Synergistic antibacterial effects of berberine, baicalin, and plumbagin against carbapenem-resistant Acinetobacter baumannii: Mechanistic exploration

2.1 Materials

Microbiological media and reagents: Mueller–Hinton (MH) agar and broth (cation-adjusted as needed; BD), phosphate-buffered saline (PBS; Sigma), crystal violet stain (Thermo Fisher), ethanol solutions (30–100%; Merck) for dehydration, glutaraldehyde (2.5%; Sigma) for fixation, and SYBR Green RT-qPCR kits (Abclonal).

Antibiotics, bacterial strains, and test compounds: Ten clinical CRAB isolates were obtained from the Taylor’s University laboratory. The isolates originated from the following departments: ICU (n = 4), respiratory (n = 2), haematology (n = 1), burn unit (n = 1), neurosurgery (n = 1), and nephrology (n = 1). Specimens included four respiratory samples (two bronchoalveolar lavage fluid and two sputum), two urine samples, one blood sample, one catheter-tip sample, one cerebrospinal fluid sample, and one wound-exudate sample (Table 1). For reference susceptibility testing, the following antibiotics were included: tigecycline, imipenem, meropenem, levofloxacin, amikacin, and polymyxin E (colistin). The test compounds were berberine (purity >98%), baicalin (purity >98%), and plumbagin (purity >98%). All standard compounds used in this study, including berberine, baicalin, and plumbagin, were purchased from the National Institutes for Food and Drug Control, China.

Computational tools: AutoDock Vina for molecular docking and GROMACS for MD simulations, PubChem for chemical information and ProTox-II for toxicity prediction.

2.2 Bacterial strains and culture

Ten carbapenem-resistant A. baumannii isolates were obtained from the university medical laboratory. Bacteria were cultured on MH agar and in broth under aerobic conditions at 35°C. Standard colony isolation procedures were used to obtain pure cultures for subsequent testing.

2.3 MIC determination

MICs of each compound and selected reference antibiotics were determined by cation-adjusted MH broth (CA-MHB) microdilution in 96-well plates, following CLSI M100 (2021). Berberine, baicalin, plumbagin, and reference antibiotics (tigecycline, imipenem, meropenem, levofloxacin, amikacin, and colistin) were prepared as twofold serial dilutions at 2× the final test concentration; 50 µL of each dilution was dispensed into each well. Next, 50 µL of bacterial suspension (1 × 10⁶ CFU/mL) was added to each well containing 50 µL of 2× drug dilution (final volume: 100 µL). This produced a final inoculum of 5 × 10⁵ CFU/mL per well, consistent with standard broth microdilution practice. Plates were incubated statically at 35°C for 18–20 h. All experiments were performed in triplicate.

2.4 Checkerboard synergy assay

To evaluate combinatorial effects, checkerboard assays were conducted for each compound pair and for the triple regimen.

Pairwise checkerboards: The concentrations of berberine and baicalin were set at 8–1024 µg/mL, and plumbagin at 2–256 µg/mL. Compounds at different concentrations were combined in pairs and added to 96-well plates containing equal volumes of CRAB suspensions. Plates were incubated at 35°C for 18 h. After incubation, FICI was calculated as: FICI = (MIC_A in combination/MIC_A alone) + (MIC_B in combination/MIC_B alone). FICI ≤ 0.5 indicated synergy; 0.5 < FICI ≤ 1 indicated an additive effect; 1 < FICI ≤ 4 indicated an indifferent interaction; and FICI > 4 indicated antagonism. All experiments were performed in triplicate, with growth controls included (Ju et al., 2022).

For the three-drug checkerboard assay, concentrations were set as follows: baicalin 0.5–64 µg/mL (8 levels), berberine 1–128 µg/mL (8 levels), and plumbagin 2–64 µg/mL (6 levels). Plumbagin was limited to six twofold dilution levels (2–64 µg/mL) because its single-agent MICs were lower than those of berberine and baicalin. Restricting plumbagin to six levels enabled complete three-drug testing in a 96-well format while still spanning sub-MIC to supra-MIC concentrations. Plates were incubated at 35°C for 18 h. Each isolate was tested in three independent experiments, including growth controls. For each triple combination, ΣFICI was calculated as: ΣFICI = (MIC_A in combination/MIC_A alone) + (MIC_B in combination/MIC_B alone) + (MIC_C in combination/MIC_C alone). Interpretation followed the criteria used for the pairwise checkerboard assay described above. An example layout of the 3D checkerboard design is shown in Fig. 3.

Close

Fig. 3.

Example of 3D checkerboard method.

Export to PPT

2.5 Time-kill curve

The bactericidal kinetics of the triple combination at the fixed synergistic ratios determined above were evaluated (Bozkurt-Guzel et al., 2020). Cultures were adjusted to approximately 5 × 10⁵ CFU/mL with CA-MHB and incubated at 35°C with shaking. Tests were performed at 0×MIC (control), 0.5×MIC, 1×MIC, 2×MIC, 4×MIC, and 8×MIC (based on the combined MIC). Samples were collected at 0, 1, 2, 4, 8, 12, and 24 h, serially diluted, and plated for colony enumeration (CFU/mL). Bactericidal activity was defined as a ≥3 log₁₀ CFU/mL reduction relative to the initial inoculum (0 h), corresponding to ≥99.9% killing.

2.6 Biofilm inhibition assay

Biofilm formation under compound exposure was quantified by crystal violet staining. Overnight cultures were diluted in fresh medium and dispensed into 96-well plates with serially diluted berberine (0–2048 µg/mL), baicalin (0–2048 µg/mL), or plumbagin (0–64 µg/mL), followed by static incubation at 35°C for 24 h to allow biofilm development. Wells were gently washed with PBS to remove planktonic cells. Adherent biofilms were fixed with methanol, stained with 0.1% crystal violet, and then rinsed. Bound dye was solubilized in ethanol, and absorbance was measured at 590 nm. Biofilm inhibition (%) was calculated as: [1 − (OD_treated/OD_control)] × 100.

2.7 SEM

For ultrastructural analysis, treated and untreated biofilms were grown on sterile coverslips and fixed with 2.5% glutaraldehyde (4°C, 12 h). SEM was performed for the untreated control and for single-compound-treated biofilms at selected concentrations. SEM of combination-treated biofilms was not performed because the triple-combination condition had already been evaluated by crystal violet biofilm quantification and time–kill assays, and SEM capacity was reserved for representative single-compound conditions. After fixation, samples were rinsed in PBS and dehydrated through a graded ethanol series (30–100%). Dried specimens were sputter-coated with gold and examined by SEM at 5–10 kV. High-magnification images were captured to assess biofilm architecture, cell morphology, and extracellular polymeric substances (EPS).

For image-based quantification, multiple random fields of view (1000×–5000×) were acquired per sample. Using ImageJ (Fiji), SEM images were converted to 8-bit, background-subtracted, thresholded (Otsu), and watershed-segmented to separate adjacent cells. Bacterial counts and total biofilm area were measured per field of view, excluding objects touching image borders. For consistency, image-analysis parameters were kept constant across all samples.

2.8 Molecular docking and dynamics

Interactions between each compound and A. baumannii OmpA were investigated in silico. Three-dimensional structures of the OmpA N-terminal β-barrel and C-terminal domain were obtained from the Protein Data Bank (PDB IDs: 1BXW and 3TD3, respectively). These Protein Data Bank entries represent individual domains rather than full-length OmpA. However, the N-terminal eight-stranded β-barrel fold is conserved among Gram-negative bacteria, providing a reasonable structural template for ligand-binding analyses. Structures of berberine, baicalin, and plumbagin were retrieved from PubChem. Docking was performed with AutoDock Vina to predict binding poses and affinities of each compound for OmpA domains. Furthermore, MD simulations (100–200 ns) were performed in GROMACS using the docked complexes to assess binding stability over time. Trajectories were analyzed for root-mean-square deviation (RMSD) and protein–ligand interactions (including hydrogen bonding and intermolecular distances), and binding free energies were estimated using Molecular Mechanics/Poisson–Boltzmann Surface Area (MM/PBSA) to compare interaction strengths among the three compounds.

2.9 RT-qPCR gene expression

CRAB cultures were treated with single compounds or the triple combination and harvested at logarithmic phase. Total RNA was extracted using FastPure (RC113-01), and cDNA was synthesized using ABScript III (RK20429) with a gDNA remover. qPCR was performed on an Archimed R4 using Genious 2× SYBR Mix (RK21205) in 20µL reactions. Cycling conditions were 95°C for 3 min; 43 cycles of 95°C for 5 s and 60°C for 30 s; followed by melt-curve analysis. The target gene was ompA, with 16S rRNA as the internal control. Primers: ompA forward 5′-TTGCACTTGCTACTATGCTTGTTG-3′, reverse 5′-TGGCTGTCTTGGAAAGTGTAACC-3′; 16S rRNA forward 5′-ACTCCTACGGGAGGCAGCAGT-3′, reverse 5′-TATTACCGCGGCTGCTGGC-3′. no-template control (NTC) and no -reverse transcription control (no-RT) controls were included, and each condition was run in technical triplicate. Relative expression was calculated by the 2^−ΔΔCt method; Statistical analyses were performed on delta Ct values (ΔCt = Ct_target − Ct_reference) using analysis of variance (ANOVA) with Tukey’s post hoc test (p < 0.05). Melt curves showed a single peak, and standard curves yielded 90–110% amplification efficiency (R² ≥ 0.99).

2.10 Toxicity prediction

In silico toxicity assessment of berberine, baicalin, and plumbagin was performed using the ProTox-II webserver (Charité University, Berlin) to estimate acute rodent toxicity, expressed as the median lethal dose (LD50), predict toxicity class (I–VI), and identify potential organ-specific toxicities. Physicochemical properties like LogP were also noted to infer absorption and distribution characteristics.

2.11 Statistical analysis

Statistical analyses were conducted in GraphPad Prism 10 with ≥3 biological replicates. Normality and homogeneity of variance were assessed using Shapiro–Wilk and Levene tests. When assumptions were met, one-/two-way or repeated-measures ANOVA with Tukey’s post hoc test was used; otherwise, Welch ANOVA with Games–Howell or Kruskal–Wallis with Dunn’s test was applied (two-tailed α = 0.05).

3.1 Antibacterial activity and synergistic effects

All three traditional Chinese medicine–derived compounds inhibited CRAB, although their single-agent activities differed. Among the clinical isolates, plumbagin showed the lowest MIC among the three compounds (16–128 μg/mL). Notably, the quality-control reference strain exhibited an MIC >512 μg/mL for plumbagin, which exceeded the MIC range observed among the clinical isolates. In contrast, berberine and baicalin required substantially higher concentrations when tested alone, with MICs around 1024 μg/mL that often approached the upper limit of the tested range, indicating relatively weak single-agent activity (Table 2).

Checkerboard assays revealed distinct interaction patterns among the compounds. Berberine plus baicalin showed consistent synergy across all isolates (FICI ≤ 0.5; Table 3). In contrast, baicalin plus plumbagin showed indifferent interactions (1.0 < FICI ≤ 4.0; Table 4), whereas berberine plus plumbagin showed predominantly additive effects in 9/10 isolates, with one isolate classified as indifferent (Table 5). No antagonism was detected for any pairwise combination under the tested conditions. The three-compound combination exhibited robust synergy across all isolates (∑FICI 0.188–0.406; Table 6). A summary of the interaction categories and combination MIC distributions is provided in Table 7.

Combination susceptibility testing showed reduced MICs for berberine/plumbagin and berberine/baicalin relative to the corresponding single-agent MICs. The most pronounced synergy occurred with the triple combination, exemplified by strain CRAB7, which yielded a ΣFICI as low as 0.188, with other strains ranging from approximately 0.25 to 0.40. Consistently, the MIC of each compound decreased substantially when used in combination. Berberine’s MIC decreased from ≥1024 µg/mL when tested alone to 32–128 µg/mL in combination. Similarly, baicalin’s MIC decreased substantially from approximately 1024 µg/mL when used individually to 32–64 µg/mL in combination, while plumbagin’s MIC values improved significantly from 16–128 µg/mL when used individually to 4–16 µg/mL in combination.

Time-kill assays conducted across ten CRAB clinical isolates revealed potent, concentration-dependent bactericidal activity of the berberine, baicalin, and plumbagin combination (Fig. 4). In the untreated control (0×MIC), all isolates showed robust growth, with counts steadily increasing over 24 h. At sub-inhibitory exposure (½×MIC), the triple combination modestly suppressed growth, followed by regrowth after an initial decline.

Close

Fig. 4.

Mean time–kill curves of the berberine + baicalin + plumbagin combination against all 10 CRAB isolates are shown. For each isolate, CFU/mL values at each time point were averaged across three independent runs prior to pooling. Each time point represents the mean CFU/mL of these isolates, and the error bar represents the standard deviation between isolates (n = 10).

Export to PPT

At 1×MIC, bacterial counts fell rapidly within 2–4 h, but regrowth occurred in most isolates after 8 h, indicating limited sustained killing at this exposure. At 2×MIC, killing was markedly enhanced: counts declined sharply, and regrowth was delayed or absent in most isolates, except isolates 4 and 7.

Notably, 4×MIC and 8×MIC achieved complete eradication in 9 isolates: counts dropped below the detection limit within 4 h and remained suppressed for 24 h. No regrowth was observed under these conditions. These findings highlight the synergistic potential of the three-compound combination at higher exposures, enabling rapid and sustained killing against multidrug-resistant isolates.

3.2 Biofilm inhibition

These compounds also significantly affected biofilm formation by CRAB. Crystal violet–based quantification showed that all three agents inhibited biofilm development, although the magnitude of inhibition differed among compounds. At comparatively low concentrations, plumbagin showed the strongest anti-biofilm effect: it significantly reduced biomass at 4–16 µg/mL and nearly abolished biofilm formation at 32–64 µg/mL. Baicalin markedly reduced biofilm biomass at 128–512 µg/mL, whereas berberine showed significant inhibition at 256–512 µg/mL. Across the tested range, plumbagin consistently reduced biofilm biomass at 4–16 µg/mL. Berberine and baicalin also inhibited biofilm formation, but only at higher absolute concentrations. Notably, plumbagin’s apparent efficacy at lower concentrations likely reflects its lower MIC and should not be interpreted as intrinsically greater anti-biofilm potency than the other compounds (Fig. 5).

Close

Fig. 5.

The inhibitory curves of berberine, baicalin and plumbagin on CRAB biofilm at different concentrations.

Export to PPT

SEM revealed significant structural differences in A. baumannii biofilms across treatment groups. In the control group, cells were densely aggregated and embedded in a thick EPS matrix, consistent with a mature biofilm (Fig. 6). A total of 71 adherent cells were observed, covering 44.05 μm², with a mean projected area of 0.620 μm² per cell, indicating the highest degree of surface colonization. In contrast, the plumbagin-treated group (8 µg/mL) showed a marked reduction in biofilm density, with only 11 loosely arranged cells detected, covering 6.84 μm². EPS was significantly diminished, and cell surfaces appeared rough and irregular, suggesting structural damage; both cell count and surface coverage were reduced by approximately 85% compared to the control. Baicalin treatment (512 µg/mL) produced a moderate reduction in biofilm formation, with 42 cells observed and a total coverage of 20.25 μm². Thinner biofilm layers, wider intercellular gaps, and reduced EPS were also evident. Berberine treatment (512 µg/mL) showed the strongest inhibition, with only 8 adherent cells and a total coverage of 3.34 μm². EPS was nearly absent, and the remaining cells appeared shrunken or ruptured, consistent with suggesting pronounced cell-envelope deformation. Overall, these SEM findings indicate that all three compounds inhibited biofilm formation, with differences in apparent potency and likely mechanisms (Table 8 and 9). Consistent with SEM quantification, adherent cell counts decreased by 40.8–88.7% and total projected biofilm area by 54.0–92.4% versus the untreated control (Table 9), providing percentage-based support for biofilm suppression.

Close

Fig. 6.

SEM images of biofilm formation after treatment with individual TCM compounds. Images were acquired at 5,000× magnification; scale bar = 10 µm (Hitachi SU8100, 3.0 kV) (white arrow).

Export to PPT

3.3 Molecular docking and mechanistic insights

To investigate how these compounds exert their effects on a molecular level, we analyzed their interactions with OmpA, a key biofilm-associated protein in A. baumannii (Figs. 7-12; Table 8). Docking suggested that all three compounds can bind OmpA, but with different affinities and preferred binding sites. Baicalin showed the strongest predicted binding, particularly to the N-terminal β-barrel of OmpA . It formed multiple hydrogen bonds and hydrophobic contacts within the barrel pore region, yielding a favorable docking score (low predicted binding energy). Berberine also adopted a favorable pose within the β-barrel, although less robust than baicalin’s. Plumbagin, in contrast, had a somewhat lower binding affinity in the simulations and tended to bind more superficially or to the edge of the barrel or the OmpA C-terminal interface. Consistent with docking, MD simulations indicated that the baicalin–OmpA complex was the most stable (Fig. 13). Hydrogen-bond occupancy and intermolecular distance analyses further supported tighter baicalin binding than the berberine or plumbagin complexes (Fig. 14). MM/PBSA analysis ranked baicalin as the most favorable binder, followed by berberine, with plumbagin least favorable (Fig. 15). RT-qPCR following sub-inhibitory exposure revealed compound-specific expression patterns (Fig. 16).

Close

Fig. 7.

Molecular docking of berberine and OmpA C-terminus. (a) Protein skeleton diagram of berberine docking with OmpA C-terminus. (b) Space filling surface. (c) 2D force analysis. (d) 3D enlarged view of binding site.

Export to PPT

Close

Fig. 8.

Molecular docking of plumbagin with the C-terminus of OmpA. (a) Protein skeleton diagram of plumbagin docking with the C-terminus of OmpA. (b) Space-filling surface. (c) 2D force analysis. (d) 3D enlarged view of the binding site.

Export to PPT

Close

Fig. 9.

Molecular docking of baicalin with the C-terminus of OmpA. (a) Protein skeleton diagram of baicalin docking with the C-terminus of OmpA. (b) Space-filling surface. (c) 2D force analysis. (d) 3D enlarged view of the binding site.

Export to PPT

Close

Fig. 10.

Molecular docking of baicalin and OmpA N-terminus. (a) Protein skeleton diagram of baicalin docking with OmpA N-terminus. (b) Space filling surface. (c) 2D force analysis. (d) 3D enlarged view of binding site.

Export to PPT

Close

Fig. 11.

Molecular docking of berberine with the N-terminus of OmpA. (a) Protein skeleton diagram of berberine docking with the N-terminus of OmpA. (b) Space-filling surface. (c) 2D force analysis. (d) 3D enlarged view of the binding site.

Export to PPT

Close

Fig. 12.

Molecular docking of plumbagin with the N-terminus of OmpA. (a) Protein skeleton diagram of plumbagin docking with the N-terminus of OmpA. (b) Space-filling surface. (c) 2D force analysis. (d) 3D enlarged view of the binding site.

Export to PPT

Close

Fig. 13.

(a) RMSD of berberine. (b) RMSD of baicalin. (c) RMSD of plumbagin. (d) RMSF of berberine. (e) RMSF of baicalin. (f) RMSF of plumbagin.

Export to PPT

Close

Fig. 14.

(a) Hydrogen bonds of berberine. (b) Hydrogen bonds of baicalin. (c) Hydrogen bonds of plumbagin. (d) MD Distance of berberine. (e) MD Distance of baicalin. (f) MD Distance of plumbagin.

Export to PPT

Close

Fig. 15.

Binding energy analysis of baicalin, berberine, and plumbagin via MM-PBSA method.

Export to PPT

Close

Fig. 16.

Modulation of ompA mRNA levels in CRAB after exposure to single compounds or their triple combination CRAB cultures were treated for 6 h with berberine (512 µg/mL), baicalin (512 µg/mL) or plumbagin (16 µg/mL), or with a triple combination containing berberine (64 µg/mL) + baicalin (64 µg/mL) + plumbagin (8 µg/mL). Untreated cells served as the control (expression = 1.0). Total RNA was isolated, reverse-transcribed, and ompA transcript abundance was quantified by RT-qPCR using 16S rRNA as the reference gene. Results are expressed as 2^⁻ΔΔCt (mean ± SEM, n = 3). Berberine alone markedly upregulated ompA (5.5-fold), whereas baicalin and plumbagin significantly downregulated expression (0.75-fold and 0.40-fold, respectively). Importantly, the triple combination suppressed ompA to an intermediate but still reduced level (0.5-fold), indicating that the repressive effects of baicalin and plumbagin override the berberine-induced up-regulation. Statistical significance: ***P < 0.001 versus the untreated control.

Export to PPT

Baicalin-treated cells showed significant downregulation of ompA mRNA (0.75-fold vs. untreated control, p < 0.01). Plumbagin treatment similarly downregulated ompA (0.4-fold, p < 0.01). In contrast, berberine exposure strongly upregulated ompA expression (5.5-fold, p < 0.01). When all three compounds were applied together, ompA expression decreased to an intermediate level (0.5-fold vs. control), suggesting that baicalin- and plumbagin-associated downregulation outweighed berberine-associated induction.

These expression changes suggest that baicalin and plumbagin may perturb regulatory pathways in A. baumannii, leading to ompA downregulation—potentially as a stress-adaptive response or signaling disruption—whereas berberine-induced ompA upregulation may reflect an alternative stress program aimed at reinforcing biofilm and envelope-associated defenses. Nevertheless, although ompA transcription increased with berberine, overall biofilm formation was still reduced, suggesting that berberine’s antibacterial activity (and possibly other target interactions) outweighed this defensive response.

3.4 In silico toxicity prediction

Finally, computational toxicity assessments provided a preliminary safety profile for the compounds. Baicalin was predicted to have an oral LD50 of 5000 mg/kg (mouse), placing it in toxicity class V, suggesting low acute toxicity in this in silico prediction. Berberine’s predicted LD50 was 200 mg/kg, corresponding to moderate toxicity (class 3). Plumbagin was predicted to be highly toxic, with an LD50 of 16 mg/kg (class 2, high toxicity). The model further flagged potential organ-specific toxicity alerts for plumbagin and berberine (berberine: neurological and respiratory; plumbagin: renal and respiratory), whereas baicalin appeared relatively safe, with only minor renal or respiratory concerns. These toxicity differences are important when considering future therapeutic use of the compounds or their combinations (Fig. 17).

Close

Fig. 17.

(a-c) Toxicity and radar profiles of baicalin, berberine and plumbagin.

Export to PPT