The phosphoethanolamine transferase PetL of Pasteurella multocida is associated with colistin resistance

Abstract

The rapid emergence and spread of colistin-resistant gram-negative bacteria has raised worldwide public health concerns, and phosphoethanolamine (PEtn) transferase modification-mediated colistin resistance has been widely documented in multiple gram-negative bacterial species. However, whether such a mechanism exists in the zoonotic pathogen Pasteurella multocida is still unknown. Recently, a novel PEtn transferase, PetL, was identified in P. multocida, but whether it is associated with colistin resistance remains to be elucidated. In this study, we found that PetL in P. multocida (PetL^PM) exhibited structural characteristics similar to those of the mobile-colistin-resistant (MCR) protein and the PEtn transferase characterized in Neisseria meningitidis. The transformation of petL^PM into E. coli or K. pneumoniae changed the phenotype of several tested strains from colistin sensitive to colistin resistant. Deletion of this gene decreased the colistin minimum inhibitory concentration (MIC) of P. multocida by 64-fold. Our extensive analysis by MALDI-TOF–MS demonstrated that PetL^PM participated in the modification of bacterial lipopolysaccharide (LPS)-lipid A. Deletion of petL^PM led to an increase in membrane charge but a decrease in cell-surface hydrophobicity and cell permeability in P. multocida. The present study is the first to report the presence of PEtn transferase-mediated colistin resistance in the zoonotic pathogen P. multocida.

Introduction

The spread of drug-resistant bacteria, including multidrug-resistant (MDR), extensive drug-resistant (XDR), and pandrug-resistant (PDR) bacteria, poses serious threats to global public health (Walsh et al. 2023). The World Health Organization (WHO) has listed several antimicrobials as critically important agents for human medicine (WHO 2023). Notably, colistin has been recognized as a “last-resort” antibiotic for treating infections associated with MDR gram-negative bacteria (Sharma et al. 2022). As a positively charged polycationic peptide, colistin mainly works by interacting with negatively charged phosphate groups of bacterial lipopolysaccharide (LPS)-lipid A, thereby disrupting bacterial permeability and leading to bacterial lysis (Poirel et al. 2017). In recent years, the worldwide increase in the emergence and dissemination of colistin-resistant bacteria, particularly those harboring the plasmid-borne mobile-colistin-resistant (MCR) gene mcr (Liu et al. 2016), has raised a global public health concern (Seethalakshmi et al. 2023). In addition, a chromosome-borne phosphoethanolamine (PEtn) transferase, which plays a role in modifying LPS lipid A by adding PEtn residues, has also been found to be associated with colistin resistance in several bacterial species, including Neisseria meningitidis, Acinetobacter baumannii and Vibrio parahaemolyticus (Huang et al. 2018). The mechanisms associated with colistin resistance in other gram-negative bacteria remain to be explored due to the importance of this agent in combating infections.

The Gram-negative zoonotic pathogen Pasteurella multocida is a leading cause of multiple economic-impact diseases in agriculture (Wilkie et al. 2012; Wilson and Ho 2013). Infections from this agent may also lead to pneumonia and other respiratory disorders, bacteremia, meningitis, sepsis, and even peritonitis in humans (Piorunek et al. 2023). In recent years, the morbidity and mortality of human Pasteurellosis have increased worldwide (Wilson and Ho 2013). This is partly due to the rapid increase in populations of companion animals (e.g., dogs and cats) and food-producing animals (e.g., pigs and cattle) worldwide to meet the nutrient and spiritual demands of humans (Peng et al. 2022). Because P. multocida commonly exists in the upper respiratory tracts of these animal species, and animal exposure, particularly biting, scratching and even licking companion animals, is the primary reason for humans acquiring pasteurellosis (Wilson and Ho 2013). Although data regarding colistin resistance in P. multocida are relatively scarce, several recent studies from other groups have shown that P. multocida strains originating from avians and ruminants exhibit high levels (between 60 and 70%) of resistance to colistin (El-Demerdash et al. 2023; Sebbar et al. 2020). However, the related mechanism remains to be elucidated. Recently, a novel PEtn-transferase, PetL, was identified in P. multocida (PetL^PM), and it is likely to play a role in transferring PEtn to LPS-lipid A (Harper et al. 2017). The aim of this study was to investigate whether PetL^PM is associated with colistin resistance.

Results

P. multocida PetL is associated with colistin resistance

Considering that there is no breakpoint available for the interpretation of colistin resistance in P. multocida, we amplified the full-length of the petL gene from P. multocida HuN001 (GenBank accession no. CP073238) and cloned this gene (petL^PM) into different plasmids, followed by transformation of the constructed petL^PM-bearing plasmids into different Klebsiella pneumoniae and Escherichia coli strains. Measurement of the minimal inhibitory concentration (MIC) of colistin on E. coli or K. pneumoniae strains containing petL^PM-bearing plasmids or empty plasmids through the broth microdilution method showed that the transformation of petL^PM led to an increase in the colistin MIC (Fig. 1A, B). In E. coli, the transformation of petL^PM caused increases of 16- and 64-fold in the colistin MICs, respectively (Fig. 1A). In K. pneumoniae, this process induced increases of 4- and 16-fold, respectively (Fig. 1B). The phenotype of several tested strains changed from colistin sensitive to colistin resistant after petL^PM transformation, as interpreted using the Clinical & Laboratory Standards Institute (CLSI) breakpoint for colistin (2 μg/mL). We next knocked out petL from HuN001 and measured the MIC of colistin on the petL-deletion strain (PM-ΔpetL), the wild-type strain (PM-WT), and the petL-complementary strain (PM-CpetL). The results demonstrated that the deletion of petL led to a 64-fold decrease in the colistin MIC in P. multocida (PM-WT vs. PM-ΔpetL: 2 μg/mL vs. 0.03125 μg/mL; Fig. 1C). Measurement of growth conditions showed that the deletion of petL did not affect the growth of P. multocida (Fig. 1D). The above findings indicated that P. multocida PetL was associated with colistin resistance.

Fig. 1

Influence of the phosphoethanolamine transferase PetL on the colistin susceptibility of Pasteurella multocida. A A column chart showing the minimal inhibitory concentration (MIC) values of colistin on different E. coli strains containing plasmid-bearing petL from P. multocida or with or without the corresponding plasmids; B A column chart showing the MIC values of colistin on different Klebsiella pneumoniae strains containing plasmid-bearing petL from P. multocida or with or without the corresponding plasmids; C A column chart showing the minimal inhibitory concentration (MIC) values of colistin on the P. multocida wild type strain (PM-WT), petL deletion strain (PM-ΔpetL), and petL complementary strain (PM-CpetL); D A line chart showing the OD₆₀₀ values of the P. multocida wild type strain (PM-WT), petL deletion strain (PM-ΔpetL), and petL complementary strain (PM-CpetL) in TSB media at different time points

P. multocida PetL participates in the modification of the bacterial LPS lipddA

To understand the working mechanism of PetL^PM, we predicted the structure of PetL^PM and compared it to the structures of MCR-1 (PDB ID code 5GOV) and PetL in N. meningitidis (Pet^NM; PDB ID code 5FGN). The results demonstrated that PetL^PM was an integral membrane protein with 5 N-terminal transmembrane helices (Fig. 2A). This topological characteristic was similar to that of MCR-2 (Sun et al. 2017). The predicted structure was highly homologous to that of MCR-1 (Fig. 2B) and Pet^NM (Fig. 2C). These findings suggest that PetL^PM may work through a similar mechanism to that of MCR. We therefore transformed petL^PM into E. coli MG1655 with the help of a P. multocida-E. coli shuttle vector pPBA1101 and isolated LPS-lipid A from MG1655 harboring petL^PM (MG1655-petL^PM) as well as MG1655 with (MG1655-pPBA) or without (MG1655) the empty vector pPBA1101 for analysis by MALDI-TOF-MS using the same model and analysis methodology for the functional characterization of MCR-2 (Sun et al. 2017). Lipid A modification signaling was detected in MG1655-petL^PM but not in MG1655-pPBA or MG1655 (Fig. 3). We also isolated LPS-lipid A from PM-WT and PM-ΔpetL for extensive analysis by MALDI-TOF–MS, but this approach still did not provide any indicative results from the mass data obtained.

Fig. 2

Predicted structural characteristics of the phosphoethanolamine transferase PetL in Pasteurella multocida. A Structural characteristics of PetL in P. multocida predicted by the TMHMM server V. 2.0; B predicted structure of PetL in P. multocida and its comparison with the structure of MCR-1 (ID code 5GOV); C predicted structure of PetL in P. multocida and its comparison with the structure of PetL characterized in Neisseria meningitidis (PetL.^NM; ID code 5FGN)

Fig. 3

Analysis of Pasteurella multocida lipopolysaccharide (LPS) and lipid A. A Separation of LPS extracted from P. multocida strain HuN001 by SDS polyacrylamide gel electrophoresis (SDS-PAGE); B Matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF-MS) analysis of lipid A extracted from E. coli MG1655; C MALDI-TOF-MS analysis of lipid A extracted from E. coli MG1655 containing plasmid pPBA1101; D MALDI-TOF–MS analysis of lipid A extracted from E. coli MG1655 containing petL from P. multocida HuN001 carried by plasmid pPBA1101

P. multocida PetL plays a role in maintaining bacterial membrane charge, cell surface hydrophobicity, and cell permeability

Considering the important role of the electrostatic interaction between the positively charged colistin and the negatively charged gram-negative bacterial lipid A in the work of colistin (Poirel et al. 2017), we measured the surface negative charge of different P. multocida strains using FITC-labeled poly-L-lysine. More negative surface charges (increased 10.55%) were detected in PM-ΔpetL than in PM-WT (Fig. 4A). In transformed E. coli strains with different petL^PM-bearing plasmids, the number of negative surface charges (DH5α (pPBA-1101) decreased by 12.24% compared to that of DH5α (pPBA-petL)), and DH5α (pMD19-T) decreased by 8.76% compared to that of DH5α (pMD19-petL)) compared to that of the control strains (Fig. 4B). Measurement of cell-surface hydrophobicity revealed that PM-ΔpetL displayed lower cell-surface hydrophobicity (decreased 19.5%) than PM-WT (Fig. 4C). In addition, PM-ΔpetL exhibited decreased cell permeability (20.71% lower) compared with that of PM-WT (Fig. 4D).

Fig. 4

Characteristics of the Pasteurella multocida wild-type strain (PM-WT), petL deletion strain (PM-ΔpetL), and petL complementation strain (PM-CpetL). A Membrane surface-negative charges of PM-WT, PM-ΔpetL and PM-CpetL; B membrane surface-negative charges of E. coli strains harboring petL from P. multocida; C cell surface hydrophobicity of PM-WT, PM-ΔpetL and PM-CpetL; D cell permeability of PM-WT, PM-ΔpetL and PM-CpetL

Discussion

As an important “last-resort” antibiotic, colistin is crucial for combating MDR-gram-negative bacteria, and the emergence of bacterial agents with decreased susceptibility to colistin has raised great public health concerns worldwide (Seethalakshmi et al. 2023). Although methodological difficulties hampering the analysis of susceptibility to colistin have led to relatively scarce data on colistin resistance in P. multocida (Kempf et al. 2016), several recent studies have shown that P. multocida strains originating from avians and ruminants exhibit high levels (between 60 and 70%) of resistance to colistin (El-Demerdash et al. 2023; Sebbar et al. 2020). These findings suggest that colistin resistance might also be a problematic phenotype in P. multocida. We believe that the susceptibility profiles of P. multocida to colistin should not be ignored due to the potential public health risks of this bacterial species (Peng et al. 2022). Considering that the treatment of human pasteurellosis mainly relies on the administration of antibiotics (Wilson and Ho 2013) and the susceptibility of P. multocida to many antimicrobial agents is decreasing (Michael 2018), colistin may play a key role in treating human pasteurellosis caused by MDR-P. multocida in specific cases, and exploring the mechanisms associated with colistin resistance in this zoonotic bacterial species is therefore of clinical significance.

In this study, we demonstrated that the newly characterized PEtn transferase PetL (Harper et al. 2017) is associated with colistin resistance, and our bioinformatics analysis revealed that PetL^NM exhibited structural characteristics similar to those of both MCR-1 and PetL^NM. Notably, the roles of PetL^NM and MCR-1 in conferring colistin resistance have been fully clarified in N. meningitidis and Enterobacteriaceae, respectively (Liu et al. 2016; Anandan et al. 2017). Therefore, the working mechanism by which PetL^PM mediates colistin resistance might also be similar to that of both MCR-1 and PetL^NM. According to published articles (Sun et al. 2017; Anandan et al. 2017), both MCR proteins and PetL^NM are indeed PEtN transferases and play roles in modifying lipid A by transferring PEtN from its primary phosphatidylethanolamine, and this role of several MCR proteins, such as MCR-2, has been experimentally demonstrated in the E. coli MG1655 model using MALDI-TOF–MS (Sun et al. 2017). Using the same model and analysis methodology, a clear curve reflecting lipid A modification in MG1655 containing petL^PM was observed in this study. However, we did not obtain any indicative results from the mass data generated by MALDI-TOF-MS analysis of LPS-lipid A from PM-WT, PM-ΔpetL or PM-CpetL. This might be because the structure of LPS of P. multocida is complex (Harper and Boyce 2017; Harper et al. 2011), and there is still a lack of reference data for the analysis. However, it is noteworthy that a recent study experimentally verified the role of PetL in transferring PEtn to lipid A in P. multocida (Harper et al. 2017). It is known that PEtn residues are positively charged; the deletion of PetL reduces the transformation of the positively charged PEtn to bacterial lipid A, and the bacterial surface-negative charge might therefore increase. This could explain why PM-ΔpetL possessed more negative charges than did PM-WT. An increase in the surface negative charge of PM-ΔpetL might be beneficial for the binding of positively charged colistin (Poirel et al. 2017). In addition, PM-ΔpetL exhibited decreased cell permeability compared with that of PM-WT. The above changes may contribute to the decreased bactericidal activity of colistin, possibly explaining why PM-ΔpetL had a lower colistin MIC than PM-WT.

In conclusion, we demonstrated that the newly identified PEtn transferase PetL of P. multocida is associated with colistin resistance. The present study is the first to report the presence of PEtn transferase-mediated colistin resistance in the zoonotic pathogen P. multocida. Although colistin is currently not frequently used for the treatment of P. multocida infections in clinical practice, our study still provides information on the selection of colistin when necessary. This might also be important due to the zoonosis of this bacterial species, as the rapid increase in the population of pets worldwide may increase the public risk of P. multocida in the future.

Materials and methods

Bacterial strains, plasmids, and bioinformatics analysis

The bacterial strains and plasmids used in this study are listed in Table 1. Unless otherwise specified, P. multocida strains were cultured as described previously (Lv et al. 2023). K. pneumoniae and E. coli strains were incubated in Luria–Bertani (LB) broth (Sigma‒Aldrich, St. Louis, US) or in agar at 37°C. When necessary, kanamycin (100 µg/mL) was given to the medium-cultured bacterial strains containing plasmid pPBA1101 (gifted by Prof. John D. Boyce), while ampicillin (100 µg/mL) was used for the culture of bacterial strains containing pMD₁₉-T. The putative structure of PetL^PM was predicted using the I-TASSER server (https://seq2fun.dcmb.med.umich.edu//I-TASSER/). The crystal structures of MCR-1 (ID code 5GOV) and PetL in N. meningitidis (PetL^NM; ID code 5FGN) were downloaded from the PDB database (www.pdb.org). Protein structures were compared using PyMOL 2.0 (https://pymol.org/2/). The characteristics of PetL^PM were analyzed using the TMHMM server V. 2.0 (http://www.cbs.dtu.dk/services/TMHMM/).

Table 1 Bacterial strains, plasmids and primers used in this study

Construction of plasmids and different types of bacterial strains

To construct petL^PM-bearing Enterobacteriaceae, full-length petL of P. multocida HuN001 (GenBank accession no. CP073238) (Lin et al. 2021) was amplified using the primers pPBA-HindIII-F/pPBA-Sac I-R or pMD19-Sal I-F/pMD19-Sac I-R, as shown in Table 1, and subsequently cloned and inserted into pPBA1101 and pMD₁₉-T to generate the recombinant plasmids pPBA-petL^PM and pMD-petL^PM, respectively. These recombinant plasmids were subsequently transformed into E. coli DH5α (pPBA-petL^PM and/or pMD-petL^PM), MG1655 (pPBA-petL^PM), and K. pneumoniae (pPBA-petL^PM) strains.

Our previously reported protocol (Lv et al. 2023) was followed to construct the P. multocida petL-deletion strain (PM-ΔpetL). Briefly, the upstream and downstream fragments of petL were amplified using the genomic DNA of P. multocida HuN001 as a template with primers (petL-L-RH-F/petL-L-RH-R, petL-R-RH-F/petL-R-RH-R), as shown in Table 1. The fragments were ligated together by overlap PCR and cloned and inserted into the plasmid pSHK5(TS)-NgAgo (Fu et al. 2019) to generate the plasmid pSHK-ΔpetL, which was subsequently transformed into competent P. multocida HuN001 cells by electroporation (2500 V, 25 μF, 600 Ω). P. multocida cells were incubated in prewarmed tryptic soy broth (TSB; Becton, Dickinson and Company, USA) supplemented with 5% newborn bovine serum (NBS; Tianhang, Hangzhou, China) and shaken at 180 rpm and 28°C for 3 h. Afterwards, the bacterial culture mixture was centrifuged at 6000 rpm for 15 min, and the pellets were suspended in 100 μL of TSB supplemented with 10% NBS. Finally, the suspensions were plated on tryptic soy agar (TSA) supplemented with 5% NBS and kanamycin (50 μg/mL final concentration), followed by incubation at 28°C for 48 h. Single colonies were picked, and the occurrence of the first crossover event was confirmed by PCR with primers petL-outer-JD-F/petL-outer-JD-R. Colonies with correct profiles were incubated on TSA containing 5% NBS at 42°C overnight to accelerate the occurrence of the second crossover event. The deletion of petL was finally confirmed by PCR with the primers petL-inner-JD-F/petL-inner-JD-R and was double confirmed by Sanger sequencing. To construct complementary strains (PM-CpetL), full-length petL was amplified via PCR with the primers pPBA-HindIII-F/pPBA-Sac I-R and subsequently cloned and inserted into pPBA1101 to generate the transforming plasmid pPBA-petL, which was subsequently transformed into competent PM-ΔpetL cells via electroporation.

Antimicrobial susceptibility testing (AST)

MIC values were measured by the broth microdilution method following the guidelines published by CLSI (CLSI 2018). Briefly, Mueller–Hinton broth (Sigma‒Aldrich, St. Louis, US) containing a series of twofold dilutions of colistin (MedChemExpress, Monmouth Junction, US) ranging from 64 µg/ml to 0.0625 µg/mL was seeded into wells (100 µL per well) in different columns of a 96-well plate (Corning, Corning, US). In parallel, bacterial strains (P. multocida, E. coli, or K. pneumoniae) were prepared by suspending overnight cultures on Mueller-Hinton agars in sterile normal saline to a turbidity of 0.5 McFarland standard. Afterwards, an additional 100-fold dilution was applied to the bacterial suspensions, which were added to the wells (100 µL per well) of the abovementioned plates containing different concentrations of colistin. The control wells included those containing Mueller–Hinton broth only (200 µL per well) and/or those containing 100 µL of broth plus 100 µL of bacterial suspension. The plate was then incubated at 37°C overnight, and parallel tests were performed to confirm the results. The CLSI breakpoint of colistin (2 µg/mL) was used for the interpretation of the testing results for E. coli and K. pneumoniae. E. coli ATCC25922 was used as a quality control.

Measurement of growth curves

Overnight cultures of P. multocida wild-type (PM-WT) and PM-ΔpetL strains were inoculated into fresh TSB supplemented with 5% NBS at a ratio of 1:100 (v/v). Thereafter, each of the bacterial inoculations (200 μL) was added to different wells of a 96-well plate (JET BIOFIL, Guangzhou, China). Growth curves for different bacterial strains were generated by measuring the OD₆₀₀ values every 1 h for 24 h using a fast automatic growth analyzer (OyGrowth Curves Ab Ltd., Finland). The experiments were repeated three times.

Measurement of bacterial surface hydrophobicity

Bacterial surface hydrophobicity was determined as previously described (Wang et al. 2014). Briefly, overnight cultures of PM-WT, PM-ΔpetL, and PM-CpetL in TSB containing 5% NBS were centrifuged at 6000 rpm for 5 min. Pellets were washed twice using PBS (pH = 7.4) and resuspended to an OD₆₀₀ of 1.0, which was recorded as A₀. Thereafter, 5 mL of bacterial culture was mixed thoroughly with 2 mL of xylene, followed by standing for 60 min at room temperature. After that, the OD₆₀₀ of the aqueous phase after extraction with xylene was also measured and recorded as A. The hydrophobicity of the bacterial cells was calculated as [(A₀ − A)/A₀] × 100 (Wang et al. 2014). The experiments were repeated three times.

Determination of outer membrane permeability

OM permeability was determined following a previously described method (Li et al. 2023). Briefly, overnight cultures of PM-WT, PM-ΔpetL and PM-CpetL in TSB containing 5% NBS were centrifuged at 6000 rpm for 5 min. The pellets were washed twice with PBS (pH = 7.4) and resuspended to an OD₆₀₀ of 0.5. Afterwards, 1.92 mL of bacterial suspension and 80 μL of N-phenyl-1-naphthylamine (NPN, Sigma‒Aldrich, St. Louis, US) solution (1 mM) were mixed rapidly. Changes in fluorescence intensity were measured using an EnSpire® Multimode Plate Reader (PerkinElmer, Waltham, US). The excitation and emission wavelengths were 350 nm and 428 nm, respectively. For each bacterial strain, five parallel tests were performed, and instruments calibrated with PBS were included as controls.

Measurement of surface negative charges

Bacterial membrane surface charges were measured as described previously (Jangir et al. 2022). Briefly, overnight cultures of PM-WT, PM-ΔpetL and PM-CpetL in TSB containing 5% NBS were centrifuged at 6000 rpm for 5 min. Pellets were washed twice using PBS (pH = 7.4) and resuspended to an OD₆₀₀ of 0.1. Instruments calibrated with PBS were included as controls. Next, FITC-labeled poly-L-lysine (Sigma-Aldrich) was added (final concentration: 6.5 µg/mL) and incubated at room temperature for 10 min. After centrifugation at 6000 rpm for 5 min, the supernatants were collected to measure the changes in fluorescence intensity using an EnSpire® Multimode Plate Reader (PerkinElmer). The excitation and emission wavelengths were 500 nm and 530 nm, respectively. Membrane surface negative charges were quantified as previously described (Jangir et al. 2022).

LPS extraction and SDS polyacrylamide gel electrophoresis (SDS‒PAGE)

LPSs of different bacterial strains were extracted using a commercial LPS Extraction Kit (iNtRON, Seongnam, South Korea) and separated using SDS polyacrylamide gel electrophoresis (SDS-PAGE). Afterwards, the gel was fixed in a prepared fixative solution (30 mL of anhydrous ethanol + 10 mL of glacial acetic acid + 40 mL of double steaming water) and shaken at 180 rpm overnight at room temperature. The solution was then discarded, and 100 mL of periodic acid solution (7 mg/mL) was added. Following shaking at 180 rpm for 10 min at room temperature, the gel was washed three times with 100 mL of double steaming water and stained with silver nitrate solution (1 mg/mL) for 30 min at room temperature. Finally, the gel was treated with 30 mg/mL sodium carbonate solution plus 0.02% formaldehyde for 20 min until the bands appeared. The reaction was stopped by the addition of 10% glacial acetic acid.

Lipid A extraction and analysis

Bacterial LPS-lipid A was isolated and purified for MALDI-TOF–MS analysis as previously described (Sun et al. 2017). Briefly, bacterial strains in the mid-log phase were inoculated on TSA supplemented with 0.8 mM IPTG and 1 µg/mL colistin and incubated at 37°C for 16- 18 h. Bacterial colonies were subsequently washed with 10 mL of Tris-HCl buffer (30 mM, pH 8.1) and subsequently centrifuged at 4°C and 5000 rpm for 15 min. The resulting pellets were subsequently resuspended in 400 µL of Tris-HCl (30 mM, pH 8.1) solution containing 20% sucrose and treated with 40 µL of lysozyme (1 mg/mL), followed by incubation in an ice bath for 30 min. The resulting bacterial suspensions were subsequently stored at -80°C for 30 min and melted at room temperature. This process was repeated twice. Then, the bacterial suspensions were mixed with 5 mL of EDTA (3 mM) and sonicated for 4 min (sonicating for one second and stopping for 3 s). The products were centrifuged at 4℃ and 5000 rpm for 15 min, followed by centrifugation of the supernatants at 4℃ and 16,800 rpm for 60 min to harvest the LPS. To obtain lipid A, the prepared LPS was dissolved in 200 µL of Tris–HCl solution (30 mM, pH 8.1) containing 0.2% SDS, which was then treated with RNase I (25 µg/mL) and DNase I (100 µg/mL) at 37°C for 2 h. The product was then treated with protease K (100 µg/mL) at 37°C for another 2 h, followed by heating at 100°C for 1 h. Thereafter, 200 µL of acidified ethanol solution (20 mL of 95% ethanol mixed with 100 µL of 4 M HCl) was added, and the mixture was mixed thoroughly. The mixture was then centrifuged at 5000 rpm for 5 min to harvest the pellets, which were then washed twice with 200 µL of 95% ethanol, followed by washing with 200 µL of absolute ethyl alcohol. Finally, lipid A was dissolved in 20 µL of chloroform–methanol solution (2/1, v/v). The extracted lipid A was extensively analyzed by MALDI-TOF–MS.

Statistical analysis

Statistical analysis was performed using multiple t tests in GraphPad Prism 8.0 (GraphPad Software, San Diego, CA). The data are presented as the means ± standard deviations (SD). The significance level was set at a P value of < 0.05 (*), a P value of < 0.01 (**), or a P value of < 0.001 (***).