Initiation of the flexirubin biosynthesis in Chitinophaga pinensis

Bacteria from the Bacteroidetes phylum are known producers of the chemotaxonomic relevant flexirubins. These orange pigments comprise a non-isoprenoid aryl-polyene carboxylic acid esterified with a dialkylresorcinol. Herein, we report a gene cluster from Chitinophaga pinensis encoding the biosynthesis of the polyene moiety and the biochemical characterization of a tyrosine ammonia-lyase and a 4-coumarate-CoA ligase responsible for the initiation of the polyene biosynthesis. Additionally, the flexirubin of C. pinensis was characterized by a combination of feeding experiments, high-performance liquid chromatography tandem mass spectrometry and matrix-assisted laser desorption/ionization mass spectrometry.


Introduction
Since the discovery of flexirubin (1) (Fig. 1) in Chitinophaga filiformis (previously Flexibacter elegans Fx e1) (Reichenbach et al., 1974;Kämpfer et al., 2006), flexirubin-type pigments have been used as chemotaxonomic markers for the bacteria of the Bacteroidetes phylum (previously called the Cytophaga-Flavobacterium-Bacteroides group). Their conserved structural feature is a ω-(4-hydroxyphenyl)-polyene carboxylic acid chromophore, esterified with a 2,5dialkylresorcinol (DAR). Known derivatives differ in methylation and/or chlorination pattern at the polyene ring and the chain lengths of the DAR alkyl chains as well as the polyene chain length (Achenbach et al., 1982). Feeding experiments with radioactive precursors combined with chemical degradation showed that tyrosine is the precursor of the polyene ring and that its methyl group is derived from methionine . As labelled acetate and malonate were also incorporated, it was proposed that fatty acid and/or polyketide biosynthesis mechanisms are involved . Whereas we could show previously that the DAR moiety is derived from a condensation of two fatty acid metabolism intermediates by DAR enzymes (Fuchs et al., 2013), the enzymes for the biosynthesis of the polyene moiety are still unknown. However, mutant strains of Flavobacterium johnsoniae UW101 revealed that the DAR biosynthesis genes from this flexirubin producer are flanked by other genes encoding proteins involved in flexirubin biosynthesis (McBride et al., 2009). One gene in this gene cluster encodes a putative enyzme from the aromatic amino acid lyase family. This enzyme family comprises histidine ammonia-lyases (HALs), phenylalanine ammonia-lyases (PALs) and tyrosine ammonia-lyases (TALs). HALs are common in bacteria and humans and deaminate histidine to urocanic acid (Michal, 1999). PALs convert phenylalanine to E-cinnamic acid (CA), whereas TALs deaminate tyrosine to 4-coumarate (4-hydroxycinnamic acid, 4CA). Whereas PAL and TAL are ubiquitous in plants and fungi, only a few examples were found in bacteria, such as Streptomyces maritimus PAL (Xiang and Moore, 2002), Photorhabdus luminescens PAL (Williams et al., 2005) Rhodobacter capsulatus TAL (Kyndt et al., 2002), Rhodobacter sphaeroides TAL (Xue et al., 2007), Streptomyces sp. Tü4128 TAL (Zhu et al., 2012) and Saccharothrix espanaensis TAL (Berner et al., 2006). 4CA or CA can be activated by 4-coumarate-CoA ligases (4CL) or E-cinnamate-CoA ligases (CCL). 4CL are common in plants where they catalyse the last reaction in the phenylpropanoid pathway and have been extensively studied in the past. Only little is known about enzymes with 4CL activity from bacteria; as to the best of our knowledge, only the enzyme ScCCL from Streptomyces coelicolor A3 (2) (Kaneko et al., 2003) and an unknown enzyme from Pseudomonas putida (Zenk et al., 1980) have been shown in vitro to have 4CL activity. Furthermore, hints for the presence of 4CL in bacterial biosynthetic pathways were gained in works about the biosynthesis of enterocin in S. maritimus (Hertweck and Moore, 2000) or the photoactive yellow protein in R. capsulatus (Kyndt et al., 2003).
Here, we describe the putative flexirubin biosynthesis gene cluster of Chitinophaga pinensis and the biochemical characterization of the TAL as well as the 4CL, both involved in the start of the polyene biosynthesis. Furthermore, we characterized the flexirubin from C. pinensis by a combination of stable isotope labelling, highperformance liquid chromatography-electrospray ionization tandem mass spectrometry (HPLC-ESI-MS) and matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS), presenting an alternative method for the detection of this pigment class in bacteria.

Results and discussion
We previously reported that heterologous expression of the genes darB (Cpin_6850) and darA (Cpin_6851) ( Fig. 2A) from C. pinensis led to the production of a DAR, which is identical to the DAR moiety in flexirubin 1 from C. filiformes (Fuchs et al., 2013). Additionally, these genes have also been found in F. johnsoniae UW101 where mutation of darB (Fjoh_1102) (Fig. 2B) resulted in a flexirubin-negative phenotype (McBride et al., 2009). In F. johnsoniae, darA and darB are part of a 37 kbp gene cluster and a spontaneous flexirubin-negative mutant was complemented by a plasmid carrying part of this cluster (Fjoh_1078 to Fjoh_1089) (Fig. 2B), which proved that additional genes from this cluster are involved in flexirubin biosynthesis (McBride et al., 2009). However, no further characterization of this gene cluster was reported. STRING analysis (Franceschini et al., 2013) of the C. pinensis genome with the F. johnsoniae flexirubin gene cluster revealed a homologous gene cluster spanning Cpin_1853 (flxA) to Cpin_1877 (flxY) and the already reported DAR gene cluster ( Fig. 2A and Tables S4 and  S5). In the gene cluster, several putative β-ketoacyl synthases, reductases, dehydratases and thioesterases are encoded, suggesting a type II fatty acid synthase-like biosynthesis of the polyene moiety (Fig. 3). As the polyene-aryl and the adjacent double bond are derived from tyrosine , we speculated that the predicted HAL FlxA (Cpin_1853) encoded in the biosynthesis gene cluster, in fact, catalyses the deamination of L-tyrosine to 4CA, which might then be activated for the polyketide synthase (PKS) machinery by adenylation through the putative acyl-CoA ligase FlxY (Cpin_1877) (Fig. 3). 4-Coumaroyl-CoA may then be used as precursor for a fatty acid-like biosynthesis of the polyene by chain   (McBride et al., 2009). Colours depict genes with the same annotation, which are connected by grey lines if their identity was ≥ 40% in a BLAST-P analysis. All genes are scaled to the depicted scale bar. elongation through the putative β-ketoacyl synthases (FlxC, FlxI, FlxL, FlxN, FlxO) followed by reduction of the β-keto function by FlxB or FlxV and generation of a double bond by a dehydratase (FlxF, FlxS). The aryl-octaene moiety may then be connected to the DAR by a ligaselike enzyme (FlxW). Flexirubin is found in the outer membrane (Irschik and Reichenbach, 1978). Thus, flxP-flxU, encoding a putative polysaccharide deacetylase, a phospholipid/glycerol acyltransferase, an outer membrane lipoprotein carrier LolA, a glycosyltransferase and a predicted exporter may be involved in the export of the pigment by yet unknown mechanisms.

Tyrosine ammonia-lyase activity
Because two motives were identified in the past that might allow the prediction of the substrate specificity of aromatic amino acid ammonia-lyases, an alignment of the FlxA primary sequence with those of eight ammonia-lyases with known substrate specificity was performed (Fig. S1). However, the amino acids at motive 1 (Watts et al., 2006) were similar to those of HAL enzymes and not TAL/PAL, whereas at motive 2 (Berner et al., 2006), the conserved residue from TAL or PAL enzymes was found. To get experimental evidence whether FlxA has ammonia-lyase activity, it was heterologously produced in Escherichia coli. Subsequent purification of FlxA (Fig. S2) allowed its incubation with putative substrates and the enzyme assays were analysed by gas chromatography-mass spectrometry (GC-MS). In assays containing FlxA and L-tyrosine, 4CA was detected ( Fig. 4A and Fig. S3A), whereas L-phenylalanine was deaminated to CA (Fig. S3B). No product formation was detectable with L-histidine, L-isoleucine and L-tryptophane as substrates (data not shown). The same results were obtained in a photometric assay (Fig. S4A). To confirm the specificity of FlxA, photometric assays with L-tyrosine or Lphenylalanine as substrate were optimized towards pH and temperature (Fig. S4B). The linear plots of the Michaelis-Menten kinetics ( Fig. S5A-D) revealed kcat/KM values of 7728 M −1 s −1 for L-tyrosine and 7.7 M −1 s −1 for L-phenylalanine. Although other bacterial TALs showed a 150-to 300-fold higher catalytic activity with L-tyrosine (Watts et al., 2006), FlxA showed a 1010-fold higher specificity towards L-tyrosine against L-phenylalanine, supporting its proposed function in the flexirubin biosynthesis.

CoA ligase activity
As the polyene-aryl and the adjacent double bond from flexirubin are derived from tyrosine , we concluded that the 4CA produced by FlxA may be activated by the putative acyl-CoA ligase FlxY to serve as substrate for the following PKS biosynthesis of the polyene. An alignment of the primary sequence from FlxY with those from ScCCL from S. coelicolor and three 4CL isoforms from Arabidopsis thaliana is shown in Fig. S6. The alignment shows that, in contrast to ScCCL (44% identity to At4Cl2), FlxY is not closely related to plant 4CLs (< 12% identity to At4CL1-3) and several residues in reported conserved domains differ from those of plant 4CLs (Schneider et al., 2003) or ScCCL (Kaneko et al., 2003), preventing an accurate prediction of the catalysed reaction. Therefore, we tested the acyl-CoA ligase activity of FlxY in vitro. After heterologous overexpression in E. coli and subsequent purification ( Furthermore, E-3-indoleacrylic acid, indole-3-propionic acid, 3-chlorocinnamic acid and 3-phenylpropionic acid were also accepted as substrates, but other carboxylic acids were not ( Fig. S8 and Table S3). To test the substrate specificity of FlxY, a coupled enzyme assay was used and optimized for product formation at different pH and temperatures ( Fig. S9A Table S6 (Zenk et al., 1980;Kaneko et al., 2003), showing that FlxY is the bacterial enzyme with the best 4CL activity reported to date.

Flexirubin structure elucidation
As C. pinensis is a close relative of C. filiformes in the Bacteroidetes phylum (Kämpfer et al., 2006), we were interested whether both strains also produce the same flexirubin derivative. Crude acetone extracts of C. pinensis cultures were yellow, and a reversible colour shift upon treatment with alkali ( Fig. S10A) suggested the presence of flexirubin as described (Fautz and Reichenbach, 1980). Subsequent purification of the yellow compound and MALDI-MS analysis led to mass of m/z 634.40217 [M] + · and a sum formula of C43H54O4 (calculated m/z 634.40166 [M] + ·; Δppm 0.8). The formation of a radical cation may be explained due to direct absorption of the laser light by the polyene instead of matrixmediated ionization mechanisms. HPLC-UV-ESI-MS (Fig. S10B) confirmed that this compound also absorbs light at 420 nm, a characteristic of flexirubins. When C. pinensis was grown in the presence of L-[methyl-2 H3]methionine, an additional signal with a mass increase of 3 Da was detected in the coloured fraction, indicating the incorporation of a methionine-derived methyl group (Fig. 5AIII). Similarly, a +3 Da mass shift was also observed when L-[ring-2 H4]tyrosine was added to the culture (Fig. 5AIII), indicating its incorporation and the loss of one deuterium at the polyene ring, as expected for a methylation during the biosynthesis. Fragmentation of m/z 634.4 [M] + · by MALDI-MS led to four main fragments (Fig. 5B), already reported for fragmentation of 1 with EI-MS , with m/z 343 and m/z 293 reflecting an ester cleavage, m/z 315.2 as loss of CO (Δ28 Da) and m/z 265.1 as loss of benzene (Δ78 Da) from m/z 343 respectively. MS 3 of m/z 293.2 resulted in a mass spectrum identical to the heterologously produced DAR  (Fuchs et al., 2013) (Fig. 5C). The detected fragments can be explained by the loss of substituents at the aromatic ring (Fig. 5C). MS 3 of m/z 343.2 gave a complex fragmentation pattern with a key fragment of m/z 121.1 (rearrangement to the tropylium cation is expected) and mass shifts of 26 and 28 Da (Fig. 5D), which turned out to be a characteristic of flexirubin polyene fragments. MS 3 of m/z 315.2 and m/z 265.1 showed similar fragments and fragmentation patterns and reflect shorter polyene fragments ( Fig. S10D and E  Therefore, from the colour of the isolated compound, positive KOH test result, incorporation of labelled precursors, and fragmentation in MALDI-MS and ESI-MS, the structure of 1 from C. pinensis is most likely similar to the reported flexirubin from C. filiformes (Achenbach et al., 1976). A remaining task would be to elucidate by means of nuclear magnetic resonance or chemical degradation if the methylation is indeed in ortho-position to the hydroxyl function of the polyene aryl, which is so far the only observed substitution pattern for flexirubins. As this is a time-consuming and challenging process for flexirubins (Fautz and Reichenbach, 1980), it would only be justified if a new derivative is expected. Nevertheless, most of the structure elucidation of 1 has been achieved without radioactive labelling and circumventing largescale fermentations as it has been necessary before (Reichenbach et al., 1974;Achenbach and Kohl, 1978).

Conclusions
In summary, we showed that the genome of C. pinensis contains a flexirubin biosynthesis gene cluster, in which flxA encodes a TAL and flxY encodes a CoA ligase with specificity for 4CA. Only a few bacterial enzymes with PAL or TAL activity were reported before, whereas HAL are widely distributed. Due to the fact that, actually from sequence information alone, the substrate specificity of these enzymes cannot be predicted, the distribution of PAL and TAL enzymes in the Bacteroidetes phylum and in bacteria in general is still unknown. Even less is known about enzymes with 4CL activity from bacteria and FlxY shows a low similarity to the well-known plant 4CL enzymes. The herein reported enzymes from C. pinensis may therefore become already characterized examples for future studies targeting the mechanisms that confer their substrate specificity and may help to develop solid predictions based upon sequence alignments. Additionally, the first two steps in the polyene biosynthesis of flexirubins were elucidated, which have been unknown before. Furthermore, the identification of flexirubin with nowadays widespread and established HPLC-ESI-MS and MALDI-MS methods was reported and might lead to the identification of additional flexirubin producers. Its column chromatography is simple because even traces of pigment can be visually detected and less than 1 μL of enriched fraction is needed for a complete MALDI-MS n analysis, enabling the characterization of flexirubins from newly described bacteria producing this chemotaxonomic marker.

Cultivation of strains
All strains (Table S1) were cultivated on solid Luria-Bertani (LB) medium (5 g l -1 of yeast extract, 10 g l -1 tryptone and 5 g l -1 NaCl) or in liquid LB medium on a rotary shaker at 200 r.p.m. For feeding experiments and extraction of chromosomal DNA, C. pinensis DSM 2588 was cultivated at 30°C and E. coli strains at 37°C. Where appropriate, kanamycin (50 μg ml -1 ) was added to the medium.

Molecular-biological methods
Molecular-biological experiments were performed according to standard procedures (Sambrook et al., 1989). Polymerase chain reactions (PCRs) were performed with the Phusion high-fidelity polymerase (Thermo Fisher Scientific, Waltham, MA) according to the manufacturer's instructions and with oligonucleotides (Table S2) from Sigma-Aldrich (St. Louis, MO, USA). DNA extraction from agarose gels was performed with the GeneJet™ Gel Extraction Kit (Thermo Fisher Scientific) and plasmid isolations with the GeneJet Plasmid Extraction Kit (Thermo Fisher Scientific).

Construction of plasmids for heterologous expression of FlxA and FlxY
The genes cpin_1853 (flxA) and cpin_1877 (flxY ) were amplified by PCR from genomic DNA isolated from C. pinensis. The resulting products, 1853fragment and 1877fragment, were digested with the restriction endonucleases BamHI/XhoI and BamHI/HindIII, respectively, and ligated into the similarly treated pCOLADuet-1, followed by transformation of E. coli DH10B cells. The plasmids pCOLA-1853 and pCOLA-1877 were isolated and the inserts were sequenced at the SeqIT GmbH (Kaiserslautern, Germany).

FlxA in vitro assays
Activity of FlxA with L-histidine, L-tyrosine, L-phenylalanine, L-isoleucine and L-tryptophan was tested in 1 ml of assays [1 μM FlxA, 1 mM L-amino acid and 50 mM Tris (pH 8.0)]. After incubation (12 h, 25°C), assay products were extracted twice with 1 volume of ethylacetate, dried and redissolved in 50 μl of CHCl3 followed by derivatization (55°C, 2 h) with 25 μl of N-methyl-N-trimethylsilyl-trifluoroacetamide. Reaction products were detected on a 7890A gas chromatograph (Agilent, Santa Clara, CA, USA) equipped with a 5975C mass selective detector (Agilent). A DB5ht column (30 m × 250 μm × 0.1 μM; Agilent) was used for the separation of the analytes, with helium as the carrier gas (1 ml min -1 ). The method parameters were as follows: injection volume: 2 μl; inlet temperature: 275°C; injection mode pulsed split (10:1); oven temperature programme: starting temperature 2 min at 130°C, then 5°C min -1 to 240°C followed by an increase to 300°C with a rate of 30°C min -1 or inlet temperature: 250°C; injection mode splitless; oven temperature programme: starting temperature 5 min at 70°C, then 5°C min -1 to 300°C. Ionization of the analyte molecules was carried out by electron impact ionization at 70 eV. For identification of the products, 'Automated Mass Deconvolution and Identification Software' (AMDIS) version 2.64 in combination with the NIST library was used.

Feeding experiments
Four hundred millilitres of liquid LB medium were inoculated 1:100 from an overnight culture of C. pinensis. After 3 h, L-[methyl-2 H3]methionine or L-[ring-2 H4]tyrosine were fed in 1 mM portions to the cultures. Feeding was repeated after 12, 24 and 48 h, resulting in a final concentration of 4 mM. As control, a culture without feeding was cultivated at the same time. Cultures were harvested 12 h after the last feeding by centrifugation at 10,000 r.p.m.

Extraction and isolation of compounds
Pelleted cells of liquid cultures were shaken for 30 min with 50 mL of acetone in the dark, followed by filtration to remove cell debris. The extracts were dried under reduced pressure. The raw extracts were fractionated using silica gel columns with hexane/ethylacetate (4:1; 1% acetic acid). Coloured fractions were dried under reduced pressure. All samples were stored dry at −20°C in the dark and solved in 10-50 μl of acetonitrile (ACN) before further analysis.

HPLC-MS
HPLC-ESI-MS analysis was performed with a Dionex UltiMate 3000 system coupled to a Bruker AmaZon X mass spectrometer and an Acquity UPLC BEH C18 1.7 μm RP column (Waters, Milford, MA, USA) as described previously (Reimer et al., 2011).

MALDI-MS
For MALDI-analysis, samples were mixed 1:2 with 1 μl of a 20 mM 4-chloro-α-cyanocinnamic acid in 70% ACN with 0.1% trifluoroacetic acid, spotted onto a polished stainless steel target and air-dried. MALDI-MS n analysis was performed with a MALDI LTQ Orbitrap XL (Thermo Fisher Scientific) equipped with a nitrogen laser at 337 nm, as described previously (Fuchs et al., 2012). Spectra were analysed and possible sum formulas were calculated using Qual Browser (version 2.0.7; Thermo Fisher Scientific).

Genome mining
For identification of DarAB gene clusters, the primary structures of the DarAB homologues from C. pinensis DSM 2588 were used for a BLAST-P identification (protein-protein-BLAST) of homologues in F. johnsoniae UW101. The reported biosynthesis gene cluster from F. johnsoniae UW101 (McBride et al., 2009) was used in a STRING analysis (version 9.0) (Franceschini et al., 2013) for the identification of the putative flexirubin polyene cluster in C. pinensis. Gene colours in Fig. 2 are based upon NCBI annotation or domain-guided annotations from BLAST-P with primary sequences of C. pinensis genes as template (see Table S4 and S5). Grey connections between genes depict an identity ≥ 40%. Fig. S1. ClustalW multiple alignment of primary sequences of FlxA from C. pinensis with those of histidine ammonialyases (HALs), phenylalanine ammonia-lyases (PALs) and tyrosine ammonia-lyases (TALs) with known substrate specificity. Sequences used for the alignment were abbreviated as follows (with GenBank accession numbers in parentheses): HAL P. putida, HAL Pseudomonas putida (P21310); HAL S. griseus, HAL Streptomyces griseus (AAA26769); PAL P. crispum, PAL Petroselinum crispum (P24481); PAL P. luminescens, PAL Photorhabdus luminescens sp. laumondii TT01 (NP_929491); PAL S. maritimus, PAL Streptomyces maritimus (AAF81735); TAL C. pinensis, TAL Chitinophaga pinensis (YP_003121550); TAL R. sphaeroides, TAL Rhodobacter sphaeroides 2.4.1 (YP_355075); TAL S. espanaensis, TAL Saccharothrix espanaensis (ABC88669); TAL S. sp. Tü 4128, TAL Streptomyces sp. Tü 4128 (AEV23249). The selectivity switch reported by Watts et al. (2006) is boxed in motive 1. The residue at the first position (here residue 140) varied between Phe (F) and His (H) in PAL sequences, whereas TAL or PAL with TAL activity have a His (H) at the first position. Furthermore, a second conserved residue was reported (Berner et al., 2006), which is boxed in motive 2 (here residue 507) and was always Glu (E) for HAL and Gln (Q) for TAL or PAL. The alignment was performed using Geneious 6.1.7 ClustalW default settings. Coloured residues indicate a similarity ≥ 75%.  Chromatogram of an assay containing FlxA and L-phenylalanine (I), the mass spectrum of the product at 6.1 min from the above chromatogram (II) and the mass spectrum of E-cinnamic acid standard (III). Fig. S4. Photometric FlxA enzyme assays. A. Endpoint assay detecting the enzymatic conversion of L-tyrosine (blue), L-phenylalanine (red) and L-histidine (green) by measuring the absorption increase of 4-coumaric acid (310 nm), E-cinnamic acid (275 nm) or urocanic acid (277 nm) respectively. B. pH dependence of FlxA activity with L-tyrosine (I) and L-phenylalanine (II). Temperature dependence of FlxA activity with K-tyrosine (III) and L-phenylalanine (IV). TAL: tyrosine ammonia-lyase; PAL: phenylalanine ammonia-lyase; HAL: histidine ammonia-lyase. Fig. S5. Michaelis-Menten kinetics of FlxA. Substrate saturation plots of FlxA with L-tyrosine (A) and L-phenylalanine (B) and their respective linearization as Lineweaver-Burk (C, D) or Hanes-Woolf plots(E, D). TAL: tyrosine ammonia-lyase; PAL: phenylalanine ammonia-lyase. Fig. S6. ClustalW multiple alignment of primary sequences of FlxY from Chitinophaga pinensis with three 4-coumarate-CoA ligases (4CL) from Arabidopsis thaliana and the E-cinnamate-CoA ligase (CCL) ScCCL from Streptomyces coelicolor A3(2). Sequences used for the alignment were abbreviated as follows (with GenBank accession numbers in parentheses): At4CL1, A. thaliana 4CL1 (AAA82888.1); At4CL2, A. thaliana 4CL2 (AF106086_1); At4CL3, A. thaliana 4CL3 (AF106088_1); ScCCL, S. coelicolor A3 (2) CCL (CAB95894); Cp4CL, C. pinensis 4CL (YP_003121574.1). Two conserved peptide boxes from 4CL are boxed in red and nine residues that are suggested to form the At4CL substrate binding pocket are marked by red triangles (Kaneko et al., 2003). The alignment was performed using Geneious 6.1.7 ClustalW default settings. Coloured residues indicate a similarity ≥ 75%. Fig. S7. HPLC-MS analysis of enzyme assays with FlxY. A. Mass spectra of CoA (I), 4-coumaroyl-CoA standard (II) and products detected in enzyme assays containing FlxY and 4-coumaric acid (III) or E-cinnamic acid (IV) as substrates. B. Theoretical fragmentation tree explaining the above mass spectra with randomly positioned charges. Fig. S8. HPLC-MS analysis of the substrate specificity of FlxY. The structures of the substrates, the EIC from CoA (green) and the EIC of expected substrate-CoA thioesters (blue) are shown. Fig. S9. Results of enzyme assays with FlxY. pH (A) and temperature optimum (B) of the coupled enzyme assay. Substrate saturation plots of FlxY with 4-coumaric acid (C) and E-cinnamic acid (D) and their respective linearization as Lineweaver-Burk (E, F) or Hanes-Woolf plots (G, H). feeding experiments with d3-methionine (II) and d4-tyrosine (III). The resulting fragments m/z 343.2 for wild type and m/z 346.2 or m/z 346.3 for the feeding experiments were further fragmented and the corresponding MS 3 mass spectra are depicted on the right side. Table S1. Strains and plasmids used in this work. Table S2. Primers and PCR products used in this study. Table S3. High-resolution MALDI-orbitrap-MS of CoAthioesters from FlxY substrate assays. Table S4. Predicted gene clusters for flexirubin biosynthesis in C. pinensis. Domain-guided annotation is based on conserved domains detected by BLAST-P of C. pinensis DSM 2588 primary sequences against the genome of F. johnsoniae UW101. Table S5. Gene cluster for flexirubin biosynthesis in F. johnsoniae UW101. Domain-guided annotation is based on conserved domains detected by BLAST-P. Table S6. Comparison of the enzymatic properties of bacterial 4CLs (4-coumarate-CoA ligase) and CCLs (cinnamate-CoA ligase). Values were obtained from assays with 4-coumarate (4CA) and E-cinnamate (CA) as substrates.