Structure and Function of a Squalene Cyclase

See allHide authors and affiliations

Science  19 Sep 1997:
Vol. 277, Issue 5333, pp. 1811-1815
DOI: 10.1126/science.277.5333.1811


The crystal structure of squalene-hopene cyclase fromAlicyclobacillus acidocaldarius was determined at 2.9 angstrom resolution. The mechanism and sequence of this cyclase are closely related to those of 2,3-oxidosqualene cyclases that catalyze the cyclization step in cholesterol biosynthesis. The structure reveals a membrane protein with membrane-binding characteristics similar to those of prostaglandin-H2 synthase, the only other reported protein of this type. The active site of the enzyme is located in a large central cavity that is of suitable size to bind squalene in its required conformation and that is lined by aromatic residues. The structure supports a mechanism in which the acid starting the reaction by protonating a carbon-carbon double bond is an aspartate that is coupled to a histidine. Numerous surface α helices are connected by characteristic QW-motifs (Q is glutamine and W is tryptophan) that tighten the protein structure, possibly for absorbing the reaction energy without structural damage.

The cyclization reactions catalyzed by squalene cyclases (S-cyclases) and 2,3-oxidosqualene cyclases (OS-cyclases) are highly complex (1) and give rise to various fused-ring compounds (2, 3). Although these enzymes have been well studied earlier (2, 4), present-day recombinant techniques have contributed much to the understanding of their function and reactivity (1, 5). Early suggestions for the reaction mechanism favored a concerted process (2), whereas current hypotheses (6, 7) dissect a series of carbocationic intermediates (Fig.1). Recently, the OS-cyclases became targets for the development of antifungal and anticholesteremic drugs (8). The integral membrane protein character of these enzymes is generally accepted (9, 10). As shown below, the S- and OS-cyclases have related sequences and should therefore have similar spatial structures. We now present the crystal structure of a bacterial S-cyclase at 2.9 Å resolution.

Figure 1

The proposed reaction steps in squalene-hopene cyclases involving carbocationic intermediates. The general acid B1:H protonates (H) squalene at C3, whereas the general base B2 deprotonates at C29 of the hopenyl cation. In a side reaction, the cation is hydroxylated forming hopan-22-ol.

Recombinant squalene-hopene cyclase from Alicyclobacillus acidocaldarius has been previously crystallized in three distinct forms (11). For phase determination by multiple isomorphous replacement (MIR), we used the crystal form A′, which contains three polypeptide chains in the asymmetric unit that show only small displacements from the threefold symmetry of crystal form A (Tables 1and 2). A molecular model was derived from the A′ crystal MIR electron density. This model was then refined in crystal form A, which shows exact threefold symmetry with only one chain in the asymmetric unit (Table 2). In agreement with our gel permeation chromatography results and those of others (12), the crystal structure reveals the presence of a dimeric enzyme. Both subunits are related by a twofold axis and form a polar interface of 850 Å2. Each subunit consists of 631 amino acid residues and has a molecular mass of 71,569 daltons; residues 1 to 9 and 629 to 631 could not be located in electron density maps. The chain is organized in two domains (Fig.2), forming a dumbbell shape. Domain 1 is an α66 barrel of two concentric rings of parallel α helices (Fig. 3) that resembles the chain fold of glucanases and a farnesyl transferase (13). Domain 2 is inserted into domain 1 and contains an α-α barrel, which appears to be an evolved version of the α66 barrel. Both barrels point with the amino ends of their inner α helices toward the molecular center, which consists of long loops from both domains forming a small β structure and enclosing a large cavity of about 1200 Å3(14).

Figure 2

Stereoview of squalene-hopene cyclase chain fold with labeled NH2- and COOH-termini (N and C), inhibitor position (L), and channel entrance (E). Color code: internal (yellow) and external (red) barrel helices, β structure (green), QW-motifs (purple), and α8 in the nonpolar plateau (white).

Figure 3

Chain fold topology showing α helices (circles), one 310-helix (hexagon), β strands (squares), and QW-motifs (zigzag) with marked chain directions upward (thin) and downward (thick). Domain border positions are given. The top sketch shows how the outer ring of the α66barrel is stabilized by the QW-motifs, one of which deviates from the others.

Table 1

MIR phasing in crystal form A′ (P32 21, a =b = 140.9 Å, c = 243.8 Å), which is closely related to form A (P321, c axis one-third) with one asymmetric chain (11). For the “native” data (NAT) and all soaks we used single and double mutants containing D376C (Asp at position 376 mutated to Cys). The soaks were as follows: M1 and M2, methylmercury acetate; T1, T2, and T3, tetrakis(acetomercury)methane; Cl, chloromercury ferrocene; and KHg, K2Hg(CN)4. Phasing was initiated with the difference-Patterson of T1 and performed with MLPHARE (25), resulting in a figure of merit of 0.424 at 3.2 Å resolution. T3 was from double-mutant C50S-D376C and lacked the major site of T1 and T2. Because of the close relation with form A, all sites are triplicate.

View this table:
Table 2

Structure refinement in crystal form A (P321, a = b = 140.7 Å,c = 81.9 Å) with one asymmetric chain (11). After MIR phasing (Table 1), solvent-flattening, and density averaging over the three asymmetric chains of A′ (25), about 90% of the sequence could be placed (26). The model was then transferred to crystal form A, where genuine native data were available, and refined with X-PLOR (27) without any cutoff of the data. At the end, B factors were refined as grouped atoms (two groups per residue). The model had cysteines at all heavy atom sites. Residual density in the central cavity was modeled as LDAO (11). The model contains residues 10 to 628 and 38 water molecules, and 90% of the residues were in the most favored regions of the Ramachandran plot (28). Values in parentheses are for the last shell (2.95 to 2.85 Å). rmsd, root-mean-square deviation.

View this table:

The aligned sequences of S-cyclase and human OS-cyclase are presented in Fig. 4. The S-cyclase contains eight QW-sequence motifs (15), five of which are also present in the OS-cyclases. Seven of the eight motifs assume virtually identical polypeptide conformations. The side chains of Q and W are stacked, forming hydrogen bonds with the amino end of the adjacent outer barrel helix and with the carbonyl end of the preceding outer barrel helix, respectively (Fig. 3). The QW-motifs connect all outer helices of the α66 barrel and several outer helices of the other α-α barrel, thus stabilizing the whole protein (Figs. 2and 3).

Figure 4

Sequence alignment of squalene-hopene cyclase from A. acidocaldarius (top row) and human liver 2,3-oxidosqualene-cyclase (bottom row) from program CLUSTAL. Secondary structures (29) and QW-motifs are marked in the color code of Fig. 2. Asterisks indicate identical residues. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

The active site is located in the large central cavity (Fig.5A), as indicated by the bound competitive inhibitor (16)N,N-dimethyldodecylamine-N-oxide (LDAO) (inhibition constant K i = 0.14 μM). This location is corroborated by a putative mechanism-based inhibitor that labeled the equivalent of Asp376 at the top of the cavity (Fig. 5A) (17) and similar experiments were done that labeled segments containing residues lining the cavity (18). Moreover, mutation of Asp376 or Asp377inactivated S-cyclase (19), and mutation of the equivalent of Asp376 (18) inactivated OS-cyclase. Finally, the active site location at the amino ends of the inner helices of the α66 barrel corresponds to those of the four enzymes with similar barrel structures (13).

Figure 5

The color-coded surface representations (30) with nonpolar (yellow), positive (blue), and negative (red) areas. (A) View similar to Fig.2 but rotated around a vertical axis and sliced. The cutting plane (checked) opens the large internal cavity with the bound inhibitor LDAO. The nonpolar channel runs to the left, opening into a nonpolar plateau. The channel constriction (C) appears closed, but it is mobile enough to be readily opened. At the upper left, hopane (two views) is shown at scale. (B) View similar to Fig. 2 directly onto the 1600 Å2 nonpolar plateau with the channel entrance (E) at its center and two nonpolar side chains pointing to the outside. This is the only large nonpolar region on the surface.

The central cavity can be accessed through a nonpolar channel between the helices of domain 2 (Fig. 5A). This domain consists of an α-α barrel including a 310-helix (Fig. 3) that leaves space for the crossing channel. Helix α8 with the following loop, the segment between α6 and α7 as well as loop α15-α16 (Figs. 2 and 3), form a large and rather mobile nonpolar plateau on the protein surface that surrounds the channel entrance (Fig. 5B).

This plateau has a solvent-accessible surface of about 1600 Å2 (14) and is encircled by a ring of positively charged residues (Fig. 5B), which suggests that the plateau plunges into the nonpolar center of a membrane, whereas the ring forms salt bridges with the displaced phospholipid and sulfolipid head groups. Such a model characterizes a monotopic membrane protein (20) and explains how the substrate squalene diffuses from the membrane interior, where it is dissolved (21) into the central cavity.

The dimer interface contains, in large part, residues from domain 2 (Fig. 6A). It orients the nonpolar plateaus in parallel (Fig. 6B). Such nonpolar plateau twins also occur with the other structurally known monotopic membrane protein (Fig. 6C), prostaglandin-H2-synthase (22). This enzyme has a completely different chain fold but also takes up its substrate from the membrane. The resemblance of these arrangements indicates a functional role as, for instance, a doubling of the membrane interaction per particle.

Figure 6

Cα backbone plots of the two structurally known monotopic membrane proteins, both showing parallel nonpolar plateaus (thick lines) that facilitate fusion with the membrane. The channel entrances (E) for substrate uptake are labeled. (A) Homodimeric S-cyclase viewed along the molecular diad. (B) S-cyclase viewed from the top of (A). (C) Homodimeric prostaglandin-H2-synthase (22) oriented like the S-cyclase in (B).

In S-cyclase the channel between the nonpolar plateau and active site contains a constriction formed by Phe166, Val174, Phe434, and Cys435 (Fig.7), which appears to block it (Fig. 5A). Residues 434 and 435, however, are on a mobile loop that contacts a surface loop around position 210 with even higher mobility (23). We therefore suggest that the constriction behaves as a gate that permits substrate passage.

Figure 7

Stereo view of the active site cavity with the inhibitor LDAO in its (2F oF c) electron density at a level of 1.3σ. The view is similar to Fig. 5A. Among the aromatic residues shown, Trp312, Phe365, Trp489, Tyr495, Phe601, Tyr609, and Tyr612 are conserved in all known sequences of the family. The top end of the cavity is highly polar, and there is a polar net at the bottom.

The central 1200 Å3 active site cavity is mainly nonpolar, but it has a highly polar patch at the top (Figs. 5A and 7). It is lined by numerous aromatic residues that could stabilize the carbocationic intermediates of the cyclization reaction by their π-electrons (7). The residues lining the cavity are well conserved but show a gradient with highest conservation at the top and lowest at the bottom. This gradient indicates that the first reaction step common for S- and OS-cyclases occurs at the polar top of the cavity and that the variable features are at the bottom. Accordingly, the initially protonating acid B1:H (Fig. 1) should be at the top; base B2 of S-cyclase should accept a proton at the bottom end of the cavity; and B2 of the OS-cyclases should be near to the center of the substrate where proton uptake from the lanosteryl cation is expected (4). Our proposal appears to be consistent with the experiments of (18) where a histidine equivalent to Trp169 near to the cavity center (Fig. 7) was labeled in later reaction steps, but it appears to disagree with the suggestion of (17) that Asp376 is near a position equivalent to ring E of hopene.

The polar top of the cavity has the sequence motif DXDD (X, any amino acid) of the S-cyclases (residues 374 to 377), which is important for catalysis (19). The OS-cyclases contain only the second aspartate residue, and this is crucial for catalysis (18). A similar catalytic motif, DDXXD, is present in farnesyl diphosphate synthase, which runs also through a cationic intermediate but appears to use this motif for metal binding (24). Because the chain folds are different, there is no evolutionary connection.

In S-cyclases Asp374 and Asp377 of the DXDD motif are hydrogen-bonded (2.6 Å distance) and close to the hydrogen-bonded (2.7 Å distance) pair Asp374:His451 (Fig. 7). We propose that the pair Asp374:Asp377 carries a negative charge that stabilizes a positive charge on Asp376:His451 and thus Asp376 in its protonated form. For the reaction, squalene diffuses into the central cavity where it assumes the required conformation (Fig. 1) with its C3 atom near to the putative proton on Asp376. We propose that Asp376 is the acid B1:H which protonates the first C-C double bond, creating a carbocation that steps through the squalene chain and converts it to hopene (Fig. 1). The protonation leaves Asp376:His451 as a stable salt bridge. The pair Asp374:Asp377 might support the initial protonation step. It is absent in the OS-cyclases, presumably because epoxides are more readily protonated than C-C double bonds.

Base B2 of S-cyclase should be located at C29 of the hopenyl cation, which is near to the last carbon of LDAO (Fig. 7) where the protein offers no suitable residue. Because the reported S-cyclase shows a side reaction (10) resulting in about 10% diplopterol (hopan-22-ol), we suggest that B2 is a water molecule polarized by other waters that are in contact with the hydrogen-bonding network of Gln262:Glu45:Glu93:Arg127, which could store a proton (Fig. 7). Diplopterol is formed if the front water adds as hydroxyl to the last carbocation instead of accepting the proton.

Taken together, squalene enters the central cavity through the constriction, which acts as a gate, and is forced into the unfavorable conformation necessary for catalysis. The scale of the required actions corresponds to the low turnover number of 0.3 s−1(12, 16). Hopene formation releases ∼200 kJ/mol at the bottom of the α66 barrel, exceeding by far the usual protein stabilization energy of ∼50 kJ/mol. The barrel does not disintegrate because its α helices at the surface are connected and thus stabilized by the QW-motifs, characteristic for this enzyme family. The resulting excitation then facilitates the return of hopene to the membrane.

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


View Abstract

Navigate This Article