Tumor Necrosis Factor-Increased Hepatic Very-Low-Density Li pop rote in Production and Increased Serum Triglyceride Levels in Diabetic Rats KENNETH R. FEINGOLD, MOUNZER SOUED, SALEH ADI, ILONA STAPRANS, JUDY SHIGENAGA, WILLIAM DOERRLER, ARTHUR MOSER, AND CARL GRUNFELD
Previous studies demonstrated that administration of tumor necrosis factor (TNF) to diabetic rats rapidly increases serum triglyceride levels and stimulates hepatic lipogenesis without affecting the activity of adipose tissue lipoprotein lipase or serum insulin levels. The purpose of this study was to determine the mechanism by which TNF increases serum triglyceride levels and stimulates hepatic fatty acid synthesis in diabetic animals. The maximal increase (-2-fold) in serum triglyceride levels in diabetic rats is seen with a dose of 10 ^g TNF/200 g body wt, and the halfmaximal effect is observed with 5 |xg TNF/200 g body wt. The clearance of labeled triglyceride-rich lipoproteins from the circulation is not affected by TNF administration (triglyceride ty2: diabetic vs. TNFadministered diabetic, 3.5 ± 0.7 vs. 4.0 ± 0.6 min, respectively; NS). The production of triglyceride, measured by the Triton WR-1339 technique, is increased twofold in diabetic animals after TNF administration. These results indicate that the rapid increase in serum triglyceride levels after TNF treatment is accounted for by increased hepatic lipoprotein secretion. TNF administration did not alter either the amount or activation state of hepatic acetylCoA carboxylase, a key regulatory enzyme in fatty acid synthesis. There was also no change in the hepatic levels of fatty acyl-CoA, an allosteric inhibitor of acetyl-CoA carboxylase. However, there was a 7 1 % increase in hepatic citrate concentrations. Citrate is an allosteric activator of acetyl-CoA carboxylase, and changes in hepatic citrate concentrations have been shown to mediate changes in the rates of fatty acid synthesis. These results suggest that the TNF-induced stimulation of hepatic lipogenesis is mediated by
From the Department of Medicine, University of California, and the Metabolism Section and Lipid Research Laboratory, Medical Service, Veterans Administration Medical Center, San Francisco, California. Address correspondence and reprint requests to Kenneth R. Feingold, MD, Metabolism Section (111F), VA Medical Center, 4150 Clement Street, San Francisco, CA 94121. Received for publication 2 February 1990 and accepted in revised form 6 August 1990.
DIABETES, VOL. 39, DECEMBER 1990
citrate activation of acetyl-CoA carboxylase. At 2 h after TNF administration, neither serum glucose nor p-hydroxybutyrate levels were adversely altered in the TNF group, indicating that the disturbances in lipid metabolism are not dependent on alterations in glycemic control. The increases in serum triglyceride levels that occur during infections or stress in diabetes may be secondary to TNF. Diabetes 39:156974, 1990
I
nfection in diabetic patients frequently results in various metabolic disturbances, including hyperglycemia, ketosis, and hyperlipidemia (1,2). The response to infection involves many cell types and is mediated by cytokines (3-5). Recent experiments by our laboratory demonstrated that the administration of the cytokine tumor necrosis factor (TNF) to diabetic animals increased serum triglyceride levels 2.4-fold at 2 h and 4.3-fold at 17 h (6). Hepatic fatty acid synthesis, measured by the incorporation of tritiated water into fatty acids, was increased 45% at 1-2 h and 87% at 16-17 h after TNF administration (6). Under these conditions, TNF treatment did not affect serum insulin levels (6). As noted by others and as found in our experiments, adipose tissue lipoprotein lipase activity in diabetic animals is markedly suppressed (6-8). The administration of TNF to diabetic animals did not produce a further decrease in adipose lipoprotein lipase activity in the diabetic animals at either 2 or 17 h after TNF administration (6). These results indicate that the TNF-induced increase in serum lipid levels does not require changes in adipose tissue lipoprotein lipase activity. The purpose of this study was to determine the mechanism by which TNF acutely increases serum triglyceride levels in diabetic animals. Additionally, the mechanism by which TNF stimulates hepatic fatty acid synthesis was explored. RESEARCH DESIGN AND METHODS
[14C]cholesterol, [14C]triolein, and [14C]sodium bicarbonate were purchased from Du Pont-NEN (Boston, MA). Ready
1569
TNF AND LIPID METABOLISM IN DIABETES
Safe scintillation fluid was purchased from Beckman (Fullerton, CA). Keto-Diastix were obtained from Ames (Elkhart, IN). Streptozocin (STZ), bovine serum albumin (BSA) fraction V, acetyl-CoA, ATP, 2-mercaptoethanol, dithiothreitol (DTT), NADH, NAD, a-ketoglutarate, a-ketoglutarate dehydrogenase, citrate, citrate lyase, malate dehydrogenase, phenylmethylsulfonyl fluoride (PMSF), benzamidine, A/-tosylL-phenylalanine chloromethylketone (TPCK), leupeptin, antipain, and pepstatin were purchased from Sigma (St. Louis, MO). Triton WR-1339 was obtained from Ruger (Irvington, NJ). Human TNF-a (sp act 5 x 107 U/mg) was kindly provided by Genentech (South San Francisco, CA). Male and female Sprague-Dawley rats (-200 g) were purchased from Bantin Kingman (Gilroy, CA). The animals were maintained on a reversed 12-h light-dark cycle (0300-1500 dark, 1500-0300 light) and were fed Simonsen (Gilroy, CA) rat chow and water ad libitum. In female rats, diabetes was induced, as in previous studies, by injecting the animals after an overnight fast with 40 mg/kg i.p. STZ in 1 M sodium citrate buffer (pH 4.5) (9). Urine samples of the animals administered STZ were periodically analyzed with Keto-Diastix, and animals were eliminated from the study if they did not have at least 1% glucosuria at all times or if they were ketonuric. The animals were injected via the tail vein with TNF in 0.5 ml of 0.9% saline or saline alone 10-14 days after STZ administration. Serum triglyceride levels were measured with diagnostic kit 405 (Sigma) after extraction with Dole's reagent. Serum 3-hydroxybutyrate levels were measured with diagnostic kit 310 (Sigma). Serum glucose levels were determined with a YSI glucose analyzer. Chylomicron clearance was measured as in our previous studies (6,10). The major mesenteric lymph duct was cannulated in male rats. After establishment of lymph flow, the animals were administered 50 |xCi of [4-i4C]cholesterol and 100 \xC\ of [3H]triolein in 2 ml of a corn oil-milk emulsion. The lymphatic drainage was collected in iced tubes for 18 h. Chylomicrons were isolated by layering 5 ml of lymph under 0.15 M NaCI and centrifuging at 5 x 106 g/min at 10°C. Labeled chylomicrons (10 mg of triglyceride) were injected via the tail vein into female diabetic rats to whom either 25 |xg TNF or saline was administered 1 h before study. Blood samples (0.075 ml) from the tail vein were obtained at 2, 4, 6, 8, and 10 min after chylomicron administration. The blood samples were extracted with Dole's reagent, and the lipid phase was counted by liquid scintillation. The t\/2 of disappearance from the circulation of both the [3H]triglyceride and [14C]cholesterol associated with the chylomicrons was calculated by linear regression analysis with Sigma-Plot (Jandel, Sausalito, CA) on an IBM AT personal computer. Hepatic triglyceride secretion was determined after Triton WR-1339 administration as in our previous studies (11). Triton WR-1339 is a nonionic detergent that traps lipoproteins in the plasma, which prevents their metabolism and thereby allows determination of the rate of secretion of lipoproteins (12-14). Female diabetic rats were administered 25 |xg TNF and 120 mg Triton WR-1339 in 0.9% i.v. saline at time 0. After 1 and 2 h, blood was obtained for the determination of serum triglyceride levels. The increase in serum triglyceride levels from 1 to 2 h is equivalent to the rate of secretion of
1570
triglyceride-containing lipoproteins. To markedly diminish the contribution of the intestine to the production of triglyceride-rich lipoproteins in these experiments, the animals were fed a fat-free high-sucrose diet consisting of 20% casein, 4% Hegsted salt mixture (ICN, Lisle, IL), 70% sucrose, and 22 g/kg of a vitamin mixture (ICN) for 16 h before study. The level of acetyl-CoA carboxylase was estimated by assaying its activity after maximal activation with citrate (15). In brief, liver was homogenized in 2 vol of 50 mM Tris, 0.5 mM EDTA, and 5 mM 2-mercaptoethanol (pH 7.5) with a Polytron homogenizer. The homogenate was sequentially centrifuged and reextracted as previously described. The supernatant from a 105,000 x g centrifugation was used. Acetyl-CoA carboxylase was maximally activated by incubation in 50 mM Tris, 20 mM citrate, 20 mM MgCI2, 1 mM DTT and 0.5 mg/ml BSA (pH 7.5) for 30 min at 37°C. AcetylCoA carboxylase activity was then immediately assayed by dilution into 100 mM Tris (pH 7.5), 1 mM DTT, 0.2 mM acetylCoA, 20 mM NaH14CO3 (0.25 ixCi/jxinol), 5 mM ATP, 20 mM citrate, 20 mM MgCI2, and BSA (0.05 mg/ml). After incubation for 5 min at 37°C, the reaction was stopped by acidification, and the samples were dried. The newly synthesized malonyl-CoA was assayed by dissolving the residue in water and counting in Ready Safe scintillation fluid. One unit of enzyme activity is defined as the amount required to form 1 ixmol of malonyl-CoA/min at 37°C. The phosphorylation state of acetyl-CoA carboxylase was estimated by assaying the ratio of activity at 0 and 2 mM citrate with the method of Jamil and Madsen (16,17). A lobe of liver was frozen in situ with a Wollenberger clamp cooled in liquid N2. The liver was powdered while frozen, then rapidly homogenized in 1.5 vol of 50 mM Tris (pH 7.5), 2 mM EDTA, 1 mM EGTA, 100 mM NaF, 1 mM DTT, 1 mg/ml BSA, 0.39 mM PMSF, 1 mM benzamidine, 0.01 mM TPCK, leupeptin (0.4 |xg/ml), antipain (0.4 |xg/ml), and pepstatin (0.4 (xg/ml) to inhibit dephosphorylation and proteolytic activation. The homogenate was centrifuged at 27,000 x g for 20 min at 0°C. The supernatant was added to a reaction mixture including 50 mM Tris (pH 7.5), 1 mM EDTA, 10 mM MgCI2, 1 mM DTT, BSA (0.75 mg/ml), 20 mM NaH14CO3 (26,000 counts/min [cpm]/(xmol), 4 mM ATP, and 0.5 mM acetylCoA in the absence or presence of 2 mM citrate in a final volume of 0.9 ml. After incubation for 4 min at 37°C, the reaction was stopped by acidification, and the samples were dried. The residue was dissolved in water and counted in scintillation fluid. Citrate levels were determined as follows. A lobe of liver was frozen in situ with a Wollenberger clamp cooled in liquid N2. Liver was powdered with a mortar and pestle while still frozen. Citrate was then extracted in 3.5 vol of 8% HCIO4 in 40% ethanol as described by Williamson and Corkey (18). The homogenate was centrifuged at 25,000 x g, and the pellet was reextracted. Both supernatants were pooled and centrifuged again at 25,000 x g for 10 min. The supernatant was adjusted to pH 6 by addition of 3 M K2CO3 containing 0.5 M triethanolamine base. Citrate was measured by incubating the extract in 50 mM triethanolamine, 10 mM MgSO4, 5 mM EDTA (pH 7.4), buffer supplemented with 10 ixg/ml NADH, and 25 jig/ml malate dehydrogenase. Absorbance was recorded at 340 nm for 5 min until stabilized, after which citrate lyase (50 |xg/ml) was added to start the
DIABETES, VOL. 39, DECEMBER 1990
K.R. FEINGOLD AND ASSOCIATES
reaction. The change in absorbance for each sample was measured every 30 s for 10 min with a Beckman DU-50 spectrophotometer with kinetics modules and compared with a standard curve of citrate. Fatty acyl-CoA levels were determined as follows. Livers were frozen in situ with a Wollenberger clamp cooled in liquid N2; the sample was powdered with a mortar and pestle, then extracted with 9 vol of 5% HCIO4 containing 2 mM EDTA (19). The homogenate was centrifuged for 10 min at 1000 x g. The pellet containing fatty acyl-CoA was suspended in 2 ml H2O and 0.8 ml of 0.85 N KOH containing 20 mM 2-mercaptoethanol and incubated for 20 min at room temperature to cleave fatty acid from CoA. The extract was acidified with 60% HCIO4, incubated for 5 min at 4°C, and then centrifuged for 10 min at 1000 x g. The supernatant containing the released CoA was adjusted to pH 6.0-6.5 by addition of 0.2 ml saturated KH2PO4, followed by titration with 10 N KOH. The sample was then centrifuged to remove potassium perchlorate. The released CoA was assayed fluorometrically (20). Samples were added to 100 mM potassium phosphate buffer (pH 7), 100 fiM NAD, 100 \x,M a-ketoglutarate, 2 mM cysteine, 2 mM MgCI2, and 1 mM EDTA. Fluorescence (excitation 345 nm, emission 455 nm) was measured before and after addition of 6 U of a-ketoglutarate dehydrogenase and compared to a standard curve of CoA. Statistical significance was determined by two-tailed t test.
TABLE 1 Effect of tumor necrosis factor (TNF) administration on indices of diabetic control in rats
Saline
Serum glucose (mM)
Serum p-hydroxybutyrate (mM)
22.0 ± 1.2
0.903 ± 0.221
22.0 ± 1.5 20.0 ± 1.8 21.0 ± 1.9
0.778 ± 0.327 1.249 ± 0.202 0.999 ±0.135
TNF (\x,g)i
5 10 25
Values are means ± SE; n - 5 for each group. Serum glucose and (3-hydroxybutyrate levels were measured 120 min after TNF administration. There were no significant differences between saline- and any TNF-administered group.
glyceride-rich lipoproteins in diabetic animals is shown in Fig. 2. Chylomicrons were labeled in triglycerides to follow triglyceride clearance and in cholesterol to follow chylomicron-remnant metabolism. The clearance of both the triglyceride and cholesterol labels from the circulation was similar in saline- and TNF-administered diabetic animals (Fig. 2).
RESULTS
The level of serum triglycerides 2 h after administration of varying doses of TNF to diabetic animals is shown in Fig. 1. As in our previous study, TNF administration produced a rapid increase in serum triglyceride levels (6). A dose of 25 ixg of TNF/200 g body wt increased serum triglyceride levels 98%. The half-maximal dose for increasing serum triglyceride levels was ~5 |xg. Administration of TNF to diabetic animals did not acutely affect either serum glucose or (3-hydroxybutyrate concentrations (Table 1). Thus, although TNF produces a rapid increase in serum triglyceride levels in diabetic animals, it has no acute effect on indices of glycemic control. The effect of TNF administration on the clearance of tri-
10 Minutes
Minutes
0.0
10
15 20 TNF Dose (/zg)
FIG. 1. Tumor necrosis factor (TNF) dose-response curve. Two hours after intravenous administration of TNF, serum was obtained for determination of triglyceride concentration. Values are means ± SE; n = 5 in each group. *P < 0.05.
DIABETES, VOL. 39, DECEMBER 1990
FIG. 2. Effect of tumor necrosis factor (TNF) on chylomicron clearance in diabetic rats. O, Diabetic (n = 5); • TNF-administered diabetic (n = 5). Mesenteric lymph from rats administered [3H]triolein {A) and [14C]cholesterol (B) was collected, and chylomicrons were isolated by centrifugation. Labeled chylomicrons (10 mg triglyceride) were injected intravenously into diabetic rats that had been administered 25 |xg TNF or saline 1 h before study. Blood samples were obtained every 90 s for 9 min, extracted with Dole's reagent, and lipid phase was counted by liquid scintillation. Values are means ± SE. tv2of disappearance from circulation was calculated as described in RESEARCH DESIGN AND METHODS (A: diabetic vs. TNF-administered diabetic, 3.5 ± 0.7 vs. 4.0 ± 0.6 min, respectively; B: diabetic vs. TNF-administered diabetic, 4.5 ± 0.7 vs 4.7 ± 0.6 min, respectively).
1571
TNF AND LIPID METABOLISM IN DIABETES
FIG. 3. Effect of tumor necrosis factor (TNF) on hepatic triglyceride secretion in diabetic rats. Diabetic rats were fed high-sucrose fat-free diet for 16 h. Animals were administered 25 |xg i.v. TNF (n = 6, hatched bar) or saline (n = 4, open bar) and 120 mg i.v. Triton WR1339 in 0.9% saline at time 0. After 1 and 2 h, blood was obtained for determination of serum triglyceride levels. Values are means ± SE and represent difference between 2- and 1-h values. P < 0.01 diabetic vs. control.
The effect of TNF administration on hepatic triglyceride production in diabetic animals is shown in Fig. 3. The production of triglyceride was increased 103% in the diabetic animals administered TNF. These results indicate that the increase in serum triglyceride levels in TNF-administered diabetic animals is accounted for by increased hepatic lipoprotein secretion. The effect of TNF administration on the hepatic activity of acetyl-CoA carboxylase, a key regulatory enzyme in fatty acid synthesis, is shown in Table 2. The level of acetyl-CoA carboxylase in the liver, determined after maximal stimulation by citrate, is similar in saline- and TNF-administered diabetic animals. Additionally, by measurement of the phosphorylation/dephosphorylation state of the enzyme, we determined whether the activation state of acetyl-CoA carboxylase could account for the acute stimulation of fatty acid synthesis. This was done by measuring the rate of enzyme activity at 0 and 2 mM citrate according to the method of Jamil and Madsen (16,17). The ratio of acetyl-CoA carboxylase activity at 0 and 2 mM citrate was similar in the saline- and TNF-administered diabetic animals (Table 2). An increase in the activity ratio would have been consistent with an increase in the activation state of the enzyme. Note that the functional phosphorylation state, which we calculated based on a nomogram published by Jamil and Madsen, also is not altered by TNF administration in diabetic animals (Table 2).
Studies have demonstrated that, under various conditions, rapid increases in hepatic fatty acid synthesis correlate with changes in the intracellular concentration of regulators of acetyl-CoA carboxylase activity (21-24). Figure 4 presents data on the effect of TNF administration on the intrahepatic concentration of the two key allosteric regulators of acetylCoA carboxylase activity. Citrate is an allosteric activator, whereas long-chain fatty acyl-CoA are inhibitors of acetylCoA carboxylase activity. Hepatic citrate concentrations increased 71% in TNF-administered diabetic animals (Fig. 4). Hepatic fatty acyl-CoA levels were not altered by TNF administration (Fig. 4). These results suggest that the rapid increase in hepatic fatty acid synthesis induced by TNF administration in diabetic animals is due to an increase in hepatic citrate levels.
DISCUSSION
Previous studies by this and other laboratories have shown that, in healthy animals, TNF administration rapidly increases serum triglyceride levels by stimulating hepatic lipogenesis and very-low-density lipoprotein (VLDL) production (6,11,25,26). In healthy animals, TNF did not alter the clearance of triglyceride-rich lipoproteins (6). The purpose of this study was to determine the mechanism by which TNF acutely increases serum triglyceride levels in diabetic animals. This study demonstrates that, in diabetic animals, serum triglyceride levels are increased approximately twofold by 120 min after TNF administration. The maximal increase in serum triglyceride levels in rats is seen with a dose of 10 |xg TNF/200 g body wt, and the half-maximal effect is observed with a dose of 5 |xg. These results should be contrasted with the effect of interleukin 1 (IL-1), another cytokine released by macrophages, on serum triglyceride levels. In nondiabetic animals, IL-1 administration induces hypertriglyceridemia (27). However, in diabetic animals, IL-1 has been reported to decrease serum triglyceride levels (28). TNF increases serum triglyceride levels in both nondiabetic and diabetic animals. Additionally, these experiments, in combination with our prior studies, demonstrate that the rapid increase in serum triglyceride levels induced by TNF administration is due to increased hepatic triglyceride production. First, the increase in serum triglyceride levels and the increase in hepatic lipogenesis both occur soon after TNF administration (6). Second, adipose tissue lipoprotein lipase activity is markedly suppressed in the diabetic animals, and TNF administration does not result in a further decrease in lipoprotein lipase
TABLE 2 Effect of tumor necrosis factor (TNF) administration on acetyl-CoA carboxylase activity
Saline TNF
Maximal activity (nmol • min" 1 • g"')
Activity ratio
Phosphorylation state (mol P/mol enzyme)
119 ± 6.8 113 ± 5.8
0.277 ± 0.036 0.228 ± 0.012
0.38 ± 0.049 0.40 ± 0.021
Values are means ± SE; n = 5 for each group. Diabetic rats were injected with saline or 25 |xg TNF. After 90 min, the livers were excised, homogenized, and assayed for acetyl-CoA carboxylase activity under conditions of maximal citrate activity by the method of Inoue and Lowenstein (15) as described in RESEARCH DESIGN AND METHODS. The activity ratio of acetyl-CoA carboxylase, measured in the absence of citrate and in the presence of 2 mM citrate, was determined by the method of Jamil and Madsen (16,17) as described in RESEARCH DESIGN AND METHODS. The phosphorylation state was calculated as described by Jamil and Madsen (16). No differences were statistically significant.
1572
DIABETES, VOL. 39, DECEMBER 1990
K.R. FEINGOLD AND ASSOCIATES
I 120100-
1
80-
60
g
40 +
|
20 +
o
o E
40 -
o
i< 30 • 20 • 10 •
FIG. 4. Effect of tumor necrosis factor (TNF) on hepatic citrate and fatty acyl-CoA levels. Diabetic rats were injected with saline (n = 5, open bars) or 25 (ig TNF (n = 5; hatched bars). After 90 min, livers were quick-frozen and extracted. Citrate and fatty acyl-CoA levels were assayed as described in RESEARCH DESIGN AND METHODS. Values are means ± SE. P < 0.01 for citrate comparisons and NS for fatty acylCoA comparisons.
activity (6). Third, TNF treatment does not alter the clearance of triglyceride-rich lipoproteins. Finally, and most important, measurement of hepatic triglyceride production with the Triton WR-1339 technique demonstrates increased hepatic triglyceride production in the TNF-administered diabetic animals. Thus, despite insulinopenia and the metabolic perturbations associated with diabetes, TNF administration stimulates hepatic triglyceride production, resulting in hypertriglyceridemia. Note that our studies focused on the mechanism of the acute increase in serum triglycerides induced by TNF. It is possible that, at later periods after TNF administration, the increase in serum lipid levels could be mediated by other mechanisms. However, previous studies in diabetic animals demonstrated that, 17 h after TNF administration, adipose tissue lipoprotein lipase activity is unchanged compared with diabetic controls, suggesting that alterations in the clearance of lipoproteins are not a major mechanism accounting for hyperlipidemia even at later periods in diabetic animals (6). It is also possible that higher doses of TNF could alter lipid metabolism by mechanisms in addition to or other than stimulating hepatic lipid secretion. The mechanism by which TNF acutely increases hepatic lipogenesis has been studied in detail in healthy rats (29). The increase in fatty acid synthesis occurs before any change in the levels of acetyl-CoA carboxylase or fatty acid synthetase activity, the key regulatory enzymes in fatty acid synthesis. This result is consistent with the long half-life of these enzymes (21-23). Additionally, there was no change
DIABETES, VOL. 39, DECEMBER 1990
in the activation state of acetyl-CoA carboxylase after TNF administration or in hepatic levels of fatty acyl-CoA, an allosteric inhibitor of acetyl-CoA carboxylase (29). However, there was an acute increase in the hepatic intracellular concentration of citrate in TNF-administered control animals (29). Citrate is an allosteric activator of acetyl-CoA carboxylase, and changes in hepatic citrate concentration have been shown to mediate acute changes in the rates of fatty acid synthesis during dietary manipulation (21-24). In diabetic animals, we also found no evidence to suggest that TNF stimulates fatty acid synthesis by altering either the amount or activation state of acetyl-CoA carboxylase or hepatic fatty acyl-CoA concentrations. Rather, as in controls, the acute stimulation of fatty acid synthesis is associated with an increase in hepatic citrate levels, which suggests that the TNF-induced stimulation of fatty acid synthesis is mediated by citrate activation of acetyl-CoA carboxylase. Moreover, this increase in hepatic citrate levels and fatty acid synthesis can occur in the absence of alterations in circulating insulin levels. Although our results clearly demonstrate that the cytokine TNF affects lipid metabolism in diabetic animals, note the absence of an acute effect on indices of glycemic control. At 2 h after TNF administration, neither serum glucose nor 3-hydroxybutrate levels were adversely affected. This indicates that the acute disturbances in lipid metabolism are not dependent on alterations in glycemic control. In previous studies, at 17 h after TNF administration, we observed a 73% increase in serum glucose levels in diabetic animals administered TNF compared with diabetic animals administered saline (6). Serum (3-hydroxybutyrate levels also increased in the diabetic animals 17 h after TNF administration (diabetic control [n = 15] vs. diabetic TNF administered [n = 14] 1.12 ±0.12 vs. 2.30 ± 0.37 mM, respectively, P
Previous studies demonstrated that administration of tumor necrosis factor (TNF) to diabetic rats rapidly increases serum triglyceride levels and stimulates hepatic lipogenesis without affecting the activity of adipose tissue lipoprotein lipase or serum insulin levels. The purpose of this study was to determine the mechanism by which TNF increases serum triglyceride levels and stimulates hepatic fatty acid synthesis in diabetic animals. The maximal increase (-2-fold) in serum triglyceride levels in diabetic rats is seen with a dose of 10 ^g TNF/200 g body wt, and the halfmaximal effect is observed with 5 |xg TNF/200 g body wt. The clearance of labeled triglyceride-rich lipoproteins from the circulation is not affected by TNF administration (triglyceride ty2: diabetic vs. TNFadministered diabetic, 3.5 ± 0.7 vs. 4.0 ± 0.6 min, respectively; NS). The production of triglyceride, measured by the Triton WR-1339 technique, is increased twofold in diabetic animals after TNF administration. These results indicate that the rapid increase in serum triglyceride levels after TNF treatment is accounted for by increased hepatic lipoprotein secretion. TNF administration did not alter either the amount or activation state of hepatic acetylCoA carboxylase, a key regulatory enzyme in fatty acid synthesis. There was also no change in the hepatic levels of fatty acyl-CoA, an allosteric inhibitor of acetyl-CoA carboxylase. However, there was a 7 1 % increase in hepatic citrate concentrations. Citrate is an allosteric activator of acetyl-CoA carboxylase, and changes in hepatic citrate concentrations have been shown to mediate changes in the rates of fatty acid synthesis. These results suggest that the TNF-induced stimulation of hepatic lipogenesis is mediated by
From the Department of Medicine, University of California, and the Metabolism Section and Lipid Research Laboratory, Medical Service, Veterans Administration Medical Center, San Francisco, California. Address correspondence and reprint requests to Kenneth R. Feingold, MD, Metabolism Section (111F), VA Medical Center, 4150 Clement Street, San Francisco, CA 94121. Received for publication 2 February 1990 and accepted in revised form 6 August 1990.
DIABETES, VOL. 39, DECEMBER 1990
citrate activation of acetyl-CoA carboxylase. At 2 h after TNF administration, neither serum glucose nor p-hydroxybutyrate levels were adversely altered in the TNF group, indicating that the disturbances in lipid metabolism are not dependent on alterations in glycemic control. The increases in serum triglyceride levels that occur during infections or stress in diabetes may be secondary to TNF. Diabetes 39:156974, 1990
I
nfection in diabetic patients frequently results in various metabolic disturbances, including hyperglycemia, ketosis, and hyperlipidemia (1,2). The response to infection involves many cell types and is mediated by cytokines (3-5). Recent experiments by our laboratory demonstrated that the administration of the cytokine tumor necrosis factor (TNF) to diabetic animals increased serum triglyceride levels 2.4-fold at 2 h and 4.3-fold at 17 h (6). Hepatic fatty acid synthesis, measured by the incorporation of tritiated water into fatty acids, was increased 45% at 1-2 h and 87% at 16-17 h after TNF administration (6). Under these conditions, TNF treatment did not affect serum insulin levels (6). As noted by others and as found in our experiments, adipose tissue lipoprotein lipase activity in diabetic animals is markedly suppressed (6-8). The administration of TNF to diabetic animals did not produce a further decrease in adipose lipoprotein lipase activity in the diabetic animals at either 2 or 17 h after TNF administration (6). These results indicate that the TNF-induced increase in serum lipid levels does not require changes in adipose tissue lipoprotein lipase activity. The purpose of this study was to determine the mechanism by which TNF acutely increases serum triglyceride levels in diabetic animals. Additionally, the mechanism by which TNF stimulates hepatic fatty acid synthesis was explored. RESEARCH DESIGN AND METHODS
[14C]cholesterol, [14C]triolein, and [14C]sodium bicarbonate were purchased from Du Pont-NEN (Boston, MA). Ready
1569
TNF AND LIPID METABOLISM IN DIABETES
Safe scintillation fluid was purchased from Beckman (Fullerton, CA). Keto-Diastix were obtained from Ames (Elkhart, IN). Streptozocin (STZ), bovine serum albumin (BSA) fraction V, acetyl-CoA, ATP, 2-mercaptoethanol, dithiothreitol (DTT), NADH, NAD, a-ketoglutarate, a-ketoglutarate dehydrogenase, citrate, citrate lyase, malate dehydrogenase, phenylmethylsulfonyl fluoride (PMSF), benzamidine, A/-tosylL-phenylalanine chloromethylketone (TPCK), leupeptin, antipain, and pepstatin were purchased from Sigma (St. Louis, MO). Triton WR-1339 was obtained from Ruger (Irvington, NJ). Human TNF-a (sp act 5 x 107 U/mg) was kindly provided by Genentech (South San Francisco, CA). Male and female Sprague-Dawley rats (-200 g) were purchased from Bantin Kingman (Gilroy, CA). The animals were maintained on a reversed 12-h light-dark cycle (0300-1500 dark, 1500-0300 light) and were fed Simonsen (Gilroy, CA) rat chow and water ad libitum. In female rats, diabetes was induced, as in previous studies, by injecting the animals after an overnight fast with 40 mg/kg i.p. STZ in 1 M sodium citrate buffer (pH 4.5) (9). Urine samples of the animals administered STZ were periodically analyzed with Keto-Diastix, and animals were eliminated from the study if they did not have at least 1% glucosuria at all times or if they were ketonuric. The animals were injected via the tail vein with TNF in 0.5 ml of 0.9% saline or saline alone 10-14 days after STZ administration. Serum triglyceride levels were measured with diagnostic kit 405 (Sigma) after extraction with Dole's reagent. Serum 3-hydroxybutyrate levels were measured with diagnostic kit 310 (Sigma). Serum glucose levels were determined with a YSI glucose analyzer. Chylomicron clearance was measured as in our previous studies (6,10). The major mesenteric lymph duct was cannulated in male rats. After establishment of lymph flow, the animals were administered 50 |xCi of [4-i4C]cholesterol and 100 \xC\ of [3H]triolein in 2 ml of a corn oil-milk emulsion. The lymphatic drainage was collected in iced tubes for 18 h. Chylomicrons were isolated by layering 5 ml of lymph under 0.15 M NaCI and centrifuging at 5 x 106 g/min at 10°C. Labeled chylomicrons (10 mg of triglyceride) were injected via the tail vein into female diabetic rats to whom either 25 |xg TNF or saline was administered 1 h before study. Blood samples (0.075 ml) from the tail vein were obtained at 2, 4, 6, 8, and 10 min after chylomicron administration. The blood samples were extracted with Dole's reagent, and the lipid phase was counted by liquid scintillation. The t\/2 of disappearance from the circulation of both the [3H]triglyceride and [14C]cholesterol associated with the chylomicrons was calculated by linear regression analysis with Sigma-Plot (Jandel, Sausalito, CA) on an IBM AT personal computer. Hepatic triglyceride secretion was determined after Triton WR-1339 administration as in our previous studies (11). Triton WR-1339 is a nonionic detergent that traps lipoproteins in the plasma, which prevents their metabolism and thereby allows determination of the rate of secretion of lipoproteins (12-14). Female diabetic rats were administered 25 |xg TNF and 120 mg Triton WR-1339 in 0.9% i.v. saline at time 0. After 1 and 2 h, blood was obtained for the determination of serum triglyceride levels. The increase in serum triglyceride levels from 1 to 2 h is equivalent to the rate of secretion of
1570
triglyceride-containing lipoproteins. To markedly diminish the contribution of the intestine to the production of triglyceride-rich lipoproteins in these experiments, the animals were fed a fat-free high-sucrose diet consisting of 20% casein, 4% Hegsted salt mixture (ICN, Lisle, IL), 70% sucrose, and 22 g/kg of a vitamin mixture (ICN) for 16 h before study. The level of acetyl-CoA carboxylase was estimated by assaying its activity after maximal activation with citrate (15). In brief, liver was homogenized in 2 vol of 50 mM Tris, 0.5 mM EDTA, and 5 mM 2-mercaptoethanol (pH 7.5) with a Polytron homogenizer. The homogenate was sequentially centrifuged and reextracted as previously described. The supernatant from a 105,000 x g centrifugation was used. Acetyl-CoA carboxylase was maximally activated by incubation in 50 mM Tris, 20 mM citrate, 20 mM MgCI2, 1 mM DTT and 0.5 mg/ml BSA (pH 7.5) for 30 min at 37°C. AcetylCoA carboxylase activity was then immediately assayed by dilution into 100 mM Tris (pH 7.5), 1 mM DTT, 0.2 mM acetylCoA, 20 mM NaH14CO3 (0.25 ixCi/jxinol), 5 mM ATP, 20 mM citrate, 20 mM MgCI2, and BSA (0.05 mg/ml). After incubation for 5 min at 37°C, the reaction was stopped by acidification, and the samples were dried. The newly synthesized malonyl-CoA was assayed by dissolving the residue in water and counting in Ready Safe scintillation fluid. One unit of enzyme activity is defined as the amount required to form 1 ixmol of malonyl-CoA/min at 37°C. The phosphorylation state of acetyl-CoA carboxylase was estimated by assaying the ratio of activity at 0 and 2 mM citrate with the method of Jamil and Madsen (16,17). A lobe of liver was frozen in situ with a Wollenberger clamp cooled in liquid N2. The liver was powdered while frozen, then rapidly homogenized in 1.5 vol of 50 mM Tris (pH 7.5), 2 mM EDTA, 1 mM EGTA, 100 mM NaF, 1 mM DTT, 1 mg/ml BSA, 0.39 mM PMSF, 1 mM benzamidine, 0.01 mM TPCK, leupeptin (0.4 |xg/ml), antipain (0.4 |xg/ml), and pepstatin (0.4 (xg/ml) to inhibit dephosphorylation and proteolytic activation. The homogenate was centrifuged at 27,000 x g for 20 min at 0°C. The supernatant was added to a reaction mixture including 50 mM Tris (pH 7.5), 1 mM EDTA, 10 mM MgCI2, 1 mM DTT, BSA (0.75 mg/ml), 20 mM NaH14CO3 (26,000 counts/min [cpm]/(xmol), 4 mM ATP, and 0.5 mM acetylCoA in the absence or presence of 2 mM citrate in a final volume of 0.9 ml. After incubation for 4 min at 37°C, the reaction was stopped by acidification, and the samples were dried. The residue was dissolved in water and counted in scintillation fluid. Citrate levels were determined as follows. A lobe of liver was frozen in situ with a Wollenberger clamp cooled in liquid N2. Liver was powdered with a mortar and pestle while still frozen. Citrate was then extracted in 3.5 vol of 8% HCIO4 in 40% ethanol as described by Williamson and Corkey (18). The homogenate was centrifuged at 25,000 x g, and the pellet was reextracted. Both supernatants were pooled and centrifuged again at 25,000 x g for 10 min. The supernatant was adjusted to pH 6 by addition of 3 M K2CO3 containing 0.5 M triethanolamine base. Citrate was measured by incubating the extract in 50 mM triethanolamine, 10 mM MgSO4, 5 mM EDTA (pH 7.4), buffer supplemented with 10 ixg/ml NADH, and 25 jig/ml malate dehydrogenase. Absorbance was recorded at 340 nm for 5 min until stabilized, after which citrate lyase (50 |xg/ml) was added to start the
DIABETES, VOL. 39, DECEMBER 1990
K.R. FEINGOLD AND ASSOCIATES
reaction. The change in absorbance for each sample was measured every 30 s for 10 min with a Beckman DU-50 spectrophotometer with kinetics modules and compared with a standard curve of citrate. Fatty acyl-CoA levels were determined as follows. Livers were frozen in situ with a Wollenberger clamp cooled in liquid N2; the sample was powdered with a mortar and pestle, then extracted with 9 vol of 5% HCIO4 containing 2 mM EDTA (19). The homogenate was centrifuged for 10 min at 1000 x g. The pellet containing fatty acyl-CoA was suspended in 2 ml H2O and 0.8 ml of 0.85 N KOH containing 20 mM 2-mercaptoethanol and incubated for 20 min at room temperature to cleave fatty acid from CoA. The extract was acidified with 60% HCIO4, incubated for 5 min at 4°C, and then centrifuged for 10 min at 1000 x g. The supernatant containing the released CoA was adjusted to pH 6.0-6.5 by addition of 0.2 ml saturated KH2PO4, followed by titration with 10 N KOH. The sample was then centrifuged to remove potassium perchlorate. The released CoA was assayed fluorometrically (20). Samples were added to 100 mM potassium phosphate buffer (pH 7), 100 fiM NAD, 100 \x,M a-ketoglutarate, 2 mM cysteine, 2 mM MgCI2, and 1 mM EDTA. Fluorescence (excitation 345 nm, emission 455 nm) was measured before and after addition of 6 U of a-ketoglutarate dehydrogenase and compared to a standard curve of CoA. Statistical significance was determined by two-tailed t test.
TABLE 1 Effect of tumor necrosis factor (TNF) administration on indices of diabetic control in rats
Saline
Serum glucose (mM)
Serum p-hydroxybutyrate (mM)
22.0 ± 1.2
0.903 ± 0.221
22.0 ± 1.5 20.0 ± 1.8 21.0 ± 1.9
0.778 ± 0.327 1.249 ± 0.202 0.999 ±0.135
TNF (\x,g)i
5 10 25
Values are means ± SE; n - 5 for each group. Serum glucose and (3-hydroxybutyrate levels were measured 120 min after TNF administration. There were no significant differences between saline- and any TNF-administered group.
glyceride-rich lipoproteins in diabetic animals is shown in Fig. 2. Chylomicrons were labeled in triglycerides to follow triglyceride clearance and in cholesterol to follow chylomicron-remnant metabolism. The clearance of both the triglyceride and cholesterol labels from the circulation was similar in saline- and TNF-administered diabetic animals (Fig. 2).
RESULTS
The level of serum triglycerides 2 h after administration of varying doses of TNF to diabetic animals is shown in Fig. 1. As in our previous study, TNF administration produced a rapid increase in serum triglyceride levels (6). A dose of 25 ixg of TNF/200 g body wt increased serum triglyceride levels 98%. The half-maximal dose for increasing serum triglyceride levels was ~5 |xg. Administration of TNF to diabetic animals did not acutely affect either serum glucose or (3-hydroxybutyrate concentrations (Table 1). Thus, although TNF produces a rapid increase in serum triglyceride levels in diabetic animals, it has no acute effect on indices of glycemic control. The effect of TNF administration on the clearance of tri-
10 Minutes
Minutes
0.0
10
15 20 TNF Dose (/zg)
FIG. 1. Tumor necrosis factor (TNF) dose-response curve. Two hours after intravenous administration of TNF, serum was obtained for determination of triglyceride concentration. Values are means ± SE; n = 5 in each group. *P < 0.05.
DIABETES, VOL. 39, DECEMBER 1990
FIG. 2. Effect of tumor necrosis factor (TNF) on chylomicron clearance in diabetic rats. O, Diabetic (n = 5); • TNF-administered diabetic (n = 5). Mesenteric lymph from rats administered [3H]triolein {A) and [14C]cholesterol (B) was collected, and chylomicrons were isolated by centrifugation. Labeled chylomicrons (10 mg triglyceride) were injected intravenously into diabetic rats that had been administered 25 |xg TNF or saline 1 h before study. Blood samples were obtained every 90 s for 9 min, extracted with Dole's reagent, and lipid phase was counted by liquid scintillation. Values are means ± SE. tv2of disappearance from circulation was calculated as described in RESEARCH DESIGN AND METHODS (A: diabetic vs. TNF-administered diabetic, 3.5 ± 0.7 vs. 4.0 ± 0.6 min, respectively; B: diabetic vs. TNF-administered diabetic, 4.5 ± 0.7 vs 4.7 ± 0.6 min, respectively).
1571
TNF AND LIPID METABOLISM IN DIABETES
FIG. 3. Effect of tumor necrosis factor (TNF) on hepatic triglyceride secretion in diabetic rats. Diabetic rats were fed high-sucrose fat-free diet for 16 h. Animals were administered 25 |xg i.v. TNF (n = 6, hatched bar) or saline (n = 4, open bar) and 120 mg i.v. Triton WR1339 in 0.9% saline at time 0. After 1 and 2 h, blood was obtained for determination of serum triglyceride levels. Values are means ± SE and represent difference between 2- and 1-h values. P < 0.01 diabetic vs. control.
The effect of TNF administration on hepatic triglyceride production in diabetic animals is shown in Fig. 3. The production of triglyceride was increased 103% in the diabetic animals administered TNF. These results indicate that the increase in serum triglyceride levels in TNF-administered diabetic animals is accounted for by increased hepatic lipoprotein secretion. The effect of TNF administration on the hepatic activity of acetyl-CoA carboxylase, a key regulatory enzyme in fatty acid synthesis, is shown in Table 2. The level of acetyl-CoA carboxylase in the liver, determined after maximal stimulation by citrate, is similar in saline- and TNF-administered diabetic animals. Additionally, by measurement of the phosphorylation/dephosphorylation state of the enzyme, we determined whether the activation state of acetyl-CoA carboxylase could account for the acute stimulation of fatty acid synthesis. This was done by measuring the rate of enzyme activity at 0 and 2 mM citrate according to the method of Jamil and Madsen (16,17). The ratio of acetyl-CoA carboxylase activity at 0 and 2 mM citrate was similar in the saline- and TNF-administered diabetic animals (Table 2). An increase in the activity ratio would have been consistent with an increase in the activation state of the enzyme. Note that the functional phosphorylation state, which we calculated based on a nomogram published by Jamil and Madsen, also is not altered by TNF administration in diabetic animals (Table 2).
Studies have demonstrated that, under various conditions, rapid increases in hepatic fatty acid synthesis correlate with changes in the intracellular concentration of regulators of acetyl-CoA carboxylase activity (21-24). Figure 4 presents data on the effect of TNF administration on the intrahepatic concentration of the two key allosteric regulators of acetylCoA carboxylase activity. Citrate is an allosteric activator, whereas long-chain fatty acyl-CoA are inhibitors of acetylCoA carboxylase activity. Hepatic citrate concentrations increased 71% in TNF-administered diabetic animals (Fig. 4). Hepatic fatty acyl-CoA levels were not altered by TNF administration (Fig. 4). These results suggest that the rapid increase in hepatic fatty acid synthesis induced by TNF administration in diabetic animals is due to an increase in hepatic citrate levels.
DISCUSSION
Previous studies by this and other laboratories have shown that, in healthy animals, TNF administration rapidly increases serum triglyceride levels by stimulating hepatic lipogenesis and very-low-density lipoprotein (VLDL) production (6,11,25,26). In healthy animals, TNF did not alter the clearance of triglyceride-rich lipoproteins (6). The purpose of this study was to determine the mechanism by which TNF acutely increases serum triglyceride levels in diabetic animals. This study demonstrates that, in diabetic animals, serum triglyceride levels are increased approximately twofold by 120 min after TNF administration. The maximal increase in serum triglyceride levels in rats is seen with a dose of 10 |xg TNF/200 g body wt, and the half-maximal effect is observed with a dose of 5 |xg. These results should be contrasted with the effect of interleukin 1 (IL-1), another cytokine released by macrophages, on serum triglyceride levels. In nondiabetic animals, IL-1 administration induces hypertriglyceridemia (27). However, in diabetic animals, IL-1 has been reported to decrease serum triglyceride levels (28). TNF increases serum triglyceride levels in both nondiabetic and diabetic animals. Additionally, these experiments, in combination with our prior studies, demonstrate that the rapid increase in serum triglyceride levels induced by TNF administration is due to increased hepatic triglyceride production. First, the increase in serum triglyceride levels and the increase in hepatic lipogenesis both occur soon after TNF administration (6). Second, adipose tissue lipoprotein lipase activity is markedly suppressed in the diabetic animals, and TNF administration does not result in a further decrease in lipoprotein lipase
TABLE 2 Effect of tumor necrosis factor (TNF) administration on acetyl-CoA carboxylase activity
Saline TNF
Maximal activity (nmol • min" 1 • g"')
Activity ratio
Phosphorylation state (mol P/mol enzyme)
119 ± 6.8 113 ± 5.8
0.277 ± 0.036 0.228 ± 0.012
0.38 ± 0.049 0.40 ± 0.021
Values are means ± SE; n = 5 for each group. Diabetic rats were injected with saline or 25 |xg TNF. After 90 min, the livers were excised, homogenized, and assayed for acetyl-CoA carboxylase activity under conditions of maximal citrate activity by the method of Inoue and Lowenstein (15) as described in RESEARCH DESIGN AND METHODS. The activity ratio of acetyl-CoA carboxylase, measured in the absence of citrate and in the presence of 2 mM citrate, was determined by the method of Jamil and Madsen (16,17) as described in RESEARCH DESIGN AND METHODS. The phosphorylation state was calculated as described by Jamil and Madsen (16). No differences were statistically significant.
1572
DIABETES, VOL. 39, DECEMBER 1990
K.R. FEINGOLD AND ASSOCIATES
I 120100-
1
80-
60
g
40 +
|
20 +
o
o E
40 -
o
i< 30 • 20 • 10 •
FIG. 4. Effect of tumor necrosis factor (TNF) on hepatic citrate and fatty acyl-CoA levels. Diabetic rats were injected with saline (n = 5, open bars) or 25 (ig TNF (n = 5; hatched bars). After 90 min, livers were quick-frozen and extracted. Citrate and fatty acyl-CoA levels were assayed as described in RESEARCH DESIGN AND METHODS. Values are means ± SE. P < 0.01 for citrate comparisons and NS for fatty acylCoA comparisons.
activity (6). Third, TNF treatment does not alter the clearance of triglyceride-rich lipoproteins. Finally, and most important, measurement of hepatic triglyceride production with the Triton WR-1339 technique demonstrates increased hepatic triglyceride production in the TNF-administered diabetic animals. Thus, despite insulinopenia and the metabolic perturbations associated with diabetes, TNF administration stimulates hepatic triglyceride production, resulting in hypertriglyceridemia. Note that our studies focused on the mechanism of the acute increase in serum triglycerides induced by TNF. It is possible that, at later periods after TNF administration, the increase in serum lipid levels could be mediated by other mechanisms. However, previous studies in diabetic animals demonstrated that, 17 h after TNF administration, adipose tissue lipoprotein lipase activity is unchanged compared with diabetic controls, suggesting that alterations in the clearance of lipoproteins are not a major mechanism accounting for hyperlipidemia even at later periods in diabetic animals (6). It is also possible that higher doses of TNF could alter lipid metabolism by mechanisms in addition to or other than stimulating hepatic lipid secretion. The mechanism by which TNF acutely increases hepatic lipogenesis has been studied in detail in healthy rats (29). The increase in fatty acid synthesis occurs before any change in the levels of acetyl-CoA carboxylase or fatty acid synthetase activity, the key regulatory enzymes in fatty acid synthesis. This result is consistent with the long half-life of these enzymes (21-23). Additionally, there was no change
DIABETES, VOL. 39, DECEMBER 1990
in the activation state of acetyl-CoA carboxylase after TNF administration or in hepatic levels of fatty acyl-CoA, an allosteric inhibitor of acetyl-CoA carboxylase (29). However, there was an acute increase in the hepatic intracellular concentration of citrate in TNF-administered control animals (29). Citrate is an allosteric activator of acetyl-CoA carboxylase, and changes in hepatic citrate concentration have been shown to mediate acute changes in the rates of fatty acid synthesis during dietary manipulation (21-24). In diabetic animals, we also found no evidence to suggest that TNF stimulates fatty acid synthesis by altering either the amount or activation state of acetyl-CoA carboxylase or hepatic fatty acyl-CoA concentrations. Rather, as in controls, the acute stimulation of fatty acid synthesis is associated with an increase in hepatic citrate levels, which suggests that the TNF-induced stimulation of fatty acid synthesis is mediated by citrate activation of acetyl-CoA carboxylase. Moreover, this increase in hepatic citrate levels and fatty acid synthesis can occur in the absence of alterations in circulating insulin levels. Although our results clearly demonstrate that the cytokine TNF affects lipid metabolism in diabetic animals, note the absence of an acute effect on indices of glycemic control. At 2 h after TNF administration, neither serum glucose nor 3-hydroxybutrate levels were adversely affected. This indicates that the acute disturbances in lipid metabolism are not dependent on alterations in glycemic control. In previous studies, at 17 h after TNF administration, we observed a 73% increase in serum glucose levels in diabetic animals administered TNF compared with diabetic animals administered saline (6). Serum (3-hydroxybutyrate levels also increased in the diabetic animals 17 h after TNF administration (diabetic control [n = 15] vs. diabetic TNF administered [n = 14] 1.12 ±0.12 vs. 2.30 ± 0.37 mM, respectively, P
