Hepatic trans-Golgi action coordinated by the GTPase ARFRP1 is crucial for lipoprotein lipidation and assembly.

The liver is a major organ in whole body lipid metabolism and malfunctioning can lead to various diseases including dyslipidemia, fatty liver disease, and type 2 diabetes. Triglycerides and cholesteryl esters are packed in the liver as very low density lipoproteins (VLDLs). Generation of these lipoproteins is initiated in the endoplasmic reticulum and further maturation likely occurs in the Golgi. ADP-ribosylation factor-related protein 1 (ARFRP1) is a small trans-Golgi-associated guanosine triphosphatase (GTPase) that regulates protein sorting and is required for chylomicron lipidation and assembly in the intestine. Here we show that the hepatocyte-specific deletion of Arfrp1 (Arfrp1(liv-/-)) results in impaired VLDL lipidation leading to reduced plasma triglyceride levels in the fasted state as well as after inhibition of lipoprotein lipase activity by Triton WR-1339. In addition, the concentration of ApoC3 that comprises 40% of protein mass of secreted VLDLs is markedly reduced in the plasma of Arfrp1(liv-/-) mice but accumulates in the liver accompanied by elevated triglycerides. Fractionation of Arfrp1(liv-/-) liver homogenates reveals more ApoB48 and a lower concentration of triglycerides in the Golgi compartments than in the corresponding fractions from control livers. In conclusion, ARFRP1 and the Golgi apparatus play an important role in lipoprotein maturation in the liver by influencing lipidation and assembly of proteins to the lipid particles.


Fig. 1. Altered plasma parameters in fasted Arfrp1
liv Ϫ / Ϫ mice. A: Free fatty acids (upper panel, n = 9-10) and total plasma triglyceride concentrations (lower panel, n = 19) of Arfrp1 liv Ϫ / Ϫ and control mice measured after overnight fasting. B: Lower triglyceride content in VLDL plasma fractions of Arfrp1 liv Ϫ / Ϫ mice after fasting determined after separation by ultracentrifugation (upper panel, pooled samples from three mice per genotype) or detected by FPLC (lower panel, n = 7-11). OD, optical density. C: Concentration of total plasma cholesterol in Arfrp1 liv Ϫ / Ϫ and control mice after fasting (n = 19). D: Shift of cholesterol content from HDL-containing to LDL plasma fractions of Arfrp1 liv Ϫ / Ϫ mice after fasting as determined after separation by ultracentrifugation (upper panel, pooled samples from three mice per genotype) or as detected by FPLC (lower panel, n = 7-11). E: Elevated ApoB48 and ApoB100 in Arfrp1 liv Ϫ / Ϫ HDL-containing plasma fractions generated by ultracentrifugation. *** P < 0.001. in the supernatant were detected with a commercial kit (Randox-TR-210, Crumlin, UK).

Lipid distribution in the lipoprotein fractions
To obtain lipoprotein fractions for fast-protein liquid chromatography (FPLC) analysis, each individual plasma sample was loaded on a fi ltration chromatography column onto a Superose 6 10/300 GL column (GE Healthcare), which separates lipoproteins according to their size, and cholesterol and triglyceride concentrations were continuously measured in the effl uent using an enzymatic assay at an optical density of 492 nm as described (16)(17)(18). This system allows separation of the three major lipoprotein classes; VLDLs, LDLs, and HDLs. Cholesterol or triglyceride concentrations were determined in the eluted fractions. Accumulated data were analyzed by the Millenium 20/0 program (Waters).
The area under each peak is proportional to the lipid concentration in the respective lipoprotein fraction. Results are expressed as lipid concentration in VLDLs, LDLs, and HDLs as previously described ( 16 ). Moreover, data are also presented as cholesteroland triglyceride-mean profi les obtained with each individual plasma sample of each group.

Determination of hepatic VLDL secretion
To determine maximum hepatic VLDL secretion, basal blood samples were collected from the tail vein after fasting overnight. Triton WR-1339 (15% in saline, 0.5 g/kg) was injected intraperitoneally to block lipoprotein lipase and blood was sampled 1, 2, 4, and 6 h after application. After centrifugation, total triglyceride content was determined using a serum triglyceride determination kit (TR0100, Sigma, USA).
To investigate VLDL composition, mice were fasted overnight and blood samples were subjected to fractionation (350,000 g , 4 h, 16°C; OptiPrep, Axis-Shield, UK) in order to separate different lipoprotein fractions. Subsequently, triglyceride and cholesterol content of fractions were determined and apolipoprotein distribution was assessed by Western blotting.

Detection of ApoB48/100 synthesis and clearance by pulse-chase experiments
Primary hepatocytes were isolated from 12-week-old mice and Arfrp1 expression was suppressed by siRNA as described ( 11 ). To determine ApoB48/100 synthesis and degradation in primary hepatocytes, intracellular pools of cysteine and methionine were depleted by incubation with Cys-free and Met-free DMEM for 30 min prior to labeling with [35S] Promix (100 Ci/500 l) for 2 h ( 19 ). For chase experiments cells were consecutively incubated with normal DMEM for 3 h. Immunoprecipitation of ApoB48/100 insulin-like growth factor IGF1 from hepatocytes ( 11 ). A lipodystrophic phenotype was initiated by a fat cell-specifi c deletion of Arfrp1. Diminished lipid droplet fusion and elevated lipolysis were responsible for smaller lipid droplets in brown adipose tissue ( 12 ). Furthermore, in a recent study it was shown that in the intestine ARFRP1 is involved in the formation and lipidation of lipid-carrying chylomicrons ( 13 ). Besides its fundamental role in whole body glucose homeostasis, the liver is further essential for the distribution of lipids in the body via synthesis and secretion of VLDLs. Because the organization of VLDLs is very similar to that of chylomicrons, the aim of this study was to clarify the impact of the GTPase ARFRP1 and the trans -Golgi for lipoprotein assembly and release in the liver.

MATERIALS AND METHODS
Liver-specifi c Arfrp1 -knockout ( Arfrp1 liv Ϫ / Ϫ ) mice were generated as described ( 11 ) and compared with their control littermates ( Arfrp1 fl ox / fl ox ). All experiments were performed at the age of 5 weeks because of an incomplete suppression of Arfrp1 expression at later time points. The animals were housed in a controlled environment (20 ± 2°C, 12/12 h light/dark cycle) and had free access to water and standard diet (V153x, Ssniff, Soest, Germany). For fasting conditions, diet was deprived overnight. All animal experiments were approved by the ethics committee of the State Agency of Environment, Health, and Consumer Protection (State of Brandenburg, Germany).

Determination of ApoC3, free fatty acids, triglycerides, and cholesterol in plasma and liver samples of fasted mice
For the quantifi cation of ApoC3 in plasma and liver samples (25 g) a rat/mouse ELISA was used and applied according to the manufacturer's instructions (Apc3, Abnova, Germany). Nonesterifi ed free fatty acid, triglyceride, and cholesterol content of plasma was determined as described ( 14 ). Plasma cholesterol was determined by enzymatic colorimetric assay (Cholesterol liquicolor; Human, Wiesbaden, Germany).

Histological staining of intracellular lipids
To visualize stored triglycerides in liver sections, Oil-Red O staining was utilized. Therefore, microscopic slides were air dried, fi xed in 4% formaldehyde, and washed in 60% isopropyl alcohol. Staining was performed for 10 min in Oil-Red O solution (0.3% Oil-Red O in 60% isopropyl alcohol) and thereafter slides were rinsed in 60% isopropyl alcohol before nuclei were counterstained with hematoxylin. The morphometric measurement of lipid droplets was performed using a Keyence BZ-9000 microscope for picture acquisition and the corresponding software for quantifi cation (Neu-Isenburg, Germany).

Data analysis
Statistical differences were determined by nonparametric Mann-Whitney-Wilcoxon test with signifi cance levels set at P < 0.05 (*), P < 0.01 (**), and P < 0.001 (***). Data are presented as means ± standard error of the mean (SEM). For statistical analysis and for graphical presentation GraphPad Prism (5.0; Graph-Pad Software, San Diego, CA) was used.

Reduced hepatic triglyceride release in
VLDLs are the main cargo of triglycerides released by the liver under fasting conditions. To investigate the synthesis and release of VLDLs from the liver, mice expressing or lacking ARFRP1 in the liver were exposed to an overnight fasting period and plasma triglycerides and apolipoproteins were investigated. The amount of free fatty acids in the plasma of Arfrp1 liv Ϫ / Ϫ and control mice was not different ( Fig. 1A ), however, plasma triglycerides of Arfrp1 liv Ϫ / Ϫ mice were signifi cantly lower compared with the control animals ( Fig. 1A ). This difference in plasma triglycerides was mainly due to a lower amount of triglycerides in the VLDL fraction of the plasma of Arfrp1 liv Ϫ / Ϫ mice [density < 1.006 g/ml in Optiprep ( 22 ) ( Fig. 1B , lower panel). The VLDL fraction of Arfrp1 liv Ϫ / Ϫ mice contained fewer triglycerides ( Arfrp1 fl ox / fl ox , 31.1 ± 3.7 mg/dl; Arfrp1 liv Ϫ / Ϫ , 19.6 ± 2.3 mg/dl; P < 0.05), whereas more triglycerides were found in the LDL fraction (density 1.01-1.030 g/ml; Arfrp1 fl ox / fl ox , 11.8 ± 1.0 mg/dl; Arfrp1 liv Ϫ / Ϫ , 17.1 ± 1.3 mg/dl; P < 0.05). from cell lysates and cultured media/supernatant was carried out, separated on a 6% gel which was dried prior to visualization on a phosphor imager system (Storm 820, Molecular Dynamics). Band intensity was quantifi ed by Image Quant 5.2 (Molecular Dynamics).

Triglyceride synthesis in primary hepatocytes
To determine the rate of de novo triglyceride synthesis, primary hepatocytes were transfected with scrambled siRNA or siAr-frp1 and then incubated with [ 14 C]-labeled glucose (1 Ci/ml) overnight ( 20 ). Triglycerides from cells were isolated by chloroform-methanol extraction and the addition of CaCl 2 to prevent contamination of the aqueous phase with lipids. Lipid extracts were dried and redissolved in chloroform prior to counting in a scintillation counter (LS6000LL; Beckman Coulter, Krefeld, Germany).

Subcellular fractionation of liver homogenates
Livers of fasted Arfrp1 liv Ϫ / Ϫ and control mice were minced in homogenization buffer [0.25 M sucrose, 1 mM EDTA, 20 mM HEPES KOH (pH 7.4), and protease inhibitor], homogenized by a potter, and fi ltered through gauze (70 m ). After removing the nuclei by centrifuging at 1,500 g for 10 min at 4°C, the homogenate was separated by Optiprep density gradient centrifugation according to the manufacturer's instructions (Axis-Shield, UK). The gradient was collected in fractions of 1 ml at the end of the centrifugation. Aliquots were used to determine triglyceride concentration in fractions (Serum Triglyceride Determination kit, Sigma, Germany). Proteins in the fractions were concentrated using trichloroacetic acid to be analyzed by Western blotting ( 13 ).

Microsomal triglyceride transfer protein activity assay
The activity of microsomal triglyceride transfer protein (MTP) was determined as described in Jaschke et al. ( 13 ). Liver tissues were homogenized and 100 g of protein were used for the assay.
A reduced plasma triglyceride accumulation in Arfrp1 liv Ϫ / Ϫ mice was accentuated by the application of the lipoprotein lipase inhibitor Triton WR-1339 after fasting, which allows determination of triglyceride release by the liver. In control mice, a strong increase of plasma triglycerides indicated an accumulation of VLDL particles in the plasma ( Fig. 2A ). In comparison, only a slight increase in triglycerides was observed in the plasma of Arfrp1 liv Ϫ / Ϫ mice. As a consequence, hepatic triglycerides revealed elevated lipid levels in Arfrp1 liv Ϫ / Ϫ mice compared with control livers ( Fig. 2B ).

Altered secretion of VLDL-associated apolipoproteins in Arfrp1 liv ؊ / ؊ mice
ApoB48/100, as the main structural lipoprotein in VLDLs, was elevated in the plasma of Arfrp1 liv Ϫ / Ϫ mice as determined by WB analysis and immunonephelometric assay ( Fig. 3A , upper and middle panel). In addition, plasma fractionation after ultracentrifugation and consecutive WB analysis of the VLDL fraction indicated an accumulation of ApoB48/100 in VLDL particles ( Fig. 3A , lower  panel). Furthermore, immunohistochemical analyses of liver sections revealed an intracellular accumulation of ApoB48/100 in Arfrp1 liv Ϫ / Ϫ livers ( Fig. 3B ). A similar increase of ApoB100 was determined in the VLDL fraction from Arfrp1 liv Ϫ / Ϫ mice after Triton WR-1335 treatment ( Fig. 3C ). In order to test whether ARFRP1 infl uences ApoB48/100 synthesis, we measured mRNA of Apob by quantitative real-time PCR without detecting differences between the genotypes ( Fig. 3D ). For studying the ApoB48/100 protein synthesis and clearance, we isolated hepatocytes from 12-week-old C57BL/6 mice, suppressed Arfrp1 expression by transfecting siRNA, and performed a pulse-chase experiment. The amount of ApoB48/100 immunoprecipitated from the lysates and from the supernatant was unaltered after depletion of Arfrp1 2 h after treatment with 35 S-labeled amino acids ( Fig. 3E , upper panel). The stability of ApoB48/100 was unaffected in siArfrp1 hepatocytes compared with control transfected cells ( Fig. 3E , lower  panel).
Besides ApoB48/100, other apolipoproteins such as ApoC3, which with 40% of protein mass in secreted VLDLs represents a major apoprotein in these lipoproteins ( 23 ), are synthesized by the liver and attached to VLDLs prior to secretion. In the plasma of fasted Arfrp1 liv Ϫ / Ϫ mice, ApoC3, that acts as an inhibitor of plasma lipoprotein lipase ( 24 ), was signifi cantly less abundant than in Arfrp1 fl ox / fl ox mice ( Fig. 3F , upper panel). This difference was most obvious in the VLDL-containing plasma part after fractionation by ultracentrifugation ( Fig. 3F , middle panel). However, ApoC3 appeared to accumulate in liver extracts of Arfrp1 liv Ϫ / Ϫ mice because the levels were signifi cantly higher in total liver extracts of these mice ( Fig. 3F , lower panel). Nevertheless, the mRNA of Apoc3 was unaltered (data not shown). IHC analysis of livers indicated an intracellular accumulation of ApoC3 in Arfrp1 liv Ϫ / Ϫ mice which was colocalized with ApoB48/100 ( Fig. 3B ).
Because ARFRP1 is known to act at the trans -Golgi ( 8 ), we next isolated intracellular fractions (ER and Golgi compartments) from liver homogenates of Arfrp1 fl ox / fl ox and Arfrp1 liv Ϫ / Ϫ mice by gradient centrifugation and detected the apoprotein and triglyceride concentrations. As expected, ARFRP1 associated with the trans -Golgi fractions that were syntaxin-6 positive ( Fig. 4A ). In the ER (positive for calnexin), levels of ApoB48 were not different between the genotypes. However, in the Golgi fractions ( trans-as well as cis/median-Golgi) of Arfrp1 liv Ϫ / Ϫ livers, we detected more ApoB48 than in the corresponding fractions from Arfrp1 fl ox/fl ox livers ( Fig. 4B ). Comparing ApoB48 levels of all fractions from Arfrp1 fl ox / fl ox and Arfrp1 liv Ϫ / Ϫ livers, it appears that the latter contain more ApoB48 (supplementary Fig. III). Measurement of the triglycerides in the intracellular fractions showed only moderate differences. In samples of Arfrp1 liv Ϫ / Ϫ liver the triglyceride concentrations were higher in the cis/median -Golgi (positive for p115) but lower in the ER fractions, whereas the trans -Golgi fractions exhibited similarly low levels in both genotypes ( Fig. 4C ).
Reduced hepatic release of the HDL-associated ApoA1 in Arfrp1 liv ؊ / ؊ mice ApoA1, as the major protein component of HDLs, is synthesized in the liver and the intestine ( 25 ). The amount of ApoA1 in the plasma of Arfrp1 liv Ϫ / Ϫ mice was reduced as determined by WB and immunonephelometric assay ( Fig. 5 , upper and middle panels). However, Apoa1 mRNA was unaltered between genotypes (data not shown), whereas total ApoA1 was markedly reduced in Arfrp1 liv Ϫ / Ϫ liver lysates ( Fig. 5 , lower panel)

Accumulation of lipids and altered lipid storage in Arfrp1 liv ؊ / ؊ livers after fasting
The isolation of triglycerides from the livers of fasted mice revealed a higher triglyceride concentration in Arfrp1 liv Ϫ / Ϫ mice ( Fig. 6B ), which was also obvious after Oil-Red O staining of liver sections ( Fig. 6A ). This elevated lipid storage in Arfrp1 liv Ϫ / Ϫ livers was accompanied by an increased hepatic expression of PLIN2 and localization of PLIN2 to the lipid droplets (supplementary Fig. I). Furthermore, these stainings indicated an alteration in lipid droplet size between the genotypes, whereas the total number of lipid droplets was unchanged between genotypes ( Fig. 6C ). Morphometric analysis of lipid droplet size in liver sections of fasted Arfrp1 liv Ϫ / Ϫ and control mice indicated a trend toward larger lipid droplets in Arfrp1 liv Ϫ / Ϫ livers ( Fig. 6D ).
glucose, isolated the lipids, and determined 14 C incorporated into triglycerides. As shown in Fig. 7B , triglyceride synthesis was not different between control and Arfrp1 -depleted hepatocytes. We furthermore studied the expression of enzymes involved in triglyceride synthesis by quantitative real-time PCR. DGAT2 catalyzes the reesterifi cation of

Hepatic lipid synthesis and VLDL lipidation in the ER in Arfrp1 fl ox / fl ox and Arfrp1 liv ؊ / ؊ mice
In order to test whether ARFRP1 infl uences the hepatic triglyceride synthesis before the lipids are incorporated into VLDL particles, we treated isolated hepatocytes transfected with scrambled or Arfrp1 -specifi c siRNA with 14 C- of ApoC3 concentration in plasma and an accumulation of VLDL particles in the Golgi. As a consequence, the lipid content and ApoC3 levels within the liver were higher in the absence of ARFRP1.
Under fasting conditions, the liver is highly important for the redistribution of lipids delivered from adipose tissue after lipolysis to peripheral organs by synthesis and secretion of VLDLs. VLDL assembly is initiated in the ER and particles are thought to be further lipidated and modifi ed in the Golgi ( 3 ). To avoid premature degradation of newly synthesized ApoB48/100, initial lipidation by MTP occurs in the ER and produces lipid-poor VLDL particles, also designated as VLDL2 ( 28 ), whereas unlipidated ApoB48/100 is directly degraded ( 28 ). However, MTP appears not to be required for fi nal bulk lipidation of VLDL particles ( 30 ). As the amount, the distribution, and the activity of MTP as well as ApoB48/100 expression were not reduced in Arfrp1 liv Ϫ / Ϫ livers, initial synthesis and lipidation of VLDL2 appeared unaffected. However, elevated plasma levels of ApoB100 and reduced triglycerides in the VLDL fraction of Arfrp1 liv Ϫ / Ϫ mice could indicate a premature release of VLDL2. ApoC3 was described to directly interact with lipids via its helix 5, thereby facilitating triglyceride incorporation into VLDLs independently of MTP activity ( 31 ). This fi nding is supported by the fact that the human ApoC3-K58E mutation located in the lipid binding domain of ApoC3 results in a marked reduction of triglycerides and VLDLs in plasma ( 32 ). Because ApoC3 concen tration in the plasma of Arfrp1 liv Ϫ / Ϫ mice was markedly reduced but accumulated in the hepatocytes, we hypothesized that ARFRP1 and its downstream effectors are needed for an triglycerides ( 26 ) and its mRNA was signifi cantly lower in livers from Arfrp1 liv Ϫ / Ϫ mice ( Fig. 7A , upper panel). De novo fatty acid synthesis appeared to be unaltered, as there was no genotype-specifi c difference in the amount of Fasn mRNA ( 27 ) ( Fig. 7A , lower panel). The initial lipidation of preVLDL particles involves the transfer of triglycerides onto ApoB48/100 particles at the ER and is facilitated by MTP ( 28 ). The measurement of MTP protein revealed no alterations in total liver lysates (supplementary Fig. IIA, upper panel) and in the Golgi and ER fractions between the genotypes (supplementary Fig. IIB). In addition, the activity of MTP in liver lysates of fasted mice indicated no infl uence of the deletion of hepatic Arfrp1 on transfer of triglycerides to preVLDL particles in the ER (supplementary Fig. IIA, lower panel). Furthermore, initial lipidation of preVLDLs can occur by lipolysis of intracellular lipid droplets. The hepatic hydrolase carboxylesterase 1/esterase-x mediates this step and knockout mice exhibit an elevated VLDL secretion and fatty liver ( 29 ). However, the total amount of carboxylesterase 1 in liver lysates was unaltered between the genotypes (supplementary Fig. IIC).

DISCUSSION
The liver plays a central role for an optimal lipid distribution in the whole body. By specifi cally deleting the GTPase ARFRP1 in hepatocytes, the impact of trans -Golgi action on the VLDL lipidation as well as on the assembly of apoproteins to the VLDL particles was demonstrated. In the absence of ARFRP1, the triglyceride content of VLDL was markedly reduced, associated with a striking reduction ApoB48/100 appears to be due to the smaller size of VLDL particles. This is also refl ected by a lower triglyceride concentration of the HDL-containing fraction that might rather be a contamination with smaller VLDL particles because it is enriched with ApoB48/100 (as shown in Fig. 1E ). However, we cannot exclude that the accumulation of ApoB48/100 is the result of an impaired hepatic clearance of VLDL remnants or LDL remnants which could be due to an impaired receptor targeting to the plasma membrane in Arfrp1 liv Ϫ / Ϫ hepatocytes based on an altered protein traffi cking through the Golgi. Thus, we assume that the tri glyceride secretion by Arfrp1 liv Ϫ / Ϫ livers is impaired according to two alterations: 1 ) by a limited lipidation of VLDL in the Golgi; and 2 ) by an impaired ApoC3 release. The overall defect is visible in the fasted state, in which the triglyceride concentration of Arfrp1 liv Ϫ / Ϫ mice was reduced by 45%. This effect was smaller (30%) under conditions that blocked the effect of ApoC3 on LPL activity. The release of limited lipidated VLDLs in Arfrp1 liv Ϫ / Ϫ mice secondarily resulted in an accumulation of triglycerides appropriate sorting of ApoC3 to the lumenal lipid droplets that are utilized as a lipid precursor for VLDL assembly. Furthermore, it was already shown by Asp et al. ( 33 ) that the Golgi-associated GTPase ARF1, which is involved in retrograde Golgi to ER and intra-Golgi traffi cking, is required for fi nal lipidation of VLDL2 particles to form mature lipid-rich VLDL particles (also designated as VLDL1), indicating the important role of the Golgi apparatus for MTP-independent lipid transfer. Furthermore, the deletion of sortilin, an intracellular sorting receptor for ApoB48/100 which is located at the trans -Golgi, resulted in a reduced secretion of ApoB48/100 ( 34 ). These studies further support the assumption that the trans -Golgi compartment is essential for hepatic lipoprotein metabolism . The elevated plasma ApoB48/100 in Arfrp1 liv Ϫ / Ϫ mice indicated on the one side that the general release of ApoB48/100-containing particles from Arfrp1 liv Ϫ / Ϫ livers is possible. On the other side, these particles, which each carry one ApoB48/100 molecule, carry less triglyceride load. Therefore, the increase in plasma liv Ϫ / Ϫ and control mice. Data are expressed as percentage of lipid droplets in categories of m 2 (n = 9-11 for liver triglycerides, n = 6 for lipid droplet distribution). * P < 0.05, *** P < 0.001. induced by prolonged fasting periods could be one explanation. Former studies in rats have demonstrated that thyroid hormone treatment ( 35 ) resulted in reduced ApoB100 synthesis. Furthermore, a prolonged fasting period (48 h) and a subsequent high carbohydrate diet resulted in complete loss of ApoB100, which was accounted for by an increase in the proportion of edited Apob mRNA ( 36 ). The accumulation of triglycerides in the liver of Arfrp1 liv Ϫ / Ϫ mice appears to be accompanied by elevated levels of PLIN2, most likely a secondary effect. However, higher expression of PLIN2 in hepatic cell lines inhibits the secretion of VLDL1 ( 37 ) and could therefore also infl uence VLDL lipidation in Arfrp1 liv Ϫ / Ϫ mice. The lack of ARFRP1 does not modulate triglyceride synthesis because in vitro triglyceride formation in isolated hepatocytes was not affected after suppression of Arfrp1 ( Fig. 7B ). The Dgat2 mRNA was signifi cantly lower in livers from Arfrp1 liv Ϫ / Ϫ mice, indicating that the reesterifi cation of triglycerides could be compensatory, suppressed most likely due to elevated accumulation of triglycerides in these livers ( Fig. 6 ). Another possible explanation for reduced Dgat2 mRNA is its regulation by low levels of plasma glucose and fasting ( 38 ). In fed Arfrp1 liv Ϫ / Ϫ mice, blood glucose levels are significantly lower in comparison to control mice ( 11 ) and could therefore be causal. Besides VLDL lipidation and therefore hepatic triglyceride release, fasting plasma triglyceride concentrations were further infl uenced by peripheral clearance mediated by lipoprotein lipase. ApoC3 is a negative regulator of lipoprotein lipase activity ( 24 ). Therefore, the reduced levels of ApoC3 in Arfrp1 liv Ϫ / Ϫ plasma could lead to a higher activity of lipoprotein lipase which further amplifi es the lowering of plasma triglycerides in fasted Arfrp1 liv Ϫ / Ϫ mice ( Fig. 3 ).
Another phenotype visible in Arfrp1 liv Ϫ / Ϫ mice is the reduced ApoA1 concentration in the plasma. ApoA1 that is synthesized by the liver (and to a lesser extent by the intestine) is the major HDL protein, and both tissues play prime roles in mediating reverse cholesterol transport, the transfer of cholesterol from peripheral tissues, such as arterial wall cells, to the liver for excretion ( 39,40 ). The impaired release of ApoA1 in the absence of ARFRP1 might be responsible for the slightly reduced cholesterol concentration in the total plasma and in the HDL-containing fraction of fasted mice.
Besides the liver, the intestine is the other tissue that produces circulating triglyceride-carrying lipoproteins ( 41 ). In a recent study, we showed that chylomicron formation, lipidation as well as apolipoprotein assembly, is dependent on the trans -Golgi and ARFRP1 ( 13 ). ARFRP1 is required for the correct recruitment and action of other GTPases (ARL1 and Rab proteins) that interact with the scaffolding protein golgin-245. Suppression of each transcript (ARFRP1, ARL1, Golgin-245, Rab2) in Caco-2 cells resulted in a marked reduction of chylomicron release. Thus, it can be speculated that impaired structural capacities at the trans -Golgi are responsible for the limited lipidation and maturation of not only chylomicrons, but also of VLDL and HDL particles. The comparison of the phenotype of the intestinespecifi c with the liver-specifi c Arfrp1 knockout identifi es strong parallels regarding lipid and lipoprotein metabolism. and of ApoB48/100 and ApoC3 under fasting conditions, indicating that ARFRP1 is needed not only for an appropriate lipidation of VLDL particles but also for their release into the plasma. In the absence of ARFRP1, the transport of VLDL particles through the Golgi does not work properly. This was shown by an accumulation of ApoB48/100 and triglycerides in the Golgi ( Fig. 4 ). We currently cannot explain why the levels of ApoB48 in the intracellular fractions are much higher than those of ApoB100, an effect that is also visible in total liver homogenates, while the ratio of both proteins was about equal in plasma samples. Differences in Apob mRNA editing In both mouse models the release of ApoB48/100-containing triglyceride-rich particles is impaired and results in lower plasma triglyceride levels and elevated plasma ApoB48/100. In addition, the release of ApoA1 is impaired leading to reduced cholesterol concentration in the plasma of knockout mice. Therefore, the lipidation and the assembly of lipoproteins with apolipoproteins is dependent on ARFRP1 actions at the trans -Golgi, a pathway that appears to be conserved in the liver and intestine.