Roles of the Active Site Water, Histidine 303, and Phenylalanine 396 in the Catalytic Mechanism of the Elongation Condensing Enzyme of Streptococcus pneumoniae*

β-Ketoacyl-ACP synthases catalyze the condensation steps in fatty acid and polyketide synthesis and are targets for the development of novel antibiotics and anti-obesity and anti-cancer agents. The roles of the active site residues in Streptococcus pneumoniae FabF (β-ketoacyl-ACP synthase II; SpFabF) were investigated to clarify the mechanism for this enzyme superfamily. The nucleophilic cysteine of the active site triad was required for acyl-enzyme formation and the overall condensation activity. The two active site histidines in the elongation condensing enzyme have different electronic states and functions. His337 is essential for condensation activity, and its protonated Nϵ stabilizes the negative charge developed on the malonyl thioester carbonyl in the transition state. The Nϵ of His303 accelerated catalysis by deprotonating a structured active site water for nucleophilic attack on the C3 of malonate, releasing bicarbonate. Lys332 controls the electronic state of His303 and also plays a critical role in the positioning of His337. Phe396 functions as a gatekeeper that controls the order of substrate addition. These data assign specific roles for each active site residue and lead to a revised general mechanism for this important class of enzymes.

␤-Ketoacyl-ACP synthases catalyze the condensation steps in fatty acid and polyketide synthesis and are targets for the development of novel antibiotics and anti-obesity and anti-cancer agents. The roles of the active site residues in Streptococcus pneumoniae FabF (␤-ketoacyl-ACP synthase II; SpFabF) were investigated to clarify the mechanism for this enzyme superfamily. The nucleophilic cysteine of the active site triad was required for acyl-enzyme formation and the overall condensation activity. The two active site histidines in the elongation condensing enzyme have different electronic states and functions. His 337 is essential for condensation activity, and its protonated N stabilizes the negative charge developed on the malonyl thioester carbonyl in the transition state. The N of His 303 accelerated catalysis by deprotonating a structured active site water for nucleophilic attack on the C3 of malonate, releasing bicarbonate. Lys 332 controls the electronic state of His 303 and also plays a critical role in the positioning of His 337 . Phe 396 functions as a gatekeeper that controls the order of substrate addition. These data assign specific roles for each active site residue and lead to a revised general mechanism for this important class of enzymes.
The condensing enzymes play a central role in fatty acid biosynthesis by elongating the growing acyl chain by two carbon atoms to initiate each elongation cycle (1,2). Somewhat uniquely in biological synthesis, the enzymes create a carbon-carbon bond via a Claisen-like condensation reaction (3). Specifically, they catalyze the condensation of malonyl-acyl carrier protein (ACP) 2 with an acyl-ACP intermediate via a two-step ping-pong kinetic mechanism. In the first step, an acyl chain from either acyl-CoA or acyl-ACP is transferred to an active site cysteine, and the cofactor is released. During the second step, malonyl-ACP binds, and a carbanion is generated on the C2 of malonate concomitant with the release of the C3 carboxyl group (4 -6). The carbanion then attacks the acyl-enzyme intermediate to produce the ␤-ketoacyl-ACP product. In the dissociated, type II synthases, the condensation reaction is carried out by monofunctional enzymes (7), and most bacteria have only a single elongation-condensing enzyme that belongs to the FabF class. In mammals and yeast, the condensing enzyme component, KS, is fused into a multidomain complex referred to as the type I or associated FAS system (8). However, it is clear from primary sequence analysis that the active site of the FAS I condensation module is very similar to the FAS II elongation enzymes (Fig. 1A). The polyketide synthases also contain a condensing enzyme module in which the same signature active site residues can be identified (Fig. 1A).
The importance of the elongation condensing enzymes in regulating fatty acid formation (7,9) and the unique chemistry of the reaction that they catalyze (3) have focused our interest on understanding the specific tasks of each active site residue in catalysis. In addition, these enzymes have emerged as attractive targets for the development of new broadspectrum antibiotics (10 -12) and anti-obesity/anti-cancer drugs (13)(14)(15)(16)(17), and there is growing interest in engineering the polyketide synthases to produce novel therapeutic agents (18). These efforts will be facilitated by a complete mechanistic understanding of the active site. At their catalytic cores, the elongation enzymes possess a Cys-His-His triad. These residues have been mutated and are thought to be critical to the overall forward condensation reaction (8, 19 -22), although the functions of these residues in the partial reactions of the catalytic cycle are not precisely known. There are also conserved lysine, glutamate and phenylalanine residues in the vicinity of the active site, and the structural and/or catalytic roles for these residues are not understood (Fig.  1A). Our systematic analysis of a panel of mutants in a model elongation condensing enzyme and the correlation of this new information with new and previously determined structures led to a reevaluation of the roles for the active site residues in the catalytic cycle. We selected SpFabF as the basis for this work, since its x-ray structure is known at the highest resolution (1.3 Å) (21). A stereo view of the SpFabF active site illustrating the orientation of the key residues investigated in this study is shown in Fig. 1B.
Construction of Mutants and Protein Purification-A DNA fragment encoding the wild-type S. pneumoniae FabF was cloned into the pET15b vector between the NdeI and BamHI sites. Mutations at the active site residues were introduced using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). All clones were checked by automated sequencing on an ABI Prism 3700 DNA Analyzer. His-tagged mutant proteins were expressed in Rossetta (DE3) cells and purified with Ni 2ϩ -nitrilotriacetic acid-agarose (Qiagen, Valencia, CA). The proper folding of the mutant proteins was confirmed by circular dichroism spectroscopy.
␤-Ketoacyl-ACP Synthase Assay-The condensation assay for EcFabB and SpFabF was described previously (29). Briefly, the assays contained 45 M myristoyl-ACP, 50 M [2-14 C]malonyl-CoA (specific activity, 52 Ci/mol), 100 M ACP, 1 g of EcFabD, and the indicated amount of EcFabB or SpFabF in a final volume of 20 l. ACP was reduced by 0.3 mM DTT before the other reaction components were added. After incubation of the sample at 37°C for 20 min, the reaction was stopped by adding 0.4 ml of reducing agent containing 0.1 M K 2 HPO 4 , 0.4 M KCl, 30% tetrahydrofuran, and 5 mg/ml NaBH 4 . The reaction mixtures were vigorously mixed and incubated at 37°C for 40 min, followed by extraction with 0.4 ml of toluene. The 14 C-isotope in the upper phase was quantitated.
Reconstituted Fatty Acid Synthase Assay-The overall fatty acid synthase activity in vitro was measured using purified fatty acid biosynthetic enzymes (30). C]malonyl-CoA (specific activity, 10.4 mCi/mmol), 2 g of EcFabD, 2 g of EcFabH, 2 g of EcFabG, 2 g of EcFabZ, 2 g of EcFabI, and 4 g wild-type or mutant SpFabF in 100 mM sodium phosphate buffer, pH 7. ACP was reduced by 1 mM ␤-mecaptoethanol for 20 min at 37°C before the remaining components were added. Following incubation at 37°C for 30 min, the reactions were stopped by placing the reaction tubes on an ice slurry. The entire mixture was loaded onto a conformationally sensitive 2 M urea, 13% acrylamide gel, and the gel was stained, dried, and exposed to a phosphor storage screen (Amersham Biosciences). The radioactivity was detected by a Typhoon 9200 PhosphoImager and quantitated using ImageQuant software.
Transacylation Assay-The assay for the transacylase activity of the condensing enzyme contained 25 M ACP, 0.3 mM DTT, 50 M [1-14 C]lauroyl-CoA (specific activity, 55 mCi/mmol) and the indicated amount of (2, 1, 0.5, 0.25, or 0.125 g) enzyme in 100 mM potassium phosphate, pH 6.8, buffer in a total volume of 40 l. ACP was reduced by DTT in buffer before the rest of the reaction components were added. Following incubation at 37°C for 30 min, the reactions were stopped by pipetting 35 l of the reaction mix onto DE81 filter discs (Whatman). The discs were then washed three times with chloroform/methanol/acetic acid (2:5:3 volume ratio) containing 0.2 M LiCl (20 min/wash with 20 ml/disc). The discs were dried and counted to quantitate [ 14 C]acyl-ACP.
Malonyl-ACP Decarboxylation Assay-The decarboxylase assay measured the conversion of malonyl-ACP to acetyl-ACP and triacetic acid lactone (TAL) (31). The reaction mix contained 50 M ACP, 0.3 mM DTT, 50 M [2-14 C]malonyl-CoA (specific activity, 52 mCi/mmol), 1 g of EcF-abD, and 10 g of wild-type or mutant SpFabF in 100 mM potassium phosphate, pH 6.8, in a final volume of 60 l (26). ACP was reduced by DTT in buffer before the rest of the reaction components were added. The reaction was started by the addition of SpFabF. Following incubation at 37°C for 40 min, the reactions were stopped by placing the reaction tubes on an ice slurry, and 35-and 25-l aliquots were removed. The 35-l portion was loaded onto a conformationally sensitive 0.5 M urea, 13% acrylamide gel. The 25-l reaction mix was spotted on a preadsorbent Silica Gel G plate (Analtech) developed in chloroform/methanol/acetic acid (24:1:1, v/v/v), dried, and exposed to a phosphor storage imager (Amersham Biosciences). The distribution of radioactivity was detected and quantitated as described above.
Malonyl-ACP Bicarbonate Exchange Reaction-The carbon exchange between malonyl-ACP and bicarbonate is the reverse reaction of the condensing enzyme (5). The reaction contained 50 M malonyl-ACP, 5 M myristoyl-ACP, 10 mM [ 14 C]bicarbonate and 8 g of wild-type SpFabF or mutants in 100 mM potassium phosphate, pH 6.8, in a final volume of 20 l. Following incubation at 37°C for 20 min, the reaction was stopped by placing the tubes in an ice slurry before being loaded onto a conformationally sensitive 0.5 M urea, 13% acrylamide gel. Radioactivity was detected and quantitated as described above.
Complementation of the fabB(Ts) Growth Phenotype-Strain CY274 (fabB15(Ts)) was unable to grow at the nonpermissive temperature (42°C) unless unsaturated fatty acids (oleate) were supplied or the fabB gene was expressed in trans. The ability of wild-type and mutant FabB to complement the fabB(Ts) phenotype was tested after transformation of strain CY274 with plasmids carrying wild-type or mutant fabB (32).
SpFabF-ACP Interactions-AlphaScreen technology is an experimental method to quantitate protein-protein interactions (33). Upon laser excitation, a chemical signal is generated on the donor beads (streptavidin-coated). When a specific interaction brings the acceptor beads (coated with a specific antibody or nickel chelater) to the proximity of the donor beads, a cascade of energy transfers take place, emitting highly amplified fluorescence (or AlphaScreen signal) at a wavelength that is lower than that of the excitation. We measured ACP affinity using the competitive binding approach. Briefly, 1 M biotinylated ACP was mixed with 1 M His-tagged wild-type SpFabF or mutant in the presence of different concentrations of regular ACP (120 nM to 10 M) in a 384-well ProxiPlate (PerkinElmer Life Sciences). After a 30-min incubation, streptavidin donor beads and Ni 2ϩ chelate acceptor beads (50 g/ml final concentration for each bead; Alphascreen Histidine Detection Kit, PerkinElmer Life Sciences) were added to the above solutions. The reaction mixtures were then incubated for 1 h before being read by Fusion TM Universal Microplate Analyzer (PerkinElmer Life Sciences), with excitation at 680 nm and emission at 600 nm.
X-ray Crystallography-Crystals of the SpFabF[H303A] mutant were grown in 20% polyethylene glycol 3350, 0.2 M potassium acetate and were in space group P2 1 2 1 2 with cell dimensions a ϭ 60.39 Å, b ϭ 88.82 Å, and c ϭ 61.05 Å. Rod-shaped crystals measuring ϳ0.25 mm on edge were harvested after 9 days and frozen in liquid nitrogen using 25% PEG 400 as a cryoprotectant. A 2.6 Å data set was collected at the SER-CAT ID beamline (Sector 22 at the Advanced Photon Source in Argonne National Laboratory), and processed using MOSFLM (34) and SCALA (35). The structure was solved by molecular replacement, using the program AMoRe (36) and the native SpFabF structure (Protein Data Bank code 1OX0) as the search model, and refined using a combination of REFMAC (37) and CNS (38). Data collection and refinement statistics are shown in Table 3. Model building was carried out using the O program (39).

RESULTS
Condensing Enzyme Activity of FabF Mutants-A subset of completed conserved residues within four protein segments in the elongation class of condensing enzymes of FAS I, FAS II, and the polyketide synthases was selected for analysis ( Fig. 1). The active site cysteine (Cys 164 ) and the two histidines (His 303 and His 337 ) that comprise the central Cys-His-His triad were obvious choices for site-directed mutagenesis. Also selected were Lys 332 and Glu 346 that have been implicated in catalysis, although Glu 346 is not conserved in FAS I and polyketide synthases (Fig. 1). Finally, Phe 396 was also included in the analysis, because it is located in a flexible region of the active site surrounded by conserved glycines. Each of these residues ( Fig. 1) was mutated to alanine in SpFabF, and the mutants were analyzed for their ability to carry out the biochemical activities that comprise the full and partial reactions of the condensing enzyme catalytic cycle.
The overall condensation reaction was determined using [2-14 C]malonyl-ACP and myristoyl-ACP as substrates. With the exception of SpFabF[H303A] and SpFabF[E346A], which retained 26 and 100% of the wild-type activity, respectively, all of the mutant condensing enzymes were deficient in their overall condensation activity ( Fig. 2A, Table 1). The function of the mutants was also evaluated in an assay to reconstitute multiple rounds of fatty acid synthesis starting with acetyl-CoA as the primer (Fig. 2B). The SpFabF[H303A] and SpFabF[F396A] mutants retained activity, whereas the other mutants were inactive in this assay format, although neither active mutant produced as much long-chain acyl-ACPs as the wild type. In order to detect very low levels of activity, the assays were repeated at higher protein concentrations, and this revealed that SpFabF[K332A] exhibited a low, but significant, activity that was estimated at 0.2% of wild-type ( Table 1). The activity of the other mutant proteins was not detected at the highest protein concentrations permitted by the assay format, setting the condensation activity of these constructs to Ͻ0.2% (Table 1).
The results with SpFabF[H303A] and SpFabF[K332A] were inconsistent with previous conclusions from work on EcFabB (40). This study concluded that the equivalent residues of His 303 and Lys 332 in EcFabB are both essential, whereas our analysis indicated that the these two residues have important, but not absolutely essential, roles in promoting the overall condensation reaction. This apparent inconsistency might be attributable to subtle differences between the FabB and FabF subfamilies. We explored this point by constructing the same panel of mutants in EcFabB as reported previously (40); all of the FabB mutants appeared to be inactive using a protein concentration in the linear range for the wild-type enzyme (not shown). However, we were able to detect a trace of activity (ϳ1% of wild type) in the EcFabB[H298A] and EcFabB[K328A] mutants when assayed at higher protein concentrations, illustrating that they retained partial condensation activity in vitro. We corroborated this conclusion using an in vivo complementation test to verify the activities of the FabB mutants ( Table  2). The panel of fabB mutant alleles was introduced into E. coli strain CY274 (fabB15(Ts)), and the resulting strains were scored for correction of the temperature-sensitive growth phenotype ( Table 2). The EcFabB[H298A] and EcFabB[K328A] mutants restored growth at the nonpermissive temperature, demonstrating that they retained condensing enzyme activity in vivo. In contrast, the inactive EcFabB[C163A] and EcFabB[H333A] mutants were unable to complement the growth phenotype. Importantly, since the condensation enzymes function as dimers, the lack of activity of the latter two mutants ruled out the possibility that complementation arose from the stabilization of the temperature-sensitive fabB allele by the expression of a properly folded protein. The results with the E. coli mutant panel ( Table 2) were qualitatively the same as with the SpFabF panel (Table 1), although the level of enzyme activity in the EcFabB[H298A] was less than the residual activity in SpFabF [H303A]. Thus, the different condensing enzyme mutant sets differed in their absolute values of activity, but the rank order of activity in the mutant proteins was the same.
Active Site Residues Participating in Transacylation-The first step in the condensing enzyme reaction is the transfer of the acyl group to the active site cysteine, and this activity was assayed by monitoring the transfer of the isotopic labeled acyl group from [1-14 C]lauroyl-CoA to ACP via the acyl-enzyme intermediate. As expected, mutants lacking this critical cysteine were completely inactive in the transacylation step (Fig. 2C) indicating that none of these residues were absolutely essential for transthioesterification. The transacylation assay can also detect accumulation of the [ 14

Condensing Enzyme Mechanism
His 337 is within hydrogen-bonding distance of Cys 164 , its electronic state is inconsistent with a role for this residue in deprotonating the active site sulfhydryl. The N␦ nitrogen of His 337 accepts a hydrogen bond from a backbone amide, meaning that the lone pair cannot be on the N nitrogen adjacent to Cys 164 (Fig. 1B). Thus, the fixed electronic state of the histidine renders it unable to facilitate the abstraction of a proton from Cys 164 . Perhaps one could postulate a conformational change, but none of the many available structures give any hint of this. The cerulenin and thiolactomycin inhibitor structures (27,41) show that the N nitrogen of His 337 donates a hydrogen bond to the carbonyl of the antibiotics, supporting the placement of the lone pair on N␦ rather than on the N adjacent to Cys 164 . In the related FabH condensing enzymes, this histidine is replaced by an asparagine residue, and the O␦1 receives the same hydrogen bond from the backbone amide as does the N␦ nitrogen of His 337 , again placing the protonated nitrogen adjacent to the active site cysteine. Although there may be mechanistic differences between the FabH and FabB/F classes of condensing enzymes, the requirement to generate the thiolate is a common feature and is attributed to the strong helix dipole conserved between the two proteins rather than the histidine, which is not conserved (26).

Residues Participating in the Decarboxylation
Step-The formation of acetyl-ACP from malonyl-ACP measures the ability of the enzymes to create the carbanion intermediate via the release of the C3 carboxyl group and occurs in the absence of an acyl-enzyme intermediate. Although this is a futile, nonproductive reaction, it independently measures the status of the decarboxylation half-reaction. This partial reaction requires that malonyl-ACP bind to the active site prior to the formation of the acyl-enzyme intermediate. This reversal of the normal order of substrate binding occurred at a low rate in the wild-type enzyme, indicating that the protein was able to limit this nonproductive reaction (Fig. 2D). However, the SpFabF[C164A] mutant exhibited robust malonyl-ACP decarboxylase activity that generated copious amounts of acetyl-ACP (Fig. 2D). A second derailment reaction that complicates the analyses of the decarboxylation activity of the SpFabF single mutants was the formation of the TAL by-product. TAL formation requires the presence of the active site cysteine (see below); thus, we eliminated TAL formation and maximized decarboxylation activity by constructing double mutants. SpFabF[C164A/H337A] did not exhibit decarboxylation activity, whereas SpFabF[C164A/H303A] retained activity, indicating that only His 337 played a critical role in the formation of the carbanion (Fig. 2D). Somewhat surprisingly, SpFabF[C164A/ K332A] was also defective in acetyl-ACP formation (Fig. 2D, Table 1), indicating that Lys 332 and His 337 were both involved in the formation of the carbanion. This was a counterintuitive result, because the crystal structure of SpFabF reveals that the side chain of Lys 332 interacts via a water molecule to His 303 rather than to His 337 (Fig. 1B).
The Formation of TAL-TAL is an abortive product of the elongation condensing enzymes (31) arising from the reverse order of substrate binding. Malonyl-ACP binds to the free enzyme and is decarboxylated. The resulting acetyl moiety is then transferred to the active site cysteine, and ACP is released. A second molecule of malonyl-ACP then binds to the acetyl-enzyme intermediate and is condensed to form ␤-ketobutyryl-ACP, which is released. In the absence of the ␤-ketoacyl-ACP reductase to convert ␤-ketobutyryl-ACP to ␤-hydroxybutyryl-ACP (43)(44)(45), the ketoacyl group is again transferred to the cysteine and condensed with a third molecule of malonyl-ACP to release 3,5-diketohexanoyl-ACP. This molecule undergoes an internal, spontaneous cyclic rearrangement to produce TAL and ACP. It was important to measure this derailment condensation reaction in our analyses of the activities of the mutants to verify that any decrease in either condensation or acetyl-ACP formation was not due to an increase in TAL formation. TAL formation requires a different assay, because it is not attached to ACP and therefore is not detectable in the gel electrophoretic assays employed in Fig. 2, B and D. TAL formation by wild-type SpFabF was low (0.122 Ϯ 0.03 pmol/g/ min) but detectable (Fig. 2E, Table 1). SpFabF[C164A], SpFabF[H337A], and the double mutants containing any one of these mutations did not produce TAL (Fig. 2E, Table 1). SpFabF[K332A] produced only 10% of the amount of TAL as wild type, consistent with its low condensation activity. However, SpFabF[H303A] displayed a robust TAL formation (Fig. 2E). This was a significant finding, which revealed that the apparent deficiency in the formation of acetyl-ACP by SpFabF[H303A] was not due to an inability to decarboxylate the substrate per se but rather due to efficient TAL production (Fig. 2E). Another important observation was that SpFabF[F396A] also produced more TAL than would be predicted based on its deficient condensation activity assayed with long-chain acyl-ACP (Table 1).
Malonyl-ACP/Bicarbonate Exchange Reaction-The diagram of the overall chemistry performed by the condensing enzymes typically shows that the C3 carboxyl group of malonate is released as CO 2 . However, the first assay developed for ACP used the condensing enzymes to carry out the malonyl-CoA/[ 14 C]bicarbonate exchange reaction (5), and this activity, which corresponded to the reverse reaction, pointed to bicarbonate and not CO 2 as the product of the reaction. Indeed, bicarbonate was recently verified as the product of the mammalian FAS I condensation step (6). We tested the ability of our SpFabF enzymes to perform the reverse exchange reaction and were only able to detect activity in wild type and SpFabF[H303A] mutant (Fig. 3A). The data suggested that SpFabF[H303A] utilized the same mechanism for the reverse reaction as the wild-type enzyme, albeit at a slower rate (Fig. 3A). Thus, not only did SpFabF[H303A] retain significant condensation activity, it was also able to carry out the reverse reaction. Conversely, the inability of the other mutants to carry out the exchange reaction was consistent with their defects in acetyl-ACP formation, TAL formation, and the overall condensation reaction (Table 1). We then tested whether bicarbonate can act as a product inhibitor of malonyl-ACP decarboxylation (Fig. 3B). The activities of SpFabF[C164A] and SpFabF[C164A/H303A] were compared with elimination of the complication of TAL formation in the decarboxylation assay, and both enzymes were equally inhibited by bicarbonate in the assay. These experiments were consistent with bicarbonate being the product of the reaction in the presence and absence of His 303 .
The crystal structure of SpFabF shows a water molecule (W1) in the active site forming a strong hydrogen bond with the side chain of His 303 (Fig. 1B). This water molecule has a B factor similar to those of the surrounding residues (21), and we hypothesize that it represents a catalytic water that is activated by the fixed electron state of His 303 to attack the carboxyl group of malonyl-ACP, leading to the release of bicarbonate. If His 303 functions in this capacity to accelerate catalysis, then SpFabF[H303A] would be more catalytically defective at acidic than alkaline pH, where hydroxide ion is present at higher concentrations. Measurement of the pH rate profiles supported this hypothesis. For the wild-type protein, the pH rate profile for the overall condensation activity was rather flat between pH 6 and 8, with a 20% decrease at pH 6.0 compared with pH 7.0. In contrast, the SpFabF[H303A] mutant activity was reduced by 75% at pH 6.0 compared with pH 7.0 (Fig. 3C). The malonyl-ACP/bicarbonate exchange of SpFabF[H303A] exhibited the same pH dependence with the lowest activity at pH 6.0 (Fig. 3D).
Interaction between ACP and SpFabF Mutants-The defects in catalytic activity in the SpFabF mutants could possibly be due to structural changes that compromised their interaction with ACP substrates. Because many of the mutants were catalytically inactive in one or more of the reactions, the proteins could not be systematically compared by kinetic analysis. Examining the crystal structures of the mutants did not give any indication that the mutations caused changes in the protein surface. This conclusion was verified by meas-uring the ability of ACP to dissociate the interaction with each Histagged SpFabF mutant protein and biotinylated ACP (Fig. 4). The base-line signal was obtained with 1 M each SpFabF and biotinylated ACP. Although there were some small differences in the ACP displacement curves between the proteins, each of the mutant proteins bound ACP with essentially the same affinity, indicating that alteration in protein-protein interactions did not contribute to the catalytic deficiencies of the SpFabF mutants.
Structure of the SpFabF[H303A] Mutant-The removal of the His 303 side chain from the active site had additional important effects on the structure of the FabF active site (Table 3; Fig. 5A). The absence of the His 303 side chain resulted in the inability to detect structured water molecules in the active site tunnel. This observation is consistent with the idea that His 303 is central to a hydrogen bond network that organizes the water molecules that fill the active site tunnel through a strong hydrogen bond interaction with W1. Also, the electron density of the Phe 396 side chain was weak (Fig. 5A). Analysis of the SpFabF[H303A] mutant using mass spectrometry confirmed that the Phe 396 residue was present; thus, the reduced electron density arose from Phe 396 adopting multiple conformations in the absence of the His 303 side chain (Fig. 5A). In the native structure (Fig. 1B), the fixed position of W1 allows only a single conformation for the Phe 396 side chain (Table 3).

DISCUSSION
Roles of His 303 , His 337 , and W1-A primary conclusion from our biochemical and structural analyses is that the elongation condensing enzymes use His 303 to activate a catalytic water molecule (Fig. 6). The active site of SpFabF (21) has a structured water molecule (W1) that is hydrogen-bonded to the N of His 303 (Fig. 1B), and we propose that His 303 activates W1 to promote a nucleophilic attack on the malonate carboxyl. Our examination of other elongation condensing enzyme structures (19,21,46) reveals that they contain water molecules in a very similar location to W1 that are hydrogen-bonded to the histidine residues equivalent to His 303 in SpFabF. In order for His 303 to activate the water molecule, the lone pair of electrons must reside on the N of the imidazole ring. The 1.3-Å crystal structure of SpFabF revealed that the N␦ of His 303 has significant additional electron density that we interpret as a hydrogen atom (21), and this places the lone pair on the N. In contrast, the N of His 337 is protonated, because the N␦ nitrogen receives a hydrogen bond from the backbone amide of Leu 339 and therefore contains the lone pair of electrons (Fig. 1B). These structural insights mesh with the activities of the SpFabF mutant panel (Table 1). His 303 is not absolutely essential for the condensation reaction in SpFabF but rather functions to accelerate the reaction rate, especially at lower pH. Wild-type SpFabF readily incorporates [ 14 C]bicarbonate into malonyl-ACP in the reverse reaction, consistent with bicarbonate as the product of the reaction. SpFabF[H303A] also performs the malonyl-ACP exchange reaction with [ 14 C]bicarbonate, albeit at a slower rate, and the presence of His 303 is therefore not essential to carry out either the forward or the reverse reaction. A role for His 303 as a base is supported by the biochemical and structural analysis of EcFabB [H298Q] and EcFabB[H298E] mutants in the decarboxylation half-reaction (47). However, the minor reduction of catalytic activity in the SpFabF[H303A] mutant argues against a major role for His 303 in the SpFabF condensing enzyme. In contrast, the severe impairment of catalytic activity following mutation of the analogous His 298 in EcFabB (40) indicates that His 298 is more critical for catalysis in this elongation condensing enzyme. It is possible that rather than acting as a base, His 303

Parameter (Unit) SpFabF͓H303A͔
Space group P2 1 2 1 2 Unit cell dimensions (Å) (a, b, c) 60. 4   functions primarily to position the water molecule for attack on the malonate carboxyl group. The differences in the effects of removing this histidine on the enzymatic activity of different elongation condensing enzymes, the fact that we do not know the rate-controlling step in this complex ping-pong reaction, and the fact that we do not have the malonyl-ACP enzyme complex structure preclude a definitive conclusion on the role of His 303 in the reaction mechanism. Our proposed mechanism is not consistent with the electronic states of the active site histidines deduced from the structures of condensing enzyme binary complexes with cerulenin and thiolactomycin (27,41). In both complexes, a carbonyl oxygen on the inhibitor receives hydrogen bonds from the N nitrogens of both His 298 and His 333 of EcFabB. The inhibitor carbonyl groups were initially interpreted as being equivalent to the thioester carbonyl group of malonate, but a close examination of the two binary complexes (27) reveals that the side chain of His 298 moves ϳ2 Å to make a hydrogen bond with the inhibitor carbonyl group and the N␦ nitrogen is thereby shifted to within 3 Å of the N nitrogen of Lys 328 . We propose that this repositioning results in the movement of the lone pair from the N nitrogen of His 298 to the N␦ nitrogen in the formation of the inhibitor complex and that the orientation and electronic states of the histidines in the drug complexes differ from those in the native protein.
Activation of Cys 164 -A key step in the transacylation reaction is the nucleophilic activation of the active site cysteine. One idea for how this occurs is that His 337 and/or His 303 facilitates the reaction by abstracting a proton from the sulfhydryl group (19,21). However, the electronic state of the N of His 337 is protonated due to the strong hydrogen bond between its N␦ with the backbone amide (Fig. 1), and none of the SpFabF active site mutants, except for SpFabF [C164A], are defective in the transfer of the primer acyl group from acyl-ACP to Cys 164 . Instead, the activation of the cysteine occurs through the dipole moment of helix N␣3, which is directed toward the sulfur atom and lowers the pK a of the cysteine sulfhydyl (21). The orientation of this helix dipole toward the active site sulfhydryl within this protein superfamily was noted in the first thiolase structure (48), and there is little doubt that it has a profound effect on the electrostatic environment of the condensing enzyme active site.
Role of Lys 332 -Lys 332 is clearly a critical component of a hydrogen bond network that includes Glu 346 and a structured water (W2) that interacts with His 303 (Fig. 1B). This network maintains the correct res-onance form of His 303 , which allows it to abstract a proton from the catalytic water, W1 (Fig. 6 (47). This structure shows that His 333 (His 337 in SpFabF) rotates into the space generated by the removal of Lys 328 (Lys 332 in SpFabF) (47) and has no effect on the position or orientation of His 298 (His 303 in SpFabF). Thus, the loss of Lys 332 has a dramatic effect on the spatial positioning and hence the function of His 337 .
Role of Phe 396 -Our studies are the first to assign a key function for Phe 396 . Phe 396 is flanked by conserved glycine residues and displays conformational variability in the available crystal structures. The structures of the FabB acyl-enzyme intermediate and the FabF-cerulenin complex (Fig. 5B) show that the aromatic ring rotates away from the active site and acts as a platform that directs the acyl chain into its binding pocket. The observation that TAL formation exceeds the forward condensation activity in the SpFabF[F396A] mutant suggests that Phe 396 acts as the "gatekeeper" that controls the order of the reaction. In the free enzyme, the Phe 396 side chain is closely packed into the active site, preventing the binding and decarboxylation of malonyl-ACP prior to the formation of the acyl-enzyme intermediate. After transacylation, the carbonyl of the thioester intermediate engages the oxyanion hole formed by the amide nitrogens of Cys 164 and Phe 396 , causing the aromatic ring of Phe 396 to rotate away from the active site, creating the necessary space to accommodate the incoming malonyl-ACP substrate. This movement is most clearly seen comparing the E. coli FabF structure with FabF with bound cerulenin (Fig. 5B). SpFabF[H303A] also exhibits enhanced TAL formation, and this is explained by the crystal structure, which reveals that the side chain of Phe 396 is disordered and creates the necessary space for the second substrate, malonyl-ACP, to bind prior to the formation of the acyl-enzyme intermediate.
A General Mechanism for Condensing Enzymes-Our analysis is consistent with the model for the catalytic mechanism of the elongation condensing enzymes shown in Fig. 6. The first step involves the binding of acyl-ACP and the transfer of the acyl group to the active site cysteine. The nucleophilicity of Cys 164 is enhanced by the strong N␣3-helix dipole effect, and the oxyanion hole formed by the backbone amides of Cys 164 and Phe 396 (hole 1) promotes the reaction by accommodating the negative charge on the thioester carbonyl that develops in the tetrahedral transition state. ACP is released, and malonyl-ACP binds to the acyl-enzyme intermediate. His 303 activates a structured catalytic water molecule to attack the carboxylate of the malonyl-ACP, and His 337 acts as a second oxyanion hole (hole 2) that promotes the formation of the carbanion at the C2 of the malonate by stabilizing the enol intermediate. The carbanion attacks the acyl-enzyme intermediate, and the tetrahedral transition state is again stabilized by the Cys 164 /Phe 396 oxyanion hole 1. The transition state resolves to release the ␤-ketoacyl-ACP product and bicarbonate. The condensation active sites of the FAS I multifunctional polypeptide and the polyketide synthases have the same Cys-His-His triad and the other key conserved residues as the FAS II elongation condensing enzymes (Fig. 1A), suggesting that the catalytic mechanisms of these condensation modules are the same as SpFabF. The role of a catalytic water in FAS I is established by the demonstration of bicarbonate as a reaction product (6).
Our model differs from that proposed by Olsen et al. (49), which suggested that the malonyl thioester engages the FabB active site with the carboxylic acid leaving group interacting with the N nitrogen of His 298 and two conserved threonines (300 and 302). This model is inconsistent with how the pantetheine arm must enter the active site tunnel, and the C2 of malonate is also too far away from the acylated cysteine (about 5 Å) for the condensation reaction to occur. In our model, the malonate binds in the opposite orientation, which brings C2 into closer proximity to the acyl-enzyme intermediate and positions the distal peptide bond of the prosthetic group in a conformation to possibly interact with the two conserved threonines that line the active site tunnel.
The use of a catalytic water probably also extends to the initiation class of condensing enzymes that possess a Cys-His-Asn catalytic triad. We have verified that bicarbonate is incorporated into the malonyl-ACP in the FabH back reaction (not shown), suggesting that bicarbonate is also the product of this reaction. Close inspection of the FabH crystal structures (26, 50 -53) reveals the presence of a catalytic water hydrogen-bonded to the active site histidine and in proximity to the cysteine sulfhydryl in an orientation very similar to that found in the elongation enzymes. This water molecule is interpreted by others as being responsible for the deprotonation of the active site sulfhydryl (53); however, our results suggest that the histidine and water may function in the same capacity as outlined in Fig. 6.
Comparison of Thiolase and Condensing Enzyme Mechanisms-The condensing enzymes belong to the thiolase superfamily, whose members catalyze the synthesis and degradation of carbon-carbon bonds. The synthetic thiolases deprotonate the ␣-carbon of acetyl-CoA, and the carbanion attacks the acetyl-enzyme intermediate to form the carbon-carbon bond. Although structurally similar, the condensing enzymes generate the same carbanion by decarboxylating a malonate thioester. There are four differences between the active sites of the two enzyme clades that reflect these distinct mechanisms. First, thiolase contains two cysteines, one that accepts the acetyl intermediate and a second that performs the deprotonation step. The deprotonating cysteine spatially corresponds to the location of the histidine-water grouping in the condensing enzymes that drives the decarboxylation step. Second, the pair of histidines in the condensing enzymes is replaced by a histidine-asparagine pair in the thiolases. The asparagine is associated with a water molecule and participates with the adjacent histidine in the formation of an oxyanion hole that stabilizes the carbanion. Third, it has been proposed that the N nitrogen of the thiolase histidine (equivalent to His 337 in SpFabF) activates the cysteine in the formation of the acetylenzyme intermediate (42). However, this is not possible in the condensing enzymes, because the lone pair of electrons is clearly located on the N␦ nitrogen (Fig. 1). We have also demonstrated that mutation of either conserved histidine does not significantly affect the transacylation step. Finally, structural studies of a number of thiolase complexes suggest that the histidine-water oxyanion hole is involved in both the transacylation and deprotonation steps. Kursula et al. (42) propose that the acetyl carbonyl oxygen engages the oxyanion hole during the transacylation step. The acetylated cysteine then rotates to engage a second oxyanion hole, making the first oxyanion hole available to participate in the condensation reaction. This is unlikely to occur in the condensing enzymes, because the long acyl chain is buried within its narrow binding pocket, preventing rotation of the acyl-enzyme intermediate.