Gram-negative and Gram-positive bacterial strains including A. baumannii ATCC 19606, P. aeruginosa ATCC 27853, E. coli ATCC 25922, K. pneumoniae ATCC 13883, methicillin-sensitive S. aureus ATCC 25923 (MSSA), methicillin-resistant S. aureus ATCC 43300 (MRSA), and several clinical strains of multidrug-resistant were kindly provided by Prof. Dai-Jie Chen from the Shanghai Jiao Tong University. Chemicals, media, and antibiotics were purchased from Sigma-Aldrich (Poole, United Kingdom) unless otherwise stated. Collismycin A and other two analogs were obtained from Alfa Chemistry (New York, United States) and Shenzhen GenProMetab Biotechnology Company.

The strain Kibdelosporangium phytohabitans XY-R10 was isolated from the root sediments (3–5 cm) of a mangrove plant Kandelia candel (L.) Druce, collected from Mai Po Inner Deep Bay Ramsar Site (E 114.05°, N22.49°; Hong Kong, China). The bacterium was cultivated in multiple 250 mL Erlenmeyer flasks each containing 80 mL of SGTPY medium (5 g starch, 5 g glucose, 1 g tryptone, 1 g peptone, 1 g yeast extract, and 17 g sea salts dissolved in 1 L of distilled water) at 28°C with agitation of 200 rpm for 6 days. The culture broth (10 L) was extracted with a double volume of ethyl acetate (EtOAc) three times. The combined EtOAc layers were dried by an evaporator to give a total crude extract (1.5 g), which was then applied to ODS column eluted with a step-gradient of water, (4:1) water/methanol (MeOH), (3:2) water/MeOH, (2:3) water/MeOH, (1:4) water/MeOH, and 100% MeOH, yielding six fractions. The fractions were tested in algicidal bioassays and the active fraction (4:1) water/MeOH was separated further using a Phenomenex Kinetex column (250 mm × 10.0 mm, 5 μm; ACN-H₂O, 25:75, 4 mL/min) on a semi-preparative HPLC (waters, 2495 series equipped with a photodiode array detector), which eventually afford the active compound 5 mg of MaiA.

To resolve the chemical structure of MaiA, NMR experiments were carried out on a Bruker Avance III 600 MHz spectrometer with DMSO-d₆ as the solvent and the data were analyzed with Bruker Topspin software. Accurate mass spectra of MaiA was recorded using an Agilent 6545 Q-TOF mass spectrometer coupled to a 1260 HPLC system. Single-crystal data were measured on an Agilent Xcalibur diffractometer with an Atlas (Gemini ultra Cu) detector.

Minimum inhibitory concentration (MIC) was determined by broth microdilution according to the Clinical and Laboratory Standards Institute guidelines. Bacteria were cultured in cation-adjusted Mueller Hinton broth at 35 ± 2°C. The MIC was defined as the lowest concentration of antibiotics with no visible growth. In order to determine the effects of combinations of MaiA with antibiotics, combinations of various dilutions of MaiA and a second drug were tested for growth inhibition by microdilution. The synergistic effect was evaluated by the fractional inhibitory concentration (FIC) index (Odds, 2003).

An anti-biofilm formation assay was performed as previously described with minor modifications (Nair et al., 2016; Park et al., 2016). Briefly, an overnight culture of each strain was diluted in MH medium (for Gram-negative strains), or LB supplemented with 0.5% glucose (for Gram-positive strains) to achieve an OD₅₉₅ value of 0.1. Then incubated statically in a 24-well flat-bottom polystyrene plate (Jet Bio-Filtration, Co., Ltd., China) with or without the varying concentrations of compounds at 37°C for 24 h. Subsequently, planktonic cells were discarded and the wells were rinsed twice with sterile PBS gently and then dried. Biofilms were stained with 0.1% crystal violet (CV) and dissolved in 30% acetic acid, and absorbance at 570 nm (OD₅₇₀) was measured to quantify biofilm biomass. The minimum biofilm inhibitory concentration (MBIC) was defined as the minimum concentration of compounds or antibiotics showing no color development (Nair et al., 2016). The effect of different concentrations of MaiA on the growth of bacteria was monitored by measuring absorbance at 595 nm using a spectrophotometer.

The metabolic activity of biofilm was examined by a metabolic dye reduction assay using MTT [3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide] according to the previous method (Yin et al., 2019). The activated succinate dehydrogenase in the viable cell of biofilm could reduce the yellow tetrazolium salt (MTT) to formazan, and the resulting product forms a blue solution when dissolved in DMSO. The metabolic activity was correlated to the optical density at 570 nm.

Synergistic activity of MaiA combined with antibiotics was measured following anti-biofilm assay.

Maipomycin A dissolved in Mili-Q water (Merck) mixed with aqueous FeSO₄ or FeCl₃ solution at the molar ratio of 0:1 and 2:1, respectively. The mixture was then subjected to HPLC and HPLC-Q-TOF MS analysis. Pure MaiA was injected as the standard.

Effect of addition of exogenous iron on the inhibition of A. baumannii ATCC 19606 biofilm formation by MaiA was investigated. A. baumannii was treated with 52 μM MaiA for 0, 3, 6, and 12 h, respectively, followed by addition of Fe(II) and Fe(III) ions. Biofilm biomass was quantified by CV stain after 24 h incubation in 24-well flat-bottom plate.

Anti-biofilm formation assay was established on glass coverslips placed in 24-well flat-bottom plates. After removal of supernatant and rinsed twice with sterile PBS, biofilm cells were stained using the BacLight Live/Dead Viability Kit (L7007, Invitrogen, Carlsbad, United States) at 37°C for 15 min in the dark and imaged with laser confocal microscope Leica TCS SP8 (excitation 488 nm, emission 588 nm) with a 20 × objective. Confocal images were processed using LAS X software to reconstruct 3D views of the biofilm. COMSTAT software was used to calculate biomass (μm³/μm²), mean thickness (μm), and other parameters (Heydorn et al., 2000).

Data were presented as the mean value ± standard deviation of three replicates. Comparison of data between treatments and controls were carried out by one-way ANOVA followed by Tukey’s HSD test using SPSS15.0 software. A P value less than 0.05 was considered statistically significant.

In this study, we found the crude extract of K. phytohabitans XY-R10 was able to inhibit the biofilm formation of A. baumannii ATCC 19606. Following bioassay-guided fractionation, MaiA with strong anti-biofilm activities was obtained. To identify the chemical structure of MaiA, high-resolution mass, and high-field NMR spectral analyses were carried out (Supplementary Figures 1–7). The ¹H NMR spectrum of MaiA showed two singlet methyl signals at δ_H 3.41, 4.16, five doublet olefinic or aromatic signals at δ_H 7.57, 8.02, 8.20, 8.42, 8.77, 8.84, and one exchangeable proton signal at δ_H 11.73. The ¹³C NMR and ¹H NMR data revealed that MaiA contained two methyl groups (δ_C 44.72, 57.25) and eleven olefinic or aromatic carbon atoms (Supplementary Table 1). These spectroscopic features suggested that MaiA belonged to the family of 2, 2′-bipyridyl, and was most similar to pyrisulfoxin A (PyrA), which was isolated as an antibiotic (Tsuge et al., 1999). The only significant differences in the NMR spectra between these two compounds were the chemical shifts of C-5 (δ_C 124.5 in 1 vs. δ_C 127.7 in PyrA) and C-9 (δ_C 44.72 in 1 vs. δ_C 39.4 in PyrA). Deduced from the positive ion [M + H]⁺ observed at m/z 308.0698 by Q-TOF MS, the molecular formula of MaiA was then determined to be C₁₃H₁₃N₃O₄S, which has one more oxygen atom than that of PyrA (C₁₃H₁₃N₃O₃S). The increase of one oxygen atom, together with the downfield shifts of C-5 and the upfield shifts of C-9, indicated the presence of a sulfonyl group in MaiA, instead of the sulfoxide group in PyrA. HMBC analysis connected the methyl protons H-9 (δ_H 3.41) to the quaternary carbon C-5 (δ_C 124.5) of the aromatic ring (Figure 1 and Supplementary Figure 1), for which the methyl sulfonyl group was allocated to the C-5 position. The planar configuration of MaiA was also confirmed by the analysis of the X-ray single-crystal diffraction data (Supplementary Tables 2–8 and Supplementary Figure 8). These data collectively suggested that the anti-biofilm compound isolated from K. phytohabitans XY-R10 was 4-methyloxyl-5-methylsulfonyl-2,2′-bipyridyl-6-carboxaldehyde oxime, a novel compound which we named maipomycin A, abbreviated as MaiA.

The antibacterial ability of Mai A was evaluated against a panel of Gram-positive and Gram-negative organisms. MIC values indicated that MaiA showed no antibacterial effect against most bacteria (MIC > 256 μg/mL), but had a weak antibacterial activity against A. baumannii (MIC = 128 μg/mL) including the reference strains and clinical isolates (Supplementary Table 9).

Maipomycin A was tested in preventing biofilm formation of pathogens using a CV staining method to quantify the production of biofilms biomass. It showed an effective inhibitory activity against the Gram-negative bacteria biofilm formation with a dose-dependent manner from 2 to 64 μg/mL (Figure 2C). The MBIC value of MaiA for A. baumannii ATCC 19606 and P. aeruginosa ATCC 27853 was 8 and 32 μg/mL, respectively. The percentage of biofilm biomass at MBIC was reduced by 84.3% for A. baumannii, and 82.6% for P. aeruginosa compared with control. It was notable that MaiA only had a slight inhibition on the planktonic cell growth of these two strains at MBIC concentration (Figures 2A,B), but it could inhibit biofilm formation effectively (Figures 2E,F).

In order to investigate whether MaiA has a broad anti-biofilm formation ability, several strains were selected for further investigation. Results displayed that MaiA could inhibit most Gram-negative bacteria biofilm formation and MBICs ranging from 8 to 32 μg/mL (Supplementary Figures 9A,D–I). However, MaiA had no anti-biofilm activity against S. aureus at concentrations up to 64 μg/ml (Supplementary Figures 9B,C).

The metabolic activity of the biofilm cells was measured by MTT staining, which is commonly used for the quantitative analysis of biofilm. The cell metabolic activity of A. baumannii and P. aeruginosa biofilm decreased significantly and dose-dependently when treated with MaiA (Figure 2D). The MBIC value of MaiA was 8 μg/mL for A. baumannii, and 32 μg/mL for P. aeruginosa determined by MTT stain. The metabolic activity of both strains was reduced by more than 80% at MBIC.

Confocal laser scanning microscopy (CLSM) was employed to analyze changes in biofilm formation. The reconstructed three-dimensional biofilm images revealed that MaiA at MBIC markedly inhibited biofilm formation compared to control (Figure 3A). The calculation results of COMSTAT software also confirmed the effect of biofilm inhibition. MaiA reduced the biomass (μm³/μm²) and mean thickness (μm) of A. baumannii ATCC 19606 biofilms by >90% and >80% for P. aeruginosa ATCC 27853 (Figure 3B). Furthermore, the increases in roughness coefficient and surface to biovolume ratio suggested the heterogeneity and incompleteness of biofilm development. These results also showed that MaiA was an effective agent to inhibit biofilm formation against both A. baumannii and P. aeruginosa.

2,2′-Bipyridine-containing compounds such as caerulomycin A (CaeA; Figure 1) exerts its immunosuppressive effect by depleting intracellular irons (Kaur et al., 2015), and ColA (Figure 1) acts as an iron chelator to inhibit tumor cell growth (Kawatani et al., 2016). We, therefore, hypothesized that MaiA might exert its antibiofilm activity through chelating irons, which were essential for both growth and biofilm formation for most bacteria (Schaible and Kaufmann, 2004). In order to prove this hypothesis, we carried out a series of experiments.

Firstly, we found that MaiA indeed chelated Fe(II) ions to form a purple complex (Figure 4A). HPLC-MS analysis further confirmed that this complex was formed by two molecules of MaiA and one molecule of Fe (Figures 4B,C). The isotope pattern of this complex in the mass spectrum was also very characteristic for compounds containing an iron atom. The confirmation of MaiA as an iron chelator suggests the bacterium might have originally produced this compound as a siderophore to acquire irons.

Secondly, the addition of exogenous iron together with MaiA could rescue A. baumannii ATCC 19606 biofilm formation in a dose-dependent manner (Figures 4D,E). MaiA significantly impaired biofilm formation without the addition of iron. The inhibition was attenuated when the mole ratio of MaiA and iron was 2:1 or 1:1 owing to MaiA completely chelated the additional iron. Moreover, the biofilm biomass did not decrease if Fe(II) was added 3 h or 6 h after treatment of MaiA (Figure 4F). The rescue experiment suggested irons were associated with the MaiA’s anti-biofilm activity.

Maipomycin A could be deactivated regardless of Fe(II) or Fe(III) ions were added. We found that MaiA forms a complex with Fe(III) ions more slowly than Fe(II) ions (Supplementary Figures 10B,C). The color and ultraviolet-visible spectra between (MaiA)₂-Fe(II) and (MaiA)₂-Fe(III) are slightly different (Figure 4A and Supplementary Figure 10). Moreover, it seemed difficult to distinguish (MaiA)₂-Fe(II) from (MaiA)₂-Fe(III) using HPLC or mass spectrum analysis (Supplementary Figure 13).

We also investigated the anti-biofilm activity of MaiA’s analogs and other iron chelators. ColA showed strong anti-biofilm activity. The MBIC values of ColA for A. baumannii ATCC 19606 and P. aeruginosa ATCC 27853 were 8 and 16 μg/mL, respectively, (Figure 5 and Supplementary Figure 11). The activity of the other analogs maipomycin B (MaiB) and 4-Methoxy-5-methylsulfanyl-[2,2′]bipyridinyl-6-carbaldehyde were much weaker than MaiA and ColA. MaiA and ColA both have a 6-carboxaldehyde oxime group, different from the 6-carbaldehyde group of MaiB and 4-Methoxy-5-methylsulfanyl-[2,2′]bipyridinyl-6-carbaldehyde. Therefore, the activity may be related to 6-carboxaldehyde oxime. As two well-known iron chelators, the biofilm inhibition activity of ethylene diamine tetraacetic acid (EDTA) and 2,2′-dipyridyl (2DP) was weaker than MaiA or ColA.

Different classes of antibiotics were combined with MaiA for screen drug interaction against a panel of Gram-positive and Gram-negative organisms. We found that MaiA itself only displayed a weak antibacterial activity but it could potentiate colistin against A. baumannii efficiently (Supplementary Table 11). A dose-response study with MaiA was performed against all four A. baumannii strains at concentrations of 32, 16, and 8 μg/mL. The MIC of colistin reduced 4–8 folds when used in combination with MaiA, and a synergistic effect between these two compounds could be observed (FICIs ≤ 0.5; Supplementary Table 12). MaiA also enhanced the anti-biofilm activity of colistin yet with only an additive effect. Furthermore, a similar synergistic effect was also observed between colistin and the other two iron chelators ColA and 2DP (Supplementary Table 13).

When MaiA and Fe were treated with a molar ratio of 2:1 after combined with colistin, the enhancement effect completely disappeared (data not shown), indicating that the iron ions might be an interrupter and the MaiA-Fe complex seems not to be the potentiator.