-CoA) (7,8) which may interfere with cellular function (9), and the entry of fatty acyl-CoA into the mitochondrion is dependent on carnitine palmitoyltransferase 1 (CPT1); malonyl-CoA, an intermediate in the DNL pathway, is a potent inhibitor of this (10). Taken together, this may lead to enhanced VLDL-TG production and a net retention of intrahepatocellular TG (4,5). Positive associations between hepatic DNL and VLDL-TG production rates have been reported (11,12).

Insulin plays a key role in regulating FA delivery to the liver. In the fasting state, plasma nonesterified FAs (NEFAs) arise predominantly from the hydrolysis of adipose tissue TG. Fasting plasma insulin concentrations have been inversely associated with NEFA release from subcutaneous abdominal adipose tissue (13). Plasma NEFA concentrations decrease after the consumption of a mixed meal due to the antilipolytic action of insulin suppressing the hydrolysis of adipose tissue TG. Spillover FAs derived from the peripheral lipoprotein lipase-mediated lipolysis of chylomicron-TG (14–16) may somewhat reduce but do not override this effect (13).

Hepatic steatosis is often seen in the context of hepatic insulin resistance, but whether hepatic steatosis causes insulin resistance or whether insulin resistance causes hepatic steatosis is unclear (17). Because insulin has the potential to influence hepatic FA synthesis and postprandial partitioning, we aimed to determine the effect of global insulin resistance on hepatic FA synthesis and partitioning in healthy men and women in the fasted state and after the consumption of a mixed test meal.

The study recruited 37 subjects from the Oxford BioBank (www.oxfordbiobank.org.uk) (18) and by advertisement; of these, 19 were considered hyperinsulinemic (HI), with a fasting plasma insulin concentration greater than the 75th centile (11.2 mU/L) of the Oxford BioBank (14), and 18 were considered normoinsulinemic (NI). All volunteers were considered nondiabetic and free from any known disease, had a BMI of <30 kg/m², were not taking medication known to affect lipid or glucose metabolism, did not smoke, and did not consume alcohol above recommended limits (2). The study was approved by Portsmouth Clinical Research Ethics Committee, and all subjects gave written informed consent. Data from a portion of the subjects reported in this work were obtained as part of another previously published study (19).

Intrahepatic lipid content was measured after an overnight fast and within 2 weeks of the metabolic study day by proton MRS (¹H-MRS) (20). As part of the ¹H-MRS spectra of liver metabolites, hepatic glycogen was measured (one peak, chemical shift 3.984), and content as a percentage of the liver water determined. Whole-body composition and fat distribution were measured using DEXA (21).

Before the study day, subjects were asked to avoid foods naturally enriched in ¹³C, alcohol, and strenuous exercise. The evening before the study day, subjects consumed deuterated water (²H₂O) (3 g/kg body water) and continued to consume ²H₂O during the course of the study day for the measurement of fasting and postprandial hepatic DNL (19). On the study day, after an overnight fast and consumption of ²H₂O, subjects came to the Clinical Research Unit. A cannula was inserted into an antecubital vein, and baseline (time 0) blood and breath samples were taken. Participants were fed a mixed test meal containing 40 g carbohydrate and 40 g fat, with 200 mg of [U¹³C]palmitic acid added to trace the fate of the dietary FA (19), and at 360 min were given a glucose drink (75 g glucose) to assess the second meal effect (22). Repeated blood and breath samples were taken during the study period. Indirect calorimetry was performed in the fasting state and then 120 min after meal consumption using a GEM calorimeter (GEMNutrition Ltd., Daresbury, Cheshire, U.K.) to determine whole-body CO₂ production, whole-body respiratory exchange ratio (RER), and basal energy expenditure.

Whole blood was collected into heparinized syringes (Starstedt, Leicester, U.K.), and plasma was rapidly separated by centrifugation at 4°C for the measurement of plasma metabolite and insulin concentrations, as previously described (14).

Separations of chylomicron of Svedberg flotation rate (S_f) >400 and VLDL-rich fraction (S_f20-400) were made by sequential flotation using density gradient ultracentrifugation, as previously described (14). The S_f 20-400 fraction was further separated by immunoaffinity chromatography to obtain a fraction completely devoid of apolipoprotein (apo)B48 and will hereafter be called VLDL (14).

Samples were taken at 0, 30, 60, 90, 120, 180, 240, 300, 360, 390, and 420 min after the consumption of the test meal for the measurement of plasma glucose, insulin, TG, NEFA, 3-hydroxybutyrate (3OHB), chylomicron-TG, and TG-rich lipoproteins (TRL)-TG and at 0, 180, 240, 300, 360, and 420 min for the analysis of VLDL-TG. Breath samples were collected at 0, 60, 90, 120, 180, 240, 300, 360, 390, and 420 min into Exetainer tubes (Labco Ltd., High Wycombe, Bucks, U.K.) for measurement of ¹³CO₂ enrichment.

To determine the specific FA composition and isotopic enrichment, total lipids were extracted from plasma and lipoproteins and FA methyl esters (FAMEs) prepared (19,23). The FA compositions (µmol/100 µmol total FA) in these fractions were determined by gas chromatography (GC), and palmitate concentrations calculated (14).

[U¹³C]palmitate enrichments were measured in plasma NEFA, TG, S_f >400 (chylomicron-TG), S_f 20-400–TG and VLDL-TG FAMEs derivatives using a Δ Plus XP GC-combustion isotope ratio mass spectrometer (Thermo Electron, Bremen, Germany) (24). The tracer-to-tracee ratio (TTR) of a baseline measurement before administration of [U¹³C]palmitate was subtracted from the TTR of each sample to account for natural abundance. The TTRs for [U¹³C]palmitate were multiplied by the corresponding palmitate concentrations to give plasma and lipoprotein tracer concentrations (25).

The ¹³C-to-¹²C ratios in breath samples and the relative rate of whole-body meal-derived FA oxidation were calculated, as previously described (24). The rate of expiration of ¹³CO₂ in breath was calculated by multiplying Vco₂ (mmol/min) by the TTR of the ¹³CO₂-to-¹²CO₂ ratio (24). To allow for sequestration of label into the bicarbonate pool, a dietary acetate recovery factor of 51% was applied (26). The data were corrected for lean mass (determined by DEXA) to account for individual differences between the NI and HI groups. Hepatic ketone body production was assessed by measuring the isotopic enrichment from [U¹³C]palmitate in 3OHB in deproteinized plasma (27).

Fasting and postprandial hepatic DNL was assessed based on the incorporation of deuterium from ²H₂O in plasma water (Finnigan GasBench II Thermo Fisher Scientific, Paisley, U.K.) into VLDL-TG palmitate using GC-mass spectrometry with monitoring ions with mass-to-charge ratios of 270 (M+0) and 271 (M+1) (28). For simplicity, DNL refers to the proportion of newly synthesized palmitate in VLDL-TG, and this represents the synthesis of FA from nonlipid precursors (29). To assess the partitioning of DNL palmitate into desaturation pathways, the ratio of [²H₂]16:1n-7 to [²H₂]16:0 in VLDL-TG was determined as a marker of hepatic stearoyl-CoA desaturase 1 (SCD1) activity (30).

HOMA of insulin resistance was calculated (31). The relative and absolute contribution of meal FA to VLDL-TG was calculated at the end of the study (time 420 min) (22).

Data were analyzed using SPSS 22 software (SPSS [UK] Ltd., Chertsey, U.K.). Data are presented as means ± SEM unless otherwise stated. Areas under the curve were calculated by the trapezoid method and have been divided by the relevant time period to give time-averaged values. For anthropometric data, comparisons between the groups were made using a general linear univariate model with sex as a covariate. All data sets, with time and group as factors, were used to investigate the change between groups over time for specific metabolites. Associations between variables were carried out using the Spearman rank correlation coefficient for the respective groups.

NI subjects were slightly older, with marginally lower amounts of total, android, and visceral fat masses (P < 0.05), despite a similar BMI, than HI subjects (Table 1). Liver fat and glycogen content was similar between the groups (Table 1). The HI group had significantly higher (P < 0.05) fasting concentrations of plasma glucose and TG than the NI group (Table 1). Fasting plasma VLDL-TG concentrations tended (P = 0.07) to be higher in the HI compared with the NI group (Table 1).

Consumption of the mixed test meal exacerbated the differences in fasting plasma glucose and insulin concentrations, with significantly higher postprandial excursions occurring in the HI group (P < 0.05) compared with the HI group (Fig. 1A and B). Fasting differences in plasma TG concentrations were not maintained over the postprandial period, with no difference between the groups (Fig. 1D). In line with the fasting data, there were no notable differences in the postprandial response in plasma NEFA or 3OHB concentrations (Fig. 1C and E). Whole-body RER tended (P = 0.07) to be higher in HI than in NI individuals in the fasting state but was significantly (P < 0.05) higher in HI than in NI subjects during the postprandial period, indicative of lower FA oxidation (Fig. 1F).

Inclusion of [U¹³C]palmitate into the mixed test meal provided the opportunity to trace the fate of dietary FA. There was no difference in the appearance of [U¹³C]palmitate in plasma chylomicron-TG between the groups (Fig. 2A). After the consumption of the second meal (at 360 min) the amount of [U¹³C]palmitate (from the first meal) incorporated into chylomicron-TG at 420 min was similar between groups, suggesting no difference in the second meal effect. The appearance of [U¹³C]palmitate in the plasma NEFA pool was significantly higher (P < 0.05) in the NI compared with the HI group (Fig. 2B), but there was no difference between the groups in the appearance of [U¹³C] in VLDL-TG (Fig. 2C). We calculated the contribution of meal-derived FA to VLDL-TG at 420 min and found a significantly (P < 0.05) lower relative contribution in the HI compared with NI group (mean ± SEM: 11% ± 1 vs. 14% ± 1, P < 0.05); this difference disappeared when expressed as an absolute concentration. In the fasting state, NI individuals had a significantly (P < 0.05) lower relative contribution of DNL to VLDL-TG compared with HI individuals. The contribution of DNL to VLDL-TG increased over the postprandial period (P < 0.01) in both groups, with the difference observed in the fasting state between the groups becoming less obvious (P = 0.07) (Fig. 2D). We assessed the [²H₂]16:1n-7–to–[²H₂]16:0 ratio in VLDL-TG, as a marker of the desaturation of DNL palmitate and found it was significantly (P < 0.01) higher in NI compared with HI individuals in the fasting state and at the end of the postprandial period (420 min) (Table 1).

We assessed dietary FA oxidation by measuring the incorporation of ¹³C (from dietary fat) in plasma 3OHB as a marker of hepatic FA oxidation and in breath CO₂ as a marker of whole-body dietary FA oxidation. We found a significantly (P < 0.05) greater incorporation ¹³C into plasma 3OHB in NI compared with HI individuals during the postprandial period (P < 0.05) (Fig. 2E). In line with this and the difference in postprandial RER, the production of ¹³CO₂ (per unit lean mass) tended (P = 0.07) to be higher in the NI group compared with the HI group (P = 0.05 for time-by-group interaction) (Fig. 2F).

To assess the effect of increased liver fat content on fasting and postprandial FA synthesis and partitioning, we compared 10 NI and 9 HI individuals with a liver fat content >3.4%, the median of groups. Fasting plasma insulin was significantly (P < 0.01) higher in the HI compared with the NI group, as were postprandial plasma insulin and glucose concentrations (Supplementary Table 1). The groups did not differ in the appearance of [¹³C] from the dietary fat into plasma chylomicron-TG, VLDL-TG, and 3OHB, or in fasting or postprandial hepatic DNL. Incorporation of [¹³C] into the plasma NEFA pool was higher (P < 0.05) in the NI compared with HI group (Supplementary Table 1).

We found a positive association between fasting plasma insulin and liver fat content in the NI and HI groups (Fig. 3A). Fasting insulin concentrations were also positively associated with liver glycogen content, but only in the NI (r_s = 0.61, P < 0.05) and not in the HI (r_s = 0.24, P = NS) group (data not shown). When combined, we found a positive association between liver fat and glycogen content (r_s = 0.38, P < 0.05, n = 35 [NI: r_s = 0.35 and HI: r_s = 0.39, P = NS for both; data not shown]). There was a significant positive association between fasting plasma insulin concentrations and fasting hepatic DNL in the HI but not the NI group (Fig. 3B). Robust inverse associations were found between fasting hepatic DNL and fasting plasma 3OHB concentrations for NI (r_s = −0.65, P < 0.01) and for HI (r_s = −0.54, P < 0.05) (data not shown).

We observed a positive association between the postprandial response in plasma VLDL-TG concentrations and hepatic DNL in the NI (P < 0.05) but not the HI group (Fig. 3C). There was a robust inverse association between the postprandial response in hepatic DNL and the incorporation of ¹³C, representing recently ingested dietary fat, in plasma 3OHB in the NI group (P < 0.001), but this association was diminished in the HI group (Fig. 3D).

Hepatic steatosis is often accompanied by hepatic insulin resistance; whether hepatic steatosis causes insulin resistance or vice versa remains unclear (17). We defined individuals as HI on the basis of their fasting plasma insulin concentration (14). We found HI individuals had marginally more total and visceral fat than NI individuals but liver fat content was similar between the groups. Despite NI and HI individuals having a similar amount of liver fat, we observed profound differences in fasting plasma glucose, insulin, and TG concentrations. Fasting and postprandial hepatic DNL were notably higher in the HI group than in the NI group. The HI subjects had significantly lower dietary FA oxidation, and the difference between the groups was augmented in the postprandial state. These observations were not evident in NI and HI individuals with a liver fat content >3.4%. Our findings demonstrate that hepatic steatosis does not need to be present to induce changes in intrahepatic FA metabolism; HI induces changes in FA partitioning that would, if maintained over a period of time, lead to accumulation of liver fat.

Within the liver, insulin integrates carbohydrate and lipid metabolism where they are directed to storage as TG and glycogen. We measured liver fat and glycogen content and found no difference between the groups in either; we did find liver fat content was positively associated with liver glycogen content. Animal studies have suggested in insulin resistance, portal hyperinsulinemia drives FoxO inactivation, leading to a decrease in the hepatic glucose 6-phosphatase catalytic subunit–to–glucokinase ratio and increased hepatic DNL, TG, diacylglycerol, and glycogen content (32). Our findings of higher hepatic DNL in HI compared with NI individuals are in line with this concept. We found a positive association between fasting plasma insulin concentrations and fasting hepatic DNL in the HI group only. Enhanced DNL leads to an increase in newly formed TG that will reside within the liver or be exported within VLDL (7,8). Hepatic DNL was positively associated with VLDL-TG concentrations in the NI but not the HI group. It could be speculated that DNL FAs were preferentially channeled toward secretion in VLDL in the NI group and toward storage in the HI group. Animal work has suggested DNL FAs exit the liver immediately as VLDL-TG rather than being stored (33). Although evidence for this is sparse in humans (34,35), hepatic DNL has been positively associated with VLDL-TG production rates (11,12).

DNL has been proposed as a pathway for sustaining metabolic homeostasis, and although an energetically inefficient way to store excess energy, it is an important mechanism for glucose disposal (36). Despite subjects consuming two test meals, we did not observe a marked divergence in postprandial hepatic DNL between the groups. It is plausible that the pattern of hepatic DNL would notably differ between the groups if followed for longer, with the HI group having a greater lipogenic response to the second meal. The induction of hepatic DNL has been suggested to contribute toward insulin resistance (6); we cannot distinguish whether an increase in DNL caused insulin resistance or vice versa.

Enhanced DNL increases the production of long-chain saturated fatty acyl-CoAs (e.g., palmitoyl-CoA) (7,8). A potential fate of newly synthesized palmitoyl-CoA is partitioning toward desaturation by SCD1 (30). We measured the ratio of [²H₂]16:1n-7 to [²H₂]16:0 in VLDL-TG as a marker of SCD1 activity and found the ratio was significantly higher in the fasting and postprandial states in NI individuals than in HI individuals. It is plausible greater desaturation of newly formed palmitate to palmitoleate would prevent accumulation of intrahepatocellular palmitoyl-CoA. Evidence from animal and cellular studies (37,38) suggests that lipotoxicity arising from the accumulation of long-chain FA is specific to saturated FA, with increased accumulation causing cell dysfunction (9). Palmitoyl-CoA can be used for the synthesis of intracellular ceramide; saturated FA and ceramides have both been suggested to upregulate proinflammatory pathways and proinsulin resistance factors (6). The factors influencing the partitioning of newly synthesized palmitoyl-CoA toward specific metabolic pathways remain unclear.

The appearance of [¹³C]palmitate in the systemic NEFA pool, consistent with spillover from chylomicron-TG hydrolysis (39), was higher in the NI than in the HI group, despite no difference in systemic NEFA concentrations. This is consistent with our observation that adipose tissue FA trafficking was downregulated in abdominally obese HI men compared with lean NI men (15). The contribution of systemic NEFA to VLDL-TG production has been reported to be similar between insulin-sensitive and insulin-resistant individuals (14) and those with and without NAFLD (40,41).

Removal of FA within the liver occurs by secretion as TG in VLDL or via oxidation. Although we did not measure VLDL-TG production rates, others (42,43) have reported VLDL apoB and TG production rates are higher in obese, insulin-resistant individuals compared with lean, insulin-sensitive individuals. In contrast, acute induction of hyperinsulinemia in healthy insulin-sensitive men suppresses the total production rate of VLDL apoB and TG (44), even in the presence of excess NEFA concentrations (45). On the basis of these observations, it could be speculated that the HI individuals in the current study had a higher VLDL-TG production rate than the NI individuals. Others have reported individuals with NAFLD have a higher VLDL-TG secretion rate than those without (41,46). We did not observe a difference in the appearance of [U¹³C]palmitate (from dietary fat) in VLDL-TG between the groups; however, our findings demonstrate that β-oxidation occurs in a proportion of dietary-derived FA entering the liver and the acetyl-CoA liberated enters a pool that is used for ketogenesis; this occurred to a greater extent in NI than in HI individuals. We did not measure the 3OHB production rate; thus, the absolute contribution of dietary FA into the ketogenic pathway cannot be determined. Production of ¹³CO₂ was also lower in HI compared with NI individuals.

In support of these differences, we found NI compared with HI individuals had a significantly lower fasting and postprandial whole-body RER, which was indicative of FA oxidation. We found a strong inverse association between postprandial hepatic DNL and the plasma [¹³C]3OHB in the NI group that was not evident in the HI group. The findings from the NI group clearly highlight a divergence in FA partitioning suggesting intracellular metabolism is being moved away from esterification toward oxidation. The switch in intracellular metabolism was not so evident in the HI group; it could be speculated that the lack of association between these pathways signals an attempt to dispose of excess intrahepatic glucose and FA.

Our study has some limitations. Although subjects consumed two test meals, peak hepatic DNL occurs ∼4–5 h after consumption of a second meal (47) and was not achieved due to the short duration of our study. We did not determine the VLDL-TG, apoB, or DNL production rate; therefore, quantitative differences in the contribution of dietary and DNL FA to VLDL-TG production cannot be determined. The production rate from dietary fat and DNL to VLDL-TG has been reported to be 0.46 µmol/min and 0.78 µmol/min in individuals without NAFLD and 0.56 µmol/min and 2.57 µmol/min in individuals with NAFLD (41). In contrast, the absolute contribution of dietary fat to VLDL-TG was notably higher than the contribution from DNL when healthy men were given a liquid formula (32% fat) via duodenal infusion over a period of 11 h (48), demonstrating the contribution of dietary fat to VLDL-TG production will be dependent on the amount and frequency of consumption.

Insulin resistance is often associated with hepatic steatosis and is therefore suggested to have a central role in the development of hepatic steatosis (49); however, whether insulin resistance causes the development of steatosis or vice versa (17) remains unclear. Our data demonstrate that notable differences in hepatic FA metabolism are evident between NI and HI individuals across a spectrum of liver fat contents. It is plausible that if maintained over a period of time or further augmented, the alterations in intrahepatic FA synthesis and partitioning reported here may lead to net liver fat accumulation.

Acknowledgments. The authors thank Louise Dennis and Rachel Craven-Todd from the Clinical Research Unit at the Oxford Centre for Diabetes, Endocrinology and Metabolism (OCDEM) for excellent nursing provision, Costas Christodoulides, Jeremy Tomlinson, and Jonathon Hazlehurst (OCDEM) for medical cover, Marje Gilbert (OCDEM) for technical assistance, Sandy Humphreys (OCDEM) for helpful statistical advice and technical assistance, and Ross Milne (Ottawa Heart Institute) for antibodies. The authors thank the volunteers from the Oxford BioBank, National Institute for Health Research (NIHR) Oxford Biomedical Research Centre, for their participation. The Oxford BioBank (www.oxfordbiobank.org.uk) is also part of the NIHR National BioResource, which supported the recalling process of the volunteers.

Funding. This study was funded by the British Heart Foundation (FS/11/18/28633 to L.H.), the Henning and Johan Throne-Holst Foundation (C.P.), and NIHR Oxford Biomedical Research Centre (F.K.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. C.P., M.P., R.B., C.A.M., and L.H. conducted the study. C.P., M.P., R.B., C.A.M., and L.H. carried out sample and data analyses. C.P., M.P., R.B., S.N., F.K., and L.H. contributed to data analysis and wrote the manuscript. S.N. and F.K. reviewed and edited the manuscript. L.H. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.