From January 2020 to December 2022, 287 non-duplicate A. baumannii isolates, including 147 CRAB and 140 carbapenem-susceptible A. baumannii (CSAB) isolates, were collected from various clinically sampled infections of inpatients at Yangjiang People’s Hospital, a tertiary hospital with 2000 beds in Guangdong Province of China. Bacterial species were identified using a VITEK 2 compact system (BioMérieux, Marcy l’Etoile, France). All strains in this study were cultured on Mueller-Hinton (MH) broth (Oxoid, Basingstoke, UK), or in MH agar plates and Luria-Bertani (LB) broth (Oxoid, Basingstoke, UK) at 37°C. All procedures performed were approved by the Ethical Committee of Yangjiang People’s Hospital and were in accordance with the 1964 Helsinki Declaration and its later amendments. The novel fluorocycline antibiotic, eravacycline (catalog no. HY-16980A), was purchased from MedChem Express (MCE, Shanghai, China). The other antimicrobials were purchased from Meilunbio (Dalian, China). Escherichia coli ATCC 25922 was used for quality control.

The minimum inhibitory concentrations (MICs) of eravacycline and several common antibiotics were determined using the broth microdilution method recommended by Clinical and Laboratory Standards Institute (CLSI) guidelines (CLSI-M100-S30). As A. baumannii MIC breakpoints for eravacycline and tigecycline have not yet been established by CLSI and FDA, this study categorized the MIC values into three levels described by prior studies (Marchaim et al., 2014; Abdallah et al., 2015): ≤2 mg/L (susceptible), 4 mg/L (intermediate), and ≥8 mg/L (resistant). The susceptibility results for other antibiotics, including meropenem, levofloxacin, gentamicin, cefepime, polymyxin B, tetracycline, doxycycline, and minocycline, were determined based on CLSI breakpoints for each antibiotic. In this study, CRAB strains were defined as those with an imipenem and/or meropenem MIC ≥16 mg/L, while CSAB strains with imipenem and meropenem MICs were both ≤4 mg/L (Fu et al., 2010).

Population analysis profiling (PAP) was performed as described previously (Abe et al., 2022). The method was used as a reference to investigate eravacycline heteroresistance among 283 clinical A. baumannii isolates with eravacycline MIC values ≤4 mg/L. In brief, 50 μL aliquots of cell suspension (corresponding to a 0.5 McFarland standard for A. baumannii cultures grown on blood agar plates for 24 h at 37°C, 1.0–1.5 × 10⁸ CFU/mL) were spread onto MH agar plates with or without eravacycline (1, 2, 4, 8, 16, and 32 mg/L). The plates were incubated for 24 h at 37°C and the number of colonies was counted. Based on the breakpoint of eravacycline MIC values for A. baumannii, eravacycline heteroresistance was defined as an eravacycline-susceptible isolate (MIC ≤ 4 mg/L) with subpopulations growing in the presence of ≥8 mg/L eravacycline at a detection threshold of 20 CFU/mL. Three colonies were selected from the 8 mg/L eravacycline concentration of the PAP test and categorized as the resistant subpopulations of each eravacycline-heteroresistant isolate (name as: strain-RS). The eravacycline-heteroresistant parental strain was considered as the susceptible subpopulation (name as: strain-HP) given that the majority subpopulations are susceptible to eravacycline (Band and Weiss, 2019). The resistant subpopulations of eravacycline MICs were reassessed after serial passaging on antibiotic-free medium to evaluate the stability of the heteroresistant phenotype.

Total DNA from eravacycline heteroresistant parental strains and their resistant subpopulations were extracted using lysis buffer for microorganisms. Polymerase chain reaction (PCR) was performed using PCR Mastermix (Thermo Fisher Scientific, Waltham, MA) according to the manufacturer’s instructions. Multilocus sequence typing (MLST) of the strains was determined by PCR and sequence alignment, as described previously (Huang et al., 2016). The carbapenemase resistance genes, bla _OXA, bla _KPC, bla _IMP, bla _VIM, bla _SIM, and bla _NDM-1 were detected by PCR (Hassan et al., 2021). Genetic mutation of the RND efflux pump regulators, adeR, adeS, adeL, and adeN, and the ribosomal protein S10 encoding gene, rpsJ, were detected by PCR amplification and sequence alignment. The primers used for PCR are listed in Supplementary Tables 1–3.

The role of the efflux pump in the eravacycline heteroresistance of CRAB isolates was detected using the efflux pump inhibitors (EPIs), Phe-Arg-β-naphthylamide (PaβN; MCE, Shanghai, China), and carbonyl cyanide m-chlorophenylhydrazone (CCCP; MCE, Shanghai, China). Eravacycline MIC was determined using the agar dilution method in the presence and absence of PAβN (50 mg/L) or CCCP (16 mg/L) in resistant subpopulations of the heteroresistant isolates, as previously described (Jo and Ko, 2021). Efflux inhibition was defined as a ≥4-fold decrease in the MICs from their original values in the presence of EPI (Tambat et al., 2022). PAP assay with PAβN (50 mg/L) or CCCP (16 mg/L) added to the MH agar plates was used to validate the potential function of the efflux pump in eravacycline heteroresistance.

The transcriptional levels of the RND-type efflux pump genes (adeB, adeG, and adeJ) and their regulators (adeS, adeL, and adeN) were quantified in eravacycline heteroresistant parental strains and resistant subpopulations using qRT-PCR as described previously (Yoon et al., 2013), with the primers listed in Supplementary Table 4. Since the three genes, adeA, adeB, and adeC, of the efflux pump, adeABC, are located within the same operon, adeB was considered representative of adeABC efflux pump expression in this study (Ju et al., 2021). The other genes shared the same principle. In brief, overnight cultures of the bacterial strains were diluted 1:100 in 10 mL of LB broth and incubated at 37°C with 220 rpm shaking until the growth reached a logarithmic phase. Total RNA was extracted using a Bacteria RNA Extraction Kit (Vazyme, Nanjing, China), and cDNA was synthesized with a HiScriptIII-RT SuperMix for qRT-PCR kit (Vazyme, Nanjing, China) according to the manufacturer’s instructions. The qRT-PCR was performed using a ChamQ Universal SYBR qRT-PCR Master Mix (Vazyme, Nanjing, China) in a LightCycler480II system (Roche, Basel, Switzerland), with the following parameters: 1 cycle at 95°C for 30 s, followed by 45 cycles of 95°C for 10 s and 60°C for 30 s. The internal control gene, rpoB, was used to normalize the expression of each candidate gene. Each sample was run in triplicate. Threshold cycle (Ct) numbers were determined using the qRT-PCR system software, and the relative transcription levels of all strains were calculated using the 2^−ΔΔCt method. CRAB-9-HP was used as the reference strain (expression = 1.0). The expression levels of the resistant subpopulation target genes were compared with their parental strains.

To confirm the role of AdeABC and AdeRS in the eravacycline heteroresistance of CRAB isolates, the efflux pump genes were silenced by antisense RNA (asRNA) in strains with a high expression of adeABC or adeRS in resistant subpopulations and their parental strains. For silencing adeABC, an adeABC asRNA expression plasmid was constructed by first amplifying the Shine–Dalgarno sequence plus ∼148 nt downstream of the start codon of adeABC and then inserting the segment between the HindIII and BamHI sites downstream of the isopropyl β-D-thiogalactoside (IPTG; MCE, Shanghai, China)-inducible promoter in pHN679 (Zheng et al., 2018a). For generating expression plasmids of adeRS asRNA, the Shine–Dalgarno sequence plus ∼146 nt downstream of the start codon of adeRS was amplified. The asRNA expression plasmids, pASadeABC and pASadeRS, were verified by PCR and DNA sequencing, separately transformed into resistant subpopulations and their parental strains by electroporation, and then verified again using PCR. The pHN679 was also introduced into the isolates as a vector control. Strains used for asRNA silencing are listed in Supplementary Table 5 and primers used for the construction and validation of transformed strains are listed in Supplementary Table 6 . The silencing efficacy of the target genes was measured using qRT-PCR as described above. For RNA isolation of the adeABC- and adeRS-silenced strains, the bacteria were pre-treated with 1 mM IPTG to induce expression of the asRNA silencing plasmids. The eravacycline MIC for the adeABC- and adeRS-silenced resistance subpopulations was performed using the broth microdilution method. The frequency of eravacycline heteroresistance in the adeABC- and adeRS-silenced parental strains was determined using the PAP test.

We used Student’s t-test for continuous variables, chi-square analysis for categorical variables. All tests were performed in IBM SPSS Statistics (version 22.0; IBM, Chicago, USA). P-values <0.05 are regarded as statistically significant.

The susceptibilities of 147 CRAB and 140 CSAB strains to eravacycline and other antibiotics are summarized in Supplementary Table 7 . Eravacycline, tigecycline and polymyxins B exhibited excellent antimicrobial activity against CRAB. The MIC₉₀ for eravacycline (1 mg/L) was lower than tigecycline (2 mg/L) and polymyxins B (2 mg/L) against these strains. Eravacycline heteroresistance was determined using the PAP test in 144 CRAB and 139 CSAB strains with an eravacycline MIC ≤4 mg/L. While 25 CRAB strains (17.36%) exhibited heteroresistance to eravacycline, no heteroresistant strains were detected in CSAB isolates. Furthermore, the frequency of eravacycline heteroresistance in CRAB isolates was increased with the patent of MIC value ( Table 1 ).

Colonies selected from the 8 mg/L eravacycline PAP test concentration were categorized as the resistant subpopulations of each eravacycline-heteroresistant isolate. The parental strains of eravacycline-heteroresistant isolates were considered susceptible subpopulations in this study since the majority subpopulations were susceptible to eravacycline. The resistant subpopulations had 8- to 128-fold higher eravacycline MICs than the parental strains ( Table 2 ). The eravacycline MICs of resistant subpopulations were 4- to 8-fold lower after 15 passages on antibiotic-free medium and were reversed to susceptible level after 30 passages ( Supplementary Table 8 ). These findings suggest that these subpopulations lacked stable mutations conferring eravacycline resistance and could be reversed to the susceptible phenotype after removal of the antibiotic pressure.

The epidemiological and molecular characteristics of eravacycline heteroresistant strains are listed in Table 2 . These strains were primarily shown to infect individuals >60 years of age (n = 22, 88%). The heteroresistant strains came from various specimens, including sputum (n = 13, 52%), wound secretions (n = 9, 36%), urine (n = 2, 8%), and blood (n = 1, 4%) that were primarily obtained from the intensive care unit (ICU) (n = 11, 44%), department of neurosurgery (n = 6, 24%) and department of burn plastic surgery (n = 4, 16%). Four patients infected with the heteroresistant strains had a fatal clinical outcome (n = 4, 16%). MLST revealed six sequence types (ST) in 25 eravacycline heteroresistant strains, with ST208 (n = 18, 72%) being the most prevalent. The predominant carbapenemase genes among these strains were bla _OXA-23 (n = 25, 100%) and bla _OXA-24 (n = 13, 52%) ( Table 2 ).

The in vitro antimicrobial activity of common antibiotics against resistant subpopulations and parental strains were compared to measure cross-resistance. The tigecycline MICs of all resistant subpopulations and the levofloxacin MICs of 21 resistant subpopulations were each more than 4-fold higher than those of parental strains ( Table 3 ). These findings indicated that resistant subpopulations had developed cross-resistance toward eravacycline, tigecycline, and levofloxacin.

To investigate the mechanisms of eravacycline heteroresistance, the effect of the EPIs on the eravacycline MIC of resistant subpopulations was determined first. The eravacycline MICs were significantly decreased by ≥4 fold with the addition of the PaβN or CCCP in 88% (22/25) or 64% (16/25) of the resistant subpopulations, respectively ( Table 4 ). These findings indicated that EPIs could potentiate eravacycline activity in heteroresistant CRAB isolates. We further examined the incidence of eravacycline heteroresistance in CRAB isolates by performing a PAP test with the EPIs. Only four eravacycline heteroresistant strains were detected from 144 CRAB isolates after the addition of PaβN, and only nine heteroresistant strains were identified with the addition of CCCP, significantly fewer than those detected in the absence of EPIs (P < 0.05) ( Table 5 ). These results suggested that EPIs can inhibit the formation of eravacycline heteroresistance in CRAB isolates.

The above results indicated that efflux pumps were involved in eravacycline heteroresistance. Thus, the relative expression of efflux pumps associated with eravacycline resistance was compared between eravacycline heteroresistant parental strains and their resistant subpopulations. The expression of adeB was significantly higher in 64% (16/25) of resistant subpopulations compared with the parental strains, and adeS expression was significantly increased in 56% (14/25) of resistant subpopulations ( Figure 1 ). No significant differences were found in the expression of other efflux pump genes. These findings suggested that high expression of the efflux pump, AdeABC, and its regulator, AdeRS, may contribute to eravacycline heteroresistance. Genetic alterations of efflux pumps and ribosomes were then investigated in resistant subpopulations. An ISAba1 insertion was detected in adeS in 40% (10/25) of resistant subpopulations ( Table 6 ), indicating that this insertion may be associated with eravacycline heteroresistance.

To further confirm the role of the efflux pump, AdeABC, and its regulator, AdeRS, in the eravacycline heteroresistance of CRAB isolates, asRNA was used to individually silence the expression of adeABC and adeRS in eravacycline heteroresistant parental strains and their resistant subpopulations. The asRNA plasmids, pASadeABC and pASadeRS, were constructed and transformed into resistant subpopulations or parental strains (five strains per gene) that showed high expression of the target gene, respectively ( Supplementary Table 5 ). The silencing efficacy of the candidate genes in the transformed strains was confirmed by qRT-PCR. After induction with 1.0 mM IPTG, the expression of adeABC and adeRS was downregulated by >80% more in the constructed strains than in the wildtype strains ( Figure 2 ). The eravacycline MICs of resistant subpopulations decreased by 4- to 8-fold after silencing adeABC or adeRS ( Table 7 ). Furthermore, PAP revealed that silencing adeABC or adeRS led to a loss of the eravacycline heteroresistant phenotype in 80% (4/5) or 60% (3/5) of parental strains, respectively ( Supplementary Table 9 ). These results confirmed that high expression of AdeABC or AdeRS contributed to the formation of eravacycline heteroresistance in CRAB isolates.