Cyclosporin Synthetase THE MOST COMPLEX PEPTIDE SYNTHESIZING MULTIENZYME POLYPEPTIDE SO FAR DESCRIBED*

Cyclosporin A and its homologues are synthesized by a single multifunctional enzyme from their precursor amino acids. Cyclosporin synthetase is a polypeptide chain with a molecular mass of approximately 800 kDa. In 3% polyacrylamide-sodium dodecyl sulfate gels it shows a single band of approximately 650 kDa, which appears to not be glycosylated. The enzyme could be purified to near-homogeneity in five steps. A 72-fold purification was obtained. All constitutive amino acids of cyclosporins are activated as thioesters via aminoadenylation by the same enzyme. Then N-methylation of the thioester-bound amino acids which are present in methylated form in the cyclosporin molecule takes place, whereby S-adenosyl-L-methionine serves as the methyl group donor. Methyltransferase activity is an integral entity of the enzyme; this could be shown by a photoaffinity labeling method. 4'-Phosphopantetheine is a prosthetic group of cyclosporin synthetase similar to other peptide and depsipeptide synthetases. Cyclosporin synthetase shows cross-reactions with monoclonal antibodies directed against enniatin synthetase.

Cyclosporin A (Fig. 1) is a cyclic undecapeptide with antiinflammatory, immunosuppressive, antifungal, and antiparasitic properties (1). It is used in transplantation surgery and in the treatment of autoimmune diseases (2,3). Cyclosporin A is produced by the fungus Beauveria nivea (previously designated Trichoderma polysporum, Tolypocludium infkztum, and Tolypocladium niveum) as the main component of 25 naturally occurring cyclosporins which have substitutions of amino acids in positions 1, 2, 4, 5, 7, and 11 and/or contain unmethylated peptide bonds in positions 1,4, 6, 9, 10, or 11 (4). Beside these naturally occurring cyclosporins, some cyclosporins differing in positions 1, 2, and 8 from cyclosporin A could be produced by feeding amino acid precursors to the fungus (5).
The structure of cyclosporin A strongly suggested a nonribosomal biosynthesis mechanism (6): three unusual amino acids (Bmt' in position 1, 2-aminobutyric acid in position 2, and D-alanine in position 8), seven N-methylated peptide bonds, and the cyclic structure of the molecule. The latter properties cyclosporin A shares with the depsipeptides enniatin and beauvericin which have been shown to be synthesized by large multienzyme complexes from their primary precursors (amino and hydroxy acids) under ATP and AdoMet consumption (7)(8)(9). Synthesis of depsipeptides involves aminoadenylation of precursors, binding of the activated precursors as thioesters, N-methylation of the corresponding enzyme-bound amino acids, elongation, and cyclization reactions (10,11).
Previous attempts to characterize the enzyme system responsible for synthesis of cyclosporins first led to the enrichment of an enzyme fraction catalyzing the synthesis of the diketopiperazine cycle-(o-Ala-MeLeu), representing a partial sequence (positions 8 and 9) of cyclosporin A (12). Although this preparation was able to activate all constitutive amino acids of cyclosporin A as thioesters via aminoadenylation, total synthesis of cyclosporin A was not observed. Further efforts guided to total in vitro synthesis of several cyclosporins by partially purified cyclosporin synthetase fractions (13) and led recently to the in vitro biosynthesis of cyclosporins not obtainable by fermentation (14). This paper describes further purification and characterization of cyclosporin synthetase and confirms that cyclosporin synthetase follows a thiotemplate mechanism (15), which has been shown previously for the biosynthesis of various other peptides and depsipeptides (16). formed was extracted as described in Ref. 13. TLC analysis was done as described (14).
For molecular weight determinations a Beckman SW 41 rotor was used; the molecular weight standards used were ribosomal 40 S subunits prepared according to Ref. 17, thyroglobulin, and katalase (Boehringer Mannheim).
Protein was determined by a dye-binding method (18) using bovine serum albumin as standard.
SDS-Gel Electrophoresis-This was done routinely in gradient gels from 15 to 2% polyacrylamide in the Laemmli system (19). For molecular weight estimations 3% polyacrylamide gels in the same system were used. Gels were fixed and stained with Coomassie blue or fixed with 50% methanol, 12% acetic acid, fluorographed with Amplify (Amersham), and autoradiographed using an x-ray film (Konica, Tokyo, Japan, or Amersham (P-Max), Braunschweig, West Germany).
Glycoproteins were stained using Schiffs reagent (Sigma) following the procedure described by . The mixtures were irradiated at 4 "C for up to 15 min with a short-wave UV light (254 nm) from a 44-watt mercury lamp from a distance of 2 cm as described previously (21,22).
Reactions were stopped by adding 1 volume of lysis buffer (0.015 M Tris-HCl, pH 6,s; 8 M urea; 1% @-mercaptoethanol; 1% SDS; 10% glycerol) and incubating the mixture for 5 min at 95 "C. Some reaction mixtures were first incubated with 1 rg of trypsin or protease from Staphylococcus aureus, strain V8 (both from Sigma) for 2 h at 25 "C and stopped thereafter. Determination of 4'-Phosphopantethein-Cyclosporin synthetase fractions (540 ~1 each) from a glycerol gradient were hydrolyzed with 1 N KOH for 1 h at 100 "C, incubated after adjusting a pH of 8 successively for 2 h at 37 "C and overnight at 4 "C with or for control without 1.3 units of bovine alkaline phosphatase (Sigma P-2276). Pantothenic acid liberated from enzyme-bound 4'-phosphopantetheine was determined microbiologically using L. plantarum (DSM 20205) as the test organism as described previously (8,23).

RESULTS
Purification of Cyclosporin Synthetase-Cyclosporin synthetase from B. niuea, strain 7939145, was purified 72-fold. The purification protocol is presented in Table I. At any step of the purification procedure, the enzyme could be stored at -80 "C for over 12 months without loss of activity. Preparation of the crude extract and precipitation with polyethyleneimine and with (NH&SO, were achieved as described in Ref. 14. The redissolved ammonium sulfate precipitation material was separated by gel filtration on a Fractogel HW-55 (F) column; the activity resided in a single peak (Fig. 2).
Due to the high molecular weight of the enzyme, further purification could be achieved by glycerol gradient ultracentrifugation (Fig. 3). Examination of the different purification steps by SDS-polyacrylamide gradient gel electrophoresis shows that cyclosporin synthetase is the major protein after the ultracentrifugation step (Fig. 4A).
Purification of the cyclosporin synthetase activity was followed by measuring the cyclosporin A synthesis rate as described under "Materials and Methods." Subsequently TLC and autoradiography were performed to confirm the cyclosporin A production (Fig. 4B). 7 ml of pooled active fractions of Fractogel chromatography were loaded onto a glycerol gradient from 50 to 25% glycerol in buffer C and centrifuged as described under "Materials and Methods." 2-ml fractions were collected; protein content (0---0) and cyclosporin synthetase activity (A-A,) of each fraction were measured.
Interestingly cyclosporin synthetase activity could also be isolated from spores of B. niuea using the extraction procedure described above.
Molecular Mass Determinations-Measurements of the molecular weight of the native cyclosporin synthetase were performed by ultracentrifugation in glycerol gradients along with standard proteins. A molecular mass of about 800 kDa was obtained. The apparent molecular weight of denatured enzyme was determined by SDS-polyacrylamide gel electrophoresis (3% gels); extrapolation of molecular masses of calibration proteins results in a molecular mass between 600 and 700 kDa for cyclosporin synthetase (Fig. 5). When protein fragments yielded from trypsin digestion of cyclosporin synthetase were separated in SDS-gradient gels, addition of their molecular masses gave a value of about 750 kDa, which is in good agreement with the findings described above. A, cyclosporin synthetase purification was followed by electrophoresis in 15-2% Laemmli polyacrylamide SDS gels. 1 ml of each of the first three purification steps was desalted by passage through PD-10 columns (Pharmacia, Freiburg, West Germany) and prepared for gel electrophoresis (see "Materials and Methods"). Lane I, 25 ~1 of crude extract; lane 2, 25 ~1 of extract after precipitation with 0.3% polyethyleneimine; lane 3, 5 ~1 of 30-50% saturated (NH,),SO, precipitation; lane 4, 25 ~1 of pooled active fractions of Fractogel HW-55 chromatography; lane 5, 25 @I of pooled active fractions of glycerol gradient ultracentrifugation.
The gel was stained with Coomassie blue; the position of cyclosporin synthetase (CySyn) is indicated. R, 100 11 each of the enzyme preparations from the purification steps mentioned in A were tested for cyclosporin synthetase activity as described in Ref. 14. The autoradiogram of the TLC separation of the ethyl acetate extracts is shown. CyA, cyclosporin A. logue D-amino acid (14); the nonchiral amino acid glycine can also substitute D-alanine),* ATP, Mg'+, and AdoMet as the methyl donor (13,14). Mg" ions can be substituted by Mn'+ ions. However, the rate of cyclosporin A formation decreases to about 50% compared to the reaction with Mg". As shown in Ref. 14, the optimal temperature for in uitro cyclosporin A synthesis is 24 "C, the reaction proceeds linearly for at least 15 min under substrate saturating conditions. Cyclosporin A synthesis is inhibited by the reaction products AMP, PP, (not by P,) and S-adenosyl-L-homocysteine, but not by cyclosporin A itself. The reaction proceeds optimally at pH 7.5, measured in Hepes buffer, which is the best buffer for in oitro cyclosporin A synthesis, followed by MOPS, Tris, TES, and, at a synthesizing enzyme, the cyclosporin synthetase described here catalyzes ATP-pyrophosphate exchange reactions dependent on all constitutive amino acids of cyclosporin A in their unmethylated form, whereas the N-methyl amino acids are not activated by the enzyme. Furthermore all amino acids required for cyclosporin C synthesis could be shown to be bound covalently as thioesters to the enzyme (Fig. 6). Cyclosporin C, in which 2aminobutyric acid in position 2 is replaced by threonine (= [Th?]cyclosporin A) was selected for these experiments because 2-aminobutyric acid was not commercially available in a W-labeled form. The same holds true for Bmt; covalent binding of this compound to the enzyme was measured indirectly by formation of [N-methyl-'*C]MeBmt using S-adenosyl[methyl-'"Clmethionine and unlabeled Bmt. With all "C-labeled amino acids used in Fig. 6, it was also possible to label cyclosporin synthetase specifically as analyzed in polyacrylamide gradient gels (not shown).

Photoaffinity
Labeling of Cyclosporin Synthetase-We were interested to clarify, whether the methyltransferase activiti is an integral part of the cyclosporin synthetase molecule or whether it is an associated but different enzyme. For this purpose we used a method for site-specific affinity labeling of methyltransferases (21), which has been previously helpful to demonstrate that the methyltransferase activity of enniatin synthetase is an integral part of the enzyme (22). By irradiation with short-wave UV light in the presence of AdoMet labeled in the methyl group various methyltransfer- were tested for their capacity to bind the constitutive amino acids of cyclosporin C as thioesters. The W-labeled amino acids were incubated together with ATP, MgCl?, and cyclosporin synthetase as described under "Materials and Methods." The protein was precipitated with 7% trichloroacetic acid, and the protein-bound radioactivity was measured. Values were corrected with results from incubations without ATP resp. Bmt. The peak fraction (when measured for in uitro synthesis of cyclosporin A) was fraction 8.
ases could be labeled covalently. Like these enzymes cyclosporin synthetase, too, was labeled when irradiated in the presence of [methyl-"'C]AdoMet or [methyl-"H]AdoMet (Fig.  7). Irradiation of the enzyme in the presence of [carboxyl-'4C] AdoMet or [U-'%]ATP did not give any labeling (not shown). When the affinity-labeled enzyme was digested either with trypsin or with S. aureus V8 protease, three radiolabeled protein bands arose (Fig. 7), suggesting the presence of more than one methyltransferase activity per cyclosporin synthetase molecule, probably three; work is in progress to determine the exact stoichiometry.

Immunological
Examinations-At the ultracentrifugation stage of purification only the major protein band and some minor bands of the preparation running just a little faster in the gel show a positive reaction with a polyclonal rabbit antiserum specifically directed against cyclosporin synthetase." These minor bands seem to represent degradation products of the enzyme, for they show a behavior very similar to cyclosporin synthetase. It is not yet clear whether they originate proteolytically or mechanically from degradation; nevertheless their concentration does not increase when preparations are standing at 4 "C, indicating the absence of proteases in the preparations.
The cyclosporin synthetase band cross-reacts in immunoblots with a polyclonal antiserum directed against enniatin synthetase as well as with the monoclonal antibodies against enniatin synthetase described in Ref. 24. With the latter antibodies, the strongest reactions could be detected with monoclonal antibodies 21.1 and 25.91, which inhibit the thioester formation with valine and recognize the denatured form of enniatin synthetase.
Furthermore, cyclosporin synthetase preparations show a very significant cross-reaction with a polyclonal antibody preparation directed specifically against pantetheine in enzyme-linked immunosorbent assay,4 suggesting the presence of 4'-phosphopantetheine as a prosthetic group similar to a ' A. Lawen  in Cyclosporin Synthetase-To confirm the assumption that 4'-phosphopantetheine forms part of cyclosporin synthetase, we performed a microbiological assay with Lactobacillu.s as a test organism. Fractions from the glycerol gradient ultracentrifugation step were analyzed in order to determine their 4'-phosphopantetheine content. As shown in Fig. 8 the synthetic activity of cyclosporin synthetase comigrates with panthotenate in the gradient. The fact, that most of the panthotenate was released after alkaline phosphatase treatment proves that it is present as 4'-phosphopantetheine in the enzyme. In addition the SDS-polyacrylamide gel electrophoresis separation shows the typical band of cyclosporin synthetase comigrating with pantothenate release and synthetic activity (Fig. 8). Further evidence for the presence of 4'-phosphopantetheine in cyclosporin synthetase was obtained from specific labeling of the enzyme by in vivo feeding of tritiated /3-alanine, which was analyzed by polyacrylamide gel electrophoresis and adjacent autoradiography (not shown).

DISCUSSION
The first attempts to establish the cell-free synthesis of cyclosporin were not successful, but led to an enzyme enrichment actively synthesizing the diketopiperazine cycle-(D-Ala-MeLeu) (12), which represents a partial sequence of cyclosporin A. Change of the cyclosporin producer strain and the buffer for enzyme preparation (Tris buffer instead of phosphate, glycerol content) resulted in successful in uitro synthesis of cyclosporin (13 Cyclosporin synthetase fractions from a glycerol gradient were hydrolyzed with 1 N KOH, incubated without (0) or with (0) alkaline phosphatase and tested in an microbiological assay with L. plantarum as described under "Materials and Methods." The pantothenate release comigrates with the in vitro synthesis of cyclesporin A (X). The Coomassie blue-stained Laemmli gel (15-2% polyacrylamide) separation of the gradient fractions is shown above; molecular masses of the standard proteins are indicated. previously used phosphate buffer; the presence of glycerol in the buffer is necessary as a stabilizer. We think that in the absence of glycerol some conformational changes of the enzyme take place which lead to the loss of its ability to produce cyclosporins.
The main reaction product of such "inactive" enzyme preparations is the diketopiperazine cyclo-(D-Ala-MeLeu)." Therefore it seems obvious that our previous preparations described in Ref. 12 contained intact but "inactive," probably conformationally changed, cyclosporin synthetase polypeptide chains. Like enniatin synthetase (8,22) which can be considered as a model system for other N-methylating peptide synthetases cyclosporin synthetase accepts only the unmethylated precursor amino acids of cyclosporins which are methylated while bound to the enzyme as thioesters as previously shown (Ref. 12). The methyltransferase(s) responsible for these Nmethylations is integral part of the enzyme as could be shown by the affinity-labeling experiments with [methyl-"'Cl AdoMet.
It is interesting that all peptide and depsipeptide synthetases from fungi (e.g. enniatin (8), beauvericin (ll), b-(L-cyaminoadipyl)-L-cysteinyl-D-valine (26), ergot peptide lactam," and cyclosporin synthetase) do not exhibit subunit structure. They consist of single polypeptide chains of molecular masses between 250 and 800 kDa, which harbor all catalytic activities necessary for peptide formation. Such enzymes are designated as "multienzyme polypeptides" in the nomenclature according to NC-IUB (27), in contrast to the "multienzyme complexes" from prokaryotes which consist of subunits (e.g. gramicidin, tyrocidin, bacitracin synthetase, for review see Ref. 16).
Experiments to determine the exact number of N-methyltransferase(s) and 4'-phosphopantheteine residues per mole of cyclosporin synthetase have been hampered by two difficulties. First, to measure exactly the absolute protein content of our preparations, because the dye-binding method we used is related to the calibration protein (bovine serum albumin in our case). Attempts to determine the protein amount gravimetrically were not successful, as we believe, due to different glycerol quantities remaining associated to the enzyme. Second, we never know the exact quantity of inactivated enzyme in our preparations. As can be seen from Table I during the last purification step, a considerable loss of specific activity is observed.
The molecular mass of cyclosporin synthetase has been determined to be between 650 and 800 kDa. In spite of this high value, we were not able to dissociate the enzyme into subunits; neither with urea nor with detergents like SDS nor with P-mercaptoethanol.
This high molecular mass is not astonishing if one realizes that there are (in the case of cyclosporin A) 7 amino acids which have to be N-methylated and in total 11 amino acids which have to be activated and combined. The overall reaction of cyclosporin synthesis can be divided in at least 40 partial reaction steps: 11 aminoadenylation reactions, 11 transthiolation reactions, 7 N-methylation reactions, 10 elongation reactions, and the final cyclization reaction); possible other transthiolation reactions from one thiol group to another are not included in this calculation, The measured molecular mass is in good agreement with a theory of Lipmann and co-workers (28), which requires a protein domain of 70 kDa for each activation site in a peptide synthetase; so, in the light of this assumption one would expect a molecular mass of 770 kDa for cyclosporin synthetase. In summary, cyclosporin synthetase appears to be the largest and most complex enzymatically active multienzyme polypeptide chain so far described and is a further example of a N-methylating peptide synthetase from eukaryotes.