CrbpI Modulates Glucose Homeostasis and Pancreas 9-cis-Retinoic Acid Concentrations

Abstract

Cellular retinol-binding protein type I (CrbpI), encoded by Rpb1, serves as a chaperone of retinol homeostasis, but its physiological effects remain incompletely understood. We show here that the Rbp1^−/− mouse has disrupted retinoid homeostasis in multiple tissues, with abnormally high 9-cis-retinoic acid (9cRA), a pancreas autacoid that attenuates glucose-stimulated insulin secretion. The Rbp1^−/− pancreas has increased retinol and intense ectopic expression of Rpb2 mRNA, which encodes CrbpII: both would contribute to increased β-cell 9cRA biosynthesis. 9cRA in Rbp1^−/− pancreas resists postprandial and glucose-induced decreases. Rbp1^−/− mice have defective islet expression of genes involved in glucose sensing and insulin secretion, as well as islet α-cell infiltration, which contribute to reduced glucose-stimulated insulin secretion, high glucagon secretion, an abnormally high rate of gluconeogenesis, and hyperglycemia. A diet rich in vitamin A (as in a standard chow diet) increases pancreas 9cRA and impairs glucose tolerance. Crbp1 attenuates the negative impact of vitamin A (retinol) on glucose tolerance, regardless of the dietary retinol content. Rbp1^−/− mice have an increased rate of fatty acid oxidation and resist obesity when fed a high-fat diet. Thus, glucose homeostasis and energy metabolism rely on Rbp1 expression and its moderation of pancreas retinol and of the autacoid 9cRA.

INTRODUCTION

Specific binding-proteins influence metabolic flux of retinoids and their physiological functions (35, 37). Diverse cell types express cellular retinol-binding protein I (CrbpI), a member of the fatty acid binding-protein gene family encoded by Rbp1, which sequesters retinol (vitamin A) and retinal with K_d values in the low nanomolar range (27, 36). The relative amounts of apo- and holo-CrbpI facilitate cellular retinol uptake, modulate retinol storage as retinyl esters (RE), and affect retinoid homeostasis by regulating enzyme activity differentially (25, 34). Retinoic acid receptor (RAR) activation depends on CrbpI-mediated retinol uptake. CrbpI mutants with low retinol-binding affinity reduce RAR function in human mammary epithelial cells, leading to a loss of differentiation (10). Nevertheless, Rbp1^−/− mice are seemingly healthy and show no gross abnormalities characteristic of overt retinoid deficiency (13). All-trans-retinoic acid (atRA) in serum and several target tissues does not differ significantly between Rbp1^−/− and wild-type (WT) mice, likely accounting for lack of gross abnormalities (20, 29). Despite these insights, physiological effects of CrbpI remain to be elucidated fully.

Retinol functions primarily through its metabolite atRA, which has diverse effects on energy metabolism. atRA induces pancreas development and differentiation into acini (18, 24, 28, 32), but restricting dietary vitamin A in diabetes-prone rats reduces diabetes and insulitis, an effect not reversed by dosing atRA, suggesting contributions of additional retinoids (9). atRA arrests differentiation of preadipocytes into mature white adipocytes, early in the differentiation process (43, 48). Ablation of Rdh1, which encodes one of several short-chain dehydrogenases that catalyze the first and rate-limiting step in atRA biosynthesis from retinol, increases adiposity in mice fed a nonobesogenic diet (50). Dosing atRA ameliorates weight gain and represses insulin resistance in mice fed an obesogenic (high-fat) diet (3). atRA induces Ucp1, which encodes an uncoupling protein expressed in brown adipose tissue that supports thermogenesis (31).

Although atRA mediates most vitamin A functions, 9-cis-RA (9cRA) occurs as an endogenous pancreas retinoid with unique actions independent of atRA (22). 9cRA, biosynthesized primarily by β-cells and detected only in pancreas, varies rapidly and inversely with serum glucose. 9cRA attenuates glucose-stimulated insulin secretion (GSIS): a decrease in 9cRA allows optimum glucose-stimulated insulin secretion, whereas dosing with 9cRA impairs insulin secretion and glucose tolerance. 9cRA acts, in part, through reduction of Glut2 and GK activities and through repressing expression of genes that function to promote insulin biosynthesis, including HNF4a and Pdx1.

We hypothesized that loss of Rbp1 would alter retinoid homeostasis and retinoid-governed energy balance. We found that Rbp1^−/− mice have robust pancreas expression of Rbp2 (encodes CrbpII), a gene normally expressed intensely only in intestinal mucosa, and elevated pancreas 9cRA, decreased pancreas expression of Pdx1, Glut2, and GK, and decreased insulin secretion but enhanced glucagon secretion after feeding. Rbp1^−/− mice are hyperglycemic, rely on increased fatty acid oxidation, and resist diet-induced obesity. These data substantiate a fundamental contribution of CrbpI to retinoid function involving pancreas 9cRA, glucose homeostasis, and whole-body energy metabolism.

MATERIALS AND METHODS

Mice.

Male C57BL/6 mice were used, unless noted otherwise, in accordance with institutional guidelines. Mice were fed ad libitum or fasted for 12 to 16 h. Seven- to twelve-week-old WT mice were purchased from the Jackson Laboratories. Seven- to twelve-week-old Rbp1^−/− mice were bred in-house from breeders obtained from Pierre Chambon and Norbert Ghyselinck. Mice were fed either a standard chow diet (Harlan Teklad 18% protein rodent diet #2018S with 30 IU of vitamin A/g and 18% of calories as fat) or an equivalent AIN93G diet with 30 IU of vitamin A/g, unless noted otherwise. We refer to diets with 30 IU of vitamin A/g as copious in vitamin A (CVA). The American Institute of Nutrition recommends that rodent diets contain 4 IU of vitamin A/g (41). Alternatively, mice were fed an CVA diet until weaning and then were fed a vitamin A-deficient (VAD) diet for 3 months. A VAD diet does not render a mouse vitamin A deficient, which was weaned from a dam fed a CVA diet, but reduces vitamin A stores and atRA levels (50). To establish diet-induced obesity (DIO), 1-month-old mice were fed a high-fat diet (HFD; 50% of calories as fat) for 5 months, with 30 IU of vitamin A/g. For tolerance tests, glucose (2 g of glucose/kg), insulin (0.5 U/kg), or sodium pyruvate (2 g/kg) was administered intraperitoneally in saline. Insulin immunofluorescence testing was performed as described previously (22). Glucagon was imaged with an anti-glucagon antibody (1:200; Dako) and an Alexa 555 secondary antibody (Invitrogen).

Retinoids and their quantification.

Retinoids were purchased from Sigma, except for 9-cis-retinol, which was synthesized (17). Retinoids were handled under yellow light. Tissue samples were harvested under yellow lights, frozen immediately in liquid nitrogen, and kept at −80°C until extraction and retinoid quantification within 1 day by liquid chromatography-tandem mass spectrometry (LC-MS/MS) for RA isomers or by high-pressure liquid chromatography (HPLC)/UV for neutral retinoids (20, 21, 23).

Serum factors.

Nonesterified fatty acids, glycerol, triglycerides, and ketone bodies were measured according to manufacturer's instructions (Waco). The percent Hb1Ac (Hb1Ac%) was measured with an A1c Now+ Multi-Test kit system (Bayer). Insulin (Crystal Chem), glucagon (Alpco Diagnostic), and growth hormone (rat/mouse growth hormone ELISA kit, Millipore catalog no. EZRNGH-45K) were measured by enzyme-linked immunosorbent assay.

Assay of metabolic values.

Male mice, 16 to 19 weeks old (WT, n = 8; Rbp1^−/−, n = 7), were housed individually in metabolic cages (Comprehensive Lab Animal Monitoring System; Columbus Instruments, Columbus, OH) for 25 h and fed ad libitum. Data collection began after 1 h to allow acclimation to the chambers. Metabolic values were recorded using packaged software (Oxymax) with an airflow rate of 0.5 liter/min. Individual cages were monitored 2 of every 10 min. The RER was calculated for each interval using the ratio CO₂ expiration to O₂ consumption. Oxygen consumption and CO₂ levels were normalized to an “effective” body weight of kg^0.75 by the software. During metabolic studies, mice were maintained at standard housing temperature and light-dark cycle. To evaluate food consumption, pelleted diet was ground by using a food processor for use in metabolic cages.

Assays with 832/13 β cells.

The pancreatic β-cell line 832/13 was cultured in RPMI 1640 with 10% fetal bovine serum, 10 mM HEPES, 2 mM l-glutamine, 1 mM sodium pyruvate, 50 μM β-mercaptoethanol, 100 U of penicillin/ml, and 100 mg of streptomycin/ml at 37°C and 5% CO₂ in 100-mm petri dishes as described previously (16). Medium was refreshed every 2 to 3 days. Cells were subcultured when they approached ≥70% confluence. Protein was quantified with the Bradford assay.

9cRA biosynthesis.

832/13 cells were cultured in 12- or 6-well plates. At confluence, the medium was replaced with serum-free medium, and retinoids (1 μM) in dimethyl sulfoxide (DMSO; 0.1% [vol/vol]) or vehicle alone were added and incubated for 1 h. Cells were lysed by using Reporter lysis buffer (Promega) and combined with their medium for retinoid quantification.

Insulin secretion.

Islets were isolated by the UCSF Diabetes and Endocrinology Research Center (San Francisco, CA). Fifteen islets of similar size were incubated at 37°C in 12-well plates with 5 mM d-glucose in RPMI 1640 supplemented with 10% fetal bovine serum, 100 U of penicillin/ml, 100 μg of streptomycin/ml, 10 mM HEPES, 2 mM l-glutamine, 1 mM sodium pyruvate, 50 μM β-mercaptoethanol, and 5% CO₂. After 16 h, islets were washed and transferred to 1.5-ml tubes in secretion medium (HBSS with 20 mM HEPES and 1% bovine serum albumin [pH 7.2]) containing 3 mM glucose. After 2 h, the secretion medium was replaced with fresh secretion medium containing 3 or 23 mM glucose with 50 or 100 nM 9cRA or vehicle alone (DMSO). Media were centrifuged 2 min at 1,000 × g to remove nonadherent cells and stored at −80°C until assay. A similar procedure was used for 832/13 cells.

mRNA expression.

Total RNA was isolated with TRIzol (Invitrogen, Carlsbad, CA). RNA was reversed transcribed with an IScript cDNA synthesis kit (Bio-Rad). TaqMan quantitative real-time PCR (qT-PCR) was done using predesigned and optimized primers (Applied Biosystems, Foster City, CA). Gene expression was measured with ABI Prism 7500 real-time PCR systems sequence detection (Applied Biosystems).

Statistics.

Data are means ± the standard errors (SE). Statistical significance was assessed by using two-tailed, unpaired Student t tests to compare two groups and by two-way analysis of variance (ANOVA) to compare two curves.

RESULTS

The Rbp1^−/− mouse has elevated pancreas 9cRA, which resists feeding-induced decreases.

We quantified multiple retinoids in serum and tissues to determine the extent of disrupted vitamin A homeostasis in Rbp1^−/− mice and observed a major effect in pancreas. Relative to WT mice, Rbp1^−/− mice had elevated pancreas 9cRA—23% during fasting and 63% in the fed state—and 9cRA did not decrease significantly after the fasted to fed transition (Fig. 1). In contrast, pancreas atRA and 9,13-di-cis-RA (9,13dcRA), an RA isomer without known biological activity, did not change after the fasting to fed transition in either genotype, and their concentrations did not differ between the two genotypes fed a diet copious in vitamin A (CVA), i.e., 30 IU of vitamin A/g. Pancreas retinal, an intermediate in the path from retinol to RA, and RE also did not change with the fed versus fasted states or between the two genotypes (Fig. 1C). Pancreas retinol did not change in either genotype after the fasted-to-fed transition but was ∼40% higher in Rbp1^−/− mice relative to the WT.

Fig. 1.

Increased 9cRA levels in the pancreases of Rbp1^−/− mice. (A) Quantification of pancreas RA isomers in fed (CVA) or 12-h-fasted mice. Values are means from two to three experiments with 8 to 10 mice/group/experiment. *, P < 0.006 versus WT mice; +, P < 0.0006 versus fasted mice. The fasted and fed Rbp1^−/− 9cRA values did not differ significantly. (B) Representative LC-MS/MS chromatograms of RA isomers in mouse pancreas (left) and liver (right). (C) Pancreas retinoids in the same mice as in panel A. Values are means of one to four experiments with 4 to 11 mice/group/experiment. *, P < 0.003 versus WT mice. (D) Retinoids in liver, adipose, and serum. *, P < 0.003 versus WT mice, 7 to 15 mice/group. Mice were fed the CVA diet ad libitum.

9cRA was not observed above the limit of detection of the LC-MS/MS assay in biological matrices (∼0.05 pmol/g of tissue) in liver, epididymal adipose, serum, or several tissues other than pancreas (Fig. 1B) (20, 23). The atRA and 9,13dcRA concentrations were similar in the liver and adipose tissues of WT and Rbp1^−/− mice, with atRA being ∼2.4-fold higher in the sera of Rbp1^−/− mice (Fig. 1D). Rbp1 ablation did not affect the retinal concentration in the liver, adipose tissue, and serum. Consistent with previous observations, retinol and RE were low (48% each) in the livers of Rbp1^−/− mice relative to WT mice (13). In contrast, adipose and serum retinol increased significantly (3-fold and 40%, respectively) in Rbp1^−/− mice relative to WT mice, whereas RE increased only in the adipose tissues of Rbp1^−/− mice (62%).

Decreasing the amount of dietary vitamin A by feeding a VAD diet to the animals for 3 months, which reduces retinoid reserves but does not render mice vitamin A deficient, decreased pancreas 9cRA, atRA, 9,13dcRA, and retinol, but not retinal, levels in both WT and Rbp1^−/− mice (Fig. 2). 9cRA, however, remained higher in the pancreases of Rbp1^−/− mice relative to WT mice, regardless of the amount of dietary vitamin A. A decrease in dietary vitamin A decreased the pancreas RE in WT mice, but not in Rbp1^−/− mice. Although the levels of pancreas retinol were similar in WT and Rbp1^−/− mice fed the VAD diet, the levels of pancreas retinol were ∼2.5-fold higher in Rbp1^−/− mice fed the CVA diet relative to WT mice. Liver retinoid stores (RE and retinol) directly reflected the amount of dietary vitamin A in both WT and Rbp1^−/− mice, with the latter having markedly lower concentrations of each.

Fig. 2.

Effects of dietary vitamin A (retinol) content on pancreas and liver retinoid concentrations. Mice were fed the CVA or the VAD for 3 months. (A) Pancreas retinoids. Values are means of one to four experiments with 4 to 10 mice/point/experiment. *, P = 0.01 versus WT mice; **, P < 0.0001 versus CVA-fed mice, two-way ANOVA; #, P < 0.013 versus WT mice; &, P < 0.02 versus CVA-fed mice; ##, P < 0.002 versus diet; ***, P < 0.009, WT mice versus Rbp1^−/− mice. (B) Liver retinol and RE, seven to nine mice/group. #, P < 0.018 versus WT mice, ***, P < 0.0001 versus diet.

Increased pancreas 9cRA biosynthesis in Rbp1^−/− mice.

We evaluated whether Rbp1^−/− mice demonstrated enhanced 9cRA biosynthesis by quantifying the concentrations of substrates and synthesis rates. Cytosol and microsomes isolated from Rbp1^−/− mouse pancreases had 53 and 16% more 9-cis-retinol than those from WT mice, respectively (Fig. 3 A). The levels of all-trans-retinol were 79 and 22% higher in Rbp1^−/− mouse cytosol and microsomes, respectively, relative to WT mice (Fig. 3B). The level of endogenous 9cRA was 74% greater in Rbp1^−/− pancreas cytosol, whereas the level of endogenous atRA in cytosol did not increase significantly. Thus, both substrates for 9cRA biosynthesis and endogenous 9cRA are increased in the Rbp1^−/− mouse pancreas.

Fig. 3.

Increased pancreas 9cRA biosynthesis in Rbp1^−/− mice. (A) On the left, HPLC/UV quantification of 9-cis-retinol in pancreas microsomes (Mcr) and cytosol (Cyt) was performed. On the right, LC-MS/MS quantification of 9cRA in pancreas cytosol was performed. *, P < 0.05 versus the WT value, three replicates/group. (B) All-trans-retinoids in pancreas subcellular fractions. On the left, HPLC/UV quantification of all-trans-retinol in pancreas microsomes (Mcr) and cytosol (Cyt) was performed. On the right, LC-MS/MS quantification of atRA in pancreas cytosol was performed. *, P < 0.05 versus WT value, three replicates per group. (C) Relative mRNA expression of retinoid binding proteins in islets isolated from WT or Rbp1^−/− mice. *, P < 0.02 versus WT mice, eight mice/group. n.d., not detected. Rbp2 encodes CrbpII, Rbp7 encodes CrbpIII, Rbp4 encodes the serum retinol binding protein, and Crabp2 encodes CrabpII. (D) Biosynthesis of RA isomers from retinol and retinal isomers bound with CrbpI or CrbpII by a combination of pancreas microsomes and cytosol. *, P = 0.0003 versus the CrbpI value, three replicates/substrate.

Rbp1^−/− homologs may compensate for the loss of CrbpI (13, 38). Therefore, we surveyed mRNA expression of retinoid binding-proteins. The level of Rbp2 mRNA (encodes CrbpII) was ∼80-fold higher in the pancreases of Rbp1^−/− mice relative to WT mice, such that its expression was similar to the Rbp1 expression in WT mice (Fig. 3C). The levels of Rbp7, which encodes CrbpIII, and of Rbp4, which encodes the serum retinol binding protein, decreased slightly. We also observed an ∼2- to 3-fold increase in Crabp2 mRNA, which encodes CrabpII, a chaperone that delivers atRA to RA receptors but also binds 9cRA with high affinity (4, 11).

CrbpII has high affinity for 9-cis-retinol, albeit ∼7-fold lower than CrbpI (K_d values of 68 and 11 nM, respectively) (19). To determine whether the nature of the Crbp expressed affects the rate of RA biosynthesis, we assessed RA biosynthesis from all-trans-retinol, 9-cis-retinol, all-trans-retinal, or 9-cis-retinal bound to either CrbpI or CrbpII (Fig. 3D). 9-cis-Retinol bound to CrbpII generated 9cRA ∼2.5-fold faster versus CrbpI, but the rate of metabolism of the retinal isomers into RAs did not depend on the specific Crbp.

Hyperglycemia and resistance to glucose-mediated 9cRA decreases in Rbp1^−/− mice.

Increased pancreas 9cRA, which did not decrease significantly after feeding, represented the most profound difference in tissue retinoids of Rbp1^−/− mice. We therefore assayed blood glucose because 9cRA functions as a pancreas autacoid that attenuates GSIS (22). The blood glucose levels in fasted Rbp1^−/− mice exceeded those in WT mice by 68%, but the fasted serum insulin values remained the same in both genotypes (Fig. 4A). The high fasting blood glucose levels in Rbp1^−/− mice did not increase significantly after feeding but were 26% higher than in CVA-fed WT mice. Serum insulin in Rbp1^−/− mice did not respond as robustly to feeding as in WT mice, reaching a level 32% lower than that of WT mice.

Fig. 4.

Effects of Rbp1 and dietary vitamin A content on glucose homeostasis. (A) Blood glucose and serum insulin values for fasted and fed mice. Values are means of five to six experiments (5 to 10 mice/group). *, P < 0.007; **, P < 0.033 versus WT values. (B) Pancreas retinoids after a 2-g/kg glucose challenge. Only 9cRA showed significant differences. Values are means of two experiments with 4 to 10 mice per point per experiment. *, P < 0.01 from zero time; **, P < 0.03 from 30 min; +, P < 0.001 versus WT mice; ***, P < 0.0001 versus WT mice, two-way ANOVA. (C) Serum insulin after the glucose challenge described in panel B. *, P < 0.01 versus WT mice. (D) Blood glucose after the glucose challenge described in panel B (seven mice/group). *, P < 0.0001, Rbp1^−/− versus WT mice, both diets (two-way ANOVA). Mice were fed the CVA or VAD diet for 3 months. (E) Lack of Rbp1 reduces Pdx-1, Glut2, and GK mRNA in isolated islets, eight replicates/group (qT-PCR). *, P < 0.001. The data were normalized to islet numbers. (F) Blood glucose after an insulin challenge. Values are means of two experiments (five mice/point).

Consistent with the data in Fig. 1A, fasted Rbp1^−/− mice had higher 9cRA levels than WT (Fig. 4B). Dosing with 2 g of glucose/kg did not decrease the amount of 9cRA in Rbp1^−/− mice after 30 min versus a corresponding 35% 9cRA decrease in WT mice. By 60 min, the levels of 9cRA in Rbp1^−/− mice decreased 28%, whereas in WT mice the 9cRA levels returned to initial values, but the decrease in Rbp1^−/− mice did not reduce 9cRA levels to the levels observed in WT mice. In contrast to 9cRA, the levels of atRA and 9,13dcRA did not change after a glucose dose in either the WT or the Rbp1^−/− mouse pancreas. Blood glucose changes in both WT and Rbp1^−/− mice did not affect retinol, retinal, or RE. The glucose challenge did not increase serum insulin levels above the baseline in Rbp1^−/− mice by 30 min after the dose; by 60 min after the dose the glucose had increased the insulin levels only modestly in Rbp1^−/− mice compared to WT mice (Fig. 4C). Consistent with their increased pancreas 9cRA level and lower serum insulin level after glucose dosing, the glucose tolerance test (GTT) revealed impaired glucose tolerance in Rbp1^−/− mice (Fig. 4D). Blood glucose in Rbp1^−/− mice persisted at abnormally high levels 270 min after glucose dosing, rivaling the values in WT mice at the serum glucose peak 30 min after dosing.

To determine whether vitamin A status affects glucose disposal, we performed a GTT with mice fed a VAD diet for 3 months, which decreases tissue retinoids but does not render mice vitamin A deficient. A VAD diet attenuated the peak concentration of blood glucose to no more than half that of Rbp1^−/− mice fed a CVA diet and hastened the clearance rate (Fig. 4D). The blood glucose levels of Rbp1^−/− mice fed a VAD decreased 50% from peak values 150 min sooner than those of Rbp1^−/− mice fed a CVA diet. WT mice also showed improved glucose tolerance when fed a VAD diet, with blood glucose values returning to baseline levels 30 min sooner than WT mice fed a CVA.

The transcription factor pancreatic and duodenal homeobox 1 (Pdx-1) stimulates pancreas glucose transporter (Glut2), glucokinase (GK), and insulin gene expression, and insulin release relies on the actions of Glut2 and GK (1, 33). Pancreas islets of Rbp1^−/− mice had Pdx-1, Glut2, and GK mRNA expression decreased 26 to 53% relative to WT mice (Fig. 4E). Dietary vitamin A content did not affect insulin sensitivity in either genotype, and Rbp1^−/− mice were not insulin resistant, as demonstrated by an insulin tolerance test (Fig. 4F).

9cRA disrupts glucose sensing and impairs insulin secretion.

To duplicate chronic elevation of 9cRA in the Rbp1^−/− mouse pancreas, we treated WT pancreas islets and the β-cell line 832/13 with 9cRA (Fig. 5). Treating 832/13 cells with 9cRA caused 9- to 15-fold decreases in Pdx-1, Glut-2, and GK mRNA expression relative to the vehicle control within 2 h and decreased GSIS in 832/13 β-cells and in isolated mouse islets.

Fig. 5.

9cRA disrupts glucose sensing and impairs insulin secretion. (A) Pdx-1, Glut2, and GK mRNA relative to vehicle controls in 832/13 β cells after 2 h of treatment with 100 nM 9cRA and 23 mM glucose. *, P < 0.04, four to six plates. (B) GSIS of WT pancreas islets or 832/13 β cells after 20 h of treatment with 9cRA or vehicle. The medium in the first 16 h had 5 mM glucose, followed by a medium change to 3 mM glucose for 2 h, followed by a medium change to 23 mM glucose for 2 h. Insulin was measured in the medium at the end of the final 2 h of incubation. Values are normalized to islet/cell numbers. *, P < 0.0001 versus control, nine replicates/group.

α-Cell infiltration, increased glucagon secretion, and enhanced gluconeogenesis in Rbp1^−/− mice.

To examine for abnormalities, we visualized islets by confocal immunocytochemical staining for glucagon (α-cell marker) and insulin (β-cell marker). Glucagon staining showed α-cell infiltration into islets, characteristic of glucagon hypersecretion (Fig. 6A) (12, 45). In fact, CVA-fed Rbp1^−/− mice had >3-fold higher serum glucagon levels than WT mice, and an insulin/glucagon ratio ∼3-fold lower than WT mice (Fig. 6B). We performed a pyruvate tolerance test (PTT) for liver gluconeogenesis, which revealed that both fasted and fed Rbp1^−/− mice had liver gluconeogenesis rates ∼40 to 50% (area under the concentration-time curve) greater than those of WT mice (Fig. 6C). In the liver there was also a 5- to 6-fold decrease in glucose-6-phosphate dehydrogenase mRNA expression in both fasted and fed Rbp1^−/− mice, relative to WT mice, a finding consistent with decreased glucose metabolism that accompanies gluconeogenesis (Fig. 6D). Circulating Hb1Ac, an indicator of prolonged hyperglycemia, was elevated 8% in Rbp1^−/− mice compared to WT (Fig. 6E).

Fig. 6.

α-Cell infiltration and increased glucagon secretion in Rbp1^−/− mice. (A) Immunohistochemistry analysis showing islet β cells (insulin staining, green) and α-cells (glucagon staining, red). The arrows indicate the α-cell infiltration into the center of the islet from Rbp1^−/− mice. (B) Serum glucagon and the insulin/glucagon ratio of fasted and fed mice (eight mice/group). *, P < 0.003 versus WT mice. (C) Blood glucose after a pyruvate challenge (five mice/group). *, P < 0.0001 versus WT mice (two-way ANOVA, genotype, both fasted and fed). (D) Liver G6PDH mRNA in fasted and fed Rbp1^−/− mice relative to WT mice (five to nine replicates/group). *, P = 0.033 (two-way ANOVA, genotype). (E) Circulating Hb1Ac% levels (8 to 10 mice/group). *, P = 0.01 versus WT mice.

Increased reliance on fat oxidation in Rbp1^−/− mice.

Glucagon stimulates fatty acid oxidation in liver (30). We performed indirect calorimetry to determine whether abnormally high serum glucagon levels in Rbp1^−/− mice affected fuel use and found that Rbp1^−/− mice have a depressed respiratory exchange ratio compared to WT mice, indicating increased fatty acid oxidation (Fig. 7A) (40). Consistently, Rbp1^−/− mice have a 38% increase in fat oxidation compared to WT mice (Fig. 7B). Rbp1^−/− mice also were 35% less active than WT mice and drank less water but ate a comparable amount of food and weighed the same as WT mice (Fig. 7C to E).

Fig. 7.

Increased fat oxidation in Rbp1^−/− mice. (A) RER per hour and as the 24-h average. *, P < 0.02 versus WT mice. (B) Fat oxidation per hour and as the 24-h average. *, P < 0.02 versus WT mice. (C) Total activity, ambulatory activity, and rearing per hour and as 24-h averages. *, P < 0.03 versus WT values. (D) Food intake and water consumption per 24 h. *, P < 0.02 versus WT mice. (E) Body weights for 4-month-old WT and Rbp1^−/− mice. For panels A to D, there were seven to eight mice/genotype; for panel E (representative data), there were five mice/genotype.

Rbp1^−/− mice resist diet-induced obesity.

We tested the effects of a high-fat diet (HFD) because Rbp1^−/− mice demonstrated increased fatty acid oxidation and decreased glucose use. Rbp1^−/− mice weighed an average 11.8 g less than WT fed an HFD for 5 months, even though weights of the genotypes did not differ with mice fed a nonobesogenic diet (CVA) (Fig. 8A). Glucose intolerance did not worsen in Rbp1^−/− mice fed the HFD relative to the nonobesogenic diet (Fig. 8B). With both diets, glucose intolerance in Rbp1^−/− mice was intermediate between results obtained with WT mice fed the nonobesogenic diet versus the HFD. The levels of 9cRA in the pancreases of Rbp1^−/− mice fed the HFD did not increase relative to the already elevated levels of Rbp1^−/− mice fed the nonobesogenic diet (Fig. 8C). The levels of 9cRA in the pancreases of WT mice fed the HFD increased ∼2-fold compared to WT mice fed a nonobesogenic diet.

Fig. 8.

Rbp1^−/− mice resist DIO. (A) Weights of mice fed the CVA diet for 6 months (18% fat, 30 IU of vitamin A/g) or an obesogenic diet (HFD, 50% fat, 30 IU of vitamin A/g) for 5 months to develop DIO, eight mice/group. *, P < 0.001 versus CVA diet; +, P = 0.003 versus WT DIO. (B) Blood glucose after a glucose challenge, 8 to 10 mice/group. *, P < 0.007 (two-way ANOVA, CVA, Rbp1^−/− mice versus WT mice); +, P < 0.0001 (two-way ANOVA, DIO, WT mice versus Rbp1^−/− mice). (C) Pancreas 9cRA in fed mice, eight mice/group. *, P < 0.003 versus WT control; +, P < 0.03 versus WT control.

DISCUSSION

We undertook this study to determine the impact of CrbpI on retinoid homeostasis in multiple adult tissues and its effects on retinoid function. Previous studies of Rbp1^−/− mice reported low liver RE, slightly decreased retinol, RE, and atRA during embryo days 10.5 through 14.5, and normal atRA in adult liver, kidney, testis, brain, and serum (13, 20, 29). We extended these findings by assaying multiple retinoids in adult mouse liver, adipose tissue, serum, and pancreas. We found elevated retinol levels in serum and in two retinoid target tissues that affect energy balance, adipose and pancreas (CVA diet), and increased pancreas 9cRA. Increased 9cRA levels seemed to be the most significant outcome observed, because 9cRA was the only known activated retinoid that differed from WT mice, and 9cRA functions as a pancreas autacoid that prevents hypoglycemia (22).

Both an increase in the RA-precursor retinol and ectopic expression of Rbp2 could prompt the increase in pancreas 9cRA in the Rbp1^−/− mouse. The increased pancreas retinol would increase the rate of 9cRA production because substrate concentration contributes to RA biosynthesis. In the absence of Rbp1, Rbp2 expression increases from <1% to ∼100% versus WT Rbp1 mRNA. This robust Rbp2 expression would contribute to increased 9cRA because CrbpII allows a faster rate of retinol dehydrogenation than CrbpI. Moreover, increased Crabp2 expression would increase the impact of increased 9cRA, since CrabpII binds 9cRA with high affinity and delivers RA to RAR and 9cRA activates RAR, as well as RXR (4, 11, 37, 42).

9cRA impairs GSIS in isolated islets, in the rat insulinoma β-cell line 832/13, and in the mouse during a GTT, whereas a glucose-prompted decrease in 9cRA fosters optimum insulin secretion (22). These actions explain hyperglycemia in the Rbp1^−/− mouse, characterized by high levels of 9cRA, lower fed serum insulin levels, a less sensitive and delayed response of 9cRA to increases in blood glucose, and lower serum insulin levels during a GTT. 9cRA impairs GSIS by decreasing Glut2 and GK activities within minutes, in the absence of short-term (15-min) changes in their mRNAs. We extended these data here by providing additional dose-response data for 9cRA in both islets and β cells and showing that 9cRA also has a longer-term effect on islet function through repressing Pdx-1, Glut2, and GK mRNA in β cells. Pdx-1 stimulates insulin biosynthesis. Glut2 and GK catalyze the uptake and rate-limiting steps, respectively, in glucose metabolism, which stimulate GSIS (47). Defects in GK and Pdx-1 cause two of the diseases known as maturity onset diabetes of the young, MODY 2 and 4, respectively (15).

Decreased insulin secretion from β cells contributes to enhanced glucagon secretion (2). Enhanced glucagon secretion during subnormal insulin secretion results in hyperglycemia (39, 44). The data support the hypothesis that relatively high 9cRA in the Rbp1^−/− mouse in the fed state attenuates insulin secretion, which raises blood glucose, and elevates glucagon secretion. Elevated glucagon secretion drives liver gluconeogenesis, as shown by the PTT, which also contributes to hyperglycemia (14, 26, 30). In addition, elevated glucagon promotes hepatic fatty acid oxidation. Thus, the Rbp1^−/− mouse suffers hyperglycemia and disproportionately relies on fatty acids, rather than glucose, as fuel. Possibly, 9cRA also may directly affect α cells, since not all 9cRA in the pancreas occurs in β cells (22). These data provide further support for a hypothesis that 9cRA attenuates hypoglycemia by reducing β-cell insulin secretion during low serum glucose, which enhances α-cell glucagon secretion, stimulating hepatic gluconeogenesis, glycogenolysis, and fatty acid oxidation.

Serum Hb1Ac levels are used to gauge the degree of prolonged hyperglycemia but were elevated only modestly in the Rbp1^−/− mouse. The level of Hb1Ac, however, does not serve as a reliable indicator of long-term hyperglycemia in all situations (8, 46). Reduced activity of glucose-6-phosphate dehydrogenase (G6PDH), the rate-limiting enzyme in the pentose phosphate pathway, produces lower Hb1Ac levels than expected during hyperglycemia through reducing glucose metabolism (5). The 5-fold-reduced G6PDH mRNA in liver of Rbp1^−/− is consistent with the reduced G6PDH activity that occurs with moderate to severe hyperglycemia in rodents (7).

Although Rbp1^−/− mice suffer glucose intolerance in the absence of obesity, they have 9cRA levels comparable to mouse models of glucose intolerance secondary to obesity: ob/ob; db/db; DIO (22). Thus, control of pancreas retinoid metabolism (via CrbpI) and increased adiposity independently influence pancreas 9cRA concentrations. The mechanism whereby adiposity increases 9cRA has not been established, but β-cell hypertrophy seems a likely contributor.

Increased use of fat as fuel (decreased RER and increased fatty acid oxidation) would contribute to protecting Rbp1^−/− mice from DIO. A reduction in RER provides a key component in protection against DIO, insulin resistance with DIO, and elevated plasma nonesterified fatty acids (6, 49). Recently, Zizola et al. also observed that Rbp1 ablation protects from adiposity, when mice are fed an HFD: these researchers attributed this to an decrease in preadipocyte differentiation into mature white adipose (51). The report did not examine CrbpI function in mice fed a nonobesogenic diet but noted lower serum insulin in Rbp1^−/− mice fed an HFD, as we report here for mice fed a nonobesogenic diet. The mechanism of CrbpI action on preadipocyte differentiation remains unresolved.

In summary, we show that copious vitamin A (as in a standard lab chow diet) impairs glucose tolerance, and CrbpI attenuates the negative impact of copious vitamin A on glucose tolerance by controlling pancreas 9cRA. These data are the first to connect CrbpI with controlling concentrations of an activated retinoid and thereby impacting the physiological function of a retinoid. Interestingly, normal endocrine pancreas function, glucose homeostasis, and energy metabolism rely on CrbpI to modulate pancreas retinol and 9cRA. Enhanced fatty acid oxidation occurs as a consequence of CrbpI absence, which would contribute to resisting weight gain in animals fed an HFD. These data offer insight into 9cRA function as a pancreas autacoid and provide elucidate the effects of CrbpI and retinoids in the maintenance of energy balance.