Endocrinology and Metabolism

Tissue triglycerides, insulin resistance, and insulin production: implications for hyperinsulinemia of obesity

Kazunori Koyama, Guoxun Chen, Young Lee, Roger H. Unger


Obesity is associated with both insulin resistance and hyperinsulinemia. Initially hyperinsulinemia compensates for the insulin resistance and thereby maintains normal glucose homeostasis. Obesity is also associated with increased tissue triglyceride (TG) content. To determine whether both insulin resistance and hyperinsulinemia might be secondary to increased tissue TG, we studied correlations between TG content of skeletal muscle, liver, and pancreas and plasma insulin, plasma [insulin] × [glucose], and β-cell function in four rat models with widely varying fat content: obese Zucker diabetic fatty rats, free-feeding lean Wistar rats, hyperleptinemic Wistar rats with profound tissue lipopenia, and rats pair fed to hyperleptinemics. Correlation coefficients >0.9 (P < 0.05) were obtained among TG of skeletal muscle, liver, and pancreas and among plasma insulin, [insulin] × [glucose] product, and β-cell function as gauged by basal, glucose-stimulated, and arginine-stimulated insulin secretion by the isolated perfused pancreas. Although these correlations cannot prove cause and effect, they are consistent with the hypothesis that the TG content of tissues sets the level of both insulin resistance and insulin production.

  • tissue fat
  • obese Zucker diabetic fatty rats
  • β-cell function

obesity is the most prevalent health problem in the United States and is currently estimated to afflict ∼75,000,000 Americans (28). Although the risk of diabetes is high, glucose homeostasis remains relatively normal for long periods of time despite insulin resistance and excessive intake of food. This ability to maintain euglycemia despite diminished insulin effectiveness indicates that insulin production is somehow matched to insulin requirements. The mechanism by which β-cells recognize the level of insulin resistance that must be overcome is unknown. Because hyperinsulinemia may be present in obesity even when glucose tolerance is normal, subtle glycemic elevations are not the signal for the enhanced insulin production at that stage of the disorder. This study was undertaken to identify a nonglycemic signal that might link insulin action to insulin production.

Tissue triglyceride (TG) content seemed to be a plausible candidate for this dual role. Long-chain free fatty acids (FFA) have long been known to interfere with insulin-mediated glucose metabolism (2, 8, 13, 15,18, 24, 25, 29) and to stimulate insulin secretion acutely (5, 11, 22,26, 31). Moreover, high FFA levels in vitro have been shown to induce in normal islets the same changes described in the compensating β-cells of obesity, such as β-cell hyperplasia and increased insulin secretion at substimulatory glucose levels (16, 26). Third, it has been shown that islets of hyperinsulinemic obese rats have an extremely high TG content compared with normal littermates (20).

There being no direct way to test the possibility that the coupling of insulin production to insulin need is mediated by tissue lipid content, we relied on correlations between β-cell function and insulin effectiveness and tissue TG content of groups of rats with a widely varying tissue fat content. At one extreme of the spectrum of fat content were obese rats, in which tissue lipids are markedly increased (20). At the other extreme we exploited a novel syndrome of profound tissue lipid depletion that is the antithesis of obesity; this syndrome was produced by inducing chronic hyperleptinemia in normal rats by means of adenovirus-leptin gene transfer (3, 34). These animals undergo rapid and selective loss of all grossly visible fat (3); additionally the TG content in skeletal muscle, liver, and pancreas declines to 1/1,000 of that of obese rats and 1/10 of that of pair-fed controls (34). Between these two extremes of severe lipopenia and obesity, we studied free-feeding normal rats and normal rats whose food intake was restricted by pair feeding to the hyperleptinemic rats. The high correlations between indexes of insulin resistance and insulin production and the TG content in the target tissues of insulin and in the tissue of insulin production are consistent with the hypothesis that tissue fat content might provide the link between insulin resistance and hyperinsulinemia.



All animals were 7–9 wk of age at the start of experiments. Obese, prediabetic Zucker diabetic fatty (ZDF) rats (fa/fa) were bred in our laboratory from ZDF/Drt-fa (F10) stock purchased from R. Peterson (University of Indiana School of Medicine, Indianapolis, IN). Male rats exhibited the previously described phenotype (30).

Male Wistar rats purchased from Charles River Breeding Laboratories (Wilmington, MA) were used as free-feeding lean controls and diet-restricted controls. Diet-restricted rats were Wistar rats pair fed to the leptin-overexpressing lipopenic rats described in the next paragraph; this resulted in a 42% reduction in their total caloric intake.

The hyperleptinemic, lipopenic animals were normal Wistar rats that had received an infusion of a recombinant adenovirus containing the leptin cDNA (AdCMV-leptin). As reported previously (3), leptin mRNA appeared in the liver in association with a rise in plasma leptin levels and a reduction in food intake and body weight (see Fig. 1). Another group of Wistar rats was infused with adenovirus containing the cDNA of an irrelevant protein, bacterial β-galactosidase (AdCMV-βGal), as a control for the viral infection.

Gene transfer studies.

To clone the rat leptin cDNA and measure its mRNA, total RNA was prepared from 1 g of epididymal adipose tissue of rats by extraction with TRIzol as recommended by the manufacturer. Oligo(dT) was used to prime first-strand cDNA synthesis by use of a cDNA synthesis kit (Clontech, Palo Alto, CA). After treatment of the first-strand cDNA with deoxyribonuclease-free ribonuclease, the leptin gene product was amplified by polymerase chain reaction (PCR) with upstream sense primer 5′-GGAGGAATCCCTGCTCCAGC-3′ and downstream antisense primer 5′-CTTCTCCTGAGGATACCTGG-3′ based on the rat leptin gene sequence (27). For both cDNA cloning and leptin mRNA measurement, amplification was performed using 1 cycle at 94°C for 3 min, followed by 35 cycles at 92°C for 45 s, at 53°C for 45 s, and at 72°C for 1 min, and then final extension at 70°C for 10 min. We also measured β-actin expression with the same amplification conditions as for leptin and a previously described oligonucleotide pair (23).

To prepare recombinant virus, a 640-bp PR fragment containing the entire leptin coding region was ligated to pCR TM 2.1 (Invitrogen, San Diego, CA) according to the manufacturer’s protocol. Sequence analysis confirmed that several clones contained the intact leptin cDNA. ABamH I- andXba I-restricted leptin cDNA fragment that included 60 bp of 5′ untranslated region and 76 bp of 3′ untranslated region was ligated to similarly treated pACCMVpLpA (10). The resulting plasmid was cotransfected with pJM17 (23) into 293 cells by calcium phosphate/DNA coprecipitation to generate the new recombinant virus AdCMV-leptin by use of previously described methods (1). Virus DNA was isolated, and the presence of the leptin gene insert was confirmed by PCR by use of the primers described above and by Southern blotting with an oligonucleotide (5′-GGATACCGACTGC GTGTGTGAAATGTCAT-3′) complementary to the rat leptin cDNA. Stocks of AdCMV-leptin were amplified and purified as described (1) and stored at −70°C in phosphate-buffered saline with 0.2% bovine serum albumin (BSA) and 10% glycerol at 1–3 × 1012plaque-forming units (pfu)/ml. A virus containing the bacterial β-galactosidase gene under control of the CMV promoter, AdCMV-βGal, was prepared and utilized as described previously (14).

The virus was infused in male Wistar rats obtained from Charles River Laboratories (Wilmington, MA). Before adenovirus infusion studies, all rats received standard rat chow (Teklad F6 8664, Madison, WI) ad libitum and had free access to water. Polyethylene tubing (PE-50, Becton-Dickinson, Franklin Lakes, NJ) was anchored in the left carotid artery of 9-wk-old Wistar rats of ∼250–300 g under pentobarbital sodium anesthesia (50 mg/kg, Abbott Laboratories, North Chicago, IL) and exteriorized via a subcutaneous tunnel. Tubing was filled with heparinized saline (1,000 U/dl) until the virus infusion was begun 3 days after surgery. Before infusion, adenovirus samples were suspended in saline and filtered through a 0.2-μm filter. Two milliliters of AdCMV-leptin or AdCMV-β-Gal containing a total of 1 × 1012 pfu were infused into anesthetized animals over a 1-h period. Animals were studied in individual metabolic cages (Nalgene, Rochester, NY), and food intake and body weight were measured daily.

Plasma measurements.

Beginning 1 day after adenovirus infusion, fasting blood samples (1–3 PM) were collected from the tail vein in capillary tubes coated with EDTA. Plasma was stored at −20°C until the time of leptin assay. Plasma leptin was assayed using the Linco leptin assay kit (Linco Research, St. Charles, MO). Plasma insulin was assayed by standard methods (37). Plasma glucose was measured by Glucose Analyzer II (Beckman, Brea, CA).

Pancreas perfusion.

Pancreata were isolated and perfused by the method of Grodsky and Fanska (12) as previously modified (17). The perfusate consisted of Krebs-Ringer bicarbonate buffer containing 4.5% Dextran T70, 5.6 mM glucose, 1% BSA, and 5 mM each of sodium pyruvate, sodium glutamate, and sodium fumarate. The flow rate was 3.0 ml/min. After a 20-min equilibration period, the pancreas was perfused for 10 min with 5.6 mM glucose. Then the glucose concentration was increased to 20 mM for 10 min. After a 10-min “rest,” during which the glucose level was returned to baseline, arginine was perfused at a concentration of 8 mM for a total of 10 min. Samples were collected at 1-min intervals for determination of insulin concentration. They were placed in chilled tubes containing 0.3 ml of 0.15 M NaCl, 0.05 M Na2EDTA, and 0.3 M benzamide and were frozen until the time of assay.

TG content of liver, skeletal muscle, pancreas, and islets.

Tissues were dissected and placed in liquid nitrogen. About 100 mg of tissues were placed in 4 ml of homogenizing buffer containing 18 mM of tris(hydroxymethyl)aminomethane ⋅ HCl (pH = 7.5), 300 mM of d-mannitol, and 5 mM of ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid and were homogenized using a hand-held polytron (Kontes Glass, Vineland, NJ) for 10 s. Lipids were extracted by the method of Folch et al. (9). Total TG were assayed by the method of Danno et al. (6).

Statistical analysis.

All data are expressed as means ± SE. Statistical difference was analyzed by unpaired t-test.P < 0.05 was considered statistically significant. Regression line was calculated using the Stat View program (Abacus Concepts, Berkeley, CA).


Clinical and laboratory findings.

AdCMV-leptin-infused rats developed hyperleptinemia averaging 13.7 ± 2.1 ng/ml during the 14 days of the study (Fig.1 A). Compared with pair-fed normoleptinemic controls, these rats appeared hyperactive and disinterested in food but in otherwise good health. Their total feed intake over the period of 14 days averaged 58% of that of βGal-treated controls (Fig.1 B). Body weight declined during the 1st wk and failed to increase during the 14-day period of study (Fig.1 C), thus confirming our previous report (3). Visible body fat was absent or profoundly reduced in all sites after 1 wk of hyperleptinemia.

Fig. 1.

Comparison of plasma leptin levels (A), body weight (B), and food intake (C) in rats infused with a recombinant adenovirus containing leptin cDNA (AdCMV-leptin), normal rats pair fed to the former, free-feeding normal rats infused with an adenovirus containing cDNA of β-galactosidase (AdCMV-βgal), and obese Zucker diabetic fatty rats [ZDF (fa/fa)].

Blood glucose levels fell below 2.6 mM and remained below normal throughout the 14 days (Fig.2 A). Fasting plasma insulin levels averaged only 33.6 ± 5.0 pM 14 days after the viral infusion, compared with 79.3 ± 5.7 and 95.7 ± 3.6 pM in pair-fed and free-feeding βGal-treated controls, respectively (Fig.2 B).

Fig. 2.

Comparison of plasma insulin (A) and glucose (B) in AdCMV-leptin-infused rats, normal rats pair fed to the former, free-feeding AdCMV-βgal-infused normal rats, and ZDF(fa/fa) rats.

Tissue lipid content in the various groups.

Two weeks after the AdCMV-leptin infusion, the tissue content of TG was measured in liver, skeletal muscle, whole pancreas, and islets of age-matched hyperleptinemic rats, pair-fed and free-feeding control rats, and obese rats. The TG content is shown in Table1, expressed both per gram of wet weight and as a percentage of the tissues of the three control groups. In the hyperleptinemic rats, the TG content of liver averaged 13.3% of that of free-feeding βGal-infused rats and was only 6.4% of that of obese rats; the fat content of their skeletal muscle was 8% of that of free-feeding βGal-infused controls. In whole pancreas, TG content of the hyperleptinemics was 5.3% of that of AdCMV-βGal controls and 12.9% of that of pair-fed controls.

View this table:
Table 1.

TG content in muscle, liver, and pancreas of AdCMV-leptin-infused rats compared with 3 control groups

In islets isolated from hyperleptinemic rats, the TG content per islet was below the sensitivity of the assay as employed. Islets of free-feeding AdCMV-βGal-infused rats contained 28 ng/islet of TG, those of pair-fed rats contained 14 ng/islet, whereas islets of obese rats averaged 990 ng/islet.

Correlations between TG content of pancreas and the product of plasma insulin and plasma glucose and insulin production by isolated pancreata.

To determine the relationship between β-cell function and islet fat content, we examined the correlations between pancreatic fat content and insulin production by isolated perfused pancreata from all groups. The perfusate consisted of either 5.6 mM glucose, 20 mM glucose, or 8 mM arginine plus 5.6 mM glucose. We also measured the correlation between pancreatic fat and plasma [insulin] × plasma [glucose], which distinguished between hypoinsulinemia that results in increased blood glucose concentrations and hypoinsulinemia associated with increased insulin effectiveness.

A highly significant relationship between the plasma [insulin] × [glucose] and TG content of tissues was observed (r = 0.93,P = 0.0001). A similar correlation was observed between plasma insulin alone and tissue TG content (r = 0.93,P = 0.0001). The relationships between TG content of the three tissues and [insulin] × [glucose] are shown in Fig. 3. Correlation with TG content of isolated islets is not shown because islets could not be weighed accurately and varied in size by a factor of 4.

Fig. 3.

Correlations between triglyceride (TG) content of all tissues surveyed (skeletal muscle, liver, and pancreas) in AdCMV-leptin-infused rats, rats pair fed to the former, free-feeding AdCMV-βgal-infused normal rats, and ZDF(fa/fa) rats and the [insulin] × [glucose] product. Correlation with plasma insulin alone gave an identicalr value (not shown).

In the studies of perfused pancreata, insulin production was lowest in the hyperleptinemic group and increased progressively as the TG content of the pancreas increased. Thus, in the hyperleptinemic group, insulin secretion at 5.6 mM glucose was low but not significantly different from that of pair-fed controls, whereas the responses to both glucose and arginine were completely absent (Fig. 4). At the other extreme of tissue TG, insulin production was greatest in the obese ZDF group and intermediate in the AdCMV-βGal-infused and pair-fed control groups. Pancreata from obese ZDF rats were highest in fat content and exhibited the highest basal insulin secretion rate and the most brisk responses to glucose and arginine in absolute terms (Fig. 4). The coefficients of correlation in TG content among the three parameters of β-cell function (basal, glucose-stimulated, and arginine-stimulated insulin secretion) ranged from 0.89 to 0.99.

Fig. 4.

Comparison of TG content of pancreas (A) and insulin production by perfused pancreata (B) isolated from hyperleptinemic rats, normal Wistar rats pair fed to the former, free-feeding AdCMV-βgal-infused rats, and ZDF(fa/fa) rats.


The goal of this study was to determine whether the fat content in the target tissues of insulin and in the pancreas is correlated with both the effectiveness of insulin and the rate of its secretion. If so, it would be consistent with the notion that the responses of β-cells are tailored to insulin need by a derivative of TG such as FFA. This would facilitate maintenance of euglycemia despite the changing body composition associated with obesity. It should, however, be stressed that correlations cannot prove cause and effect and that alternative interpretations of these correlations are equally possible. For example, primary hyperinsulinemia could increase lipogenesis and cause increased tissue TG, which could in turn give rise to insulin resistance (24, 25).

We employed both obese ZDF rats and lean Wistar rats. The availability of a unique “fat-free” animal, the hyperleptinemic rat, a novel model of extreme lipopenia that is the antithesis of obesity, made it possible to examine these correlations over the broadest possible spectrum of TG content, ranging from extreme lipopenia to severe obesity. In acute perfusion experiments, leptin had no effect on insulin production (data not shown), but in chronic studies of cultured islets, leptin abolished the insulin responses to both glucose- and arginine-stimulated insulin secretion while depleting islet TG; FFA restored the insulin response almost immediately (19).

We observed that the production of the fasting plasma insulin and blood glucose levels, a crude index of insulin sensitivity, was significantly correlated with TG content, as was the fasting plasma insulin level by itself. Insulin sensitivity was markedly enhanced (low insulin-glucose product) in the lipopenic hyperleptinemic group. Fasting glucose levels in these groups were in a hypoglycemic range despite reduction in fasting plasma insulin levels; the insulin-glucose product of hyperleptinemic rats averaged 13% of normal controls and 4% of obese rats, evidence of exquisite insulin sensitivity. Clearly lipid content in target organs was correlated with this index of insulin sensitivity;r values ranged from 0.96 to 0.99 and were statistically significant (P < 0.05). However, because serum cortisol levels were not measured, a role of hypocortisolism in the insulin sensitivity of the hyperleptinemic rats cannot be excluded.

Pancreatic TG content was correlated with β-cell function as reflected by the fasting insulin level (r = 0.97;P = 0.037). All phases of insulin secretion by the isolated perfused pancreas, basal and both glucose- and arginine-stimulated insulin secretion, also varied directly with the TG content of the pancreas.

The results are consistent with, but do not prove, the hypothesis that tissue TG content determines the level of insulin production and matches it to the level of insulin effectiveness, perhaps by providing an intracellular source of FFA retrieved from intracellular (TG) storage sites. Because the TG content of skeletal muscle and pancreas is also positively correlated, it is likely that it reflects generalized changes in tissue fat. Teleologically, this postulated linkage between insulin requirement and insulin secretion would serve to reduce the risk of diabetes in obesity and to minimize the possibility of hypoglycemia in undernutrition.

These observations follow more than three decades of research linking fatty acid metabolism to carbohydrate metabolism and diabetes (2, 4, 5,7, 8, 11, 13, 15, 16, 18, 20-22, 24-26, 29, 31-35, 38). Most attention has been directed at the effects of plasma lipids in reducing the insulin sensitivity of target tissues (3, 8, 13, 18, 23). However, the initial reports of stimulatory effects of FFA on islets (5, 11, 22) and, more recently, their inhibitory actions on β-cell function (2, 7, 16, 38) have laid the groundwork for the present studies. The main difference is the emphasis on tissue TG content rather than perfusing levels of FFA and TG.


We thank Kay McCorkle, who provided technical support, and Sharryn Harris, who provided secretarial assistance.


  • Address for reprint requests: R. H. Unger, Center for Diabetes Research, Univ. of Texas Southwestern Medical Center, 5323 Harry Hines Blvd., Dallas, TX 75235–8854.

  • This study was supported by National Institute of Diabetes and Digestive and Kidney Diseases Grant DK-02700–37, the National Institutes of Health/Juvenile Diabetes Foundation Diabetes Interdisciplinary Research Program, and the Department of Veterans Affairs Institutional Research Support Grant SMI 821–109.


View Abstract