Skip to main content
Proceedings of the National Academy of Sciences of the United States of America logoLink to Proceedings of the National Academy of Sciences of the United States of America
. 2001 Dec 18;98(26):14808–14813. doi: 10.1073/pnas.011399198

Crystal structure of the macrocycle-forming thioesterase domain of the erythromycin polyketide synthase: Versatility from a unique substrate channel

Shiou-Chuan Tsai *,†, Larry J W Miercke *, Jolanta Krucinski *, Rajesh Gokhale , Julian C-H Chen *, Paul G Foster *, David E Cane , Chaitan Khosla †,§,¶,, Robert M Stroud *,
PMCID: PMC64940  PMID: 11752428

Abstract

As the first structural elucidation of a modular polyketide synthase (PKS) domain, the crystal structure of the macrocycle-forming thioesterase (TE) domain from the 6-deoxyerythronolide B synthase (DEBS) was solved by a combination of multiple isomorphous replacement and multiwavelength anomalous dispersion and refined to an R factor of 24.1% to 2.8-Å resolution. Its overall tertiary architecture belongs to the α/β-hydrolase family, with two unusual features unprecedented in this family: a hydrophobic leucine-rich dimer interface and a substrate channel that passes through the entire protein. The active site triad, comprised of Asp-169, His-259, and Ser-142, is located in the middle of the substrate channel, suggesting the passage of the substrate through the protein. Modeling indicates that the active site can accommodate and orient the 6-deoxyerythronolide B precursor uniquely, while at the same time shielding the active site from external water and catalyzing cyclization by macrolactone formation. The geometry and organization of functional groups explain the observed substrate specificity of this TE and offer strategies for engineering macrocycle biosynthesis. Docking of a homology model of the upstream acyl carrier protein (ACP6) against the TE suggests that the 2-fold axis of the TE dimer may also be the axis of symmetry that determines the arrangement of domains in the entire DEBS. Sequence conservation suggests that all TEs from modular polyketide synthases have a similar fold, dimer 2-fold axis, and substrate channel geometry.


Modular polyketide synthases (PKSs) are a family of multienzyme complexes that synthesize the polyketide cores of biologically active compounds, including natural products that have become important pharmaceuticals, such as erythromycin, rifamycin, FK506, rapamycin, and avermectin (1). Their remarkable combination of substrate tolerance and selectivity is largely because of their modular architecture, in which different catalytic domains are combined into “modules” (Fig. 1), such that each module contains several enzymes and adds one additional building block to a growing polyketide chain. 6-Deoxyerythronolide B synthase (DEBS) is a modular PKS that catalyzes the biosynthesis of 6-deoxyerythronolide B (6-dEB, 1, Fig. 1), the macrocyclic core of the antibiotic erythromycin (2, 3). The entire DEBS is a homodimer containing two copies each of 28 catalytic domains organized into a loading didomain, six extension modules (each composed of several domains), and a terminal thioesterase (TE) that cyclizes and releases the final product, 6-dEB (1, Fig. 1). Combinatorial substitution of enzymes of DEBS, by others from rapamycin PKS, gave rise to over 100 novel compounds at varying yields (4). To date, no structure of a modular PKS component has been reported. When the structures of individual domains are determined, there will be opportunity to apply structure-based principles of protein engineering in next-generation efforts to rationally design novel polyketide products.

Figure 1.

Figure 1

The modular architecture of DEBS, its intermediates, and its final product as the 14-membered ring of 6-dEB. AT, acyltransferase; ACP, acyl carrier protein; KS, ketosynthase; KR, ketoreductase; DH, dehydratase; ER, enoylreductase; TE, thioesterase.

The TE domain of DEBS catalyzes the cyclization and release of a highly functionalized heptaketide chain via lactonization (Fig. 1). This domain is covalently linked to the rest of DEBS via the acyl carrier protein on the sixth module (ACP6), which transfers the final linear form of the polyketide substrate to the TE via a phosphopantethienyl arm attached to ACP6. DEBS TE is a homodimer with a molecular mass of 66 kDa. At least 12 other modular PKSs are known to harbor homologous terminal TE domains, with 29–75% sequence identity compared with DEBS TE. Chain release by modular PKS TEs involves regioselective cleavage of an acyl-enzyme intermediate by a specific hydroxyl group on the polyketide substrate, leading to the formation of a macrocyclic product (Fig. 1). Although the natural product of DEBS TE is a 14-membered ring, this enzyme has been shown to support the formation of alternatively functionalized 6-, 8-, 12-, 14-, and 16-membered ring systems (5, 6). These observations underscore the tolerance of DEBS TE for the generation of a spectrum of cyclic products. Given the importance of this reaction in synthetic chemistry (and natural product chemistry in particular), understanding the mechanism and substrate specificity of DEBS TE presents new opportunities for chemoenzymatic synthesis of complex molecules by modification of the enzyme and substrates.

To date, four crystal structures of hydrolytic TEs (involved in fatty acid synthesis and metabolism) have been solved (710). These enzymes have only 2–8% sequence identity with macrocycle-forming TEs such as DEBS TE. The crystal structure of the DEBS TE reported here reveals an unusual hydrophobic leucine-rich dimer interface and an open substrate channel that passes through the entire enzyme. The structure provides a basis for understanding the substrate specificity of this enzyme (11, 12), as well as the pathway by which polyketide chains are channeled into its active site. It also defines specific amino acids that could be mutated in attempts to generate novel polyketides. Lastly, docking studies between the TE dimer and the preceding domain, ACP6, suggest that the 2-fold axis of the TE dimer and the entire DEBS homodimeric assembly coincides. The implications of this architecture for interdomain chain transfer are discussed.

Methods

Expression, Purification, and Crystallization.

Recombinant DEBS TE was expressed in Escherichia coli and purified as described (11). Selenomethionine-substituted DEBS TE was generated by using the method of Van Duyne (13) and purified as above. TE eluted from a Ni-agarose column was further purified by using a Poros HQ column (PerSeptive Biosystems, Framingham, MA) and eluted at 0.8 M NaCl.

Crystals of DEBS TE were grown in hanging drops at 4°C by vapor diffusion. The well buffer was 30% polyethylene glycol 400/100 mM Hepes, pH 7.5/2 mM DTT/100 mM MgCl2. Protein concentration was 10 mg/ml in 20 mM Hepes, pH 7.5/2 mM DTT. Drops were generated by mixing 4 μl of the purified protein solution with 4 μl of well buffer. Trigonal TE crystals, 300 × 300 × 200 μm, grew in several weeks.

Data Collection.

Crystals were flash cooled to −170°C. Native data to 2.8-Å resolution were collected on an ADSC Quantum-4 charge-coupled device detector at beamline 5.0.2 of the Advanced Light Source (Berkeley, CA). The space group was P3121 with unit cell dimensions a = 130.5 Å, c = 208.5 Å. The density as measured by floatation in a Ficoll gradient (14) shows 76% solvent content, corresponding to three TE monomers per asymmetric unit. Intensities were integrated and scaled by using the hkl package (15) with no σ cutoff. The data from Sm3+- (4.0 Å) and Ni2+- (4.3 Å) soaked crystals were collected on a MAR-Research (Hamburg, Germany) image plate at beamline 9–1 of the Stanford Synchrotron Radiation Laboratory (SSRL). Multiwavelength anomalous dispersion (MAD) data of Sm3+-soaked (4.0 Å) and Se-derivatized (4.0 Å) crystals were collected on an ADSC Quantum-4 charge-coupled device detector at beamline 9–2 of SSRL. Statistics are shown in Table 1.

Table 1.

Structure determination and refinement

Data collection statistics
Crystal Wavelength, Å Resolution, Å Observations Unique Completeness,* % Rsym,* % II*
Native 1.0000 2.8 190,617 50,859 99.1 (93.1) 3.1 (61.7) 22.4 (1.6)
SmCl3 1.0800 3.8 389,060 20,793 99.0 (100) 8.6 (30.8) 20.8 (6.9)
NiCl2 0.9800 4.3 206,467 14,966 98.6 (99.7) 8.9 (38.7) 11.3 (3.4)
SeMet w1 0.9794 4.0 582,831 17,970 100 17.2 (64.2) 12.0 (3.2)
SeMet w2 0.9792 4.0 582,987 17,970 100 17.4 (60.1) 11.0 (2.7)
SeMet w3 0.9566 4.0 582,596 17,970 100 16.2 (59.5) 11.7 (2.7)
SmCl3 w1 1.8465 4.0 518,279 18,738 100 18.5 (39.5) 11.8 (4.5)
SmCl3 w2 1.8451 4.0 517,367 18,689 100 19.1 (41.7) 10.1 (4.4)
SmCl3 w3 1.2398 4.0 518,525 18,914 100 15.1 (27.6) 12.3 (5.2)
Phasing statistics
Derivative Resolution cutoff, Å Number of sites Phasing power (iso) Phasing power (ano) RCullis Figure of merit

SmCl3 4.3 7 1.5 1.6 0.7 0.4
NiCl2 4.3 5 0.8 0.8 0.9 0.2
SeMet MAD 4.0 15 1.08, 0.96, 0.22 1.03, 0.93, 0.23 0.77, 0.79, 0.94 0.5
SmCl3MAD 4.0 9 1.66, 1.66, 0.77 1.57, 1.59, 0.74 0.62, 0.62, 0.88 0.5
Combined figure of merit (MIR/MAD) = 0.8 (DM) = 0.9
Refinement statistics
No. of reflections Protein atoms Water atoms Resolution, Å Rcryst, % Rfree, % Average B factor,2

By cns 50,859 (2,542) 5,988 439 30–2.8 24.1 25.1 76.9, 160.3, 68.9
By refmac5 (TLS) 48,255 (2,573) 5,988 418 30–2.8 23.9 27.6 52.9, 42.1, 52.5
rms bond lengths, Å = ±0.02 rms bond angles, ° = ±1.95 Luzzati coordinate error, Å = 0.43
Ramachandran Core 84.3% Allowed 14.3% Generous 1.4%
*

Statistics for the highest-resolution bin are in parentheses. 

Values are presented without averaging the Friedel pairs. 

For monomers A, B, and C. 

MIR and MAD Phasing, Model Building, and Refinement.

Phases were calculated by a combination of multiple isomorphic replacement (MIR) and MAD. Two derivatives were identified from 250 heavy atom soaks. Ni2+ (a = 131.0, c = 209.0 Å) was found to bind to the TE at its C-terminal His6 tag; Sm3+ (a = 131.4, c = 209.5 Å) was also found to bind to the TE and was used to collect both isomorphous and MAD data. The MIR derivative data sets were scaled to the native and were isomorphous with the native data; the MAD derivative data sets were scaled to the far remote wavelength, and heavy atom parameters were refined with the program cns (16). Seven sites in the Sm3+ derivative were identified in difference Patterson maps and used to phase all other derivatives, revealing five Ni sites, six Sm sites in the MAD data set of Sm3+ derivative, and 21 Se sites in the MAD data set of Se derivative. The 21 Se sites correspond to seven sites per monomer that account for all methionines except the first methionine (Met-1), which is disordered. The phases from MIR and MAD were combined by using cns, and the initial 4.3-Å MIR–MAD map was used as a starting point for “solvent flipping” in cns, followed by phase extension (0.0025 Å per cycle of density modification) to 2.8 Å (Table 1).

The initial map showed a single continuous density for the main chain with strong side-chain features that allowed unambiguous sequence assignment for 90% (residues 15–281) of the 294 residues in DEBS TE. The model was built by using quanta, and the initial model gave an R factor of 42.2% over the resolution range 20–2.8 Å. Refinement in cns by a combination of simulated annealing, energy minimization, and B factor refinement reduced the Rcryst and Rfree values to 27.3 and 29.4%, respectively, without including any solvent. Water molecules were added and edited based on Fo − Fc maps. Subsequent refinement resulted in Rcryst and Rfree values of 24.1 and 25.1%, respectively.

Two of the three monomers in the asymmetric unit (monomers A and C) are almost identical, with a rms deviation for backbone atoms of <0.5 Å. Monomers A and C each interact with three other surrounding monomers. The density for monomer B is much less well defined, indicating its relative disorder, reflected in the much higher B factors than for A and C (Table 1). For this reason, noncrystallographic symmetry averaging was not used. We tested whether this is because monomer B interacts only with monomer A, which leads to a higher rigid-body displacement of monomer B in the lattice. The displacement parameters of the three monomers were parameterized by allowing separate relative rigid-body displacements of each monomer separately, as being a component of the individual B factors within each monomer. This was carried out by using REFMAC 5 in the CCP4 suite with additional refinements on the T, L, and S displacement tensors (17). The T, L, and S displacements for monomer B are 10–20 times larger than for A and C, and the rigid body approximation supported an R factor of 23.9% (Rfree = 27.6%) to 2.8 Å. Correction of the TLS-refined model for the underlying rigid body displacement resulted in an ordered model of monomer B that has comparable B factors to monomers A and C (Table 1). Therefore, the high B factors of monomer B are because of high displacement instead of high thermal disorder. Water molecules were then reinspected and edited after TLS refinement.

Results and Discussion

Overall Structure.

The overall structure of the DEBS TE offers the first structural observation of a modular PKS domain. Each TE monomer has the features of the α/β hydrolase fold (18, 19); however, this TE has an unusual substrate channel that passes through the entire protein, with the active site (Asp-169–His-259–Ser-142) located in the middle of the substrate channel (Fig. 2). DEBS TE consists of a central seven-stranded β-sheet with the second strand (β2) antiparallel to the remaining strands (Fig. 2A). The β7 strand is shorter than β8, and the C terminus of β7 is twisted 90° to form part of the substrate channel (Fig. 2A). The β strands are flanked on either side by α helices, two on one side and four on the other. Two additional α helices at the N-terminal end of the TE make up the dimer interface (Fig. 2A).

Figure 2.

Figure 2

(A) Overall structures of DEBS TE dimer. (B) The hydrophobic leucine-rich dimer interface of DEBS TE, monomer A in red (residues labeled in black), monomer B in blue (residues labeled in purple). (C) Molecular surface of DEBS TE dimer, monomer A in green, monomer B in yellow, the electropositive groove (proposed ACP6-binding site) in blue. (D) The most favorable geometry between a homology model of ACP6 and DEBS TE, predicted by FTDOC. A global view shows that the arginine groove is the primary docking site for ACP6. A monomer of ACP6 (red) was docked against the TE dimer (green). It forms a dimer with the ACP6 dimer partner (brown), generated by rotation of the docked ACP6 around the TE dimer 2-fold axis. This suggests that the 2-fold axis of TE dimer may be the 2-fold axis of the entire modular PKS. The proposed overall quaternary structure of DEBS based on the coincidence of 2-fold axes for the TE and modeled ACP6 domains are shown as a miniature blue and red dimer (Left). The direction of the DEBS polypeptide chains from N- to C-terminal as indicated suggests that adjacent reactions may be specified to take place from either one of the polypeptide chains. (E) The docked ACP–TE interface shows Asp-45, Ser-46, Glu-71, His-72, and Pro-73 (purple) of ACP6 interacting with the Arg-19, Arg-23, Arg-193, and Arg-197 (blue) of TE. Hydrogen bonds are colored in green. The figure was generated by using GRASP (27) and SWISS PDB VIEWER rendered with POV-RAY (21).

Significantly, residues in each secondary structural feature are highly conserved (especially in regions surrounding the catalytic triad) among TEs that catalyze the release of polyketide chains from modular PKSs. Many of these conserved residues (Fig. 3, purple F) are presumably important in maintaining the TE fold as a result of their location at conserved turns or through hydrogen bond networks and packing between different regions of the protein. Hydrophobic residues that form the active site and those that line the channel are also conserved among PKS TEs (Fig. 3, purple S). Together, these observations suggest that the fold, the channel, and the catalytic triad are conserved among TEs from different modular PKSs.

Figure 3.

Figure 3

Sequence alignment among PKS TEs. Abbreviations indicate the final macrolide product for the corresponding modular PKSs. DEBS, 6-deoxyerythronolide B synthase; OLEANDO, oleandolide; METHY, 10-deoxymethynolide; TYLACTON, tylactone aglycone; AVERMEC, avermycin aglycone. Symbols: @, helix and #, β sheets above the alignment indicate secondary structures. Purple symbols under the alignment indicate homologous residues of the following nature: R, proposed arginine-rich ACP-binding site; D, hydrophobic leucine-rich dimer interface; S, substrate channel lining; F, fold maintenance. Sequence alignment was prepared by MULTALIN (28).

The Hydrophobic Leucine-Rich Dimer Interface and the 2-Fold Axis.

Modular PKSs, including the TE domains, function as homodimers; therefore, the 2-fold axis of the DEBS TE dimer may provide insights into the 2-fold axis of symmetry and the quaternary structural design of the entire DEBS. DEBS TE was confirmed to be a monodispersed dimer in solution, as measured by dynamic light scattering (DLS) and size exclusion chromatography. One dimer (C− and its partner C′) lies on the crystallographic 2-fold axis. The noncrystallographic 2-fold dimer A−B has the same dimer interface, suggesting that the interface is retained in intact DEBS. The dimer interface, with a buried surface area of 1,726 Å2 (Fig. 2B), consisted of the two N-terminal α-helices α1 and α2 (Fig. 2 AC). A total of eight conserved leucines (Leu-18, -37, -38, and -41) and two phenylalanines (Phe-44) interact at the dimer interface (Fig. 2B), which is highly unusual in α,β-hydrolases, whose dimer interfaces are generally of an electrostatic nature (18). Because the leucines that form the hydrophobic leucine-rich dimer interface (Fig. 3, purple D) are conserved among different TE domains from modular PKSs, these TEs may also have a similar hydrophobic dimer interface and 2-fold axis. Preliminary structural data indicate that this is indeed the case for the crystal structure of the terminal TE domain from the picromycin synthase (S.-C.T., R. Lu, D. E. Cane, C.K., and R.M.S., unpublished results).

ACP6 Docking and the 2-Fold Axis.

To investigate the possible association with the upstream domain that transfers the substrate to the TE, the crystal structure of DEBS TE dimer was docked against a homology model of the ACP6 monomer. ACP6 is covalently linked to the TE in DEBS. A homology model of ACP6 was generated by appropriate sequence substitutions into an averaged NMR structure of the actinorhodin ACP (2AF8) (20), which shares 48% sequence similarity with ACP6. [The model was obtained via the Swiss Model server (21)]. Previous structural studies of several ACPs have revealed characteristic electronegative patches on their surfaces, which are thought to facilitate protein–protein interaction between ACPs and other PKS domains or subunits, including the TE (20, 22). Computer-simulated docking between the ACP6 model and the DEBS TE structure was carried out by using the program ftdock (23), revealing that the most probable docking site for ACP6 on the TE is an electropositive groove (Fig. 2 C and D), which includes five highly conserved arginine residues (Fig. 3, purple R). The most favorable configuration predicted by ftdock (Fig. 2 D and E), based on binding energy and surface shape complementation, is plausible in light of three observations. First, in this configuration, extensive protein–protein interactions covering a surface area of 1,147 Å2 are predicted at the ACP6–TE interface, in which negatively charged residues of ACP6 interact extensively with the arginine groove of TE (Fig. 2E). Second, in the crystal structure of the TE, the arginine groove is located close to the N terminus (≈5 Å to residue 15, Fig. 2 C and D). Because the C-terminal end of ACP6 is connected to the N-terminal end of the TE without an intervening linker region, the docking sites in the TE and ACP6 should bring these two close to each other. Third, the 2-fold axis of the TE dimer is used to generate the dimeric partner of the docked ACP6 monomer (Fig. 2D), and the two ACP6 monomers themselves reveal a complementary interface. This result is consistent with the DLS measurements of an isolated ACP domain from DEBS, showing that it exists as a dimer. Thus, the observed 2-fold axis of the TE dimer may be the 2-fold axis of the entire DEBS dimer with adjacent catalytic domains arranged sequentially along the axis (Fig. 2D Left).

A Unique Substrate Channel.

The substrate channel of DEBS TE is unique not only in its openness on both sides but also in its shape, which must contribute to the observed substrate specificity of DEBS TE. None of the four fatty acid synthase TE structures previously solved has a substrate channel. It is rare to observe such an open substrate channel for an enzyme belonging to the α/β hydrolase family (19). This unique substrate channel is lined by highly conserved hydrophobic residues (Fig. 3, purple S) and nine nonconserved hydrophilic residues.

The active side triad, His-259–Asp-169–Ser-142, is located in the middle (7.5 Å from the N-terminal surface) of the substrate channel. The site of the covalent attachment of the acyl chain to the active site during catalysis is at catalytic residue Ser-142, located in the nucleophilic elbow between β6 and α5, a feature that is characteristic of α/β hydrolases (18). Asp-169, His-259, and Ser-142 form the hydrogen-bonded catalytic triad. This catalytic triad agrees with the general mechanism for TEs of this family, in which His-259, stabilized by Asp-169, acts as the catalytic base that accepts the proton from Ser-142, as it attacks the ACP6-bound thioester substrate, to yield an acyl-O-serine intermediate. The NH of Ala-143 may serve as the oxyanion stabilizing residue. The enzyme-bound intermediate is subsequently attacked either by water (hydrolysis) or in this case by a hydroxyl group on the polyketide chain (cyclization) (Fig. 1).

The substrate channel runs through the protein over a distance of ≈20 Å, with a mouth diameter of 8 Å on the side containing the N terminus (referred to as the N-side) and 5.5 Å on the side containing the C terminus (referred as the C-side). From the exterior (Fig. 2C), the substrate channel appears to be a long narrow channel that can accommodate only a linear polyketide chain. However, the interior of the substrate channel reveals a flat wide cavity around the catalytic triad, with a maximum diameter of 15 Å (Fig. 4A). Manual docking of the product 6-dEB (Fig. 1) into this cavity, followed by energy minimization, demonstrates that 14-membered macrolides can fit into the cavity without altering the protein conformation. When the C-1 carbonyl carbon of 6-dEB is constrained to be covalently bonded to the hydroxyl group of Ser-142, as must occur in the acyl-O-serine intermediate, the macrocycle can be accommodated in only one stereochemical orientation within the active site (Fig. 4). The validity of this model for TE–macrocycle association is reinforced by two observations. First, in this orientation, the C-13 ethyl side chain of 6-dEB points toward the channel opening (Fig. 4) and should therefore tolerate other substituents, and DEBS TE can generate erythronolide derivatives with an extremely broad range of C-13 substituents (24). Second, the cavity is wide enough to accommodate a slightly larger macrocycle, a feature that is consistent with the proven ability of DEBS to catalyze the biosynthesis of 16-membered macrocycles (6).

Figure 4.

Figure 4

(A) The molecular surface of DEBS TE sliced at the open substrate channel (entrance on the left, exit on the right), indicating the link between the unique substrate specificity and the unique shape of the channel. The product 6-dEB was docked in the channel. Residues forming hydrogen bonds with the product are labeled on the surface. (B) Manual docking followed by CNS energy minimization between 6-dEB (purple) and the DEBS TE. Hydrogen bonds between the active site residues and 6-dEB are indicated in green. Figures were prepared by INSIGHTII and SWISS PDB VIEWER rendered with POV-RAY (21).

By probing the substrate channel with the docking simulation, several predictions can be made. The model suggests that the backbone NH of Ala-143 lines the oxyanion hole and acts as a hydrogen bond donor to the C-1 carbonyl in the presence of an attacking nucleophile. Second, the bridging oxygen in the lactone (O-13) is within hydrogen-bonding distance of His-259, suggesting that His-259 promotes deprotonation of the corresponding hydroxyl group in the acyl-enzyme intermediate. Several key residues in the substrate channel are within hydrogen-bonding distance of nonreactive functional groups on the polyketide backbone and are likely to influence its substrate specificity. In particular, C-3—OH hydrogen bonds to Asn-180 and the backbone carbonyl of Tyr-171, C-5—OH hydrogen bonds to Asn-180 and Glu-184, and both C-9⩵O and C-11—OH hydrogen bond to the side-chain of Thr-76 and the backbone NH of Ala-77 (Fig. 4B). These extensive hydrogen-bonding interactions steer the polyketide chain such that only C-13—OH is close to C-1⩵O for lactonization to take place.

Consistent with our prediction, earlier studies have shown that simple analogs of the natural substrate of the DEBS TE that lack these functional groups can be hydrolyzed but not cyclized by the TE (11, 25), presumably because of the lack of hydrogen bonds to steer the substrate in a conformation and orientation favorable for cyclization. Further, DEBS TE has a marked preference for the (2R, 3S) diastereomer of a 2-methyl-3-hydroxy-acyl chain substrate over its (2S, 3R) diastereomer (11). This is consistent with the hydrogen-bonding pattern at C-3—OH (Fig. 4B), because the C3—OH—Asn-180 hydrogen bond can form only between TE and the (2R, 3S) diastereomer. Lastly, the higher hydrolysis turnover for longer polyketide substrates versus shorter ones (11) can also be rationalized as a result of the hydrophobic interactions that could be expected to occur between a long alkyl chain and the active site.

Two key residues in the hydrogen-bonding scheme (Fig. 4B), Asn-180 and Glu-184, are located on the α6 helix, which has the highest B factors. Not only does this helix influence the shape of the substrate channel, but it is also the least conserved region among TEs from different modular PKSs (Fig. 3, residues 177–190). It is possible that this α helix may be flexible and adapt to different polyketide substrates. The analysis of the substrate channel therefore offers strategies to systematically alter the substrate specificities of TEs by mutation of residues involved in defining the substrate channel.

The Mechanism and Its Implications.

On the basis of the crystal structure of the DEBS TE, associated modeling studies, and available experimental data, the following mechanism is proposed for the TE-catalyzed release of 6-deoxyerythronolid B from DEBS (Fig. 5). ACP6 binds to TE on the N-side via the assistance of a covalent linkage and interactions at the arginine groove of the TE. The polyketide chain is inserted into the long substrate channel of the TE, where it is transacylated from the phosphopantetheinyl arm of ACP6 to the nucleophilic Ser-142 in the active site of the TE. The accurate orientation of the polyketide chain in the active site residues and several functional groups (C-1⩵O, C-3—OH, C-5—OH, C-9⩵O, C-11—OH, C13—OH) on the polyketide chain (Fig. 5A). This results in positioning the C-13—OH next to the acyl-O-serine linkage. The active site is shielded from external water molecules as the competing nucleophile, on the N-side (diameter 8.0 Å) by the entire modular DEBS and on the C-side (diameter 5.5 Å) by the anchored polyketide chain (diameter 7.0 Å, Fig. 5B). Deprotonation of C-13—OH is facilitated by the basic His-259, leading to the displacement of the enzyme to yield the final product, 6-dEB (Fig. 5C). The hydrogen-bonding scheme in the active site also ensures that a unique 14-membered ring product is formed between C-1⩵O and C-13—OH instead of other hydroxyl groups on the substrate. The product may dissociate from the substrate channel via the C-side, in which case protein conformational change to widen the C-side may occur, possibly via the movement of helix-α6, which has the highest B factors of TE and is held by only two loops on either side of the helix (Fig. 2A). Subsequent enzyme turnover can then continue by the binding of substrate from the N-side.

Figure 5.

Figure 5

The proposed mechanism of the chain termination reaction catalyzed by DEBS TE. (A) ACP6, which binds to the TE on the N-side via the arginine groove, and a covalent linkage at the N terminus, presents the full-length polyketide as a phosphopentathienyl-bound acyl chain. (B) The functional groups C-1⩵O, C-3—OH, C-5—OH, C-9⩵O, C-11—OH, and C13—OH on the polyketide chain anchor the polyketide chain in the active site by forming seven hydrogen bonds with the protein. Upon nucleophilic attack of Ser-142, a covalently bound intermediate is formed. (c) Subsequent attack on C-1⩵O by C13—OH, which is anchored near C-1⩵O by the hydrogen-bonding scheme and whose nucleophilicity is enhanced by His-259, results in the formation of the macrolactone. Product can then be released out of the C-side exit.

The mechanism, based on the crystal structure of DEBS TE, can be used for three purposes. First, it allows the design of libraries of acyclic substrates that, on presentation to the DEBS TE, should undergo regioselective cyclization to generate conformationally constrained products. Efficient synthesis of such macrocycles is an important objective in synthetic chemistry, and the resulting small molecule libraries could provide clinically useful leads. Second, it provides a rational basis for the choice of DEBS TE (or any of its homologs) as a terminal TE in the design of engineered modular PKSs. Given the enormous potential of PKS modules to process natural and unnatural substrates (26), the installation of terminal TE domains that are kinetically competent for chain release (especially if it involves concomitant cyclization) is important for future efforts in combinatorial biosynthesis by using modular PKSs. Finally, it is possible that the polyketide chain may be processed by passing from ACP6 to the N-side for cyclization and may exit through the C-side of DEBS TE. This hypothesis, coupled with the common 2-fold axis of the TE and the modeled ACP6 dimer, suggests a quaternary structure of DEBS (Fig. 2D), in which the covalently linked domains are arranged predominantly as homodimers around a common 2-fold axis that may run through the entire DEBS homodimer. Such an arrangement would allow successive catalytic events to be carried out by domains that may lie on either of the 2-fold related chains.

Acknowledgments

We thank Janet Finer-Moore, Demetri Moustakas, Christopher Reyes, and Kinkead Reiling for advice and assistance during the course of this work. This work was supported by grants from the National Institutes of Health to R.M.S. (CA 63081), C.K. (CA 66736), and D.E.C. (GM22172).

Abbreviations

TE

thioesterase

DEBS

6-deoxyerythronolide B synthase

PKS

polyketide synthase

MAD

multiwavelength anomalous dispersion

MIR

multiple isomorphous replacement

6-dEB

6-deoxyerythronolide B

Footnotes

This paper was submitted directly (Track II) to the PNAS office.

Data deposition: The atomic coordinates reported in this paper have been deposited in the Protein Data Bank, www.rcsb.org (PDB ID code 1KEZ). They can also be obtained directly from C.K. or R.M.S.

References

  • 1.Cane D E, Walsh C T, Khosla C. Science. 1998;282:63–68. doi: 10.1126/science.282.5386.63. [DOI] [PubMed] [Google Scholar]
  • 2.Donadio S, Staver M J, McAlpine J B, Swanson S J, Katz L. Science. 1991;252:675–679. doi: 10.1126/science.2024119. [DOI] [PubMed] [Google Scholar]
  • 3.Cortes J, Wiesmann K E H, Roberts G A, Brown M J, Staunton J, Leadlay P F. Science. 1995;268:1487–1489. doi: 10.1126/science.7770773. [DOI] [PubMed] [Google Scholar]
  • 4.McDaniel R, Thamchaipenet A, Gustafsson C, Fu H, Betlach M, Ashley G. Proc Natl Acad Sci USA. 1999;96:1846–1851. doi: 10.1073/pnas.96.5.1846. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Kao C M, Luo G, Katz L, Cane D E, Khosla C. J Am Chem Soc. 1995;117:9105–9106. [Google Scholar]
  • 6.Jacobsen J R, Hutchinson C R, Cane D E, Khosla C. Science. 1997;277:367–369. doi: 10.1126/science.277.5324.367. [DOI] [PubMed] [Google Scholar]
  • 7.Bellizzi J J, III, Widom J, Kemp C, Lu J Y, Das A K, Hofmann S L, Clardy J. Proc Natl Acad Sci USA. 2000;97:4573–4578. doi: 10.1073/pnas.080508097. . (First Published April 18, 2000; 10.1073/pnas.080508097) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Devedjiev Y, Dauter Z, Kuznetsov S R, Jones T L, Derewenda Z S. Structure Folding Des. 2000;8:1137–1146. doi: 10.1016/s0969-2126(00)00529-3. [DOI] [PubMed] [Google Scholar]
  • 9.Lawson D M, Derewenda U, Serre L, Ferri S, Szittner R, Wei Y, Meighen E A, Derewenda Z S. Biochemistry. 1994;33:9382–9388. doi: 10.1021/bi00198a003. [DOI] [PubMed] [Google Scholar]
  • 10.Li J, Derewenda U, Dauter Z, Smith S, Derewenda Z S. Nat Struct Biol. 2000;7:555–559. doi: 10.1038/76776. [DOI] [PubMed] [Google Scholar]
  • 11.Gokhale R S, Hunziker D, Cane D E, Khosla C. Chem Biol. 1999;6:117–125. doi: 10.1016/S1074-5521(99)80008-8. [DOI] [PubMed] [Google Scholar]
  • 12.Rowe C J, Bohm I U, Thomas I P, Wilkinson B, Rudd B A, Foster G, Blackaby A P, Sidebottom P J, Roddis Y, Buss A D, et al. Chem Biol. 2001;8:475–485. doi: 10.1016/s1074-5521(01)00024-2. [DOI] [PubMed] [Google Scholar]
  • 13.Van Duyne G D, Standaert R F, Karplus P A, Schreiber S L, Clardy I. J Mol Biol. 1993;229:105–124. doi: 10.1006/jmbi.1993.1012. [DOI] [PubMed] [Google Scholar]
  • 14.Matthews B W. Methods Enzymol. 1985;114:176–187. doi: 10.1016/0076-6879(85)14018-8. [DOI] [PubMed] [Google Scholar]
  • 15.Otwinoswski Z, Minor W. Methods Enzymol. 1997;276:307–326. doi: 10.1016/S0076-6879(97)76066-X. [DOI] [PubMed] [Google Scholar]
  • 16.Brunger A T, Adams P D, Clore G M, Delano W L, Gros P, Grosse-Kunstleve R W, Jiang J-S, Kuszewski J, Nilges N, Pannu N S, et al. Acta Crystallogr D. 1998;54:905–921. doi: 10.1107/s0907444998003254. [DOI] [PubMed] [Google Scholar]
  • 17.Winn M, Isupov M, Murshudov G N. Acta Crystallogr D. 2001;57:122–133. doi: 10.1107/s0907444900014736. [DOI] [PubMed] [Google Scholar]
  • 18.Nardini M, Dijkstra B W. Curr Opin Struct Biol. 1999;9:732–737. doi: 10.1016/s0959-440x(99)00037-8. [DOI] [PubMed] [Google Scholar]
  • 19.Schrag J D, Cygler M. Methods Enzymol. 1997;284:85–107. doi: 10.1016/s0076-6879(97)84006-2. [DOI] [PubMed] [Google Scholar]
  • 20.Crump M P, Crosby J, Dempsey C E, Parkinson J A, Murray M, Hopwood D A, Simpson T J. Biochemistry. 1997;36:6000–6008. doi: 10.1021/bi970006+. [DOI] [PubMed] [Google Scholar]
  • 21.Guex N, Peitsch M C. Electrophoresis. 1997;18:2714–2723. doi: 10.1002/elps.1150181505. [DOI] [PubMed] [Google Scholar]
  • 22.Holak T A, Kearsley S K, Kim Y, Prestegard J H. Biochemistry. 1988;27:6135–6142. doi: 10.1021/bi00416a046. [DOI] [PubMed] [Google Scholar]
  • 23.Jackson R M, Gabb H A, Sternberg M J E. J Mol Biol. 1998;176:265–285. doi: 10.1006/jmbi.1997.1519. [DOI] [PubMed] [Google Scholar]
  • 24.Jacobsen J R, Keatinge-Clay A T, Cane D E, Khosla C. Bioorg Med Chem. 1998;6:1171–1177. doi: 10.1016/s0968-0896(98)00081-9. [DOI] [PubMed] [Google Scholar]
  • 25.Aggarwal R, Caffrey P, Leadlay P F, Smith C J, Staunton J. J Chem Soc Chem Commun. 1995;N15:1519–1520. [Google Scholar]
  • 26.Wu N, Tsuji S, Cane D E, Khosla C. J Am Chem Soc. 2001;123:6465–6474. doi: 10.1021/ja010219t. [DOI] [PubMed] [Google Scholar]
  • 27.Nicholls A, Honig B. J Comp Chem. 1991;12:435–445. [Google Scholar]
  • 28.Corpet F. Nucleic Acids Res. 1998;16:10881–10890. doi: 10.1093/nar/16.22.10881. [DOI] [PMC free article] [PubMed] [Google Scholar]

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences

RESOURCES