Introduction

Ovariectomy increases body weight and basal glucose level, and leads to IGT in mice [1, 2]. Oestrogen replacement therapy normalises these abnormalities, which implies that oestrogens play an important role in glucose metabolism [1, 2]. Several models of oestrogen insufficiency have been used to examine the mechanism(s) of these regulatory events. In male aromatase gene knockout mice, glucose intolerance and insulin resistance developed after 12 weeks of age, and were accompanied by an increase in body weight [3]. The effects of oestrogens are mediated by two receptors: ERα and ERβ. In ERα-knockout (ERKO) mice, but not ERβ-knockout (BERKO) mice, body mass, fat deposition and cholesterol levels were increased [4]. Both male and female ERKO mice had IGT, supporting the hypothesis that the glucose-lowering effect of oestrogens is mediated by ERα [5].

The mechanism(s) behind the IGT in ERKO mice is unknown. IGT and diabetes may develop due to decreased insulin sensitivity in the liver and/or in extra-hepatic tissues (muscle, adipose tissue) [6, 7] in combination with decreased insulin secretion from pancreatic beta cells [8, 9]. Therefore, in the present work we first determined glucose tolerance in ERKO and BERKO mice. Second, we studied insulin responsiveness to glucose in isolated islets in vitro. Third, we measured glucose turnover during an euglyacemic–hyperinsulinaemic clamp in female ERKO mice to estimate hepatic and peripheral insulin sensitivity. Fourth, we determined insulin-mediated glucose uptake in vitro in isolated soleus and EDL muscles. Since insulin resistance was largely localised in the liver, we analysed genome-wide expression profiles in female ERKO and wild-type mouse livers, using high-density oligonucleotide microarrays representing ∼20,000 genes to investigate the possible molecular mechanisms underlying the observed phenotypes.

Materials and methods

Animals

All animal experiments were approved by the local ethical committee. Three-month-old female and male ERKO mice, female BERKO mice and their respective controls were obtained from Taconic M&B (Ejby, Denmark). All mice used in this study were of C57BL/6J background [10, 11]. ERKO mice did not express the wild-type gene encoding ERα; however, a small splice variant of the disrupted ERα gene along with a residual level of ERα binding was present [12].

Insulin secretion in vitro

Islets were isolated from pancreas of wild-type and female ERKO mice by collagenase digestion [13]. Groups of five islets were preincubated at 37°C for 1 h in KRB, pH 7.4, containing 3.3 mmol/l glucose. The islets were then incubated in KRB for 1 h at 37°C with 3.3, 8.3 and 16.7 mmol/l glucose, or with 3.3 mmol/l glucose and 20 mmol/l arginine. After incubation, the supernatant fractions were stored at −20°C for insulin assay. Insulin was measured by RIA [14].

Euglycaemic–hyperinsulinaemic clamp

Mice were anesthetized with midazolam (0.25 mg/mouse; Dormicum, Hoffman-LaRoche, Basel, Switzerland) and a combination of fluanisone (0.5 mg/mouse) and fentanyl (0.02 mg/mouse; Hypnorm, Janssen, Beers, Belgium). Thereafter, the right jugular vein and the left carotid artery were catheterised. Thirty minutes after introduction of the catheters (t= −100 min), a bolus injection of [3-3H]glucose (3 MBq; Amersham Pharmacia Biotech, Amersham, UK) was given, followed by a continuous infusion of 0.056 MBq kg−1 min−1 [3-3H]glucose. This infusion was continued throughout the 190 min study period, at which point a steady state was reached. At t=0 min, a blood sample was taken for the determination of insulin and [3-3H]glucose concentrations, followed by an insulin infusion at the rate of 20 mU kg−1 min−1 (Actrapid, Novo Nordisk, Denmark). Blood glucose concentrations were then determined at 5 min intervals and were maintained at a concentration of 6.5 mmol/l by infusion of a solution of 2.2 mol/l glucose at a variable rate. Blood samples were taken at 60 and 90 min for the determination of insulin, and at 90 min for the determination of [3-3H]glucose. Whole-body insulin sensitivity was calculated as the 60–90 min glucose infusion rate divided by the mean of the 60 min and 90 min insulin levels. The blood samples (100 μl) taken at 0 and 90 min were deproteinised, evaporated, and resuspended in deionised water for the determination of radioactivity and glucose levels. Basal endogenous glucose production (EGP) was calculated by dividing the rate of infusion of [3-3H]glucose by the plasma glucose specific activity (i.e. dpm per min divided by dpm per mg glucose). The glucose appearance at 90 min was measured by dividing the infusion rate in dpm by the plasma glucose specific activity at this time point. EGP at this time was calculated by subtracting the glucose infusion rate from the glucose appearance rate. Finally, the glucose disposal rate was calculated as the glucose appearance rate divided by the glucose concentration.

Skeletal muscle incubation procedure and assessment of glucose transport

Female ERKO and wild-type mice were anaesthetised via i.p. injection of 2.5% avertin (0.02 ml/g body weight), and EDL and soleus muscles were removed for in vitro incubation. Isolated muscles were incubated for glucose uptake essentially as described for the rat epitrochlearis muscle [15].

Glucose transport was assessed using 2-deoxyglucose as described [16]. Muscles were transferred to vials containing glucose-free krebs–Henseleit bicarbonate buffer (KHB) containing 20 mmol/l mannitol for 10 min. Thereafter, muscles were incubated in KHB containing 1 mmol/l 2-deoxy[3H]glucose (3.7 GBq/ml) (American Radiolabeled Chemicals, St Louis, MO, USA) and 19 mmol/l [14C]-mannitol (2.59 GBq/ml) (Moravec Biochemicals, Brea, CA, USA) for 20 min and then immediately frozen in liquid nitrogen. The extracellular space and intracellular 2-deoxyglucose concentrations were determined as previously described [15, 16]. Glucose transport activity was expressed as nmol/l 2-deoxyglucose mg−1 muscle h−1.

Assays for plasma leptin, adiponectin and resistin

Plasma leptin levels were analysed by double-antibody RIA techniques using rabbit antimouse leptin antibodies, 125I-labelled leptin and mouse leptin as standard (Linco Research, St Charles, MO, USA) [17]. Plasma adiponectin was measured using RIA kit with multispecies rabbit anti-adiponectin antiserum and 125I-labelled murine adiponectin as tracer (Linco Research) [18]. Recombinant mouse adiponectin was used as standard. Free and bound radioactivity was separated by use of an anti-IgG (goat anti-guinea pig) antibody. The sensitivity of the assay was 0.8 ng/ml and the CV was <8.2% at both low and high levels. Plasma resistin levels were measured by double-antibody RIA kit (Linco Research) [19], using 125I-labeled murine resistin, a mouse resistin rabbit antiserum and mouse resistin as standard. The sensitivity was 0.78 ng/ml with an intra-assay CV of 3.6%.

RNA preparation

Frozen livers were homogenized and total RNA was purified using the TRIzol reagent (Invitrogen, Carlsbad, CA, USA), followed by RNeasy kits (Qiagen, Valencia, CA,USA). RNA quality was assayed using the Agilent 2100 Bioanalyzer (Agilent, Palo Alto, CA).

Microarray experiment

Labelled cRNA was synthesised from total RNA according to the standard Affymetrix protocol (Affymetrix, Santa Clara, CA, USA) and 15 μg of cRNA were hybridised to Mouse 430 A gene chips, washed and scanned.

Data analysis

Scanned data files were analyzed using MAS 5.0 software from Affymetrix (Affymetrix). A ‘coincidence call’ was used to identify genes with different expression levels between ERKO and wild-type mouse livers. The coincidence call was calculated from the number of changed calls comparing all ERKO mouse samples with all wild-type samples (nine comparisons in total). Genes that were consistently changed in at least five comparisons (226 increased and 350 decreased) were selected as candidate genes predicted to be differently expressed in ERKO mouse livers compared with wild-type mouse livers. Interestingly, Scd1 and Lepr genes were changed in eight and nine comparisons, respectively. The fold change reported for genes, as based on the Affymetrix analysis, is derived from average signal log ratio for all comparisons.

The over-representation analysis approach [20] was used to test sets of related genes that might be systematically altered in ERKO livers. First, 226 increased and 350 decreased genes in ERKO vs wild-type mice were selected according to the criteria above. High-Throughput GoMiner (htgm) was employed to find enrichment of changed genes involved in a particular function, using all the a priori defined gene ontology (GO) categories (http://www.geneontology.org). The original GoMiner software has been described by Zeeberg et al. [21]. Enrichment of changed genes, involved in a priori defined GO categories, is determined by a one-sided Fisher’s exact test. Significantly changed GO categories with between 10 and 250 genes are selected to exclude very small and general pathways, respectively. The p value from the one-sided Fisher’s exact test is reported, as well as the FDR test,which shows the estimated false discovery rate after multiple-comparison correction based on re-sampling technique.

Quantitative real-time RT-PCR

Total RNA (2 μg) from each individual animal were reverse-transcribed into cDNA using superscript II (Invitrogen) with random hexamer primers. The expression of Scd1 and Lepr was quantified using a SYBR green real-time PCR (RT-PCR) protocol with normalisation to 18S (Applied Biosystems, Foster City, CA, USA). The sequences of the primers employed to target the genes were are follows: Scd1: forward: GACCTGAAAGCCGAGAAGC; reverse: ATGAAGCACATCAGCAGGAGG; Lepr: forward: GAGCAGGCGTGCCATC; reverse: GTACCCGTCAGTTTCACATGATATATTG; 18S: forward: GCTTAATTTGACTCAACACGGGA; reverse: AGCTATCAATCTGTCAATCCTGTCC.

Three individual animals were analyzed in each group. Analysis of melting curves demonstrated amplification of one specific gene product for each primer pair.

Statistics

The results are expressed as means ± SEM. An unpaired Student’s t-test was used to assess differences between wild-type and ERKO mice. All statistical tests were performed with Sigma Stat for Windows Version 1.0 (Jandel Scientific Software, Erkrath, Germany). A P value of p<0.05 was considered significant.

Results

Characteristics

Female ERKO mice from our colony had 20% higher body weights than wild-type mice (28.3±0.7 vs 23.8±0.5 g, respectively, p<0.001). In contrast, male ERKO and wild-type mice had similar body weights (33.2±0.9 vs 33.1±0.7 g, respectively). Female BERKO mice had normal body weight (21.1±0.4 vs 22.1±0.4 g in controls). ERKO female and male mice had higher fasting blood glucose levels and IGT (data not shown). In contrast, fasting blood glucose and glucose tolerance were normal in BERKO mice (data not shown).

Insulin secretion in vitro

In islets from wild-type and ERKO mice, glucose (3.3, 8.8, 16.7 mmol/l) stimulated insulin release in a dose-dependent manner (Fig. 1a). Insulin release was similar in wild-type and ERKO mouse islets at all glucose concentrations. At 3.3 mmol/l glucose, in the presence of arginine (20 mmol/l), insulin release was also similar between genotypes. There was no difference in insulin secretion in the presence of glucose or arginine between BERKO and wild-type mouse islets (Fig. 1b).

Fig. 1
figure 1

Insulin secretion in isolated islets from ERKO female (open bars) (a), BERKO female (open bars) (b) and corresponding wild-type mice (black bars) at different glucose concentrations and in the presence of arginine. Data are presented as the means ± SEM of eight separate experiments

Euglycaemic–hyperinsulinaemic clamp

Whole-body insulin sensitivity and glucose turnover were determined in female ERKO and wild-type mice using the euglycaemic hyperinsulinaemic clamp technique (Table 1). In the basal state, plasma insulin levels were higher in ERKO than in wild-type mice, whereas EGP was similar between the groups, suggesting the presence of insulin resistance in ERKO mice. During the clamp, mean blood glucose and plasma insulin levels were similar between genotypes, whereas the glucose infusion rate needed to maintain euglycaemia was significantly lower in ERKO mice. Whole-body insulin sensitivity was markedly decreased in ERKO mice. Furthermore, suppression of EGP by the increased insulinaemia was impaired in ERKO mice. The peripheral glucose clearance rate was similar between the groups. Thus, in vivo ERKO mice display impaired whole-body insulin sensitivity, largely explained by an impaired suppression of EGP by insulin.

Table 1 Euglycaemic–hyperinsulinaemic clamp in female ERKO mice

Glucose uptake in soleus and EDL muscles

Basal glucose uptake was similar between wild-type and female ERKO mice. Insulin (0.18 and 12 nmol/l) increased 2-deoxyglucose uptake in soleus and EDL skeletal muscle in wild-type and ERKO mice (Fig. 2a,b). However, insulin-stimulated glucose uptake was suppressed in ERKO mice at both insulin concentrations (Fig. 2a).

Fig. 2
figure 2

Glucose uptake in skeletal muscle. a Soleus or b EDL muscles from wild-type (black bars) or female ERKO (open bars) mice were incubated for 30 min in the absence (basal) or presence of insulin (0.18 or 12 nmol/l), and 2-deoxyglucose transport was assessed as described in “Material and methods”. Values represent means ± SEM for 6–10 muscles, **p<0.01 and ***p<0.001 vs wild-type

Plasma leptin, adiponectin and resistin

Circulating leptin levels were significantly higher in female ERKO mice than wild-type mice (Table 2), whereas plasma adiponectin levels were significantly lower in ERKO mice. Resistin levels did not differ between ERKO and wild-type mice.

Table 2 Plasma leptin, adiponectin and resistin levels in female ERKO mice
Table 3 Significantly changed gene ontology (GO) categories identified by High-Throughout GoMiner

Differences in hepatic gene expression profiles between ERKO and wild-type mice

Importantly, we identified Scd1 and Lepr among a set of genes that were differently expressed. The expression level of Scd1 was increased two-fold, whilst Lepr expression was decreased eight-fold (data not shown). The differential expression of these two genes derived from the Affymetrix platform was confirmed by real-time PCR (Fig. 3).

Fig. 3
figure 3

RNA samples from individual mice prepared for gene profiling experiments were analysed by semi-quantitative real-time RT-PCR for Lepr and Scd1 expression levels (wild-type: black bar; ERKO: open bar). Each sample was analysed in triplicate. The variation within triplicate was less than 5%. mRNA levels are expressed relative to the mean wild-type values. Data are presented as means ± SEM of three individual livers. *p<0.05 and **p<0.01 vs control

To get a better understanding of overall changes in gene expression, we used an over-representation analysis procedure [20] to detect coordinate changes in the expression of groups of functionally related genes or pathways employing the web-based tool ‘htgm’. As shown in Table 3, ‘htgm’ identifies 22 GO categories enriched for increased genes in ERKO compared with wild-type mouse livers and 15 GO categories enriched for decreased genes in ERKO livers compared to wild-type mouse livers. GO categories significantly enriched for increased genes include pathways involved in steroid biosynthesis (GO:0006694), lipid biosynthesis (GO:0008610) and lipid metabolism (GO:0006629). GO categories significantly enriched for decreased genes include pathways involved in electron transport (GO:0006118), steroid metabolism (GO:0008202) and lipid transport (GO:0006869). The enrichment of increased or decreased genes was not found in the GO categories involved in glucose metabolism, such as gluconeogenesis (GO: 0006094), glucose metabolism (GO:0006006) and glucose transport (GO:0015758).

It was found that amongst the genes involved in steroid and lipid biosynthesis, the expression of Hsd3b5 was dramatically increased. Also, Elovl3 expression was significantly elevated in categories of lipid biosynthesis and metabolism. The expression of Fmo3, Cyp3a41 and Cyp2b9, which are involved in electron transport, as well as Sth2, which is involved in steroid metabolism, was significantly decreased.

The list of hepatic genes involved in the different categories is presented in Table 4.

Table 4 Genes with changed expression in ERKO compared to wild type liver in identified gene ontology (GO) categories by High-Throughout GoMiner

Discussion

ERKO mice, irrespective of gender, developed fasting hyperinsulinaemia, hyperglycaemia and IGT. In contrast, fasting blood glucose levels and glucose tolerance were normal in BERKO mice, clearly illustrating the difference in the functions of these two oestrogen receptors in this respect. ERKO mice exhibited reduced whole-body insulin sensitivity. Insulin action on EGP was markedly impaired, while glucose uptake in muscle was slightly decreased.

Hyperglycaemia in ERKO mice can be due to impaired insulin secretion from pancreatic beta cells. Thus, in islets from ovariectomized female mice insulin secretion was significantly less in response to glucose or arginine than in control mice and this effect was abolished when the animals were treated with oestradiol and progesterone [1]. We did not find any changes in either basal or glucose- or arginine-induced insulin release in isolated ERKO mouse islets. Importantly, obese mice fed a high-fat diet maintained normoglycaemia by increasing insulin levels when compared with mice fed a low-fat diet [22]. Hence, the absence of increased glucose-stimulated insulin secretion in the presence of insulin resistance might suggest an islet dysfunction in vivo in ERKO mice.

To determine whether the insulin resistance was of hepatic and/or extra-hepatic origin, we measured glucose turnover with [3-3H]glucose during basal and hyperinsulinaemic conditions. This study revealed that insulin resistance in ERKO mice could be mainly accounted for by the defective suppression of EGP. This is most likely due to insulin resistance in the liver, since other possible sources (splanchnic area, kidney) contribute only slightly to EGP [23, 24]. In agreement with this finding, oestrogens have been shown to increase hepatic insulin sensitivity by decreasing gluconeogenesis and glycogenolysis in virgin female rats and ovariectomised mice [25, 26]. Oestrogen administration decreased the expression of phosphoenolpyruvate carboxykinase, a rate-limiting enzyme in the regulation of gluconeogenesis [26]. Oestrogen treatment of ob/ob mice decreased the activity and expression of hepatic glucose-6-phosphatase, contributing to normalization of blood glucose levels [27]. Furthermore, estrogen treatment of rats significantly decreased the hepatic activity and expression of 11β-hydroxysteroid dehydrogenase type 1, an enzyme that catalyses the conversion of inactive 11-dehydrocorticosterone to corticosterone [28].

Our clamp experiments suggest that extra-hepatic insulin resistance plays a negligible role in the development of whole-body insulin resistance in ERKO mice. However, in isolated skeletal muscle studies, insulin-mediated glucose uptake was decreased in ERKO mice, supporting the notion that ERKO mice also exhibit extrahepatic insulin resistance. Since our ERKO mice demonstrate fasting hyperglycaemia and IGT, it is possible that the insulin resistance in isolated muscle may be accounted for by glucotoxicity. In earlier studies, we have demonstrated that normalisation of blood glucose levels in diabetic GK rats restores insulin sensitivity, which we interpreted as indicating that glucotoxicity accounts for insulin resistance in the muscle [29]. Since ERKO mice show slightly increased NEFA levels [4], it is possible that lipotoxicity may also play a role in this context.

In order to study the mechanisms behind the hepatic insulin resistance in ERKO mice, we analysed genome-wide expression profiles in female ERKO and wild-type mouse livers using a high-density oligonucleotide microarray technique. This study revealed that the expression of gene categories involved in lipid biosynthesis was increased, while the expression of gene categories involved in lipid transport was decreased in ERKO mice. This is an important finding, since increased levels of lipids in the liver contribute to hepatic insulin resistance [30].

In ERKO mice, we found an increased expression of Scd1, which encodes an enzyme that is required for the biosynthesis of monounsaturated fatty acids. Monounsaturated fatty acids are used as substrates for the biosynthesis of triglycerides and membrane phospholipids [31]. Changes in the ratio of saturated/unsaturated fatty acids may result in alterations in membrane lipid composition and fluidity, which can modify G-protein receptor and tyrosine kinase-linked receptor signalling [32]. Recent studies suggest that in mice, increased expression of Scd1 leads to alterations in hepatic lipid metabolism and development of the metabolic syndrome [33]. On the other hand, mice with a targeted disruption of Scd1 gene exhibit increased insulin sensitivity and fatty acid metabolism [34]. The activity and expression of Scd1 were highly elevated in ob/ob mice and were normalized by leptin treatment [35]. Also, the hepatic Scd1 mRNA level was significantly increased in leptin-resistant fa/fa ZDF rats [36]. While plasma leptin levels were significantly increased in ERKO mice, we also found a decreased hepatic expression of Lepr in ERKO mice. Similarly, ovariectomised rats demonstrated increased plasma leptin levels, as well as a decrease in the expression of leptin receptor in the hypothalamus and adipose tissues, which was normalised by oestradiol treatment [37]. Oestradiol treatment of ovariectomised mice significantly increased hepatic Lepr mRNA level, even after 2 h, suggesting a direct role of estrogens on the regulation of Lepr (Gao H, Dahlman-Wright K, Gustafson JA, unpublished results). It is therefore possible that the reciprocal expression of Lepr and Scd1 may constitute an important mechanism underlying hepatic insulin resistance in ERKO mice. Ohlsson et al. [4] demonstrated slightly increased plasma NEFA levels in ERKO mice. Therefore it is possible that the increased access of NEFA to the liver may also contribute to hepatic insulin resistance.

Our microarray study revealed significant increases in hepatic expression of the Elovl3 gene in ERKO mice. Elov13 is necessary for the synthesis of very-long-chain fatty acids. Changes in the expression of Elovl3 may lead to membrane defects, including the modification of phospholipid composition and signal transduction [38]. On the other hand, expression of Sth2 (steroid metabolism pathway) was markedly decreased. This enzyme regulates the bioavailability and activity of steroid hormones, particularly, levels of hepatic oestrogens [39].

The most dramatic changes were found in the expression of Hsd3b5 (upregulated) and Fmo3, Cyp3a41 and Cyp2b9 (down regulated). Hsd3b5 catalyses the inactivation of steroid hormones while genes involved in electron transport such as Fmo3 and cytochrome P450 family play important roles in drug metabolism and detoxification in the liver. Hsd3b5 is not expressed in control female mice [40], while Fmo3, Cyp3a41 and Cyp2b9 are only expressed in female mice [41, 42]. Therefore, the very robust changes in expression of these genes in female ERKO mice probably reflect alterations in oestrogen signalling. Importantly, the relation of the genes discussed above to hepatic glucose production is not known. In ERKO mice the expression of genes participating in regulation of gluconeogenesis was unaltered.

Mice fed a high fat-diet developed hepatic insulin resistance which was accompanied by an increase in plasma resistin levels and which was completely reversed by antisense resistin [43]. In our experiment, plasma resistin levels were normal. On the other hand, adiponectin levels were decreased in ERKO mice. It has been shown that ERKO mice had eight-fold elevated oestrogen levels compared with wild-type mice [44] and that estrogen decreases serum adiponectin levels in mice [45]. Therefore, it is possible that in our animals, oestrogens decrease adiponectin levels by acting through ERβ. Low levels of adiponectin are associated with obesity, insulin resistance and type-2 diabetes [46], while transgenic mice overexpressing adiponectin showed increased adiponectin levels and improved insulin sensitivity [47]. Adiponectin activates AMP-activated protein kinase resulting in phosphorylation of acetyl CoA carboxylase and thereby increased fatty acid oxidation and glucose uptake in the muscle and decreased hepatic gluconeogenesis [48]. Hence, decreased adiponectin levels may contribute to insulin resistance in the liver and/or muscle in ERKO mice.

In conclusion, our study demonstrates that the oestrogen-signalling pathway is an important regulatory factor in glucose homeostasis. Lack of ERα, but not ERβ, caused glucose intolerance and insulin resistance in both female and male mice. Thus, the ER subtypes have specialised roles in maintaining glucose homeostasis. ERKO mice display profound hepatic insulin resistance, concomitant with a moderate defect in glucose uptake in skeletal muscle. We propose that hepatic insulin resistance results from the upregulation of lipogenic genes via the suppression of Lepr expression.