Structure of Human Methionine Aminopeptidase-2 Complexed with Fumagillin

See allHide authors and affiliations

Science  13 Nov 1998:
Vol. 282, Issue 5392, pp. 1324-1327
DOI: 10.1126/science.282.5392.1324


The fungal metabolite fumagillin suppresses the formation of new blood vessels, and a fumagillin analog is currently in clinical trials as an anticancer agent. The molecular target of fumagillin is methionine aminopeptidase-2 (MetAP-2). A 1.8 Å resolution crystal structure of free and inhibited human MetAP-2 shows a covalent bond formed between a reactive epoxide of fumagillin and histidine-231 in the active site of MetAP-2. Extensive hydrophobic and water-mediated polar interactions with other parts of fumagillin provide additional affinity. Fumagillin-based drugs inhibit MetAP-2 but not MetAP-1, and the three-dimensional structure also indicates the likely determinants of this specificity. The structural basis for fumagillin's potency and specificity forms the starting point for structure-based drug design.

Angiogenesis, the growth of new blood vessels, is a pathological determinant in tumor progression, diabetic retinopathy, and rheumatoid arthritis (1). The serendipitous discovery that fumagillin, a fungal metabolite, potently inhibits angiogenesis initiated the systematic development of small molecule angiogenesis inhibitors (2, 3) (Fig. 1). One semisynthetic derivative of fumagillin, TNP-470, is in clinical trials as an anticancer agent (Fig. 1) (3, 4). Fumagillin-based affinity reagents identified MetAP-2 as the specific cellular target of fumagillin, and this specificity was confirmed with genetically altered yeast strains (5, 6). The correlation between the antiproliferative activity of several fumagillin analogs with their ability to inhibit MetAP-2 activity in vitro suggests that MetAP-2 is the physiologically relevant target of fumagillin-based therapeutic agents (6). This suggestion is strengthened by a recent report that human endothelial cells are especially sensitive to fumagillin and that proliferation of these cells can be blocked by human MetAP-2 antisense oligonucleotides (7). MetAPs, which remove NH2-terminal methionines from proteins in a nonprocessive manner, are highly conserved in sequence (8) (Fig. 2). These cobalt-containing metalloproteases are divided into two families, type 1 (MetAP-1) and type 2 (MetAP-2), and fumagillin inhibits the in vivo activity of only MetAP-2 (5, 6).

Figure 1

The chemical structure of fumagillin and TNP-470.

Figure 2

Structure-based sequence alignment of selected MetAP sequences based on the three-dimensional structure of E. coli MetAP-1 (M1-ec) (19), P. furiosus MetAP-2 (M2-pf) (20), and human MetAP-2 (M2-hu). Numbers on top are sequence alignment numbers and those on the bottom are the sequence numbers of HsMetAP-2 used here. Yeast MetAP-1 (M1-y) and MetAP-2 (M2-y) sequences are from Swiss-Prot (ID numbers Q01662 and P38174, respectively). Similar and identical sequences are shaded by light blue and yellow, respectively. Red triangles are residues involved in metal coordination, and blue triangles are residues in contact with fumagillin. Figure prepared with Alscript (17).

To investigate fumagillin's inhibitory mechanism, we determined the crystal structure of human MetAP-2 with and without bound fumagillin. Human MetAP-2 (HsMetAP-2) was expressed in Sf21 insect cells (9). Crystals of HsMetAP-2 were prepared, diffraction data were collected at the F1 station of CHESS, and the structure was solved by the molecular replacement method (10, 11). The structure has been refined to a final R factor of 0.183 and 0.194 for native and fumagillin-complexed HsMetAP-2, respectively, for the 1.8 to 25.0 Å data (Fig. 3A) (11).

Figure 3

(A) Electron density (|F complexF native|, native phases, 3σ contour level) of fumagillin in the catalytic pocket of human MetAP-2. An atomic model of the final structure is embedded in this electron density. (B) Overall structure of the complex between human MetAP-2 (red, green, and blue) and fumagillin (yellow and red ball and stick). The two metals at the catalytic site are dark blue spheres partly obscured by fumagillin. Secondary structural elements of MetAP-2 are labeled. Drawing prepared with Molscript (18).

Earlier crystallographic studies of the MetAP-1 from Escherichia coli (EcMetAP-1) (12) and the MetAP-2 fromPyrococcus furiosus (PfMetAP-2) (13) have defined the overall topology of the MetAP family. Like EcMetAP-1 and PfMetAP-2, HsMetAP-2 has a central β sheet with an active site located roughly at the center of the sheet's concave face (Fig. 3B). Two pairs of α helices (α1-α2, α3-α4) and a short COOH-terminal tail cover the sheet's convex face (Fig. 3B).

HsMetAP-2 has several features that distinguish it from the prokaryotic MetAPs (Fig. 2). HsMetAP-2, unlike EcMetAP-1 and PfMetAP-2, has a 165-residue NH2-terminal extension, which is not essential for aminopeptidase activity (6). In the structure described here, the NH2-terminal extension is largely disordered, and clear electron density begins at Lys110 with a disordered loop from residues 138 to 153 (Fig. 3B). The visible portion of the NH2-terminal extension lies on the convex surface near helices α1 and α2 and far from the active site.

Residues 381 to 444, the long insertion that distinguishes the MetAP-2 family from the MetAP-1 family (Fig. 2), comprise the end of β7, α5, α6, α7, and the beginning of β8 and form a compact domain that does not interact significantly with the rest of the protein (Fig. 3B). A small insertion, which includes β4 and the beginning of β5 (residues 312 to 319), distinguishes eukaryotic from prokaryotic MetAP-2. Neither of these insertions disrupts the MetAP-1 secondary structure (Figs. 2 and 3B). The insertions and NH2-terminal extension in HsMetAP-2 break the pseudo twofold symmetry of EcMetAP-1 (12).

The active site is a deep pocket with two cobalts at its base and a completely covered side pocket that presumably serves as the specificity pocket for the NH2-terminal methionine side chain of natural substrates. Reorientation of the Tyr444side chain and some water molecules opens this pocket to solvent. The Tyr444 residue is completely conserved in the MetAP-2 family and comes at the end of the long insertion that distinguishes the MetAP-2 family (Fig. 2). In the absence of fumagillin, the cobalts are coordinated by Asp251, Asp262, His331, Glu364, Glu459, and a water molecule. The Asp262 and Glu459 residues are bidentate ligands coordinating both cobalts, Asp251 is bidentate with cobalt 1, Glu364 and His331 are monodentate with cobalt 2, and one clearly defined water molecule interacts with cobalt 2. All of the residues that coordinate the metals are on β strands near the center of the β sheet and are conserved in all MetAP sequences (Fig. 2).

The electron density for fumagillin was clearly visible in the difference electron density synthesis (Fig. 3A). Fumagillin has several structural components arrayed around its conformationally fixed cyclohexane ring, and a comparison of the active site of HsMetAP-2 with and without fumagillin shows how each component contributes to binding (Fig. 4, A and B). Earlier work had established that covalent bond formation causes fumagillin's irreversible inhibition, and the x-ray structure shows a covalent bond between an imidazole nitrogen (Nɛ2) atom of His231 and the carbon of the spirocyclic epoxide (5, 6) (Fig. 3B). The formation of this C–N bond, although not predicted, is analogous to the alkylation of a catalytic histidine by α-chloroketone inhibitors of serine proteases (14). Histidine-231 does not move significantly upon bond formation; its nucleophilic imidazole nitrogen is perfectly positioned to bond with the methylene of the epoxide (Fig. 4B). The oxygen liberated from the breaking of the epoxide is coordinated with cobalt (3.28 Å), and it occupies the approximate position of the cobalt-associated water molecule in the uncomplexed structure. A water that is equidistant from both cobalts forms a hydrogen bond with this fumagillin oxygen (Fig. 4A). The only residue that moves significantly upon fumagillin complexation is His339, which rotates its side chain to avoid close contacts with fumagillin (Fig. 4B).

Figure 4

(A) LIGPLOT of fumagillin in the binding pocket (19). Black dashed lines indicate hydrogen bonds, and red radial lines indicate hydrophobic contacts. Water molecule number 556 is 2.89 Å from the methoxyl oxygen and 2.88 Å from the epoxide oxygen. Water molecule number 642 is 2.74 Å from a fumagillin oxygen, and 2.1 and 2.3 Å from each of the metals. (B) Fumagillin in the active site of HsMetAP-2. Fumagillin and the covalently attached His231 are depicted as ball and stick with carbon yellow, oxygen red, and nitrogen blue. Side chains that interact with fumagillin are drawn as sticks, and the color represents native protein (green) or complex (blue). (C) A hypothetical model of fumagillin in the binding pocket of EcMetAP-1. The side chains of MetAP-1 that would prevent fumagillin's binding in different models are shown in blue; residues that would interact with fumagillin are in green. The long unsaturated side chain has been omitted for clarity. Drawing was prepared with Molscript (18).

The epoxide-bearing side chain of fumagillin occupies the completely covered pocket near the active site (Fig. 4B). It has hydrophobic contacts with His331 at the mouth of the pocket, Tyr444, Ile338, His339, and Phe219 (Fig. 4B). A well-defined water molecule forms hydrogen bonds with the side chain epoxide and the methoxyl group at C5 (Fig. 4A). The long unsaturated side chain protrudes from the binding pocket and makes two hydrophobic contacts with Leu328 and Leu447, and both residues are conserved in the MetAP-2 family (Figs. 3 and 4A). Leucine-447 lies near the end of the insertion that defines the MetAP-2 family, and the constriction formed by the Leu447-Leu328 pair provides a structural basis for the MetAP-2 family's requirement for a substrate with a small (<1.3 Å radius of gyration) side chain at P2 (15). The terminal carboxyl of the side chain makes a hydrogen bond with Asp376. Fumagillin can be analyzed as a substrate mimic with the epoxide-bearing side chain resembling methionine's side chain and the opened epoxide substituting for a nucleophilic water near the scissile carbonyl bond. The long unsaturated side chain protruding from the pocket mimics the COOH-terminal peptide chain.

The ability of fumagillin and related compounds to covalently inhibit MetAP-2 is even more remarkable in light of their specificity for MetAP-2 over MetAP-1 because the two enzymes have very similar active sites (Fig. 2). In yeast, either MetAP-1 or MetAP-2 function can be eliminated and the remaining enzyme will compensate (16). Elimination of both MetAP-1 and MetAP-2 function is lethal (16). Features that might generate such specificity can be highlighted by superimposing the 81 central core residues of HsMetAP-2 and EcMetAP-1 (0.83 Å root mean square deviation for main chain atoms) (Fig. 4C). In this superimposed model, EcHis79, the residue that would covalently bond fumagillin, is too far away to form a bond. Two compensatory modifications are possible: moving fumagillin by about 1.6 Å toward EcHis79or moving EcHis79 toward fumagillin. Moving fumagillin leads to severe steric clashes of C5 and the C6 methoxyl group with EcTyr168 and of C7 and C8 with EcCys78 (Fig. 4C). Tyrosine-168 is conserved in the MetAP-1 family, and its smaller counterpart, Leu328, is conserved in the MetAP-2 family (Fig. 2). In addition to the steric clashes, the fumagillin oxygen that interacts with cobalt 2 would be more than 5 Å away from either cobalt, a distance greatly in excess of a metal-water interaction.

Moving EcHis79 to bond with fumagillin also has troublesome features. The first is the assumption that EcHis79 can be moved because its position appears to be fixed by a series of hydrogen bonds involving both main and side chain atoms between β1 and β2. The positions of EcHis79 and HsHis231 are also influenced by the adjacent residues, which are strongly conserved in both families: Ala-Ala-His-Tyr/Phe in MetAP-2 and Val/Ile-Cys-His-Gly in MetAP-1 (Fig. 2). If EcHis79 could be moved to bond with fumagillin, there would be problems accommodating fumagillin's side chain in the specificity pocket. In MetAP-1, the size of the conserved Phe177(HsIle338) makes the pocket substantially narrower (Figs. 2and 4C). The relatively slim side chain of methionine can be accommodated in this narrower pocket whereas the bulky and conformationally restricted side chain of fumagillin cannot. Because MetAP-1 lacks Tyr444 and has an open specificity pocket, the narrower pocket would be easily accessible. Thus, fumagillin's inability to inhibit MetAP-1 can be traced to the position of the nucleophilic His in MetAP-1, the difficulty of repositioning this residue because of consistent size differences in the adjacent residues, and a narrowing of the specificity pocket.

Our results provide a structural framework for understanding the relation of human MetAP-2 to prokaryotic and other eukaryotic MetAPs, fumagillin's ability to inhibit MetAP-2, and the basis of fumagillin's specificity. Insights from this analysis will also be useful in structure-based drug design. Fumagillin-based therapeutics such as TNP-470 share fumagillin's conformationally rigid template and key features. The two drugs differ in the side chain at C6, a region that shows few, if any, ligand interactions. Thus, TNP-470 is likely to inhibit MetAP-2 by occupying the active site in the same fashion as fumagillin.

  • * To whom correspondence should be addressed. E-mail: jcc12{at}


View Abstract

Navigate This Article