A new substrate specificity for acyl transferase domains of the ascomycin polyketide synthase in Streptomyces hygroscopicus.

Ascomycin (FK520) is a structurally complex macrolide with immunosuppressant activity produced by Streptomyces hygroscopicus. The biosynthetic origin of C12-C15 and the two methoxy groups at C13 and C15 has been unclear. It was previously shown that acetate is not incorporated into C12-C15 of the macrolactone ring. Here, the acyl transferase (AT) of domain 8 in the ascomycin polyketide synthase was replaced with heterologous ATs by double homologous recombination. When AT8 was replaced with methylmalonyl-CoA-specific AT domains, the strains produced 13-methyl-13-desmethoxyascomycin, whereas when AT8 was replaced with a malonyl-specific domain, the strains produced 13-desmethoxyascomycin. These data show that ascomycin AT8 does not use malonyl- or methylmalonyl-CoA as a substrate in its native context. Therefore, AT8 must be specific for a substrate bearing oxygen on the alpha carbon. Feeding experiments showed that [(13)C]glycerol is incorporated into C12-C15 of ascomycin, indicating that both modules 7 and 8 of the polyketide synthase use an extender unit that can be derived from glycerol. When AT6 of the 6-deoxyerythronolide B synthase gene was replaced with ascomycin AT8 and the engineered gene was expressed in Streptomyces lividans, the strain produced 6-deoxyerythronolide B and 2-demethyl-6-deoxyerythronolide B. Therefore, although neither malonyl-CoA nor methylmalonyl-CoA is a substrate for ascomycin AT8 in its native context, both are substrates in the foreign context of the 6-deoxyerythronolide B synthase. Thus, we have demonstrated a new specificity for an AT domain in the ascomycin polyketide synthase and present evidence that specificity can be affected by context.

The macrolactone precursor of ascomycin is biosynthesized by a large PKS complex consisting of a loading module for a shikimate-derived starter unit; 10 modules for malonyl, methylmalonyl, or other PKS extender units; and a peptide synthetase module for addition of pipecolate (8 -13). FK520 and FK506 have methoxy groups at C13 and C15, which could be derived by post-PKS hydroxylation followed by O-methylation or by direct incorporation of an extender unit with an oxygen at the ␣ carbon. Feeding of [ 13 C 2 ]acetate was reported to label C8-C9 and C20-C23 of the macrolactone ring, as expected, but not C12-C15 in either FK520 or FK506 (8). A [1-13 C]erythrose feed, used to establish that the dihydroxycyclohexane starter unit is derived from shikimate, unexpectedly labeled C12 and C14, implying that the extender unit is more readily derived from erythrose than from acetate.
Sequence analysis of the ascomycin gene cluster from Streptomyces hygroscopicus (14) showed that it contains close homologues of the genes in the FK506 cluster (9 -11) that encode the three PKS subunits, the peptide synthetase, the lysine cyclodeaminase, the C9 hydroxylase, the 31-O-methyltransferase, and the putative 9-hydroxyl oxidase. In the flanking regions of the FK520 cluster, additional genes were found with proposed roles in synthesis of precursors, including a putative methoxymalonyl-ACP precursor for the 13-and 15-methoxy groups of FK520 (14).
The choice of extender unit by a modular PKS system is determined by the acyl transferase (AT) domain of each module (6,(15)(16)(17). Comparison of methylmalonyl-CoA-and malonyl-CoA-specific AT domain sequences shows that they cluster into two groups with sequence motifs that diverge according to the specificity of the domain (15). More recently, AT domains specific for ethylmalonyl units have been identified, and they are most closely related to the methylmalonyl-specific domains (18). In addition, there is circumstantial evidence that some domains are specific for an extender unit with oxygen on the ␣ carbon (19 -21). Despite the availability of the Escherichia coli malonyl-CoA:ACP transacylase crystal structure (22) and extensive biochemical experiments with both the rat fatty acid synthase and several bacterial PKSs (23)(24)(25)(26)(27), the structural basis for AT domain substrate specificity has not been established.
Here we describe replacement of the ascomycin AT8 domain with AT domains specific for malonyl or methylmalonyl ex-tender units, resulting in the production of analogues modified at C13 and showing that ascomycin AT8 is specific for an extender unit bearing oxygen on the ␣ carbon. Precursor labeling experiments showed that ascomycin PKS modules 7 and 8 use an extender unit that can be derived from glycerol. In addition, we show that the specificity of this AT domain can be altered by placing it into a foreign PKS context.

EXPERIMENTAL PROCEDURES
Strains and Media-The ascomycin-producing strain, S. hygroscopicus ATCC 14891, was plated on SY medium (10 g of soluble starch, 2 g of yeast extract, and 20 g of agar per liter) to prepare spore stocks and grown in TSBGM (tryptic soy broth supplemented with 50 mM TES buffer, pH 7, and 1% glucose) for production of ascomycin or its analogues. Growth and transformation of Streptomyces lividans K4-114 has been described previously (28). Use of the phage vector KC515 followed established procedures (28).
Preparation of Phage Carrying AT Replacement Cassettes-PKS module 8 was isolated as a 4.6-kb SphI fragment from cosmid pKOS65-C31 (14) and cloned into Litmus28 (New England Biolabs), and the orientation with a unique SacI site proximal to the SpeI site of the polylinker was used (pKOS60-21). A synthetic linker was ligated between the SpeI and SacI sites and then subsequently ligated between the SphI and AflII sites to give pKOS60-29, which was used as a template to isolate regions flanking AT8 by PCR. The PCR mixtures for all reactions described herein contained (50 l, total volume) 1 l of template (either diluted plasmid DNA, undiluted genomic DNA, or high-titer phage stock), 10ϫ Pfu buffer, 10ϫ z-deoxynucleotide triphosphate mix (200 M each deoxynucleotide triphosphate except 100 M dGTP and 100 M 7-deaza-dGTP; Roche Molecular Biochemicals), 10% Me 2 SO, 400 nM of each primer, and 1 l of cloned Pfu polymerase (Stratagene). Reactions were cycled 25-30 times at 95°C for 30 s, 60°C for 30 s, and 72°C for 2 min. To amplify the 5Ј-flanking region, the forward primer was 5Ј-CGACTCACTAGTGGGCAGATCTGGC-3Ј, and one of two reverse primers was used: 1) 5Ј-CACGCCTAGGCCGGTCG-GTCTCGGGCCAC-3Ј, which introduced an AvrII site near the 3Ј end of the ketoacyl synthase domain, or 2) 5Ј-GCGGCTAGCTGCTCGC-CCATCGCGGGATGC-3Ј, which introduced an NheI site near the 5Ј end of the AT boundary. These PCR products were cloned as SpeI to AvrII and SpeI to NheI fragments into Litmus28 and Litmus38, respectively, to give pKOS60-37-4 and pKOS60-37-2. The 3Ј-flanking region was isolated using a forward primer (5Ј-GATGTACAGCTCGAGTCG-GCACGCCCGGCCGCATC-3Ј) that introduced an XhoI site at the AT/ dehydratase boundary plus a BsrGI site to facilitate the construction; the reverse primer was 5Ј-CGACTCACTTAAGCCATGCATCC-3Ј. This PCR product was isolated as a BsrGI to AflII fragment and ligated into pKOS60-37-4 cut with Acc65I and AflII or pKOS60-37-2 cut with BsrGI and AflII, giving pKOS60-39-13 and pKOSS60-39-1, respectively.
Heterologous AT domains were obtained by PCR using primers that introduced AvrII and XhoI or NheI and XhoI sites at the 5Ј and 3Ј ends, respectively. Each heterologous AT was cloned into pKOS60-39-13 (AvrII-XhoI) or pKOS60-39-1 (NheI-XhoI) to create a series of AT replacement cassettes. Each replacement cassette was isolated as a BglII to NsiI fragment and ligated to KC515 DNA cut with BamHI and PstI, and the ligation mixture was introduced into S. lividans TK24 by transfection. The resulting plaques were purified, and high-titer phage stocks were prepared as described previously (28). Recombinant phage preparations were checked by PCR using a pair of primers annealing within the dehydratase/keto reductase-flanking sequence.
Construction of Strains with AT8 Replaced by Heterologous ATs-Spores of S. hygroscopicus (10 7 ) were suspended in Difco nutrient broth, heat-shocked at 50°C for 10 min, and mixed with an equal number of plaque-forming units of recombinant phage in Difco nutrient broth. The mixture was spread on R2YE plates (28), and after an overnight incubation at 30°C, plates were overlaid with 1.25 mg of thiostrepton suspended in 1 ml of H 2 O. After 7-10 days, sporulating colonies were streaked on SY agar containing 50 g/ml thiostrepton. Colonies were picked from these plates after about 5 days and macerated in 200 l of H 2 O using a microfuge pestle. Half the suspension was inoculated into a 25 ϫ 150-mm tube with two 4-mm glass beads and 5 ml of TSBGM; the other half was inoculated into the same medium with 50 g/ml thiostrepton. After 2-3 days, a 1.5-ml sample of each thiostreptoncontaining culture was harvested for total DNA isolation, and a 0.75-ml sample was prepared for LC-MS analysis by adding an equal volume of MeCN and clarifying by centrifugation. DNA was isolated by the SDS, proteinase K, phenol method (28) and analyzed by Southern blot hybridization using the digoxigenin labeling and detection kits supplied by Roche Molecular Biochemicals. Each culture grown without thiostrepton was spread on an SY agar plate to obtain spores to screen for the second recombination event.
Selected spore preparations from the nonselective propagation plates above were titered and spread on several plates at about 100 -200 colonies/plate. When colonies were sufficiently sporulated, they were replica-plated onto SY agar plates containing 50 g/ml thiostrepton. After 4 -5 days, putative thio S recombinants were identified and clonally isolated. These were grown in 5-ml TSBGM tube cultures for 3 days, and the broth was extracted as described above for LC-MS analysis (see below). Strains producing 13-DMA or 13-MDMA were designated KOS45-170 and KOS60-135, respectively.
Precursor Labeling Studies-[ 13 C 2 ]Glycine and [ 13 C 3 ]glycerol were obtained from Cambridge Isotope Laboratories. [ 13 C 2 ]Glycolate, a mixture of ethyl [ 13 C 2 ]bromoacetate (Aldrich; 98% 13 C; 1.0 g), sodium acetate (1.0 g), and anhydrous N,N-dimethylformamide (15 ml), was stirred at 80°C for 3 h. Water was added, and the mixture was extracted three times with Et 2 O. The combined organic layers were washed twice with H 2 O, washed once with brine, and dried over Na 2 SO 4 . After removal of solvent, a clear oil (0.45 g) was obtained. 1 H NMR (CDCl 3 ): ␦ 4.59 (2H, dd, J ϭ 149, 4.8 Hz), 4.23 (2H, qd, J ϭ 7.2, 3.2 Hz), 2.16 (3H, s), 1.29 (3H, t, J ϭ 7.2 Hz). The acetate was dissolved in MeOH (15 ml), 0.5 M LiOH (15 ml) was added, and the mixture was stirred overnight at room temperature. After neutralization with 2 N HCl, the mixture was extracted twice with ether. The aqueous layer was dried under high vacuum to give [ 13 C 2 ]lithium glycolate (508 mg; containing some sodium acetate and NaCl). 1  [methyl-13 C]Methoxyacetic acid:ethyl diazoacetate (1.8 ml) was added dropwise to a stirred solution of Rh 2 (OAc) 4 (10 mg) in [ 13 C]methanol (Isotec; 99 atom% 13 C; 1.0 g). After 4 h, the mixture was diluted with 10 ml of hexanes and filtered through a 1 cm plug of silica gel. The silica was washed with 3:1 hexanes/ether, and the eluates were combined and concentrated to give 1.35 g of ethyl [methyl-13 C]methoxyacetate as a colorless liquid. 1  2 Hz). The ester was dissolved in 2 ml of MeOH and treated with 10 ml of 6 N NaOH for 12 h. The mixture was extracted three times with CH 2 Cl 2 , and then the aqueous phase was acidified to below pH 2 using concentrated HCl and diluted with 100 ml of saturated aqueous NaCl before continuous extraction with ether for 24 h. The ether extract was dried over MgSO 4 , filtered, and evaporated to yield 1.5 g of an oil. Bulb-to-bulb distillation under reduced pressure yielded clean [methyl-13 C]methoxyacetic acid (0.70 g). 1  [methyl-13 C]Methoxymalonic acid:5-diazo-Meldrum's acid (2 g), [ 13 C]methanol (2 g), Rh 2 (OAc) 4 (250 mg), and toluene (25 ml) were mixed in an Ace pressure tube. The capped tube was heated at 140°C for 2 h, the reaction was filtered, and the filtrate was evaporated. The residue was dissolved in H 2 O and extracted three times with EtOAc. The combined organic layers were dried over MgSO 4 , filtered, and evaporated to yield an oil (1.3 g), which was identified as a mixture of dimethyl methoxymalonate and monomethyl methoxymalonate. The oil was dissolved in MeOH (5 ml), 6 N NaOH (10 ml) was added, and the mixture was left overnight at room temperature. After cooling on ice, the mixture was acidified to pH Ͻ 1 with HCl and filtered. The filtrate was dried under high vacuum to yield a yellow solid (1 g). To remove the inorganic salt, the solid was extracted three times with acetone; the acetone extracts were evaporated to give methoxymalonic acid as a solid (0.65 g). 1 H NMR (acetone-d 6 ): ␦ 4.47 (1H, d, J ϭ 4 Hz), 3.46 (3H, d, J ϭ 143 Hz). 13 C NMR (acetone-d 6 ): ␦ 167.0, 79.9, 57.5 (enriched).
Cultures grown in TSBGM were harvested at 24 h, washed twice, and resuspended in 100 mM MES, pH 6.0, 1% glucose to the original culture volume. The resting cell suspension was shaken in baffled flasks at 30°C and 175 rpm. The 13 C-labeled precursors were added in three equal portions at 24, 36, and 48 h to obtain 0.5 g/liter final concentration. Resting cell cultures were harvested at 56 h by adding half the volume of the culture of MeOH. The resulting broth/extract was clarified by centrifugation and loaded at 25 ml/min onto a column of Diaion HP-20ss pre-equilibrated in MeOH-H 2 O (2:1). The column volume was 1 ⁄20 the volume of the broth/extract. The column was eluted at 8 ml/min with MeOH-H 2 O at ratios of 1:2 (2 column volumes), 1:1 (4 column volumes), 7:3 (4 column volumes), and 9:1 (4 column volumes). The 9:1 eluent was concentrated to give crude ascomycin, which was chromatographed over Bond-Elut ODS solid phase extraction cartridges by eluting with MeOH-H 2 O (9:1). This material was suitable for NMR analysis.
Expression of DEBS with AT6 Replaced by Ascomycin AT8 -The Streptomyces expression vector pKAO127 containing the SCP2* replicon and the DEBS genes expressed via the actI promoter and actII-ORF4 transcription activator were described previously (29). Constructs with SpeI and PstI sites engineered at the eryAT6 boundaries have also been described previously (6). The ascomycin AT8 was isolated by PCR with primers that introduced a SpeI site at the 5Ј boundary and an NsiI site at the 3Ј boundary. The PCR product was cloned into Litmus28 to give pKOS38-178, which was checked by sequencing, and subsequently inserted between the engineered SpeI and PstI sites flanking DEBS AT6. This was assembled into the DEBS expression construct pKOS38-187. The construct was introduced into S. lividans K4-114 by standard protoplast transformation (28), and the transformants were cultured in R5 medium for production of the 6-DEB analogue.
LC-MS Analysis-Samples were analyzed by on-line extraction by LC-MS using a system comprised of a 10 port, 2 position switching valve/injector, Beckman System Gold HPLC, an Alltech evaporative light scattering detector, and a PE SCIEX API100 LC-mass spectrometer equipped with an atmospheric pressure chemical ionization source. For ascomycin analogue analyses, whole cultures were extracted by adding 1 volume of MeCN and clarified by centrifugation. Sample (250 l) was loaded onto an Upchurch 4.3 ϫ 10 mm ODS guard column that had been pre-equilibrated for 1 min with 0.1% acetic acid at 1 ml/min, with the eluate being diverted to waste. At 30 s postinjection a linear gradient to 1:1 MeCN-H 2 O (0.1% acetic acid) over 1 min was started, with the eluate still diverted to waste. At 2 min, the direction of flow through the guard column was reversed, and the eluent was diverted onto a Metachem Inertsil ODS-3 column (5 m; 4.6 ϫ 150 mm) at 50°C pre-equilibrated with 1:1 MeCN-H 2 O (0.1% acetic acid). A linear gradient from 50% to 100% MeCN (0.1% acetic acid) at 1 ml/min over 5 min and then 100% MeCN (0.1% acetic acid) for 4 min was monitored by evaporative light scattering and mass spectrometry. For 6-DEB and 2-demethyl-6-DEB, 100 l of clarified whole broth was loaded onto the guard column after a 1 min pre-equilibration with H 2 O at 1 ml/min. At 30 s postinjection, a linear gradient to 15% MeCN over 1 min was initiated. At 2 min, the direction of flow through the guard column was reversed, and the eluent was diverted onto a Metachem Inertsil ODS-3 column (5 m; 4.6 ϫ 150 mm) pre-equilibrated with 15% MeCN. A linear gradient from 15% to 100% MeCN at 1 ml/min over 6 min and then 100% MeCN for 3 min was monitored by evaporative light scattering and mass spectrometry.

Replacement of AT8 in the Ascomycin PKS-
We previously hypothesized that AT8 of the ascomycin PKS selects the unusual precursor methoxymalonyl-ACP leading to direct incorporation of the methoxy group at C13 (14). To explore this, ascomycin AT8 was replaced with heterologous ATs of known specificity by double homologous recombination using the phage vector KC515 as shown in Fig. 2. Between 10 and 100 thiostrepton-resistant colonies were obtained after infection with recombinant phages, whereas KC515 alone gave no thiostrepton-resistant colonies. Ten isolates each of lysogens from the rapamycin AT3 or erythromycin AT2 construct were analyzed in detail. None produced detectable ascomycin or related products, consistent with insertion of the phage into a gene essential for ascomycin production. Southern blot hybridization experiments (using XhoI or Acc65I digestion of genomic DNA) showed that of the erythromycin AT2 lysogens, seven arose by recombination at the ketoacyl synthase 8 sequence, and three arose by recombination at the dehydratase 8/keto reductase 8 sequence, whereas of the rapamycin AT3 lysogens, eight recombined at the dehydratase 8/keto reductase 8 sequence, one recombined at the ketoacyl synthase 8, and one apparently recombined at the dehydratase 1/keto reductase 1 sequence, which is 98% identical over 1 kb with keto reductase 8/dehydratase 8. Three lysogens from the malonyl-specific rapamycin AT12 construct also did not produce FK520, and the expected first crossover event was verified using PCR.
After growth in the absence of selection, thio S colonies appeared at a frequency of about 0.3%, half of which produced ascomycin and had therefore reverted to wild-type. With the exception of one thio S recombinant that produced no ascomycin-related compound, the remaining thio S recombinants produced a compound with an LC retention time and atomic mass consistent with either 13-DMA (for the rapamycin AT12 replacement) or 13-MDMA (for the rapamycin AT3 and erythromycin AT2 replacements). The overall statistics indicated little bias for recombination via one flanking sequence over the other. After growing selected strains in laboratory-scale stirred-tank fermenters and purifying the ascomycin analogues, structures were verified by mass spectrometry and NMR analyses. 2 For the rapamycin AT3 replacement strain, only 13-MDMA was produced, and the relative stereochemistry of the 13-methyl group was the same as that of the 13-methoxy group of ascomycin. Thus, the ␣-methyl epimerization activity that occurs in rapamycin module 3 does not appear to reside on the AT domain, consistent with previous work (23). The 13- FIG. 2. Diagram showing how S. hygroscopicus strains were engineered to replace AT8 in the ascomycin PKS with AT domains of other specificity. Each AT replacement cassette was delivered on the KC515 phage vector, and thiostrepton-resistant recombinants were selected. Isolates that had crossed over at either the ketoacyl synthase 8 or dehydratase 8 sequence were observed by Southern blot analyses. Analysis of the thio S isolates indicated that both types of second crossover events occurred, giving wild-type revertants or strains containing the heterologous AT replacement, which produced the predicted ascomycin analogue. See "Experimental Procedures" for details.
DMA and 13-MDMA analogues were produced at levels about 5% of parental ascomycin titers (50 mg/liter), which is better than most PKS engineering results (6). For the rapamycin AT12 replacement strain, 13-DMA was the predominant product, but a small amount (ϳ5% of the total) of ascomycin was also observed, even after two rounds of clonal isolation. A summary of the products identified in cultures of the engineered strains is presented in Table I.
Incorporation of 13 C-labeled Precursors into Ascomycin-It was reported previously that [1,2-13 C]acetate can be incorporated into the macrolactone ring carbons C8, C9, and C20-C23 but not C12-C15 in either ascomycin or FK506 (8). To explore the origin of C12-C15 further, the incorporation of potential 13 (Table  II). No significant enrichment of the O-methyl carbon atoms was seen if either [methyl-13 C]methoxyacetate or [methyl-13 C]methoxymalonate was fed. In addition, no detectable coupled signals at C12-C15 were seen after feeding glycolate. Glycine gave weak coupled signals at C12-C15 and additional weak signals for the O-methyl carbons (at C13, C15, and C31), presumably because of metabolism via the one-carbon pool. On the other hand, glycerol feeding gave unequivocal signals resulting from 1 J C-C for C12-C13 and C14-C15, indicating intact incorporation of two-carbon units at these positions (Table II).
Expression of DEBS with AT6 Replaced by Ascomycin AT8 -Replacement of DEBS AT6 by the malonyl-specific rapamycin AT2 and expression in S. lividans was previously shown to give production of 2-demethyl-6-DEB instead of 6-DEB (6). To fur-ther study the ascomycin AT8 domain, the same fusion junctions engineered in this previous construct were used to replace DEBS AT6 with ascomycin AT8. LC-MS analysis of S. lividans cultures expressing this construct revealed that 6-DEB and 2-demethyl-6-DEB (Fig. 3) were produced at ϳ7 and 3 mg/liter, respectively, or 10 -20% the level of 6-DEB produced by S. lividans cultures expressing the wild-type DEBS genes. No 2-methoxy-2-demethyl-6-DEB was observed. Thus, ascomycin AT8 in this foreign PKS context, expressed in a strain that does not contain the methoxymalonyl precursor, selects either malonyl-CoA or methylmalonyl-CoA as alternative substrate relatively efficiently.

DISCUSSION
The results of AT8 replacement in the ascomycin PKS have proven that module 8 uses a precursor other than malonyl-CoA or methylmalonyl-CoA. Moreover, we have shown that C12-C15 arise by sequential incorporation of an extender unit that can be derived from glycerol, consistent with our previous proposal, based on gene homologies, that the direct precursor comes from the glycolytic pathway (14). Recent results have confirmed that replacement of AT7 with ATs specific for malonyl, methylmalonyl, or ethylmalonyl extenders gave the predicted 15-desmethoxy analogue. 3 Two possible reasons glycolate was not incorporated are that it was not taken up by the cells or that one of the activities needed to convert it to glycerate (e.g. tartronate semialdehyde synthase) was absent from S. hygroscopicus under the growth conditions used.
Our results are consistent with the proposal that methoxymalonyl-ACP is the substrate used by modules 7 and 8 (14). The substrate is more likely to be a methoxymalonyl thioester rather than a hydroxymalonyl thioester for the following reasons. First, aside from fkbM, the assigned 31-O-methyltransferase gene, fkbG is the only other methyltransferase gene found in the cluster (14). It is unlikely that FkbG methylates both the 13-and 15-hydroxyl groups of a post-PKS intermediate because methyltransferases are known to have very tight substrate specificity. Second, the sequence of FkbG is most similar to two methyltransferases encoded in clusters for macrolides with a methoxy group at an ␣ carbon and is more similar to plant caffeoyl-CoA methyltransferases than to any of the enzymes known to methylate a post-PKS macrolide intermediate (14). This is consistent with FkbG methylating a precursor to the extender unit before its incorporation into the 3 A. Schirmer, W. P. Revill, and L. Katz, unpublished data.   polyketide chain. Finally, if hydroxymalonyl units were incorporated at these positions, the polyketide would rearrange to form the hemiketal isomers known to be favored following demethylation of these methoxy groups by mammalian liver CYP3A (4,5), and this could interfere with subsequent biosynthetic steps. Our previous proposal that the substrate is methoxymalonyl-ACP instead of methoxymalonyl-CoA was based on the presence of an unusual ACP gene (fkbJ) in the set of five genes believed to encode the synthesis of this extender unit (14). If this hypothesis is correct, methoxymalonyl-specific AT domains may recognize features of the unusual ACP as an additional way to maintain tight substrate specificity. The choice of precursor by a given AT domain is governed by its relative selectivity for the available substrates and the effective concentrations of those substrates. Clearly, there is an adequate supply of both malonyl-CoA and methylmalonyl-CoA in S. hygroscopicus during ascomycin production because these precursors are incorporated by other modules of the ascomycin PKS. Despite this, AT8 in its native ascomycin PKS context discriminates against these substrates because the corresponding analogues have never been observed from the wild-type strain. In addition, when the fkbG gene was disrupted, no ascomycin-related structures were detected, 4 showing that AT8 (and AT7) discriminates against malonyl-CoA or methylmalonyl-CoA in its native context.
Extender unit specificity is usually tight, although there are examples of relaxed specificity, such as module 4 of the epothilone PKS, which incorporates either malonyl or methylmalonyl units (30,31), and modules in the ascomycin and monensin PKSs, which can incorporate ethylmalonyl or methylmalonyl units (8,32). Rapamycin AT12, which is specific for malonyl-CoA in its native context, accepted the methoxymalonyl substrate to some extent, i.e. about 5% of the product from the engineered strain was ascomycin in the ascomycin PKS context. When ascomycin AT8 was placed into the context of DEBS module 6, both methylmalonyl-CoA and malonyl-CoA were used efficiently as substrates, even though the domain rejects these substrates in the native context, as discussed above. Recently, genes encoding methoxymalonyl-ACP biosynthesis from the ansamitocin gene cluster of Actinosynnema pretiosum were co-expressed along with the engineered DEBS genes in S. lividans, and the resulting strain produced 2-methoxy-2-demethyl-6-DEB. 5 This shows that ascomycin AT8 maintains its original substrate preference in this foreign context. On the other hand, our results also suggest that changing the context of an AT domain can affect substrate selectivity. We hypothesize that external steric forces on a domain after folding and assembly of the PKS complex can change the conformation of a domain sufficiently to affect its specificity. Replacement of a short segment of amino acids toward the C terminus of AT domains, which lies outside the homology with E. coli transacylase and is probably on the exterior of the domain, also has been shown to alter specificity (24). Perhaps this region is important for the interaction of the AT domain with the rest of the PKS complex and thus can affect specificity by the mechanism hypothesized here.
Although primary sequence motifs have been identified that play a role in substrate specificity (15,33), it has not yet been possible to identify sequence motifs correlated with specificity for the methoxymalonyl substrate. Indeed, AT domains specific for methoxymalonyl substrates appear to have evolved from either malonyl-specific domains, as in the case of the ascomycin PKS, or from methylmalonyl-specific domains, as in the case of the niddamycin and soraphen PKSs (18,34).