The Impact of Colistin Resistance on the Activation of Innate Immunity by Lipopolysaccharide Modification

ABSTRACT

Colistin resistance is acquired by different lipopolysaccharide (LPS) modifications. We proposed to evaluate the of effect in vivo colistin resistance acquisition on the innate immune response. We used a pair of ST11 clone Klebsiella pneumoniae strains: an OXA-48, CTX-M-15 K. pneumoniae strain susceptible to colistin (CS-Kp) isolated from a urinary infection and its colistin-resistant variant (CR-Kp) from the same patient after prolonged treatment with colistin. No mutation of previously described genes for colistin resistance (pmrA, pmrB, mgrB, phoP/Q, arnA, arnC, arnT, ugdH, and crrAB) was found in the CR-Kp genome; however, LPS modifications were characterized by negative-ion matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry. The strains were cocultured with human monocytes to determine their survival after phagocytosis and induction to apoptosis. Also, monocytes were stimulated with bacterial LPS to study cytokine and immune checkpoint production. The addition of 4-amino-4-deoxy-l-arabinose (Ara4N) to lipid A of CR-Kp accounted for the colistin resistance. CR-Kp survived significantly longer inside human monocytes after being phagocytosed than did the CS-Kp strain. In addition, LPS from CR-Kp induced both higher apoptosis in monocytes and higher levels of cytokine and immune checkpoint production than LPS from CS-Kp. Our data reveal a variable impact of colistin resistance on the innate immune system, depending on the responsible mechanism. Adding Ara4N to LPS in K. pneumoniae increases bacterial survival after phagocytosis and elicits a higher inflammatory response than its colistin-susceptible counterpart.

INTRODUCTION

Lipopolysaccharide (LPS) is a glycolipid from the outer membrane of Gram-negative bacteria that comprise a lipid A domain (endotoxin), a core oligosaccharide, and a repetitive glycan polymer (O antigen) that projects above the cell surface (1). LPS is the target for colistin (polymyxin E) (2), an old antibiotic that is currently often the only therapeutic option for carbapenemase-producing and antibiotic multidrug-resistant pathogens (3, 4). The electrostatic interaction between polycationic colistin and negative charges of LPS on the surface of Gram-negative bacteria leads to bacterial cell membrane destabilization, triggering cell death (5).

The colistin resistance rate is increasing worldwide (6, 7), mainly as a consequence of LPS modifications following chromosomal mutations in the phoP/phoQ, pmrA/pmrB, and mgrB genes or by horizontal acquisition of a plasmid carrying the mcr-1 gene (8). These mutations lead to membrane modifications, mainly the addition of cationic phosphoethanolamine (PEtN) or 4-amino-l-arabinose (Ara4N) on lipid A, decreasing the negative charge on the bacterial surface and subsequently lowering colistin binding (5, 9, 10). Beyond the LPS structural modifications, a significant impact on bacterial fitness of both chromosomal (11, 12), and plasmid-borne (13, 14) mutations has been reported in several bacterial species.

LPS is one of the major pathogen-associated molecular patterns involved in the activation of the host innate immune system in the early stages of infection (15–18), and its alterations dampen immune recognition (19, 20). After LPS binds to Toll-like receptor 4 (TLR-4), monocytes and macrophages activate an inflammatory response through the NF-κB pathway leading to cytokine production (17, 21). Lipid A modifications have been suggested to trigger a lower inflammatory response by decreased TLR-4 binding; however, available data are controversial. While some studies showed that colistin-resistant strains with pEtN and Ara4N additions prompt attenuated host responses (13, 22, 23), others revealed higher inflammatory cytokine production in bacteria with these modifications (24–26). The explanation underlying these divergent results could be the use of bacterial strains with differences in other biological features such as the above-mentioned fitness (13, 14).

LPS also triggers the expression of some immune checkpoints and other factors, leading to an immunosuppressive status known as endotoxin tolerance (27, 28). Several reports point to immune checkpoints as both therapeutic targets and outcome biomarkers in cancer and infectious diseases (29). However, their modulation after LPS exposure has scarcely been explored, particularly for LPS variants of colistin-resistant microorganisms (18).

Here, we aim to assess the immunological impact of colistin resistance in terms of phagocytosis, inflammatory cytokine production, and immune checkpoint expression, using an isogenic pair (colistin-susceptible and -resistant ST11 clones) of Klebsiella pneumoniae clinical isolates.

RESULTS

Lipid A modification in CR-Kp.

Structural variations of colistin-resistant Klebsiella pneumoniae (CR-Kp) lipid A with respect to that from a colistin-susceptible strain (CS-Kp) were determined by negative matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) mass spectrometry. CS-Kp exhibited the 2 major ions of m/z 1,814 and 1,841 (Fig. 1A), corresponding to the hexa-acyl species of lipid A, with a C′-2 acyl-oxo-acyl chain A hydroxylation (-OH), as previously described (22). CR-Kp lipid A exhibited an increase of relative intensity of the m/z 1,814 peak compared with the CS-Kp. Moreover, CR-Kp had a further ion of m/z 1,971 (Fig. 1B), resulting from the addition of Ara4N by glycosylation in the C′-1 phosphate group of the m/z 1,841 structure with a polymyxin resistance ratio (PRR) of 0.29. This ion was described as one of the major causes of colistin resistance in K. pneumoniae (30). We found some mutations in the CR-Kp genome (see Table S1 in the supplemental material), highlighting wcaJ and opgE as genes previously associated with colistin resistance; however, none were in genes identified for Ara4N addition.

FIG 1.

Differential lipid A composition of CS-Kp and CR-Kp strains. Representative negative-ion MALDI-TOF mass spectra of lipid A from the colistin-susceptible (A) and colistin-resistant (B) K. pneumoniae clinical strains. Proposed chemical structures are shown of the most relevant lipid A ions, including the 4-amino-4-deoxy-l-arabinose (Ara4N) addition, explaining their colistin resistance.

CR-Kp evades host intracellular killing and enhances apoptosis.

Although the initial rates of phagocytosis (Fig. 2A) were similar, CR-Kp exhibited a significant intracellular killing evasion, staying alive longer inside the monocytes than CS-Kp (Fig. 2B). CR-Kp also induced significantly higher monocyte apoptosis, particularly total apoptosis, than did its CS-Kp counterpart (Fig. 3).

FIG 2.

The CR-Kp strain exhibited higher intracellular killing evasion ability in a phagocytosis assay. (A) Schematic representation of procedures for phagocytosis and intracellular killing assays. Human monocytes were infected with separate strains at a multiplicity of infection of 10 (10 bacteria per monocyte) for 30 min. Then, the culture was disrupted with lysozyme and washed to discard nonphagocytosed bacteria. (B) Total CFU of K. pneumoniae inside the monocyte at various time points after phagocytosis. **, P < 0.01 in Mann-Whitney U test for colistin-resistant (CR) and CS comparison. n = 4 independent biological replicates with two technical replicates each.

FIG 3.

The CR-Kp strain induced higher apoptosis in human monocytes. (A) Representative gating strategy followed in the apoptosis assay. FSC, forward scatter; SSC, side scatter. (B to D) Percentage of early (B), late (C), and total (D) apoptosis in gated CD14⁺ monocytes at various time points from coculture at a multiplicity of infection of 10 (10 bacteria per monocyte) with colistin-resistant (CR) and colistin-susceptible (CS) K. pneumoniae. **, P < 0.01 in Mann-Whitney U test for CR and CS comparison. n = 4 independent biological replicates.

CR-Kp LPS triggered increased cytokine production and immune checkpoints.

We next assessed whether the lipid A-modified LPS elicited a different immune response. We found LPS from CR-Kp induced an enhanced inflammatory response in human monocytes by higher production of tumor necrosis factor alpha (TNF-α), interleukin-6 (IL-6), and C-X-C motif chemokine ligand 10 (CXCL10) than that by CS-Kp (Fig. 4). Other cytokines, such as IL-1beta, IL-10, and interferon gamma, showed a similar pattern but without statistical significance (Fig. S1).

FIG 4.

LPS from CR-Kp induced higher inflammatory cytokine production. Levels of TNF-α (A), IL-6 (B), and CXCL10 (C) in monocyte supernatant after 24 h of stimulation with 10 ng/mL of commercial LPS (LPS, from E. coli O111:B4) or the same concentration of isolated LPS from K. pneumoniae strains. *, P < 0.05; and **, P < 0.01, in Friedman test followed by Dunn test for multiple group comparison and Wilcoxon test for paired CS and CR comparison. n = 6 independent biological replicates (monocytes from 6 different heathy donors). Dots and their line connectors indicate paired samples from the same healthy donor.

Several reports have indicated the potential role of immune checkpoints in the context of sepsis, given that LPS can induce their expression. We studied their changes in monocyte supernatant after stimulation with the purified LPS of our strains. A significantly higher overall production of galectin-9, sTim-3, and sCD86 was observed for CR-Kp LPS than for that from the susceptible variant. (Fig. 5 and Fig. S2).

FIG 5.

LPS from CR-Kp strain induced higher immune checkpoint production. Levels of soluble Tim-3 (A), soluble CD86 (B), and galectin-9 (C) in monocyte supernatant after 24 h of stimulation with 10 ng/mL of commercial LPS (LPS, from E. coli O111:B4) or the same concentration of isolated LPS from all K. pneumoniae strains. *, P < 0.05; **, P < 0.01, in Friedman test followed by Dunn test for multiple group comparison and Wilcoxon test for paired CS and CR comparison. n = 6 independent biological replicates (monocytes from 6 different heathy donors). Dots and their line connectors indicate paired samples from the same healthy donor.

DISCUSSION

The aim of this study was to decipher the impact of colistin resistance on the innate immune system caused by lipid A modification from isogenic pairs of clinical K. pneumoniae strains and their purified LPS in monocytes. Previous studies had revealed that the mcr-1-gene-harboring bacteria induced less inflammation than the wild type (13, 23), although its effect on bacterial phagocytosis was not defined. A more recent study focusing on the mcr-3 gene with combined PEtN and Ara4N additions revealed protection against phagocytosis by macrophages. Here, we found the CR-Kp strain exhibited clear resistance against intracellular killing that was considerably higher than that of its susceptible counterpart, although it also induced phagocytosis at a higher rate than its susceptible counterpart. Negative-ionization MALDI-TOF mass spectrometry revealed that CR-Kp exhibited an addition of Ara4N to hydroxylated lipid A. The PRR of our CR-Kp was 0.29, a value similar to that observed in Ara4N lipid A modification from colistin-resistant OXA-48-producing Escherichia coli that lacked mcr genes (31).

Resistance to colistin is associated with LPS modifications altering the immunogenic nature of lipid A as a result of enzyme synthesis mutations (13, 18, 25, 32). It is important to note that we did not find plasmid genes, nor mutations on the pmrB, pmrA, or mgrB genes, nor other reported mechanisms, including phoP/Q mutations in CR-Kp (33); thus, the subjacent genotype for this modification remains elusive. Dogan et al. described that the iron uptake systems by kfu and ybtS genes in colistin-resistant K. pneumoniae clones enhanced survival against the phagocytosis of neutrophils (33). Similarly, the absence of WcaJ glycosyltransferase in K. pneumoniae with the subsequent reduction of colanic acid in LPS reduced phagocytosis ability (34). Along these lines, although we performed a paired study with a pair of clinically isolated isogenic strains, a possible study limitation is that we have not previously found a genetic mechanism for Ara4N acquisition (including pmrA/B, mgrB, phoP/Q, arnA/C/T, ugdH, and crrAB genes). Among the single nucleotide polymorphisms (SNPs) and mutations found, we could highlight mutations in opgE and wcaJ genes as they have been associated with resistance to colistin and macrophage activation differences in K. pneumoniae (34, 35). However, opgE is associated with the addition of PEtN instead of Ara4N, and in the case of wcaJ, the resistance mechanism remains unknown (34, 35). The effect on the immune response in other forms of resistance acquisition should be evaluated in future studies.

The LPS molecule binds TLR-4 in innate immune cells, triggering the inflammatory response through cytokine production via the NF-κB pathway. Infection with recombinant mcr-1-expressing bacteria significantly downmodulated p38–mitogen-activated protein kinase (MAPK) activation and NF-κB nuclear translocation, leading to a decrease in the proinflammatory cytokines TNF-α, IL-12, and IL-1β and lower caspase 1 activity with respect to the colistin-susceptible strains (23). However, these CR mutants showed reduced fitness and attenuated virulence in a Galleria mellonella model (13). In contrast, our K. pneumoniae strains had a completely opposite effect characterized by higher inflammatory cytokines as well as higher apoptosis induction in the CR variant. Our K. pneumoniae results are closer to those observed in Acinetobacter baumannii, in which multiresistant mutants induced more inflammation and resisted phagocytosis more than wild-type variants (36). Moreover, modifications of lipid A in the different bacterial species could be different. The Ara4N addition in our clinical K. pneumoniae strain appears to be responsible for colistin resistance. This same modification in Burkholderia lipid A is described to enhance inflammatory responses in macrophages (26, 37), similarly to what we found in K. pneumoniae.

Immune checkpoints either activate or inhibit the immune system (29, 38, 39); their blockade is a successful treatment for lung cancer and melanoma (39, 40). Although bacterial LPS had been shown to also induce the expression of immune checkpoints (27, 28, 41) the effects of LPS modifications remained unexplored, particularly those related to colistin resistance. Our study demonstrated that colistin resistance through in vivo LPS modification induced higher galectin-9, sTim-3, and sCD86 production, probably causing detrimental effects on the host immune response. For their part, galectin-9 and sTim3 could play a role in immune system anergy, according to data indicating that these receptors are linked to secondary infection, septic shock, and poor outcomes in infectious diseases (42, 43). Therefore, the relationship between colistin resistance and immune checkpoint blockade and activation syndrome in infectious diseases should be further explored.

In conclusion, despite their isogenic background (ST11 K. pneumoniae clones), a different immune activation profile in terms of phagocytosis, intracellular killing evasion, inflammatory response, and immune checkpoint expression was demonstrated for K. pneumoniae strains. Colistin resistance acquisition in K. pneumoniae significantly delayed internal bacterial clearance by human monocytes, and their LPS induced a higher inflammatory response and immune checkpoint expression. Our data described the immunological cost of colistin resistance by our specific LPS modification, although these effects should be individually studied, taking into account each type of LPS modification in different bacterial species.

MATERIALS AND METHODS

Bacterial strains.

The set of isogenic bacterial strains and their MIC colistin values consisted of K. pneumoniae susceptible (CS-Kp, 0.5 mg/L) and resistant (CR-Kp, 32 mg/L) ST11 clone clinical strains. The CS-Kp strain was a clinical OXA-48- and CTX-M-15-producing K. pneumoniae strain obtained from the urine of a 79-year-old hospitalized woman with diabetes mellitus and stage IV chronic kidney disease who had recurrent urinary tract infections. Although the patient was successfully treated with meropenem plus amikacin plus colistin, she relapsed 2 months later with a colistin-resistant variant, the CR-Kp isolate. The CS-Kp and CR-Kp were submitted to whole-genome sequencing on an Illumina MiSeq platform (2 × 300 bp), and the sequence was deposited in BioProject under accession no. PRJNA914628. Each genome was assembled with VelvetOptimiser 2.2.5 (44) and SPAdes v3.9.0 (45). We used Prokka 1.14.6 (46) to annotate the CS-Kp assembled genome, which was selected as the reference. SNP variant calling for CR-Kp was assessed by Snippy software v4.6.0 (47). Finally, the CARD database (48) and manual inspection were used to identify colistin resistance-associated genes and mutations. The mutation list is included in Table S1 in the supplemental material. Genetic relatedness was analyzed in a previous work by multilocus sequence typing (MLST) and pulsed-field gel electrophoresis (PFGE), finding that both isolates belonged to the ST11 K. pneumoniae clone (32). In the same line, the fitness cost of the colistin resistance acquisition was previously tested, finding a relative growth rate of 0.973 and a fitness decrease of 2.66% (P value of 0.059) in CR-Kp compared with CS-Kp.

Mass spectrometry characterization of K. pneumoniae lipid A.

Lipid A from the K. pneumoniae clinical strains was extracted with the MBT Lipid Xtract kit (Bruker Daltonics, Germany), following the manufacturer’s instructions. Briefly, bacteria were grown overnight in Mueller-Hinton agar, picked up by a 1-μL inoculation loop, and mixed in 50 μL of MTB Lipid Xtract hydrolysis buffer in a low-binding tube. Then, 44 μL of the cell suspension was discarded, and the remaining 6 μL was submitted to a heating process at 90°C for 10 min. The tubes were left for 2 min with the lid open to completely evaporate the buffer. The dried pellets were washed with 50 μL of washing buffer without dissolving the pellet. The total volume of the washing buffer was discarded by pipetting. Finally, 5 μL of matrix was pipetted up and down for 15 to 20 s to resuspend the dried pellet, and 2 μL was spotted on an MTP 364 polished steel MALDI target plate. Mass spectrometry runs were performed with a matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) Autoflex maX spectrometer (Bruker Daltonics) in linear negative-ion mode. Spectra were analyzed using FlexAnalysis v.3.4. software (Bruker Daltonics). The polymyxin resistance ratio (PRR) was estimated from the acquired spectra by summing the intensities of the lipid A peaks attributable to the addition of Ara4N (m/z 1,971) and dividing this number by the intensity of the peak corresponding to native lipid A (m/z 1,814) as described by Furnis et al. (31).

LPS isolation.

The LPS was purified following a modified version of a previously described method (49). Briefly, 1.5 mL of overnight cultures at 37°C in Luria–Bertani (LB) broth (Difco, USA) of both the susceptible and resistant strains (supplemented with 6 mg/L of colistin) adjusted to an optical density at 600 nm (OD₆₀₀) of 0.5 was centrifuged at 10,600 × g for 10 min. The pellet was suspended in 200 μL of sodium dodecyl sulfate (SDS) buffer dyed with bromophenol (a solution of 4% β-mercaptoethanol, 4% SDS, and 20% glycerol in 0.1 M Tris-HCl, pH 6.8, diluted 1:1 in sterile distilled water) and then boiled for 15 min. After room tempering, 10 μL of a proteinase K solution (10 mg/L) was added and incubated at 59°C for 3 h. Afterward, 200 μL of ice-cold Tris-saturated phenol was added to each sample, with vigorous vortexing. Then, samples were incubated at 65°C for 15 min, being shaken occasionally. Again, at room temperature 1 mL of diethyl ether was added to each sample, which was then vortexed and centrifuged at 16,000 × g for 10 min. Finally, the bottom blue layer containing the LPS was extracted. The product was extracted once again, starting from the SDS step, to obtain a purified solution of LPS. Purity was assessed by SDS gel and Coomassie blue staining as previously described (50). LPS concentration was determined by a quantitative chromogenic endpoint test, the QCL-1000 assay (Lonza Walkersville, Inc., Walkersville, MD, USA).

Ethics statement.

The Ethics Committee of Ramón y Cajal University Hospital approved the epidemiological study of the multiresistant antibiotic isolates. The human blood buffy coats from healthy anonymous donors were obtained from the blood donor service of La Paz University Hospital. All the participants provided written consent in accordance with the ethical guidelines of the 1975 Declaration of Helsinki and the Committee for Human Subjects of La Paz University Hospital (PI-4100).

Peripheral blood monocyte isolation.

Healthy donors were recruited from the blood donor service of La Paz University Hospital, and fresh blood from venipuncture was collected in K2EDTA tubes (BD Vacutainer, USA). The blood was added to Ficoll-Paque Plus (Cytiva, MA, USA) according to the manufacturer’s protocol. The peripheral blood mononuclear cells were counted, and the monocyte percentage was checked by flow cytometry (BD FACSCalibur). The monocyte population was enriched by an adherence protocol after 1 h of culture in serum-free Roswell Park Memorial Institute (RPMI) liquid medium as previously described (28, 51).

Phagocytosis and host intracellular killing assay.

Phagocytosis was performed following previously described reports (51, 52). The schematic procedure is shown in Fig. 2A. Briefly, isolated human monocytes were exposed to the bacteria at a multiplicity of infection (MOI) of 10 (10 bacteria per monocyte) in RPMI medium without antibiotics for 30 min. The monocytes were washed three times with phosphate-buffered saline (PBS), treated with lysozyme (1 mg/mL) for 10 min, and washed another three times to remove the nonphagocytosed bacteria. Supernatants after these washes were seeded on LB agar plates (Difco) at 37°C for 24 h as negative controls. To determine the intracellular bacteria, 0, 1, 2, and 4 h after phagocytosis, monocytes were washed twice to discard possible extracellular bacteria and subjected to soft lysing with 0.5% Triton for 10 min. Aliquots from monocyte lysates were seeded on LB agar plates (Difco) at 37°C for 24 h for counting CFU.

Flow cytometry staining and apoptosis assay.

Monocytes resulting from the host intracellular killing assay described above were used to analyze apoptosis induction. Briefly, monocytes were washed twice with phosphate-buffered saline (PBS) after lysozyme treatment and cultivated for 1 and 4 h. Then, cells were collected and labeled with CD14 allophycocyanin (APC) (Immunostep, Spain). Apoptosis was determined employing the fluorescein isothiocyanate annexin V apoptosis detection kit (Immunostep) following the manufacturer’s protocol. Additionally, an apoptosis assay was performed on monocytes from healthy donors by stimulation for 1 and 4 h with 100 ng/mL of LPS purified from the bacterial strains.

Cytokine and immune checkpoint quantification.

The supernatants of the monocyte cultures were analyzed after stimulation with 10 ng/mL of the different purified LPSs for 24 h. Commercial E. coli O111:B4 LPS (Sigma-Aldrich, MA, USA) was used as a positive control. The inflammatory cytokines and the soluble immune checkpoint production in supernatant were quantified by a multiplex cytometric bead array (LEGENDplex HU essential immune response panel and LEGENDplex HU immune checkpoint panel 1, respectively). Samples were acquired in a BD FACSCalibur cytometer and analyzed by the LEGENDplex Data analysis software suite (Qognit, Inc., CA, USA).

Statistical analysis.

The results are expressed as mean ± standard error of the mean. The sample size (n) as well as the statistical test for each experiment is shown in the figure legends. All experiments were repeated at least twice, in different replicteas and using monocytes from at least 4 unrelated donors. Briefly, for the phagocytosis experiments, data were first compared using the Kruskal-Wallis test, followed by the Mann-Whitney U test. For cytokine and immune checkpoint production after stimulation with purified LPS from the CS and CR variants, a Friedman test and Dunn’s multiple-comparison test, as well as a Wilcoxon signed-rank test, were performed. P values of <0.05 were considered significant at a 95% confidence interval. Statistical analyses were performed with IBM SPSS 23 (Armonk, NY, USA) and GraphPad Prism 8 (San Diego, CA, USA) software.

Data availability.

Whole-genome sequences were deposited in the NCBI database under BioProject number PRJNA914628.