|Home | About | Journals | Submit | Contact Us | Français|
X-ray crystal structural determination of FABP4 in complex with four inhibitors revealed the complex binding modes, and the resulting observations led to improvement of the inhibitory potency of FABP4 inhibitors. However, the detailed structure–activity relationship (SAR) could not be explained from these structural observations. For a more detailed understanding of the interactions between FABP4 and inhibitors, fragment molecular orbital analyses were performed. These analyses revealed that the total interfragment interaction energies of FABP4 and each inhibitor correlated with the ranking of the Ki value for the four inhibitors. Furthermore, interactions between each inhibitor and amino acid residues in FABP4 were identified. The oxygen atom of Lys58 in FABP4 was found to be very important for strong interactions with FABP4. These results might provide useful information for the development of novel potent FABP4 inhibitors.
Fatty acid-binding proteins (FABPs) are cytoplasmic proteins that bind to hydrophobic ligands, such as long chain fatty acids, in a noncovalent and reversible manner. It has been suggested that FABPs act as chaperone proteins for fatty acids and play important roles in homeostasis of fatty acids and the lipid signaling pathway.1 The protein FABP4 (aka aFABP or aP2), a member of the FABP family, is a 14.6 kDa cytosolic protein mainly expressed in macrophages and adipocytes. Previous studies have reported that FABP4-deficiency in genetic and diet-induced obese mice show protection against the development of hyperinsulinemia and insulin resistance.2,3 An FABP4-deficiency in apolipoprotein E-deficient mice also shows protection against the development of atherosclerosis.4 Furthermore, treatments with prototype small molecular inhibitors for FABP4 attenuate fatty infiltration in livers of ob/ob mice.5 These studies have demonstrated that FABP4-deficiency is related to several diseases, including diabetes, atherosclerosis, and liver disease, and that potent, selective, small molecular inhibitors for FABP4 might be potential therapeutic agents for these diseases. Recently, a short series of FABP4 inhibitors have been reported, with most of them identified by a structure-based drug design (SBDD) method.6−10 Thus, the crystal structures of several ligand-bound human FABP 4s have been determined,5,6,11 which provided insights into the structural foundation underlying the binding modes of endogenous ligands and small molecular inhibitors in the FABP4 binding site. Structurally, FABP4 contains 10 antiparallel β-strands forming a β-clam shell structure, and its characteristic structure provides a substrate binding site.12−19 Considering these results, ligands lie in the FABP4 internal pocket and usually have polar interactions between FABP4’s Arg126 and Tyr128 side chains and the ligand carboxylic acid. In addition, the importance of a negatively charged substituent, such as a CO2–, SO3–, or PO3– group, in these inhibitors has been clarified.20
The purpose of this study was to identify an FABP4 inhibitor as a new drug candidate. Initially, our in-house compound library was screened using a human FABP4 binding assay. We identified a hit compound that possessed an indole core structure (data not shown). Furthermore, 300 compounds were designed based on this compound and assayed for their binding activity against FABP4. Novel FABP4 inhibitors were thus identified (Table 1). The complex crystal structures of FABP4 bound with four novel inhibitors were then determined, and this information led to selection of improved FABP4 inhibitors, compounds 1 and 2 (Table 1). Among the four inhibitors, compounds 1, 2, and 3 possessed activities sequentially from high to low, and compound 4 showed the lowest activity (Table 1, Ki values). As a control, the Ki value of the reference compound, BMS309403,5 was 0.16 μM using the same assay.
However, it was difficult to understand the structure–activity relationship (SAR) by observation of only hydrogen bond interactions. For more detailed consideration, these interactions were investigated using the fragment molecular orbital (FMO) method,21−24 which has been developed for accurate analysis of large biomolecules, such as protein–ligand complexes. Using the calculated interfragment interaction energies (IFIE)22−28 between these four inhibitors and FABP4 amino acid residues, the interaction differences were described in detail to allow the development of new FABP4 inhibitors.
Before the crystallized complex structures were determined, the question was examined as to whether FABP4 maintained its structure and binding activity to palmitate, the native lipid substrate, when bound to compound 3, the lead candidate inhibitor identified by nuclear magnetic resonance (NMR). A 1H–15N heteronuclear single quantum correlation (HSQC) spectrum (Figure S1) of apo-FABP4 showed approximately 120 signals, which equaled the number of FABP4 amino acid residues. The well-dispersed signals in the spectrum suggested that conformation was maintained and implied a β-sheet-rich structure, as has been reported for the apo-FABP4 structure (PDB code, 3RZY).29 The HSQC spectrum of FABP4 with palmitic acid showed that the addition of palmitic acid caused chemical shift perturbations in almost half of the FABP4 signals. These changes might have arisen from changes in the intramolecular hydrogen-bond network beyond the binding interface, as has been described in the NMR evaluation of FABP4 with R- and S-ibuprofen.30 This suggested that palmitic acid bound to FABP4 through the same binding mode as the ibuprofens. The lead inhibitor candidate, compound 3, was subsequently subjected to an FABP4/compound 3 complex NMR analysis. In the resulting HSQC spectrum, almost half of FABP4’s signals were shifted, which strongly suggested that compound 3 bound to FABP4 in the same manner as palmitic acid (Figure Figure11). From these results, FABP4/compound 1–4 complexes were prepared and used for crystallization analysis.
Next, the crystallographic structure of FABP4 was solved as complexes with the four inhibitors (compounds 1–4), and the interactions analyzed to provide useful information for the development of novel potent FABP4 inhibitors. The 31 residues located within 5.0 Å of any atom of compound 1, the most potent inhibitor, were aligned. Among the complexes, the binding sites were found to be very similar, as demonstrated by the root-mean-square deviation value of ~0.3 Å. All four compounds possessed a carboxyl group that generally formed hydrogen bonds with FABP4’s Arg126 and Tyr128 (Figure Figure22), but the other moieties also showed distinct interactions. These observations clearly indicated that FABP4 had more interactions with compounds 1 and 2 than with compounds 3 and 4. The effective interactions of FABP4 with inhibitors were further elucidated by observing in detail the interactions of FABP4 and the characteristic moieties of compounds 1 and 2. Two hydrogen bonds with Ser55 and Lys58 and four hydrophobic interactions with Ala36, Pro38, Phe57, and Ala75 were observed in compound 1’s pyrazole substructure (Figure Figure22a). Meanwhile, one hydrogen bond with Ser55 and five hydrophobic interactions with Ala33, Ala36, Pro38, Lys58, and Ala75 were observed in compound 2’s 2-methoxypyridine moiety (Figure Figure22b). Observing only the numbers of hydrogen bonds and hydrophobic interactions in these crystal structures was insufficient to explain the precise activity differences between compounds 1 and 2. More information regarding binding modes and potencies was obtained via the FMO method, which was expected to allow calculation of the precise interaction energies in these complexes.
The validity of the FMO method was confirmed by analyzing the correlation between IFIE and Ki values. The IFIE is the sum of the interaction energy components, as electrostatic, exchange repulsion, charge transfer and mixing term, and dispersion energies (ES, EX, CT + mix, and DI, respectively), and was calculated for each amino acid residue. IFIEs of all amino acid residues were totaled to create the IFIE sum. Negative values represented effective interaction energies for binding an amino acid residue with a ligand. The IFIE sum and Ki of the four inhibitors for FABP4 are shown in Table 1. The correlation coefficient between log(Ki) and IFIE sum was 0.915, indicating that they correlated completely. Therefore, the FMO method was considered to be a valid argument for assessing the activities observed in this study.
IFIEs between each compound and FABP4 amino acid residues are shown in Figure Figure44. An amino acid residue was treated as a single fragment by separating the carbonyl group and α-carbon. The main interactions of residues from Lys52 to Asp77 are shown magnified in Figure Figure55.
Compound 1 exhibited strong interactions with Ser55, Phe57, and Lys58 (fragment 59), which enabled compound 1 to have the smallest Ki value among the four inhibitors. Notably, compound 1 formed hydrogen bonds with the Ser55 side chain oxygen atom and Lys58 main chain (Figure Figure33a), and the related IFIEs were the lowest observed, at −12.33 and −14.96 kcal/mol, respectively (Table S1). A hydrogen bond with Lys58 has been described in the docking model of Pimozide.31 It is notable that this interaction was detected here in the X-ray crystal structure, indicating its importance in the present molecular binding as well as the utility of the FMO method. Phe57 also formed pi–pi T-shaped interactions (Figure Figure22a), but the related strengths were not clear. The IFIE of Phe57 revealed strong interactions, at −9.79 kcal/mol. In addition, the IFIEs of fragment 76 (Asp76) represented the largest repulsive force of all IFIEs. Among Asp76 IFIEs, the IFIE between compound 1 and Asp76 was the smallest. The moieties of compound 1 near Asp76 were cyclopropanes, according to the X-ray crystal structure, and cyclopropane was a common moiety in compounds 1, 2, and 3. Comparing the positions of each cyclopropane, the distance between compound 1 cyclopropane and Asp76 was ~0.4 Å longer than that of compounds 2 and 3 (Figure S2). This greater distance was probably because compound 1 was anchored by hydrogen bonds with Ser55 and Lys58 and assumed to be the main reason the repulsive energy between compound 1 and Asp76 was the lowest.
The situation for compound 2 was a bit complicated because this compound hydrogen-bonded with Ser55 in the crystal structure (Figure Figure33b) and the IFIE was −7.93 kcal/mol. However, the largest IFIE for compound 2 was −9.83 kcal/mol, with FABP4 fragment 58 (Table S1). While, the Lys58 oxygen atom formed a CH···O hydrogen bond32 with compound 2 in the crystal structure (Figure Figure22b), the amine group of Lys58’s main chain interacted strongly, as ES energy (−7.30 kcal/mol). Furthermore, Phe57’s IFIE was also large (−5.39 kcal/mol), as determined from FMO measurements (Figure Figure55), but no interactions were recognized from X-ray structure analyses (Figure Figure22b). As FMO analysis calculated interaction energies based on electronic states, the interactions could not be recognized using only distances and degrees. In summary, compound 1’s IFIE sum was smaller than that of compound 2.
In this way, the SARs of compounds 1 and 2 could be explained on the basis of interactions with amino acid residues and energies calculated from FMO analyses. These results showed that an FABP4 inhibitor required a strong interaction with Lys58, and thus, compound 1, which possessed strong interactions with Lys58, also possessed high inhibition activity.
Adding the feature of FMO analysis, aP2 fragment 53 (Ser53) exhibited the largest IFIE (−9.64 kcal/mol; Figure Figure55, green bar) with compound 3 although no interactions with Ser53 were recognized from X-ray crystal structural analysis (Figure Figure22c). As the main IFIE component with compound 3 and Ser53 was DI energy (−6.04 kcal/mol, Table S1), this energy could not be recognized from distances and angles of X-ray crystal structures. In addition, Ser53’s IFIE was greater than that of Arg126 and Tyr128 (−2.16 and −1.21 kcal/mol, respectively), which unexpectedly formed hydrogen bonds.
In conclusion, aP2 crystallographic structures were solved as complexes with four inhibitors and the interactions were described by FMO analyses. These analyses were very helpful in determining why compound 1 possessed the most potent inhibitory activity. Furthermore, the importance of the interaction with Lys58 was revealed, contributing to the understanding for the design of a strong inhibitor, which will be useful for future inhibitor identification efforts.
We extend our deepest appreciation to Hitoshi Wakatsuki for their technical assistance. We thank the staff of the Photon Factory (Tsukuba, Japan) for assistance during data collection. We also thank Naoyuki Fukuchi, Ajinomoto Pharmaceuticals Co., Ltd. for helpful discussions. Finally, we thank Kaori Fukuzawa, Nihon University, and Takayuki Tsukamoto, Mizuho Information & Research Institute.
All authors have given approval to the final version of the manuscript.
The authors declare no competing financial interest.