Utility of epimerization domains for the redesign of nonribosomal peptide synthetases Daniel B. Stein, Uwe Linne and Mohamed A. Marahiel Fachbereich Chemie ⁄ Biochemie, Philipps-Universita ¨ t Marburg, Germany d-Configured amino acids are an important particular- ity occurring in various natural bioactive peptide com- pounds. These are, for example, peptides secreted by certain animals such as amphibians and spiders [1] or the antimicrobial lantibiotics which are synthesized ribosomally in bacteria [2]. Another group of numer- ous pharmacologically interesting peptides containing d-amino acids is produced in bacteria and fungi by a class of huge multienzymes, the modular nonribosomal peptide synthetases (NRPSs) [3,4]. Generally, NRPS assembly lines are constructed by a certain sequence of domains with specific catalytic functions. Essential core domains are the adenylation (A) domain for the selective acivation of amino acids Keywords epimerization; multienzyme; nonribosomal peptide synthetase; protein engineering Correspondence M. A. Marahiel, Fachbereich Chemie ⁄ Biochemie, Philipps-Universita ¨ t Marburg, Hans-Meerwein-Straße, 35032 Marburg, Germany Fax: +49 6421 2822191 Tel: +49 6421 2825722 E-mail: marahiel@chemie.uni-marburg.de (Received 23 May 2005, revised 11 July 2005, accepted 18 July 2005) doi:10.1111/j.1742-4658.2005.04871.x Many pharmacologically important agents are assembled on multimodular nonribosomal peptide synthetases (NRPSs) whose modules comprise a set of core domains with all essential catalytic functions necessary for the incorporation and modification of one building block. Very often, d-amino acids are found in such products which, with few exceptions, are generated by the action of NRPS integrated epimerization (E) domains that alter the stereochemistry of the corresponding peptidyl carrier protein (PCP) bound l-intermediate. In this study we present a quantitative investigation of substrate specificity of four different E domains (two ‘peptidyl-’ and two ‘aminoacyl-’E domains) derived from different NRPSs towards PCP bound peptides. The respective PCP-E bidomain apo-proteins (TycB 3 -, FenD 2 -, TycA- and GrsA-PCP-E) were primed with various peptidyl-CoA precur- sors by utilizing the promiscuous phosphopantetheinyl transferase Sfp. PCP bound peptidyl-S-Ppant epimerization products were chemically cleaved and analyzed for their l ⁄ d-ratios by LCMS. We were able to show that all four E domains tolerate a broad variety of peptidyl-S-Ppant-sub- strates as evaluated by k obs values and final l ⁄ d-product equilibria deter- mined for each reaction. The two C-terminal amino acids of the substrate seem to be recognized by ‘peptidyl-’E domains. Interestingly, the ‘amino- acyl-’E domains GrsA- and TycA-E were also able to convert the elongated intermediates. All four E domains accepted an N-methylated precursor as well and epimerized this substrate with high efficiency. Finally, we could demonstrate that the condensation (C) domain of TycB 1 is also able to process peptidyl substrates transferred by TycA. In conclusion, these find- ings are of great impact on future engineering attempts. Abbreviations A, adenylation domain; aminoacyl- or peptidyl-S-Ppant, aminoacylated thioester form of cofactor Ppant bound to the strictly conserved serine residue of PCPs; C, condensation domain; DKP, diketopiperazine; E, epimerization domain; ICR, ion cyclotron resonance; NRPS, nonribosomal peptide synthetase; PCP, peptidyl carrier protein (also refered to as T); PCP C , PCP normally localized in front of a C domain; PCP E , PCP naturally connected to an E domain; Ppant, 4¢ -phosphopantetheine; SIM, single ion mode; T, thiolation domain – also refered to as PCP but used for protein descriptions (‘one letter–one domain’ nomenclature of NRPSs); TE, thioesterase domain; TFA, trifluoroacetic acid. 4506 FEBS Journal 272 (2005) 4506–4520 ª 2005 FEBS as aminoacyl-O-AMP under ATP consumption [5], and the following peptidyl carrier protein [PCP, often also denoted as thiolation (T) domain] [6,7]. Post- translational introduction of an 4¢-phosphopantetheine (Ppant) cofactor forms the active holo-PCP which accepts the adenylated amino acid to yield an amino- acyl-S-Ppant-PCP [8,9]. Other core domains are the condensation (C) domain carrying out the peptide bond formation between two PCP bound S-Ppant intermediates [10], and a C-terminally located termin- ation domain for product release which is in most systems a thioesterase (TE) domain [11]. Besides other optional modifications, the incorpor- ation of d-amino acids into the product is a special fea- ture appearing in nonribosomally synthesized peptides. For NRPS d-configured amino acids can be provided by external racemases, then selectively be recognized, and activated by a specific A domain. Nevertheless, more commonly epimerization (E) domains [12,13] integrated in a module catalyze the conversion of a PCP bound, Ppant thioesterified l ⁄ d-amino acid (in ini- tiation modules) or l ⁄ d-peptidyl (in elongation mod- ules) moiety by de- and reprotonating the C a atom of the substrate. The epimerization reaction is reversible and finds its end in the adjustment of an equilibrium between both isomers [14]. However, only the d-ami- noacyl- or peptidyl-S-Ppant is singled out by the stereo- selective donor site of the downstream C domain for condensation with the next building block [15,16]. Bac- terial NRPS systems most often consist of several dis- tributed enzymes that, with only few exceptions, carry E domains at their C-terminal end. Such E domains were shown to be involved in the specific and ordered recognition of synthetases within NRPS assembly lines [17], a finding that makes them interesting candidates for the engineering of new hybrid enzymes. Short com- munication-mediating (COM) domains as part of E domains were recently identified to be responsible for this selective interaction [18]. Although the function of E domains was studied extensively by a mutational approach, only a little insight was afforded into the exact mechanistic process of the catalysis [14]. Other studies addressed the portability of E domains by con- structing fusion proteins of the type A ⁄ PCP-E. The main limitation of this immense time consuming genetic approach is the low throughput. Though, it led to the information that the E domain originating from the Phe-activating intiation module TycA is also able to convert Trp, Ile and Val with slightly decreased effi- ciency [13]. Aminoacyl-pantetheine derivatives were reported to be accepted as soluble substrates by the E domain of GrsA-ATE (PheATE) [19]. In doing so, very high K m values were determined indicating that it is not possible to gain information about the native substrate tolerance of E domains this way. Supplementary knowledge about E domains is limited to timing of condensation and epimerization [20,21] and to the ability of the TycB 3 -E domain, originally embedded in a Phe-activating elongation module of the tyrocidine synthetase, to epimerize the aminoacyl-Phe-S-Ppant substrate instead of the cognate enzyme bound tetra- peptide with lower efficiency [20]. On the other hand, nothing is known about the ability of an aminoacyl- S-Ppant epimerizing E domain to convert peptidyl- S-Ppant substrates. This could be of special interest for the engineering of NRPSs by module extension. As in all previous approaches only the amino acid being epimerized was varied and mainly aminoacyl sub- strates were investigated, it remains unclear if there is a strong specialization of E domains for recognizing distinct sites of their cognate substrate and therefore if one can strictly distinguish between ‘aminoacyl-’ and ‘peptidyl-’E domains. In this study, we accomplished an investigation of the substrate specificity of E domains by using various peptidyl-CoA precursors. The peptidyl-S-Pant moiety of these CoAs was transferred onto four different PCP-E bidomain constructs (TycB 3 -, FenD 2 -, TycA- and GrsA-PCP-E) (known to be the smallest working system for E domains [14]) under exploitation of the promiscuity shown by the 4¢-Ppant transferase Sfp (Fig. 1). Subsequently, velocity of the catalyzed epi- merization and the final l to d equilibrium were ana- lyzed by chemical cleavage of the reaction products from the enzyme after quenching the reaction. By this chemoenzymatic approach, which forms a new mini- mal system for the investigation of native E domain specificity, we could reveal a broad substrate tolerance of the C-terminal E domains used here. Most interest- ingly, we observed the ability of E domains, naturally located in initiation modules, to epimerize peptidyl- S-Ppant substrates. Likewise, we detected acceptance and elongation of peptidyl substrates, first loaded onto TycA-PCP-E and allowed to epimerize, by the TycB 1 - C domain which naturally follows an initiation module and there exclusively connects two aminoacyl residues. Results Generation and purification of the recombinant enzymes A set of four recombinant PCP-E bidomain proteins derived from different NRPS systems was constructed (Fig. 2). The enzymes chosen for this study harbour E domains which are all located at the C termini of the D. B. Stein et al. Epimerization in nonribosomal peptide synthesis FEBS Journal 272 (2005) 4506–4520 ª 2005 FEBS 4507 NRPSs. Two constructs thereof [TycA-PCP-E (65.6 kDa) and GrsA-PCP-E (67.6 kDa)] contain E domains from initiation modules (both activating and incorporating Phe) which only epimerize aminoacyl- S-Ppant substrates in their natural context. The two other constructs [TycB 3 -PCP-E (63.9 kDa) and FenD 2 - PCP-E (66.8 kDa)] in contrast contain E domains from elongation modules (TycB 3 is also activating and incorporating Phe, FenD 2 is activating and incorpor- ating Thr) both naturally epimerizing tetrapeptidyl- S-Ppant substrates. The individual enzymes were successfully expressed as C-terminal His 6 -tagged fusions in the heterologous host Escherichia coli M15 ⁄ pREP4 and could be purified to homogeneity by single-step Ni 2+ affinity chromatography as confirmed by SDS ⁄ PAGE (data not shown). Use of synthetic peptidyl-CoAs with Sfp and PCP-E bidomain constructs for assaying epimerization activity The first aim of this study was to develop an assay for the investigation of E domain substrate specificity with synthetic CoA precursors and HPLC based analysis of Fig. 1. Schematic representation of the assay system for the investigation of E domain substrate specificity. The PCP dom- ain of the PCP-E bidomain protein is primed with the peptidyl-Ppant moiety of the CoA substrate under catalytic action of Sfp. This initiates the catalytic conversion by the E domain leading to the formation of an equilibrium between the two enzyme bound L ⁄ D-peptidyl-S-Ppant intermediates (the Ppant moiety is illustrated by the wavy line). Fig. 2. Schematic representation of the biosynthetic operons for (1) tyrocidine from B. brevis ATCC 8185 (2) gramicidin from B. brevis ATCC 9999 and (3) fengycin from B. subtilis F29-3. The gene fragments cloned for construction of the recombinant PCP-E-bidomain proteins used in this study are presented in consideration of their relative location underneath each system. Epimerization in nonribosomal peptide synthesis D. B. Stein et al. 4508 FEBS Journal 272 (2005) 4506–4520 ª 2005 FEBS reaction products. The preparation and use of pep- tidyl- instead of aminoacyl-CoAs was preferred because E domains had not been investigated with elongated S-Ppant-substrates before. In addition, proceeding this way should enable one to identify amino acid residues of the substrate, other than the one actually epimer- ized, which might reveal an influence on the epimeriza- tion reaction. Initial experiments showed that the peptidyl-S-Ppant moiety of synthetically prepared peptidyl-CoAs (for all precursors synthesized in this study see Fig. 3) can be loaded onto the recombinant PCP-E constructs util- izing 15 lm of the 4¢-Ppant transferase Sfp for a fast modification (as described in Experimental Proce- dures). The reaction was quenched at defined time points after the addition of Sfp by adding 10% (v ⁄ v) tricholoroacetic acid (TCA) to the samples. This led to precipitation of the proteins including enzyme bound peptidyl-S-Ppant intermediates. After separation of excess substrate by washing, reaction products were set free under alkaline conditions and could be applied to HPLC-MS. First measurements revealed that the load- ing reaction was almost completed after 15 s (earliest point of time assessed) as estimated by comparing the intensity of MS signals belonging to peptides regained from samples which were taken at different time points. When Sfp was omitted in the reaction mixture, no detectable amount of the corresponding peptide was regained by the sample work up (data not shown). Both diastereomers resulting from the epimerization reactions were separated by reversed phase HPLC (Fig. 4). The relative values of l- ⁄ d-isomers were cal- culated easily by determining the area of the ion extracted MS signals. Therefore, the use of an internal standard was not necessary for quantification. Pro- ceeding like this, it was possible to resolve formation of converted peptidyl substrates in dependence of time (Fig. 5) (the margin of error, as estimated by up to four independent determinations of each time curve, was approximately ± 5%). Apparent k obs values could be calculated from these data in analogy to the radio- TLC assays earlier reported [13,14]. The decisive advantage of this chemoenzymatic approach is the bypassing of natural substrate activation and conden- sation which allows loading of any peptidyl precursor. With these fundamental notices it was possible to establish this new minimal system for investigation of E domain specificity. Assuming that ‘peptidyl-’E domains mainly recog- nize the C-terminal amino acid residue of their pep- tidyl-S-Ppant substrate (the directly thioester bound and actually epimerized moiety), in a set of peptidyl- CoAs ) besides the cognate precursor for TycB 3 -PCP- E, fPFF-CoA (1)[d-Phe-l-Pro-l-Phe-l-Phe-S-Ppant; we will use the shorter one letter abbreviation for amino acids (capital letter for l-amino acids, lower- case letter for d-amino acids) throughout this text], and the corresponding shortened one, FF-S-Ppant (2) ) this decisive site (R 2 ) was varied while keeping the N-terminal Phe (R 1 ) constant. To test if the final equilibrium position reached in the reaction with FF- CoA (2, conversion from l to d) is the same when converting from d to l, Ff-CoA (3) was designed. Fur- ther synthesized CoAs include variations of aromatic [FY-CoA, F(p-fluoro) F-CoA, FW-CoA (4–6)], alipha- tic [FL-CoA (7)], neutral hydrophilic [FN-CoA, FS- CoA, FT-CoA (8–10)], acidic [FD-CoA (11)], and Fig. 3. Presentation of all peptidyl-CoAs synthesized in this study. Compound 1 is the mimic of the natural substrate for TycB 3 -E and 2 the corresponding shortened dipeptidyl-CoA. In precursor 3, the C-terminal Phe (R 2 )isD-configured. While R 1 was kept constant (Phe) in compounds 4–13, R 2 was varied. In contrast, CoA 14 con- tains a constant Phe for R 2 and a Ser for R 1 . Compound 15–17 were designed for investigation of E domain tolerance towards N-methylated substrates. D. B. Stein et al. Epimerization in nonribosomal peptide synthesis FEBS Journal 272 (2005) 4506–4520 ª 2005 FEBS 4509 basic [FH-CoA (12), FK-CoA (13)] amino acid resi- dues for R 2 . Then, for evaluating the influence of the N-terminal substrate part in SF-CoA (14) the C-ter- minal Phe (R 2 ) was kept constant while residue R 1 was changed to Ser. The peptidyl-CoAs synthesized in this study (Fig. 3) were basically designed to be suitable for investigating the TycB 3 -E domain specificity (construct TycB 3 -PCP- E). In its natural environment within the tyrocidine biosynthetic template this E domain is responsible for epimerizing the tetrapeptidyl substrate fPFF-S-Ppant. For keeping synthesis efforts as simple as possible we wanted to restrict ourselves to the utilization of dipept- idyl-CoA precursors. Consequently, we first tested if the cognate fPFF-S-Ppant and the shortened one FF- S-Ppant are converted with comparable efficiencies from l to d. The corresponding CoA substrates 1 and 2 were used to load fPFF-S-Ppant and FF-S-Ppant onto TycB 3 -PCP-E with the help of Sfp, after which the epimerization reaction was followed up. It could be observed that both velocity of the l to d conversion and the portion of d-isomer regained from the enzyme after equilibration are comparable (data below and Table 1). This indicated that at least the two N-ter- minal amino acid residues of the original tetrapeptidyl substrate are not significantly involved in substrate recognition by the TycB 3 -E domain, allowing us to pursue our work with dipeptidyl-CoAs as model precursors. The E domain of FenD 2 is originally embedded in a Thr activating module and thus should represent an epimerase with an opponent substrate specificity in comparison with TycB 3 -E [expected to be Phe (R 2 ) optimized]. FenD 2 -E in its natural environment within the fengycin synthetic template [22] (Fig. 2) converts the tetrapeptidyl substrate EoYT-S-Ppant (l-Glu-d- Orn-l-Tyr-l-Thr-S-Ppant). Thus, the shortened dipep- tidyl-mimic (two C-terminal residues) of the natural substrate for FenD 2 -E is YT-S-Ppant. Because Tyr and Phe are analogues belonging to the group of aro- matic amino acids, the synthesized dipeptidyl-CoAs AB Fig. 4. HPLC-MS analysis of dipeptides regained from the PCP-E-enzyme exemplified by the reaction of TycB 3 -PCP-E with Sfp and FF-CoA (1). After priming of the protein with the precursor utilizing Sfp the reaction was quenched in time dependent manner by the addition of 10% TCA. Reaction products were released from the enzyme by thioester cleavage with 0.1 M KOH and applied to HPLC. (A) Chromato- grams of the extracted [M + H] + mass signal, which is shown in (B), of separated L-Phe-L-Phe and L-Phe-D-Phe. The traces show analyses of hydrolysed FF-CoA and samples taken at time points listed in the figure. Fig. 5. 2D-Plot for illustration of epimerization activity exemplified by the reaction of TycB 3 -PCP-E with different substrates. After loading of the corresponding peptidyl-S-Ppant moiety of the CoA- precursors (shown in the legend underneath the plot) by the help of Sfp, reactions were quenched at different points of time with 10% TCA. Products were regained by alkaline cleavage of the thio- ester and analysed by HPLC. The amount of each isomer was quantified and the formation of the D(R 2 )-peptides (in percentage) presented in dependence of time. Epimerization in nonribosomal peptide synthesis D. B. Stein et al. 4510 FEBS Journal 272 (2005) 4506–4520 ª 2005 FEBS with N-terminal Phe should also be suitable for inves- tigating this system. Nothing was known so far about the ability of E domains deriving from initiation modules to convert peptidyl-S-Ppants. In general, loading the peptidyl- S-Ppant moiety of the synthesized CoAs onto PCP-E constructs is an elegant method of determining toler- ance of the TycA- and GrsA-E domains (cognate sub- strate Phe-S-Ppant) for elongated substrates. In addition, the use of CoA precursors enabled us to investigate the tolerance of E domains for N-methyla- ted (alkylated) peptidyl substrates. So far, this had been tested with fusion proteins derived from the actinomycin biosynthetic system [23] which did not lead to the detection of converted product. In this study we applied the tripeptidyl precursor FFMeF-CoA (17) because utilizing the dipeptidyl substrates FP-CoA (15) and FMeF-CoA (16) resulted in exceedingly fast forma- tion of the corresponding DKPs (diketopiperazines) for noted reasons [24] as soon as loaded onto the protein with Sfp or when hydrolysed in control reactions. Epimerization activity of TycB 3 - and FenD 2 -PCP-E The cognate substrate of E domains localized in elon- gation modules (‘peptidyl-’E domains) is a nonribo- somally assembled peptidyl-S-Ppant intermediate. At this, catalytic conversion from l to d only concerns the C-terminal amino acid residue of the enzyme bound substrate. It had already been shown that the TycB 3 -E domain (probed with TycB 3 -ATE) converts an l-Phe-S-Ppant substrate with reduced efficiency resulting in about 40% d-Phe [20]. The epimerization velocity of TycB 3 -E with its cognate substrate (fPFF- S-Ppant) had not been published so far but the final l ⁄ d ratio of products was reported to be around 1 : 1 [16]. We tested the epimerization activity of both the TycB 3 - and FenD 2 -E domain with various peptidyl- CoA precursors to acquire a conception of ‘peptidyl-’E domain substrate tolerance and to investigate the influ- ence of additional N-terminal amino acids on substrate recognition. The results are summarized in Table 1. First investigations with the TycB 3 -PCP-E bidomain (as described above) revealed that the cognate tetra- peptidyl-S-Ppant substrate and its shortened dipeptidyl mimic are epimerized equally fast (both 1.9 min )1 ) and final d-portions of the regained peptides were also found to be in corresponding ranges (fPFf, 56%; Ff, 63%). With the same velocity a comparable equilib- rium (61% regained Ff) is reached when TycB 3 -E converts substrate Ff-S-Ppant (loading of 3) from d to l. By utilization of precursors FY-CoA (4) and Table 1. Results of epimerization activity assay. nd, not determined. Precursor a Bidomain system TycB 3 -PCPE FenD 2 -PCPE TycA-PCPE GrsA-PCPE k obs (min )1 ) D ⁄ (D + L) (%) b k obs (min )1 ) D ⁄ (D + L) (%) k obs (min )1 ) D ⁄ (D + L) (%) k obs (min )1 ) D ⁄ (D + L) (%) 1 fPFF-CoA 1.9 56 1.8 56 1.0 59 1.5 62 2 FF-CoA 1.9 63 1.6 58 1.8 54 1.7 55 3 Ff-CoA 1.9 61 1.6 58 1.8 71 1.7 61 4 FY-CoA 2.0 62 1.5 62 2.0 60 > 2.0 60 5 F(p-fluoro) F-CoA 1.9 63 1.8 56 1.9 58 1.7 54 6 FW-CoA 1.2 59 0.8 55 0.8 35 1.3 46 7 FL-CoA 1.4 63 1.3 58 1.4 60 1.3 60 8 FN-CoA 1.2 60 1.2 59 1.9 68 1.9 68 9 FS-CoA 1.5 60 1.8 61 > 2.0 63 > 2.0 68 10 FT-CoA nd nd 1.3 41 nd nd nd nd 11 FD-CoA + c nd + nd + nd + nd 12 FH-CoA + nd + nd + nd + nd 13 FK-CoA d 0.8 30 0.9 26 1.0 52 0.9 54 14 SF-CoA 1.5 59 0.9 55 > 2.0 52 1.7 49 15 FP-CoA nd nd nd nd nd nd nd nd 16 FMeF-CoA nd nd nd nd nd nd nd nd 17 FFMeF-CoA > 2.0 74 > 2.0 70 1.6 71 > 2.0 68 a Shown in Fig. 4. b Measured after 15 min incubation. c +, epimerization activity could be detected but not be quantified due to analytical reasons. d As a consequence from the basic sample work up, the products detected here were not the linear peptides FK and Fk but the cyclic peptides resulting after nucleophilic attack of the Lys-e-amino group to the carbonyl group of the peptide-S-Ppant thioester. Yet this has no effect on the observed Ca stereochemistry. D. B. Stein et al. Epimerization in nonribosomal peptide synthesis FEBS Journal 272 (2005) 4506–4520 ª 2005 FEBS 4511 F(p-fluoro)F-CoA (5) it was demonstrated that TycB 3 - E is also tolerant towards unusual substituted Phe ana- logues. On the other hand, decreased activity (1.2 min )1 ) was observed when TycB 3 -PCP-E was primed with FW-CoA (6) and Sfp indicating that there is no general preference for aromatic amino acid resi- dues at the C-terminal position (R 2 ) of the substrate. Instead, the sterical pretension of Trp might be respon- sible for slowing down the conversion to the final equi- librium (59% Fw). Likewise, other variations for R 2 affected the epimerization velocity [1.2–1.5 min )1 when using precursors FL-CoA (7), FN-CoA (8), and FS-CoA (9)] but not the equilibrium position ( 60% d-isomer in all cases). Similar results obtained by using SF-CoA (14) permit speculations that the second amino acid residue of the peptidyl-S-Pant substrate (R 1 ) is also involved in the recognition by this E domain to some extent. Peptides regained from the enzyme after assaying the epimerization activity with the S-Ppant precursors FD-CoA (11) and FH-CoA (12) could only be analyzed qualitatively by the des- cribed HPLC-MS methods. It was possible to detect converted products but poor separation and low signal quality did not allow for a quantification of the results. The strongest decrease of efficiency was observed with FK-S-Ppant (loading of 13, 0.8 min )1 ) even having an effect on the portion of d-isomer regained from the enzyme after equilibration (only 30% Fk). This clearly indicates that a charged residue for R 2 somehow blocks up substrate recognition or correct binding in the active site of TycB 3 -E and substrate tolerance is limited. Because FenD 2 -E derived from a Thr activating module, it was expected to have a preference for substrates with hydrophilic C-terminal amino acid residues. Nevertheless, the catalytic conversion of fPFF-S-Ppant, after loading of substrate 1 onto FenD 2 -PCP-E, was carried out with an efficiency (1.8 min )1 , 56% fPFf) comparable to that of TycB 3 -E (see above). Surprisingly, epimerization activity tested with the precursors containing C-terminal variations of aromatic amino acids (2–5) was only slightly decreased [1.5–1.8 min )1 , about 60% regained d(R 2 )-peptides] in comparison to TycB 3 -E. However, when FW-CoA (6) was loaded onto FenD 2 -PCP-E (Trp for R 2 ) the effi- ciency of catalytic conversion leading to the final equi- librium (55% Fw) again was lowered to 0.8 min )1 [see assay with TycB 3 -PCP-E (1.2 min )1 )]. Unexpectedly, FenD 2 -E showed no preference towards substrates with C-terminal variations of hydrophilic and aliphatic amino acids [use of FN-CoA (8) and FL-CoA (7), 1.2–1.3 min )1 , 55% regained d-form], only FS-S-Ppant [FS-CoA (9)] was converted more effectively (1.8 min )1 , 61% Fs) when compared to TycB 3 -E. Nev- ertheless, FenD 2 -E showed even less efficient epimeri- zation (1.3 min )1 ) with the mimic of its cognate substrate FT-S-Ppant [FT-CoA (10)] resulting in only 41% Ft. Apparently, substrate tolerance of ‘peptidyl-’E domains does not totally conform to the specificity of the module from which the E domain derives. Activity after loading of FD-CoA (11) and FH-CoA (12) could be observed but not be quantified (see above). As seen with TycB 3 -PCP-E, catalysis of epimerization after loading FK-CoA (13) onto FenD 2 -PCP-E was strongly impaired (0.9 min )1 , 26% Fk). Also, a much slower reaction was observed when using SF-CoA (14) (0.9 min )1 ) resulting in 55% Sf-S-Ppant. Here, catalysis seems to be affected due to the fact that both amino acid residues of the substrate are not cognate, which underlines the presumption that the two C-terminal amino acids are involved in recognition by ‘peptidyl-’E domains. Epimerization activity of TycA- and GrsA-PCP-E The cognate substrate of E domains deriving from ini- tiation modules is a PCP bound aminoacyl-S-Ppant precursor. It has been reported that TycA-E and especially GrsA-E, investigated with rapid quench methods, are very efficient enzymes converting l-Phe- S-Ppant enzyme exceedingly fast to a final 2 : 1 (d- ⁄ l-S-Ppant) equilibrium [13,14,19]. Still, up to now it has not been reported if the function of these ‘ami- noacyl-’E domains is decided by their position within the NRPS system or if their substrate specificity really differs from ‘peptidyl-’E domains. In this study, we addressed this important question by loading peptidyl- S-Pant moieties of CoAs onto TycA- and GrsA-PCP- E and were able to show epimerization activity of ‘aminoacyl-’E domains with elongated enzyme bound S-Ppant substrates. This can be of great impact for future engineering approaches. The results are summarized in Table 1. In general, TycA- and GrsA-E behave very simi- larly. Both were able to convert FF-S-Ppant [loading of FF-CoA (2)] almost as efficiently as TycB 3 -E (TycA-E, 1.8 min )1 ; GrsA-E, 1.7 min )1 ) equilibrating to a slightly reduced amount of 55% Ff-S-Ppant when compared to the d- ⁄ l-ratio with their cognate substrate Phe (see above). Surprisingly, a shifted equilibrium position was reached when the ‘amino- acyl-’E domains epimerize from d to l after loading of Ff-CoA (3) onto the PCP-E bidomains (TycA, 71% Ff; GrsA, 61% Ff). The catalytic efficiency was further impaired by longer peptidyl substrates as seen when using fPFF-CoA (1) (TycA, 1.0 min )1 ; Epimerization in nonribosomal peptide synthesis D. B. Stein et al. 4512 FEBS Journal 272 (2005) 4506–4520 ª 2005 FEBS GrsA, 1.5 min )1 ) but this finding did not concern the final product ratio (about 60% d-form). Although the tested ‘aminoacyl-’E domains were act- ive with peptidyl substrates, their tolerance towards distinct S-Ppant precursors apparently deviates from that of ‘peptidyl-’E domains. Both constructs very efficiently converted FY-S-Ppant (about 2.0 min )1 )to yield 60% Fy-S-Ppant and accepted F(p-fluoro)F- S-Ppant (substrate 5) as well. The use of FW-CoA (6) not only showed a strong effect on the epimeri- zation velocity (TycA, 0.8 min )1 ; GrsA, 1.3 min )1 ) but also on produced amount of the corresponding d-isomer (TycA, 35% Fw; GrsA 46% Fw). The ‘aminoacyl-’E domains seem to be impaired even more by the bulky C-terminal Trp residue of the substrate than the tested ‘peptidyl-’E domains. When converting enzyme bound FL- S -Ppant, the same velocity (1.3–1.4 min )1 ) and final equilibrium of 60% Fl-S-Ppant were observed, compared to TycB 3 - and FenD 2 -E. In contrast to TycB 3 -E, TycA- and GrsA- E are more tolerant towards hydrophilic amino acids for R 2 , displayed in strong activity with FN- (both TycA- and GrsA-E, 1.9 min )1 , 68% Fn) and FS- S-Ppant intermediates (both > 2.0 min )1 ; TycA, 63% Fs; GrsA, 68% Fs). Activity was also detected when FD-CoA (11) and FH-CoA (12) were applied in the assay, but results again could not be quantified due to reasons described before. As observed after load- ing FK-CoA (13), the velocity of the catalytic con- version was decreased (TycA, 1.0 min )1 ; GrsA, 0.9 min )1 ) as seen before with the ‘peptidyl-’E domains. Nevertheless, conversion by the ‘aminoacyl-’E domains resulted in an increased portion of d-isomer after equilibration (TycA, 52% Fk; GrsA, 54% Fk). Obviously, TycA- and GrsA-E, in contrast to TycB 3 - and FenD 2 -E, are preferably able to coordi- nate and epimerize substrates with charged C-ter- minal amino acid residues. The epimerization activity with SF-S-Ppant-enzyme [loading of SF-CoA (14)] varied from results with FF-CoA (TycA, > 2.0 min )1 ; GrsA, 1.7 min )1 , about 50% regained Sf) in approximately the same order of magnitude as when using the precursor FS-CoA (9). This indicates that variations in both parts of the substrate can equally affect the epimerization activity of ‘aminoacyl-’E domains. Epimerization activity with N-methylated peptidyl substrates Out of previous studies with fusion proteins [23] the question arose if E domains are compatible with pre- ceding N-methylation. Theoretically, a methyl (alkyl) group attached to the nitrogen atom of the first pep- tide bond within the substrate should not cause hindrance of C a -proton abstraction needed for the catalytic conversion. Although a prolific interaction of E domains with M domains on the enzymatic level remains an open question we could show that epime- rization is occurring after loading the Ppant moiety of precursor FFMeF-CoA (17, i.e. l-Phe-l-Phe-N- Me-l-Phe-CoA) onto all PCP-E-bidomain constructs utilized in this study (Table 1 and Fig. 6). The N-methylated S-Ppant intermediate was converted very effectively by all enzymes (> 2.0 min )1 ); only TycA-E showed a slightly lowered activity (1.6 min )1 ). Additionally, a high portion of d-isomer (FFMef) was regained from the enzymes after adjust- ment of the final equilibrium (about 70% FFMef in all cases). Obviously, E domains are compatible with this class of precursors and furthermore N-methyla- tion of peptidyl-S -Ppant substrates seems even to support the catalytic conversion. Elongation of peptidyl substrates by the C domain of TycB 1 Our results showed that E domains deriving from initi- ation modules are able to convert peptidyl-S-Ppant substrates instead of the cognate aminoacyl intermedi- ate. To evaluate this finding in the context of biocom- binatorial approaches, an interesting question was if a B A Fig. 6. Separation of the N-methylated trip- eptides FFMeF ⁄ FFMef by chiral HPLC. The peptides were regained from TycB 3 -PCP-E after quenching the reaction at the points of time following each trace. Illustrated in (A) are the corresponding signals for the mass ([M + H] + ) shown in (B) obtained by SIM-MS analysis. D. B. Stein et al. Epimerization in nonribosomal peptide synthesis FEBS Journal 272 (2005) 4506–4520 ª 2005 FEBS 4513 downstream C domain interacting in the native system with an ‘aminoacyl-’E domain is able to accept and elongate these enzyme bound peptidyl precursors. The upstream electrophilic donor site of the TycB 1 -C domain showed in previous studies a relaxed specificity concerning the transferred amino acid substrate, but d-configuration is measurably preferred [15,25]. In this study, we also could reveal tolerance of the TycB 1 -C domain for peptidyl substrates by using the well estab- lished system TycA ⁄ TycB 1 [10]. TycA-PCP-E therefore was primed with the peptidyl-S-Ppant moiety of selec- ted CoAs with Sfp and preincubated (see Experimental procedures) for epimerization of the bound intermedi- ates. Separately, the acceptor enzyme TycB 1 -CAT ⁄ TE srf , chosen because it seemed to be a promising candidate for fast product release [24], was preincubat- ed to activate and covalently load l-Pro. Product for- mation was initiated by mixing equal volumes of the protein solutions and stopped by enzyme precipitation with methanol. Products contained in the supernatant were identified by HPLC-MS analysis. All detected products (as expected after condensa- tion with Pro) are shown in Table 2. As expected, only one product was observed when TycA-PCP-E was loa- ded with either FF- or Ff-CoA (2 and 3) and peptides transferred to TycB 1 -CAT ⁄ TE srf for elongation. Cata- lytic release presumably yielded FfP with a detected mass of 410.3 gÆmol )1 ([M + H] + ). No products were found when ATP was omitted from the assay (Fig. 7). Also, formation of only one distinct product was detected when TycB 1 -CAT ⁄ TE srf processed the pep- tides transferred by TycA-PCP-E after loading of FL- and fPFF-CoA (7 and 1). Thus, TycB 1 -C is able to accept and elongate all of these peptidyl-S-Ppants up to a length of four amino acids (substrate 1 was the longest used in our study). Although we have not com- pared the formed tripeptides and pentapeptides to chemical standard compounds, the detection of only one product indicates that the stereo-selectivity of the donor site is retained when TycB 1 -C accepts these peptidyl substrates. Yet, the C-terminal part of the peptidyl-S-Ppant substrate accepted by TycB 1 -C seems to be involved in recognition, as seen before with the tested E domains. After using FT-CoA (10) and FK-CoA (13) in the elongation assay, no detectable products with the expected mass (FtP, [M + H] + ¼ 364 gÆmol )1 ; FkP, [M + H] + ¼ 391 gÆmol )1 ) were released by TycB 1 -CAT ⁄ Te srf , clearly indicating that Table 2. Identification of elongated peptidyl products investigated in the system TycA ⁄ TycB1. nd, no product detected. Precursor for TycA-PCPE a Expected product Ionization method Species Observed mass (calculated mass) [g mol )1 ] FF-CoA (2) FfP ESI [M + H] + 410.3 (410.2) Ff-CoA (3) FfP ESI [M + H] + 410.3 (410.2) FL-CoA (7) FlP ESI [M + H] + 376.3 (376.2) FS-CoA (9) FsP ESI [M + H] + 350.1 b (350.2) FT-CoA (10) FtP ESI [M + H] + nd (364.2) FK-CoA (13) FkP ESI [M + H] + nd (391.2) SF-CoA (14) SfP ESI [M + H] + 350.1 d (350.2) fPFF-CoA (1) fPFfP ESI [M + H] + 654.3 (654.3) a Shown in Fig. 4. b A product mixture of 1 : 1 (presumably FsP ⁄ FSP) was detected. c A product mixture of 5.4 : 1 (presumably SfP ⁄ SFP) was detected. AB Fig. 7. HPLC analysis of peptidyl substrates elongated by the C domain of TycB 1 . As an example, the formation of the tripeptide FfP is illus- trated in (A). TycA-PCP-E was primed with FF-S-Ppant by help of Sfp and FF-CoA (2), allowed to equilibrate, and incubated with TycB 1 -CAT ⁄ TE srf [24], which had been preincubated to load Pro. FF was presumably produced by in trans action of the thioesterase of TycB 1 -CAT ⁄ TE srf with FF-CoA. Nevertheless, TycB 1 -C accepts and elongates Ff to form FfP (black trace) that was released by the acceptor enzyme and could be identified by detection of the corresponding [M + H] + mass signal shown in (B). No product was detected when ATP was omitted in the assay (grey trace). Epimerization in nonribosomal peptide synthesis D. B. Stein et al. 4514 FEBS Journal 272 (2005) 4506–4520 ª 2005 FEBS the TycB 1 -C donor site discriminates noncognate C-terminal amino acid residues of the S-Ppant sub- strate. The assay with FS-CoA (9) resulted in a 1 : 1 mixture of products with the corresponding mass (m ⁄ z) of 350 ([M + H] + ) as expected for FsP ⁄ FSP. Likewise, a 5.4 : 1 mixture of products was obtained after transfer of SF [loading of SF-CoA (14) onto TycA-PCP-E] to TycB 1 -CAT ⁄ TE srf . Although, due to missing chemical standards, the actual stereochemistry of these products remains obscure, considering the results of the epimerization assay with TycA-PCP-E, we expect the portion of 5.4 (84%) to be SfP. This indicates that the second amino acid residue (R 1 )of the peptidyl-S-Ppant substrate coming from TycA- PCP-E influences stereo-selection by the TycB 1 -C domain donor site only slightly. Discussion In this study, we gained new information on substrate recognition and specificity of E domains focusing onto aspects of potential decisive sites (amino acid residues) of a peptidyl-substrate recognized by an E domain with the aim to utilize E-domains for the redesign of non- ribosomal peptides. The main issue to be addressed was if the substrate specificity of E domains is concordant with the specificity of the module (activated amino acid) it originates from. Besides this, an interesting question was if E domains of initiation modules are generally able to epimerize peptidyl- instead of their cognate aminoacyl-substrates (‘aminoacyl-’ vs. ‘peptidyl-’E domains). For this purpose, the use of CoA precursors and different apo-PCP-E bidomain constructs turned out to be very useful for following up epimerization reactions and evaluating the activities of E domains. The approach had already been used in similar manner for the characterization of C [15,16] and TE domains [26]. We accomplished covalent loading of the peptidyl- S-Ppant moiety of the peptidyl-CoAs onto the invariant Ser of each PCP domain by exploiting the promiscuous Ppant transferase Sfp [8,9]. The advantage of this new minimal system for the investigation of E domain spe- cificity is not only the circumvention of natural sub- strate activation but also the HPLC based analysis of reaction products in contrast to the previously used radio-TLC assay [13,14,20]. Also, the E domains stay in native connection to their respective PCP E , which was shown to be essential for E domain activity [13], and consequently should behave as if they are in their native enzymatic environment. We investigated substrate recognition and tolerance of two ‘peptidyl-’E domains (proteins TycB 3 - and FenD 2 -PCP-E) and two ‘aminoacyl-’E domains (proteins TycA- and GrsA-PCP-E), respectively. As expected, TycB 3 -E shows a preference for peptidyl-S- Ppant substrates with C-terminal Phe and Phe ana- logues. Nevertheless, the use of precursor 6 with a Trp in position R 2 revealed a discrimination of this aroma- tic residue. This finding was surprising because the module TycB 3 was reported to be activating Trp as well with great efficiency [27]. Therefore, one would have concluded that TycB 3 -E should be similarly able to epimerize a peptidyl-S-Ppant substrate with a C-ter- minal Trp. However, TycB 3 -E is tolerant towards a broad variety of peptidyl-S-Ppant substrates with altered C-terminal amino acid residues. Assuming that the specificity of an E domain (recog- nition of the C-terminal site of the substrate) is identi- cal with the specificity of the module it originates from, we decided to investigate FenD 2 -E as first exam- ple of an E domain contained in a module which is not naturally Phe but Thr specific. As described earlier, the synthetic CoA precursors were also suitable for this system. Interestingly, FenD 2 -E showed no general enhanced activity with substrates containing C-ter- minal hydrophilic amino acid residues. Instead, even a slightly reduced rate (1.3 min )1 ) was detected when we used the mimic-precursor of the cognate substrate (FT-CoA, 10). Moreover, this reaction resulted in only 41% Ft. Seemingly, there is no direct correlation between the substrate specificity of E domains and the specificity of the module they originate from. In sum- mary, the substrate tolerances of the two different ‘peptidyl-’E domains investigated in this study conform more with each other and the substrate specificities are less evolved than expected. In general, both E domains tolerate a broad variety of altered substrates. Within the scope of the investigations reported here, we also discovered acitvity of the two ‘aminoacyl-’E domains (TycA- and GrsA-E) with peptidyl-S-Ppant substrates. In contrast, applying another strategy with fusion proteins and natural substrate activation as reported before for the investigation of tolerance towards altered aminoacyl substrates [13] would require the construction of large and complex protein systems. This is again emphasizing the convenience of our minimized assay approach utilizing PCP-E bi- domains with synthetic peptidyl-CoAs and Sfp. Both PheATE initiation modules, TycA- and especially GrsA-ATE [19] have been very well characterized in the past, so far most often investigated in the so called DKP assay [10], and known to resemble each other in many characteristics. Thus, we expected and observed great resemblance between their two ‘aminoacyl-’E domains. Two striking differences were found in con- trast to the tested ‘peptidyl-’E domains. First, when D. B. Stein et al. Epimerization in nonribosomal peptide synthesis FEBS Journal 272 (2005) 4506–4520 ª 2005 FEBS 4515 [...]... the C domain as indicated by a 5.4 : 1 mixture (l-Ser-d ⁄ l-Phe-l-Pro supposed) of tripeptidyl product detected after elongation of Sf- ⁄ SF-S-Ppant (loading of substrate 14) by the TycB1-C domain Although the focus of this study is not the engineering of hybrid–NRPS enzymes the results obtained on the substrate specificity of E domains could pave the way to engineering of biosynthesis of valuable nonribosomal. .. Aminoacyl-CoAs as probes of condensation domain selectivity in nonribosomal peptide synthesis Science 284, 486– 489 4519 Epimerization in nonribosomal peptide synthesis 16 Clugston SL, Sieber SA, Marahiel MA & Walsh CT (2003) Chirality of peptide bond-forming condensation domains in nonribosomal peptide synthetases: The C5 domain of tyrocidine synthetase in a DCL catalyst Biochemistry 42, 12095–12104 17... & Walsh CT (2001) Generality of peptide cyclization catalyzed by isolated thioesterase domains of nonribosomal peptide synthetases Biochemistry 40, 7099– 7108 Dokel S & Marahiel MA (2000) Dipeptide formation ¨ on enigneered hybrid peptide synthetases Chem Biol 7, 373–384 Linne U & Marahiel MA (2004) Reactions catalyzed by mature and recombinant nonribosomal peptide synthetases Methods Enzymol 388,... Schauwecker F, Pfennig F, Grammel N & Keller U (2000) Construction and in vitro analysis of a new bi-modular polypeptide synthetase for synthesis of N-methylated acyl peptides Chem Biol 7, 287–297 24 Schwarzer D, Mootz HD & Marahiel MA (2001) Exploring the impact of different thioesterase domains for the design of hybrid peptide synthetases Chem Biol 8, 997–1010 25 Ehmann DE, Trauger JW, Stachelhaus T & Walsh... domains These findings and the observed ability of the TycB1-C domain to elongate peptidylsubstrates reveal new possibilities for the portability of E domains and rules for use in biocombinatorial rearrangements of NRPSs FEBS Journal 272 (2005) 4506–4520 ª 2005 FEBS Epimerization in nonribosomal peptide synthesis Experimental procedures Cloning of PCP-E genes, expression and protein purification The pQE... Portability of epimerization domain and role of peptidyl carrier protein on epimerization activity in nonribosomal peptide synthetases Biochemistry 40, 15824–15834 14 Stachelhaus T & Walsh CT (2000) Mutational analysis of the epimerization domain in the initiation module PheATE of gramicidin S synthetase Biochemistry 39, 5775–5787 15 Belshaw PJ, Walsh CT & Stachelhaus T (1999) Aminoacyl-CoAs as probes of condensation... intermediate of the substrate during the catalytic epimerization by E domains and promote the reaction this way In addition to these new insights into E domain activity, we gained important information about the substrate tolerance of the TycB1-C domain’s electrophilic donor site concerning the possibility of elongating enzyme bound peptidyl substrates provided by TycA This question consequently arises from the. .. min, the precipitates were separated by centrifugation (15 700 g, 4 °C) for 45 min The pellets were washed with 750 lL ether ⁄ ethanol (3 : 1) and 750 lL ether, followed each time by centrifugation (15 700 g, 4 °C) for 5 min, and dried for 5 min at 37 °C The products were released from the proteins by alkaline cleavage of the thioester upon the addition of 100 lL 0.1 m KOH and incubation at 70 °C for. .. consequently arises from the ability of ‘aminoacyl-’E domains to convert peptidylS-Ppant substrates In previous studies it was demonstrated that there is a relaxed specificity of the C domain’s donor site (of C domains following E domains in the natural context) towards various aminoacyl substrates, yet there is a high stereo-selectivity and preference of the d isomer provided by the upstream module [15,16,25]... for detection of peptidyl-S-Ppant epimerization To follow up epimerization of the first, directly thioester bound, amino acid residue of the peptidyl-S-Ppant substrates, priming of the PCP-E enzymes was initiated as described above (final volume 600 lL) At defined time points after the addition of Sfp 100 lL aliquots were taken and immediately quenched in 1 mL 10% (w ⁄ v) TCA After incubation on ice for . to utilize E -domains for the redesign of non- ribosomal peptides. The main issue to be addressed was if the substrate specificity of E domains is concordant with the specificity of the module (activated. enzymes the results obtained on the substrate specificity of E domains could pave the way to engineering of biosynthesis of valuable non- ribosomal peptides in vivo. With the discovery of the rules for. Utility of epimerization domains for the redesign of nonribosomal peptide synthetases Daniel B. Stein, Uwe Linne and Mohamed A. Marahiel Fachbereich