A novel small RNA PhaS contributes to polymyxin B-heteroresistance in carbapenem-resistant Klebsiella pneumoniae

ABSTRACT

In recent years, polymyxin has been used as a last-resort therapy for carbapenem-resistant bacterial infections. The emergence of heteroresistance (HR) to polymyxin hampers the efficacy of polymyxin treatment by amplifying resistant subpopulation. However, the mechanisms behind polymyxin HR remain unclear. Small noncoding RNAs (sRNAs) play an important role in regulating drug resistance. The purpose of this study was to investigate the effects and mechanisms of sRNA on polymyxin B (PB)-HR in carbapenem-resistant Klebsiella pneumoniae. In this study, a novel sRNA PhaS was identified by transcriptome sequencing. PhaS expression was elevated in the PB heteroresistant subpopulation. Overexpression and deletion of PhaS were constructed in three carbapenem-resistant K. pneumoniae strains. Population analysis profiling, growth curve, and time-killing curve analysis showed that PhaS enhanced PB-HR. In addition, we verified that PhaS directly targeted phoP through the green fluorescent protein reporter system. PhaS promoted the expression of phoP, thereby encouraging the expression of downstream genes pmrD and arnT. This upregulation of arnT promoted the 4-amino-4-deoxyL-arabinosaccharide (L-Ara4N) modification of lipid A in PhaS overexpressing strains, thus enhancing PB-HR. Further, within the promoter region of PhaS, specific PhoP recognition sites were identified. ONPG assays and RT-qPCR analysis confirmed that PhaS expression was positively modulated by PhoP and thus up-regulated by PB stimulation. To sum up, a novel sRNA enhancing PB-HR was identified and a positive feedback regulatory pathway of sRNA-PhoP/Q was demonstrated in the study. This helps to provide a more comprehensive and clear understanding of the underlying mechanisms behind polymyxin HR in carbapenem-resistant K. pneumoniae.

Introduction

In recent years, the widespread of carbapenem-resistant Enterobacteriaceae (CRE) has become a serious problem. CRE has been listed as an urgent antibiotic resistance threat by the Centers for Disease Control and Prevention of the United States and in the World Health Organization’s priority list of bacteria for which new antimicrobials are urgently needed [1]. A multicenter study covering 25 cities and provinces in China showed that carbapenem resistant Klebsiella pneumoniae (CRKP) accounted for the highest proportion of CRE strains (66.7%) [2]. An observational study involving 3195 CRKP-infected patients showed that the mortality rate associated with CRKP infection was high (37.2%) [3]. Besides, the mortality rate of CRKP infected patients (42.14%) was significantly higher than that of carbapenem-sensitive K. pneumoniae infected patients (21.16%) [4]. The high mortality rate of CRKP infections is mainly due to the multidrug resistance of CRKP and the very limited antibiotics available for clinical treatment. Polymyxin has been considered as the last resort treatment option for CRE infections [5,6]. However, the recent emergence of heteroresistance (HR) to polymyxin in CRKP has posed a diagnostic and therapeutic dilemma for clinicians.

HR refers to the existence of subpopulations with inconsistent sensitivity to a particular antimicrobial agent of a strain from the same clonal origin, that is, the phenomenon where resistance and sensitivity to a certain antibiotic occur at the same time [7,8]. The emergence of HR makes clinical treatment increasingly difficult. HR can interfere with antimicrobial susceptibility testing in clinical microbiology laboratories. Even when drug sensitivity test indicates sensitivity, antibiotic treatment can still turn out to be ineffective due to the existence of a resistant subpopulation [9].

However, the mechanisms behind polymyxin HR remain unclear. Polymyxin mainly interacts with the conserved lipid A in the outer membrane of Gram negative bacteria, converting Gram negative bacteria to lyse bacterial cells and producing bactericidal effects [10]. Bacteria can protect themselves from polymyxin lysis by modifying or strengthening lipid A. The reported mechanism behind HR to polymyxin is mainly related to an increase in lipid A modification mediated by the activation of two-component system (TCS). The PhoP/Q, PmrA/AB and CrrA/B TCSs regulate eptA or arnBCADTEF operon, which promotes the phosphoethanolamine (pEtN) or 4-amino-4-deoxyL-arabinosaccharide (L-Ara4N) modification of lipid A [11]. This increases the positive charge of the cell membrane and reduces the electrostatic adsorption of the cationic antimicrobial peptide polymyxin, mediating polymyxin resistance or HR.

Extensive research showed that regulatory small non-coding RNAs (sRNAs) could modulate antibiotic resistance and sensitivity [12,13]. sRNAs are key players in regulating various biological and physiological processes of cells under specific conditions. Typically, sRNAs function by either binding to a protein thereby modifying its activity, or as antisense regulators by base-pairing with their target mRNAs. The base-pairing sRNAs can regulate transcription, translation, and mRNA stability [14].

Our previous study found that CRKP has HR phenomenon to polymyxin [15], but the mechanism is still unclear. In this paper, we investigated the mechanism under the polymyxin B (PB)-HR in CRKP strains by whole genome sequencing (WGS) and transcriptome sequencing (RNA-Seq) and discovered a novel sRNA by bioinformatics analysis. We demonstrated the positive feedback regulatory pathway of sRNA PhaS-PhoP/Q TCS in CRKP strains. This study refines the mechanism of sRNA function in bacterial drug resistance.

Materials and methods

Bacterial strains, plasmids, and growth conditions

All strains and plasmids used in this study are listed in Table S1. MGH78578 was a gift from Prof. Tieli Zhou (The First Affiliated Hospital of Wenzhou Medical University). Unless indicated otherwise, bacteria cultures were grown at 37°C in LB medium or on LB plates containing 1.5% agar. Apramycin (50 mg/mL) was required when the strain contained pSTVA, pSTVA-psRNA686, pSTVA-phoP, or pIJ773 plasmids. Hygromycin (100 mg/mL) was required when the strain contained pACBSR-Hyg or pFLP-Hyg plasmids. Gentamicin (30 mg/mL) was required when the strain contained pUCP32T-gfp, pUCP32T-phoP-gfp or pUCP32T-phoP^mut-gfp plasmids. Tetracycline (50 mg/mL) was required when the strain contained pME6522 or pMEO-PsRNA686 plasmids. The pACBSR-Hyg, pFLP-Hyg, and pIJ773 plasmids were gifts from Prof. Yunsong Yu (Sir Run Run Shaw Hospital of Zhejiang University School of Medicine).

Antimicrobial susceptibility testing

Fifty-four CRKP strains were isolated from clinical samples. All strains were first identified by the Vitek 2 Compact (bioMérieux, France) according to standard culture and isolation procedures, followed by storage of the strains in 25% glycerol-LB broth at −80°C, and resuscitation on LB plates before the experiment. A Vitek 2 system (bioMérieux, France) was used for antimicrobial susceptibility testing (AST). The broth microdilution method was used to determine the minimum inhibitory concentrations (MICs) of CRKP strains to PB according to the standard protocols of the Clinical and Laboratory Standards Institute (CLSI) guidelines.

Polymyxins population analysis profiles

PB (Ape$\times$Bio Tech, United States) was freshly prepared for each experiment. Mueller-Hinton broth (MHB) (Oxoid, United Kingdom) and Mueller-Hinton agar plates (MHA) (Oxoid, United Kingdom) were used for susceptibility testing. Population analysis profilings (PAP) were performed to investigate the presence of PB-HR [7]. Overnight cultures were subcultured 1:15 into MH broth for 2 h to an optical density at 600 nm of 0.65–0.7 (∼10⁹ CFU/mL). Cultures were diluted to 10⁸, 10⁷, 10⁶, 10⁵, and 10⁴ CFU/mL in physiological saline. 20 μl of diluents were plated on MHA with or without PB at the following concentrations: 2, 4, 8, 16, 32, 64, and 128 mg/L. Colonies were counted after overnight incubation at 37℃. Following CLSI M709 documents, MHA was prepared for MIC determination in agar dilution assays. The strain was considered heteroresistant when MIC was >8-fold higher than the highest noninhibitory concentration (HNIC). Escherichia coli strain ATCC 25922 was used as quality control strain.

Whole-genome sequencing and assembly based analysis

Genomic DNA was extracted from CRKP isolates using the MiniBEST kit (Takara Bio Inc.). DNA libraries were prepared using QIAseq FX DNA Library Kits (QIAGEN, Hilden, Germany) and sequenced on an Illumina NextSeq 500 platform (Illumina, San Diego, CA, United States). Adaptor, unreliable reads, and low-quality bases of raw sequencing reads were trimmed using fastp v0.21.0 [16]. More than 752 million clean reads and 1.12 Gb of sequencing data were obtained, and more than 90.42% of reads had a quality score of Q30. Paired reads were assembled using SPAdes v3.13.0 [17]. All genomes were annotated by Prokka v1.14.6 [18]. In silico multilocus sequence typing (MLST) analysis was conducted using mlst v2.16.2 (https://github.com/tseemann/mlst) based on the Pasteur scheme (https://pubmlst.org/). Antimicrobial resistance genes and virulence-associated genes were identified using ABRicate v0.8.13 (https://github.com/tseemann/abricate).

Pan-genome analysis and phylogenetic analysis

WGS draft genome of 55 strains and a reference genome (Accession number: NC009648) were included. Pan-genome analysis was performed using Roary with the annotated assemblies in GFF3 format (output of Prokka). A concatenated alignment of genes shared among ≥99% of all isolates (core genome) was aligned by Mafft 7.407 [19]. Iqtree2 was used to construct a maximum likelihood phylogenetic tree from the variable positions in these core genome alignments [20]. MLST-type, and antimicrobial resistance genes were labelled on the core genome phylogeny with iTOL (https://itol.embl.de).

RNA-sequencing

RNA-sequencing (RNA-seq) was used to screen for the differentially expressed transcripts between the native population (NP) and the heteroresistant subpopulation (HSP) for three typical HR strains (B1, D1, and D4), respectively. We performed a PAP test to obtain the NP strains and HSP strains. A single clone from M-H plates without PB was selected as the NP strain. NP strains were grown in 3 mL CAMHB at 37℃ 200 rpm until the logarithmic growth stage. A single clone from M-H plates with PB (16 mg/mL) was selected as the HSP strain. HSP strains were treated with PB (16 mg/mL) for 24 hours and continuously passaged for 3 days without PB pressure. Then the cultures were plated on MHA with PB (16 mg/L) for overnight culture. Single clones were randomly selected from PB plates and inoculated into 3 ml CAMHB at 37℃ 200 rpm until the logarithmic growth stage as described in Fig. S1. Total RNA was extracted using NEB Next® Ultra^TM Directional RNA Library Prep Kit for Illumina® (New England Biolabs, MA, USA) according to the manufacturer’s instructions. Fragmentation was performed using RT Buffer interruption within the ALFA-SEQ Directional RNALib Prep Kit. The library construction and Illumina NovaSeq 6000 system-based sequencing were performed by the Guangdong Magigene Biotechnology Co., Ltd. (Guangzhou, China). Raw sequencing reads were cleaned by fastp and then mapped to the reference genome sequence of K. pneumoniae strain HS11286 (GenBank accession number NC_016845) with HISAT2 [21]. Next, the alignments, together with the given annotation file, were passed to StringTie v2.1.4 for transcript assembly [22]. Afterward, a matrix of read counts was produced by using prepDE.py3. Finally, to calculate gene expression changes treated against control conditions, differential expression analysis was performed separately for the bacterial and the host matrices by using the edgeR R Package on Windows Platform. Genes with a false discovery rate (FDR, q-value) < 0.05 in differential expression and |log2(Fold Change) |≥1 were considered significantly different. These genes were used for subsequent analysis. The KEGG Pathways (Kyoto Encyclopedia of Genes and Genomes) and GO terms (Gene Ontology) enrichment analysis were performed with R package clusterProfiler 4.2.2 [23]. GO terms and KEGG Pathways with FDR ≤ 0.05 were screened for significant enrichment.

sRNA prediction and homologue identification

Transcript boundaries and novel transcripts such as sRNAs were identified using the Rockhopper software, which takes RNA sequencing reads output by high-throughput sequencing technology as input. New transcripts were compared with the protein library and the sRNA library (sRNAMap, sRNATarBase, SIPHT and Rfam) to select novel candidate sRNA. To obtain the start and end positions of sRNA686 transcription, Samtools v1.13 was used to measure the read depth encompassing the sRNA686 locus at each nucleotide position. The read depth was then normalized using the estimateSizeFactors function in DESeq2 before plotting. Rockhopper software predicted that new transcripts were located in intergenic regions, which were between 100 bp downstream of the 3’end of one gene and 100 bp upstream of the 5’end of the next gene. A 2780-mer long nucleotide sequence of sRNA686 and its neighbouring genes was extracted from K. pneumoniae strain HS11286 (NC_016845) to create a BLAST database. Genomic sequencing assembly sets of 3612 K. pneumoniae strains (3557 strains from NCBI RefSeq database and 55 strains from this study) were used as query sequences for blastn analysis. Orthologues identified from complete bacterial genomes with E-values of < 1×10⁻⁴, percentage identity > 90%, and query coverage values of > 90% were retained for further analysis. Mafft 7.407 was used to compare the blastn output sequences, while SNP-sites were used to quickly extract SNP information from the aligned fasta sequences. SNP statistics were used for R plotting and post-modification with Adobe Illustrator. Orthologues identified from complete bacterial genomes with E values of <1×10⁻⁴, percentage identity >90%, and query coverage values of >90% were retained for further analysis.

Construction of the overexpression strains

The plasmid pSTVA was derived from pSTV28 which can replicate stably in E. coli under the strong promoter PlacZ and was saved in our lab. The plasmid pSTVA was identical to pACBSR, except that the chloramphenicol resistance marker was replaced by an apramycin resistance marker. pSTVA skeleton was amplified by primers pSTV28-F and pSTV28-R using pSTV28 as a template. The primers used here are all listed in Table S2. The fragment of the apramycin resistance gene was amplified by primers In28-Apr-F and In28-Apr-R using pCasCure-oriT-GFP plasmid as a template. Subsequently, PCR products were ligated using a Ready-to-Use Seamless Cloning Kit (BBI Life Science, Shanghai, China) to obtain the pSTVA plasmid. To construct the sRNA686 overexpression plasmid, the whole sequence of sRNA686 containing the promoter was amplified by PCR from K13 genomic DNA using the cloning primers INA-686-F and INA-686-R, and the pSTVA plasmid was amplified by primers InA-F2 and InA-R2. The sequence of sRNA686 from K13 is identical to KPHS11286. Then, PCR products were ligated. Similarly, to construct the phoP overexpression plasmid, the whole sequence of the phoP gene was amplified by PCR from WT genomic DNA using the cloning primers INA-PhoP-F and INA-PhoP-R. After that, PCR product and EcoR I-digested pSTVA plasmid were gel extracted and then ligated. The constructed sRNA686 and phoP overexpressing plasmid was confirmed by direct sequencing and then transformed into K. pneumoniae D1, K13, and MGH78578 to generate the overexpression strain. RT-qPCR was performed to verify the overexpression of sRNA686 and phoP.

Construction of the sRNA686 and phoP knockout strains

Gene knockout of the sRNA686 and phoP was performed in K. pneumoniae D1, K13, and MGH78578 using λ Red recombinase system [24,25]. Briefly, primers containing upstream and downstream 80 bp homologous arms of the target gene were designed to amplify the apramycin resistance cassette and FRT locus with pIJ773 as a template, as shown in Table S2. pACBSR-Hyg was used to express the λ Red recombinase induced by L-arabinose. FRT-flanked apramycin resistance cassette with up-downstream homology sequence was then transformed into competent cells by electroporation, and the resistance marker was removed by pFLP-Hyg. The constructed sRNA686 and phoP knockout strains of D1, K13, and MGH78578 were confirmed by direct sequencing and RT-qPCR. After that, pSTVA was electrotransformed to obtain the corresponding knockout strains carrying the empty vector (EV).

Growth curve and time-kill assays

Growth curve analysis under PB pressure was performed. Overnight cultures were subcultured 1:15 into MH broth for 2 h to OD₆₀₀= 0.7 (∼10⁹ CFU/mL). Then 20μl of cultures was added into MHB with or without PB (4 mg/L). 200μl of diluents was added to a 96-well plate, and subsequently incubated at 37°C with gentle shaking. OD₆₀₀ was continuously monitored using BioTek Synergy H1 microplate reader. To determine the extent of microbial elimination by antimicrobial compounds, time-kill experiments were performed as previously described with a slight modification [26]. Overnight cultures were subcultured 1:15 into MH broth for 2 h to OD₆₀₀= 0.7 (∼10⁹ CFU/mL). 700μl of diluents was centrifuged and resuspended with MHB containing PB (16 mg/L). 200μl of diluents was added to 96-well plate, and subsequently incubated at 37°C with gentle shaking. OD₆₀₀ was continuously monitored using BioTek Synergy H1 microplate reader.

RT-qPCR

The PB-HSPs were collected as previously described. Cultures of NPs and HSPs were grown overnight at 37℃ in MHB medium without PB. The 200μl culture then was subcultured in a fresh MHB medium with or without PB (2 mg/L) (cell density equivalent) for the indicated time before harvesting. Isolation of total RNA and synthesis of cDNA were performed as previously described. Briefly, the total RNA of strains was extracted by Trizol Reagent (Invitrogen, Carlsbad, CA). Reverse transcription was then performed with PrimeScript RT reagent kit (TaKaRa, Dalian, China) to get cDNAs. Following that, quantitative PCR (qPCR) was performed in triplicate with each cDNA template using SYBR green Premix Pro Taq HS qPCR Kit (Accurate Biology, Changsha, China), and then on a ViiA^TM 7 Dx system (Applied Biosystems, Foster, CA). The primers for the PCR amplification of cDNA were designed using the primer3 programme (Table S2). Quantification cycle values were normalized to the gene 23S using the relative threshold cycle (2^-ΔΔCt) method.

Fluorescent reporter assays

A GFP reporter plasmid (pUCP32T-gfp) from our lab was used. The WT phoP mRNA sequence containing putative binding sites of sRNA686 was amplified by PCR from K13 WT genomic DNA, using the cloning primers described in Table S2. The sequence was then inserted into the Xba I/Nco I sites upstream of the first codon of GFP in pUCP32T-gfp plasmid to gain a pUCP32T-phoP-gfp plasmid, which was confirmed by direct sequencing. Next, this constructed plasmid was transformed into E. coli DH5a containing pSTVA-psRNA686 or pSTVA. For mutant (Mut) plasmid, pUCP32T-phoP^mut-gfp, which carried a mutated sequence in the complementary sites of sRNA686 was generated using fusion PCR. The sequence chosen for mutation was CTTTATTTAACAGCCGTTTATATTTTGCGTATATAATGAGCGT, and it was mutated to GAAATAAATTGTCGGCAAATATAAAACGCATATATTACTCGCA. After co-transformation, the cultured E. coli DH5a strains were collected by centrifugation, washed, and resuspended in 0.9% NaCl. Finally, 200 μl of the strain solution was transferred to a black polystyrene 96-well plate. The fluorescence intensity (F485/535) was measured by a BioTek Synergy H1 microplate reader (BioTek, Winooski, VT). GFP activity was expressed in arbitrary units (AU) as F485/535. 10 μl of the strain solution was also added to a slide to observe the fluorescence by Nikon ECLIPSE Ti2-U fluorescence microscope (Nikon, Tokyo, Japan).

Extraction and mass spectrometry analysis of lipid A

Lipid A profiling by mass spectrometry was performed as previously described [27]. NP and HSP strains of K13-EV, K13-PhaS⁺, and K13-ΔPhaS were obtained by PAP tests. The overnight cultures were collected and lyophilized. 10 mg of lyophilized bacteria were washed twice with a mixture of chloroform and methanol (1:2, by vol.) and then washed again with a mixture of chloroform, methanol, and water (3:2:0.25, by vol.). The bacteria were then suspended in 400 μl of a mixture of isobutyric acid and 1 M ammonium hydroxide (5:3, by vol.) and heated for 2 h at 100°C. After cooling and centrifuging at 2000 × g for 15 min, the supernatants were diluted with water (1:1, by vol.) and lyophilized overnight. The lipid extracts were washed with 400 μl of methanol twice and extracted with 100 μl of a mixture of chloroform, methanol, and water (3:1.5:0.25, by vol.). Saturated 2,5-dihydroxybenzoic acid (DHB) was prepared with chloroform/methanol/water (3:1:5:0.25) as solvent. Mass spectra were detected in negative ion mode by the UltrafleXtreme MALDI-TOF/TOF (Bruker, Germany).

β-galactosidase assays

The plasmid pME6522 containing a promoter-less lacZ reporter was used for sRNA686 promoter analysis. A 191 bp long (−113 to +78 bp) promoter fragment of sRNA686 was inserted into EcoR I/Pst I sites upstream of the lacZ reporter in pME6522 to generate pMEO-PsRNA686. All constructs were confirmed by direct sequencing. Primers used in this study are listed in Table S2. β-galactosidase assays were carried out by the Miller method [28,29]. Cultures were collected and resuspended with a Z-buffer. 500 μl suspension were added with 500 μl Z-buffer containing β-mercaptoethanol (0.28%). 100 μl chloroform and 50 μl SDS were used to permeabilize cells. Then 200 μl o-nitrophenyl-β-_D-galactoside (ONPG) was added. The reaction was stopped with 500 μl Na₂CO₃. β-galactosidase activities = 1000×(OD₄₂₀-1.75×OD₅₅₀)/OD₆₀₀×time×volume (time /min, volume /ml).

Statistical analysis

For each separate set of assays, at least three independent experiments were performed. Data is shown as mean ± SEM by using GraphPad Prism 8 (GraphPad Software, San Diego, CA). Statistical significance was determined by the Student’s t-test between two groups, with P values represented as *, P < 0.05, **, P < 0.01, and ***, P < 0.001.

Ethics statement

The study was approved by the local Research Ethics Committee for Clinical Research and Animal Trials of the First Affiliated Hospital of Sun Yat-sen University (Approval No. [2019]483).

Results

Genomic characterization of PB-HR CRKP

To investigate the HR to PB in CRKP and the genomic characterization of HR strains, we collected 54 clinical CRKP and one standard strain MGH78578 and performed WGS. All strains were resistant to carbapenem antibiotics but not to PB (Table S3). The results of PAPs indicated that 90.47% (49/54) isolates exhibited PB-HR (Figure 1 and Fig. S3). The WGS based MLST analysis showed that most of the PB-HR strains were ST11 (43/49, 87.7%), and the others belonged to four different STs [ST45 (4%), ST1 (2%), ST307 (2%), ST412 (2%), and a new ST type (2%)]. Five strains that were not PB-HR belonged to ST11. 89.7% (44/49) PB-HR clinical strains carried bla_KPC-2 gene, two of which co-harboured bla_OXA-23 or bla_OXA-66 genes. 10.2% (5/49) PB-HR strains carried bla_NDM-1 or bla_NDM-5 gene, two of which co-harboured bla_NDM and bla_OXA genes (Figure 1). All ST11 isolates carried bla_KPC-2 gene, most of ST11 isolates harboured streptomycin and spectacycin resistance genes aadA2, fosfomycin resistance genes fosA6, quinolone resistance genes qnrS1, oqxAB and tetracycline resistance genes tet(A). Mobile colistin resistance gene mcr was not found in PB-HR strains.

Figure 1.

Phylogenetic tree, PB HR, and genomic characteristics of CRKP isolates. Iqtree2 was used to construct a maximum likelihood phylogenetic tree from the variable positions in 55 CRKP core genome alignments. The existence of PB HR was labelled by a colour star. In the heatmap, blue blocks denote the presence of an antimicrobial resistance gene.

Analysis of differentially expressed genes by RNA-Seq

To explore the PB-HR mechanism, we focused on the transcriptomics changes of genes. RNA-Seq was performed on three heteroresistant resistance subpopulations (HSP) and native populations (NP) of CRKP strains (B1, D1, and D4), respectively. B1, D1 and D4 out of 49 strains demonstrated typical PB-HR. Results showed that there were 252 co-variated genes between B1-HSP and B1-NP, D1-HSP and D1-NP, D4-HSP and D4-NP (Figure 2A). The expression levels of phoP, pmrD, arnT, and arnA were significantly increased in HSP strains (Figure 2B). Functional clustering results showed that different expressions of genes enriched in histidine metabolism, lipopolysaccharide biosynthesis, cationic antimicrobial peptide (CAMP) resistance, type II secretion system secreting secreted proteins, outer membrane protein synthesis, and glycoside hydrolase (Figure 2C). The expression of phoP, pmrD, arnBCADTEF, and lipid A biosynthetic lauryl transferase lpxL in B1-HSP, D1-HSP, D4-HSP was significantly higher than those of NP strains (Table 1). Gene Ontology (GO) term analysis showed that genes were enriched in cellular components, protein-containing complexes, localization, metabolic processes, and cellular processes (Figure 2D). These significant changes in PB resistance associated genes might be the potential mechanisms behind polymyxin HR.

Figure 2.

Transcriptomic changes in HSP strains. (A) Venn diagram. Genes differently expressed between HSP and NP strains of B1, D1, and D4. (B)Volcano plot analysis of RNA-Seq results. (C) Bubble charts of functional enrichment analysis of differentially expressed genes in HSP strains. (D) GO enrichment analysis of co-upregulated and co-downregulated genes in cellular component, molecular function biological process, and their P value.

Table 1.

Common differentially expressed genes associated with CAMP observed in HSP- and NP- strains of B1, D1 and D4.

Gene	log2FC
	B1-HSP vs B1-NP	D1-HSP vs D1-NP	D4-HSP vs D4-NP
phoP	1.36	1.76	2.00
pmrD	2.71	3.59	2.98
arnF	1.56	1.55	1.55
arnT	1.97	2.56	2.37
arnA	2.59	2.77	2.34
arnD	3.01	2.7	2.36
arnC	2.69	2.37	2.18
arnB	3.31	3.14	2.81
degP	1.41	2.35	2.56
lpxL	1.22	1.60	1.40

Identification of a novel sRNA highly expressed in HSP strains by RNA-Seq

In addition to the expression changes of genes, we also focused on changes in sRNAs when analysing transcriptome results. 42 new transcripts in the intergenic region were found to demonstrate different expressions in HSP of three strains, B1, D1, and D4. Among the 42 candidates, a novel sRNA (sRNA686) significantly increased in HSP strains (Figure 3A and Figure 2B), which was confirmed by RT-qPCR (Figure 3B). In addition, a high level of sRNA686 was found in the clinically isolated mucous-type CRKP strain K13 and the standard strain MGH78578 by RT-qPCR (Figure 3B), both of which were previously shown to be the PB-HR type. sRNA686 was also overexpressed in HSP strains of other clinical CRKP (Fig. S4).

Figure 3.

sRNA686 highly expressed in HSP. (A) sRNA686 highly expressed in HSP detected by RNA-Seq. (B) The expression of sRNA686 in NP strains and their HSP strains was detected by RT-qPCR. (C) Sequence conservation of sRNA686 and its neighbouring genes among 3612 K. pneumoniae isolates. Below, the percentage of orthologues identical to the sRNA686 allele from the reference strain at each nucleotide is shown. (D) The secondary structure of sRNA686 predicted by the web tool RNAfold. Data is shown as mean ± SEM of at least three independent experiments. *, P < 0.05, **, P < 0.01, ***, P < 0.001, ns, non-significant.

To investigate the sequence conservation of sRNA686 in K. pneumoniae strains, the 3557 K. pneumoniae’s genome sequence from the NCBI RefSeq database and the 55 strains genome from this study were used for orthologues identification. The locus of sRNA686 was highly conserved (Figure 3C), suggesting that sRNA686 had a conserved function. There are sRNA686 homologs in E. coli and in other Klebsiella species (Table S4). Visual analysis of RNA-seq reads aligned to the reference genome file in Integrative Genomics Viewer (IGV) showed that sRNA686 was a ∼239 nt long transcript on the negative strand of the genome (NC_016845, from 4389232 to 4389471). The secondary structure of sRNA686 was predicted by the web tool RNAfold, showing that it contained four stable stem loops (Figure 3D).

sRNA686 enhances polymyxin HR in CRKP

The increased expression of sRNA686 in HSP strains suggests that it may be involved in PB-HR of CRKP. To further investigate, we constructed sRNA686 deficient strains (ΔsRNA686) and overexpression strains (sRNA686⁺) in PB-HR strains. Restricted to the resistance of plasmid, D1, K13 (clinical multidrug-resistant strain), and MGH78578 (a standard strain) were selected as the study object (Fig. S2). Bacterial growth curves showed that neither deletion nor overexpression of sRNA686 affected the growth of K. pneumoniae strains D1, K13, and MGH78578 (Figure 4A). Then, we explored the effects of sRNA686 on the PB-HR of K. pneumoniae by PAP analysis. The results showed that overexpression of sRNA686 could improve PB-HR of the D1, K13, and MGH78578 strains, which was manifested by the presence of a subpopulation of resistant cells with an increased MIC obtained in PAP experiments. We also determined the frequency of the resistant subpopulation. The results showed that the frequency of the resistant subpopulation was increased in K13-sRNA686⁺ and decreased in K13-ΔsRNA686 and MGH78578-ΔsRNA686 (Figure 4B). Next, we performed time-kill assays on sRNA686 overexpressed and deficient strains. Results showed that the growth curve of those WT (pSTVA) strains of D1, K13, and MGH78578 decreased initially and increased after 6 hours under 16 mg/L of PB, which was consistent with the characteristics of HR strains (Figure 4C). The regrowth of the sRNA686⁺ strains was faster than the WT (pSTVA) strains, while the deficiency of sRNA686 restricted the regrowth in the D1 strain. Moreover, we investigated the effect of sRNA686 on the growth of K. pneumoniae bacteria under PB pressure using growth curve tests. Results showed that strains overexpressing sRNA686 were slightly faster than WT (pSTVA) at the exponential phase and stationary phase under 4 mg/L PB stress (P < 0.05), while the growth of strains knocking down sRNA686 (the K13, D1 strains) was significantly limited under PB stress (P< 0.05). Meanwhile, the phoP gene knockout strain was barely able to grow. The above results suggested that sRNA686 enhanced the resistance of K. pneumoniae to PB (Figure 4D). The results of PAP, growth curve, and time killing analysis reflected that sRNA686 contributed to PB-HR in CRKP. Thus, sRNA686 was hereafter referred to as PB HR associated sRNA PhaS.

Figure 4.

sRNA686 enhanced polymyxin HR in CRKP. (A) Growth curve of the EV, sRNA686+, ΔsRNA686, and ΔphoP strains in antibiotic-free MHB medium. (B) Population analysis profiling (PAP) of PB HR in sRNA686 overexpressing and deficient strains (K13, D1, and MGH78578). phoP deficient strains were also detected. (C)The time-kill assay of the EV, sRNA686+, ΔsRNA686, and ΔphoP strains treated with 16 mg/L PB. (D) Growth curve of the EV, sRNA686+, ΔsRNA686, and ΔphoP strains under 4 mg/L PB stress. Data is shown as mean ± SEM of at least three independent experiments. *, P < 0.05, **, P < 0.01, ***, P < 0.001, ns, non-significant.

Involvement of PhaS in regulating PhoP/Q and the LPS modification

To elucidate the intricate molecular mechanism behind the enhancement of PB-HR by PhaS, we initially employed the IntaRNA software to predict the potential genetic targets for PhaS. Notably, a potential binding site for PhaS was found within the 5'-UTR region of the phoP gene, exhibiting a strong interaction energy of −15.45 kcal/mol, indicating a significant affinity between the two molecules (Figure 5A). The second structure of the 5'-UTR region of the PhoP gene was shown in Figure 5A. Then, we used RT-qPCR to detect the effect of PhaS on phoP at the transcriptional level in NP- and HSP-group of strains K13 and MGH78578, respectively. An upregulation of phoP was found in PhaS overexpressing strains, while a decrease was observed in PhaS deletion strains (Figure 5B). Meanwhile, the expressions of pmrD and arnT, two PhoP regulated genes, were obviously increased in PhaS overexpressing strains, and the decrease of pmrD and arnT were not obvious in the PhaS deletion strains. We hypothesized that PhaS directly targeted phoP, promoting the expression of both phoP and the downstream genes. To further confirm this, a green fluorescent protein (GFP) reporter system was developed. The predicted PhaS-targeted phoP sequence was inserted upstream of the gfp gene to construct pUCP32T-phoP-gfp plasmid, facilitating the investigation of PhaS's regulatory effects on phoP expression. Then the plasmid was co-transformed into E. coli DH5α with PhaS overexpressing plasmid (pSTVA-pPhaS) or pSTVA (EV) and the fluorescence intensity was measured by a microplate reader and fluorescence microscope (Figure 5C). As shown in Figure 5D, overexpression of PhaS resulted in a significant enhancement of GFP intensity when the wild-type (WT) phoP sequence was present. However, this increase was not observed when a mutant sequence of PhoP lacking the PhaS-PhoP base-pairing site was utilized, suggesting that the interaction between PhaS and the specific base-pairing site was crucial for PhaS's regulatory effect on phoP expression.

Figure 5.

PhaS directly targeted PhoP. (A) PhaS and its putative binding sequence in the 5’-UTR of phoP. The secondary structure of the phoP mRNA region spanning −104 to +3. (B) The expression levels of phoP, pmrD, and arnT were measured by RT-qPCR in PhaS⁺ and ΔPhaS strains. (C) A green fluorescent protein (GFP) reporter system was constructed to investigate the direct interactions between sRNA and its potential target. (D) The plasmid pSTVA (EV) or pSTVA-pPhaS was cotransformed with a GFP reporter plasmid (pUCP32T-gfp) containing a wild-type (WT) or mutant (Mut) sequences of PhoP mRNA into E. coli DH5a, then the fluorescence was measured by microscopy, and the intensity was detected by a BioTek Synergy H1 microplate reader and expressed in AU as F485/535. Data is shown as mean ± SEM of at least three independent experiments. *, P < 0.05. **, P < 0.01. ***, P < 0.001. ns, non-significant.

To explore the involvement of PhaS in lipid A modification, we extracted lipid A from the EV, PhaS⁺ and ΔPhaS (K13 strain) in both NP and HSP groups to perform mass spectrum analysis. Compared to NP strains (Figure 6A-C), the m/z peak distributions representing the modification of lipid A structure were observed in HSP strains as indicated in Figure 6E-F. The predominant peaks at m/z 1761, 1825 and 1841 were presented in all the samples. As previous studies [27,30] reported, the predominant peak at m/z 1841 represented the hexa-acylated lipid A molecule with hydroxylation on the fatty acyl chain and peak at m/z 1761 represented the dephosphorylated form of lipid A at m/z 1841. Mass shift, m/z 1761–1892 (Δm/z +131) and m/z 1841–1972 (Δm/z +131) represented an L-Ara4N addition to the lipid A phosphate group. A mass shift of m/z 1841–1972 (Δm/z +131) was observed in the HSP strain of EV, PhaS⁺, and ΔPhaS. The relative intensity of peak at m/z 1892 and peak at m/z 1972 were 4.89% and 4.06% in PhaS⁺-HSP, which showed a significant increase compared to those in EV-HSP strain (0.53% and 1.14%). Concurrently, the peaks at m/z 1861, 1994 and 2125 exhibited a notable elevation in PhaS⁺ strains. However, the precise molecular configuration of those lipid A species remains unexplored and has not been previously reported. It is worth noting that the relationship between m/z 1861, 1994 and 2125 increases with a step size of 131, which happens to be the molecular weight of L-Ara4N.

Figure 6.

Mass spectra of lipid A in K13-NP and K13-HSP strains of EV, PhaS⁺, and ΔPhaS. (A) MS analysis of the NP strain of K13-EV. Peaks at m/z 1761, m/z 1825, and m/z 1841 represent a base lipid A, hexa-acylated lipid A, and hydroxylation hexa-acylated lipid A, respectively. (B) MS analysis of the NP strain of K13-PhaS⁺. (C) MS analysis of the NP strain of K13-ΔPhaS. (D) MS analysis of the HSP strain of K13-EV. Mass shift, m/z 1841–1972 (Δ m/z + 131), corresponding to an L-Ara4N modification of lipid A. (E) MS analysis of the HSP strain of K13-PhaS⁺. Mass shift, m/z 1761–1892 (Δ m/z + 131) and m/z 1841–1972 (Δ m/z + 131), corresponding to an L-Ara4N modification of lipid A. (F) MS analysis of the HSP strain of K13-ΔPhaS.

Phas was activated by PhoP under PB treatment

To further explore how PhaS increased expression under PB pressure, transcription factor (TF) motifs scanning was performed on the upstream regions (500 bp) of PhaS using the MEME Suite FIMO Software. Results showed that two PhoP motifs were located at −35 bp ∼ −52 bp on the negative strand and −188 bp ∼ −171 bp on the positive strand, indicating that PhaS could be activated by PhoP (Figure 7A). To investigate whether PhaS was regulated by the PhoP/Q two-component system, we first constructed phoP overexpressing and deficient strains in D1, K13, and MGH78578. RT-qPCR results showed that PhaS expression increased in phoP overexpression strains and decreased in phoP-deficient strains (Figure 7B). We subsequently evaluated the influence of PhoP on the transcription activity of the PhaS promoter. A PhaS-lacZ reporter fusion in a promoter-less pME6522 plasmid was constructed. The plasmid pMEO-PPhaS-lacZ was then electrotransformed into K13-PhoP⁺ and K13-ΔPhoP, and the β-galactosidase activity was detected. Overexpression of phoP significantly enhanced PhaS promoter activity with 2 mg/L PB stimulation or without PB treatment, while the PhaS promoter activity was significantly weakened in ΔPhoP strain (Figure 7C). Results demonstrated that PhaS was upregulated by PhoP, thus responded to PB stimulation.

Figure 7.

PhaS was activated by PhoP. (A) Diagram of PhoP binding motif in the PhaS, pmrD, arnB, and lpxL. A consensus motif for the PhoP binding site was generated using the MEME software tool. (B) The expression level of PhaS was measured by RT-qPCR in phoP overexpressing and deletion strains. (C) β-galactosidase activity assay to explore the influence of PhoP on the promoter activity of PhaS. The K13 strains carrying the reporter plasmid pME6522-PPhaS-lacZ combined with pSTVA-phoP or pSTVA (EV) were grown with or without PB (2 mg/L), subjected to β-galactosidase activity assay. Data is shown as mean ± SEM of at least three independent experiments. *, P < 0.05. **, P < 0.01. ***, P < 0.001. ns, non-significant.

Discussion

The problem of multidrug resistance in K. pneumoniae is becoming increasingly severe, especially for the resistance rate to carbapenems [31]. Polymyxin is considered a last-resort treatment option for CRKP infections [32]. However, the emergence of HR to polymyxin in recent years has made the situation even more difficult [33]. In this study, we explored the PB-HR mechanism in CRKP and discovered a novel sRNA involved in regulating the PhoP/Q TCS and enhancing HR to polymyxin in CRKP.

In this study, we uncovered a widespread occurrence of the PB-HR phenomenon among clinical strains of CRKP, which exhibited a higher PB-HR rate, surpassing the previously reported figure of 76.62% [34]. In another multicenter study conducted in China focusing on the PB-HR phenomenon in CRKP strains, the observed frequency of colistin heteroresistance exhibited a range spanning from 2.9% to 13.9%, among which Guangzhou demonstrated the highest rate (13.9%)[35]. The extremely high PB-HR rate in this study might be because all CRKP strains in our study were resistant bacteria from Guangzhou. The increasing PB-HR rate deserved our attention. The problem of bacterial drug resistance became increasingly serious, posing significant threat to public health. It appeared that PB monotherapy may not be effective in the treatment of CRKP infection, and it was better to use combination therapy [36].

WGS analysis revealed that ST11 KPC producers were still the predominant clones (43/49, 87.7%), which was consistent with previous studies (60.7%) in China [35]. ST11 was the predominant hospital acquired CRKP clone in China due to its ability to carry drug resistance or virulence plasmids [37]. Victor I. Band et al. found that the PB-HR rate of K. pneumoniae was 8.4% in the United States, among which ST258 was the predominant clone and contained similar antibiotic resistance genes [38]. Therefore, the ST types of PB-HR strains were still associated with global pandemic strains in local areas. Notably, we detected PB-HR in a hypervirulent K. pneumoniae strain (K13) belonging to ST412. Furthermore, Li et al. identified a hypervirulent K. pneumoniae belonging to ST23 from an individual in the community, providing valuable insights into the existence and potential pathogenicity of this particular bacterial strain in the community setting [10]. The recent emergence of hypervirulent and heteroresistant bacteria will make the infection uncontrollable and cause serious consequences [39,40].

The main target of PB was the lipopolysaccharide (LPS) of the bacterial outer membrane [41]. Gram negative bacteria escaped the bactericidal effect of polymyxins mainly by replacing the phosphate groups of lipid A of LPS by the cationic 4-amino-4-deoxy-Larabinose (L-Ara4N) and/or phosphoethanolamine (PEtN) moieties, thus hindering the binding of polymyxins [42,43]. The L-Ara4N and PEtN modification of lipid A was most regulated by the PmrA/B and PhoP/Q systems [44]. The activated sensor kinases, PmrB and PhoQ, could phosphorylate PmrA and PhoP, respectively. PmrA upregulated pmrC and arnBCADTEF operon which catalyzed the synthesis and transfer of L-Ara4N to lipid A. In K. pneumoniae, PhoP could directly activate the arnBCADTEF operon and indirectly via PmrD-dependent activation of the PmrA to activate the arnBCADTEF operon. PhoP could directly activate the arnBCADTEF operon and indirectly activate the arnBCADTEF operon via PmrA in a PmrD-dependent manner in K. pneumoniae. The mutations of PmrA/B or PhoP/Q genes ultimately led to the upregulation of arnBCADTEF and the L-Ara4N modification of lipid A was observed in HR in K. pneumoniae [45], E. coli [46], Enterobacter cloacae [47] and Acinetobacter baumannii [48,49].

Our previous study found the existence of PB-HR in clinically isolated CRKP, which was presumably due to the overexpression of the PmrA/B and PhoP/Q TCS in the HSP [15]. RNA-Seq was a crucial method for probing gene expression changes and mining sRNAs [50,51]. To further explore the mechanism under the HR, we performed transcriptome analysis by RNA-Seq between HSP and NP strains. 252 differentially expressed common genes in three HSP strains were enriched in CAMP resistance pathways and TCS, among which phoP, pmrD, and arnBCADTEF were highly expressed in HSP strains. It was consistent with reported studies that these genes were upregulated in colistin HR K. pneumoniae strains [35,45].

Several studies have shown that sRNAs participated in the regulation of PB resistance by interacting with genes involved in the processes of LPS synthesis and modification. In E. coli, Sigma E dependent sRNA MicA could downregulate PhoP/Q expression and several members of the PhoP/Q regulation, thus inhibiting the LPS modification and resistance to cationic antimicrobial peptides [11,52]. sRNA MgrR was found in E. coli, P. aeruginosa, and Salmonella species, which were positively regulated by PhoP and in turn, downregulated the LPS modification gene eptB [53]. sRNA Sr006 positively regulated the expression of lipid A deacylation gene pagL in Pseudomonas aeruginosa, promoting polymyxin resistance [54].

While most of these sRNAs were extensively studied in E. coli, Salmonella, and Pseudomonas aeruginosa, there is scarce research about sRNA involving in K. pneumoniae. In this study, we discovered a novel sRNA PhaS contributed to PB-HR in CRKP through the regulation of PhoP/Q TCS. PhaS was analyzed as a ∼239 nt long transcript on the negative strand of the genome in K. pneumoniae. A similar and conservative sRNA sequence was found in various Klebsiella species and certain isolates of E. coli. This suggested that PhaS might have a similar and vital function. sRNA usually plays a regulatory role through complementary pairing with target genes [55], such as sRNA ReaL could positively regulate the pqsC gene in Pseudomonas aeruginosa [56]. The IntaRNA predicted that PhaS targeted the 5'-UTR of phoP, and its binding site was located in the −10 to −35 promoter regions. We confirmed that PhaS directly bound to PhoP by the GFP reporter system. The phoP promoter region was predicted to form a hairpin structure by itself. PhaS cloud complementarily pair with the PhoP promoter region, destabilizing the hairpin structure, exposing ribosome binding sites, and facilitating translation. The results of qPCR showed that PhaS promoted the expression of phoP, thereby upregulating pmrD and arnT. The products of arnT gene catalyzed the transfer of L-Ara4N to lipid A, and the levels of PB resistance depended on the extent of modified lipid A molecules that reduce the electronegative potential of LPS [57]. Mass spectrometry analysis showed an increase of the L-Ara4N modification of lipid A in PhaS⁺-HSP strain. Therefore, PhaS contributed to PB-HR by upregulating the expression of phoP and increasing the modification of lipid A.

PhoP/Q was a widely regulated TCS that regulated both gene expression and sRNA expression [58]. PhoP displayed a divergent transcriptional function on its recognized site [44,59]. With regards to the upregulation mechanisms of PhaS in HSP strains, a PhoP recognition site was discovered on the reversed strand upstream of the PhaS transcription start point. It was found that PhaS was upregulated in the phoP overexpression strains and β-galactosidase assays confirmed that PhaS was positively regulated by PhoP/Q, suggesting a divergent regulatory effect of PhoP.

In summary, our study disclosed a novel sRNA contributing to PB-HR in CRKP and a positive feedback regulatory pathway of sRNA PhaS-PhoP/Q TCS in CRKP (Figure 8). Under PB pressure, CRKP sensed environmental changes and activated the PhoP/Q TCS which promotes PhaS expression. PhaS also enhances phoP expression by complementarily pairing with the promoter region of phoP, thereby enhancing the expression of downstream genes and modification of lipid A. It realized the positive feedback regulation effect of CRKP under PB pressure, improved the ability of CRKP to cope with PB, and promoted the PB-HR of CRKP. Thus, our study found a novel sRNA contributing to PB-HR in CRKP and advanced the understanding of the mechanisms of HR.

Figure 8.

Schematic representation of a novel sRNA PhaS-PhoP regulatory network. PhoQ in the inner membrane sensed PB and was activated. Activated PhoQ promoted the phosphorylated state of PhoP, which subsequently stimulated the expression of sRNA PhaS. PhaS, by targeting phoP, upregulated the expression of phoP, consequently promoting the expression of pmrD and arnBCADTEF, thereby promoting the modification of lipid A and facilitating PB-HR. A positive regulatory network of PhaS-PhoP was established inCRKP. Furthermore, PB activated PhoP/Q TCS, which also could enhance the expression of downstream molecules.

Funding Statement

This work was supported by the Natural Science Foundation of Guangdong Province (Grant No. 2021A1515010423; Grant No. 2023A1515010607) and Guangzhou major science and technology special project (Grant No. 202103000026).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Author contributions

Z.Z., Y.T. and G.X. performed the majority of the experiments. S.Z., Y.C. and G.Z. also conducted the experiments. J.P. and S.C participated in the experimental design, data analysis and wrote the manuscript. J.Z., C.C. and H.B. conceived the project, designed and undertook experiments, interpreted data, and wrote the manuscript. All authors read and approved the manuscript.

Data availability statement

The genome assemblies of K. pneumoniae reported in this study have been deposited in the NCBI GenBank genomic DNA database under BioProject accession number PRJNA1050405.