Mapping and Molecular Modeling of S -Adenosyl- L -methionine Binding Sites in N -Methyltransferase Domains of the Multifunctional Polypeptide Cyclosporin Synthetase*

We employed a highly specific photoaffinity labeling procedure, using 14 C-labeled S -adenosyl- L -methionine (AdoMet) to define the chemical structure of the AdoMet binding centers on cyclosporin synthetase (CySyn). Tryptic digestion of CySyn photolabeled with either [ methyl - 14 C]AdoMet or [ carboxyl - 14 C]AdoMet yielded the sequence H 2 N-Asn-Asp-Gly-Leu-Glu-Ser-Tyr-Val-Gly-Ile- Glu-Pro-Ser-Arg-COOH (residues 10644–10657), situated within the N -methyltransferase domain of module 8 of CySyn. Radiosequencing detected Glu 10654 and Pro 10655 as the major sites of derivatization. [ carboxyl - 14 C]AdoMet in addition labeled Tyr 10650 . Chymotryptic digestion generated the radiolabeled peptide H 2 N-Ile-Gly-Leu-Glu-Pro-Ser-Gln-Ser-Ala-Val-Gln-Phe-COOH,correspondingtoaminoacids2125–2136ofthe N -meth-yltransferase domain of module 2. The radiolabeled amino acids were identified as Glu 2128 and Pro 2129 , which are equivalent in position and function to the modified residues identified with tryptic digestions in module 8. Homology modeling of the N -methyltrans-ferase domains indicates that these regions conserve the consensus modifications to the protocol. 25 g of lyophilized mycelium was ground to a fine powder and extracted with 500 ml of buffer A (200 ml of 1 M Tris-HCl, pH 7.8; 150 ml of 2 M KCl; 40 ml of 0.1 M EDTA, pH 7, 10 m M dithioerythritol, 40% (w/v) glycerol) and centrifuged for 10 min at 10,000 rpm and 4 °C. Nuclear material was precipitated from the su- pernatant with 0.3% polyethyleneimine and pelleted by centrifugation for 10 min at 10,000 rpm. The protein precipitated between 30 and 50% saturated (NH 4 ) 2 SO 4 was collected by centrifugation and resolubilized in buffer B (100 ml of 1 M Tris-HCl, pH 7.8, 40 ml 0.1 M EDTA, pH 7, 15% (w/v) glycerol, 4 m M dithioerythritol). The protein suspension was resolved by gel filtration on Fractogel EMD BioSEC, pore size 0.2–0.4 mm (Merck). Active fractions were pooled and concentrated. Aliquots (1 ml) of the concentrate were loaded onto a 50–25% (w/v) stabilized glycerol gradient in buffer C (0.1 m M Tris-HCl, pH 7.8, 50 m M KCl, 4 m M EDTA, 4 m M dithioerythritol). The samples were subjected to ultracen-trifugation, first for 6 h at10,000 rpm and subsequently for an addi- tional 48 h at 22,000 rpm. Active fractions are pooled and once again subjected to glycerol gradient ultracentrifugation, yielding fractions of CySyn of electrophoretic purity (Coomassie Brilliant Blue G-250 staining). thin (HPTLC) of bovine serum albumin as a standard fraction V, globulin-free; Sigma). Sequence Alignment and Homology Analysis— The sequence alignment of the seven N -methyltransferase domains of CySyn was per- formed using ClustalW The protein sequence of CySyn (accession number Z28383) was retrieved from the EMBL nucleotide/protein se- quence data base. Model Generation and Molecular Docking— A structural model of the AdoMet binding site of the N -MTase domain of module 8 ( N -MET M8) of CySyn was constructed using the experimental structure of catechol- O -methyltransferase (RCSB Protein Data Bank www.rcsb.org/pdb; PDB ID-1VID) as the modeling template. The model was constructed using the Swiss-Model and Deep-Viewer environment (36–38). The quality of the model was evaluated with the internal WHATIF-Check program High resolution images of the model were generated with the ray-tracer program POV-Ray version 3.01 (by Christopher Cason; available on the World Wide www.povay.org).

Biologically active, low molecular weight peptides of microbial origin have found a niche in medicine, agriculture, and biological research, by virtue of their enormous structural and functional diversity. A veritable plethora of these compounds exhibit structural modifications such as N-methylated peptide bonds. The natural cyclosporins are products of cyclosporin synthetase (CySyn), 1 a very complex high molecular mass (1.7-MDa) single polypeptide chain nonribosomal peptide synthetase (NRPS) (1,2). The major natural product of the enzyme, cyclosporin A, is a cyclic undecapeptide that contains three nonproteogenic amino acids (D-alanine, (4R)-4-[(E)-2-butyl]-4methyl-L-threonine (Bmt), and L-2-aminobutyric acid) and seven N-methylated peptide bonds (3). To date, it remains the drug of choice used in transplantation surgery and in the treatment of autoimmune diseases (4,5). In addition to its immunosuppressive action, cyclosporin A exhibits several other biological activities, including anti-fungal (6), anti-inflammatory (7,8), anti-parasitic (9), and anti-human immunodeficiency viral (10,11) properties.
The nonribosomal biosynthetic assembly of cyclosporins proceeds via a protein template-driven mechanism (12)(13)(14). CySyn catalyzes a total of 40 reactions in the synthesis of the cycloundecapeptide product (1,2,12). CySyn is composed of 11 conserved modules, which are each arranged into functional domains that catalyze the activation, modification, and polymerization of the constituent amino acids of the peptide product (15)(16)(17)(18). The order of these semiautonomous modules is co-linear with the sequence of the peptide product (14,18). The fundamental module unit consists of catalytic domains responsible for substrate amino acid activation (A-domain), thiolation (T-domain, synonymous with peptidyl carrier protein), and condensation (C-domain) (19,20). Together, these domains form the functional NRPS repeating unit, which represents the minimal domain set necessary for peptide elongation. Often these modules carry an additional modification domain of ϳ430 amino acids between the A-and T-domains. Functional and sequence analyses have indicated that this region represents an N-methyltransferase (N-MTase) function in CySyn (1,12,18,21). In the process of nonribosomal peptide assembly, activation of the substrate amino acid occurs at the A-domain through the generation of a transient aminoacyl adenylate that is in turn covalently attached via a thioester linkage to the thiol of the 4Ј-phosphopantetheinyl prosthetic cofactor of the T-domain in the respective module (18,22,23). The peptide is elongated in a linear NH 2 -to COOH-terminal direction with successive rounds of trans-thiolations and transpeptidations between each activated amino acid facilitated by the interplay of 4Ј-phosphopantetheinyl internal carriers. This sequence of events is repeated in a stepwise fashion until the consummation of peptide assembly. In the case of cyclosporin synthesis, release of the full-length peptide ensues via cycliza-tion. Prokaryotic NRPSs usually employ a thioesterase domain (TE-domain) situated at the extreme COOH terminus of the last module (24,25), whereas most fungal NRPSs, such as CySyn, utilize a unique COOH-terminal C-domain for cyclization and release of the mature peptide (26).
CySyn contains four basic C-A-T modules (modules 1, 6, 9, and 11); the other seven modules (modules 2, 3, 4, 5, 7, 8, and 10) display the additional N-MTase domain insert between the A-and T-domains ( Fig. 1) (18). Whereas analysis of the mechanism of action of the domains forming the catalytic triad has progressed well during recent years, characterization of the modification domains, and especially the N-MTase domain, is still in its infancy. Methyltransferases (MTases) are enzymes that catalyze the transfer of the methyl group from the ubiquitous cofactor S-adenosyl-L-methionine (AdoMet) to nitrogen, oxygen, sulfur, or carbon atoms of a range of small molecules and macromolecular species (27). These enzymes can be divided into families according to four major substrate classes, protein, RNA, DNA, and small molecule (27). The N-MTase domains of NRPS catalyze the bimolecular transfer of the Smethyl group of AdoMet to the ␣-nitrogen of the thioesterified amino acid, releasing S-adenosyl-L-homocysteine as a reaction product ( Fig. 1) (1,12). Accordingly, these N-MTase domains can be assigned as small molecule MTases.
Most AdoMet-dependent MTases have a bilobial structure and are organized into an AdoMet binding domain, which provides cofactor binding contacts and catalytic residues, and a second domain mainly responsible for conferring substrate specificity (27). Despite the poor overall sequence identity across AdoMet-dependent MTases and varied substrate specificity, these enzymes display a structurally conserved cofactorbinding domain termed the "AdoMet-binding fold" (27)(28)(29). Moreover, a structural conservation of the mode of cofactor binding is observed across most structurally defined members of this superfamily; this has led to the delineation of a set of consensus AdoMet binding interactions (28). In most AdoMetdependent MTases, the AdoMet binding region is constructed from a set of noncontiguous sequence elements. The low overall level of sequence identity across MTases has complicated the assignment of specific motifs to AdoMet binding. Kagan and Clarke (30) defined a set of three conserved sequence motifs, which are apparent in most non-nucleic acid MTases (Fig. 2). Subsequently, a structure-guided sequence alignment of several structurally characterized AdoMet-dependent MTases revealed, motifs I and II participate directly in cofactor binding, whereas motif III serves a structural role, forming the core of ␤-strand 5 of the AdoMet-binding fold (Fig. 2) (27). More importantly, this comparison indicated that the key residues governing the interactions with the cofactor molecule are localized within four motifs, I-IV (27). In this paper, we shall adopt the AdoMet binding motif nomenclature as defined by Faunam et al. (27). In most MTases, these motifs occur in the same linear order and are separated by similar intervals in the polypeptide (27,30,31).
Although much progress has been made toward the understanding of the catalytic mechanism of various protein and DNA methyltransferases, the assignment of unambiguous, specific amino acid sequence motifs to AdoMet binding and catalysis of N-methylation have not been established for NRPS. We have previously demonstrated that [methyl-14 C]AdoMet can be utilized to directly photolabel the N-MTase domains of CySyn in a highly specific manner. 2 In this study, stochiometric photoaffinity labeling demonstrated that CySyn is photolabeled with AdoMet in a molar ratio of ϳ7:1 (AdoMet/CySyn), which is in good agreement with the seven N-MTase domains identified in the CySyn cDNA sequence. The specificity of photolabeling was demonstrated by competitive displacement by unlabeled AdoMet and its congeners S-adenosyl-L-ethionine, S-adenosyl-L-homocysteine, and sinefungin. Photolabeling was only tenable with native CySyn and was not detected with denatured enzyme or with proteins, which do not bind AdoMet. The photolabeling of the AdoMet binding sites displayed homotropic negative cooperativity, characterized by a curvilinear Scatchard plot with upward concavity. Here we have employed a combination of classical photoaffinity labeling and peptide mapping techniques to define the specific amino acid sequences involved in the cofactor interaction. We also present a three-dimensional homology model of the cofactor binding interaction with the N-MTase domains of CySyn. The proposed bimolecular interaction model provides a satisfactory structure-function rationale for the photoaffinity labeling data and conforms agreeably to homologous cofactor interactions defined from structurally characterized AdoMet-dependent MTases. To our knowledge, this is the first report of a structure-related characterization of the N-MTase domains of a NRPS. Due to the high degree of sequence identity conserved across NRPS, the information garnered from this study can be generalized to all NRPS family members expressing an N-MTase function. Growth of Organism and Enzyme Preparation-Tolypocladium inflatum Syn. Beauveria nivea, strain 7939/45 was donated by Dr. M. Dreyfuss (Novartis) and cultured as previously described (32). CySyn was purified to near homogeneity as described earlier (1)   A module (1100 -1500 aa) consists of all of the functional domains required to incorporate a residue into the peptide chain. All CySyn modules contain domains responsible for the activation (A; ϳ550 aa), thiolation (T; ϳ100 aa), and condensation (C; ϳ450 aa) reactions of peptide assembly. In addition to this essential catalytic domain triad, N-methylating modules exhibit an N-MTase (M; ϳ420 aa) inserted between the A-and T-domains responsible for the transfer of the methyl moiety of AdoMet to the cognate amino acid of the module, whereas it is covalently tethered to the respective modular 4Ј-phosphopantetheinyl prosthetic group.

Materials-S-Adenosyl-L-[methyl-
modifications to the protocol. 25 g of lyophilized mycelium was ground to a fine powder and extracted with 500 ml of buffer A (200 ml of 1 M Tris-HCl, pH 7.8; 150 ml of 2 M KCl; 40 ml of 0.1 M EDTA, pH 7, 10 mM dithioerythritol, 40% (w/v) glycerol) and centrifuged for 10 min at 10,000 rpm and 4°C. Nuclear material was precipitated from the supernatant with 0.3% polyethyleneimine and pelleted by centrifugation for 10 min at 10,000 rpm. The protein precipitated between 30 and 50% saturated (NH 4 ) 2 SO 4 was collected by centrifugation and resolubilized in buffer B (100 ml of 1 M Tris-HCl, pH 7.8, 40 ml 0.1 M EDTA, pH 7, 15% (w/v) glycerol, 4 mM dithioerythritol). The protein suspension was resolved by gel filtration on Fractogel EMD BioSEC, pore size 0.2-0.4 mm (Merck). Active fractions were pooled and concentrated. Aliquots (1 ml) of the concentrate were loaded onto a 50 -25% (w/v) stabilized glycerol gradient in buffer C (0.1 mM Tris-HCl, pH 7.8, 50 mM KCl, 4 mM EDTA, 4 mM dithioerythritol). The samples were subjected to ultracentrifugation, first for 6 h at 10,000 rpm and subsequently for an additional 48 h at 22,000 rpm. Active fractions are pooled and once again subjected to glycerol gradient ultracentrifugation, yielding fractions of CySyn of electrophoretic purity (Coomassie Brilliant Blue G-250 staining).
Thin Layer Chromatography-Two-dimensional high performance thin layer chromatography (HPTLC) of peptide digests was performed according to Pataki (33). For developing the chromatograms, butanol/ acetic acid/water (4:1:1) (solvent A) was used in the first dimension, and phenol/water (75:25) (solvent B) was used in the second dimension (33). Amino acid standards were detected with ninhydrin reagent (0.3% in acetone); for development, the chromatogram was sprayed and heated at 60°C for 20 min. The radiolabeled material was visualized by autoradiography, and the developed film was in turn superimposed onto the ninhydrin-stained HPTLC sheet to allow for correlation with the position of amino acid standards. Where specified, material was scraped from the plates and analyzed by electrospray ionization-mass spectrometry (ESIMS  beled protein was first partially digested by incubation with sequencing grade modified trypsin or chymotrypsin (Sigma) at a protease/substrate ratio of 1:125 (w/w) at 37°C for 1.5 h. The reaction was terminated by the addition of a protease inhibitor mixture (Sigma catalog no. P2714) and an equal volume of SDS-PAGE sample buffer. The resultant fragments were resolved by SDS-PAGE on a 12% polyacrylamide gel. The gel was dried, and photolabeled fragments were visualized by phosphorimaging. The apparent molecular masses of the photolabeled fragments were determined by comparison with the mobility of molecular mass standards (Bio-Rad). Prominent photolabeled bands generated from the partial tryptic and chymotryptic digests of ϳ56 and ϳ60 kDa, respectively, were excised from dried gels using the phosphor image as a template. They were rehydrated with 300 l of elution buffer (0.1 M NH 4 HCO 3 , 0.1% SDS) and finely crushed. The labeled protein was passively eluted from the macerated gel pieces into 5 ml of elution buffer for 3 days with agitation. The eluent was separated from the gel by centrifugation to prevent the protein from reentering the gel. The elution was repeated three times to maximize the recovery of protein from the gel. The eluents were filtered with a 0.45-m syringe filter (Pall Gelman Sciences, Sydney, New South Wales, Australia) and combined, and the protein was precipitated with 5 volumes of acetone at Ϫ20°C overnight to concentrate and remove excess SDS. The protein precipitate was pelleted by centrifugation at 13,000 ϫ g for 15 min, and the supernatant was discarded. The pellet was resuspended in digestion buffer (0.1 M NH 4 HCO 3 , 2 M urea, pH 8.0). To ensure complete reaction, the radiolabeled protein fragment was reduced by incubation with 45 mM dithiothreitol at 55°C for 30 min, the mixture was then cooled to room temperature, and iodoacetamide was added to a final concentration of 100 mM. The mixture was then incubated for 30 min in the dark to alkylate the cysteinyl thiol groups. The protein fragment was then exhaustively digested with either trypsin or chymotrypsin 1:25 (w/w) for 2 days at 37°C. After 4 and 20 h of incubation, the initial amount of protease was added again to the reaction. Preparative purification of the photolabeled tryptic peptides was achieved by several successive reverse phase-high performance liquid chromatography (HPLC) runs. The digested sample was dried, resuspended in 0.05% trifluoroacetic acid in water, and subjected to reverse phase HPLC on a Zorbax 300-Å C 18 column (3.5 m; 4.6 mm ϫ 250 mm; Agilent Technologies, Melbourne, Victoria, Australia) using 0.05% trifluoroacetic acid in water (solvent C) and 0.05% trifluoroacetic acid in acetonitrile (solvent D). The peptides were eluted by a linear gradient from 0 to 80% solvent D in 80 min at a constant flow rate of 0.75 ml/min. The eluent was monitored at 214 nm, fractions were collected, and aliquots of each fraction were assayed for radioactivity by liquid scintillation counting. The radioactive peak fractions were dried and reconstituted in 0.05% trifluoroacetic acid in water and rechromatographed. The purified radioactive peptide fractions from successive runs were pooled, concentrated, and subjected to NH 2 -terminal sequence analysis.
Purified [ 14 C]AdoMet-modified peptides were also digested with pronase in 100 mM triethanolammonium acetate buffer, pH 6.8, for 12 h at 37°C. The mixture was then boiled, eluted through a Sephadex G-25 M column, and concentrated. The concentrate was redissolved and digested for another 12 h at 25°C with carboxypeptidase Y (400 g/ml) in 100 mM triethanolammonium acetate buffer, pH 6.0. The digestion mixture was then boiled, eluted through a Sephadex G-25 M column, and concentrated. The concentrate was subjected to another round of proteolysis with proteinase K in 10 mM Tris-HCl buffer, pH 7.5, for 12 h at 37°C. The digest was eluted through a Sephadex G-25 M column, dried and redissolved in 60 l of water, and resolved by HPTLC.
Sequence Analysis-Amino-terminal sequencing was performed on a Procise TM (Applied Biosystems, Melbourne, Victoria, Australia) protein sequencer. Sequencing was performed at the Australian Proteome Analysis Facility. The dissolved peptide sample was applied onto a precycled Biobrene-treated trifluoroacetic acid glass fiber disk (Applied Biosystems) and subjected to cycles of Edman degradation following a modified pulse-liquid protocol. Phenylthiohydantoin (PTH)-derivative mass values were derived from a correlation of chromatographic peak heights of the PTH-derivatives released with each cycle with a standard 10-pmol mixture of the 20 proteogenic amino acids (Applied Biosystems). The theoretical initial yield and repetitive yield were calculated from linear least squares analysis of the semilogarithmic plot of the quantity (pmol) of PTH-derivative released with each cycle versus the cycle number. The low yielding residues Ser, Thr, and Trp were omitted from the analysis.
An aliquot of the PTH-derivatives released from each cycle was counted for 14 C radioactivity; the remainder of the material was injected into the on-line HPLC PTH-derivative analysis system for identification.
ESIMS-Mass values of purified photolabeled proteolytic peptides and modified amino acids were determined on a Micromass (Manchester, UK) Platform II liquid chromatography/mass spectrometry system.
Protein Assay-Protein was determined by the dye-binding method of Bradford (34) using bovine serum albumin as a standard (prepared from fraction V, globulin-free; Sigma).
Sequence Alignment and Homology Analysis-The sequence alignment of the seven N-methyltransferase domains of CySyn was performed using ClustalW (35). The protein sequence of CySyn (accession number Z28383) was retrieved from the EMBL nucleotide/protein sequence data base.
Model Generation and Molecular Docking-A structural model of the AdoMet binding site of the N-MTase domain of module 8 (N-MET M8) of CySyn was constructed using the experimental structure of catechol-O-methyltransferase (RCSB Protein Data Bank www.rcsb.org/pdb; PDB ID-1VID) as the modeling template. The model was constructed using the Swiss-Model and Deep-Viewer environment (36 -38). The quality of the model was evaluated with the internal WHATIF-Check program (39,40). High resolution images of the model were generated with the ray-tracer program POV-Ray version 3.01 (by Christopher Cason; available on the World Wide Web at www.povay.org).

Identification of the AdoMet Binding
Region-In order to define the amino acid sites at which AdoMet interacts with the N-MTase domains, samples of CySyn were photolabeled with either [methyl-14 C]AdoMet or [carboxyl-14 C]AdoMet on a preparative scale. The sites of photoinsertion were localized by sequential peptide mapping and Edman sequencing of isolated radioactive peptides. The high specificity of the photolabeling reaction has been demonstrated elsewhere. 2 Initially, partial proteolytic digestion of the photolabeled enzyme with either trypsin or chymotrypsin was performed. The resultant fragments were resolved by SDS-PAGE and visualized by phosphorimaging. Limited tryptic digestion generated a series of photolabeled fragments migrating with apparent molecular masses ranging from 35 to 220 kDa (Fig. 3a). The peptide corresponding to the 56-kDa radiolabeled band was excised, eluted, and digested to completion with trypsin. The resultant peptides were resolved by reverse phase HPLC as described under "Experimental Procedures." It was previously reported that the decomposition of ligand-protein photoadducts during reverse phase HPLC is somewhat dependent on the flow rate of separation (41). To circumvent this shortcoming, a slow flow rate (0.75 ml/min) was employed in the reverse phase HPLC separations of the proteolytic digests of photolabeled protein fragments. Aliquots of each collected fraction were analyzed for radioactivity (Fig. 3b). In the two independent preparative scale experiments performed, radioactivity consistently eluted in three peaks (labeled A-C). The major peak, C, corresponded to fractions eluting between 43 and 45 min, whereas the two minor peaks, B and A, eluted at 32-35 min and with the void volume, respectively. Comparable elution profiles were obtained with both [methyl-14 C]AdoMet-and [carboxyl- 14 C]AdoMet-photolabeled samples. The radioactive fractions corresponding to radioactive peak C (indicated by the arrow in Fig. 3a) were pooled and rechromatographed until a single peptide peak was obtained (Fig. 3c). The material from each of the radioactive peaks was subjected to NH 2 -terminal sequence analysis. Radioactive peaks A and B did not yield any sequence information and therefore most likely represent unbound radioligand and/or decomposition products of the photoadduct, which may be unstable under reverse phase HPLC conditions.
Edman sequencing of the purified material from radioactive peak C isolated from either [methyl-14 C]AdoMet-or [carboxyl-14 C]AdoMet-photolabeled samples yielded a single 14-residue sequence, H 2 N-Asn-Asp-Gly-Leu-Glu-Ser-Tyr-Val-Gly-Ile-Glu-Pro-Ser-Arg-COOH. By comparison with the established cDNA sequence of CySyn, the sequence could be unambiguously assigned to residues 10644 -10657, situated in N-MET M8 (Fig.  2). The specific residues modified in the peptide were identified from the release of radioactivity during each cycle of the Edman degradation (Fig. 4, a and b). In both [methyl-14 C]AdoMetand [carboxyl-14 C]AdoMet-photolabeled preparations, a significant radioactivity release was observed with cycles 11-14. The majority of radioactivity release coincides with cycles 11 and 12, consistent with photoincorporation at Pro 10655 and, to a lesser extent, Glu 10654 . A substantial decrease in the yield of the PTH-derivative of the corresponding residues is also observed (Fig. 4, a and b). Furthermore, the output chromatogram from the on-line HPLC PTH-derivative analyzer obtained from cycles 11 and 12 from both samples shows an extraneous peak lagging close to the representative PTH-derivative peak of the Glu and Pro residues (data not shown). Moreover, radioactive peak C, corresponding to the labeled peptide, displays a lagging shoulder; this may represent a distinct photoadduct, which would account for the minor level of photoincorporation detected in the Glu 10654 . In the case of the [carboxyl-14 C]AdoMet-photolabeled sample, a significant release of radioactivity also occurred in cycle 7, corresponding to Tyr 10650 . An attenuated yield of the PTH-derivative of this residue was also observed; in addition, the output HPLC chromatogram of the PTH-derivatives from this cycle displayed an extraneous peak that does not correspond to any of the PTH-derivative standards (data not shown). The radioactivity coinciding with this cycle is significantly less than the radioactivity release observed with cycles 11 (Glu 10654 ) and 12 (Pro 10655 ), suggesting that these sites are preferentially labeled as opposed to the Tyr 10650 position. The lower degree of photoincorporation observed at the Tyr and Glu positions may be due to the adduct at these sites being more labile to the conditions of Edman sequencing or possibly due to a lower inherent propensity of photoinsertion at these sites. These findings indicate that the photoinsertion of [ 14 C]AdoMet is heterogeneous, with modification occurring on at least three disparate sites.
It is also noteworthy that these residues may not represent the exclusive sites of photoincorporation, since a considerable proportion of radioactivity was lost during both reverse phase HPLC and sequencing processes. However, in comparison with the proteolytic peptides derived from [methyl-14 C]AdoMetmodified preparations, far less background radioactivity was released with each cycle from [carboxyl-14 C]AdoMet-modified peptide preparations. The low levels of radioactivity release observed with cycles 13 and 14 can be attributed to carryover from unreacted NH 2 -terminal residue from cycles 11 and 12, which is one of the primary complications associated with the Edman procedure. Taking into consideration the observed loss of radioactivity during reverse phase HPLC and the background release of radioactivity during Edman degradation, it appears that the photolabeled amino acid derivatives are slowly hydrolyzed at the pH conditions of HPLC and Edman sequencing.
The purified peptides were further analyzed by ESIMS in positive ionization mode. The observed mass of each species corresponds to the calculated monoisotopic mass of each nonlabeled peptide species plus the mass of one AdoMet molecule (Table I). Although the nature of the modification remains obscure, these data confirm that the entire AdoMet molecule is photocross-linked to the enzyme.
To confirm the above results, the proteolytic mapping experiment was repeated with chymotrypsin. The different specificity of this protease led to the isolation of a peptide derived from N-MET M2 of CySyn. Mild chymotryptic cleavage of the [ 14 C]AdoMet-photolabeled enzyme generated fragments ranging in molecular mass from 26 to 200 kDa (Fig. 5a). A ϳ60-kDa fragment was isolated and further digested with chymotrypsin. The resultant peptides were separated by reverse phase HPLC. Aliquots of the peak fractions were assayed for radioactivity. Consistent with the results of the tryptic digestion, the elution profile displayed three radioactive peaks (Fig. 5b). The fractions corresponding to the prominent peak (peak C) were pooled from successive runs, concentrated, and purified to homogeneity by rechromatography (Fig. 5c). NH 2 -terminal sequence analysis of peaks A and B did not yield any sequence, The ϳ56-kDa fragment, indicated by the arrow, was eluted from the gel, reduced, alkylated, and exhaustively digested with trypsin. The reverse phase HPLC fractionation of the resultant tryptic peptides was performed as described under "Experimental Procedures." Peptides were eluted at a flow rate of 0.75 ml/min, and the UV absorption of the eluent was monitored at 214 nm. Peak fractions were collected, and aliquots were counted for radioactivity. The position of peak C is indicated by the arrow. b, radioactivity profile of the eluent indicated three prominent radioactive peaks labeled A-C. c, radioactive peak fraction C yielded the maximal [methyl-14 C]AdoMet-associated radioactivity and was subjected to further treatment. Inset, peak fractions corresponding to peak C were pooled, concentrated, and rechromatographed.
The radioactivity profile of each peptide indicates that the radioactivity release occurs primarily during cycle 5, corresponding to Pro 2129 , and to a lesser extent in cycle 4, corre-sponding to Glu 2128 (Fig. 4, c and d). A concomitant decrease in the yields of the PTH-derivatives of these residues is also observed (Fig. 4, c and d). Moreover, the HPLC profiles for these residues displayed an extraneous peak not corresponding to any of the standard amino acids (data not shown). These data indicate Pro 2129 as the primary and Glu 2128 as the secondary site of [ 14 C]AdoMet photoincorporation on the peptide, suggesting that these residues form an intimate interaction with the cofactor. The results confirmed the photolabeling data obtained from tryptic peptides and showed that photolabeling occurs at the same residue positions in different N-MTase domains of CySyn.
ESIMS analysis of the purified peptides indicated that the mass of each species corresponds to the calculated monoisotopic mass of each nonlabeled peptide species plus the mass of one AdoMet molecule (Table I).
Two-dimensional HPTLC Analysis-To confirm the sites of photoinsertion, the photolabeled peptides were digested to composite amino acids with a combination of proteases and analyzed by two-dimensional HPTLC and ESIMS. The digest of the [methyl-14 C]AdoMet-photolabeled tryptic peptide yielded two major radioactive spots, whereas the digest of the [carboxyl-14 C]AdoMet-photolabeled tryptic peptide yielded three radioactive species. The migration of these compounds did not correlate with any of the 20 proteogenic amino acid standards (data not shown). The radioactive compounds were extracted from the plates, rechromatographed, and analyzed by ESIMS. The observed masses of the radioactive material from the three spots derived from tryptic digestion of the [carboxyl-

FIG. 4. Radioactivity released during sequential Edman degradation of the purified 14 C-AdoMet-modified proteolytic peptides.
The proteolytic peptides were immobilized onto glass fiber filters and subjected to cycles of Edman degradation. An aliquot of each PTH-derivative was sampled and assayed for radioactivity (gray bars); the remaining material was injected into the on-line HPLC system for identification and quantification of amino acid yield (black bars). The identity of each amino acid is denoted on the abscissa. The same amino acid sequence and pattern of radioactivity release were observed in two independent experiments. a, [methyl- 14 (Table II). The amino acids Asp and Pro were also obtained from the chymotryptic peptides. These data are in conformity with the radiosequencing results and further substantiate the identity of the modified residues Model Generation-The conserved tertiary topology of the AdoMet binding fold and equivalence of the order and spacing of the AdoMet binding motifs allow for the prediction of the structure of the AdoMet binding site of the N-MTase domains of CySyn. The N-MTase domains of CySyn were modeled based on the consensus folding topology of rat catechol-O-methyltransferase (27,42). The homologous architecture of the AdoMet binding region in catechol-O-methyltransferase exhibits a doubly wound open ␣/␤ Rossmann-type fold topology (27,42,43). The three-dimensional model of the N-MET M8 region of CySyn is presented (Fig. 6a); comparable results were obtained with modeling of the other six N-MTase domains of CySyn. The model was generated from the sequence region spanning residues 10,603-10,780 based on a structure-guided alignment of consensus motifs I-IV in the template and target sequence. The root mean square deviation of superimposed C ␣ atoms was 0.5 Å, indicating that the structures are almost indistinguishable; thus, the model fits the experimentally derived template structure very well. The modeled structure accommodates the same secondary structural elements and tertiary positioning of AdoMet binding contact points from consensus regions I-IV. The consensus structural topology consists of a seven-stranded ␤-sheet of mixed strand polarity sand-wiched by six ␣-helices (Fig. 6b). The central ␤-sheet consists of five parallel ␤-strands with a topological organization (5141112131) and two anti-parallel ␤-strands (6172). Three of the helices (A-C) are at the front of the sheet, and three (D-F) are at the back of the sheet, creating an ␣/␤/␣ secondary structural succession. Moreover, the structure shows internal symmetry; the ␤-␣-␤-␣ segment encompassing ␤-strands 1 and 2 and ␣-helices B and C can be superimposed onto that encompassing ␤-strands 4 and 5 and ␣-helices E and F. This secondary structural arrangement creates a topological switch point between ␤-strands 1 and 4 where the strand order reverses, producing a cleft that forms the cofactor binding pocket (27). The walls of the binding pocket are formed by the ␤ to ␣ adjoining loops diverging from the COOH termini of ␤-strands 1-4. Since the structure of catechol-O-methyltransferase was employed as the modeling template, the cofactor molecule was docked into the macromolecular surface in the extended conformation. However, the cofactor is bound in a distorted conformation in some MTases; hence, the extended conformational interaction should not be taken as absolute (44).
The consensus AdoMet binding motifs are present in the N-MTase domains of CySyn; motif I is clearly discernible, whereas the unambiguous assignment of residue blocks to motifs II-IV is not plausible without structural guidance due to the sequence heterogeneity in these regions (Fig. 2). The spatial location of the AdoMet binding motifs in the N-MTase domains of CySyn is also consistent with the spacing of these elements in other small molecule MTases (Fig. 2) (27, 30). On the primary level, motif I is characterized by a glycine-rich stretch; the consensus sequence is represented by GXGXG in non-nucleic acid AdoMet-dependent MTases (30). This motif is highly conserved and is the signature sequence for AdoMet-dependent MTases. In the proposed model, this motif forms the adjoining loop (termed the G-loop (31)) between ␤-strand 1 and The ϳ60-kDa fragment, indicated by the arrow, was treated as described in the legend to Fig. 3. The elution position of the peptide corresponding to the major release of radioactivity is denoted as radioactive peak C and indicated by an arrow (42-43 min). b, radioactivity profile of the eluent; c, radioactive peak fraction C yielded the maximal [methyl-14 C]AdoMet-associated radioactivity and was subjected to further treatment. Inset, peak fractions corresponding to peak C were pooled, concentrated, rechromatographed, and submitted for sequence analysis.
␣-helix B, which accommodates the methionine moiety of AdoMet. This region represents a common structural element of the AdoMet binding cleft of these enzymes (28). The conserved glycines allow for hydrogen bonding of the main chain atoms to the carboxyl and amino groups of the AdoMet methionine (Fig. 7). Within the cleft of the G-loop, the amino group of the AdoMet methionine is hydrogen-bonded by the main chain oxygen of Gly 10631 , whereas the main chain nitrogen of Gly 10635 hydrogen-bonds with the methionine carboxyl.
Motif II is more heterogeneous on the primary level but invariably contains an acidic residue followed by a hydrophobic or aliphatic residue, which is commonly involved in stabilizing the binding of the adenine ring of AdoMet by van der Waals packing in a face-to-face or edge-to-face fashion. In most structurally defined AdoMet-dependent MTases, the highly conserved acidic residue plays a direct role in AdoMet binding by hydrogen-bonding with the hydroxyls of the adenine ribose (27,28,31). In the homology model, motif II forms the loop (termed the D-loop (31)) at the COOH terminus of ␤-strand 2; the conserved acidic residue, represented by the photolabeled Glu 10654 , is centered at the end of ␤-strand 2 and forms hydrogen bonds to the ribose hydroxyls of AdoMet (Figs. 6 and 7). In the predicted structure of N-MET M8, the binding of the adenine ring is stabilized by van der Waals packing in a perpendicular fashion with Pro 10655 adjacent to the conserved Glu 10654 of the D-loop (Fig. 7). Considering that the photolabeled Pro 10655 is in close proximity to the adenine ring, it is possible that the observed modification at this site may have resulted from destabilization of the -bond state of the adenine ring.
Motif III contributes hydrogen-bonding contacts to the N-1 and N-6 atoms of the adenine ring of AdoMet. In the proposed model, motif III is centered at the COOH terminus of ␤-strand 3, the exocyclic amino group (N-6) of the AdoMet adenine ring hydrogen-bonds to the side chain of the conserved Asp 10686 , with its main chain nitrogen hydrogen-bonding to the N-1 atom of adenine (Fig. 7).
Motif IV is located in the loop region at the COOH terminus of ␤-strand 4. In nucleic acid MTases, this region is well conserved (45)(46)(47), whereas in non-nucleic acid AdoMet-dependent MTases the region is poorly conserved and no consensus has been defined. In NRPS N-MTase domains, this region is represented by the conserved sequence block NSV(A/V)QYFP (Fig.  2) (48). The model implies that the conserved Gln 10705 forms hydrogen bonds with the amino and carboxylate of the AdoMet methionine, whereas the adjacent Tyr 10706 and Phe 10707 make edge-to-face van der Waals contacts with opposite sides of the adenine ring (Fig. 7). DISCUSSION Very little is known about the chemical mechanisms of cofactor binding and the steric course of N-methyl transfer in NRPS systems, and to date no structural data have been garnered. We have employed a highly specific photoaffinity labeling procedure in combination with proteolytic mapping and Edman degradation to define the primary structural elements involved in AdoMet binding in the N-MTase domain regions of CySyn. Our data suggest that the cofactor binds to topologically equivalent sites on the seven N-MTase domains of CySyn. Radiosequencing of the isolated photolabeled peptides indicates a heterogeneous photolabeling with mutually exclusive modification at three amino acid positions. This was confirmed by proteolytic digestion of the peptides and ESIMS analysis of the composite residues. The [ 14 C]AdoMet peptides obtained from both tryptic and chymotryptic fragmentations correspond to the consensus motif (Gly-Leu-Xaa-Xaa-Tyr-Val-Gly-Leu-Asp/Glu-Pro), which is conserved across the N-MTase domains of several NRPS systems from both eukaryotic and prokaryotic organisms (48). This region is homologous to the small molecule AdoMet-dependent MTases motif II consensus sequence element (P/G)(Q/T)(F/Y/A)DA(I/V/Y)(F/I)(C/V/L) (30). In an early photoaffinity labeling study of the N-MTase domain of enniatin synthetase (49), which fell short of identifying the labeled amino acid(s), AdoMet photolabeling was proteolytically mapped to a nonapeptide fragment ( 2098 H 2 N-Asn-Leu-Asn-Pro-Gly-Leu-Asn-Ser-Tyr-COOH 2106 ). This peptide is homologous to the photolabeled proteolytic peptides identified in this study. Considering the data together, it is evident that the cofactor interacts with homologous sequence elements within the cofactor-binding site of the N-MTase domains of distinct NRPS systems. Furthermore, both studies are in line with a study of rat guanidinoacetate MTase, which identified a Tyr residue within the AdoMet-binding motif II region of this enzyme as forming the covalent adduct with AdoMet (50,51). The recent resolution of the crystal structure of rat guanidinoacetate MTase revealed that this residue is proximal to the AdoMet binding pocket of this enzyme (52). The primary structure of these peptides also contains homologous residues, identified as the photocross-linking sites in this study. The COOHterminal Tyr 2106 of the peptide from enniatin synthetase corresponds to the modified Tyr 10650 observed in the [carboxyl-14 C]AdoMet-photolabeled tryptic peptide fragment from CySyn. Moreover, mutation of Tyr 2106 to Ala in the recombinant enniatin synthetase N-MTase domain resulted in a pronounced reduction in AdoMet photolabeling (48), suggesting that the side chain of this residue is involved in integral contacts with the cofactor and/or maintenance of the binding site. The presence of several contact points, in addition to Tyr 10650 , identified in this study would account for the inability of this substitution to completely abolish AdoMet binding. Although our experimental results concur with these data, the homology model of the N-MET M8 of CySyn implies that the modified Tyr 10650 serves a structural role and is situated within ␤-strand 2 of the AdoMet binding fold. The observed photoincorporation at the Tyr 10650 position can be construed in terms of the vectorial structural movements initiated with AdoMet photolabeling. 2 It is likely that these conformational transitions move the Tyr 10650 sufficiently close to the ligand to undergo photocrosslinking. Such vectorial interdomain movements are commonplace in multifunctional polypeptides (53,54).
Although it is not represented in the homology model, it is tenable that the aromatic ring of the conserved Tyr 10650 position acts to stabilize cofactor binding via a cation-interaction with the electron-deficient sulfonium center of the AdoMet molecule (Fig. 8a). If the Tyr 10650 side chain does indeed stabilize AdoMet binding by an electrostatic interaction with the sulfonium center, then the polarization of the positively charged sulfonium sulfur would facilitate the methyl transfer reaction (55). Evidence for a similar mechanism has been proposed for the reaction mechanisms of the DNA MTases M.TaqI (56) and M.HhaI (57). This aromatic-cation interaction would also serve to stabilize the possible cationic N-methyl-amino acid reaction intermediate prior to subsequent proton abstraction (Fig. 8b) (56). This postulate is substantiated by evidence that the stabilization of the cationic transition state through cation-interactions accelerates catalysis of biomimetic methylation reactions (58,59). Moreover, this is consistent with the proposed interaction of the AdoMet molecule with the conserved motif IV region, which is rich in aromatic residues (Fig.  2). The involvement of cation-aromatic interactions in ligand binding to proteins has been reported in a number of systems (60 -63). Aromatic residues in proteins are believed to act as attractive cation binding sites by providing stabilizing interactions with the positive charge of ligands conferred through their electron-rich systems.
It is conceivable that the steric hindrance imposed by the modification at the corresponding Tyr position (Tyr 2124 ) in N-MET M2 would inhibit chymotryptic cleavage of the peptide backbone at this site. However the positional heterogeneity of photoinsertion and the very low yield of the Tyr-labeled fragment would account for the isolation of the [ 14 C]AdoMet-labeled H 2 N-Ile-Gly-Leu-Glu-Pro-Ser-Gln-Ser-Ala-Val-Gln-Phe-COOH peptide as the major photolabeled chymotryptic fragment derived from N-MET M2. Therefore, the labeled peptide mixture isolated via reverse phase HPLC predominantly consists of the Glu and Pro derivatized species.
ESIMS analysis of the purified peptides demonstrates that the entire AdoMet molecule is covalently attached to the photoreactive residues as opposed to the composite adenosine or methionine or the transmethylation of the S-methyl group from AdoMet to a specific site on the protein. This result accounts also for the fact that we were able to label CySyn with [carboxyl- 14 C]AdoMet in contrast to our earlier results (1). This difference is likely to be due to a slightly changed methodology employed in two different laboratories. These data also indicate that the observed labeling at the different residue positions on each peptide fragment is mutually exclusive with a stoichiometry of one AdoMet molecule per peptide.
Although our findings demonstrate an exclusive stoichiometry of photolabeling, the region of the AdoMet molecule, which is involved in the formation of the covalent cross-link and the precise photochemistry of the reaction, remains enigmatic. Early studies on the photochemical cross-linking of purines have shown that substitution/addition reactions can occur either across the 1,6 double bond (64) or C-8 position (65, 66) of the adenine moiety. It is tenable to imagine that a comparable photochemical reaction occurs in the case of the photoaffinity cross-linking of [ 14 C]AdoMet to the N-MTase domains of CySyn.
The AdoMet binding domain of MTases exhibits a comparable three-dimensional folding (27)(28)(29). This structural similarity extends to the tertiary positions of the amino acid side chains interacting with the AdoMet molecule. These residue positions are located in motif elements that are highly conserved within a given MTases subfamily but only exhibit marginal sequence identity to the corresponding regions of other MTases subfamilies. Based on this common AdoMet binding domain architecture and equivalent motif positioning, we have constructed three-dimensional models of the AdoMet binding region of the N-MTase domains of CySyn. The architecture of the putative AdoMet binding domain predicts that the N-MTases domains of CySyn conserve the consensus AdoMet binding contacts defined from comparisons of several structurally characterized AdoMet-dependent MTases (28). The molecular determinants of the AdoMet binding site in the N-MTase domain regions can all be mapped to four noncontiguous sequence elements, which display positional and distant sequence identity to the cofactor binding motif regions of non-nucleic acid MTases. Interpretation of the photoaffinity labeling data within the context of the presented N-MET M8 homology model of the cofactor interaction satisfactorily illustrates the role of the modified residues in the binding of the cofactor molecule.
The model indicates that the [ 14 C]AdoMet-modified peptides build a loop region between the ␤-strand 2 and ␣-helix C secondary structural elements. The heterogeneity of the photoadduct can be construed in terms of the inherent flexibility of polypeptides in loop segments. The photolabeled amino acids are conserved within the N-MTase domains of CySyn, and identical or equivalent residue positions are observed within the N-MTase domain regions of heterogeneous NRPS systems, which further supports the importance of these residues for the recognition and binding of the chemical structure of the AdoMet molecule. The model predicts the modified Pro makes FIG. 8. Hypothetical model for the transmethylation mechanism of the N-MTase domains of CySyn. a, the putative interaction between the positive (red sphere) sulfonium center of AdoMet and the electron-rich (blue sphere) side chain of Tyr 10650 would produce a polarization of the sulfonium methyl, thereby facilitating transfer to the target nucleophile. b, methyl group transfer possibly proceeds via a cationic N-methyl-amino acid transition state (2) prior to subsequent proton abstraction (3) (56). c, sequence analysis indicates the N-MTase domains of CySyn are functionally arranged into AdoMet binding, catalytic, and substrate recognition regions from the NH 2 to COOH terminus. edge-to-face contacts with the AdoMet adenine, whereas the side chain of the modified Glu hydrogen bonds to the ribose hydroxyls.
Many MTases are bilobial and, in addition to the core AdoMet binding region, display structurally unique accessory domains that function in substrate recognition. In comparison with the NH 2 -terminal AdoMet binding region, the COOH-terminal regions of the N-MTase domains of CySyn are more variable. Given that three of the seven N-MTase domains of CySyn N-methylate different amino acid substrates, it is likely that these COOH-terminal regions form the substrate recognition component. All AdoMet-dependent MTases possess AdoMet binding, substrate recognition, and methyl transfer functions. The NH 2 -to COOH-terminal order of these regions varies between MTase classes (31). Based on this common architecture, we suggest that NRPS N-MTase domains are functionally organized into AdoMet-binding catalytic substrate recognition regions from the NH 2 to COOH terminus (Fig. 8c).
The significant level of sequence conservation seen among a given subfamily of MTases allows for the allocation of specific motifs that are inherent to that subfamily to either cofactor binding or substrate recognition. In addition to the four cofactor binding motifs, the N-MTase domains of CySyn display several unique conserved sequence blocks that are innate only to the N-MTase domains of NRPS (Fig. 2). The functional significance of these conserved residues remains unknown, but site-directed mutagenesis of these specific sites should in the future allow us to further refine the contemporary model.
The present consensus is that the evolution of MTases with differing substrate specificities resulted from the formation of hybrid enzymes (28,31). It is believed that the COOH-terminal regions have been exchanged, allowing for the recognition of new substrate targets without modification of the AdoMet binding scaffold. The N-MTase domains of NRPS may have originated from monomeric MTases that fused with the peptide synthetase genes. Given the diverse forms of modifying domains that have been identified in NRPS systems (67), it seems that the alteration of the secondary metabolite biosynthetic genes to confer changes in the structure and bioactivity of the product is a common phenomenon.
The present work presents the first structural information on the identity and location of the AdoMet binding site within the N-MTase domains of CySyn. The homology model presented provides a satisfactory structure-function rationale for the photoaffinity labeling data and a valuable insight into the bimolecular interaction. The model also presents the basis for the analysis and design of site-directed N-MTase mutants to elaborate structure-function studies. The results of this study indicate that the binding of AdoMet to the N-MTase domains of CySyn obeys the consensus AdoMet binding pattern observed in most structurally characterized MTases. It appears the binding of AdoMet is governed by interactions within four noncontiguous sequence elements, which are conserved across members of this MTase subfamily. Given the high degree of sequence homology shared among NRPS N-MTase domains, the findings presented herein serve as a paradigm of AdoMet binding in these systems.