Halogenated Metal-Binding Compounds from Shipworm Symbionts.

Bacteria use small molecules to impose strict regulation over the acquisition, uptake, and sequestration of transition metal ions. Low-abundance nutrient metals, such as Fe(III), need to be scavenged from the environment by high-affinity chelating molecules called siderophores. Conversely, metal ions that become toxic at high concentrations need to be sequestered and detoxified. Often, bacteria produce a suite of compounds that bind various metal ions at different affinities in order to maintain homeostasis. Turnerbactin, a triscatecholate siderophore isolated from the intracellular shipworm symbiont Teredinibacter turnerae T7901, is responsible for iron regulation and uptake. Herein, another series of compounds are described that complex with iron, copper, and molybdenum in solution. Teredinibactins belong to a class of metal-binding molecules that utilize a phenolate-thiazoline moiety in the coordination of metal ions. In contrast to other compounds in this class, such as yersiniabactin, the phenyl ring is decorated with a 2,4-dihydroxy-3-halo substitution pattern. UV-vis absorption spectroscopy based titration experiments with CuCl2 show the formation of an intermediate complex at substoichiometric concentrations and conversion to a copper-bound complex at 1:1 molar equiv.

Metal ion regulation is essential for bacteria. Transition metal ions are necessary as cofactors for essential cellular processes, but they can also become toxic at high concentrations.[1,2] This balance requires systems for the acquisition and export of metal ions to maintain ideal physiological concentrations. The most thoroughly described systems in bacteria involve iron regulation. Fe(III) is essential for bacterial survival, but bioavailable iron(III) is limited in the environment due to poor solubility.[3] Low molecular weight metabolites that chelate Fe(III), known as siderophores, are able to scavenge the nutrient with extremely high affinity, then are recognized for uptake into the cell by specific cell surface receptors. These siderophore systems are upregulated in Fe(III)-limited environments, which allows the organism to maintain suitable levels for biological processes including DNA and RNA synthesis and electron transfer.[4−8] A number of compounds originally described as iron-complexing siderophores have subsequently been found to complex copper, zinc, cobalt, or other transition metals as well.[9]

Though less extensively researched, complete systems for the regulation of non-iron transition metal ions have also been described. These were largely studied in the context of copper detoxification in model organisms, rather than the acquisition and storage of ions in environmental or clinical settings.[10] However, the discovery of an analogous copper acquisition system surrounding the small-molecule methanobactin in the methane-oxidizing microbe Methylosinus trichosporium OB3b was a major step in recognizing the importance of copper-chelating compounds.[11] Subsequent studies have elucidated the integral role small-molecule metabolites play in the regulation of Cu(II) ions, particularly in relation to pathogenesis at host–microbe interfaces. For instance, the phenolate-thiazoline compound yersiniabactin was originally isolated as an iron-complexing molecule described as a virulence factor in Yersinia enterocolytica,[12] but has since been extensively researched for its role in protecting pathogens from copper toxicity during host infection.[10] Despite the growing evidence for the role of siderophore and copper-binding molecules in pathogenesis, little attention has been paid to metal regulation in commensal or nutritional symbioses.

Wood-boring shipworms are marine bivalve mollusks that host a distinct symbiotic microbiota in their gills. One of the members of many shipworm gill microbial communities, Teredinibacter turnerae, has become recognized as a prolific producer of natural products.[13,14] One of the described molecules, turnerbactin, is a trischatecholate NRPS product that was chemically characterized as a siderophore.[15] Many microbes are known to synthesize and secrete multiple small molecules with a range of affinities for different metal ions in response to their environment. Herein, a second metal-binding compound secreted by T. turnerae T7901 is described. Teredinibactin A (1) forms complexes with copper, iron, and molybdenum in solution and is a member of the phenolate-thiazoline structural family, which includes pyochelin and yersiniabactin.

Results and Discussion

Isolation and Structure Elucidation

Teredinibacter turnerae T7901 was grown in Shipworm Basal Media (SBM)[16] (8 L) for 7 days. The cells were removed by centrifugation, and the culture medium was extracted with hydrophobic polystyrene resin. The resin was washed to remove salts and media components, then eluted by a stepwise gradient with aqueous MeOH. The 75% and 100% elutions contained a series of shared peaks by HPLC-DAD and LCMS analysis and were combined and subjected to semipreparative HPLC to isolate individual compounds.

HRESIMS of 1 suggested a molecular formula of C12H11ClN2O5S, based on the presence of a protonated molecule at m/z 331.0153 and an isotopic distribution indicating incorporation of a chlorine (Figure S7). The 1H NMR spectrum showed nine distinct signals and one severely broadened signal. The gHSQCAD correlations indicated that there were five protonated carbons, including two in the aromatic range, two methylene bridge carbons, and one methine proton in the range of an amino acid alpha proton (δH 5.33). Remaining proton signals were not coupled to carbon signals and had chemicals shifts consistent with amine NH (δH 8.53) and phenolic or carboxylic acid (δH 13.09, 12.6) functional groups.

The gCOSY spectrum showed correlations making up three separate spin systems. A correlation from the triplet NH (δH 8.53, J = 5.9 Hz) proton to a doublet signal with an integration of 2 at δH 3.81 was consistent with a glycine residue. This was supported by gHMBCAD correlations from the methylene protons to two carbonyl carbons at δC 169.6 and 170.9. The next spin system included two diastereotopic methylene protons at δH 3.66 and 3.54 coupled to the triplet methine proton at δH 5.33, which was a likely alpha proton. gHMBCAD correlations from the signal at δH 5.33 to the carbonyl at δC 169.6 and a deshielded carbon at δC 172.4, along with the known incorporation of a sulfur atom from the chemical formula, was consistent with a thiazoline ring.

The carbonyl at δC 169.6 that showed correlations in the HMBC to both the thiazoline methine and glycine methylene was used to link these two spin systems together into a modified dipeptide substructure. The carbonyl at δC 170.9 only has a correlation from the glycine residue, and its chemical shift is consistent with that of a terminal carboxylic acid, which is also supported by the broadened singlet proton at δH 13.09. This substructure has a formula of C6H7N2O3S, leaving a remainder of six carbons, four protons, two oxygens, and one chlorine.

The last spin system included the two aromatic protons at δH 6.6 and 7.25. Interestingly, the carbon chemical shifts of these two positions are δC 107.6 and 129.6, respectively, according to the gHSQCAD. Furthermore, the gHMBCAD and 13C spectra showed that the four remaining aromatic carbons were at chemical shifts of δC 108.7, 107.0, 156.3, and 157.9. A gHMBCAD correlation from the proton at δH 7.25 to the thiazoline carbon at δC 172.4 placed the previously described substructure on the adjacent carbon. The only way to explain the remaining gHMBCAD correlations, as well as the interesting shielded aromatic carbon signals in the δC 107–109 range, was with a 2,4-dihydroxy, 3-chloro substitution pattern on the phenyl ring. This substitution pattern was unexpected, as most known natural products with a similar moiety include the 2,3- or 3,4-dihydroxy pattern, as derived from 2,3-dihydroxybenzoic acid (DHBA). Only one of the phenolic protons (δH 11.04) shows a sharp signal, while the other is absent or severely broadened.

The placement of these functional groups was further supported by the isolation and structure elucidation of 2. The HRESIMS data indicated a formula of C12H12N2O5S based on the protonated molecule observed at m/z 297.0544 (Figure S15). This formula suggested a nonhalogenated form of 1, and the isotopic distribution supports this.

The 1D and 2D NMR signals arising from the dipeptide substructure of 1 are also present in the spectra for 2, indicating that this portion of the molecule is the same in both. However, in 2 there are three protonated carbons in the aromatic region. The proton at δH 7.27 (d, J = 8.6 Hz) shows a gCOSY correlation to δH 6.38 (dd, J = 8.6, 2.3 Hz), which also shows a weak gCOSY correlation to a proton at δH 6.31 (d, J = 2.3 Hz). This small 2.3 Hz coupling constant indicates a long-range four-bond coupling, supporting a ortho, meta arrangement of protons on the aromatic ring. This pattern is further supported by the gHMBC correlations from each of the aromatic protons, as shown in Figure .

Figure 1

COSY and key HMBC correlations in 2D NMR experiments of 1 and 2.

A third analogue was isolated in trace quantity based on the similarity in UV–vis spectral profile in the HPLC. HRESIMS indicated a molecular formula of C12H11BrN2O5S due to the isotopic distribution consistent with a brominated compound and a protonated molecule observed with m/z 374.9640 (Figure S16). While insufficient material was collected to complete a full NMR characterization of this compound, the high-resolution mass, retention time, and identical UV–vis spectrum compared to 1 are consistent with the bromine atom replacing the chlorine without further structural modifications.

Finally, the only stereocenter (C-9) was resolved by comparison of calculated ECD spectra with experimentally acquired values. ECD spectra for the 9R and 9S stereoisomers were predicted by calculating transitions using TD-DFT calculations in Gaussian and importing these transitions into SpecDis for visualization (Figure ). The experimental ECD result clearly matched the spectrum calculated for the R configuration of C-9, completing the absolute structure of 1.

Figure 2

Experimental vs calculated ECD measurement of 1. The experimental data closely match the calculated spectrum for the R configuration of C-9.

When compared to other known phenolate-thiazoline compounds, the 2,4-dihydroxy-3-halo substitution on the phenyl ring is unique. Despite rigorous bioinformatics investigation into the genome of the producing strain, no obvious biosynthetic gene cluster has been identified. However, it could be hypothesized that the compound comes from the coupling of a cysteine-glycine dipeptide with a reactive intermediate of the aromatic moiety. Similar chemical reactions were shown to be done through an alpha-keto nitrile intermediate in Arabidopsis.[17] Interestingly, a number of copper-chelating compounds with isonitrile functionalities have been isolated from Streptomyces thioluteus and Aspergillus fumigatus,[18,19] and fungal isocyanide synthase enzymes are closely associated with copper homeostasis.[20]

Spectroscopic Analysis of Metal Ion Binding Properties of Teredinibactin A (1)

Examination of the HRESIMS spectrum of 1 showed an ion cluster with a base peak at m/z = 713.9346. The elemental formula predictor in MassLynx calculated a molecular formula of C24H20Cl2S2N4O1054Fe+, representing a [2M – 2H + Fe]+ complex. The isotope model for this chemical formula matches the experimental data, which suggests that two molecules of 1 may be able to complex an iron ion (Figure S8). This finding, as well as similarities in the structure to known metal-binding compounds, prompted an investigation into a potential metal chelation role for this molecule.

The UV–vis absorption spectrum of 1 shows clear changes subsequent to incubation with various metal ions and metallo-organic complexes. Incubation with three molar equivalents (MEs) of ferric ammonium citrate (FAC) for 1 h resulted in a clear change in the UV–vis spectrum, specifically an increased relative absorbance around 360 nm, while incubation with Fe-EDTA did not result in any changes (Figure A). A change was also observed after incubation with Fe(III)Cl3, along with an obvious change in the color of the solution from clear to a rust-orange, but not with Fe(II)Cl2 (Figure B). Typically, siderophores have much stronger affinity for Fe(III) relative to Fe(II).

Figure 3

UV–vis absorbance spectra of teredinibactin A (1) complexed with various metal ions. (A) 1 was incubated with organic iron complexes, and a clear change in UV absorption occurs after incubation with ferric ammonium citrate, but not ferric EDTA. (B) 1 demonstrates an obvious change in UV–vis absorbance following incubation with ferric chloride, but not ferrous chloride. (C) A panel of metals used to supplement the growth medium of T7901 shows potential complex formation between 1 and molybdenum and copper ions. (D) A titration experiment indicates that an end point complex is formed with copper at 2 ME of copper to 1.

In order to look for additional potential metal ion interactions, a mixture of metals used in the preparation of artificial seawater for the growth medium of T. turnerae T7901 was incubated with the compound. A clear change in the UV–vis spectrum was observed; thus each individual metal was prepared separately. After incubation with each metal at three MEs for 30 min, CuSO4 and Na2MoO4 were the only two metals that caused UV–vis spectral changes, indicating potential binding of these metals with 1 (Figure C). The spectral change from Na2MoO4 incubation is minor and consists of an increased absorbance around 270 nm. The spectral changes following CuSO4 were the most robust and matched the changes seen following incubation with the metal mix used in media production, so a titration from 0.1 to 3 ME of copper to 1 was performed.

Following incubation with 0.1 ME of CuSO4, the total absorbance of the compound is decreased, but there is no change to the wavelength of each absorbance. Interestingly, a clear transition is observed at 0.5 ME, which results in an absorption spectrum that resembles an intermediate point between the metal-free compound and the 3 ME trace. At a stoichiometric equivalent of 1 and CuSO4, the UV–vis absorbance indicates the formation of a distinct complex. Adding excess CuSO4 at 1.5, 2, and 3 ME indicates that the final end point complex is formed at 2 ME (Figure D). These data indicate the formation of an intermediate complex at substoichiometric concentrations of CuSO4 and a conversion to a copper-bound form at equivalent molar concentrations that is saturated with excess copper at 2 ME. While most metal-binding assays were focused on 1, which is the major analogue, 2 also forms a clear complex after incubation with CuCl2 (Figure S18).

In order to investigate whether the production of teredinibactins is regulated by metal concentrations in the environment, growth experiments were done over a range of copper sulfate concentrations and teredinibactin production was monitored by DAD-HPLC. Unexpectedly, T7901 growth and teredinibactin production were unaffected in media containing 0, 0.1, 0.5, and 1 μM CuSO4 (Figure S17). At high copper concentrations ([CuSO4] = 5 μM), T7901 was unable to grow, and no compound was detected in the growth medium. This is in contrast to control experiments in which T7901 was unable to grow in iron-depleted media, and robust production of the siderophore turnerbactin was observed in iron-limited ([FAC] = 1 μM) conditions. The production of teredinibactin was difficult to ascertain in the iron-limited condition, as the production of turnerbactin dominated the secondary metabolite profile by UV–vis HPLC and LCMS. It is thus difficult to ascribe a physiological role for teredinibactin either as a chalkophore or for the detoxification of free Cu2+, as neither limiting nor toxic levels of copper ions stimulated an overproduction. It is possible that the regulation of teredinibactin production is more complex, involving an interplay of concentrations of iron, copper, turnerbactin, and other trace transition metals.

Experimental Section

General Experimental Procedures

Optical rotations were obtained using a PerkinElmer PE-343 polarimeter. UV–vis spectra were obtained using a Molecular Devices SpectraMax M2 spectrophotometer. ECD spectra were obtained on an AVIV Biomedical, Inc. CD Spec model 410 (Lakewood, NJ, USA). NMR data were collected using a Varian 500 MHz NMR spectrometer with 5 mm Varian Oneprobe (1H 500 MHz, 13C 125 MHz). The residual solvent signals from DMSO-d6 at δH 2.50 and δC 39.52 were used for referencing. High-resolution mass spectra were acquired using a Waters Xevo G2-XS QTof mass spectrometer equipped with a Zspray ESI source and fed by an Acquity H class UPLC system with a Waters Acquity CSH C18 column (2.1 × 50 mm, 1.8 μm). Semipreparative HPLC was performed on a Thermo UltiMate 3000 system with a DAD detector.

Biological Materials

Teredinibacter turnerae T7901 was originally isolated from a sample of the shipworm Bankia gouldi as described in Waterbury et al. 1983.[16] Glyercol stocks were obtained from the public biorepository operated by Ocean Genome Legacy, Inc. This strain is also available from the American Type Culture Collection (accession number 39867).

Fermentation, Extraction, and Isolation

Teredinibacter turnerae T7901 was inoculated from seed cultures (4 mL) into six 2.8 L baffled Fernbach flasks each containing 1 L of SBM[16] and cultured for 7 days at 30 °C with shaking at 180 rpm. The fermentation broth was centrifuged at 4,731g at 4 °C to pellet the cells. The supernatant was extracted with a 1:1:1 mix of Diaion HP20, Sepabeads SP850, and Sepabeads SP825L resins for 4 h at room temperature (rt). The resin was recovered by filtration over clean cotton and washed with ddH2O to remove salts and 25% MeOH to remove media components, then eluted with 50%, 75%, and 100% MeOH. The 75% and 100% MeOH fractions were combined (102.9 mg) based on similarity in the analytical HPLC traces. This fraction was subjected to semipreparative HPLC using a Thermo Synchronis-Aq column (C18, 250 × 10 mm, 5 μm, 3.5 mL/min) and mobile phases of H2O + 0.1% TFA (solvent A) and MeCN (solvent B) with the following gradient profile:

Teredinibactin A (1):

4.2 mg; orange amorphous solid; [α]20D −1.8 (c 0.0005, MeOH); UV (1% MeOH in H2O) λmax (log ε) 275 (3.65), 305 (3.56), 375 (3.08), Figure ; ECD (0.001 M, MeOH) λmax (Δε) 221 (−8.1), 274 (+5.1), 323 (−1.4), 337 (−0.72) nm.; FTIR (polyethylene card) νmax 3303, 3065, 1647, 1539, 1462, 1203, 1138, 1064, 719; 1H NMR and 13C NMR (Tables and S1); HRESIMS m/z 331.0153 [M + H]+ (calcd for C12H12ClN2O5S+, 331.0155).

Table 1

1H and 13C NMR Data for Teredinibactin A (1) and Dechloroteredinibactin A (2) (1H 500 MHz, 13C 125 MHz; DMSO-d6)

	1		2
position	δC, type	δH (J in Hz)	δC, type	δH (J in Hz)
phenyl
1	108.7, C		108.3, C
2	129.6, CH	7.25, d (8.7)	132.2, CH	7.27, d (8.6)
3	107.6, CH	6.60, d (8.7)	107.9, CH	6.38, dd (2.3, 8.6)
4	156.3,a C		160.2,b C
4-OH		11.04, s		10.28, s
5	107, C		102.3, C	6.31, d (2.3)
6	157.9,a C		162.3,b, C
6-OH		12.60,c brs		12.18,d brs
7	172.4, C		172.0, C
8a	33.2, CH2	3.66, t (10.4)	33.1, CH2	3.62, dd (9.7, 10.9)
8b		3.54, dd (8.0, 11.1)		3.48, dd (8.0, 11.1)
9	76.9, CH	5.33, d (8.7)	77.2, CH	5.27, t (8.8)
10	169.6, C		169.6, C
Gly
NH		8.53, t (5.9)		8.47, t (5.9)
11	41, CH2	3.81, d (5.9)	40.9, CH2	3.81, d (5.9)
12	170.9, C		170.9, C
12-OH		13.09,c brs		12.58,d brs

These assignments may be interchanged.

Dechloroteredinibactin A (2):

4.35 mg; orange amorphous solid; [α]20D −1.6 (c 0.0005, MeOH); UV (1% MeOH in H2O) λmax (log ε) 270 (3.9), 306 (3.76), 374 (2.67); 1H and 13C NMR (Tables and S2); HRESIMS m/z 297.0544 [M + H]+ (calcd for C12H13N2O5S+, 297.0545).

ECD Calculations

The 9R and 9S stereoisomers of 1 were modeled separately in ChemDraw 3D and imported into Spartan. A Boltzmann distribution was created from the free energy of all conformers of each stereoisomer, and those that represented a combined >99% of the population were imported to Gaussian 16. These were geometry optimized by DFT calculations at a theory level of b3lyp/6-31g(d,p), and transitions for each conformer of each stereoisomer were obtained using TD-DFT calculations using a theory level of b3lyp/6-31g(d′,p′). Transitions were then imported into SpecDis 1.71 for the generation of calculated ECD spectra. Individual spectra were weighted according to the Boltzmann distribution and combined to create a single, averaged spectrum for each stereoisomer.

UV–Vis-Based Metal Binding Assay

Each metallo-organic or metal salt was prepared fresh in acid-washed glassware. Incubations were performed at a volume of 750 μL in sterile, 1.7 mL microcentrifuge tubes. To each tube were added 660 μL of ddH2O, 15 μL of 1 dissolved in 50% MeOH (final concentration (f/c) 0.3 mM) or 15 μL of vehicle (f/c 1% MeOH), and 75 μL metal solution (f/c 0.03–1 mM). The volumes of ddH2O and metal solutions were adjusted for titration experiments, but final concentrations and total volume remained constant. Solutions were incubated at room temperature for 1 h, then diluted 3-fold into ddH2O in cuvettes for reading in the spectrophotometer.

Growth Experiments

Modified SBM, in which no copper or iron was added, was prepared fresh in an acid-washed bottle. Medium (4 mL) was added to sterile, acid-washed culture tubes, and final concentrations of iron and copper were adjusted through the addition of sterile FAC and CuSO4 solutions. Two-day-old T7901 cultures (4 μL) were used to seed each tube. Cultures were incubated with shaking at 30 °C for 5 days. Cultures were centrifuged to remove cells, and the supernatant was extracted 1:1 with EtOAc. The extracts were dried in vacuo and resuspended in neat MeOH, then filtered and subjected to analytical HPLC.