Discovery and Biosynthesis of Bolagladins: Unusual Lipodepsipeptides from Burkholderia gladioli Clinical Isolates

Abstract Two Burkholderia gladioli strains isolated from the lungs of cystic fibrosis patients were found to produce unusual lipodepsipeptides containing a unique citrate‐derived fatty acid and a rare dehydro‐β‐alanine residue. The gene cluster responsible for their biosynthesis was identified by bioinformatics and insertional mutagenesis. In‐frame deletions and enzyme activity assays were used to investigate the functions of several proteins encoded by the biosynthetic gene cluster, which was found in the genomes of about 45 % of B. gladioli isolates, suggesting that its metabolic products play an important role in the growth and/or survival of the species. The Chrome Azurol S assay indicated that these metabolites bind ferric iron, which suppresses their production when added to the growth medium. Moreover, a gene encoding a TonB‐dependent ferric‐siderophore receptor is adjacent to the biosynthetic genes, suggesting that these metabolites may function as siderophores in B. gladioli.


Introduction
Iron is an essential element for most organisms.I n biological systems,i ron is mainly present in two oxidation states:ferrous and ferric. [1] Theinterconversion of the ferrous and ferric oxidation states is essential for av ariety of redox reactions and electron transfer processes in cells. [2] However, despite the relatively high abundance of iron in the Earths crust, it is not readily bioavailable.T his is because it is predominantly in the ferric form, which has low aqueous solubility. [3] Moreover,s equestration by proteins such as hemoglobin, ferritin, transferrin, and lactoferrin also contributes to low iron availability in mammals. [3] As aconsequence, microorganisms have evolved several strategies for obtaining iron from the environment and their hosts.O ne strategy involves the production of siderophores,s pecialised metabolites with ah igh affinity for ferric iron. [4] In Gram-negative bacteria, ferric-siderophore complexes are typically transported into the periplasm by TonB-dependent outer membrane receptors.Periplasmic binding proteins then shuttle the ferric-siderophore complexes to the inner membrane,w here ATP-binding cassette (ABC) transporters transfer them into the cytoplasm. [5] Once in the cytoplasm,iron is released from the ferric-siderophore complex by reduction to the ferrous form and/or hydrolytic cleavage of the ligand. [5a] Several investigations of pathogenic microorganisms have identified as trong link between siderophore production and virulence. [6] In the early stages of infection, siderophores play ak ey role in sequestering iron from host tissues.T he identification of siderophores produced by pathogenic microorganisms is therefore an important field of study because interference with siderophore-mediated iron uptake is an attractive strategy for attenuating virulence. [7] Thed iscovery of novel siderophores is also relevant to other therapeutic applications.F or example,d esferrioxamine B( marketed as Desferal), asiderophore produced by numerous Streptomyces species, [8] is used to treat iron overload resulting from multiple blood transfusions and aluminium toxicity due to dialysis. However,b ecause Desferal is administered by injection and has multiple side effects,o rally available alternatives with fewer side effects are actively being sought. [9] In addition to clinical uses,s iderophores show promise for environmental and biotechnological applications. [10] Burkholderia is ag enus of highly diverse Gram-negative bacteria. Species belonging to this genus have been isolated from soil, water, plants,a nimals and humans. [11] Several investigations have shown that they are prolific producers of specialised metabolites, [12] and genome sequence analyses have highlighted their underexplored biosynthetic potential. [13] To date,p athogenic Burkholderia species have been reported to produce siderophores belonging to four distinct structural classes (ornibactins/malleobactins,p yochelin, cepabactin, and cepaciachelin;F igure 1). [14] Members of the cepacia complex and pseudomallei group produce one or more of these siderophores.Incontrast, Burkholderia gladioli does not produce any of these metabolites and was long believed to employ siderophore-independent mechanisms for iron acquisition. [15] However,H ertweck and co-workers recently reported that three B. gladioli environmental isolates produce the diazeniumdiolate siderophore gladiobactin. [16] We have recently discovered that B. gladioli BCC0238, isolated from the sputum of ac hild with cystic fibrosis (CF), produces several specialised metabolites.T hese include gladiolin, anovel macrolide with potent activity against drug resistant Mycobacterium tuberculosis clinical isolates, [12i] and icosalide A1, [17] an asymmetric lipopeptodiolide antibiotic originally isolated from an Aureobasidium species fungus, [18] but subsequently shown to be produced by various strains of B. gladioli including one associated with the fungus. [17,19] Analysis of the B. gladioli BCC0238 complete genome sequence identified several gene clusters encoding cryptic nonribosomal peptide synthetase (NRPS), polyketide synthase (PKS) and hybrid NRPS-PKS assembly lines. [12g] Here we report the discovery of bolagladins Aa nd B, novel lipodepsipeptides containing au nique citrate-derived fatty acid and ar are dehydro-b-alanine residue,a st he metabolic products of one of these gene clusters.Acombination of comparative bioinformatics analyses,targeted gene deletions, and enzyme activity assays elucidated several key steps in the biosynthesis of these unusual natural products,w hich we propose based on several lines of evidence may function as siderophores.

Results and Discussion
Isolation and Structure Elucidation of the Bolagladins B. gladioli BCC0238 produces gladiolin and icosalide A1 when grown on as olid minimal medium containing glycerol as the sole carbon source. [12i, 17] Twon ovel metabolites were identified by UHPLC-ESI-Q-TOF-MS analysis of extracts from cultures of ag ladiolin nonproducing mutant of B. gladioli BCC0238 (with an in-frame deletion in the region of gbnD1 encoding the ER domain) grown on BSM medium containing ribose as an additional carbon source alongside glycerol. Them etabolites were isolated as amorphous solids from ethyl acetate extracts of 3day cultures (1 Lt otal volume) using semipreparative HPLC,a nd their planar structures were elucidated using ac ombination of HR-ESI-MS and 1D/2D NMR experiments ( Figure 2; Tables S2 and S3, Figures S1-S14  9.91, and 9.99, respectively) in the COSY spectrum of bolagladin A 1.F urther analysis of 1 HNMR and COSY spectra identified four a-amino acid spin systems,w hich in conjunction with HSQC and HMBC  data, were assigned to av aline residue,ahomoserine (Hse) residue and two serine residues in the sequence Ser-Hse-Val-Ser ( Figure 2). A 3 J CH correlation from the b-protons of the Nterminal Ser (N-Ser) residue (d H 4.73 and 4.90) to the carbonyl carbon of the C-terminal Ser (C-Ser) residue (d C 170.0) established the presence of at etradepsipeptide (Figure 2). Analysis of the COSY and HMBC data also revealed ad ehydro-b-alanine (Dba) residue,w ith ac oupling constant of 14.0 Hz between the a and b protons indicating that the double bond is E-configured. Theonly other natural products reported to contain aD ba residue are the enamidonins isolated from Streptomyces species. [20] 2 J CH correlations were observed between the NÀHproton of the N-Ser residue in the tetradepsipeptide (d H 9.99) and the carbonyl carbon of the Dba residue (d C 168.3), and the N À Hp roton of Dba residue (d H 11.56) and the carbonyl carbon of afatty acid residue (d C 165.1). This indicates that the Dba residue links the fatty acid residue to the tetradepsipeptide.The fatty acid residue is very unusual, with Z-configured double bonds spanning C2-C3 and C11-C12 (based on CH=CH coupling constants of 11.5 and 13.5 Hz, respectively), and two carboxylic acid groups and amethoxy group appended to its tail ( Figure 2).
Theabsolute configurations of the a-amino acid residues in bolagladin A 1 were determined as l-Ser-d-Val-l-Hse-l-Ser using Marfeysm ethod [21] ( Figure S15). Comparison of the NMR spectroscopic data for bolagladins A 1 and B 2 showed that the Va lresidue in the former is substituted by an Ile residue in the latter, consistent with the difference of CH 2 in the molecular formulae.The absolute configurations of the a-amino acids in bolagladin B (2)were similarly shown to be l-Ser-d-allo-Ile-l-Hse-l-Ser by Marfeysm ethod (Figure S15), using aC 3c olumn instead of aC 18 column to separate the Marfeysd erivatives of d-Ile and d-allo-Ile ( Figure S15). [22] Overall, the structures of bolagladins A 1 and B 2 are highly unusual, containing au nique fatty acid residue with several polar functional groups appended to its tail, linked via arare Dba residue to adepsitetrapeptide.W ethus sought to develop an understanding of bolagladin biosynthesis,focusing on the origin of the starter unit for the fatty acid assembly and the mechanism of Dba incorporation.

Bolagladin Biosynthesis
Bioinformatics analysis of the B. gladioli BCC0238 genome sequence identified the putative bolagladin (bol) biosynthetic gene cluster (Figure 3a nd Table S4), which contains ag ene (bolH)e ncoding at etramodular NRPS. Sequence analysis of the adenylation (A) domains in this NRPS indicated that modules 1-4 activate l-Ser, l-Val/l-Ile, l-Hse and l-Ser,r espectively (Table S5), [23] suggesting the NRPS assembles the depsipeptide portion of the bolagladins. Similarly,phylogenetic analysis of the BolH condensation (C) domains indicated that they belong to the following groups. [24] Module 1: chain initiating (C I -links external acyl donor and l-configured acyl acceptor);modules 2and 4: chain elongating ( L C L -links l-configured acyl donor and acceptor);a nd module 3: bifunctional condensation/epimerization (C E -epimerizes acyl donor to d-configuration and links with lconfigured acyl acceptor; Figure 3). This is consistent with the structure elucidation data, which indicate that the bolagladins contain l-Ser, l-Hse and d-Val/d-allo-Ile.
To verify the involvement of this gene cluster in bolagladin biosynthesis,weinactivated bolH in B. gladioli BCC0238 DgbnD1_ER via insertional mutagenesis.L C-MS analysis confirmed that the production of bolagladins A 1 and B 2 was abolished in the mutant (Figure 4).
Putative functions were assigned to the proteins encoded by the bolA-bolT genes flanking bolH on the basis of sequence analyses (Table S4). This enabled us to propose aplausible pathway for bolagladin biosynthesis (Figure 3). To validate this pathway experimentally,weenvisioned examining the effect of in-frame deletions in key biosynthetic genes on bolagladin production. However,t his proved to be challenging and time consuming in B. gladioli BCC0238 (see Supporting Information), so we screened other genomesequenced B. gladioli isolates containing the bol locus (see below) to establish whether they i) produce the bolagladins and ii)are amenable to construction of in-frame deletions. B. gladioli BCC1622 (also isolated from aC Fl ung infection) was found to meet both criteria and an analogous gladiolin nonproducing mutant of this strain (B. gladioli BCC1622 DgbnD1_ER)w as created to enable in-depth functional studies of selected bolagladin biosynthetic genes.BolR shows sequence similarity to citrate synthase,akey enzyme in the Krebs cycle,which catalyzes the condensation of acetyl-CoA with oxaloactetate and hydrolysis of the resulting thioester to form citrate (Figure 5a). [25] This suggests that citryl-CoA may serve as the starter unit for assembly of the unusual fatty acid residue incorporated into the bolagladins (Figure 3). To verify the function of BolR, we overproduced it in E. coli as an Nterminal His 6 -fusion and purified it to homogeneity (see Supporting Information). Incubation of the purified protein with oxaloacetate (4)a nd acetyl-CoA (5)f or 2h at room temperature resulted in complete conversion to products with molecular formulae corresponding to citric acid (6)a nd coenzyme A 7 (Figure 5b).
To probe the role played by citrate in bolagladin biosynthesis,w ec onstructed an in-frame deletion in bolR. Ther esulting mutant was unable to assemble bolagladins A and B, but produced an ew metabolite with the molecular formula C 37 H 59 N 5 O 12 (calculated for C 37 H 60 N 5 O 12 + :7 66.4233, found:7 66.4238;F igures 6a nd S16). Theproduction level of this metabolite was insufficient for NMR spectroscopic analysis.W etherefore conducted LC-MS/MS analyses,which indicated the metabolite has the same depsipeptide core as bolagladin B 2 (molecular formula C 40 H 63 N 5 O 14 ;F igures 6 and S16) and that the structural differences must lie in the fatty acid and/or Dba residues.H owever, because the new metabolite contains three fewer carbon atoms,f our fewer hydrogen atoms and two fewer oxygen atoms than bolagladin B, but the same number of nitrogen atoms,t hese differences cannot be due to modification or loss of the Dba residue.T he most plausible structure for this metabolite is 3, in which the fatty acid residue is derived from an oxaloacetyl-CoA starter unit that is subsequently reduced (Figure 6). This suggests that BolR may catalyze the condensation of acetyl-CoA with the keto group of oxaloacetyl-CoA in bolagadin biosynthesis ( Figure S17), but further experiments will be required to confirm this.
We postulated that BolS,which shows sequence similarity to S-adenosyl-l-methione (SAM) dependent methyltransferases,c atalyzes methylation of the hydroxyl group in the citrate-derived starter unit at some point during bolagladin biosynthesis.T ot est this hypothesis,w ec onstructed an inframe deletion in bolS.  (Figures 7a nd S16). LC-MS/MS analyses indicated that these compounds are congeners of 1 and 2,l acking aC H 2 group from the fatty acid or Dba residues ( Figure S17). Although production levels of these metabolites were low,w ew ere able to purify as ufficient quantity of the bolagladin Bcongener for NMR spectroscopic analysis (Table S6, Figures S18-S23). Differences were ob- served in the 13 Cchemical shifts of the resonances due to C16, C17, C18, C19 and C20 in the fatty acid residue and the signals attributed to the O-methyl group in bolagladin Bwere absent from the 1 Ha nd 13 Cs pectra. We therefore conclude that the two metabolites accumulated in the bolS mutant are des-methyl-bolagladins A 8 and B 9.T he low levels of 8 and 9 produced by the mutant suggest O-methylation of the citratederived starter unit is an early step in bolagladin biosynthesis, rather than al ate-stage modification. However,t he precise timing of this reaction remains to be determined.
Citrate (or its O-methylated derivative) is proposed to be converted to the corresponding coenzyme At hioester by BolB,w hich shows similarity to acyl-CoAs ynthetases (Figure 3and Table S4). Either BolM or BolP,both of which show similarity to b-ketoacyls ynthase III (KAS III) enzymes that typically initiate fatty acid biosynthesis in bacteria (Table S4), [26] could then catalyse the elongation of citryl-CoA (or its O-methylated derivative) with am alonyl group attached to the primary metabolic fatty acid synthase (FAS) acyl carrier protein (ACP) (Figure 3). Interestingly,t he conserved active site Cys residue is mutated to Ser in BolM and ThrinBolP.The functional significance of this is unclear, but an analogous Cys to Ser mutation is observed in DpsC, aK AS III homologue that has been reported to initiate assembly of the daunorubicin polyketide chain using ap ropionyl-CoA starter unit. [27] Further processing of the bketothioester resulting from elongation of (O-methyl)-citryl-   CoA with malonyl-ACP by the primary metabolic FASwould afford the saturated O-methyl-citrate-primed fatty acyl-ACP thioester 10 ( Figure 3). The1 1, 12 and 2, 3d ouble bonds are likely introduced into this thioester by BolF and BolL, which are similar to membrane-associated fatty acid desaturases and acyl-ACPd esaturases,r espectively (Figure 3a nd Table S4), completing the assembly of the unusual fatty acid residue incorporated into the bolagladins.
TheDba residue of the bolagladins is postulated to derive from l-aspartate,which we propose undergoes adenylation of its b-carboxyl group catalyzed by BolO,f ollowed by transfer onto BolC,aputative freestanding ACP ( Figure 3). BolO shows similarity to VinN, [28] which catalyzes adenylation of the b-carboxyl group of b-methyl-aspartate and subsequent transfer to the freestanding ACPV inL in the biosynthesis of vicenistatin. Moreover,t he Asp230 and Ser299/Arg331 residues in VinN,w hich are proposed to play ak ey role in recognition of the a-amino and a-carboxyl groups of bmethyl-aspartate,r espectively,a re conserved in BolO (Figure S24). In vicenistatin biosynthesis,the pyridoxal phosphate (PLP)-dependent enzyme VinO catalyses decarboxylation of the b-methyl-isoaspartyl-VinL thioester to form the corresponding a-methyl-b-alanyl thioester. [24] We propose that BolN,w hich shows sequence similarity to BtrK, aP LPdependent enzyme that catalyzes decarboxylation of an isoglutamyl-ACP thioester in butirosin biosynthesis,p lays an analogous role in bolagladin biosynthesis (Figure 3). In-frame deletion of bolN abolished bolagladin production and it could not be restored by feeding b-alanine to the mutant. This is consistent with the proposed roles of BolO,B olC and BolN and rules out b-alanine as an intermediate in the biosynthetic pathway.T he amino group of the b-alanyl-BolC thioester is proposed to be condensed with the citrate-primed fatty acyl thioester by one of the KAS III homologues BolM or BolP (Figure 3).
Thes equence similarity of BolQ to acyl-CoA dehydrogenases suggested it might be responsible for formation of the Dba residue via desaturation of the N-acyl-b-alanyl-BolC thioester. To test this hypothesis,w ec reated an in-frame deletion in bolQ,w hich abolished the production of bolagladins Aa nd Ba nd led to the accumulation of two new metabolites with the molecular formulae C 39 H 63 N 5 O 14 and C 40 H 65 N 5 O 14 (Figure 8). Purification and NMR spectroscopic analysis showed these metabolites are dihydro-bolagladins A 11 and B 12,inwhich the Dba residue has been replaced by balanine (Tables S7 and S8, and Figures S25-S35). This is consistent with the proposed function of BolQ as a N-acyl-balanyl thioester desaturase,a lthough on the basis of these data the possibility of BolQ converting the b-alanine residue to Dba at al ater stage in bolagladin biosynthesis cannot be ruled out.

The bol Locus is Widely Conserved in B. gladioli
Al ocal nucleotide BLAST search of 1318 genomes representing Burkholderia, Paraburkholderia and Caballeronia species showed B. gladioli is the only species containing the bolagladin biosynthetic gene cluster (as indicated by the presence of bolH). [29] Read mapping of paired-end Illumina reads from 234 B. gladioli strains against the bol gene cluster revealed that it is present in 105 strains.H owever,t here is evidence of gene substitutions and deletions in the regions  flanking the cluster (Figure 9). The bolU gene,e ncoding ap utative TonB-dependent ferric-siderophore outer membrane receptor,i sa bsent from five of the 105 B. gladioli genomes (BCC1837, BCC1843, BCC1861, BCC1870 and BCC1871) containing the bolagladin biosynthetic gene cluster and is present in 52 %(67 of 129) of genomes lacking it.

Biological Function of the Bolagladins
Bolagladins A( 1)a nd B( 2)s howed no activity at aconcentration of 64 mgmL À1 in disc diffusion assays against any of the ESKAPE panel of bacterial pathogens, [30] Mycobacterium bovis BCG,o rCandida albicans. The3 -methoxy-1,4-dicarboxylate moiety of the bolagladins is reminiscent of ferric-iron-chelating citrate residues in several siderophores, such as achromobactin, vibrioferrin, and staphyloferrin B. [31] Moreover, bolU flanking the left end of the bolagladin biosynthetic gene cluster encodes aputative TonB-dependent ferric-siderophore outer membrane receptor.W et hus hypothesised that the bolagladins could function as siderophores.A lthough ferric complexes of the bolagladins were too labile to observe in ESI-Q-TOF MS analyses,the chrome azurol S( CAS) assay indicated they are able to sequester ferric iron, albeit at higher concentrations than typical trishydroxamate siderophores ( Figure S36). [4] Siderophore production in microbes is usually regulated by iron availability.I ni ron-deficient media, production is upregulated, whereas in iron replete media it is suppressed. Accordingly,the level of bolagladin production by B. gladioli BCC1622 DgbnD1_ER decreased as increasing concentrations of ferric iron were added to the medium ( Figure S36). Combined with the results of the CASa ssay,t hese data indicate that the bolagladins may function as siderophores in B. gladioli.
Siderophores often contribute to virulence in pathogenic bacteria. Thus,w eu sed a Galleria mellonella wax moth infection model to investigate whether the bolagladins are av irulence factor in B. gladioli. Ornibactin was shown to contribute to virulence in Burkholderia cenocepacia using this and other models. [15,32] However,n oa ttenuation of virulence towards the wax moth larvae was observed in the DbolN mutant of B. gladioli BCC1622 DgbnD1_ER,r elative to the parental strain ( Figure S37).
As imilar observation was reported for pyochelin deficient mutants of B. cenocepacia H111, [32] leading to the hypothesis that some Burkholderia siderophores may play ar ole in metal homeostasis rather than virulence. [33] To investigate this hypothesis, B. cenocepacia H111 pyochelin and ornibactin nonproducing mutants were grown alongside the wild-type strain in media containing various metal salts (10-50 mM). Them etal salts did not affect the wild-type strain, but were toxic to the siderophore nonproducing mutants,i ndicating that pyochelin and ornibactin protect B. cenocepacia against metal ion toxicity. [33] This finding was further supported by supplementing the medium with pyochelin and ornibactin, which reversed toxicity in the siderophore nonproducing mutants. [33] We thus compared the tolerance of the B. gladioli BCC1622 DgbnD1_ER and DgbnD1_ER-DbolN strains to salts of aluminium, zinc,cobalt, copper, cadmium, nickel and lead at various concentrations up to 100 mM. No decrease in metal ion tolerance was observed for the bolagladin nonproducing mutant relative to parent strain and the results of these experiments indicated that B. gladioli is much less susceptible to metal ion toxicity than B. cenocepacia. Further experiments demonstrated that B. gladioli could readily grow in media containing 1mMN i 2+ ,A l 2+ and Zn 2+ .A dditional work is required to establish the genetic basis for this high degree of metal ion tolerance.

Conclusion
Bacteria belonging to the Burkholderia genus are increasingly being recognised as an underexplored source of novel natural products. [34] Using ac ombination of carbon source modification and inactivation of the biosynthetic pathway for gladiolin, which is produced at high titre and interferes with the detection of lower abundance metabolites,w eh ave identified bolagladins A(1)and (2)asthe metabolic product of acryptic nonribosomal peptide biosynthetic gene cluster in B. gladioli. Similar approaches may prove useful for identifying the products of other cryptic Burkholderia biosynthetic gene clusters.
Theb olagladins have very unusual structures,c onsisting of au nique fatty acid residue with several polar functional groups appended to its tail, linked via avery rare Dba residue to ad epsitetrapeptide.I dentification of the bolagladin biosynthetic gene cluster in B. gladioli BCC0238 and BCC1622 enabled us to probe the origin of these unusual structural features using ac ombination of detailed bioinformatics analysis,g ene deletions and enzyme activity assays. Based on these studies,wepropose that the fatty acid residue originates from ac itryl-CoA starter unit. To our knowledge, there is no precedent for the utilisation of citryl-CoA as astarter unit in fatty acid or polyketide biosynthesis.W ealso hypothesise that the Dba residue derives from loading of aspartate onto af reestanding ACP. Ther esulting isoaspartyl thioester is proposed to undergo decarboxylation and Nacylation, prior to being desaturated by af lavin-dependent dehydrogenase,r esulting in an N-acyl-Dba thioester starter unit for the NRPS that assembles the depsitetrapeptide. While loading of aspartate/b-methyl-aspartate onto as tandalone ACP, followed by decarboxylation and N-acylation, is aw ell-established mechanism for provision of b-aminoacyl starter units to type Im odular PKSs, [35] to our knowledge there is no precedent for this mechanism being employed to provide such starter units to NRPSs.T hus,o ur findings expand the scope of b-aminoacyl starter unit biosynthetic machinery,w hich may prove useful in bioengineering approaches to natural product diversification.
Theb olagladin biosynthetic gene cluster is present in approximately 45 %ofB. gladioli genomes,but was not found in other Burkholderia species.T his suggests the bolagladins play an important role in the adaption of B. gladioli to its environmental niches.S everal lines of evidence indicate that the bolagladins may function as siderophores,f urther challenging the assumption that B. gladioli employs siderophoreindependent mechanisms for iron acquisition. [15][16] While the mode of ferric iron binding to the bolagladins remains to be established, it seems likely that the two citrate-derived carboxy and methoxy groups in the unusual fatty acid residue, and one or both of the side-chain hydroxyl groups in the depsipetide are involved. Siderophores are known to play roles in virulence and metal ion tolerance in other Burkholderia species,b ut a Galleria mellonella infection model did not provide evidence that the bolagladins contribute to virulence in B. gladioli and bolagladin-deficient mutants did not show reduced tolerance towards metal ions.F urther studies are therefore required to develop ab etter under-standing of the adaptive benefit the bolagladins confer on B. gladioli.