GFJ was squeezed from fresh California Ruby Red grapefruit provided by the California Grapefruit Growers Cooperative, centrifuged at 10,400×g for 10 min at 4°C to remove pulp, amended with 0.15% saccharin (w/v), divided into 25 ml aliquots, and stored at -20°C [13], [14]. The pH of this clarified preparation (cGFJ) was 3.5, compared to 5.5 for the sweetened water used as control. We determined that the caloric content of the cGFJ was 1335 J/ml by bomb calorimetry of a lyophilized sample as previously described [15]. Control mice were given water with 4% glucose (w/v) and 0.15% saccharin (hereafter called control or control water), so that all groups consumed isocaloric liquids with the same amount of saccharin.

Procedures were approved by the University of California-Berkeley Animal Care and Use Committee and were done according to AAALAC guidelines. Four-week-old male C57BL/6J mice were purchased from Jackson Laboratories (catalog # 000664). Mice were housed individually and were fed purified diets upon arrival (unless noted otherwise) with either 10% fat (LFD) (Research Diets Cat. # D12450B) or 60% fat (HFD) (Research Diets Cat. #D12492). Any stress induced by housing mice in isolation was normalized by equivalent and concurrent treatment of mice in each experiment.

Mice were weighed three times per week. Food consumption was monitored twice per week. Mice were divided randomly into groups of six (unless noted otherwise): controls (water with 4% glucose and 0.15% saccharin); 50% cGFJ (50% cGFJ/water with 0.15% saccharin); 25% cGFJ (25% cGFJ/water with 4% glucose and 0.15% saccharin); naringin (0.72 mg/day in water with 4% glucose and 0.15% saccharin); metformin (7.5 mg/day metformin with 4% glucose and 0.15% saccharin); metformin + cGFJ (7.5 mg/day metformin with 0.15% saccharin in 50% cGFJ). Liquids were given in volumetric bottles (Med Associates, cat # PHM-127-15) to quantify consumption and were replaced daily.

Glucose was measured Monday, Wednesday, and Friday between 9 and 11 AM with a NovaMax blood glucose monitor in blood from a tail prick (AmericanDiabetesWholesale). Glucometer values were corrected using a glucose enzymatic assay kit (Sigma, cat # GAHK20-1KT).

For the GTT, mice were fasted overnight and injected i.p. with 0.2 ml of glucose in sterile water to deliver 2 g/kg glucose. For the ITT, mice were fasted 4 hr and injected i.p. with 0.75 units of insulin/kg. For the PTT, mice were fasted overnight and injected i.p. with 0.2 ml of pyruvate in sterile PBS to deliver 2 g/kg.

Insulin concentrations were determined with a high-range insulin ELISA kit (ALPCO cat# 80-INSMSH-E01, E10) in blood taken retro-orbitally after an overnight fast. Mice were allowed access to food 4 hr and were re-sampled.

Protein concentrations were assayed with a BCA protein assay kit (Thermo Scientific cat# 23227). TG concentrations were assayed with the Infinity TG kit (Thermo Scientific cat# TR2241).

Livers were fixed 1 hr at 4°C with 4% paraformaldahyde, and were incubated overnight at 4°C with a cryopreservation medium of 30% sucrose, 20% Optimal Cutting Temperature medium (VWR cat# 25608-930), and 50% Superblock consisting of Block plus 2% normal donkey serum. Block consisted of 50 ml 10× Hanks balanced salt solution, 50 ml fetal calf serum, 5 g bovine serum albumin, and 0.25 g saponin in 500 ml. Blocks were sectioned into 8 µm strips at −23°C. Sections were stained 1 hr at room temperature with a nonpolar BODIPY probe (Molecular Probes cat# D-3922). Slides were mounted with DAPI/glycerol mounting medium (Life Technologies cat# S36938) and stored at −20°C until imaging.

Real-Time PCR was performed using the TaqMan Universal Master Mix II (Applied Biosystems). Primers were purchased from Integrated DNA Technologies (Table 1).

Livers were homogenized with a Polytron PT2100 in radio immunoprecipitation lysis buffer containing protease and phosphatase inhibitors (Sigma cat# P8340 and cat# P5726) and centrifuged 5 min at 3220×g. Protein (50 µg) was loaded onto a 4–20% Tris-glycine gel. Antibodies were purchased from Cell Signaling. Signals were quantified with a LI-COR Odyssey gel analysis system and normalized to β-tubulin.

At 4-weeks-old, mice (7 per group) were fed a HFD for 2 wk while drinking 50% cGFJ or control water ad libitum. Mice were fasted overnight and gavaged with 740 kBq [¹⁴C]oleate in 200 µL olive oil, or 740 kBq [³H]2-deoxy-D-glucose in 200 µl sterile PBS containing 2.5 g/kg glucose, or 740 kBq [¹⁴C]taurocholic acid in 500 µM taurocholic acid in sterile water. Blood was taken retro-orbitaly 15, 60, 120, 180, and 240 min after dosing. Radioactivity was measured in 10 µL serum.

Mice were assayed individually by indirect calorimetry (Columbus Instruments, Columbus Ohio, US) during a fast or after fasting 7 hr and re-feeding 1.1 g of the HFD, followed by fasting overnight. Experimental analyses were started between 3–4 PM and continued for ∼23 hr. Activity was monitored in 10 min intervals.

Total liver FA concentrations (C16:0, C16:1, C18:0, C18:1 and C18:2) were determined by gas chromatography-flame ionization detection [15]. Palmitate synthesis was measured by analysis of stable isotope incorporation. On day 0 mice were injected i.p. with 100% D₂O (Sigma cat # 151890) containing 0.9% NaCl (0.35 ml/g body weight). Mice were given 8% D₂O in their drinking solutions for 17 d. Deuterium incorporation into serum and liver was determined by GC/MS analysis [16]–[18]. Palmitate synthesis was calculated as the fraction of newly synthesized palmitate × total mg palmitate.

Statistical analysis was performed as described in the figure legends. Data are means ± SE. Statistical significance was determined by two-tailed, unpaired t-tests.

Based on average daily liquid consumption, mice were adverse to drinking unsweetened 100% cGFJ, sweetened 100% cGFJ, or saccharin/cyclamate sweetened 50% cGFJ/water (v/v) (Fig. 1A-C). In contrast, cGFJ consumption was comparable to control-group liquid consumption when mice were given 50% GFJ sweetened with 0.15% saccharin (Fig. 1D).

Mice were fed a either a LFD or a HFD for 100 d with access to “control water” (see Material and Methods) or 50% cGFJ as their sole sources of liquids. cGFJ intake did not affect average daily nor cumulative food consumption during a LFD (Fig. 2A, B). Total liquid consumed by LFD-fed mice was 141±1.1 ml water vs. 135±0.5 ml 50% cGFJ (P<0.05). This 6 ml difference in liquid consumed over 100 days represents an energy intake difference of <8 kJ or ∼0.002% of total caloric intake. Total calories consumed by the cGFJ group were 4822±83 kJ vs. 5023±163 kJ for controls (P>0.05). No differences occurred in weight between the GFJ and control (Fig. 2C). Consistent with similar weights, no differences occurred in epididymal fat pads for LFD-fed mice (Fig. 2D).

cGFJ intake also did not affect average daily nor cumulative liquid and food consumption during feeding a HFD (Fig. 2E, F). The 50% GFJ group consumed 137±0.5 ml vs. 140±2.2 ml by controls (P>0.05). Cumulative food consumption was 5580±193 kJ for the GFJ group vs. 5684±155 kJ for controls (P>0.05). In contrast to the LFD-fed mice, the HFD-fed mice with access to 50% GFJ weighed 18.4% less than controls at the end of the 100 d: 31.4±0.7 g vs. 38.5±2.8 g, P<0.05 (Fig. 2G). Body weight trended lower within 15 d after initiating GFJ access and became statistically significant by day 78. Epididymal fat pads of the cGFJ group weighed 50% less than those of the control group (Fig. 2H).

The caloric value of feces collected over the final 24 hr of the 100-day-study from the cGFJ group was similar to controls, as measured by bomb calorimetry, even though average daily fecal mass was ∼23% lower in the cGFJ group (Fig. 2I).

After two weeks feeding a HFD, radiolabeled metabolites in serum (AUC) of mice gavaged with [³H]glucose, [¹⁴C]oleic acid or [¹⁴C]taurocholic acid did not differ between cGFJ and controls during a 240 min assay (data not shown).

Indirect calorimetry of fasted mice revealed no significant differences in 24 hr energy expenditure (VO₂ and VCO₂), substrate use (respiratory exchange ratio), heat production or activity between the HFD-fed GFJ and control groups (data not shown).

At the end of the LFD study, no significant difference in fasting blood glucose levels occurred between cGFJ group and control (Fig. 3A). In the fed state, cGFJ had no effect on serum insulin levels in mice fed a LFD (data not shown). cGFJ produced no significant differences in the GTT or ITT at either time (data not shown). Even without a high fat challenge, however, fasting serum insulin levels were 2-fold lower in the cGFJ vs. the control group fed a LFD (Fig. 3B).

At week 7 of the HFD, the GTT blood glucose area under the curve (AUC) was 6% lower (P<0.05) for the HFD-fed GFJ group compared to controls (data not shown). By week 11 this difference increased to 11% (Fig. 3C, D). In an initial ITT done at week 9, the blood glucose AUC was ∼17% lower (P<0.05) in the cGFJ group compared to controls (data not shown). This difference was maintained at week 13 (Fig. 3E, F). At the end of the HFD study, fasting blood glucose values were 13% lower in cGFJ mice compared to controls (Fig. 3G). Fasting serum insulin levels were 72% lower in the cGFJ group compared to controls (Fig. 3H). Fed insulin levels were not different between the two groups (data not shown).

Improved insulin sensitivity as a result of cGFJ supplementation was confirmed by evaluating activation of AKT, the insulin receptor downstream kinase, in fasted mice. GFJ produced a 3-fold and 1.4-fold increase, respectively in p-AKT/total AKT ratios in the quadricep muscle and liver, compared to controls (Fig. 4A).

Consumption of 50% cGFJ reduced the amount of TG in livers of mice fed a HFD by 38% compared to controls, and reduced the numbers and sizes of lipid droplets after ten days (Fig. 4B). Using an in vivo heavy water labeling approach [18], we determined that total fatty acid de novo synthesis in liver did not differ significantly between cGFJ and controls, and the synthesis rates of specific fatty acids were similar to control, except for oleate (Fig. 4C). In contrast to fatty acid synthesis, a PTT showed that the 10-day intervention in HFD-fed mice produced a 9% decrease in gluconeogenesis (Fig. 4D).

To determine the impact of cGFJ on mice with diet-induced obesity, animals were fed a HFD 10 wk and then allowed access to 50% GFJ, while continuing the HFD. Introduction of cGFJ did not change daily or cumulative liquid or calorie consumption (Fig. 5A, B). By the end of this experiment on day 55, the 50% cGFJ group weighed ∼8% less than controls (GFJ, 33.4±1 g vs. control, 36.4 g±1.9 g, P<0.05) (Fig. 5C). Body weights had become significantly different starting on day 9 (P<0.05). A 13% decrease in blood glucose occurred as early as day 10 post intervention (Fig. 5D). Final resting serum glucose levels in the cGFJ group were 110±1 mg/dL (96.1±1 mM) compared to the control value of 119±1 mg/dL (6.6±1 mM) (P<0.5). A GTT at week 6 post intervention revealed a 12.5% decrease (P<0.05) in the AUC, consistent with increased glucose tolerance (Fig. 5E). This observation was augmented by an ITT at week 7, which showed an AUC for the cGFJ group 9.5% lower than control (Fig. 5F).

We compared the metabolic effects of GFJ with those of naringin, a bioactive compound in cGFJ, and metformin, a drug used widely to treat type 2 diabetes and nonalcoholic steatohepatosis during 106-day of feeding a HFD [19], [20]. Liquid and calorie consumption was comparable among all four groups (Fig. 6A, B). Body weights of the 50% cGFJ group, but not of the metformin- or naringin-supplemented groups, were significantly lower compared to controls at the end of the study (control, 32.9±0.5 vs. cGFJ, 28.2±g, P<0.05) (Fig. 6C). All three intervention groups had a statistically significant drop in blood glucose compared to controls by day 8 (113 to 110 mg/dL or 6.3 to 6.1 mM), which continued on days 10 (119 to 114 mg/dL or 6.6 to 6.3 mM) and 17 (117 to 111 mg/dL or 6.5 to 6.2 mM) (Fig. 6D). Blood glucose at the end of the study was ∼20% lower in the three treatment groups compared to the control.

The AUC of a GTT at week 7 was not significantly lower for the cGFJ, naringin, or metformin groups than the control group (data not shown). By week 13, however, the AUC values for the cGFJ, naringin, and metformin groups were 8, 7 and 12% lower than the control group, respectively (P<0.05 for all) (data not shown). An ITT at week 14 revealed AUC values 9, 8 and 15% lower for the GFJ, naringin, and metformin groups, respectively, relative to the control group (P<0.05 for all) (data not shown).

To determine the possibility of synergistic or additive effects on blood glucose, mice fed a HFD were allowed access to metformin in a solution of 50% cGFJ for 17 d. In the same experiment a second group of mice was allowed access to 25% GFJ. The combination of metformin and 50% cGFJ produced no significantly different effect on blood glucose relative to either alone (Fig. 7A). Blood glucose in the 25% cGFJ group decreased comparably to the 50% cGFJ, metformin, and metformin plus 50% GFJ groups. The final blood glucose value of each treatment group was 11–14% lower than control (P<0.05).

We compared the effects of metformin, naringin and 50% cGFK on p-AMPK levels in liver and muscle of mice fed a HFD for 15 weeks. cGFJ decreased the ratio p-AMPK/AMPK in liver 25%, compared to a 52% decrease in response to naringin, and to a 1.6-fold increase in response to metformin (Fig. 7B). We assayed p-acetyl-CoA-carboxylase (ACC), because AMPK deactivates ACC via phosphorylation [21], [22]. p-ACC was unchanged in the GFJ and naringin groups, but metformin increased p-ACC in liver 2-fold (Fig. 7C). In muscle, only metformin increased p-AMPK (2.7-fold) (Fig. 7D) and p-ACC/AC (2.6-fold) (Fig 7E).

In a further attempt to determine the mechanism(s) of cGFJ effects we assessed changes in expression of select metabolic genes in liver and small intestine of mice fed a HFD for 17 days (Fig. 8A, B) or after 100 d (Fig. 8C, D). In the short-term experiment, the only significant change in livers of mice treated with cGFJ was a 50% decrease in SHP, and the only significant change in intestine was a 9.5-fold increase in CYP7A1. In the long-term experiment, liver showed a 52% decrease in CYP7a1, a 41% decrease in FAS, a 37% decrease in SREBP1c, and a 35% decrease in PGC1α. In the small intestine, the only significant change was a 46% decrease in SHP.