This study does not involve any human or animal participants nor does the study involve any invasion of privacy or accessing confidential information of individuals. The ethical committee of Yeungnam University has granted the exemption of ethical approval.

A. baumannii ATCC 17978 and eight clinical A. baumannii isolates (ATCC BAA-1709, A 550, A 578, A 553, A 556, A 580, A 571, A 564) were obtained from burns patients at the National Rehabilitation Institute of Mexico; A. baumannii ATCC 17978 was used as a reference strain (Cruz-Muniz et al., 2017). For the dual biofilm experiment, we used C. albicans DAY185 (obtained from the Korean Culture Center of Microorganisms¹) and A. baumannii ATCC 17978. All experiments were conducted at 37°C, and trypticase soy broth (TSB) and potato dextrose broth (PDB) media were used for the biofilm assay, Luria-Bertani (LB) medium for the pellicle assay, and motility agar (MA) medium in the motility experiment. Chemicals including twelve flavonoids viz. flavone (99%), 6-aminoflavone (97%), 6-hydroxyflavone (98%), apigenin (97%), chrysin (97%), curcumin (94%), daidzein (98%), fisetin (98%), genistein (98%), luteolin (98%), phloretin (99%), and quercetin (98%), gallium nitrate (99.9%), and crystal violet (90%) were purchased from Sigma-Aldrich Co. (MO, United States). The structures of these flavonoids are provided in Figure 1A. TSB, PDB, LB media, and ethanol (95%) were purchased from Becton Dickison and company (NJ, United States) and dimethyl sulfoxide (DMSO) from Duksan Pure Chemicals (Daegu, South Korea), respectively. All 12 flavonoids solutions were prepared by diluting them in DMSO that was also used as a negative control.

A. baumannii initially was streaked from −80°C glycerol stock on trypticase soy agar (TSA) plates, and a single fresh colony was inoculated in TSB (2 ml) in 14-ml tubes and incubated at 37°C and 250 rpm (Lee et al., 2012). Overnight cultures were reinoculated at 1:100 dilution in TSB. For cell growth measurements, a 1:100 inoculum was cultured in TSB (25 ml) in 250 ml flasks and incubated at 37°C overnight with agitation at 250 rpm. Optical densities were then measured at 600 nm using a spectrophotometer (Optizen 2120 UV, Mecasys, South Korea). Streaking and subculturing of C. albicans DAY185 was performed using potato dextrose agar (PDA) or PDB, unless otherwise specified (Manoharan et al., 2017). C. albicans was preserved at −80°C in 1 ml of PDB supplemented with 30% glycerol, and when needed, streaked on PDA plates. Plates were incubated for 48 h at 37°C and a fresh single colony was then inoculated into 25 ml of PDB and cultured overnight at 37°C. A minimum two independent cultures were used for each experiment.

MICs were determined (Betts and Wareham, 2014; Singh, 2014; Singh et al., 2017) using the National Committee for Clinical Laboratory Standards (2002) susceptibility testing guidelines in 96-well microtiter plates (SPL Life Sciences, Pocheon, South Korea). Briefly, an overnight culture at a dilution of 1:100 inoculated in TSB was cultured for 24 h in the presence of curcumin (100, 200, or 500 μg/ml) at 37°C. After incubation cultures were spread on TSA plates, incubated for 24 h at 37°C, and cell colonies were counted. Experiments were performed using at least two independent cultures.

Static biofilm formation was assayed in 96-well polystyrene plates as previously reported (Lee et al., 2014). Briefly, cells in TSB (total volume 300 μl) inoculated at initial turbidity of 0.05 at 600 nm (OD₆₀₀) were cultured with or without for 24 h without shaking at 37°C. The same amount of TSB was added to peripheral wells of 96-well plate to avoid edge effects. To quantify total biofilm formation, biofilms in 96-well plates were stained with 0.1% crystal violet for 20 min, dissolved in 95% ethanol, and absorbances were measured at 570 nm (OD₅₇₀). Cell growth in 96-well plates was also measured at 620 nm (OD₆₂₀). For initial anti-biofilm screening, we tested all twelve flavonoids at a concentration of 50 μg/ml; results presented are averages of at least six replicate wells. Another static biofilm formation assay was performed in 14 ml polyethylene tube, as previously described (Pour et al., 2011). A. baumannii ATCC 17978 cells were inoculated 1:100 in 5 ml of TSB medium with curcumin at 0, 10, 20, 50, and 100 μg/ml and incubated for 24 h without shaking. Ring biofilms in 14 ml polyethylene tubes were stained with crystal violet and results presented are the averages of at least three repetitions.

Acinetobacter baumannii was inoculated at an OD₆₀₀ of 0.05 in 3 ml of TSB in glass bottomed confocal dishes (SPL life Sciences, Pocheon, South Korea) for 24 h at 37°C with curcumin at 0, 10, and 50 μg/ml without shaking. To visualize biofilm structures, cells were stained with carboxyfluorescein diacetate succinimidyl ester (Invitrogen, Molecular Probes, Inc, Eugene, OR, United States). Biofilm structures were evaluated by CLSM (Nikon Eclipse Ti, Tokyo, Japan) (Lee et al., 2014), and their spatial characteristics were quantified using COMSTAT biofilm program² by analyzing at least four random positions in three independent cultures. To measure biofilm formation, color confocal images (20 image stacks) were converted to gray scale using ImageJ program³. COMSTAT biofilm software was used to determine biomasses (μm³ per μm²), mean thicknesses (μm), and substratum coverages (%) (Runci et al., 2017).

A. baumannii can form pellicles more readily at air-liquid interfaces than other pathogenic Acinetobacter species (Chabane et al., 2014). The pellicle formation assay used was a modification of a previously described protocol (Marti et al., 2011). In brief, overnight bacterial cultures were diluted 1:100 in 5 ml of LB broth and grown in glass tubes for 72 h at 25 and 37°C in the dark without agitation (Mussi et al., 2010). Amounts of pellicle material were assessed by adding 1 ml of ethanol to tube underneath pellicle material, removing floating pellicles, and resuspending them in phosphate buffer saline (PBS) as previously reported (Giles et al., 2015). OD₆₀₀ values were measured using a spectrophotometer (Optizen 2120 UV, Mecasys, South Korea). Experiments were conducted in triplicate on three different days.

To assess surface motility with different concentrations of agar, MA containing 0.4% agarose, 1% tryptone, and 0.5% yeast extract was used (Clemmer et al., 2011), and MA medium supplemented with 0.25% agar (Eijkelkamp et al., 2011). Curcumin at 10 and 50 μg/ml concentration was added to MA, and DMSO (0.1%) was used as a negative control. Overnight grown ∼0.2 μl cultures of A. baumannii ATCC 17978 and three other multi-drug resistant clinical isolates (A 550, A 556, and A 580) were placed on motility plates using a sterile pipette tip. Sizes of halos produced by cells traveling across agar plates were measured after 9 h of incubation at 37°C. Each experiment was performed using at least three independent cultures.

Because of its antibiofilm activity against C. albicans (Alalwan et al., 2017; Tan et al., 2018), we speculated curcumin would exhibit antibiofilm activity against a mixed culture of C. albicans and A. baumannii. Cells were inoculated together in PDB and TSB (50:50) mixed for C. albicans and A. baumannii at 1:50 (CFU ∼1 × 10⁷) and 1:100 (CFU ∼7 × 10⁷) dilution ratios from each with overnight cultures, respectively. Pure cultures of C. albicans and A. baumannii were tested at the same time. Biofilms in 96-well plates were stained with 0.1% crystal violet, dissolved 95% ethanol, and OD₅₇₀ values were used to quantify total biofilm formation. Cell growth in 96-well plate were determined using OD₆₂₀ values. Hyphal formation by C. albicans and mixed biofilm formation were assessed by SEM as previously described (Lee et al., 2014). Briefly, small pieces (0.5 cm × 0.5 cm) of nylon filter were placed in wells of 96-well plates containing 300 μl cells/well. Cells were incubated in the absence or presence of curcumin at 37°C for 24 h without shaking. Prior to observation, biofilm samples were fixed with 2.5% glutaraldehyde and 2% formaldehyde for 24 h, serially post fixed in PBS and osmium tetroxide, and dehydrated using an ethanol series (50, 70, 80, 90, 95, and 100%) and isoamyl acetate. After critical-point drying, cells on filters were sputter-coated with palladium/gold and observed under an S-4100 scanning electron microscope (Hitachi, Tokyo, Japan) at magnifications ranging from x 1,000 to 10,000 using an accelerating voltage of 15 kV.

The molecular docking assays was conducted as previously described (Russo et al., 2016; Rajasekharan et al., 2017). Docking studies were performed to evaluate interactions between all twelve flavonoids and BfmR binding sites of A. baumannii (Protein Data Bank 6BR7). The three-dimensional structure of the beryllium fluorinated (BeF₃^–) receiver domain of A. baumannii BfmR resolved at 1.86 Å was used for docking simulation. This BfmR domain consisted of two chains (A and B) and has a sequence length of 133 amino acids. For grid generation, beryllium fluorinated ligand was used as the centroid and ligands were docked at positions proximal to BeF₃^– binding pockets (Draughn et al., 2018) using Schrodinger software 11.4 (Cambridge, MA, United States). A BfmR inhibitor 2-aminoimidazole and two biofilm inhibitors, virstatin and LED 209, were also docked at active sites. The glide score value and more specific target binding interactions to Asp15 and Asp58 were recorded.

The C. elegans killing assay used was a modification of a previously described protocol (Beceiro et al., 2014). Briefly, non-infected nematodes (∼20–30) [fer-15(b26);fem-1(hc17)] were pipetted into 96-well plate containing M9 buffer and overnight curcumin (50 μg/ml) treated and untreated with A. baumannii and/or C. albicans cells. As a second dose curcumin was added to respective wells to make final concentration 50 μg/ml (total volume 300 μl). Nematodes were incubated at 25°C and viabilities were determined as previously described (Rajsekharan et al., 2018), by exposing them to LED or UV LED lights for 10–30 s using an iRiS^TM Digital Cell Imaging System (Logos BioSystems, South Korea). Three independent experiments (n = ∼20–30) were conducted.

Replication numbers for assays are provided above and results are expressed as means ± standard deviations. The statistical analysis was performed by one-way ANOVA followed by Dunnett’s test using SPSS version 23 (SPSS Inc., Chicago, IL, United States). P values of < 0.05 were regarded significant and asterisks are used to indicate significant differences between treated and untreated samples.

The effects of the 12 flavonoids (Figure 1A) were initially investigated on A. baumannii ATCC 17978 biofilm formation in 96-well polystyrene plates using a crystal violet assay. Of the 12 flavonoids, luteolin, genistein, quercetin, fisetin, phloretin, and curcumin at 50 μg/ml exhibited biofilm inhibition, whereas 6-aminoflavone, apigenin, 6-hydroxyflavone, and chrysin increased biofilm formation, and the backbone flavone and daidzein had little effect (Figure 1B). Three flavonoids, that is, fisetin, phloretin, and curcumin, significantly and dose-dependently reduced biofilm formation, for example, these three flavonoids at 10 and 100 μg/ml reduced biofilm formation by A. baumannii ATCC 17978 by >45 and >86%, respectively (Figure 1C).

Next, we investigated the antibiofilm activities of fisetin, phloretin, and curcumin against eight clinical A. baumannii strains (Cruz-Muniz et al., 2017). Of these strains, ATCC BAA-1709, A 550, A 578, and A 553 were highly biofilm-forming, A 556, A 580, A 571 were intermediate, and A 564 had poor biofilm forming ability. Interestingly, fisetin, phloretin, and curcumin at 50 μg/ml all inhibited biofilm formation by the multidrug resistance (MDR) strains A 550, A 556, A 580, and A 564, and the antibiotic-sensitive ATCC BAA-1709 and A 571 strains. On the other hand, biofilm formation by two MDR strains, that is, A 553 (sensitive to colistin and amikacin), and A 578 (sensitive to colistin, imipenem, and meropenem) were not affected by fisetin, phloretin, or curcumin (Figure 2). Of these three flavonoids, curcumin was the most effective biofilm inhibitor against tested A. baumannii strains, and thus, it was the focus of subsequent studies conducted using the reference A. baumannii ATCC 17978 strain.

The antibiofilm activity of curcumin was compared with that of gallium nitrate (a known biofilm inhibitor) (Runci et al., 2017). Both curcumin and gallium nitrate dose-dependently inhibited biofilm formation by the ATCC 17978 strain, though curcumin was superior to gallium nitrate at same concentrations (Figures 3A,B). For example, curcumin at 20 or 100 μg/ml decreased biofilm formation in polystyrene 96-well plate by 46 and 93%, respectively, whereas gallium nitrate at these concentrations decreased biofilm formation by 24 and 67%, respectively. Biofilm formation was also assessed in polyethylene tubes, as previously described (Pour et al., 2011), and curcumin was found to dose-dependently inhibit ring biofilm formation by A. baumannii (Figure 3D). We also examined the effect of curcumin on planktonic cell growth. The MICs of curcumin and gallium nitrate against ATCC 17978 strain were determined to be >500 and >1000 μg/ml, respectively. It was difficult to determine an exact MIC as curcumin precipitated from solution at higher concentrations, as previously reported (Betts and Wareham, 2014). Curcumin at concentrations up to 200 μg/ml slightly reduced (by ≤38%) the planktonic cell growth of A. baumannii under shaking conditions in a flask (Figure 3C). These results indicated that antibiofilm activity of curcumin was not due to its antimicrobial activity, indicating curcumin may less prone to the development of drug resistance.

The antibiofilm effect of curcumin was further confirmed by CSLM and COMSTAT analysis. Interestingly, A. baumannii ATCC 17978 formed relatively thin surface biofilms on glass surfaces (Figure 3E) but robust ring biofilms on polyethylene (Figure 3D). Furthermore, the antibiofilm activity of curcumin was more marked on glass than polystyrene (Figure 3A) and polyethylene (Figure 3D) surfaces. For example, curcumin at 10 μg/ml markedly reduced surface biofilm formation on glass (Figure 3E), whereas at 100 μg/ml curcumin was more effective at preventing ring biofilm formation on polyethylene (Figure 3D). Biofilm reduction was also confirmed by COMSTAT analysis, which showed curcumin at 10 or 50 μg/ml significantly reduced biofilm biomasses, average thicknesses, and substrate coverage (Figure 3F). Specifically, biofilm biomass, thickness, and substrate coverage were reduced by curcumin at 10 μg/ml by >95% versus untreated controls.

A. baumannii colonizes the upper surfaces of static liquids and form biofilms at air-liquid interfaces by a process called pellicle formation, which is a type of biofilm formation (Kentache et al., 2017). The effect of curcumin on pellicle formation of ATCC 17978 and three other multi-drug resistant clinical isolates (A 550, A 556, and, A 580) were measured at 25 and 37°C in LB medium without shaking. After 24 h, a thin pellicle started to form at the liquid surface, and by the end of the third day, an opaque, solid pellicle covered the entire liquid surface (Figures 4A,B and Supplementary Figure S1D). Pellicle growth was found to be greater at 25°C, which concurs with the results of Marti et al. (2011), and to be significantly inhibited by curcumin at 50 μg/ml (Figure 4C and Supplementary Figure S1F).

Acinetobacter baumannii biofilm formation depends on the synthesis of pili, which are structures assembled by the csuA/BABCDE chaperone-usher secretion system (Luo et al., 2015). Curcumin at 10 μg/ml reduced surface motility on 0.4% agarose and 0.25% agar (Figures 5A,B). Surface motility of ATCC 17978 with 0.4% agarose measuring mean halo diameters at 9 h were 1.4 ± 0.6 cm for curcumin treatment and 5.6 ± 1.6 cm for the non-treated control. Similarly, surface motilities measured in 0.25% agar using mean halo diameters were 0.3 ± 0.2 cm for curcumin at 10 μg/ml and 6.6 ± 2.3 cm for the control. Also, three clinical isolates (A 550, A 556, and A 580) were motile, which was significantly inhibited by curcumin (Supplementary Figures S1A–C,E). This is interesting since pili play a role both in biofilm formation and motility that were markedly abolished by curcumin.

Biofilm formation is a survival policy for bacteria and fungi in challenging environments (Wu et al., 2015). Crystal violet biofilm and SEM assays were used to examine the inhibitory effects of curcumin on mixed biofilms of C. albicans and A. baumannii. To form these dual biofilms, we used a (50:50) mixed medium of PDB and TSB to enable C. albicans and A. baumannii growth, respectively. Under these conditions, decent biofilm formation (1.0∼3.0 at OD₅₇₀) of individual C. albicans and A. baumannii strain and also co-culture of two species was observed (Figures 6A,B). As previously reported by Alalwan et al. (2017), curcumin dose-dependently inhibited biofilm formation by C. albicans (Figure 6A), for example, at 10 μg/ml curcumin reduced its biofilm formation by >80%. Importantly, curcumin at 20 μg/ml reduced mixed biofilm formation by >85% (Figure 6B). SEM analysis showed that in mixed biofilms of C. albicans and A. baumannii, C. albicans formed large hyphae and few yeast cells, which were much larger than A. baumannii cells, and A. baumannii cells appeared to be encased in C. albicans hyphae (Figure 6C).

Biofilm inhibition in A. baumannii is under the control of a BfmR/S, which is a two-component system (Liou et al., 2014). In this part of the study, we investigated interactions between several flavonoids and the BeF₃^– domain of BfmR (Figure 7 and Supplementary Figure S2). The positive control, 2-aminoimidazole (Thompson et al., 2012) exhibited hydrogen bond interactions with Asp15 and Asp58 and a Pi-Pi stacking with Lys107 (Figure 7A). Also, two known biofilm inhibitors (virstatin and LED209) interacted with Asp16 (adjacent to Asp15) and Lys107 (Figures 7B,C). Curcumin was found to interact well with the active site with a binding energy of −38.7 kcal/mol (Supplementary Table S1). It formed backbone H-bonds with negatively charged Asp15 and two H-bonds with non-polar Val109 (Figure 7D). Fistein and quercetin also formed two H-bonds with Asp15 and Pi-Pi stacking with Lys107 (Figure 7F, Supplementary Figure S2, and Supplementary Table S1), while phloretin was found to interact better than fistein or quercetin with a binding energy of −41.8 kcal/mol, resulting from the formation of H-bonds with Asp15 and Asp58 and Pi-Pi stacking with Lys107 (Figure 7E, Supplementary Figure S2, and Supplementary Table S1). Several other flavonoids (6-aminoflavone, apigenin, 6-hydroxyflavone, luteolin, chrysin, flavone, and daidzein) were also tested which did not interact with the active site and showed poor interaction patterns (Supplementary Figure S2). Furthermore, our in vitro studies, showed these flavonoids did not inhibit biofilm formation (Figure 1B). Overall, BfmR binding efficacies of all 12 flavonoids were correlated with their antibiofilm efficacies. Based on in vitro and in silico findings, we speculate that the interaction of curcumin with BfmR could be one of the possible causes for its antibiofilm activity. However, further in vitro studies are required to confirm the curcumin/BfmR interaction.

Since A. baumannii and C. albicans kills the nematode C. elegans (Beceiro et al., 2014), a C. elegans killing assay was performed to examine the protective effect of curcumin. A. baumannii infection caused 80% C. elegans fatality (20% survival) in 5 days (Figure 8A), but the presence of curcumin at 50 μg/ml reduced this to 35% (Figure 8A and Supplementary Figure S3), while the cell numbers of A. baumannii are similar (Supplementary Figure S4). Also, curcumin significantly attenuated the virulence of C. albicans in the nematode. These results show that curcumin effectively reduced the virulence of A. baumannii and C. albicans in our nematode model. Interestingly, mixed infection of C. albicans and A. baumannii showed much less virulence than the single pathogenic infection on C. elegans. This result confirms that C. albicans and A. baumannii are antagonistic each other and reduce their virulence against C. elegans as reported in references (Kostoulias et al., 2016). In addition, we investigated the chemical toxicity of curcumin against uninfected C. elegans. After 4 days trial, curcumin treated nematodes showed similar trends like the non-treated controls (Figure 8B), confirming that curcumin was nontoxic to worms, and did not affect the survival rate.