Crystal Structure of the ECH2 Catalytic Domain of CurF from Lyngbya majuscula

Curacin A is a mixed polyketide/nonribosomal peptide possessing anti-mitotic and anti-proliferative activity. In the biosynthesis of curacin A, the N-terminal domain of the CurF multifunctional protein catalyzes decarboxylation of 3-methylglutaconyl-acyl carrier protein (ACP) to 3-methylcrotonyl-ACP, the postulated precursor of the cyclopropane ring of curacin A. This decarboxylase is encoded within an “HCS cassette” that is used by several other polyketide biosynthetic systems to generate chemical diversity by introduction of a β-branch functional group to the natural product. The crystal structure of the CurF N-terminal ECH2 domain establishes that the protein is a crotonase superfamily member. Ala78 and Gly118 form an oxyanion hole in the active site that includes only three polar side chains as potential catalytic residues. Site-directed mutagenesis and a biochemical assay established critical functions for His240 and Lys86, whereas Tyr82 was nonessential. A decarboxylation mechanism is proposed in which His240 serves to stabilize the substrate carboxylate and Lys86 donates a proton to C-4 of the acyl-ACP enolate intermediate to form the Δ2 unsaturated isopentenoyl-ACP product. The CurF ECH2 domain showed a 20-fold selectivity for ACP-over CoA-linked substrates. Specificity for ACP-linked substrates has not been reported for any other crotonase superfamily decarboxylase. Tyr73 may select against CoA-linked substrates by blocking a contact of Arg38 with the CoA adenosine 5′-phosphate.

Tyr 73 may select against CoA-linked substrates by blocking a contact of Arg 38 with the CoA adenosine 5-phosphate.
Polyketides and nonribosomal peptides are important secondary metabolites possessing an array of biological activities and broad chemical diversity (1)(2)(3)(4). Type I modular polyketide synthase (PKS) 5 and nonribosomal peptide synthetase (NRPS) systems sequentially elongate and modify a growing ketide or peptide chain as it is passed from module to module. The biosynthetic machinery generates tremendous structural variation in these compounds by use of specific reductive domain combinations within modules and/or gene cassettes that encode proteins capable of producing unique functionalities.
The natural product curacin A is a mixed polyketide/nonribosomal peptide produced by the marine cyanobacterium Lyngbya majuscula. Curacin A contains a unique combination of functionalities, including a cyclopropyl ring, a cis-vinyl thiazoline heterocycle, and a terminal alkene (Fig. 1A). Curacin A was first shown in 1994 to possess anti-mitotic activity by the inhibition of tubulin polymerization, cell cycle arrest at the G 2 /M transition, and anti-proliferative activity against colon, renal, and breast cancer-derived cell lines (5). Further studies have demonstrated rapid, essentially irreversible binding of curacin A to tubulin with potent inhibition of colchicine binding (6).
The biosynthetic gene cluster (cur) for curacin A was recently identified and characterized, revealing that the metabolite is generated by a hybrid type I PKS/NRPS (7). The first six genes in the cur cluster (curA-curF) have been implicated in the formation of the cyclopropyl ring as well as the thiazoline ring of curacin A. Genes within this region include an "HCS cassette" that encodes an acyl carrier protein (ACP) (CurB), a putative ketosynthase (CurC), a 3-hydroxy-3-methylglutaryl (HMG)-CoA synthase (HCS) (CurD), and two domains containing similarity to enoyl-CoA hydratases (CurE (ECH 1 ) and the N-terminal ϳ260 residues of CurF (ECH 2 )) ( Fig. 2).
HCS cassettes are important sources of ␤-branches in polyketides and allow for the generation of wide chemical diversity at the branch point (Fig. 1). Among the functional groups generated are the cyclopropyl ring of curacin A (7,8), the vinyl chloride of jamaicamide (9), the exomethylene group of the pederin family of polyketides (10), and the pendant methyl (or ethyl) groups of bacillaene (11,12), mupirocin (pseudomonic acid A) (13), myxovirescin A (14,15), and virginiamycin M (16). Subsets of the HCS-containing gene cassette are found in gene clusters that specify production of the bryostatins (17) and leinamycin (18).
Among PKS gene clusters with HCS cassettes, at least six encode ECH 1 /ECH 2 -like enzyme pairs. Recently, the unique catalytic activities of CurE ECH 1 and the CurF ECH 2 domain have been ascertained (8). CurE ECH 1 catalyzes conversion of (S)-HMG-ACP to 3-methylglutaconyl-ACP. Subsequently, the N-terminal ECH 2 domain of CurF catalyzes decarboxylation to 3-methylcrotonyl-ACP, the postulated precursor of the cyclopropane ring of curacin A (Fig. 1A). The analogous activities of two other ECH 1 /ECH 2 -like enzyme pairs was demonstrated for the HCS cassettes of myxovirescin A and bacillaene (11,14). Based on the structures of pathway end products, decarboxylation by ECH 2 -type enzymes are postulated to be followed by proton donation at C-4 leading to a ⌬ 2 unsaturated product in curacin A, bacillaene, virginiamycin M, mupirocin, or myxovirescin A or at C-2 leading to a ⌬ 3 unsaturated product in jamaicamide and pederin (Fig. 1). Thus, ECH 2 represents a critical point of divergence for the generation of correct downstream functionalities in the corresponding natural product. Understanding how these enzymes control product double bond regiochemistry will facilitate the prediction of enzymatic activity for this class of enzymes, as well as aid protein engineering attempts to tailor chemistry for the production of novel natural products.
Enoyl-CoA hydratases belong to the crotonase superfamily, consisting of a wide variety of mechanistically diverse enzymes that exhibit various activities, such as hydratase (19), dehalogenase (20), decarboxylase (21,22), isomerase (Refs. 22-25; Protein Data Bank code 2F6Q), hydrolase (22), and carbon-carbon bond formation (22) or cleavage (27,28). Because of this catalytic diversity, the low sequence identity superfamily lacks an invariant catalytic residue, and superfamily membership is established through crystal structures. The common mechanistic theme of the superfamily is the stabilization of an enolate anion intermediate of phosphopantetheine-linked substrates by two backbone amide groups forming an oxyanion hole (29). CurF ECH 2 has less than 20% sequence identity to identified members of the superfamily. A unique feature of the ECH 2 -like enzymes involved in natural product biosynthesis is that they are thought to depend exclusively on ACP-linked substrates in vivo. Structural clues that enable these enzymes to discriminate between CoA and ACP-linked substrates will increase our understanding of the evolutionary changes that have occurred to favor one substrate over another, as well as to control functional group diversity in the final natural product.
In this paper we report the 1.85 Å crystal structure of the wild type, N-terminal ECH 2 domain of CurF and the 1.65 Å structure of the corresponding Y82F variant. Modeling of the substrate of CurF ECH 2 was performed, and Tyr 82 , Lys 86 , and His 240 were identified as potential catalytic or substrate-binding residues. Site-directed mutagenesis in a coupled ECH 1 /ECH 2 dehydration/decarboxylation assay with ACP-linked substrates demonstrated that CurF ECH 2 His 240 and Lys 86 are critical to catalysis. Finally, we demonstrated that the 3-methylglutaconyl-ACP substrate is preferred over 3-methylglutaconyl-CoA, a preference that may be controlled by a conserved bulky amino acid residue in the HCS cassette class of decarboxylases.

Cloning, Site-directed Mutagenesis, and Protein Expression-
The plasmid pMCSG7::CurFd17 was generated by PCR amplification of coding sequence corresponding to residues 17-257 of CurF from the pML9 cosmid DNA (7) and inserted into the vector pMCSG7 (30). The plasmid pMCSG7::CurFd17 was transformed into BL21(DE3) and grown at 37°C in 2ϫYT medium to an A 600 of 0.6 -0.8 in 2L baffle flasks. The cultures were adjusted to 18°C, and isopropyl ␤-D-thiogalactopyranoside was added to final concentration of 0.4 mM and allowed to grow 8 -16 h with shaking. The cells were harvested by centrifugation, and cell pellets were frozen immediately at Ϫ20°C. Selenomethionine-labeled protein was produced in BL21(DE3) using selenomethionine minimal medium according to the protocol of Guerrero et al. (31). Site-directed mutagenesis was performed by the QuikChange method (Stratagene) (see supplemental materials). All of the constructs were verified by DNA sequencing. CurA (2057-2146)-ACP(II) with a C-terminal His 6 tag was constructed by inserting synthetic DNA (a kind gift from Christopher Calderone and Christopher T. Walsh, Harvard Medical School) into pET29a using NdeI and XhoI restriction sites. Protein overexpression and purification of CurA-ACP(II) were performed as described previously for CurB (8).
Protein Purification-All of the steps were performed at 4°C. The cell pellet from 1 liter of culture was resuspended in 35 ml of buffer A (20 mM Tris, pH 7.9, 500 mM NaCl, 10% glycerol, 20 mM imidazole). Cellytic Express (200 -300 mg) was added prior to lysis by sonication (Sigma-Aldrich). The lysate was cleared by centrifugation at Ͼ38,000 ϫ g for 45 min. Supernatant was filtered by a 0.45-m filter, loaded onto a 5-ml HisTrap (GE Healthcare) nickel-nitrilotriacetic acid resin column, and washed with 10 column volumes of buffer A. The protein was eluted with ϳ200 mM imidazole by a linear gradient of buffer B (20 mM Tris, pH 7.9, 500 mM NaCl, 10% glycerol, 750 mM imidazole). For His tag removal, the fractions were pooled, and buffer was exchanged with buffer A containing 1 mM dithiothreitol and incubated at 4°C for 24 -72 h with 2% (w/w) Histagged tobacco etch virus protease. The reaction mixture was reloaded onto the HisTrap column, and flow-through fractions were pooled, concentrated, and loaded onto a HiLoad 16/60 Superdex 75 (GE Healthcare) column equilibrated with 20 mM Tris, pH 7.9, 500 mM NaCl, and 10% glycerol. The fractions were then combined, concentrated to 20 mg/ml, and either flash frozen in liquid N 2 or stored at 4°C.
Crystallization-The crystals were grown in 24 -72 h at 4°C by microseeding in hanging drops using the vapor diffusion method. Equal volumes were mixed of protein solution and mother liquor containing 1.3-1.8 M sodium malonate, pH 7.0, 50 mM HEPES, pH 6.8, and 0 -25 mM 2-ethanamidoethyl 3-methylbut-2-enethioate. The crystals were harvested in loops and directly frozen by plunging into liquid N 2 . Both trigonal (P321) and rhombohedral (R32) crystals grew in the above crystallization condition from the same seed stock.
Data Collection and Structure Determination-The diffraction data were collected at 100 K on GM/CA-CAT beamlines 23ID-B and 23ID-D at the Advanced Photon Source in the Argonne National Laboratory (Argonne, IL). The data were processed using the HKL2000 suite (32). Initial phasing was performed using a three-wavelength multiple wavelength anomalous dispersion data set from a single trigonal wild type selenomethionine-labeled protein crystal. SOLVE was used to find the four selenium sites (average figure of merit (FOM) ϭ 0.47, score ϭ 25.07) and for multiple wavelength anomalous dispersion phasing (overall FOM ϭ 0.57) (33). RESOLVE was used for density modification (overall FOM ϭ 0.69) and partial automated model building (34,35). In both crystal forms, one ECH 2 subunit was present in the asymmetric unit. Modeling was completed manually using COOT (36). Refinement was performed using REFMAC5 of the CCP4 suite with TLS (37-39). The refined model was used as a probe structure for molecular replacement using PHASER with data from rhombohedral crystals (40,41).
Sequence and Structure Analysis-Sequence alignments were performed by T_COFFEE (42). The figures and structure alignments were generated with PyMOL (43). Surface area calculations were calculated using AREAIMOL (44). Identification of structural neighbors was performed by a DALI search (45).
Structure Alignment and Substrate Modeling-C␣ atoms of the crotonase core domain of CurF ECH 2 (residues 18 -222) were aligned with the analogous core region of liganded crotonase superfamily members carboxymethylproline synthase (Ref. 22 S2). Initial atomic coordinates and topology files for the substrate 3-methylglutaconyl moiety attached to 4-phosphopantetheinic acid were generated using the PRODRG2 server (47). Using the conformation of 4-phosphopantethenic acid bound to rat ECH as a guide (Protein Data Bank code 1DUB), the substrate thioester oxygen was fixed at the position of a water molecule in the oxyanion hole of wild type CurF ECH 2 . Three water molecules overlapping the modeled substrate were removed. The model was refined by energy minimization using the program CNS (48) in 500 steps of conjugate gradient minimization with no experimental energy terms, no crystallographic symmetry restraints, and fixed main chain positions after the addition of polar hydrogens. ECH 1 /ECH 2 -coupled Enzymatic Assay-The activities of the CurF ECH 2 wild type and variants were measured in the ECH 1 -ECH 2 coupled assay, as previously reported (8). In brief, 50 M (R,S)-HMG-CurA-ACP(II) was incubated with 2 M ECH 1 and ECH 2 (wild type or variants) in 50 mM Tris-HCl, pH 7.5, at 37°C for 1 h. The reactions were terminated by 10% formic acid immediately before loading the reaction mixture on the Jupiter C4 (5, 300 A) reverse phase column (Phenomenex), and the proteins were eluted with CH 3 CN (0.05% HCOOH and 0.05% CF 3 COOH)/H 2 O (0.05% HCOOH and 0.05%CF 3 COOH). The chromatogram peaks were normalized by 32 Karat software (Beckman Coulter) and subjected to base-line subtraction before peak area calculations.
Mass Spectrometry Analysis-Multiply protonated CurA-ACP(II) with or without treatment with ECH 1 and/or ECH 2 was generated by electrospray ionization (ESI) at 70 l/h (Apollo ion source; Bruker Daltonics, Billerica, MA) of a solution containing 2 M CurA-ACP(II) (55:45 CH 3 CN:H 2 O with 0.05% HCOOH and 0.05% CF 3 COOH). All of the mass spectra were collected with an actively shielded 7 Tesla Fourier transform ion cyclotron resonance (FTICR) mass spectrometer with a quadrupole front end (APEX-Q, Bruker Daltonics). Ions produced by ESI were externally accumulated in a hexapole for 1 s, transferred via high voltage ion optics, and captured in an ICR cell by gated trapping. This accumulation sequence was looped six times. The ESI capillary voltage was set to Ϫ3.8 kV. Nitrogen drying gas (200 -250°C) was employed to assist desolvation of ESI droplets. All of the data were acquired with XMASS software (version 6.1; Bruker Daltonics) in broadband mode from m/z ϭ 200 -2000 with 512,000 data points and summed over 10 scans. The mass spectra were analyzed with MIDAS analysis software (49).

RESULTS
Structure Determination-Initial crystal screening with a polypeptide including residues 1-257 of CurF produced crystals that diffracted to only ϳ3.8 Å, but an N-terminal truncation including residues 17-257 yielded crystals that diffracted to beyond 2 Å. The structure was solved by multiple wavelength anomalous dispersion phasing using selenomethionyl CurF ECH 2 ( Table 1). The resulting model was refined against the 2.0 Å data set in crystal form I and used to solve the structures by molecular replacement in crystal form II ( Table 2).
Quaternary Structure of the ECH 2 Domain of CurF-The trimer structure fundamental to the crotonase superfamily occurs in both crystal forms of ECH 2 (Fig. 3B). Subunit contacts are virtually identical in the two crystal forms. A total of 15% of the solvent-accessible surface area of the monomer is buried in the ECH 2 trimer (total buried surface area per monomer ϭ 1050 Å 2 ). The extensive buried surface and the lack of water in the subunit interface together indicate that the trimeric association observed in the crystal structure reflects a true quaternary structure for the protein. However, in solution CurF ECH 2 also displayed concentration-dependent dissociation (supplemental Fig. S1), indicating a dynamic equilibrium between trimeric and lower oligomeric states.
Active Site-Despite less than 20% sequence identity with other crotonase superfamily members, strong structural similarity of the CurF ECH 2 domain to crotonase superfamily members enabled us to align the structures and to identify critical elements of the active site, including the substrate-binding tunnel, active site oxyanion hole, and active site chamber. Structural alignments with several other crotonase superfamily members (supplemental Fig. S2) clearly indicate that the backbone amides of residues Ala 78 and Gly 118 form the oxyanion hole. These residues follow conserved Gly 77 and Gly 117 , which have backbone conformations only accessible to glycine. In this manner, the peptide planes of residues 77-78 and 117-118 are oriented so their amides can stabilize the proposed enolate anion intermediate by hydrogen bonding. In the crystal structures, a water molecule occupies the oxyanion hole (Fig. 4A). Attempts to obtain crystal structures of complexes with product analogues were unsuccessful. The substrate of CurF ECH 2 was then modeled into the active site (see "Experimental Procedures"). Only three polar side chains  (Tyr 82 , Lys 86 , and His 240 ) are present within the primarily hydrophobic active site chamber (Fig. 4B).
Decarboxylase Activity of Active Site Variants with an ACPlinked Substrate-Based on the active site structure and substrate modeling, the three polar residues in the active site chamber, Tyr 82 , Lys 86 , and His 240 , were tested by site-directed mutagenesis. Proteins possessing Y82F, K86A, K86Q H240A, or H240Q were produced and activity toward ACP-linked substrates was evaluated in a coupled ECH 1 /ECH 2 assay (8).
FTICR-MS confirmed that the ECH 1 and ECH 2 products were produced from the ACP-linked substrate ( Fig. 5C and supplemental Fig. S3) 8). The effect of each amino acid substitution was identical in assays with ACP-and CoA-linked substrate (Ref. 8 and data not shown). Substitution of Tyr 82 by Phe led to 2-fold reduction in product formation in comparison to wild type (Fig. 5), indicating that Tyr 82 plays little or no role in catalysis. Consistent with this conclusion, the Tyr 82 phenyl ring in the 1.65 Å crystal structure of the Y82F FIGURE 3. CurF ECH 2 structure. A, stereo diagram of the monomer of the CurF ECH 2 domain with the active site chamber indicated by an asterisk. B, trimeric structure of the CurF ECH 2 domain with the active site chamber indicated by an asterisk. C, multiple sequence alignment of ECH 2 -like decarboxylases from gene clusters encoding the biosynthesis of curacin (CurF), bacillaene (PksI, 44% identity), virginiamycin M (VirE, 36% identity), myxovirescin A (TaY, 43% identity), mupirocin (MupK, 39% identity), jamaicamide (JamJ, 59% identity), pederin (PedI, 36% identity), and sequences of the two structurally characterized crotonase superfamily decarboxylases (CarB [17% identity] and MMCD [15% identity]). Secondary structure elements of CurF are shown above the alignment, residues of the oxyanion hole are indicated by triangles below the alignment, invariant residues are red, sites of conservative substitution are blue, and similar residues are green. Every tenth residue is underlined. DECEMBER  Decarboxylase Preference for ACP-linked Substrates over CoAlinked Substrates-CoA-and ACPlinked substrates for ECH 2 were prepared by enzymatic synthesis to evaluate the substrate preference of the decarboxylase without the complication of a coupled ECH 1 /ECH 2 assay. CurF ECH 2 had a 20-fold preference for the ACP-linked substrate under our standard assay conditions. CurF ECH 2 decarboxylated Ͼ70% of 3-methylglutaconyl-ACP but only 3% of 3-methylglutaconyl-CoA (Fig. 6). A preference for ACPlinked substrates has not been reported for any member of the crotonase superfamily. Although it presumably has an ACP-linked substrate in vivo, CurE ECH 1 differed from CurF ECH 2 and showed no substrate preference by providing similar yields of 3-methylglutaconyl-ACP from (R,S)-HMG-ACP and of 3-methylglutaconyl-CoA from (R,S)-HMG-CoA (Fig. 6).

DISCUSSION
The N-terminal ECH 2 domain of CurF possesses the crotonase superfamily fold. As is seen with the two other structurally characterized crotonase superfamily members possessing biotin-independent decarboxylase activity  . ECH 1 /ECH 2 coupled enzymatic assays for ECH 2 wild type and mutants. A, HPLC chromatograms showing the (R,S)-HMG-CurA-ACP(II) and associated species before and after ECH 1 /ECH 2 coupled reactions at 37°C for 1 h; the peaks correspond to (R,S)-HMG-CurA-ACP(II) (peak 1), 3-methylglutaconyl-CurA-ACP(II) (peak 2), and 3-methylcrotonyl-CurA-ACP(II) (peak 3). B, comparison of the peak areas of 3-methylcrotonyl-CurA-ACP(II) for ECH 2 wild type and mutants. The data were subjected to normalization and base-line subtraction before the peak area calculation, and the assays were duplicated. C, FTICR-MS identification of the ACP-linked substrate and products. Only the preferred substrate for ECH 1 , (S)-HMG-ACP, is shown in the reaction scheme (8).
(CarB (22) and MMCD (21)), the C-terminal helical domain packs against the monomer in a self-association fold (23). Thus, the active site chamber is formed exclusively by residues within one monomer of the trimer.
The diversity of chemical reactions catalyzed by crotonase superfamily members and their highly divergent sequences result in little if any conservation within the active site chamber. Nevertheless, the close structural similarity with other crotonase superfamily members suggests strongly that Ala 78 and Gly 118 form the oxyanion hole of CurF ECH 2 (supplemental Fig. S2). The site-directed mutagenesis results show that Lys 86 and His 240 play crucial roles in catalysis (Fig. 5). We propose a decarboxylation mechanism in which the oxyanion hole anchors the substrate in the active site and stabilizes the enolate anion, His 240 hydrogen bonds with the substrate carboxylate, and Lys 86 is the proton donor to the C-4 position of the decarboxylated substrate (Fig. 7). The hydrophobic environment of the active site chamber enhances the reactivity of these groups.
Other than Lys 86 and His 240 , Tyr 82 is the only polar side chain within the active site chamber. However, elimination of the phenolic hydroxyl by substitution with Phe had only a 2-fold effect on product formation, indicating that Tyr 82 does not play a major role in substrate hydrogen bonding or proton donation. Interestingly, Tyr 140 of MMCD was proposed to hydrogen bond with the carboxylate group of methylmalonyl CoA in a mechanism analogous to that outlined above (21). Tyr 82 of CurF ECH 2 and Tyr 140 of MMCD are not in analogous parts of the protein structures. Significantly, Tyr 82 is not conserved among sequences of decarboxylases in the crotonase superfamily (Fig. 3C).
His 240 is conserved in all decarboxylases encoded by HCS cassettes and also in CarB (Fig. 3C). His 240 is in an excellent position to assist catalysis by hydrogen bonding with the substrate carboxylate, perhaps orienting the carboxyl optimally to be a leaving group (Fig. 7A). In liganded CarB, His 229 is in an identical position to His 240 of CurF ECH 2 and also was proposed to function by stabilizing the substrate carboxylate (22). The importance of His 240 in catalysis was demonstrated in the H240A and H240Q variants of CurF ECH 2 , in which catalysis was severely impaired (Fig. 5). His 240 of CurF ECH 2 is the only invariant polar residue within the active site chamber, and we propose that it serves to stabilize the substrate carboxylate here and in other ECH 2 -like decarboxylases of HCS cassettes (Fig. 3C). An analogous His does not exist in MMCD. The equivalent helix to ␣10 of CurF ECH 2 , which contains His 240 , is further from the active site chamber in MMCD and forms extensive trimer contacts with the adjacent monomer.
After decarboxylation, proton donation to the substrate C-4 carbon to form isopentenyl-ACP is necessary for enolate collapse to the ⌬ 2 unsaturated product (Fig. 7B). Lys 86 is the most likely proton donor in CurF ECH 2 , given the relatively fixed position of the enolate anion in the oxyanion hole during substrate binding and the hydrophobic nature of the active site chamber. Lys 86 resides on helix ␣2 and has the necessary flexibility and reach to accommodate proton donation at the C-4 carbon of the substrate (Fig. 4B). His 240 is not a candidate pro-   DECEMBER 7, 2007 • VOLUME 282 • NUMBER 49 ton donor because it appears to be unable to reach the C-4 carbon of the decarboxylated intermediate, as required for subsequent product formation, and also unable to move closer to the C-4 carbon. Its backbone is held in place as part of the long helix ␣10, and its side chain is positioned by a hydrogen bond with the carbonyl of Thr 147 (Fig. 4A) deep within the active site chamber. In contrast, Lys 86 is located on helix ␣2, the most mobile part of the CurF ECH 2 structure as seen in distinctly higher crystallographic temperature factors. This region of the crotonase fold is generally mobile and is completely disordered in some structures.

Structure of CurF Decarboxylase
Proton addition to the C-4 carbon following decarboxylation should produce a ⌬ 2 unsaturated product, whereas proton addition to C-2 should yield the ⌬ 3 product (Fig. 7). Examples of both reaction routes exist, based on the structures of the PKS products (Fig. 1). Thus, the mechanism of ECH 2 is a key determinant of the regiochemistry of the ultimate product of each pathway. The active site chambers of other ECH 2 -like decarboxylases appear equally as hydrophobic as the CurF ECH 2 active site with the exception of the loop between helices ␣2 and ␣3 (residues 89 -95 of CurF; Fig. 3C). This region covers the active site ( Fig. 4B) but differs in length and sequence among the ECH 2 decarboxylases (Fig. 3C). It is uncertain whether all residues aligned with CurF 89 -95 constitute the ␣2-␣3 loop in these decarboxylases or whether some of them are part of a longer helix ␣3, as in CarB and MMCD (supplemental Fig. S2). Variability in the ␣2-␣3 loop is expected among the ECH 2 decarboxylases. Each enzyme must possess a unique conformation to accommodate its substrate, often bulkier than the methyl group of the CurF ECH 2 substrate (Fig. 1). In addition, a unique proton donor must be positioned to generate specifically a ⌬ 2 or ⌬ 3 unsaturated product.  (Fig. 3C).
The CurF ECH 2 domain is the first crotonase superfamily member shown to act preferentially upon ACP-linked substrates. Product formation in vitro with an ACP-linked substrate was 20-fold greater than with a CoA-linked substrate (Fig. 6). Despite the strong preference for ACP-linked substrates by CurF ECH 2 , structure alignment with other crotonase superfamily enzymes having CoA ligands (supplemental Fig. S2) revealed remarkably few structural changes in the CoAbinding region. Structural motifs that recognize the CoA adenine ring in other crotonase enzymes and the overall backbone conformation that forms the CoA adenosine-binding site are identical in CurF ECH 2 . One notable difference in CurF ECH 2 is the position of the Tyr 73 side chain. In the CoA-dependent enzymes, a basic side chain occupies this space and interacts with the 4-phosphate of pantothenic acid and the 5Ј-phosphate of the CoA nucleotide. CurF Tyr 73 is conserved as a bulky phenylalanine or tyrosine residue in the ECH 2 -like decarboxylases from HCS cassettes with presumed specificity for ACP-linked substrates (Fig. 3C), whereas the analogous position is always a smaller alanine, serine, or valine residue in the CoA-dependent enzymes. Thus, it appears that CoA is a poorer substrate of CurF ECH 2 in part because the Tyr 73 side chain blocks a basic side chain (Arg 38 in CurF) from entering the CoA-binding site. CurF ECH 2 containing an alanine substitution at the Tyr 73 site was insoluble and has frustrated efforts to test this hypothesis directly. The structural conservation in this region implies a relatively recent evolution of protein function toward ACPlinked substrates.
An intriguing outcome of the CurF ECH 2 structure is the apparent symmetry mismatch between its trimeric structure and the fundamentally dimeric structure of several downstream domains in the CurF polypeptide (enoyl reductase, ketosynthase, and dehydratase) (7,51). The CurF ECH 2 domain possesses the crotonase self-association fold with the active site fully formed by each monomer and also demonstrates a capacity for trimer dissociation in solution (supplemental Fig. S1). Thus, the ECH 2 domain may be monomeric in the context of full-length CurF, similar to the bacterial fatty acid ␤-oxidation multienzyme complex in which an N-terminal monomeric crotonase domain is fused to a dimeric dehydrogenase domain (50). Nevertheless, existence of the classical crotonase trimer in the isolated ECH 2 domain (Fig. 3B) is strongly suggestive of a trimeric ECH 2 within full-length CurF. This could be accomplished in either of two ways. CurF could be a hexamer containing two ECH 2 trimers flexibly tethered to three dimers of the downstream domains. Alternatively, CurF could be a dimer in which ECH 2 forms a heterotrimer with ECH 1 (CurE). Determination of the oligomeric organization of CurF is an on-going investigation.
In summary, the crystal structure of the N-terminal ECH 2 domain of CurF PKS-NRPS multifunctional protein from L. majuscula possesses a crotonase fold. The backbone amides of Ala 78 and Gly 118 form an oxyanion hole. The hydrophobic active site chamber includes only three polar side chains. Of these, Lys 86 and His 240 are critical for catalytic activity, but Tyr 82 is not. Based on assay results and comparisons with other crotonase decarboxylases, His 240 is proposed to stabilize the substrate carboxylate, and Lys 86 is proposed to donate a proton to the C-4 position of product. These structural features enable CurF ECH 2 to generate specifically the ⌬ 2 alkene regiochemistry in formation of the key isopentenyl-ACP product during curacin A biosynthesis. CurF ECH 2 is highly selective for ACP-linked substrates, whereas CurE ECH 1 is nonselective. Strong sequence conservation of both the ECH 1 dehydratases (35-95% identity) and the ECH 2 decarboxylases (35-60% identity) encoded by HCS cassettes suggests that these domains have similar substrate preferences to the curacin ECH 1 and ECH 2 .