PRODUCTION OF BETA-LACTAM ANTIBIOTICS

Field of the invention

The present invention relates to the production of β-lactam antibiotics.

Background of the invention β-Lactam antibiotics are the largest family of secondary metabolites produced in nature by microorganisms. The most important classes of the β-lactam antibiotics both clinically and economically are the penicillins (penam) and cephalosporins (cephem). Their biosynthesis occurs via a complex pathway of enzymatic steps. The first two steps are the key steps in the biosynthetic pathways of the penam and cephem classes of β-lactam antibiotics. After these two steps the biosynthetic pathways to the penicillins and cephalosporins diverge. The first step in the biosynthesis of the penicillin, cephalosporin and cephamycin antibiotics is the condensation of the L-isomers of three amino acids, L-α-amino adipic acid (A), L-cystein (C) and L-valine (V) into a tripeptide, δ-(L-α-aminoadipyl)-L- cysteinyl-D-valine or ACV. This step is catalyzed by δ-(L-α-aminoadipyl)-L-cysteinyl-D-valine synthetase or ACVS. In the second step, the tripeptide ACV is oxidatively cyclised by the action of lsopenicillin N synthase (hereinafter referred to as IPNS) or cyclase. The product of this reaction is lsopenicillin N; this compound contains the typical β-lactam and thiazolidine ring structures and possesses antibacterial activity. From lsopenicillin N the penicillins G or V are formed by exchange of the hydrophilic α-aminoadipyl side chain by a hydrophobic side chain. The side chains commonly used in industrial processes are either phenylacetic acid (PA), yielding penicillin G, or phenoxyacetic acid (POA), yielding penicillin V; this exchange reaction is catalyzed by the enzyme acyltransferase (AT).

Due to the substrate specificity of the enzyme AT, it is not possible to exchange the α-aminoadipyl side chain for any side chain of interest, although it was shown that adipic acid and certain thio-derivatives of adipic acid could be exchanged (see WO 95/04148 and WO 95/04149). In particular, the side chains of industrially important penicillins and cephalosporins cannot be directly exchanged via AT. Consequently, most of the β-lactam antibiotics presently used are prepared by semi-synthetic methods. These semi-synthetic β- lactam antibiotics are obtained by modifying an N-substituted β-lactam product by one or more chemical and/or enzymatic reactions. These semi-synthetic

methods have the disadvantage that they include many steps, are not environmentally friendly and are rather costly. It would therefore be highly desirable to avail of a completely fermentative route to β-lactam antibiotics, for instance to amoxicillin, ampicillin, cefadroxil and cefalexin. Various options can be thought of for a completely fermentative route to semisynthetic penam and cephem antibiotics.

For instance, one could focus at exchanging the α-aminoadipyl side chain of lsopenicillin N for the appropriate side chain of interest, e.g. the α-amino-p- hydroxyphenylacetyl side chain in case of amoxicillin. This would require modification of the substrate specificity of the enzyme AT, since the native enzyme has a rather narrow substrate specificity and is not capable of catalyzing such an exchange. In addition, this would require modification of the enzyme CoA ligase that activates the side chain to be exchanged.

Alternatively, one could focus at modifying the first two steps in the penicillin biosynthetic route in such a way that amoxicillin is directly synthesized and secreted. However, this would require substantial modification of the ACVS and IPNS enzymes.

For instance, for amoxicillin, it firstly would require the production of a tripeptide producing the amoxicillin side chain, i.e. the tripeptide D-p-hydroxyphenylglycyl-L-cysteinyl- D-valine, instead of ACV. Secondly, it would require an enzyme that is able to cyclise this tripeptide.

ACVS is a non-ribosomal peptide synthetase (NRPS) that catalyses the formation of the tripeptide LLD-ACV (δ-(L-α-aminoadipyl)-L-cysteinyl-D-valine). In this tripeptide, a peptide bond is formed between the δ-carboxylic group of L-α-aminoadipic acid and the amino group of L-cystein, and additionally the stereochemical conformation of valine is changed from L to D.

In recent years, several laboratories and institutions probed the options for the targeted engineering of NRPS. As main approaches, domains and modules were exchanged and as a consequence new specificities were introduced, leading to changed peptide products. In most cases, the engineering approaches were restricted to the enzyme fragments of the same organism (or even the same NRPS), severely limiting the options for enzyme engineering. A main problem of recent NRPS engineering addressed in the literature is the change of stereochemistry of an amino acid of the peptide, i.e. a change from L- to D- stereochemistry. A peptide bond-forming condensation domain immediately

downstream of an epimerization domain is D-specific for the peptidyl or aminoacyl donor and L-specific for the amino acyl acceptor. Such a condensation domain is represented as ^D Cι_. An ^L Cι_ domain at this position does not yield condensation of the donor and acceptor moieties (Clugston S.L. et al. 2003, Biochemistry, 42, 12095-12104). A further problem that may limit options for engineering is that C and A domains are regarded as an inseparable couple (Mootz H. D. et al. 2000, Proc. Natl. Acad. ScL USA, 97, 5848-5853).

It is now surprisingly found that engineering of ACVS is feasible to provide an engineered enzyme that is capable of catalysing the formation of a tripeptide HpgCV or PgCV, wherein in the N-terminal peptide bond an α-carboxylic group is used instead of the δ-carboxylic group that is used in the natural tripeptide ACV and, additionally, that is capable of modifying the L stereochemical configuration of the first amino acid to a D configuration.

Several groups have further investigated the substrate specificity of the enzyme IPNS. See for reviews on this topic Baldwin and Bradley, Chem. Rev. 1990, 90, 1079-1088 and Huffman et al. J. Med. Chem. 1992, 35, 1897-1914. For instance, a m-COOH-DL- phenylglycyl-L-cysteinyl-D-valine and a p-hydroxyphenylacetyl-L-cysteinyl-D-valine substrate were mentioned not to result in antibiotic activity (Huffman et al, supra).

It is also surprisingly found that the native enzyme IPNS is capable of acting on the tripeptides Hydroxyphenylglycyl-Cysteinyl-Valine (HpgCV) and Phenylglycyl-Cysteinyl- Valine (PgCV), more particularly on the DLD-variants of these tripeptides. The finding of this feature of IPNS opens up the road to further developments, among which the screening of IPNS variants to isolate IPNS molecules with an improved activity on these non-native peptides. The findings of the present invention for the first time enable a completely fermentative production of antibiotics that formerly could be obtained via semi-synthetic ways only.

Detailed description of the invention

Thus, in a first aspect, the present invention provides a process for the production of an N-α-amino-hydroxyphenylacetyl or an N-α-aminophenylacetyl β-lactam antibiotic comprising contacting a tripeptide hydroxyphenylglycyl-cysteinyl-valine (HpgCV) or a tripeptide phenylglycyl-cysteinyl-valine (PgCV) with an IPNS to effect formation of the N-α-amino-hydroxyphenylacetyl or the N-α-aminophenylacetyl β-lactam antibiotic.

The present invention surprisingly shows for the first time that the enzyme IPNS is able to convert the non-native tripeptide substrates HpgCV or PgCV, respectively, to β- lactam compounds with a N-α-amino-hydroxyphenylacetyl or an N-α-aminophenylacetyl side chain, respectively, such as amoxicillin or ampicillin, respectively. According to the invention, an "IPNS enzyme" or an "enzyme with IPNS activity" is an enzyme that cyclises a tripeptide HpgCV or PgCV to a penam antibiotic molecule. Such IPNS enzymes typically have a degree of identity to the Emericella (Aspergillus) nidulans IPNS amino acid sequence of SEQ ID NO: 1 of at least 50%.

The amino acids hydroxyphenylglycine (Hpg) and phenylglycine (Pg) in these tripeptides may be in the D- as well as the L-form, but preferably are in the D-form. The hydroxyphenylglycine preferably is p-hydroxyphenylglycine.

The detection of this non-native tripeptide-cyclising activity of IPNS is done using a bioassay and/or using LC/MS analysis.

The development of the bioassay required various choices to be made. For instance, a suitable microorganism needed to be selected for testing of antibiotic activity. In addition, a suitable IPNS needed to be selected, since IPNS enzymes from different species appear to differ in specific activity and stability.

Now that the non-native tripeptide cyclising activity of IPNS as mentioned above has been detected, it is possible to optimise the reaction conditions and to screen IPNS variants for enzymes with improved specific activity and/or altered substrate specificity.

The in vitro cyclisation of the non-native tripeptide to give a penam antibiotic typically is done using the reaction conditions as mentioned below.

The pH of the reaction mixture may be between 6 and 8, using any suitable buffer.

The reaction mixture should further contain a suitable amount of Fe(II) ions. Finally, presence of a reducing agent, for instance DTT or TCEP (Tris(2-carboxyethyl)phosphine), is necessary to keep the precursor tripeptide in a reduced state. Preferably, the precursor tripeptide is pretreated with such a suitable reducing agent.

Optionally, the formed penam antibiotic, such as amoxicillin or ampicillin, may be further converted to produce a cephem antibiotic compound. This conversion requires at least the enzyme expandase, for instance to form cefadroxil or cefalexin. Optionally, other cephem compounds may be formed by further enzyme conversions using for instance hydroxylase and acetyl transferase enzyme activities or hydroxylase and carbamoyl transferase enzyme activities. The enzymes necessary for these conversions are suitably

obtainable from cephem-producing microorganisms, such as Streptomyces clavuligerus, Nocardia lactamdurans or Acremonium chrysogenum. See for a recent review Liras and Martin, International Microbiology (2006) 9: 9-19.

In a preferred embodiment of the invention, an IPNS enzyme is used as is described herein below.

In one embodiment, custom made HpgCV and PgCV tripeptides are used as starting compounds in the process. In another embodiment, the tripeptides HpgCV and PgCV are produced in a process comprising i) contacting the amino acids hydroxyphenylglycine (Hpg) or phenylglycine (Pg), cystein (C) and valine (V) with a non- ribosomal peptide synthetase (NRPS) to effect formation of the tripeptide hydroxyphenylglycyl-cysteinyl-valine (HpgCV) or phenylglycyl-cysteinyl-valine (PgCV).

The NRPS for use in this embodiment is a non-natural NRPS with a modular structure comprising three modules, a first module M1 specific for Hpg or Pg, a second module M2 specific for C and a third module M3 specific for V, as is described herein below. The overall process for the production of an N-α-amino-hydroxyphenylacetyl or an

N-α-aminophenylacetyl β-lactam antibiotic from its precursor amino acids thus comprises i) contacting the amino acids hydroxyphenylglycine (Hpg) or phenylglycine (Pg), cystein (C) and valine (V) with a non-ribosomal peptide synthetase (NRPS) to effect formation of the tripeptide hydroxyphenylglycyl-cysteinyl-valine (HpgCV) or phenylglycyl-cysteinyl-valine (PgCV), and ii) contacting said tripeptides with IPNS to effect formation of- the N-α-amino- hydroxyphenylacetyl or the N-α-aminophenylacetyl β-lactam antibiotic.

The processes may be performed in vivo using a suitably engineered microbial strain, as described herein below.

In a second aspect of the invention, variant IPNS polypeptides are provided that are modified as compared to a parent IPNS polypeptide and that have an improved cyclising activity on a tripeptide HpgCV or PgCV as compared to the parent IPNS.

An "improved activity" of a variant IPNS enzyme according to the invention is an activity that ensures the production of at least 2 times the amount of antibiotic activity from the precursor tripeptide as compared to the parent IPNS, preferably at least 5 times the amount, more preferably at least 10 times the amount. If the parent IPNS has an undetectable activity on the precursor peptide, an improved activity encompasses any measurable activity above the detection limit.

Suitable positions for modification in a parent IPNS may be selected based on the criteria of a) creating space in the active center, and/or b) weakening of the binding of the C- terminus to the active site. The C-terminus of the IPNS molecule acts as a quasi-substrate when the enzyme is not loaded with a substrate and thus competes with the substrate for active site binding. Upon approach of the substrate to the active center, the C-terminus withdraws, creating space for the substrate. It may thus be beneficial to shift this competition in favour of the binding of the substrate.

The parent IPNS may originate from any suitable microbial source, as mentioned herein below. A particular variant IPNS enzyme of the invention, when aligned to an IPNS of

SEQ ID NO: 1 , is modified as compared to a parent IPNS enzyme in at least one of the positions 75, 91 , 183, 185, 287, 321 , 324, 331 , and is capable of converting the tripeptide HpgCV or PgCV, respectively, to the antibiotic amoxicillin or ampicillin, respectively. Modifications at these positions create more space in the peptide side chain binding pocket to accommodate the more bulky Hpg and Pg side chains of the HpgCV and PgCV tripeptides as compared to the linear and flexible α-amino-adipyl side chain of ACV. A preferred variant contains a modification on all of the above positions. Preferred modifications at these positions are the modifications 75LVT, 91 FH, 183ACGTVL, 185STGAC, 287HSDQMK, 321 FSAVTQEM, 324NDSTVA, 331 GASCVDN. Especially preferred modifications are 75VT, 183AG, 185GA, 287HQ, 321AVTM, 324NSA, 331ASVDN.

A modification may be a substitution of a particular amino acid at the indicated position(s) for a different amino acid, may be an insertion of an amino acid at the indicated position(s) or may be a deletion of an amino acid from the indicated position(s).

Another particular variant IPNS enzyme of the invention, when aligned to an IPNS of SEQ ID NO: 1 , is modified as compared to a parent IPNS enzyme in at least position 185, and optionally in at least one of the following positions, using the position numbering of SEQ ID NO: 1 : 91 , 104, 183, 190, 321 , 324, 325, 331 , and is capable of converting the tripeptide HpgCV or PgCV, respectively, to the antibiotic amoxicillin or ampicillin, respectively. Modifications at these positions shift the competition between the substrate and IPNS -C- terminal residues -Q330-T331 for binding in the active site in favour of the substrate by weakening the binding of the C-terminus to the active site and/or by stabilising the

conformation in which the active site is available for substrate binding. It was surprisingly shown that the modification in at least position 185 confers a highly improved cyclisation activity to the IPNS enzyme containing the modification as compared to a parent IPNS not containing the modification.

A preferred variant according to the invention comprises the modification 185RKH, more preferably the modification 185RK, most preferably the modification 185R. The original amino acid at position 185 depends on the parent IPNS that is used, for instance may be S, T or V.

In one embodiment, the 185RKH modification is combined with a modification in at least one of the positions 91 , 104, 183, 190, 321 , 324, 325, 331 , as mentioned in the table below. In particular, R and K can interact favourably with the C-terminal T331 when the C-terminal region N-G-Q-T331 has adopted the active conformation. Surprisingly, the interaction between 185RK and T331 is very beneficial for increasing the activity for HpgCV and PgCV. Likely 185RK favours the formation of the active conformation using HpgCV and PgCV as substrates.

^* At those positions where an amino acid is specified, the particular amino acid is conserved within all native IPNS enzymes known to date.

In the present invention, a denotation like e.g. "185RK" means that the amino acid in position 185 of the parent IPNS in question is substituted with either R or K, or that at position 185 of the parent IPNS either R or K are inserted when in the parent IPNS no amino acid is present at this position. The nature of the original amino acid residue will depend on the parent IPNS that is used. A denotation like e.g. "VST185RK"

means that a specific amino acid residue at position 185 present in the parent IPNS, in this example V, S or T, is substituted for a different amino acid, in this example R or K.

A suitable parent IPNS is an IPNS polypeptide obtainable from a fungal or bacterial organism capable of producing β-lactams. A preferred parent IPNS is an IPNS obtainable from Cephalosporium acremonium, Penicillium chrysogenum, Aspergillus (Emeήcella) nidulans, Streptomyces jumonjinensis, Nocardia lactamdurans, Streptomyces microflavis, Lysobacter lactamgenus, Flavobacterium species, Streptomyces clavuligerus, Streptomyces griseus and/or Streptomyces cattleya. An especially preferred parent IPNS is obtainable from Streptomyces clavuligerus and/or Aspergillus (Emericella) nidulans. Typically, such a parent IPNS is a polypeptide with IPNS activity that has a degree of identity of at least 50%, preferably at least 55%, more, preferably at least 60%, more preferably at least 65%, more preferably at least 70%, more preferably at least 75%, or most preferably at least 80% to the amino acid sequence of SEQ ID NO: 1. In addition to the modifications as set out above, preferably the modification

185RKH, the IPNS polypeptide may comprise additional modifications that concern positions in the polypeptide wherein a modification does not substantially affect the folding or activity of the polypeptide. Typically, such modifications may be conservative modifications, for instance substitutions wherein a non-polar, polar uncharged, polar charged or aromatic amino acid is substituted for a different amino acid from the same category, or may be due to intra-strain or intra-species variation. Polypeptides having such additional modifications typically have a degree of identity to the sequence of SEQ ID NO: 1 of at least 50%.

Thus, in one embodiment, the polypeptide having IPNS activity and comprising at least one of the modifications as set out above has a degree of identity of at least 50%, preferably at least 55%, more, preferably at least 60%, more preferably at least 65%, more preferably at least 70%, more preferably at least 75%, or most preferably at least 80% to the amino acid sequence of SEQ ID NO: 1.

The terms "homology" or "percent identity" are used interchangeably herein. For the purpose of this invention, it is defined here that in order to determine the percent identity of two amino acid sequences, the sequences are aligned for optimal comparison purposes (e.g., gaps can be introduced in each sequence for optimal alignment). The amino acid residues at corresponding amino acid positions are then

compared. When a position in the first sequence is occupied by the same amino acid residue as the corresponding position in the second sequence, then the molecules are identical at that position. The percent identity between the two sequences is a function of the number of identical positions shared by the sequences (i.e., % identity = number of identical positions/total number of positions (i.e. overlapping positions including gaps) x 100). Preferably, the two sequences are the same length.

The skilled person will be aware of the fact that several different computer programs are available to determine the homology between two sequences. For instance, a comparison of sequences and determination of percent identity between two sequences can be accomplished using a mathematical algorithm. In a preferred embodiment, the percent identity between two amino acid sequences is determined using the Needleman and Wunsch (J. MoI. Biol. (48): 444-453 (1970)) algorithm which has been incorporated into the GAP program in the Accelrys GCG software package (available at http://www.accelrys.com/piOducts/gcg/), using either a Blossom 62 matrix or a PAM250 matrix, and a gap weight of 16, 14, 12, 10, 8, 6, or 4 and a length weight of 0.5, 1 , 2, 3, 4, 5, or 6. The skilled person will appreciate that all these different parameters will yield slightly different results but that the overall percentage identity of two sequences is not significantly altered when using different algorithms. Preferably, the matrix is a Blossom 62 matrix with a gap weight of 10.0 and a length weight of 0.5. The protein sequences of the present invention can further be used as a

"query sequence" to perform a search against public databases to, for example, identify other family members or related sequences. Such searches can be performed using the blastp, psi-blast, phi-blast and tblastn programs (version 2.0) of Altschul, et al. (1990) J. MoI. Biol. 215:403-10. When utilizing blastp, psi-blast, phi-blast and tblastn programs, the default parameters of the respective programs (e.g., blastp, psi-blast, phi-blast and tblastn programs) can be used. See the homepage of the National Center for Biotechnology Information at www-ncjg^

In a third aspect, a polypeptide is provided that is a Non-Ribosomal Peptide Synthetase comprising three modules, a first module M1 specific for Hpg or Pg, a second module M2 specific for C and a third module M3 specific for V. With the term "specific for" an amino acid to characterise a module is meant that the particular module enables incorporation of the indicated amino acid. The first module M1 enables incorporation of a first amino acid L-p-hydroxyphenylglycine or L-phenylglycine and, preferably, its conversion

to the corresponding D-amino acid. The second module M2 enables incorporation of the amino acid L-cystein while being coupled to the amino acid Hpg or Pg. In particular, when the amino acid Hpg or Pg is in its D-form, the M2 module specific for C comprises a ^D C _L domain that is fused to an A domain that is heterologous thereto. The third module M3 enables incorporation of the amino acid L-valine and its conversion to the corresponding D- amino acid. In this way, the peptide synthetase catalyzes the formation of a DLD-tripeptide Hydroxyphenylglycyl-Cysteinyl-Valine (HpgCV) or Phenylglycyl-Cysteinyl-Valine (PgCV) from its L-amino acid precursors Hpg or Pg, C and V.

The term "module" as used in the present invention defines a catalytic unit that enables incorporation of one peptide building block, usually an amino acid, in the product, usually a peptide, and may include domains for modifications like epimerisation and methylation.

Each module of an NRPS is composed of so-called "domains", each domain being responsible for a specific reaction step in the incorporation of one peptide building block. Each module at least contains an adenylation domain (A domain), responsible for recognition and activation of a dedicated amino acid, and a thiolation domain (T or PCP domain), responsible for transport of intermediates to the catalytic centers. The second and further modules further contain a condensation domain (C domain), responsible for formation of the peptide bond, and the last module further contains a termination domain (TE domain), responsible for release of the peptide. Optionally, a module may contain additional domains such as an epimerization domain (E-domain), responsible for conversion of the L-form of the incorporated amino acid to the D-form. See Sieber S. A. et al. 2005, Chem. Rev., 105, 715-738 for a review of the modular structure of NRPS.

A suitable source for the M1 module of the hybrid peptide synthetase is an NRPS enzyme catalyzing formation of a peptide comprising the amino acid X, wherein X is Hpg or Pg, to be incorporated as first amino acid in the XCV tripeptide. Thus, a suitable M1 module is selected taking into account the nature of the amino acid to be incorporated as first amino acid of the tripeptide. In particular, the A domain of a module determines selectivity for a particular amino acid. Thus, an M1 module may be selected based on the specificity of an A domain for the amino acid to be incorporated. Such a selection may occur according to the specificity determining signature motif of A domains as defined by Stachelhaus T. et al. 1999, Chem. & Biol., 8, 493-505.

The M1 module does not need to contain a C-domain and a TE domain, as being the first module of the NRPS. Thus, if present in the source module, a C and/or a TE domain may suitably be removed, to obtain a first module M1 without a C and/or a TE domain. In addition to an A and a T domain, the module M1 of the NRPS should contain an E-domain, if an L-amino acid needs to be converted to a D-amino acid. Thus, if not present in the source module, an E domain is fused to the T domain of the source module, to obtain a first module M1 containing an A, T and E domain.

In general, the specificity of module M1 of the NRPS for p-hydroxyphenylglycine or phenylglycine may be judged by experimental data and/or may be based on the specificity determining signature motif of A domains published by Stachelhaus T. et al. 1999, Chem. & Biol., 8, 493-505.

Preferably, a first module M1 with p-hydroxyphenylglycine specificity is obtainable from a CDA (Calcium-Dependent Antibiotic) Synthetase, in particular is the sixth module of a CDA synthetase (numeration of CDA Synthetase modules is given as published by Hojati Z. et al. 2002, Chem. & Biol., 9, 1175-1187) . Preferably, the CDA synthetase is obtainable from Streptomyces coelicolor. Alternatively, Hpg-specific modules may be obtainable from a Chloroerenomycin Synthetase, in particular are the fourth and fifth module of a Chloroerenomycin Synthetase, preferably a Chloroerenomycin Synthetase obtainable from Amycolatopsis orientalis (Trauger J.W. et al. 2000, Proc. Nat. Acad. Sci. USA, 97, 3112- 3117); or from a Complestatin Synthetase, in particular is the seventh module of a Complestatin Synthetase, preferably a Complestatin Synthetase obtainable from Streptomyces lavendulae (Chiu HT. et al. 2001 , Proc. Nat. Acad. Sci. USA, 98, 8548-8553). All these modules have specifity for L-Hpg and convert it into the D-stereoisomer.

Preferably, a first module M1 with phenylglycine specificity is obtainable from a Pristinamycin Synthetase, in particular is the C-terminal module of the SnbD protein of Pristinamycin Synthetase, as published in Thibaut, D. et al. 1997, J. Bad, 179, 697-704. Preferably, the Pristinamycin Synthetase is obtainable from Streptomyces pristinaspiralis.

The C-terminal source module from Pristinamycin Synthetase contains a TE domain and does not contain an E domain. To prepare a module functioning as a first module in the peptide synthetase of the invention, the TE domain suitably is removed from the C-terminal source module and an E domain is fused to the T domain of the thus- modified C-terminal module. An E domain may be obtainable from any suitable NRPS, for instance from another module of the same NRPS enzyme, or also from a module of a

different NRPS enzyme with similar (e.g. p-hydroxyphenylglycine or phenylglycine) or different amino acid specificity of the adenylation domain. Preferably, the E domain is obtainable from a CDA Synthetase from Streptomyces coelicolor, preferably from the sixth module, as specified above. Thus, in this embodiment,the module M1 of the NRPS is a hybrid module.

The second module M2 of the peptide synthetase should enable incorporation of the amino acid cystein as second amino acid of the tripeptide DLD-XCV, wherein X is Hpg or Pg,. Selection of this module may be based on the specificity determining signature motif of A domains as published by Stachelhaus T. et al. 1999. To enable coupling of the L-cysteinyl acceptor to the D-X-aminoacyl donor, the

C domain of the M2 module is a ^D C _L domain (as mentioned above and as explained in Clugston S. L. et al. 2003). This ⁰ C _L domain is fused to an A domain that is heterologous thereto. The term "heterologous" as used in this context means that the C and A domains are from different modules. These different modules may be from the same enzyme or may be from different enzymes. Preferably, the A domain is obtainable from the second module of an ACVS. Surprisingly, the hybrid M2 module comprising such a ^D C _L -A domain configuration appears to be capable of incorporation of the amino acid cystein.

In a preferred embodiment, the ^D C _L domain of the M2 module is obtainable from the module immediately downstream of the module that is the source of the first module M1 of the peptide synthetase of the invention. For instance, the ^D C _L domain of the M2 module of the peptide synthetase is the ⁰ C _L domain of the seventh module of the CDA synthetase that is the source of the first module M1.

In another embodiment, the ⁰ C _L domain of the M2 module of the peptide synthetase is the ^D C _L domain of the second module of the Bacillus subtilis RB14 lturin Synthetase Protein ItuC, as defined by Tsuge K. et al. 2001 , J. Bad, 183, 6265-6273.

In a preferred embodiment of the invention, the second module M2 of the peptide synthetase is at least partly obtainable from the enzyme that is the source of the third module M3 of the peptide synthetase. In particular, the A and T domains of the M2 module of the peptide synthetase are obtainable from the module immediately upstream of the module that is the source of the third module of the peptide synthetase of the invention. For instance, the A and T domains of the M2 module of the peptide synthetase may be the A and T domains of the second module of an ACVS.

The third module M3 of the peptide synthetase should enable incorporation of the amino acid valine as the third amino acid of the tripeptide, as well as its conversion to the D- form, to yield the tripeptide DLD-XCV.

In a preferred embodiment of the invention, the third module of the peptide synthetase is obtainable from an ACVS, in particular is the third module of an ACVS.

The ACVS as mentioned above preferably is a bacterial or fungal ACVS, more preferably a bacterial ACVS obtainable from Nocardia lactamdurans or a fungal ACVS obtainable from a filamentous fungus such as Penicillium chrysogenum, Acremonium chrysogenum, Aspergillus nidulans. The modules M1 , M2 and M3 of the peptide synthetase may have the amino acid sequences as shown below. However, these sequences are merely shown as examples and are not intended to limit the scope of the invention. The skilled person will appreciate that NRPS A domains for instance share about 30-60% amino acid sequence identity, even A domains with specificity for the same amino acid but from a different source, and comprise several core motifs among which a specificity determining signature motif (Stachelhaus T. et al. 1999).

The M1 module of the peptide synthetase for instance has an amino acid sequence according to SEQ ID NO: 2 or SEQ ID NO: 4 or an amino sequence with a percentage identity of at least 30%, more preferably at least 40%, even more preferably at least 50%, most preferably at least 60% to the amino acid sequence of SEQ ID NO: 2 or SEQ ID NO: 4. Such polypeptide modules with a percentage identity of at least 30% are also called homologous sequences or homologues.

The M2 module of the peptide synthetase for instance has an amino acid sequence according to SEQ ID NO: 6 or to SEQ ID NO: 8 or an amino sequence with a percentage identity of at least 30%, more preferably at least 40%, even more preferably at least 50%, most preferably at least 60%, to the amino acid sequence of SEQ ID NO: 6 or

SEC ID NO: 8.

The M3 module of the peptide synthetase for instance has an amino acid sequence according to SEQ ID NO: 10 or an amino sequence with a percentage identity of at least 30%, more preferably at least 40%, even more preferably at least 50%, most preferably at least 60%, to the amino acid sequence of SEQ ID NO: 10.

The modules of the peptide synthetase may be obtained from natural NRPS enzymes as specified above or may de derived from such natural NRPS enzymes by

mutagenesis techniques, such as random and/or site-directed mutagenesis and/or gene shuffling. If necessary, the source module may further be engineered to add necessary domains, delete unnecessary domains or substitute a domain for a corresponding domain from another module. Typically, the A domain of a module determines specificity for a particular amino acid, whereas E and C domains may be obtained form any module of choice. For instance, in a situation that the selected source module for incorporation of the first amino acid X is not a first module (M 1 Module) in the source enzyme and thus comprises a C domain, the C domain of the source module may be deleted. In a situation that the source module does not contain an E domain while epimerization of the incorporated amino acid is desired, a suitable E domain may be added.

Engineered NRPS enzymes may be constructed by fusion of the appropriate domains and/or modules in the appropriate order. It is also possible to exchange a module or domain of an enzyme for a suitable module or domain of another enzyme. This fusion or exchange of domains and/or modules may be done using genetic engineering techniques commonly known in the art. Fusion of two different domains or modules may typically be done in the linker regions that are present in between modules or domains. See for instance EP 1 255 816 and Mootz H. D. et al. 2000 disclosing these types of constructions. Part or all of the sequences may also be obtained by custom synthesis of the appropriate polynucleotide sequence(s). For instance, the fusion of an ATE tri-domain fragment from a Hpg-specific NRPS module to the bi-modular Cys-Val-specific fragment of an ACVS may be done as follows. The ATE fragment of a Hpg specific module may be isolated by restriction enzyme digestion of the corresponding NRPS gene at the linker positions, more specifically, between the C domain and the A domain of the Hpg specific module, in case of a C-terminal module or between the C domain and the A domain of the Hpg specific module and between the E domain and the subsequent domain (C or TE domain), in case of an internal elongation module. The bi-modular Cys-Val fragment of ACVS may be obtained by 1 ) leaving the C- terminus intact, and 2) exchanging the C domain of the Cys specific module 2 for a C- domain which has ^D C _L specificity. In analogy to isolation of the ATE fragment, an ATEC four-domain fragment may be isolated including the C domain of the adjacent downstream module. The latter is fused to the bi-modular Cys-Val fragment of ACVS without the upstream C domain.

The NRPS enzymes as described herein may be suitably subjected to mutagenesis techniques, e.g. to improve the catalytic properties of the enzymes.

Polypeptides as described herein may be produced by synthetic means although usually they will be made recombinantly by expression of a polynucleotide sequence encoding the polypeptide in a suitable host organism.

In a fourth aspect, polynucleotides (e.g. isolated and/or purified) are provided comprising a polynucleotide sequence encoding the variant IPNS or the NRPS polypeptides of the previous aspects of the invention. The polynucleotides of the present invention further include any degenerate versions of a polynucleotide sequence encoding the polypeptide. For example, the skilled person may, using routine techniques, make nucleotide substitutions that do not affect the protein sequence encoded in the polynucleotides of the invention to reflect the codon usage of any particular host organism in which the polypeptides of the invention are to be expressed. Preferably, coding sequences in the polynucleotides are optimized by codon-pair optimization. The polynucleotide sequence of the invention may be RNA or DNA and includes genomic DNA, synthetic DNA or cDNA. Preferably, the polynucleotide is a DNA sequence.

The polynucleotides encoding the modules M1 , M2 and M3 of the NRPS enzyme of the first aspect may for instance have a nucleotide sequence according to SEQ ID NO: 3, SEQ ID NO: 5, SEQ ID NO: 7, SEQ ID NO: 9 and SEQ ID NO: 1 1 , encoding the amino acid sequences of SEQ ID NO: 2, SEQ ID NO: 4, SEQ ID NO: 6, SEQ ID NO: 8 and SEQ ID NO: 10, or may be nucleotide sequences encoding homologues of the above mentioned amino acid sequences, as defined above.

Polynucleotides may be synthesized according to methods well known in the art. They may be produced by combining oligonucleotides synthesized according to and along the nucleotide sequence of the polynucleotide of the invention. Alternatively, they may be synthesized by mutagenising a parental polynucleotide at any desired position. Polynucleotides may further be used to obtain polynucleotides encoding a further modified polypeptide, e.g. by subjecting polynucleotides to additional mutagenesis techniques.

Thus, polypeptides with improved activity typically are obtained by a method comprising the steps of subjecting a polynucleotide encoding the polypeptide to

mutagenesis, screening the obtained population of variant polypeptides for desired activity and isolating variants with an improved activity.

The mutagenesis may be done using any suitable technique known to the person skilled in the art. The mutagenesis may encompass subjecting a polynucleotide to random mutagenesis as well as site-directed mutagenesis. When site-directed mutagenesis is used, it is preferably combined with saturation mutagenesis at the selected position(s), enabling the substitution of the original amino acid for any other amino acid. Polynucleotide shuffling (gene shuffling) technology (for instance as disclosed in WO95/22625, WO98/27230, WO98/01581 , WO00/46344 and/or WO03/010183) may be used to obtain variants with a random combination of any variant position present in any member of a starting population of molecules. The starting population may further include one or more variants according to the invention.

The invention also provides vectors comprising a polynucleotide of the invention, including cloning and expression vectors or cassettes. In an expression vector or cassette, the polynucleotide is operably linked to a regulatory sequence that is capable of providing for the expression of a polypeptide from its coding sequence by the host cell. The term "operably linked" refers to a juxtaposition wherein the components described are in a relationship permitting them to function in their intended manner. A regulatory sequence such as a promoter, an enhancer or another expression regulatory signal "operably linked" to a coding sequence is positioned in such a way that expression of a polypeptide from its coding sequence is achieved under conditions compatible with the regulatory sequences.

Promoters/enhancers and other expression regulatory signals may be selected to be compatible with the host cell for which the expression cassette or vector is designed. If the polypeptide is produced as a secreted protein, the polynucleotide sequence encoding a mature form of the polypeptide in the expression cassette is operably linked to a polynucleotide sequence encoding a signal peptide.

The DNA sequence encoding the polypeptide of the invention is preferably introduced into a suitable host as part of an expression cassette. For transformation of the suitable host with the expression cassette, transformation procedures are available which are well known to the skilled person. The expression cassette can be used for transformation of the host as part of a vector carrying a selectable marker, or the expression cassette may be co-transformed as a separate molecule together with the

vector carrying a selectable marker. The vector may comprise one or more selectable marker genes.

For most filamentous fungi and yeasts, the expression construct is preferably integrated in the genome of the host cell in order to obtain stable transformants. In that case, the constructs are either integrated at random loci in the genome, or at predetermined target loci using homologous recombination.

Thus, a fifth aspect of the invention provides host cells transformed with or comprising a polynucleotide or vector of the invention.

Suitable host cells are host cells that allow for a high expression level of a polypeptide of interest. Such host cells are usable in case the polypeptides need to be produced and further to be used, e.g. in in vitro reactions. A heterologous host may be chosen wherein the polypeptide of the invention is produced in a form that is substantially free from other polypeptides with a similar activity as the polypeptide of the invention. This may be achieved by choosing a host that does not normally produce such polypeptides with similar activity.

Suitable host cells also are cells capable of production of β-lactam compounds, preferably host cells possessing the capacity to produce β-lactam compounds in high levels. The host cells may for example be prokaryotic (for example bacterial), fungal or yeast cells. The host may be selected based on the choice to produce a penam or a cephem compound. When production of a cephem compound is envisaged, the host may natively contain the necessary genes of the biosynthetic pathway leading to a cephem compound, in particular genes encoding expandase activity, and, optionally, hydroxylase and acetyl transferase activity. Alternatively, one or more genes of the biosynthetic pathway leading to a cephem compound may be transformed into a host cell devoid of these genes. It is thereby known to the skilled person that the enzymes expandase and expandase/hydroxylase are capable of expanding ampicillin and amoxicillin (Chin et al. 2003, FEMS Microbiol. Lett., 218, 251-257; Lloyd et al. 2004, J. Biol. Chem. 279, 15420-15426).

In one embodiment, a suitable host cell is a cell wherein the native genes encoding the ACVS and/or IPNS enzymes are inactivated, for instance by insertional inactivation. It is also possible to delete the complete penicillin biosynthetic cluster comprising the genes encoding ACVS, IPNS and AT. In this way the production of the β-lactam compound of interest is possible without simultaneous production of the

natural β-lactam. lnsertional inactivation may thereby occur using a gene encoding a NRPS and/or a gene encoding an IPNS as described above. In host cells that contain multiple copies of β-lactam gene clusters, host cells wherein these clusters are spontaneously deleted may be selected. For instance, the deletion of β-lactam gene clusters is described in patent application PCT/EP2007/054045.

Another suitable host cell is a cell that is capable of synthesising the precursor amino acids Hpg or Pg. Heterologous expression of the genes of the biosynthetic pathway leading to Hpg or Pg is disclosed in WO 02/34921. The biosynthesis of Pg or Hpg is achieved by withdrawing phenylpyruvate (PP) or p-hydroxyphenylpyruvate (HPP), respectively, from the aromatic amino acid pathway, converting PP or HPP to mandelic acid (MA) or p-hydroxymandelic acid (HMA), respectively, converting MA or HMA to phenylglyoxylate (PGL) or p-hydroxyphenylglyoxylate (HPGL), respectively, and finally converting PGL or or HPGL to D-Pg or D-Hpg, respectively. Example 15 exemplifies expression of the Hpg or Pg biosynthetic pathway in Penicillium chrysogenum.

Another suitable host cell is a cell that (over)expresses a 4'-phospho- pantetheine transferase (PPTase). 4'-Phosphopantetheine (PPT) is an essential prosthetic group of amongst others acyl-carrier proteins of fatty acid synthases and polyketide synthases, and peptidyl carrier proteins of NRPS's. The free thiol moiety of PPT serves to covalently bind the acyl reaction intermediates as thioesters during the multistep assembly of the monomeric precursors, typically acetyl, malonyl, and aminoacyl groups. The PPT moiety is derived from coenzyme A (CoA) and posttranslationally transferred onto an invariant serine side chain. This Mg ²⁺ -dependent conversion of the apoproteins to the holoproteins is catalyzed by the 4'- phosphopantetheine transferases (PPTases). It is advantageous to (over)express a PPTase with a broad substrate specificity. Such a PPTase is for instance encoded by the gsp gene from Bacillus brevis (Borchert et al. 1994, J. Bacteriology, 176, 2458- 2462).

A host may suitably include one or more of the modifications as mentioned above. A preferred host is a strain of Penicilium chrysogenum.

In a further aspect the invention provides a process for preparing a polypeptide according to the invention by cultivating a host cell (e.g. transformed with an expression vector or cassette as described above) under conditions conducive to expression (by

the vector or cassette) of the polypeptide according to the invention, and optionally recovering the expressed polypeptide. The polypeptide may be produced as a secreted protein in which case the polynucleotide sequence encoding a mature form of the polypeptide in the expression construct is operably linked to a polynucleotide sequence encoding a signal peptide. The polypeptide may also be produced as a fusion protein, i.e. fused to (part of) another polypeptide, for instance fused to maltose-binding protein.

For in vitro reactions, a secreted polypeptide or a cell-free extract comprising the polypeptide may be used. Optionally, the polypeptide may be (partially) purified prior to its use. In a further aspect, the invention provides a process for preparing a β-lactam compound by cultivating a host cell of the previous aspect under conditions to provide for expression of the IPNS and/or NRPS polypeptide(s) as described herein and conducive to the production of a β-lactam compound, and optionally recovering the β- lactam compound. According to the invention, the β-lactam compound that is produced is an N-acylated β-lactam compound, wherein the N-acyl side chain is a N-α-amino- hydroxyphenylacetyl or an N-α-aminophenylacetyl side chain. Advantageously, the present invention discloses the fully fermentative production of such β-lactam antibiotics. Examples of such β-lactam antibiotics are amoxicillin, ampicillin, cefadroxil and cefalexin. The host cells according to the invention may be cultured using procedures known in the art. For each combination of a promoter and a host cell, culture conditions are available which are conducive to expression of the polypeptide of the invention. After reaching the desired cell density or titre of the polypeptide the culture is stopped and the polypeptide is recovered using known procedures. Additionally, fermentation conditions may be established conducive to the production of a β-lactam.

The fermentation medium may comprise a known culture medium containing a carbon source (e.g. glucose, maltose, molasses), a nitrogen source (e.g. ammonium sulphate, ammonium nitrate, ammonium chloride, organic nitrogen sources e.g. yeast extract, malt extract, peptone), vitamins, and other inorganic nutrient sources (e.g. phosphate, magnesium, potassium, zinc, iron, trace elements, etc.). Optionally, an inducer may be included.

The selection of the appropriate medium may be based on the choice of expression host and/or based on the regulatory requirements of the expression

construct. Such media are known to those skilled in the art. The medium may, if desired, contain additional components favouring the transformed expression hosts over other potentially contaminating microorganisms. It may also be necessary to supplement the medium with precursor compounds for the β-lactam compounds to be produced. For instance, it may be necessary, depending on the host that is used, to include the amino acids Hpg or Pg in the culture medium. If so, these amino acids are preferably added as a separate feed.

The fermentation can be performed over a period of 0.5-30 days. It may be a batch, continuous or fed-batch process, suitably at a temperature in the range of between 0 and 45°C and, for example, at a pH between 2 and 10. Preferred fermentation conditions are a temperature in the range of between 20 and 37°C and/or a pH between 3 and 9. The appropriate conditions are usually selected based on the choice of the expression host and the protein and/or β-lactam compound to be expressed. After fermentation, if necessary, the cells can be removed from the fermentation broth by means of centrifugation or filtration. After fermentation has stopped and/or after removal of the cells, the polypeptide of the invention or the produced β-lactam compound may be recovered using conventional means. Recovery may include purification and/or extraction and/or crystallization steps.

Conveniently, the polypeptide of the invention or the β-lactam compound is combined with suitable (solid or liquid) carriers or diluents including buffers to produce a polypeptide or β-lactam compound composition. The polypeptide or the β-lactam compound may be attached to or mixed with a carrier, e.g. immobilized on a solid carrier. Thus the present invention provides in a further aspect a composition comprising a polypeptide of the invention or a β-lactam compound. This may be in a form suitable for packaging, transport and/or storage, preferably where the activity of the polypeptide is retained. Compositions may be of paste, liquid, emulsion, powder, flake, granulate, pellet or other extrudate form.

EXAMPLES

General materials and methods Standard DNA procedures were carried out as described elsewhere

(Sambrook, J. et al., 1989, Molecular cloning: a laboratory manual, 2 ^nd Ed., Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York) unless otherwise stated. DNA

was amplified using the proofreading enzyme Phusion polymerase (Finnzymes). Restriction enzymes were from Invitrogen or New England Biolabs. Fungal growth was performed in a mineral medium, containing (g/L): glucose (5); lactose (80); urea (4.5); (NhU) ₂ SO ₄ (1.1 ); Na ₂ SO ₄ (2.9); KH ₂ PO ₄ (5.2); K ₂ HPO ₄ (4.8) and 10 mL/L of a trace element solution A containing citric acid (150); FeSO ₄ JH ₂ O (15); MgSO ₄ JH ₂ O (150); H ₃ BO ₃ (0.0075); CuSO ₄ .5H ₂ O (0.24); CoSO ₄ JH ₂ O (0.375); ZnSO ₄ JH ₂ O (5); MnSO ₄ -H ₂ O (2.28); CaCI ₂ .2H ₂ O (0.99); pH before sterilization 6.5. As rich medium YEPD was used, containing (g/L): yeast extract (10); peptone (10); glucose (20). The cultures are incubated at 25°C in an orbital shaker at 280 rpm for 72-168 h. Penicillium chrysogenum transformations were carried out with the strain

Wisconsin 54-1255 (ATCC 28089). Other strains of P. chrysogenum, including mutants of strain Wisconsin 54-1255, having an improved β-lactam yield, are also suitable. An example of such a strain is CBS 455.95. Furthermore, P. chrysogenum strains with mutations or deletions leading to absence or a decrease of β-lactam production are also suitable as transformation hosts, for instance strains wherein the β-lactam gene clusters are deleted as described in patent application PCT/EP2007/054045.

The transformation of P. chrysogenum was performed as described in WO2004/106347.

Example 1

Cultivation of Penicillium chrysogenum and analysis of non-natural tripeptides such as DLD-Hpg-Cys-Val or DLD Pg-Cys-Val

Penicillium chrysogenum was cultivated in mineral medium (25 ml.) for 96 h at 25°C. Optionally, cultures were grown with or without amino acids such as L-Hpg (1 mM final concentration) or L-Pg (1 mM final concentration). At the end of the fermentation, the mycelium was removed by centrifugation or filtration and the mycelium was washed with physiological salt. Both the mycelium and the medium were assayed for tripeptides such as DLD-Hpg-Cys-Val and DLD-Pg-Cys-Val, using MS techniques such as LC- MS/MS. As standards, custom synthesized DLD-Hpg-Cys-Val and DLD-Pg-Cys-Val tripeptides were used. As a result, neither in the mycelium nor in the medium, tripeptides such as DLD-Hpg-Cys-Val and DLD-Pg-Cys-Val tripeptides were identified. As a conclusion, P. chrysogenum does not produce these non-natural tripeptides.

Example 2 Cloning of the Non-Ribosomal Peptide Synthetase module and domain cassettes

2 a. Cloning of NRPS module fragments which catalyze the incorporation and subsequent epimerization of L-Hpg to D-Hpg

Chromosomal DNA was isolated from Streptomyces coelicolor A3(2). Subsequently, PCR reactions were carried out, adding 3-8% DMSO to the reaction mixtures, furthermore, employing standard conditions. The following oligonucleotides were used: HpgFwi (caccatggcgacgctgccggaactgttc; SEQ ID NO: 12) and HpgRevi (tcaattaattaaccactc-ggactcaaggtcggac; SEQ ID NO: 13). The resulting PCR fragment Hpg1 , yielding a ATE tridomain of module 6 of CDA I Synthetase from S. coelicolor, was cloned into a pENTR/SD/D-Topo vector (Invitrogen, The Netherlands), resulting in plasmid pHpgi .

2 b. Custom synthesis of a synthetic NRPS module fragment which catalyzes the incorporation and subsequent epimerization of L-Pg to D-Pg.

A synthetic DNA fragment consisting of the DNA sequence according to SEQ ID

NO: 5 was custom-synthesized (DNA 2.0, Menlo Park, CA, U.S.A.). The DNA fragment harbors a ATE tridomain. The AT fragment originated from the C-terminal module of the SnbD protein of S. pristinaspiralis Pristinamycin Synthetase, The E-domain was taken from the S. coelicolor CDA Synthetase module 6.

The DNA fragment was cloned into the vector pCR-Blunt (Invitrogen), using the Protocol delivered by the supplier, yielding vector pPg1.

2 c. Cloning of the bi-modular NPRS fragments specific for Cysteine and Valine incorporation

For this experiment, the genes of ACV (δ(L-α-aminoadipate)-L-Cysteine, D-Valine) Tripeptide Synthetase from Penicillium chrysogenum and Nocardia lactamdurans were taken as templates. PCR amplifications using chromosomal DNA templates of either organism were carried out. For P. chrysogenum, the following oligonucleotides were employed: The oligo pair PcM2M3fw1 (caccttaattaatgaggagaatgcagcaacggac; SEQ ID NO: 14) and PcM2M3rev1 (tcaatagcgagcgaggtgttc; SEQ ID NO: 15), resulting in Fragment PcM2M3_1 (domain organization CATCATETe), the oligo pair PcM2M3fw2

(caccactagtctggagtatctctcatctatc; SEQ ID NO: 16) and PcM2M3rev1 (tcaatagcgagcgaggtgttc; SEQ ID NO: 15), resulting in Fragment PcM2M3_2 (domain organization ATCATETe). Fragment PcM2M3_1 was cloned into pENTR/SD/D-Topo yielding pPcM2M3_1 , and cloning of fragment PcM2M3_2 in the same vector gave plasmid pPcM2M3_2. In an analogues way, PCR amplifications with chromosomal DNA of Nocardia lactamdurans was carried out. Reactions using the oligonucleotides NIM2M3fw1 (caccttaattaatgccgaagaggtcaccacc; SEQ ID NO: 17) and NIM2M3rev1 (ggtcatcgctccctagg; SEQ ID NO: 18) gave DNA fragment NIM2M3_1 (domain organization CATCATETe), while reations with the oligos NIM2M3fw2 (caccactagtctcatctcaccgtcgatg; SEQ ID NO: 19) and NIM2M3rev1 (ggtcatcgctccctagg; SEQ ID NO: 18) resulted in fragment NIM2M3_2 (domain organization ATCATETe). Both fragments were finally cloned into pENTR/SD/D-Topo vectors, yielding plasmids pNIM2M3_1 and pNIM2M3_2, respectively.

2 d. Cloning of the NRPS condensation domain fragment specific for the peptide bond formation between a D- and a L-amino acid

Two different condensation domains with ⁰ C _L stereochemistry were cloned. The first domain was PCR-amplified from CDAI Synthetase using Streptomyces coelicolor chromosomal DNA, employing the oligo-nucleotides sc_C_fw(caccttaattaatagccagcaaccgacccgtgcg; SEQ ID NO: 20) and sc_C-rev (ttactagtgtcgcgggcgtacgcctcgtc; SEQ ID NO: 21 ), yielding fragment sc_C. The second fragment which consisted of a NRPS ^D C _L C-domain from the B. subtilis lturin Synthetase ItuC module C2, according to SEQ ID NO: 22, was custom synthesized and named Bs_C. Both fragments were cloned into the vector pENTR/SD/D-Topo, resulting in the plasmids pSC_C and pBS_C, respectively.

Example 3

Construction of Tripeptide Non-Ribosomal Peptide Synthetase genes by fusion of the cloned NRPS module and domain cassettes

The plasmid pHpgi was digested with the restriction enzymes Pad and Ascl, resulting in a 5.9 kb DNA fragment which contains the Hpg-specific NRPS module gene and most of the vector sequence. The plasmids pPcM1 M2_1 was cut with Ascl and

Pad, showing a 8.5 kb DNA fragment. Ligation of this fragment and the 5.9 kb fragment from pHpgi gave vector pSCPCHybridi . Plasmids pPcM1 M2_2, pSC_C were

restricted with Spel and Ascl, resulting in 7.0 kb and 1.4 kb fragments, respectively. Both fragments were ligated and gave plasmid pPcM1 M2_3. Plasmids pPcM1 M2_3 and plasmid pHpgi were cut with Pcil and Pad, giving 12 kb and 4.5 kb DNA fragments, respectively. Ligation of both fragments resulted in the final construct pSC_SC_C_PCM2M3Hybrid1. Further restriction of pSC_SC_C_PCM2M3Hybrid1 and pBS_C with Pad and Spel led to fragments of 16 kb (pSC_SC_C_PCM2M3Hybrid1 without the C-domain of module 2 of ACV Synthetase) and 1.4 kb (Bacillus subtilis C- domain), respectively. Ligation of both fragments gave plasmid pSC_BS_C_PCM2M3Hybrid2. Restriction of pHpgi and pNIM1 M2_1 with Pad and AfIII gave 3.7 kb and 8.8 kb DNA fragments. Ligation of both fragments yielded plasmid pSCNLHybridi . Plasmids pSC_SC_C_PCM2M3Hybrid1 and pNIM1 M2_2 were restricted with Spel and AfIII, resulting in 4 kb (ATEC tetradomain organization from Hpg module and (D)C(L) domain) and 7 kb (ATCATETe organization) fragments, respectively. Ligation of both fragments gave plasmid pSC_SC_C_NLM2M3Hybrid1. In analogy, plasmids pSC_BS_C_PCM2M3Hybrid1 and pNIM1 M2_2 were restricted with Spel and AfIII, resulting in 4 kb (ATEC tetradomain organization from Hpg module and (D)C(L) domain) and 7 kb (ATCATETe organization) fragments, respectively. Ligation yielded plasmid pSC_BS_C_NLM2M3Hybrid2. For the construction of a Phenylglycin specific Non-Ribosomal Peptide

Synthetase fusion gene, the plasmids pPg1 and pSC_SC_C_NLM2M3Hybrid1 were restricted each with the restriction enzymes AfIII and Pad. The 3.4 kb DNA fragment (harboring the ATE NRPS tridomain) of pPg1 was ligated into the 14 kb vector fragments obtained by restriction of pSC_SC_C_NLM2M3Hybrid1 , resulting in plasmid pPg_SC_C_NIM2M3Hybrid1. In analogy, if vector pSC_SC_C_NLM2M3Hybrid1 was replaced by vector pSC_BS_C_NLM2M3Hybrid2, the resulting plasmid was pPg_BS_C_NIM2M3Hybrid2.

Example 4 Construction of a Penicillium chysogenum Gateway® expression destination vector for hybrid-Non-Ribosomal Peptide Synthetases

Plasmid pPT12, harboring the Notl-flanked Promotor-Terminator cassette from the IPN Synthase from Penicillium chrysogenum, was digested with Notl (for vector

description, see Theilgaard et al. 2000, Biotechnology and Bioengineering, 72, 380- 387). The so obtained IPNS Promoter-Terminator cassette was cloned into pBluescriptll SK (Stratagene, The Netherlands), that was also treated with Notl. Ligation led to pProducti . Plasmid pProduct 1 was restricted with BamH1 and Xhol. Additionally, a DNA fragment which harbored a Phleomycin resistance cassette under the control of a gpdA promoter, was custom synthesized. The cassette has a sequence according to SEQ ID NO: 23.

After restriction of this custom synthesized DNA fragment with BamHI and Xhol the resulting DNA was ligated into the vector pProducti (which was opened with BamHI , Xhol). The resulting plasmid was named pProduct2. In a final cloning step, a DNA fragment containing a attR1-cat-ccdB-attR2 cassette (Chloramphenicol resistance and a toxicity gene for E. coli) was PCR-amplified from a Gateway ^® Destination vector such as pET-DEST42 (Invitrogen, Carlsbad, CA, USA). The following oligo-nucleotides were used: fw: ttatcgatttgcataaaaaacagac (SEQ ID NO: 24), rev: ttatcgatgcttaccttcaagcttcg(SEQ ID NO: 25). The resulting DNA fragment was restricted with CIaI and subsequently cloned in pProduct2, which was previously opened by digestion with CIaI as well. Here, plasmids were screened which harbored the attR1 - cat-ccdB-attR2 fragment in such an orientation that attR1 is fused next to the IPNS promoter P| _PN s- The resulting Gateway ^® destination vector was named pDEST-PcexpM .

Example 5

Construction of P. chrysogenum expression constructs for Non-Ribosomal

Peptide Synthetases

For the construction of P. chrysogenum expression constructs, the Gateway ^® LR-Clonase reaction was carried out using standard conditions as described by

Invitrogen. For detailed protocols, see manuals for the Gateway ^® technology at litiβj/M!Wλ ^f ^LQyltroaθA-C _. orn. For the reaction, the entry vectors pSCNLHybridi , pSC_SC_C_NLM2M3Hybrid1 and pSC_BS_C_NLM2M3Hybrid2 were each incubated with the Gateway ^® destination vector pDEST-Pcexpri . See table 1 for the resulting expression plasmid names.

Alternatively, for the construction of P. chrysogenum expression constructs, the constructs containing the engineered tripeptide NRPS-genes were functionally linked to the λ/ofl-flanked promoter-terminator cassette from the pαbC-gene from

Penicillium chrysogenum (for vector description, see Theilgaard et al. 2000, Biotechnology and Bioengineering, 72, 380-387) by fusion-PCR. See Table 1 for the resulting expression plasmid names.

Table 1

Example 6

Transformation of Penicillium chrysogenum with linearized plasmid pEXPR- NRPS4, cultivation and tripeptide analysis

In independent experiments, Penicillium chrysogenum was transformed with pEXPR-NRPS4 that was linearized prior to transformation with suitable enzymes that cut in the vector backbone, such as Psil or Pcil. Selection of transformants was done on mineral medium agar plates with 50 g/mL Phleomycin and 1 M Saccharose.

Phleomycin resistant colonies appearing on these protoplast regeneration plates were re-streaked on fresh phleomycin agar plates without the saccharose and grown until sporulation. The phleomycin resistant transformants were screened via colony PCR for the presence of the NRPS genes. For this, a small piece of colony was suspended in

50 L TE buffer (Sambrook et al., 1989) and incubated for 10 min at 95 C. To discard the cell debris, the mixture was centrifuged for 5 min at 3000 rpm. The supernatant (5

L) was used as a template for the PCR reaction with Super-Taq from HT Biotechnology Ltd. The PCR reactions were analyzed on the E-gel96 from Invitrogen.

The following oligo-nucleotides were used in the colony-PCR screen (see SEQ ID NO: 26 and SEQ ID NO: 27):

Transformed plasmid: pEXPR-NRPS4 Oligo-nucleotides for colony PCR: Fw: GCCTGGTGCCTGATGC

Rev: GGTGTGGTCGGAGACG

The expected size of these PCR reactions was 303 bp.

The positive clones underwent one further purification step on fresh phleomycin agar plates without the saccharose. Subsequently, they were grown on mineral medium as described in "General Materials and Methods". Additionally, the amino acid L-Hpg (4-Hydroxyphenylglycine) or L-Pg (Phenylglycine) were added to a final concentration of 1 mM, depending on what tripeptide was analyzed. Cultivation was carried out on shake flask scale or on 24-well micro titer plates. At the end of the fermentation, the mycelium was removed by centrifugation or filtration and the mycelium was washed with physiological salt. Both the mycelium and the medium were assayed for tripeptides such as DLD-Hpg-Cys-Val (in case L-Hpg was added to the medium) and DLD-Pg-Cys-Val (in case L-Pg was added to the medium), using MS techniques such as LC-MS/MS. As standards, custom synthesized DLD-Hpg-Cys-Val and DLD-Pg-Cys-Val tripeptides were used. As a result, neither in the mycelium nor in the medium, tripeptides such as DLD-Hpg-Cys-Val and DLD-Pg-Cys-Val tripeptides were identified. As a conclusion, P. chrysogenum does not produce these un-natural tripeptides, if NRPS expression cassettes constructs such as pEXPR-NRPS4 were integrated.

Example 7

Transformation of Penicillium chrysogenum with linearized plasmids pEXPR- NRPS5 and pEXPR-NRPS6, cultivation and DLD-Hpg-Cys-Val tripeptide production

In independent experiments, Penicillium chrysogenum was transformed with either pEXPR-NRPS5 or pEXPR-NRPS6 that were linearized prior to transformation with suitable enzymes that cut in the vector backbone, such as Psil or Pcil. The transformation of P. chrysogenum was carried out under conditions e.g. described in WO2004/106347. Selection of transformants was done on mineral medium agar plates with 50 g/mL Phleomycin and 1 M Saccharose. Phleomycin resistant colonies appearing on these protoplast regeneration plates were re-streaked on fresh phleomycin agar plates without the saccharose and grown until sporulation. The phleomycin resistant transformants were screened via colony PCR for the presence of the NRPS genes. For this, a small piece of colony was suspended in 50 L TE buffer (Sambrook et al., 1989) and incubated for 10 min at 95 C. To discard the cell debris, the mixture was centrifuged for 5 min at 3000 rpm. The supernatant (5 L) was used as

a template for the PCR reaction with Super-Taq from HT Biotechnology Ltd. The PCR reactions were analyzed on the E-gel96 from Invitrogen.

The following oligo-nucleotides were used in the colony-PCR screen (see SEQ ID NO: 26 and SEQ ID NO: 27):

Transformed plasmid pEXPR-NRPS 5 or pEXPR-NRPS6

Oligo-nucleotides for colony PCR Fw: GCCTGGTGCCTGATGC

Rev: GGTGTGGTCGGAGACG

The expected size of these PCR reactions was 303 bp.

The positive clones underwent one further purification step on fresh phleomycin agar plates without the saccharose. Subsequently, they were grown on mineral medium as described in "General Materials and Methods". Additionally, the amino acids L-Hpg (4-Hydroxyphenylglycine) or L-Pg (Phenylglycine) were added to a final concentration of 1 mM, depending on what tripeptide was analyzed. Cultivation was carried out on shake flask scale or on 24-well micro titer plates. At the end of the fermentation, the mycelium was removed by centrifugation or filtration and the mycelium was washed with physiological salt. Both the mycelium and the medium were assayed for tripeptides such as DLD-Hpg-Cys-Val (in case L-Hpg was added to the medium) and DLD-Pg-Cys-Val (in case L-Pg was added to the medium), using MS techniques such as LC-MS/MS. As standards, custom synthesized DLD-Hpg-Cys-Val and DLD-Pg-Cys-Val tripeptides were used. The mycelium and the supernatant showed significant levels of the tripeptide DLD-Hpg-Cys-Val tripeptide. The negative control, for which an untransformed P. chrysogenum strain Wisconsin 54-1255 (ATCC 28089) was employed, showed no formation of DLD-Hpg-Cys-Val tripeptide. This result gives proof that the transformed engineered Non-Ribosomal Peptide Synthetases produce the unnatural tripeptide DLD-Hpg-Cys-Val.

Example 8 Transformation of Penicillium chrysogenum with linearized plasmid pEXPR-

NRPS7 and pEXPR-NRPS8, cultivation and DLD-Pg-Cys-Val tripeptide production

In independent experiments, Penicillium chrysogenum was transformed with either pEXPR-NRPS7 or pEXPR-NRPSδ that were linearized prior to transformation with suitable enzymes that cut in the vector backbone, such as Psil or Pcil. The

transformation of P. chrysogenum was carried out under conditions e.g. described in WO2004/106347. Selection of transformants was done on mineral medium agar plates with 50 g/mL Phleomycin and 1 M Saccharose. Phleomycin resistant colonies appearing on these protoplast regeneration plates were re-streaked on fresh phleomycin agar plates without the saccharose and grown until sporulation. The phleomycin resistant transformants were screened via colony PCR for the presence of the NRPS genes. For this, a small piece of colony was suspended in 50 L TE buffer (Sambrook et al., 1989) and incubated for 10 min at 95 C. To discard the cell debris, the mixture was centrifuged for 5 min at 3000 rpm. The supernatant (5 L) was used as a template for the PCR reaction with Super-Taq from HT Biotechnology Ltd. The PCR reactions were analyzed on the E-gel96 from Invitrogen.

The following oligo-nucleotides were used in the colony-PCR screen(see SEQ ID NO: 26 and SEQ ID NO: 27): Transformed plasmid pEXPR-NRPS 7 or pEXPR-NRPSδ

Oligo-nucleotides for colony PCR Fw: GCCTGGTGCCTGATGC

Rev: GGTGTGGTCGGAGACG

The expected size of these PCR reactions was 303 bp. The positive clones underwent one further purification step on fresh phleomycin agar plates without the saccharose. Subsequently, they were grown on mineral medium as described in "General Materials and Methods". Additionally, the amino acids L-Hpg (4-Hydroxyphenylglycine) or L-Pg (Phenylglycine) were added to a final concentration of 1 mM, depending on what tripeptide was analyzed. Cultivation was carried out on shake flask scale or on 24-well micro titer plates. At the end of the fermentation, the mycelium was removed by centrifugation or filtration and the mycelium was washed with physiological salt. Both the mycelium and the medium were assayed for tripeptides such as DLD-Pg-Cys-Val (in case L-Pg was added to the medium), using MS techniques such as LC-MS/MS. As standards, custom synthesized DLD-Hpg-Cys-Val and DLD-Pg-Cys-Val tripeptides were used. The mycelium and the supernatant showed significant levels of the tripeptide DLD-Pg-Cys-Val tripeptide. The negative control, for which an untransformed P. chrysogenum strain Wisconsin 54-1255 (ATCC 28089) was employed, showed no formation of DLD-Pg-Cys-Val tripeptide. This result gives proof

that the transformed engineered Non-Ribosomal Peptide Synthetases produce the unnatural tripeptide DLD-Pg-Cys-Val.

Example 9 In vitro production of amoxicillin by conversion of DLD-HpgCV

Three different constructs, containing the pαbC-genes of Penicillium chrysogenum (pAJL-pcbC), Nocardia lactamdurans (pAJL-pcbC-NL) and Aspergillus nidulans (pXTN313) were inoculated from a glycerol stock and grown overnight at 37°C, in 2xTY and 35 μg chloroamphenicol per ml. The respective pcbC-genes encode IPNSs which are being produced as fusion proteins with MBP (maltose binding protein), encoded by the ma/E-gene from Escherichia coli. The malE-pcbC fusion genes are under control of the IPTG-inducible føc-promoter.

Plasmid DNA was isolated from the overnight cultures. The isolated DNA was used to transform E. coli TOP10 (Invitrogen). One isolated colony was used to inoculate 10 ml of 2xTY supplemented with 35 μg chloroamphenicol per ml. After O/N incubation at 37°C and 280 rpm, the OD ⁶⁰⁰ was measured. The strains were diluted to 100 ml of fresh medium, to a final OD ⁶⁰⁰ of 0.015. Growth at 37°C and 280 rpm was allowed until the OD ⁶⁰⁰ reached a value between 0.4 and 0.6. IPTG was added to a final concentration of 0.5 mM. The incubation was continued at 22°C and 220 rpm, overnight.

Cells were harvested by centrifugation and washed with 0.9% NaCI in milliQ water. The cell pellets were resuspended in 1.5 ml extraction buffer (50 mM Tris.HCI pH 7.5; 0,5 mg lysozyme/ml, 5 mM DTT). Sonification was performed in order to lyse the cells. After centrifugation for 10 minutes at 14000 rpm in an Eppendorf centrifuge, 200 μl aliquots were collected and frozen in liquid nitrogen. The frozen cell free extracts were stored at -80 ⁰ C.

IPNS activity assays were performed with ACV and HpgCV as substrates. Apart from the standard assay conditions for the IPNS activity assay, modifications were applied to the assay mix in order to maximize the chance of success.

Reaction mixes were as follows:

Reaction mix 1 (standard assay) 30 - 50 mM Tris.HCI pH 8.0 (preferably 36 mM) 1 - 6.7 mM ascorbate (preferably 3 mM) 50 μM to 2 mM FeSO ₄ (preferably 86 μM) 0.3 - 2 mM tripeptide (preferably 1.5 mM for ACV and 0.75 mM for HpgCV) 0.75 - 4 mM DTT (preferably 3 mM)

The bis-ACV was reduced to ACV by mixing two aliquots of 7.5 mM bis-ACV with one aliquot of 60 mM DTT, followed by an incubation of 5 to 25 minutes at room temperature. Likewise, HpgCV was reduced prior to the actual assay.

Reaction mix 2 (alternative assay) 30 - 50 mM HEPES pH 7.0 (preferably 36 mM) 0.1 - 0.2 mM ascorbate (preferably 0.1 mM) 1- - 50 μ M FeSO ₄ (preferably 25 μ M)

0.3 - 2 mM tripeptide (preferably 1.5 mM for ACV and 1.5 mM for HpgCV) 0.2 - 1.2 mM TCEP (Tris(2-Carboxyethyl) phosphine) (preferably 1 mM)

Bis-ACV and HpgCV were pre-treated with TCEP in order to break S-S bonds in the dimers. To this end, two aliquots of 7.5 mM bis-ACV, or HpgCV, were mixed with one aliquot of 20 mM TCEP, followed by an incubation of 5 to 25 minutes at room temperature.

5 μl of the CFEs, containing the MBP-IPNS fusion proteins, were incubated with 595 μl of the reaction mixes mentioned above. After 10 minutes of incubation at 25°C, the reaction was stopped by adding 125 μl of the reaction with 50 μl ice-cold methanol. After incubation at -20 ⁰ C for at least one hour, the samples were analysed by

LC/MS.

The LC/MS analysis was performed on an LCQ (Thermo Scientific), using the following settings:

Eluents Eluens A: 0.1 % FA in MiIIiQ water Eluens B: 0.1 % FA in MiIIiQ water

Gradient T(min) flow (ml/min) %A %B

0.0 0.2 98 2

5.0 0.2 98 2

20.0 0.2 78 22

20.1 0.2 98 2

30.0 0.2 98 2

Column Varian lnertsil 3 ODS 3 CP22568 150 ^* 2.1 mm 3 μm Column Temp. 55°C Flow 0.2 ml/min

Injection volume 25 μl Tray Temp. 4°C

Retention time: IPN Approx. 12.2 min

Amoxicillin Approx. 4.9 min Ampicillin Approx. 16.5 min

MS

Instrument LCQ (SM05)

LC/MS OFF axis ESI/pos Probe height nr 6, depth nr 3

LC/MS/MS 360 iw = 5.0 aa = 30%

366 iw = 5.0 aa = 30%

350 iw = 5.0 aa = 30%

LC/MS/MS/MS 349 iw = 5.0 aa = 30% m/z range MS: 200 - 1000 micro scans 1 inject time 500 ms

The samples were diluted 10 times with milliQ water and 25 μl was injected into the LC/MS system.

Substrates (ACV, HpgCV, PgCV) and products (IPN, amoxicillin, ampicillin) were identified based on retention time (LC), the mass over charge (m/z) value (MS) as well as the fragmentation pattern upon fragmentation in the MS ⁿ -mode.

The results are summarized in the table below.

Table 2: Conversion of the substrates ACV and HpgCV into IPN and amoxicillin respectively by the MBP-IPNS fusion proteins from different sources

Surprisingly, under the experimental conditions of reaction mix 2, MBP-IPNS is capable of converting HpgCV into amoxicillin, while no amoxicillin is converted under the conditions of reaction mix 1 .

Example 10

Development of a bioassay allowing the detection of the conversion of DLD- HpgCV to amoxicillin

The bacterial strains Escherichia coli ESS, Bacillus subtilis ATCC6633 and

Micrococcus luteus ATCC9341 were grown overnight in 2xTY-medium at 30 ⁰ C and 280 rpm. After 16 hours of growth, the OD ⁶⁰⁰ was measured. Typically, the OD ⁶⁰⁰ amounts 5.0, ranging from 2.0 to 8.0. A solution of amoxicillin was made at a concentration of 7 mg/ml. Subsequently, serial dilutions were made ranging from 1 mg/l down to 0.001 mg/l. MiIIiQ water served as a negative control.

In a 96 wells microplate, 25 μl of the serial dilutions were pipetted in the wells.

Each microplate was prepared in duplicate. The overnight bacterial cultures were diluted in fresh 2xTY-medium to a final OD ⁶⁰⁰ of 0.01. To each well, 150 μl of the diluted bacterial cultures was added. The microplates were covered with a lid, and incubated overnight at

25°C and the duplicate microplate at 37°C.

After 16 hours of incubation, the OD ⁶⁰⁰ was read in a microplate reader. From the results, the concentration of amoxicillin still allowing uninhibited growth of the test microorganisms can be read.

Table 3: Concentration of amoxicillin still allowing uninhibited growth in 2xTY medium

Example 11

Development of a bioassay allowing the detection of the conversion of DLD-PgCV to ampicillin

The bacterial strains Escherichia coli ESS, Bacillus subtilis ATCC6633 and Micrococcus luteus ATCC9341 were grown overnight in 2xTY-medium at 30 ⁰ C and 280 rpm. After 16 hours of growth, the OD ⁶⁰⁰ was measured. Typically, the OD ⁶⁰⁰ amounts 5.0, ranging from 2.0 to 8.0. A solution of ampicillin was made at a concentration of 7 mg/ml. Subsequently, serial dilutions were made ranging from 1 mg/l down to 0.001 mg/l. MiIIiQ water served as a negative control.

In a 96 wells microplate, 25 μl of the serial dilutions were pipetted in the wells. Each microplate was prepared in duplicate. The overnight bacterial cultures were diluted in fresh 2xTY-medium to a final OD ⁶⁰⁰ of 0.01. To each well, 150 μl of the diluted bacterial cultures was added. The microplates were covered with a gas permeable adhesive seal, and incubated overnight at 25°C and the duplicate microplate at 37°C.

After 16 hours of incubation, the OD ⁶⁰⁰ was read in a microplate reader. From the results, the concentration of ampicillin still allowing uninhibited growth of the test microorganisms can be read.

Table 4: Concentration of ampicillin still allowing uninhibited growth in 2xTY medium

Example 12

Screening of pcbC-genes from different genera for the in vitro conversion of DLD-

HpgCV to amoxicillin The coding regions of pαbC-genes from 1 1 species were ordered at DNA 2.0

(1430 O'Brien Drive, Suite E, Menlo Park, CA 94025, USA). To this end, the DNA sequence of the coding region was taken as a basis. The following species were selected: Cephalosporium Acremonium, Penicillium chrysogenum, Aspergillus (Emericella) nidulans, Streptomyces jumonjinensis, Nocardia lactamdurans, Streptomyces microflavis, Lysobacter lactamgenus, Flavobacterium species, Streptomyces clavuligerus, Streptomyces griseus, Streptomyces cattleya.

The sequence of the start-codon was changed to CATATG, the recoginition site of λ/αtel, in order to facilitate future cloning steps. Likewise, a λ/s/ϊ-site (ATGCAT) was introduced immediately behind the stop-codon. Internal λ/s/ ^' l- and λ/αtel-sites were removed as much as possible.

The plasmids carrying the pαbC-genes were cut with λ/s/ ^' l and Nde\ and the coding region of the pαbC-gene was isolated and subcloned into vector pSJ127, which was cut with the same enzymes. In this way, the coding region of the pαbC-gene is cloned in frame with the ma/E-gene encoding maltose binding protein, under control of an inducible promoter. Upon induction of expression, a fusion protein will be synthesized consisting of the maltose binding protein at the N-terminus of the fusion protein and IPNS at the C-terminus.

E.coli TOP10, transformed with the ma/E-pαbC-plasmids, were grown in 2xTY- medium containing 35 μg chloroamphenicol/ml, overnight at 37°C and 280 rpm. The next day, the OD ⁶⁰⁰ was measured and cells were diluted in fresh medium to a final OD ⁶⁰⁰ of 0.015. The cells were grown further (37°C, 280 rpm) until the OD ⁶⁰⁰ reached a value of 0.4 - 0.6. IPTG was added to a final concentration of 0.5 mM. Growth was proceeded at 22°C, 280 rpm overnight. Cells were harvested by centrifugation for 10 minutes at 5000 rpm and 4°C. Cells were washed with a 0.9% NaCI-solution, and pelleted again by centrifugation. The cell pellets were frozen at -20 ⁰ C for at least 16 hours.

Cell pellets were resuspended in 1.5 ml extraction buffer (50 mM Tris.HCI pH 7.5; 5 mM DTT) and extracts were made by sonification. The extracts were centrifuged

in order to remove cellular debris in an Eppendorf centrifuge (10 minutes, 14000 rpm, 4°C). The supernatant (cell free extract, CFE) was aliquoted into fresh Eppendorf tubes and frozen in liquid nitrogen, and subsequently stored at -80 ⁰ C.

The conversion of HpgCV to amoxicillin was assayed as follows. Firstly, the substrate HpgCV was reduced by mixing 500 μl of 200 mM HpgCV and 500 μl 20 mM

TCEP, followed by incubation at 25°C for 10 minutes. Secondly, the following assay mixture was made (for 10 reactions): 300 μl 0.5 mM FeSO ₄ .7H ₂ O; 480 μl 1.25 mM ascorbate; 4270 μl 50 mM HEPES (pH 7.0). The assay was started by mixing 90 μl reduced substrate, 505 μl assay mixture and 5 μl of the CFE, containing the MBP-IPNS fusion protein.

The reaction mixture was incubated at 25°C for 10 minutes.

25 μl of the reaction mixture was mixed with 150 μl of a diluted overnight culture of Micrococcus luteus in 2xTY (OD ⁶⁰⁰ =0.01 ), in a microplate. In order to verify that any possible inhibition of growth was caused by the formation of amoxicillin, each sample was also tested in the presence of β-lactamase (Penase ^® from BBL Difco), which cleaves the β-lactam ring and thereby inactivates the antibiotic activity.

The microplate was covered with a lid and incubated overnight at 30 ⁰ C and 550 rpm. The OD ⁶⁰⁰ was read on a microplate reader.

Surprisingly, the MBP-IPNS fusion proteins originating from the pαbC-genes from A. nidulans as well as from S. clavuligerus both showed inhibition of the growth of M. luteus, which was grown fully in the presence of Penase ^® , indicating that the inhibition was due to the formation of amoxicillin from DLD-HpgCV.

This was confirmed by LC-MS/MS analysis. To this end, the reaction was stopped by adding 125 μl reaction mixture with 50 μl methanol (-20 ⁰ C). The samples were incubated for at least 1 hour at -20°C. A centrifugation step (14000 rpm for 15 minutes at 4°C) was performed, followed by LC-MS/MS analysis (see Example 9). The results confirmed the results of the bio-assay.

As a control, LLD-ACV was included. Indeed, as expected, almost all MBP- IPNS fusion proteins formed IPN (isopenicillinN). Addition of Penase ^® to the reaction mixture deleted the observed effect. Again, the results were confirmed by LC-MS/MS.

Example 13 Engineering pcbC-genes

The clones pSJ-AnlPNS (pαbC-gene from A. nidulans) and pSJ-SclPNS

(pcbC-gene from S. clavuligerus) were subjected to error prone PCR (EP-PCR) using the Diversify PCR kit from Clontech according to the guidelines of the supplier. In short, the number of errors introduced by PCR can be modified by applying different PCR conditions.

The amplified pαbC-inserts were subcloned into the expression vector pBADMHmalEDEST, using the restriction enzymes Nde\ and λ/s/ϊ. Also in this plasmid, IPNS will be produced as an MBP fusion protein.

Of each condition applied, 10 clones were sequenced. The libraries in which a mutation rate of approximately 1 mutation per kb was found were selected for the screening for improved variants with respect to the conversion of HpgCV into amoxicillin. The selected A. nidulans- and S. clavuligerus pcbC libraries were grown, and the insert was subcloned into the expression vector pBADMHmalEDEST, using the restriction enzymes Nde\ and λ/s/ϊ. Also in this plasmid, IPNS will be produced as an MBP fusion protein.

A third library was constructed, in which mutations were introduced at specific positions within the pαbC-gene from A. nidulans (site directed approach). These positions were selected on two criteria: a) the creation of space in the active center, and b) weakening the C-terminus binding.

The creation of space in the active center may be beneficial for improving the activity of the native IPNS on HpgCV, since the Hpg-moiety is larger than the α- aminoadipate-moiety in ACV.

The C-terminus of the IPNS molecule acts as a quasi-substrate when the enzyme is not loaded with a substrate. Upon the approach of the substrate to the active center, the C-terminus withdraws, creating space for the substrate. In order to prevent this competition between C-terminus and substrate, alterations were introduced here as well.

Eight amino acid positions were selected for modification: I75, Y91 , S183, V185, N287, L321 , L324 and T331.

Primers were designed, 33-mers, in which the three middle nucleotides (encoding the amino acid to be modified) were randomized (NNN), thus creating the possibility for any amino acid to be present in the resulting IPNS molecule at the indicated positions. All eight positions were thus modified by fusion PCR, resulting in eight sub- libraries, each with single site mutations. One additional library was made in which all the mutations were combined in a random fashion.

Also in this case, the inserts of the thus obtained library was subcloned in the expression vector pBADMHmalEDEST, using the restriction enzymes Nde\ and λ/s/ϊ. The three libraries were plated from glycerol stocks on selective agar-medium, containing 50 μg zeocin per ml. Of each EP-PCR library, 2500 independent clones were picked and used to inoculate 150 μl 2xTY-medium containing 50 μg zeocin per ml, in 96 wells plates. Of the site directed library, 200 clones were picked from each sub- library for inoculation of 150 μl 2xTY-medium containing 50 μg zeocin per ml, in 96 wells plates.

After growth overnight at 37°C and 280 rpm, 5 μl of the confluent cultures was transferred to 1 ml fresh 2xTY-medium containing 50 μg zeocin per ml, in a deepwell microplate. Growth was allowed at 37°C and 280 rpm, until the OD ⁶⁰⁰ reached a value between 0.4 and 0.6, on average across the microplate. Cells were induced to produce the MBP-IPNS fusion protein by the addition of 40 μl 5% L-arabinose to each of the individual wells. Subsequently, the cells were grown for 24 hours at 22°C and 220 rpm. Cells were harvested by centrifugation. The supernatant was discarded and the cell pellets were frozen overnight at -20 ⁰ C.

To each well, containing a cell pellet, 100 μl Lysis Buffer (50 mM HEPES pH 7.0; 0.5 mg/ml lysozyme; 0.1 mg/ml DNAsel; 5 mM DTT; 5 mM MgSO4) was added, followed by an incubation at room temperature for 30 minutes.

The microplates were spun, and 50 μl of the supernatant was transferred to a fresh microplate. 50 μl assay-mix (50 mM HEPES pH 7.0; 1 mM TCEP; 100 μM Na- ascorbate; 25 μM Fe-sulfate; 5 mM HpgCV) was added to the supernatant, followed by an incubation for 30 minutes at 25°C. 125 μl of a diluted M. /t/tet/s-culture in 2xTY (final OD ⁶⁰⁰ is 0.01 ) was added. The plate was incubated overnight at 30 ⁰ C. The OD ⁶⁰⁰ was read using a microplate reader.

18 clones, 16 originating from the V185 sub-library and 2 from the recombined sub-library were selected from the site directed A. nidulans library. All these clones showed a complete inhibition of the growth of M. luteus, indicating that a compound was formed inhibiting the growth of the indicator strain. This inhibition was retested according to the above protocol, and again positive. Moreover, the inhibition was absent upon the addition of Penase ^® , indicating that the inhibitory compound is a β- lactam.

The positive clones were grown in 10 ml 2xTY with 50 μg zeocin per ml. As a control source of IPNS, pBAD-AnpcbC was included in the assay. After overnight growth at 37°C and 280 rpm, the OD ⁶⁰⁰ was determined. 100 ml 2xTY medium containing 50 μg zeocin per ml was inoculated to an initial OD ⁶⁰⁰ of 0.015. The cells were grown at 37°C and 280 rpm until the OD ⁶⁰⁰ reached a value of 0.4 to 0.6. Production of the fusion protein MBP-IPNS was induced by the addition of L-arabinose to a final concentration of 0.2%. Growth was maintained at 22°C and 220 rpm, overnight. Cells were harvested and washed and frozen at -20 ⁰ C overnight.. Cell lysates were prepared by the addition of 1.5 ml Lysis Buffer (50 mM HEPES pH 7.0; 0.5 mg/ml lysozyme; 0.1 mg/ml DNAsel; 5 mM DTT; 5 mM MgSO4), followed by an incubation at room temperature for 30 minutes. The extract was centrifuged and the supernatant (CFE) was pipetted to a fresh tube. Two bioassays were done:

1 ) A liquid bio-assay: 8 μl of the CFE was mixed with 592 μl of assay mix (50 mM HEPES pH 7.0; 1 mM TCEP; 100 μM Na-ascorbate; 25 μM Fe-sulfate; 5 mM HpgCV). After incubation for 10 minutes at 25°C, 125 μl of a diluted M. /t/tet/s-culture in 2xTY (final OD ⁶⁰⁰ is 0.01 ) was added. As control substrates, ACV was used (both LLD- and DLD-PgCV may be used, although DLD-PgCV is the preferred substrate). In addition, for each sample an extra control was included, in which Penase ^® was added to the assay-mix. After overnight growth at 30 ⁰ C, the OD ⁶⁰⁰ was read.

2) A bio-assay on agar-plates: 8 μl of the CFE was mixed with 592 μl of assay mix (50 mM HEPES pH 7.0; 1 mM TCEP; 100 μM Na-ascorbate; 25 μM Fe-sulfate; 5 mM HpgCV). After incubation for 10 minutes at 25°C, 50 μl of the assay-mix was spotted in an agar-plate in which wholes were made. Of each sample an extra control was included, in which Penase ^® was added to the assay-mix. The agar consisted of 2xTY-agar supplemented with M. luteus to a final OD ⁶⁰⁰ of 0.01. After overnight

incubation at 30 ⁰ C, the presence or absence of cleared zones, due to the action of β- lactams, was scored.

Table 5 shows an overview of the activity of some of the mutants. Each mutant represents a larger group of mutants having a comparable IPNS activity on HpgCV as a substrate. The activity of the wild type IPNS enzymes from A. nidulans en S. clavuligerus was below the detection limit in these screening assays.

Table 5: Overview of the bioassays on the IPNS activity of the selected clones and controls.

All the positive clones in the screening were subjected to sequence analysis and representative clones were subjected to LC/MS analysis. LC/MS analysis was performed allowing the identification of the formed products based on retention time, the mass of the compound and the fragmentation pattern of the product. The amount of amoxicillin that is produced was determined by the standard addition method and is expressed in AU/mg protein, wherein AU represents the amount of amoxicillin (in ng) formed in 10 minutes under the specified reaction conditions. Table 6 summarizes the clones, the changed amino acid residues and the activity of the clones on HpgCV (in AU amoxicillin/mg protein in the E.coli CFE).

Table 6: Overview of the mutations and activity on HpcCV of the active IPNS mutants

n.d. = not detectable

Example 14

Conversion of DLD-PgCV to ampicillin

The activity of the mutant pαbC-clones 36A4 (V185R), 35A12 (V185K), 35F1 (V185K, V293A), 36F3 (T1 1 1A, H135R, V185R), 35H12 (V185K, P190H), 36C4 (V185H, T331 C) and the wild-type pcbC-gene from A. nidulans was determined on DLD-PgCV as a substrate. To this end, the clones were cultured in 10 ml 2xTY with 50 μg zeocin per ml. After overnight growth at 37°C and 280 rpm, the OD ⁶⁰⁰ was determined. 100 ml 2xTY medium containing 50 μg zeocin per ml was inoculated to an initial OD ⁶⁰⁰ of 0.015. The cells were grown at 37°C and 280 rpm until the OD ⁶⁰⁰ reached a value of 0.4 to 0.6. Production of the fusion protein MBP-IPNS was induced by the addition of L-arabinose to a final concentration of 0.2%. Growth was maintained at 22°C and 220 rpm, overnight. Cells were harvested and washed and frozen at -20 ⁰ C overnight. Cell lysates were prepared by the addition of 1.5 ml Lysis Buffer (50 mM HEPES pH 7.0; 0.5 mg/ml lysozyme; 0.1 mg/ml DNAsel; 5 mM DTT; 5 mM MgSO4), followed by an incubation at room temperature for 30 minutes. The extract was centrifuged and the supernatant (CFE) was pipetted to a fresh tube.

8 μl of the CFE was mixed with 592 μl of assay mix (50 mM HEPES pH 7.0; 1 mM TCEP; 100 μM Na-ascorbate; 25 μM Fe-sulfate; 5 mM PgCV). After incubation for 10 minutes at 25°C, 100 μl of a diluted M. /t/tet/s-culture in 2xTY (final OD ⁶⁰⁰ is 0.01 ) was added to 50μl of the reaction mixture. As a control substrate, ACV was used. In addition, for each sample an extra control was included, in which Penase ^® was added

to the assay-mix. After overnight growth at 30 ⁰ C, the OD ⁶⁰⁰ was read. All samples, even the wild-type A.nidulans IPNS fusion protein to MBP, resulted surprisingly in inhibition of growth of M. luteus.

LC/MS analysis was performed on these samples, allowing the identification of the formed products based on retention time, the mass of the compound and the fragmentation pattern of the product. The amount of ampicillin that was produced was determined by the standard addition method. In table 7, the activity on PgCV has been summarized.

Table 7: Activity of the mutant and wild-type MBP-IPNS fusion proteins on PgCV (in ng product per mg total protein, per 10 minutes) as determined by LC-MS analysis

Example 15 Construction of a Penicillium strain producing D-Hpg or L-Hpg

15a Construction of Penicillium expression plasmids

The genes of the hydroxymandelate synthase from Amycolatopsis oήentalis and Streptomyces coelicolor (hmaS), and the genes encoding hydroxyphenylglycine aminotransferase (hpgAT from Pseudomonas putida and hpgT from Streptomyces coelicolor respectively) were cloned in plasmid plPCLTA by PCR amplification of these genes from the different corresponding pBAD plasmids described in WO 02/34921. The gene mdlB, encoding mandelate dehydrogenase from Pseudomonas putida, was amplified from pGEM-Bldm. The insertion in the Nde\ and λ/s/ ^' l sites of of plPCLTA was chosen, leading to an ATG fusion of the appropriate gene with the Penicillium promoter. All genes were amplified by PCR, and the amplification primers were constructed by

introducing a Nde\ restriction site in the upstream primer and a λ/s/ϊ restriction site in the downstream primer. Because the genes hpgA T from Pseudomonas putida contains an internal λ/s/ϊ-site, an alternative approach was carried out, in which the insert was amplified using primers which introduced Nde\- and Ssal-sites. The PCR fragment with the additional Nde\/Bsa\ sites was subcloned as a blunt end fragment in the cloning vector pCR-blunt-TOPO (Invitrogen) leading to the construction of pCR-bl-hpgAT _Pp . After sequence determination revealed that it contained the correct insert including the flanking Nde\/Bsa\ sites, this plasmid was used for the construction of plPphpgATgWA by cloning the Nde\/Bsa\ fragment containing the hpgAT _Pp gene in the Nde\/Ppu'\ 0\ site of pi PCLTA.

In plPCLTA, the genes introduced in this vector are under control of the pcbC- promoter of Penicillium chrysogenum. In addition, the transcription terminator of the penDE-gene, encoding acyltransferase, has been used in these constructs.

The following expression cassettes were constructed:

The following primers were used in the cloning amplification of the genes mentioned above:

A: Amycolatopsis orientalis genes Construction of plAohmaSgWA hmaS-Ao-Nde: δ'-CAGGAGGAATTACATATGCAGAATTTCGAG (SEQ ID NO: 28)

hmaS-Ao-Nsi: δ'-CGGCCAGGGATGCATACGTCATCGCCGAGC (SEQ ID NO: 29)

B: Streptomyces coelicolor genes Construction of pIScohmaSgWA hmaS-Sc-Nde: δ'-CAGGAGGAATTACATATGCCGCCCAGTGAC (SEQ ID NO: 30) hmaS-Sc-Nsi: δ'-GAATTCCCATATGCATCCAGGTCATCGGCC (SEQ ID NO: 31 )

Construction of pIScohpgTgWA hpgT-Sc-Nde: δ'-CAGGAGGAATTACATATGACCACCACCACC (SEQ ID NO: 32) hpgT-Sc-Nsi: δ'-TCCCATATGCATCCTCAACCGTTAGACGCC (SEQ ID NO: 33)

C: Pseudomonas putida genes Construction of pIPpmdIBgWA hpgT-Sc-Nde: δ'-GTGAGGTAACATATGAGCCAGAATCTCTTT (SEQ ID NO: 34) hpgT-Sc-Nsi: δ'-GTAATCAATGCATCACTCATGCGTGTGTTC (SEQ ID NO: 35)

Construction of pIPphpgATgWA hpgAT-Nde δ'-CAGGAGGAATTACATATGTCTATTTATAGC (SEQ ID NO: 36) hpgAT-Bsal 5'-GTCCTCGGTCTCATGCλT CTCGAGTTAGCCCAGGAGGT (SEQ ID NO: 37)

15b Transformation of Pencillium chrysogenum

The plasmids plAohmaSgWA, pIScohmaSgWA, pIPpmdIBgWA, pIPphpgATgWA and pIScohpgTgWA were cut with Not\, which excises the expression cassette from the vector.

Penicillium chrysogenum DS12975, an industrial penicillin-producing strain from DSM, was transformed with 4 μg of DNA (i.e. the Notl-digests) and 0.25 μg of a selection marker fragment. In this case, a fragment was used which confers resistance to phleomycin.

The following combinations of expression cassettes were transformed:

Selection of transformants was done on mineral medium agar plates with 1 gram per litre and 1 M saccharose. Phleomycin resistant colonies appearing upon protoplast regeneration were re-streaked on fresh phleomycin agar plates without saccharose and grown until sporulation.

Transformants were streaked on the same medium and integration of the expression cassettes was verified by PCR, using specific primers for the individual genes.

Transformants containing the three intended genes were streaked on agar slant tubes and sporulated on rice.

The following strains were selected for further analysis:

15c Fermentative production of D-Hpg by Penicillium transformants

Penicillium chrysogenum transformants AFF108, AFF133, AFF146 and AFF152 were cultivated on shake flasks for 6 days. Intracellular and extracellular products were determined by NMR.

Table 8: Intracellular D-Hpg concentrations detected by NMR. Cells were grown for 3 to 6 days.

Cells were grown for 3, 4, 5, and 6 days to find an optimum for D-Hpg production. The concentration after 3 days was high enough to detect peaks in the NMR spectrum and to quantify the D-Hpg content. All selected transformants produced detectable levels of D-Hpg but the highest D-Hpg concentration was observed in AFF133 (80 μg/ g DW). The results are listed in table 8.

15d Fermentative production of L-Hpg by Penicillium transformants

Penicillium chrysogenum strains AFF158, AFF162, AFF164, AFF191 , AFF193 and AFF195, containing the L-Hpg biosynthesis genes, were cultivated on shake flasks for 3, 4, 5 and 6 days. Intracellular and extracellular products were determined by NMR. Again, no extracellular product could be detected but intracellular^, substantial levels of L-Hpg were perceived (see table 9).

The intracellular L-Hpg levels could be quantified for all strains and are listed in table 6. Unlike the D-Hpg levels, the L-Hpg levels increased in time and reached an optimum around 5 or 6 days of growth. In AFF195 mycelium the highest concentration was measured which was 2,6 mg L-Hpg per gram dry weight. The L-Hpg producing strains gained an approx. 30 times higher intracellular concentration than the AFF133, the highest D-Hpg producer.

Table 9: Intracellular L-Hpg concentrations detected by NMR.

In a next experiment the cells were cultivated for 5, 6 and 7 days however, after 5 or 6 days the L-Hpg concentration did not increase further but started to decrease (data not shown).

Example 16 Fermentative production of amoxicillin by P. chrysogenum

A Pencillium chrysogenum strain devoid of the penicillin biosynthetic clusters as discribed in patent application PCT/EP2007/054045 was transformed with combinations of 1 ) the plasmid pEXPR-NRPS6 (see table 1 ) harbouring an NRPS-gene encoding

HpgCVS (DLD-Hpg-Cys-Val Synthetase), 2) a plasmid harbouring the gene encoding the IPNS mutant V185R and 3) a plasmid harbouring the gsp-gene from Bacillus brevis, encoding a Ppant-transferase (PPT), under control of the gpαW-promoter. Selection of transformants was done on mineral medium agar plates with 50 μg phleomycin per ml and 1 M saccharose. Phleomycin resistant colonies appearing upon protoplast regeneration were re-streaked on fresh phleomycin agar plates without saccharose and grown until sporulation.

Of each transformation about 100 colonies were grown on mineral medium. Additionally, the amino acid L-Hpg was added at a concentration varying from 1 mM to 10 mM. Cultivation was carried out on shake flask scale or in 24-well microtiter plates. Incubation was done at 25°C for one week at 200 rpm.

At several time points (4 days tot 7 days), samples were taken. The broth was centrifuged in order to remove the mycelium. A total broth sample was removed and

freeze-dried. Both the freeze-dried broth sample and the supernatant were analyzed for the presence of amoxicillin.

In case of the supernatant, 10 and 50 μl of the supernatant was added to 190 and 150 μl respectively of a diluted M. luteus culture, as described in example 2. As a control, Penase ^® was added to the sample, which degrades amoxicillin. Several transformants showed an inhibition of growth of M. luteus, indicating the action of amoxicillin produced by the transformant. The addition of Penase ^® prevented the inhibition of growth of M. luteus.

The freeze-dried broth samples were treated with 100 - 500 μl of hot milliQ water (90 ⁰ C; volume depending on the amount of biomass), in order to make a lysate. 10 μl and 50 μl of the lysate were mixed with 150 μl and 190 μl of a diluted M. luteus culture respectively, as described above. A Penase ^® control was added as well. Also in this case, several samples did inhibit the growth of M. luteus. The positive extracts corresponded to the supernatant samples. The identity of the β-lactam that was formed was confirmed to be amoxicillin by LC/MS/MS, as described in example 9.