Site-specific mutagenesis and domain substitutions in the loading module of the nystatin polyketide synthase, and their effects on nystatin biosynthesis in Streptomyces noursei.

The loading module for the nystatin polyketide synthase (PKS) in Streptomyces noursei is represented by the NysA protein composed of a ketosynthase (KS(S)), acyltransferase, dehydratase, and an acyl carrier protein. The absolute requirement of this protein for initiation of nystatin biosynthesis was demonstrated by the in-frame deletion of the nysA gene in S. noursei. The role of the NysA KS(S) domain, however, remained unclear, since no data on the significance of the "active site" serine (Ser-170) residue in the loading modules of type I PKSs were available. Site-specific mutagenesis of Ser-170 both in the wild-type NysA and in the hybrid loading module containing malonyl-specific acyltransferase domain from the extender module had no effect on nystatin biosynthesis. A second mutation (S413N) of the NysA KS(S) domain was discovered that completely abolished the ability of the hybrids to restore nystatin biosynthesis, presumably by affecting the ability of the resulting proteins to catalyze the required substrate decarboxylation. In contrast, NysA and its Ser-170 mutants bearing the same S413N mutation were able to restore nystatin production to significant levels, probably by using acetyl-CoA as a starter unit. Together, these data suggest that the KS(S) domain of NysA differs from the KS(Q) domains found in the loading modules of several PKS type I systems in that the active site residue is not significant for its activity.

Modular (type I) polyketide synthases (PKSs) 1 are multifunctional proteins responsible for the biosynthesis of structurally diverse natural products, macrolides, with a wide range of pharmacological applications. PKSs catalyze decarboxylative condensations of simple carboxylic acids into the growing polyketide chain by a mechanism similar to the fatty-acid synthases. PKSs, however, are more diverse in their catalytic reactions, including the use of different primer and extender molecules. Each PKS module is a collection of domains with distinct catalytic functions during PKS catalysis, one module being responsible for one condensation step in the biosynthetic pathway. The minimal domains necessary for condensation are ketosynthase (KS), acyltransferase (AT), and acyl carrier protein (ACP). In addition, the domains ketoreductase, dehydratase (DH), and enoyl reductase responsible for different degrees of the reduction state of every ␤-carbon in the polyketide chain may be present in various combinations. It follows that the chemical structure of the final product is largely reflected in the modular organization of the PKS proteins (1).
A distinctive feature of modular PKS is the presence of a loading module for chain initiation, which is usually fused to the first extender module. The loading module of the 6-deoxyerythronolide B synthase catalyzing the biosynthesis of the erythromycin macrolactone ring in Saccharopolyspora erythraea is composed of the AT and ACP domains. Although several models for chain initiation have been proposed for this system (2,3), evidence accumulated so far indicates that monocarboxylic acyl-CoA species are utilized as substrates. However, recent characterization of the new modular PKS proteins has shown that multidomain loading modules are more common. Such loading modules typically possess, in addition to AT and ACP, a domain designated KS Q , which is similar to the chain length factor (CLF) of type II PKSs (4). Both CLF and KS Q are similar to the KS domains, while having a glutamine residue substituting cysteine in the active site. In the actinorhodin type II PKS CLF forms a heterodimer with KS, and by mutational studies it has been demonstrated that both domains are competent to catalyze substrate decarboxylation, as long as at least one of the domains has a glutamine residue positioned in its active site (4). A similar property has been suggested for the KS Q domains of some modular PKS, such as monensin PKS loading module, where a Q177A substitution in the KS Q domain resulted in a 10-fold reduction of the decarboxylase activity in vitro (4). In addition, KS Q is typically accompanied by a distinct AT domain, with proposed high specificity for dicarboxylic substrates compared with monocarboxylic substrates (3,5). According to a model suggested by others (4), AT loads a dicarboxylic acid onto the ACP, whereas the KS Q domain governs decarboxylation in a way resembling the catalysis performed by extender modules. However, the residue in an "active site" position of the KS might be not the only one crucial for decarboxylation. Mutational analysis of the actinorhodin PKS, for example, demonstrated that additional KS residues might be critical, indicating that the precise mechanism for the KSdependent decarboxylation still remains unclear (6).
We have reported previously (7) cloning and analysis of the complete gene cluster governing biosynthesis of the polyene antibiotic nystatin in Streptomyces noursei, and we showed that the deduced nys-PKS loading module (NysA) possesses some unusual features. NysA is a single module polypeptide composed of four catalytic domains (see Fig. 1) and thus resembles a typical extender module devoid of ketoreductase domain. It contains a KS-like domain with a serine residue (KS S ) instead of a cysteine in the proposed active site position 170. The DH domain does not seem to have any function in NysA, except probably for maintenance of a protein structural integrity, and most likely represents an evolutionary remnant. The primary sequence of the AT domain (AT0) displays an acetate-specific signature (8) and contains the conserved Arg-117 residue (Escherichia coli AT numbering) typical for such domains with high selectivity for CoA esters of dicarboxylic acids (3). It has been shown recently (9, 10) that two of these features, the KS S domain and AT domain with the conserved Arg-117 residue, are shared by other polyene type I PKSs. However, in view of unclear significance of the KS S domain for the NysA function, it remained unknown whether malonyl-CoA or acetyl-CoA is used as primer for nystatin synthesis (Fig. 1, A and B).
In the present study we have analyzed the in vivo role of NysA for the initiation of nystatin synthesis in S. noursei. In particular, the significance of the KS S and the AT0 domains was investigated by site-directed mutagenesis and construction of the hybrid loading modules. Our data indicate that the KS S domain is important, but not essential, for the initiation of nystatin biosynthesis.

Bacterial Strains, Media, and Growth Conditions
Bacterial strains, plasmids, and recombinant phages used in this study are listed in Table I. S. noursei strains were maintained on ISP2 agar medium (Difco), and E. coli strains were grown in L broth or L agar (11). Conjugal plasmid transfer from E. coli ET 12567 (pUZ8002) and the gene replacement procedure were performed as described previously (12,13). Transformation of E. coli was performed as described elsewhere (14). Agar and liquid media, when appropriate, were supplemented with antibiotics using the following concentrations: chloramphenicol, 20 g/ml; kanamycin, 50 g/ml; ampicillin, 100 g/ml; apramycin, 50 g/ml. Analysis of the nystatin production by S. noursei strains was performed in SAO-23 liquid medium as reported previously (13,15).

DNA Manipulations, DNA Sequencing, and PCR
General DNA manipulations were performed as described previously (11). DNA fragments from agarose gels were purified using the Qiaex II kit (Qiagen, Germany). Southern blot analysis was performed with the digoxigenin-11-dUTP High Prime labeling kit (Roche Molecular Biochemicals) according to the manufacturer's instructions. Oligonucleo-tide primers were purchased from Amersham Biosciences, and DNA sequencing was performed using the ABI Prism Big Dye Terminator Cycle sequencing kit and the Genetic Analyzer 3100 (Applied Biosystems, Inc.). The PCRs were performed with the Expand High Fidelity PCR System (Roche Molecular Biochemicals) using the Eppendorf Mastercycler (Eppendorf, Germany), as described previously (15).

Construction of Vectors for In-frame Deletion
and Chromosomal Integration of nysA nysA In-frame Deletion Vector-The 11.0-kb BamHI fragment from phage N76 was cloned into pGEM3Zf(Ϫ), and from this construct (pL76B11) the 5.83-kb PstI-SacI DNA fragment encompassing the coding region of the nysA gene and surrounding sequences was ligated into the corresponding sites of plasmid pGEM3Zf(Ϫ). From the resulting plasmid (pL76PS5.8) the 1.68-kb MscI-PstI and the 2.79-kb SacI-SmaI DNA fragments were excised and ligated into the PstI/SacI-digested pGEM3Zf(Ϫ). From this construct, designated pNAD1, the 4.49-kb EcoRI-HindIII insert was ligated together with the 3.0-kb EcoRI-Hin-dIII fragment from plasmid pSOK201, resulting in the nysA replacement vector pNAD2.
nysA Integration Vector-The 4.63-kb EcoRI-SpeI fragment from plasmid pL76B11 (including the coding region of nysA and 157-bp upstream sequences) was ligated into the EcoRI/XbaI sites of pGEM3Zf(Ϫ), as well as into the EcoRI-SpeI sites of pLITMUS28, resulting in plasmids pL76ES4.6 and pLIT76ES4.6, respectively. From pL76ES4.6 the 4.63-kb EcoRI-HindIII insert was excised and ligated into the EcoRI/HindIII sites of vector pSOK804, resulting in the nysA integration vector pKS S AT0. Plasmid pSOK804 is a mobilizable vector capable of site-specific integration into the S. noursei chromosome recently constructed in or laboratory. 2

PCR Amplification of DNA Regions Encoding Individual PKS Domains
The nys-PKS regions encoding individual domains were PCR-amplified and subcloned. Primers for the introduction of unique restriction sites were typically used to facilitate further assembly of the resulting fragments into hybrid genes. The new restriction sites introduced and used for the assembly procedure were in all cases positioned in the inter-domain linker regions to avoid any unwarranted structural effects for the resulting hybrid proteins. All the resulting constructs were verified by DNA sequencing. 2 O. Sekurova, personal communication. The in situ decarboxylation, presumably catalyzed by the KS S , takes place before the resulting acetyl unit is transferred to NysB for condensation with the first propionate extender. B, alternative model. An acetyl unit is recruited directly from acetyl-CoA, and no substrate decarboxylation is required before condensation.
PCR Amplification and Subcloning of DNA Regions Encoding Variant KS Domains-The 1.75-kb DNA fragments covering the coding region of the variant NysA KS Q , KS C , and KS S domains, including the nysA promoter region, were PCR-amplified from the plasmid pLIT76SS2.8 and its mutated derivatives (see above) DNA using the following primer pair: NAKS-F, 5Ј-TACGACTCACTAGTCTTCG-GCGCGCG-3Ј (sense), and NAKS-R, 5Ј-GGAGGATCCTGGTGGGCCG-CACCGC-3Ј (antisense). The latter primer contains a BamHI recognition site (underlined) used for the hybrid assembly procedure (see below). The resulting PCR fragments were blunt-ligated into the SmaI site of pUC18, and from the resulting constructs the 1.75-kb EcoRI-BamHI inserts were excised and ligated into the corresponding sites of vector pGEM3Zf(Ϫ), resulting in the constructs pGEM-KS Q , pGEM-KS C , and pGEM-KS S , respectively (see Table I). In addition, a derivative of pGEM-KS C harboring a second point mutation (S413N) of the KS coding region was identified and denoted pGEM-KS CN . This mutation most probably resulted from an erroneous PCR amplification. The 1.0-kb MluI-SpeI fragment from plasmid pGEM-KS CN (including the S413N but not the S170C mutation) was used to substitute the corresponding fragments in plasmids pGEM-KS Q and pGEM-KS S , resulting in plasmids pGEM-KS QN and pGEM-KS SN , respectively (see Table I).
PCR Amplification and Subcloning of the nysA DH ϩ ACP Coding Region-The 1.63-kb DNA fragment encompassing the continuous coding region of the NysA DH ϩ ACP domains was PCR-amplified from the plasmid pL76B11 DNA, using the following primer pair: NADH-F, 5Ј-GGAGCATGCCTTCTACGCCGGCAC-3Ј (sense), and NADH-R, 5Ј-GCGAAGCTTGGTCAGTCCTCGTCATC-3Ј (antisense). Underlined in the primers are the restriction sites (HindIII and SphI, respectively) used for end digestion and cloning of the PCR fragment into the corresponding sites of pGEM3Zf(Ϫ), resulting in plasmid pGEM-DHACP. PCR Amplification and Subcloning of the NysB AT1 and the NysC AT3 Coding Regions-The 0.99-kb DNA fragment including the coding region of the Nys-PKS module 1 AT domain (AT1) was PCR-amplified from the pL76B11 DNA using the following primer pair: NBAT1-F, 5Ј-GGAGGATCCGACGGTGTTCGTGTTCC-3Ј (sense), and NBAT1-R, 5Ј-GCAGCATGCCAGTCCACGTCCACAC-3Ј (antisense). Similarly, the 0.99-kb DNA fragment covering the coding region of the Nys-PKS module 3 AT domain (AT3) was PCR-amplified from the phage N14 DNA using the following primer pair: NCAT3-F, 5Ј-GCAGGATCCCT-GCGCCGTCCTCTTC-3Ј (sense), and NCAT3-R, 5Ј-GCAGCATGC-CAGCGGACCGTGACGC-3Ј (antisense). Both the resulting PCR products were end-digested with BamHI and SphI (recognition sites underlined in each primer) and ligated into the corresponding sites of pGEM3Zf(Ϫ), resulting in the constructs pGEM-AT1 and pGEM-AT3, respectively.

Assembly of DNA Fragments Encoding Individual PKS Domains into Integration Vectors Expressing Hybrid-loading Modules
One of the variant 1.75-kb EcoRI-BamHI KS-encoding fragments from the construct pGEM-KS C , pGEM-KS Q , pGEM-KS S , pGEM-KS CN , and pGEM-KS QN , or pGEM-KS SN , the 1.6-kb SphI-HindIII DH ϩ ACP insert from pGEM-DHACP, and one of the variant 0.99-kb BamHI-SphI AT inserts from pGEM-AT1 or pGEM-AT3 were ligated into the EcoRI/ HindIII sites of the vector pSOK804. By using this strategy the following nine different integration plasmids were obtained: pKS S AT1, pK-S C AT1, pKS Q AT1, pKS S AT3, pKS C AT3, pKS Q AT3, pKS SN AT3, pKS CN AT3, and pKS QN AT3 (see Tables I and II).

Sequence Alignment and Phylogenetic Analysis
Alignment of the amino acid sequences was done using the ClustalW 1.82 software provided on-line by the European Bioinformatics Institute (www.ebi.ac.uk/clustalw). The alignment was visualized using the BoxShade 3.21 on-line software from the Swiss node of EMBnet (www.ch.embnet.org/software/BOX_form.html).
The phylogenetic tree was drawn by the TreeView 1.6.6 software developed by R. D. M. Page at the University of Glasgow (taxonomy. zoology.gla.ac.uk/rod/treeview.html) using the distances calculated after the ClustalW alignment at the EBI server (see above).

RESULTS
In-frame Deletion and Complementation of nysA-In a previous report (7) we described the nysA gene disruption abolishing all nystatin synthesis in S. noursei. However, to rule out the possibility of any polar effects that might have been imposed by this mutation, we constructed a nysA in-frame deletion mutant NDA59 by double homologous recombination, using the gene replacement vector pNAD2. The 1.36-kb in-frame deletion affecting the nysA gene within this plasmid eliminated the coding sequence for the C-terminal part of KS S and most of the AT0 domain (see Fig. 2). Analysis of NDA59 showed that it produces no nystatin (Fig. 2B), thus confirming that the NysA loading module is essential for initiation of nystatin synthesis in S. noursei. In an attempt to complement this mutant, plasmid pKS S AT0 was chromosomally integrated in the NDA59 strain. This plasmid harbors a DNA fragment encompassing the nysA gene and a 157-bp sequence upstream of its coding region, presumably containing the nysA promoter. Nystatin synthesis was restored to wild-type levels (Fig. 2C) in the NDA59 (pKS S AT0) strain, demonstrating that the nysA gene expressed in trans from its endogenous promoter can efficiently complement the nysA deficiency in the NDA59 mutant. We next integrated plasmid pKS S AT0 into the wild-type strain in order to test whether expression of an additional copy of nysA may stimulate nystatin production. This manipulation had no detectable effect on the nystatin synthesis (data not shown) suggesting that the NysA expression level is not a limiting factor for the production of this antibiotic under the conditions tested.
Replacement of the Proposed Active Site Serine (Ser-170) in the NysA KS S Domain Has No Effect on Nystatin Synthesis in Vivo-It is generally accepted that the initial condensation step a Description of the integration vectors used is given in Table I. b KS domains harboring between zero and two amino acid substitutions were used, and the particular mutations are indicated. c The variant AT domains used are as follows: AT0, NysA; AT1, NysB module 1; AT3, NysC module 3. d Nystatin production was measured as described under "Experimental Procedures." All experiments were performed in triplicate, and the mean values with the corresponding S.E. are given for all cases, 100% represents the production levels obtained in NDA59 (pKS S AT0), and 0 nystatin is below 0.05%. during the PKS-catalyzed polyketide synthesis takes place on the KS domain of the first extender module. Loading modules usually either lack the KS domains completely or have KS Q domains with a glutamine instead of a cysteine in the active site position. These KS Q domains in the loading modules are believed to be responsible for decarboxylation of the starter dicarboxylic acid CoA esters prior to condensation. The KS S domain of NysA has a serine residue in this position (Fig. 3A), and the role of this domain for initiation of nystatin synthesis was unknown. We proposed that if AT0 recruits malonyl-CoA, then KS S should possess a decarboxylase activity, whereas no such activity is needed if acetyl-CoA is used (see Fig. 1). To test whether the Ser-170 in KS S is important for the NysA activity, this residue was substituted with glutamine (S170Q), cysteine (S170C), and the catalytically "silent" glycine (S170G). Integration vectors expressing the resulting mutant nysA genes (plasmids pKS Q AT0, pKS C AT0, and pKS G AT0, respectively) were introduced into the S. noursei mutant NDA59, and the resulting recombinant strains were analyzed for nystatin production. Surprisingly, similar nystatin production by the NDA59 (pKS S AT0) and NDA59-based recombinant strains expressing NysA with mutagenized KS domain was observed (Table II), suggesting that the Ser-170 residue is not important for NysAmediated initiation of the nystatin biosynthesis under these conditions.

Construction and Expression of Hybrid Loading Modules with Mutant KS and Alternative AT Domains in the NDA59
Mutant-If one assumes that NysA is using malonyl-CoA as a starter, then the results described above would imply that the supposed KS-mediated decarboxylation of malonyl-CoA is independent of the type of residue in the active site position 170. Alternatively, the NysA might utilize acetyl-CoA as a starter, suggesting that decarboxylation is not at all required in the initiation process, and the KS S domain has no catalytic role (Fig. 1). To resolve these issues, we constructed hybrid NysA derivatives where the AT0 domain was substituted with the AT domains from the downstream PKS modules that must have high specificity for the dicarboxylic acid substrates.
Based on the known nystatin structure and sequence analysis of its biosynthetic genes, the AT domains of the nys-PKS modules 1 (AT1) and 3 (AT3) are clearly specific for methylmalonyl-and malonyl-CoA, respectively (7). Both these domains were individually used to substitute for AT0 in NysA and its KS mutants. The purpose of these experiments was to unravel whether NysA can utilize dicarboxylic acid substrates in vivo, and whether this ability is dependent on the type of residue in position 170 of the loading module. Care was taken to keep the polypeptide length, the interdomain linkers, and the domain order of each hybrid similar to the native NysA protein. The C-terminal part of NysA encompassing DH and ACP domains was kept intact in all cases. The resulting vectors were introduced into S. noursei mutant NDA59, and the recombinant strains were analyzed for nystatin production. None of the integration vectors expressing the AT1 hybrids (pKS S AT1, pKS C AT1, and pKS Q AT1) could restore synthesis of either nystatin or related polyenes when introduced into the NDA59 mutant (Table II). However, the integration vectors expressing the AT3 hybrids (pKS S AT3, pKS C AT3, and pKS Q AT3) were able to restore nystatin production in NDA59, although not as efficiently as the pKS S AT0 expressing the wild-type nysA (Table II). These results indicate that NysA is capable of using malonyl-CoA as a starter unit in a manner that seems to be independent of the type of residue in position 170. In this way our data contradict previous reports (3,4) stating that the type of residue in this particular position is critical for decarboxylase activity and suggest that other residues within NysA may be essential. The lower production levels obtained with the AT3 hybrids (between 11 and 17%) compared with those obtained with the wild-type NysA may be due to several reasons, the improper folding of the hybrid proteins being most likely. This result may also in part be explained if the hybrids have narrower substrate specificity, i.e. if they are capable of accepting only malonyl-CoA, whereas both acetyl-CoA and malonyl-CoA may be used by the wild-type NysA.
A Novel KS Mutation (S413N) Abolishes the in Vivo Activity Exclusively in the AT3 Hybrids-During the hybrid construction procedure (see "Experimental Procedures"), an unwarranted mutation was discovered, resulting in substitution of the highly conserved serine residue in position 413 of the NysA KS domain with an asparagine (S413N). First, this mutation was introduced into the integration vectors expressing the AT3 hybrids, and the resulting plasmids (pKS SN AT3, pKS CN AT3, and pKS QN AT3) were used to complement NDA59. Surprisingly, none of the resulting recombinant strains was able to produce nystatin (Table II), suggesting that Ser-413 was critical for the in vivo activity of the AT3 hybrids. Sequence comparisons (Fig. 3A) revealed that this particular residue is highly conserved in the KSs of single polypeptide loading modules as well as extender module KSs. We reasoned that if this mutation somehow affects the decarboxylase activity linked to KS S , and wild-type NysA can use both acetyl-and malonyl-CoA as starter units, the effect of such a mutation would be less dramatic on the AT0 background. Therefore, the S413N mutation was introduced into the integration vectors expressing NysA and its Ser-170 mutants, and the resulting plasmids (pKS SN AT0, pKS CN AT0, and pKS QN AT0) were introduced into the NDA59 mutant. Remarkably, the nystatin synthesis in the resulting recombinant strains was restored to 19 -36% compared with the NDA59 (pKS S AT0) (Table II). Together, these results suggest that the observed effect of the S413N mutation on the ability of the loading modules to initiate nystatin biosynthesis depends on the type of AT domain present (see "Discussion").

DISCUSSION
Modular PKSs have proven to be exciting targets for the rational genetic engineering with the aim of producing new macrolide antibiotics. Several strategies such as single domain inactivation, domain swapping and insertion, and module rearrangements have led to synthesis of novel polyketides (16). The process of initiation of polyketide synthesis governed by the distinct loading modules may also be a target for genetic manipulations aimed at the synthesis of new products (5,17). The majority of the modular PKS loading modules can be divided into two classes, depending on their domain organization and, consequently, their mode of initiation of polyketide synthesis. One class is represented by the loading modules having, in addition to the AT and ACP domains, a KS Q domain with suggested decarboxylase activity. These modules, represented by oleandomycin, niddamycin, and pikromycin PKSs, accept dicarboxylic acid-CoA esters such as malonyl-CoA and methylmalonyl-CoA as starters. Such starter molecules must be decarboxylated before they are loaded onto the first extender module, and the KS Q domain seems to be crucial for this process (4). Yet another class of loading modules found, for example, in the erythromycin and avermectin PKSs is characterized by having the AT and ACP domains only. These modules apparently use monocarboxylic acid-CoA esters such as propionyl-CoA and isobutyryl-CoA to load them onto the first extender module of the corresponding PKS. Therefore, no decarboxylation is required during initiation of polyketide synthesis in the latter systems.
The deduced loading module for the nystatin PKS, NysA, is composed of the KS S , AT0, DH, and ACP domains (see Fig. 1). By constructing a nysA in-frame deletion mutant, we have demonstrated that NysA is essential for in vivo production of nystatin in S. noursei. This result suggests that the first extension module of NysB cannot be loaded directly by a starter carboxylic acid, as was apparently the case for erythromycin PKS, where inactivation of the 6-deoxyerythronolide B synthase loading module did not abolish erythromycin production completely (2,18). In ery-PKS, the direct loading of the KS domain of the first extension module was proposed as an alternative mechanism for chain initiation (3).
During the polyketide chain elongation the KS domains of each extender module supposedly play two roles: (i) they catalyze condensation of the extender unit onto the polyketide chain, and (ii) simultaneously promote the in situ decarboxylation of the recruited dicarboxylic acyl-CoA esters during the condensation process. These two functions seem to be inseparable in the extender KSs, since the condensation process strictly depends on decarboxylation (19,20). The condensation invariantly requires the conserved cysteine in the KS active site, whereas the mechanism and amino acid residue(s) involved in decarboxylation are still not completely unraveled (4,6). It is suggested that there exists a cooperation between the KS and AT domains on each module for the discriminative recruitment and concomitant decarboxylation of dicarboxylated substrates (3,5). In the complex loading modules containing KS Q domains, the active site glutamine has been shown to be important for decarboxylase activity (4). In such loading modules the AT domains are found to be structurally and functionally similar to their counterparts found in the extender modules, with high specificity for the dicarboxylic acyl-CoAs (3). NysA has an extender-type AT domain suggesting that malonyl-CoA might be its preferred substrate. However, this assumption heavily relies on the decarboxylase activity of the KS S , analogous to what has been demonstrated for the KS Q in other systems. At least in the type II PKS system, the CLF with serine residue in the active site position retained most of its decarboxylase activity (6).
One obvious strategy to test whether the KS S is at all required for initiation of nystatin biosynthesis would be to make an in-frame deletion within the KS S -coding region of nysA. However, previous attempts on using such strategies for inactivation of the nys-PKS reduction domains typically abolished all polyene production, presumably by disturbing the folding of the resulting mutant proteins. 3 Our mutagenesis results clearly show that the type of residue in the active site position 170 of KS S has no significance for the in vivo activity of NysA. However, the successful complementation of the nysA-deficient mutant NDA59 using the AT3 hybrids indicates that NysA is somehow capable of catalyzing the required decarboxylation of malonyl-CoA. Although the type of residue at position 170 had little effect on the ability of the hybrid proteins to initiate nystatin biosynthesis, the second S413N mutation completely aborted this function in such hybrids. The most probable explanation is that the latter mutation has abolished decarboxylase activity of these proteins preventing proper priming with the malonyl-CoA recruited by the AT3. Alternatively, this substitution might have caused unwarranted structural changes in the AT3 hybrids leading to loss of all in vivo activity. The latter explanation seems unlikely, since introduction of the S413N mutation in the wild-type NysA, as well as in the NysA KS C and KS Q mutants, did not abolish their ability to initiate nystatin synthesis, although it has been diminished by 64 -81%. By assuming that the S413N mutation abolishes decarboxylase activity in NysA, the residual in vivo activity observed for NysA and its active site mutants bearing this mutation could be due to the utilization of acetyl-CoA as a starter. This notion is in agreement with a recent report showing that loading module AT of the ery-PKS has, in contrast to the extender AT, a relaxed specificity toward acyl-CoA substrates and can accept both propionyl-CoA, acetyl-CoA, and butyryl-CoA (21). No clues as to which amino acid residues might be responsible for the proposed broad substrate specificity of AT0 in NysA could be found by inspecting the alignment of the AT amino acid sequences (data not shown). Loading modules of the polyene PKSs seem to be rather unique, as they are represented by separate polypeptides, whereas in other modular PKS systems loading modules are fused to the first extender module. This feature may in principle provide a greater degree of freedom in folding of the polyene PKS loading modules that, in turn, might affect the choice of the starter unit by these proteins.
The inability of the AT1 hybrids to complement nystatin synthesis in NDA59 could be due to several reasons. Based on the results obtained with the AT3 hybrids, we suggest that these observations cannot be solely explained by the lack of the decarboxylase activity in the hybrid proteins. A possible explanation may be that the downstream nys-PKS module in NysB cannot use the propionate primer generated by these engineered proteins, and further elongation is aborted. Alternatively, the heterologous AT1 domain might be the cause of the misfolding of the hybrid modules and the subsequent loss of activity. Similar problems are documented by others (22), and in such cases site-directed mutagenesis has successfully been applied as an alternative strategy for changing the substrate specificity of AT domains.
Alignment of the amino acid sequences for the KS-like domains in the PKS type I loading modules clearly shows that all KS Q domains contain a glycine at position 413 (NysA numbering), whereas the single polypeptide loading modules contain highly conserved serine (Fig. 3A). A threonine residue in the PimS0 loading module is structurally very similar to serine and thus might well serve as its functional substitute. Interestingly, the alignment of the KS domains from the extender modules of 13 PKS type I systems (encompassing together over 100 KS domains) shows that the Ser-413 residue is highly conserved among these domains, with only one exception; KS in the module 2 of rifamycin PKS (data not shown). Ser-413 was also found to be highly conserved in the extender KSs of the PKS type I systems discovered in such diverse bacterial genera as Mycobacterium, Nostoc, Stigmatella, Desulfovibrio, and Planktothrix (data not shown). Finally, the phylogenetic tree based on the KS sequence alignments clearly shows that KS S cluster with the extender KS domains, whereas the KS Q domains represent a separate branch (Fig. 3B). All these data further suggest that single polypeptide loading modules differ from their KS Q counterparts, which could explain the discrepancy in the data obtained upon the active site residue mutagenesis reported here and by others (4).