WO2000052152A1

WO2000052152A1 - Non-ribosomal peptide synthetases, method for producing same and the use thereof

Info

Publication number: WO2000052152A1
Application number: PCT/EP2000/001652
Authority: WO
Inventors: Mohamed A. Marahiel; Torsten Stachelhaus; Henning Mootz; Dirk Konz
Original assignee: Marahiel Mohamed A; Torsten Stachelhaus; Henning Mootz; Dirk Konz
Priority date: 1999-03-03
Filing date: 2000-02-28
Publication date: 2000-09-08
Also published as: CN1342201A; EP1159416A1; CA2364569A1; AU3425700A; JP2002537806A; IL145004A0; DE19909146A1

Abstract

The present invention relates to a method for producing modified non-ribosomal peptide synthetases (NRPS). The invention also relates to the use thereof for producing native or artificial peptides or proteinogenes. In particular, a desired peptide can be obtained by means of a novel non-ribosomal code.

Description

description

NON-RIBOSOMAL PEPTIDE SYNTHETICS, METHOD FOR THE PRODUCTION AND USE THEREOF

The present invention relates to a method for producing modified non-ribosomal peptide synthetases (NRPS) and the use of those for producing native or artificial peptides and proteinogens.

Non-ribosomal peptide synthetases (NRPS or peptide synthetases) are modular enzymes with complex structures and important biological functions. Numerous peptides with pharmaceutical and biotechnological interest - in particular those such as biosurfactants, siderophores, antibiotics, cytostatics, immunosuppressants or antitumor agents - are produced by large enzyme complexes, the so-called NRPS (Marahiel et al. (1997), Chem. Rev. 97: p.2651- 2673) biosynthesized. These include well-known drugs such as cyclosporin A and vancomycin. The large variety of bioactive peptides produced in this way is a result of the large structural diversity of the NRPS. In addition to the 20 proteinogenic amino acids, NRPS often build special residues, such as α-hydroxy amino acids or non-proteinogenic amino acids. The individual residues are linked to one another via peptide bonds or the formation of esters and lactones. The residues can also be further modified by N-methylation, heterocyclic ring formation or epimerization. Many of the peptides synthesized by natural NRPS are also cyclized by ester or peptide bonds. Generally speaking, NRPS can play a key role in the synthesis of complex peptide bio-compounds.

It has been found that the multifunctional protein templates have an essentially modular structure. The module is the catalytic unit that incorporates a specific amino acid into the product (peptide) (Marahiel et al. (1997), Chem. Rev. 97: S.2651-2673).

The individual modules are composed of so-called domains, each of which is responsible for a specific reaction. The adenylation domain (A- Domain) determines the entry of the substrate into the peptide by the A domain selecting and adenylating the substrate, usually an amino acid. The activated amino acid is then bound via a thioester bond to the cofactor 4'-phosphopantetin in a thiolation domain (T). From there, the aminoacyl or peptidyl residues are condensed to a neighboring module. This reaction is catalyzed by the condensation domain (C domain). These three, the C, A and T domains form the basic unit of the multimodular NRPS. The last module of an NRPS usually contains a termination domain (Te domain) which is responsible for the release of the synthesis product. The order of the modules within the NRPS determines the structure of the peptide.

Modification domains are built into the corresponding module at locations where a modified amino acid is incorporated.

In EP 0789078 A2, the exchange of domains for the variation of known compounds was merely proposed; there is only one technical teaching there with regard to the basic control of the peptide chain termination reaction.

In detail: An adenylation (A) domain, for example, recognizes and activates its cognate substrate amino acid as an enzyme-associated amino acyl adenylate with simultaneous ATP hydrolysis. These relatively unstable intermediates are then stabilized by transfer to the thiol group of a 4'-phosphopantethein cofactor (4'-PAN) covalently linked to a thiolation (T) domain. With the participation of condensation (C) domains, a sequential chain growth up to the finished peptide product then takes place in an ordered sequence of transpeptidations of the amino acids activated in this way. The structure of the products formed can also be changed by modifying the incorporated monomers (for example epimerization or N-methylation) or the main peptide chain (for example acylation or glycosylation). Such functionalizations are catalyzed by special domains or polyketide synthase (PKS) modules within the NRPS matrix. It could be shown that A domains determine the primary structure (sequence) of the non-tribosomal peptide formed via their substrate specificity and relative sequence within a specific NRPS system. Due to the resolution of the crystal structure of two members of the superfamily of adenylate-forming enzymes, this was the luciferase from Photinus pyralis [Conti et al., 1996, Structure 4: 287-298] and the A domain of the gramicidin S- Synthetase 1 (hereinafter referred to as PheA) from Bacillus brevis [Conti et al., 1997, EMBO J. 16: 4174-4183], could be shown that although both enzymes have only 16% identity with regard to their primary structure, they express one homologous folding topology. The structure of PheA (Figure 1A), which was determined in the presence of its substrates ATP and phenylalanine, provided first insights into the molecular mechanisms of substrate recognition and activation.

Sequence comparisons of NRPS-A domains identified highly conserved sequence areas, the so-called core motifs, which occur within the A domains with almost unchanged localization and amino acid sequence. Previous mutation analyzes and now also the crystal structure of PheA have shown that the vast majority of these core sequences are essentially involved in ATP binding and hydrolysis. In addition, some core motifs also have functions in binding the substrate amino acid. For example, the α-amino group and α-carboxy group of the substrate amino acid L-phenylalanine in PheA are coordinated electrostatically by the residues Asp235 (Core A4) and Lys 517 (Core A10) [Marahiel et al., 1997, Chem. Rev. 97: 2651-2673]. The actual substrate specificity-mediating region of PheA, on the other hand, is located in an approximately 100 amino acid-long region which has a reduced conservation within the group of the A domains. This region forms the binding pocket for the side chain of the phenylalanine substrate (FIG. 1B), which is built up on one side by the residues Ala236, Ile330 and Cys331, and on the other side by the residues Ala322, Ala301, He 299 and Thr278 . Both sides of the bag are separated by the indole ring of the Trp239, which is located on the bottom of the bag. From previous studies in silico (computer simulations) on NRPS-A domains, it could be concluded that the substrate-specific-determining region of A domains is probably located between the core motifs A3 and A6. This region shows a comparatively reduced conservation, whereas an increased homology relationship was postulated for domains with the same substrate specificity. Accordingly, domains that activate the same substrates should appear grouped in phylogenetic trees. Such correlations, however, could not be verified in detailed analyzes. Rather, it was shown that the evolutionary origin (host organism) of the A domains had a greater influence on their bundling than their substrate specificity.

To derive the consensus sequences, we determined the corresponding ten residues of the putative binding pockets of 160 A domains accessible in public sequence databases by means of sequence comparisons with PheA and luciferase. In the following, only these ten residues were used as an independent amino acid sequence for each A domain in order to carry out alignments and phylogenetic tests with the MegAlign program from Lasergene (available, for example, from GATC, GmbH. Fritz-Arnold Str. 23, D-78467 Konstanz) . By means of these investigations, a phylogenetic family tree was obtained (FIG. 2), in which a grouping of the 10 amino acid sequences according to the specificity of their domains of origin is revealed. This result gives a clear indication that the selected residues are actually involved in the substrate recognition. Based on these investigations, discrepancies between postulated and observed substrate specificities of A domains can be explained and the specificities of NRPS systems can be predicted.

The object underlying the present invention was a method for modifying A domains for the purpose of providing and producing appropriately modified non-ribosomal peptide synthetics for the production of new peptides. The invention allows the targeted production of a desired non-ribosomal peptide. Surprisingly, consensus sequences could be derived from the phylogenetic studies, which are to be understood as the non-tribosomal code of the NRPS (Table 1). Similar to the ribosomal code, this code is degenerate (redundant), and so far, for example, four different codes for Leu-activating domains, three codes for Val and two for Cys-activating domains have been identified. It can be assumed that there are also several substrate detection strategies for other substrates which can be determined with the aid of the inventive method.

Therefore, the invention also relates to the non-ribosomal code, represented in Table 1, coding for a non-ribosomal peptide.

The general shape of substrate binding pockets can be predicted on the basis of the “codons” determined by sequence comparisons. The putative binding pockets for the substrate amino acids Asp, Orn and Val are shown as examples in FIG. 3. While Asp235 and Lys517 are keys in all A domains - To mediate interactions with the amino and carboxy groups of the substrate and are therefore highly conserved, it is assumed that the remaining residues are largely responsible for determining the specificity for a certain amino acid side chain: the corresponding residues (items 236 to 331) in the three given Examples support this theory and perfectly reflect the needs regarding the polarity and size of the activated substrates.

(1) Asp activation (codon 'Asp'): The basic residue His322 (possibly also Lys278) establishes a polar interaction with the acidic side chain, whereas the rest Leu236 closes the binding pocket for the acceptance of longer-chain substrates (e.g. Glu).

(2) Orn activation (codon Orn (2) '): The acidic residues Glu278 and Asp331 seem to play a key role in the recognition of the large basic side chain. (3) Val activation (codon Nal (3) '): The entire binding pocket is built up by hydrophobic residues. The relatively large space requirement resulting from the branching of the β position is taken into account (1) by using smaller side groups in the upper area, and (2) spatially demanding residues in the lower area, which expand the binding pocket.

In addition to degeneration, another homology to the ribosomal code is the appearance of flexible (wobble-like) positions. As shown in Table 1, positions 278, 299 and 331 are those wobble-like residues that have increased flexibility within certain codons. In general, it is observed that binding pockets which recognize smaller amino acids have a higher flexibility at positions near the bottom of the pocket, whereas with large substrates a greater variability is observed in the upper region (see Table 1).

The results of the in silico studies, which are summarized in Figure 3 and Table 1, show that the individual constituents of a binding pocket have different significance for the formation of a specific substrate specificity. To confirm this observation, the remains of 160 binding pockets were examined in more detail. The results shown in Figure 4 and Table 2 show that the ten constituents of a binding pocket can be divided into three groups. Their uneven variability probably reflects their different meanings in conveying substrate specificity:

(1) 'Invariant' positions (pos. 235 and 517): Asp235 is only absent in A domains that activate substrates without an α-amino group. Examples include the α-aminoadipate-activating domains of the AcvA synthetases or the carboxyacid-activating domains of the Enterobactin (Escherichia coli) and Yersiniabactin systems (Yersinia pestis). Lys517, on the other hand, is absolutely invariant and binds both the α-carboxy group of the cognate amino acid and the 04 'and 05' of the ATP / AMP-Ribose unit, which presumably brings both substrates into a position of optimal interaction. Both residues convey important Interaction with the (largely) unchangeable α-amino and α-carboxy groups of the substrate, which explains their own invariance.

(2) 'Conditional variant' positions (pos. 236, 301 and 330): These residues show only slight variability within all examined domains and in 93% of all cases hydrophobic amino acids are realized at these positions. Taking into account the low flexibility of these positions and the disproportionately low occurrence of charged or polar side chains, an essential importance in conveying the substrate specificity can be excluded.

(3) 'Highest variant' positions (pos. 239, 278, 299, 322 and 331): These positions have the greatest flexibility with regard to the amino acids used. Apart from items 299 and 331, which generally appear to be very adaptive and wobble-like (see Table 1), all positions are predestined for the establishment of substrate specificity. Taking into account their relative location within the binding pocket and the data shown in Tables 1 + 2 and 3, it can be assumed that position 322 has a greater influence on smaller substrate residues, whereas positions 239 and 278 rather affect larger residues. This hypothesis is e.g. supported by the codons of the Asp (His322), Glu (Lys239) and Orn (Glu278) activating A domains (Table 1 +2). In all of the examples mentioned, the polarity of the recognized substrate and the 'highly variant' positions of the binding pocket are perfectly coupled (see Table 2; dark gray background).

The modified DNA is implanted chromosomally or extrachromosomally in the host organism according to methods known per se to the person skilled in the art (Sambrook, J., Fritsch, E.F. & Maniatis, T. (1989), Molecular Cloning: A Laboratory Manual, Cold Spring Harbor Laboratory Press).

The present invention therefore relates to the embodiments set out in the patent claims. For experimental confirmation, we mutated several of the specificity-mediating amino acid constituents of the binding pocket of PheA and examined the mutants for resulting changes in (1) the rate of ATP-pyrophosphate exchange (enzyme activity) and (2) the activation of mis-cognate amino acids (Enzyme specificity). With regard to the latter point, it should not be left unmentioned that the wild-type protein showed a significant activation of the mis-cognate substrates Trp (18%) and Leu (7%) from the outset. The results of our PheA mutagenesis experiments are summarized in Table 3 and can be generalized as follows:

(1) The substitution for somewhat larger amino acids was inevitably at the expense of the space available in the binding pocket. As a result, the ATP pyrophosphate exchange rate of these mutants decreased as their specificity for the cognate substrate phenylalanine increased (e.g. T278M, T278Q, A301V, A322S and C331 L). In all cases, the page specificity for the miscognate substrates Trp and Leu was reduced by a factor of 20.

(2) The exact opposite was observed for exchanges for somewhat smaller groups, which expanded the binding pocket slightly. In these mutants, the catalytic activity of the wild-type protein was rescued, while the ability to discriminate against mis-cognate substrates by up to 5-fold (e.g. A301G, A322G and I330V).

(3) The mutation of the 'highest variant' side chains (items 239, 278, 322 and 331) caused an up to 5-fold loss of enzyme activity (e.g. W239L, T278M and C331L). The introduction of polar or charged side groups changed the overall polarity of the binding pocket, which virtually completely inactivated the corresponding enzyme (e.g. A322D, A322K and A322N).

(4) The observed increase in activation of a particular mis-cognate substrate was predictable based on the specificity code (Table 1). For example, the activity and specificity for Phe in the mutant PheA (1330V) remained unaffected, but this construct revealed a significant increase in the activation of the mis-cognate substrates Trp and Tyr. Interestingly, all Trp and Tyr activating domains have a valine at position 330. A similar relationship could also be observed for mutations towards the 'Leu (4)' codon (Table 1), as shown below.

'Phe' and 'Leu (4)' codons (Table 1) have approximately 60% homology and the main differences concern positions 278, 301 and 322 (note: two-out of three positions are 'highest variant' groups) ). Accordingly, it was assumed and subsequently also confirmed experimentally that point mutations in the direction of the 'Leu (4)' codon (→ T278M and A301 G) show increased activation of leucine (Table 3). Interestingly, for the mutant PheA (T278M) the catalytic efficiency for the unspecific activation of L-Leu is virtually unaffected (same Vmax), while at the same time the activity for the cognate substrates (D- and L-Phe) is approximately 3-fold lower learned. The exact opposite was observed for the mutant PheA (A301G). Here the Vmax for the activation of phenylalanine remained unaffected, while at the same time the catalytic activation of L-Leu increased almost twice. All of these changes in substrate specificity and activity were significant, reproducible, and well above the standard deviation. How only three differences in the codon allow a distinction between phenylalanine and leucine is currently still unclear, but at least for item 301 it can be assumed that due to differences in the spatial extent of a CH (Phe) versus CH3 group (Leu ) is discriminated in the δ position of the corresponding amino acid. All domains that activate substrates with sterically demanding γ or δ positions have the smallest possible side group (Gly) in position 301 (Tables 2 and 3).

To completely change the substrate specificity of PheA to leucine, we constructed the double mutant PheA (T278M / A301G). The codon of this mutant shows a similarity of approx. 80% to the 'Leu (4)' codon (see Table 1) and is in the phylogenetic tree in the immediate vicinity of Leu-activating domains to find (Figure 2; medium gray background). As shown in FIG. 5A, this construct actually actually preferentially activates Leu with a catalytic efficiency which is in no way inferior to or even exceeds the wild-type enzyme (activity of the mutant for L-Leu: kcat / Km = 86 mM- 1 min-1, wild-type activity for L-Phe: 69 mM-1 min-1; Table 4). An at least 30-fold increase in the L-Leu-specific activation efficiency was found. Amazingly, the double mutant was only slightly impaired in the activation of D-Phe, whereas the catalytic efficiency for the activation of the other stereoisomer, L-Phe, experienced a significant, approximately 7-fold reduction (FIG. 5A and Table 4). In contrast, for the wild-type protein PheA it could be shown that the interactions of the protein with both ligands, L- and D-Phe, are almost identical. As shown in Figure 1B, the benzyl ring of the side group of D-Phe is rotated about 30 ° relative to the L-Phe ring, resulting in a slight shift in the relative position of the β-C- but not the α-C -Atoms results.

We changed the specificity of an A domain according to the method of the invention, the crystal structure of which has not yet been elucidated. The starting point of this experiment was the finding that the codons of Asp and Asn activating domains found are very similar (approx. 80%; see Table 1), and the main differences are positions 322 (His vs. Glu; a 'highest Variant 'group) and 330 (Val vs. Ile; a' conditional variant 'group). Accordingly, we chose the Asp-activating A domain of the extensively characterized surfactin biosynthesis complex [Cosmina, P. et al. 1993, Mol. Microbiol. 8: 821-831] and attempted to change their specificity towards asparagine. For the wild-type protein AspA, a specific activation of L-Asp, but not L-Asn, with moderate activity (wild-type activity for L-Asp: kcat / Km = 10 mM-1 min-1; cf. FIG. 5B and Table 4) can be confirmed. In contrast, the point mutant AspA (H322E) revealed a high selectivity for L-Asn (FIG. 5B), although this change in specificity was associated with its decrease in catalytic activity (factor 10) (activity of the mutant for L-Asn: kcat / Km = 1mM-1 min-1; Table 4). Thus it could be shown with the construct AspA (H322E) that a complete change in substrate specificity can be achieved by introducing a single point mutation. In addition, the mutant AspA (H322E) also showed an approximately 5-fold increase in the activation efficiency for similarly large, aliphatic Asn analogues, such as, for example, Ile, Leu and Val.

The approximately 70-fold increase in the efficiency of L-Asn-specific activation was further increased by introducing a He330Val point mutation. As a result, catalytic activity of the now activating Asn domain was achieved, which was in the same order of magnitude as the original AspA domain.

Analysis of the codons of Glu and Gln activating domains showed that these domains also implement very similar codons (Table 1). The main difference concerns position 239 (Lys vs. Gin; a 'highest variant' group). We therefore chose the glu-activating domain of the surfactin biosynthesis complete (SrfA-A1) and tried to change the specificity towards glutamine. For the wild-type protein GluA, a specific activation of L-Glu, but not L-Gln, was confirmed with moderate activity (kcat / Km = 25 mM-1 min-1). In contrast, the point mutant GluA (K239Q) revealed a high selectivity for L-Gln, with at the same time only a slightly reduced catalytic activity (kcat / Km = 20 mM-1 min-1). This result could be expected due to the codons of the mutant and gin domains and shows that even with the construct GluA (K239Q) a complete change in substrate specificity could be achieved by introducing a single point mutation.

It is shown below that the changed substrate specificity of an A domain can also be used for the in vivo synthesis of a peptide with a modified primary structure. As an example, the gene fragment of the Asn-specific double mutant AspA (1330V / H322E) was transferred into the chromosome of the surfactin producer B. subtilis ATCC 21332 by means of homologous recombination. This integration was carried out using a two-stage recombination method [Schneider, A. et al. 1998, Mol. Gen. Genet. 257: 308-318] and led to the substitution of aspA 'against aspA (H322A / l330V)' in the srfA-B gene of the surfactin biosynthesis operon. The identity of this clone, B. subtilis Asn-5, was verified by PCR amplification of the mutated asp (H322E / l330V) 'gene fragment from its chromosome and subsequent DNA sequence analysis of the amplified product.

The biosurfactant surfactin is a lipoheptapeptide with a variable ß-hydroxy fatty acid content (6-9 methylene groups), which is produced at the transition to the stationary growth phase and is secreted into the culture medium. Accordingly, the B. subtilis strains ATCC 21332 and Asn-5 were fermented to detect the modified lipoheptapeptide, and wild-type and [Asn-5] surfactin were prepared from their culture supernatants. The subsequent HPLC-MS analysis showed that the [Asn-5] surfactin was produced in the same quantity compared to the wild type (1), (2) had a longer retention time, and (3) in the ESIMS analysis a mass difference of 1 Da showed. This mass difference was significant, and in the negative ion mode, 14 series of ions (variable fatty acid content) could be determined for the wild type at m / z 992 to 1,034 and for the mutant at m / z 991 to 1,033. This finding, together with the reduced hemolytic activity of the [Asn-5] surfactin, is to be seen as a clear indication of the substitution of a carboxy (Asp-5; wild type) for a carbamide group (Asn-5; mutant) in the surfactin. In summary, these results show that by changing the substrate specificity of an A domain within a catalytic NRPS template, the corresponding peptide product could also be changed in vivo.

Basically, the inventive method for the targeted synthesis of peptides looks like this:

With the aid of the method according to the invention, it is possible, starting from a naturally occurring, functionally active NRPS system, to specifically change each amino acid position of the non-tribosomal product peptide formed by directed point mutations within the A domains. For this purpose, the DNA of the A domain to be changed is first extracted from the chromosome of the producer organism amplified and cloned into an E. coli His-tag expression vector. With the help of this construct it is now possible to produce sufficient amounts of functionally active A-domain protein, to purify it and to examine it in vitro for its substrate specificity. In the following, point mutations can be introduced into the recombinant A-domain DNA, which lead to a change in the substrate specificity according to Table 1). To verify this change, the mutated A domains are first overexpressed in E. coli, purified and examined for their substrate specificity. If this enables the expected change in specificity to be confirmed, the mutated A-domain DNA is exchanged for the native A-domain DNA in the chromosome of the producer organism via homologous recombination. The producer strain with a mutated A domain thus generated can now be used to produce the modified non-tribosomal peptide.

The method according to the invention can also be used to influence the specificity and / or activity of known biologically active peptides. For this purpose, certain A domains are changed on the basis of the code found so that the peptides synthesized with the aid of the modified NRPS have the desired properties. An example of this is the improvement in solubility which can be achieved by replacing hydrophobic with hydrophilic amino acids and vice versa.

The results according to the invention also allow a targeted prediction of substrate specificities of already sequenced NRPS genes, the function of which has not yet been determined. This is becoming increasingly important in the context of the steadily increasing sequence quantities due to various genome sequencing projects. So it is with the method described here e.g. possible to predict the putative structure of the product peptide formed from the DNA sequence of an NRPS cluster. Subsequent structural-functional analysis of such genes is thus made much easier.

To derive the substrate specificity of an NRPS-A domain with known DNA Sequence, a sequence comparison is first carried out with the DNA sequence of PheA. In this way, the corresponding ten residues of the putative binding pocket of the A domain to be examined can be identified. These ten residues are now used as an independent amino acid sequence in alignments and phylogenetic studies with the ten residues of the binding pocket of all other known A domains. This results in a strict breakdown of the sequences with regard to the substrate specificity of the associated A domains. The substrate specificity of the A domain to be examined can thus be determined via the grouping behavior.

Using the knowledge according to the invention, it is also possible from the known DNA sequences for certain A, C and T domains, and if necessary. other domains - as described in Mahariel et al. (supra) - synthesize NRPS that are not derived from natural synthetases.

The following examples serve to explain the invention in more detail without restricting it.

Examples

example 1

PCR amplification and cloning of the PheA and AspA mutants. All PheA mutants were generated by inverse PCR by directed mutagenesis in the plasmid pPheA. The PCR amplification of the entire plasmid was carried out using the expand long-range PCR system (Boehringer Mannheim; Mannheim, Germany, catalog number: 1681842) in compliance with the manufacturer's protocol. The following 5'-phosphorylated oligonucleotides (MWG-Biotech, Ebersberg, Germany) were used (the mutated codons are underlined): (Seq ID-NO: 1) 3'-A236: 5'-ATC AAA AGA GAT GCT GGC A-3 '; (Seq ID-NO: 2) 5'-A236L: 5'-TTA TCT GTA TGG GAG ATG TTT ATG-3 '; (Seq ID-NO: 3) 3'-W239: 5'-TAC AGA TGC ATC AAA AGA G-3 "; (Seq ID-NO: 4) 5'-W239G: 5'-GGA GAG ATG TTT ATG GCT TTG TTA AC-3 '; (Seq ID-NO: 5) 5'-W239L: 5'-TTA GAG ATG TTT ATG GCT TTG TTA-3 '; (Seq ID-NO: 6) 3'-T278: 5'-AAT AAC AGT GAT TTC CTT TTG G-3 '; (Seq ID-NO: 7) 5'-T278M: 5'-ATG CTG CCA CCT ACC TAT GTA GTT-3 '; (Seq ID-NO: 8) 5'-T278Q: 5'-CAG CTG CCA CCT ACC TAT GTA GTT-3 '; (Seq ID-NO: 9) 3'-1299: 5'-TAA CGT TTG TAT CGA TAA AAT AC-3 '; (Seq ID-NO: 10) 5'-I299T: 5'-ACT ACA GCA GGC TCA GCT AC-3 '; (Seq ID-NO: 11) 3'-A301G: 5'-TCC TGT AAT TAA CGT TTG TAT CG-3 '; (Seq ID-NO: 12) 3'-A301V: 5'-AAC TGT AAT TAA CGT TTG TAT CG-3 '; (Seq ID-NO: 13) 5'-A301: 5'-GGC TCA GCT ACC TCG CCT-3 '; (Seq ID-NO: 14) 3'-A322: 5'-AAT GTA AGT TAC TTT CTC CTT C-3 '; (Seq ID-NO: 15) 5'-A322D: 5'-AAT GAT TAT GGC CCT ACG GAA ACA-3 '; (Seq ID-NO: 16) 5'-A322E: 5'-AAT GAA TAT GGC CCT ACG GAA ACA-3 '; (Seq ID-NO: 17) 5'-A322G: 5'-AAT GGC TAT GGC CCT ACG GAA ACA-3 '; (Seq ID-NO: 18) 5'-A3221: 5'-AAT ATT TAT GGC CCT ACG GAA ACA-3 '; (Seq ID-NO: 19) 5'-A322K: 5'-AAT AAG TAT GGC CCT ACG GAA ACA-3 '; (Seq ID-NO: 20) 5'-A322N: 5'-AAT TTG TAT GGC CCT ACG GAA ACA-3 '; (Seq ID-NO: 21) 5'-A322Q: 5'-AAT CAG TAT GGC CCT ACG GAA ACA-3 '; (Seq ID-NO: 22) 5'-A322S: 5'-AAT AGC TAT GGC CCT ACG GAA ACA-3 '; (Seq ID-NO: 23) 3'-1330: 5'-AGT TGT TTC CGT AGG GC-3 '; and (Seq ID-NO: 24) 5'-1330V: 5'-GTT TGT GCG ACT ACA TGG G-3 '.

The PCR amplificates were purified using the 'QIAquick-spin PCR purification' system (Qiagen; Hilden, Germany, catalog number: 28104), their cohesive ends were smoothed and then ligated intramolecularly. The addition of the restriction endonuclease Dpnl (Amersham / Buchler; Braunschweig, Germany; order no .: E0302Z), which requires a methylated recognition sequence, allowed a targeted degradation of the PCR template (pPheA; purified from a dam + - Escherichia coli strain) and thus avoiding false positive clones. Standard methods were used for all DNA manipulations and the preparation of the recombinant plasmids from E. coli XL-1-blue (Stratagene, Heidelberg, Germany, order no .: 200268) [Sambrook et al., 1989, Molecular Cloning: A Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY]. The following plasmids could be obtained by cloning the PCR amplificates: pPheA (A236L) (by using the oligonucleotides Seq ID-NO: 1 and Seq ID-NO: 2), pPheA (W239G) (Seq ID-NO: 3 plus Seq ID-NO: 4), pPheA (W239L) (Seq ID-NO: 3 plus Seq ID-NO: 5), pPheA (T278M) (Seq ID-NO: 6 plus Seq ID-NO: 7), pPheA (T278Q) (Seq ID-NO: 6 plus Seq ID -NO: 8), pPheA (l299T) (Seq ID-NO: 9 plus Seq ID-NO: 10), pPheA (A301 G) (Seq ID-NO: 11 plus Seq ID-NO: 13), pPheA (A301V ) (Seq ID-NO: 12 plus Seq ID-NO: 13), pPheA (A322D) (Seq ID-NO: 14 plus Seq ID-NO: 15), pPheA (A322E) (Seq ID-NO: 14 plus Seq ID-NO: 16), pPheA (A322G) (Seq ID-NO: 14 plus Seq ID-NO: 17), pPheA (A322l) (Seq ID-NO: 14 plus Seq ID-NO: 18), pPheA (A322K ) (Seq ID-NO: 14 plus Seq ID-NO: 19), pPheA (A322N) (Seq ID-NO: 14 plus Seq ID-NO: 20), pPheA (A322Q) (Seq ID-NO: 14 plus Seq ID-NO: 21), pPheA (A322S) (Seq ID-NO: 14 plus Seq ID-NO: 22) and pPheA (1330V) (Seq ID-NO: 23 plus Seq ID-NO: 24). The identity of all constructs was checked and confirmed by DNA sequencing on an 'ABI prism 310 Genetic Analyzer'(ABI; Weiterstadt, Germany, order no .: 310-A).

The DNA fragment coding for the aspartate-activating A domain from SrfA-B was amplified by means of PCR from the chromosomal DNA from Bacillus subtilis JH642 with the following oligonucleotides: (Seq ID-NO: 25) δ'-AspA: 5'- AAT CCA TGG CGA ACG TTC GGC TGT CTG-3 'and (Seq ID-NO: 26) 3'-AspA: 5'-AAT GGA TCC GGC CAA GGC CTT GCC-3'. The resulting PCR amplificate was purified (see above), digested with the restriction enzymes Ncol and BamHI (Amersham / Buchler; Braunschweig, Germany; order no .: E1160Z and E1010Y) and digested in the similarly prepared 'is-tag' vector pQE60 (Qiagen; Hilden, Germany, order no .: 33603). This cloning gave the plasmid pAspA, which could subsequently be used as a template for the generation of pAspA (H322E) by means of inverse PCR (see above): (Seq ID-NO: 27) 5'-AspA (H322E): 5'-TAA TGA GTA CGG CCC GAC AGA AGC-3 '(Seq ID-NO: 28) and 3'-AspA (H322E): 5'-ATA AAT TCG GTA TGT CCA TAC-3'. The resulting plasmid pAspA (H322E) subsequently served as a template for the generation of pAspA (H322E / l330V) by means of inverse PCR with the following primers (Seq ID-NO: 29) 5'-AspA (l330V): 5'-TAC CGG CCC ACA GAA GCA ACG GTC GGC-3 'and (Seq ID-NO: 30) 3'-AspA (I330V): 5'-CTG TGG GCC CGT ACT CAT TAA TAA ATT CGG-3'. The identity of all constructs was again checked and confirmed by DNA sequencing. The DNA fragment coding for the glutamate-activating A domain from SrfA-A was amplified by PCR from the chromosomal DNA of Bacillus subtilis JH642 with the following oligonucleotides: (Seq ID-NO: 31) 5'-GluA: 5'-TAT GGA TCC ATT GAT GAA TTA ACA CTG-3 'and (Seq ID- NO: 32) 3'-GluA: 5'-TAT GGA TCC GAT TGC TTT TTC AGT -3 '. The resulting PCR amplificate was purified (see above), digested with the restriction enzyme BamHI (Amersham / Buchler; Braunschweig, Germany; order no .: E1010Y) and with Bglll (Amersham / Buchler; Braunschweig, Germany; order no. No .: E1021Y) cloned his-tag vector pQE60 (Qiagen; Hilden, Germany, order no .: 33603). This cloning gave the plasmid pGluA, which could subsequently be used as a template for the generation of pGluA (K239Q) by means of inverse PCR (see above): (Seq ID NO: 33) 5'-GluA (K239Q): 5'-GTG CAG CAA ATC TTC GCG TCG CTT C-3 'and (Seq ID-NO: 34) 3'-GluA (K239Q): 5'-TGA CGC ATC AAA GTG GAA C-3'. The identity of all constructs was again checked and confirmed by DNA sequencing.

Example 2

Expression and purification of adenylation domains (wild type and mutants). The expression of the (mutated) gene fragments and the purification of the resulting His6 fusion proteins were described as described elsewhere [Mootz et al., 1997, J. Bacteriol. 179: 6843-6850]. The ligation into the BamHI site of pQE60 resulted in the fusion of the amino acid sequence "GSRSHHHHHH" to the carboxy terminus of the respective recombinant protein. Using SDS-PAGE it could be shown that most proteins could be purified almost homogeneously using Ni2 + affinity chromatography; however, two constructs were insoluble (see Table 3). Fractions containing the respective recombinant protein were pooled and dialyzed against assay buffer (50 mM HEPES, pH 8.0, 100 mM sodium chloride, 10 mM magnesium chloride, 2 mM dithioerythritol (DTE) and 1 mM EDTA). After adding 10% glycerol (v / v), the proteins could be stored at - 80 ° C without any noticeable loss of their catalytic activity. The protein concentration of the different solutions was determined using the calculated extinction coefficients at 280 nm (A280nm): 64060 M-1 cm-1 for PheA and all PheA mutants except PheA (W239Xaa), 58370 M-1 cm-1 for PheA (W239Xaa), and 39780 M-1 cm-1 for AspA and AspA (H322E) and 35490 M-1 cm-1 for GluA and GluA (K239Q).

Example 3

ATP pyrophosphate exchange reaction.

The ATP-pyrophosphate exchange reaction was carried out to determine both the catalytic activity and the specificity of the purified, recombinant A domains [Mootz & Marahiel, 1997, J. Bacteriol. 179: 6843-6850]. The specificity was checked with all 20 proteinogenic amino acids, as well as L-ornithine and D-phenylalanine. The reaction mixtures contained (final volume: 100 μL): 50 mM HEPES, pH 8.0, 100 mM sodium chloride, 10 mM magnesium chloride, 2 mM DTE, 1 mM EDTA, 0 to 2 mM amino acid, and 250 nM enzyme. The various reactions were initiated by adding 2 mM ATP, 0.2 mM sodium pyrophosphate and 0.15 μCi (16.06 Ci / mmol) sodium [32P] pyrophosphate (NEN / DuPont; Germany; order no .: NEX019) and for 10 min Incubated at 37 ° C. The exchange reactions could then be stopped by adding 0.5 mL stop solution: 1.2% (w / v) activated carbon, 0.1 M sodium pyrophosphate and 0.35 M perchloric acid. The activated carbon was separated by centrifugation, washed once with 1 ml of double-distilled water and resuspended in 0.5 ml of water. After adding 3.5 mL scintillation liquid (Rotiscint Eco Plus; Roth; Art.-No. 0016.2), the carbon-bound radioactivity could be determined on a scintillation measuring device (1900CA Tri-Carb 'liquid scintillation analyzer'; Packard).

Example 4

Computer analysis to determine the putative substrate binding pockets. Protein sequences from 160 A domains were obtained from publicly accessible databases (eg http // www.ncbi. Nlm.nih.gov/Entrez/nucleotide. Html) and on the approximately 100 amino acid region between the core motifs A4 and A5 reduced [Marahiel et al., 1997, Chem. Rev. 97: 2651-2673]. These sequences were subsequently aligned, using the MegAlign subroutine from the DNAStar program package, with respect to sequence homologies, the method 'clustal' was used in its basic setting. The sole purpose of this first alignment was to facilitate the subsequent assignment of the putative substrate binding pocket constituents. This assignment could ultimately be made by taking into account their position, relative to highly conserved sequence motifs (core motifs A4 and A5, as well as the motifs TPS 'and' GE ';structural'anchors'). All sequences were subsequently trimmed to the determined constituents of the binding pockets, and new alignments and phylogenetic studies were carried out with these only ten amino acid-sized regions. A grouping of the coding sequences according to the substrate specificity of their origin domains could be achieved with every method and every parameter set. The phylogenetic tree shown in FIG. 2 was determined using the Jotun Hein method using the following parameters: gap penalty 12, gap length penalty 6, and Ktuple 2.

Example 5

Integration of the AspA (H322E / l330V) coding gene fragment into the chromosome of B. subtilis AS20.

In order to demonstrate the potential of our method for the in vivo synthesis of directionally modified peptide antibiotics, the coding gene fragment of the AspA double mutant (Asp (H322E / l331V) was inserted into the srfA biosynthesis operon from B. subtilis [Cosmina, P. et al., 1993, Mol. Microbiol. 8: 821-831]. In order to construct the integration plasmid required for this purpose, the corresponding srfA-B gene fragment coding for the Asp-activating module (base pair 14195-18200; [Cosmina , P. et al. 1993, Mol. Microbiol. 8: 821-831]), by means of PCR from the chromosome of B. subtilis ATCC 21332 (underlined: restriction endonuclease cleavage sites): (Seq ID-NO: 35) 5 '-homo Asp (Clal): 5'-TAA ATC GAT GGA GGC TGC CAA GG -3'; and (Seq ID-NO: 36) 3'-homo Asp (Spei): 5'-TAA ACT AGT CAG TAA ATC CGC CCA GT -3 '. The resulting amplificate was purified with the restriction enzymes Clal and Spei (Amersham / Buchler; Braunschweig, Germany; order no .: E11034Y and E1086Z) digested and cloned into the equally prepared vector pBluescript SK (-) (Stratagene, Heidelberg, Germany, order no .: 212206). This cloning provided the plasmid pSK-homoAsp, from which subsequently an approximately 1.3 kb aspA fragment (base pair 15212.) with the restriction endonucleases EcoRI and Pstl (Amersham / Buchler; Braunschweig, Germany; order no .: E1040Y and E1073Y) -16559; [Cosmina, P. et al. 1993, Mol. Microbiol. 8: 821-831]) was cut out and replaced with the complementary, double mutated fragment from pAspA (H322E / 1333V). The identity of the resulting integration plasmid pSK-homoAsp (H322E / 1330V) could be confirmed by DNA sequencing.

A proven co-transformation technique ('congression') was used to integrate the mutated aspA 'fragment into the surfactin biosynthesis operon [Schneider, A. et al. 1998, Mol. Gen. Genet. 257: 308-318]. The B. subtilis strain AS20 [Schneider, A. et al. 1998, Mol. Gen. Genet. 257: 308-318], in which the aspA 'fragment (base pair: 15245-17177; [Cosmina, P. et al. 1993, Mol. Microbiol. 8: 821-831]) is deleted and substituted for the chloramphenicol transferase gene has been. B. subtilis AS20 was made competent for DNA uptake and co-transformed with the integration vector pSK-homoAsp (H322E / l330V) and the helper plasmid pNEXT33A [Schneider, A. et al. 1998, Mol. Gen. Genet. 257: 308-318]. The latter helper plasmid initially mediated a positive selection via its neomycin resistance gene, and the clones obtained could then be examined for the loss of their chloramphenicol cassette and thus the integration of the aspA (H322E / l330V) 'gene fragment. The identity of the constructed clone B. subtilis Asn-5 could be verified by PCR amplification of this area from the chromosome of the selected clones and subsequent DNA sequencing.

Example 6

Surfactin preparation and analysis.

The preparation of the surfactin derivative from B. subtilis Asn-5 was carried out according to a

Standard procedures [Schneider, A. et al. 1998, Mol. Gen. Genet. 257: 308-318]. to

The methanolic solution obtained was analyzed using a ΗP1100 Series

LC / MSD 'HPLC system using a PepRPC HR 5/5 column column (Pharmacia, Freiburg, Germany, order no .: 18-0383-01). The separation of the individual components was achieved at a flow rate of 0.7 mL min-1 by applying a linear gradient from 35% to 70% acetonitrile (v / v) in 0.1% trifluoroacetic acid (v / v). The detection was carried out at a wavelength of 214 nm and by means of an online mass spectrometric analysis of the eluted substances. As a control, a preparation from the wild-type surfactin producer B. subtilis ATCC 21332 was processed and analyzed analogously.

Description of the figures:

Figure 1: Structural basis for the detection and activation of phenylalanine (from Conti et al. [Conti et al., 1997, EMBO J. 16: 4174-4183]). (A) The band diagram shows that PheA consists of two folding domains, a large N-terminal (below) and a smaller C-terminal (above). AMP and the substrate amino acid phenylalanine are shown in black. (B) The binding pocket for phenylalanine is formed from ten amino acids. Asp235 and Lys517 mediate electrostatic interactions (dashed lines) with the α-amino and α-carboxyl group of the substrate. The actual binding pocket, which enables the specific recognition of the phenylalanine side group, is formed by Ala236, Ile330 and Cys331 on the one hand, and Ala322, Ala301, 1299 and Thr278 on the other. Both sides are kept separate at the lower end by the indole ring of the Trp239. This architecture allows the PheA binding pocket to accept both stereoisomers, L-Phe (light gray) and D-Phe (dark gray) without noticeable change in conformation.

Figure 2: Phylogenetic tree constructed with the ten specificity-mediating amino acids of the A domains. The phylogenetic examination reveals a grouping of the codons determined, according to the specificity of their domains of origin (light gray boxes). This confirms the correctness of the concept of structural homology between A domains and shows that the ten amino acids determined in each case really form the codon for the recognition of the cognate substrate amino acid. Examples are given which show that the specificity of newly discovered domains can be predicted using this technique and that the grouping is guided by the experimentally observed specificity (dark gray background). This also applies to the constructed mutants PheA (T278M / A301G) and AspA (H322E), which show a Leu or Asn specificity (medium gray background) in the ATP-pyrophosphate exchange reaction.

Figure 3: Schematic representation of the postulated binding pockets of three adenylation domains. The codons determined from three different substrates ('Asp', Orn (2) 'and Nal (3)'; see Table 1) were projected into the representation of the binding pocket of PheA shown in FIG. 1B. Aliphatic (light gray), polar (striped), acidic (white) and basic (dark gray) side groups are shown schematically. In all three cases, Asp235 and Lys517 mediate interactions with the α-amino and α-carboxyl group of the respective substrate, while all other residues (Xaa236 to Xaa331) mediate the actual recognition of the side chain of the substrate. A perfect correlation between the polarity of the binding pocket and the substrate can be observed for all the examples shown.

Figure 4: observed variability of the constituents of different substrate binding pockets. The constituent amino acids of the various positions in 160 postulated substrate binding pockets were examined for their variability. The proportional distribution of the nature of the substrates in the examined A domain is shown in the middle gray box. The light gray boxes, which are connected to each position of the binding pocket, indicate the number of different residues observed (frequency: = 1%) and the proportional occurrence of hydrophobic, polar, acidic and basic side groups at these points. Based on the results shown, the ten constituent amino acids can be classified into three subgroups: (1) Positions 235 (Asp: acid, white) and 517 (Lys: basic, dark gray) are 'invariant'. (2) Items 236, 301 and 330 are only 'conditionally variant'. They reveal a below average occurrence of charged and polar side groups and the vast majority (93%) of the A domains examined use aliphatic residues (gray) at these positions. As a 'highest variant' they can Positions 239, 278, 299, 322 and 331 are considered, which show a high degree of variability with regard to the use of different amino acids.

Figure 5: Targeted change in the substrate specificity of PheA and AspA. The substrate specificity of wild type (white) and mutants (different shades of gray) were examined with the help of ATP-pyrophophate exchange reactions. The substrates used are shown on the abscissa and the maximum observed activity for each protein was set at 100%. (A) Point mutations in PheA towards the 'Leu (4)' codon (T278M and A301 G) slightly increase the site specificity for Leu, while the preferred substrate is still Phe. In contrast, the corresponding double mutant (PheA (T278M / A301G); dark gray) preferentially activates Leu with a catalytic activity that is in no way inferior to the efficiency of the wild-type enzyme PheA for Phe. (B) A H322E point mutation in AspA was sufficient to completely shift the substrate specificity of the mutant from Asp to Asn. The observed activation patterns of the mutants correspond to the position of their codons in the phylogenetic tree shown in FIG. 2 (medium gray background).

Figures 6-9 stand for Tables 1-4.

Claims

Claims:

1. A method for the targeted non-ribosomal synthesis of peptides of a desired structure, wherein in a given DNA sequence coding for a non-ribosomal peptide synthetase, one or more A-domain coding sections in accordance with the non-ribosomal code presented in Table 1) be changed so that the expression product of the changed DNA sequence expresses the peptide of the desired structure.

2. The method according to claim 1, wherein in at least one A domain to be changed at most 10, preferably at most 6, particularly preferably at most 3, very particularly preferably one amino acid is exchanged.

3. The method according to claim 1, wherein in the given DNA sequence A domains coding sections are added or removed for the purpose of producing longer or shorter peptides.

4. The method according to at least one of claims 1 to 3, wherein the synthesized peptides have antibiotic, immunosuppressive, cytostatic, antiviral, antihelmitic, fungicidal or surface-active action.

5. The method according to at least one of claims 1 to 4, wherein at least one modified non-ribosomal peptide synthetase is contained in a host organism.

6. The method according to claim 5, wherein the host organism is selected from the group: bacteria, fungi, yeasts, particularly preferred are Bacilli, Actinomycetes, Myxobacteria, Pseudomonas and Escherichia Coli, very particularly preferred is Bacillus subtilis.

7. The method of claim 5, wherein the host organism is a plant.

8. The method of claim 5, wherein the host organism is a mammal.

9. The method of claim 8, wherein the host organism is a human.

10. The method according to at least one of claims 1 to 4, wherein the modified a non-ribosomal peptide synthetase is used in isolated form (in vitro).

11. The method of claim 10, wherein the altered non-ribosomal peptide synthetase is immobilized.

12. The method of claim 11, wherein the altered non-ribosomal peptide synthetase is bound to a support.

13. DNA sequence coding for a non-ribosomal peptide synthetase, in which one or more A domains have been changed using the non-ribosomal code in Table 1 such that the expression product of the changed DNA sequence is the peptide of a predetermined structure, which is preferably not corresponds to a natural peptide.

14. DNA sequence according to claim 13, wherein in at least one A domain to be changed at most 10, preferably at most 6, particularly preferably at most 3, very particularly preferably one amino acid is exchanged.

15. Use of the DNA sequence according to one of claims 13 or 14 for the synthesis of peptides in one of the methods according to claims 1 to 11.

16. Peptide obtainable using the non-ribosomal code according to Table 1, Table 1 being part of the claim.