Absence of adipose differentiation related protein upregulates hepatic VLDL secretion, relieves hepatosteatosis, and improves whole body insulin resistance in leptin-deficient mice.

We previously showed that adipose differentiation related protein (Adfp)-deficient mice display a 60% reduction in hepatic triglyceride (TG) content. In this study, we investigated the role of ADFP in lipid and glucose homeostasis in a genetic obesity model, Lep(ob/ob) mice. We bred Adfp(-/-) mice with Lep(ob/ob) mice to create Lep(ob/ob)/Adfp(-/-) and Lep(ob/ob)/Adfp(+/+) mice and analyzed the hepatic lipids, lipid droplet (LD) morphology, LD protein composition and distribution, lipogenic gene expression, and VLDL secretion, as well as insulin sensitivity of the two groups of mice. Compared with Lep(ob/ob)/Adfp(+/+) mice, Lep(ob/ob)/Adfp(-/-) mice displayed an increased VLDL secretion rate, a 25% reduction in hepatic TG associated with improvement in fatty liver grossly and microscopically with a change of the size of LDs in a proportion of the hepatocytes and a redistribution of major LD-associated proteins from the cytoplasmic compartment to the LD surface. There was no detectable change in lipogenic gene expression. Lep(ob/ob)/Adfp(-/-) mice also had improved glucose tolerance and insulin sensitivity in both liver and muscle. The alteration of LD size in the liver of Lep(ob/ob)/Adfp(-/-) mice despite the relocation of other LDPs to the LD indicates a nonredundant role for ADFP in determining the size and distribution of hepatic LDs.

Nonalcoholic fatty liver disease (NAFLD) has reached epidemic proportions in the US ( 1 ) and is especially common in patients with obesity and type 2 diabetes. NAFLD is characterized by the accumulation of triglyceride (TG) and the appearance of lipid droplets (LDs) in hepatocytes. LDs encompass a neutral lipid core, mainly of TG and cholesteryl ester, enveloped by a phospholipid monolayer ( 2 ). Also on the surface are LD proteins (LDPs) that consist of a family of proteins called PAT-domain proteins ( 3-5; see below ), along with other proteins with varied functions, such as lipid metabolic enzymes and proteins involved in vesicle traffi cking (6)(7)(8). The presence of these specialized proteins suggests that LDs are not merely passive storage receptacles of cytosolic lipids but highly organized organelles involved in cellular metabolism (9)(10)(11)(12).
The PAT-domain proteins, named after Perilipin (Plin1), adipose differentiation related protein (Adfp; Plin2), and Tip47 (Plin3), are the predominant LDPs. A new Plinbased nomenclature (as shown in parenthesis when it fi rst appears in the text) for these LDPs was recently proposed ( 13 ), but we have followed the conventional nomenclature in this paper. Whereas perilipin expression is confi ned to adipose and steroidogenic tissues, Adfp and Tip47 are present in multiple tissues. Perilipin-knock-out mice are almost completely devoid of body fat because of high basal lipolysis; they are resistant to diet-induced as well as genetic obesity, pinpointing the role of perilipins in the regulation of lipolysis ( 14,15 ). Adfp ablation results in a marked reduction in hepatic TG ( 16 ), reduced cholesterol Abstract We previously showed that adipose differentiation related protein ( Adfp )-defi cient mice display a 60% reduction in hepatic triglyceride (TG) content. In this study, we investigated the role of ADFP in lipid and glucose homeostasis in a genetic obesity model, Lep ob/ob mice. We bred Supplementary key words Adrp • lipid droplet • Tip47 • ob/ob • fatty liver MRI Whole body compositions of mice were analyzed by EchoMRI (Echo Medical Systems) according to the manufacturer's instructions.

Plasma chemistry analysis
We collected blood from the orbital plexus under isofl urane (Vedco) anesthesia. Plasma was frozen in aliquots at Ϫ 20°C or used immediately after collection. We used enzymatic kits for determination of serum NEFA (Wako), glycerol (Sigma Aldrich), cholesterol, and TG (Infi nity). Plasma glucose was monitored by glucometer and insulin was measured by ELISA (Mercodia).

Liver lipid analysis
We homogenized 200 mg liver tissues in 2 ml of PBS, extracted lipids from these homogenate according to Bligh and Dyer ( 27 ), and fractionated different lipid species by one dimensional TLC (silica Gel-60, Analtech), using petroleum ether/ether/glacial acetic acid (85:25:1). Lipids were visualized by incubating the TLC plate in saturated iodine chamber. For quantitative analysis of hepatic lipids, we followed the method of Schartz and Wolins ( 28 ) by fi rst partitioning lipids from water-soluble components of the tissue extracts with organic extraction followed by colorimetric enzymatic detection kits for TG (Infi nity), cholesterol, and cholesterol ester (BioVision).

Determination of rate of VLDL secretion in vivo
We quantifi ed the rate of VLDL secretion in vivo by injecting intraperitoneally Pluronic F-127 (BASF Corporation; 2 mg/g body weight in PBS), a lipoprotein lipase inhibitor ( 29 ), and monitored the plasma TG before, and 1, 2, and 3 h afterwards using an enzymatic kit (Infi nity).

Quantitative RT-PCR
We isolated RNA using an RNeasy mini-kit (Qiagen) and treated all samples (10 µg) with RNase-free DNase-I before using Superscript-II First Strand kit (Invitrogen) and oligo-dT primer to synthesize fi rst strand cDNA (in 100 µL reaction volume). Two micoliters of these samples were used for quantitative PCR using iQ-SYBR Green Supermix (Bio Rad) under MX3000P system (Stratagene). Primer sequences used in this study are listed in supplementary Table I . We used geNorm algorithm ( 30 ) to determine the most stably expressed housekeeping reference genes as controls.

Western blot analysis
For Western blotting, equivalent amounts of protein homogenate were resolved by 4-15% SDS-PAGE, transferred to nylon membrane, and probed with specifi c antibodies for visualization by enhanced chemiluminescence (SuperSignal kit, Pierce). Semiquantitative protein analysis on Western blots was done using Image J Software.

Histology and immunohistochemistry
We removed a piece of liver and fi xed it in formalin overnight before dehydration and paraffi n embedding. Five micron sections were stained with hematoxylin and eosin or further processed for immunofl uorescence staining by simultaneously labeling proteins with specifi c primary antibodies (listed in "Chemicals and Reagents") and subsequently labeled with fl uorescence-tagged secondary antibodies as described previously ( 31 ). Images were analyzed under a Zeiss Axioplan-2 Imaging System.
The reduction of hepatic TG in Adfp-null mice presents us with an opportunity to study the role of Adfp in the development of NAFLD and its metabolic consequences. Lean C57Bl/6J mice, however, do not spontaneously develop NAFLD, so we decided to use the obese mouse model of NAFLD. Leptin -defi cient ( Lep ob/ob /Adfp +/+ ) mice are a model for NAFLD and type-2 diabetes. They are hyperphagic due to a mutation in the Leptin gene whose protein product is normally secreted from the adipose tissue to communicate with the brain as a satiety signal ( 23 ). Loss of functional Leptin causes Lep ob/ob /Adfp +/+ mice to become obese. Although Leptin also regulates other bodily functions ( 23 ), the severe obesity, hyperglycemia, hyperinsulinemia, and fatty liver development occur early in these mice. We hypothesized that absence of Adfp in the Leptindefi cient mice would reduce the hepatic TG and ameliorate hepatic fat accumulation and its metabolic consequences without altering their body weight and obesity. To test this hypothesis, we bred Adfp Ϫ / Ϫ mice into Lep ob/ob /Adfp +/+ mice to produce Lep ob/ob /Adfp Ϫ / Ϫ . We examined in these animals how the presence and absence of ADFP modulates cellular lipid homeostasis at the transcript, protein, and lipid substrate levels as well as the distribution/expression of other LDPs. Finally, we took advantage of this mouse model to study the relationship between fatty liver and insulin resistance ( 24,25 ).

Chemicals and reagents
All chemicals were purchased from Sigma Chemical, except lipid standards (Avanti Polar Lipid). Primary antibodies were purchased from Chemicon (GAPDH) and Progen (Guinea Pig anti-ADFP, GP40 mN1). Anti-ADFP, TIP47, and LSDP5 were generated through Strategic Diagnostics by immunizing rabbits with respective 6His-tagged, full-length recombinant proteins. Rabbit anti-ABHD5 was a gift from Dr. Takashi Osumi at University of Hyogo, Japan. Alexa 488-tagged anti-rabbit, anti-guinea pig, Alexa 555-tagged anti-rabbit, and anti-guinea pig antisera were obtained from Invitrogen. Mouse microsomal TG transfer protein (MTTP) antibody was generated previously ( 26 ).

Mice
Adfp Ϫ / Ϫ mice ( 16 ) were crossed to Lep ob/ob mice to generate Lep ob/ob /Adfp Ϫ / Ϫ mice. Lep ob/ob / Adfp +/+ mice were used as controls. Mice were maintained in a temperature-controlled facility with 12 h light/dark cycles and free access to regular chow and water. Male mice of 8-12 weeks old were used throughout this study unless otherwise indicated, and all were in the C57BL/6J genetic background. All studies were conducted according to the "Principles of Laboratory Animal Care" (NIH publication No. 85023, revised 1985) and the guidelines of the IACUC of Baylor College of Medicine. the liver in Lep ob/ob /Adfp +/+ mice showed considerable pallor, consistent with fatty liver ( Fig. 1A ). The liver of the Lep ob/ob /Adfp Ϫ / Ϫ mice, however, displayed a relatively normal reddish hue ( Fig. 1A ). There was no difference in liver weight. Essentially all hepatocytes of Lep ob/ob /Adfp +/+ mice were tightly packed with numerous moderate-sized LDs taking up the entire cytoplasmic space. In contrast, about one-half of the hepatocytes of Lep ob/ob /Adfp Ϫ / Ϫ mice were notable for the presence of single (unilocular) or few (oligo-locular), huge-sized LDs with preservation of a small amount of cytoplasm ( Fig. 1B ). The Lep ob/ob /Adfp Ϫ / Ϫ cells that did not contain these huge LDs contained far fewer LDs that were generally much smaller than those in hepatocytes of Lep ob/ob /Adfp +/+ mice. We extracted total lipids from the liver and analyzed them by TLC, which showed that hepatic total TG was reduced in the Lep ob/ob /Adfp Ϫ / Ϫ compared with the Lep ob/ob / Adfp +/+ mice. We used enzymatic kits to determine hepatic TG content, which was found to be reduced by ‫ف‬ 25% in Lep ob/ob /Adfp Ϫ / Ϫ ( Fig. 1D ). There was a ‫ف‬ 50% increase in hepatic cholesterol in the Lep ob/ob /Adfp Ϫ / Ϫ mice ( Fig. 1D ).
Although cholesteryl ester was also increased in the Lep ob/ob / Adfp Ϫ / Ϫ mice, the difference did not reach statistical signifi cance.

Lep ob/ob /Adfp ؊ / ؊ liver displays increased VLDL secretion and MTTP expression
We measured VLDL secretion using the lipase inhibitor, Pluronic F-127 ( 29 ), and found a signifi cantly increased rate of VLDL secretion in Lep ob/ob /Adfp Fig. 2A ). We next quantifi ed the level of MTTP, the rate-limiting enzyme for VLDL assembly and secretion by Western blot analysis and found a markedly increased level of hepatic MTTP protein in Lep ob/ob /Adfp Ϫ / Ϫ compared with Lep ob/ob /Adfp +/+ mice ( Fig. 2B ). To determine if the increased MTTP expression was also observed at the transcript level, we measured the relative level of MTTP mRNA by RT-PCR and found it to be the same in mice of the two genotypes (supplementary Fig. II), indicating that the MTTP protein overexpression in Lep ob/ob /Adfp Ϫ / Ϫ mice occurred at a posttranscriptional level. By fast performance liquid chromatography analysis, we found that both HDL (cholesterol) and VLDL (TG) were increased in Lep ob/ob / Adfp Ϫ / Ϫ mice compared with Lep ob/ob /Adfp +/+ mice ( Fig.  2C, D ), though total plasma TG and cholesterol were not different between the two groups ( Table 1 ).

Lep ob/ob /Adfp ؊ / ؊ mice display improved glucose tolerance compared with Lep ob/ob /Adfp +/+ mice
We examined the degree of glucose intolerance in these mice by oral GTT. In response to oral glucose, plasma glucose increased and peaked at 15 min in both Lep ob/ob / Adfp Ϫ / Ϫ and Lep ob/ob /Adfp +/+ and tapered off in the next 2 h ( Fig. 3A ). During the test, the absolute glucose levels were signifi cantly lower in every sampling point (except at 30 min) in Lep ob/ob /Adfp Ϫ / Ϫ as compared with Lep ob/ob /Adfp +/+ mice ( Fig. 3A ). The corresponding insulin levels were not different between the two groups ( Fig. 3B ). mice were gavaged with glucose (1.5 g/kg of body weight) after 4 h fasting. For ITT, fasted mice were injected (i.p.) with insulin (5 U/kg body weight, Humulin R; Eli Lily). Experiments were performed between 10 AM and 12 PM. Blood was taken before and 15, 30, 60, and 120 min after treatment for determination of glucose and insulin levels.

Isolation of LD fraction
We removed 200 mg of liver tissue, which was cut into tiny pieces, and soaked in 3.5 ml of cold 250-STMDPS (250 mM sucrose, 50 mM Tris-HCL, 5 mM MgCl 2 , 1 mM DTT, 0.5 mM PMSF, and Spermidine 25µg/ml) buffer ( 33 ) in a nitrogen bomb (Parr Instrument). Tissue was homogenized and LD fraction isolated as described by Liu et al. ( 34 ). Briefl y, total homogenate was centrifuged at 10,000 g for 10 min to remove debris. The supernatant was mixed well and an aliquot was saved for total liver protein analysis, and the rest was subjected to further ultracentrifugation at 45,000 g for 1 h to separate and recover LD fraction fl oated on the top, the aqueous cytosolic fraction in the middle, and the microsomal pellet at the bottom.

Statistical analysis
Student's t -test was used for statistical analysis. The Mann-Whitney test was used when sample size was small (n < 5). Differences were considered signifi cant when P < 0.05.

Altered abundance of LD-associated proteins in different cellular compartments but no change in their mRNA levels in the absence of ADFP
We previously showed that Adfp defi ciency in vivo does not elicit a compensatory upregulation of other LDPs in primary mouse fi broblasts ( 16 ). However, we found in clonal cells derived from embryonic fi broblasts upregulated Tip47 gene expression in Adfp Ϫ / Ϫ cells upon lipid loading ( 35 ). Therefore, we examined whether LDPs and their transcripts were altered in the liver of the Lep ob/ob / Adfp Ϫ / Ϫ mice. As many of the LDPs are localized both in the cytosol and LD, we determined by Western blotting the relative abundance of these proteins in total, cytosolic, and isolated LD fractions. PLIN, CIDE-A, -B and -C, ATGL, ADFP, TIP47, LSDP5, and ABHD5 were examined; however, only the last four were detectable in the liver by Western blotting ( Fig. 5 ). These blots allowed us to compare the relative protein abundance in Lep ob/ob /Adfp Ϫ / Ϫ and Lep ob/ob /Adfp +/+ liver from a particular subcellular fraction. We note that it would be inappropriate to compare the protein bands across different blots, because different antibodies and exposure times were used and protein recovery in individual fractions was not identical.
In the total protein extracts, we did not fi nd any difference in the band intensity in TIP47. However, LSDP5 and ABHD5 were more abundant in the liver of Lep ob/ob /Adfp

Lep ob/ob /Adfp
؊ / ؊ and Lep ob/ob /Adfp +/+ mice display a similar glucose response during ITT We performed an ITT as an initial test of insulin sensitivity. As shown in Fig. 3C , Lep ob/ob /Adfp +/+ mice responded to the insulin injection with a steeper drop in plasma glucose than the Lep ob/ob /Adfp Ϫ / Ϫ mice ( Fig. 3C ), though there was no difference between the two groups when we expressed the results as percent drop in plasma glucose ( Fig. 3D ).

Hyperinsulinemic-euglycemic clamp reveals improved insulin action in Lep ob/ob /Adfp
As a more defi nitive test of insulin sensitivity in vivo, we performed a hyperinsulinemic-euglycemic clamp on the mice. Basal hepatic glucose production rate was similar in Lep ob/ob /Adfp Ϫ / Ϫ and Lep ob/ob /Adfp +/+ mice ( Fig. 4A ). During the clamp, the Lep ob/ob /Adfp Ϫ / Ϫ mice required a much higher glucose infusion rate to maintain euglycemia ( Fig.  4B ), indicating enhanced insulin sensitivity. They also exhibited an increased glucose disposal rate ( Fig. 4C ) compared with Lep ob/ob /Adfp +/+ mice, indicating a higher rate of peripheral glucose uptake in the Lep ob/ob /Adfp Ϫ / Ϫ mice in response to insulin infusion, which also resulted in a better suppression of hepatic glucose production in Lep ob/ob / Adfp Ϫ / Ϫ compared with Lep ob/ob /Adfp +/+ mice ( Fig. 4D ).

Gene transcript levels of metabolic enzymes
To determine if the reduced hepatic TG in Lep ob/ob / Adfp Ϫ / Ϫ mice was a result of changes in gene expression involved in carbohydrate or lipid metabolic pathways, we isolated RNA from the liver of these mice and performed quantitative RT-PCR analysis. We examined the following genes involved in lipogenesis (Srebp-1c, Acc1, Fasn, and of Lep ob/ob /Adfp Ϫ / Ϫ than Lep ob/ob /Adfp +/+ liver ( Fig. 5C ). GAPDH was not present in the LD fraction ( Fig. 5C ).
To determine if the change in protein abundance was the result of a change in mRNA expression, we measured the hepatic mRNA level of Tip47, Lsdp5, and Abhd5 genes by quantitative RT-PCR and found that the mRNA level for each of these proteins was similar between Lep ob/ob /Adfp Ϫ / Ϫ and Lep ob/ob /Adfp +/+ mice ( Fig. 5D ).

LDs in Lep ob/ob /Adfp ؊ / ؊ hepatocytes are enriched in TIP47
To further explore the difference in protein expression in some of the LDPs detected by Western blotting ( Fig. 5 ), we performed immunofl uorescence microscopy to determine the physical localization of ADFP, TIP47, LSDP5, and ABHD5 proteins in liver sections of Lep ob/ob /Adfp Ϫ / Ϫ and Lep ob/ob /Adfp +/+ mice. Unfortunately, in our hands, the antibodies against LSDP5 and ABHD5 did not work for immunofl uorescence staining, and we had to confi ne the immuno-morphological analysis to TIP47.
We isolated LD fractions from the same liver samples and reconstituted the samples directly into protein gel loading buffer, because conventional aqueous buffer would not dissolve the hydrophobic proteins that co-purifi ed with the lipids. Among the four proteins, we found an abundant amount of ADFP in Lep ob/ob /Adfp +/+ liver extracts, which was absent in Lep ob/ob /Adfp Ϫ / Ϫ liver extracts ( Fig. 5C ).
The amount of TIP47 in the LD fraction was signifi cantly upregulated in Lep ob/ob /Adfp Ϫ / Ϫ compared with Lep ob/ob / Adfp +/+ liver ( Fig. 5C ). Similarly, we found that LSDP5 and ABHD5 were also much more abundant in the LD fraction absence of detectable changes in lipogenesis, lipid uptake, utilization, or transport ( 16 ). Importantly, we detected in these mice an increase in liver MTTP protein without any changes in mRNA expression. However, using Triton WR1339 to inhibit the vascular lipases, we previously found that the rate of VLDL secretion from the liver of Adfp Ϫ / Ϫ mice was similar to that of Adfp +/+ mice. Recently, another lipase inhibitor called Pluronic F-127, also known as poloxamer 407, was reported to be less toxic and a better lipase inhibitor than Triton WR1339 for quantifying VLDL secretion in rodents in vivo ( 29 ). We therefore repeated the measurement using this new reagent and found that, indeed, Adfp Ϫ / Ϫ mice on a wild-type C57BL/6J background have a VLDL secretion rate signifi cantly higher than that in Adfp +/+ mice (supplementary Fig. III) despite Adfp Ϫ / Ϫ mice having a markedly reduced cytosolic TG content. Thus, we conclude that unbridled VLDL oversecretion occurs in Adfp Ϫ / Ϫ mice (in C57BL/6J genetic background), contributing to TG depletion in the liver.
We found in this investigation that, like wild-type mice, absence of ADFP in Lep ob/ob mice leads to increased hepatic MTTP expression at the protein but not mRNA level, as well as reduced TG content. We reasoned that a similar and diffusely distributed in the cytoplasm ( Fig. 6 , top panels). In contrast, in Lep ob/ob /Adfp Ϫ / Ϫ hepatocytes, ADFP was not detectable (because of the Adfp gene knock-out), while Alexa555-labeled TIP47 protein prominently lined the surface of LDs, in both large unilocular and smaller LDs ( Fig. 6 , bottom panels). This physical appearance is consistent with the results obtained from Western blot analysis.

DISCUSSION
By breeding Adfp Ϫ / Ϫ into Lep ob/ob mice, we produced a mouse model that exhibits attenuated hepatosteatosis in the presence of severe obesity compared with Lep ob/ob / Adfp +/+ mice. This improvement in fatty liver was associated with substantial changes in LD size and LDP distribution. The loss of ADFP also modulated the concentration and distribution of select LDPs, while it improved glucose tolerance and insulin insensitivity that plague Lep ob/ob mice.
We previously reported that Adfp defi ciency in wild-type (lean) mice had an ‫ف‬ 60% TG reduction in the liver in the fp -defi cient (deletion of exons 2 and 3) mice. This N-terminal truncated ADFP protein seemed to be able to substitute for the function of the dominant form of ADFP in the mammary gland ( 18 ), but not in the retina function ( 19 ), perhaps due to an extremely low level of expression in the latter tissue. The N-terminal truncated form of ADFP was not sensitive to proteasome degradation as the full-length ADFP was ( 36 ). It also did not prevent TIP47 from accessing to the LD surface. In agreement with the previous study ( 18 ), using a combination of Western, Northern, and RT-PCR analyses, we did not detect the expression of the N-terminal truncated ADFP protein or mRNA in the liver of wild-type or knock-out mice (supplementary Fig. IV). The short truncated form seems to be specifi c to the mammary gland, and RT-PCR data suggests that an alternative transcription initiation site may reside in exon 4 (supplementary Fig. IVC).
Using an Adfp antisense oligonucleotide, Imai et al. ( 37 ) knocked down ADFP expression in the liver of Lep ob/ob mice and observed lowering of hepatic TG in treated mice. However, the mechanism seemed to be quite different from that in the knock-out model, because they observed signifi cantly downregulated lipogenic gene expression in the liver of treated mice, contrary to our results in Adfp Ϫ / Ϫ mice with wild-type or Lep ob/ob background. In addition, they also observed a downregulation of VLDL secretion ( 37 ) instead of the upregulation that we found in both wild-type and Lep ob/ob mice that lack ADFP (supplementary Fig. III; Fig. 2A ).
Hepatosteatosis is commonly associated with insulin resistance ( 38,39 ). It has been suggested that obesity leads to chronic infl ammation in adipose tissue, which releases pro-infl ammatory cytokines that trigger insulin resistance that may somehow contribute to hepatosteatosis ( 40 ). Adfp defi ciency in either wild-type ( 16 ) or Lep ob/ob background (current study) is not associated with a change in adipose or lean body mass (supplementary Fig. IB). It does not alter adipocyte differentiation or lipolysis in vitro or in mechanism likely underlies the improvement in the hepatosteatosis in these animals, a hypothesis that was confi rmed by direct measurement of VLDL secretion using Pluronic F-127 ( Fig. 2 ). We think that this is a key fi nding that underscores the important function of ADFP in lipid economy and homeostasis in the liver.
In addition to the decreased hepatic TG, we also found that the hepatic cholesterol was increased in the liver of There is evidence that tissues other than the liver also display little redundancy in Adfp function as a key protein for effi cient LD formation. We have demonstrated that macrophages export cholesterol more readily in Adfp defi ciency, making them more resistant to foam cell formation. Furthermore, apolipoprotein E-defi cient mice with ADFP defi ciency are more resistant than ADFP-replete apolipoprotein E-defi cient animals to atherosclerosis development, presumably because of the impaired foam cell formation in these animals ( 17 ). In our two previous studies ( 16,17 ) and in the current investigation, we did not fi nd any difference in lipid-related metabolic gene expression in tissues that are lipid-depleted because of ADFP defi ciency. In addition to TG and cholesterol homeostasis, Imanishi et al. ( 19 ) showed that Adfp defi ciency also caused abnormalities in retinyl ester homeostasis in the retinal pigmented epithelium, leading to delayed clearances of all-trans -retinal and all-trans -retinol from rod photoreceptor cells and resulting in impaired dark adaptation in the eye.
With respect to mammary gland function, Adfp-deficient mice coped well without ADFP in milk fat storage and secretion. An N-terminal truncated ADFP protein and its mRNA were found in the mammary gland ( 18 ), as well as retinal pigmented epithelium ( 19 ) of wild-type and Ad- ance and insulin sensitivity. Interestingly, hyperinsulinemiceuglycemic clamps indicate that skeletal muscle also exhibits increased glucose uptake in Lep ob/ob /Adfp Ϫ / Ϫ compared with Lep ob/ob /Adfp +/+ mice. Although muscle TG levels are the same between these two groups of mice (data not shown), the muscular tissues are totally surrounded by fat in the obese mice, which may affect the accuracy of TG quantifi cation.
We have previously examined the effect of Adfp ablation on LD size and distribution in the liver of C57BL/6 mice following high fat diet-induced hepatosteatosis ( 16 ). We found that the total number of LDs identifi able by light microscopy was signifi cantly reduced in Adfp Ϫ / Ϫ compared vivo ( 16 ) and is not associated with changes in macrophage pro-infl ammatory cytokine gene expression ( 17 ). In this study, the level of liver enzymes (AST and ALT) was not different in Lep ob/ob /Adfp Ϫ / Ϫ and Lep ob/ob /Adfp +/+ mice. It is interesting that reduced hepatic TG content of the Lep ob/ob / Adfp Ϫ / Ϫ mice per se, without apparent changes in liver enzymes, appears to be associated with improved glucose intolerance and insulin sensitivity in Lep ob/ob /Adfp Ϫ / Ϫ mice compared with the Lep ob/ob /Adfp +/+ mice. However, in addition to the quantitative change in TG, there were changes in abundance and distribution involving some of the PAT proteins when ADFP was absent. It is unclear if such changes contribute to the improvement of glucose intoler- Because several proteins have very similar molecular weights, multiple protein gels were used. The numbers denoted on the right side of each gel image indicates the gel IDs. A lighter band immediately above the ADFP band in A and B represents cross-reacting material unrelated to ADFP. The very faint ADFP bands in the A and B blots from Lep ob/ob /Adfp Ϫ / Ϫ mice were the result of slight spill-over from the neighboring Lep ob/ob /Adfp +/+ sample wells. Rabbit anti-full-length ADFP and TIP47 proteins (Strategic Diagnostics) were used in these Western blot analyses. Other antibodies were as specifi ed in "Materials and Methods." GAPDH proteins are used as loading control to estimate (using Image J software) the relative protein abundance, shown at the bottom of each panel, between Lep ob/ob /Adfp Ϫ / Ϫ and Lep ob/ob /Adfp +/+ mice. * P < 0.05; ** P < 0.01. GAPDH is not detectable in the Western blot of proteins isolated from LD fraction. Relative protein abundance in LD fraction was therefore estimated using blots that were produced from the identical amount of tissue sample without a non-LDP control (as they were not present in these blots Although it is tempting to postulate the coalescence of LDs, which has been observed in Drosophila S2 cells ( 41 ), as the origin of these huge droplets, with ADFP playing an inhibitory role in their formation, we have no direct evidence for this process happening. Importantly, in both lean and Lep ob/ob mice, lack of ADFP lowers total hepatic TG content by ‫ف‬ 25-60%. ADFP is the fi rst LDP that appears during adipocyte differentiation in vitro ( 42 ). Interestingly, the expression of Tip47 , Lsdp5 , and Abhd5 transcripts in the liver of Lep ob/ob / Adfp Ϫ / Ϫ mice was not different from that in Lep ob/ob /Adfp +/+ mice. The two types of mice also expressed similar amounts of TIP47 at the protein level. TIP47 protein was also unaltered in the absence of ADFP in the retinal pigmented epithelium ( 19 ). The concentration of LSDP5 and ABHD5 proteins was, however, increased in the liver of the Lep ob/ob / Adfp Ϫ / Ϫ mice. It has been shown that ADFP and PERILIPIN are both regulated at the posttranslational levels via degradation through the ubiquitin-mediated proteasome pathway ( 43,44 ). It appears likely that LSDP5 and ABHD5 are also regulated posttranscriptionally in response to changes in the protein composition of LDs or to lipid abundance.
TIP47 is normally distributed mainly in the cytoplasm and is redistributed to LDs when intracellular lipids become abundant ( 35 , 45 ). In this study, we showed that TIP47 redistributed to the LD surface in the liver when ADFP was absent. This redistribution of TIP47 was evident in Western blots of subcellular fractions as well as in immunofl uorescence microscopy ( Figs. 5 and 6). The situation appears analogous to THP-1 macrophages in which TIP47 relocated from the cytoplasmic compartment to the surface of LDs when ADFP protein expression was knocked down ( 46 ).
TG secretion was reduced in rat hepatoma cells when ABHD5 was knocked down by short hairpin RNA ( 47 ), and lipolysis of cytosolic stored TG was part of this process. We showed that ABHD5 is enriched in the LDs of Lep ob/ob / Adfp Ϫ / Ϫ hepatocytes, and this is accompanied by an increase of VLDL secretion in vivo. However, we did not ob-with Adfp +/+ mouse liver sections (3,247 Fig. V). Thus, absence of ADFP precludes the formation of LDs of moderate sizes (up to 15-17 m in about 10% of the hepatocytes) in lean C57BL/6J mice with diet-induced fatty liver, whereas the extreme fatty liver of Lep ob/ob /Adfp Ϫ / Ϫ mice (also in C57BL/6J genetic background) is associated with a reduced number of LDs (that are much smaller than those in Lep ob/ob /Adfp +/+ mice) in many hepatocytes. However, the absence of ADFP in this model also leads to the appearance of giant LDs in some other hepatocytes ( Fig. 1B ; supplementary Fig. V). serve a change in liver lipase activity (supplementary Fig.  VI), although TG hydrolysis assay was done using the liver homogenate that may not refl ect the in vivo TG turnover rate. TG synthesis was not altered in primary hepatocytes isolated from Lep ob/ob /Adfp Ϫ / Ϫ mice compared with that from Lep ob/ob /Adfp +/+ hepatocytes (supplementary Fig. VII). In addition, we also did not observe any signifi cant change in gene expression of key enzymes involved in lipid metabolisms (supplementary Fig. II). The detailed molecular interaction between ABHD5, ADFP, and other LDPs in hepatic TG homeostasis remains to be determined.
In conclusion, we created Lep ob/ob / Adfp Ϫ / Ϫ mice to investigate the role of ADFP and other LDPs in hepatic TG dynamics and whole body glucose homeostasis. We showed that ADFP plays a key role in modulating LD formation and size distribution. In the absence of ADFP, several other LDPs are upregulated in LDs in the liver ( Figs. 5C and 6 ). The presence in a fraction of the Lep ob/ob / Adfp Ϫ / Ϫ hepatocytes of large uni-and oligo-locular LDs in the absence of ADFP, despite the relocation of other LDPs to the LD, indicates a nonredundant role for ADFP in determining the size of hepatic LDs via mechanisms yet to be determined. The effect of ADFP expression on LD size distribution is complex, and the phenotypic consequence appears to depend in part on the severity of the fatty liver. We further showed that in mice of both wildtype C57BL/6J and Lep ob/ob background, ADFP regulates VLDL secretion in vivo. Moreover, the improvement in fatty liver in the absence of ADFP attenuates the insulin resistance associated with Lep ob/ob mice, both in the liver and in skeletal muscle, underscoring the importance of the degree of fatty liver in regulating whole body glucose homeostasis.