enzyme technologies for pharm and biotech app

Becker Lilly Research Laboratories, Eli Lilly and Company, napolis, Indiana India-Gary M.. Birch Lilly Research Laboratories, Eli Lilly and Company, olis, Indiana Indianap-Ridong Chen De

This book is printed on acid-free paper.

Headquarters

Marcel Dekker, Inc

270 Madison Avenue, New York, NY 10016

Copyright2001 by Marcel Dekker, Inc All Rights Reserved.

Neither this book nor any part may be reproduced or transmitted in any form or by anymeans, electronic or mechanical, including photocopying, microfilming, and recording,

or by any information storage and retrieval system, without permission in writing fromthe publisher

Current printing (last digit):

10 9 8 7 6 5 4 3 2 1

PRINTED IN THE UNITED STATES OF AMERICA

Genes related to human and animal health are being discovered at an ing rate As gene products, enzymes are being explored for their function and

ever-increas-application in a rapidly emerging field that has been termed functional genomics Enzyme Technologies for Pharmaceutical and Biotechnological Applications fills

a unique niche for a comprehensive account of certain important enzymes inhuman and animal health Readers can also gain important insights into enzymetechnologies in both the pharmaceutical and biotechnological industries The pri-mary aim of this book is to highlight how, what, and where enzymes have becomecritically important or are rapidly emerging in these two overlapping and interde-pendent industries

As a state-of-the-art work on enzyme technologies, the book covers fourbasic principles and applications in (1) antibiotic biosynthesis, (2) biocatalysis,(3) modern screening/optimization, and (4) emerging new technologies In Part

I, on biosynthesis, the emphasis is placed on both improvements in antibioticyield and ways to increase antibiotic structural diversity by modifications of thebiosynthetic pathways from diverse microorganisms Here, the emphasis is onusing genes to deliver enzymes and to thereby perform metabolic engineeringincluding precursor-directed biosynthesis or mutasynthesis The use of recombi-nant techniques to generate protein products that are unnatural to the microbialworld is also discussed, using specific examples of challenging problems in thisarea

Part II on biocatalysis, covers the direct application of enzymes as chemicaltools in manipulating small- to medium-sized synthetic organic compounds Ma-nipulation of the enzyme tools by genetic engineering is described Chapter 8

iii

discusses how a novel form of enzymes, cross-linked enzyme crystals (CLCs), areespecially useful as chemical catalysts An example of a large-scale application ofthese chemical tools brings the area into focus by leaving the laboratory andentering the manufacturing plant.

Part III on screening for and optimization of enzyme inhibitors describesintegrated approaches in therapeutic discovery research using enzyme targets rel-evant to human and animal diseases For high-throughput screening, the activityassays for the enzyme targets adopt both conventional (colorimetry, spectrometry,and radioactivity) and contemporary methodologies (fluorescence) A selectiveenzymatic assay maximizes validated hits from large diversified libraries of sam-ples derived from natural products and synthetic compounds, including thosearising from combinatorial chemistry The chapters on screening concentrate ondevelopment of effective enzymatic assays, each of which represents specific,kinetic, and molecular interactions between the enzyme and its substrate as well

as inhibitors and thus reflects the pharmacological and chemical interplay at thetargeted enzyme The primary screening goal is the production of manageablenumbers of hits that ultimately generate high-quality lead compounds As a prac-tical rule, optimization of those lead compounds by medicinal chemists is thecritical follow-up step required for the discovery of viable drug candidates Theprocess of lead compound optimization for an enzyme inhibitor, often referred

to as structure-activity relationship studies or drug design, is dictated by standing the molecular and kinetic interactions between the enzyme and its inhibi-tor These insights are typically gained by analyzing X-ray crystallographic depic-tions and by elucidating the kinetic behavior of the enzyme-inhibitor complex

under-to improve potency and selectivity and under-to understand mechanisms of interactions.The chapters on inhibitor screening/optimization emphasize the synergistic im-portance of high-throughput screening and structure-function based optimizationstudies for therapeutic discovery programs

Finally, Part IV on emerging technologies examines some non-traditionalmethods by which enzymes may play important new roles in the drug discoveryprocesses of the future The present ability to completely locate and sequencethe gene clusters responsible for the multistep biosyntheses of complex naturalproducts has spawned new technologies Such technologies can precisely and/

or deliberately modify certain parts of gene clusters within organisms or, tively, can interchange portions of gene clusters between organisms In each in-stance, new unnatural natural products may be formed by fermentation of thenew genetically modified microorganisms The exchange of genetic material can

alterna-be logically extended into a combinatorial paradigm called combinatorial thesis or combinatorial enzymology, thereby leading to even larger numbers of

biosyn-new natural products Extensive interdisciplinary collaboration between biosyn-new get identification and screening laboratories, medicinal chemists, and molecularmodeling/computational chemists will become even more essential in the futurefor rapid discovery of useful new entities to evaluate in the field or clinic

tar-signing a precise function to genes ( functional prediction) and redetar-signing the function of enzymes (enzyme engineering) can play increasingly significant roles

in drug discovery Also, the utility of functional genomics in identifying relevant enzyme targets depends closely on the molecular understanding of these

disease-targets under physiological and pathological conditions ( functional proteomics) Enzyme Technologies for Pharmaceutical and Biotechnological Applica- tions is informative, practical, timely, and applicable worldwide to the pharma-

ceutical and biotechnological industries Real-world examples provided out the book are important for discriminating between the use of enzymes solelyfor academic studies and the practical use of enzymes in industrial applications.The reader will acquire a better understanding of applied sciences in the field.Areas that have been extensively covered in reviews and the general literature(such as use of natural lipases in organic synthesis) have been minimized here

through-By focusing on real-world applications, the reader will obtain a clearer standing of what is new and relevant in the field

under-The book is intended primarily for industrial and research scientists withinterests in adopting and maximizing enzyme technologies for pharmaceuticaldiscovery, development, and manufacturing The book can also be used by gradu-ate and postdoctoral students in practical enzymology, biochemistry, microbiol-ogy, molecular biology, and biochemical engineering, as well as by students ingraduate-level courses covering practical enzymology and enzyme biochemistry

Herbert A Kirst Wu-Kuang Yeh Milton J Zmijewski, Jr.

Preface iii

I Biosynthesis

1 δ-(L-α-Aminoadipyl)-L-Cysteinyl-D-Valine Synthetase as a

Hans von Do¨hren, Wibke Kallow, Mary Anne Tavanlar, Torsten

Schwecke, Ralf Dieckmann, and Volker Uhlmann

2 Metabolic Engineering for Cephalosporin C Yield Improvement

Joe E Dotzlaf, Steven W Queener, and Wu-Kuang Yeh

3 Bioconversion of Penicillins to Cephalosporins 61

Arnold L Demain, Jose L Adrio, and Jacqueline M Piret

4 Direct Fermentative Production of Acyltylosins by Genetically

Akira Arisawa and Hiroshi Tsunekawa

vii

5 Engineering Streptomyces avermitilis for the Production of

Novel Avermectins: Mutant Design and Titer Improvement 113

Claudio D Denoya, Kim J Stutzman-Engwall, and Hamish

Jon D Stewart, Sonia Rodrı´guez, and Margaret M Kayser

8 Cross-Linked Enzyme Crystals: Biocatalysts for the Organic

11 Penicillin-Binding Proteins as Antimicrobial Targets:

Expression, Purification, and Assay Technologies 263

Genshi Zhao, Timothy I Meier, and Wu-Kuang Yeh

12 Development of a High-Throughput Screen for Streptococcus

pneumoniae UDP-N-Acetylmuramoyl-Alanine: d-Glutamate

Ligase (MurD) for the Identification of MurD Inhibitors 289

Michele C Smith, James A Cook, Gary M Birch, Stephen A.

Hitchcock, Robert B Peery, Joann Hoskins, Paul L Skatrud,

Raymond C Yao, and Karen L Cox

13 Purification and Assay Development for Human Rhinovirus

Q May Wang and Robert B Johnson

15 Screening for Inhibitors of Lipid Metabolism 343

Hiroshi Tomoda and Satoshi O ¯ mura

16 Design and Development of a Selective Assay System for the

Hsiu-Chiung Yang, Marian Mosior, and Edward A Dennis

17 Understanding and Exploiting Bacterial Polyketide Synthases 397

Robert McDaniel and Chaitan Khosla

18 Polyketide Synthases: Analysis and Use in Synthesis 427

Kira J Weissman and James Staunton

19 Enzymatic Synthesis of Fungal N-Methylated Cyclopeptides and

Mirko Glinski, Till Hornbogen, and Rainer Zocher

20 New Strategies for Target Identification, Validation, and Use of

Joaquim Trias and Zhengyu Yuan

21 Use of Genomics for Enzyme-Based Drug Discovery 515

24 Proteomics: Chromatographic Fractionation Prior to

Two-Dimensional Polyacrylamide Gel Electrophoresis for Enrichment

of Low-Abundance Proteins to Facilitate Identification by Mass

Srinivasan Krishnan, John E Hale, and Gerald W Becker

Jose L Adrio Department of Biochemistry, Antibioticos, S.A.U., Leo´n, Spain

Akira Arisawa Biochemistry Laboratory, Central Research Laboratories, cian Corporation, Fujisawa, Japan

Mer-Gerald W Becker Lilly Research Laboratories, Eli Lilly and Company, napolis, Indiana

India-Gary M Birch Lilly Research Laboratories, Eli Lilly and Company, olis, Indiana

Indianap-Ridong Chen Department of Biochemistry, University of Saskatchewan, katoon, Saskatchewan, Canada

Sas-Andrew R Cockshott Lilly Research Laboratories, Eli Lilly and Company,Indianapolis, Indiana

James A Cook Lilly Research Laboratories, Eli Lilly and Company, olis, Indiana

Indianap-Karen L Cox Lilly Research Laboratories, Eli Lilly and Company, lis, Indiana

Indianapo-xi

Arnold L Demain Department of Biology, Massachusetts Institute of nology, Cambridge, Massachusetts

Tech-Edward A Dennis Department of Chemistry and Biochemistry, Revelle lege and School of Medicine, University of California, San Diego, California

Col-Claudio D Denoya Bioprocess Research, Global Research and Development,Pfizer, Inc., Groton, Connecticut

Ralf Dieckmann Biotechnology Center, Technical University Berlin, Berlin,Germany

Joe E Dotzlaf Lilly Research Laboratories, Eli Lilly and Company, lis, Indiana

Indianapo-Timothy G Geary Discovery Research, Pharmacia Animal Health, zoo, Michigan

Kalama-Mirko Glinski Max Volmer Institute for Biophysical Chemistry and istry, Technical University Berlin, Berlin, Germany

Biochem-Michael D Grim Westboro, Massachusetts

John E Hale Lilly Research Laboratories, Eli Lilly and Company, lis, Indiana

Indianapo-Stephen A Hitchcock Lilly Research Laboratories, Eli Lilly and Company,Indianapolis, Indiana

Till Hornbogen Max Volmer Institute for Biophysical Chemistry and chemistry, Technical University Berlin, Berlin, Germany

Bio-Joann Hoskins Lilly Research Laboratories, Eli Lilly and Company, olis, Indiana

Indianap-Robert B Johnson Lilly Research Laboratories, Eli Lilly and Company, napolis, Indiana

India-Wibke Kallow AnagnosTec GmbH, Luckenwalde, Germany

Margaret M Kayser Department of Chemistry, University of New wick, Saint John, New Brunswick, Canada

Bruns-Adam J Kreuzman Lilly Research Laboratories, Eli Lilly and Company, anapolis, Indiana

Indi-Srinivasan Krishnan Lilly Research Laboratories, Eli Lilly and Company, dianapolis, Indiana

In-Hamish A I McArthur Bioprocess Research, Global Research and ment, Pfizer, Inc., Groton, Connecticut

Develop-Robert McDaniel Kosan Biosciences, Inc., Hayward, California

Timothy I Meier Lilly Research Laboratories, Eli Lilly and Company, napolis, Indiana

India-Marian Mosior Lilly Research Laboratories, Eli Lilly and Company, olis, Indiana

Indianap-Satoshi O ¯ mura Research Center for Biological Function, The Kitasato

Insti-tute and Graduate School of Pharmaceutical Sciences, The Kitasato InstiInsti-tute andKitasato University, Tokyo, Japan

Ramesh N Patel Process Research and Development, Bristol-Myers SquibbCompany, New Brunswick, New Jersey

Robert B Peery Lilly Research Laboratories, Eli Lilly and Company, napolis, Indiana

India-Jacqueline M Piret Biology Department, Northeastern University, Boston,Massachusetts

Steven W Queener Lilly Research Laboratories, Eli Lilly and Company, anapolis, Indiana

Indi-Sonia Rodrı´guez Department of Chemistry, University of Florida, Gainesville,Florida

Siddhartha Roychoudhury Discovery-Biology, Procter & Gamble ceuticals, Mason, Ohio

Pharma-Molly B Schmid Microcide Pharmaceuticals, Inc., Mountain View, California

Torsten Schwecke Biotechnology Center, Technical University Berlin, Berlin,Germany

Paul L Skatrud Lilly Research Laboratories, Eli Lilly and Company, napolis, Indiana

India-Michele C Smith Lilly Research Laboratories, Eli Lilly and Company, napolis, Indiana

India-James Staunton Department of Chemistry, University of Cambridge, bridge, United Kingdom

Cam-Jon D Stewart Department of Chemistry, University of Florida, Gainesville,Florida

Kim J Stutzman-Engwall Bioprocess Research, Global Research and opment, Pfizer, Inc., Groton, Connecticut

Devel-Mary Anne Tavanlar National Institute of Molecular Biology and ogy, University of the Philippines Los Ban˜os, Laguna, Philippines

Biotechnol-Hiroshi Tomoda Research Center for Biological Function, The Kitasato tute and Graduate School of Pharmaceutical Sciences, The Kitasato Institute andKitasato University, Tokyo, Japan

Insti-Joaquim Trias Department of Microbiology, Versicor, Inc., Fremont, fornia

Cali-Hiroshi Tsunekawa Pharmaceuticals and Chemicals Division, Mercian ration, Tokyo, Japan

Corpo-Volker Uhlmann Department of Histopathology, St James’s Hospital, Dublin,Ireland

Hans von Do¨hren Biotechnology Center, Technical University Berlin, Berlin,Germany

Q May Wang Lilly Research Laboratories, Eli Lilly and Company, lis, Indiana

Indianapo-Hsiu-Chiung Yang Lilly Research Laboratories, Eli Lilly and Company, anapolis, Indiana

Indi-Raymond C Yao Lilly Research Laboratories, Eli Lilly and Company, napolis, Indiana

India-Wu-Kuang Yeh Lilly Research Laboratories, Eli Lilly and Company, napolis, Indiana

India-Zhengyu Yuan Versicor, Inc., Fremont, California

Genshi Zhao Lilly Research Laboratories, Eli Lilly and Company, lis, Indiana

Indianapo-Rainer Zocher Max Volmer Institute for Biophysical Chemistry and istry, Technical University Berlin, Berlin, Germany

Biochem-Valine Synthetase as a Model

Tripeptide Synthetase

Hans von Do ¨ hren, Torsten Schwecke, and Ralf Dieckmann

Technical University Berlin, Berlin, Germany

Wibke Kallow

AnagnosTec GmbH, Luckenwalde, Germany

Mary Anne Tavanlar

University of the Philippines Los Ban˜os, Laguna, Philippines

Volker Uhlmann

St James’s Hospital, Dublin, Ireland

I INTRODUCTION

A Nonribosomal Peptide-Forming Systems and Penicillin

As beta-lactam antibiotics continue to be a major contributor to human healthpreservation, research on the biosynthesis of penicillin, an almost ancient drug,continues to open up roads to new technologies and perspectives The provision

of precursor peptides to be transformed enzymatically with chemically achieved efficiency into mono- or bicyclic antibiotics has been termed by JackBaldwin and colleagues ‘‘the irreversible commitment of metabolic carbon tothe secondary metabolism’’ [1] The synthesis of such peptides is indeed per-formed by a remarkable class of synthetases which, in contrast to the protein-synthesizing machinery, have been termed a nonribosomal system or nonribo-somal peptide synthetases (NRPS) [2] These peptide synthetases have beenshown to catalyze the irreversible synthesis of peptides differing both in sequenceand structural variability, thus extending the scope of directly gene-encoded poly-

un-1

peptides The synthetic principle, unravelled mainly by Kiyoshi Kurahashi, So¨renLaland, Fritz Lipmann, and later Horst Kleinkauf and colleagues (reviewed in[3]), consists of a process of amino acid selection, activation of carboxyl groups

as adenylates, formation of stable thioester intermediates, and successive sations in an assembly-line process with no free intermediates The enzymaticmachinery thus resembles a kind of self-feeding solid-phase process Similar se-quential condensation processes are known from polyketides and related acetate-derived compounds, and many compounds combining the respective buildingblocks in mixed biosynthetic systems are known

conden-A major step in our continuing understanding of these remarkable multistepsystems goes back to the cloning of the peptide synthetase producing the penicil-lin precursor tripeptide δ-(L-α-aminoadipyl)-L-cysteinyl-D-valine (ACV) Thefirst sequences of this remarkable and quite large gene became available in 1990,

derived from Penicillium chrysogenum [4,5], Aspergillus nidulans [6], nium chrysogenum [7], and Nocardia lactamdurans [8] These efforts immedi-

Acremo-ately led to the revision of the classical thiotemplate model to the multiple-carrierthiotemplate model [9,10], which was later verified experimentally [11,12] Atthe same time, the ACV synthetase sequences were the first to introduce us tothe multidomain structure of this class of enzymes, and thus initiated the still-continuing efforts to understand the molecular principles of multistep synthetasesand synthases

B From Tripeptide to Synthetase

The formation of the tripeptide ACV (the Arnstein tripeptide) upon feeding of

labeled valine to mycelia of P chrysogenum had already been shown by Arnstein

and co-workers in 1960 [13,14], and demonstrated later in cell-free extracts byBauer [15] In more refined work, Abraham and Loder established in extracts

of A chrysogenum the MgATP2 ⫺-dependent formation of ACV from δ-(L-αaminoadipyl)-L-cysteine (AC) and L-valine [16,17], while no product was ob-served with L-α-aminoadipate (Aad) and L-cysteinyl-L-valine (LL-CV), or ACand D-valine The stereochemistry of the tripeptide has been established as LLD[18,19], and the biosynthesis from the respective L-configured amino acids wasagain demonstrated in a beta-lactam-defective mutant by Shirafuji and colleagues

-[20] This mutant Takeda N2 of A chrysogenum, later shown to carry a point

mutation in the isopenicillin N synthase gene, was shown by Adlington and workers to produce both AC and ACV [21] Among tripeptide analogs isolatedfrom fermentation broth were Aad-Ala-D-Val (D configuration not proven), Aad-Ser-D-Val, and Aad-Ser-isodehydro-Val [22] At that time it was generally ac-cepted that the Arnstein tripeptide would be the product of two consecutive reac-tions catalyzed by the hypothetical AC and ACV synthetases [23] Thus Lara and

co-colleagues, studying the stimulation of penicillin formation in P chrysogenum by

L-glutamate, determined the hypothetical level of AC synthetase by phosphate

The possible analogy of ACV formation to the well-established plate processes had earlier led Feodor Lynen (H Bergmeyer, unpublished data)and then Arny Demain to take a closer look at these synthetases Employing the

thiotem-industrial strain C-10 of A chrysogenum (ATCC 48272), Banko in Demain’s

laboratory succeeded in isolating an AC-forming enzyme [25], and later showedthe rate of ACV production by this preparation from the three amino acids to befaster than from AC and L-valine [26] Similar results were obtained by Jensen

in Westlake’s laboratory with the actinomycete Streptomyces clavuligerus [27],

and ACV synthesis was also achieved with an immobilized enzyme preparationbound to an anion-exchange resin [28] The isolation and purification of an ACV

synthetase has first been reported with the A nidulans enzyme [29], later to be followed by synthetases from A chrysogenum [30–32] and S clavuligerus [33–

36] These studies established a single multienzyme to be responsible for thesynthesis of ACV from Aad, Cys, Val, and MgATP2 ⫺ An associated protein

in the S clavuligerus preparation [34] was later shown to be proclavaminate amidinohydrolase [37] Thus, possible in vivo interactions with other proteins in

this biosynthetic process are lost during standard purification operations

II ACV SYNTHETASES: ORGANIZATION, PURIFICATION,

AND CHARACTERIZATION

A Overall Organization

Protein sequence data of prokaryotic and eukaryotic ACV synthetases reveal asimilar organization of three adenylate domains, three thiolation domains (carrierproteins), two condensation domains, one epimerization domain, and one thioes-terase domain These domains have been well characterized from multiple se-quence data of various NRPS systems [38–40] and data are easily accessable,

with the terms AMP binding domain for adenylate domains (IPR000873, PS00012, PF00501); pp-binding, phosphopantetheine, or acp domain for thiola-

tion domains (IPR000255, PS00455, PF00550); condensation and epimerization

domains as DUF4 (IPR001240) with unknown function; and thioesterase

do-mains (IPR001031) A primary approach in the identification of dodo-mains is thedetection of highly conserved core sequences Such sequences have been derivedfrom selected NRPS data [38] The respective ACV synthetase cores show anexcellent agreement with little deviation (Table 1) Inspection of the primarydata shows alterations especially in the C-terminal subdomains of the adenylatedomains (motifs A8–A10), and in the condensation domains Domain boundariescan be predicted from multiple alignments, which help to identify linker regions[41] The unique availability of both prokaryotic and eukaryotic sequences mayprovide additional insights not yet evaluated by data analysis These sequences

include the far N-terminal regions of about 40 amino acid residues present infungal synthetases, which points also to some insecurity with respect to initiationsites and the actual terminal sequences Another interesting region is the N-terminal region of about 200 amino acids preceding the first adenylation domain,which shows significant similarities to an internal region of condensation do-mains The recent identification of terminal and linker-associated sequences in-volved in domain interactions has not yet been investigated in NRPS systems[42,43].

As the second elementary step of data analysis, the comparative multiplealignment of the identified domains is conducted Complete adenylate domainsequences provide an ACV synthetase cluster, with subclusters of the individualadenylate domains in positions 1, 2, and 3 [44] This clustering indicates a similar-ity of ACV synthetase domains compared to other NRPS systems, and a likewisesimilarity of the domains in their respective positions This clustering has beentaken as evidence for the evolution of NRPS by gene duplication within individ-ual systems [45], and would as well support the horizontal gene transfer hypothe-sis of beta-lactam biosynthetic clusters [46] A different picture, however,emerges when the sequences of thiolation or condensation domains are compared:the carrier domains of the first modules are found in a different cluster than those

of the second and third module [47] Likewise, a CLUSTAL alignment of thetwo condensation domains places them in different contexts, which have beenspeculatively correlated with the specific type of condensation reaction (linking

aδ-carboxyl group with anα-amino group, or linking a cysteine carboxyl groupwith an α-amino group) [47] The condensation domain-related epimerizationdomain is found within the respective cluster of epimerization domains Thus, ifapplied to thiolation, condensation, or epimerization domains, the system-specificclustering observed for adenylation domains is not evident, possibly due to themore diverse differences in functions More detailed structural investigations areneeded to interpret the functional significance of these observations

The three-module system found in all ACV synthetases correlates well withthe sequential condensation, epimerization, and hydrolytic release of the tripep-tide At first sight, the arrangement of domains and modules fits perfectly thecollinearity rule found in the majority of NRPS systems, which implies that thelinear sequence of domains corresponds to the sequence of reactions of the path-way [40,45]

B Purification Work and Protein Studies

Studies have been performed on the purification and characterization of

synthe-tases from A nidulans, A chrysogenum, S clavuligerus, Flavobacterium sp [48],

N lactamdurans [49], and P chrysogenum [50] Except for Flavobacterium and S clavuligerus, the complete gene sequences are available, which reveal

Module 2

Module 3

Module 2

Module 3

Module 1

Module 2

Module 3

Domain 2

Domain 1

Domain 2

Domain 2

a Domains and their respective core sequences are taken from Ref 38; for clarity, multiple residues in one position have been listed on a second line.

b Bold residues are altered with respect to the defined core sequences.

extensive similarities discussed in detail below Surprisingly, the isolation of

the synthetase from strains of P chrysogenum proved most difficult Steps

fol-lowing salting out by ammonium sulfate fractionation were found to be followed

by proteolytic cleavage and inactivation (von Do¨hren et al., unpublished data).These problems were only recently overcome by Theilgaard and colleagues[50]

In perspective, the lessons learned from other NRPS systems have beenconfirmed Due to the large size, ACV synthetases may be salted out at fairlylow concentrations of ammonium sulfate and are easily enriched by steps such

as gel filtration or purified by gradient centrifugation methods Strains to be

ap-plied should be industrial high producers, such as A chrysogenum C-10, or strains containing high levels of enzyme, such as S clavuligerus So far, overexpressing

constructs which might permit easy access to fungal NRPS are missing An

ap-proach for heterologous expression of the N lactamdurans ACV synthetase in Streptomyces lividans proved successful, but only about 0.25% of the total protein

constituted the synthetase, and about 30% of the activity was recovered after twocolumn purifications [49]

Comparative assessments of purification protocols are generally difficult,and wide ranges of activities have been reported ACV synthetases have beenassayed by either employing a peptide adsorption resin (Porapak) [29,48] or highperformance liquid chromatography (HPLC) detection of the derivatized tripep-tide [26,27] The adsorption assay, employing radiolabeled amino acids, has been

Table 2 Purification of ACV Synthetases: Activities and Recovery

Activitya Protein(turnover recoveryb Activity

a Activity determined by adsorption assay (A) or HPLC (H).

b Protein recovery with respect to total protein (T) or ammonium sulfate fraction (A).

limited use, due to traces of unlabeled substrates diluting the label and tion of amino acids and by-products, if no further purification by, e.g., thin-layerchromatography (TLC) is performed Estimates from optimized assays relying

coadsorp-on the HPLC detecticoadsorp-on of ACV show comparable turnover numbers in the range

of 4–5 min⫺1 for the S clavuligerus enzyme, and 8–11 min⫺1 for the A chrysogenum enzyme (Table 2, [51]) These rates compare well with those estab-

lished for other NRPS systems, such as gramicidin S, cyclosporin, or enniatinsynthetases This rate, however, is the sum of 10 to 40 individual enzymaticreactions leading to a complex product The purification protocols illustrate wellthe balance between activity and purity, considering the decline in activity due

to proteolytic degradation and oxidation of these large proteins [54,55] Thus,effective purification should not exceed two column steps, and conditions pro-moting enzyme damage should be reduced to a minimum

Table 3 ACV Synthetases: Protein Properties

Molecular mass Molecular mass, Molecular mass

Lysobacter sp. 411,593 Not available Not available 56

Flavobacterium Not available Not done 300 (G)d

48sp

aData are available under the accession numbers: X54853 (Aspergillus); (a) X54296 (Penicillium, Turner et al.); (b) P26046 (Penicillium, Martin et al.); P258464 (Acremonium); P27743 (Nocardia); D50308 (Lysobacter); assumed holo-enzyme data should contain additional 3⫻ 340 Da for cofac- tors.

b Polyacrylamide gel electrophoresis under denaturating conditions employing dodecylsulfate.

c Estimations done either using native gel filtration (G) or native polyacrylamide gel electrophoresis (P).

d Reference value has been corrected with respect to gramicidin S synthetase 2 (512 kDa).

Identity of the isolate with the corresponding genes has been verified for the

synthetases from A nidulans [6] and A chrysogenum [30] by sequence analysis of

tryptic peptides Estimation of the molecular mass of purified ACV synthetasesagree with the predicted reading frames Only few marker proteins are available

in the respective region of 200 to 1000 kDa, and evaluations of electrophoreticmobility as a function of size in the presence of dodecylsulfate have not beenclearly established (Table 3) The more essential question of whether the synthe-tases associate into dimers has been approached by electrophoresis under nativeconditions, gel filtration, and sedimentation studies Again, the uncertainties inthese migration studies are high, and shape factors matter even more than for

protein–dodecylsulfate complexes Except for the A chrysogenum enzyme, no

significant evidence has been seen for dimerization No enzyme concentrationdependence of ACV synthesis has been found, however, for a functional support

of this finding [51]

III BASIC CATALYTIC PROPERTIES

A Enzyme Catalytic Studies

1 Basic Reaction Scheme

The evaluation of kinetic properties of NRPS systems is a problem of generallyunderestimated complexity The basic path was established in 1971, defining acti-vation, thiolation, and peptidyl transfers as basic reactions The further refinementfrom structural data to establish the multiple carrier model, and now to tackledomain interactions, has added some precision to the questions asked However,

we have not yet arrived at a complete kinetic description of even the simpletripeptide synthetase The ACV synthetase operates with four different substrates

at six binding sites, releasing 3 moles of AMP and 3 moles of MgPPi for eachACV formed at optimal conditions [51] A sequence of 10 reactions has beenproposed in analogy to other NRPS-systems:

E1-Spl-Aad⫹ E2-Sp2-CysC1→ E2-Sp2-Cys-Aad⫹ El-SplH (5)

E3-Sp3-Val-Cys-Aad→Ep3 E3-Sp3-D-Val-Cys-Aad (9)

→TE Aad-Cys-D-Val⫹ E3-Sp3H (10)The synthetase consists of the three modules E1, E2, and E3 (for a completedescription, see Sec II.A) Each module is composed of an activation site formingthe acyl or aminoacyl adenylate, a carrier domain which is posttranslationallymodified with 4′-phosphopantetheine (SP), and a condensation domain (C1, C2)

or, alternatively, a structurally similar epimerization domain (Ep) Activation ofaminoadipate (Aad) leads to an acylated enzyme intermediate, in which Aad isattached to the terminal cysteamine of the cofactor (E1-Sp1-Aad) [reactions (1)and (2)] Likewise, activation of cysteine (Cys) leads to cysteinylated module 2[reactions (3) and (4)] For the condensation reaction to occur between aminoadi-pate as donor and cysteine as acceptor, both intermediates are thought to react

at the condensation site of module 1 (C1) Each condensation site is composed,

in analogy to ribosomal peptide formation, of an aminoacyl and a peptidyl site

In this case of initiation, the thioester of Aad enters the P-site, while the thioester

of Cys enters the A-site Condensation occurs and leaves the dipeptidyl ate Aad-Cys at the carrier protein of the second module [reaction (5)] The thirdamino acid valine is activated on module 3, and Val is attached to the carrierprotein 3 [reactions (6) and (7)] Formation of the tripeptide occurs at the secondcondensation site C2, with the dipeptidyl intermediate entering the P-site and thevalinyl-intermediate the A-site [reaction (8)]

intermedi-Finally, epimerization of the tripeptide (or dipeptide) intermediate occurs

at the epimerization site of module 3 (Ep3) [reaction (9)], and the stereospecificpeptide release is controlled by the thioesterase (TE) [reaction (10)]

Early work summarized in Sec I.B had shown that ACV is made fromL-Aad, L-Cys, and L-Val, thatδ-(L-α-Aad)-L-Cys (AC) may be converted intoACV, but L-Cys-D-Val is not converted Likewise, D-Val is not a substrate forACV synthetase, contrary to the first-characterized NRPS systems of gramicidin

S and tyrocidine [2,3] These observations are in agreement with the scheme,except for the incorporation of AC This dipeptide was later shown to be activated

as an adenylate

2 Adenylate Formation

Activation of the amino acid carboxyl groups is the only set of clearly reversiblereactions catalysed by NRPS So far these reactions have been studied in thecase of ACV synthetases only by the amino acid-dependent ATP–PPi exchangereaction [reaction (11)]:

E⫹ RCO⫺

2 ⫹ MgATP2 ⫺s E{RCOAMP} ⫹ MgPPi2 ⫺ (11)

In case of equilibrium, addition of labeled PPi yields labeled ATP and this productcan be employed to detect NRPS or related enzymes The respective reactionrates provide information on adenylate formation/pyrophosphorylysis, apparent

Kmof substrates and substrate analogs, and with some enzyme kinetic efforts,substrate affinities and the patterns of substrate binding may be deduced Theease and the sensitivity of the procedure makes it the primary method of investi-gation of NRPS substrate specificity

There are several remarkable features of this isotope-exchange assay Ifthe reaction progress is followed, equilibrium of isotope exchange is attained at

a defined distribution of the label between ATP and PPi This has been shown

in the case of ACV synthetase by Baldwin and colleagues [57], but is rarely donebecause of the restrictions of enzyme stability Evaluation of this equilibrium,

which has been reached in the case of the A chrysogenum enzyme in about 3

hr, has never been attempted, but the concentrations roughly reflect the initialrates of the exchange reactions

Following structure–activity studies, the adenylate is thought to be lized within a cleft formed between the two subdomains of the activation domain[41,58] The rate is thus related to the formation, presence, and stability of thismixed anhydride with respect to PPi, and at high MgATP2⫺concentrations, withrespect to ATP in the formation of diadenosine tetraphosphate (A2P4) Thus ahigh rate of the amino acid-dependent isotope exchange does not necessarilyreflect the efficiency of adenylate formation, and certainly not the efficiency ofincorporation of an amino acid into peptidyl intermediates or the final pro-duct

stabi-As most NRPS multienzymes are multidomain proteins with multiple vation domains, multiple sites may participate in the reactions assayed, and noclear result concerning a single specific site may result In ACV synthetases, the

acti-nonadditivity of the initial rates has been observed in the S clavuligerus enzyme [35] and the A chrysogenum enzyme [1] Two or more site activations of one

substrate amino acid could be expected to depend on different binding constants,and thus be detectable by kinetic analysis So far, however, no evidence for mixedtypes of concentration dependence has been found It is thus not yet clear ifnonadditivity results from misactivation or alteration of kinetic properties in thepresence of multiple substrates In the case of gramicidin S synthetase 2, evidencefor misactivations has been reported [59]

A fairly large number of potential amino acids has been assayed for ate formation These amino acids can be grouped into compounds presumablybeing processed by just one of the three domains, or by several of the domains.Some of the data have been compiled in Table 4, and the main questions to beaddressed are: (1) Do we find significant differences in fungal and bacterial ACV

adenyl-Substrate ACVS A chrysogenuma ACVS S clavuligerusbGroup 1: predicted Aad analogs

b Data have been taken mainly from Ref 51.

c Note that there are large differences in the rates of Aad activation between the fungal and bacterial enzymes.

d n.a., not available.

Source: From Refs 47, 51; von Do¨hren et al., unpublished.

synthetases with respect to site specificity? and (2) How does adenylate formationreflect the uptake into the peptide products?

Early investigations missed the detection of 2-aminoadipate activation in

the S clavuligerus ACV synthetase [1] Thus adenylate formation had been

ques-tioned, and even a study titrating the release of PPi from gamma-labeled ATP[60] showed less than 1% yield compared to the amount of enzyme applied [60].Employing Aad concentrations in the range of 5–20 mM and substituting Trisbuffer with MOPS buffer, Schwecke finally did demonstrate an Aad-dependent

ATP-PPi exchange [60] with a K mof about 10 mM Another substantial difference

of the Aad domains is activation of L-Glu by the fungal enzyme, while no tion was detectable in the bacterial one Another significant difference concernsthe activation of the dipeptide Aad-Cys, which again does not promote the isotope

activa-exchange in S clavuligerus ACV synthetase As a third major difference,

aro-matic amino acids are well activated by the bacterial ACV synthetase, but not

as well by the A chrysogenum enzyme A surprising result has been reported with ACV synthetase from N lactamdurans expressed in S lividans: the enzyme

converted 6-oxo-piperidine-2-carboxylic acid and cystathionine into ACV [49]

As only nanomoles of labeled ACV were formed from millimolar amounts ofprecursors, the data need to be substantiated in more detail No adenylate forma-

tion has been found with cystathionine and the A chrysogenum enzyme (von

Do¨hren et al., unpublished)

pockets This analysis has led both Stachelhaus and Challis and colleagues tothe proposal of a nonribosomal code [45,62] Aligning the polypeptide sequencesbetween the core motifs A3 and A6 (see Sec II.A), eight or nine pocket-liningresidues are predicted Identical or similar residues permit the prediction of aminoacid specificity with remarkable accuracy The bacterial and fungal ACV synthe-tase domains each show identical sets of residues, which, however, do not matchany set from other known NRPS systems (Table 5) Stachelhaus et al [62] listthe derived ACV synthetase codes as defined templates for Aad, Cys (type 2),and Val (type 2) Most striking is the presence of charged side chains in all ofthese signatures, which would not be neutralized by substrate charges Challis et

al [45] include only some of the ACV synthetase data in their analysis, and

Table 5 Predicted Amino Acid-Binding Side Chains for ACV Synthetases

and Other NRPS Systems

fungalACV synthetase, EPRNLVEA von Do¨hren et al.bacterial

Lys2, various eu- DPRHFVMQ von Do¨hren et al.karyotes

fungalACV synthetase, DHESDIGI von Do¨hren et al.bacterial

fungalACV synthetase, DFESLAAY von Do¨hren et al.bacterial

comment on the proposed Cys pocket as being close in structure to the Val pocket,and differing from other NRPS due to a speculative isolated evolution by geneduplication The predictive structure-based model of amino acid recognition, de-rived largely by comparative analysis of bacterial NRPS domains, clearly is notapplicable for ACV synthetases A possible explanation could be found in a dif-fering architecture of these substrate-binding sites Thus an alternative selection

of the critical substrate-lining residues might define differing contact sites withoutthe problem of unbalanced charges

To derive binding constants, the multiplicity of substrates has to be ered Most work has settled for a constant MgATP2 ⫺concentration to approxi-

consid-mate a K m with simple Michaelis-Menten kinetics It is not clear, however, if

such K mvalues derived for either ACV synthesis or the amino acid-dependentATP-PPi exchange are comparable The available data have been compiled inTable 6

3 Adenylate Domains

To definitely assign specific catalytic activities to defined regions, the dissection

of ACV synthetases has been attempted by proteolytic and molecular genetictechniques (Fig 1) Several quite unexpected results have been obtained As

shown in Fig 1, limited proteolysis of A chrysogenum ACV synthetase by

pro-teinase K and subtilisin produced two fragments of 119 and 95 kDa, respectively[63] The 119-kDa fragment (fragment 1) contained the second adenylate domainand specifically catalyzed a cysteine-dependent ATP–PPi exchange reaction The

Table 6 Kinetic Constants (Km) of ACV Synthetases

Figure 1 Structural organization of ACV synthetases and proteolytic (1,2) as well asexpressed fragments (3–7) Top: domain structure, A1–A3 adenylate domains of the threemodules, T1–T3 thiolation domains, C1 and C2 condensation domains, E epimerizationdomain, TE thioesterase, and N aminoterminal domain Triangles indicate identified subtil-

isin and proteinase K cleavage sites of the A chrysogenum enzyme leading to fragment

1 (activating cysteine) and fragment 2 (activating 2-aminoadipate) [63] Fragment 3 is the

complete A3 domain of P chrysogenum expressed in A nidulans, which activates leucine

and 2-aminoadipate, but not valine [63] Fragment 4 is the amino-terminal region of the

S clavuligerus enzyme, expressed in insoluble form in E coli, and solubilized All three

substrates, aminoadipate, cysteine, and valine, form adenylates, but exclusively aminoadipate is bound as thioester [60,64] Fragment 5 is the C-terminal fragment of the

2-A chrysogenum enzyme, expressed in E coli in insoluble form, solubilized, and shown to

activate leucine, valine, and 2-aminoadipate (von Do¨hren et al., unpublished) Fragments 6

and 7 of the P chrysogenum enzyme were expressed in soluble form in A nidulans as

β-galactosidase fusion proteins, and activate 2-aminoadipate and valine, and cysteine, line, and 2-aminoadipate, respectively [65]

va-95-kDa fragment containing the third adenylate domain yielded a second age product of 47 kDa (fragment 3), which surprisingly activated Aad Activation

cleav-of Val was completely lost upon proteinase treatment

In molecular genetic approaches, a 110-kDa N-terminal fragment of the S clavuligerus ACV synthetase has been expressed in Escherichia coli (fragment

4) [60] The insoluble protein obtained was dissolved in urea and renatured, but

activation of all three amino acids was detected, with Val giving the highestreaction rate Although only a fraction of the expressed protein is posttransla-tionally modified by pantetheine, the formation of thioesters had been investi-gated, and only Aad was detected [60,64] This experiment demonstrated a selec-tive step following adenylate formation for the first time It further demonstratesthat although misactivation may proceed in a fragment of the synthetase, a furtherselection of intermediates may take place.

As these experiments required the solubilization of denaturated tides, some doubt as to their validity is justified To overcome the folding andsolubility problems, Turner et al accomplished the homologous expression of

polypep-ACV synthetase fragments from P chrysogenum in A nidulans [65,66] A strain carrying an acvA deletion was used to expressβ-galactosidase fusions of the firstmodule and a fragment containing the second and third module, respectively(fragments 6 and 7, β-galactosidase fusion not shown) The respective fusionproteins were obtained only in low yield, and proteolysis precluded their completepurification However, the first domain was shown to activate Aad, Val, Cys,isoleucine, allo-isoleucine,α-aminobutyrate, S-carboxymethyl-cysteine, and glu-tamic acid, which is in complete agreement with the activity data of the bacterialfragment The C-terminal fragment did activate Cys, which also was shown toform a thioester, but in addition produced adenylates with Val, isoleucine, leu-cine, and α-aminobutyrate No activation of Aad was detected These resultsagain supported the assignment of Aad activation to the first adenylate domain.The third adenylate domain has been further investigated by fragment ex-

pression A C-terminal 136-kDa fragment of the A chrysogenum ACV synthetase was expressed in E coli and the resulting protein pellet solubilized and refolded

(von Do¨hren et al., unpublished, fragment 5) Activation of leucine, Val,

α-aminobutyrate, Aad, aminocaproic acid, and norvaline was detected To

achieve expression of a soluble enzyme, the adenylate domain of the P chrysogenum synthetase was excised as a 501-amino acid fragment by polymer- ase chain reaction (PCR) and expressed in A nidulans, again as aβ-galactosidasefusion protein (fragment 3) [63] The purified protein showed adenylate formationwith leucine, 2-amino-ethyl-cysteine, Aad, S-carboxymethyl-Cys, but surpris-ingly not with Val The interpretation of the third-domain data implies that theVal-dependent ATP–PPi exchange activity diminished with the reduction of frag-ment size, and finally disappeared in the fusion protein Aad and leucine activa-tion have been found in the smaller fragments

These results show a clear distortion of substrate binding and catalytic tivities upon fragmentation of the multienzymes The substrate pocket architec-ture seems to depend on the context of adjacent domains as well Although ques-tions remain, the linearity rule of NRPS holds in ACV synthetases Openquestions remain on the fate of possibly misactivated amino acids in the terminal

ac-4 Peptide Synthesis and Adenylate Intermediates

In a remarkable series of experiments, Baldwin and colleagues have studied

pep-tide bond formation in the A chrysogenum ACV synthetase [1,57,67–69] In the

presence of glutamate, Cys, and Val, L-cysteinyl-D-valine is recovered sively As has been stated above, glutamate does form an adenylate, but fails to

exclu-be incorporated into peptides The glutamate adenylate is thus not accessable forcondensation to proceed at the first condensation domain Presumably its pres-ence, due to an induced conformational change [41], enhances peptide bond for-mation at the second condensation domain between the thioesters of Cys and Val,followed by epimerization of the dipeptidyl intermediate and hydrolytic release[reactions (12) and (13), with Cys for A2]

E2-Sp2-A2⫹ E3-Sp3-Val→E3-Sp3-Val-A2⫹ E2-Sp2H (12)E3-Sp3-Val-A2→Ep3 E3-Sp3-D-Val-A2 →TE A2-D-Val ⫹ E3-Sp3H (13)Likewise, the assumed O-methyl-serinyl-thioester intermediate reacts onlyslowly with the Aad thioester, but is readily released by aminolysis of free Val[reaction (14), with A2 either Cys or O-methyl-serine] or may react with theVal-thioester intermediate to be further processed [reactions (12) and (13), withO-methyl-serine for A2]

Following the dipeptide synthesis from O-methyl-serine and Val by termining the loss18O from di[18O]valine, evidence for both reactions was ob-tained, but a direct reaction of Cys-AMP with free Val cannot be excluded [reac-tion (15)]

The dipeptide L,L-O-(methyl-serinyl)valine was formed without significantloss of label, and at the same time, no label was observed in the AMP released.This result thus excludes a thioester intermediate, which by thiolysis of the aden-ylate would have led to an even distribution of the18O between AMP and Val.However, the isomeric dipeptide L,D-O-(methyl-serinyl)valine was recoveredwith all possible labeling patterns of18O18O,16O18O, and16O16O [69] To explainthe retainment of label in the epimerized dipeptide is not easy Baldwin andcolleagues propose an alternative direct acyl transfer mechanism operating withdipeptidyl adenylates of the type Cys-Val, being epimerized, and transferred to

a thiol group to undergo peptide bond formation with the Aad thioester followed

by hydrolysis [1] This is a possible interpretation of the data, but the rates offormation of the dipeptide shunt or byproducts in the 1–2% range compared to

ACV formation may exclude this path from in vivo or optimized in vitro

condi-tions Any direct acyl transfer mechanism would be based on the surface tion of free adenylates between the respective condensation domains, which isboth unlikely and not required to interpret the available experimental data [67–69] Evidence for the collinearity rule in the catalytic sequence has been obtained,omitting Val from the tripeptide biosynthetic assay: the dipeptide Aad-Cys hasbeen shown to accumulate as a thioester intermediate [70]

migra-To demonstrate peptide bond formation from aminoacyl adenylates, mann [71] has applied the isolated adenylate domain of tyrocidine synthetase 1

Dieck-to generate various dipeptides These were obtained from phenylalanyl adenylatewith alanine, leucine, leucineamide, and phenylalanine [reaction (16), with

AA⫽ amino acid or amine acceptor]

Further studies are needed on the kinetics of aminolysis of adenylates andthioesters and the possible migration of adenylates to contact reactive intermedi-ates Such surface diffusion or tunneling has to compete with the effectivepantetheine-mediated covalent transport system

5 Thiolation

Aminoacylation or acylation of the ‘‘swinging arm’’ cofactor 4′eine is considered as the covalent transport principle in NRPS and polyketidesynthases (PKS) Experimental procedures to establish the presence of thio-ester intermediates have largely relied on the demonstration of acid stable andperformic acid-cleavable radiolabeled amino acids This approach has recentlybeen extended by the mass spectrometric detection of cleaved intermediates andcofactor-containing enzyme fragments [12,70] Any radiolabeling procedure ofenzyme-bound intermediates requires free pantetheine thiol groups, but these may

-phosphopanteth-be acylated in the respective enzyme preparation Stabilities of aminoacyl- andpeptidylthioesters depend on the type of acyl compound involved Rates of hy-drolytic cleavage have been estimated in the gramicidin S system [72] At 3°C,half-lives for aminoacyl- or peptidylthioesters were between 1 and 90 hr Reducedstabilities of 0.4–0.5 hr were observed for thioesters of ornithine or ornithyl-peptides due to the cyclization to 3-amino-2-piperidone The 2-aminoadipylintermediate apparently does not cyclize effectively, since 6-oxo-piperazine-2-carboxylic acid has not been detected upon incubation of ACV synthetase from

P chrysogenum in the presence of Aad and MgATP2[50,73] It is thus not prising that aminoacylation levels are often well below the expected estimates

sur-ylate and thiolation domains, or possibly of nonadjacent domains in trans [75].

Carrier domains have a conserved structure of four helices with a loopcontaining the cofactor attachment site [76,108] 4′-Phosphopantetheine-proteintransferases (PPT) from eukaryotes have so far not been identified, but partially

purified from A nidulans, Fusarium scirpi, and Tolypocladium niveum (von

Do¨h-ren et al., unpublished) AppaDo¨h-rently the beta-lactam biosynthetic cluster does notcontain the respective gene, but at least some of the producer strains have beenshown to contain additional NRPS clusters The quantitation of the cofactor inisolated synthetases represents an unsolved analytical problem Microbiologicaldeterminations are not reliable, as underestimates are generally obtained [1,30]

A possible approach to determine the state of posttranslational modification, theratio of apo- to holoenzyme, or the cofactor stoichiometry is to measure the 4′-phosphopantetheine transfer by the pantetheine-protein transferase assay [77–

79] A PPT fraction obtained from A nidulans, which modified apo-tyrocidine synthetase 1 from Bacillus brevis, did not transfer significant amounts of pante-

theine from labeled CoA to ACV synthetase, indicating the absence of synthetase So far no fungal PPT-genes have been identified

apo-The central role of the thiolation domain is evident from their multipleinteractions with adjacent and nonadjacent domains The current multiple-carrierthiotemplate model predicts successive contacts of the first thiolation domain ofACV synthetase with the adjacent aminoadipate adenylate domain and the firstcondensation domain; the second thiolation domain to interact with the cysteineadenylate domain, and both the adjacent and nonadjacent condensation domains;the third thiolation domain then interacts with the valine adenylate domain, thenonadjacent second condensation domain, the adjacent epimerization domain,and the nonadjacent thioesterase domain (Fig 2) Protein regions involved inthese successive protein–protein interactions involve the highly conserved carrierdomain structures of only about 80 amino acids So far, only two detailed struc-tures of the respective NRPS domains are available [61,108]

6 Peptide Bond Formation

Condensation reactions require a domain of about 450 amino acids, which hasbeen functionally identified employing gramicidin S/tyrocidine synthetase sys-tems [80] The current functional interpretation proposes, in analogy to the ribo-somal system, an aminoacyl site and a peptidyl site to enter the activated interme-diates [47] The acylated carrier proteins would thus resemble charged tRNAs,and the condensing site the peptidyl transferase As condensing domains showstructural differences, the question of control of incoming substrates may be rele-vant It has been shown by Belshaw and colleagues with a condensing domain