Biochemical Evidence for an Editing Role of Thioesterase II in the Biosynthesis of the Polyketide Pikromycin*

The pikromycin biosynthetic gene cluster contains the pikAV gene encoding a type II thioesterase (TEII). TEII is not responsible for polyketide termination and cyclization, and its biosynthetic role has been unclear. During polyketide biosynthesis, extender units such as methylmalonyl acyl carrier protein (ACP) may prematurely decarboxylate to generate the corresponding acyl-ACP, which cannot be used as a substrate in the condensing reaction by the corresponding ketosynthase domain, rendering the polyketide synthase module inactive. It has been proposed that TEII may serve as an “editing” enzyme and reactivate these modules by removing acyl moieties attached to ACP domains. Using a purified recombinant TEII we have tested this hypothesis by using in vitro enzyme assays and a range of acyl-ACP, malonyl-ACP, and methylmalonyl-ACP substrates derived from either PikAIII or the loading didomain of DEBS1 (6-deoxyerythronolide B synthase; ATL-ACPL). The pikromycin TEII exhibited highK m values (>100 μm) with all substrates and no apparent ACP specificity, catalyzing cleavage of methylmalonyl-ACP from both ATL-ACPL(k cat/K m 3.3 ± 1.1m −1 s−1) and PikAIII (k cat/K m 2.9 ± 0.9m −1 s−1). The TEII exhibited some acyl-group specificity, catalyzing hydrolysis of propionyl (k cat/K m 15.8 ± 1.8m −1 s−1) and butyryl (k cat/K m 17.5 ± 2.1m −1 s−1) derivatives of ATL-ACPL faster than acetyl (k cat/K m 4.9 ± 0.7m −1 s−1), malonyl (k cat/K m 3.9 ± 0.5m −1 s−1), or methylmalonyl derivatives. PikAIV containing a TEI domain catalyzed cleavage of propionyl derivative of ATL-ACPL at a dramatically lower rate than TEII. These results provide the first unequivocal in vitro evidence that TEII can hydrolyze acyl-ACP thioesters and a model for the action of TEII in which the enzyme remains primarily dissociated from the polyketide synthase, preferentially removing aberrant acyl-ACP species with long half-lives. The lack of rigorous substrate specificity for TEII may explain the surprising observation that high level expression of the protein inStreptomyces venezuelae leads to significant (>50%) titer decreases.

Polyketides are a large and structurally diverse class of natural products that possess a wide range of biological activities (1). These compounds are used throughout medicinal and agricultural fields as antimicrobials, immunosuppressants, antiparasitics, and anticancer agents. Despite their structural diversity, polyketides are assembled by a common mechanism of decarboxylative condensations of simple malonate derivatives by polyketide synthases (PKSs) 1 in a manner very similar to fatty acid biosynthesis (2,3). Type I PKSs are a family of PKSs that are analogous to vertebrate fatty acid synthase that catalyze the biosynthesis of the polyketide moieties of various secondary metabolites in Streptomyces (4 -6). They are gigantic multifunctional modular proteins. Each module is responsible for one cycle of polyketide chain elongation and contains a set of discrete catalytic domains of ketosynthase (KS), acyltransferase (AT), and acyl carrier protein (ACP). Ketoreductase, dehydratase (DH), and enoyl reductase domains may also be present, allowing structural variation in the level of processing of the ␤-ketoacyl chain (4,6). The fully extended polyketide chain bound to the PKS as an acyl-ACP thioester is often released and cyclized by a thioesterase domain (TEI) covalently linked to the last extending module of the PKS (7). In many cases, additional genes encoding a TE have been found within a polyketide biosynthetic gene cluster, for example, the tylosin PKS of Streptomyces fradiae (8), pikromycin PKS of Streptomyces venezuelae (5), rifamycin PKS of Amycolatopsis mediterranei (9), and the erythromycin PKS (DEBS) of Saccharopolyspora erythraea (10). These genes encoding a discrete protein were named as TEII to differentiate from the chain releasing TEI domains in modular polyketide synthases. Discrete TEII enzymes are also associated with bacterial nonribosomal peptide synthases, responsible for the production of macrocyclic peptide compounds (11), and animal fatty acid synthases (12). Sequence analysis has revealed that these thioesterases are probably structurally and evolutionarily related (13). They have a common serine esterase motif, GXSXG, ϳ100 residues from the N terminus and another conserved downstream motif, GXH, of which histidine might accept a proton from the hydroxyl group of the active site serine, thereby increasing its nucleophile character (13). Because TEI domains have been shown to be a necessary and sufficient factor for the release and cyclization of the polyketide chain in vivo and in vitro (7, 14 -16), the role of the TEII enzyme encoded within many PKS gene clusters has presented an intriguing question. By far most of the information about the role of TEII in polyketide biosynthesis comes from gene disruption and complementation studies. When the rifR gene (encoding a TEII) was deleted from the rifamycin PKS, production of the polyketide rifamycin B dropped to 30 -60% of the normal yields in A. mediterranei (17). Disruption of the TEII-encoding gene (tylO) in the tylosin producer S. fradiae also resulted in 85% reduction of polyketide production (18). Production in the disrupted strain was restored by complementation with the intact tylO gene. More interestingly, it was reported that heterologous TEIIs (nbmB from S. narbonensis and scoT from Streptomyces coelicolor) could also successfully complement the inactivated TEII gene by restoring polyketide production (18,19). In contrast to these observations it has been reported that deletion of the TEII-encoding pikAV gene and srfA-TE gene does not lead to decreases in the production of the aglycones produced by the pikromycin PKS (20) and surfactin by the non-ribosomal peptide synthases (21), respectively. These observations have led to the suggestion that the TEII might play a beneficial but not essential role common to most if not all biosynthetic processes catalyzed by modular PKSs (19). One proposal has been an editing role in which TEII removes aberrant groups attached by thioester linkage ACP domains within the PKS extension modules, which might otherwise block the normal chain elongation process (18).
Recent studies on KS domains show they can catalyze a decarboxylation reaction of the ACP-bound dicarboxyl extender unit even when the KS active site cysteine is not primed with incoming acyl intermediates ( Fig. 1) (22,23). Polyketide chains grow by decarboxyl condensations between a ␤-ketoacyl chain bound to the KS domain and a carboxylated extender unit (malonyl-, methylmalonyl-, ethylmalonyl-ACP, etc.). If the extender unit is decarboxylated to a nonactivated acyl group (acetyl-, propionyl-, butyryl-ACP, etc.), the intermediate acyl chain cannot be processed to downstream domains through the normal condensation reaction. The polyketide biosynthetic process would thus be derailed, resulting in low yields of the fully extended polyketide product (24). TEII-catalyzed cleavage of these ACP-bound acyl residues would allow the activated dicarboxylic acid extender units to be loaded onto the PKS, restoring polyketide production. Consistent with such a role is the observation that the homologous TEII and TEI from an animal fatty acid synthases have substrate specificity for medium length fatty acid chains (C8, C10, and C12) and longer chain fatty acids (C14, C16, and C18), respectively (25). Recent studies on TEII (tylO) of the TylPKS show that it could hydrolyze synthetic acyl (acetyl-, propionyl-, and butyryl-)-p-nitrophenyl esters and N-acetylcysteamine thioesters, providing additional preliminary support for the proposed editing role of TEII (24).
In this study, we have generated recombinant TEII (PikAV) from S. venezuelae and performed the first catalytic study of such an enzyme with physiologically relevant substrates, various acyl-and carboxylated acyl-ACP thioesters. The pikromycin TEII exhibits a very high K m for all ACP substrates and likely is predominantly dissociated from the PKS, minimizing its potential to block the normal biosynthetic process. TEII is active with a wide range of ACP-bound thioesters, consistent with an editing role for all of the modules within a PKS. However, this lack of rigorous substrate specificity allows the TEII to remove acyl-ACP thioesters in the PKS-loading module required to initiate the process or the activated carboxylated acyl-ACP thioesters in the extension modules. Thus, either an excess or lack of TEII may lead to decreases in polyketide production. In vivo experiments with overexpression of pikAV in pikromycin-producing cultures of S. venezuelae are shown to be consistent with this observation. DL-2-[2 14 C]methylmalonyl-CoA (60 mCi/mmol, 10 Ci/ml) was purchased from American Radiolabeled Chemicals, Inc. All other chemicals were purchased from Sigma and were of the highest available grade. Standard molecular cloning techniques were employed (26). PCR amplifications were performed using standard PCR strategies with Pfu Turbo DNA polymerase (Stratagene) (27).

Materials and General
Protein Expression and Purification-TEII was obtained by expression of pikAV of S. venezuelae in Escherichia coli. The pikAV gene (GenBank TM accession number AF79138: 36989 -37834) was amplified from a genomic library of S. venezuelae by PCR using the following primers designed with NdeI and HindIII sites at the 5Ј and 3Ј ends, respectively (5Ј-AAGCGGGCATATGACCGACAGACCTCTGAA-3Ј and 5Ј-TCGTAAGCTTCCGTGGGTTCTGCCATCT-3Ј). The PCR product was ligated into the pET21c expression vector to give pBK18, which was maintained in E. coli TG2. E. coli BL21-CodonPlus-RP was used as a host for the expression of PikAV. Overexpression (2-4 mg/liter) was accomplished by induction using 0.8 mM isopropyl-1-thio-␤-D-galactopyranoside at A 600 ϭ 0.5-0.6 and incubation of the culture at 28°C overnight. Cells were harvested by centrifugation at 3,000 ϫ g and resuspended with buffer A (50 mM Tris (pH 8.0), 2.5 mM dithiothreitol (DTT), 10% glycerol). After disruption of the cells using French press at 10,000 p.s.i., the lysate was clarified by centrifugation at 10,000 ϫ g. The supernatant was loaded onto nickel nitrilotriacetic acid column previously equilibrated with 50 mM Tris (pH 8.0), and washed with 10 mM imidazole in the same buffer. The pikromycin TEII was eluted with buffer A containing 100 mM imidazole. Fractions containing the protein were pooled and concentrated with a Centricon TM (Amicon, Inc.) concentrator and subsequently diluted with buffer A containing 5 mM imidazole. The concentrated protein was further purified using a Hi-Trap chelating high performance column (5 ml, Amersham Biosciences) with a gradient elution from 5 to 250 mM imidazole in buffer A. The purification was monitored spectrophotometrically by absorbance at 280 nm. The pooled peak fractions were concentrated, and buffer was subsequently changed to buffer B (100 mM Na 2 HPO 4 (pH 7.2), 2.5 mM DTT, 1 mM EDTA, 20% glycerol). The concentration of the resulting pure TEII solution was determined using the dotMETRIC TM assay kit from Bioworld.
The last module of PikPKS, PikAIV, containing a TEI domain was expressed in E. coli. After several steps of PCR and subcloning, the pikAIV gene (GenBank TM accession number AF79138: 32952-36992) was cloned into pET24b as an NdeI and HindIII fragment to give pDHS4188, which expresses pikAIV as a C-terminal His-tagged fusion protein. The plasmid pDHS4188 was transferred into BL21(DE3) strain harboring the sfp gene (pRSG56) (28) and induced by 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside at 0.6 absorbance. After overnight culture at 25°C, the cells were harvested and disrupted using a French press at 10,000 p.s.i. The PikAIV-TEI was expressed at level of 8 -12 mg/liter and purified as described above for TEII. The protein (2.5 mg/ml) was kept in 100 mM NaH 2 PO 4 (pH 7.3), 1 mM EDTA, 0.2 mM DTT, 10% glycerol after dialysis in the same buffer. The activity of TEII and TEI was confirmed by hydrolysis assay using the p-nitrophenyl esters of acetate, propionate, and butyrate in sodium phosphate buffer (pH 7.5), 100 M EDTA as described previously (24). The substrates were dissolved into assay buffer using dimethyl sulfoxide.
To obtain the loading domain (AT L -ACP L ) of DEBS1 with a C-terminal hexahistidine tag, the corresponding gene (GenBank TM accession number 63677) was cloned as an XbaI-HindIII fragment into pET21c (Novagen) overexpression vector to give pBK12. The PCR amplification was from a DEBS1 construct (pBK3) (29) using the primers 5Ј-GAAT-TCGAGCATATGGACGCGTGGCGGACC-3Ј and 5Ј-TTCGTTAAGCTT-GCGGGTTTCCCGTTG-3Ј. For expression of the holo-AT L -ACP L didomain, pBK12 was introduced into E. coli BL21-CodonPlus-RP (Stratagene) containing the plasmid pRSG56, which carries a kanamycin gene and the sfp gene. BL21-CodonPlus-RP/pBK12 was cultured in LB supplemented with 50 mg/ml kanamycin and 100 mg/ml ampicillin at 37°C. When the growth reached A 600 ϭ 0.5-0.6, expression of the didomain (16-18 mg/liter) was induced by 0.8 mM isopropyl-1-thio-␤-Dgalactopyranoside followed by 12 h of incubation at 30°C. Cells were harvested by centrifugation at 3000 ϫ g and resuspended in 50 mM Tris (pH 8.0). The resuspended cells were lysed by a French press at 10,000 p.s.i. and pelleted by centrifugation at 10,000 ϫ g. The published purification procedure was then used to obtain 8.6 mg/ml concentration of holo-AT L -ACP L (30).
The pikAIII gene (GenBank TM accession number AF79138: 28161-32849) was cloned into pET28b as an NdeI-HindIII fragment after multi-step PCR reactions and subcloning steps. The resulting construct (pDHS4405) was used to express PikAIII (8 -12 mg/liter) as a fusion protein with an N-terminal His tag in E. coli. Expression and purification were analogous to that described for PikAIV.
Autoradiography Demonstrating Deacylation of Propionyl AT L -ACP L by TEII-To generate a radiolabeled acyl-AT L -ACP L , holo-AT L -ACP L (50 M) was incubated at room temperature in a reaction mixture containing 180 M [1-14 C]propionyl-CoA, 100 mM NaH 2 PO 4 (pH 7.2), 2.5 mM DTT, 1 mM EDTA, and 20% glycerol. After a 45-min incubation, the reaction mixture was concentrated, and the excess propionyl-CoA was removed by exchanging buffer using a Centriprep TM concentrator (Amicon). The radiolabeled propionyl-AT L -ACP L (20 M) was incubated with 5 M TEII in 50 mM NaH 2 PO 4 (pH 8.0) at room temperature for 6 h. A control incubation in the absence of TEII was carried out simultaneously. The reaction was quenched by the addition of 10% trichloroacetic acid and incubated for 30 min in ice. The proteins were pelleted and rinsed with acetone. The pellet was redissolved into 20 l of protein gel-loading buffer. The mixture was loaded onto a 1.5-mm 12% polyacrylamide gel and run at 70 V. When the marker dye reached the bottom of the gel, it was stained, destained, and dried on filter paper. The dried gel was subjected to PhosphorImager analysis (Molecular Dynamics) for 48 h.
Enzyme Assays-The enzyme activity of TEII was measured on several radiolabeled acyl-ACP substrates. Typical reactions to make radiolabeled acyl-ACPs contained 50 M either AT L -ACP L or PikAIII in a mixture of 1:9 or 1:5 radiolabeled and nonradiolabeled acyl-CoA, malonyl-CoA, or methylmalonyl-CoA (1 mM), 100 mM NaH 2 PO 4 (pH 7.2), 2.5 mM DTT, 1 mM EDTA, and 20% glycerol. After a 1-3-h incubation, the reaction mixture was concentrated, and the buffer was exchanged to 50 mM phosphate buffer (pH 8.0) using a Centriprep TM concentrator. A time course experiment was performed using a reaction mixture (25 l) Triplicate analyses were carried out for each concentration. The reactions were quenched by 100 l of 10% trichloroacetic acid, and the proteins were pelleted by centrifugation at 10,000 ϫ g. Each pellet was rinsed with an excess of 10% trichloroacetic acid and dissolved into 200 l of SDS solution (2% SDS, 20 mM NaOH). The protein suspension was mixed with 5 ml of liquid scintillation fluid, and the radioactivity was quantified by liquid scintillation counting. The loss of radioactivity compared with the control reaction without TEII/PikAIV-TEI was used to calculate the amount of hydrolyzed acyl chain from the ACP domain.
Effect of High Level Expression of TEII on Aglycone Production in S. venezuelae-The TEII gene (pikAV) was PCR-amplified as a XbaI and HindIII fragment and cloned into a high copy plasmid pSE34, a derivative of pWHM3 with the ermE* promoter (31), to give pSC70. The pikAV gene was also cloned into a derivative of pDHS702 (16) (a low copy number plasmid based on SCP2* origin of replication) for expression of TEII under the control of the pikAI promoter. Expression of the TEII from the resulting pSC38 plasmid resembles a natural expression pattern of pikA cluster in S. venezuelae. A sitedirected mutation was introduced to pSC70 to produce an expression plasmid of inactive TEII protein (pBK56). Ser-99 was changed to Ala using QuikChange TM XL mutagenesis kit (Stratagene) and primers 5Ј-CGGGCACGCCCTCGGCGCTAGCGTCGCCTTCGAGACG-3Ј and 5Ј-CGTCTCGAAGGCGACGCTAGCGCCGAGGGCGTGCCCG-3Ј (the alanine codon is underlined). The TEII expression plasmids are shuttle vectors having a ColE E. coli origin of replication. They were introduced to S. venezuelae SC1022 strain, which has an in-frame deletion of 687 bp of pikAV (32). Each of the transformants was cultured in 50 ml of SCM medium (20) supplemented with thiostrepton and kanamycin at 25 g/ml in a 250 ml flask. After 3 days of incubation at 30°C, the culture broth was clarified by centrifugation at 4000 ϫ g, and the supernatant was extracted with chloroform twice. The aglycone production was quantified by HPLC using a C18 column with monitoring at 230 nm as described previously (20).

Generation of Recombinant TEII and Acyl-ACP Substrates-
The TEII protein encoded by pikAIV was expressed from plasmid pBK18 in E. coli as a fusion protein tagged with a Cterminal hexahistidine and purified to homogeneity using Ni 2ϩ affinity column chromatography (Fig. 2). The recombinant protein was stable in 50 mM phosphate buffer with 1 mM EDTA during the purification procedures. Overnight dialysis or high salt concentration buffer, however, resulted in partial precipitation of the TEII. The esterase activity of the TEII was confirmed by hydrolysis assay using p-nitrophenyl ester derivatives (data not shown), which are well known substrates for serine proteases and have been used for thioesterase studies in the animal fatty acid synthase (33) and the TEII associated with the tylosin PKS (24). The TEII showed significant hydrolytic rates with acetyl, propionyl, and butyryl esters. However, the substrate specificity of TEII with these model systems could not readily be addressed due to high background hydrolysis rates and limited solubility of the substrates under the assay conditions. Similar observations have been made with these substrates using the TEII (TylO) encoded within the tylosin biosynthetic gene cluster (24).
To further probe the role of TEII in polyketide biosynthesis, we examined catalytic activity with a more physiologically relevant substrate, a PKS ACP thioester of monocarboxylic acid, as might be generated by aberrant decarboxylation of malonate or malonate derivatives bound to ACP domain. We selected the loading domain of DEBS (AT L -ACP L ) because this is known to have relaxed substrate specificity accepting acetyl-, propionyl-, butyryl-, malonyl-, and methylmalonyl-CoA (30), allowing a clear comparison of TEII hydrolysis of thioesters of these groups on the same ACP.
The gene encoding the AT L -ACP L didomain region of DEBS1 was PCR amplified as an NdeI and HindIII fragment and cloned into an expression vector pET21c to give pBK12. The didomain fragment includes the N-terminal 107 extra amino acids that have been shown recently to be essential for activity of this didomain (30). The gene encoding the Sfp (a phosphopantetheine transferase from Bacillus subtilis) was co-expressed to ensure posttranslational modification of the apo-AT L -ACP L to the holo form (28). The holo-AT L -ACP L was expressed as a C-terminal hexahistidine-tagged fusion protein and purified by nickel affinity chromatography to 99% purity.
The AT L -ACP L was incubated with [1-14 C]propionyl-CoA, leading to rapid acylation. Other acyl-AT L -ACP L derivatives were also generated in the same reaction mixture using different radiolabeled CoA derivatives (acetyl-, butyryl-, malonyl-, and methylmalonyl-CoA). Prolonged incubations (up to 3 h) ensured that greater than 95% of the protein was converted to the thioester, even with substrates such as malonyl-CoA.
Thioesterase Activity of TEII on Propionyl-AT L -ACP L -The radiolabeled propionyl-AT L -ACP L was incubated at room temperature with or without TEII for 6 h. After trichloroacetic acid precipitation, the reaction mixture was analyzed by SDS-PAGE and autoradiography. The reaction without TEII showed a clear radioactive protein band at the position of AT L -ACP L protein, whereas the corresponding band was barely detected in the reaction containing TEII protein (Fig. 3), indicating hydrolysis of the radioactive propionyl residue from the protein. A kinetic analysis of this phenomenon using liquid scintillation counting showed that the loss of radioactivity of propionyl-AT L -ACP L was proportionally correlated with the TEII concentration, indicating the deacylation is the result of enzyme reaction by TEII. These results provide the first clear evidence of thioesterase activity of TEII on the acyl thioester of a PKS ACP domain, as might be generated by aberrant decarboxylation of extending units during polyketide biosynthesis.
A comparison of the thioesterase activity of PikAV TEII was made with that of TEI at the C terminus of PikAIV. The pikromycin PikAIV-TEI was overexpressed in E. coli and purified as a C-terminal His-tag fusion protein (Fig. 2B). Numerous in vivo and in vitro studies have previously shown that an equivalent TEI domain at the end of the erythromycin PKS can catalyze hydrolysis and cyclization reactions with shorter chain acyl-ACPs than the natural length of acyl-ACP (34,35). A time course experiment comparing catalytic activity of TEII and the PikAIV-TEI multifunctional protein revealed that the propionyl group was removed from the ACP domain more rapidly by the TEII (Fig. 4A). The concentration dependence of the propionyl-AT L -ACP L hydrolysis by TEII and PikAIV-TEI (Fig. 4B) was determined. TEII removed propionyl groups from the ACP domain about 8 times more efficiently than the TEI.
Substrate Specificity of TEII-More quantitative informa- tion concerning the substrate specificity of TEII was obtained by calculation of k cat /K m for several acyl-ACP substrates such as propionyl-, acetyl-and butyryl-AT L -ACP L (potentially generated by decarboxylation of the malonate-derived extender units such as methyl malonate, malonate, or ethyl malonate bound to an ACP domain). All three acyl-ACP substrates tested were hydrolyzed by the TEII at a significant rate (Fig.  5). In all cases the high K m for the substrates (all reaction rates exhibited a linear dependence on substrate concentration to 150 M) coupled with their limited solubility precluded us from obtaining true saturation kinetics. Instead, the apparent k cat /K m was used for the comparison of enzyme efficiency and specificity and calculated on the assumption that at substrate concentrations significantly below the K m most of the enzyme is free ([E] t is approximately [E]). Under these circumstance, the ratio k cat /K m behaves as a second-order rate constant for the reaction between substrate and free enzyme (k cat /K m ϭ V/[E] [S]). The catalytic efficiency for butyryl-and propionyl-ACP was 17.5 Ϯ 2.1 and 15.8 Ϯ 1.8 M Ϫ1 s Ϫ1 , respectively. The acetyl-ACP was a poorer substrate under these conditions (4.9 Ϯ 0.7 M Ϫ1 s Ϫ1 ). Carboxylated acyl substrates were also examined for the first time with a TEII. Because these are the correct substrate for PKS elongation modules, we expected to see significant substrate discrimination by TEII. The malonyl and methylmalonyl derivatives of the AT L -ACP L (3.9 Ϯ 0.5 and 3.3 Ϯ 1.1 M Ϫ1 s Ϫ1 , respectively) were hydrolyzed less efficiently than the propionyl and butyryl derivatives but were substrates nonetheless.
ACP Specificity of TEII-The use of the DEBS PKS AT L -ACP L -loading domain and its established broad substrate specificity allowed comparison of the relative activities of pikromycin TEII of a range of acyl and carboxylated substrates bound to the same ACP. To investigate the ACP specificity of the pikromycin TEII, the hydrolysis rate of the methylmalonyl thioesters from DEBS AT L -ACP L -loading domain and PikAIII (a single modular protein of PikPKSs and, thus, a potential physiological substrate for TEII) were compared (Fig. 6). The TEII showed significant and indistinguishable k cat /K m (3.3 Ϯ 1.1 and 2.9 Ϯ 0.9 M Ϫ1 s Ϫ1 for methylmalonyl thioesters of AT L -ACP L and PikAIII, respectively).

Effect of TEII Expression on Aglycone Production in Pikromycin PKS-
We carried out a complementation study of TEII in S. venezuelae strain SC1022, which has an in-frame deletion of pikAV. In previous studies, this strain has been shown to produce only aglycones, as the deleted region of pikAV contains a transcription unit essential for expression of the des genes responsible for the production of the final glycosylated products (32). The titer of the polyketide products was apparently not affected by deletion of the TEII gene (yields were comparable with that observed for a desVI mutant) (20). Complementation of strain SC1022 with the TEII gene expressed from the pikAI promoter in a low copy number (pSC38) plasmid had no affect upon aglycone levels (10-deoxymethynolide and narbonolide levels of 9 Ϯ 2 and 49 Ϯ 4 mg/liter), comparable with that observed with SC1022. Complementation of SC1022 with pSC70 (a derivative of the high copy number plasmid pSE34, which expresses TEII constitutively from the ermE* promoter) led to a consistent and dramatic 50 -70% decrease in aglycone production (4 Ϯ 3 mg/liter 10-deoxymethynolide and 12 Ϯ 8 mg/liter narbonolide, respectively) (Fig. 7). No suppression of aglycone production was observed with SC1022 carrying either pBK56, a pSE34 derivative expressing the S99A TEII mutant, or pSE34. DISCUSSION KS enzymes catalyze the elongation step in the biosynthesis of fatty acids and polyketides in a multi-step process involving decarboxylation of the elongation unit (typically malonyl-or methylmalonyl-ACP) and attack of the resulting carbanion on an acyl group covalently bound to the active site cysteine residue (36). Evidence indicates that the decarboxylation reaction is most efficient when its active site cysteine is thus modified, thereby minimizing nonproductive decarboxylation of the activated extender units (36). Nonetheless, studies have shown that decarboxylation of malonyl-ACP without prior acylation of the active site, so-called aberrant decarboxylation, could take place, albeit at a lower rate. The KS domain of the yeast fatty acid synthase and the KS components of type II fatty acid synthase system can catalyze malonyl-ACP decarboxylation in the absence of an acyl-enzyme intermediate (37,38). Examples of extender unit decarboxylation in the absence of the acylated KS have also been seen in modular PKSs. Under normal conditions, DEBS3 from S. erythraea can catalyze the last two extending reactions in 6-deoxyerythronolide B biosynthesis us- ing methylmalonyl-CoA as an extending unit. In a cell-free extract, DEBS3 is able to synthesize triketide lactone (without a loading domain) from just methylmalonyl-CoA, indicating that the KS domain was primed with a propionyl starting unit through an aberrant decarboxylation of methylmalonyl-ACP (39). Similarly, decarboxylation of methylmalonyl-ACP within the first extension module may explain why a site-specific mutant of the loading domain ACP in DEBS1-TE (resulting in the absence of 4Ј-phosphopantetheine prosthetic group) can still produce 2-5% of wild type levels of triketide lactone (22). These results suggest that as a result of this activity of the KS domain, a low rate of aberrant decarboxylation is inevitable during polyketide synthesis. Although such aberrant decarboxylation may occur at low frequency, it inevitably may lead over a multi-day fermentation to significant decreased efficiency of a PKS (24). Hydrolysis of the acyl-ACP groups by a thioesterase (TEII) would restore the PKS efficiency and maximize polyketide production.
For such a role TEII would need to be an individual protein (not covalent linked to module) able to hydrolyze short-chain acyl-ACP thioesters. Previous studies on the tylosin TEII have demonstrated using N-acetylcysteamine thioesters a preference for short-chain acyl groups over longer chain acyl groups or diketide derivatives. In this study we have demonstrated for the first time the ability of a PKS TEII to hydrolyze acyl groups attached to PKS ACP domains. The observation that the TEII enzyme catalyzed propionyl-ACP hydrolysis for DEBS1 AT L -ACP L significantly faster than PikAIV containing TEI is also consistent with the differing roles of the enzymes. The PikAIV-TEI is covalently linked to the last extension module of the PKS and is, thus, presented with substrate for efficient cyclization via an intramolecular process. In addition, recent crystallographic studies of the homologous TEI domain from the erythromycin PKS have revealed how these domains are ideally designed for catalyzing the cyclization process with longer polyketide chains rather than thioester hydrolysis (40). Indeed, the TEI domain attached to DEBS1-TE catalyzes cyclization reactions several orders of magnitude faster (k cat of Ͼ5 min Ϫ1 ) than observed for intermolecular acyl-ACP cleavage by Pi-kAIV-TE in the current study (7).
As an editing enzyme, TEII would need to scan all of the ACP domains within a given PKS and would not be expected to exhibit significant ACP specificity, accounting for the observation that methyl malonate units were removed from ACP domains of DEBS loading domain and PikAIII at nearly identical rates. Such low ACP specificity for TEII may also explain why heterologous expression of the TEII gene (nbmB) from S. narbonensis in S. fradiae increased tylactone production in a TEII (tylO) mutant from 15% to that observed in the wild type strain (18). A similar observation involving complementation of the tylO mutant strain with a TEII (scoT) obtained from a putative type I modular PKS gene cluster of S. coelicolor A3 (2) has also been reported (15,19). In fact it appears that the ACP group may not contribute significantly to substrate binding and catalysis with TEII. The hydrolysis rates reported for TylO with propionyl-and butyryl-N-acetylcysteamine thioesters (12.9 and 6.5 M Ϫ1 s Ϫ1 , respectively) (24) are comparable with our observations of the PikAV (TEII) with the corresponding thioesters of the DEBS AT L -ACP L loading domain (17.5 Ϯ 2.1 and Although TEII is not predicted to have ACP specificity, it might be expected to hydrolyze acyl-ACP units formed by aberrant decarboxylation more efficiently than with the corresponding carboxylated acyl-ACP substrates (required for the extension steps in polyketide biosynthesis). Consistent with this prediction, the pikromycin TEII hydrolyzed propionyl-and butyryl-ACP at a rate higher than for malonyl-and methylmalonyl-ACP derivatives. At odds with the prediction was the observation that acetyl-ACP was cleaved at a rate comparable with malonyl-ACP (the correct substrate for the PikAII-catalyzed second extension step in pikromycin biosynthesis) and significantly slower than butyryl-ACP (the ethylmalonyl-ACP, which would decarboxylate to generate this substrate, is not used in pikromycin biosynthesis). Preferential hydrolysis of propionate-and butyrate-p-nitrophenol esters (data not shown) was also observed with the TEII, suggesting this observation was not an artifact associated with use of the DEBS AT L -ACP L . In a previous observation, the tylosin TEII was shown to prefer the hydrolysis of an N-acetyl cysteamine thioesters over that observed for either propionate and acetate thioesters despite tylosin being generated using methylmalonyl-, ethylmalonyl-, and malonyl-ACP extender units (24). Thus, there are some small differences in the acyl group specificities of the tylosin and pikromycin TEII, which do not readily correlate with the substrate specificities of the extension modules within the two corresponding PKSs.
The pikromycin TEII exhibited a very high K m (Ͼ100 M) with all thioesters of PikAIII and DEBS AT L -ACP L . High K m values have previously been reported for the tylosin TEII using propionyl (37.9 mM)-and butyryl (28.2 mM)-N-acetyl cysteamine thioesters (24). Because concentrations of each of the PKS polypeptides within a cell are unlikely to even approach 100 M, the pikromycin TEII would appear to operate under nonsaturating conditions. The pikromycin TEII would, thus, be predominantly dissociated from the PKS and, as such, more able to scan all of the ACP domains within a PKS. A greater capacity of TEII to bind ACP would presumably give rise to a higher affinity for acyl-ACP and a more efficient enzyme. However, a concomitant increase in affinity for the carboxylated acyl-ACP derivatives or even holo-ACP domains would be predicted. Not only might this interfere with an ability to scan all of the PKS ACP domains, but a tightly bound TEII could also interfere with the normal processing of the PKS. A low affinity for the PKS and a relatively slow rate of reaction may thus FIG. 7. Effect of TEII overexpression on polyketide production in S. venezuelae. The aglycones (10-deoxymethynolide and narbonolide) produced by pSC38/SC1022 (solid line) and pSC70/SC1022 (dashed line) were quantified by HPLC analysis. pSC70 is a high copy number TEII (PikAV) expression plasmid derived from pWHM3 containing the ermE* promoter (pSE34). The pSC38 plasmid derived from pDHS702 is based on SCP2* origin of replication with TEII expression under control of the pikAI promoter, which allows regulated expression in S. venezuelae. Fermentations of SC1022 carrying pBK56 (expressing an inactive TEII), pSE34 (an empty plasmid control), or pSC38 were indistinguishable. The fermentation and analysis conditions were as described under "Experimental Procedures." ensure that TEII only becomes involved once an aberrant decarboxylation has occurred and the biosynthetic process has stalled (providing acyl-ACP substrates with significant halflives). As such, carboxylated acyl-ACP units, although substrates for TEII, should be processed readily by the PKS and not readily hydrolyzed in vivo. The k cat of 5 Ϯ 1 min Ϫ1 for triketide production by DEBS1-TE (7) is much faster than the observed rate of 0.15 min Ϫ1 for propionyl-ACP (150 M) hydrolysis observed for TEII under nonsaturating conditions. The editing function of TEII might, thus, be dependent not only upon its substrate specificity but also a low affinity for ACP thioester substrates and a significant difference in the reaction it catalyzes with that of the normal functioning PKS.
An editing role for TEII is consistent with previous observations that deletion of this gene generally leads to loss of production of polyketide or polypeptide. This is not the case for the pikromycin PKS, where deletion of the gene in strain SC1022 has previously been reported to have no affect upon polyketide yields. No increase in aglycone production in SC1022 was observed with the TEII complementation plasmid, consistent with this previous observation. The reasons why the TEII encoded by pikAV are not required for maximal polyketide production in SC1022 are unclear but may be due to the presence of another enzyme within S. venezuelae with similar activity. Such a possibility is quite reasonable given that putative TEII genes are often associated with modular PKS gene clusters, and many streptomycetes (41) including S. venezuelae contain several distinct clusters.
An editing role for TEII suggests that manipulation of its expression level or properties might have significant benefits for commercially important fermentation processes (24). It has recently been shown that expression of the TEII for S. erythraea leads to an 80% increase in production of the erythromycin aglycone, 6-deoxyerythronolide B, in cultures of recombinant E. coli (42). It has also been suggested that in genetically engineered PKSs, aberrant decarboxylation may be more prevalent, and TEII may help boost polyketide production (although levels of polyketides produced by hybrid pikromycin PKSs in S. venezuelae are very low despite the presence of the pikromycin TEII and possibly an additional TEII (29,43)). However, the observation that TEII removes an acyl group from a loading domain ACP (the DEBS AT L -ACP L ) and indications that it is unable to rigorously differentiate between an acyl-ACP substrate and its carboxylated derivative on an extension module suggest high level expression of TEII could suppress, not enhance, polyketide production in vivo. The decreased yields of aglycone production with plasmid-based high level expression of the pikromycin TEII in strain SC1022 are consistent with this prediction. The exact reasons for this decrease are unknown. However, the observation that high level expression of an inactive TEII does not suppress titers suggests that it is a result of the catalytic activity of TEII rather than simply binding to the PKS. Thus, it would appear that in addition to substrate specificity and kinetic properties, control of the timing and level of TEII expression are important for the enzyme to function appropriately as an editing enzyme. Within this context we note that recent S1 nuclease mapping studies of the TEII (scoT) promoter region in S. coelicolor have shown that transcription is dependent on growth phase and is only detectable during the late transition period. A regulatory system presumably prevents expression of the gene throughout most of the growth cycle and coordinates expression with the associated modular PKS gene cluster (19). Indeed, expression profiling of the entire pik gene cluster will determine the coordinated regulation of all pikA biosynthetic genes, including pikAV. Note Added in Proof-An elegant biochemical characterization of two TEIIs associated with nonribosomal peptide synthases was published (44) after acceptance of this manuscript.