Interaction of Antibiotics with Innate Host Defense Factors against Salmonella enterica Serotype Newport

It is becoming increasingly understood that the current paradigms of in vitro antimicrobial susceptibility testing may have significant shortcomings in predicting activity in vivo. This study evaluated the activity of several antibiotics alone and in combination against clinical isolates of Salmonella enterica serotype Newport (meningitis case) utilizing both conventional and physiological media. In addition, the interactions of these antibiotics with components of the innate immune system were evaluated. Azithromycin, which has performed quite well clinically despite high MICs in conventional media, was shown to be more active in physiological media and to enhance innate immune system killing. Alternatively, chloramphenicol did not show enhanced immune system killing, paralleling its inferior clinical performance to other antibiotics that have been used to treat Salmonella meningitis. These findings are important additions to the building understanding of current in vitro antimicrobial assay limitations that hopefully will amount to future improvements in these assays to better predict clinical efficacy and activity in vivo.

ever, rare cases have been described in immunocompetent adults, including in one neurosurgical patient (6,7). Outcomes are generally poor, and treatment is not standardized due to the small number of reported cases (8). Nontyphoidal Salmonella meningitis in immunocompetent adults is exceedingly rare, and as a result, therapy is even less defined. Given the very complex pharmacodynamic interactions between coadministered antibiotics and endogenous immune factors, including the potential selection of antimicrobial resistance (9), we examined how different antibiotics influence Salmonella susceptibility to killing by cathelicidin LL-37, a critical immune component in meningeal infection (10,11), as well as whole blood and neutrophils.
Kill curves at 2ϫ, 4ϫ, and 8ϫ MIC were performed for CRO, CIP, and CM in both CA-MHB and RPMIϩ10% LB and for AZM in RPMIϩ10% LB for the CSF Salmonella isolate. The results in Fig. S1 in the supplemental material demonstrate bactericidal activity for all drugs except CM, which was bacteriostatic.
Checkerboard studies between CRO, CIP, and AZM are shown in Table 2, with synergy exhibited between (i) CRO and CIP in RPMIϩ10% LB, (ii) CRO and CIP in CA-MHB, and (iii) CRO and AZM in RPMIϩ10% LB based on fractional inhibitory concentration index (FICI). Additionally, time-kill assays using both standard bacteriologic CA-MHB and supplemented mammalian tissue culture medium RPMIϩ10% LB were performed to assess synergy and bactericidal activity for CIPϩCRO, and the results are shown in Fig. 1. For both CSF and blood Salmonella isolates, the combination CROϩCIP in both CA-MHB and RPMIϩ10% LB demonstrated bactericidal activity (Ͼ3-    Table 1).
We next assessed the influence of overnight pretreatment of Salmonella enterica with sub-MICs of CRO, CIP, or CM on subsequent bacterial killing by LL-37. Figure 3 demonstrates that growth in sub-MIC CRO greatly sensitizes the pathogen to killing by LL-37 in both strains (P ϭ 0.002 for the CSF isolate and P ϭ 0.006 for the blood isolate versus the paired antibiotic-untreated control). However, this phenomenon was not observed for CIP or CM, which attenuated the killing activity of LL-37. This achieve significance only with chloramphenicol against the CSF strain (P ϭ 0.04 versus control).
Antibiotic exposure and sensitization to innate immune-mediated killing. Neutrophil killing assays of CSF and blood Salmonella isolates were performed following exposure to antibiotic-free medium (untreated) or pretreatment with 1/4 MIC of CRO, 1/4 MIC of CIP, or 1/4 MIC of CM. CRO sensitizes Salmonella enterica serotype Newport to neutrophil killing, whereas CM and CIP do not (Fig. 4). Sub-MIC concentrations of CRO and CROϩCIP sensitize Salmonella to neutrophil killing (Fig. 5A) and whole-blood killing (Fig. 5B), whereas CIP alone does not. Additionally, neutrophil killing assays of the Salmonella isolates pretreated with 1/4 MIC LL-37 were performed following exposure to antibiotic-free medium (untreated) or 1ϫ MIC of CRO, 1ϫ MIC of CIP, or 1ϫ MIC of CM (conditions predetermined by pilot studies). Pretreatment with LL-37 notably sensitized the patient's Salmonella enterica serotype Newport isolates to neutrophil killing in the presence of 1ϫ MIC of CRO, 1ϫ MIC of CIP, and 1ϫ MIC of CM (Fig. 6).

DISCUSSION
Limitations of in vitro antimicrobial susceptibility testing methods as predictors of in vivo treatment efficacy are increasingly being recognized (12,13). Such limitations in assays traditionally performed in bacterial growth media stem from two basic interrelated concepts: (i) the inability of bacterial media to replicate the in vivo environment  (e.g., pH, availability of nutrients, and drug permeability) and (ii) the absence of host innate immunity factors that likely have important pharmacodynamic interactions with administered antibiotics. For example, we have previously shown the clinical importance of ␤-lactams as adjunctive therapy in treating ␤-lactam-resistant Gram-positive pathogens, such as methicillin-resistant Staphylococcus aureus (14,15) and vancomycin-  resistant Enterococcus (16), and that the macrolide antibiotic AZM may demonstrate excellent in vivo activity against some resistant Gram-negative bacteria (17,18). A recent study of Salmonella that demonstrated tissue-specific selection of antimicrobialresistant subpopulations that may serve as a reservoir of treatment failure that escapes detection by standard bacteriological media (9) provides further evidence that the paradigms of antimicrobial susceptibility testing established decades ago require a reevaluation for improvement.
While antimicrobials used to treat invasive S. enterica serovar Typhimurium infections have been examined extensively in vitro and in animal models of infection (19), the role of host immunity factors in enhancing their activities has not been examined. In this study, we found (i) consistent synergy between CRO and CIP (the agents used to successfully treat the patient) against both the CSF and blood isolates in physiologic RPMIϩ10% LB by checkerboard and time-kill assays, (ii) synergistic killing of Salmonella by combinations of sub-MICs of LL-37 and CRO, CIP, and AZM, with a most pronounced effect between CRO and LL-37, and (iii) sensitization to killing of Salmonella by human neutrophils and whole blood after growth in 1/4 MICs of CRO and CROϩCIP, but not CIP or CM. CM, a bacteriostatic agent against the Salmonella strains tested, showed no synergy with LL-37.
Due to its rarity, the treatment of Salmonella meningitis is far from standardized. In the largest assessment to date, Owusu-Ofori and Scheld describe 126 cases: most infections were due to Salmonella Typhimurium, and all but 2 occurred in children less than 2 years of age (4). In this retrospective analysis, clinical outcomes were significantly better in patients receiving 3rd-generation cephalosporins, quinolones, or imipenem compared to conventional therapies such as CM, clotrimoxazole, or ampicillin. Among the latter group, CM (the only bacteriostatic agent examined in this study) fared the most poorly in that 35 out of 54 patients (65%) relapsed or died (4). This compared to successful treatment in 83% and 89% of patients receiving 3rd-generation cephalosporins and quinolones, respectively (4). Paralleling this inferior clinical performance, the results from this study with CM showed (i) bacteriostatic activity at 2ϫ, 4ϫ, and 8ϫ MIC in both RPMIϩ10% LB and CA-MHB media and (ii) additive activity with LL-37 (FICI ϭ 1). This contrasted with CRO and CIP, which were each noted to have bactericidal activity at 2ϫ, 4ϫ, and 8ϫ MIC in both RPMIϩ10% LB and CA-MHB media and synergy in combination with LL-37 (FICI ϭ 0.5). While prior work has shown antagonism of LL-37 activity by bacteriostatic antibiotics against Escherichia coli and S. aureus (20), the lack of synergy of CM with LL-37 demonstrates another collateral benefit of bactericidal antibiotics in that they enhance bacterial killing by innate immune mechanisms. These data highlight the importance of antimicrobial activity beyond the "susceptible versus resistant" designation reported by clinical microbiology laboratories when treating serious infections such as meningitis.
The effects of various antibiotics on the sensitivity of Salmonella to killing by components of human innate immunity were of interest. Cathelicidin LL-37 is a critical host defense antimicrobial peptide in the pathogenesis of bacterial meningitis, and therefore cooperative activity between LL-37 and the antibiotics used to treat meningitis is anticipated to be an important parameter to ensure successful treatment (10). While this has not been extensively studied, we previously found that ceftaroline demonstrated this synergy with LL-37 to a much greater degree than vancomycin, paralleling its superior performance in a difficult case of Streptococcus pneumoniae meningitis (11). In the present study, we found that pretreatment of the Salmonella isolates with 1/2 MIC of CRO, CIP, or AZM enhanced killing by LL-37 but resulted in either stasis or growth alone (Fig. 2).
It is worth mentioning that AZM has been used successfully to treat enteric fever due to Salmonella (21). While the relationship of MICs measured in standard microbiological media such as CA-MHB to clinical outcome is not entirely clear, the demonstration of a severalfold reduction in AZM MIC in RPMIϩ10% LB medium suggests that further clinical-microbiological correlation studies with AZM are warranted. A recent study demonstrated that the addition of bicarbonate to standard bacteriological media approximates susceptibility testing results obtained in physiologic media (13). The same investigators demonstrated that the susceptibility results obtained using bicarbonate-buffered bacteriological media or physiologic media were more reflective of antimicrobial performance in vivo using animal models of infection (13). It appears the in vivo activity of AZM against Salmonella may be vastly underappreciated using conventional antimicrobial susceptibility methods, a theme that appears to be recurring with Gram-negative bacteria given previously published data (17,18,22). A more accurate assessment of antibiotic activity in vivo may be appreciated by considering antibiotic interactions with various components of the innate immune system, understanding variability in antibiotic activity based on bacterial inoculum, and utilizing susceptibility testing conditions that better recapitulate the host environment, including medium type and pH.

MATERIALS AND METHODS
Bacterial strains. The Salmonella enterica serotype Newport strains examined in this study were the bloodstream and cerebrospinal fluid (CSF) isolates from a patient who was successfully treated and whose lumboperitoneal shunt was salvaged without removal using CRO and CIP.
Antimicrobial susceptibility testing. CIP, CRO, and CM were purchased from Sigma-Aldrich (St. Louis, MO). AZM in 500-mg lyophilized vials was purchased from Baxter Healthcare (Deerfield, IL). Human cathelicidin LL-37 was purchased from AnaSpec, Inc. (Fremont, CA), diluted to a stock concentration of 640 M, and stored as frozen single-use 50-l aliquots. Antimicrobial susceptibility testing and checkerboard assays were performed in cation-adjusted Mueller-Hinton broth (CA-MHB) and in RPMI 1640 medium supplemented with 10% Luria-Bertani broth (RPMIϩ10% LB) according to CLSI methods using freshly thawed bacteria from the Ϫ80°C freezer plated on Todd-Hewitt agar (THA) (23). Checkerboard assays (8 by 8) were run twice, and the results were interpreted by calculating the fractional inhibitory concentration index (FICI). The FICI is calculated using the formula FICabx 1 ϩ FICabx 2 ϭ FICI, where FICabx 1 ϭ the MIC of antibiotic 1 in combination divided by the MIC of antibiotic 1 alone and FICabx 2 ϭ the MIC of antibiotic 2 in combination divided by the MIC of antibiotic 2 alone. An FICI of Ͻ0.5 defines synergy, Ͼ0.5 to Ͻ1.0 additivity, Ͼ1 to Ͻ4.0 indifference, and Ͼ4.0 antagonism (24). In vitro killing assays were performed with single drugs and in various combinations using CA-MHB or RPMIϩ10% LB with starting inocula of approximately 5 ϫ 10 6 CFU/ml, and sampling was performed at 4 and 24 h. LL-37 killing assays were performed in RPMIϩ10% LB using starting inocula of 5 ϫ 10 5 CFU/ml as previously described. Bacterial survival was assessed at 2 h in the LL-37 killing assays on antibioticexposed bacteria and at 6 h in the studies involving LL-37 combined with antibiotics. For the experiments examining the effects of antibiotics on sensitization to LL-37 killing, bacteria were grown overnight in CA-MHB, pelleted, washed once in phosphate-buffered saline (PBS), and then resuspended in fresh RPMIϩ10% LB for addition to LL-37 killing assays. Bacterial counts were enumerated on THA plates.
Neutrophil killing assays. Human neutrophils were isolated from healthy donors using the Poly-morphPrep system (Axis-Shield) under protocols approved by the University of California San Diego Human Subjects Institutional Review Board for use in established bacterial killing assays with minor Antibiotic Interaction with Innate Immunity modifications (14). Neutrophils were used to seed a 96-well plate (2 ϫ 10 5 cells/well). In experiment A, cells were infected at a multiplicity of infection (MOI) of 10 with Salmonella CSF or blood isolates that were grown overnight without antibiotic or with antibiotics (1/4 MIC of CRO, 1/4 MIC of CIP, 1/4 MIC of CM, or 1/4 MIC CRO ϩ 1/4 MIC of CIP). In experiment B, cells were infected at a multiplicity of infection (MOI) of 10 with Salmonella CSF or blood isolates grown to stationary phase overnight with 1/4 MIC of LL-37 and then exposed to neutrophils without antibiotics or with antibiotics (1ϫ MIC of CRO, 1ϫ MIC of CIP, or 1ϫ MIC of CM). After incubation for 15 min at 37°C and 5% CO 2 , the cells were lysed with 0.025% Triton X-100 and serially diluted in phosphate-buffered saline (PBS), and the total number of remaining bacteria was enumerated on Luria agar (LA) plates. The percentage of bacterial survival at 15 min was calculated as the mean Ϯ standard error of the mean (SEM) of the initial inoculum.
Whole-blood killing assay. Blood from healthy donors was collected using heparinized syringes. Salmonella CSF or blood isolates were grown to stationary phase overnight without antibiotic or with antibiotic (1/4 MIC of CRO, 1/4 MIC of CIP, 1/4 MIC of CM, or 1/4 MIC of CRO ϩ 1/4 MIC of CIP). A total of 2 ϫ 10 3 CFU of the stationary-phase bacteria were mixed with 200 l of whole human blood in siliconized tubes and incubated at 37°C with orbital rotation for 60 min, 25-l aliquots were removed, and cells were lysed with 0.025% Triton X-100, serially diluted in PBS, and plated onto LA for enumeration of surviving CFU (25,26). These studies were approved by the University of California San Diego Human Research Protections Program.