Structure Elucidation and Biosynthesis of Fuscachelins, Peptide Siderophores from the Moderate Thermophile Thermobifida Fusca

Home , Bacillibactin

Structure elucidation and biosynthesis of fuscachelins, peptide siderophores from the moderate thermophile Thermobifida fusca

Eric J. Dimise, Paul F. Widboom, and Steven D. Bruner*

Department of Chemistry, Boston College, Eugene F. Merkert Chemistry Center, Chestnut Hill, MA 02467

Edited by Christopher T. Walsh, Harvard Medical School, Boston, MA, and approved August 20, 2008 (received for review June 4, 2008) Bacteria belonging to the order Actinomycetales have proven to be an biosynthetic genes with the product, prediction of the peptide important source of biologically active and often therapeutically structure is possible. Despite several recent examples of this ap- useful natural products. The characterization of orphan biosynthetic proach, there remain aspects of NRP biosynthesis that are difficult gene clusters is an emerging and valuable approach to the discovery to predict, including uncommon amino acid incorporation, domain of novel small molecules. Analysis of the recently sequenced genome skipping/repeating, and macrocyclization. of the thermophilic actinomycete Thermobifida fusca revealed an Thermobifida fusca is a moderately thermophilic actinomycete orphan nonribosomal peptide biosynthetic gene cluster coding for an widely studied as a model organism for thermostable extracellular unknown siderophore natural product. T. fusca is a model organism cellulases (16–20). The genomic sequence of T. fusca YX was for the study of thermostable cellulases and is a major degrader of reported recently (21). There are no characterized secondary plant cell walls. Here, we report the isolation and structure elucidation metabolites from this actinomycete and few characterized natural of the fuscachelins, siderophore natural products produced by T. products from any thermophilic bacteria or archaea (22). One fusca. In addition, we report the purification and biochemical char- recent example is the elucidation of benzodiazepine biosynthesis in acterization of the termination module of the nonribosomal peptide Streptomyces refuineus (23). Here, we describe a family of structur- synthetase. Biochemical analysis of adenylation domain specificity ally novel nonribosomal peptide siderophores, termed fuscachelins,

supports the assignment of this gene cluster as the producer of the produced by an orphan gene cluster from T. fusca. The elucidated BIOCHEMISTRY fuscachelin siderophores. The proposed nonribosomal peptide bio- biosynthetic pathway contains many unusual aspects that were not synthetic pathway exhibits several atypical features, including a predictable by bioinformatic analysis. In addition, structure eluci- macrocyclizing thioesterase that produces a 10-membered cyclic dep- dation of the fuscachelins revealed a molecular architecture not sipeptide and a nonlinear assembly line, resulting in the unique observed in iron-chelating siderophore secondary metabolites. heterodimeric architecture of the siderophore natural product. Results ͉ ͉ natural product isolation nonribosomal peptide biosynthesis Siderophore Gene Cluster in T. fusca. An uncharacterized gene genome mining cluster in T. fusca contains genes corresponding to a multimodular NRP synthetase secondary metabolite biosynthetic pathway [see ron is a nutrient that is required by virtually all organisms to supporting information (SI) Fig. S1]. Three NRP synthetase genes Iconduct essential life processes. Under aerobic conditions, the designated fscGHI are contiguous in the cluster and correspond to ferric oxidation state predominates as the extraordinarily water- five peptide extension modules (Fig. 1A). The first gene, fscG, insoluble Fe(OH)3 salt (1). These environmentally limiting condi- encodes a 390-kDa protein comprising three NRP synthetase tions have placed selective pressure on organisms to develop modules, each containing the core elongation domains: condensa- controlled and specific mechanisms to acquire iron. Siderophores tion (C), adenylation (A), and peptidyl carrier protein (PCP). The are secondary metabolites used to scavenge ferric ion selectively first module contains a predicted epimerization (E) domain, sug- through the formation of soluble chelation complexes (2, 3). This gesting that the stereochemistry of the corresponding amino acid in structurally diverse group of small molecules contains metal- the product has the D-configuration. The second gene, fscH, chelating motifs that commonly include hydroxamates, catechols, contains a single elongation module and is followed by a gene for ␣-hydroxyacids, and heterocycles to bind iron with high affinity. the termination module, FscI, which contains a C-terminal thioes- Iron uptake is frequently a limiting factor for growth, including in terase (TE) domain. Upstream of fscGHI are genes with sequence human hosts, making siderophores virulence factors in a variety of homology to characterized 2,3-dihydroxybenzoic acid (Dhb) bio- human pathogens and a target for antimicrobial therapy (4–6). synthetic genes. FscA, FscB, and FscD are homologous to the well The constantly expanding pool of microbial genomic sequence characterized catecholate biosynthetic enzymes: isochorismate syn- data has prompted the isolation of new natural products through thase, isochorismatase, and 2,3-dihydro-Dhb dehydrogenase (24– the identification of orphan biosynthetic gene clusters (7–9). Ex- 26). An adenylation domain (FscC) with predicted specificity for ploiting the predictive nature of biosynthetic pathways, natural Dhb and a dedicated aryl-carrier protein (FscF) are present as products can be isolated and characterized by using an assay-guided stand-alone domains to incorporate Dhb as the starter building fractionation approach. Nonribosomal peptide (NRP) biosynthetic block. Additional proximal genes are present that are homologous machinery is often used to construct siderophores (10, 11). Peptide- based architectures allow for the incorporation of common iron- chelating functionalities. Structure elucidation of nonribosomal Author contributions: E.J.D., P.F.W., and S.D.B. designed research; E.J.D. and P.F.W. per- peptide natural products is particularly amenable to a genome formed research; E.J.D., P.F.W., and S.D.B. analyzed data; and E.J.D., P.F.W., and S.D.B. mining approach because the assembly-line nature of the enzymatic wrote the paper. machinery leads to predictable products. NRP synthetases are The authors declare no conﬂict of interest. large, multidomain enzymes catalyzing the assembly of peptides by This article is a PNAS Direct Submission. a thioester-templated mechanism (10). Identification of amino acid *To whom correspondence should be addressed. E-mail: [email protected]. building blocks is possible from analysis of the sequence of the NRP This article contains supporting information online at www.pnas.org/cgi/content/full/ synthetase adenylation domains (12–15). By using this information 0805451105/DCSupplemental. in combination with the frequently observed colinearity of the © 2008 by The National Academy of Sciences of the USA

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0805451105 PNAS ͉ October 7, 2008 ͉ vol. 105 ͉ no. 40 ͉ 15311–15316 Downloaded by guest on September 26, 2021 sponding to fuscachelin B) was further purified by reverse-phase chromatography, and the structure was determined. 1H, TOCSY, and COSY experiments established the amino acid content of the peptide by identifying the individual amino acid spin systems: Arg, Gly, Ser, HOOrn in addition to Dhb (Figs. S3 and S4). Unexpect- edly, 13C and 15N gHMBC experiments to confirm the amino acid connectivity revealed a heterodimeric peptide with the sequence Dhb-Arg-Gly-Gly-Ser-HOOrn-Gly-Gly-Arg-Dhb (Table 1 and Fig. S5). The mass of the isolated product supports this structure with a measured exact m/z of 1048.4448 ([MϩH]ϩ, calculated 1048.4448) and a molecular formula of C42H62N15O17 (Fig. S6). Close exam- ination of the 1H NMR spectra reflects a heterodimeric structure. For example, the integration of the amide and ␣-C protons is 2:4:1:1, Arg:Gly:Ser:HOOrn (see Fig. S3), and fine double signals are evident in the asymmetric halves of the molecule (for example, the protons of Arg and Dhb). To confirm the NMR structural assignment, MALDI-TOF/TOF fragmentation (Fig. 3 and Fig. S6) was performed, and all fragments are consistent with the predicted structure. The determination of amino acid chirality was conducted by using Marfey’s method (36), indicating the presence of D-Arg, Gly, L-Ser, and L-HOOrn in a Ϸ2:4:1:1 ratio based on peak integration (Fig. S7). An additional chromatographic peak from the T. fusca preparation, eluting just before fuscachelin B, exhibited NMR spectra very similar to fuscachelin B. The measured m/z of this distinct product was 1047.4614 ([MϩH]ϩ, calculated 1047.4608, C42H63N16O16) corresponding to a change to an NH versus an O in fuscachelin B. Inspection of the 1H NMR and mass spectral fragmentation data showed that this difference was localized at the HOOrn residue, and additional NMR experiments, in particular 1H/15N gHSQC, are consistent with an ␣-amide structure (termed Fig. 1. The predicted gene products of a nonribosomal peptide biosynthetic fuscachelin C) as shown in Fig. 2B (Fig. S8). cluster in T. fusca.(A) Schematic organization of the nonribosomal peptide A third CAS-positive fraction (fuscachelin A) was evident in the assembly line components. (C, condensation; A, adenylation; ArCP, aryl carrier protein; E, epimerization; PCP, peptidyl carrier protein; and TE, thioesterase chromatographic separation (see Fig. 1A). This fraction was un- domains). (B) Comparison of the adenylation domain speciﬁcity with known nonribosomal peptide natural products (12, 30–34).

to genes traditionally associated with siderophore production and utilization. These include an L-ornithine hydroxylase (FscE), membrane proteins for siderophore export and uptake, and a ferric iron reductase. The amino acid specificity of the individual modules can be predicted by using methodology that compares active-site residues of known NRP synthetase A domains (Fig. 1B) (27–29). The specificity of the first module of FscG did not correspond convinc- ingly to any characterized domains but suggested activation of a basic amino acid. The second and third modules of FscG are highly similar to each other (Ϸ85% identity over the adenylation domains), and both are predicted to activate glycine. Analysis of FscH ␦ and FscI suggests activation of L-serine and L-N -hydroxyornithine (HOOrn), respectively. Based on this analysis, the peptide product can be predicted as a pentapeptide; N-capped with Dhb, a structural architecture unlike any characterized siderophore.

Structure Solution of the T. fusca Siderophores, Fuscachelins. To resolve the structure of the T. fusca siderophore, the natural product was isolated from the producing bacterium and characterized with NMR and mass spectrometry. T. fusca was grown in iron-depleted Ha¨gerdalmedium at 55°C, and the siderophore was extracted from pelleted cells by using methanol. Siderophore activity was monitored throughout production and purification by using the chrome azurol S (CAS) assay for iron-binding activity (35). In addition, fractionation by HPLC was monitored by absorption at 320 nm, characteristic of the predicted catechol functionality. Four separate peaks were isolated from the preparation (Fig. 2A and Fig. S2). A Fig. 2. Puriﬁcation and structures of fuscachelins A–C. (A) HPLC chromatographic minor peak, with the longest retention time, was the free acid trace illustrates four distinct products from T. fusca containing the catechol function- 2,3-Dhb as determined by NMR analysis. A major peak (corre- ality. Dhb, dihydroxybenzoic acid. (B) Chemical structures of fuscachelins A–C.

15312 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0805451105 Dimise et al. Downloaded by guest on September 26, 2021 Table 1. 1H and 13C NMR spectral data for fuscachelin B

Position ␦H, multiplicity J,Hz ␦C Position ␦H, multiplicity J,Hz ␦C 1 148.3 30 1.61 m 29.1 2 146.1 31 4.28dd 4.6, 5.3 54.0 3 7.05 d 7.9 121.2 32 176.6 4 6.82 t 8.0 121.2 33 7.93 d 7.7 5 7.24 d 8.1 120.9 34 172.3 6 117.9 35 3.97 m 43.9 7 171.6 36 8.33 8 8.71 d 6.7 37 173.1 9 4.53dd 7.7 55.3 38 3.95 m 43.7 10 1.98 m 29.5 39 8.78 d 1.89 m 40 176.3 11 1.71 m 25.9 41 4.53dd 7.7 55.3 12 3.21 t 42.0 42 1.98 m 29.5 13 7.15 brs 1.89 m 14 158.2 43 1.71 m 25.9 15 6.59 brs 44 3.21 t 42.0 16 176.3 45 7.15 brs 17 8.78 d 46 158.2 18 4.03 m 44.1 47 6.59 brs 19 172.8 48 8.71 d 6.7 20 8.33 49 171.6 21 4.01 m 43.6 50 117.9 22 171.7 51 148.3 23 8.03 d 6.6 52 146.1 24 5.00dd 5.6, 5.0 54.0 53 7.05 d 121.2 25 3.77 m 62.0 54 6.82 t 121.2 26 171.6 55 7.24 d 120.9 27 28 3.58 m 49.2 3.50 m

29 1.78 m 25.8 BIOCHEMISTRY

stable during the process of purification and appeared to decom- hydroxyornithine side-chain ␦-protons signals were pronounced. pose slowly to fuscachelin B. The mass of this peak is 18 Da less than These changes suggested a substantial change in environment of fuscachelin B ([MϩH]ϩ, m/z predicted: 1030.4342, m/z observed: these two central amino acids in the fuscachelins. Furthermore, this 1030.4344; see Fig. S6), suggesting a hydration/dehydration process. change was capable of being communicated to the Dhb moiety in Aspects of the NMR spectra of fuscachelin A were very similar to the extremities of the molecule, as is evidenced by the increased those described for fuscachelin B, with the D-Arg and Gly signals complexity of the splitting patterns and broadened peaks of the remaining largely identical. There were, however, distinct differ- aromatic protons. Over time, upon standing at ambient tempera- ␦ ences in the L-N -hydroxyornithine, L-serine, and Dhb signals. In ture (unbuffered D2O), these changes in proton signals began to particular, the ␤-hydrogens of serine were shifted to 4.48 and 4.36 disappear as signals for fuscachelin B concurrently grew into the ppm (compared with 3.77 ppm in fuscachelin B), suggesting the spectrum (Fig. S9). These observations, in concert with the ob- ␦ presence of a seryl ester. In addition, shifts in the L-N - served mass difference, suggested the presence of a macrolactone

Fig. 3. MALDI-TOF/TOF fragmentation of fuscachelin B.

Dimise et al. PNAS ͉ October 7, 2008 ͉ vol. 105 ͉ no. 40 ͉ 15313 Downloaded by guest on September 26, 2021 The amino acid specificity of the module was determined by using the pyrophosphate exchange assay of adenylation domain function. L-HOOrn was chemically synthesized by using published protocols from N-Boc-L-ornithine (37). This predicted amino acid was as- sayed along with the 20 proteinogenic amino acids and L-ornithine. The results show a significant preference for L-HOOrn with the next best amino acid substrate the isosteric L-lysine (Fig. 5B). Discussion Identifying orphan biosynthetic gene clusters through genome mining has proven to be a successful approach toward the discovery of novel small molecule secondary metabolites. The moderately thermophilic actinomycete, T. fusca, contains one gene cluster belonging to the NRP family of natural products. The cluster contains the hallmarks of siderophore-producing enzymes, with Fig. 4. Proposed origin of fuscachelin B from the product of the biosynthetic biosynthetic pathways to the iron-chelating 2,3-dihydroxybenzyl gene cluster, fuscachelin A. HPLC analysis showing the proposed hydrolytic and hydroxamide moieties. Although the enzymatic components of conversion of fuscachelin A to fuscachelin B. the fuscachelin cluster have homologs in other siderophore NRP synthetases, the combination and organization of the fsc genes are involving the serine side chain and the ␣-carboxyl of L-HOOrn (Fig. unique. Siderophore natural products most often bind to extracel- 2B). The unusual 10-membered cyclodepsipeptide core structure in lular ferric ions through the formation of a hexacoordinate, octa- fuscachelin A was confirmed by additional NMR-based measure- hedral chelation complex. The T. fusca siderophore is able to ments (Fig. S9). Further evidence for the presence of a macrolac- achieve this through the production of a heterodimeric nonriboso- tone and the degradation of fuscachelin A to B was provided by the mal peptide with two terminal catechols and an internal hydrox- alkali treatment of purified fuscachelin A (Fig. 4). The results show amate. The biosynthetic pathway for this siderophore, termed that fuscachelin A is converted to fuscachelin B based on HPLC fuscachelin A, is unusual, and prediction of the structure of the natural product would not be possible based on analysis of the analysis. As discussed below, the macrolactone fuscachelin A is the genomic sequence alone. Namely, unambiguous assignment of the most likely initial biosynthetic product of the T. fusca fuscachelin adenylation domains, prediction of macrocyclization, and the un- gene cluster. Energy minimization of fuscachelin A results in a usual nonlinear biosynthetic pathway were unexpected. structure in which the two catechols are in proximity to the central We initially characterized a siderophore natural product (fus- hydroxamate, and modeling of a metal chelate is easily accommo- cachelin B) from T. fusca as a linear octapeptide N-capped with two dated (Fig. S10). 2,3-dihydroxybenzoic acids and an internal ␣,␧-linked HOOrn. The Siderophore production and utilization depend on cellular re- structure was determined with detailed NMR and mass spectro- sponses to iron in the environment. The concentration of ferrous metric characterization of the isolated product. A second fraction iron in the growth medium affects the expression of bacterial in the purification (fuscachelin C) exhibited characteristics nearly siderophore biosynthetic genes (1). To provide evidence that the identical to fuscachelin B, with the notable exception of a 1-Da mass fuscachelins are the biologically relevant siderophore for T. fusca, difference, corresponding to a change of an oxygen atom to an NH. the bacteria were grown in minimal medium with and without iron, This change was localized by NMR and mass spectral fragmenta- and the production of fuscachelin was monitored by HPLC analysis. tion data to the L-HOOrn subunit, specifically as an ␣-carboxamide As predicted, the presence of iron suppressed fuscachelin produc- group. A third product was isolated from T. fusca that displayed tion (Fig. S2). In addition, the fuscachelin B–iron complex was properties similar to those of fuscachelin B and C and slowly prepared and characterized by MALDI-TOF mass spectral analysis decomposed over time, or rapidly with alkali treatment, to fuscach- to demonstrate that the natural product provides a scaffold for iron elin B. This product (fuscachelin A) was characterized as a mac- ϩ ϩ binding. The results ([M Fe] m/z 1101.3600, C42H59N15O17Fe) rolactone, and, based on the predicted biosynthetic pathway, it is agree well with the predicted value of 1101.3557 (Fig. S6). most likely the initial natural product of the T. fusca NRP synthetase machinery. We believe that fuscachelins B and C are Cloning, Purification, and Biochemical Characterization of the Termi- degradation products of fuscachelin A, resulting from nucleophilic nal Module FscI. To provide supporting evidence that the fsc genes opening of the 10-membered macrolactone: hydrolysis to form are responsible for the in vivo production of the fuscachelins in T. fuscachelin B and aminolysis (the growth medium contains 24 mM fusca, the terminal module, FscI, was characterized in vitro. The ammonium sulfate) to form fuscachelin C. 4.0-kb gene for fscI was cloned from T. fusca genomic DNA into an The predicted biosynthetic pathway (Fig. 6) for fuscachelin A Escherichia coli expression vector using PCR-based methods. The contains well characterized genes for the biosynthesis of a peptidic four-domain protein was overexpressed in E. coli to high levels and siderophore, although there are several unusual features. The purified to homogeneity as judged by SDS/PAGE analysis (Fig. 5A). condensation domain of FscI couples both Dhb-Arg-Gly-Gly and

Fig. 5. Characterization of FscI. (A) SDS/polyacryl- amide gel illustrating overexpressed and puriﬁed FscI (148 kDa). (B) Relative activity of FscI with various amino acids as judged by the pyrophosphate exchange assay for adenylation function.

15314 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0805451105 Dimise et al. Downloaded by guest on September 26, 2021 Fig. 6. Proposed biosynthetic pathway to fuscachelin A.

Dhb-Arg-Gly-Gly-Ser to the ␦-and ␣-nitrogens of L-HOOrn bound system equipped with a Shimadzu SPD-10A UV-visible detector. A detection to the PCP domain of FscI. The two coupling partners originate wavelength of 320 nm was chosen to monitor elution of the 2,3-dihydroxybenzyl from the carrier domains of FscG and FscH. The second coupling moiety. A linear gradient of 2–50% methanol in 0.1% trifluoroacetic acid and Ϫ1 (as diagrammed in Fig. 6) between FscI and FscG involves the water was run over 30 min at 10 ml min . CAS-positive fractions were collected between 23 and 28 min and lyophilized. The lyophilized material was dissolved in skipping of module FscH, an example of nonlinear peptide assem- 5 ml of 23% methanol and water. One-milliliter portions were injected onto the bly, observed in a limited number of NRP pathways (38). The column and purified by using a linear gradient of 23–33% methanol in water and coupling of two PCP-bound acyl intermediates to L-HOOrn and the 0.1% trifluoroacetic acid over 60 min at 8 ml minϪ1. Three CAS-positive fractions

predicted module skipping is reminiscent of the chemistry of the were collected at 17.1, 18.6, and 22.8 min. Lyophilization of the CAS-positive BIOCHEMISTRY NRP synthetase CchH in coelichelin biosynthesis (12). After the fractions afforded 2.0, 3.0, and 2.3 mg of fuscachelins C, B, and A, respectively, as tandem couplings to the HOOrn core, the thioesterase domain of white solids. The fourth CAS-positive fraction (eluting at 24.8 min) was free FscI performs a macrocyclization to form the 10-membered dep- 2,3-dihydroxybenzoic acid as judged by 1H NMR. Samples for 15N NMR analysis 15 sipeptide ring of fuscachelin A. The FscI TE domain shows signif- were prepared in an identical fashion with the exception that ( NH4)2SO4 (99% 15N, Cambridge Isotope Laboratories) was included in the Ha¨gerdal medium as icant homology to the thioesterase of DhbF of bacillibactin bio- the sole nitrogen source. synthesis (30). DhbF catalyzes a related reaction, the cyclotrimerization of Dhb-Ser-Gly to form a 12-membered macro- NMR Structure Elucidation. All NMR spectra were acquired by using a D2O lactone siderophore. To confirm the activity of the gene cluster, fscI susceptibility matched 5-mm Shigemi advanced NMR microtube. NMR solvents was cloned from T. fusca and overexpressed in E. coli. The were purchased from Cambridge Isotope Laboratories. 1H, 13C, 15N, and 1H-13C adenylation domain of the purified protein was biochemically correlation spectra were acquired on a Varian INOVA 500 MHz (1H) NMR spec- demonstrated to prefer the nonproteinogenic amino acid L- trometer with VNMR 6.1C software. Spectra were processed by using MestReC HOOrn. This result both establishes the role and position of the version 4.9.9.8 software. Spectra acquired in D2O were referenced to the residual NRP synthetase FscI in the pathway and demonstrates that the HOD peak. In cases where an H2O/D2O (4:1) solvent mixture was used to observe backbone amide protons and exchangeable side chain protons, the intense flavin monooxygenase (FscE) acts on the free amino acid L- solvent peak was suppressed by using the PRESAT and solvent suppression pa- ornithine and not the mature peptide. rameters in the VNMR 6.1C software. 1H, TOCSY (512 increments, 64 transients, Here, we present the structure elucidation of a siderophore from 80-ms mixing time), gCOSY (512 increments, 48 transients), gHSQC (512 incre- the thermophilic actinomycete T. fusca. The natural product was ments, 32 transients), and gHMBC (512 increments, 64 transients) spectra were discovered by using a genome-mining approach. Fuscachelin A is acquired by using a Varian 500 ID/PFG 50-202-MHz inverse probe. 13C and 15N the first secondary metabolite isolated from this bacterium, and the spectra were acquired by using a Varian 500 SW/PFG 50-202-MHz broadband 13 described natural product gene cluster is one of only a few from any probe. C spectra were referenced to CD3OD, which was used as an internal 15 15 standard. N spectra were externally referenced to 2.9 M NH4Cl/1 M DCl in D2O thermophilic species. The proposed biosynthetic pathway contains 1 15 unusual aspects that demonstrate the flexibility of NRP assembly- (Cambridge Isotope Laboratories). H- N correlation spectra (gHSQC, 256 increments, 4 transients; and gHMBC, 128 increments, 16 transients) were acquired on line chemistry to construct biologically active peptides, and the a Varian VNMRS 600-MHz (1H) NMR spectrometer equipped with a 5-mm HCN structure of fuscachelin A represents a new molecular architecture AutoX inverse probe. for the chelation of iron. MS Analysis. High-resolution ESIϩ single-mass analysis was performed on an Materials and Methods Agilent LC/MSD TOF mass spectrometer equipped with Agilent Technologies Purification of Fuscachelins A–C. T. fusca spores (ATCC 2773) were used to 1100 series cap-LC pumps. MS/MS experiments were performed on an Applied inoculate 5-ml liquid cultures grown in LB broth at 55°C and 150 rpm in a Thermo Biosystems/MDS SCIEX 4800 MALDI-TOF/TOF mass spectrometer. A ferric– Electron Corporation Forma Orbital Shaker. After 48 h, the cells were thoroughly fuscachelin B complex was prepared by mixing purified fuscachelin B with 1% exchanged into 5 ml of iron-deficient Ha¨gerdal medium (39) by repeated cen- FeCl3 in a 1:1 siderophore:Fe ratio (40). The solution was mixed at room trifugation and used to inoculate 1-liter liquid cultures, also in iron-deficient temperature for 2 h and lyophilized. The dried sample was subjected to ESIϩ Ha¨gerdal medium. One-liter cultures were grown for 7 days at 55°C and 150 rpm analysis by using a Micromass LCT-TOF mass spectrometer. in a Thermo Electron Corporation Forma Orbital Shaker. Cells were pelleted at 10,000 rpm and 4°C for 30 min by using a Beckman–Coulter J2-HS centrifuge. Cell Marfey’s Method Analysis of Amino Acids. Five hundred micrograms of fuscach- pellets were collected and extracted five times with 30-ml portions of methanol. elin B was dissolved in 400 ␮l of 6 M HCl and heated at 110°C for 24 h. The mixture The methanol extracts were concentrated in vacuo to a small volume (Ϸ5 ml), and was lyophilized, and the residue was dissolved in 10 ␮l of water, 20 ␮lof1M ␣ siderophore activity was confirmed by using the CAS assay of Schwyn and Nei- NaHCO3, and 170 ␮l of 1% Marfey’s reagent [N -(2,4-dinitro-5-fluorophenyl)-L- lands (35). CAS-positive extracts were concentrated to dryness and subjected to alaninamide, Sigma–Aldrich] acetone solution and heated at 37°C for 1 h. The two rounds of preparative HPLC by using a Vydac 218TP1022 protein and peptide reaction was quenched with 20 ␮l of 1 M HCl and the mixture lyophilized. The C18 column (250 ϫ 22 mm, 10 ␮m) on a Shimadzu LC-6AD liquid chromatography dried products were dissolved in 1:1 water:acetonitrile and 0.1% trifluoroacetic

Dimise et al. PNAS ͉ October 7, 2008 ͉ vol. 105 ͉ no. 40 ͉ 15315 Downloaded by guest on September 26, 2021 acid solution to a final volume of 400 ␮l and analyzed on a Vydac 218TP54 protein TCG (HindIII). The PCR products were purified through agarose gel electrophore- and peptide C18 HPLC column (250 ϫ 4.6 mm, 5 ␮m) on a Shimadzu LC-6AD liquid sis and gel extraction (Qiagen) and cleaved with the EcoRI and HindIII restriction chromatography system equipped with a Shimadzu SPD-10A UV-visible detector. endonucleases. The fscI gene was then ligated into the plasmid pET30a. The The separation was performed by using a linear gradient of 0- 52.5% buffer B (10 plasmid was transformed into E. coli BL21(DE3) cells for gene expression. Cultures mM ammonium formate, 1% methanol, 60% acetonitrile, pH 5.2) in buffer A (10 (1 liter) were grown to A600 ϭ 0.5–0.7 at 37°C, at which point the shaker was mM ammonium formate, 1% methanol, 5% acetonitrile, pH 5.2) over 45 min by cooled to 18°C, and overexpression was initiated by the addition of 50 ␮M IPTG. using detection wavelengths of 340 and 220 nm, an injection volume of 5 ␮l, and Cultures were continued for 18 h and were harvested by centrifugation, followed a volume flow rate of 1 ml minϪ1 (36). Stereochemical assignments were made by by resuspension in 500 mM NaCl, 20 mM Tris⅐HCl (pH 7.5) and lysed by passage comparison with the retention times of Marfey’s derivatives prepared from through a French pressure cell at 1,000 psi. Lysate was centrifuged at 10,000 rpm authentic amino acid standards of D/L-arginine, D/L-serine, D/L-ornithine, and for 20 min in a Beckman Coulter J2-HS centrifuge. The supernatant was incubated glycine (Sigma–Aldrich). In addition, the peak for L-HOOrn was identified by using for 1 h with 1 ml of metal-affinity resin (Talon resin; Clontech). Resin was washed ϫ ⅐ the Marfey’s derivative of a synthetic standard of L-HOOrn that was chemically with 4 10 ml of 500 mM NaCl, 20 mM Tris HCl (pH 7.5), and protein was eluted ϫ ⅐ synthesized by using an established protocol (37). Marfey’s derivatives of HPLC with 2 10 ml of 500 mM NaCl, 20 mM Tris HCl (pH 7.5), 250 mM imidazole. ⅐ amino acid standards were prepared by mixing 50 ␮l of 50 mM (aq) amino acid, Protein was dialyzed against 100 mM NaCl, 20 mM Tris HCl (pH 7.5), 1 mM ␤-mercaptoethanol, 10% (vol/vol) glycerol, concentrated to 17 ␮M, and flash 100 ␮l of 1% Marfey’s reagent/acetone solution, and 20 ␮l of 1 M NaHCO3, then heating at 37°C for 1 h, lyophilizing, and dissolving in 1:1 water:acetonitrile and frozen. 0.1% trifluoroacetic acid solution to a final volume of 400 ␮l. Separation of the standard mixture was optimized to the conditions outlined above. Pyrophosphate Exchange Assay. Amino acid-dependent ATP-sodium pyrophosphate assays were performed as follows. A 100-␮l reaction contained 75 ⅐ 32 Fuscachelin A HPLC Peak Shift Analysis. Twenty microliters of 100 ␮M standards mM Tris HCl (pH 8.0), 10 mM MgCl2,5mMDTT,5mMATP,1mMNa4 P2O7, ␮ Ϫ1 ␮ of purified fuscachelins A and B were run on a Vydac 218TP54 protein and 100 g/ml BSA, 1 mM amino acid, 2 M FscI. Reactions were initiated by addition of enzyme and incubated at 30°C for 0.5 h. The reaction was peptide C18 HPLC column (250 ϫ 4.6 mm, 5 ␮m) by using a linear gradient of quenched by the addition of 500 ␮l of 3.5% charcoal, 1.6% perchloric acid, 200 23–28% B (methanol) in A (water and 0.1% trifluoroacetic acid) over 30 min mM Na P O . The charcoal was centrifuged and resuspended twice with 500 at1mlminϪ1. HPLC traces were acquired at 220 and 335 nm. Then, a 300-␮l 4 2 7 ␮l of 1.6% perchloric acid, 200 mM Na P O . After washing, the charcoal was sample of 100 ␮M fuscachelin A in 100 mM phosphate buffer (pH 10) was 4 2 7 mixed with 3 ml of scintillation fluid and read by a Beckman–Coulter LS 6500 shaken at 37°C for 2 h. The reaction mixture was lyophilized to dryness, taken scintillation counter. All reactions were performed in triplicate. up in 23% methanol and water to a final volume of 300 ␮l, and subjected to the HPLC conditions outlined above. ACKNOWLEDGMENTS. We thank Michael Thomas for helpful discussions and John Boylan, Rebecca Butcher, and Charles Sheahan for invaluable assistance Cloning, Expression, and Purification of FscI. The gene for fscI was amplified by on NMR experiments. We acknowledge experimental contributions by Joanne using PCR from T. fusca genomic DNA with the following primers: 5Ј-GCG GAA Kehlbeck. This work was supported by the National Science Foundation Grant TTC ACC ACC GCA GCC GCG GGT (EcoRI), 5Ј-GCG AAG CTT CTA GCT GTG TCC GGA CAREER0645653 (to S.D.B.).

1. Sigel A, Sigel HE (1998) Metal Ions in Biological Systems: Iron Transport and Storage 23. Hu Y, et al. (2007) Benzodiazepine biosynthesis in Streptomyces refuineus. Chem Biol in Microorganisms, Plants, and Animals (Dekker, New York). 14:691–701. 2. Drechsel H, Jung G (1998) Peptide siderophores. J Pept Sci 4:147–181. 24. Liu J, Duncan K, Walsh CT (1989) Nucleotide sequence of a cluster of Escherichia coli 3. Fischbach MA, Lin H, Liu DR, Walsh CT (2006) How pathogenic bacteria evade mam- enterobactin biosynthesis genes: Identification of entA and purification of its product malian sabotage in the battle for iron. Nat Chem Biol 2:132–138. 2,3-dihydro-2,3-dihydroxybenzoate dehydrogenase. J Bacteriol 171:791–798. 4. Schaible UE, Kaufmann SH (2004) Iron and microbial infection. Nat Rev Microbiol 25. Liu J, Quinn N, Berchtold GA, Walsh CT (1990) Overexpression, purification, and 2:946–953. characterization of isochorismate synthase (EntC), the first enzyme involved in the 5. Rodriguez GM, Smith I (2003) Mechanisms of iron regulation in mycobacteria: Role in biosynthesis of enterobactin from chorismate. Biochemistry 29:1417–1425. physiology and virulence. Mol Microbiol 47:1485–1494. 26. Rusnak F, Liu J, Quinn N, Berchtold GA, Walsh CT (1990) Subcloning of the enterobactin 6. Perry RD, Balbo PB, Jones HA, Fetherston JD, DeMoll E (1999) Yersiniabactin from Yersinia pestis: Biochemical characterization of the siderophore and its role in iron biosynthetic gene entB: Expression, purification, characterization, and substrate spec- transport and regulation. Microbiology 145:1181–1190. ificity of isochorismatase. Biochemistry 29:1425–1435. 7. Wilkinson B, Micklefield J (2007) Mining and engineering natural-product biosynthetic 27. Stachelhaus T, Mootz HD, Marahiel MA (1999) The specificity-conferring code of pathways. Nat Chem Biol 3:379–386. adenylation domains in nonribosomal peptide synthetases. Chem Biol 6:493–505. 8. Challis GL (2008) Genome mining for novel natural product discovery. J Med Chem 28. Challis GL, Ravel J, Townsend CA (2000) Predictive, structure-based model of amino acid 51:2618–2628. recognition by nonribosomal peptide synthetase adenylation domains. Chem Biol 7:211–224. 9. Gross H (2007) Strategies to unravel the function of orphan biosynthesis pathways: 29. Rausch C, Weber T, Kohlbacher O, Wohlleben W, Huson DH (2005) Specificity predic- Recent examples and future prospects. Appl Microbiol Biotechnol 75:267–277. tion of adenylation domains in nonribosomal peptide synthetases (NRPS) by using 10. Sieber SA, Marahiel MA (2005) Molecular mechanisms underlying nonribosomal pep- transductive support vector machines (TSVMs). Nucleic Acids Res 33:5799–5808. tide synthesis: Approaches to new antibiotics. Chem Rev 105:715–738. 30. May JJ, Wendrich TM, Marahiel MA (2001) The dhb operon of Bacillus subtilis encodes 11. Fischbach MA, Walsh CT (2006) Assembly-line enzymology for polyketide and nonriboso- the biosynthetic template for the catecholic siderophore 2,3-dihydroxybenzoate- mal peptide antibiotics: Logic, machinery, and mechanisms. Chem Rev 106:3468–3496. glycine-threonine trimeric ester bacillibactin. J Biol Chem 276:7209–7217. 12. Lautru S, Deeth RJ, Bailey LM, Challis GL (2005) Discovery of a new peptide natural 31. Moffitt MC, Neilan BA (2004) Characterization of the nodularin synthetase gene cluster product by Streptomyces coelicolor genome mining. Nat Chem Biol 1:265–269. and proposed theory of the evolution of cyanobacterial hepatotoxins. Appl Environ 13. Gross H, et al. (2007) The genomisotopic approach: A systematic method to isolate Microbiol 70:6353–6362. products of orphan biosynthetic gene clusters. Chem Biol 14:53–63. 14. Ishida K, et al. (2007) Biosynthesis and structure of aeruginoside 126A and 126B, 32. Balado M, Osorio CR, Lemos ML (2006) A gene cluster involved in the biosynthesis of cyanobacterial peptide glycosides bearing a 2-carboxyl-6-hydroxyoctahydroindole vanchrobactin, a chromosome-encoded siderophore produced by Vibrio anguillarum. moiety. Chem Biol 14:565–576. Microbiology 152:3517–3528. 15. de Bruijn I, et al. (2007) Genome-based discovery, structure prediction, and functional 33. Guenzi E, Galli G, Grgurina I, Gross DC, Grandi G (1998) Characterization of the analysis of cyclic lipopeptide antibiotics in Pseudomonas species. Mol Microbiol syringomycin synthetase gene cluster: A link between prokaryotic and eukaryotic 63:417–428. peptide synthetases. J Biol Chem 273:32857–32863. 16. Bachmann SL, McCarthy AJ (1991) Purification and cooperative activity of enzymes 34. Hoffmann D, Hevel JM, Moore RE, Moore BS (2003) Sequence analysis and biochemical constituting the xylan-degrading system of Thermomonospora fusca. Appl Environ characterization of the nostopeptolide A biosynthetic gene cluster from Nostoc sp Microbiol 57:2121–2130. GSV224. Gene 311:171–180. 17. Wilson DB (1992) Biochemistry and genetics of actinomycete cellulases. Crit Rev 35. Schwyn B, Neilands JB (1987) Universal chemical assay for the detection and determi- Biotech 12:45–63. nation of siderophores. Anal Biochem 160:47–56. 18. Sakon J, Irwin D, Wilson DB, Karplus PA (1997) Structure and mechanism of endo/ 36. Bhushan R, Bruckner H (2004) Marfey’s reagent for chiral amino acid analysis: A review. exocellulase E4 from Thermomonospora fusca. Nat Struct Biol 4:810–818. Amino Acids 27:231–247. 19. Hilge M, et al. (1998) High-resolution native and complex structures of thermostable 37. Lin YM, Miller MJ (1999) Practical synthesis of hydroxamate-derived siderophore ␤-mannanase from Thermomonospora fusca: Substrate specificity in glycosyl hydro- components by an indirect oxidation method and syntheses of a DIG-siderophore lase family 5. Structure 6:1433–1444. conjugate and a biotin–siderophore conjugate. J Org Chem 64:7451–7458. 20. Larsson AM, et al. (2005) Crystal structure of Thermobifida fusca endoglucanase Cel6A in complex with substrate and inhibitor: The role of tyrosine Y73 in substrate ring 38. Haynes SW, Challis GL (2007) Nonlinear enzymatic logic in natural product modular distortion. Biochemistry 44:12915–12922. mega-synthases and -synthetases. Curr Opin Drug Discov Dev 10:203–218. 21. Lykidis A, et al. (2007) Genome sequence and analysis of the soil cellulolytic actino- 39. Ha¨gerdal BG, Ferchak JD, Pye EK (1978) Cellulolytic enzyme system of Thermoactino- mycete. Thermobifida fusca YX. J Bacteriol 189:2477–2486. myces sp grown on microcrystalline cellulose. Appl Environ Microbiol 36:606–612. 22. Donadio S, Monciardini P, Sosio M (2007) Polyketide synthases and nonribosomal peptide 40. Stintzi A, Barnes C, Xu J, Raymond KN (2000) Microbial iron transport via a siderophore synthetases: The emerging view from bacterial genomics. Nat Prod Rep 24:1073–1109. shuttle: A membrane ion transport paradigm. Proc Natl Acad Sci USA 97:10691–10696.

15316 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0805451105 Dimise et al. Downloaded by guest on September 26, 2021