Post-translational Modifications of the Four Conserved Lysine Residues within the Collagenous Domain of Adiponectin Are Required for the Formation of Its High Molecular Weight Oligomeric Complex*

Adiponectin is a multifunctional adipokine that circulates as several oligomeric complexes in the blood stream. However, the molecular basis that regulates the production of the adiponectin oligomers remains largely elusive. We have shown previously that several conserved lysine residues (positions 68, 71, 80, and 104) within the collagenous domain of adiponectin are modified by hydroxylation and glycosylation (Wang, Y., Xu, A., Knight, C., Xu, L. Y., and Cooper, G. J. (2002) J. Biol. Chem. 277, 19521–19529). Here, we investigated the potential roles of these post-translational modifications in oligomeric complex formation of adiponectin. Gel filtration chromatography revealed that adiponectin produced from mammalian cells formed trimeric, hexameric, and high molecular weight (HMW) oligomeric complexes. These three oligomeric forms were differentially glycosylated, with the HMW oligomer having the highest carbohydrate content. Disruption of hydroxylation and glycosylation by substitution of the four conserved lysines with arginines selectively abrogated the intracellular assembly of the HMW oligomers in vitro as well as in vivo. In type 2 diabetic patients, both the ratios of HMW to total adiponectin and the degree of adiponectin glycosylation were significantly decreased compared with healthy controls. Functional studies of adiponectin-null mice revealed that abrogation of lysine hydroxylation/glycosylation markedly decreased the ability of adiponectin to stimulate phosphorylation of AMP-activated protein kinase in liver tissue. Chronic treatment of db/db diabetic mice with wild-type adiponectin alleviated hyperglycemia, hypertriglyceridemia, hepatic steatosis, and insulin resistance, whereas full-length adiponectin without proper post-translational modifications and HMW oligomers showed substantially decreased activities. Taken together, these data suggest that hydroxylation and glycosylation of the lysine residues within the collagenous domain of adiponectin are critically involved in regulating the formation of its HMW oligomeric complex and consequently contribute to the insulin-sensitizing activity of adiponectin in hepatocytes.

Adiponectin is a multifunctional adipokine that circulates as several oligomeric complexes in the blood stream. However, the molecular basis that regulates the production of the adiponectin oligomers remains largely elusive. We have shown previously that several conserved lysine residues ( Chem. 277, 19521-19529). Here, we investigated the potential roles of these post-translational modifications in oligomeric complex formation of adiponectin. Gel filtration chromatography revealed that adiponectin produced from mammalian cells formed trimeric, hexameric, and high molecular weight (HMW) oligomeric complexes. These three oligomeric forms were differentially glycosylated, with the HMW oligomer having the highest carbohydrate content. Disruption of hydroxylation and glycosylation by substitution of the four conserved lysines with arginines selectively abrogated the intracellular assembly of the HMW oligomers in vitro as well as in vivo. In type 2 diabetic patients, both the ratios of HMW to total adiponectin and the degree of adiponectin glycosylation were significantly decreased compared with healthy controls. Functional studies of adiponectin-null mice revealed that abrogation of lysine hydroxylation/glycosylation markedly decreased the ability of adiponectin to stimulate phosphorylation of AMP-activated protein kinase in liver tissue. Chronic treatment of db/db diabetic mice with wild-type adiponectin alleviated hyperglycemia, hypertriglyceridemia, hepatic steatosis, and insulin resistance, whereas full-length adiponectin without proper post-translational modifications and HMW oligomers showed substantially decreased activities. Taken together, these data suggest that hydroxylation and glycosylation of the lysine residues within the collagenous domain of adiponectin are critically involved in regulating the formation of its HMW oligomeric complex and consequently contribute to the insulin-sensitizing activity of adiponectin in hepatocytes.
Adiponectin, a hormone synthesized by adipocytes, is an abundant serum adipokine with potent insulin-sensitizing activity (1)(2)(3). Unlike most other adipokines, the plasma levels of adiponectin are significantly decreased in obese individuals and patients with insulin resistance, type 2 diabetes mellitus (T2DM), 2 and cardiovascular diseases (4 -7). Elevation of circulating adiponectin by either transgenic overexpression or direct supplementation with recombinant adiponectin can alleviate many metabolic abnormalities associated with various insulin-resistant and/or diabetic animal models (8 -12). The globular domain of adiponectin decreases postprandial blood glucose, enhances lipid clearance, and increases insulin sensitivity by enhancing fatty acid ␤-oxidation in skeletal muscle (8). On the other hand, full-length adiponectin generated from mammalian cells enhances the sensitivity of insulin to inhibit hepatic glucose production by suppressing the expression of several key enzymes involved in gluconeogenesis, including phosphoenolpyruvate carboxylase and glucose-6-phosphatase (10).
In addition to its metabolic effects, adiponectin also possesses antiatherogenic, anti-inflammatory, and hepatoprotective functions. This adipokine can prevent atherosclerotic lesion formation in apolipoprotein E-deficient mice, possibly by decreasing endothelial expression of inflammatory adhesion molecules, blocking foam cell formation, and inhibiting proliferation of aortic smooth muscle cells (11,(13)(14)(15)(16). It can also alleviate both alcoholic and nonalcoholic steatohepatitis and liver injuries induced by toxins such as carbon tetrachloride and lipopolysaccharides (17)(18)(19)(20). More recently, adiponectin has been proposed to play a protective role against cardiac hypertrophy and ischemia/reperfusioninduced myocardial infarction through the activation of AMP-activated protein kinase (AMPK) in myocardium (21)(22)(23).
Adiponectin belongs structurally to the complement C1q-like protein family, members of which consist of an N-terminal collagenous domain comprising of 22 Gly-X-Y repeats and a C-terminal globular head domain (3,24,25). In the circulation, adiponectin is predominantly present as several characteristic oligomeric complexes (26 -29). The basic building block of the adiponectin complex is a trimer or low molecular weight (LMW) oligomer, which is formed via hydrophobic interactions within its globular domain. Two trimers self-associate to form a disulfide-linked hexamer or middle molecular weight (MMW) * This work was supported by grants from the Marsden Fund of the Royal Society of New oligomer, which further assembles into a bouquet-like high molecular weight (HMW) multimeric complex that consists of 12-18 monomers. Different oligomeric forms of adiponectin might activate different signaling pathways and exert distinct functions on its target tissues (16, 27, 30 -32). The globular domain of adiponectin, which can form only trimers, is more potent than the other two oligomers in inducing AMPK activation and fatty acid ␤-oxidation in skeletal muscles (33). On the other hand, the HMW complex is the most active form of adiponectin in suppressing hepatic glucose production (26,30). Furthermore, only the HMW adiponectin, but not the trimeric and hexameric forms, can protect endothelial cells from apoptosis (34).
It has been proposed recently that the HMW/total adiponectin ratios, but not the absolute amounts of adiponectin, are closely correlated with insulin sensitivity in humans and rodents (30). The percentage composition of the HMW form in db/db diabetic mice is much lower than in wild-type littermates, despite similar levels of total adiponectin in the circulation (29,30). Clinical studies have also demonstrated that T2DM and coronary heart disease are associated with selective reduction of the HMW oligomeric adiponectin (35,36). On the other hand, weight reduction or treatment with the insulin-sensitizing drug rosiglitazone preferentially elevates the HMW form of adiponectin without significant effects on the other two oligomeric complexes (30,37). Interestingly, two rare missense mutations (G84R and G90S) identified in T2DM patients can form trimers and hexamers, but lack the capacity to form the HMW complex (36). These data collectively suggest that impaired formation of the HMW adiponectin might be causally associated with insulin resistance and T2DM.
The mechanism that regulates the formation and distribution of the oligomeric adiponectin complexes remains largely obscure at this stage. We have shown previously that adiponectin is modified at the posttranslational level during its secretion from adipocytes (38). Several conserved lysine residues (positions 68, 71, 80, and 104) within the collagenous domain of murine adiponectin are hydroxylated and glycosylated by an ␣-1,2-glucosylgalactosyl group, which might in turn contribute to its insulin-sensitizing activity in hepatocytes (28,38). In this study, we investigated the potential roles of these post-translational modifications in regulating the oligomeric complex formation of adiponectin in cell culture systems, animal models, and human subjects.

EXPERIMENTAL PROCEDURES
Materials-Minoxidil, dexamethasone, 3-isobutyl-1-methylxanthine, FLAG peptide, and anti-FLAG antibody M2 affinity gel were purchased from Sigma. Dulbecco's modified Eagle's medium (DMEM) and Lipofectamine transfection reagent were from Invitrogen. The QuikChange site-directed mutagenesis kit was from Stratagene. The Adeno-X expression system and the Immun-Blot kit for glycoprotein detection were from Clontech and Bio-Rad, respectively. The antibodies against total AMPK␣ and phospho-Thr 172 AMPK␣ were obtained from Cell Signaling Technology, Inc. Complete protease inhibitor mixture tablets were from Roche Diagnostics.
Site-directed Mutagenesis-The expression vector pcDNA-Ad-F, which encodes full-length murine adiponectin with a FLAG epitope tag at its C terminus (38), was used as a template to construct vectors encoding adiponectin variants in which the four lysines (residues 68, 71, 80, and 104) were replaced with arginines using the QuikChange sitedirected mutagenesis kit. The oligonucleotide primers used for mutation of each lysine residue are listed in Table 1. Expression vectors encoding various FLAG-tagged adiponectin variants with one, two, three, or all four lysine residues substituted with arginines were con-structed by sequential mutation of each site. All mutations were confirmed by DNA sequencing.
Expression and Purification of Recombinant Murine Adiponectin and Its Variants-The vectors for expression of wild-type adiponectin or its variants were transfected into HEK293 cells using Lipofectamine reagent according to the manufacturer's instructions. The transfected cells were incubated in serum-free DMEM containing 0.2% vitamin C and 0.2% bovine serum albumin for 48 h. The FLAG-tagged recombinant proteins were purified using the anti-FLAG monoclonal antibody M2 affinity gel as we described previously (38). The cells were solubilized with lysis buffer containing 1% Triton X-100, 25 mM HEPES, 5 mM EDTA, and 100 mM NaCl in the presence of protease inhibitor mixture and stored at Ϫ80°C until used.
Construction and Production of Recombinant Adenoviruses Encoding Wild-type or Mutant Adiponectin-Adenovirus expression vectors that encode FLAG-tagged adiponectin and its variants were generated using the Adeno-X expression system as we described previously (39). The recombinant virus was packaged and amplified in HEK293 cells and purified by CsCl density gradient centrifugation. The titers of the recombinant virus were determined by plaque assay.
Quantification of Different Oligomeric Forms of Adiponectin by Gel Filtration Chromatography and Enzyme-linked Immunosorbent Assay (ELISA)-Serum samples, conditioned culture medium, or cell lysates were diluted with phosphate-buffered saline to a final volume of 1 ml, loaded onto an ⌬KTA Explorer fast protein chromatography system, fractionated through a HiLoad 16/60 Superdex 200 column (GE Healthcare), and eluted with phosphate-buffered saline at a flow rate of 1 ml/min. Each 1-ml fraction was collected and subjected to ELISA analysis for murine adiponectin to determine the concentrations of each oligomeric form of adiponectin as we have described elsewhere (29).
Western Blot Analysis-For detection of different oligomeric adiponectin complexes, mouse serum samples, cell lysates, or conditioned culture medium was incubated with nonreducing sample buffer (1% SDS, 5% glycerol, and 10 mM Tris-HCl (pH 6.8)) at room temperature for 10 min, separated by 4 -20% gradient SDS-PAGE, transferred to polyvinylidene difluoride membranes, and immunoblotted with affinity-purified rabbit anti-mouse adiponectin IgG as we described previously (29).
For detection of total or phosphorylated AMPK, 50 g of protein homogenates from liver or gastrocnemius muscle was separated by 10% SDS-PAGE and probed with anti-total AMPK or anti-phospho-AMPK antibody as described (33). For glycoprotein detection, proteins separated by SDS-PAGE were transferred to nitrocellulose membranes, and glycoproteins were detected using the commercial Immun-Blot kit according to the manufacturer's instructions (38).
Pulse-Chase Experiment-HEK293 cells were transfected with expression vectors that encode wild-type adiponectin or its variants. 24 h after transfection, cells were starved for 1 h in methionine-and  (29,38). Cells were washed with phosphate-buffered saline and solubilized in immunoprecipitation buffer (25 mM HEPES (pH 7.5), 5 mM each EDTA and EGTA, 100 mM NaCl, 10% glycerol, and 1% Triton X-100 plus protease inhibitor mixture), and the cell debris was removed by centrifugation at 13,000 ϫ g for 5 min. All cell lysates were adjusted to an equal protein concentration (0.5 g/l) with immunoprecipitation buffer. Cell lysates (500 l) were preincubated with 50 l of protein A/G beads (GE Healthcare) for 1 h to remove nonspecific binding. The supernatants were incubated with 10 g of rabbit anti-mouse adiponectin IgG overnight at 4°C with shaking, and 50 l of protein A/G beads was then added to the supernatant for another 1 h. The beads were precipitated and washed with immunoprecipitation buffer three times, and the immunoprecipitated complexes were eluted by incubation with 100 l of SDS-PAGE loading buffer. The eluted samples were separated by 12% SDS-PAGE and then analyzed by phosphorimaging (Fujifilm BAS-2000). The radioactivity associated with each band was quantified as we described previously (29).

Lysyl Hydroxylase Inhibitor Study in Rat Primary
Adipocytes-White fat precursor cells were isolated from the epididymal fat pads of 6-weekold male Wistar rats and differentiated into mature adipocytes as described previously (29). 10 days after differentiation, ϳ80% of the cells became mature adipocytes filled with lipid droplets as assessed by Oil Red O staining. The cells were treated with either vehicle (0.1% ethanol) or the lysyl hydroxylase inhibitor minoxidil at the indicated concentrations for 2 days in the presence of 10% fetal bovine serum. The cells were then incubated with serum-free medium containing 0.2% bovine serum albumin and the inhibitor for another 24 h. Both cells and the conditioned medium were collected for analysis of oligomeric adiponectin complexes as described above.
Animal Studies-C57BL/6J adiponectin knockout mice (40) were housed under controlled light/dark cycle (12/12 h) and temperature conditions with free access to water and a standard mouse diet. 8 -10week-old male mice weighing 20 -25 g were administered 5 ϫ 10 7 plaque-forming units of various recombinant adenoviruses via tail vein injection. At different time points, small drops of blood were taken from the tails for analysis of the composition of different oligomeric forms of adiponectin. For the acute studies, the animals were fasted overnight and intravenously injected with various forms of recombinant adiponectin (2 mg/kg of body weight). After 10 min, mice were killed; blood was collected by cardiac puncture; and tissues were quickly dissected and snap-frozen in liquid nitrogen for further analyses.
Male C57BKS db/db mice (10 -11 weeks old; The Jackson Laboratory) were used to evaluate the chronic effect of various forms of adiponectin. The protein was continuously delivered to the mice using ALZET osmotic pumps (DURECT Corp., Cupertino, CA) as we described previously (18,19). The pumps were filled with various protein solutions or saline as a control. Mice were anesthetized, and ϳ0.5-cm incisions were made in the lower backs of the animals. Small subcutaneous pockets were made to insert the pumps, which delivered the protein solutions at a constant rate for a period of 2 weeks.
Measurement of Blood Parameters, Intraperitoneal Glucose Tolerance Test, and Insulin Tolerance Test-Fasting (16 h) plasma glucose levels and triglyceride (TG) concentrations were measured using a glucose assay kit (Sigma) and triglyceride reagent (Pointe Scientific, Inc., Canton, MI), respectively. Fasting serum insulin concentrations were quantified using the commercial ELISA kits from Mercodia AB (Uppsala, Sweden). Serum total cholesterol free fatty acids was analyzed using kits from Wako Chemicals USA, Inc. (Richmond, VA), and Roche Diagnostics, respectively. TG contents in liver and gastrocnemius muscle were analyzed as described (18,39).
For intraperitoneal glucose tolerance testing, mice fasted overnight (16 h) were given a glucose load by intraperitoneal injection (1 g of glucose/kg of body weight). For insulin tolerance testing, mice were starved for 6 h and then intraperitoneally injected with insulin (1 unit/kg of body weight). Plasma glucose levels were measured at different time points as indicated.
Studies of Human Subjects-Serum samples from 12 male patients with T2DM and 12 age-matched male healthy controls were selected from our previous study (39). The clinical characteristics of these subjects are listed in Table 2. The plasma levels of total adiponectin and the ratios of HMW to total adiponectin were determined by an in-house human adiponectin ELISA method and gel filtration chromatography as we described previously (29). To determine the carbohydrate contents of endogenous human adiponectin, serum samples (1 ml) were depleted of albumin and IgG using the ProteoExtract albumin/IgG removal kit (Calbiochem). The remaining supernatant was then incubated with 100 l of Sepharose beads coupled with mouse non-immune IgG to remove nonspecific binding. The clarified supernatant was subsequently incubated with 100 l of Sepharose beads coupled with mouse anti-human adiponectin monoclonal IgG (29) overnight at 4°C. The beads were washed with Tris-buffered saline, and the bound protein complexes were eluted with 0.1 M glycine HCl (pH 2.5). The eluted fractions were concentrated, and the adiponectin concentrations were quantified by ELISA. Samples containing equal amounts of adiponectin were separated by 10% SDS-PAGE and subjected to Immun-Blot analysis for carbohydrate content as described above. The intensity of each band was quantified using ImageMaster software (GE Healthcare).
Statistical Analysis-All values are presented as the means Ϯ S.E. Differences between groups were determined by Student's t test or oneway analysis of variance. In all statistical comparisons, p Ͻ 0.05 was considered statistically significant.

RESULTS
The Three Oligomeric Complexes of Adiponectin Produced from Mammalian Cells Are Differentially Glycosylated-Our previous study demonstrated that both endogenous adiponectin and recombinant adiponectin produced from mammalian cells are post-translationally modified by hydroxylation and glycosylation (38). To investigate the roles of these post-translational modifications in oligomeric complex formation of adiponectin, we generated recombinant adiponectin from both prokaryotic and HEK293 cells. Gel filtration analysis revealed that adiponectin produced from HEK293 cells formed HMW, MMW, and LMW oligomeric complexes, which were eluted at 45.5-49, 52.5-56.5, and 58.5-62 min, respectively (Fig. 1A). Non-heating and nonreducing SDS-PAGE analysis resolved this recombinant protein into three dis-tinct bands with apparent molecular weights of Ͼ250,000, ϳ180,000, and ϳ90,000, which are equivalent to the HMW, MMW, and LMW oligomeric forms, respectively (Fig. 1B). On the other hand, adiponectin derived from Escherichia coli, which is not glycosylated, formed only LMW and MMW oligomers, but lacked the capacity to further assemble into the HMW oligomeric complexes. Carbohydrate detection revealed that the three oligomeric forms of adiponectin derived from mammalian cells were differentially glycosylated (Fig. 2). The carbohydrate contents of the HMW adiponectin were significantly higher than those of the MMW and LMW oligomeric complexes. These results suggest that the post-translational modifications might be important for the production of the HMW oligomeric complex of adiponectin.
Substitution of the Four Lysine Residues with Arginines Causes Decreased Secretion of Adiponectin and Impaired Intracellular Assembly of Its HMW Complexes-We have reported previously that glycosylation of adiponectin occurs primarily at the four hydroxylysine residues (positions 68, 71, 80, and 104) within its collagen-like domain (28,38). Compared with wild-type adiponectin, the adiponectin mutant with the four lysines replaced with arginines (ADN-K/R) is much less effective in enhancing insulin's ability to suppress hepatic glucose production (38). Here, we investigated whether the post-translational modifications of these four lysines of adiponectin play any role in its secretion and/or oligomeric complex formation in HEK293 cells transfected with different expression vectors. Quantitative ELISA analysis revealed that the concentration of wild-type adiponectin released into the culture medium was much higher than that of the ADN-K/R mutant (Fig. 3A), although the steady-state adiponectin mRNA levels were comparable (data not shown). Pulse-chase experiments with 35 S demonstrated that the secretion rate of wild-type adiponectin was much faster than that of the ADN-K/R mutant (Fig. 3, B and C). 15 h after chasing, the majority of wild-type adiponectin synthesized during a 2-h pulse period was   released into the medium, whereas ϳ40% of the ADN-K/R mutant was still retained inside the cells. These results suggest that the four hydroxylysines and their attached glycosides might be important for efficient secretion of adiponectin.
Non-heating and nonreducing gel electrophoresis and Western blot analysis showed that wild-type adiponectin formed all three oligomers (Fig. 4). The HMW, MMW, and LMW oligomeric complexes accounted for ϳ37, ϳ42, and ϳ19%, respectively, inside the cells as well as in the conditioned culture medium. On the other hand, both intracellular and extracellular ADN-K/R mutants were present predominantly as the LMW and MMW oligomeric forms, whereas the HMW oligomers were barely detectable. Notably, ADN-K/R oligomers migrated slightly faster than wild-type adiponectin, suggesting the loss of the carbohydrate moieties. These data demonstrate that post-translational modifications of these four conserved lysine residues are critical for intracellular assembly of the HMW oligomeric adiponectin complexes.
To further determine the contribution of the post-translational modifications of each of the four lysine residues to the secretion of adiponectin and its oligomeric complex formation, we generated a series of vectors for expression of the adiponectin variants with one, two, or three of the lysines substituted with arginines. These expression vectors were transiently transfected into HEK293 cells to assess the capacity of these adiponectin variants to form the HMW oligomeric complex. Quantitative ELISA analysis revealed that adiponectin released into the conditioned medium progressively decreased with sequential arginine substi-tution of the four lysines (data not shown). Gel filtration analysis showed that a single mutation of each of the four lysines (K68R, K71R, K80R, and K104R) caused only a slight but not significant decrease in the formation of the HMW oligomeric adiponectin (Fig. 5). On the other hand, the ability of the double mutants (K68R/K71R and K80R/ K104E) and triple mutants (K68R/K71R/K80R and K68R/K71R/K104R) to assemble into the HMW oligomeric complex was significantly attenuated. Notably, the percentage composition of the HMW oligomeric form versus total adiponectin was gradually decreased following sequential mutation of one, two, three, and all four lysine residues, suggesting that the hydroxylysines and the attached glycosides might function in a cooperative manner to facilitate the formation of the HMW oligomeric adiponectin.

Inhibition of Lysyl Hydroxylases by Minoxidil Decreases the Formation of HMW Adiponectin Oligomers in Rat
Adipocytes-A family of lysyl hydroxylases (EC 1.14.11.4), consisting of lysyl hydroxylase-1, -2a, -2b, and -3, is responsible for catalyzing the conversion of lysine into hydroxylysine, which provides attachment sites for further glycosylation (41). Minoxidil has been shown to inhibit the activity of lysyl hydroxylase by decreasing the gene expression of all three isoenzymes (42,43). We therefore tested the effect of minoxidil on the oligomeric complex formation of endogenous adiponectin in primary rat adipocytes. In line with our previous report (29), we found that the HMW oligomer accounted for ϳ35% of the total adiponectin in the conditioned medium of adipocytes, which was significantly lower than that inside the cells (Fig. 6), suggesting that secretion of the HMW oligomer from adipocytes is relatively slower than that of the other two oligomeric forms. Incubation of mature rat adipocytes with 500 M minoxidil for 72 h led to an ϳ70% reduction of lysyl hydroxylase activity (data not shown). The percentage composition of the HMW complex versus total adiponectin in both cell lysates and the conditioned medium of adipocytes was substantially decreased following treatment with this compound. Treatment with 500 M minoxidil also decreased the percentage composition of HMW versus total adiponectin in HEK293 cells transfected with a vector that encodes wild-type adiponectin (data not shown).

. Comparison of the oligomeric complex distribution between wild-type adiponectin and the adiponectin variant ADN-K/R expressed in HEK293 cells.
HEK293 cells were transfected with vectors encoding either wild-type adiponectin (ADN-WT) or ADN-K/R. 48 h after transfection, both cell lysates and the conditioned medium were harvested and subjected to ELISA analysis for total mouse adiponectin. 200 ng of total adiponectin from cell lysates or the conditioned medium was separated by nonreducing and non-heating SDS-PAGE and immunoblotted with anti-mouse adiponectin antibody (A). The conditioned medium was fractionated, and the fractions corresponding to the three oligomeric forms of adiponectin were pooled (B). The percentage composition of each oligomeric form versus the total was determined as described under "Experimental Procedures." *, p Ͻ 0.01 versus wild-type adiponectin (n ϭ 4). Note that the percentage composition of each oligomeric form in the cell lysates was similar to that in the conditioned medium (data not shown).

Disruption of Hydroxylation and Glycosylation of the Four Lysine Residues Impairs the Production of the HMW Oligomeric Adiponectin in
Vivo-To further evaluate the effect of lysine hydroxylation/glycosylation on oligomeric complex formation in vivo, we generated recombinant adenoviruses that encode wild-type adiponectin or its variant ADN-K/R. The recombinant adenovirus was introduced into adiponectin-null mice via tail vein injection. Reverse transcription-PCR analysis revealed that the adiponectin gene was expressed primarily in liver tissue following injection with the recombinant adenoviruses. Despite having a similar mRNA expression level in liver tissue, the protein concentrations of wild-type adiponectin in both liver tissue and the circulation were substantially higher than those of the adiponectin variant ADN-K/R (Fig. 7), suggesting that lysine hydroxylation and glycosylation might be involved in the stabilization of adiponectin in vivo. In serum as well as in liver lysates, the predominant oligomeric form of wild-type adiponectin was the HMW oligomer, which accounted for ϳ65% of the total adiponectin (Fig. 8). In contrast, the HMW oligomeric complex accounted for only ϳ9% of the total adiponectin variant ADN-K/R, suggesting that hydroxylation/glycosylation of lysine residues is critical for the formation of HMW adiponectin oligomers in vivo.

Hydroxylation/Glycosylation and Formation of the HMW Complexes Are Critically Involved in Adiponectin-induced Activation of AMPK in
Liver, but Not in Skeletal Muscle-Adiponectin exerts its multiple metabolic effects primarily via inducing phosphorylation and activation of AMPK in liver and skeletal muscle (33). To investigate the potential roles of lysine hydroxylation/glycosylation and the subsequent formation of HMW oligomers in regulating the metabolic activities of adiponectin, we next supplemented the adiponectin-null mice with different forms of adiponectin with varying degrees of post-translational modifications. Intravenous injection of different forms of recombinant adiponectin into adiponectin-null mice at 2 mg/kg of body weight raised the plasma levels of adiponectin from 0 to ϳ7-10 g/ml within 5 min. In liver tissue, HEK293 cell-produced wild-type adiponectin, which contains properly hydroxylated/glycosylated lysines and also forms the HMW oligomeric complexes, potently induced phosphorylation of AMPK (Fig. 9). Disruption of lysine hydroxylation/glycosylation either by substitution of the four lysines with arginines or by the lysyl hydroxylase inhibitor minoxidil significantly attenuated the activities of adi-   . Liver homogenates or serum samples were fractionated by gel filtration, and the percentage composition of the HMW oligomer versus total adiponectin was determined as described in the legend to Fig. 4 (B). *, p Ͻ 0.01 versus the control (n ϭ 4).
ponectin on phosphorylation of AMPK. Bacterially generated fulllength or globular adiponectin, both of which lack the post-translational modifications and cannot form the HMW oligomeric complex, had no obvious effects. In skeletal muscle, globular adiponectin potently induced AMPK phosphorylation, whereas the same dosage of fulllength adiponectin produced from different sources exhibited little activity.
Lysine Hydroxylation/Glycosylation and HMW Oligomeric Complexes Are Important for the Antidiabetic Activities of Full-length Adiponectin in db/db Diabetic Mice-We next investigated the chronic effects of various forms of adiponectin in db/db mice, an established obese/diabetic mouse model with insulin resistance. To this end, mice were surgically implanted with an ALZET osmotic pump, which delivered various forms of adiponectin (1.5 mg/kg of body weight/day) or physiological saline at a constant rate. Delivery of various forms of adiponectin protein at this dosage caused an ϳ2.0 -2.5-fold elevation of the circulating adiponectin levels throughout the 2-week treatment period. These treatments had no obvious effect on food intake, body weight gains, and serum levels of insulin and total cholesterol ( Table 2). HEK293 cell-produced wild-type adiponectin, which possesses properly hydroxylated/glycosylated lysines and also forms HMW oligomeric complexes, significantly decreased blood glucose levels and serum TG concentrations and also caused a drastic reduction of TG accumulation in liver tissue and a modest decrease in TG content in skeletal muscle (Table 2). In addition, treatment with wild-type adiponectin alleviated glucose intolerance and insulin resistance in db/db mice, as determined by intraperitoneal glucose tolerance test and insulin tolerance test analyses (Fig. 10). The adiponectin mutant ADN-K/R and full-length adiponectin produced from E. coli, both of which lack proper post-translational modifications and HMW oligomeric complexes, exhibited substantially decreased activities compared with wild-type adiponectin. Treatment of db/db mice with globular adiponectin also decreased circulating TG concentrations and alleviated glucose intolerance and insulin resistance. However, globular adiponectin had no significant effect on fasting blood glucose levels. On the other hand, compared with wildtype adiponectin produced from HEK293 cells, globular adiponectin showed much greater effects in decreasing TG accumulation in skeletal muscle, but was less potent in reducing hepatic TG contents (Table 2). These data further support several previous reports showing that globular adiponectin increases insulin sensitivity primarily through reduction of muscular lipid accumulation, whereas full-length adiponectin with proper post-translational modifications and HMW oligomeric complexes exerts its beneficial metabolic effects mainly via its hepatic actions (16,30,31).
Both the HMW/Total Adiponectin Ratios and the Degree of Adiponectin Glycosylation Are Decreased in T2DM Patients-To explore the clinical relevance of the above findings, we next investigated the relationship between the post-translational modifications and oligomeric complex formations in 12 T2DM patients and 12 age/sex-matched healthy controls. The clinical characteristics of these subjects are summarized in Table 3. Compared with the healthy controls, the plasma levels of total adiponectin as well as the percentage composition of HMW versus total adiponectin were significantly decreased (Fig. 11). Furthermore, the carbohydrate content of adiponectin in T2DM patients was also significantly lower than that in healthy controls. There was a strong positive correlation between the ratios of HMW to total adiponectin and the carbohydrate content of adiponectin (r ϭ 0.213, p Ͻ 0.05). These results indicate that decreased glycosylation of hydroxylysines might be causally linked to the impaired formation of the HMW adiponectin oligomers in certain pathological conditions. FIGURE 9. Differential effects of different forms of adiponectin on phosphorylation of AMPK in liver and gastrocnemius muscle. Various forms of recombinant adiponectin were intravenously injected into C57BL/6J adiponectin knockout mice at 2 mg/kg of body weight. 10 min after injection, mice were killed, and livers (A) and gastrocnemius muscles (B) were dissected. 50 g of protein homogenates from these tissues was separated by SDS-PAGE and immunoblotted with anti-total AMPK (AMPK␣) or anti-phospho-AMPK (pAMPK) antibody as indicated. AND-WT, wildtype adiponectin; ADN-WTϩminoxidil, wild-type adiponectin produced in HEK293 cells in the presence of 500 M minoxidil; fAd-E.coli: full-length adiponectin expressed in E. coli; gAd-E.coli, globular head of adiponectin produced in E. coli. Representative immunoblots are shown for each quantitative analysis performed in experiments repeated at least three times. *, p Ͻ 0.05 compared with the control.

DISCUSSION
Regulation of oligomeric complex formation is increasingly recognized to be an important mechanism that modulates the pleiotropic biological functions of adiponectin. Several recent studies have demonstrated that the ratio of HMW to total adiponectin is more closely associated with glucose tolerance and insulin sensitivity than the plasma levels of total adiponectin (30,36,37). The HMW/total adiponectin ratio is decreased in obese subjects and in patients with T2DM and coronary heart disease, and this decrease is reversed following moderate weight reduction and treatment with thiazolidinediones (30,35,44). We (29) and others (26) have found that the sexual dimorphism of adiponectin in both humans and rodents is attributed primarily to the difference in the HMW adiponectin, with females having significantly higher levels of this oligomer compared with males. Although these data support the role of the HMW adiponectin as an endogenous insulin sensitizer, the molecular basis that mediates the formation of the HMW oligomeric complexes at the post-translational level remains poorly understood.
Several previous studies have demonstrated the necessity of disulfide bonds mediated by Cys 39 in the formation of hexameric adiponectin and HMW adiponectin, but have not shown whether these disulfide bonds are sufficient for the high order structural formation of adiponectin. In this study, we have provided both in vivo and in vitro evidence to support the notion that post-translational modifications, specifically the hydroxylation and further glycosylation of several lysine residues within the collagenous domain, are required for intracellular assembly of the HMW adiponectin oligomers. First, bacterially generated full-length adiponectin, which lacks post-translational modifications, could not form the HMW oligomers (Fig. 1). Second, ablation of hydroxylation and glycosylation by substitution of the four lysines (residues 68, 71, 80, and 104) with arginines impeded the intracellular assembly of the HMW adiponectin in both the cell culture system and mice (Figs. 4 and 8). In addition, treatment with minoxidil, a lysyl hydroxylase inhibitor, also resulted in a marked reduction of the HMW adiponectin (Fig. 6). Consistent with our results, a recent study on lysyl hydroxylase-3-null mice showed that ablation of this enzyme leads to a total absence of hydroxylysine and its attached carbohydrates in type IV collagen and impaired high order structure formation of this protein (43). Notably, in addition to its lysyl hydroxylase activity, lysyl hydroxylase-3 also possesses relatively low levels of glucosyltransferase and galactosyltransferase activities, suggesting that this enzyme alone is sufficient for catalyzing all three steps, including lysine hydroxylation and its further attachment with an ␣-1,2-glucosylgalactosyl group (45). Further study is warranted to investigate whether this enzyme is involved in regulating the HMW oligomeric complex formation of adiponectin via catalyzing lysine hydroxylation and glycosylation of adiponectin in adipocytes.
The intracellular assembly and secretion of oligomeric adiponectin complexes are complex processes that might vary in different cell types. The ratio of HMW to total adiponectin inside rat primary adipocytes was ϳ65%, which is significantly higher than that expressed in HEK293 cells (ϳ35%). On the other hand, the percentage composition of HMW versus total adiponectin in the culture medium of adipocytes was much lower than that in cell lysates, suggesting that the HMW adiponectin is selectively retained inside the cells. We have reported recently that the secretion of the HMW adiponectin from adipocytes is much slower FIGURE 10. Chronic effects of various forms of adiponectin on glucose tolerance and insulin sensitivity in db/db diabetic mice. The intraperitoneal glucose tolerance test (ipGTT) and the insulin tolerance test (ITT) were conducted at days 11 and 13 after various treatments. In the insulin tolerance test study, plasma glucose levels were normalized to those at t ϭ 0 min in each treatment group. ADN-WT, wild-type adiponectin; fAd-E.Coli: full-length adiponectin expressed in E. coli; gAd-E. Coli: globular head of adiponectin produced in E. coli. *, p Ͻ 0.05 versus the saline-treated group (n ϭ 4 -5). FIGURE 11. Carbohydrate content of adiponectin and ratio of HMW to total adiponectin in T2DM patients and healthy individuals. 2 g of human adiponectin purified from each serum sample by affinity chromatography was subjected to SDS-PAGE and carbohydrate estimation analysis as described under "Experimental Procedures." A, representative gels are shown. B, the results from quantitative analysis of carbohydrate contents are shown. C, serum samples from these subjects were fractionated by gel filtration chromatography, and each fraction was analyzed using an in-house ELISA method for human adiponectin to determine the ratio of the HMW oligomers as described. *, p Ͻ 0.01 versus healthy controls. than that of the other two oligomeric complexes and that testosterone treatment further decreases the secretion of this oligomeric form (29). Acute injection of insulin and glucose has been shown to selectively decrease the circulating levels of HMW adiponectin in mice, although the underlying mechanism remains to be determined (30). In this study, we have shown that the decreased ratio of HMW to total adiponectin in type 2 diabetic patients is closely associated with the reduced degree of adiponectin glycosylation (Fig. 11), implying that impaired posttranslational modifications of adiponectin might be causally linked with T2DM by disrupting the high order structure formation of this adipokine. Several previous studies of Scherer and co-workers (3,10,46) suggest that the HMW oligomer of adiponectin is the major active form responsible for its insulin-sensitizing effect in hepatocytes. The HMW adiponectin oligomers can decrease blood glucose levels by activation of AMPK, which in turn inhibits hepatic glucose production by decreasing the expression of key gluconeogenic genes such as phosphoenolpyruvate carboxykinase and glucose-6-phosphatase (10,30,33). Consistent with these reports, our study has also demonstrated that acute injection of wild-type adiponectin produced from mammalian cells can induce phosphorylation of AMPK in liver tissue (Fig. 9). Chronic treatment of db/db diabetic mice with wild-type adiponectin led to a significant decrease in blood glucose and serum TG concentrations and also improved glucose tolerance and insulin sensitivity ( Fig. 10 and Table 2). Notably, these changes were associated with a marked reduction of hepatic TG accumulation. All these beneficial metabolic effects of fulllength adiponectin were significantly attenuated upon the decrease in the composition of the HMW oligomers and were abolished upon the depletion of the HMW oligomers by ablation of lysine hydroxylation and glycosylation. Therefore, the magnitude of AMPK phosphorylation in liver tissue and the metabolic effects of adiponectin in db/db diabetic mice appear to correlate well with the composition of HMW adiponectin oligomers. In contrast to their actions in liver, none of the full-length adiponectins, regardless of the composition of the HMW oligomers, had any obvious effect on AMPK phosphorylation in skeletal muscle. These findings are in agreement with recent clinical reports showing that the plasma levels of adiponectin HMW oligomers are strongly correlated with hepatic insulin sensitivity, but are less relevant to the insulin sensitivity in muscle (44,47,48). In line with previous reports (32,33), our results also demonstrate that globular adiponectin, which forms exclusively as trimers, had potent effects on activation of AMPK and on reduction of lipid accumulation in skeletal muscle, which was associated with a significant alleviation of hypertriglyceridemia and insulin resistance. Nevertheless, globular adiponectin lacks the ability to activate AMPK in liver and to decrease fasting blood glucose levels in db/db diabetic mice ( Fig. 9 and Table 2). Taken together, these results support the notion that globular adiponectin and full-length adiponectin exert distinct metabolic actions in different target tissues (16,30,31).
In summary, this study provides evidence demonstrating that hydroxylation and glycosylation of several conserved lysine residues within the collagenous domain of adiponectin are required for formation of its HMW oligomeric complexes, which in turn confer the insulin-sensitizing effects of this adipokine in hepatocytes. In addition, decreased levels of the HMW adiponectin and reduced carbohydrate content of this protein are concurrently observed in patients with T2DM, suggesting that these events might be causally linked with this disease. Further investigation is needed to identify the precise enzyme that mediates lysine hydroxylation and glycosylation of adiponectin in adipocytes and to elucidate the underlying mechanisms that regulate these events.