E. piscicida PPD130/91 and Edwardsiella ictaluri (E. ictaluri) were kindly provided by Dr. Haixia Xie (Institute of Hydrobiology, Chinese Academy of Sciences, Wuhan, Hubei, China). E. piscicida LY-2019 and ZX-1 were isolated from naturally infected fish. Mutant strains ΔarnT and Δugd were generated from the parental strain PPD130/91 following the method described previously (25). All aforementioned bacterial strains were preserved in tryptic soy broth (TSB) supplemented with 20% glycerol at −80 °C in our laboratory. In vivo infection assays were performed using the PPD130/91 strain of E. piscicida. Detailed information on the bacterial strains and reagents used in this study is listed in Supplementary Text S1 and Supplementary Table S2, respectively.

The antibiotic colistin sulfate (colistin) was obtained from MedChem Express (Shanghai, China). All solvents and diluents used in this study were prepared in accordance with the Clinical and Laboratory Standards Institute (CLSI) 2022 guidelines. E. piscicida isolates were grown for 12 h in liquid TSB at 28 °C. The MIC of each tested antibiotic was determined using the broth microdilution method as previously described (29). Briefly, overnight bacterial cultures were adjusted to a final concentration of approximately 1 × 106 colony-forming units per milliliter (CFU/ml) and inoculated into 96-well microplates containing twofold serial dilutions of colistin. The plates were then incubated for 18–20 h at 28 °C, and the optical density at 540 nm (OD₅₄₀) was measured using a microplate spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). The MIC value was defined as the minimal concentration that completely inhibited visible bacterial growth. All the experiments were performed at least in triplicate.

The synergistic antibacterial activity of the combination of colistin and PFK-158 was evaluated using the checkerboard assay. Different concentrations of colistin and PFK-158 were prepared with a two-fold serial dilution method (8 × 12 matrix). Bacterial strains were routinely propagated in tryptic soy broth (TSB) medium prior to experimental use. For the assay, bacterial culture was inoculated into a sterile 96-well plate at a final density of 1 × 10⁶ CFU/ml and incubated for 18–20 h at 28 °C in three different types of media: CAMHB, host-mimicking Dulbecco's Modified Eagle Medium (DMEM) and LPM [5 mM KCl, 7.5 mM (NH₄)₂SO₄, 0.5 mM K₂SO₄, 80 mM MES (pH 5.8), 0.1% casamino acids, 0.3% (v/v) glycerol, 24 μM MgCl₂, and 337 μM PO43−]. Optical density at 540 nm was recorded for each culture as previously described (30). The antibacterial effects of colistin and PFK-158 were quantified as follows using the fractional inhibitory concentration index (FICi), as described earlier (30):

In this study, MIC_a represents the MIC of colistin alone; MIC_ab represents the MIC of colistin in combination with PFK-158; MIC_b represents the MIC of PFK-158 alone; MIC_ba represents the MIC of PFK-158 in combination with colistin. A FICi value of ≤ 0.5 indicates a synergistic effect, whereas 0.5 < FICi ≤ 1 denotes a partial synergistic effect, and a FICi value > 1 is considered as antagonistic interaction. Each experiment was performed at least in triplicate.

The time-kill assay was conducted to evaluate the synergistic antibacterial activity of colistin in combination with PFK-158. Briefly, an overnight bacterial culture was inoculated into fresh TSB (0.1 %, v/v) and incubated for 24 h at 28 °C. The resulting bacterial suspension was treated with colistin (final concentration of 8 or 16 μg/ml), or PFK-158 (4 μg/ml) alone, or with the combined colistin/PFK-158 treatment for 24 h at 28 °C. During incubation, 100 μl aliquots were collected from each tube at 0, 1, 3, 5, 7, 9, 12, and 24 h. Ten microliter aliquots of each serially diluted sample were plated onto TSB agar plates for colony counting, and the viable bacterial count (CFU/ml) was determined. Synergism between colistin and PFK-158 was considered as a ≥ 2 log₁₀ reduction in CFU/ml compared with the most effective mono-treatment. Each experiment was performed at least in triplicate.

Outer membrane permeability was assayed using the 1-N-phenylnaphthylamine (NPN) dye following the manufacturer's instructions and previously described protocols (25, 30). Briefly, overnight bacterial cultures were washed and resuspended in 5 mM HEPES (pH 7.2) containing 10 μM NPN, adjusted to an OD₅₄₀ of 0.5 and then treated with PFK-158 (0–8 μg/ml) or colistin (16 μg/ml) at 28 °C. After 4 h of incubation in the dark, the fluorescence intensity of NPN was measured (excitation at 355 nm/emission at 420 nm) using a Synergy NEO2 multifunctional microplate reader (BioTek, Vermont, USA). Each experiment was performed in triplicate.

Inner membrane integrity was evaluated by staining with propidium iodide (PI; 20 μM) following a previously published method (25). Bacterial cells were cultured and treated as described above, and the fluorescence intensity of PI-labeled cells was measured (excitation at 488 nm/emission at 615 nm) using a microplate reader. Each experiment was performed at least in triplicate.

2′7′-dichlorodihydrofluorescein diacetate (DCFH-DA) probe was used to detect intracellular ROS levels in E. piscicida as previously described (29). Briefly, pre-treated E. piscicida suspensions were incubated with a final concentration of 5 μM DCFH-DA in the dark for 2 h with shaking, after which the fluorescence intensity was measured (excitation at 488 nm/emission at 525 nm) using a microplate reader. Each experiment was performed at least in triplicate.

The morphological changes of treated bacterial cells were observed using SEM and TEM according to the methods described (30, 31). Briefly, treated E. piscicida PPD130/91 cells were collected after 24 h by centrifugation (6,000 g, 5 min, 4 °C) and fixed for an additional 24 h at 4 °C in 2.5% (v/v) glutaraldehyde solution. The cells were then washed three times with phosphate-buffered saline (PBS, 0.1 M), and fixed for another 2 h in osmic acid solution. Subsequently, the fixed E. piscicida cells were dehydrated through an ethanol series (30%−100%). Finally, a portion of the treated cells was sputtered with gold for SEM observation. The remaining dehydrated cells were embedded, and ultrathin sectioned using an ultramicrotome and scanned by a TEM after staining with 2% uranyl acetate and Reynolds' lead citrate.

Accumulation assays using Hoechst 33342 (H33342) were performed to evaluate the inhibitory effect of PFK-158 on bacterial efflux pump activity following a previous study (32) with slight modifications. First, the cytotoxicity of H33342 was assessed to determine a non-toxic working concentration. E. piscicida at the mid-logarithmic growth phase was incubated with 2.5, 10, and 50 μM H33342, and samples were collected at 0, 30, 60, and 120 min for colony counting on TSB agar plates. The preliminary results indicated that 2.5 μM H33342 was an appropriate concentration for subsequent experiments. Hence, E. piscicida at the logarithmic growth phase (OD₅₄₀ ≈ 0.5) were adjusted to OD₅₄₀ ≈ 0.1, and treated with PFK-158 at a concentration ranging from 2 to 8 μg/ml. Aliquots (180 μl each) were transferred to a 96-well black micro-plate and H33342 was added to achieve a final concentration of 2.5 μM. The fluorescence intensity of H33342 was detected (excitation at 355 nm/emission at 460 nm) using a microplate reader for 30 cycles with 75 s intervals between readings.

The intracellular concentration of colistin was measured using an enzyme-linked immunosorbent assay (ELISA) kit (Abebio, Wuhan, China) following the manufacturer's instructions. Briefly, E. piscicida in mid-logarithmic growth phase was treated with colistin alone (8 or 16 μg/ml) or in combination with PFK-158 (4 μg/ml) at 28 °C for 6 h. Bacterial cells were collected by centrifugation (6,000 × g, 5 min, 4 °C) and washed three times with 0.9% (w/v) sterile normal saline (NaCl). The pellets were resuspended in normal saline and adjusted to an OD540 of 1.0 to normalize bacterial cell densities across all experimental samples. Finally, bacterial cells were disrupted by ultrasonication on ice (200 W, 5 s on/5 s off, 3 min total), and the supernatants obtained after centrifugation (10,000 × g, 5 min, 4 °C) were used for quantification of intracellular colistin. Each experiment was performed at least in triplicate.

E. piscicida PPD130/91 cells in the mid-logarithmic growth phase (1 × 10⁷ CFU/ml) cultured in CAMHB medium were treated with colistin alone or in combination with PFK-158 for 6 h, with the colistin only treatment serving as the control group. The bacterial cells were collected by centrifugation (6, 000 × g, 5 min, 4 °C) and washed three times with PBS. Finally, three independent replicates were prepared, and bacterial pellets were immediately frozen in liquid nitrogen and promptly shipped to Majorbio Biotechnology Co., Ltd. (Shanghai, China) for transcriptome sequencing and bioinformatic analysis. Genes exhibiting an absolute fold change (|FC|) ≥ 2 and a Bonferroni-corrected p value (p_adj) of < 0.05 were defined as differentially expressed genes (DEGs). Quantitative real-time polymerase chain reaction (qRT-PCR) analysis was used to validate the expression patterns of selected DEGs. 16S rRNA was used as the reference gene, and the relative expression levels were calculated using the 2^−ΔΔCT method (33). Additional details regarding the transcriptomic and qRT-PCR analysis are provided in Supplementary Text S3 and Supplementary Table S3.

Healthy adult zebrafish (Danio rerio) with a standard length of 2.5–3.0 cm and body weight of 0.20–0.25 g were randomly allocated into five groups (26–27 fish per group). For infection, zebrafish in all treatment groups were intraperitoneally (i.p.) injected with 10 μl of E. piscicida PPD130/91 suspension at a dose of 2 × 105 CFU per fish, according to previous studies (25, 32). Fish in the negative control group (1) received an i.p. injection of 10 μl of normal saline only, without bacterial challenge. One hour after bacterial infection, fish were intraperitoneally administered with (2) normal saline (0.9%, NaCl; vehicle group), (3) colistin (8 mg/kg body weight), (4) PFK-158 (10 mg/kg body weight), or (5) colistin combined with PFK-158 (8 mg/kg colistin + 10 mg/kg PFK-158). After 24 h post infection, a subset of zebrafish (n = 6 per group) were euthanized with a rapid cooling method (2–4 °C, ice–water bath) as previously described (32). Tissue samples of the liver, spleen, kidney, intestine, and gills were immediately dissected from euthanized fish for subsequent bacterial load quantification. The survival status of the remaining zebrafish in each group was monitored daily for 7 consecutive days post infection. Zebrafish were resumed on feeding at 48 h post-infection to maintain normal physiological conditions during the survival assay.

Statistical analysis was performed using GraphPad Prism 9.0 and SPSS software. All results are presented as mean ± standard deviation (SD). For data following a normal distribution (verified using the Shapiro–Wilk test), one-way or two-way analysis of variance (ANOVA) was conducted. For non-normally distributed data, the Mann–Whitney U test was applied. Differences in zebrafish survival were evaluated using the log-rank (Mantel-Cox) test. Statistical significance was defined as p < 0.05, with significance levels represented as ^*p < 0.05, ^**p < 0.01, and ^***p < 0.001.

Primary screening of six candidate adjuvants showed MICs of each E. piscicida strain against these compounds ranged from 2 to 256 μg/ml. All isolates exhibited high sensitivity to carbonyl cyanide m-chlorophenylhydrazone (CCCP) but displayed markedly greater resistance to PFK-158 and the remaining candidate compounds (Table 1 and Supplementary Table S1). PFK-158 alone exhibited no detectable antimicrobial activity against any of the tested isolates (MIC values were all ≥ 256 μg/ml; Table 1 and Supplementary Table S1). Checkerboard assays indicated that the colistin/PFK-158 combination reduced the MIC_colistin of E. piscicida PPD130/91 from 64 to 8 μg/ml (an eight-fold reduction; Table 2). Likewise, this combination also decreased the MIC_colistin of strains LY-2019 and ZX-1 from 32 to 8 μg/ml and from 64 to 4 μg/ml, respectively (Table 2). The corresponding FICi values for strains PPD130/91, LY-2019, and ZX-1 were 0.133 (synergistic), 0.156 (synergistic), and 0.094 (synergistic), respectively (Figure 1; Table 2). The colistin/CCCP combination also exhibited significant synergistic activity against E. piscicida (Table 2). In contrast, the remaining candidate adjuvants, epigallocatechin-3-gallate (EGCG), quercetin, myricetin, and Phe-Arg-β-naphthylamide dihydrochloride (PAβN), failed to exhibit significant synergistic effects when combined with colistin (Table 1 and Supplementary Table S1). Given that CCCP functions as a protonophore with multiple off-target effects and no established therapeutic efficacy, whereas PFK-158 is an antitumor agent that has successfully completed a Phase I clinical trial with minimal observed toxicity, PFK-158 was selected as the most promising adjuvant for subsequent mechanistic and in vivo investigations.

Strains PPD130/91, LY-2019, and ZX-1 exhibited robust growth profiles when treated individually with colistin (2–16 μg/ml) or PFK-158 (2 or 4 μg/ml) in TSB medium (Figure 2). No statistically significant differences in growth were observed among these groups after 24 h (Figure 2). In contrast, the treatment with a combination of colistin (16 μg/ml) and PFK-158 (4 μg/ml) reduced the surviving populations of PPD130/91, LY-2019, and ZX-1 by 4.67, 4.21, and 2.99 log₁₀ CFU/ml, respectively, after 24 h (Figures 2D–F). The colistin/PFK-158 combination also exhibited strong synergistic effects against E. ictaluri, Edwardsiella anguillarum (E. anguillarum), Vibrio parahaemolyticus, Vibrio vulnificus, and Escherichia coli (E. coli) DH5α (MCR-1; Supplementary Figures S1A–G), except for the empty vector containing strain E. coli DH5α (pJN105; Supplementary Figure S1H). These results suggest that PFK-158 in combination with colistin effectively reverse colistin resistance phenotypes among various fish pathogens in vitro, showing broad-spectrum synergistic antibacterial activity.

Checkerboard assays were further performed using host-mimicking culture media, including DMEM and the LPM medium. The colistin/PFK-158 combination significantly reduced the MIC_colistin of E. piscicida PPD130/91 from 64 to 16 μg/ml in DMEM medium and from 32 to 2 μg/ml in LPM medium (Supplementary Figure S2). The FICi values were 0.281 and 0.094 (both < 0.5), suggesting an excellent synergistic interaction between colistin and PFK-158 under both conditions. Therefore, this combination shows promising application potential for the treatment of MDR bacterial infections in vivo.

The results show that, neither PFK-158 (4 μg/ml) nor colistin (16 μg/ml) administered alone induced any obvious morphological changes in E. piscicida PPD130/91 cells compared with the control group (Figures 3A, C, E). In contrast, treatment with the colistin-PFK-158 combination elicited varying degrees of concave deformation and surface shrinkage in the bacterial cells (Figure 3G). Notably, TEM revealed no discernible morphological changes in the E. piscicida cell membrane under any of the tested treatment conditions (Figures 3B, D, F, H). Biochemical assays, however, demonstrated that combined treatment with 4 μg/ml PFK-158 and colistin increased outer membrane (OM) permeability by 20.95, 64.56, and 37.05% in strains PPD130/91, ZX-1, and LY-2019, respectively, after 4 h of incubation, compared with colistin monotherapy (Figures 4A–C). When the concentration of PFK-158 was increased to 8 μg/ml, the combination further enhanced OM permeability after 4 h by 20.84, 47.72, and 27.70% in PPD130/91, ZX-1 and LY-2019 (Figures 4A–C), respectively, relative to colistin alone. In stark contrast, the colistin/PFK-158 combination did not significantly enhance the IM permeability (Figures 4D–F) compared with the control or single agent treatments. These data suggest that PFK-158 significantly enhance the OM-damaging activity of colistin against E. piscicida isolates, a mechanism that likely contributes to the enhanced bactericidal efficacy of the colistin-PFK-158 combination.

Bacterial efflux activity was evaluated using the H33342 accumulation assay. As shown in Figures 5A–C, significant accumulation of H33342 was observed in all three E. piscicida strains following treatment with PFK-158 alone, demonstrating that PFK-158 exerts a potent inhibitory effect on bacterial efflux pump function. Concordantly, combined treatment with the colistin-PFK-158 combination resulted in a statistically significant increase in intracellular colistin concentrations by 24.97, 29.06, and 19.16% in strains PPD130/91, LY-2019, and ZX-1, respectively, relative to colistin monotherapy (Figures 5D–F), compared with the colistin monotherapy.

In parallel, co-treatment with colistin and 4 μg/ml PFK-158 induced marked overproduction of ROS, with ROS levels increasing by 37.12, 37.65, and 35.84% in PPD130/91, LY-2019, and ZX-1 (Figures 5G–I), respectively, relative to colistin alone treatment. Increasing the PFK-158 concentration to 8 μg/ml further augmented ROS levels by 62.88, 63.74, and 61.56% (Figures 5G–I) in the corresponding strains. Notably, supplementation with 5 mM N-acetyl-L-cysteine (NAC) effectively abrogated ROS overproduction, restoring levels to those comparable with the untreated control groups (Figures 5G–I). Collectively, these findings elucidate a sequential mechanistic cascade: PFK-158 inhibits bacterial efflux pump activity, thereby enhancing intracellular colistin accumulation; this, in turn, promotes excessive ROS generation and induces oxidative damage in bacterial cells.

Considering that the combination of colistin and PFK-158 exhibited a much stronger antibacterial effect than colistin monotherapy group (serving as the control), we focused on identifying DEGs between the colistin/PFK-158 treated and colistin alone (control) groups. The DEGs were significantly enriched in pathways pertaining to bacterial secretion and virulence systems, flagellar assembly, LPS modification, biofilm formation and two-component system (TCS) pathways (Figure 6 and Supplementary Figure S3). Specifically, genes involved in the bacterial type III secretion system (T3SS; e.g., esaJ, eseG, and eseE) and type VI secretion system (T6SS; e.g., evpP, evpC, and evpJ) were remarkably downregulated in the combination treatment group (Figures 6A, B). Furthermore, the combination also significantly suppressed expression of the arnBCADTEF operon and ugd (encoding UDP-glucose dehydrogenase), which are responsible for LPS biosynthesis and modification (Figure 6D). Additionally, the expression of several biofilm-associated genes (e.g., rpoS, bcsA, ETAE-1937, ETAE-3304, and ETAE-2682) was simultaneously downregulated (Figure 6E), and biofilm formation was remarkably inhibited by the combination treatment (Supplementary Figure S4). The expression levels of 25 representative DEGs (e.g., ugd, arnT, evpP, and rpoS) were validated via qRT-PCR (Supplementary Figure S4C), and their expression profiles were consistent with those derived from RNA-sequencing (RNA-seq).

Consistently, PFK-158 significantly enhanced the binding affinity of colistin for the bacterial cell membrane in a concentration-dependent manner (Supplementary Figure S5). For instance, as the PFK-158 concentration increased from 0 to 2, 4, 8, and 16 μg/ml, the membrane-bound colistin level rose concomitantly from 19.97 to 30.35, 43.85, 57.87, and 81.84 ppb (Supplementary Figure S5), respectively. When the colistin concentration was 32 μg/ml, PFK-158 still enhanced the membrane binding of colistin by 1.17-, 1.69-, 2.00-, and 2.38-fold at PFK-158 concentrations of 2, 4, 8, and 16 μg/ml, respectively, compared to the control group (Supplementary Figure S5).

The therapeutic efficacy of the colistin/PFK-158 combination was evaluated using a zebrafish infection model. Infected zebrafish exhibited typical clinical manifestations of edwardsiellosis, including external signs (skin ulceration and hemorrhage) and internal pathologies (abdominal distension and visceral organ damage; Supplementary Figures S6B–D). In stark contrast, combined treatment with colistin (8 mg/kg) and PFK-158 (10 mg/kg) markedly ameliorated these clinical signs (Supplementary Figure S6E) and significantly improved the survival rate of infected zebrafish (Figure 7A). By comparison, monotherapy with either PFK-158 or colistin failed to confer any protective effect against E. piscicida infection. Specifically, only one zebrafish survived until 7 days post-infection in both the saline-treated control group (5.0%, 1/20) and the PFK-158 monotherapy group (5.0%, 1/20). Similarly, merely two zebrafish remained alive in the colistin monotherapy group (9.5%, 2/21). In stark contrast, the combination treatment group achieved a markedly higher survival rate of 40.0% (8/20) over the same observation period (Figure 7A). Furthermore, the combined colistin/PFK-158 treatment also significantly decreased the bacterial loads in the liver, spleen, kidney, intestine, and gill of infected zebrafish by 2.29-, 2.73-, 2.16-, 2.02-, and 1.79 –log₁₀ CFU/ml, respectively (Figure 7B) compared with the vehicle control. Taken together, these findings indicate that PFK-158 effectively restores the in vivo efficacy of colistin against intrinsically colistin-resistant E. piscicida, thereby validating the robust therapeutic potential of this combination strategy.