Introduction
Allergic asthma affects over 100 million people worldwide and its prevalence is increasing, particularly among children in industrialized countries [
1]. This growing epidemic has been attributed in part to improved sanitation methods and widespread antibiotic use, which have in turn reduced childhood exposures to microbial antigens critical for proper immune development [
2]. Blaser and Falkow [
3] have recently revisited the hygiene hypothesis, suggesting that it might not be a decline in childhood infections that is important, rather, it is how modern societal practices are causing the disappearance of ancestral species of our indigenous microbiota that might confer benefits beyond our current understanding. The loss of these ancestral species could be causing a rapid reorganization of the microbial hierarchy in our guts faster than our immune systems can adapt, promoting the emergence of dysregulated immunological disorders like allergic asthma.
Over the last several decades, evidence from human and mouse studies indicate that colonization of the gut early in life has a substantial role in directing immune system development. Disrupting this normal ‘immunological program’ could increase the likelihood of developing atopic disorders like asthma. Epidemiological data show gut microbiota differs between asthmatic and non‐asthmatic infants [
4], early life exposure to environmental microorganisms is protective [
5] and early life as well as prenatal antibiotic exposures increase the risk of allergic asthma [
6,
7,
8].
Antibiotic or probiotic administration as a means of microbiota manipulation in experimental models of murine allergic airways disease has provided further insights into what is now referred to as the ‘gut–lung axis’. Noverr
et al [
9,
10] describe how antibiotic treatment followed by oral administration of
Candida albicans renders immunocompetent hosts susceptible to allergic airways disease. Similarly, oral administrations of
Mycobacterium vaccae [
11],
Helicobacter pylori [
12] as well as conventional probiotic strains [
13,
14] have been shown to ameliorate symptoms of allergic airways disease in mice. Several of these studies highlight the necessity of neonatal administration rather than adult, emphasizing that immune modulation is most effective during the developmental period [
12,
14]. Despite this growing body of evidence that supports a role for the gut microbiota in allergic disease, there is little known about how antibiotic‐induced changes in microbial composition affect the immune status of the host and influence disease susceptibility.
The human gut is colonized by ∼10
14 bacteria, and contains upwards of 1,000 bacterial species [
15]. Specific subsets of the microbiota have been shown to differentially regulate immune function. For example,
Bacteroides fragilis modulates T‐helper‐type 1/2 (Th1/Th2) balance [
16], segmented filamentous bacteria direct Th17 cell differentiation [
17] and
Clostridium species induce regulatory T‐cell (Treg) production [
18]. Antibiotics have the ability to create global shifts in microbial communities, and are valuable tools for studying how changes in the gut microbiota influence disease states [
19].
Here we characterize the effects of two antibiotics on resident murine gut flora and the subsequent susceptibility of these mice to experimental allergic asthma. Although it is known that this model only recapitulates some aspects of the human disease (particularly when mice are only given a single or limited number of antigen challenges) it does recapitulate the airway hyperresponsiveness (AHR) and some of the early phase responses [
20]. It has therefore become a very popular model for evaluating susceptibility to Th2‐driven allergic airways disease. Our data show that neonatal, but not adult, exposure to certain antibiotics promotes enhanced susceptibility to experimental allergic asthma.
Discussion
Over the last several decades, substantial effort has been focused on identifying bacterial populations that correlate with the development of allergy‐related disorders [
4,
25]. The results differ significantly depending on the study, which likely stems from differences in sample populations and methods of microbiota analysis. Nevertheless, most of these studies point towards the first 6 months of life as a ‘critical window’, during which the infant gut microbiota might influence key immunological events that alter allergic sensitization [
4]. In the current study we have addressed the same question using a different approach: by creating defined shifts in the microbiota with antibiotics early in life, we have identified members of the microbiota that differ between more and less susceptible individuals.
In this study, animals treated neonatally with vancomycin showed increases in all major indicators of experimental allergic asthma. Correspondingly, vancomycin treatment markedly shifted the composition of the gut microbiota, compared with both streptomycin‐treated and control populations. Adult vancomycin treatment also altered the microbiota, however, to a different extent, which could be why differences in disease susceptibility ensued. Because our neonatal antibiotic model encompassed several critical time points (that is, prenatal, perinatal, weaning), it will be important to look at how antibiotic treatment at each time point individually affects disease outcome.
Microbial diversity was also reduced after vancomycin, but not streptomycin treatment, despite having similar numbers of total bacteria. There is evidence that reduced diversity in the infant gut microbiota correlates with increased risk of allergic disorders [
26]. On the basis of the extent of microbial reorganization in the gut after neonatal antibiotic treatment, several bacterial populations were positively and negatively associated with disease severity.
Vancomycin‐shifted microbiota could alter host immune development in several ways such as inducing Th2 hyperreactivity or reducing key tolergenic regulatory mechanisms. Consistent with our report, Atarashi
et al [
18] observed that vancomycin treatment reduces colonic CD4
+Foxp3
+ Tregs. They suggest that this observation is due to a reduction in
Clostridium species (clusters IV and XIVa) that they identify as potent inducers of Tregs. We also found that depletion of
Clostridiales with vancomycin correlated with a reduction in cells expressing a CD4
+CD25
+Foxp3
+ Treg phenotype; however, vancomycin treatment also reduced other bacterial phyla. Unlike the colon, we were not able to detect significant changes in CD4
+CD25
+Foxp3
+ Treg numbers in the lung. Perhaps a more detailed look at Treg populations in the lung and mediastinal lymph nodes over the course of disease induction, measuring Treg suppressive capacity or analysing other regulatory cell populations (for example, interleukin 10 producing Tr1 cells), may provide new insights.
Surprisingly,
Bacteroidetes were almost entirely depleted after vancomycin treatment, despite not being directly targeted by the antibiotic; they were replaced by an overgrowth of
Lactobacilli. Several
Bacteroides strains have been implicated in Treg differentiation [
18,
23] and their depletion might also explain the altered susceptibility to experimental allergic asthma. Epidemiological data indicate that antibiotic use in early infancy as well as infants born by caesarean section show similar decreases in
Bacteroides species [
27], both of which are factors associated with increased childhood asthma [
8,
28].
Lactobacilli were negatively correlated with Treg abundance in this study, despite their previous associations with Treg induction in probiotic studies [
29]. These regulatory effects are likely strain‐dependent, and perhaps there are other species or strains of
Lactobacilli that have reciprocal effects that remain uncharacterized. Interestingly,
Lactobacilli actually only represent a small proportion of total bacteria in the healthy mammalian intestine [
21] despite their potential benefits, which suggests that an overgrowth of particular
Lactobacilli species might do more harm than good. Colonizing gnotobiotic animals with subsets of the bacteria discussed here might help identify whether they have specific roles in the development of experimental allergic asthma.
In our study, two events were required to significantly affect the outcome of experimental asthma: (1) global restructuring of the intestinal microbiota with vancomycin and (2) altering the microbiota during infancy, a time point when key immunological thresholds are established. There is substantial evidence that microbial signals are required for lymphoid tissue development and maintenance in the gut (reviewed in Hill and Artis [
30][30]). Vancomycin‐shifted microbiota likely produces an entirely different set of microbial signals compared with healthy controls that could alter intestinal epithelial cell signalling cascades and dysregulate innate and adaptive immune responses. It is already known that mice treated with vancomycin have fewer Th17 cells in the small intestine [
31] and fewer CD4
+ Treg cells in the colon [
18]. Studies in germ‐free animals show that the regulation of immune cell subsets (monocyte/macrophages, B cells, CD4
+Treg/Th17 and CD8
+ T cells) by gut microbes is not restricted to the intestinal compartment [
30]. Similarly, mice deficient in the innate recognition receptors TLR4 or GPR43 show exacerbated asthma responses [
32,
33]. Insights from antibiotic use in other disease models suggest mechanisms mediated by skewed Th2 immune responses [
34] and elevated serum IgE levels [
35]. Dendritic cells have also been implicated in the pathogenesis of asthma [
36], and are thought to be involved in bridging the gut–lung axis [
37], however whether or not signals from intestinal bacteria influence systemic dendritic cell populations remains to be tested directly. Although the link between gut and lung mucosal immune responses remains unclear, we propose that vancomycin administered early in life selects for a community of microbes that disrupt the balance of proinflammatory and regulatory immune responses occurring at local and distant mucosal sites.
In summary, using relevant doses of antibiotics we have found that neonatal, but not adult, treatment with select antibiotics can have profound effects on susceptibility to experimental murine allergic asthma. While antibiotic treatment did not substantially deplete bacterial numbers, it resulted in profound shifts in microbial composition. This was correlated with the loss of a key regulatory subset of immune cells in the colon that could, by way of the gut–lung axis, promote enhanced susceptibility to inflammatory diseases in the lung by a mechanism that remains undefined. This work provides a new framework for interrogating the altered communities responsible for heightened Th2 immune responses in human disease, including asthma.
Methods
Mice. C57BL/6J mice (Jackson Laboratories, Bar Harbor, ME) were bred and maintained in a specific pathogen‐free facility at The Biomedical Research Center. All experiments were in accordance with the University of British Columbia Animal Care Committee guidelines.
Antibiotic treatment. C57BL/6J breeding pairs were given vancomycin or streptomycin (Sigma‐Aldrich, St Louis, MI) at 200 mg l−1 in drinking water. Pups born from respective breeding pairs were reared on antibiotic‐treated water with their littermates for the duration of the experiment. Hence, our term ‘neonatal exposure’ refers to antibiotic exposure both in utero and after birth. Alternatively, 7‐week‐old C57BL/6J female mice were given vancomycin or streptomycin (Sigma) at 200 mg l−1 in drinking water for 2 days before sensitization with OVA, hence ‘adult exposure’.
Vancomycin, an antibiotic that directly targets Gram‐positive bacteria, and streptomycin, an antibiotic that directly targets both Gram‐positive and Gram‐negative bacteria, were chosen because both are used in clinical practice and are poorly absorbed when given orally, minimizing the risk of systemic effects on the host. The rationale behind the clinically relevant concentrations used has been described in detail previously [
38].
Ovalbumin model of experimental murine allergic asthma. Experimental murine allergic asthma was induced as previously described [
39] with minor modifications. Although this model does not fully recapitulate the phenotype of human allergic asthma (particularly after limited numbers of antigen challenge [
20]), it is a useful model for evaluating many aspects of this Th2‐driven lung inflammatory disease. Mice were sensitized intraperitoneally with 200 μg OVA and 1.3 mg alum (both from Sigma) on days 0 and 7. On days 21, 22, 23 and 25, mice were challenged intranasally with 1 mg OVA in PBS. On day 26, mice were anaesthetized with 200 mg kg
−1 ketamine per 10 mg kg
−1 xylazine and blood was collected by cardiac puncture. After killing, BALs were performed by 3 × 1 ml washes with PBS. Total BAL cells were counted by hemocytometer and eosinophils were quantified from cytospins (Thermo Shandon, Pittsburg, PA) stained with HemaStain (Fisher Scientific), based on standard morphological criteria.
Determination of serum IgE. OVA‐specific IgE in serum was measured by enzyme‐linked immunosorbent assay (Chondrex, Redmond, WA).
Airway hyperresponsiveness. Airway hyperresponsiveness was measured on day 26. Mice were anaesthetized, tracheotomized, and intubated with a cannula. Following paralysis with pancuronium bromide (0.5 mg kg−1), AR was measured with a flexiVent apparatus (SCIREQ, Montreal, QC, Canada). Increasing concentrations of MCh (0–1.5 mg kg−1) were administered via the jugular vein. MCh response curves were generated by calculating the change in AR relative to baseline (PBS) for each MCh dose.
Histology. Lungs were collected and fixed in 10% formalin, embedded in paraffin, cut longitudinally into 5‐μm sections and stained with haematoxylin and eosin. Inflammation was blindly assessed from five fields per section, each graded on a scale of 0–5 (0=no signs of disease, 5=severe disease) for each of the following parameters: (1) peribronchial infiltration, (2) perivascular infiltration, (3) parenchymal infiltration and (4) epithelium damage for a maximum score of 20.
Microbial analysis. Bacteria in stool pellets of 7‐week‐old naive mice were stained with SYBR green and counted as previously described [
40]. For composition analyses, stool pellets and ileal contents from naive or OVA‐challenged mice were homogenized using a bead‐beating method (FastPrep instrument, MP Biomedicals, Solon, OH) and total DNA was extracted (Ultra Clean Fecal DNA kit, Mo Bio Laboratories, Carlsbad, CA). 16S rRNA gene fragments were amplified using 33 nucleotide‐bar‐coded primer pairs (27F; 5′‐AGAGTTTGATCMTGGCTCAG‐3′), (519R; 5′‐GWATTACCGCGGCKGCTG‐3′). PCR products were gel‐purified (QIAquick gel extraction kit, Qiagen, Valencia, CA). Each amplicon (100 ng) was pooled and pyrosequenced using a 454 Titanium platform (Roche, Branford, CT).
Bioinformatics. High‐quality sequence reads were determined using MOTHUR[
41] and V‐XTRACTOR [
42]. Each read was assigned to the GREENGENES database [
22] using a naive Bayesian classifier [
43]. Global community structure comparisons from faecal and ileal samples were made using PCO [
44], permutational multivariate analysis of variance [
45] and Simpson's diversity index [
46] implemented in
primer6+ [
47]. Taxa‐treatment association analysis [
48] in conjunction with additional statistical analyses (
supplementary methods online) were performed to determine antibiotic‐related indicator species (faecal samples only). Raw, unfiltered sequences were submitted to the SRA of NCBI.
Isolation of lymphocytes and flow cytometry. Colon and lung tissues were collected from 7‐week‐old naive mice and cells were isolated as previously described [
49]. Cells were stained with fluorochrome‐conjugated antibodies against CD45, CD4, CD25 and Foxp3 (BD Biosciences, Franklin Lakes, NJ). Flow cytometry was performed using an LSR II (BD Biosciences) and data were analysed with FlowJo 8.7 software (TreeStar, Ashland, OR).
Statistics. Differences between control and experimental groups were compared using an unpaired Student's t‐test to calculate statistical significance (GraphPad Prism software, version 4.0, San Diego, CA).
Acknowledgements
We thank Michael Hughes, Jami Bennett, Roland Scholz and Marie‐Claire Arrieta for their helpful advice; Kay Jian for her technical support, Pedro Dimitriu for pyrosequencing support; and members of The Biomedical Research Center Animal Facility for expert animal care. This work was funded by a Canadian Institutes of Health Research (CIHR) Catalyst Grant and a CIHR Emerging Team Grant in partnership with Genome BC and the AllerGen NCE.
Author contributions: S.L.R: study concept and design, data acquisition and analysis; M.J.G.: experimental asthma data acquisition and analysis; M.H.: microbial ecology‐related bioinformatics and statistical analyses; B.P.W.: intellectual content, study design, data acquisition and analysis; L.T., M.W., N.G.: intellectual content, data acquisition and analysis; M.‐R.B.: study design; W.W.M.: intellectual content; K.M.M. and B.B.F.: supervision, study design and interpretation.