Title: Crystal structure of the biotin carboxylase domain of human acetyl‐CoA carboxylase 2
Abstract: Acetyl-CoA carboxylase (ACC) catalyzes the carboxylation of acetyl-CoA to produce malonyl-CoA. In prokaryotes, ACCs are multisubunit enzymes consisting of biotin carboxylase (BC), biotin carboxyl carrier protein (BCCP), and carboxyl transferase (CT) as separate proteins, whereas eurokaryotic ACCs are multidomain enzymes containing the three domains of each catalytic activity within a single polypeptide chain in order of BC, BCCP, and CT domain.1 The BC domain catalyzes the ATP-dependent carboxylation of the biotin moiety incorporated in the BCCP domain and the CT domain transfers the carboxyl group from the biotin moiety to acetyl-CoA to produce malonyl Co-A.2 Mammalian ACC exists as two tissue-specific isozymes: ACC1 present in lipogenic tissues (liver and adipose) and ACC2 present in oxidative tissues (liver, heart and skeletal muscle).3-5 ACC1 is involved in the biosynthesis of fatty acids and the malonyl-CoA produced by ACC1 is used as a building block to extend the chain length of fatty acids by fatty acid synthase (FAS).1, 6, 7 In contrast, malonyl-CoA produced by ACC2 acts as an inhibitor of carnitine palmitoyltransferase (CPT I), an enzyme that imports fatty acids into the mitochondria for β-oxidation to acetyl-CoA.6, 8-11 Experimental results with ACC2 knockout mice suggest that the stimulation of β-oxidation by reducing malonyl-CoA in oxidative tissues can be a good strategy for the treatment of obesity and type-2 diabetes. The reduced level of malonyl-CoA caused continuous fatty acid oxidation and reduced body fat and body weight of the mice, in spite of more food consumption. In addition, these animals were protected against diabetes and obesity induced by high-fat/high-carbohydrate diets. These observations suggest that ACC2 inhibitors may be potential therapeutic agents for the treatment of obesity, diabetes, and metabolic syndrome.12-15 In mammals, ACC activities of both isoforms are known to be controlled by AMP-activated protein kinase (AMPK). ACCs are phosphoproteins and their regulation by reversible phosphorylation is crucial to the control of fatty acid synthesis and oxidation.16 AMPK inactivates ACCs in vivo by phosphorylating specific serines in the N-terminus.17, 18 In humans, the major AMPK phosphorylation sites in ACC1 and ACC2 are Ser117 and Ser222, respectively [Fig. 1(C)].7, 19 The major role of AMPK in the cell is to monitor energy status and regulate the ATP-consuming and ATP-producing pathways accordingly. Since ACCs are involved in energy-related pathways of fatty acid synthesis and oxidation by consuming ATP, they are prime physiological targets for AMPK.6 Here, we report the crystal structure of the biotin carboxylase (BC) domain of human ACC2 (residues, 217–775) containing Ser222, the AMPK phosphorylation site. It is the first structure of a BC domain from a mammalian ACC. In eukaryotes, only one BC domain structure has been reported to date, that of yeast.20 However, yeast ACC is not regulated by AMPK and does not have a serine residue for phosphorylation by AMPK. The N-terminal region of the BC domain of human ACC2 is disordered in the structure; however, this flexibility implies a critical role for this region in the AMPK regulation. Structure of the BC domain of human ACC2. (A) The BC domain is shown in a ribbon model and the disordered regions (residues, 217–219, 225–238, 412–417, 657–665, and 687–698) shown as dotted lines. The disordered Ser222 is labeled. The four domains are colored individually. The N- and C-terminal regions are also indicated. (B) Structural comparison between the soraphen A binding site of human BC domain without soraphen A (purple) and that of yeast BC in complex with soraphen A (green). (C) Structural comparison between the BC domains from human ACC2 (purple) and yeast ACC (green), when the A- and C-domains of the two proteins are superimposed. To emphasize the structural differences, the B-domain, AB linker, and N-terminus are colored more strongly and labeled. The arrow indicates the putative ATP-binding site. (D) Sequence alignment of the BC domains of human ACC2, ACC1, rat ACC2, and yeast ACC. Conserved residues are shown in red. The serine residues of mammalian ACCs that are phosphorylated by AMPK are shown in green. The disordered residues of the N-terminus in the structure of the human ACC2 BC domain are indicated by a yellow box. The DNA encoding human acetyl-CoA carboxylase (ACC)2 BC domain (residues, 217–775) was amplified by the standard PCR-based cloning strategy. The PCR product was annealed into a modified pET21b expression vector encoding additional residues MRGSGS at the N-terminus and LEHHHHHH at the C-terminus, and its identity was confirmed by sequencing. This plasmid was transformed into BL21 Escherichia coli cells. The transformant cells were grown in LB medium at 37°C up to an A600 nm of 0.6. The protein expression was induced by adding 1 mM IPTG, and the cells were grown for 18 h. Cells were harvested by centrifugation and the cell pellet was resuspended in the lysis buffer (20 mM Tris-HCl pH 7.5, 200 mM NaCl, 1 mM PMSF, 14 mM β-mercaptoethanol, and 10% glycerol) and subjected to repeated sonication with intervals of cooling. The clarified cell lysate was applied to Ni-NTA resin (Qiagen), washed with buffer A (20 mM Tris-HCl pH 7.5, 200 mM NaCl, 10 mM imidazole, 7 mM β-mercaptoethanol, and 10% glycerol), and the bound protein was eluted with buffer A plus 500 mM imidazole. The eluted protein was further purified by gel filteration using Superdex-200 (Pharmacia) equilibrated against the buffer containing 25 mM HEPES (pH 7.5), 150 mM NaCl, 5% Glycerol, and 2 mM DTT. The purity of the pooled fractions was checked by SDS-PAGE, and the protein sample was concentrated to 12 mg/mL for crystallization. The protein was crystallized at 22°C by vapor diffusion using hanging-drop method and a protein to reservoir solution ratio of 1:1, with the reservoir solution containing 1.5–1.8M ammonium sulfate and 0.1M MES (pH 6.0–6.8). Prior to X-ray data collection, crystals were transferred to the cryoprotectant solution containing the reservoir solution plus 25% (v/v) glycerol for a few seconds, then looped from the drop and flash-frozen in liquid nitrogen. The X-ray diffraction data for the crystal were collected at 4A beam line of Pohang Light Source (PLS) and processed using HKL2000.21 The crystal belongs to the space group P3221 and the unit cell parameters are a = b = 75.80 Å, c = 189.01 Å. The predicted Matthews coefficient is 2.46 Å3/Da, assuming a monomer per asymmetric unit, corresponding to ∼50% solvent content by volume. Crystallographic data collection statistics are given in Table I. The structure of the BC domain of human ACC2 was solved by molecular replacement (MR) using the program CNS22 with the structure of yeast BC domain (PDB entry 1W96) as a search model. The model shares 64% sequence identity with the BC domain of human ACC2. Using the data with a resolution range of 20.0–2.5 Å, the MR solution structure was refined with the CNS package and manual model building was performed using the program QUANTA [Accelrys (San Diego)]. Subsequent rounds of model adjustment, simulated annealing, and thermal parameter refinement were performed. During refinement, 5% of reflection data chosen randomly from the observed data were used for cross-validation with the Rfree value. Water molecules were automatically picked up by the CNS package, and confirmed based on peak height and distance criteria in the Fo − Fc and 2Fo − Fc maps. The quality of the refined structure was verified with MolProbity.23 The coordinates and structure factors have been deposited in PDB (http://www.rcsb.org/pdb) under the accession code of 2HJW. The crystal of human ACC2 BC domain belongs to the space group P3221 with unit cell parameters of a = b = 75.80 Å, c = 189.01 Å, containing one monomer in the asymmetric unit. The refined model of human ACC2 BC domain gave Rfactor and Rfree values of 0.212 and 0.248, respectively. The final structure contains one molecule of human ACC2 BC domain with residues 239–411, 418–656, 666–686, 699–759, and 87 water molecules. Table I summarizes the statistics for structure refinement. The overall architecture of human BC domain is similar to that of yeast BC domain, which is consistent with their high sequence homology (64% identity). The overall structure of human BC domain consists of three subdomains: A-domain (residues, 217–375), B-domain (residues, 436–496), and C-domain (residues, 497–775), as well as the AB linker (residues, 376–435) which connects the A- and B-domains [Fig. 1(A)]. The structures of prokaryotic BC subunit24, 25 and eukaryotic BC domain20 have the ATP-grasp fold, which is common in several enzymes with biochemical functions related to ATP-dependent acylation.26, 27 The putative active site of the human BC domain, where the ATP hydrolysis occurs, is located between the B- and C-domain. The B-domain is a lid over the active site and is thought to undergo a large conformational change during catalysis. In the structure of E. coli BC, the conformational change of the B-domain occurs and the active site is closed upon ATP binding, whereas the apo form of the enzyme shows an open active site and a more disordered conformation of the B-domain.24, 25 In the crystal structure of yeast BC domain, the B-domain is mostly in the closed conformation even in the absence of ATP in the active site.20 Here, the ATP-free human BC domain assumes a more closed conformation of B-domain than that of yeast BC domain [Fig. 1(C)]. This might be an artifact of the crystal packing, since there are several interactions between the B-domain and a symmetry related C-domain. Ser482 of the B-domain makes H-bond with the backbone amide group of Arg710 of a symmetry related C-domain (3.2 Å). Lys426 makes H-bonds with the backbone carbonyl groups of the symmetry related Thr537 (3.2 Å) and Ile538 (2.8 Å). In addition, Ile475 and Leu476 make hydrophobic contact with the symmetry related Leu643 (3.9–4.3 Å). Another difference between the structures of the human and yeast BC domains is at the N-terminus. Sequence analysis shows that the homology between yeast ACC and mammalian ACCs is very low in this region [Fig. 1(D)], implying structural differences between their N-termini. The N-terminus of yeast BC domain does not contain a serine residue for phosphorylation by AMPK due to the lack of a regulatory mechanism by AMPK in yeast or bacterial ACCs unlike mammalian ACCs. The disorder of the N-terminus (residues, 217–238) in the structure of human BC domain may be due to the high flexibility of the loop containing the phosphorylation site, whereas the N-terminus of the yeast BC domain is well-structured and an integral part of the A-domain. The high content of basic residues in the loop connecting the phosphorylation site might be responsible for its high flexibility. This high flexibility could play a critical role when Ser222 is phosphorylated by AMPK by facilitating a structural change for the regulation of activity. Of course, we cannot exclude the possibility that the disorder of the N-terminus of human ACC2 BC domain may be an artifact of the construct, which lacks the rest of the protein. The N-terminus could make specific interactions in the full length ACC2 with the BC, BCCP, and CT domains. The bacterial BC subunits are dimers but human ACC2 BC domain does not show a similar dimer interface in the crystal packing. Like the yeast BC domain, human ACC2 BC domain shows large conformational differences in the residues of the region as compared to the bacterial BC subunits, which can explain why mammalian BC domains are monomeric in solution and catalytically inactive.20, 28 In addition, the flexible N-terminus might inhibit dimer formation. The crystal structure of the yeast BC domain in complex with soraphen A, a natural product with a potent inhibitory activity against eukaryotic ACCs,29-31 showed that the inhibitor binds to the putative dimer interface of yeast BC domain and so might inhibit the BC activity allosterically by disrupting dimer formation.20 Because of the conservation of the residues involved in binding soraphen A, the structure of the soraphen A binding site of yeast BC is well superimposed to that of the putative soraphen A binding site in human ACC2 BC. However, the side chain conformations of Trp681 and Met594 in human ACC2 BC are unfavorable for soraphen A binding due to a steric clash, whereas the corresponding residues in yeast BC (Trp487 and Met393) assumed suitable conformations for recognizing soraphen A prior to binding [Fig. 1(B)]. The Kd value of soraphen A for the BC domains of human ACC1 and ACC2 is ∼1 nM.28 This high binding affinity is mainly due to the extensive interactions between soraphen A and the human BC domain as in the case of yeast BC domain in complex with soraphen A. Accordingly, conformational changes of Trp681 and Met594 in human ACC2 BC are required to accommodate soraphen A with the high binding affinity. The structure of the human ACC2 BC domain highlights both similarities and important differences from those of the yeast BC domain and the bacterial BC subunit, and further biochemical and structural studies can elucidate the mechanism of AMPK regulation of ACC2. ACCs have been highlighted as therapeutic targets for obesity and diabetes as they play crucial roles in fatty acid metabolism. This structure could facilitate the discovery of human ACC2 inhibitors for the treatment of obesity and diabetes. We thank Dr. H. S. Lee and G. H. Kim for assistance during the data collection at the beamline 4A of Pohang Light Source (PLS).