Peptide Bond Formation in Nonribosomal Peptide Biosynthesis

Recently, considerable insight has been gained into the modular organization and catalytic properties of nonribosomal peptide synthetases. However, molecular and biochemical aspects of the condensation of two aminoacyl substrates or a peptidyl and an aminoacyl substrate, leading to the formation of a peptide bond, have remained essentially impenetrable. To investigate this crucial part of nonribosomal peptide synthesis, an in vitro assay for a dipeptide formation was developed. Two recombinant holomodules, GrsA (PheATE), providing d-Phe, and a C-terminally truncated TycB, corresponding to the first, l-Pro-incorporating module (ProCAT), were investigated. Upon combination of the two aminoacylated modules, a fast reaction is observed, due to the formation of the linear dipeptided-Phe-l-Pro-S-enzyme on ProCAT, followed by a noncatalyzed release of the dipeptide from the enzyme. The liberated product was identified by TLC, high pressure liquid chromatography-mass spectrometry, 1H and 13C NMR, and comparison with a chemically synthesized standard to be the expectedd-Phe-l-Pro diketopiperazine. Further minimization of the two modules was not possible without a loss of transfer activity. Likewise, a mutation in a proposed active-site motif (HHXXXDG) of the condensation domain giving ProCAT(H147V), abolished the condensation reaction. These results strongly suggest the condensation domain to be involved in the catalysis of nonribosomal peptide bond formation with the histidine 147 playing a catalytic role.

Microorganisms use the nonribosomal pathway to synthesize a large group of structurally diverse and often complex secondary metabolites with biological activities, including antibiotics, siderophores, biosurfactants, immunosuppressants, and antitumor and antiviral agents. These low molecular weight peptide-based compounds are synthesized by means of multifunctional enzymes, the peptide synthetases, which can recruit not only proteogenic amino acids but also a large number of unusual amino acids and hydroxy acids that form peptide and ester bonds. The incorporated constituents can be further altered by epimerization or N-methylation, and the peptide backbone can be acylated, glycosylated, or cyclized (1)(2)(3). Study of the peptide synthetases is crucial to understanding the complex biosynthesis of bioactive peptides and should produce useful tools for synthesizing novel compounds (1,(3)(4)(5)(6). In recent years, considerable progress has been made in elucidating the modular structure of peptide synthetases by sequencing several of the biosynthetic gene clusters. The results have already enabled us to create novel products with biological activity through genetic engineering (4,5). Also, recombinant proteins have been utilized to gain insight into the domain architecture and the mechanisms of substrate activation (6 -11). On the other hand, most of the mechanistic aspects underlying the processes of condensation, epimerization, termination and interaction between synthetases have remained essentially elusive on the molecular level (1,2,12).
Peptide synthetases are composed of modules, which contain all enzymatic activities to incorporate one constituent into the product. The modules are in co-linear arrangement with the primary structure of the product (see Fig. 1), and they can be subdivided into different domains responsible for the single chemical reactions (1,8). The adenylation domain recognizes a substrate amino acid and activates it in the form of an acyl adenylate at the expense of ATP (8,9,11,13,14). Subsequently, the activated substrate is covalently bound to the thiolation domain or peptidyl carrier protein by a thioester linkage of the acyl group with the thiol moiety of the co-factor 4Ј-phosphopantetheine (P-pant) 1 (8,15,16). The latter is covalently attached to the side-chain hydroxyl moiety of an invariant serine residue within the peptidyl carrier protein (17). According to the multiple carrier thiotemplate model (18), elongation then occurs by swinging of the loaded P-pant arm to a donor site, where a peptide bond is formed between the activated acyl group and the free amino (or imino, hydroxy) group of a second amino acid, which is loaded in a similar manner onto the next module and positioned at an acceptor site (2,12,18). Passed on in this fashion from the first to the last module of the enzyme complex, the peptide chain grows in an N-to C-terminal direction from the starter amino acid until it reaches its final length. During the synthesis, all intermediates remain covalently bound to the enzyme complex (19). Upon release, which is in most cases thought to be mediated by a thioesterase domain, the linear peptide can also be branched, cyclized, or oligomerized (1, 2). * The work was supported by the Deutsche Forschungsgemeinschaft, EG project Cell Factories and the Fonds der Chemischen Industrie. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  In our effort to understand the mechanistic details of the manifold reactions catalyzed by peptide synthetases, we have been particularly interested in the elongation process, because the reaction of peptide bond formation and the responsible site(s) have not yet been characterized. Sequence analysis, however, has identified the so-called condensation domain as a good candidate for this site (20). This domain is found as a part of the repetitive modules and coincides in frequency with the number of peptide bonds that have to be formed for the linear peptide of final length. Condensation domains are always fused to the N-terminal end of modules accepting acyl groups from the preceding module (elongation modules), and they are absent in modules activating the first acyl constituent to be incorporated (initiation modules) (see Fig. 1) (1,20). With respect to the multiple carrier model, the condensation domains are expected to harbor an acceptor and a donor site for the loaded P-pant carriers of the two adjacent modules (12,18). A data base search for homologues of the condensation domain (which is about 450 amino acids in length) reveals at first sight no significant similarities to protein sequences other than peptide synthetases (for one possible homologue, see "Discussion"). However, the domains contain a sequence motif (HHXXXDG, His motif) that resembles the active-site motif of well studied acyl transferases, such as dihydrolipoyl transacetylase and chloramphenicol acetyltransferase (1,20). For these enzymes, the second histidine has been shown to act as the catalytic base in the course of the respective reaction (21)(22)(23)(24)(25)(26).
Attempts to study the reaction of peptide bond formation are confronted by some experimental difficulties. It is one of the reactions performed in multiple number, and it requires the interaction of at least two modules. Furthermore, the peptide synthetases are members of the secondary metabolism, which makes them hard to purify in large quantities, and their enormous size makes them difficult to handle. On the other hand, recombinant enzymes (fragments) were found not to be modified into their active holo-forms due to a missing P-pant transferase activity in the heterologous host (8,10,27). Very recent discovery of a superfamily of P-pant transferases, including those specific for modifying peptide synthetases, has provided us with the tools to investigate the complex mechanisms of nonribosomal peptide synthesis using recombinant proteins (12,17,28).
In this work, we have developed a system to study the process of nonribosomal peptide bond formation. It is based on the initiation reaction of gramicidin S and tyrocidine synthesis but has been simplified by enzyme fragmentation to a single condensation event. We show that the condensation domain indeed plays a catalytic role and that the His motif is part of (one of) the active site(s).
PCR products were purified with the QIAquick-spin PCR purification kit (Qiagen), digested with NcoI and BamHI, and ligated into the His 6 -tag vector pQE60, which was cut in the same manner. Standard procedures were applied for all DNA manipulations (29) and the preparation of the recombinant plasmids using Escherichia coli strain XL1-Blue (30). Cloning in pQE60 yielded the plasmids pPheAT (using oligonucleotides 1 and 2), pPheATE (1 plus 3), pProCA (4 plus 6), pProCAT (4 plus 7), and pProAT (5 plus 7). The plasmid pProCAT(H147V) was obtained by site-directed mutagenesis of pProCAT using inverse PCR. The following primers were 5Ј-phosphorylated and used to amplify the entire plasmid pProCAT: 5Ј-H147V, 5Ј-TGA TCC TCA TGG ACG GCT GG-3Ј; 3Ј-H147V, 5Ј-CGT GAA AGC TCC AGA TGA CC-3Ј. The mutations (underlined) introduced with the primers effect an H(CAC)147V-(GTG) substitution within the corresponding gene product and a silent H(CAT)146H(CAC) mutation to generate a PmaCI restriction site that allowed a simple detection of the mutated plasmid. The fusion sites between vector and inserts of the constructed plasmids as well as the mutation of the plasmid pProCAT(H147V) were confirmed by sequencing using the ABI prism 310 Genetic Analyzer (ABI).
Expression and Purification of Functional Holopeptide Synthetase Fragments-To achieve in vivo production of functional holopeptide synthetase modules, the P-pant transferase gene gsp (31) was co-expressed. For this purpose, the PvuII fragment of pGsp ϩ (31), containing the gsp gene under the control of the T7 promoter, was cloned into the SmaI site of the plasmid pREP4, yielding the helper plasmid pREP4/ gsp. Normally, pREP4 facilitates a constitutive production of the lac repressor protein LacI in order to ensure tight regulation of pQE expression vectors. Both pREP4/gsp and the pQE60 derivatives carrying the peptide synthetase fragments were transformed for protein expression into E. coli BL21(DE3). This strain bears the isopropyl-␤-D-thiogalactopyranoside-inducible gene of the T7 polymerase on a prophage DE3 within its chromosome (32). This arrangement allowed the coexpression of peptide synthetase fragments and the P-pant transferase Gsp. Expression of the recombinant constructs and purification of the His 6 -tagged proteins was carried out as described previously (9,16). The fusion of a His 6 tag results in appending the amino acid sequence "GSRSHHHHHH" at the C terminus of the recombinant protein. As judged by SDS-polyacrylamide gel electrophoresis (33), the proteins could be purified to apparent homogeneity using single-step Ni 2ϩ affinity chromatography. Fractions containing the recombinant proteins were pooled and dialyzed against assay buffer (50 mM HEPES, pH 8.0, 100 mM sodium chloride, 10 mM magnesium chloride, 2 mM dithioerythritol, and 1 mM EDTA). After the addition of 10% glycerol (v/v), the proteins could be stored at Ϫ80°C with no observable loss of activity even after several months. Protein concentrations were determined using the calculated extinction coefficients for the absorbance of these proteins at 280 nm (  Peptide synthetase are a set of modules, each of which incorporates one amino acid constituent into the product. The modules are in a colinear arrangement with the primary sequence of the synthesized peptide. Substrate amino acids are activated and bound to the P-pant co-factor of each module, from where elongation occurs. The modules can be subdivided into domains. An initiation module consists of an adenylation (black circle) and a thiolation (striped) domain. An elongation module is built of an N-terminal condensation (white), an adenylation, and a thiolation domain. Additional domains for substrate modification, such as epimerization (gray), are optional (as shown for the initiation module).
was performed as described above using the following 5Ј-modified oligonucleotides: 5Ј-Bli-SphI, 5Ј-ATA GCA TGC AAG CAG AAA AAA TAT CTC ATG-3Ј; 3Ј-Bli-BglII, 5Ј-ATA AGA TCT TGA CAG TTC AGC GCA CG-3Ј. The amplified DNA was purified, digested with SphI and BglII, and ligated into the His 6 tag vector pQE70. The resultant expression plasmid, pBli, was transformed in E. coli M15[pREP4], and subsequent expression and purification of the recombinant protein was carried out as described previously (16). To maintain the overexpressed protein in solution, the sonication buffer was modified and contained 50 mM HEPES (pH 8.0), 500 mM sodium chloride, and 10 mM ␤-mercaptoethanol. As judged from a Coomassie-stained SDS-polyacrylamide gel, the 28-kDa protein Bli could be purified to near homogeneity. The relevant fractions were pooled and dialyzed against modified assay buffer (50 mM HEPES, pH 8.0, 250 mM sodium chloride, 10 mM magnesium chloride, 10 mM ␤-mercaptoethanol, and 1 mM EDTA); the protein concentration was then determined using the calculated A 280 of Bli (31,750 M Ϫ1 cm Ϫ1 ). In vitro incorporation of the co-factor P-pant was performed as described elsewhere (17,35) in assay buffer containing 500 nM peptide synthetase (fragment), 10 nM Bli, and 0.1 mM coenzyme A (CoASH). The extent of apo to holo conversion was judged by the rate of thioester formation prior to and after the in vitro modification.
ATP-pyrophosphate Exchange Assay-The ATP-pyrophosphate exchange reaction was carried out to confirm the adenylation activity of all recombinant peptide synthetase fragments purified. Reaction mixtures contained the following (final volume of 100 l): 50 mM HEPES, pH 8.0, 100 mM sodium chloride, 10 mM magnesium chloride, 2 mM dithioerythritol, 1 mM EDTA, 0.5 mM amino acid, and 250 nM enzyme. The reaction was initiated by the addition of 2 mM ATP, 0.2 mM tetrasodium pyrophosphate, and 0.15 Ci (16.06 Ci/mmol) of tetrasodium [ 32 P]pyrophosphate (NEN Life Science Products) and incubated at 37°C for 10 min. Reactions were quenched by adding 0.5 ml of a stop mix containing 1.2% (w/v) activated charcoal, 0.1 M tetrasodium pyrophosphate, and 0.35 M perchloric acid. Subsequently, the charcoal was pelleted by centrifugation, washed once with 1 ml of water, and resuspended in 0.5 ml of water. After the addition of 3.5 ml of liquid scintillation fluid (Rotiscint Eco Plus; Roth), the charcoal-bound radioactivity was determined by liquid scintillation counting (LSC) using a 1900CA Tri-Carb liquid scintillation analyzer (Packard).
Thioester Formation: Radioassay for the Detection of Covalent Amino Acid Incorporation-Thioester formation to monitor the amount of covalent amino acid binding was basically performed as described previously (16). Reaction mixtures in assay buffer contained 500 nM enzyme, 2 mM ATP, and 3 Ci of L-[4-3 H]phenylalanine (27.0 Ci/mmol) or L-[5-3 H]proline (24.7 Ci/mmol) purchased from Amersham/Buchler. The assay was initiated by the addition of the cognate radiolabeled amino acid. For the determination of enzyme kinetics, samples (100 L) were removed at various time points from 5 to 600 s, and the reaction was quenched immediately by the addition of 800 l of chilled 10% (w/v) trichloroacetic acid. The samples were collected on mixed ester filters (ME25; Schleicher & Schü ll) and washed with 10% trichloroacetic acid, and the acid-stable label was then quantified by LSC.
Dipeptide Formation: Radioassay for the Detection of D-Phe-L-Pro Dipeptide Formation-The formation of the dipeptide was analyzed using different combinations of derivatives of GrsA and the first module of TycB. To ensure a complete acylation of the peptide synthetase fragments with their cognate amino acids, a preincubation was carried out for 3 min at 37°C; 1 M PheATE or PheAT was incubated in assay buffer containing 2 mM ATP and 4.4 M of L-[4-3 H]phenylalanine (27.0 Ci/mmol) (mixture A); ProCAT, ProCA, ProAT, or ProCAT(H147V) was incubated in assay buffer containing 1 M enzyme, 2 mM ATP and 0.5 mM proline (mixture B). Alternatively, labeled phenylalanine was replaced by 0.5 mM of the unlabeled amino acid, and/or the unlabeled proline was substituted by 4.8 M of L-[5-3 H]proline (24.7 Ci/mmol). The condensation reaction was initiated by the addition of 1 volume of mixture B to 1 volume of mixture A in the different combinations delineated in Figs. 4 and 6. At time points from 15 to 600 s, 100-l samples were taken and immediately quenched by the addition of 0.8 ml of chilled 10% trichloroacetic acid. Trichloroacetic acid-precipitable label was collected and quantified as described above.
In order to analyze the nature of the product(s), samples (100 L) were removed at time points from 1 to 60 min, diluted by adding 400 l of water, and immediately extracted with butanol/chloroform (4:1 (v/v)). The organic phases were transferred to fresh tubes and washed once with 500 l of 0.1 M sodium chloride. Nonradioactive samples were prepared in a 10-fold up-scale of the radioactive assay. After removal of the solvent under vacuum, the residue was further investigated by TLC, HPLC-MS, and 1 H and 13 C NMR (see below).
Identification of the Extracted Product Diketopiperazine-For TLC separation, radiolabeled samples were resolved in 20 l of 90% ethanol (v/v), applied to silica plates (Silica gel 60 F 254 ; Merck) and run in solvent system A: ethyl acetate/pyridine/acetic acid/water (90:30:9:16 (v/v)). Labeling was visualized by autoradiography (X-Omat AR; Kodak) and/or by scanning on a LB2723 thin layer scanner II (Berthold). Nonradioactive samples prepared were resolved in 10% of buffer B and analyzed by HPLC using a C18 reversed-phase column (Nucleosil 3 ϫ 250 mm, pore size 120 Å, particle size 3 mm; Macherey & Nagel) on a Pharmacia HPLC system monitoring at detector wavelengths of 214 and 256 nm simultaneously. The following gradient profile was used at a flow rate of 0.35 ml min Ϫ1 : 10 s loading (10% buffer B), linear gradient up to 30% buffer B in 1 min, followed by a linear gradient to 100% buffer B in 20 min, and then holding 100% buffer B for 10 min (buffer A, 0.05% formic acid in H 2 O; buffer B, 0.04% formic acid in methanol).
HPLC-MS of nonradioactive samples was performed on a Gynkotek HPLC system and a ESI source (Analytica of Branford Inc.) as an interface to a HP 5989 Hewlett Packard mass spectrometer in negative mode. Sample preparation and conditions for the HPLC were the same as described above.
NMR spectra of nonradioactive samples in CDCl 3 were obtained on a Bruker AMX 500 spectrometer. The methods described above for product analysis were also carried out using an authentic standard of D-Phe-L-Pro-diketopiperazine, which was chemically synthesized by Dr. P. Quadflieg at DSM Research.

Strategy and Construction of Recombinant Peptide
Synthetase Fragments-In our search for a system to study peptide bond formation in nonribosomal peptide synthesis we demanded two criteria: 1) only a single elongation event was desired, and 2) the product should be liberated from the enzyme(s) after condensation to ensure multiple turnovers and thereby to facilitate product detection and analysis. All nonribosomal systems sequenced so far, however, assemble more than two constituents, and termination and product release are not well understood. Taking advantage of a peculiarity previously reported in gramicidin S and tyrocidine synthesis, we attempted to construct a recombinant enzyme system that would fulfill our criteria. Gramicidin S is formed on the pentamodular GrsAB complex via the linear pentapeptide D-Phe-Pro-Val-Orn-Leu, which is then dimerized head-to-tail to the final product (36). The cyclic decapeptide tyrocidine is formed by cyclization of the linear intermediate D-Phe-Pro-Phe-D-Phe-Asn-Gln-Tyr-Val-Orn-Leu on the decamodular complex Ty-cABC (37,38). GrsA and TycA are single module peptide synthetases incorporating D-Phe (39,40). The first module of GrsB (which consists of four modules) and TycB (three modules; TycC contains the remaining six modules) incorporates L-Pro (9,41). The initiation reaction in both systems is the formation of the first peptide bond between D-Phe and L-Pro (42). Interestingly, in both cases, this dipeptide intermediate is cleaved off the enzyme template by intramolecular cyclization when the systems are not supplied with the substrate amino acid of the next module (43,44). In a reaction that is probably noncatalytic, the cyclic diketopiperazine (DKP; 3-benzyl-(8-ar)-hexahydro-pyrrolo[1,2-␣]pyrazine-1,4-dione) is formed by an intramolecular attack of the amino group of D-Phe on the thioesteractivated carboxyl of Pro. We therefore reasoned that it was possible to construct an appropriate system based on the two modules required for this initiation reaction. Since it was shown that TycA and GrsA are exchangeable during this reaction (45), the entire GrsA, which activates and epimerizes L-Phe, and the first module of TycB, which activates L-Pro, were used (Fig. 2).
PheATE, the recombinant His 6 tag derivative of GrsA, consists of one module containing three domains (residues 1-1098) (39). It harbors an N-terminal Phe-activating adenylation (A) domain, a peptidyl carrier protein, or thiolation (T) domain, and at its carboxyl terminus an epimerization (E) domain, which probably catalyzes L to D conversion of the thioesterified L-Phe (8). TycB, consisting of three modules (3588 amino acids), was dissected to obtain the enzyme fragment ProCAT, which corresponds to the first module (residues 1-1049) and includes a postulated N-terminal condensation (C) domain, an L-Proactivating adenylation (A) domain, and a thiolation (T) domain (9). Truncated derivatives of these two modules were constructed as illustrated in Fig. 2A. Deletion of the epimerization domain of PheATE yielded the protein PheAT (residues 1-612), comprising only the adenylation and the thiolation domain. The proteins ProAT (residues 442-1049) and ProCA (residues 1-989) lack the condensation domain and the thiolation domain, respectively, of the single-module ProCAT. Additionally, the mutant ProCAT(H147V) was constructed, exhibiting a valine substitution of a presumed active-site histidine residue (position 147) in the condensation domain. All recombinant proteins carry a C-terminal His 6 tag and were obtained using PCR as described under "Experimental Procedures." In Vivo Apo to Holo Conversion of Peptide Synthetase Fragments-Peptide synthetase fragments expressed in a heterologous host like E. coli lack the P-pant modification to their active holo-form, a circumstance that was recently explained by the lack of a cognate P-pant transferase (17). To overcome this problem, we integrated the gsp gene, encoding the specific P-pant transferase Gsp of the gramicidin S biosynthetic system (31), in the helper plasmid pREP4. Expression of Gsp in the co-expression system has been proven by SDS-polyacrylamide gel electrophoresis and Western blotting (data not shown). The stoichiometric conversion of the peptide synthetase fragments from apo to holo form was tested by additional incubation with co-substrate CoASH and Bli, the P-pant transferase associated with the bacitracin biosynthetic system (34). Nonincubated and incubated samples were applied to the thiolation assay and examined for a detectable increase in the formation of a covalent, acid-stable complex with [ 3 H]Phe or [ 3 H]Pro. No additional label could be incorporated into the holoenzymes, indicating their complete phosphopantetheinylation under the given expression conditions (data not shown). As a control, the assay was performed with PheAT expressed without Gsp, which revealed a low level of [ 3 H]Phe incorporation (approximately 8%). After in vitro holo-conversion of this substoichiometrically modified PheAT (using Bli and CoASH), an increase in the formation of acid-stable label (over 85%) could be observed, indicating that PheAT can be modified by Bli. It is remarkable that peptide synthetase fragments can be phosphopantetheinylated by peptide synthetase-modifying enzymes of different origin (namely Sfp, Gsp, and Bli), whereas other Ppant-requiring enzymes possess a cognate partner enzyme responsible for apo to holo conversion, and frequently various P-pant transferases and P-pant-requiring enzymes are not compatible (17).
Expression, Purification, and Investigation of the Enzymatic Properties of Holopeptide Synthetase Fragments-Utilization of the helper plasmid pREP/gsp allowed the co-expression of peptide synthetase fragments and Gsp and ensured their stoichiometric modification. All recombinant proteins, namely PheATE (127 kDa), PheAT (71 kDa), ProCAT (120 kDa), ProCA (114 kDa), ProAT (70 kDa), and ProCAT(H147V) (120 kDa), were purified to apparent homogeneity using single-step Ni 2ϩ affinity chromatography. As shown in Fig. 2B, their observed sizes were consistent with the calculated molecular masses. Prior to studying nonribosomal peptide bond formation, the functionality and enzymatic properties of the purified peptide synthetase fragments were confirmed. For this purpose, they were first investigated in amino acid-dependent ATP-pyrophosphate exchange reactions. PheATE and PheAT exhibited high exchange activities for both Phe isomers in a range of approximately 25 nanokatals nmol Ϫ1 (taken as 100%; 1 nanokatal is the amount of enzyme catalyzing the incorporation of 1 nmol of pyrophosphate into ATP per second). Highest nonspecific activities were obtained for L-Trp (13%) and L-Leu (6%), whereas all other substrates (including L-Pro) and the control (without amino acid) were found to be below 2%. TycB derivatives exclusively activated L-Pro with an overall activity of about 12 nanokatals nmol Ϫ1 ; all non-cognate substrates (including Phe) were at a background level of under 1%. These data are in good agreement with previous studies, verifying results obtained for different GrsA deletion mutants (8) and TycB1(ProA) (9). Covalent aminoacylation was investigated under essentially the same conditions as used for the subsequent dipeptide formation. As an example, Fig. 3 shows the time courses for the thiolation of PheATE and ProCAT. Half-life times (t1 ⁄2 ) were found in the range of 5 s for PheATE to 30 s for ProCAT, and no significant alterations in the velocity of covalent substrate binding could be observed for the other peptide synthetase fragments investigated (PheAT, ProAT, and ProCAT(H147V); data not shown). In contrast, as expected due to the deleted peptidyl carrier protein (co-factor binding site), no significant incorporation of label could be detected in the case of ProCA. These results are consistent with prior investigations (46) and support the preceding data for the adenylation reaction, revealing an approximately 2-fold faster activation in the case of the GrsA derivatives. In conclusion, based on the domain structure of the peptide synthetase fragments constructed, all postulated activities could be attributed to the distinct fragments, and no significant influences of neighboring deletions on these particular functions have been observed.
Elongation Assay and Product Detection-Besides restriction to a single condensation event and release of the product, an additional advantage of investigating the initiation reaction of the tyrocidine biosynthetic pathway is that peptide bond formation can exclusively take place after combining both enzymes. We separately preincubated PheATE and ProCAT with their cognate amino acid and ATP to achieve a complete aminoacylation of both modules. After 3 min, a time interval that ensures a complete loading of both enzymes (see kinetic studies above and Fig. 3), we combined the two reaction mixtures and monitored formation of the dipeptide. Under these conditions, peptide bond formation should occur with the postulated transfer of the phenylalanyl moiety from PheATE to ProCAT, freeing up the thiolation site on PheATE, which could then be rapidly reacylated by phenylalanyl adenylate. We therefore speculated that simultaneous labeling of both PheATE and ProCAT would be detectable using [ 3 H]Phe. Fig. 4 shows the kinetics of increasing and decreasing levels of acid-stable [ 3 H]Phe over a time course of 13 min in the condensation assay. After autonomous thiolation of PheATE and ProCAT (stage I) and mixing of both enzymes, a fast reaction took place, nearly doubling the amount of acid-stable label within a period of 45 s (stage II) and with a t1 ⁄2 of approximately 11 s. As shown in control experiments, this reaction required the presence of both acylated enzymes and did not occur if any of the substrates (ATP, Phe, or Pro) was omitted from the reaction (data not shown). The additional label was acid-stable, indicating either the formation of a peptide bond or the covalent linkage to a thiol group (P-pant or cysteine residue) of the enzyme. After 45 s, the described reaction reached its maximum, and the label then began to be released from the enzyme(s) (stage III). The half-life time of this slower and presumably noncatalyzed reaction was estimated to be 180 Ϯ 5 s. In a parallel experiment using radiolabeled Pro, this label was also released from ProCAT at the same time and with the same velocity as observed before (not shown), indicating the formation and liberation of a product that is constituted of both phenylalanine and proline.
Product Analysis by TLC-To investigate the product formed and released by the interaction of PheATE and ProCAT, we carried out the condensation assay as described, diluted the samples at appropriate time intervals by the addition of water, and instantly extracted the radiolabeled product(s) using butanol/chloroform (4:1 (v/v)). Organic phases were prepared as described, applied to silica plates, and separated by TLC. Radiolabeled products were identified by autoradiography and radioscanning on a TLC scanner. As a control, we first investigated PheATE and ProCAT separately. ProCAT alone did not form any by-product that could be detected in the organic phase, whereas PheATE alone yielded at least three minor by-products (R F values of 0.15, 0.58, and 0.92 in solvent A; not shown), whose formation depended solely on the presence of phenylalanine. PheATE and ProCAT incubated together yielded a major spot with an R F of about 0.74 in addition to the by-products of PheATE. The appearance of this spot required the presence of all substrates, ATP, Phe, and Pro. In a radioactivity scan, the intensity of this spot was at least 1000 times higher than detected for the by-products of PheATE, and it could be increased when both amino acids were radiolabeled ([ 3 H]Phe and [ 3 H]Pro), indicating a product constituted of both amino acids. This product was already detectable after 1 min and rapidly accumulated in the organic phase during advanced incubation. After 45 min, more than 90% of the radiolabel in the reaction mixture could be recovered in the organic phase; thus, the occurrence and concentration of the detected product correlate with the disappearance of acid-stable label in the condensation assay (compare Fig. 4, stage III).
Product Analysis by HPLC-MS-To analyze the putative dipeptide by HPLC, a nonradioactive condensation assay was performed. Samples were extracted as described and applied to a reversed-phase chromatography. The HPLC chromatogram showed a peak at a retention time of about 18.8 min (Fig. 5A). The corresponding compound could be identified as Phe-Pro-DKP by comparison with the retention time of a chemically synthesized standard, as well as by its mass spectrum obtained on-line using ESI MS (Fig. 5C). This assignment was also supported by kinetic studies (Fig. 5B), because the intensity of the peak correlated with the disappearance of acid-stable label in the dipeptide reaction (compare Fig. 4) and is in agreement with the results obtained by TLC analysis. The steady increase of the Phe-Pro-DKP peak also reveals a constant rate of product formation, indicating multiple turnovers. The compound eluting at a retention time of 12 min was identified as Phe. The other signals may be assigned to the matrix but have not been further identified.
In the mass spectrum, the two major signals at m/z ϭ 245 and m/z ϭ 489 correspond to the monomer [M ϩ 1] and the dimer [2M ϩ 1], respectively, and confirmed the calculated molecular mass of Phe-Pro-DKP (244.28 g mol Ϫ1 ). Formation of a noncovalent dimer is a common phenomenon in this analyt- The absence of a third product with a molecular mass corresponding to a dichlorinated compound confirms the assignment of the signal at m/z ϭ 489 as the noncovalent dimer [2M ϩ 1] rather than a cyclic tetrapeptide. A mixture of both compounds, a cyclic di-and tetrapeptide, could also be excluded, since the HPLC analysis clearly showed only one product peak.
NMR Studies of the Product-For further identification of the product, fractions eluting from the HPLC column at the corresponding retention time were collected from several runs to give approximately 2 mg of nonradiolabeled Phe-Pro-DKP. The dried product was dissolved in 200 l CDCl 3 and subjected to 1 H and 13  Dipeptide Formation Depends on the Presence of a Functional Condensation Domain-To determine whether all the domains of PheATE and ProCAT are required to catalyze peptide bond formation, we utilized some further deletion mutants of both modules in the condensation assay. In a first experiment, the reaction was performed without the condensation domain, using PheATE and ProAT. As shown in Fig. 6B, no product was formed, indicating the general importance of the condensation domain for peptide bond formation. To clarify whether this finding resulted from a loss of required protein-protein interactions or from a lack of a catalytic activity, a single mutation (H147V) within the highly conserved His motif (HHXXXDG) of the condensation domain of ProCAT was generated and used in the condensation assay. Although retaining the catalytic activities of L-Pro adenylation and thioesterification, the mutant protein ProCAT(H147V) failed to catalyze the D-Phe to L-Pro transfer, when incubated with PheATE. We therefore can assign the condensation domain, by direct biochemical evidence, a catalytic role in the process of peptide bond formation. The histidine residue 147, which is strictly conserved within all condensation domains, is essential for this catalysis.
In the following experiments, we sought to investigate the participation of two other domains within the modules PheATE and ProCAT in the condensation process. The deletion mutant ProCA lacks the thiolation domain and was confirmed to be unable to thioesterify Pro. Considering the multiple carrier thiotemplate mechanism, this protein was predicted to be inactive in the condensation assay. Indeed, upon incubation with PheATE no product formation could be observed (Fig. 6C). The marginal decrease of acid-precipitable label observed (Fig. 6C) suggests a very slow hydrolysis of the PheATE-bound phenylalanine in the presence of ProCA. This may indicate an activity of ProCA (i.e. the condensation domain) to destabilize the thioester linkage between P-pant and phenylalanine. We did not further investigate this phenomenon.
Finally, the epimerization domain of GrsA was deleted, yielding the truncated module PheAT. Likewise, no product formation was detectable when the condensation assay was performed with this protein and ProCAT (Fig. 6D). A control using D-Phe instead of L-Phe as a substrate for PheAT showed that an epimerized Phe in thioester linkage is not sufficient to restore peptide bond formation. It was demonstrated previ-

FIG. 5. HPLC and HPLC-MS analysis of the D-Phe-L-Pro-diketopiperazine.
A, nonradioactive samples were prepared as described under "Experimental Procedures" and analyzed by HPLC using a C18 reverse-phase column. The product corresponding to the peak at a retention time of 18.8 min is only formed by PheATE and ProCAT together and cannot be detected using truncated modules. B, time dependence of the product formation correlates with the disappearance of acid-stable label in the condensation assay (compare Fig. 4) and is in agreement with the kinetic TLC data obtained. C, the mass spectrum of the peak at 18. ously that the D-isomer is an appropriate substrate for GrsA and derivatives (8,36). We therefore suspect that the deletion of the epimerization domain has led to a lack of accurate protein-protein interactions with ProCAT or interrupted other essential functions. DISCUSSION We present here a system to study the condensation reaction, i.e. the central reaction of peptide bond formation, in nonribosomal peptide synthesis on the molecular level using genetically engineered recombinant proteins. Our assay is based on the similar initiation reaction of the gramicidin S and tyrocidine biosynthesis. For these systems, it was known that omission of the third amino acid in order of their synthesis leads to a premature cleavage of the D-Phe-L-Pro dipeptide intermediate (43,44). Although this reaction is probably noncatalytic, it is favored by the conformational stability (49) of the prolinecontaining diketopiperazine and allows a slow turnover of dipeptide formation. The same reaction has also been observed with proteolytic fragments probably corresponding to the first two modules of the wild-type enzymes (43,44). The tetramodular GrsB was treated with proteases, and a fragment with L-Pro-activating activity was isolated. This fragment, which was shown to contain the authentic N terminus of GrsB, was estimated to be 114 kDa in size and accepted D-Phe from GrsA to form the D-Phe-L-Pro diketopiperazine (50). Likewise, a L-Pro-activating fragment of tyrocidine synthetase was obtained by prolonged incubation of crude cell extracts and upon incubation with TycA yielded the same product (44). Consistent with the similar initiation reaction, the corresponding modules in gramicidin S and tyrocidine synthesis are equivalently composed in terms of domain organization and share an especially high degree of sequence similarity (9,39,40,51). It has been shown that GrsA can replace TycA in tyrocidine biosynthesis (45).
These observations have led us to use recombinant GrsA, which is designated PheATE, and a recombinant protein corresponding to the first module of TycB, here referred to as ProCAT. To obtain the heterologously expressed modules in their active holo-form, the P-pant transferase Gsp was coexpressed from a second plasmid in E. coli. This method was recently reported (27) and similarly developed in our laboratory after the discovery of the superfamily of P-pant transferases, which render peptide synthetases catalytically active through post-translational modification with the co-factor P-pant (17).
Our results indicate that Gsp, which is the P-pant transferase of the gramicidin S biosynthesis system (31), apparently completely converts the expressed peptide synthetase modules into their holo-form, since a subsequent P-pant-transferase treatment in vitro did not further increase enzyme activity. Incubation of the two holo-modules with all substrates indeed resulted in dipeptide formation and subsequent cleavage of the D-Phe-L-Pro diketopiperazine (Fig. 7). The product was purified and identified by TLC, HPLC-MS analysis, 1 H and 13 C NMR studies, and comparison with a chemically synthesized standard. The results showed that the system consisting of PheATE and ProCAT is an appropriate model system for studying determinants of the D-Phe-to-L-Pro transfer reaction.
The main goal of our study was to determine whether or not the condensation domain is catalytically involved in peptide bond formation. For this purpose, we first deleted the condensation domain of ProCAT. The N-terminally truncated module ProAT did not form any dipeptide when incubated with PheATE. This result showed the indispensability of the condensation domain for dipeptide formation but could still be explained by the loss of protein-protein interactions. We therefore mutated the strictly conserved second histidine of the His motif HHXXXDG within the condensation domain of ProCAT to valine, giving the protein ProCAT(H147V). The mutant protein retained the catalytic activities to adenylate and thioesterify L-Pro, but it failed to accept D-Phe from PheATE in the condensation assay to form the D-Phe-L-Pro dipeptide. This finding is clear evidence for the catalytic role of the condensation domain in the process of peptide bond formation. It further provides precise experimental support for the importance of the second histidine of the His motif, which may act as a base during the nonribosomal condensation of amino acids (1,20). Mutations in the His motif in the fourth module of surfactin synthetase were recently shown to abolish the overall surfactin production in vivo and in vitro with purified enzymes (52). However, in this study it was not further analyzed which process exactly was interrupted.
The involvement of the His motif in the active site of condensation domains was previously suggested by sequence analysis (20). This motif was discovered to be a common feature with other families of acyltransferases, such as chloramphenicol acetyltransferase and the dihydrolipoyl transacetylases of the 2-oxoacid dehydrogenase complexes, suggesting a similar mechanism of acyl transfer (20). Nevertheless, the similarities with these acyltransferases are restricted to the His motif. Mutational analysis and crystallographic studies on these families indicated that the second histidine is in the active site acting as base to promote the nucleophilic attack of the C3hydroxyl moiety of chloramphenicol or the thiol moiety of CoASH on the thioester carbon atom of acetyl-CoA or the dihydrolipoamide acceptor, respectively (21,(23)(24)(25)(26). By analogy, histidine 147 of ProCAT would abstract the proton of the imino group of the thioesterified L-Pro to promote the nucleophilic attack on the carbonyl group of D-Phe bound to the P-pant of PheATE (Fig. 7). Given the similarity of all condensation domains yet reported and the different acyl precursors that are incorporated by these enzymes, such as amino, imino, hydroxy, and aryl acids, a common mechanism for amide, peptide, and ester bond formation in nonribosomal synthesis can be assumed. In general, the second histidine would then be involved in catalyzing the nucleophilic attack of the amino, imino, or hydroxy group of the activated acyl constituent in the acceptor site of the condensation domain on the carbonyl group of the incoming precursor at the donor site.
Interestingly, the condensation domain of peptide synthetases has only a few homologues in the data bases. One of these is an open reading frame, ORF2, of unknown function in Mycobacterium tuberculosis BCG (accession number Q02279), which displays about 25% similarity with the condensation domain of ProCAT. 2 ORF2 is located adjacent to the gene encoding mycocerosic acid synthase (54). Since this enzyme lacks the thioesterase domain, one can speculate that the gene product of orf2 functions in acyl transfer of the mycocerosic acid to phenolphthiocerol (55). In nonribosomal synthesis, the analysis of the sequenced biosynthesis clusters suggests that both condensation and thioesterase domains can catalyze amide, peptide, or ester bond formation (1,56). A condensation domain is located between each consecutive pair of activation units (adenylation and thiolation domains) and therefore seems to be always utilized for condensation steps required to build the linear intermediate. This supports the hypothesis of an acceptor and a donor site at the condensation domain to bind the P-pant carriers of the adjacent modules. The structurally and chemically more specialized termination reactions, such as cyclization or branching via amide (e.g. bacitracin), peptide (e.g. tyrocidine, gramicidin S, HC-toxin, cyclosporin), or ester bond (e.g. surfactin, pristinamycin, rapamycin); multimerization via ester bond (e.g. enterobactin, enniatin); or cleavage of the linear intermediate (e.g. ␦-(L-␣-aminoadipoyl)-L-cysteinyl-D-valine (ACV)) can apparently be carried out either by a thioesterase domain (all bacterial systems and ACV synthetase) or by a condensation domain (HC-toxin, cyclosporin, enniatin, rapamycin) (1). A novel class of specialized condensation domains was very recently discovered. This so-called cyclization domain is thought to catalyze both peptide bond and thiazoline ring formation (57).
In the system for peptide bond formation, we further asked the question of whether all the other domains of the two modules would be necessary for dipeptide formation. Since in the sequential order of reactions L-Phe epimerization to D-Phe occurs prior to peptide bond formation (only the D-isomer is transferred (58)), we deleted the epimerization domain of PheATE. The resulting PheAT lacks epimerization activity of L-Phe (8). PheAT also failed to trigger dipeptide formation when incubated with ProCAT. This is, however, not due to the inability to incorporate D-Phe, since PheAT supplied with D-Phe as a substrate was equally unable to initiate the reaction. In fact, D-Phe as a substrate is activated and thioesterified (8), and it also triggers dipeptide formation in the case of PheATE. Thus, we assume that the epimerization domain is involved in protein-protein interactions and/or in correct positioning of the D-Phe-P-pant carrier (Fig. 7).
Finally, we deleted the thiolation domain of ProCAT. This domain contains the attachment site of the co-factor P-pant (17,18). Correspondingly, the truncated fragment ProCA lost the ability to thioesterify its substrate L-Pro. As expected, this deletion also abolished dipeptide formation. This result is consistent with the multiple carrier model for nonribosomal peptide synthesis, according to which each module requires a thiolation domain equipped with the co-factor (18). In a similar context, it was shown for several modules of surfactin synthetase that mutation of the conserved serine residue within the thiolation domain, which serves as P-pant attachment site, abolishes surfactin production (59).
Our condensation assay indicates a stoichiometric thioesterification of all thiolation sites prior to peptide bond formation, which agrees with the multiple carrier thiotemplate model. In contrast, we cannot find any support for the alternative mechanism of peptide bond formation proposed for the synthesis of the tripeptide ACV by ACV synthetase (60,61). In this work, it was reasoned that peptide bond formation occurs without thioesterification of the valine residue (60). With respect to the primary structure of ACV synthetase, a typical peptide synthetase containing adenylation, thiolation, condensation, epimerization, and thioesterase domains, the same general mechanism of product formation should be expected.
The modular architecture of peptide synthetases allows one to attempt the targeted reprogramming of the enzyme templates by genetic means with the goal of synthesizing novel compounds (1). The inherent potential to do so has been demonstrated in initial work (4,5,53). Further exploitation of this potential will depend on a better understanding of the genetics and enzymology of the peptide synthetases. The molecular mechanism employed by the condensation domain in cata-2 V. de Crécy-Lagard, unpublished results. lyzing the elongation process in nonribosomal peptide synthesis will be a central aspect of future investigations.