The A. baumannii strains used in this study were listed in Table 1 , among which the type strain ATCC 19606 and 17978 are polymyxin-susceptible, six clinical strains were isolated from Qilu Hospital of Shandong University and identified as polymyxin-resistant A. baumannii isolates, and 10213 was obtained by ATCC 19606. All A. baumannii strains were routinely grown at 37°C in tryptone soy broth (TSB, Beijing Solarbio Science & Technology Co., Ltd., China). Four different varieties of aloe plants leaves, i.e., Aloe arborescens Miller (AA), Aloe barbadensais Miller (AB), Aloe ferox Miller (AF), and Aloe vera L. var. chinensis (Haw.) Berger (AL), were purchased from Jiangsu Shuyang Meiruoxiahua Co., Ltd., China. All antibiotics and compounds used in this study were obtained from Shanghai Yuanye Bio-Technology Co., Ltd. (Shanghai, China).

Choline chloride, D-(+)-glucose, malic acid, citric acid and lactic acid were used to prepare the natural deep eutectic solvents (NADESs). As previously reported (Dai et al., 2013; Wu et al., 2018), different NADESs were employed in this study with the molar ratios as choline chloride:D-glucose (1:1), choline chloride:malic acid (2:1), choline chloride:citric acid (2:1), and lactate:D-glucose (5:1), respectively. The mixtures were slowly heated in a water bath at a temperature of 80°C with continuous stirring for a period of 30 to 60 minutes until a uniform, transparent, and clear liquid was formed.

The fresh leaves of aloe were washed with water, cut into small pieces and homogenized for crushing. The resulting mixture was then filtered through a clean muslin cloth to remove any residual plant matter, and the supernatant was collected. Subsequently, the liquid extract was then pre-frozen at -80°C for 12 h, followed by freeze-drying using a Crystal Alpha freeze dryer (Germany). The freeze-drying process consisted of a 40-minute freezing step, an 8-hour primary drying step, and a 4-hour secondary drying step, yielding a fine powder. Ultrasound-assisted extraction was performed in a 20 mL round-bottom flask with 0.25 g of aloe powder and 10 mL of NADES solvent. The mixture was subjected to ultrasonication (KQ5200B, 200 W, 40 kHz, KunShan, China) at 80°C for 30 min. After extraction was completed, the solutions were collected and centrifuged at 10000 rpm/min for 15 min. All crude extracts used in the experiments were stored at -20°C under light-protected conditions. The extracts used in subsequent experiments were all taken from the same batch of samples to minimize the impact of batch-to-batch variation on the experimental results. The supernatant was diluted 100-fold with methanol and then filtered using a 0.22 µm organic filter membrane for subsequent HPLC injection. After being concentrated by rotary evaporation, the crude extracts were dissolved in methanol and subsequently subjected to separation using Sephadex LH-20 gel column chromatography. Methanol was used for elution at a controlled flow rate of 30 s per drop under atmospheric pressure, with fractions collected at 10-mL intervals. Based on the colorimetric results of thin-layer chromatography, the fractions were combined and then concentrated by evaporation. Compounds in the fractions were further separated by semipreparative high performance liquid chromatography (HPLC, ThermoFisher) equipped with a 250 × 10 mm Luna C₁₈ column (Phenomenex) eluting with gradient eluent of methanol in water from 60% to 100% at 30°C at a flow rate of 1.8 mL/min with 254 nm UV spectrum.

HPLC analysis was performed with a Thermo Scientific Ultimate 3000 HPLC system using a column (Thermo Fisher Scientific, C₁₈, 4.6 × 250 mm, 5 μm) equipped with a guard column. The binary mobile phase consisted of water (A) and methanol (B) in a linear gradient program from 5% B to 100% B in 46 min at a flow rate of 0.8 mL/min. The methanol concentrations were changed as follows: 5 ~ 100% B (0 ~ 30 min), 100% B (30 ~ 40 min), 5% B (40 ~ 46 min). The column temperature was as 30 °C and the sample volume injected was 10 μL. Coupled MS analysis was performed on an Impact HD Q-TOF (Bruker, Germany) equipped with an ESI source. Diode array detector spectra were acquired at 254 nm over a scan range of 190 ~ 400 nm. ESI source parameters were used as follows: negative ion mode, heater temperature 200°C, nitrogen (purity ≥ 99.99%) pumped into the ion source at a rate of 4 L/min, sheath gas flow rate 50 arb, spray voltage 3.5 kV, and scanning ranges of m/z 50 ~ 1500 for both MS and MS/MS mass spectra.

Broth cultures of A. baumannii COLR 10213 in the logarithmic growth phase were collected and resuspended to 1.0×10⁸ CFUs/mL. A sterile cotton swab was soaked in A. baumannii COLR 10213 culture until fully absorbed. Then the surface of the TSB plate was coated evenly by the sterile cotton. Subsequently, 100 μL aloe liquid extracts (25 mg/mL) was added in the sterile Oxford cup and placed on the TSB agar plate. The same volume of solvent or solution containing 10 mg/L CST were used as controls. The plates were gently placed into an incubator at 37°C for 24 h until the clear inhibition zone were observed and their diameters were recorded.

Minimum inhibitory concentrations (MICs) of all compounds were determined by the broth microdilution assay according to the Clinical and Laboratory Standards Institute (CLSI) recommendations for A. baumannii (CLSI, 2021). Briefly, drugs or crude extracts were 2-fold diluted in Mueller-Hinton broth (Oxoid, UK) mixed with bacterial suspensions containing approximately 5 × 10⁵ colony forming units (CFUs)/mL in a sterilized 96-well microliter plate (NEST Biotechnology, China). After 18 ~ 24 h incubation at 37°C, the MIC values were defined as the lowest concentrations of drugs with no visible growth of bacteria. After the MIC determination, 50 μL aliquots from all the wells which showed no visible bacterial growth were seeded on TSB agar plates and incubated at 37°C for 24 h. Minimum bactericidal concentration (MBC) was recorded as the lowest concentration killing 99.9% of the bacterial population (CLSI, 2019).

A microdilution checkerboard method was used to determine the potential effects of individual compound combinations (Ju et al., 2022). The interactions were evaluated using the fractional inhibitory concentration index (FICI). The FICI was defined as (MIC_{A in combination/}MIC_{A alone}) + (MIC_{B in combination/}MIC_{B alone}). The interaction inferred from the resulting FICI values was assessed according to the following criteria: synergy, ≤ 0.5; additivity, > 0.5 to ≤ 1; indifference (no interaction), > 1 to ≤ 4; antagonism, > 4.

Bacterial cultures in the exponential phase were diluted into MHB media to a final concentration of 5 × 10⁵ CFUs/mL. The bacterial suspensions were supplemented with varying concentrations of the compounds, either alone or in combination. The mixture was incubated at 37°C for 0, 2, 4, 6, 12, and 24 h, respectively. Following a 10-fold serial dilution on MHB plates, the number of surviving bacterial colonies was counted after overnight incubation.

A. baumannii 19606 and 17978 were cultured in TSB overnight at 37°C. After adjusting the concentration of bacteria suspensions to 1 × 10⁹ CFUs/mL, 300 μL of 1 × 10⁹ CFUs/mL were added to TSB plates containing 10 mg/L PMB or CST supplemented with or without varying concentrations of EAR and 256 mg/L EDTA. The TSB plates were incubated at 37°C for 24 h and the number of colonies grown was recorded.

To determine the cytotoxicity of reagents, the human hepatic (HEP3B) and renal (HEK293) cells were maintained in Dulbecco’s Modified Eagle Medium supplemented with 10% fetal bovine serum (NEST Biotechnology, Wuxi, China) and 1% penicillin/streptomycin (Beijing Solarbio Science & Technology Co., Ltd.). A 250 µL of cell culture solution with a cell density of 5 × 10⁴ cells/mL was transferred to each well of a 48-well plate and incubated at 37°C in 5% CO₂ atmosphere for 24 ~ 48 h. Test compounds at various concentrations (256 mg/L EDTA, 0 ~ 128 mg/L EAR, and 0 ~ 16 mg/L PMB or CST) with incubation medium were supplemented and cultivated for additional 24 h in the same conditions. Finally, 25 μL of the CCK-8 Cell Proliferation and Cytotoxicity Assay Kit (Beijing Solarbio Science & Technology Co., Ltd.) was added into each well, and OD at 450 nm was measured using a multifunction microplate reader (Infinite M200 Pro, Tecan) after incubation for 1 h at 37°C. The cell suspension supplemented with CCK-8 reagent was used as the control group, while the culture medium supplemented with CCK-8 reagent served as the black group. Cell viability was calculated using the following formula: cell viability (%) = (A_sample - A_blank)/(A_control - A_blank) × 100, where A presented absorbance at 450 nm.

A mice skin wound model of polymyxin-resistant A. baumannii infection was developed as previously reported with some modification (Wang et al., 2024). Briefly, after removing the dorsal hair of mice and sterilizing with cotton swabs dipped in 75% ethanol solution, a sterile disposable biopsy punch (Rapid Core, Ted Pella, Inc., USA) was used to produce a full-thickness excision wound of 11-mm in diameter on each BALB/c mouse (Jinan Pengyue Experimental Animal. Breeding Co., Ltd, China). A sterile cotton swab was soaked in A. baumannii 1415E culture (1 × 10⁸ CFUs/mL) until fully absorbed. The bacterial culture was gently applied to the mouse skin wound using a wet swab and allowed to air dry at room temperature, with this procedure repeated for three consecutive days to establish continuous infection. PBS buffer was used as negative control. On the 5^th and 6^th d after the infection, the mice were euthanized by cervical dislocation. The wound cavity was collected and homogenized, followed by dilution with sterile saline and spreading onto agar plates. The plates were subsequently incubated at 37°C for 24 h to facilitate bacterial growth, enabling the determination of bacterial load through colony counting.

After successful establishment of the infection by A. baumannii 1415E, mice were randomly assigned to groups with each group consisting of six mice. Subsequently, the wound sites were gently swabbed daily using cotton swabs soaked in a solution containing 32 mg/mL EDTA, 4 mg/mL EAR, and 2 mg/mL PMB. The PBS solution was utilized as the blank control group, whereas 2 mg/mL PMB was employed as the comparative group. The progress of wound healing in mice was observed daily, alongside measurements of wound diameter to calculate wound area. On the 12^th d of treatment, the bacterial load in wound cavity was detected as described above.

Each measurement was performed with a minimum of three replicates. The data are expressed as Mean ± SD (standard deviation). P-values were calculated using Student’s t-test for comparisons between two groups, the significance levels were denoted as follows: *, P < 0.05, **, P < 0.01, ***, P < 0.001. NS, not significant. Unless specified otherwise, the comparisons were made against the control group. The statistical analysis was performed using GraphPad Prism 9.5 software.

The fresh leaves of four different varieties of aloe plants (AA, AB, AF, and AL) were extracted using water (Aqueous), methanol (MeOH), or NADESs (ChCl-Glu, ChCl-Ma, ChCl-Ca, and Lac-Glu). The crude extracts were assessed for their antimicrobial activity using ZOI assay, employing the high-level polymyxin-resistant A. baumannii COLR10213 as the indictive strain. As shown in Figures 1A, B , the inhibition zones generated by NADES extracts were larger than the water or alcohol extracts, among which the inhibition by Lac-Glu extracts was the most pronounced. The extracts by Lac-Glu from AA and AB exhibited a larger zone against COLR10213 (with diameters of 50.04 ± 0.7 mm and 52.2 ± 0.5 mm, respectively), compared to those from AF and AL (with diameters of 39.0 ± 0.4 mm and 38.3 ± 0.4 mm, respectively). Next, the major components in crude extracts from AA and AB by NADESs were analyzed using HPLC. As shown in Figures 1C, D and Supplementary Figure 1 , the midsection distribution of chromatographic peaks for different NADES extracts was generally consistent. Compared with other NADES extracts, Lac-Glu extract of AA (LacGlu-AA) and AB (LacGlu-AB) exhibited three prominently different peaks with relatively high contents between retention times of 24 and 28 minutes in the latter half, respectively denoted as peaks 1 to 3. A distinct differential peak was detected in the relatively polar front half with a retention time of approximately 10 minutes, denoted as peak 4. According the tested HPLC retention time ( Supplementary Figure 2 ) of specific chemical components in aloe (Liang et al., 2021), peaks 1-4 were preliminarily identified aloe emodin, rhein, emodin, and aloesin, respectively. Through high-resolution MS ( Figure 1E ) and MS/MS ( Supplementary Figure 3 ) analysis, these identifications were further confirmed. Peaks 1-4 were subsequently purified from the crude extracts following protocols described in the method, and compounds were further confirmed by comparing their retention times and ultraviolet absorption spectroscopy against standard aloe emodin, aloesin, emodin, and rhein ( Supplementary Figure 4 ).

The antibacterial and bactericidal activities of the isolated compounds were analyzed against seven polymyxin-resistant A. baumannii strains, including 10213 ( Table 1 ). Among the four compounds, aloe emodin, emodin, and rhein showed pronounced antibacterial (MIC ranged in 0.5 ~ 2 mg/L) and bactericidal (MBC 2 ~ > 1024 mg/L) activities compared to aloesin (MIC > 1024 mg/L, MBC > 1024 mg/L). Aloe emodin, emodin, and rhein demonstrated higher antibacterial activity (MIC 0.5 ~ 32 mg/L) against A. baumannii 10213, 6588E, 038E, and 245B but lower antibacterial activity (MIC 256 ~ 1024 mg/L) against A. baumannii 07AC366, 189B, and 1415E. Considering the potential synergy of active ingredients in medicinal plants (Williamson, 2001; Hu et al., 2023), a mixture named as EAR was prepared by combining equal mass ratios of aloe emodin, emodin, and rhein. In addition to maintaining the high antibacterial and bactericidal activities against A. baumannii 10213, 6588E, 038E, and 245B (MIC 1 ~ 2 mg/L, MBC 2 ~ 64 mg/L), EAR notably enhanced its effectiveness against A. baumannii 07AC366 and 189B (MIC 4 ~ 16 mg/L, MBC 2 ~ >256 mg/L). However, EAR did not exhibit any significant activity against A. baumannii 1415E ( Table 1 ). These findings highlighted the potential variability in EAR’s efficacy across different polymyxin-resistant A. baumannii strains.

Acknowledging the well-established potential of natural products to synergize with polymyxins and enhance their antibacterial effects against multidrug-resistant bacteria (Lin et al., 2021), we attempted to assess the effects of aloe emodin, emodin, rhein, EAR and crude extracts (LacGlu-AB as the representative) when combined with polymyxins. Through checkerboard assay, significant synergistic effects (FICI < 0.5) of the tested compounds were observed that resulted for enhancing the sensitivity of certain polymyxin-resistant A. baumannii strains to PMB and CST ( Figure 2A -left panel and Supplementary Table 1 ). Specifically, exposure to a 1 mg/L concentration of EAR reduced the MIC of polymyxins against A. baumannii 189B from an initial value exceeding 512 mg/L to below 0.03 mg/L ( Table 1 ; Supplementary Table S1 ). However, it was noteworthy that aside from strain 189B, the MBC of polymyxins against other A. baumannii strains remained considerably high, ranging from 4 to 512 mg/L ( Figure 2B -left panel). This outcome suggested the necessity for further optimization strategies to achieve higher antibacterial activities across diverse strains of polymyxin-resistant A. baumannii. To facilitate this objective, EDTA was chosen for incorporation due to its previously indicated capability to disrupt the Gram-negative bacterial outer membrane by chelating Mg²⁺ and Ca²⁺, consequently increasing the permeability of the outer membrane (Nabil et al., 2000).

Using A. baumannii strain 1415E as an example, a notable decrease in the effective concentration of polymyxins was observed when combined with EDTA and aloe emodin, emodin, rhein, EAR, or crude extracts, respectively ( Supplementary Table 2 ). Particularly, in the presence of 256 mg/L EDTA, the MIC of colistin combined with EAR against the strain 1415E decreased by up to 16,384-folds. Therefore, the antibacterial and bactericidal activities of polymyxins in combination with aloe emodin, emodin, rhein, EAR, or crude extracts against seven polymyxin-resistant A. baumannii strains were further analyzed in the presence of 256 mg/L EDTA. As expected, the results revealed that the MIC of combination treatments achieved a decrease of 4 to 16384 folds compared to polymyxins treatments ( Figure 2A ; Supplementary Table 3 ), and the MBC of combination treatments reduced to 0.03 ~ 64 mg/L ( Figure 2B ; Supplementary Table 3 ). Remarkably, EAR exhibited unique advantages when combined with 256 mg/L EDTA and polymyxins, leading to a significant reduction in both MIC and MBC, and demonstrating clear synergistic effects (FICI < 0.5). Therefore, subsequent research primarily focuses on combination with the EAR mixture rather than aloe emodin, emodin, rhein or crude extracts.

Through comprehensive analysis of MBC data ( Figure 2 ; Supplementary Tables 1-3 ), the concentrations of EDTA and PMB (or CST) in the combination treatment were determined to be 256 mg/L and 2 mg/L, respectively. To determine the optimal working concentration of EAR in the combination treatment, gradient concentrations of EAR at 8 mg/L, 16 mg/L, and 32 mg/L was supplemented in the combination to measure its time-kill curve against A. baumannii 1415E, respectively. As shown in Figure 3A , 256 mg/L EDTA and 2 mg/L polymyxins had no obvious impact on the growth of 1415E when used individually. Combining EDTA with polymyxins significantly inhibited the growth of 1415E, and the total killing of 1415E was observed in the EDTA-CST group after 24 h of treatment. When EAR was added in the combination, the bactericidal efficiency against 1415E was greatly enhanced, especially the supplement of 32 mg/L EAR achieved the complete sterilization within 6 h and 4 h in the PMB and CST group, respectively. Therefore, 256 mg/L EDTA, 2 mg/L PMB or CST, and 32 mg/L EAR were used in the subsequent combination treatment. As expected, the combination treatments showed rapid killing effects against all the tested polymyxin-resistant strains in a short period of time less than 6 h ( Figure 3B ). Notably, A. baumannii 10213, 189B, and 6588E were completely eradicated within 2 h, meeting the objective of rapid sterilization.

High-concentration polymyxin treatment (≥4 MIC) A. baumannii populations evolved high-level and fixed resistance (Zhao et al., 2022). Given that EDTA in combination with EAR restored polymyxins sensitivity in polymyxin-resistant A. baumannii, was it possible for this combination therapy to further increase polymyxins sensitivity in polymyxins-sensitive A. baumannii? To answer this question, determination of the frequency of polymyxin-resistant mutation were performed with both EAR and EDTA in combination with the polymyxins and reference bacterial strains A. baumannii ATCC 19606 and 17978. The frequency of resistant mutations in strains 19606 and 17978 to polymyxins was assessed under the selective pressure of 10 mg/L PMB or CST ( Figure 4 ). The polymyxins-resistant mutation frequency was determined to be 10^-6 ~ 10^-5 for strain 19606 and 10^-7 ~ 10^-6 for strain 17978. Notably, the presence of EAR led to a significant reduction in the mutation frequency of strains 19606 and 17978 to polymyxins in a concentration-dependent manner (P < 0.001). Specifically, when the concentration of EAR exceeded or equaled 50 mg/L, the polymyxins-resistant mutation frequency in strain 19606 dropped below 10^-9. Moreover, in the presence of 256 mg/L EDTA, EAR at a concentration of 25 mg/L was sufficient to achieve a polymyxins-resistant mutation frequency below 10^-9. Analogously, a concentration of EAR equal to or exceeding 25 mg/L was observed to attenuate the mutation frequency conferring polymyxins resistance in strain 17978 to below 10^-9. Furthermore, in the presence of 256 mg/L EDTA, the required EAR concentration was reduced to 5 mg/L.

Cytotoxicity assays, pivotal in assessing the clinical development potential of therapeutic strategies, were performed on kidney (HEK293) and liver (HEP3B) cell line to evaluate the safety profile of the drug combination. As shown in Figure 5A , in the presence of 256 mg/L EDTA, EAR not exceeding 32 mg/L were found to have no significant effect on the viability of HEK293 cells, while doses ranging from 4 to 32 mg/L of EAR were observed to result in a marginal reduction (ranging from 6.31% to 13.60%, P < 0.05) in the viability of HEP3B cells. Upon combining 256 mg/L EDTA and 32 mg/L EAR with polymyxins ( Figure 5B ), various concentrations of PMB (ranging from 2 to 16 mg/L) showed no significant influence on the viability of either HEK293 or HEP3B cells. In contrast, higher concentrations of CST (16 mg/L) led to a notable decrease in the viability of HEK293 cells by 9.39% and of HEP3B cells, ranging from 11.96% (at 4 mg/L CST) to 19.29% (at 16 mg/L CST), which is consistent with the previous report that CST was cytotoxic at relatively high concentrations (Ritesh and Arun, 2018). Given the comparatively lower cell toxicity observed with PMB compared to CST, the combination strategy involving PMB was chosen for subsequent mouse skin infection treatment. Considering the excellent performance of Aloe in treating skin diseases and promoting skin wound healing (Liang et al., 2021), the skin wound infection in mice was used to evaluation efficacy of the drug combination. Due to its representative nature characterized by strong resistance to polymyxins and EAR in vitro ( Table 1 ), the clinically isolated A. baumannii 1415E was employed to establish the infection. On the 5^th and 6^th d post-infection ( Figure 6A ), the bacterial burden of 1415E at the wound site of the mice stabilized at approximately 8×10⁶ CFUs/mL, indicating the successful establishment of the mouse skin wound infection, despite that the bacterial load slightly decreased compared to the initial load (2 × 10⁷ CFUs/mL). After 12 d of continuous combination therapy (32 mg/kg EDTA + 4 mg/kg EAR + 2 mg/kg PMB), the wound area of the mice gradually decreased from the initial 94.98 mm² to 7.98 ± 3.91 mm², showing a trend of gradual healing ( Figures 6B, C ). At the same time, the wound healing both in the mock treatment (PBS buffer) group and the PMB monotherapy group (2 mg PMB/kg) was remarkably slower, with wound areas of 33.36 ± 8.33 mm² and 41.61 ± 7.66 mm², respectively. Moreover, the colony counting of the wound skin tissue of the mice was performed after 12 d of treatment, and the results ( Figure 6D ) showed that the bacterial burden of the combination therapy group was 10 ~ 30 CFUs/mL, significantly (P < 0.001) lower than that of the PMB monotherapy group (4.20 × 10⁴ ~ 1.03 × 10⁵ CFUs/mL) and the mock treatment group (9.00 × 10³ ~ 5.20 × 10⁴ CFUs/mL). The results indicated that the drug combination exhibited great potential to treat the skin infections caused by polymyxin-resistant A. baumannii.