In an initial search for antimicrobial compounds we isolated strain S2753 related to Photobacterium halotolerans [29]. Subsequently, the known antibiotic, holomycin, was identified as responsible for its growth inhibitory activity [45]. When investigating ethyl acetate extracts of S2753 in an agar diffusion assay monitoring expression of the S. aureus virulence genes hla, rnaIII, and spa [28], we observed an increased expression of spa and decreased expression of hla and rnaIII. The inverse effect of the extracts on spa and hla/rnaIII expression, respectively, indicates the presence of at least one compound that interferes with the S. aureus agr QS system [28]. Secondary screening of the extract by explorative solid-phase extraction (E-SPE) [48] detected the potential QSI activity in a fraction that did not inhibit growth of S. aureus or V. anguillarum (data not shown). Bioassay-guided fractionation by diol and C-18 columns resulted in the isolation of two compounds active in the S. aureus agar diffusion assay (Figure 1). The activity of the pure compounds matched the initial activity of the extract, confirming that these compounds are responsible for the observed changes in gene expression.

The solonamides were isolated as white powder with respective molecular formulas C₃₀H₄₆N₄O₆ (A) and C₃₂H₅₀N₄O₆ (B) as determined by HRMS (1 ppm mass accuracy). Analysis of NMR data characterized the structures of the solonamides as cyclodepsipeptides consisting of four amino acids and a 3-hydroxy fatty acid (Figure 2). The amino acid composition was elucidated as alanine, phenylalanine, and two leucines for both peptides based on DQF-COSY and HSQC NMR data.

The spin systems of the amino acids were confirmed through strong and unambiguous H2BC correlations [49] and the carbonyl signals assigned by HMBC correlations. Through careful inspection of the DQF-COSY and H2BC NMR data, solonamide A was found to contain a 3-hydroxyhexanoic acid (Hha), while solonamide B contained a 3-hydroxyoctanoic acid (Hoa). Long-range HMBC and NOESY correlation data allowed the sequence of amino acids to be established as fatty acid-Phe-Leu-Ala-Leu (Figure in Supplementary Information). This was corroborated by MS-MS experiments giving the exact molecular formula of the fragments (2 ppm mass accuracy, see Supplementary Information). The signal from one oxygen-bearing carbon with a high carbon shift indicated an ester linkage. The ring closure linkage was secured by HMBC correlations from H-3 in the fatty acid moiety to the carbonyl in Leu and a weak NOESY correlation from H-3 to the Leu amide and Hα protons. In total, this accounted for the ten degrees of unsaturation resulting from the macrocyclic ring, five carbonyls, and the phenyl group.

The absolute configurations of the individual amino acids were established by acid hydrolysis and Marfey’s method with UHPLC analysis. Both peptides were found to contain L-Phe, D-Ala, and an enantiomeric pair of L-Leu and D-Leu. Acid hydrolysis of the reduced linear peptides and subsequent Marfey’s derivatization specified the stereochemistry of the two Leu, exchanging the L-Leu peak (RT 3.77 min) with a new peak (RT 3.73 min), attributable to the corresponding alcohol. Thus, L- and D-stereochemistry was assigned as fatty acid-L-Phe-D-Leu-D-Ala-L-Leu in both solonamide A and B.

The absolute configuration of the fatty acid was established by NMR spectroscopic analysis of the ¹H and ¹⁹F chemical shift differences (Δδ^SR) in the (R)- and (S)-Mosher’s esters analysis of solonamide A and B. The absolute stereochemistry of C-3 in the 3-hydroxy fatty acid was established as (R) in both depsipeptides (Figure 3).

The yield of solonamides was up to 10 mg L⁻¹, which is a high organic yield compared to that of other γ-proteobacteria [50,51]. This suggests that they could be storage compounds accumulated during growth. However, the solonamides were also produced on a chitin based minimal medium (Supplementary Information) which indicated that these compounds may be produced in the natural habitat of vibrios, such as chitinous zooplankton. In addition, D-alanine and L-leucine are incorporated in the structure; amino acids that are not present in the laboratory medium. Thus, the Photobacterium seems to produce these specific stereoforms rather than incorporating the available amino acids.

To test whether the solonamides are also produced by strains related to our Photobacterium isolate, we compared strain S2753 with P. halotolerans LMG 22194^T, P. rosenbergii LMG 22223^T, and P. angustum S14 [52]. None of these strains produced solonamides (as confirmed by LC-UV/MS), and none affected virulence gene expression in the gene-reporter agar diffusion assay. None of the three strains inhibited growth of V. anguillarum, and holomycin, the antibiotic of S2753 [45], was not detected.

To the best of our knowledge, only two species of Photobacterium have been investigated for their secondary metabolites so far [51,53,54]. Oku et al. [51] isolated unnarmicin A and C from a marine Photobacterium strain MBIC06485 related to P. leiognathi. Like the solonamides, unnarmicin A and C consist of four amino acids (L-Phe, L-Leu, D-Phe, L-Leu) and a 3-hydroxyoctanoic and 3-hydroxyhexanoic fatty acid, respectively. The finding of the unnarmicins in another marine Photobacterium sp. indicates that production of such peptides could be a common feature in this group of bacteria, despite the absence of solonamides in any related strains that we investigated.

To verify that the purified solonamides do in fact cause transcriptional changes in virulence gene expression and to assess if the effect is strain specific, we isolated mRNA from S. aureus 8325-4 and the community-acquired strain, USA300 at different stages of growth following solonamide exposure and monitored gene expression by Northern blot analysis. Solonamide B dramatically reduced the expression of both hla and rnaIII while increasing expression of spa, strongly indicating that the compound interferes with agr regulation (Figure 4). The decreased expression of rnaIII was even more pronounced in the highly virulent USA300 strain where high agr activity is suspected to be a main contributor to the virulence of the strain [55,56]. The solonamides did not affect the growth rate of the liquid S. aureus cultures. Solonamide A was able to increase spa expression, but caused only a marginal reduction of hla and rnaIII expression in both 8325-4 and USA300. The discrepancy between the Northern blot analysis and the agar diffusion assay (mainly with regard to hla) may be rooted in the much higher concentrations of solonamides that are used in the agar diffusion assay as compared to the Northern blot analysis. Also, the Northern blot analysis directly monitors the amount of mRNA shortly after solonamide addition, whereas the agar diffusion assay relies on the accumulation of β-galactosidase enzyme over a period of 15 or 35 h.

The structural similarity of the solonamides and the AIPs (Figure 2) suggest that they may be competitive inhibitors of the agr system. Unlike the AIPs, the solonamides are cyclized through a 3-hydroxy fatty acid forming a lactone rather than a thiolactone. However, synthetic lactone and lactam variations of natural AIPs have been found to have antagonistic activity [19,23,57], which our study corroborates. While inhibition of agr by AIPs is more tolerant of sequence and structural diversity than is activation [23], Mayville et al. [14] found that the presence of the hydrophobic leucine and phenylalanine residues is important for the inhibition of the agr response. Both solonamides contain a leucine and phenylalanine; however, the reduced activity of solonamide A indicates that the overall hydrophobicity of the depsipeptides affected by the varying length of the fatty acid moiety might be an important factor influencing activity.

The solonamides are the first reported antagonists produced by a natural source with a structure resembling that of native S. aureus AIPs. Kiran et al. (2008) [25] identified hamamelitannin from Hamamelis virginiana (witch hazel) as an inhibitor of RNAIII and δ-hemolysin production in S. aureus 8325-4, USA300, and clinical S. epidermidis isolate MH. Also, ambuic acid from an unidentified fungal strain was found to attenuate agr [26].Given the relatively low abundance of staphylococci in the marine environment, it seems unlikely that the Photobacterium sp. S2753 produces solonamides as part of a deliberate strategy to interfere with this specific type of bacteria. However, the solonamides might be targeted at other Gram-positive bacteria in the marine environment, such as bacilli and actinobacteria, though little is known about quorum sensing pathways in marine Gram-positive bacteria. A large number of different QS systems have been characterized from Gram-negative bacteria in the marine environment [57,58,59,60,61,62]. We speculate that the solonamides could also affect such systems despite sharing little structural similarity to agonists and antagonists of the systems identified to date [63]. Acylated homoserine lactones, the most widely researched type of QS molecules in Gram-negatives, can serve as both agonists and antagonists in different systems [27], and thus the solonamides may also serve as quorum sensing signals for the Photobacterium itself. However, we did not detect solonamides or compounds with similar QSI activity in any of the related strains.

Our findings suggest that quorum sensing inhibition could be an alternative therapeutic strategy for treatment of MRSA S. aureus infections; however, the effect in an in vivo infection model needs to be tested. Many genes are under QS control both in Gram-positive and Gram-negative bacteria, and thus it is not a simple drug target [64,65]. For example, biofilm formation in S. aureus has been linked to low QS activity [66], and so there is a risk that the use of QS inhibitors could lead to decreased susceptibility of traditional antibiotics. Also, it is still unknown how QS inhibition will affect the overall fitness of a pathogen under in vivo conditions and thus pose a selective pressure for development of resistance [67] or enhanced virulence [68,69].

Bacterial strain S2753 was isolated from a mussel surface collected in the tropical Pacific (9.1°S 156.8°E) during the Danish Galathea 3 expedition [29]. S2753 was assigned to the Vibrionaceae by 16S rRNA gene sequence similarity [29] and identified as being closely related to Photobacterium halotolerans based on recA and rpoA gene sequences, with homologies of 87% (recA) and 94% (rpoA) [45]. BLAST analyses showed that other closely related species were Photobacterium rosenbergii and Photobacterium angustum S14 [52].

S2753 was cultured in each 30 mL of (i) marine minimal medium (MMM) [70] containing 0.4% glucose and 0.3% casamino acids; (ii) Marine Broth 2216 (MB; Difco 2216); (iii) Sigma Sea Salts (SSS; Sigma S9883; 40 g L⁻¹) containing 0.4% glucose and 0.3% casamino acids, and (iv) MMM containing 0.2% colloidal chitin [71,72] to investigate the best conditions for production of antibacterial compounds. All cultures were incubated aerated (200 rpm) for 72 h at 25 °C. All cultures were extracted with an equal volume of ethyl acetate (EtOAc). The extract was evaporated under nitrogen until dryness and redissolved in 300 μL 80% v/v ethanol (EtOH) in water. 20 μL of the extract was tested in an agar diffusion assay where expression from promoters of hla, rnaIII, and spa is monitored [28].

For further screening, the culture broth of S2753 grown in MMM (1 L, 72 h, 25 °C, 100 rpm) was extracted directly with sterile Diaion HP20SS (12 g L⁻¹, 24 h) (Sigma-Aldrich, St. Louis, MO). The resin was filtered off and washed with 80% (v/v) acetonitrile (MeCN)/water (300 mL) and the extract evaporated until dryness on a rotary evaporator. From this extract, 10 mg dry material was subjected to an explorative solid-phase extraction (E-SPE) protocol [48]. This yielded 15 fractions for re-testing in the agar diffusion assay as described above. The E-SPE fractions were also tested for antibacterial activity against Vibrio anguillarum strain 90-11-287 and S. aureus 8325 in a well diffusion agar assay [73].

Using 10 L glass fermentors, S2753 was cultured in 5 × 4 L SSS (iii, above) containing 0.4% glucose and 0.3% casamino acids (25 °C, 72 h, 100 rpm) as this medium gave comparable yields to that of MMM (i) but at a lower cost. The broth was extracted directly with Diaion HP20SS (12 g L⁻¹) as described above. The extract (3.4 g) was redissolved in EtOAc, absorbed onto 5 g Isolute diol (Biotage, Uppsala, Sweden), and added to a glass column with pure diol (95 g). A total of 12 fractions were collected from the diol column (100 g, 20 × 350 mm) ranging from heptane, dichlormethane (DCM), EtOAc, to pure methanol (MeOH), running under gravity. The fractions containing the QSI compounds (fraction 5, 20:80 (v/v) EtOAc/DCM and fraction 6, 40:60 (v/v) EtOAc/DCM) were separated on Sepra ZT C₁₈ (Phenomenex, Torrance, CA)(10 g SNAP) on an Isolera automated flash system (Biotage) using a MeCN/water gradient 25–75% over 20 min (12 mL min⁻¹). Pure compounds were obtained directly: Solonamide A (17 mg) and B (201 mg). Activity of pure compounds was tested in the agar diffusion assay as described above (20 μL per well), with a final concentration of compounds of 5 mg mL⁻¹ in dimethyl sulfoxide (DMSO).

NMR spectra were recorded on a Varian Unity Inova 500 MHz spectrometer equipped with a 5 mm probe using standard pulse sequences. The signals of the solvent were used as internal references (δ_H 2.49 and δ_C 39.5 ppm for DMSO). Carbonyl shifts were confirmed with ¹³C 1D on a Bruker Avance 800 MHz spectrometer with a 5 mm TCI cryoprobe at the Danish Instrument Center for NMR Spectroscopy of Biological Macromolecules.

LC-MS and LC-MS/MS analyses were performed on a maXis quadrupole time of flight mass spectrometer (Bruker Daltonics, Bremen, Germany) equipped with an electrospray (ESI) ion source. The MS was connected to an Ultimate 3000 UHPLC system (Dionex, Sunnyvale, CA) equipped with a diode-array detector. Separation was performed at 40 °C on a 150 mm × 2.1 mm ID, 2.6 µm Kinetex C₁₈ column (Phenomenex) using a linear water/MeCN (both buffered with 20 mM formic acid) gradient starting from 15% MeCN and increased to 100% in 13 min at a flow of 0.4 mL min⁻¹. The MS and MS/MS experiments were performed in ESI⁺ with a data acquisition range of m/z 100–1200 with collision energy of 27 V. The MS was calibrated using sodium formate automatically infused prior to each analytical run, providing a mass accuracy of below 1 ppm in MS mode and 2 ppm in MS/MS mode.

The absolute configuration of the amino acids were found using acid hydrolysis (6 M HCl, 110 °C, 20 h) [74] and derivatisation with Marfey’s reagent (1-fluoro-2,4-dinitrophenyl-5-l-alanine amide, FDAA, Sigma-Aldrich) following the protocol by Bonnard et al. [75]. Ultra-high liquid chromatography-diode array (UHPLC-UV) analyses of the amino acids were done on a Dionex RSLC Ultimate 3000 (Dionex) equipped with a diode-array detector. Separation was obtained on a Kinetex C₁₈ column (150 × 2.10 mm, 2.6 μm, Phenomenex) maintained at 60 °C using a linear gradient starting from 25% MeCN in water (both buffered with 50 ppm TFA) increasing to 27% MeCN over 6 min at a flow rate of 0.8 mL min⁻¹. Retention times of the FDAA amino acid derivatives used as standards were as follows (maximum standard deviation ± 0.002 min): FDAA (1.50 min), L-Ala (1.14 min), D-Ala (1.61 min), L-Phe (3.58 min), D-Phe (5.04 min), L-Leu (3.77 min), D-Leu (5.49 min), comparable to the observed retention times from the solonamide-derived amino acids.

To specify the stereochemistry of enantiomeric amino acids, the depsipeptides were reduced by LiBH₄. The resulting linear peptides were subjected to the above mentioned acid hydrolysis and Marfey’s derivatisation. Details are given in the Supplementary Information.

For the absolute configuration of the fatty acid residues, the (R)- and (S)-Mosher’s esters were prepared for both depsipeptides, and the stereocenters were assigned based on their ¹H and ¹⁹F chemical shift differences (Δδ^SR) [76,77]. Details are given in the Supplementary Information.

Solonamide A: white amorphous solids; UV (MeCN/H₂O) λ_max 200 nm (100%); for NMR data refer to Table 1; HRESIMS m/z 558.3486 (calcd for C₃₀H₄₆N₄O₆, 558.3496).

Solonamide B: white amorphous solids; UV (MeCN/H₂O) λ_max 200 nm (100%); for NMR data refer to Table 1; HRESIMS m/z 586.3725 (calcd for C₃₂H₅₀N₄O₆, 586.3730).

To investigate the potential production of agr inhibitors in related Photobacterium strains, the metabolite profile of S2753 was compared by liquid chromatography-diode array/mass spectrometry (LC-UV/MS) to P. halotolerans (LMG 22194^T), P. rosenbergii (LMG 22223^T), and P. angustum (S14) described by de Nyset al. [52]. All strains were grown in 30 mL MMM containing 0.4% glucose and 0.3% casamino acids (25 °C, 72 h, 200 rpm). Cultures were extracted with an equal volume of EtOAc and evaporated under nitrogen. Residues were redissolved in MeOH for LC-UV/MS analyses and in 80% EtOH for bioassay testing. Inhibition of agr was tested as described above. Extracts were also tested against V. anguillarum strain 90-11-287 for growth inhibition. LC-UV/MS analyses were performed on an Agilent 1100 liquid chromatograph with a diode array detector (Agilent, Waldbronn, Germany) coupled to an LCT TOF mass spectrometer (Micromass, Manchester, UK) using a Z-spray ESI source. Separation was obtained on a Luna II C₁₈ column (50 × 2 mm, 3 μm, Phenomenex) fitted with a security guard system using a linear gradient starting from 15% MeCN in water (both buffered with 20 mM formic acid) increasing to 100% MeCN over 20 min at a flow rate of 0.3 mL min⁻¹.

Northern blot analysis was performed as described previously [78]. The strains used were S. aureus FPR 3757 [55], a CA-MRSA USA300 obtained from ATCC (Boras, Sweden), and 8325-4 [17]. Samples for RNA purification were taken from cultures in Tryptone Soya Broth (TSB, Oxoid, Greve, Denmark) shaking at 185 rpm at 37 °C in a water bath (10 mL culture in 100 mL Erlenmeyer flask). Growth was monitored by measuring optical density at OD₆₀₀. Start inoculum was OD₆₀₀ = 0.03. Solonamides were added at OD₆₀₀ = 0.4. Samples for RNA purification were taken at OD₆₀₀ = 0.7 and 1.7. Probes targeting rnaIII, spa, and hla transcripts were amplified by PCR using primers hla forward (5′-GGG TTA GCC TGG CCT TCA GCC-3′), hla reverse (5′-GGG TGC CAT ATA CCG GGT TC-3′), spa forward (5′-GGG GGT GTA GGT ATT GCA TCT G-3′), spa reverse (5′-GGG GCT CCT GAA GGA TCG TC-3′), rnaIII forward (5′-GGG GAT CAC AGA GAT GTG ATG-3′), and rnaIII reverse (5′-GGG CAT AGC ACT GAG TCC AAG G-3′)(TAG Copenhagen A/S, Denmark). The resulting PCR fragments were 311 bp (hla), 719 bp (spa), and 316 bp (rnaIII), respectively.