Objective The gut microbiota is essential for the development of the intestinal immune system. Animal models have suggested that the gut microbiota also acts as a major modulator of systemic innate immunity during sepsis. Microbiota disruption by broad-spectrum antibiotics could thus have adverse effects on cellular responsiveness towards invading pathogens. As such, the use of antibiotics may attribute to immunosuppression as seen in sepsis. We aimed to test whether disruption of the gut microbiota affects systemic innate immune responses during endotoxemia in healthy subjects.
Design In this proof-of-principle intervention trial, 16 healthy young men received either no treatment or broad-spectrum antibiotics (ciprofloxacin, vancomycin and metronidazole) for 7 days, after which all were administered lipopolysaccharide intravenously to induce a transient sepsis-like syndrome. At various time points, blood and faeces were sampled.
Results Gut microbiota diversity was significantly lowered by the antibiotic treatment in all subjects. Clinical parameters, neutrophil influx, cytokine production, coagulation activation and endothelial activation during endotoxemia were not different between antibiotic-pretreated and control individuals. Antibiotic treatment had no impact on blood leucocyte responsiveness to various Toll-like receptor ligands and clinically relevant causative agents of sepsis (Streptococcus pneumoniae, Klebsiella pneumoniae, Escherichia coli) during endotoxemia.
Conclusions These findings suggest that gut microbiota disruption by broad-spectrum antibiotics does not affect systemic innate immune responses in healthy subjects during endotoxemia in humans, disproving our hypothesis. Further research is needed to test this hypothesis in critically ill patients. These data underline the importance of translating findings in mice to humans.
Trial registration number ClinicalTrials.gov (NCT02127749; Pre-results).
- ENTERIC BACTERIAL MICROFLORA
- IMMUNE RESPONSE
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Significance of this study
What is already known on this subject?
Almost three-quarters of patients on the intensive care units receive antibiotics.
The intestinal microbiota has been shown to act as a modulator of systemic immunity during sepsis in mice.
Antibiotic microbiota disruption may thus lead to decreased immune responses in critically ill patients.
The effect of microbiota disruption on systemic immune responses during endotoxemia in humans has not been studied.
What are the new findings?
This study shows that gut microbiota disruption by broad-spectrum antibiotics does not affect systemic innate immune responses during human endotoxemia.
How might it impact on clinical practice in the foreseeable future?
The intestinal microbiota is viewed upon as a possible therapeutic target in many diseases. This study is the first to translate preclinical findings on the effect of the gut microbiota on the immune system during sepsis to a clinical setting. These data underline the importance of translating findings in mice to humans.
The intestinal microbiota fulfil many essential functions, such as digesting nutrients and stimulating development of both the intestinal and systemic immune system in mice.1 ,2 The important contribution of the intestinal microbiota in the host defence against pathogens is illustrated by the increased risk of Clostridium difficile infections after broad-spectrum antibiotics and has been confirmed by the success of faecal transplantation for recurrent C. difficile infections.3 Recently, the first case studies of successful therapeutic use of faecal microbiota transplantation in patients with therapy refractory sepsis have been reported.4 ,5 Antibiotic use is widespread in intensive care units (ICUs): the Extended Prevalence of Infection in Intensive Care (EPIC II) study, a 1-day, prospective, point prevalence study among 14 414 patients in 1265 ICUs across the world, estimated that on any given day 71% of all ICU patients receive antibiotics.6 The potential clinical relevance of the devastating effect of antibiotics on the gut microbiota has become a much-debated topic only recently, now that culturing techniques have been replaced by sequencing technologies.7 A simple course of ciprofloxacin, for example, disrupts the ecosystem rapidly and drastically, resulting in a quick drop in microbial diversity and a shift in community composition.8
It has recently been shown in murine studies that a healthy gut microbiota has a protective role during systemic inflammation and sepsis.9–14 A well-balanced microbiota provides a set of signalling compounds such as peptidoglycan to systemic innate immune cells, making them more prepared for combating invading pathogens.10–12 ,14 Mice that are born in sterile conditions or that are treated with antibiotics appear to have fewer neutrophils produced by their bone marrow, which in addition seem less capable of phagocytosing pathogens such as Streptococcus pneumoniae, Staphylococcus aureus and Escherichia coli.11–13 ,15 Alveolar macrophages too are proposed to be less capable of phagocytosing pathogens and to be more prone to an allergy-promoting phenotype in the absence of a healthy gut microbiota.10 ,14 ,16 In humans, the existence of such systemic immunomodulation by the intestinal microbiota has never been investigated. If microbiota disruption by broad-spectrum antibiotics has adverse effects on systemic cellular responsiveness towards invading pathogens, the use of antibiotics may, for example, contribute to sepsis-associated immunosuppression and the occurrence of secondary infections in the critically ill.17
It is not feasible to study the effect of the gut microbiota on the immune system in critically ill patients in a controlled setting. Therefore, in the present study we used the well-characterised model of human inﬂammation induced by intravenous injection of low-dose lipopolysaccharide (LPS) to evaluate the effect of antibiotic-induced microbiota depletion on systemic inflammation.18 ,19 Based on the murine data described above, we hypothesised that disruption of the intestinal microbiota by a course of broad-spectrum antibiotics would attenuate innate immune responses during endotoxemia in healthy subjects.
Subjects and ethics statement
Healthy, non-smoking, Caucasian male subjects (aged 18–25 years) were recruited by advertising. Screening, consisting of a questionnaire, physical examination, routine blood and urine investigation and electrocardiogram, did not reveal any abnormalities. The study was approved by the institutional ethics and research committee.
Sixteen subjects were randomised into two groups, which received either no treatment or broad-spectrum antibiotics (ciprofloxacin 500 mg twice a day, vancomycin 500 mg thrice a day and metronidazole 500 mg thrice a day orally) for 7 days in order to disrupt the intestinal microbiota. After a 36-hour washout period, all subjects were given an intravenous bolus infusion of LPS (from E. coli O113 Reference Endotoxin, CC-RE lot 3, kindly provided by Dr. Anthony Suffredini (National Institutes of Health, Bethesda, Maryland, USA)) at 2 ng/kg bodyweight.20 ,21 Temperature, blood pressure and heart rate were measured every 30 min during the first two hours after LPS challenge, later at a decreased frequency. EDTA-anticoagulated and heparin-anticoagulated venous blood was sampled before antibiotics (day 0) and on the study day (day 9) before LPS infusion (t=0) and at t=0.5, 1, 1.5, 2, 3, 4, 6 and 8 hours after LPS infusion. EDTA-anticoagulated plasma was collected by centrifugation at 1750×g for 10 min at 15°C and stored at −20°C until assays were performed. Heparin-anticoagulated whole blood was stimulated directly ex vivo for 4 hours with various Toll-like receptor (TLR) agonists (LPS 100 ng/mL (from E. coli O111:B4, ultrapure)), PAM3CSK4 1 μg/mL or flagellin 0.1 μg/mL (from Salmonella typhimurium, ultrapure; all from Invivogen, San Diego, California, USA) and clinically relevant bacterial pathogens (heat-killed S. pneumoniae, Klebsiella pneumoniae or E. coli; all 1×108 CFU/mL).
Faecal samples were collected at home and stored at −20°C and within 24 hours samples transported to the study centre for storage at −80°C. DNA was extracted using a bead-beating protocol. Analysis of microbiota diversity and composition was performed by Illumina Miseq sequencing of 16S rRNA genes from extracted faecal bacterial DNA with primers 27F-DegS and 338R as described before,22 followed by analysis using the QIIME software package as described.7 ,23 All samples were analysed in one sequence run. Principal component analysis was performed in an unconstrained way using the Canoco 5 software package (Biometris, Wageningen, the Netherlands).
Tumour necrosis factor (TNF)-α, interleukin (IL)-1β, IL-6, IL-8, IL-10 and IL-12p70 were measured by cytometric bead array (BD Bioscience, San Jose, California, USA) using a FACS Calibur (BD Biosciences, Mountain View, California, USA). E-selectin was measured by ELISA (R&D systems, Minneapolis, Minnesota, USA). Myeloperoxidase (MPO), D-dimer, plasminogen activator inhibitor-1 (PAI-1), tissue-type plasminogen activator (tPA), soluble intercellular adhesion molecule (ICAM)-1 and soluble vascular cell adhesion molecule (VCAM)-1 were measured by Luminex multiplex assay (Affymetrix eBioscience, Santa Clara, CA) using a BioPlex 200 (BioRad, Hercules, California, USA).
All values are presented as mean±SEM. Differences between groups were analysed using a repeated measurements two-way analysis of variance, using GraphPad Prism 5 (GraphPad Software, San Diego, California, USA). A p value of <0.05 was considered to represent statistically significant difference. The number of eight per group is based on previous studies using the human endotoxemia model, from which we know that these sample sizes result in statistically significant differences if relevant changes are to be found.20 ,24 ,25
A detailed description of subjects, stimulations and microbiota analysis can be found in the online supplementary material.
Antibiotic perturbation of the microbiota
The effect of the 7-day course of ciprofloxacin, vancomycin and metronidazole on the intestinal microbiota was profound and resulted in a loss of diversity and a shift in faecal community composition, as determined by sequencing of bacterial 16S rRNA genes. Administration of broad-spectrum antibiotics disrupted the intestinal microbiota in all treated subjects, lowering the median diversity expressed as Shannon index from 4.9 to 1.8, whereas the microbiota of control subjects remained the same (figure 1A). Further analysis of the bacterial communities by principal component analysis confirmed this and revealed all samples from non-antibiotic-treated subjects to cluster together while those of the antibiotic-treated subjects formed a clearly separate cluster (figure 1B). In the microbiota of antibiotic-treated subjects, 23 bacterial groups were differentially present compared with before treatment (figure 1C). Several relevant bacterial groups, such as Bifidobacterium and Roseburia, were significantly and over 100-fold reduced by the antibiotic treatment. Moreover, some taxa were induced by the antibiotic treatment, including several Streptococci and Lactobacilli, which are often endogenously resistant to vancomycin.
LPS injection induced symptoms consisting of fever, chills, headache and/or nausea in all subjects. A temporary rise in temperature, heart rate and blood pressure was observed in all subjects (figure 2; blood pressure: data not shown). Body temperature peaked at 3 hours post LPS infusion (38.1±0.3°C in control subjects vs 38.0±0.2°C in antibiotic-treated subjects); heart rate peaked after 4 hours (86±3 bpm in control subjects vs 91±4 bpm in antibiotic-treated subjects). Microbiota perturbation by antibiotics did not influence any of these signs and symptoms during endotoxemia (p=0.35 and 0.97, respectively).
Endotoxemia reproducibly induces neutrophilia, enabling us to test the first of our subhypotheses, namely that antibiotic gut microbiota disruption diminishes the number of neutrophils that are released from the bone marrow into the systemic circulation. At 4 hours post LPS infusion, a significant increase in peripheral blood leucocyte counts was observed (figure 3A, both groups p<0.0001 vs baseline). Peak values were 7.5±0.5×109/L in controls vs 9.2±0.9×109/L in antibiotic-treated subjects. The increase in leucocyte counts was mainly caused by a rise in neutrophils in both groups (figure 3B, 6.5±0.4×109/L in control subjects vs 8.4±0.9×109/L in antibiotic-treated subjects). LPS administration leads to degranulation of neutrophils, as reflected by increased plasma concentrations of MPO (figure 3C). In both groups, MPO levels increased steadily up to 8 hours post LPS infusion (p<0.0001 vs baseline). Overall, no effect of pretreatment with broad-spectrum antibiotics was seen on any of these neutrophil-related parameters (all p≥0.13).
To further study the systemic innate immune response upon LPS injection, cytokine release was measured at the protein level at various time points post challenge. As expected, LPS administration was followed by a transient rise in plasma cytokine concentrations (figure 4). An early TNF-α peak (90 min post LPS infusion) and subsequent peaks for IL-6, IL-8 and IL-10 (2 or 3 hours post infusion) were detected (all p<0.0001 vs baseline). Throughout the experiment, antibiotic treatment did not influence the plasma concentrations of proinflammatory TNF-α, IL-6 or IL-8 (all p≥0.61) or anti-inflammatory IL-10 (p=0.60).
Coagulation and fibrinolysis
Activation of the coagulation and fibrinolytic system is another hallmark sign of human endotoxemia.20 All subjects demonstrated evidence of activation and inhibition of fibrinolysis during endotoxemia, as reflected by elevated plasma concentrations of tPA (figure 5A; peak 13 421±2514 pg/mL in control subjects vs 18 137±2319 pg/mL in antibiotic-treated subjects) and PAI-1 (figure 5B; peak 3368±467 pg/mL in control subjects vs 3930±1108 pg/mL in antibiotic-treated subjects). In addition, plasma levels of D-dimer were strongly elevated upon LPS injection with maximum values seen 8 hours post challenge (figure 5C; p<0.0001 compared with baseline for both groups). However, overall no significant effect of antibiotic microbiota disruption was seen on tPA (p=0.43), PAI-1 (p=0.93) or D-dimer (p=0.88) concentrations.
To assess the influence of gut microbiota disruption on endothelial cell activation during inflammation, we measured plasma levels of soluble ICAM-1, VCAM-1 and E-selectin (figure 6). As anticipated, LPS administration elicited profound rises in plasma levels of these proteins (all p<0.0001 compared with baseline), which is in correspondence with earlier reports.20 ,26 Maximum values were reached 6–8 hours after LPS injection (figure 6). However, none of these markers for endothelial activation were different between antibiotic-pretreated subjects and controls (all p≥0.14).
Lastly, LPS infusion leads to tolerance upon ex vivo restimulation of blood with LPS or other TLR ligands.27 ,28 This phenomenon, also called LPS tolerance or immunosuppression, is seen in patients with sepsis, burns and trauma and is strongly associated with poor outcome.29 ,30 In view of the literature, the absence of signals from a healthy gut microbiota might contribute to LPS tolerance, or worsen it. In order to investigate this, we tested the capacity of whole blood, sampled at t=0 (pre-LPS infusion) and t=4 (post-LPS infusion) to release pro-inflammatory and anti-inflammatory cytokines upon stimulation with different TLR ligands and common causative agents of sepsis (S. pneumoniae, K. pneumoniae and E. coli). Whole blood obtained 4 hours after LPS injection released less TNF-α and IL-1β than blood drawn before LPS infection upon ex vivo stimulation with LPS (figure 7). Stimulation with TLR2 or TLR5 ligands (see online supplementary figure S1) or bacteria (S. pneumoniae, K. pneumoniae or E. coli; online supplementary figure S1) gave parallel results. In concordance with the in vivo findings, no effect of microbiota disruption was found on the induction of LPS-tolerance or cross-tolerance during human endotoxemia.
In this study, we demonstrate that antibiotic microbiota disruption does not affect systemic innate immune responses during endotoxemia in healthy adults. No effect of antibiotic pretreatment was observed on clinical signs and symptoms, neutrophil recruitment, cytokine production, induction of coagulation and fibrinolysis, endothelial activation and LPS tolerance upon LPS injection. To the best of our knowledge, this study is the first to investigate whether the intestinal microbiota influences the innate immune system in a human model of systemic inflammation.
A limited number of studies have investigated the effect of a course of antibiotics on the composition of the gut microbiota. A simple course of ciprofloxacin or treatment with metronidazole/clarithromycin for eradication of Helicobacter pylori will result in a rapid drop of microbiota diversity levels and large shifts in bacterial communities.8 ,31 ,32 In general, microbiota diversity levels will recover within weeks, although with different bacterial compositions than before antibiotic treatment.8 ,31 ,32 Our results support these findings, with drastically lower diversity levels and a decrease in absolute numbers of bacteria considered to be beneficial such as Bifidobacteria, abundantly found in early-life microbiota and used in probiotic formulations, and Roseburia, known to include butyrate-producing taxa.33 However, the clinical relevance of these marked shifts following antibiotic treatment remains to be determined. A recent study among 142 Finnish children showed a clear distinction in microbial composition after recurrent use of antibiotics that was associated with higher risk of obesity and asthma.7 This study showed a differential effect of antibiotics with macrolides being more pervasive than penicillins. This was also observed in a meta-analysis where some but not all antibiotics were found to be associated with IBD.34 Another prospective study among all Danish singleton children born from 1995 to 2003 showed a strong association between antibiotic use and incidence of Crohn's disease.35 Considering the fact that humans have evolved in mutual dependency with their intestinal bacterial ecosystem, it is hard to imagine destruction of this ecosystem by antibiotics not having effects on human physiology. Murine experiments support this idea, showing, for example, that antibiotic treatment of pregnant dams leads to decreased bacterial diversity in their offspring.12 In a granulocyte-colony stimulating factor-dependent way, these mice displayed granulocytopenia with subsequently an increased susceptibility to sepsis induced by E. coli or K. pneumoniae.12 The present study is the first to test the proposed effect of the intestinal microbiota on the systemic innate immune system in humans and to not only look at the effect of broad-spectrum antibiotics on the intestinal microbiota but also investigate whether this has any consequences for human physiology.
Our findings in humans are not in line with previous murine studies that describe a protective role of the gut microbiota in various models of systemic inflammation2 ,10–14 and contradict the hypothesis that broad-spectrum antibiotic-induced microbial disturbances result in decreased systemic immune responses during human endotoxemia. We put special emphasis on two potential adverse effects of microbiota disruption on systemic immune responses that have emerged from preclinical murine studies. First, several mouse studies showed a protective effect of a healthy gut microbiota during sepsis by stimulating bone marrow neutrophil production.11–13 ,15 However, we did not find any effect of antibiotic-induced gut microbiota disturbance on systemic neutrophil influx or degranulation during endotoxemia in healthy subjects. Second, we expected cytokine production during endotoxemia to be diminished following treatment with broad-spectrum antibiotics, based on findings of our own and others in murine models of sepsis.12 ,14 On the contrary, we here demonstrate that antibiotic pretreatment does not affect systemic cytokine levels during human endotoxemia.
Nevertheless, these results do not rule out subtle immunomodulation by the intestinal microbiota. The innate immune system has a great level of redundancy, which may compensate for relatively small deficits of certain cell types. It should be emphasised that the in vivo response to LPS does not directly translate to the control of pathogens. Also, age may be an influential factor: the bone marrow of neonates receiving antibiotics may very well be affected by the absence of sufficient signals derived from the gut microbiota,12 whereas that of young adults is not. Elderly people may also be more prone to these kinds of disturbances in microbiota-derived signals, with less compensatory mechanisms present due to comorbidities and medications.36 Still, demonstrating such a subtle effect in a heterogeneous patient population may be quite hard. Small descriptive studies of faecal microbiota in critically ill patients have associated higher bacterial diversity with better outcome37 ,38 and a large retrospective cohort study-associated dysbiosis in the gut with higher incidence of subsequent severe sepsis.39 The latter identified 43 095 hospitalisations among 10 996 patients and divided them into three diagnosis groups based on likelihood of intestinal microbiota disturbance (non-infection: not likely, infection: likely, C. difficile infection: very likely). Retrospectively, the likelihood of intestinal microbiota disturbance was associated with an increased risk of subsequent severe sepsis rehospitalisation, which was not the case for non-sepsis rehospitalisations.39 Although no causal relationships can be deduced from this study, it supports our initial hypothesis.
There are several shortcomings to the present study. First, results obtained from healthy young adults cannot be extrapolated to patients on the ICU. Also, in a clinical setting, infectious processes will already be ongoing when antibiotics are started, while we presently administered the antibiotics prior to activating the immune system with LPS. Still, in this proof-of-concept study it is important to use a model that is free from as many confounders as possible in order to be able to dissect the interplay between intestinal microbiota and innate immune system during systemic inflammation. Third, due to the combination of antibiotics we used, we cannot distinguish effects of single antibiotics or bacterial strains. We opted for this broad-spectrum antibiotic regimen given its similarity to the ones used in the landmark murine studies on the role of the gut microbiota in systemic inflammation and sepsis10–12 ,14 ,40 and the considerable overlap of its spectrum with commonly used antibiotics in everyday clinical practice in the ICU. The 36-hour interval between antibiotics and LPS infusion was chosen to avoid direct interference of the antibiotics with any of the measurements. Although the microbiota was strongly altered compared with before treatment, a partial recovery may have taken place during this time. Also, it should be kept in mind that antibiotics may have direct immunomodulatory effects, as was shown for macrolides in chronic obstructive pulmonary disease.41 ,42 Lastly, the diversity in microbiota composition between individuals may be a confounding variable and has been associated with health parameters.43 It has recently been suggested however that on a functional level most people have a similar intestinal microbiota with equal numbers of bacterial metabolic genes in important pathways such as carbohydrate metabolism and vitamin biosynthesis.44
This study shows that, contrary to results from recent murine studies, gut microbiota disruption by broad-spectrum antibiotics does not affect systemic innate immune responses during human endotoxemia. The translatability of mouse models for inflammation and sepsis has been under debate recently.45 ,46 Experimental murine models are of importance in deciphering potential mechanisms, testing new hypotheses and investigating novel therapeutic strategies, but one should aim to translate—and verify—promising findings to the human situation at an early stage.
The authors would like to thank Guido Clerx for his clinical assistance, Steven Aalvink for his help with the workup of intestinal microbiota samples and Dr Rene Lutter, Barbara Dierdorp and Tamara Dekker for their help with the Luminex assays. Special thanks to Rene Lutter for critically reading the manuscript and giving them the opportunity to perform Luminex assays.
Contributors JML and DRC performed clinical work; JML and CB performed analyses; JML, AFV and WJW designed the study; JML and WJW wrote the manuscript; WMV, TP and WJW supervised and critically revised the work.
Funding This work was supported by the Netherlands Organization for Health Research development (ZonMw, grant nr. 90700424 to WJW) and the Netherlands Organization for Scientific Research (NWO, Spinoza and SIAM Gravitation Grant no. 024.002.002 to WMV).
Competing interests None declared.
Patient consent Obtained.
Ethics approval All subjects gave written informed consent and research was conducted in accordance with the declaration of Helsinki. The study was approved by the Ethics and research committee of the Academic Medical Center, Amsterdam. Netherlands Trial Registry number NTR4549; ClinicalTrials.gov identifier NCT02127749.
Provenance and peer review Not commissioned; externally peer reviewed.