China is a major producer of waterfowl. To investigate the prevalence of bla_NDM-5 carried by bacteria in waterfowl, our research team established close contact with numerous domestic farms from 2021 to 2023. This collaboration enabled us to collect a significant number of disease samples from the Sichuan, Shaanxi, Chongqing, Xinjiang, and Anhui regions, which are provincial-level administrative units in China. A total of 431 carcasses of diseased ducks were sent by farmers to the laboratory under cold conditions for testing. We obtained intestinal fecal contents by dissecting the carcasses in a sterile laboratory to ensure that the isolated bacteria accurately represent the local situation without contamination from various environmental factors. The intestinal content samples were initially inoculated in Luria-Bertani (LB) broth (HOPEBIO, Qingdao, China, #HB0128) at 37°C for enrichment culture, followed by inoculation onto MacConkey agar (HOPEBIO, Qingdao, China, #HB6238) containing 4 mg/L of imipenem (Meilun Biotechnology Co., Ltd., Dalian). Pink colonies were subjected to further identification using 16S rRNA sequencing with universal primers. The identified E. coli isolates were preserved in LB broth supplemented with 30% glycerol and stored at −80°C for subsequent analyses. Imipenem-resistant strains were selected and subjected to PCR amplification and sequencing. Universal 16S rRNA and bla_NDM-5 identification primers were used to identify bla_NDM-5-positive E. coli (Hornsey et al., 2011; Zhang et al., 2021b).

The modified Hodge test (MHT), modified carbapenem inactivation test (mCIM), and EDTA-modified carbapenem inactivation test (eCIM) for all bla_NDM-5-positive E. coli isolates were performed according to the standards of the American Clinical and Laboratory Standards Institute (CLSI) (Humphries et al., 2021). Isolates were resuspended in 1.5 mL of LB and then treated with 2 mL of LB containing 5 mM EDTA (15 μL of 0.5 M). Meropenem tablets (Oxoid Ltd., London, England, # CT0774B) were incubated at 35°C for 4 h ± 15 min and then placed on MH agar (Hope Bio-Technology Co., Ltd., Qingdao, #HB6231) inoculated with carbapenem-sensitive E. coli ATCC 25922. After 16–20 h of incubation, the mCIM results were interpreted as follows: an inhibition zone of ≥19 mm = negative; 6–15 mm = positive; and 16–18 mm with a tip-sized colony = intermediate (positive). A ≥5 mm increase in the eCIM inhibition zone compared to the mCIM indicates a positive result for metallo-carbapenemase; an increase of <4 mm indicates a negative result for metallo-carbapenemase. Escherichia coli ATCC 25922 served as the control. All experiments were conducted in triplicate.

DNA extraction was performed using the heating extraction method as previously described (Zhang et al., 2021b), and the extracted DNA was stored at −80°C. All isolates were assessed for antibiotic sensitivity to 11 antibiotics, following the methodology recommended by CLSI. In brief, bacterial cultures were prepared and dispensed into sterile 96-well polystyrene plates at a volume of 200 μL per well. The concentrations of antibiotics in wells 1 to 11 were 512, 256, 128, 64, 32, 16, 8, 4, 2, 1, and 0.5 mg/L, respectively. The 12th well served as a blank control containing Mueller-Hinton medium without antibiotics. The 11 antibiotics tested included imipenem (IMP, #S24020), tetracycline (TE, #S17051), norfloxacin (NFX, #S17080) (Yuanye Biotechnology Co., Ltd., Shanghai), azithromycin (AZM, #MB1024), kanamycin (KAN, #MB1130), trimethoprim (SXT, #MB1317), polymyxin B (PB, #MB1188) (Meilun Biotechnology Co., Ltd., Dalian), chloramphenicol (C, #IC0320) (Solarbio Science & Technology Co., Ltd., Beijing), aztreonam (ATM, #A801653), cefotaxime (CTX, #C804340), and ampicillin (AMP, #A830931) (Macklin Biochemical Technology Co., Ltd., Shanghai). Escherichia coli ATCC25922 was utilized as the negative control strain.

Based on the MIC results, it was necessary to detect the presence of ARGs in the isolates. PCR was used to detect ARGs associated with different classes of antibiotics, including β-lactams (bla_TEM, bla_SHV, bla_CTX-M, bla_NDM, bla_KPC), quinolones (qnrA, qnrB, and qnrS), aminoglycosides [aac(6′)-Ib-cr, aadA1, and aac(3)-I], sulfonamides (sul1, sul2, and sul3), tetracyclines (tetA, tetB, tetC, and tetM), glucocorticoids (floR, cat1), macrolides (ermA, ermB), and polymyxin (mcr-1). The primers used for PCR are detailed in Supplementary Table S1. All positive PCR products were sent to Tsingke Biotechnology Co., Ltd. for DNA sequencing. The resulting sequences were analyzed using the BLAST tool available at the online gene database.¹

We used a bla_NDM-5-positive E. coli strain as the donor and NaN₃-resistant E. coli J53 as the recipient. The donor and recipient bacteria were cultured on LB agar plates containing 4 mg/L IMP, 150 mg/L NaN₃, and 4 mg/L IMP + 150 mg/L NaN₃, respectively. After incubating for 18 h in a constant temperature incubator at 37°C, we observed the results. Both the donor and recipient bacteria were combined in a new tube at a ratio of 1:4 and incubated at 37°C for 24 h, followed by a 10-fold dilution in sterile saline to achieve the appropriate concentration. A volume of 100 μL of the mixed bacterial solution was then plated on LB agar containing 4 mg/L IMP + 150 mg/L NaN₃ and incubated at 37°C for 24 h. Single colonies from the 4 mg/L IMP + 150 mg/L NaN3 LB agar were selected for bla_NDM-5 detection, and bla_NDM-5-positive strains were identified as transconjugants.

For testing conjugational frequency, both donor and recipient bacteria were activated and diluted to an OD600 of 0.5. They were then mixed in a new tube at a ratio of 1:1 and incubated at 37°C for 24 h, followed by a 10-fold dilution with sterile saline to reach the desired concentration. Subsequently, 100 μL of the mixed bacterial solution was plated on LB agar containing 4 mg/L IMP + 150 mg/L NaN₃, while 100 μL of E. coli J53 was spread on LB agar with 150 mg/L NaN₃. Both were incubated at 37°C for 24 h. Single colonies on the 4 mg/L IMP +150 mg/L NaN₃ agar were counted as transconjugants (designated as T), while single colonies on the 150 mg/L NaN₃ agar were counted as receptors (designated as R). The conjugation frequency was calculated as T/R. All experiments were conducted in triplicate.

The PCR-based replicon typing (PBRT) technique was used to detect plasmid incompatibility groups in transconjugants, as previously described with specific primers (Carattoli et al., 2005; Johnson et al., 2007).

Salmonella H9812 served as the reference strain for molecular quality standards. For the analyses, the bla_NDM-5-positive strains were embedded in 1% SeaKem Gold Agarose (Lonza Group Co., Ltd.) and subjected to enzymatic digestion using S1 nuclease (Takara Biomedical Technology Co., Ltd., Beijing) at 37°C for 4 h. Electrophoresis was subsequently performed using 1% agarose gel under specific conditions: an operating voltage of 6 V/cm and a pulse time ranging from 6.76 s to 35.38 s, utilizing the CHEF DR III system for a total duration of 18 h at 14°C.

The transconjugants were cultured in LB broth with 4 mg/L of imipenem (IMP) added continuously, with transfers to fresh medium every 12 h. Stability was measured every 24 h. Bacteria were diluted 10-fold to the appropriate concentration using sterile saline, and 100 μL of the diluted bacteria was applied to LB agar containing 4 mg/L of IMP, as well as to regular LB agar. The plates were then incubated in a constant temperature incubator at 37°C for 18 h. The number of single colonies on the LB agar with 4 mg/L of IMP indicated the number of bla_NDM-5-positive strains, while the number of single colonies on regular LB agar represented the total number of strains. The genetic stability of the bla_NDM-5-positive plasmid was calculated as the ratio of bla_NDM-5-positive strains to the total number of strains. All studies were performed in triplicate.

To amplify the complete bla_NDM-5 sequence, primers bla_NDM-5-F (5′-CGCCATATGGCGATGGAATTGCCCAAT-3′) and bla_NDM-5-R (5′-CGCGGATCCGCGTCAATGATGATGATGAGATGGCGCAGCTTGTC-3′) were employed, incorporating Nde I and BamH I restriction sites at their respective termini. The amplification reaction mixture had a total volume of 20 μL and included 10 μL of 2 × Taq PCR Green Mix, 1 μL of each primer (bla_NDM-5-F and bla_NDM-5-R), 2 μL of DNA, and 6 μL of ddH₂O. The PCR cycling conditions were as follows: initial denaturation at 95°C for 15 s, annealing at 55°C for 15 s, and extension at 65°C for 1 min, for a total of 32 cycles. The PCR products were analyzed using 1% agarose gel electrophoresis (120 V for 20 min) and purified using a PCR purification kit (Tiangen Biotech Co., Ltd., Beijing). A digestion reaction was performed with 5 μL of the purified PCR product (bla_NDM-5) and the pET32a(+) vector containing 6 × His. The ligation was allowed to proceed overnight at 4°C. After sequencing the recombinant plasmid and confirming its accuracy through BLAST alignment, the plasmid was transformed into E. coli-BL21(DE3) and E. coli DH5α.

Escherichia coli-BL21(DE3)-pET32a(+)-bla_NDM-5 was cultured with an optical density (OD600) of 0.6, and 0.5 M isopropyl-β-D-thiogalactoside (IPTG) (Solarbio Science & Technology Co., Ltd., Beijing) was added for induction at 16°C overnight. The bacteria were then collected and centrifuged. The resulting pellet was resuspended in 0.05 M Tris–HCl solution and disrupted using ice bath ultrasound. Subsequently, purification was performed according to the Ni-NTA Beads 6FF kit (Changzhou Smart-Lifesciences Biotechnology Co., Ltd.). The purified samples were then analyzed using sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) (Li L. et al., 2024).

To evaluate the changes in MIC values of the tested strains against imipenem in the presence of NDM-5 protein, a reaction system was prepared consisting of 30 μg/mL of total protein and 11 concentrations of imipenem (512 mg/L, 256 mg/L, 128 mg/L, 64 mg/L, 32 mg/L, 16 mg/L, 8 mg/L, 4 mg/L, 2 mg/L, 1 mg/L, and 0.5 mg/L). The reaction was incubated at 37°C in triplicate alongside a negative control group that lacked protein. The changes in OD600 of the strains were measured using a microplate reader every 4 h over a total period of 24 h (at 4 h, 8 h, 12 h, 16 h, 20 h, and 24 h) to determine alterations in MIC values against imipenem.

To investigate the fitness associated with bla_NDM-5-positive plasmids, we employed a combination of growth curve analyses, biofilm formation assessments, and in vitro competition assays. These experiments utilized the conjugants E. coli DH5α-pET32a(+)-bla_NDM-5 and E. coli DH5α-pET32a(+), following previously described methods (Lu et al., 2021; Huo et al., 2024). All experiments were conducted in triplicate.

A low-level imipenem-resistant strain carrying bla_NDM-5 was selected and cultured to an OD600 of 1. Based on the MIC results for imipenem, a concentration of 1 × MIC was chosen for induction on the first day. The concentration of imipenem was doubled daily, following the sequence: 1 × MIC, 2 × MIC, 4 × MIC, 8 × MIC, and so on. Each day, 500 μL of the culture was transferred to 20 mL of LB broth supplemented with the corresponding concentration of imipenem, and the culture was incubated at 37°C for 24 h until no further growth was observed.

The expression of bla_NDM-5 in high-level resistant strains after induction was assessed using qRT-PCR. The fluorescent quantitative primers for bla_NDM-5 were selected based on a previous study (Zhao et al., 2023). The TAKARA RNAiso Easy Kit (Takara Biomedical Technology Co., Ltd., Beijing) was used to isolate mRNA from imipenem-exposed strains, and the Hifair ® II 1st Strand cDNA Synthesis SuperMix Kit (Yeasen Biotechnology Co., Ltd., Shanghai) was used for reverse transcription to synthesize cDNA. The expression levels of bla_NDM-5 were quantified using cDNA as a template, with the 16S rRNA gene serving as the internal reference. The results were expressed as the ratio of the expression levels of the bla_NDM-5 target gene compared to the internal reference gene (Zhao et al., 2023).

The MIC of imipenem was determined using the broth microdilution method, with 5 mg/L of carbonyl cyanide m-chlorophenylhydrazine (CCCP) (Solarbio Science & Technology Co., Ltd., Beijing) as an efflux pump inhibitor. This method followed the same protocol as the antibiotic sensitivity test.

To study the adaptive cost and stability of the altered strains, a combination of growth curve analysis, in vitro competition tests, and stability assessments was employed. The experiment utilized strains both before and after induction, based on previous references (Lu et al., 2021; Huo et al., 2024). All experiments were conducted in triplicate.

Statistical analyses were performed using SPSS 27.0 software. Pearson’s chi-square test and Fisher’s exact test were utilized to compare the antimicrobial resistance of 103 E. coli isolates and evaluate the correlation among different ARGs, or between AMR phenotypes and their corresponding ARGs. Odds ratios (OR) and their 95% confidence intervals were calculated, with OR < 1 indicating a negative correlation and OR > 1 indicating a positive correlation. A p value of <0.05 was considered statistically significant.

Upon transporting ducks from selected areas to the laboratory, researchers collected intestinal contents in a clean and disinfected animal facility to ensure non-repetitive samples of intestinal feces from each duck. A total of 431 intestinal fecal samples were collected from each duck. From a total of 431 intestinal feces collected from waterfowl sources in Sichuan, Anhui, Shaanxi, Xinjiang, and Chongqing, 103 strains were identified as bla_NDM-5-positive E. coli, resulting in a positive rate of 30.5%. The majority of bla_NDM-5-positive E. coli were isolated from Sichuan Province, specifically from areas such as Deyang, Meishan, Xinjin, Chongzhou, Jintang, Suining, Dayi, and Leshan. Overall, with the exception of Shaanxi Province, the farms from which the positive bacterial strain samples were collected are located between 20 and 300 km from our laboratory. Specific details are provided in Table 1.

All bla_NDM-5-positive E. coli strains tested positive in the MHT, mCIM, and eCIM tests. The results of the MHT are presented in Supplementary Figure S1, while the mCIM and eCIM results are shown in Supplementary Table S2 and Supplementary Figure S2. Specific values can be found in the Appendix.

The detection of ARGs in 103 bla_NDM-5-positive E. coli strains is presented in Figure 1. Among these, the chloramphenicol resistance gene floR had the highest detection rate (n = 91, 88.3%), followed by the β-lactam resistance gene bla_CTX-M (n = 90, 87.4%), the tetracycline resistance gene tetA (n = 87, 84.5%), the sulfonamide resistance gene sul2 (n = 84, 81.6%), the aminoglycoside resistance gene aadA1 (n = 72, 69.9%), and the sulfonamide resistance gene sul1 (n = 58, 56.3%). The detection rates of the other ARGs were less than 50%.

The analysis results indicate that there are significant correlations, primarily positive, among the ARGs carried by the 103 bla_NDM-5-positive E. coli strains. As shown in Supplementary Table S3, a total of 45 pairs of ARGs exhibited significant positive correlations, while 23 pairs showed significant negative correlations. Among these, tetB has the strongest positive correlation with sul3 or ermB (OR, 101.000; 95% CI, 3.370–3027.447). In terms of negative correlations, sul2 and tetC displayed the strongest negative correlation (OR, 0.030; 95% CI, 0.006–0.156). Detailed analyses of the other correlation results are presented in Supplementary Table S3.

Except for strain DY51, all strains exhibited high-level resistance to imipenem, with MIC values ranging from 64 to ≥512 mg/L (Supplementary Table S4). The resistance rates for the remaining 10 antibiotics were as follows: AMP (n = 103, 100%), CTX (n = 102, 99.0%), and NFX (n = 100, 97.1%) showed the highest resistance rates, followed by SXT (n = 92, 89.3%), KAN (n = 92, 89.3%), and C (n = 92, 89.3%). Finally, the resistance rates for AZM (n = 28, 27.2%) and PB (n = 16, 15.5%) were relatively low.

These results indicate that all 103 bla_NDM-5-positive E. coli strains are MDR, with resistance levels ranging from 3 to 8 antibiotics, primarily six antibiotics (n = 57, 55.3%). The specific antimicrobial resistance profiles are presented in Table 2.

The correlation analysis between resistance genes and AMR phenotypes revealed significant correlations between the resistance genes and AMR in the 103 bla_NDM-5-positive E. coli strains from waterfowl, primarily negative correlations. As shown in Supplementary Table S5, a total of seven pairs of genes exhibited significant positive correlations, while nine pairs showed significant negative correlations. Among these, KAN and aadA1 exhibited the strongest positive correlation (OR, 13.696; 95% CI, 2.757–68.036), while NFX and emrB displayed the strongest negative correlation (OR, 0.020; 95% CI, 0.001–0.446). Other analyses are detailed in Supplementary Table S5.

Fifty-three bla_NDM-5-positive E. coli strains were able to transfer imipenem resistance to the recipient strain J53, with the MIC of the recipient strain increasing from 0.5 mg/L to either 4 mg/L or 512 mg/L. PCR and DNA sequencing confirmed that all 53 transconjugants were bla_NDM-5-positive. By counting the number of colonies of the 53 bla_NDM-5-positive transconjugants on double-resistant plates and the colonies of E. coli J53 on NaN3-resistant plates, the range of conjugative transfer frequency was calculated to be 5.22 × 10⁻⁷ to 9.48 × 10⁻⁴ (Figure 2). The plasmid profiles of randomly selected donor and recipient bacteria were analyzed using S1-PFGE (Supplementary Figure S3). The donor bacteria carried 3–4 plasmids, while the recipient bacteria carried 1 plasmid, indicating that bla_NDM-5 was located on a mobile plasmid. Furthermore, replicon typing of all transconjugants was performed via PCR and sequencing; the results indicated that the 53 bla_NDM-5-positive transconjugants harbored 13 different replicon types, namely, IncX1, IncX3, HI2, I1, N, FIA, FIB, Y, P, FIC, FIIS, K, and B/O. Notably, IncX1 (28.3%) was the dominant type, followed by HI2 (26.4%) and K (15.1%). The plasmid profiles of randomly selected donor and recipient bacteria were shown by S1-PFGE (Supplementary Figure S3). The donor bacteria carried 3–4 plasmids, while most recipient bacteria carried one plasmid, indicating that bla_NDM-5 was located on a mobile plasmid.

To investigate whether the bla_NDM-5-positive plasmids acquired by the transconjugants possess stable genetic characteristics, this study conducted a plasmid stability test on 53 strains carrying bla_NDM-5-conjugative plasmids. During the experiment, the genetic stability of these plasmids was compared under conditions of both imipenem selection pressure and in its absence, as illustrated in Figure 3. In the absence of antibiotic selection pressure, the plasmid stability of the 53 transconjugants showed a slight decline with increasing passage numbers, while the plasmid stability rate remained above 75%. Under a selection pressure of 4 mg/L imipenem, the plasmid stability rate of the 53 transconjugants was even higher than that observed without antibiotics, maintaining above 92%, indicating stronger stability. This suggests that the stable presence of bla_NDM-5 conjugative positive plasmids is influenced not only by the stable genetic characteristics of the gene itself but also by the presence of antibiotics in the environment. These findings imply that residual antibiotics in the environment can contribute to the stable presence of bla_NDM-5-positive plasmids in resistant strains, thereby enhancing the dissemination of antibiotic resistance.

Full-length amplification of bla_NDM-5 was performed on the transformant E. coli DH5α carrying the pET32a(+)-bla_NDM-5 plasmid, and the electrophoresis gel results are presented in Supplementary Figure S4. The band size is approximately 813 bp, which is consistent with the expected size. The amplification product was sequenced, and the sequencing results were subjected to BLAST alignment, confirming the successful construction of the recombinant plasmid.

Total protein was extracted from the modified E. coli-BL21(DE3) strain harboring the pET32a(+)-bla_NDM-5 plasmid. The bla_NDM-5 protein was subsequently purified using nickel column affinity chromatography. After SDS-PAGE analysis, a clear protein band was observed at the 28.5 kDa position, as shown in Supplementary Figure S5, aligning with the expected size of the NDM-5 protein.

Initially, the concentration of the purified NDM-5 protein was determined. Based on this concentration, the activity of the purified NDM-5 protein was assayed. Table 3 and Figure 4 described the changes in the MIC values of imipenem against the low-level resistant strain DY51 in the presence of NDM-5 protein. Due to the hydrolytic activity of the NDM-5 protein on imipenem, the resistance of the low-level resistant strain DY51 and the non-bla_NDM-5-carrying strain E. coli-BL21(DE3) to imipenem gradually increased, resulting in rising MIC values over time. This clearly demonstrates that the purified NDM-5 protein exhibits significant activity, capable of hydrolyzing imipenem and enabling the strains to survive in the presence of high antibiotic concentrations.

The OD₆₀₀ values of 53 transconjugants, the transformant E. coli DH5α-pET32a(+)-bla_NDM-5, E. coli DH5α-pET32a(+), and the recipient bacteria E. coli J53 were measured at various time points to plot the changes in growth curves before and after the transfer of bla_NDM-5-positive plasmids. The results are presented in Figure 5. Among the 53 transconjugants, 32 of them (DY3-2-J, DY4-7-J, dy8-3-J, SN18-2-J, dy21-5-J, dy21-6-J, XJ22-4-J, XJ22-6-J, XJ23-2-J, XJ23-6-J, MS25-1-J, SX28-1-J, SX28-2-J, SX28-6-J, XJ31-4-J, XJ31-6-J, SX34-1-J, CZ37-2-J, CZ37-3-J, CZ37-6-J, CZ37-8-J, CZ37-11-J, CZ37-13-J, CZ37-14-J, CZ37-15-J, SX34-6-J, MS38-4-J, SN42-1-J, SN42-7-J, DY43-2-J, LS44-4-J, and LS44-5-J) exhibited slower growth rates after acquiring the bla_NDM-5-positive plasmid, while the growth rates of the remaining 21 transconjugants (DY3-3-J, MS20-1-J, MS20-2-J, MS20-4-J, XJ22-3-J, SX34-5-J, DY35-17-J, DY35-18-J, CZ37-1-J, CZ37-9-J, SN42-10-J, SN42-3-J, SN42-4-J, SN42-5-J, SN42-6-J, DY43-1-J, DY43-6-J, LS44-3-J, LS44-6-J, DY16-10-J, and DY51-J) were comparable to that of E. coli J53. In the transformant group, the growth rate of E. coli DH5α-pET32a(+)-bla_NDM-5, which acquired the bla_NDM-5 gene, was slower compared to E. coli DH5α-pET32a(+).

The biofilm formation ability of the 53 bla_NDM-5 transconjugants and E. coli J53 was assessed by measuring the OD value at a wavelength of 570 nm. A higher OD₅₇₀ value indicates stronger biofilm formation ability. The results are displayed in Figure 6. Among the 53 bla_NDM-5 transconjugants, 40 transconjugants (XJ22-4-J, SX34-5-J, DY35-17-J, CZ37-11-J, CZ37-13-J, CZ37-14-J, CZ37-15-J, DY3-2-J, DY3-3-J, SN18-2-J, MS20-2-J, MS20-4-J, SX28-6-J, MS38-4-J, dy21-5-J, XJ23-2-J, SX34-1-J, DY35-18-J, CZ37-2-J, CZ37-3-J, CZ37-6-J, dy21-6-J, XJ22-3-J, MS25-1-J, SX28-1-J, XJ31-6-J, SN42-1-J, SN42-10-J, DY51-J, SN42-3-J, SN42-4-J, SN42-5-J, SN42-6-J, SN42-7-J, DY43-1-J, DY43-2-J, DY43-6-J, LS44-3-J, LS44-5-J, and LS44-6-J) demonstrated significantly enhanced biofilm formation ability after acquiring the bla_NDM-5 conjugative plasmid. In contrast, 13 transconjugants (DY4-7-J, SX34-6-J, CZ37-9-J, DY16-10-J, MS20-1-J, CZ37-1-J, DY8-3-J, XJ22-6-J, CZ37-8-J, XJ23-6-J, SX28-2-J, XJ31-4-J, and LS44-4-J) showed no significant difference in biofilm formation ability compared to the recipient bacteria E. coli J53 after acquiring the bla_NDM-5 plasmid.

To investigate the effect of bla_NDM-5 on the fitness of transformants, an in vitro competition experiment was conducted between E. coli DH5α-pET32a(+)-bla_NDM-5 and E. coli DH5α-pET32a(+). As shown in Supplementary Figure S6, the relative fitness (RF) values were less than 1 at 24, 48, and 72 h, indicating that the acquisition of bla_NDM-5 imposed a certain fitness cost on the transformants, although this cost was relatively small.

Using MIC testing, this study identified a strain DY51 with low-level resistance to imipenem (MIC = 4 mg/L). Furthermore, the recombinant strain E. coli DH5α-pET32a(+)-bla_NDM-5 constructed in this study also exhibited a MIC value of 4 mg/L for imipenem. Based on these results, the low-level resistant strains DY51 and E. coli DH5α-pET32a(+)-bla_NDM-5 were selected for an in vitro antibiotic induction experiment with imipenem. After exposure to varying doses of imipenem, two high-level imipenem-resistant strains, namely, DY51-I and E. coli DH5α-pET32a(+)-bla_NDM-5-I, were obtained. The MIC value of strain DY51-I for imipenem increased to 512 mg/L, representing a 128-fold increase. Similarly, the MIC value of E. coli DH5α-pET32a(+)-bla_NDM-5-I for imipenem also rose to 512 mg/L, also a 128-fold increase.

Since there is a significant relationship between the expression level of the bla_NDM-5 gene and bacterial resistance, qRT-PCR was utilized to determine the expression levels of the bla_NDM-5 gene in E. coli DH5α-pET32a(+)-bla_NDM-5 and DY51 strains before and after induction. As illustrated in Figures 7A,B, the expression level of the bla_NDM-5 gene in DY51 increased following exposure, with the expression level in the exposed resistant strain DY51-I rising approximately 8-fold compared to the pre-exposed strain, showing a significant difference (p < 0.05). Similarly, the bla_NDM-5 expression level in the exposed resistant strain E. coli DH5α-pET32a(+)-bla_NDM-5-I increased approximately 4-fold in comparison with the pre-changed strain, also showing a statistically significant difference (p < 0.05). This suggests that the high-level resistance to imipenem in the exposed resistant strains E. coli DH5α-pET32a(+)-bla_NDM-5-I and DY51-I may be attributed to a substantial increase in the expression level of the bla_NDM-5 gene.

Overexpression of efflux pumps can decrease the sensitivity of E. coli to imipenem. Therefore, the MICs of the highly imipenem-resistant changed strains DY51-I and E. coli DH5α-pET32a(+)-bla_NDM-5-I were determined following the addition of the efflux pump inhibitor CCCP. As indicated in Table 4, the MIC values of the highly imipenem-resistant changed strains decreased in the presence of CCCP, suggesting that the high resistance to imipenem in these strains is related to the overexpression of efflux pumps.

To investigate the stability of high-level imipenem resistance in induced strains, resistance stability tests were conducted on the induced resistant strains E. coli DH5α-pET32a(+)-bla_NDM-5-I and DY51-I. As shown in Figure 7C and Supplementary Table S6, the MIC values of the induced strains E. coli DH5α-pET32a(+)-bla_NDM-5-I and DY51-I against imipenem remained at 512 mg/L. However, it is noteworthy that under conditions without antibiotic pressure, the induced strain E. coli DH5α-pET32a(+)-bla_NDM-5-I gradually lost the blaNDM-5-positive plasmid. In contrast, the high resistance to imipenem in the induced strain DY51-I could be stably maintained for a considerable period without antibiotic pressure.

To investigate the growth characteristics of the induced resistant strains, growth curve experiments were performed on E. coli DH5α-pET32a(+)-bla_NDM-5-I and DY51-I. The results, as shown in Figures 7D,E, indicate that both induced resistant strains exhibited faster growth rates compared to their non-changed counterparts.

Pairwise in vitro competition experiments were conducted between the high-level resistant strains E. coli DH5α-pET32a(+)-bla_NDM-5-I and DY51-I and their pre-induced counterparts E. coli DH5α-pET32a(+)-bla_NDM-5 and DY51. As shown in Supplementary Figure S10, after 24 h, the high-resistant strains E. coli DH5α-pET32a(+)-bla_NDM-5-I and DY51-I displayed relatively strong competitive advantages over the pre-induced strains. However, at 48 h, the high-resistant strains demonstrated absolute competitive superiority.