High prevalence of mgrB-mediated colistin resistance among carbapenem-resistant Klebsiella pneumoniae is associated with biofilm formation, and can be overcome by colistin-EDTA combination therapy

The global prevalence of colistin-resistant Klebsiella pneumoniae (ColRkp) facilitated by chromosomal and plasmid-mediated Ara4N or PEtN-remodeled LPS alterations has steadily increased with increased colistin usage for treating carbapenem-resistant K. pneumoniae (CRkp). Our study demonstrated the rising trend of ColRkp showing extensively and pandrug-resistant characteristics among CRkp, with a prevalence of 28.5%, which was mediated by chromosomal mgrB, pmrB, or phoQ mutations (91.5%), and plasmid-mediated mcr-1.1, mcr-8.1, mcr-8.2 alone or in conjunction with R256G PmrB (8.5%). Several genetic alterations in mgrB (85.1%) with increased expressions of Ara4N-related phoPQ and pmrK were critical for establishing colistin resistance in our isolates. In this study, we discovered the significant associations between extensively drug-resistant bacteria (XDR) and pandrug-resistant bacteria (PDR) ColRkp in terms of moderate, weak or no biofilm-producing abilities, and altered expressions of virulence factors. These ColRkp would therefore be very challenging to treat, emphasizing for innovative therapy to combat these infections. Regardless of the underlying colistin-resistant mechanisms, colistin-EDTA combination therapy in this study produced potent synergistic effects in both in vitro and in vivo murine bacteremia, with no ColRkp regrowth and improved animal survival, implying the significance of colistin-EDTA combination therapy as systemic therapy for unlocking colistin resistance in ColRkp-associated bacteremia.


The abilities to produce biofilms varied between XDR and PDR ColRkp isolates. The majority
of ColRkp isolates (95.7%) produced biofilms, with 76.7% of XDR ColRkp produced strong biofilms, 11.6% had moderate biofilms, 9.3% formed weak biofilms and 2.3% developed no biofilms (Fig. 4a,b). Meanwhile, 75% of PDR ColRkp developed strong biofilms and 25% had no biofilms (Fig. 4b). Significant associations existed between XDR and PDR ColRkp isolates in terms of moderate or weak abilities to produce biofilms, which were higher in XDR ColRkp than in PDR isolates. Interestingly, no biofilm-producing ability was higher in PDR than XDR ColRkp ( Fig. 4b; p < 0.0001). It was also observed that all five colistin-susceptible K. pneumoniae (ColSkp) strains developed significantly stronger biofilms (100%) when compared to XDR and PDR ColRkp (Supplementary results Table 1; p < 0.0001).

Colistin-EDTA combination showed remarkable synergistic effects against ColRkp in vitro.
The adjuvant-EDTA was discovered to have inhibitory effects on ColRkp at concentrations of 3-24 mg/mL (Table 2). When we performed checkerboard assays on ColRkp isolates (n = 45), colistin MIC was lowered to 0.25 mg/L when given in conjunction with 12 mg/mL EDTA, and this colistin-EDTA combination displayed substantial synergistic effects (FICI ≤ 0.5) on all tested 45 ColRkp isolates (  (Fig. 6a-k). Despite bacterial regrowth after 6 h of colistin and EDTA monotherapy, colistin-EDTA combination exhibited remarkable synergistic activities in reducing > 3log 10 of bacteria starting 2 h after treatment and produced prolonged bactericidal effects with no regrowth until 24 h, in all tested XDR and PDR ColRkp isolates (n = 11), regardless of their underlying colistin resistance mechanisms (Fig. 6a-k).

Administration of colistin-EDTA combination showed potent synergistic activities in murine
ColRkp-associated bacteraemia. When compared to colistin and EDTA monotherapy, single intraperitoneal administration of colistin-EDTA combination significantly reduced bacterial burden in murine ColRkpassociated bacteraemia induced by strong biofilm-producing XDR ColRkp with inactivated mgrB by IS1-like element with the presence of evaluated virulence factors ( Fig. 7a;  www.nature.com/scientificreports/ bination therapy once a day significantly improved the survival of treated mice with ColRkp-associated bacteraemia as compared to their monotherapy ( Fig. 7b; p < 0.0001).

Discussion
Klebsiella pneumoniae is a pathogen causing severe untreatable hospital-acquired infections in immunocompromised patients owing to increasing rates of antimicrobial resistance 1 . With the rising prevalence of CRkp, colistin has emerged as a feasible therapeutic option due to paucity of effective therapeutic alternatives and restrictions in novel antibiotic development 3 . Consequently, the global prevalence of ColRkp has steadily increased as a result of expanded usage of colistin, revealing the significant threat for the emergence and spread of extensively and pandrug-resistant strains around the world 3 . In our study, a total of 165 CRkp isolates were observed to exhibit different antibiotic susceptibilities in both planktonic and biofilm environments. Among CRkp, we discovered a 28.5% prevalence of XDR and PDR ColRkp (n = 47), which increased over time from 14.9% in 2016 to 36.2% in 2021.The colistin-resistant rate in our clinical setting is comparable to India (30%) 25 and Italy (22.4%) 26 , but higher than other clinical settings in Thailand (6.6%) 27 , Nigeria (9.1%) 28 and other regions of the world 29 . Despite the lack of clinical data on colistin use in our hospital, the rising trend of ColRkp is almost certainly due to selective pressure from increased colistin use in clinical settings for increasing CRkp burden, and in poultry industry setting as a short-term colistin preventive strategy for Gram-negative bacterial infections 30,31 . This could lead to the emergence and colonization of ColRkp among patients, healthy adults, and food animals 27,30,31 , causing further circulation of colistin resistance with a higher regional colistin-resistant prevalence in Thailand. Because colistin resistance is continuously growing and varying between countries over time 3,31 , addressing the underlying colistin-resistant mechanisms has become critical to deduce. Among ESBL and carbapenemaseproducing ColRkp in this study, mgrB alteration played a significant role (85.1%), with inactivation by IS1-like, ISkpn14-like, IS3-like, IS5-like and IS1380-like elements (61.7%), point mutation (12.8%) and deletion (10.6%), which is consistent with several studies around the world [32][33][34] . Downregulated expressions of mgrB were observed in ColRkp with IS integration in promoter region and G3A mgrB, as proposed previously 32,35 . Moreover, as evidenced in previous studies, expressions of Ara4N-related and PEtN-related LPS modification genes were significantly upregulated in ColRkp isolates with different underlying colistin resistance mechanisms [35][36][37] . Because the majority of ColRkp (85.1%) had inactivated mgrB, several genetic alterations in mgrB with upregulated Ara4N-related LPS modification genes were considered to be crucial in establishing colistin resistance in our isolates, as previously demonstrated 32,35,36 . In this study, 4.25% of isolates exhibited the T157P PmrB, along with www.nature.com/scientificreports/ overexpression of PEtN-related pmrCAB operons, which has previously been proved to cause colistin resistance in K. pneumoniae 38 . Amino acid substitution-R256G PmrB with significant pmrCAB transcription, was revealed as a combined colistin-resistant mechanism in 4.25% of our isolates (n = 2) harboring mcr-1.1 or mcr-8.2 genes. Although the PROVEAN bioinformatic tool anticipated a deleterious effect of R256G PmrB on its protein function, this substitution was discovered as lineage-specific mutations in both polymyxin-susceptible as well as resistant K. pneumoniae, and it has been confirmed by others to be unrelated to colistin resistance in K. pneumoniae [39][40][41][42] . Furthermore, PhoQ E82K substitution with significant expression of phoQ was identified in 2.1% of ColRkp isolates, which is consistent with prior research 43 . This study showed an 8.5% prevalence of plasmid-mediated mcr genes, which is higher than reports described in earlier studies in Thailand (< 1 to 3.2%) [41][42][43][44][45] . Until recently, K. pneumoniae of livestock origins from different regions of Thailand have been documented to harbor plasmid-mediated mcr-8 phosphoethanolamine transferase 27,41 . Our study is the first to show the presence of mcr-8.1 and mcr-8.2 in ESBL and carbapenemaseproducing ColRkp isolates from human clinical samples in Thailand. These findings revealed that mcr-8 and its variants have been existing for a period and are widely disseminated among K. pneumoniae of both human and animal origins in this region, suggesting a growing threat of antibiotic resistance in the years 46 . The judicious use www.nature.com/scientificreports/ of colistin, as well as continuous monitoring of mcr genes transferability and stability, will tremendously help in the prevention and control of antimicrobial resistance 47 . Previous studies reported the establishment of hypervirulent ColRkp with diverse virulence characteristics 8-11 and the potential roles of mgrB, pmrAB and phoPQ in supporting bacterial virulence [19][20][21] , highlighting the importance of exploring the association between colistin resistance and other virulence factors that influence bacterial pathogenicity. Biofilm formation and diverse virulent factors including mrkD, kfu, ybtS, ompK35, ompK36, uge, wabG and luxS have been implicated in bacterial colonization, invasion, and pathogenicity within the host [12][13][14][15][16][17][18] . In this study, we discovered that majority of ColRkp produced biofilms, whereas significant associations existed between XDR and PDR ColRkp in terms of moderate, or weak biofilm-producing abilities, which were higher in XDR than in PDR isolates. Interestingly, no biofilm-producing ability was higher in PDR than XDR ColRkp. These findings suggest that not only PDR ColRkp, but XDR ColRkp may also have increased biofilm-mediated antibiotic tolerance which could enhance their abilities to resist the antibiotics effects in order www.nature.com/scientificreports/ to produce untreatable infections 48 . Additionally, all of the evaluated virulence genes were encoded in 12.8% of XDR ColRkp stains, and other virulence gene combinations were encoded in varying frequencies in other XDR and PDR ColRkp isolates. There were significant overexpression of ompK35, ompK36, kfu, uge, and luxS as well as lower expression of wabG in ColRkp compared to ColSkp, demonstrating the coexistence and altered expression of bacterial virulence factors in ColRkp. Due to their resistance to last resort colistin therapy, increased biofilm-producing abilities, coexistence and altered expression of virulence factors, these isolates would be very challenging to treat and it emphasizes the critical requirement for innovative therapy to combat these infections in healthcare settings 8,11,22 . Consistent with earlier findings, a colistin-EDTA combination demonstrated in vitro potent synergistic effects at significantly lower colistin concentrations with no bacterial regrowth, implying lower probabilities of developing colistin toxicities under this combination therapy, and inferring fewer resistance concerns after its prolonged therapy 6 . Despite previous studies highlighting colistin-EDTA combination as lock therapy in localized catheter-associated infections 6,49 , administration of this combination in our murine bacteremia model with ColRkp resulted in a reduction of bacterial burden and increased animal survival, indicating their potent www.nature.com/scientificreports/ in vivo efficacy as systemic therapy in overcoming ColRkp-associated bacteremia. The potent synergistic effects of colistin-EDTA against XDR and PDR ColRkp isolates could be attributed to EDTA ions sequestration activities, which increase bacterial outer membrane permeabilities and then sensitize as well as synergize with colistin to regain colistin efficacy of increased permeabilizations, leading to enhanced intracellular content release and bacterial death 3,50 . EDTA chelation could augment colistin's entry into bacteria to exert bactericidal effects by blocking intracellular targets of colistin-essential respiratory enzymes, thereby unlocking colistin resistance regardless of underlying colistin-resistant mechanisms 51 .
In conclusion, these data revealed a rising trend of both chromosomal and plasmid-mediated colistin resistance in K. pneumoniae isolated from Chulalongkorn Memorial Hospital, Thailand between 2016 to 2021, which was also linked to altered bacterial virulence factors. Potent synergistic effects of colistin-EDTA combination against ColRkp-associated bacteremia suggest their promising application as systemic therapy in unlocking colistin resistance of untreatable superbugs, regardless of underlying colistin-resistant mechanisms, but more clinical trials are necessary to further evaluate their clinical efficacy, tolerance, and safety. The effects of altered virulence gene expression in ColRkp will need to be investigated further to learn more about how colistinresistant bacteria modulate their pathogenicity inside the host, which will support the implementation of more effective targeted strategies to overcome and mitigate their infectivity.

Materials and methods
Bacterial isolates and antimicrobial susceptibility testing. A total of 165 CRkp clinical isolates which showed resistance to either imipenem or meropenem or both, were obtained from Chulalongkorn Memorial Hospital, Thailand during 2016 to 2021 after approved by the Institutional Review Board (IRB) of the Faculty of Medicine, Chulalongkorn University, Bangkok, Thailand. All isolates are identified by 16srRNA sequencing.
To determine planktonic susceptibilities (MIC) to colistin and EDTA, drugs were serially diluted two-fold in 96-well microtiter plates using standard broth microdilution according to criteria in EUCAST (criteria for Enterobacteriaceae for colistin only) 52 and CLSI 53 . Planktonic susceptibilities to other antibiotics including imipenem, meropenem, ceftazidime, ciprofloxacin, amikacin and fosfomycin supplemented with glucose-6-phosphate were determined by agar dilution 53 . MIC was determined as the lowest concentration that inhibited the visible growth of the bacteria. According to antibiotic susceptibilities, ColRkp were categorized to XDR (non-susceptibility to at least one agent in all but two or fewer antimicrobial categories) and PDR (non-susceptibility to all agents in all antimicrobial categories) 54 .

Characterization of colistin resistance mechanisms.
Chromosomal-mediated colistin resistance mechanisms were analyzed by targeted amplification and sequencing of mgrB, pmrA, pmrB, phoP and phoQ using previously described primers 35,38 (Supplementary Methods). The resulting nucleotide and amino acid sequences were analyzed by Basic Local Alignment Search Tool (BLAST) and multiple sequence alignment by Florence Corpet (http:// multa lin. toulh ouse. inra. fr/ mutal in/ multa lin. html) was used to compare mutations observed in ColRkp nucleotide and amino acid sequences to reference sequences of K. pneumoniae subsp. pneumoniae MGH 78578 (GenBank accession number. CP_000647.1), Insertion Sequences (ISs) were analyzed using the IS finder web site (www-is.biotoul.fr). For determination of plasmid-mediated colistin resistance, mcr-1-9 genes were screened by PCR using primers as established previously and confirmed by subsequent sequence analysis using the primers that target for amplification of mcr gene of interest 55,56 . The PROVEAN tool v.1.1.5 (http:// prove an. jcvi. org/ index. php) was used to predict the effect of amino acid substitutions on protein function 57 . PROVEAN score ≤ − 2.5 was deleterious for protein function, and a score > − 2.5 was considered to have a neutral effect on protein function.
Determination of extended spectrum β-lactamase (ESBL) and carbapenemase genes. Presence of ESBL (CTXM,TEM,OXA,SHV) and carbapenemase genes (KPC, NDM, OXA-48, IMP, VIM) conferring resistance to broad range of β-lactam and carbapenem antibiotics were analyzed among ColRkp isolates using primers as previously reported 58,59 . Determination of the expressions of LPS modification genes associated with ColRkp. By using specific primers as previously described 60,61 , Quantitative RT-PCR (qRT-PCR) was used to assess the expression levels of LPS modification genes among ColRkp isolates with various colistin-resistant mechanisms. These genes include Ara4N-related pmrK and phoPQ, connector pmrD, and PEtN-related pmrCAB that are known to be involved in establishing Ara4N-related and PEtN-related LPS modification for colistin resistance. The tested isolates were ColRkp isolates with IS1-like, G60A and deleted mgrB, T157P PmrB, E82K PhoP and combined presence of mcr-8.1 and R256G PmrB. Briefly, Monarch Total RNA Miniprep Kit (Biolabs, New England) was used to extract total RNA from tested bacterial cultures grown in Luria-Bertani broth (Merck, Darmstadt, Germany) during the mid-logarithmic growth phase. These DNase-treated purified RNA was subsequently reversetranscribed into cDNA and qRT-PCR expression assays were performed using cDNA of tested ColRkp and ColSkp clinical isolates.
Determination of in vitro biofilm-mediated colistin tolerance. Using crystal violet assay, the amounts of biofilms produced by ColRkp were determined for assessing their biofilm-producing abilities and biofilm-mediated antibiotic tolerance 62 . Briefly overnight cultures of bacteria were standardized with an OD 600 of 0.02 at 600 nm (5 × 10 7 CFU mL −1 ) and 100 μL aliquots are added in triplicate to flat-bottomed 96-well polystyrene microtiter plates (SPL Life Sciences). Plates were then incubated at 37 °C for 24 h. Adherent biofilms were fixed with crystal violet (0.1%) and stained biofilms were solubilized with 30% acetic acid. Absorbance Animal study. 6-8-week-old C57BL/6 male mice were purchased from Nomura Siam International (Pathumwan, Bangkok, Thailand) and were used in all experiments. Animals were at rest for 1 week in the animal facility before use. Animals received food and water ad libitum and were housed at a maximum of 2 mice per cage, weighed and closely monitored for any signs of distress throughout experimental periods. The animal study was conducted according to guidelines and protocols approved by the Institutional Animal Care and Use Committee of the Faculty of Medicine, Chulalongkorn University, Bangkok, Thailand, based on the National Institutes of Health (NIH), USA.
In vivo murine bacteraemia model. To establish ColRkp-associated bacteremia in vivo, the previously published murine bacteremia model through intraperitoneal inoculation was performed 69 . We used a clinical XDR ColRkp isolate with the presence of evaluated virulence factors collected from the blood of bacteremia patients. This ESBL and carbapenemase-producing ColRkp also had inactivated mgrB from IS1-like insertion between nucleotides + 71 and + 72, which was one of the most common colistin-resistant mechanisms observed in this study. Immunocompetent male C57BL/6 mice were inoculated intraperitoneally with 1 × 10 6 CFU of bacterial suspension with 5% porcine mucin (Sigma-Aldrich) and murine ColRkp-associated bacteremia was allowed to develop for 1 h as previously reported 69 . To evaluate in vivo effects of colistin, EDTA and colistin-EDTA combination, the animals with ColRkp-associated bacteremia were given a single dose of PBS (control), colistin (20 mg/kg), EDTA (40 mg/kg) and colistin-EDTA (20 mg/kg + 40 mg/kg) intraperitoneally, for a total of 4 groups with 10 animals in each group. All mice were then euthanized at 14 h post-infection. Peritoneal fluid was collected by injecting 2 mL sterile saline solution into the peritoneum, followed by gentle massage and aspiration. Peritoneal fluid samples were then serially diluted and plated on nutrient agar for counting of bacterial load. For analysis of survival in mice, the same treatment procedure was repeated once daily, and survival of mice were monitored until clinical endpoint or experimental endpoint was reached. Clinical endpoint was determined using a five-point body condition score analysing weight loss, decrease in body temperature, respiratory distress, hampered mobility, and hunched posture. Experimental endpoint was defined as 10 days post infection for mice not reaching clinical endpoint.