A novel serum protein similar to C1q, produced exclusively in adipocytes.

We describe a novel 30-kDa secretory protein, Acrp30 (adipocyte complement-related protein of 30 kDa), that is made exclusively in adipocytes and whose mRNA is induced over 100-fold during adipocyte differentiation. Acrp30 is structurally similar to complement factor C1q and to a hibernation-specific protein isolated from the plasma of Siberian chipmunks; it forms large homo-oligomers that undergo a series of post-translational modifications. Like adipsin, secretion of Acrp30 is enhanced by insulin, and Acrp30 is an abundant serum protein. Acrp30 may be a factor that participates in the delicately balanced system of energy homeostasis involving food intake and carbohydrate and lipid catabolism. Our experiments also further corroborate the existence of an insulin-regulated secretory pathway in adipocytes.

Insulin-induced glucose transport occurs in heart, striated muscle, and fat tissue. In adipocytes, glucose uptake increases 20-to 30-fold in the presence of insulin. Glucose transport is mediated by the sodium-independent facilitative glucose transporters GLUT1 and GLUT4, which, in response to insulin, translocate from an intracellular compartment to the plasma membrane (1,2). GLUT4, which is expressed only in fat and skeletal and cardiac muscle, is the primary transporter involved in this process and is the predominant transporter expressed in these tissues (3,4). Insulin also causes translocation of several receptor proteins from intracellular membranes to the plasma membrane (5)(6)(7). Adipocytes are a principal storage depot for triglycerides and express a specific transport protein allowing them to import free fatty acids (8).
Adipocytes also secrete several proteins potentially important in homeostatic control of glucose and lipid metabolism. Adipsin, equivalent to Factor D of the alternative complement pathway (9), is synthesized exclusively in adipocytes, and its secretion is enhanced severalfold by insulin (10). The function of adipsin in fat cell biology is not known, nor are the roles of complement factors C3 and B that are also secreted by adipocytes (11). Tumor necrosis factor ␣, also secreted by adipocytes, is a key mediator of insulin resistance in animal models of non-insulin-dependent diabetes mellitus. Tumor necrosis factor ␣ directly interferes with the signaling of insulin through its receptor and consequently blocks biological actions of insulin including insulin-stimulated glucose uptake (reviewed in Ref. 12). Adipocytes are the only cell type known to secrete the ob gene product (13). In the absence of ob (ob Ϫ /ob Ϫ mice) or its presumed receptor, the db gene product (db Ϫ /db Ϫ mice) the mice overeat and become obese and diabetic (14). Here we describe another novel protein, Acrp30, 1 that is exclusively synthesized in adipose tissue and secreted into serum. Like adipsin, secretion of Acrp30 is enhanced severalfold by insulin. While we do not know the function of this protein, its sequence and structural resemblance to complement factor C1q is intriguing. Importantly, our experiments confirm the existence of an insulin-regulated secretory pathway in adipocytes.

EXPERIMENTAL PROCEDURES
Cloning of Acrp30 and DNA Analysis-A full-length cDNA library templated by mRNA from 3T3-L1 adipocytes at day 8 of differentiation (15) was screened with a digoxygenin-labeled cDNA fragment obtained from the random sequencing screen. Labeling, hybridization, and detection were performed according to the manufacturer's instructions (Boehringer Mannheim). One of the positive clones obtained was subjected to automated sequencing on a Applied Biosystems 373-A Sequencer. The entire 1.3-kb insert was sequenced at least 2 independent times on one and once on the complementary strand. Sequence analysis was performed with the DNAstar package and showed an open reading frame of 741 bp encoding a protein of 28 kDa. The sequence has been submitted to GenBank TM and has the accession number U37222. Homology searches were performed at NCBI using the BLAST network service, and alignments were performed with the Megalign program from DNAstar using the Clustal algorithm. Only the globular domain for the type X collagen was used for the alignment (residues 562-680).
Generation of Specific Antibodies-A peptide corresponding to the sequence, EDDVTTTEELAPALV (residues 18 -32), was used to generate specific anti-Acrp30 antibodies in rabbits (multiple antigen peptide technology, Research Genetics).
mRNA Isolation and Analysis-Isolation of mRNA from tissues and from 3T3-L1 cells at various stages of differentiation was as described in Ref. 15, as was 32 P labeling of DNA, agarose gel electrophoresis of mRNA, and its transfer to nylon membranes.
Pulse-Chase Experiments and Immunoprecipitations-3T3-L1 adipocytes were starved for 30 min in Dulbecco's modified Eagle's medium (ICN) lacking cysteine and methionine and then labeled for 10 min in the same medium containing 0.5 mCi/ml Express Protein Labeling Reagent (1000 Ci/mmol) (DuPont NEN). The cells were then washed twice with Dulbecco's modified Eagle's medium supplemented with unlabeled cysteine and methionine, and then fresh growth medium containing 300 M cycloheximide was added. At the indicated time points, the medium was collected. Insoluble material from the medium was removed by centrifugation (15,000 ϫ g for 10 min); the supernatants were precleared with 50 l of Protein A-Sepharose for 30 min at 4°C and then immunoprecipitated with 50 l of affinity-purified anti-Acrp30 antibody for 2 h at 4°C. Immunoprecipitates were washed 4 times in lysis buffer (1% Triton X-100, 60 mM octyl glucoside, 150 mM NaCl, 20 mM Tris, pH 8.0, 2 mM EDTA, 1 mM phenylmethylsulfonyl * This work was supported in part by National Institutes of Health Grant DK 47618 and a grant from Pfizer Corp. (to H. F. L.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Supported by a fellowship from the Swiss National Science Cross-linking of Acrp30 -A 10-cm plate of day 8 3T3-L1 adipocytes was labeled overnight with [ 35 S]methionine and cysteine as described above. The medium was collected and, by means of several spins in a Centricon 10 microconcentrator, the buffer was replaced with 150 mM NaCl, 50 mM KP i , pH 8.5. A stock solution of 200 mg/ml bis(sulfosuccinimidyl)suberate (BS 3 , Pierce) in dimethyl sulfoxide was prepared fresh and added to the final concentrations indicated in the figure legends. Reactions were allowed to proceed for 30 min on ice, and excess cross-linker was quenched by addition of 500 mM Tris buffer, pH 8.0. Samples were diluted 1:1 with lysis buffer and subjected to immunoprecipitation with anti-Acrp30 antibodies.
Other Methods-Separation of proteins by SDS-PAGE, fluorography, immunoblotting, protein determinations, and densitometric scanning of the gels were performed as described previously (16).

RESULTS
In order to identify novel adipocyte-specific proteins, we have randomly sequenced portions of 1000 clones from a subtractive cDNA library enriched in mRNAs induced during adipocyte differentiation of 3T3-L1 fibroblasts (15). Northern blot analysis using one ϳ250-bp clone showed a marked induction during adipocyte differentiation, and thus a full-length cDNA was isolated and sequenced. The encoded protein, Acrp30, was novel; it contained 247 amino acids with a predicted molecular mass of 28 kDa. Acrp30 consists of a predicted amino-terminal signal sequence, followed by a stretch of 27 amino acids that does not show significant homology to any protein in the data base and then by 22 perfect Gly-X-Pro or Gly-X-X repeats (Fig.  1, A and B). The carboxyl-terminal globular domain exhibits striking homology to a number of proteins, such as the globular domains of type VIII and type X collagens (17), the subunits of complement factor C1q (18) and a protein found in the serum of hibernating animals during the summer months (19). Structurally, albeit not at the primary sequence level, the protein resembles the lung surfactant protein (20) and the hepatocyte mannan-binding protein (21), both of which have collagen-like domains and globular domains of similar size.
Northern blot analysis shows that Acrp30 is expressed exclusively in adipocytes. It is not expressed in 3T3-L1 fibroblasts and is induced over 100-fold during adipocyte differentiation. Induction occurs between days 2 and 4, at the same time as other adipocytespecific proteins such as GLUT4 (22) and Rab3D (15) (Fig. 1C). These results were confirmed by Western blot analysis (data not shown). The amount of Acrp30 mRNA may decline somewhat from Day 6 to Day 8 (Fig. 1C), but this drop is not reproducible in other experiments. In any case, we have not studied the accumulation or stability of Acrp30 mRNA after Day 8.
An antibody raised against a peptide corresponding to the unique amino-terminal domain of Acrp30 recognized a 3T3-L1 adipocyte protein of approximately 28 kDa (not shown). Acrp30 contains one potential N-glycosylation site, within the collagen domain, but apparently is not glycosylated; Endo H treatment did not cause a shift in molecular mass of Acrp30 at any time during a metabolic pulse-chase experiment (not shown). Acrp30 does become modified post-translationally, since, after 20 min of chase, there was a small but reproducible reduction in gel mobility. This most likely represents hydroxylation of collagen domain proline residues in the endoplasmic reticulum or Golgi compartments, by analogy to a similar modification in the structurally related mannan-binding protein (23). In 3T3-L1 adipocytes unstimulated by insulin, 50% of newly made Acrp30 is secreted into the medium at 2.5 to 3 h of chase as judged by densitometric scanning of the immunoprecipitates of intracellular and extracellular 35 S-labeled Acrp30. Indeed, Acrp30 can be detected by Western blotting in normal mouse serum. A protein of the identical molecular weight can be detected by Western blot analysis of 3T3-L1 adipocytes (not shown). The anti-peptide antibody is specific for the mouse homologue, as it does not cross-react with bovine, human, or rabbit serum (Fig. 2).
To examine effects of insulin on Acrp30 secretion, we monitored the discrete population of newly made protein generated in a short pulse with labeled amino acids followed by inhibition of further protein synthesis by cycloheximide. This offers increased sensitivity compared to examining secretion of the entire cellular complement of Acrp30, particularly in light of the very long t1 ⁄2 for secretion of Acrp30. Fig. 3 shows that, during the first 60 min of chase, insulin causes a 4-fold increase in secretion of newly made Acrp30. After 60 min, the rates of Acrp30 secretion are the same in unstimulated and insulinstimulated cells. Similarly, insulin causes a 4-fold increase in adipsin secretion during the first 30 min of chase, but, afterwards, the rate of adipsin secretion is the same in control and insulin-treated cells (Fig. 3 and Ref. 10). The ability of insulin to abolish the lag in adipsin secretion has been seen in several separate experiments. We hypothesize that a fraction of newly made adipsin and Acrp30 are sorted, probably in the trans-Golgi reticulum, into regulated secretory vesicles whose exocytosis is induced (in an unknown manner) by insulin, whereas the balance is sorted into vesicles that are constitutively exocytosed. Partial sorting of protein hormones into regulated secretory vesicles has been seen in other types of cultured cells (24,25). We do not know how insulin causes an increase in protein secretion; insulin could cause a more efficient overall processing of secretory proteins in 3T3-L1 adipocytes. We are currently isolating other adipocyte-specific secretory proteins to study this process in detail.
Complement factor C1q consists of three related polypeptides that form heterotrimeric subunits containing a threestranded collagen "tail" and three globular "heads"; six of these subunits generate an 18-mer complex often referred to as a "bouquet of flowers" (reviewed in Ref. 26). The experiments in Fig. 4 show that Acrp30 has a similar oligomeric structure, but is composed of a single type of polypeptide chain. When analyzed by velocity gradient sedimentation analysis, Acrp30 in blood serum migrates as two species of apparent molecular masses of 90 kDa and 300 kDa (Fig. 4C). Disregarding the presumably nonglobular shape of the complex that could lead to a slight distortion of the molecular weight determination, the former is probably a trimer and the latter could be a nonamer or dodecamer. Isoelectric focusing followed by SDS-PAGE of 35 S-Acrp30 secreted by 3T3-L1 adipocytes reveals only a single polypeptide (not shown), suggesting that Acrp30 forms homo-oligomeric structures. Chemical cross-linking using low concentrations of BS 3 of 35 S medium from 3T3-L1 adipocytes, followed by specific immunoprecipitation and SDS-PAGE under reducing conditions, shows mainly dimers and trimers (lanes 1-5, Fig. 4A). Larger concentrations of the BS 3 cross-linking agent generated Acrp30 proteins that migrated as hexamers as well as yet larger species. As extensively crosslinked proteins migrate aberrantly upon SDS-PAGE, it is difficult to determine the exact size of the high molecular mass form indicated by an asterisk. It could represent either a nonamer or a dodecameric structure. Together, panels A and C of Fig. 4 show that Acrp30 forms homotrimers that interact together to generate nonamers or dodecamers. Nonreducing SDS-PAGE reveals that two of the subunits in a trimer are FIG. 2. Acrp30 is a secretory protein found in blood. Acrp30 can be detected by Western blotting in serum from mice; the antibody does not cross-react with calf, human, or rabbit serum. One microliter of fetal calf, rabbit, mouse, and human serum was boiled for 5 min in 2 ϫ sample buffer and analyzed by SDS-PAGE and Western blotting with the anti-Acrp30 antibody according to standard protocols. Antibody was visualized with an anti-rabbit IgG antibody coupled to horseradish peroxidase using a chemiluminesence kit from DuPont NEN.

FIG. 3.
A, insulin stimulation of Acrp30 and adipsin secretion by 3T3-L1 adipocytes. Two 10-cm dishes of day 8 3T3-L1 adipocytes were labeled for 10 min in medium containing [ 35 S]methionine and cysteine as described under "Experimental Procedures." The cells were then incubated in growth medium containing cycloheximide (to prevent further protein biosynthesis) and containing or lacking 100 nM insulin. Every 30 min, the culture medium was removed and replaced with fresh, prewarmed medium containing or lacking 100 nM insulin. The media were subjected to sequential immunoprecipitations with anti-Acrp30 and anti-adipsin antibodies as described under "Experimental Procedures" and analyzed by electrophoresis through a 12% polyacrylamide gel containing SDS. As Acrp30 and adipsin contain a comparable number of cysteine and methionine residues (7 and 9, respectively) and equal exposures of the autoradiograms were used, one can determine from the intensities of the bands that approximately equal amounts of the two proteins are secreted. As judged by the amount of 35  disulfide-bonded together (Fig. 4B), similar to other proteins containing a collagen domain, including the macrophage scavenger receptor (27). Besides being a homo-oligomer, Acrp30 differs from C1q in containing an uninterrupted stretch of 22 perfect Gly-X-X repeats; this suggests that Acrp30 has a straight collagen stalk as opposed to the characteristic kinked collagen domain in C1q caused by imperfect Gly-X-X repeats in two of the three subunits (26). DISCUSSION We do not yet know the function of Acrp30. However, its expression exclusively in adipocytes, its enhanced secretion by insulin, and its presence in normal serum, suggests that it is, like the ob protein, involved in the control of the nutritional status of the organism. Acrp30 is a relatively abundant serum protein, accounting for up to 0.05% of total serum protein as judged by quantitative Western blotting using recombinant Acrp30 as a standard (data not shown). Even though we have no evidence at this stage, we cannot exclude the possibility that Acrp30, like C3 complement released by adipocytes (28), is converted proteolytically to a bioactive molecule.
Our experiments also corroborate the existence of a regulated secretory pathway in adipocytes. We do not yet know whether adipsin and/or Acrp30 are in the same intracellular vesicles that contain GLUT4 and that fuse with the plasma membrane in response to insulin, or whether they are in different types of vesicles. Adipocytes express two members of the Rab3 family, Rab3A and Rab3D (29); these are found in vesicles of different density. Rab3s are small GTP-binding proteins involved in regulated exocytic events. Except for adipocytes, Rab3A is found only in neuronal and neuroendocrine cells; in neurons, Rab3A is localized to synaptic vesicles and is important for their targeting to the plasma membrane (30). An attractive hypothesis under test is that, in adipocytes, Rab3A is localized to vesicles containing Acrp30 and/or adipsin, and that possibly Rab3D mediates insulin-triggered exocytosis of vesicles containing GLUT4. In any case, the mechanism of signal transduction from the insulin receptor to regulated exocytosis of intracellular vesicles remains an important unsolved problem.