The marine fungus 067bN1.2 was isolated from a dead pine (Pinus sp.) collected in the splash zone in Kongsfjord, Berlevåg Norway in 2010. The isolate grew from a small wooden cube plated onto agar medium (specified below) during a campaign to study wood-inhabiting fungi of 50 intertidal and sea-floor logs along the Northern Norwegian coast, where Lulworthiales was one of the five most frequent orders isolated (Rämä et al., 2014). The fungus was subcultured and DNA sequenced, and the fungus was phylogenetically placed in the Lulworthiales order (isolate TR498 represents 067bN1.2 in Rämä et al., 2014). At the time of the publication (2014), the closest match from Blast, based on a 5.8S/large ribosomal subunit (LSU) dataset, was Lulworthia medusa (LSU sequence: AF195637). The following primer pairs were used for the internal transcribed spacer (ITS), LSU and small ribosomal subunit (SSU) sequencing, respectively: ITS5-ITS4 (White et al., 1990), LR0R-LR5 (Vilgalys and Hester, 1990; Rehner and Samuels, 1994), and NS1-NS4 (White et al., 1990). The ITS, LSU, and SSU sequences are deposited in GenBank under the following accessions: MW377595, MW375591, and MW375590. The mycelium of the fungus was preserved on pieces of agar in 20% glycerol solution at −80°C.

To identify the isolate 067bN1.2 growing as an asexual morph in culture and determine its systematic position within the order Lulworthiales, a phylogenetic analysis was run using a dataset consisting of nrSSU, nrITS, and nrLSU sequences. The reference sequences included in the analyses were sampled based on recent phylogenetic studies focusing on Lulworthiales (Azevedo et al., 2017; Poli et al., 2020) and retrieved from Genbank (Supplementary Table 1). Sequences for each gene were aligned individually using the E-INS-I and G-INS-I algorithms of MAFFT v7.388 (Katoh et al., 2002; Katoh and Standley, 2013) in Geneious Prime v.11.0.4 followed by manual adjustment. The concatenated dataset consisting of SSU, 5.8S, and LSU sequences and having a length of 2,270 nt was run through PartitionFinder v2.1.1 (Lanfear et al., 2017) to test for best-fit partitioning schemes and evolutionary models with the following settings: models MrBayes, linked branch lengths, greedy search, and AIC and BIC model selection (Lanfear et al., 2012). This suggested three partitions with varying models: symmetrical model with equal base frequencies and gamma distributed rate variation among sites without (SYM+G) and with (SYM+I+G) invariable sites and general time reversible model with variable base frequencies and gamma distributed rate variation among sites (GTR+G). A phylogenetic analysis was set up applying suggested models using Parallel-MPI MrBayes v3.2.7a with beagle, and was run for 5,000,000 generations or until average standard deviation of split frequencies was below 0.0009 with sampling each of the 2,500 generations (Ronquist et al., 2012). In addition, RAxML in Geneious v10.2.3 was run with the same partitions under GTRCAT and GTRGAMMA using rapid-bootstrapping algorithm with 2,000 replicates with search for best scoring ML tree (Stamatakis, 2006). The resulting MrBayes tree was similar to the RAxML tree, excluding some of the basal nodes within Lulworthiaceae shown as polytomies in the MrBayes tree.

For the purpose of this study, the fungal isolate was plated from glycerol stock and grown on nutrient-poor malt agar with sea salts [4 g/L malt extract (Moss Malt Extrakt, Jensen & Co AS), 40 g/L sea salts (S9883, Sigma-Aldrich), 15 g/L agar (A1296, Sigma-Aldrich) and Milli-Q^® H₂O] until the growth covered the entire agar plate (approximately 40 days). Milli-Q^® H₂O was produced with the in-house Milli-Q^® system. One-half of the agar plate covered in mycelium was used to inoculate each liquid culture, in malt medium with added sea salts (4 g/L malt extract, 40 g/L sea salts). Two cultures of 200 ml were inoculated and incubated for 107 days at static conditions and 13°C. Before the addition of resin for extraction, mycelium was taken from the culture for inoculation of another round of cultures. The second cultivation contained four cultures with 250 ml of malt extract medium supplemented with sea salts and cultivated under the same conditions for 83 days. The total culture volume used for the extraction of 1 was 1.4 L. The cultures were extracted using Diaion HP-20 resin (13607, Supelco) and methanol (20864, HPLC grade, VWR) as described previously (Kristoffersen et al., 2018; Schneider et al., 2020). The extract was dried in a rotary evaporator at 40°C under reduced pressure and stored at −20°C.

As part of our ongoing search for antimicrobial compounds, extracts of marine microorganisms are fractioned into six fractions using flash chromatography, as previously described (Schneider et al., 2020). When we investigated the antibacterial potential of fractions produced from several understudied marine fungi, one fraction from isolate 067bN1.2 piqued our interest due to its antibacterial activity. In the active fraction, 1 was the dominating peak. The monoisotopic mass, calculated elemental composition and fragmentation pattern of 1 was determined using UHPLC-ESI-HRMS. UHPLC-ESI-HRMS was performed with positive ionization mode, using an Acquity I-class UPLC with an Acquity UPLC C18 column (1.7 μm, 2.1 mm × 100 mm), coupled to a PDA detector and a Vion IMS QToF (all from Waters). Compounds were eluted with a gradient over 12 min, from 10 to 90% acetonitrile (LiChrosolv, 1.00029, Supelco) with 0.1% formic acid (Sigma-Aldrich) in Milli-Q H₂O and a flow rate of 0.45 ml/min. Waters UNIFI 1.9.4 Scientific Information System was used to process and analyze the data. Elemental compositions of compounds in the samples were used to search relevant databases, such as Chemspider, in order to identify known compounds. Since the calculated elemental composition gave no hits in database searches, 1 was nominated for isolation.

Initial attempts to isolate 1 was performed using mass guided preparative HPLC. This strategy proved difficult due to extensive binding of the compound to an Atlantis Prep C18 (10 μM, 10 × 250 mm) (Waters) column, leading to inefficient isolation and column contamination. The preparative system and mobile phases used were as previously described (Schneider et al., 2020). The resulting sample (referred to as compound 2) was later used to assist in structure elucidation of compound 1.

To avoid wear and tear of the preparative HPLC system, attempts were made to isolate 1 using flash chromatography. The dried extract was dissolved in 90% methanol, and 2 g of Diaion HP-20SS (13615, Supelco) was added before removing the solvent under reduced pressure. Flash columns were prepared as previously described (Kristoffersen et al., 2018). The column was equilibrated using 5% methanol, before the dried extract-Diaion HP-20SS mixture was applied to the top of the column (maximum 2 g of extract per round). The fractionation was performed on a Biotage SP4^TM system (Biotage) with a flow rate of 12 ml/min and a stepwise gradient from 5 to 100% methanol over 32 min. The following stepwise elution method was used: methanol:water (5:95, 25:75, 50:50, 75:25, 6 min per step, resulting in 12 fractions) followed by methanol (100% over 12 min, resulting in six fractions). The MeOH fractions were analyzed using UHPLC-ESI-HRMS. In the second fraction eluting at 100% MeOH, 1 was the dominating peak and was submitted for NMR and bioactivity analysis. The sample of 1 was therefore produced by pooling the second fraction eluting at 100% MeOH from multiple rounds of flash fractionation and drying the resulting volume under reduced pressure.

The structure of 1 was established by 1D and 2D NMR experiments. NMR spectra were acquired in DMSO-d₆ and methanol-d₃ on a Bruker Avance III HD spectrometer operating at 600 MHz for protons, equipped with an inverse TCI probe cryogenically enhanced for ¹H, ¹³C, and ²H. All NMR spectra were acquired at 298 K, in 3-mm solvent matched Shigemi tubes using standard pulse programs for proton, carbon, HSQC, HMBC, HMQC (J = 4–5 Hz), COSY, NOESY, ROESY and 1,1-ADEQUATE experiments with gradient selection and adiabatic versions where applicable. ¹H/¹³C chemical shifts were referenced to the residual solvent peak (δ_H = 2.50 PPM, δ_C = 39.52 PPM for DMSO). All data were acquired and processed using Topspin 3.5pl7 (Bruker Biospin) including the structure elucidation module CMC-se v. 2.5.1. ¹³C prediction was done using Mestrelabs MestReNova software version 14.2.0-26256 with the Modgraph NMRPredict Desktop. Optical rotation data were obtained using an AA-10R automatic polarimeter (Optical Activity LTD).

Lulworthinone (1): green colored film. [α]²⁰_D -120 ± 0.02 (c 0.2 DMSO). ¹H and ¹³C NMR spectroscopic data, Supplementary Table 3. HRESIMS m/z 741.2204 [M+H]⁺ (calculated for C₃₇H₄₁O₁₄S, 741.2217).

The Minimal Inhibitory Concentration (MIC) of 1 against a panel of Gram-positive and Gram-negative reference bacteria was determined by broth microdilution, at final concentrations 0.2–100 μg/ml (twofold dilution series). The experiments were performed with three technical replicates. The panel of reference bacteria consisted of the following strains: S. aureus (ATCC 25923), MRSA (ATCC 33591), Escherichia coli (ATCC 25922), Pseudomonas aeruginosa (ATCC 27853), Enterococcus faecalis (ATCC 29212), and Streptococcus agalactiae (ATCC 12386), all strains from LGC Standards (Teddington). Briefly, the bacteria were inoculated from freeze stock onto blood agar plates (University Hospital of North Norway) and transferred to liquid medium for overnight incubation at 37°C. S. aureus, E. coli, and P. aeruginosa were grown in Brain Heart Infusion medium (BHI, 53286, Sigma-Aldrich), and E. faecalis and S. agalactiae were grown in Difco^TM Mueller Hinton medium (MH, 275730, BD Biosciences). After overnight incubation in the respective media, the bacteria were brought to exponential growth by addition of fresh media, and incubated to reach a turbidity of 0.5 McFarland standard. The bacteria were diluted in their respective media 1:1,000 prior to addition. Subsequently, the bacteria were added to 96-well microtiter plates at 50 μl/well. A mixture of 50 μl of autoclaved Milli-Q^® H₂O and 50 μl fresh autoclaved media was used as negative control, and 50 μl of autoclaved Milli-Q^® H₂O was added to 50 μl of bacteria suspension as growth control. The compound was diluted in DMSO and autoclaved Milli-Q^® H₂O (highest concentration of DMSO in the assay was 0.5%), and 50 μl was added to the bacterial suspension. Final volume in the wells was 100 μl. The plates were incubated overnight at 37°C. After incubation, growth was measured by absorbance at 600 nm with 1420 Multilabel Counter VICTOR^3TM (Perkin Elmer). Assay controls with gentamicin in a dilution series are routinely run, as well as routine counting of CFUs for each bacterium. For the strains where the compound displayed activity, the MIC was determined with three biological replicates each containing three technical replicates (n = 9). The lowest concentration of 1 that completely inhibited the growth of the bacteria was determined as the MIC.

To investigate if 1 had a bacteriocidal or bacteriostatic effect on S. aureus and S. agalactiae, the compound was inoculated together with the bacteria, as described above, and after overnight incubation, the inoculum was plated onto agar and incubated overnight at 37°C. The experiment was done with 12.5 and 25 μg/ml concentrations of 1 in triplicate, with two biological replicates (n = 6). Inspired by Zheng et al. (2007), we tested 1, together with reserpine (broad spectrum efflux pump inhibitor) against the Gram-negative reference strains E. coli and P. aeruginosa. The assay was conducted as described above, with reserpine (L03506, Thermo Fisher Scientific) added to a final concentration of 20 μg/ml.

Initial testing of 1 was conducted against a panel containing clinically relevant antibiotic-resistant bacteria: Gram-positive MRSA, vancomycin-resistant Enterococcus faecium (VRE), and Gram-negative bacteria resistant to extended-spectrum beta-lactamases as well as carbapenemases (ESBL-Carba) (detailed information about the clinical isolates can be found in Supplementary Table 2). The initial testing was conducted at one concentration, 100 μg/ml.

The final antibacterial testing of 1 was executed using the five clinical MRSA isolates and the VRE isolates (Supplementary Table 2). The isolates were tested by broth microdilution according to the Clinical Laboratory Standard Institute (CLSI) (2012) method MO7-A9. In brief, 1 was solubilized with 100% DMSO and diluted with autoclaved Milli-Q^® H₂O to prepare a 200 μg/ml working solution. The final DMSO concentration did not exceed 1% to exclude any artificial influence on the assay. The bacterial inoculum was prepared to contain 1 × 10⁶ CFU/ml in cationic-adjusted BBL^TM Mueller-Hinton II broth (BD). The inoculum was mixed in a 1:1 ratio with the working solution of 1 (twofold dilutions, ranging from 0.2 to 100 μg/ml) for a final amount of 5 × 10⁵ CFU/ml in each well of a 96-well round-bottom polypropylene plate (Greiner Bio-One GmbH). Growth control (without compound) and sterility control (without bacteria) were included for each strain. Each strain was tested in three independent biological replicates with four technical replicates on consecutive days. As quality assurance for the assay, the protocol was also performed with E. coli ATCC 25922 using Gentamicin (Merck Life Science) as a reference antibiotic. The 96-well plates were incubated at 37°C for 24 h without shaking. The MIC values were defined as the lowest concentration of 1 resulting in no visual bacterial growth, determined by visual inspection and 600 nm absorbance measurements with CLARIOstar plate reader (BMG LABTECH).

Inhibition of biofilm production by 1 of Staphylococcus epidermidis (ATCC 35984, LGC Standards) was determined at final concentrations 0.2–100 μg/ml (twofold dilution series). Briefly, the bacteria were inoculated from freeze stock onto blood agar plates (University Hospital of North Norway) and transferred to tryptic soy broth (TSB, 22092, Sigma-Aldrich) for overnight incubation at 37°C. The overnight cultures were subsequently diluted 1:100 in fresh TSB with 1% glucose and added to 96-well microtiter plates, 50 μl/well. Positive control was S. epidermidis in fresh media with glucose, and negative control was a non-biofilm producing Staphylococcus haemolyticus (clinical isolate 8-7A, University Hospital of North Norway) in fresh media with glucose. The compound was diluted in DMSO and autoclaved Milli-Q^® H₂O (highest concentration of DMSO in the assay was 0.5%), and 50 μl was added to the bacterial suspension. Final volume in the wells was 100 μl. The plates were incubated at 37°C overnight. Growth inhibition of the bacterium was determined by visual inspection of the plates prior to further treatment. The bacterial suspension was poured out and the biofilm was fixated by heat, before adding 70 μl of 0.1% crystal violet solution (V5265, Sigma-Aldrich) and staining for 5 min. The crystal violet solution was removed and the wells were washed with water before the plates were dried by heat. The bound crystal violet was dissolved in 70 μl of 70% ethanol, and the presence of violet color, indicating biofilm formation, was measured at 600 nm absorbance using a 1420 Multilabel Counter VICTOR^3TM reader. Percent biofilm formation was calculated using the equation below. The data were visualized using GraphPad Prism 8.4.2, and the built-in ROUT method was used to detect and remove outliers from the dataset (Q = 1%).

To determine whether 1 could eradicate biofilm established by S. epidermidis, a modified biofilm inhibition assay protocol was performed. Here, the bacteria were grown overnight in a microtiter plate to allow the biofilm to be established prior to the addition of 1. After addition of 1, the plates are incubated overnight. Following this, the biofilm was fixated and colored and results were read as stated above. The experiment was conducted once with three technical replicates with concentrations of 0.2–100 μg/ml (twofold dilution series).

The antiproliferative activities of 1 was evaluated against the melanoma cell line A2058 (ATCC, CRL-11147^TM), the hepatocellular carcinoma cell line HepG2 (ATCC, HB-8065^TM), and the non-malignant lung fibroblast cell line MRC5 (ATCC, CCL-171^TM) in a MTS in vitro cell proliferation assay. The compound was tested in concentrations from 6.3 to 100 μg/ml against all cell lines, with three biological replicates each containing three technical replicates (n = 9). A2058 was cultured and assayed in Dulbecco’s Modified Eagle’s Medium (D-MEM, D6171, Sigma-Aldrich). HepG2 was cultured and assayed in MEM Earle’s (F0325, Biochrom) supplemented with 5 ml of non-essential amino acids (K0293, Biochrom) and 1 mM sodium pyruvate (L0473, Biochrom). MRC5 was cultured and assayed in MEM Eagle (M7278, Sigma-Aldrich) supplemented with 5 ml of non-essential amino acids, 1 mM sodium pyruvate, and 0.15% (w/v) sodium bicarbonate (L1713, Biochrom). In addition, all media were supplemented with 10% fetal bovine serum (FBS, S1810, Biowest), 10 μg/ml gentamicin (A2712, Biochrom), and 5 ml of glutamine stable (200 mM per 500 ml medium, X0551, Biowest). Briefly, the cells were seeded in 96-well microtiter plates (Nunclon Delta Surface, VWR) at 2,000 cells/well for A2058, 4,000 cells/well for MRC5, and 20,000 cells/well for HepG2. After incubation for 24 h in 5% CO₂ at 37°C, the media was replaced and compound was added, generating a total volume of 100 μl/well. A2058 and MRC5 were incubated for 72 h before assaying, and HepG2 for 24 h. Subsequently, 10 μl of CellTiter 96 AQueous One Solution Reagent (G358B, Promega) was added to each well and the plates were incubated for 1 h at 37°C. Following this, the absorbance was measured at 485 nm with a DTX 880 multimode detector (Beckman Coulter). Negative controls were cells assayed with their respective cell media, and positive controls were cells treated with 10% DMSO (D4540, Sigma-Aldrich). Percent cell survival was calculated using the equation below. The data were visualized using GraphPad Prism 8.4.2 and IC₅₀ was calculated. The built-in ROUT method was used to detect and remove outliers from the dataset (Q = 1%).

The MIC of 1 was determined by broth microdilution against C. albicans (ATCC 90028, LGC Standards), at final concentrations of 0.2–100 μg/ml (twofold dilution series). The experiment was performed as one biological replicate, with three technical replicates (n = 3). Briefly, the fungus was inoculated from freeze stock onto potato dextrose agar [24 g/L potato dextrose broth (P6685, Sigma-Aldrich), 15 g/L agar (A1296, Sigma-Aldrich)] and incubated overnight at 37°C. From the overnight culture, five to eight colonies were transferred to 5 ml of sterile 0.9% NaCl, before the cell density was adjusted to 1–5 × 10⁶ cells/ml by adding 0.9% NaCl. The cell density was evaluated with 0.5 McFarland standard (Remel 0.5 McFarland Equivalence Turbidity Standard, 10026732, Thermo Fisher Scientific). The fungal suspension was further diluted 1:50, and then 1:20 (1–5 × 10³ CFU/ml) in RPMI medium (R7755, Sigma-Aldrich) with 0.165 mol/L MOPS (M3183, Sigma-Aldrich) and 10.25 ml of L-glutamine. The compound was added to the microtiter plate together with the fungal suspension (1:1), to a final volume of 200 μl. The final concentration of fungal cells was 0.5–2.5 × 10³ CFU/ml. Absorbance in the wells was measured with 1420 Multilabel Counter VICTOR^3TM right after addition of compound, after 24 h and after 48 h. The plates were incubated at 37°C. Amphotericin B was used as negative control at final concentration 8 μg/ml. Growth control contained fungal suspension and autoclaved Milli-Q^® H₂O.

Due to lack of distinct morphological characters of the cultured asexual morph and closely related reference sequences in GenBank, the fungus is identified to family level, as Lulworthiaceae sp., for the purpose of this study. A phylogenetic study was carried out with 28 taxa (including outgroups and isolate 067bN1.2), all representing different species, as shown in Figure 1. The combined dataset of 5.8S, SSU, and LSU had an aligned length of 2,270 characters, and phylogenetic inference was estimated using both Maximum Likelihood and Bayesian Inference criteria. The isolate producing 1, 067bN1.2, was placed on its own branch within the Lulworthiaceae, forming a sister clade to the clade including Halazoon fuscus, Lulworthia medusa, Lulworthia cf. purpurea and Halazoon melhae. Sequences of Koralionastes ellipticus were included to exclude the possibility that the isolate 067bN1.2 is part of the family Koralionastetaceae. Koralionastes ellipticus was placed outside of Lulworthiaceae.

Compound 1 was selected for isolation due to its antibacterial activity in an initial screen of fractions from several understudied marine fungi. Compound 1 was the dominating peak in the active fraction from fungal isolate 067bN1.2 Lulworthiaceae sp., and subsequently the fungus was re-cultivated, cultures were extracted, and the compound was isolated using RP flash chromatography. The extraction of 1.4 L of fungal culture yielded 1,017.2 mg of extract.

Initially, attempts were made to isolate the compound using preparative HPLC. This strategy had several drawbacks, including unfavorable behavior of the compound in the preparative column. This resulted in the compound eluting over several minutes (band broadening) and carryover. A batch of the compound was, however, retrieved using this strategy, resulting in a compound later determined to be a structural isomer and artifact of compound 1 (referred to as 2 throughout this article), produced due to the acidic conditions in the mobile phase. The structures of 1 and 2 can be seen in Figure 2.

Flash chromatography was better suited for the isolation of 1. This isolation strategy yielded 63.8 mg of 1, corresponding to a yield of ∼45 mg/L culture medium. Compound 1 was obtained as a green colored substance. The molecular formula was calculated to be C₃₇H₄₀O₁₄S by UHPLC-ESI-HRMS (m/z 741.2204 [M+H]⁺) (calculated as C₃₇H₄₁O₁₄S, 741.2217), suggesting 18 degrees of unsaturation. The low-energy collision mass spectrum of 1 can be seen in Supplementary Figure 2. MS signals of a neutral loss of 80 Da (ESI+) was observed, indicating the presence of a sulfate group in the structure. The UV absorption maxima were 224, 260, and 373 nm, which corresponded well with the previously published dinapinones (Kawaguchi et al., 2013). The UV-vis spectrum for 1 can be seen in Supplementary Figure 3. The IR spectrum of 1 displayed absorption bands for sulfoxide (S=O, 1,002 cm^–1), aromatic alkene (C=C, 1542 and 1,618 cm^–1), carbonyl (C=O, 1,645 cm^–1), alkane (C-H, 2,857 cm^–1), aromatic alkene (C-H, 2926 cm^–1), and hydroxyl (C-OH, 3455 cm^–1) bonds. After isolation, the structure of 1 (Figure 2) was elucidated by 1D and 2D NMR experiments (Supplementary Figures 4–16).

Initial structure elucidation was made on the sample isolated by preparative HPLC with formic acid present in the mobile phases (compound 2). The established molecular formula suggested a highly conjugated system. The purity of 2 was estimated to be ∼80% from a quantitative proton spectrum with respect to non-solvent impurities (Supplementary Figure 4). Four singlet protons were identified in the aromatic region, along with three O-CH signals at ∼4.5 ppm with complex couplings along with a methoxy singlet at 3.77 ppm. Furthermore, five hydroxyl protons were identified; three between 9.5 and 10.0 ppm, and two between 13.5 and 14.0 ppm. The deshielded nature of the latter sets them apart from the other hydroxyls and suggests they may be involved in an angled intramolecular hydrogen bond, which is commonly seen for keto-enol pair configurations such as this. All 37 carbons could be identified by 1D ¹³C NMR (Supplementary Figure 5), which showed 2 to contain a large number of aromatic quaternary carbons, two ester-like carbonyls, along with 10 peaks in the aliphatic region (Table 1).

HSQC, HMBC, and 1,1-ADEQUATE spectra (Supplementary Figures 6, 7) allowed the identification of two substituted napthopyrone-like moieties, as well as two five-membered aliphatic chains (denoted C15-C11 and C15’-C11’, respectively), which were fully assigned using a combination of HSQC-TOCSY, TOCSY, COSY, and HMBC (Figure 3i). The aliphatic chains were determined to be attached at the C10 position of the napthopyrone-like moieties by tracing the spin system into H9 and H9’, respectively, and supported by multiple long-range ¹H-¹³C correlations. The C2 and C2’ carbonyls could be directly assigned from long-range couplings from the 10/10’ position, but the hydroxyl carrying carbons in positions 3/3’ and 4/4’ could only be assigned through weak ⁴J_CH correlations from the aromatic protons (Figure 3iii).

The OH-4 and OH-6 could be assigned based on NOE correlations between OH-6 and both H5 and H7, while OH-4 only displayed correlations with H5. The OH-3 and OH-3’ are predicted to have more deshielded chemical shifts due to their proximity to the carbonyl moiety and a probable intramolecular hydrogen bond—however, it was not possible to individually distinguish OH3 and OH-3’ due to the absence of any correlations in NOESY, ROESY, and HMBC spectra. Thus, four fragments could initially be elucidated (Figure 3i). A weak ⁴J_C8H7’ correlation could be detected, linking fragment A to fragment B (Figure 3i) at the C8 and C5’ positions, respectively, and thus the only remaining ambiguity is the position of the -SO^3– group vis-à-vis the remaining -OH in the 9’ or 4’ positions. The absence of NOEs and COSY correlations between OH-4’ and H9’ suggests that it is positioned at C4’ with the sulfate positioned at C9’ (Figure 3ii). The ³J_HH coupling constant between H9’ and H10’ was measured to be 2.0 Hz from line shape fitting the splitting of H9’, indicating that these protons are at a significantly offset dihedral angle to one another—thus suggesting a relative R/S or S/R configuration of 9’ and 10’. ¹³C prediction was consistent with the structure of 2 (Supplementary Figure 9), with a mean error of 2.79 ppm between the observed and predicted ¹³C shifts.

A second isolation where no acidic conditions were used, yielding 1, was also examined. ¹H NMR revealed significantly perturbed chemical shifts as well as line broadening and heterogeneity throughout the spectra (Supplementary Figure 11). Multiple resonances in the carbon spectrum (Supplementary Figures 12, 13), especially for two resonances in the carbonyl area (presumably C3 and C3’), are heterogenous, reflecting the nuclei existing in several stable, but slightly different micro environments. The same observation is made in the proton spectrum (Supplementary Figure 11) for H9’, OMe-6’, H5, H7, 4’-OH, and 4-OH. A major difference was observed in the non-acidic preparation (1), compared to 2, the presence of a 9’-OH. At ∼15 ppm, two heterogeneous OH protons were observed, deshielded by approximately 1 ppm compared to the OH-3’s in the original sample preparation, while the three hydroxyls at ∼10 ppm could no longer be detected (Supplementary Figures 8–13). Thus, the detectable aromatic hydroxyl groups, identified as OH-4’ and OH-4, appeared to be involved in (stronger) hydrogen bonding, while three aromatic hydroxyls, the remaining OH-6, OH-3’ and OH-3, were unaccounted for. At the same time, the majority of all other nuclei in the molecule are shielded by approximately 0.5 ppm. Together, these observations suggest that the neutral pH preparation resulted in a different molecule, 1, that formed loose aggregates in DMSO and methanol, stabilized by both hydrogen bonding (deshielding) and stacking (shielding) interactions. Overall, worse spectral quality resulted in that the C2 and C3 from 2 could not be individually assigned in 1, although they must correspond to the two chemical shifts of 169.4 and 173 ppm by the logic of elimination. A number of the carbons show heterogenic peaks (notably the presumed C3 and C3’), most likely as the result of through space proximity to the sulfate group and sensitivity to its different possible conformation (details in section “Discussion”).

The identity of 1 was established to be identical to 2 with the only difference being that the sulfate group was attached to C6 instead of C9’, supported by the loss of the OH correlating with H5 and H7, and the appearance of an OH correlating with H9’ through a ³J_HH. There is furthermore a heterogeneity and chemical shift perturbation hotspot (vis-à-vis 2) around the C6 position to support the assignment of a C6 sulfate. All chemical shifts and correlations are summarized in Supplementary Table 3. The data do not unambiguously prove whether the 3-OH’s are deprotonated or if the signal is lost due to rapid exchange, but the fact that the OH-9’ is observable under the same conditions is an indicium for the OH-3’s to be deprotonated in 1. No plausible resonance structures to explain the deprotonation and deshielding that does not involve the oxidation, and thus change in mass, have been found.

The non-aggregated 2 could be scavenged by lowering the pH of 1 with the addition of hydrochloric acid, upon which ¹H and HSQC spectra of the two samples of 2 show a great resemblance (Supplementary Figure 10). The molecular formula of 2 and 1 as well as the scavenged 2 were identical in the two preparations, as no change in mass was observed by high-resolution mass spectrometry.

Compound 1 was tested against six reference bacteria (four Gram-positive and two Gram-negative strains). The compound was active against two of the Gram-positive reference strains, S. aureus and S. agalactiae, with MIC values of 6.25 and 12.5 μg/ml, respectively. No activity was observed against the Gram-negative strains, E. coli and P. aeruginosa, or the Gram-positive E. faecalis or MRSA strain (Supplementary Table 4). As bacterial resistance toward available antibiotics is the main challenge in future treatment of pathogenic diseases, 1 was tested against a panel of drug-resistant clinical strains (Supplementary Table 2). The panel included five MRSA and six VRE strains. Compound 1 was also tested in a pre-screen against four Gram-negative clinical bacterial strains: E. coli, Klebsiella pneumoniae, Acinetobacter baumanii, and P. aeruginosa (all ESBL-Carba). No activity was detected against the Gram-negative bacteria (Supplementary Table 4). Compound 1 showed activity against the MRSA strains with MICs in the 1.56–6.25 μg/ml (2.12–8.44 μM) range, see Table 2. The activity of the compound was significantly less profound against the VRE strains (MIC = 50 μg/ml or higher) (Supplementary Table 4).

To investigate if 1 has bacteriostatic or bacteriocidal effects on the two reference strains S. aureus and S. agalactiae, both were incubated with the compound at 12.5 and 25 μg/ml overnight and subsequently plated onto agar. For S. aureus, there was no growth on the plates after overnight incubation, indicating a bacteriocidal effect of 1. For S. agalactiae, one of the parallels at 12.5 μg/ml (MIC of 1 against this bacterium) displayed growth on the agar plate, which was expected as visual growth could also be seen in the microtiter plate for this parallel. The remaining five parallels at this concentration, and the concentration above, had no growth in the microtiter plates, or on agar after overnight incubation. This strongly indicates that 1 also has bacteriocidal effect on S. agalactiae. Compound 1 was also tested together with the efflux pump inhibitor reserpine to see if the lack of activity toward Gram-negative strains was caused by efflux of 1, but no activity was obtained.

The ability of 1 to inhibit biofilm production by S. epidermidis and to remove established S. epidermidis biofilm was assessed. In the biofilm inhibition assay, the biofilm production was completely inhibited (below 5% biofilm formation) down to 12.5 μg/ml (Figure 4). Clear inhibition of the bacterial growth could also be observed to 25 μg/ml by visual inspection of plates before fixation of biofilm, raising the question if the biofilm inhibition is mainly caused by growth inhibition of the bacterium. To further evaluate the potential biofilm activity, removal of established biofilm was assessed. There was no activity of 1 at concentrations up to 100 μg/ml against the established biofilm, further supporting the hypothesis that the biofilm inhibition is mainly due to growth inhibition of the bacterium.

The antiproliferative activities of 1 was assessed against human melanoma cells (A2058), human non-malignant lung fibroblasts (MRC5), and human hepatocellular carcinoma cells (HepG2), in a concentration range of 6.25–100 μg/ml. The non-malignant cell line was included as a test for general toxicity, while the other cell line was included to assess possible anti-cancer activities. Antiproliferative activity was observed against all cell lines, with IC₅₀ values of 15.5, 32, and 27 μg/ml against A2058, MRC5, and HepG2, respectively (Table 3). Compound 1 was also assayed for antifungal activity against C. albicans at concentrations up to 100 μg/ml, and no activity was seen.