Hepatic Fatty Acid Balance and Hepatic Fat Content in Humans With Severe Obesity

Abstract

Objective

Nonalcoholic fatty liver disease can lead to hepatic inflammation/damage. Understanding the physiological mechanisms that contribute to excess hepatic lipid accumulation may help identify effective treatments.

Design

We recruited 25 nondiabetic patients with severe obesity scheduled for bariatric surgery. To evaluate liver export of triglyceride fatty acids, we measured very-low-density lipoprotein (VLDL)–triglyceride secretion rates the day prior to surgery using an infusion of autologous [1-¹⁴C]triolein-labeled VLDL particles. Ketone body response to fasting and intrahepatic long-chain acylcarnitine concentrations were used as indices of hepatic fatty acid oxidation. We measured intraoperative hepatic uptake rates of plasma free fatty acids using a continuous infusion of [U-¹³C]palmitate, combined with a bolus dose of [9,10-³H]palmitate and carefully timed liver biopsies. Total intrahepatic lipids were measured in liver biopsy samples to determine fatty liver status. The hepatic concentrations and enrichment from [U-¹³C]palmitate in diacylglycerols, sphingolipids, and acyl-carnitines were measured using liquid chromatography/tandem mass spectrometry.

Results

Among study participants with fatty liver disease, intrahepatic lipid was negatively correlated with VLDL-triglyceride secretion rates (r = −0.92, P = 0.01) but unrelated to hepatic free fatty acid uptake or indices of hepatic fatty acid oxidation. VLDL-triglyceride secretion rates were positively correlated with hepatic concentrations of saturated diacylglycerol (r = 0.46, P = 0.02) and sphingosine-1-phosphate (r = 0.44, P = 0.03).

Conclusion

We conclude that in nondiabetic humans with severe obesity, excess intrahepatic lipid is associated with limited export of triglyceride in VLDL particles rather than increased uptake of systemic free fatty acids.

Nonalcoholic fatty liver disease (NAFLD) is defined as excess accumulation of intrahepatic lipid (IHL), primarily triglycerides (TGs). NAFLD may progress to nonalcoholic steatohepatitis (NASH), characterized by inflammation, oxidative stress, and fibrosis. Patients with NAFLD often have other metabolic diseases, including insulin resistance, obesity, type 2 diabetes, and adverse plasma lipid profiles. It has been reported that 84% to 93% of patients with severe obesity, including those with diabetes and other comorbidities, undergoing bariatric surgery have some degree of hepatosteatosis (1, 2).

The underlying mechanisms that contribute to excess hepatic lipid accumulation are likely multifactorial, but inevitably must relate to the balance of fatty acid uptake/synthesis and fatty acid disposal. In patients with NAFLD/NASH, more than half of hepatic TG is derived from circulating free fatty acids (FFAs) (3). In metabolically healthy, nonobese adults hepatic FFA uptake is dependent on the total plasma FFA concentration (4), but decreasing hepatic FFA uptake does not reduce liver fat (4). This may not be surprising, because even in lean adults the liver contains certain amounts of structural lipids, including phospholipids, TGs, cholesterol, and sphingolipids (5). Under normal circumstances alterations in hepatic fatty acid uptake are eventually offset by changes in fatty acid disposal. However, in humans with NAFLD it remains unclear whether greater fatty acid uptake or limited fatty acid disposal plays a primary role in the pathology of excess IHL.

If excess IHL in cytoplasmic lipid droplets arises from reduced ability to dispose of fatty acids, this could be due to impaired export of these fatty acids in very-low-density lipoprotein (VLDL)–TG or reduced fatty acid oxidation. In healthy adults, the export of TGs (and cholesterol ester) from the liver as VLDL particles is induced by fasting and acutely suppressed by insulin through both direct and indirect effects (6, 7). Reports of VLDL production in humans with NAFLD are somewhat conflicting, likely because of the differing methodology used and different study populations (8–10).

Although much of the literature regarding NAFLD has focused on TGs and cholesterol, other classes of lipids, such as diacylglycerols (DGs) and sphingolipids, may contribute to the pathogenesis of NAFLD (11, 12). Both DGs (13) and sphingolipids accumulate in steatotic human livers (11, 14), and these lipids exist as constituents of circulating VLDL particles (15–17)

Patients undergoing bariatric surgery represent a unique population to interrogate the relationship between hepatic fat content and the balance between uptake and disposal of fatty acids. They have a greater probability of NAFLD and surgery is scheduled on an elective basis, allowing for complex studies to be meshed with their clinical care. Herein we report our findings of the relationship between total hepatic lipid content, VLDL-TG secretion rates, hepatic FFA uptake rates, and the contribution of systemic FFA to intrahepatic TGs, DGs, acyl-carnitines, and sphingolipids among patients with severe obesity without diabetes.

Experimental Procedures

Participants

Twenty-five adults 18 to 55 y of age, BMI >35 kg/m², scheduled for elective bariatric surgery participated in this study. Patients undergoing bariatric surgery at the Mayo Clinic are not mandated to undergo a prior weight loss regimen, and thus all participants were weight stable for 3 months prior to surgery. All participants were white. Patients taking lipid-lowering medications (e.g., fibrates, statins, niacin) could participate only when their primary health care provider deemed it safe for them to discontinue their use 4 weeks prior to the study. Likewise, potential volunteers receiving beta-blockers must have been able to safely discontinue their use 3 days prior to the study to participate. Exclusion criteria included type 2 diabetes mellitus (T2DM), type 1 diabetes, presence or a history of liver disease other than NAFLD, use of nicotine, alcohol consumption >20 g/d, smoking, and the use of any medication known to affect fatty acid metabolism that could not be discontinued prior to the study. The study was approved by the Mayo Clinic Institutional Review Board, and informed, written consent was obtained from all volunteers.

Protocol

The protocol for the VLDL-TG turnover portion of the study has been described (18). Briefly, 1 week prior to the inpatient study visit, which was scheduled to occur the day prior to surgery, a 100-mL fasting blood sample was obtained for ex vivo VLDL-TG labeling with [1-¹⁴C]triolein (PerkinElmer, Boston, MA). At this time a dual-energy x-ray absorptiometry scan (Lunar iDXA, GE Healthcare, Madison, WI) and a single slice CT scan of the abdomen were performed to measure body composition (19). The following week participants were admitted to the Mayo Clinic inpatient clinical research unit the evening before the study. After fasting overnight, an IV catheter was placed in the forearm vein for tracer infusions and a retrograde IV catheter was placed in the contralateral hand vein to allow for collection of arterialized blood using the heated (55°C) hand vein technique (20). Following a baseline blood sample, a primed, continuous infusion of the volunteer’s own ex vivo–labeled [1-¹⁴C]VLDL-TG was started and maintained for 4 hours (Fig. 1). Resting energy expenditure (REE) and respiratory exchange ratios (RERs) were measured using indirect calorimetry at three time points during the [1-¹⁴C]VLDL-TG infusion. Blood samples were collected every 30 minutes during the infusion (excluding the 120 minute time point) to measure VLDL-TG turnover (21). Two participants were unable to complete the VLDL-TG turnover portion of this study. Plasma insulin concentrations were measured on the baseline sample and the sample was collected at 240 minutes. After stopping the infusion, additional blood samples were collected during a 2-hour period to assess the plasma β-hydroxybutyrate response to continued fasting. The participants were then provided with meals that approximated their usual food intake for the remainder of that day and then fasted overnight in preparation for surgery the next day.

The following morning a continuous infusion of [U-¹³C]palmitate (Isotec, Sigma-Aldrich, Miamisburg, OH) was started around 6:30 am in the preoperative waiting area and continued during surgery to measure FFA-palmitate rate of appearance and disappearance (flux). For the first six studies we used an infusion rate of 2 nmol·kg of fat-free mass⁻¹·min⁻¹, but then increased the rate to 8 nmol·kg of fat-free mass⁻¹·min⁻¹ to achieve greater enrichment in liver lipids. A radial artery catheter was placed by the anesthesiologist as part of the surgical procedure and was used for blood collection. To measure hepatic FFA uptake we gave an IV bolus of [9,10-³H]palmitate (100 µCi; PerkinElmer, Boston, MA) at the direction of the surgeon, who timed it such that the liver biopsy could be collected 30 minutes later. The 30-minute interval was chosen because this allows the tracer to be cleared from circulation and yet not appear in appreciable amounts in VLDL-TG (22); thus, the tissue tracer content reflects uptake and esterification. Arterial blood was collected at the time of the ³H-palmitate bolus (minute 0) and at 15 and 30 minutes after the bolus for measurement of palmitate concentration and enrichment.

Ex vivo VLDL-TG tracer preparation

The details regarding ex vivo labeling of VLDL-TG with [1-¹⁴C]triolein have been previously described in detail (18, 23). Briefly, an average of 38 ± 4 µCi of [1-¹⁴C]triolein was added to the plasma obtained from the 100-mL blood sample collected during the screening visit. After mixing the sample, the VLDL fraction was isolated using density gradient ultracentrifugation (50.3 Ti rotor, Optima™, LE-80K, Beckman Instruments, Spinco Division, Palo Alto, CA) and mixed with normal saline to achieve the final volume for infusion. All preparation procedures were performed under sterile conditions, and the final samples were tested for pyrogens and sterility prior to infusion.

VLDL-TG concentration and specific activity

As previously described (18), the VLDL fraction from plasma samples obtained during the [1-¹⁴C]VLDL-TG infusion were isolated by density gradient ultracentrifugation. The supernatant containing the VLDL particles was used to measure specific activity and TG concentration (mmol/L; Cobas Integra^® 400 Plus, Roche Diagnostics, Indianapolis, IN). Total VLDL-TG specific activity and concentration were determined from average, steady-state values during the last 2 hours of the infusion.

Plasma insulin concentrations, ApoB100 concentrations, and palmitate enrichment and concentration

Plasma insulin concentrations were measured using a chemiluminescent immunoassay (Sanofi Diagnostics Pasteur, Chaska, MN). ApoB100 concentrations were measured using a sandwich ELISA, with goat anti-human ApoB antibody (catalog no. 20A-G1b; Academy Biomedical, Houston, TX) and ApoB100 standard (Kamiya Biomedical, Seattle, WA) (24). ApoB100 concentrations were unavailable for two participants. Infusate and plasma [U-¹³C]palmitate enrichments were determined by liquid chromatography–mass spectrometry monitoring for [M+2 – H]⁻ and [M+16 – H]⁻ as previously described (25). Plasma palmitate concentration was determined by HPLC (26).

Intrahepatic lipid

Approximately 200 to 500 mg of liver tissue was harvested and weighed. A portion of the sample was quickly divided into several aliquots and frozen immediately in liquid nitrogen. The remainder of the biopsy was homogenized and lipids were extracted using chloroform/methanol (2:1). The organic layer from the lipid extraction was placed in a preweighed vial, dried, and reweighed to yield lipid weight (mg of lipid/g of liver). After the vial was weighed, scintillation cocktail was added and specific activity was measured using dual-channel liquid scintillation counting.

The β-hydroxybutyrate concentration in response to fasting was used as an indirect measure of ketone production and was measured using a commercially available colorimetric assay kit (Cayman Chemical, Ann Arbor, MI).

Hepatic palmitoyl-carnitine, ceramide, sphingolipid, and DG concentrations and enrichment

Hepatic palmitoyl-carnitine concentrations and enrichments were measured using liquid chromatography–tandem mass spectrometry (LC-MS/MS) as previously described (27). Hepatic ceramide and sphingolipid concentrations, as well as [U-¹³C]palmitate enrichment of C16-ceramide, were determined by LC-MS/MS (28). Hepatic DG concentrations as well as [U-¹³C]palmitate enrichment in hepatic 16/16 and 16/18:1 DGs were measured by LC-MS/MS (29). The following DG species concentration standards were used: 16/18:2, 18:1/18:2, 16/16, 16/18, 16/18:1, 18:1/18:1, and 18/18:1.

Calculations and statistical analysis

Data are presented as means ± SD. Because DGs with saturated fatty acids are proposed to be more potent signaling molecules, we calculated the DG saturation index as the ratio of the 16/16 plus 16/18 DG concentrations to the total DG concentrations. IHL >5.56% is considered an accepted definition for hepatic steatosis (30). Therefore, the data were analyzed using percentage IHL as a continuous variable and by group using 5.56% IHL as a cut-point for hepatic steatosis. Systemic palmitate flux was calculated by dividing the [U-¹³C]palmitate infusion rate by the average of the plasma palmitate enrichment for the last 30 minutes prior to the liver biopsy. Hepatic palmitate uptake rates (µmol/g of tissue) were calculated by multiplying the fraction of the tracer present in each gram of liver (³H disintegrations per minute per gram of liver tissue ÷ dose of ³H-palmitate) by systemic palmitate flux. The VLDL-TG concentration was divided by the ApoB100 concentration in the VLDL fraction to estimate the amount of TGs per VLDL particle. Because we measured the VLDL-TG secretion rates, we could then calculate the VLDL-ApoB100 secretion rates by multiplying the VLDL-TG secretion rates by the ratio of VLDL-TG to VLDL-ApoB100. Pearson correlation analysis was used to determine associations for variables of interest with a normal distribution. A Spearman rho was used to evaluate data that did not meet assumptions for parametric tests.

Results

Subject characteristics

Participant demographic, body composition, REE, VLDL-TG and palmitate kinetics, and plasma concentration data are provided in Table 1, and hepatic lipid data are provided in Table 2. Because all volunteers were undergoing bariatric surgery, the BMI of our participants averaged ∼48 kg/m²; 60% of volunteers were women. IHL content averaged 5% and the values ranged from 1% to 15%. Eight participants had >5.56% IHL. As expected, in this population of bariatric surgery patients with obesity, VLDL-TG and plasma palmitate concentrations were greater than those we have measured in nonobese populations and in class I populations with obesity (22). Plasma palmitate concentrations at the beginning of the surgery were 162 ± 43 µmol/L compared with 141 ± 25 µmol/L (P < 0.01) the previous morning during the VLDL turnover study. There was a good correlation (r = 0.63, P < 0.001) between concentrations observed on the 2 days, suggesting that the values we observed on the day of surgery were representative of the usual patterns of adipose tissue lipolysis for these patients. Plasma palmitate concentrations increased by ∼20% during the surgical procedure; the average concentrations and palmitate disappearance rates during the 30-minute interval following the [³H]palmitate bolus are provided in Table 1.

Hepatic lipid content, VLDL-TG secretion, and hepatic VLDL-TG storage

When analyzed by group (NAFLD and non-NAFLD) using the predetermined cut-point value of 5.56%, IHL was negatively correlated with VLDL-TG secretion rate (Fig. 2A, r = −0.92, P = 0.01) for those with NAFLD. There was no significant correlation between percentage hepatic lipid and VLDL-TG secretion rate for those without NAFLD. Given the appearance of the relationship (Fig. 2A and 2B), we conducted a post hoc secondary analysis that fitted a quadratic function to the data; there was a parabolic relationship (R² = 0.28) between VLDL-TG secretion rates and percentage IHL when data from all participants were analyzed together (Fig. 2B). The vertex of the fit for the x-axis was 7.9 (e.g., the amount of IHL at which the VLDL-TG secretion rate begins to decrease). Regardless of whether the data were analyzed using the 5.56% cut-point or as a quadratic using all data, our results indicate that decreased VLDL-TG secretion rates are associated with greater amounts of IHL for those with NAFLD. As expected, VLDL-TG secretion rates were positively correlated with plasma VLDL-TG concentrations (Fig. 2C, r = 0.48, P < 0.02). Plasma total TG concentrations were not correlated with percentage IHL (Fig. 2D). Plasma VLDL-TG concentrations tended to correlate with ApoB100 concentrations in the group as a whole (r = 0.34, P = 0.10), with no differences in this relationship when examined by IHL status (data not shown). ApoB100 secretion rate was not correlated with percentage IHL for the group as a whole (r = 0.07, P = 0.75), or for the groups with IHL <5.56% (r = 0.08, P = 0.75) or >5.56% (r = 0.51, P = 0.30).

We detected the [1-¹⁴C]oleate from the VLDL-TG tracer in hepatic lipids. The dose-adjusted hepatic content of ¹⁴C-labeled lipid 24 hours after the tracer infusion was positively associated with percentage IHL (Fig. 2E, r = 0.69, P < 0.001).

Hepatic palmitate uptake and IHL

To understand the factors relating to differences in hepatic FFA storage, we also examined the relationship between tracer-measured storage rates of systemic palmitate and arterial plasma palmitate concentrations. There was a positive correlation between plasma palmitate concentrations and hepatic storage of systemic palmitate (Fig. 3A, r = 0.51, P < 0.02).

Given the findings regarding IHL content and VLDL-TG secretion rates, we also examined the relationship between hepatic palmitate uptake and IHL separately for those with >5.6% and <5.56% IHL. For those with IHL <5.56%, uptake of systemically derived palmitate was positively associated with IHL content (Figure 3B) (ρ = 0.52, P < 0.03). However, there was no relationship between these two parameters for those with IHL >5.56% (Fig. 3C), and there was no linear relationship when the data from all subjects were examined together (Fig. 3D).

Hepatic markers of lipid oxidation

We measured hepatic palmitoyl-carnitine concentrations and the β-hydroxybutyrate response to 6 hours of fasting as potential proxy measures of hepatic fatty acid oxidation. Hepatic palmitoyl-carnitine concentrations were negatively correlated with the average fasting RER from the previous day (r = −0.44, P < 0.03), but not with IHL, plasma concentrations of β-hydroxybutyrate, or fasting plasma insulin concentrations. The β-hydroxybutyrate response to fasting was not associated with percentage IHL, fasting insulin concentrations, or fasting RER.

Non-TG hepatic lipids

The values for non-TG hepatic lipids are provided in Table 3. To determine whether percentage IHL, hepatic palmitate storage rates, or VLDL-TG secretion rates were associated with lipid metabolites, we measured hepatic DG, ceramide, and sphingolipid concentrations as well as the enrichment in C16:0/16:0 and C16:0/18:1 DG, C16-ceramide, and ceramide sphingoid backbone resulting from the incorporation of [U-¹³C]palmitate. As shown in Fig. 4A, hepatic total DG concentrations were significantly positively correlated with percentage IHL (r = 0.46, P = 0.02); the DG saturation index was not associated with IHL. VLDL-TG secretion rates were not associated with total hepatic DG concentrations, but they were associated with the DG saturation index (r = 0.48, P = 0.02; Fig. 4B).

Hepatic palmitate uptake rates were positively associated with the fraction of hepatic 16:0/16:0 DG derived from plasma palmitate (r = 0.56, P = 0.004), but not with hepatic 16:0/16:0 DG concentrations, suggesting that greater FFA uptake rates facilitate more rapid equilibration with hepatic DG pools (Fig. 4C).

No significant associations were observed for percentage IHL or hepatic palmitate uptake rate with any of the intrahepatic ceramide and sphingolipid concentrations or enrichment relative to plasma (data not shown). Of interest, VLDL-TG secretion rates were significantly correlated with hepatic sphingosine-1-phosphate (S1P) concentration (r = 0.44, P = 0.03; Fig. 4D).

Discussion

The primary aim of this study was to understand whether excess IHL in severe obesity is related to greater uptake or limited disposal of fatty acids. To measure hepatic uptake of systemic FFAs, we used a timed, IV bolus of a radiolabeled palmitate tracer combined with an intraoperative liver biopsy while simultaneously measuring total palmitate flux using a stable isotope tracer. To assess liver disposal of fatty acids, we measured VLDL-TG secretion rates using the patient’s own labeled VLDL as well as intrahepatic palmitoyl-carnitine concentrations and the β-hydroxybutyrate response to short-term fasting. Finally, we quantified the hepatic retention of VLDL-TG derived fatty acids after ∼24 hours. By assessing each aspect of liver fatty acid balance we were able to demonstrate that high IHL is unrelated to excess hepatic uptake of systemic FFAs in nondiabetic patients with severe obesity. The patients in our study with greater amounts of IHL had lesser VLDL-TG secretion rates, but not reduced hepatic palmitoyl-carnitine concentrations or ketone body responses to fasting. Taken together, our results point toward defects in VLDL-TG secretion as contributing to maintenance of excess IHL in nondiabetic patients with severe obesity.

Because increased IHL is consistently associated with insulin resistance and increased rates of VLDL secretion, our finding of a negative correlation between total IHL and VLDL-TG secretion rate in participants with IHL >5.56% (Fig. 2A) might be unexpected. However, Mittendorfer et al. (9), studying >200 participants (lean and with obesity), found a positive linear relationship between VLDL-TG secretion rate and hepatic TGs up to 6% IHL, and no relationship with hepatic TGs >6%. Unfortunately, without direct measures of hepatic FFA uptake (9), they could not determine the potential role of this process. Reanalyzing our smaller data set using their 6% cutoff revealed a trend (r = 0.46, P = 0.059) toward a positive linear relationship between VLDL-TG secretion rate and percentage IHL in those participants with IHL <6%, which would be consistent with the observations of Mittendorfer et al. (9). Poulsen et al. (10) found that men with NAFLD (n = 18) have a higher basal VLDL-TG secretion rate and fail to suppress mean VLDL-TG secretion in response to a hyperinsulinemic clamp compared with men without NAFLD (n = 9). In this study (10), the analysis binned men into NAFLD positive and negative, which may have obscured the nuance of VLDL secretion nearest to the level at which secretion rates may be the greatest, as demonstrated by both our results and those of Mittendorfer et al. (9). Adiels et al. (8), using a multicompartmental modeling approach, found increased VLDL-TG production rates in men with T2DM compared with nondiabetic men with obesity and noted that VLDL-TG production was weakly but positively correlated with liver fat content. Because we excluded patients with T2DM and had a population with only a modest burden of metabolic disease, it is not possible to directly compare the results of that study (8) with our results. We did not find that ApoB100 secretion rate was related to liver fat content, whereas there did seem to be relationships between liver fat and VLDL-TG secretion rates. This suggests that any differences in VLDL-TG secretion between those with varying amounts of IHL in our study is primarily related to changes in particle size rather than VLDL particle number.

Sources of hepatic TG include FFAs, fatty acids from lipoprotein particles, dietary fat, and de novo synthesis from glucose and amino acids. Unlike the somewhat dichotomous relationships between IHL (>5.56% and <5.56%) and VLDL-TG secretion and hepatic palmitate uptake, our data indicate a more continuous relationship between retention of fatty acids from VLDL and IHL (Fig. 2E). Without knowing the initial hepatic clearance of VLDL particles it is difficult to discern whether this relationship is purely a reflection of reduced ability of those with NAFLD to export or oxidize fatty acids derived from lipoproteins or some combination of differences in initial uptake and reduced disposal in those with NAFLD. These data do indicate that hepatic retention of VLDL-TG–derived fatty acids can extend for considerable periods of time, which might confound studies that assume uniformly rapid turnover of hepatic fatty acids.

Because alterations in mitochondrial structure and function may occur in NAFLD, there could be defects in fatty acid disposal via oxidative pathways in persons with excess IHL. Using stable isotope techniques, Sunny et al. (31) provided evidence that flux of lipid through the tricarboxylic acid cycle is increased in patients with NAFLD. As assessed by high-resolution respirometry, mitochondrial respiration rates are increased in patients with insulin resistance who are obese with NAFLD, but decreased mitochondrial function is seen with more advanced NASH (32). Our measures of fatty acid oxidation (hepatic palmitoyl-carnitine content and the β-hydroxybutyrate response to 6 hours of fasting) were unrelated to IHL. The finding that hepatic palmitoyl-carnitine concentrations are inversely related to average fasting RER may reflect the relatively large (∼20% to 35%) (33, 34) contribution of liver to REE, perhaps enough to influence whole-body RER. Although imperfect, these measurements do not support the concept of reduced fatty acid flux through oxidative pathways in this cohort.

Patients with NAFLD typically display some degree of hepatic insulin resistance (35) and frequently have high circulating FFAs. This is consistent with adipose tissue insulin resistance, resulting in increased FFA delivery to the liver (36). Patients with NAFLD also have higher hepatic mRNA and protein content of CD36, a transport protein for fatty acids (37), although we found that hepatic palmitate uptake is independent of CD36 content (38). Donnelly et al. (3) reported that 59% of intrahepatic TG is derived from plasma FFAs. We found a positive association between hepatic palmitate uptake rates and plasma palmitate concentrations (r = 0.52, P = 0.03) in those with <5.56% hepatic lipid, which would support a role for systemic FFA delivery in determining hepatic lipid content within the normal range.

We found that hepatic total DG concentrations were positively correlated with IHL. It is unclear whether this accumulation of total DGs with increasing lipid content is pathogenic or whether this accumulation is simply a marker of the larger perturbations occurring in hepatic lipid metabolism because DGs are necessary intermediates in the formation of TGs for storage. Hepatocyte lipid droplet DG content has been described as a strong predictor of hepatic insulin resistance (13). Of more interest, the fraction of hepatic 16:0/16:0 DGs that were derived from plasma was positively correlated with hepatic palmitate uptake in our study. This suggests that greater delivery of palmitate, and probably other FFA species, drives DG synthesis. In addition to acting as precursors to TGs, DGs serve as second messenger signaling molecules (39). Increased hepatic content of total DGs (but not other bioactive lipid metabolites) was reported to correlate with protein kinase C-ε–mediated disruption of canonical insulin signaling and positively correlated with Homeostatic model assessment of insulin resistance (13). We confirmed this observation in our study; that is, total hepatic DG was positively correlated with Homeostatic model assessment of insulin resistance (r = 0.43, P = 0.03). The observation that VLDL-TG secretion rate is positively correlated with greater amounts of saturated DGs relative to total DGs, but not total DG content itself, suggests that there may be some differential regulation by saturated DGs. Future studies should address the stereospecificity of DG species in relationship to regulation of VLDL-TG synthesis and secretion.

Although there was no relationship between hepatic content of S1P and IHL (data not shown), there was a positive relationship (Fig. 4D) between hepatic S1P concentrations and VLDL-TG secretion rate. S1P has been described as a bioactive signaling molecule with diverse functions (e.g., wound healing, immune system, and hepatic fibrosis) and can uniquely act as an intracellular second messenger as well as an extracellular (via G-protein–coupled receptors) signaling molecule (40). There is a paucity of evidence with regard to the role that S1P serves in healthy liver function, and, to our knowledge, this study is the first description of a relationship between hepatic levels of S1P and VLDL-TG secretion. Further studies should interrogate this relationship.

Recruiting people with class II/III obesity (mean BMI of 47.8 kg/m²) without T2DM allowed us to dissociate hyperglycemia and diabetes medications from the pathology of hepatic steatosis itself. The average fasting insulin concentrations in this cohort were relatively normal, but with a wide range (Table 1). We found no relationship between fasting plasma insulin concentrations and IHL (r = 0.16, P = 0.46) or plasma insulin and VLDL-TG rate of appearance (r = 0.20, P = 0.37). Our data are consistent with relatively intact hepatic insulin sensitivity to suppress VLDL synthesis and secretion (41–43). One limitation in translating our findings more broadly and within the context of existing literature is that this cohort was not as metabolically unhealthy as a typical bariatric surgery population. Furthermore, our findings cannot be extrapolated to patients who are less severely obese with overt T2DM, NAFLD, and other metabolic perturbations, who may have different metabolic regulation of pathways associated with these diseases. However, our findings should prompt consideration of these processes as potential pathways to accumulating excess liver fat. We also note that our cohort of patients who underwent bariatric surgery had a lesser prevalence of fatty liver (32%) than previously reported for this degree of obesity, which might be attributable to this population being free of T2DM. The method we used to measure hepatic FFA storage is only able to measure the fate of systemic FFAs. We have previously reported that larger amounts of visceral fat are associated with a greater proportion of hepatic FFA delivery originating from visceral adipose tissue lipolysis (44), and (absent injection of tracer into the portal vein) it is not possible to determine what the contribution of visceral adipose-derived FFAs is to hepatic fatty acid metabolism. Finally, the study was designed to primarily assess hepatic lipid content; unfortunately, limitations on biopsy sample size did not allow us to assess either fibrosis or inflammation in this cohort.

In summary, we provide strong evidence that in nondiabetic patients with class II/III obesity, hepatic lipid in the NAFLD range is most strongly related to decreased export of TGs in VLDL rather than an increase in uptake of systemic FFAs or reduced hepatic fat oxidation. The amounts of DGs with saturated fatty acids and the hepatic content of S1P may play roles in both VLDL-TG secretion and possibly the underlying pathology of fatty liver disease.

Acknowledgments

We express our sincere appreciation to the individuals who participated in this study. Additionally, we thank our research and laboratory team, Barb Norby, Carley Vrieze, Christy Allred, Deb Harteneck, Darlene Lucas, and Lendia Zhou, as well as the CTSA Clinical Research Unit staff and core laboratories. We also thank Dr. Frank Sacks at the Harvard T. H. Chan School of Public Health for providing the protocol and insight into measuring ApoB100 concentrations.

Financial Support: This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases Grants DK40484, DK45343, and DK50456 and the National Center for Research Resources and the National Center for Advancing Translational Sciences, National Institutes of Health, through Grants 1 UL1 RR024150-01 and UL1 TR000135. K.A.L., N.C.B., and K.C.H. received fellowship funding from National Institutes of Health Training Grant T32-DK07352 and American Diabetes Association Fellowship Grant 7-112-MN-36. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Author Contributions: Conceived of and designed the study (N.C.B. and M.D.J.), performed experiments (N.C.B., J.M.T., N.W.G., T.A.K., M.L.K., J.M.S., and K.C.H.), analyzed data (K.A.L., N.C.B., J.M.T., and M.D.J), interpreted data (K.A.L., N.C.B., J.M.T., K.C.H., and M.J.D.), drafted manuscript (K.A.L., N.C.B., and M.D.J.), edited and revised (K.A.L., N.C.B., J.M.T., N.W.G., K.C.H., and M.D.J.), and approved final copy (all authors).

Additional Information

Disclosure Summary: The authors have nothing to disclose.

Data Availability: The datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.

Abbreviations:

Abbreviations:
 
• BMI

  body mass index

 
• DG

  diacylglycerol

 
• FFA

  free fatty acid

 
• IHL

  intrahepatic lipid

 
• LC-MS/MS

  liquid chromatography–tandem mass spectrometry

 
• NAFLD

  nonalcoholic fatty liver disease

 
• NASH

  nonalcoholic steatohepatitis

 
• REE

  resting energy expenditure

 
• RER

  respiratory exchange ratio

 
• S1P

  sphingosine-1-phosphate

 
• T2DM

  type 2 diabetes mellitus

 
• TG

  triglyceride

 
• VLDL

  very-low-density lipoprotein

References and Notes

Author notes