CXC Chemokines Exhibit Bactericidal Activity against Multidrug-Resistant Gram-Negative Pathogens

ABSTRACT The continued rise and spread of antimicrobial resistance among bacterial pathogens pose a serious challenge to global health. Countering antimicrobial-resistant pathogens requires a multifaceted effort that includes the discovery of novel therapeutic approaches. Here, we establish the capacity of the human CXC chemokines CXCL9 and CXCL10 to kill multidrug-resistant Gram-negative bacteria, including New Delhi metallo-beta-lactamase-1-producing Klebsiella pneumoniae and colistin-resistant members of the family Enterobacteriaceae that harbor the mobile colistin resistance protein MCR-1 and thus possess phosphoethanolamine-modified lipid A. Colistin-resistant K. pneumoniae isolates affected by genetic mutation of the PmrA/PmrB two-component system, a chromosomally encoded regulator of lipopolysaccharide modification, and containing 4-amino-4-deoxy-l-arabinose-modified lipid A were also found to be susceptible to chemokine-mediated antimicrobial activity. However, loss of PhoP/PhoQ autoregulatory control, caused by disruption of the gene encoding the negative regulator MgrB, limited the bactericidal effects of CXCL9 and CXCL10 in a variable, strain-specific manner. Cumulatively, these findings provide mechanistic insight into chemokine-mediated antimicrobial activity, highlight disparities amongst determinants of colistin resistance, and suggest that chemokine-mediated bactericidal effects merit additional investigation as a therapeutic avenue for treating infections caused by multidrug-resistant pathogens.

IMPORTANCE As bacterial pathogens become resistant to multiple antibiotics, the infections they cause become increasingly difficult to treat. Carbapenem antibiotics provide an essential clinical barrier against multidrug-resistant bacteria; however, the dissemination of bacterial enzymes capable of inactivating carbapenems threatens the utility of these important antibiotics. Compounding this concern is the global spread of bacteria invulnerable to colistin, a polymyxin antibiotic considered to be a last line of defense against carbapenem-resistant pathogens. As the effectiveness of existing antibiotics erodes, it is critical to develop innovative antimicrobial therapies. To this end, we demonstrate that the chemokines CXCL9 and CXCL10 kill the most concerning carbapenem-and colistin-resistant pathogens. Our findings provide a murine (EC 50 , 1.6 M) CCL28 (5,12,13). Our findings demonstrate that CXCL9 and CXCL10 kill carbapenem-resistant bacterial isolates that also maintain broad coresistance against additional classes of antibiotics. Given the increasingly limited options available to treat infections caused by MDR bacterial pathogens, this novel observation highlights the potential utility of chemokine-mediated antimicrobial activity as a foundation for developing innovative therapeutic strategies to counter a range of antibiotic-resistant pathogens.
CXC chemokines mediate antimicrobial activity against mcr-1-harboring bacteria. CST is a cationic lipopeptide that targets Gram-negative bacteria through electrostatic interactions with phosphate groups on the lipid A component of lipopolysaccharide (14). While the genetic basis of CST resistance is diverse, the principal mechanisms are analogous; chemical modification of lipid A phosphates mask their negative charge, thereby decreasing affinity between CST and the bacterial cell envelope (4). Most concerning among determinants of CST resistance is mcr-1, a horizontally transmissible, plasmid-borne gene that encodes a phosphoethanolamine (pEtN) transferase that catalyzes the addition of pEtN onto phosphate moieties of lipid A (3). MCR-1 has been shown to reduce CST susceptibility in a number of important human pathogens, including E. coli, K. pneumoniae, Pseudomonas aeruginosa, and Acinetobacter baumannii (15).
To define the capacity of CXCL9 and CXCL10 to kill mcr-1 ϩ CST-resistant bacteria, three E. coli isolates were used, FI-4531, an mcr-1 ϩ clinical isolate; J53, a laboratory reference strain; and J53 mcr-1, a J53 transconjugant carrying the mcr-1-containing plasmid from FI-4531. The CST resistance of mcr-1 ϩ isolates was confirmed using a modified Etest (Fig. 2A). To verify the presence or absence of pEtN-modified lipid A in the outer membrane of the above bacterial strains ( Fig. 2B; see Fig. S1 in the supple- , and trimethoprim-sulfamethoxazole (SXT). Bacteria were treated with 48 g/ml CXCL10 (B) or CXCL9 (C), and survival was measured by CFU determination (limit of detection, 500 CFU/ml; n.d., none detected). Data are expressed as percentages of the respective untreated-control value and represent the mean Ϯ the standard error of the mean (n ϭ 3). *, P Ͻ 0.05; **, P Ͻ 0.01 (compared to ATCC 43816). (D) CRE isolate BL12125 was treated with increasing concentrations of CXCL10. Data are expressed as percentages of the untreated-control value and represent the mean Ϯ the standard error of the mean (n ϭ 3). ***, P Ͻ 0.001 (compared to the untreated control). mental material), lipid extracts from each were prepared and analyzed by matrixassisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry (Fig. 2C). The canonical, unmodified bis-phosphorylated hexa-acylated base peak was observed in all mass spectra and was the only major ion detected for J53. In addition to the base peak, both FI-4531 and J53 mcr-1 displayed major ions corresponding to pEtN-modified lipid A. Thus, the CST resistance of mcr-1 ϩ strains was appropriately associated with pEtN-modified lipid A.
Exposure of FI-4531, J53, or J53 mcr-1 to 4.0 g/ml human CXCL10 resulted in approximately 2-log killing, as measured by CFU determination, without regard to mcr-1 status (Fig. 2D). Treatment with 12.0 g/ml CXCL10 consistently resulted in the death of the initial inoculum of each E. coli strain below the level of detection (500 CFU/ml). CXCL9 was also observed to mediate similar, marked bactericidal effects was observed in all spectra. Major ions at m/z 1,919 and 1,839, corresponding to pEtN-modified lipid A and pEtN-modified lipid A lacking the second phosphate moiety, were detected only in spectra generated from mcr-1 ϩ isolates. E. coli isolates were treated with 4 or 12 g/ml CXCL10 (D) or CXCL9 (E); K. pneumoniae strains were treated with 48 g/ml CXCL10 (F). Different chemokine concentrations were assayed on the basis of the inherent susceptibilities of these organisms. Survival was measured by CFU determination (limit of detection, 500 CFU/ml; n.d., none detected). Data are expressed as percentages of the respective untreated-control value and represent the mean Ϯ the standard error of the mean (n ϭ 3). ns, not significant.
against the E. coli strains examined (Fig. 2E). Furthermore, introduction of a mcr-1carrying recombinant plasmid into CST-susceptible K. pneumoniae ATCC 13883 did not promote resistance to chemokine-mediated antimicrobial activity (Fig. 2F), despite conferring CST resistance (MIC, 32 g/ml) (15). These data demonstrate that CXCL9 and CXCL10 are able to effectively kill mcr-1 ϩ CST-resistant pathogens that maintain pEtN-modified lipid A in their outer membrane.
Disparate chromosomal determinants of CST resistance differentially impact the bactericidal activity of CXC chemokines. Similar to pEtN, the modification of lipid A with 4-amino-4-deoxy-L-arabinose (L-Ara4N) reduces the net charge of the bacterial cell envelope, thereby promoting CST resistance. The PmrA/PmrB two-component regulatory system governs the expression of chromosomal loci that encode the enzymes responsible for L-Ara4N synthesis and incorporation into lipid A (16). Discrete, nonsynonymous mutations in pmrA or pmrB are associated with the constitutive activation of this signaling pathway and CST resistance (4). Particular mutations in PhoP/PhoQ, an upstream activator of the response regulator PmrA (17), and genetic disruption of MgrB, a negative regulator of PhoP/PhoQ signaling (18), also result in the activation of PmrA and CST-resistant phenotypes (4,19). While not horizontally transmissible, these determinants have been identified in a number of human pathogens and present an emergent clinical challenge (19).
To test the effects of chromosomal determinants of CST resistance on chemokinemediated antimicrobial activity, we examined MDR, mcr-1 negative, CST-resistant K. pneumoniae clinical isolates (Fig. 3A). Bacteria were treated with 48 g/ml CXCL10, and survival was measured by CFU determination (Fig. 3B). Curiously, CXCL10 exhibited strain-specific bactericidal activity against the isolates examined, i.e., three susceptible   (Fig. 3C). MALDI-TOF analysis demonstrated that all of the bacterial isolates examined possessed similar relative amounts of L-Ara4N-modified lipid A in their outer membrane (Fig. S2); thus, the reduction in anionic surface charge afforded by L-Ara4N modification of lipid A does not appear to explain the differences in susceptibility to chemokine-mediated antimicrobial activity. This observation is consistent with our finding that pEtN modification of lipid A also does not reduce susceptibility to CXCL9 or CXCL10. Taken together, these results support the notion that CXCL9 and CXCL10 kill bacterial pathogens via a mechanism distinct, at least in part, from those reliant upon electrostatic interaction as proposed for a number of cationic antimicrobial peptides, including CST.
We next examined the possibility that disparate chromosomal determinants conferring CST resistance account for the noted variability in CXCL10 susceptibility. To identify the specific mutations/disruptions that underlie CST resistance in the above clinical isolates, five loci (mgrB, pmrA, pmrB, phoP, and phoQ) previously reported to be associated with CST-resistant phenotypes were examined by amplicon sequencing and analysis of whole-genome sequence data (4,20). All of our CST-resistant, CXCL10susceptible K. pneumoniae isolates (BA2664, BL8800, and BA2880) possessed a nonsynonymous pmrB point mutation resulting in a Thr157Pro amino acid substitution. This particular substitution has been proposed to impact dimerization and phosphotransferase by PmrB, leading to constitutive activation of PmrA and, consequently, CST resistance (21). In contrast, CST-resistant organisms that exhibited partial or complete resistance to CXCL10 were found to possess mgrB loci disrupted by insertion sequences (MS84, IS5-like element; BA3783, IS1-like element; BU19801, ISKpn25 element) or to be missing the mgrB gene entirely (BL849) (Fig. 3D). Functional inactivation of mgrB eliminates the negative feedback loop governing PhoP/PhoQ signaling, resulting in both direct and indirect (via PmrD-dependent activation of PmrA) transcriptional activation of the biosynthetic machinery necessary for L-Ara4N modification of lipid A in K. pneumoniae (4,22). No isolates were found to have genetic aberrations previously associated with CST resistance in the pmrA, phoP, or phoQ loci.
That impairment of PhoP/PhoQ autoregulatory control curtails the antimicrobial activity of CXCL10, but direct disruption of PmrA/PmrB does not, reasonably suggests that disparate chromosomal determinants of CST resistance inequitably affect CXCL10 susceptibility. Indeed, while constitutive activation of PmrA, either directly or indirectly, would be expected to result in the incorporation of L-Ara4N-modified lipid A into the bacterial outer membrane, disruption of PhoP/PhoQ autoregulation would also be expected to influence the expression and/or activity of a number of additional enzymes and regulatory systems central to defensive alteration of the bacterial outer membrane (23). These findings highlight the underappreciated prospect that distinct chromosomal determinants of CST resistance differentially impact outer membrane physiology and thus variably affect microbial traits beyond CST resistance.
Restoration of PhoP/PhoQ autoregulatory control abolishes protection from CXCL10. To substantiate that mgrB disruption and the resultant loss of PhoP/PhoQ autoregulation account for reduced CXCL10 susceptibility, we performed genetic complementation and susceptibility determination. For these experiments, native K. pneumoniae mgrB was introduced into isolate MS84. This MDR isolate was chosen for mgrB complementation since it demonstrated reduced CXCL10 susceptibility yet was still susceptible to TET and thus amenable to screening selection.
Complementation of isolate MS84 with pACYC::mgrB restored CST susceptibility (Fig. 4A), elicited the loss of L-Ara4N-modified lipid A ( Fig. 4B and C), and markedly increased susceptibility to CXCL10 (Fig. 4D). In contrast, the empty-vector control was equivalent to the MS84 parent strain in all of the regards examined. These data were validated by using an additional set of three K. pneumoniae isolates, KKBO1, KKBO4, and KKBO4 mgrB. KKBO1 and KKBO4 are sequentially collected, K. pneumoniae carbapen-emase (KPC)-producing clinical isolates that are CST susceptible or resistant, respectively. The CST resistance of KKBO4 has previously been shown to be caused by insertional inactivation of mgrB (24). Genetic complementation with pACYC::mgrB reestablished CST susceptibility in KKBO4 (Fig. 4E) and eliminated L-Ara4N-modified lipid A (Fig. S2). Determination of chemokine susceptibility demonstrated that only K. pneumoniae isolates harboring intact mgrB alleles were fully sensitive to the bactericidal effects of CXCL10 (Fig. 4F). These observations demonstrate that diminished CXCL10 susceptibility is associated with the functional inactivation of mgrB.
A pair of additional observations can be made from the lipid A mass spectra presented in this work. In K. pneumoniae, the PhoP/PhoQ-activated loci lpxO and pagP encode lipid A-modifying enzymes responsible for the hydroxylation of 2=-linked Additional ions observed at m/z 1,840, 2,063, and 2,079 are indicative of lipid A hydroxylation, palmitoylation, or both, respectively. (D) Bacteria were treated with 48 g/ml CXCL10. Survival was measured by CFU counting (limit of detection, 500 CFU/ml; n.d., none detected). Data are expressed as percentages of the respective untreated-control value and represent the mean Ϯ the standard error of the mean (n ϭ 3). (E) CST susceptibility testing by modified E test. Images are representative of two or three separate tests. (F) Bacteria were treated with 48 g/ml CXCL10. Survival was measured by CFU counting (limit of detection, 500 CFU/ml; n.d., none detected). Data are expressed as percentages of the respective untreated control and represent the mean Ϯ the standard error of the mean (n ϭ 3).

CXC Chemokines Kill Multidrug-Resistant Pathogens
® myristate (LpxO) and the addition of palmitate (C16:0) onto the hydroxyl group of hydroxymyristate (PagP) (25). The reinforcement of hydrogen bond networks by hydroxylated lipid A and the strengthening of hydrophobic interactions by palmitoylated lipid A are proposed to fortify outer membrane barrier function and impede penetration by antimicrobial peptides. Although hydroxylation and palmitoylation of lipid A have each been reported to promote bacterial resistance to antimicrobial peptides (26,27), the presence of these PhoP/PhoQ-regulated modifications did not appear to account for CXCL10 resistance in isolates harboring disruptions in mgrB. This conclusion is evidenced by the detection of 2-hydroxylated lipid A (12/12 isolates) and palmitoylated lipid A (6/12 isolates) in both CXCL10-susceptible and -resistant K. pneumoniae strains ( Fig. 4; Fig. S2).
Further investigation is required to clarify the molecular basis of PhoP/PhoQdependent protection from the bactericidal effects of CXCL9 and CXCL10 and to determine whether this limitation can be overcome. Of interest, K. pneumoniae isolate BU19801 demonstrated mgrB disruption yet was only marginally resistant to CXCL10 (Fig. 3). This phenotype is distinct from the pronounced CXCL10 resistance exhibited by other clinical isolates harboring interrupted mgrB and may provide an opportunity to elucidate the molecular basis of protection from CXCL10.
While it is unknown whether antimicrobial chemokines act through a common mechanism, they are generally considered to target bacteria via electrostatic interactions between positively charged topological "patches" present on the chemokine and negatively charged constituents of the bacterial cell envelope, notably, the lipid A component of lipopolysaccharide in Gram-negative bacteria (7,28). This basic mechanism is analogous to those proposed for a number of cationic antimicrobial peptides, including CST, and is in agreement with many experimental results (6, 29). However, pEtN or L-Ara4N modification of lipid A phosphate moieties and consequent reduction of the bacterial surface charge were not found to specifically diminish the bactericidal effects of CXCL9 and CXCL10. These findings suggest that factors other than chargebased interactions may be critical for the bactericidal action of these CXC chemokines. This notion is supported by previous findings that CXCL10 may interact directly with the conserved bacterial component FtsX, which displays homology with CXCR3, the cognate cellular receptor of CXCL10 (30,31). Elucidation of the underlying principles of chemokine-mediated antimicrobial activity will inform the development of innovative strategies for treating infections caused by antimicrobial-resistant pathogens.

MATERIALS AND METHODS
Bacterial strains and culture conditions. The bacterial isolates used in this study and their derivatives are summarized in Table 1 (20). K. pneumoniae ATCC 43816 was used as a species-matched control for phenotypic assays. Sequence typing of the above isolates was performed and assignments were made in accordance with the Institut Pasteur K. pneumoniae multilocus sequence typing scheme (http://bigsdb.pasteur.fr/). K. pneumoniae KKBO1 and KKBO4 (24), and E. coli FI-4531 were obtained in Italy (32). E. coli J53 and K. pneumoniae ATCC 13883 were used as host backgrounds for mcr-1-containing plasmids. E. coli DH5␣ was used as an intermediate in molecular cloning. Positive control strains for antimicrobial susceptibility testing included Staphylococcus aureus ATCC 29213, Enterococcus faecalis ATCC 29212, E. coli ATCC 25922, and P. aeruginosa ATCC 27853. Unless otherwise noted, bacteria were cultured in Luria-Bertani (LB) medium at 37°C with continuous shaking. E. coli FI-4531 and J53 mcr-1 were grown in the presence of 1 g/ml CST freshly prepared. K. pneumoniae MS84 derivatives and KKBO4 mgrB, each harboring pACYC constructs, were grown in the presence of 25 g/ml TET. K. pneumoniae strains derived from ATCC 13883 and containing pBCSK constructs were grown in the presence of 50 g/ml chloramphenicol.
CST Etest. While CST resistance was established by broth microdilution, the maintenance of resistant phenotypes was confirmed with a modified Etest. Standardized inocula were prepared by diluting overnight cultures 1:35 (vol/vol) in sterile LB broth and incubating them at 37°C with shaking. When an optical density at 600 nm (OD 600 ) of 0.6 was reached, a 1.5-ml volume was removed and centrifuged at 16,000 relative centrifugal force for 1 min. The supernatant was discarded, and 7 l of wet pellet was transferred into 5 ml of 0.9% sterile saline in a 15-ml glass tube. The suspensions were adjusted to a bacterial concentration equivalent to a 1.0 McFarland turbidity standard and inoculated onto LB agar with sterile polyester-tipped swabs. Plates were allowed to dry for 10 min prior to the application of CST Etest strips (BioMérieux). Sample plates were imaged after 20 h of incubation at 37°C. K. pneumoniae Etest results not presented here are included in Fig. S3.
Chemokine treatment. Recombinant human CXCL9 and CXCL10 were purchased from PeproTech (Rocky Hill, NJ) and reconstituted at 1 mg/ml in distilled water containing 0.3% human serum albumin. Overnight bacterial cultures were used to inoculate sterile LB medium at an OD 600 of 0.1 to 0.2. Subcultures were incubated at 37°C with continuous shaking until an OD 600 of 0.6 was attained. Bacteria were then diluted to a concentration of 3 ϫ 10 6 CFU/ml in 10 mM potassium phosphate buffer (pH 7.4) supplemented with 1% trypticase soy broth and treated with the chemokine indicated or an equal volume of 0.3% human serum albumin (untreated control) in the wells of a 96-cluster plate (12). After 2 h of incubation at 37°C, serial dilutions were prepared and inoculated onto LB agar plates that were incubated overnight at room temperature prior to colony enumeration.
Lipid A isolation from whole cells. Membrane lipids were extracted, and lipid A isolated, by an optimized small-scale hot ammonium isobutyrate-based protocol (37). Briefly, bacteria were streaked onto LB agar to obtain individual colonies. From these colonies, cultures were prepared in 5 ml of LB medium. Following overnight incubation at 37°C with continuous shaking, bacteria were harvested by centrifugation (4,000 relative centrifugal force for 10 min) and pellets were treated with a 5:3 mixture of 70% (vol/vol) isobutyric acid and 1 M ammonium hydroxide (400-l total volume). Samples were then incubated at 100°C for 30 to 45 min. After centrifugation (2,000 relative centrifugal force for 15 min) to remove cell debris, the supernatants were transferred to fresh tubes, combined in a 1:1 ratio of distilled water, frozen on dry ice, and lyophilized overnight. The resulting dry pellets contained whole-cell extracts of membrane lipids.
MALDI-TOF mass spectrometry. Dry lipid extracts were washed twice with 1 ml of methanol and then resuspended in 200 l of a 2:1:0.25 chloroform-methanol-water solvent mixture. Aliquots of 1 l each of norharmane matrix (10 mg/ml in 2:1 chloroform-methanol) and then analyte were spotted directly onto stainless steel target plates in duplicate. Mass spectra were recorded in negative-ion mode with a Bruker Microflex LRF MALDI-TOF mass spectrometer (Bruker Daltonics, Billerica, MA) operated in reflectron mode. The instrument was equipped with a 337-nm nitrogen laser, and analyses were performed at 39.5% global intensity. Typically, 900 laser shots were summed to acquire each spectrum and at least three mass spectra were acquired per spot. Electrospray tuning mix (Agilent, Palo Alto, CA)  Fig. S1. K. pneumoniae lipid A mass spectra not presented here are in Fig. S2. Genetic complementation and molecular cloning. Complementation of isolate MS84 with mgrB was accomplished as previously reported for genetic complementation in KKBO4 (24). Briefly, native mgrB and flanking regions were amplified from the genomic DNA of isolate BL12125 by high-fidelity PCR with primers AvaI_mgrB_F1 (5=-GACTAGCTCGGGAACACGTTTTGAAACAAGTCGATGATTC-3=) and Eco-RI_mgrB_R1 (5=-CTAGTCGAATTCCACCACCTCAAAGAGAAGGCGTTC-3=). PCR mixtures were prepared with Phusion DNA polymerase and incubated at 98°C for 30 s prior to 32 cycles of 98°C for 10 s, 62°C for 15 s, and 72°C for 20 s, followed by a final extension at 72°C for 7 min. The mgrB amplicon was purified with the MinElute PCR purification kit (Qiagen, Hilden, Germany) and sequence verified. The mgrB insert was cloned into pACYC184 (New England Biolabs, Ipswich, MA) following double restriction digestion of each with AvaI and EcoRI-HF, and ligation with T4 DNA ligase. The resulting construct, pACYC::mgrB, was electroporated into E. coli DH5␣. Following sequence validation, pACYC::mgrB was transformed into freshly prepared, electrocompetent MS84 cells (1.8 kV, 200 ⍀, 25 F). An empty-vector control strain was also generated by transformation.
The introduction of mcr-1 from isolate FI-4531 into E. coli J53 (F Ϫ met pro Azi r [azide resistant]) was accomplished by conjugation at a donor/recipient ratio of 1:10 (38). Bacteria were incubated at 35°C for 16 h, and transconjugants were selected on Mueller-Hinton agar supplemented with CST (2 g/ml) and sodium azide (150 g/ml). Transfer of the mcr-1 resistance determinant was confirmed by PCR. For mcr-1 expression by K. pneumoniae ATCC 13883, the mcr-1 gene and its native promoter were amplified by PCR and cloned into pBCSK. The generation of ATCC 13883 harboring pBCSK::mcr-1 or the empty vector alone has been described previously (15).
Examination of chromosomal loci associated with CST resistance. PCR amplification of chromosomal loci was performed by using primer pairs and cycling conditions previously reported for phoP, phoQ, pmrA, and pmrB (21) or for mgrB (19). PCR amplicons were visualized by gel electrophoresis, and the nucleotide sequences determined were compared to those of K. pneumoniae reference strains of the same sequence type to identify mutations/disruptions. The status of phoP/phoQ, pmrA/pmrB, and mgrB maintained by the 10 MDR K. pneumoniae isolates from Pakistan was also examined by using available draft whole-genome sequences available under accession numbers MAGC00000000 and MAGE00000000-MAGM00000000 (20). The genome of K. pneumoniae HS11286 was used for comparison (39). Where appropriate, the absence of mcr variants was confirmed on assembled genomes with CLC Genomics Workbench 9.5.4 by using the reference sequences NG_050417.1 (mcr-1), NG_051171.1 (mcr-2), and KX236309 (mcr-1.2). Insertion sequences were identified with ISfinder (40). For K. pneumoniae isolate BL849, no mgrB sequence or flanking regions could be identified by either in silico PCR or read mapping.
Statistical analysis. Statistical analysis and graphing were performed with GraphPad Prism 7 software; a P value of Ͻ0.05 was considered significant. Statistically significant differences among treatment groups were determined by one-way analysis of variance with Bonferroni's multiplecomparison test or, when only two groups were experimentally assayed, an unpaired Student's t test. The EC 50 of CXCL10-mediated bactericidal activity was determined by using the sigmoidal dose-response equation of nonlinear regression and is presented as the EC 50 and the 95% confidence interval.