Liver-Restricted Repin1 Deficiency Improves Whole-Body Insulin Sensitivity, Alters Lipid Metabolism, and Causes Secondary Changes in Adipose Tissue in Mice

Previously, we identified a quantitative trait locus (QTL) for body weight, serum fasting insulin, and triglycerides (TGs) on rat chromosome 4 (1–3). Replication initiator 1 (Repin1) emerged as a potential positional candidate gene within the QTL region considering associations of metabolic alterations in rats with a single nucleotide polymorphism (449C/T) in the Repin1 coding region and with the size of a triplet repeat in the 3′-untranslated region of the Repin1 gene (4). Repin1 was initially discovered as replication initiation region protein 60 kDa (RIP60) in a study investigating DNA binding proteins involved in replication activation of the Chinese hamster dihydrofolate reductase gene (dhfr) (5). Repin1 binds to two ATT-rich sites in oriβ, a short region 3′ to the dhfr gene, acting as an enhancer of DNA bending during initiation of DNA synthesis (6,7). Plasmid replication assays demonstrated only weak replication enhancing activity, and thus Repin1 may act as an accessory factor in origin recognition prior to the assembly of preinitiation complexes (8). After it was first cloned in 2000, characterization of DNA binding and bending properties revealed the first structural insight into Repin1/RIP60 function as a polydactyl zinc finger protein of the Cys₂-His₂ type (8).

Repin1 is ubiquitously expressed with the highest expression levels in adipose tissue and the liver (4). Moreover, hepatic expression levels of Repin1 are significantly associated with genotype and serum lipid profiles of different rat strains (4). Recently, we showed that Repin1 plays a role in the regulation of lipid accumulation in 3T3-L1 adipocytes as knockdown of Repin1 by small interfering RNA (siRNA) resulted in reduced palmitate uptake and altered mRNA expression of fatty acid transporter Cd36 and genes involved in lipid droplet formation (Vamp4 and Snap23), adipogenesis, and glucose transport (9). A balance between storage and release of lipids by adipose tissue is essential for maintenance of normal energy homeostasis and prevention of ectopic lipid accumulation in peripheral tissues, such as liver, pancreas, or skeletal muscle (10).

Based on our previous findings, we hypothesize that Repin1 is involved in glucose homeostasis and hepatic lipid storage and may contribute to alterations in glucose homeostasis and lipid metabolism associated with obesity. We therefore tested the in vivo consequences of Repin1 gene disruption in mice with a liver-restricted, Cre-loxP–mediated Repin1 deletion (LRep1^−/−).

All animal studies were approved by the local authorities of the state of Saxony, Germany, and of Mecklenburg-West Pomerania, Germany, as recommended by the responsible local animal ethics review board (Regierungspräsidium Leipzig, TVV20/08, T02/13, TVV15/14, and Rostock LALLF M-V/TSD/7221.3-1.1-099/12, Germany). All mice were housed in pathogen-free facilities in groups of three to five at 22 ± 2°C on a 12-h light/dark cycle. Animals were bred and kept in the animal laboratories at the University of Leipzig and were fed a standard chow diet (Altromin GmbH, Lage, Germany). A subgroup of eight LRep1^−/− and eight wild-type (WT) mice were kept on a high-fat diet (HFD) containing 55.2% of calories from fat (C1057; Altromin). Animals had ad libitum access to water at all times, and food was only withdrawn if required for an experiment.

The Repin1 gene in the liver was inactivated using conditional gene–targeting strategies. Floxed Rep1^−/− mice were generated by Artemis Pharmaceuticals (Köln, Germany). Exon 2 was flanked by loxP sites, and two positive selection markers were used in order to increase corecombination frequency of both loxP sites (Fig. 1A). The selection markers are flanked by frt (Neo) or F3 sites (Puro). The conditional knockout (KO) occurs after in vivo Flp-mediated removal of selection markers and constitutive KO by Cre-mediated deletion of exon 2. The deletion of exon 2 resulted in loss of function by removing the complete open reading frame.

Mouse genomic fragments were ET subcloned using RP23 BAC library and recloned into the basic targeting vector harboring the indicated features. Mice homozygous for the loxP-flanked Rep1 allele (Rep1^flox/flox) were crossed with mice expressing a Cre recombinase under control of the albumin (Alb) promoter (C57BL/6-TgN(AlbCre)21Mgn, stock 003574; The Jackson Laboratory). In the liver, Cre recombinase mediates the deletion of all floxed alleles. LRep1 mice were on C57BL/6N background.

Genotyping was performed by PCR using genomic DNA isolated from the tail tip. Genomic DNA was prepared by using the DNeasy Kit (Qiagen, Hilden, Germany). The following two primer pairs were used to genotype LRep1 loxP sites: 5′-CCCAACACTGATTACAGATCC-3′ (forward) and 5′-GTGGGATCAGATAG AACTTAGC-3′ (reverse) as well as the AlbCre recombinase 5′-GCGGTCTGGCAGTAAAAACTATC-3′ (forward) and 5′-GTGAAA CAGCATTGCTGTCACTT-3′ (reverse). PCR was performed for 35 cycles of 95°C (loxP sites) or 94°C (AlbCre), 60°C (30 s, loxP sites), or 51°C (60 s, AlbCre) and 72°C (60 s each) using the Fermentas DreamTaq Polymerase (Fermentas GmbH, St. Leon-Rot, Germany) and a Peltier Thermal Cycler PTC-200 (Bio-Rad, Hercules, CA). DNA from WT mice produced a 292-bp band, and a 484-bp band was detected in LRep1^AlbCre lox mice.

Twelve male mice of each genotype (LRep1^−/− and control littermates [Rep1^flox/flox, Rep1^flox/+, and WT]) were studied from the age of 6 weeks up to 40 weeks. Body weight was recorded weekly, and body length (naso-anal length) was measured once at an age of 32 weeks (n = 10 per genotype).

Intraperitoneal glucose tolerance tests (GTTs) and insulin tolerance tests (ITTs) were performed in males at the age of 12, 24, and 40 weeks as described previously (11). In brief, GTT was performed after an overnight fast of 14 h by injecting 2 g glucose per kg body weight into LRep1^−/− and littermate controls. Blood samples for glucose measurements were taken at different time points after 0, 15, 30, 60, and 120 min as described previously (11).

ITT was performed in random-fed animals by injecting 0.75 units/kg body weight human regular insulin (40 units Insuman Rapid; Sanofi, Frankfurt/Main, Germany). Glucose levels were determined in blood collected from tail tip immediately before and 15, 30, and 60 min after the intraperitoneal injection.

Indirect calorimetry was assessed by a calorimetry module (CaloSys V2.1; TSE Systems, Bad Homburg, Germany) at an age of 30 weeks. After 2 h of acclimatization, mean oxygen consumption (VO₂) as well as spontaneous activity (XYZ cage movement) and ability to run on a treadmill were recorded for 72 h. At an age of 16 weeks, a subgroup of 20 (n = 10 per genotype) mice underwent a food intake measurement over a time period of 1 week. The daily food intake was calculated as the average intake of chow within the time stated. Rectal body temperature was measured at an age of 32 weeks. Whole-body composition (fat mass, lean mass, and total body water) was determined in awake mice by using nuclear magnetic resonance technology with EchoMRI-700 instrument (Echo Medical Systems, Houston, TX) in control and LRep1^−/− mice at 3, 8, 16, 24, and 80 weeks of age. At least four animals per genotype and time point were measured. Data were analyzed by the manufacturer’s software.

Mice were killed at the age of 32 weeks by an overdose of anesthetic (isoflurane; Baxter, Unterschleißheim, Germany). Liver, heart, brain, lung, spleen, pancreas, kidney, muscle, and subcutaneous (SC) and epigonadal adipose tissue (Epi) were immediately removed. The organs (liver, brown adipose tissue [BAT], and Epi) were weighed, and organ mass was related to whole-body mass to obtain relative organ weights.

Blood glucose values were determined from whole–venous blood samples using an automated glucose monitor (FreeStyle Mini; Abbott GmbH, Ludwigshafen, Germany). Insulin, leptin, and adiponectin serum concentrations were measured by ELISA using mouse standards according to the manufacturer’s guidelines (Mouse/Rat Insulin ELISA and Mouse Leptin ELISA; Crystal Chem Inc., Downers Grove, IL; and Mouse Adiponectin ELISA; Adipogen International, Incheon, Korea). Serum concentrations of alanine aminotransferase (ALT), aspartate aminotransferase (AST), Alb, free fatty acids (FFAs), TGs, LDL cholesterol, and HDL cholesterol were analyzed by an automatic chemical analyzer in our Institute of Laboratory Medicine and Clinical Chemistry. Serum glycerol concentration was measured using Adipolysis Assay Kit (Merck Millipore, Billerica, MA) in male LRep1^−/− and controls at an age of 30 weeks.

Catheters were implanted in the left jugular vein and hyperinsulinemic-euglycemic clamps of six males of each genotype were performed at the age of 20 weeks. Clamp was performed as described previously (12–14).

Lipids were extracted from mouse hepatocytes using the Folch et al. (15) protocol with minor modification (16). Molecular lipid species were identified and quantified using LipidXplorer software (17) developed by MPI CBG (Dresden, Germany). Species were quantified by comparing the intensities of their peaks to peaks of spiked internal standards; lipid quantities determined in individual samples were normalized by the total protein content determined by Bradford assay. Cholesterol was quantified as previously described (18).

RNA isolation and quantitative real-time PCR were performed as previously described (11). mRNA expression of genes listed in Supplementary Table 3 was determined. Specific mRNA expression was calculated relative to 18s RNA, which was used as an internal control due to its resistance to glucose-dependent regulation (19).

In vivo lipogenesis was performed as previously described in detail (20,21). To measure hepatic TG production rate, mice were intraperitoneally injected with Poloxamer 407 (p407; Sigma-Aldrich) in saline ∼4 h into the light cycle, and plasma TG was measured over a 4-h period as described elsewhere (22). At an age of 16 weeks, after an overnight fast, 200 μL olive oil was administrated by intragastic gavage feeding tube. Blood samples were taken by submandibular bleeding at 0, 1, 2, 3, and 4 h after fat load for TG measurements. Eight male mice per genotype were studied.

The determination of glucose transport, lipolysis, palmitate uptake, and adipocyte size distribution was performed as previously described (9,23).

For Western blot analysis, tissues were removed and homogenized in homogenization buffer with tissue-mill homogenizer (MM 400; Retsch GmbH, Haan, Germany), proteins were isolated using standard techniques, and Western blot analysis was performed with antibodies raised against Repin1 (N-20, 1:1,000; Santa Cruz Biotechnology, Santa Cruz, CA), Pparγ (1:100, #351540; Antikörper Online), Irs1 and pIrs1 (1:1,000; Cell Signaling Technology), Akt and pAkt (1:1,000; Cell Signaling Technology), Glut2 (1:200, ab85715, Abcam, Cambridge, U.K.), Pparγ (1:1,000; Cell Signaling Technology), ACC (1:1,000; Cell Signaling Technology), lipocalin2 (Lcn2) (1:1,000; Abcam), Cd36 (1:500; Antibodies Online), and GAPDH antibody (1:3,000; Research Diagnostics, Flanders, Netherlands) as loading control. For Cd36 protein detection, supernatant was centrifuged at 100,000g for 45 min at 4°C. To get the membrane and cytosol fractions, the pellet was suspended in ice-cold sucrose buffer and taken to Cd36 analysis in Western blot.

Mice were anesthetized by intraperitoneal injection, and adequacy of the anesthesia was ensured by loss of pedal reflexes. The abdominal cavity of the mice was opened, and 125-μL samples containing 5 units regular human insulin diluted in 0.9% saline were injected into the vena cava inferior. Sham injections were performed with 125 μL of 0.9% saline. Samples of liver tissue were harvested 10 min after injection, respectively, and proteins (Akt, pAkt, Irs1, and pIrs1) were extracted from tissues for Western blot analysis.

RNA from liver samples of three male WT and three male LRep1^−/− mice was used for microarray RNA analysis. Analysis of RNA integrity and RNA concentration as well as probe synthesis, hybridization, and scanning was performed as previously described (24).

Histology and immunohistochemistry were performed as previously described (23,25).

Data are given as means ± SE. Data sets were analyzed for statistical significance using a two-tailed unpaired Student t test or Mann-Whitney U test. P values <0.05 were considered significant.

In order to generate LRep1^−/− mice, mice homozygous for the loxP-flanked Rep1 allele (Rep1^flox/flox) were crossed with mice expressing the Cre recombinase under control of the liver-specific Alb promoter. The targeting strategy is shown in Fig. 1A. LRep1^−/− mice obtained with the expected Mendelian frequency were fertile. Efficiency and specificity of the Repin1 KO were examined in tissue lysates from control and LRep1 mice by Western blot analyses (Fig. 1B and C). Western blot analysis of liver lysates from LRep1^−/− mice confirmed reduced Repin1 expression by ∼85% in livers (Fig. 1C). Since the Cre recombinase is only active in hepatocytes, and since hepatocytes make up ∼85% of the total number of cells in the liver, it is likely that the minimal Repin1 protein in LRep1^−/− mice is derived from vascular endothelial cells, Kupffer cells, and other nonparenchymal cells. Immunohistochemistry of Repin1 in the liver was performed to test whether nonparenchymal cells are responsible for the remaining Repin1 expression in LRep1^−/− mice. The immunohistochemical images of the liver of LRep1^−/− mice demonstrate the absence of Repin1 in hepatocytes but positive Repin1 staining for endothelial cells, Kupffer cells, and hepatic stellate cells (Fig. 1D). In contrast, hepatocellular cytoplasm of WT mice showed positive staining for Repin1 (Fig. 1D). Repin1 expression in all other tissues was indistinguishable between LRep1^−/− and control littermates (Fig. 1B).

LRep1^−/− mice had normal body weight compared with control littermates (Rep1^flox/flox, Rep1^flox/+, and WT) up to an age of 28 weeks when LRep1^−/− mice become significantly leaner with significantly less total body fat content up to an age of 40 weeks (Fig. 1E and F). This may be due to significantly higher VO₂ consumption and spontaneously higher activity (z-axis day) in LRep1^−/− mice (Fig. 1G–I). Homozygous deficiency of Repin1 had no significant influence on body length (Supplementary Fig. 4A), daily food intake (Supplementary Fig. 4B), or relative tissue weights of livers (WT, 5.2%; LRep1^−/−, 5.3%).

To further investigate the role of Repin1 on glucose homeostasis, we characterized the physiological consequences of reduced liver Repin1 expression on insulin action and glucose metabolism. Hyperinsulinemic-euglycemic clamp studies revealed significantly higher whole-body insulin sensitivity in LRep1^−/− compared with control mice (Fig. 2A). At the steady state, glucose infusion rate (GIR) was ∼60% higher in LRep1^−/− compared with control mice (Fig. 2A). LRep1^−/− mice showed a higher rate of hepatic glucose production during the basal period as compared with control animals (Fig. 2B), whereas insulin suppressed hepatic glucose production significantly better in LRep1^−/− (−98%) compared with control mice (−51%) (Fig. 2C). Circulating insulin concentrations achieved in the steady state of the clamp were not significantly different between WT (9.1 ± 1.4 ng/mL) and LRep1^−/− (8.3 ± 1.6 ng/mL) mice.

In addition, we monitored blood glucose, insulin, as well as adiponectin serum concentrations (Table 1) and performed serial GTTs and ITTs over an age range from 12 to 40 weeks (Fig. 2D and E and Supplementary Fig. 4C and D). Independent of age, intraperitoneal GTTs revealed normal glucose tolerance in LRep1^−/− mice (Fig. 2D). At an age of 24 weeks, LRep1^−/− mice showed significantly improved insulin sensitivity compared with controls (Fig. 2E), which was even more pronounced at a higher age (Fig. 2F–H). Significantly lower HbA_1c level confirmed the long-term glucose metabolism in LRep1^−/− compared with control mice (Fig. 2I). Moreover, insulin secretion in response to the intraperitoneal glucose load in 40-week-old LRep1^−/− mice showed an improved insulin secretion response to glucose reflected by a higher insulin peak at 15 min and a faster decline in insulin serum concentrations than in controls (Fig. 2G).

Taken together, reduced Repin1 expression in liver results in improved whole-body insulin sensitivity, subsequently contributing to better glycemic control in LRep1^−/− mice.

We further sought to investigate the consequences of disrupted Repin1 in the liver and improved insulin sensitivity by analyzing key insulin signaling molecules in livers of LRep1^−/− and control mice after insulin stimulation. Supporting a role of Repin1 in the regulation of hepatic insulin sensitivity at the molecular level, we find a significant upregulation of phosphorylated Akt and Pparγ expression in LRep1^−/− compared with control mice (Fig. 3A). These molecular changes may underlie significantly higher Glut2 expression (Fig. 3A), which may explain higher basal glucose uptake in the hyperinsulinemic-euglycemic clamp in LRep1^−/− compared with control mice.

To determine the physiological consequences of reduced liver tissue Repin1 expression, we monitored total serum cholesterol, TG serum concentrations, FFA, and liver function tests such as circulating serum ALT, AST, glutamate dehydrogenase (GLDH), and Alb concentrations. Fasted TGs and FFAs were significantly higher in LRep1^−/− mice at an age of 32 weeks (Table 1). ALT, AST, and Alb were not affected by Repin1 KO, suggesting that lack of Repin1 has no adverse effect on liver function (Table 1). Also hematoxylin-eosin (H-E) staining of control and LRep1^−/− mice showed normal hepatic architecture without signs of hepatocyte injury (Fig. 3D and Supplementary Figs. 2 and 3).

Liver lipidomics analyses revealed significantly lower liver content triacylglycerides (TAGs) in LRep1^−/− compared with control mice (Table 2). The amount of TAGs in liver was significantly decreased in LRep1^−/− mice by ∼40% compared with controls (Table 2 and Fig. 3C). Hepatic cholesterol and cholesterolester were unchanged (Table 2). Lower lipid accumulation in the liver of LRep1^−/− mice may at least in part explain improved whole-body insulin sensitivity in these mice.

To identify potential mechanisms underlying reduced TG storage in the liver and elevated circulating TGs and FFAs in LRep1^−/− mice, we examined hepatic expression of fatty acid transporters Cd36, Fatp1, Fatp2, Fatp4, and carnitine palmitoyltransferase 1 (Ctp1) as the first component and rate-limiting step in β-oxidation and Cpt2 (Supplementary Table 3). We performed in vivo VLDL production assays, in vivo lipogenesis, and oral fat load tests. Injection of p407 resulted in a linear increase in serum TG concentration without significant differences between both experimental groups, which suggests no influence of hepatic VLDL synthesis on altered lipid content in LRep1^−/− mice (Fig. 3G). Further, to investigate de novo lipogenesis, we injected mice both with tritiated water and [U-¹⁴C]-glucose (Fig. 3J). Total rate of hepatic fatty acid synthesis was comparable between controls and LRep1^−/− mice, and the incorporation of [U-¹⁴C]-glucose into de novo fatty acids was similar as well (Fig. 3I and J). Expression of a key lipogenic enzyme, ACC (pACC), was not changed in livers of LRep1^−/− mice compared with littermate controls (Fig. 3K). Cd36 was significantly reduced in LRep1^−/− mice by ∼50% at both protein and mRNA levels (Fig. 3E and F).

In a subgroup of eight animals of each genotype, we performed an HFD study starting at 6 weeks of age until 16 weeks of age. Weight gain after HFD was not different between all groups (Fig. 4A). Relative liver weight was slightly reduced in LRep1^−/− mice compared with controls (data not shown). Also ALT and AST as well as Alb levels did not differ among genotypes (data not shown), and Epi mass was comparable between LRep1^−/− and control mice (data not shown). Despite indistinguishable relative adipose tissue mass, LRep1^−/− mice showed decreased maximal adipocyte diameters compared with controls in response to HFD (Fig. 4B–F). Maximal epigonadal adipocyte diameter is significantly larger in controls compared with LRep1^−/− mice (Fig. 4E and F). Adipocyte frequency size distribution confirmed these findings (Fig. 4C and D). These differences are more pronounced in epigonadal than in SC fat. To elucidate in more detail adipocyte function, we analyzed glucose uptake and glycerol release in isolated adipocytes. Here, we detected a significant elevated glucose uptake and glycerol release under basal conditions in SC adipocytes of LRep1^−/− mice (Fig. 4I and J). Insulin-stimulated glucose uptake was comparable between the experimental groups (Fig. 4J). To elucidate in more detail elevated basal glycerol release, we performed gene expression analysis in adipose tissue. Here, we detected an elevation of all lipolysis enzymes (Lpl, Atgl,and Hsl) in SC adipose tissue in LRep1^−/− mice (Supplementary Table 5), indicating enhanced lipolysis. Taken together, LRep1^−/− mice were protected against HFD-induced adipocyte hypertrophy.

To identify Repin1-regulated target genes, we measured mRNA expression in the liver of LRep1^−/− and control mice using a microarray approach. Lcn2, also known as neutrophil gelatinase–associated lipocalin, was the strongest regulated gene in response to reduced Repin1 expression (Supplementary Table 2). We further found reduced levels of clathrin-coated vesicle transport protein (ap3m2), vesicle-associated membrane proteins, oxidative stress genes such as metallothein1 (Mt1) and 2 (Mt2), as well as LDL receptor (Ldlr) and Rarres 2 (chemerin). Exocytosis factor titin, mitochondrial protein Letm2, and kallikrein inhibitor (serpinA4-ps1) were expressed higher in livers of LRep1^−/−mice (Supplementary Table 2). In addition to these genes, we detected a number of Repin1-regulated genes including various biological processes and molecular functions (Supplementary Tables 1 and 2). We could confirm the expression microarray data for Lcn2 and chemerin in a bigger liver tissue cohort (see Supplementary Table 4). Circulating Lcn2 concentrations and liver Lcn2 expressions were lower in LRep1^−/− compared with littermate control mice (P = 0.2 serum; P = 0.06 liver protein) (Supplementary Fig. 5).

Array mRNA expression changes together with the results of the hepatic lipidomic screen indicated an alteration in lipid droplet fusion and formation. We therefore investigated mRNA levels of proteins involved in the fusion process of lipid droplets in more detail. Vesicle-associated membrane protein 4 (Vamp4) and synaptosomal-associated protein, 23 kDa (Snap23), were significantly decreased in livers of Repin1-deficient mice. Vamp4 was reduced by 80% and Snap23 by ∼40% in LRep1^−/− (Supplementary Table 3).

We propose the following model of how liver-restricted knockdown of Repin1 may cause the observed phenotype (Fig. 5). Hepatic Repin1 deficiency causes improved liver and whole-body insulin sensitivity most likely through transcriptional regulation of genes involved in insulin signaling (Pparγ and Akt), glucose transport (Glut2), lipid uptake (Cd36), and lipid droplet formation (Vamp4 and Snap23). These changes result in less body weight gain with aging, higher energy expenditure, and reduced liver fat accumulation in LRep1^−/− compared with controls. Lower liver lipid content may be primarily caused by significantly reduced Cd36 expression in LRep1^−/− mice. Improved whole-body insulin sensitivity of LRep1^−/− mice is likely a consequence of reduced liver fat, decreased hepatic glucose production, and eventually lower expression of hepatokines such as Lcn2, which are associated with insulin resistance.

Our data provide the first in vivo evidence that Repin1 deletion in liver leads to lower body weight, reduced hepatic steatosis, increased energy expenditure and physical activity, and improved insulin sensitivity both at the organ and whole-body level. Lower body weight in LRep1^−/− compared with control mice may be due to several mechanisms. LRep1^−/− mice are characterized by higher VO₂ consumption and, at least in one dimension, higher spontaneous activity with unaltered food intake compared with WT mice. We tested the hypothesis that the increased energy expenditure in LRep1^−/− compared with WT mice is a result of increased BAT mass or activity. Since mean body temperature, relative and absolute BAT mass, as well as expression of BAT marker genes in white adipose tissue are not different between LRep1^−/− and control mice, we exclude an effect of liver Repin1 deficiency on BAT or browning of white adipose tissue. Taken together, increased energy expenditure and activity of LRep1^−/− mice contribute to lower body weight.

Importantly, deletion of Repin1 in liver does not cause an increase in inflammatory marker genes or immune cell infiltration into livers of LRep1^−/− mice. Moreover, AST, ALT, and GLDH serum concentrations were not significantly different between WT and LRep1^−/− mice, excluding a hepatotoxic effect of Repin1 deletion.

Reduced liver fat is most likely due to significant changes in the expression of Repin1 target genes such as Cd36 and fatty acid binding proteins (FABPs). In the context of these (beneficial) metabolic consequences of reduced liver-specific Repin1 expression, elevated circulating TGs and FFAs in LRep1^−/− mice seem to be contradictory to the phenotype. We therefore propose a model (Fig. 5) of how these changes in circulating lipids may be the result of changes in expression of key molecules in fatty acid uptake, transport, and lipid droplet formation in the liver. Elevated TGs and FFAs are most likely due to decreased hepatic lipid uptake capacity as a result of lower Cd36 expression in LRep1^−/− mice. In addition to reduced Cd36 expression, several mechanisms may contribute to the observed reduction in TG storage in LRep1^−/− mice. Both de novo lipogenesis and VLDL production were not significantly different between LRep1^−/− and control mice, suggesting that these processes do not cause reduced TG storage in livers of LRep1^−/− mice. Moreover, we did not find differences in [¹⁴C]palmitate oxidation or the expression of key enzymes of β-oxidation between the genotypes, suggesting that Repin1 KO in liver does not significantly affect fatty acid oxidation in liver. Key gene expressions of TG synthesis (Fasn, Scd1, and Scd2) were not altered by Repin1 deletion in liver. However, since we did not directly assess TG synthesis in hepatocytes, we cannot exclude that this mechanism may contribute to reduced TG storage in livers of LRep1^−/− mice. In accordance with altered liver lipid transporter expression, we detected lower concentrations of TAGs in the liver of LRep1^−/− mice. These in vivo results strongly support our previous data on siRNA-mediated Repin1 knockdown in 3T3-L1 cells, which caused significantly reduced mRNA expression of fatty acid transporter Cd36 and lipid droplet genes (Snap23 and Vamp4) in 3T3-L1 cells (9). Moreover, the circulating lipid profile of LRep1^−/− mice closely reflects that of CD36 KO mice, which are also characterized by elevated fasting FFAs and triacylglycerol, as well as cholesterol serum concentrations (26). Interestingly, both LRep1^−/− and CD36 KO mice are protected against impaired glucose homeostasis, despite these alterations in lipid metabolism (26). Therefore, Repin1 deficiency in the liver provides a model to dissect the effects of liver fat accumulation from those of increased circulating lipids on whole-body insulin sensitivity. The insulin-sensitive phenotype of LRep1^−/− mice further suggests that hepatic steatosis may represent a crucial mechanism in the development of insulin resistance independently of the increased circulating FFAs and TGs. Noteworthy, in mice with liver-specific insulin resistance due to a targeted disruption of the hepatic insulin receptor (LIRKO mice), circulating TGs and FFAs are ∼40–50% lower compared with controls (27). This opposite phenotype further suggests that improved insulin sensitivity in the liver of LRep1^−/− mice is the primary effect of reduced Repin1 expression and changes in circulating lipids are secondary to that.

Kennedy et al. (28) demonstrated that adipose tissue from Cd36 KO mice was more insulin sensitive and had lower levels of inflammatory markers (i.e., IFN-γ and MCP-1) as compared with WT mice.

By several measures, we found beneficial effects of Repin1 deletion in liver on insulin sensitivity both under chow- and HFD-fed conditions. Noteworthy, insulin secretion dynamic in 40-week-old LRep1^−/− mice suggests that Repin1 KO in the liver may cause secondary changes in islets that are in line with the observed improvements in whole-body insulin sensitivity. Whether the insulin secretion profile of LRep1^−/− mice is due to improved insulin sensitivity or the result of a hepatic factor directly affecting islets needs to be explored in subsequent studies. Improved insulin sensitivity in LRep1^−/− mice could be the result of improved activation of the insulin signaling cascade (e.g., Akt phosphorylation), increased Pparγ expression, lower liver fat content mediated by reduced expression of fatty acid transport proteins in LRep1^−/− mice, lower total body fat content, and lower expression of insulin resistance–associated hepatokines (e.g., Lcn2 and chemerin).

With regard to the latter mechanism, mRNA expression array identified Lcn2 as the strongest potential candidate. Elevated Lcn2 serum concentrations are associated with obesity, dyslipidemia, and insulin resistance (29). In livers of the LRep1^−/− mice, we detected a 60-fold reduction of Lcn2 mRNA level and reduced liver protein levels (P = 0.06) compared with controls. There was a trend for lower circulating Lcn2 in LRep1^−/− mice. Lcn2, also known as neutrophil gelatinase–associated lipocalin, is a lipocalin subfamily member and has been recently identified as an adipose tissue–derived cytokine (30). Lcn2 is an extracellular lipocalin and has structural similarity with FABPs, and both are members of the multigene family of up and down β-barrel proteins (31). Both intracellular FABPs and the extracellular lipocalins have a clearly defined β-barrel motif that forms either an interior cavity (FABP) or a deep pit (lipocalin) that constitutes the lipid binding domain (31). Because of the unique structure, the lipocalins function as efficient transporters for a number of different hydrophobic ligands in extracellular milieus, including a variety of retinoids, fatty acids, biliverdin, pheromones, porphyrins, odorants, steroids, and iron (32). In vitro, it was shown that exogenous Lcn2 promotes insulin resistance in cultured hepatocytes (33). It is likely that Lcn2 as retinoic acid could potentially be involved in or mediate lipid storage effects in both liver and adipose tissue. Moreover, reduced Lcn2 in LRep1^−/− mice may contribute to the observed alterations in adipose tissue with a higher number of smaller adipocytes in LRep1^−/− compared with control mice upon HFD. Liver Repin1 deficiency leads to secondary changes in adipose tissue with a reduction in adipocyte size under HFD in homozygous LRep1^−/− mice, indicating a protection against hypertrophy of adipocytes. However, since we did not biopsy adipose tissue during HFD, we do not provide direct evidence for increased adipogenesis in LRep1^−/− mice. Protection of LRep1^−/− mice against adipocyte hypertrophy could contribute to significantly better insulin sensitivity as measured by ITT after 10 weeks of HFD.

Our previous in vitro studies demonstrated that Repin1 expression increases during adipogenesis and that RNA interference–based Repin1 downregulation in mature adipocytes significantly reduces adipocyte size (9). We found significant correlations between Repin1 mRNA expression and total body fat mass as well as adipocyte size in human paired visceral and SC adipose tissue, suggesting a potential role for Repin1 in the regulation of adipocyte size (9). However, it is not clear how a lack of Repin1 in liver might influence adipocyte size. One possibility is that absence of Repin1 in liver generates signals that either restrict adipocyte lipid load or increase adipogenesis. For the latter mechanism, we consider reduced Lcn2 expression in livers of LRep1^−/− mice as a candidate molecule.

This hypothesis is supported by data that Lcn2 deficiency protects mice from developing aging- and obesity-induced insulin resistance by modulating insulin resistance factors in adipose tissue (34). Lcn2-disrupted mice are partly protected from HFD-induced insulin resistance. In this context, it has been shown that other members of the lipocalin family, in particular lipocalin-type prostaglandin D synthase, may impair the adipogenesis program (35). Although we do not have direct evidence for increased adipogenesis in LRep1^−/− mice, we speculate that changes in adipose tissue morphology may be due to reduced Lcn2 levels with subsequent disinhibition of adipogenesis.

In contrast to our observation marked Lcn2 deficiency leads to enlarged adipocytes and improved adipose tissue function, which has been related to reduced inflammation in adipose tissue of these mice (34). We cannot exclude that changes in adipose tissue of LRep1^−/− mice are at least in part mediated by reduced inflammatory activation in adipose tissue. In addition, reduced chemerin expression may lead to smaller adipocyte size. Indirect evidence from insulin-sensitive obese individuals supports the hypothesis that lower circulating chemerin (in insulin-sensitive, healthy obese) is associated with smaller adipocyte size (36). Whether additional mediators contribute to the specific response to HFD in adipose tissue of LRep1^−/− mice should be further studied. Adipocyte-specific insulin sensitivity was not altered in LRep1^−/− mice. Hepatocyte-derived factors may have further contributed to increased lipolysis in adipocytes of LRep1^−/− mice, which has been suggested by higher basal glycerol release and higher expression of key lipolysis enzymes (Hsl, Atgl, and Lpl) in LRep1^−/− adipocytes.

To further elucidate potential mechanisms for improved liver insulin sensitivity and reduced liver fat content in LRep1^−/− mice, we performed a hepatic lipidomics screen as well as mRNA expression array analysis in liver samples. The liver lipidomics data demonstrate that Repin1 modulates the overall profile of liver lipid species. Here, the main finding of the screen was that hepatic deletion of Repin1 had a significant effect on TAG amount in livers, suggesting altered lipid storage in liver. Thus, reduced liver lipid storage ability could explain significantly higher serum TG levels in LRep1^−/− mice. Since LRep1^−/− hepatocytes did not undergo typical ballooning under HFD conditions, we examined whether genes involved in lipid droplet fusion as well as lipid transport proteins are regulated by Repin1 deficiency. We detected significantly reduced levels of Vamp4 and Snap23 in livers of Repin1-deficient mice, suggesting a disturbance in lipid vesicle formation and hepatic lipid upload. Boström et al. (37) demonstrated that Snap23 and Vamp4 are functional in the fusion between lipid droplets and are essential for the growth of lipid droplets. Our results are in accordance with our previous finding that downregulation of Repin1 in 3T3-L1 adipocytes by siRNA leads to significantly reduced lipid droplet size (9). In parallel to the mRNA expression data in the liver of Repin1-deficient mice, we found reduced Snap23 and Vamp4 expression in 3T3-L1 cells upon downregulation of Repin1 (9). Knockdown of VAMP4 and SNAP23 has been recently shown to decrease the rate of lipid droplet fusion and the size of the lipid droplets (33). Noteworthy, decreased VAMP4 and SNAP23 gene expression in response to Repin1 knockdown seems to be specific, since other lipid droplet proteins, including perilipin and syntaxin5, were not regulated by Repin1 (9).

In conclusion, we provide a model of how Repin1 may, through regulation of gene expression, lead to lower body weight, improved liver and whole-body insulin sensitivity, and alterations in lipid metabolism. We propose that hepatic Repin1 deficiency primarily leads to improved insulin sensitivity and reduced liver fat content with secondary changes in serum lipid profile due to alterations in hepatic lipid transport. Furthermore, hepatic deletion of Repin1 causes altered expression of molecules, including Lcn2 and chemerin, which may contribute to reduced adipocyte size in response to HFD, which may subsequently further contribute to beneficial effects on whole-body insulin sensitivity in LRep1^−/− mice.

Our findings indicate that Repin1 contributes to insulin sensitivity and glucose and lipid metabolism by regulating the expression of key molecules of these processes. Therefore, alterations in Repin1 expression may contribute to the pathogenesis of insulin resistance and dyslipidemia and subsequent impairment of glucose homeostasis.

Acknowledgments. The authors thank Eva Böge, Manuela Prellberg, Viola Döbel, Anne Kunath, and Daniela Kern (University of Leipzig) for technical assistance. Furthermore, the authors thank Professor Fitzl (Institute of Anatomy, University of Leipzig) for help with liver staining.

Funding. This work was supported by the clinical research group “Atherobesity” (KFO 152, project KL-2346) and the SFB 1052: B1 (to M.B.), B3 (to P.K.), and B4 (to N.K.) founded by Deutsche Forschungsgemeinschaft, formel1 program, and the Federal Ministry of Education and Research, Germany, FKZ: 01EO1001 (N.K.). J.T.H. is funded by the European Union and the Free State of Saxony.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. M.K. researched data and wrote the manuscript. J.K., N.H., and I.K. researched data and performed Western blots. J.B. and J.T.H. performed Western blots and histological experiments. G.F. measured serum concentrations. P.K. reviewed the manuscript and contributed to discussion. M.M.-S., R.G., S.S., and A.S. performed lipidomics studies. K.K. performed Affymetrix GeneChip analysis. K.A. researched data and performed immunostaining. M.S. conceived the research ideas and reviewed and edited the manuscript. M.B. reviewed and edited the manuscript. N.K. conceived the research ideas, supervised the project, and wrote the manuscript. N.K. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.