Colistin Combined With Tigecycline: A Promising Alternative Strategy to Combat Escherichia coli Harboring blaNDM–5 and mcr-1

Infections due to carbapenem-resistant NDM-producing Escherichia coli represent a major therapeutic challenge, especially in situations of pre-existing colistin resistance. The aim of this study was to investigate combinatorial pharmacodynamics of colistin and tigecycline against E. coli harboring blaNDM–5 and mcr-1, with possible mechanisms explored as well. Colistin disrupted the bacterial outer-membrane and facilitated tigecycline uptake largely independent of mcr-1 expression, which allowed a potentiation of the tigecycline-colistin combination. A concentration-dependent decrease in colistin MIC and EC50 was observed with increasing tigecycline levels. Clinically relevant concentrations of colistin and tigecycline combination significantly decreased bacterial density of colistin-resistant E. coli by 3.9 to 6.1-log10 cfu/mL over 48 h at both inoculums of 106 and 108 cfu/mL, and were more active than each drug alone (P < 0.01). Importantly, colistin and tigecycline combination therapy was efficacious in the murine thigh infection model at clinically relevant doses, resulting in >2.0-log10cfu/thigh reduction in bacterial density compared to each monotherapy. These data suggest that the use of colistin and tigecycline combination can provide a therapeutic alternative for infection caused by multidrug-resistant E. coli that harbored both blaNDM–5 and mcr-1.


INTRODUCTION
Infections caused by carbapenem-resistant Enterobacteriaceae (CRE), especially the New Delhi metallo-β-lactamases (NDM)-producing Escherichia coli, have become a global therapeutic challenge in clinical and public health settings (Perez and Bonomo, 2018). In general, isolates carrying bla NDM tend to carry other resistance genes thus limiting treatment options (Falagas et al., 2014;Liu et al., 2019). Currently, the polymyxin antibiotics (polymyxin B and colistin) have reemerged as the last-line therapy against CRE. However, the clinical efficacy of polymyxin antibiotics has been significantly compromised by the widespread emergence of mobile colistin resistance gene mcr-1 . Worryingly, the MCR-1-producing E. coli that coexist with NDM-1, NDM-5, and NDM-9 have been recently reported worldwide, and these isolates possess resistance to fluoroquinolones, sulfonamides, β-lactams, tetracycline, and aminoglycosides (Du et al., 2016;Yao et al., 2016). Fortunately, the level of mcr-1-mediated colistin resistance is moderate , thus the use of colistin-based combinations would be of considerable clinical significance.
Tigecycline is the first of glycylcycline class that exhibited mainly bacteriostatic activity (Meagher et al., 2005). Of note, the decreased clinical efficacy and increased mortality have been previously reported regarding tigecycline monotherapy in the treatment of severe infections (Yahav et al., 2011). Therefore, clinicians should avoid tigecycline monotherapy to reserve it as another last-resort drug.
In this study, we systemically investigated the activity of colistin and tigecycline combination at the clinically achievable concentrations in vitro and in a murine thigh infection model against carbapenem-resistant E. coli harboring bla NDM-5 , especially in situations of pre-existing the mcr-1 gene and high bacterial burdens. Additionally, we explored the underlying mechanisms of this combination (Figure 1) by determination of bacterial out-membrane integrity and tigecycline accumulation.

Combinatorial Susceptibility Testing
The MICs of colistin for each E. coli strain were determined in the absence and presence of twofold increasing tigecycline concentrations (0.13-0.5 mg/L) using a modified broth microdilution method (Wiegand et al., 2008). The interaction of this combination was evaluated in duplicate for each isolate with a checkerboard assay (CST range 0.25-32 mg/L; TGC range 0.015-32 mg/L). Inhibition was read visually to calculate the fractional inhibitory concentration index (FICI), with an FICI ≤ 0.5 deemed synergistic. In addition, cell density was assessed using a spectrometer to estimate cell densities for MacSynergy II analysis (Prichard and Shipman, 1990). The MacSynergy II program uses the Bliss independence algorithm to generate a 3-dimensional response profile of the synergy-antagonism landscape by representing the theoretical indifferent surface. Peaks and troughs represent synergy and antagonism, respectively, and the extents of these were defined using interaction volumes (µM 2 ): <25, additive; 25 to 50, minor but significant; 50 to 100, moderate; and >100, strong synergy or antagonism (Deshpande et al., 2016;Lai et al., 2016). The results were expressed as the mean interaction volumes calculated at the 95% confidence level from three independent experiments.

Assessment of Colistin-Induced Outer-Membrane Disruption
The 1-N-phenylnaphthylamine (NPN) assay was performed to assess bacterial outer-membrane permeability to colistin as previously described (Buyck et al., 2012). Uptake of NPN by E. coli cells was a measure of the degree of permeability, and the subsequent fluorescence indicated a permeability breakdown (Macnair et al., 2018). Thus, NPN uptake was used to quantitatively indicate the colistin-induced outer membrane disruption. Mid-logarithmic cultures of E. coli strains were washed and suspended in PBS to a density of 10 9 cfu/mL (i.e., OD 600nm = 1.0). Bacterial cells were added to PBS containing NPN (10 µM) and varying concentrations of colistin in black 96well microplates. After 1 h of incubation at 37 • C, fluorescence was read using an EnSight multimode plate reader (PerkinElmer, Waltham, MA, United States) at 355 nm excitation and 405 nm emission wavelengths. NPN uptake (%) was calculated for each E. coli strain as described elsewhere (Macnair et al., 2018). Full NPN uptake (100%) was achieved by adding 100 mg/L of colistin.

Intracellular Accumulation of Tigecycline
The levels of tigecycline accumulation by mcr-1-positive andnegative E. coli strains in the absence and presence of colistin were determined as our previously described . Overnight cultures of E. coli strains were diluted to 10 9 cfu/mL into CAMHB and grown in the same medium for 20 min with 10 mg/L of tigecycline alone and in combination with 2 mg/L of colistin. Bacterial cells were collected by centrifugation at 3000 × g for 10 min, washed with sterile normal saline and dried to obtain the dry weight. Bacteria cells were lysed by sonication for 15 min and then centrifuged at 3000 × g for 10 min to remove the cell debris. Tigecycline concentrations in the resulting cell extracts were determined by a LC-MS/MS method (Sun et al., 2019; details are given in the Supplementary Material). All experiments were performed at least five independent biological replicates. Results were expressed as amount of tigecycline incorporated per dry weight of bacteria.

In vitro Time-Kill Experiments
In vitro time-kill experiments were conducted to characterize the activity of the colistin and tigecycline combination using previously described methods (Rao et al., 2016). In brief, overnight E. coli cultures (∼10 6 or 10 8 cfu/mL) were exposed to colistin (2 and 8 mg/L) alone and in combination with tigecycline (0.25 mg/L) over a period of 48 h. The choice of colistin concentrations was based on the clinically achievable serum steady-state concentration (C ss ) and peak concentration (C max ) FIGURE 1 | Graphic potential mechanisms for increased activity of colistin in combination with tigecycline against E. coli harboring bla NDM-5 and mcr-1: colistin-induced bacterial outer-membrane perturbation and tigecycline accumulation. CST, colistin; TGC, tigecycline. in humans, while the tigecycline concentration was chosen to simulate the average C ss at the clinical dose of 50 mg every 12 h (Van Wart et al., 2006;Tran et al., 2016;Nation et al., 2017). Serial samples were obtained at 0, 1, 3, 6, 9, 12, 24, 27, 30, 33, and 48 h after incubation at 37 • C. Bacterial counts were determined based on the quantitative cultures on MHA plates. Historical time-kill data of colistin alone for portion of study strains were obtained from our previous report (Zhou et al., 2017).

In vitro Pharmacodynamic (PD) Analysis
The concentration-effect curves were used to quantitatively evaluate the potency of colistin and tigecycline combination against E. coli strains harboring bla NDM-5 and mcr-1, at initial inoculums of 10 6 and 10 8 cfu/mL, respectively. The testing procedure consisted of four groups, and each group included tubes with twofold increasing concentrations of colistin from 0.5 to 16 mg/L, in the absence and presence of tigecycline at 0.13, 0.25, and 0.5 mg/L. After 48 h of incubation, the microbiological response was measured by the log 10 change in bacterial density vs. pre-exposure at 0 h. The relationships between colistin concentrations and antibacterial response to single and combination therapies were fit to the Hill sigmoid E max model: , where E 0 is the log 10 change in bacterial count without colistin, E max is the maximal effect, EC 50 is the colistin concentration required to achieve 50% of E max and N is the slope of concentration-effect curve. The PD analysis was carried out by the non-linear leastsquares regression in WinNonlin software Version 6.1 (Pharsight, Sunnyvale, CA, United States) (Zhou et al., 2017). The coefficient of determination (R 2 ) was used to estimate the variance of PD regression analysis. Mann-Whitney test was used to compare the parameters of E max and EC 50 between mcr-1-positive and -negative strains. Differences of PD parameter at 10 6 vs. 10 8 cfu/mL inoculum were determined using Wilcoxon signed-rank test in GraphPad Prism 8 software (San Diego, CA, United States) and a P value of <0.05 was considered significant.

Murine Thigh Infection Model and Treatment Regimens
All animal experimental protocols were approved by South China Agricultural University (SCAU) Institutional Animal Ethics Committee (Guangzhou, China) and performed in accordance with the SCAU Institutional Laboratory Animal Care and Use guidelines. Six-week-old, 25-27 g, specificpathogen-free, female ICR mice (Hunan SJA Laboratory Animal, Changsha, China) were rendered neutropenic by administration of cyclophosphamide intraperitoneally as previously described (Zhou et al., 2018). Thigh infections with each E. coli were produced by injecting 0.1 mL of bacterial suspension in normal saline (10 6.5 and 10 8.5 cfu/mL). At 2 h after infection, mice were randomized to receive (i) no therapy (control), (ii) colistin at 7.5 mg/kg intraperitoneally (i.p.) twice a day (bid), (iii) tigecycline at 5 mg/kg subcutaneously (s.c.) bid, or (iv) combination of colistin and tigecycline. The current usual doses of colistin (3 MIU, equivalent to 240 mg, every 8 h) and tigecycline (100 mg initially, then 50 mg bid) were acceptable for the treatment of severe infections in humans (Meagher et al., 2005;Docobo-Perez et al., 2012). In this study, the drug doses in mice were selected to mimic the pharmacokinetic profiles of human clinical doses of 300 and 200 mg, respectively (Meagher et al., 2005;Karnik et al., 2013;Zhou et al., 2017;Zhao et al., 2018). Control and antibiotictreated mice were sacrificed at 24 h after start of therapy. Thigh muscles were aseptically removed, homogenized and bacteria were cultured quantitatively using the plate counting method, and results were expressed as the log 10 cfu/thigh. Three mice (i.e., six thighs) were included in each group. The Mann-Whitney U-test was used to compare bacterial densities in target tissue between mono-and combination therapies.

In vitro Susceptibility and Interaction Assessment
The carbapenem-resistant E. coli strains were highly resistant to almost all tested antibiotics ( Table 1). As expected, E. coli strain 2630 lacking mcr-1 was susceptible to colistin, with an MIC of 0.5 mg/L in the absence of tigecycline (Table 1). However, the strains that harbored bla NDM-5 and mcr-1 were resistant both to meropenem (MIC ≥ 16 mg/L) and colistin (MIC ≥ 4 mg/L). Interestingly, colistin MICs for mcr-1-positive CRE strains decreased to 1/4 to 1/16 of the original levels as tigecycline concentration was raised from 0 to 0.5 mg/L ( Table 1). This was confirmed using the checkerboard assay that showed synergistic effects of the colistin and tigecycline combination. The FICI values varied from 0.38 to 0.5 for all except the colistin susceptible strain 2630 (Table 1). In particular, E. coli 1320 that carried both bla NDM-5 and mcr-1 displayed a highly significant synergistic response to this combination across the range of drug concentrations tested, with a clear peak at 0.5 mg/L tigecycline and 1 or 2 mg/L colistin (Figure 2A). Different degrees of synergy were observed for all study E. coli strains with synergy volumes that ranged from 36.9 to 183 µM 2 ( Figure 2B).

Colistin-Induced Outer-Membrane Perturbation and Tigecycline Accumulation
Carriage of mcr-1 in carbapenem-resistant E. coli strains increased their resistance to colistin-induced outer-membrane disruption as expected. NPN uptake in mcr-1-harboring E. coli was significantly less than E. coli 2630 after exposure to colistin at 0.78 to 12.5 mg/L ( Figure 2C; P < 0.05), with corresponding colistin MIC increases from 8-to 16-fold ( Table 1). The colistin concentrations required to achieve the comparable levels of NPN uptake increased eightfold in mcr-1-positive compared to -negative E. coli strains. For example, 45% of NPN uptake was observed at 0.78 mg/L colistin for colistin-susceptible E. coli 2630, while similar NPN uptake (38% to 53%) occurred at 6.25 mg/L colistin for mcr-1-harboring strains ( Figure 2C). It seems that the additional levels of outer-membrane perturbation in a colistin-susceptible strain can be achieved by increasing the concentration of colistin eightfold in mcr-1-harboring E. coli strains. Importantly, when combined with the clinically relevant concentration of colistin at 2 mg/L, intracellular accumulations of tigecycline markedly increased in all study E. Frontiers in Microbiology | www.frontiersin.org coli strains (P < 0.01; Figure 2D). Although the concentration of 2 mg/L colistin alone was insufficient to inhibit growth of E. coli harboring both bla NDM-5 and mcr-1 (Figures 3H-J), it provided sufficient outer-membrane perturbation to facilitate tigecycline uptake and subsequent tigecycline-induced growth inhibition ( Figure 2D).

In vitro Time-Kill Experiments
At a low inoculum (10 6 cfu/mL), colistin alone at 2 mg/L achieved complete the bactericidal activity (>6.3-log 10 reduction) over 24 h against colistin-susceptible strain 2630. The activity was not further improved at higher colistin levels or in combination with tigecycline ( Figure 3B). Against the colistin-resistant E. coli 1320, the clinically achievable concentrations of colistin resulted in early bactericidal activity only, with a 1.3-to 3.2-log 10 reduction in bacterial density, followed by rapid regrowth beyond 6 h. However, complete bacterial eradication was attained with the combination of 8 mg/L colistin and 0.25 mg/L tigecycline ( Figure 3D). Similarly, in the presence of 0.25 mg/L tigecycline, substantial killing of E. coli 2610 was achieved with >2 mg/L colistin ( Figure 3I). Interestingly, despite the lack of activity that was observed for all colistin monotherapies against E. coli 2121, tigecycline displayed the ability to increase killing activity over 48 h of exposure to colistin ( Figure 3J). Monotherapy with a high colistin concentration (8 mg/L) or the combination of 0.25 mg/L tigecycline and 2 mg/L colistin exhibited sustained bactericidal activity at the high inoculum (10 8 cfu/mL) of E. coli 2630 ( Figure 3F). However, even the high colistin levels of 8 mg/L were inactive for the colistin-resistant strains, whereas in combination with 0.25 mg/L tigecycline resulted in a 2.1-to 3.9-log 10 reduction in bacterial density ( Figures 3H,K-L). Tigecycline monotherapy at 0.06 or 0.25 mg/L performed no different from the growth control against all study E. coli at both low and high inoculums (Figure 3).

Concentration-Effect Relationships
The concentration-effect relationship was fitted to a Hill-type equation (R 2 > 0.95), and the PD parameter of EC 50 representing colistin potency was significantly greater in mcr-1-harboring strains compared with E. coli 2630 (P < 0.01; Table 2). In addition, the EC 50 values at 10 8 cfu/mL inoculum were 1.5-to 18.4-times higher than those at 10 6 cfu/mL inoculum (mean = 5.3, P < 0.001). In the three strains that harbored bla NDM-5 and mcr-1, a clear tendency toward higher E max values were seen with a 10 8 cfu/mL inoculum, whereas no significant difference was noted at 10 6 cfu/mL ( Table 2).
Overall, we found similar dose-dependent shifts with increasing tigecycline levels to a lower colistin concentration required to suppress the growth of E. coli at both inoculums (Figures 3M,N). For example, at 10 6 cfu/mL, inhibition of E. coli 2630 occurred at the colistin concentration of 0.75 mg/L and decreased threefold to 0.25 mg/L in the presence of tigecycline (Supplementary Figure S1C). Carriage of mcr-1 increased the colistin concentration required for growth inhibition to 8 mg/L, which was 11-fold greater than for E. coli 2630 ( Figure 3M). However, in combination with tigecycline from 0.13 to 0.5 mg/L, the colistin levels for growth inhibition were only 0.75 mg/L or twofold and fourfold greater than the concentration needed to synergize with tigecycline against E. coli 2630 (Figure 3M and Supplementary Figure S1C). It seems that the mcr-1 gene only provided protection against colistin monotherapy, but not an ability to resist the colistin and tigecycline combination therapy.

In vivo Efficacy of Mono-and Combination Therapies
During thigh infection with a low initial burden, colistin monotherapy led to decreased bacterial density by 1.62log 10 cfu/thigh for colistin-susceptible E. coli 2630, compared to the untreated control at 0 h ( Figure 4B). However, for colistinresistant strains, neither colistin nor tigecycline monotherapy showed a significant reduction in bacterial density after 24 h of therapy. Interestingly, colistin and tigecycline combination proved efficacious, resulting in >2.0 log 10 cfu/thigh reduction compared to each monotherapy (P < 0.0001, Mann-Whitney U-test; Figures 4C-F). The high initial burden in the murine thigh infection model was used to stimulate the severe infections that result in high mortality, and the effectiveness of combination therapy is a general proof of principle. Monotherapy with colistin or tigecycline did not achieve notable antibacterial effects against E. coli harboring bla NDM-5 and mcr-1 at the high initial inoculum (Figure 5). Importantly, the combination of colistin and tigecycline significantly increased killing activity , 3112 (C) and strains harboring bla NDM-5 and mcr-1 (D-F) in the murine thigh infection model with a low initial burden of 10 6 cfu/thigh. Colistin (7.5 mg/kg i.p. bid) and tigecycline (5 mg/kg s.c. bid) and the combination were administrated at 2 h post-infection. Horizontal lines represent the mean and standard deviation of bacterial densities for each group (n = 6). Colistin and tigecycline combination therapy resulted in a >2.0 log 10 cfu/thigh reduction relative to each monotherapy (P < 0.0001, Mann-Whitney U-test).

DISCUSSION
Treatment options for carbapenem-resistant E. coli infections are very limited especially if the mcr-1 gene is also present in the infecting strains. Tigecycline and colistin are currently the last-resort antibiotics for the treatment of severe infections (Sun et al., 2019). However, tigecycline demonstrates mainly bacteriostatic activity with low serum levels (Van Wart et al., 2006). Concerns have been raised regarding the efficacy of tigecycline monotherapy in the light of decreased clinical success rates (Yahav et al., 2011). Indeed, in the current study, tigecycline monotherapy did not achieve positive outcomes in a murine thigh infection model when the study E. coli strains harbored both bla NDM-5 and mcr-1, despite the fact that most of strains (5/6) remained susceptible to tigecycline except the strain 1320. Fortunately, the presence of mcr-1 only slightly increased the MIC of colistin (Zhou et al., 2017).
Consequently, there was a compelling reason to use colistin and tigecycline in combination. Colistin and tigecycline combination therapy against CRE infection had varying outcomes from synergy to indifference (Bercot et al., 2011;Karaoglan et al., 2013;Rao et al., 2016;Cai et al., 2017;Ku et al., 2017). In this study, combination of clinically achievable concentration of colistin and tigecycline produced a synergistic activity in vitro against E. coli harboring bla NDM-5 and mcr-1, resulting in a >4.0-log 10 cfu/mL reduction by 48 h. An additional in vivo synergistic effect was indeed observed in the murine thigh model, at both low and high inoculums. Supporting our findings, colistin displayed a similar synergistic interaction with tigecycline for carbapenem-resistant A. baumannii and K. pneumoniae (Pournaras et al., 2011;Karaoglan et al., 2013;Ku et al., 2017). Data from previous case reports also showed beneficial activity of tigecycline and colistin combination therapy against K. pneumoniae bacteremia (Cobo et al., 2008). Interestingly, the higher dose of tigecycline has been shown to be associated with better synergistic outcomes against multidrug-resistant CRE, compared with the conventional dosing regimen (De Pascale et al., 2014;Cai et al., 2017). In contrast, a potential trend toward antagonism was observed at lower tigecycline concentrations (Albur et al., 2012). , 3112 (C) and strains harboring bla NDM-5 and mcr-1 (D-F) in the murine thigh infection model with a high initial burden of 10 8 cfu/thigh. Colistin (7.5 mg/kg i.p. bid) and tigecycline (5 mg/kg s.c. bid) and the combination were administrated at 2 h post-infection. Horizontal lines represent the mean and standard deviation of bacterial densities for each group (n = 6). Colistin and tigecycline combination therapy resulted in a >2.5 log 10 cfu/thigh reduction relative to each monotherapy (P < 0.0005, Mann-Whitney U-test).
Of note, previous studies that used this combination employed different methods, and the isolates were not well-described genotypically, making the results difficult to generalize. Here, we demonstrated increased activity of colistin in combination tigecycline against E. coli strains that harbored bla NDM-5 and mcr-1, including the pandemic clonal complex ST648 (Hornsey et al., 2011). The clinical impact of infections due to colistin-resistant NDM-5-producing E. coli is currently unknown, but our findings provide an alternative approach to combat such resistant strains. In support of this view, a recent report indicated that colistin and tigecycline combination was able to prevent the emergence of high-level resistance to these antibiotics (Cai et al., 2017).
The potentiation effect of this combination is most likely related to their different mechanisms of action at separate bacterial targets. Tigecycline acts in the cytoplasm by binding to the ribosomal complex that requires drug to enter the bacterial cells first (Bauer et al., 2004). In general, uptake of tigecycline across the bacterial cell wall and cytoplasmic membrane includes two ways: passive diffusion and an energy-dependent active transport system (Schnappinger and Hillen, 1996;Chopra and Roberts, 2001). In Gram-negative bacteria, the cell wall is surrounded by the outer-membrane and tigecycline moves through membranes via porin channels in the absence of colistin (Roberts, 2003). Colistin resulted in bacterial outer-membrane disruption and instable regions in cytoplasmic membrane that may facilitate tigecycline passive accumulation (Macnair et al., 2018). Supporting this speculation, our NPN uptake and intracellular tigecycline accumulation assays demonstrated that exposure to colistin did promote tigecycline uptake and subsequent tigecycline-induced growth inhibition independent of mcr-1 expression. This scenario has been reported for colistin in combination with minocycline, the prodrug of tigecycline (Liang et al., 2011). However, the precise details of how colistin affects the energy-dependent transport of tigecycline still remain unclear.
Owing to the paucity of novel antibiotics, colistinbased combination therapy was therefore regarded as an alternative approach to combat colistin-resistant CRE infections. A synergistic effect of colistin with amikacin, rifampicin, and osthole has been reported (Lagerback et al., 2016;Zhou et al., 2017Zhou et al., , 2019. However, systemic administration of colistin is associated with nephrotoxicity despite the fact that toxicity is dose-dependent and reversible on discontinuation of treatment (Biswas et al., 2012). Therefore, the clinical utility of colistin should be prudent when used in combination with other nephrotoxic antibiotics such as gentamicin and amikacin. Previous nephrotoxicity studies in mice indicated that only mild kidney damage was observed until an accumulated dose of 72 mg/kg colistin, and suggested an acceptable colistin single dose ranges within 40 mg/kg in mice (Cheah et al., 2015;Roberts et al., 2015). Therefore, the much lower colistin dose (7.5 mg/kg) that used in this study should be safe for mice by comparison. In fact, many previous studies have employed 7.5 mg/kg colistin to carry out in vivo efficacy studies in mice Zhou et al., 2017;Macnair et al., 2018). In the present study, tigecycline demonstrated bactericidal activity against E. coli harboring bla NDM-5 and mcr-1 when combined with the clinically relevant concentration of colistin at 2 mg/L, which is considered as the appropriate partnered concentration to avoid renal impairment (Tran et al., 2016). Importantly, the combination of tigecycline with colistin we studied here may allow lower colistin dose sparing regimens that reduce nephrotoxicity for treating colistinresistant CRE infections. Previous comparative observational studies also showed a lower-than-expected toxicity for tigecycline and colistin combination therapy (Zhang et al., 2013). Even patients with kidney disease could benefit from colistin-based combination therapy, when provided with a lower daily dose of colistin achieving comparable efficacy (Falagas et al., 2006;Biswas et al., 2012). In addition, a retrospective cohort study indicated that colistin is a valuable antibiotic with acceptable nephrotoxicity (<7%) and considerable efficacy that depends on daily dose (Falagas et al., 2010).
Our investigation has several limitations. For example, the combination was evaluated in a small number of strains despite the different clonal types. In addition, the murine thigh model is a local infection model, and additional study is needed to evaluate the usefulness of this combination in the clinical setting. Moreover, based on our current results, we do not know whether the colistin-induced increased accumulation of tigecycline in bacterial cells is "drug specific" or more broad range for other antibiotics. Although this is beyond the scope of this study, future studies should examine this potential mechanism.
In summary, this study demonstrated increased activity of colistin and tigecycline combination against E. coli harboring bla NDM-5 and mcr-1. Importantly, a potentiation effect occurred at the clinically relevant concentrations of colistin and tigecycline, and was efficacious in the murine thigh infection model. In addition, we demonstrated for the first time that colistin permeabilization of the bacterial outer-membrane facilitates the uptake of tigecycline, contributing to increased activity of the combination.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation, to any qualified researcher.

ETHICS STATEMENT
This study was carried out in accordance with the recommendations of ethical guidelines of South China Agricultural University. All animal experimental protocols and isolation procedures for E. coli strains were reviewed and approved by the South China Agricultural University Institutional Animal Ethics Committee (2019B161 and 2018B095). Individual written informed consent for the use of isolates was obtained.

AUTHOR CONTRIBUTIONS
Y-HL and Y-FZ designed the study and wrote the manuscript. Y-FZ, PL, and C-JZ carried out the experiments. JS and X-PL analyzed the data. All authors read and approved the final manuscript.