Source control and antibiotic treatment are the mainstays in the treatment of infectious disease. Interestingly, antibiotics not only directly reduce bacterial growth; they also modulate cellular host defence, especially the antimicrobial functions of monocytes and granulocytes 1. Some antibiotics enhance cellular host defence mechanisms, but others have adverse effects. How these alterations occur remains poorly understood, but cytokines and chemokines are throught to be important regulators of antimicrobial functions 2.

Granulocyte colony-stimulating factor (G-CSF) is a glycopeptide which stimulates the granulocyte lineage and stem cells. G-CSF also suppresses release of pro-inflammatory cytokines such as TNF-α by macrophages 3. G-CSF stimulates neutrophil recruitment and chemotaxis 4, increases phagocytosis of bacteria and production of free radicals 5, and retards normal granulocyte apoptosis 6. The beneficial effects of G-CSF have been demonstrated in different animal models of sepsis 789101112. In clinical trials, recombinant G-CSF (filgrastim) proved to be effective in diabetic foot ulcers 13 and high-risk febrile neutropenia 14, but not for community acquired pneumonia 15.

To evaluate the interaction of different antibiotics with G-CSF, we used clinic modelling randomised trials (CMRTs). The inclusion of an i.v. antibiotic prophylaxis is one important feature of CMRTs for modeling clinical complexity. Others include appropriate use of anaesthesia, volume loading, laparotomy, peritoneal contamination and infection with human stool bacteria (PCI), suitable postoperative analgesia, and complicating risk factors such as hypothermia and blood loss which we have assessed previously 1116. Furthermore it was validated and confirmed extensively by our group in terms of microbiologic characterization and reproducibility, as shown by dose-mortality relationship 17. In addition characteristics of randomised clinical trials such as sample size calculation, randomization, double-blind design, analysis of morbidity, intent to treat analysis and adequate statistics are included in CMRTs. CMRT characteristics are summarised in a Table [see Additional file 1].

Using this model, we have evaluated cephalosporins, aminoglycosides, fluoro-quinolones, and carbapenemes in combination with G-CSF. All antibiotics are routinely used for clinical peritonitis treatment at our institution. Interestingly, 7 antibiotic/G-CSF combinations improved survival, but 4 others did not 1217. For combination, one of the best antibiotics (coamoxiclav) and one of the worst antibiotics (Cef/met) with regard to survival were selected for further investigations analyzing PMN function and pro-inflammatory organ cytokine expression and systemic concentrations. We hypothesised that positive antibiotic/G-CSF interactions resulting in a reduced mortality rate after sepsis depend on improved PMN function and reduced pro-inflammatory organ cytokine expression and reduced systemic cytokine levels.

The study was performed with permission of the animal welfare committee in Giessen, Hessen, Germany. 179 male Wistar rats, 220–280 g (Charles River Wiga, Sulzfeld, Germany) were used. They were given standard diet (Altromin, Lage, Germany) and water ad libitum.

The cytokine G-CSF is a glycopeptide that antagonizes the neutrophil depression associated with bone marrow transplantation or sepsis 2223. Cef/met alone reduces the mortality rate in rats with peritoneal contamination and infection, but no additive effect was obtained by combining it with G-CSF. Only the combination coamoxiclav plus G-CSF was highly effective. This result is consistent with a previous report 24.

No differences were observed in the white blood cell count or PMN ratio after G-CSF prophylaxis in combination with the two antibiotics. The absolute number of the white blood cells thus does not appear to be a crucial factor explaining the efficacy of various antibiotics in combination with G-CSF.

Antimicrobial PMN functions, namely the migratory activity and superoxide production are differently regulated. In obstetric patients without antibiotic treatment reduced PMN migration was predictive for infections in cases of preterm premature rupture of membranes 25. From the literature it is known that G-CSF stimulates cellular immune defence mechanisms and alters the cytokine network. G-CSF induces the granulocyte recruitment and enhances their function at the site of infection, e.g. peritoneally. Especially the production of superoxide anions could be normalized in septic animals by G-CSF treatment 26, but these results did not explain the antibiotic differences.

Unfortunately, only very few results are available assessing the antimicrobial activity and interaction of G-CSF with different antibiotics. Beneficial effects of G-CSF in combination with fluoro-quinolones such as ofloxacin on the antimicrobial functions were reported 27. With the antibiotics alone several investigations were performed. Coamoxiclav increases the phagocytotic activity of PMNs in vitro tested with un-opsonized and opsonized Staphylococcus aureus 28. The super oxide formation was unchanged in this trial. Only very high coamoxiclav concentrations (100 mg/l) modestly suppress this function 28. In contrast, for ceftriaxone it was shown that the PMN migration induced by the peptide FMLP (formyl-methionyl-leucyl-phenylalanine) was suppressed 29. This was also the case for the phagocytosis of adherent PMNs but not for PMNs in solution 30. The respiratory burst was unchanged in this group. In vitro metrodinazole seems to have no influence to the migratory activity, phagocytosis and the respiratory burst 3132. Thus although ceftriaxone is a potent antibiotic, its effects on PMN antimicrobial functions are less impressive. The difference in interaction of G-CSF with the two antibiotics most probably results from differences in phagocyte activity, although a direct influence on the antibiotic activity can not be ruled out.

There is a close interaction of the antimicrobial PMN functions and the cytokine cascade activated during infections. In our trial, only the combination of G-CSF and coamoxiclav decreased systemic TNF-α. In the liver, the cytokine mRNA expression showed a heterogeneous picture. TNF-α was also significantly reduced by the combination coamoxiclav plus G-CSF compared to antibiotic only. This was not the case with the Cef/met, but in the group without G-CSF prophylaxis, a lower expression of TNF-α was observed after Cef/met prophylaxis. IL-1 mRNA levels were not different from control after prophylaxis with coamoxiclav plus G-CSF, but after prophylaxis with Cef/met plus G-CSF they remained still elevated. The combinations showed no difference between the two antibiotics regarding the cytokines MIP-2 (macrophage inflammatory protein-2) and IL-10. MIP-2 is a chemokine predominantly produced by mononuclear leucocytes which attracts PMNs and IL-10 is a well known anti-inflammatory cytokine. Surprisingly these two factors seem to be less important for the different antibiotic effect in combination with G-CSF, at least as measured in the liver. We can not exclude that they may play an important role in other organs or systemically.

Antibiotics have different effects on Cytokines. A recent in vitro study demonstrated that coamoxiclav reduces mRNA expression of pro-inflammatory cytokines such as TNF-α, IL-1, IL-6 from PMNs stimulated with Klebsiella pneumoniae; in contrast, expression of the anti-inflammatory cytokine IL-10 did not increase 1. In vitro work also demonstrates that TNF-α release is increased by ceftriaxone, whereas other cytokines including IL-1, IL-6, and IL-8 were unchanged 3334. TNF-α is able to enhance the undirected migration and to prime PMNs for a second stimulus 35. Grutoski et al showed that the effect of TNF-α is much more complex, since TNF-α induces a factor which suppresses migration and antimicrobial activity of PMNs in a paracrine manner 36. Elevated TNF-α levels are regarded to be adverse for the antimicrobial functions of PMNs and consequently for the outcome from major infections such as sepsis. In contrast G-CSF has anti-inflammatory properties 37 and the potential to enhance migration and the oxidative burst in septic patients 38. The antibiotic differences have also implications for the possible clinical use of G-CSF. In a recent randomised clinical trial with a G-CSF prophylaxis for improvement of the outcome of high risk patients after colorectal cancer surgery also two different antibiotics (cefuroxime/metronidazole and ofloxacin/metronidazole) were used 39. In this trial again one antibiotic (cefuroxime/metronidazole) in combination with G-CSF was more effective to reduce the complication rate and length of hospital stay and to improve quality of life. The effects from cytokines and PMN functions are differentially regulated by antibiotics.

There are substantial differences in the interaction of antibiotics with G-CSF. G-CSF was only able to improve survival in combination with coamoxiclav. Increased survival may be explained by an improvement of antimicrobial PMN functions and by an altered cytokine expression which regulate these functions. Therefore, the selection of the antibiotic for combination with G-CSF in sepsis treatment should be guided not only by the bacteria to be eliminated, but also by the effects on antimicrobial functions of PMNs and the cytokine response. The antimicrobial activity will be optimised by a positive combination of direct antibiotic effects and indirect effects of the innate immune system like the PMNs.

AB: antibiotic; Cef/met: ceftriaxone/metrodinazole; CMRT: clinic modelling randomized trial; FMLP: formyl-methionyl-leucyl-phenylalanine; G-CSF: granulocyte-colony stimulating factor; IL: interleukin; MIP: macrophage inflammatory protein 2; PCI: peritoneal contamination and infection; PMN: polymorph nuclear granulocyte; TNF-α: tumour necrosis factor alpha; WBC: white blood cell count.

The authors declare that they have no competing interest. The work was in part supported by a grant of the Deutsche Forschungsgemeinschaft (DFG).

AB designed the study, supervised the experiments and wrote the manuscript in most parts. MH performed the experiments in most parts. DS helped to improve the manuscript. WL had the study idea and gave input to the study design and experimental procedure. AT was involved in data analysis, experimentation and writing of the manuscript