Crystal Structure of Inhibitor-Bound Human 5-Lipoxygenase-Activating Protein

See allHide authors and affiliations

Science  27 Jul 2007:
Vol. 317, Issue 5837, pp. 510-512
DOI: 10.1126/science.1144346


Leukotrienes are proinflammatory products of arachidonic acid oxidation by 5-lipoxygenase that have been shown to be involved in respiratory and cardiovascular diseases. The integral membrane protein FLAP is essential for leukotriene biosynthesis. We describe the x-ray crystal structures of human FLAP in complex with two leukotriene biosynthesis inhibitors at 4.0 and 4.2 angstrom resolution, respectively. The structures show that inhibitors bind in membrane-embedded pockets of FLAP, which suggests how these inhibitors prevent arachidonic acid from binding to FLAP and subsequently being transferred to 5-lipoxygenase, thereby preventing leukotriene biosynthesis. This structural information provides a platform for the development of therapeutics for respiratory and cardiovascular diseases.

Leukotrienes are lipid mediators of inflammation that are involved in the pathogenesis of respiratory and cardiovascular diseases (1, 2). Cellular activation by immune complexes and other inflammatory stimuli results in an increase of intracellular calcium and the translocation of cytosolic phospholipase A2 (cPLA2) and 5-lipoxygenase (5-LO) from the cytosol to the nuclear membrane (3). In the presence of the 5-lipoxygenase–activating protein (FLAP), arachidonic acid (AA) released from the nuclear membrane by cPLA2 is delivered to 5-LO for conversion to 5-(S)-hydroperoxy-6,8,11,14-eicosatetraenoic acid (5-HpETE) and then leukotriene A4 (LTA4) (4). Membrane interaction of 5-LO with FLAP is essential for leukotriene biosynthesis (35). FLAP is an integral membrane protein that belongs to the MAPEG (membrane-associated proteins in eicosanoid and glutathione metabolism) superfamily (68). In contrast to other MAPEGs, FLAP has not been shown to have enzymatic activity or to be functionally modulated by glutathione.

The FLAP inhibitor MK-591 inhibits cellular biosynthesis of the neutrophil activator leukotriene B4 (LTB4) and the bronchoconstrictive and vasoactive leukotrienes C4, D4, and E4 (LTC4, LTD4, and LTE4, collectively known as cysteinyl leukotrienes) and has been shown to be efficacious in allergen challenge and chronic asthma clinical trials (9, 10). FLAP inhibitors have also been shown to be protective in acute and chronic animal cardiovascular models, and human polymorphisms have been linked to increased risk of myocardial infraction and stroke (1, 11). These studies suggest that, in addition to their therapeutic benefit in respiratory disease, FLAP inhibitors will have an important future role in the prevention and treatment of cardiovascular disease (1, 10, 11).

FLAP was discovered because it binds MK-886, a leukotriene biosynthesis inhibitor that prevents the binding of AA to FLAP in a concentration-dependent manner (12) and that also blocks the association of FLAP with 5-LO at higher concentrations (13, 14). FLAP functions both as a membrane anchor for 5-LO (3) and as an AA-binding protein (12). How FLAP activates 5-LO is not understood, but it involves a physical interaction between FLAP and 5-LO (15). To understand FLAP's role in leukotriene production and how inhibitors regulate its function, we determined the x-ray crystal structure of FLAP in complex with two leukotriene biosynthesis inhibitors.

The crystal structures of FLAP in complex with MK-591 and an iodinated analog of MK-591 (Fig. 1, A and B) (10, 13) were determined at 4.0Å and 4.2Å resolution (figs. S1 to S5 and table S1). FLAP contains four transmembrane helices (α1 to α4) that are connected by two elongated cytosolic loops (C1 and C2) and one short lumenal loop (L1) (Fig. 1, C and D). The orientations of the cytosolic and lumenal loops of FLAP are consistent with previous studies on MAPEGs (1618). Lumenal helix αL (residues 3 to 8) precedes helix α1 (residues 10 to 37) and is linkedtohelix α1 by a single residue at the lower leaflet of the membrane (Fig. 1C). Cytosolic loop C1 (residues 38 to 47) connects the first helix to the second (residues 48 to 77), and both protrude from the upper leaflet of the membrane. Helix α2 has a distinctive bend at residue P65 at the midpoint of the membrane. Loop L1 (residues 78 to 80) connects the second to the third helix (residues 81 to 101) which is perpendicular to the plane of the bilayer. Loop C2 (residues 102 to 115) connects the third helix to the fourth (residues 116 to 138). Helix α4 begins in the middle of the membrane and continues into the lumen. The C terminus of FLAP is largely disordered beyond residue G140. This region is highly conserved among other FLAP proteins but not in other MAPEGs, suggesting that it may have a role in the biological activity of FLAP. Analysis of sequence similarity indicates that the secondary structure of FLAP is probably shared among other MAPEGs, but FLAP does not display significant structural homology with other protein containing transmembrane helical bundles. Structural comparisons between human FLAP and rat microsomal glutathione S-transferase–1 (mGST-1) produces a root mean square deviation of 3.1 Å for 107 aligned residues (fig. S2, A to C). The proposed glutathione binding site in rat microsomal glutathione transferase 1 (19) is not conserved in FLAP, and there is no electron density in that region of the molecule, consistent with the lack of evidence that FLAP is modulated by glutathione.

Fig. 1.

(A) Chemical structures of MK-591 and (B) an iodinated analog of MK-591. (C) The FLAP monomer viewed parallel to the nuclear membrane with red helices and green loops. The unstructured C terminus of FLAP (residues 141 to 161) extends beyond G140. (D) The FLAP trimer with monomers colored green, cyan, and magenta. The view is given parallel to the nuclear membrane. Bound inhibitor molecules are shown as stick models with yellow carbon atoms, blue nitrogen atoms, red oxygen atoms, and purple iodine atoms.

FLAP crystallizes as a homotrimer (Fig. 1D), consistent with previous studies on MAPEGs (1921). The trimer has a flattened cytosolic top (Figs. 1D and 2B) and a slightly pointed lumenal base (Figs. 1D and 2C). Electron density and temperature factors indicate that the cytosolic loops are flexible. The three-fold axis of the trimer is perpendicular to the plane of the membrane, and it resembles a cylinder 60 Å high and 36 Å wide.

Fig. 2.

Electrostatic surface of FLAP. The view is given (A) parallel to the nuclear membrane, (B) from the cytosol, and (C) from the lumen. One of the three surface grooves has been circled. The cytosolic and lumenal ends of the trimer are positively charged and negatively charged, respectively. (D) Central pocket of FLAP as calculated by CASTp (Computed Atlas of Surface Topography of proteins) (28). Bound inhibitor molecules are shown as stick models with green carbon atoms, blue nitrogen atoms, red oxygen atoms, and purple iodine atoms, and the trimer is shown as a white coil. The front of this pocket has been removed for clarity. The lumenal entrance to this pocket is formed by negatively charged helices. The negatively charged constriction within the membrane is formed by residues Q58 and D62. The surfaces are colored by electrostatic potential with blue and red corresponding to +40 kT and –40 kT.

There are extensive intersubunit contacts, with each monomer burying ∼4900 Å2. On both sides of the membrane and within the bilayer, cytosolic loops C1 and C2, lumenal helix αL, and helices α1, α2, and α4 form an elaborate network of intermolecular contacts. Unexpectedly, helix α3 and the lumenal loop do not form any intersubunit contacts within the membrane. There are also three polar interactions (N23-T66, Q58-Q58 and Q58-D62) between adjacent monomers near the three-fold axis. Hydrophobic residues line the membrane-embedded portion of the intersubunit interfaces and form extensive nonpolar contacts. There is also nonprotein electron density in the lipid-exposed regions that most likely represents bound detergent or lipid molecules.

Three ∼750 Å3 grooves are located between adjacent monomers on the lipid-exposed surface of the trimer (Fig. 2A). Within these grooves, there is clear connected nonprotein electron density and peaks in anomalous difference maps that are consistent with three bound inhibitor molecules (figs. S3 to S5). Inhibitors intercalate between monomers in the surface grooves and form van der Waals interactions with residues V20, V21, G24, F25, and A27 from helix α1; Y112 and I113 from cytosolic loop C2; A63 from helix α2; and I119, L120, and F123 from helix α4 of the adjacent monomer (Fig. 3, A and B). A limited number of weak polar interactions are also formed with residue N23 from helix α1, D62, and T66 from helix α2, and K116 from helix α4 of the neighboring subunit (Fig. 3, A and B).

Fig. 3.

The inhibitor binding site of FLAP in complex with (A) the iodinated analog of MK-591 and (B) MK-591. Monomers are colored green, cyan, and magenta. All side chains within these inhibitors are shown as sticks. Inhibitor molecules are shown as stick models with yellow carbon atoms, blue nitrogens, red oxygens, purple iodines, green chlorines, and orange sulfurs.

Given the resolution of our electron density maps and the unexpected location of the inhibitor binding site, we performed direct mutagenesis to confirm the location of the site by measuring the binding of radiolabeled iodinated analog of MK-591 to mutant forms of FLAP. Mutating residues close (≤4 Å) to the inhibitor cause a dramatic decrease in binding affinity (Fig. 3A and table S2). Unexpectedly, several directed mutations increase inhibitor binding affinity (V20A, I113A, K116A, and Δ1-10). The insensitivity of mutations distant from the bound inhibitor (>4 Å) suggests that the observed loss in binding affinity is a direct result of mutating proximal residues and not due to aberrant folding. Collectively, these data support our identification of the inhibitor binding site.

Previous studies defined regions of FLAP that are important for binding of inhibitors (table S3) (2224). When viewed in the context of the FLAP structure, most of the loss-of-function deletions are at the cytosolic surface. Jointly removing the cytosolic portions of helices α1 and α2 and cytosolic loop C1 would substantially alter the FLAP structure, providing a likely explanation for the observed loss in inhibitor binding. In contrast, the mutations that do not significantly affect activity involve removing a short cytosolic piece of helix α1 and/or cytosolic loop C1 or three residues from cytosolic loop C2. These mutations would cause a minimal structural perturbation and would not affect inhibitor binding.

Perhaps the most unexpected feature of the FLAP structure is an internal ∼3,200 Å3 pocket at the bottom of the trimer and open to the lumen (Fig. 2, C and D). The entrance to this pocket has an internal diameter of ∼6 Å and is formed by the lumenal helices and loops that define three distinct entry crevices on the lumenal surface of the trimer (Fig. 2, C and D). The pocket extends to about the middle of the membrane, where it is constricted by a symmetric ring of P65 residues (minimum radius of ∼0.5 Å) (25), which coincides with the bend in helix α2. Proline-induced distortions in transmembrane helices are known to be involved in transmembrane signal transduction and solute transport by acting as molecular hinges (26). An additional symmetric ring, formed by residues Q58 and D62 from helix α2, sits above this point. The asymmetric distribution of the electrostatic potential of this pocket is striking: Its lumenal entrance and membrane-embedded constriction point are negatively charged, whereas the interior lining is largely hydrophobic (Fig. 2, C and D).

By combining the structure presented here with previous data, we can propose a model to explain the selective transfer of AA to 5-LO by FLAP and how FLAP inhibitors prevent this process. Cellular activation initiates the calcium-dependent translocation of cPLA2 and 5-LO from the cytosol to the nuclear membrane. The N terminus of 5-LO contains a calcium-binding domain that presumably anchors it to the nuclear membrane adjacent to FLAP. It is reasonable to assume that the C-terminal catalytic domain of 5-LO binds to the cytosolic loops of FLAP. Previous FLAP mutagenesis studies show partial overlap between the AA and inhibitor binding sites of FLAP (2224). Our structure shows that the inhibitor binding sites are located within nuclear membrane, which provides an appropriate environment for the lateral diffusion of AA molecules to FLAP. This event may be coupled to structural changes in FLAP, specifically the relative conformation of helix α2, the constriction formed by the symmetrical ring of P65 residues, and the negatively charged symmetrical ring of Q58 and D62 residues. We hypothesize that one 5-LO molecule binds to each FLAP trimer and that AA molecules are sequentially transferred to 5-LO. When loaded with AA, 5-LO converts AA to HpETE and then to LTA4, which can be either converted to LTB4 by leukotriene A4 hydrolase, exported from cells for transcellular metabolism, or released back into the nuclear membrane for conversion to LTC4 by leukotriene C4 synthase (2, 4, 5). The lateral transfer of LTA4 from FLAP to leukotriene C4 synthase within the nuclear membrane is consistent with recent observations showing that these MAPEGs form a macromolecular complex (27).

Leukotrienes have an established pathological role in allergic and respiratory diseases. Animal and human genetic evidence suggests they may also have an important role in atherosclerosis, myocardial infarction, and stroke. The structure of FLAP provides a tool for the development of novel therapies for respiratory and cardiovascular diseases and for the design of focused experiments to probe the cell biology of FLAP and its role in leukotriene biosynthesis.

Supporting Online Material

Materials and Methods

Figs. S1 to S5

Tables S1 to S3


References and Notes

View Abstract

Navigate This Article