Improved Insulin Sensitivity Is Associated With Restricted Intake of Dietary Glycoxidation Products in the db/db Mouse

Female 4-week-old db/db (+/+) C57/ BL/KsJ (n = 20) and control mice (n = 16) were purchased from The Jackson Laboratories (Bar Harbor, MN). Mice were given pelleted food and water ad libitum in a specific pathogen-free environment at the Center for Laboratory Animal Science, Mount Sinai School of Medicine. Animal care and experimental procedures conformed with the “Guide for the Care and Use of Laboratory Animals” (Department of Health, Education, and Welfare, publication no. NIH 78-23, 1996). After 1 week of adjustment, db/db and control mice were randomly divided into two groups and placed for 20 weeks on two commercially available standard rodent diets that were similar in nutritional and caloric content (Table 1) but had different AGE content based on enzyme-linked immunosorbent assay (ELISA) (see below). AIN-93G (Bio-Serv, Frenchtown, NJ) was used for the low-AGE diet (LAD), and PicoLab Rodent Diet 20 (Labdiet; Purina Mills, St. Louis, MO), with a 3.4-fold higher AGE content, was used for the high-AGE diet (HAD) (Table 2). Fasting blood glucose, insulin levels, and body weights were monitored biweekly. After 1 week of adjustment, food intake (in grams of food) of individual mice was recorded daily for 1 week. Intravenous assays were performed in all mice in groups of five. Because statistical significant difference was reached with five mice per group, we report our results with this number of animals.

AGE concentrations in diets (LAD and HAD) and in mouse serum were determined by ELISA using a monoclonal antibody raised against AGE-KLH (4G9; Alteon, Ramsey, NJ), as previously described (29–31). The antibody is highly reactive with BSA modified by CML (32), an established biomarker for AGEs (10,18,22, 33). We also assessed MG derivatives in the diets as well as in pooled mouse serum by ELISA using a monoclonal anti-MG-BSA antibody (MG3D11) raised against MG-modified BSA (21.8 MG-modified Arg/mol BSA, by HPLC [high-performance liquid chromatography]; obtained from Dr. Y. Al-Abed, The Picower Institute, Manhasset, NY) (34). MG3D11 was found to be strongly immunoreactive against MG-ovalbumin and AGE-BSA, but not with CML-BSA or unmodified BSA.

Blood glucose was determined with an Elite Glucometer (Bayer, Mishawaka, IN). Serum insulin levels were measured with the Ultra-Sensitive Rat Insulin ELISA Kit (Alpco Diagnostics, Windham, NH) using rat standards (100% cross-reactivity with mouse insulin). Intravenous glucose tolerance test (IVGTT) was performed by tail vein injection of glucose (1 g/kg i.v.) after an overnight fast (13–15 h). Blood samples were taken before and at intervals between 2 and 60 min after glucose infusion. Insulin tolerance test (ITT) was performed by intraperitoneal injection of human regular insulin (0.75 units/kg) (Novolin R; Novo Nordisk, Clayton, NC) after a 6-h fast. Blood samples were collected before and 15, 30, 60, 90, and 120 min after insulin administration.

Serum leptin was measured using the mouse leptin ELISA kit (Crystal Chemical, Chicago). Serum free fatty acids were measured using the NEFA (nonesterified fatty acid) C kit (Wako Chemicals, Neuss, Germany) with oleic acid as standard. Serum triglycerides and glycerol were measured using the GPO-Trinder colorimetric assay kit (Sigma Diagnostics, St. Louis, MO). Total plasma cholesterol and HDL cholesterol were determined using the infinity cholesterol reagent colorimetric assay kit (Sigma Diagnostics). For plasma HDL cholesterol measurement, plasma samples (60 μl) were fractionated by two sequential steps of ultracentrifugation using a Beckman TLA-100 rotor at 70,000 rpm for 3 h at 4°C followed by centrifugation with KBr (1.12 density) at 70,000 rpm at 4°C overnight. After centrifugation, the lower portion of the samples was used for the determination of HDL cholesterol.

For monitoring insulin levels, 40-μl whole-blood samples were taken from the retroorbital sinus by a capillary pipette containing heparin. Serum was separated from blood cells by centrifugation (4,000 rpm for 15 min) and stored at −20°C until measurement. During the IVGTT and ITT, blood was collected from the tail vein. For the lipid profile, 200 μl whole blood was collected from the retroorbital sinus and 5 μl of 0.5 mol/l EDTA, pH 8, was immediately added. Plasma was separated by centrifugation (4,000 rpm for 15 min.) and stored at 4°C; measurements were completed within 2 days of blood withdrawal. In all the assays described above, samples were not pooled but processed and analyzed individually. However, for lipoprotein separation, it was necessary to pool samples (0.5 ml) from eight animals per group. The pooled plasma from each group was subjected to fast-performance liquid chromatography (FPLC) gel filtration on two Superose 6 columns. Ninety-five fractions were collected, and the total cholesterol concentration of each fraction was assessed enzymatically using the Sigma kit.

After a 5-h fast, [1,2-³H]deoxy-d-glucose (2-[³H]DG) (NEN Life Science Products, Boston, MA) (0.5 μCi/g body wt) was administered as an intravenous bolus. Sixty minutes after 2-[³H]DG administration, the animals (n = 5) were killed. Abdominal adipose tissue and the livers were removed, weighed, and solubilized with NCS^R-II Tissue Solubilizer (Amersham, Arlington Heights, IL). Radioactivity was counted in the samples using a liquid scintillation counter (Model 1219; Perkin Elmer, Gaithersburg, MD).

After the animals were killed, pancreata were removed and fixed overnight in 10% formalin. Fixed tissues were processed for paraffin embedding. Sections were cut at a thickness of 4 μm, mounted on glass slides, and stained with hematoxylin and eosin. For detection of insulin, 10 consecutive paraffin-embedded sections of the pancreata were stained with a guinea pig polyclonal anti-swine insulin antibody (Dako, Carpinteria, CA). The immunoreactivity was detected by the Multi-link-HRP Supersensitive System (BioGenex, San Ramon, CA).

Twenty-four-hour urine was collected from all mice, which were housed singularly in metabolic cages. Total urine glucose was measured by Elite Glucometer (Bayer, Mishawaka, IN), total urinary creatinine content was assessed with the Reflotron creatinine reagents in the Reflotron system (Roche, Indianapolis, IN), and urinary albumin was measured by an ELISA kit (Alpco Diagnostics, Windham, NH).

Numerical data are expressed as means ± SE. An unpaired two-tailed Student’s t test was used for statistical analysis. P values <0.05 were considered significant.

Based on published daily requirements, both standard diets (LAD and HAD) were nutritionally equivalent and well balanced with supplemental micronutrients against possible antioxidant vitamin losses (Table 1). As shown in Table 2, they differed by 3.4-fold in AGE content, assessed as CML-like immunoepitopes. Because the daily food intake was equal between HAD- and LAD-fed db/db mice (Fig. 1D), HAD-fed db/db mice ingested ∼3.4-fold more CML than the LAD-fed db/ db mice. This was reflected in part in the fasting serum CML levels, which were twofold lower in LAD-fed animals than in HAD-fed animals (P = 0.02) (Table 2). A similar pattern between dietary and pooled mouse serum MG levels was obtained using a MG-specific ELISA (LAD 0.7 μg/mg vs. HAD 2.3 μg/mg; LAD-fed db/db mice 22.2 nmol/ml vs. HAD-fed db/ db mice 34.4 nmol/ml; LAD-fed control mice 9.3 nmol/ml vs. HAD-fed control mice 27.6 nmol/ml; n = 8–10/ group). Thus, steady-state serum levels of two distinct AGEs, CML and MG, were consistent with the estimated content of the respective AGEs in the dietary formulas (Table 2).

As expected, all db/db mice exhibited marked hyperinsulinemia compared with controls throughout the study. Within 6 weeks of treatment, LAD-fed db/db mice displayed significantly lower fasting serum insulin levels, which, by 14 weeks, remained one-quarter of those in HAD-fed db/db mice (P = 0.01) (Fig. 1B). In the control mice of either diet group, fasting glucose and fasting serum insulin levels remained similar regardless of diet (Fig. 1A and B). After 14 weeks, fasting insulin levels in HAD-fed db/db mice began to decline, possibly due to the expected pancreatic islet atrophy in the db/db (+/+) C57/BL/KsJ strain (28).

No differences in body weight and growth rate were noted up to week 8, when a plateau was reached in the LAD-fed db/db mice. Thereafter, LAD-fed db/db mice stopped gaining and began to lose weight, reaching significantly lower levels compared with the HAD-fed db/db mice (P = 0.001) (Fig. 1C) by week 14. In addition, at the end of the study, despite identical food intake, serum leptin levels were significantly lower in LAD-fed versus HAD-fed db/db mice (36 ± 2.5 vs. 48 ± 4 ng/ml; P = 0.02). Control mice from either diet group showed no significant differences in body weight (Fig. 1C) or in serum leptin concentrations (2.1 ± 0.4 vs. 2.7 ± 0.4 ng/ml). After week 15, body weight began to decline in both db/db diet groups, consistent with previous observations (28).

To determine possible changes in insulin sensitivity versus IR, we performed an ITT at week 11. As shown in Fig. 2A, LAD-fed db/db mice revealed a significant improvement in glycemic response compared with that of HAD-fed db/db mice (at 60 min, P = 0.007). Of interest, ITT data from the control LAD-fed mice were significantly different, with a better glucose disposal rate than that of the HAD-fed control mice (at 60 min, P = 0.0003) (Fig. 2B).

Based on the above findings, the effect of LAD on glucose tolerance was tested by IVGTT in all groups. At week 18 of the study, glucose tolerance was significantly improved in LAD-fed db/db mice compared with HAD-fed db/db mice (at 40 min, P = 0.004) (Fig. 2C). A similar tendency (but not statistically significant) in glucose tolerance was obtained in control mice (Fig. 2D).

To investigate insulin action on peripheral tissues, glucose uptake by white adipose tissue and liver was compared in the two diet groups at week 20. 2-[³H]DG uptake in adipose tissue from LAD-fed db/db mice increased by twofold compared with HAD-fed db/db mice (P = 0.02) (Fig. 3A). Adipose tissue 2-[³H]DG uptake in HAD-fed db/db mice was 6.2-fold lower than in age-matched control animals. However, LAD-fed db/db mice showed a threefold lower 2-[³H]DG uptake than the respective control group. 2-[³H]DG uptake by adipose tissue was similar in control animals on either diet. In comparison, liver tissue 2-[³H]DG uptake in LAD- and HAD-fed db/db mice showed no discernible difference (Fig. 3B). Of note, age-matched HAD-fed control mice exhibited a significantly lower liver tissue 2-[³H]DG uptake than LAD-fed control mice (P = 0.05) (Fig. 3B).

Plasma lipids were measured after an overnight fast (13–15 h) at week 17. Total cholesterol levels in db/db mice from both dietary groups were significantly higher than in control animals (Table 3). In LAD-fed db/ db mice, they were significantly higher than in HAD-fed db/db mice (P < 0.0001); a minor but significant difference related to dietary AGE intake was seen in the control groups (P = 0.02) (Table 3). Although db/db mice exhibited higher plasma triglyceride and FFA levels than the control groups (by 1.5-fold, P < 0.001), no difference was noted between the LAD and HAD groups (Table 3).

Lipoprotein fractions in pooled plasma from each group of mice were resolved by FPLC (35). Mouse plasma appeared to contain the majority of cholesterol in the form of HDL (Fig. 4, fraction 58–70). As expected, the HDL peak was higher in db/db mice than in the corresponding control mice, regardless of diet (36). Most notable, however, was a marked rise in the HDL peak of LAD-fed mice (Fig. 4). In addition, in these animals, we detected significant peaks in the intermediate-density lipoprotein/LDL range (Fig. 4, fraction 18–21). The distribution of plasma cholesterol in the various lipoprotein fractions was also examined by sequential ultracentrifugation of HDL. By this method, consistent with the FPLC profile, HDL cholesterol levels were increased in all db/db mice compared with the respective control animals. However, in LAD-fed db/db mice, HDL cholesterol was twofold higher than in the HAD-fed db/db mice (P < 0.0001) (Table 3). Furthermore, control LAD-fed mice showed a tendency toward higher HDL cholesterol levels than control HAD-fed mice (Table 3).

Histological examination and insulin immunostaining were performed in pancreatic sections from each group at the end of the study. Islets of HAD-fed db/db mice exhibited hyperplasia and hypertrophy combined with loss of islet structure and cellular homogeneity, changes seen only infrequently in islets from LAD-fed db/db mice. In addition, islet degeneration and insulin degranulation, including glucagon-producing cell displacement from the periphery to the center of the islets, were more evident in HAD-fed db/db mice than in LAD-fed db/db mice. Overall, LAD-fed db/ db mice presented better preserved, compact islets with intensely positive staining for insulin (Fig. 5). Comparing 10 consecutive pancreatic sections per mouse from the LAD- (n = 5) and HAD-fed (n = 5) control mice, we could not observe a discernible difference in islet structure and insulin immunostaining.

Total 24-h urinary glucose excretion was comparable in LAD- and HAD-fed db/db mice (1.6 ± 0.15 vs. 1.4 ± 0.2 mg/24 h; n = 8–10/group), consistent with similar blood glucose levels, whereas the albumin/creatinine ratio revealed a clear difference (1.1 ± 0.1 vs. 2.1 ± 1.1; n = 8– 10/group). These data, although not reaching statistical significance in the present study, were consistent with earlier observations (37).

The present study reveals a significant relationship between dietary AGE content, circulating AGEs, and insulin sensitivity in a murine model of type 2 diabetes and obesity. It furthermore illustrates the protective effects of a standard rodent diet with low levels of certain AGEs against the progression of IR and possibly against obesity in the same model.

The effects demonstrated in the present study were not attributable to quantitative differences in basic nutrients, vitamins, or energy profiles of the two diets (Table 1). Instead, the findings were consistent with the dietary content in protein/lipid glycoxidation derivatives, herein reflected in levels of CML and MG immunoepitopes. In this study, serum levels of CML and MG correlated with the respective LAD or HAD given to each experimental group, whether diabetic or nondiabetic.

CML and MG have been recognized as indicators of glycoxidant burden in vivo because both are well-characterized products of protein and lipid glycoxidation reactions (10,16,17), correlate with age (22,23) and with chronic hyperglycemia (9, 10,16), and are thought to contribute to intracellular oxidative stress and tissue damage (16,18). Based on the vast heterogenity of AGEs, however, the identity of the actual AGE related to the observed metabolic effects remains speculative because the pathogenic structure of the majority of these remains elusive. A similar positive relationship between ingested and circulating AGE levels has been observed previously, whether in connection to a defined oral test prepared with or without carbohydrate (14), to a single rodent diet (AIN93G) prepared under different temperatures (27,37), or to complex foods consumed by human subjects (13). The present study used two standard diets that were nutritionally equivalent despite being from different vendors—thus more closely resembling the naturally mixed human diets—and provides further support that food-based AGEs may be important contributors to the circulating AGEs.

More importantly, the present study reveals a novel finding in that certain diet-related glycoxidant substances influence the progression of IR in C57/ BL/KsJ db/db mice, in which obesity and type 2 diabetes are caused by a single gene mutation (38).

Despite equal food intake in the two groups, LAD-fed db/ db mice, already obese and diabetic, ceased to gain weight within 10 weeks of study, and, by 20 weeks, they exhibited a significant weight loss. Throughout this period, fasting serum insulin levels, a well-established indicator for the progression of IR (39), were significantly lower in LAD-fed db/db mice than in HAD-fed db/db mice, without the characteristic rise and fall typically seen in the C57/ BL/KsJ db/db strain (36) and in the HAD-fed db/db mice herein. A further indication of improved peripheral glucose disposal and insulin sensitivity in the LAD-fed db/db mice was provided by the nearly complete restoration of glycemic response to IVGTT and ITT.

The insulin kinetics seen in the LAD-fed mice were consistent with the well-preserved pancreatic islet morphology and function demonstrated in this group at the end of the study. This may have resulted from a reduced demand for insulin production, which is attributable to improved peripheral glucose utilization in this group. In contrast, islets of age-matched HAD-fed db/db mice displayed the expected hypertrophy coupled with a loss of insulin-producing cells and the normal islet architecture (40). These results, together with the metabolic findings, reinforced the protective effects related to the LAD diet.

Glucose uptake in abdominal adipose tissue revealed that LAD improved IR by partially restoring the severely impaired insulin-stimulated glucose uptake by adipocytes in LAD-fed db/db mice. Due to the relatively high fat content in muscle samples of the obese db/db mice, accurate determination of 2-[³H]DG specific uptake by skeletal muscle cells was questionable. Thus, this study relied on glucose uptake by abdominal adipose tissue, an equally important contributor to IR (41). The doubling of glucose uptake by fat tissue in LAD-fed db/db mice, although far from that of controls, together with the significantly improved IVGTT and ITT, supported the hypothesis that dietary AGE intake exerts direct effects at sites of peripheral insulin action; a restriction of AGE intake clearly results in improved glucose uptake. In the context of equal food intake, the body weight loss in the LAD-fed db/db mice is also consistent with a shift in endogenous glucose metabolism, resulting in a more efficient glucose disposal rate.

By comparison, hepatic glucose output in the db/db mice, assessed as fasting glycemia, was not affected by the LAD diet. Increased hepatic glucose output is primarily responsible for fasting hyperglycemia in diabetes. A relative insulin deficiency, along with an increase in the counterregulatory hormones glucagon and catecholamines, is involved in the pathogenesis of enhanced hepatic gluconeogenesis. IR and increased hepatic gluconeogenesis are already present in young db/db C57/BL/KsJ mice at a time when they are still normoglycemic (42,43). A possible explanation for the enhanced hepatic gluconeogenesis, not influenced by AGE intake in this study, is the increased secretion of glucagon and the glucagon/insulin ratio in db/ db C57/BL/KsJ mice (44,45). Further supportive evidence for increased gluconeogenesis in type 2 diabetes animal models was provided by increased liver mRNA levels of the transcription factor PGC-1 and the gluconeogenic enzyme PEPCK (46).

As expected, all db/db mice developed the dyslipidemia previously described in this model (36), with elevated serum FFA, tryglyceride, total plasma cholesterol, and HDL cholesterol levels. As previously noted, the increase in HDL cholesterol in db/db mice is in contrast to the decreased levels frequently found in humans with obesity or diabetes but is consistent with another mouse model of obesity, the ob/ob mouse (47). In those mice, it was found that there was a decrease in hepatic apoAI mRNA that resulted in decreased HDL production, but there was an even greater decrease in HDL plasma clearance, which explained the elevated HDL cholesterol level. In the present study, there were significantly higher total serum cholesterol levels in the LAD-fed db/db mice than in the control and HAD-fed mice, attributable almost exclusively to a marked increase in HDL cholesterol concentrations. This finding constitutes an additional novel observation and may be in part related to the improvement in IR found in the LAD-fed db/db mice, which may further elevate the normally high level of plasma HDL cholesterol in db/db mice. The molecular mechanism for the further increase is not known but may be related to our previous findings in diabetic rats (48). In those studies, we also found (as in the ob/ob mouse) decreased hepatic apoAI mRNA levels that were increased by insulin treatment and that would be expected to result in increased production of HDL. Although as noted above, unlike db/db or ob/ob mice, obese and diabetic patients have depressed plasma HDL cholesterol levels but, like the db/db mice in the present study, they also exhibit elevations in HDL cholesterol levels as insulin sensitivity increases (49, 50). In addition, treatment of type 2 diabetic patients with the insulin sensitizer rosiglitazone is thought to favorably affect the LDL and HDL cholesterol heterogeneity (51). Given the substantial clinical implications of a potentially similar effect on human HDL levels imparted by the modulation of dietary AGEs, further studies with regard to the mechanisms involved in the elevation in the LAD db/db mice are warranted.

From the present studies, it can be inferred that a diet rich in AGEs may constitute a critical environmental modulator of endogenous metabolism in diabetic mice.

This work was supported in part by National Institutes of Health Grants AG09453 to H.V. and HL61814 and HL70524 to E.A.F. S.M.H. is a recipient of a fellowship award from Deutsche Forschungsgemeinschaft (DFG).

The authors thank Jenny Suriawinate for technical support on immunohistochemistry, Anthony Doré for assistance with the animal surgical procedures, and Sabina Omerhodzic for the lipoprotein assays.