Observations that the airway microbiome is disturbed in asthma may be confounded by the widespread use of antibiotics and inhaled steroids. We have therefore examined the oropharyngeal microbiome in early onset wheezing infants from a rural area of tropical Ecuador where antibiotic usage is minimal and glucocorticoid usage is absent.
Materials and Methods
We performed pyrosequencing of amplicons of the polymorphic bacterial 16S rRNA gene from oropharyngeal samples from 24 infants with non-infectious early onset wheezing and 24 healthy controls (average age 10.2 months). We analyzed microbial community structure and differences between cases and controls by QIIME software.
We obtained 76,627 high quality sequences classified into 182 operational taxonomic units (OTUs). Firmicutes was the most common and diverse phylum (71.22% of sequences) with Streptococcus being the most common genus (49.72%). Known pathogens were found significantly more often in cases of infantile wheeze compared to controls, exemplified by Haemophilus spp. (OR = 2.12, 95% Confidence Interval (CI) 1.82–2.47; P = 5.46×10−23) and Staphylococcus spp. (OR = 124.1, 95%CI 59.0–261.2; P = 1.87×10−241). Other OTUs were less common in cases than controls, notably Veillonella spp. (OR = 0.59, 95%CI = 0.56–0.62; P = 8.06×10−86).
The airway microbiota appeared to contain many more Streptococci than found in Western Europe and the USA. Comparisons between healthy and wheezing infants revealed a significant difference in several bacterial phylotypes that were not confounded by antibiotics or use of inhaled steroids. The increased prevalence of pathogens such as Haemophilus and Staphylococcus spp. in cases may contribute to wheezing illnesses in this age group.
Citation: Cardenas PA, Cooper PJ, Cox MJ, Chico M, Arias C, Moffatt MF, et al. (2012) Upper Airways Microbiota in Antibiotic-Naïve Wheezing and Healthy Infants from the Tropics of Rural Ecuador. PLoS ONE 7(10): e46803. doi:10.1371/journal.pone.0046803
Editor: Susanne Krauss-Etschmann, Ludwig-Maximilians-University Munich, Germany
Received: April 6, 2012; Accepted: September 10, 2012; Published: October 5, 2012
Copyright: © Cardenas et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The cohort and collection of samples in Ecuador was supported by the Wellcome Trust (grant number, 088862/Z/09/Z). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Asthma is a chronic disease of the airways that is characterized by an abnormal mucosa, intermittent airway inflammation and symptoms of wheezing, dyspnea and cough. The syndrome results from a complex interplay between genetic and environmental factors .
A worthwhile understanding of the causes of asthma needs to reconcile consistent epidemiological indications of the importance of the microbiome (also known as microbiota) to the disease . These include the protection afforded by a rich microbial environment in early life , , observations that the bronchial tree contains a characteristic flora that is disturbed by the presence of pathogens such as Haemophilus infuenzae in asthma , , birth cohort studies showing that the presence of the same pathogens in throat swabs predicts the later development of asthma  and recognition that these bacteria have consistently been associated with exacerbations of asthma .
Of potential importance is the finding that organisms commonly found in healthy airways and mucosal surfaces are significantly reduced in asthmatic airways . Investigations of inflammatory bowel disease have shown that a normal bacterial flora is essential in maintaining a healthy mucosa  and similar mechanisms are likely to be important in the airways –. Murine studies have shown that sterility of the airways and intestinal tract results in enhanced inflammatory responses to a variety of stimuli , .
Ninety percent of the cells in the human body are microorganisms including bacteria, parasites and archaea . These microorganisms are commensal on body surfaces exposed to the external environment including the gut, respiratory tract and skin. Allergy, and other immune diseases are associated with differences in microbial communities, but it is unclear if these differences are causes or consequences of disease , , . Although most bacteria are not cultivable with standard methods , the membership of complex microbial communities can be quantified and classified by DNA sequencing of the conserved bacterial 16S rRNA gene , . Bacteria are classified by these sequences into Operational Taxonomic Units (OTUs). OTUs approximate closely but not completely to taxonomy derived from classical techniques, and sequences of other regions may be necessary for precise discrimination at the species level.
Previous studies of the airway microbiome in healthy and in diseased subjects have been carried out in Westernized societies – where antibiotic use and the prescription of inhaled corticosteroids is almost ubiquitous and confounds understanding of the microbiome. We have therefore carried out a sequence-based study of the upper airway microbiome in children from the Esmeraldas province in rural Ecuador who have had minimal exposure to antibiotic medications and no exposure to inhaled steroids.
Ecuador has strong regional differences in the prevalence of wheeze in rural compared to urban areas. The prevalence of wheezing in children has been estimated to be 16.6% in urban areas  while in rural areas of the Pichincha province the rate of current wheezing has been estimated between 0.8% and 2.2% , . Factors that may be protective against asthma in this region include exposures associated with living in a rural environment, low antibiotic usage and a high rate of geohelminth parasitic infections. We therefore sought to compare and contrast the airway microbiome in infants with non-infective wheeze and healthy controls, and to relate our findings to surveys of the airway microbiome in European children.
Materials and Methods
A case-control study was designed to investigate the upper airway microbiota profiles of early onset non-infectious wheezing infants (cases) and healthy infants (controls). The project used DNA extracted from oropharyngeal swabs samples collected from the hypopharynx of 48 infants (average age 10.2 months) that were recruited as part of a birth cohort (the ECUAVIDA cohort) in the Esmeraldas Province in Ecuador. The aim of the ECUAVIDA cohort study is to investigate the effects of early infant infections on the development of immunity, allergic sensitization and allergic disease and the methodology has been previously described in detail , . The study is an unselected population-based birth cohort that has recruited 2,403 newborns in the rural District of Quininde in the Esmeraldas Province, Ecuador. Detailed data has been collected from the mothers at the time of the first antenatal visit using questionnaires and environmental sampling. The protocol for the ECUAVIDA cohort was approved by the Ethical Committees of the Hospital Pedro Vicente Maldonado and Universidad San Francisco de Quito, Quito, Ecuador.
Twenty-four infants were selected with early onset multiple-trigger non-infectious wheezing according to the GINA guidelines (http://www.ginasthma.org/). Wheezing illness was diagnosed by a physician. Twenty-four healthy controls (no history of wheezing, current respiratory disease, chronic disease or current infections) were selected and paired by age range to cases. The samples were collected when the cases and controls did not have any evidence of a current airway infection (cold symptoms and fever). The infants in both groups had a minimal history of antibiotic use, with 87% never being exposed to antibiotics, and the whole group receiving on average 0.12 courses of antibiotics per child (Table 1). None of the infants had received antibiotics for any reason for at least two weeks prior to sampling. None of the subjects had ever received corticosteroids. All subjects were of the same mixed ethnic background, lived in the same town and had access to the same basic electricity, water, and sanitation services. All subjects had received vaccines recommended by the Ecuadorian Ministry of Public Health. None had received anti-Streptococcal vaccination.
Sample Collection and Storage
Throat swabs were collected by a physician using sterile cotton swabs and were placed in collection tubes (Qiagen, UK). Sampling was performed carefully without touching any surface other than the oropharynx and using a tongue depressor. Each swab was rubbed approximately five times around the oropharynx. An even pressure was applied and the swab was rotated without interruption. Post sampling, the swab was immediately placed back into the collection tube and stored at −20°C and subsequently within the next 24 hours at −80°C. Samples were shipped to Imperial College, London on dry ice.
Total sequence counts for individual operational taxonomic units (OTUs) are shown in the right column. Taxonomy assignments at the phylum level are shown in the inner column and colour coded. Intrusions of different colours within particular phyla indicate discrepancies between the phylogenetic and database classifications.
Bacterial DNA Extraction from Throat Swabs
Bacterial DNA was extracted from the throat swabs using a modified protocol of the commercial QIAmp DNA Mini Kit (Qiagen). Additional steps at the beginning of the protocol were included to improve the lysis of Gram positive bacteria .
Each swab head was transferred into a 2 ml microcentrifuge tube, and 432 µl TE (Tris EDTA pH 8.0)+18 µl 4X lysozyme solution was added (4X lysozyme in a concentration of 1000 U/µl was prepared from lysozyme stock at 30,000 U/µl Ready-Lyse™ Lysozyme Solution of EPICENTRE, UK).
Samples were incubated for 1 hour at 37°C; to allow improved lysis of Gram +ve bacterial walls. During this hour, samples were vortexed for 20 seconds at intervals of 15 minutes. Next 30 µl of Proteinase K and 450 µl of Buffer AL were added to the tube and samples were incubated at 56°C for 30 minutes. To terminate the Proteinase K step, samples were then incubated for 5 min at 95°C.450 µl of Ethanol (96–100%) was added to the sample. In order to obtain a homogenous solution this was mixed by vortexing. This solution was then applied to the QIAamp Spin Column as per the manufacturer’s protocol. In the final step 40 µl of nuclease free water was added instead of the elution buffer supplied by the kit. If the DNA was not being used immediately after extraction samples were stored at −20°C until required.
Multiple rarefaction curves were collated from each sample’s Shannon diversity index. The graphic shows the estimated diversity plotted against the number of sequences per sample. Each line represents one sample. The plateau in each estimated diversity curve indicates the minimum number of sequences to capture diversity. For all samples the plateau was achieved at approximately 320 sequences (red vertical line), well below our chosen rarefaction threshold of 969 sequences (blue line).
Amplification of 16S rRNA Gene
Polymerase chain reaction (PCR) was used to amplify the variable regions 3 to 5 (V3–V5, Primers 454B_357F and 454A_926R ) of the gene that encodes for 16S rRNA in bacteria. Samples were multiplexed using sample specific barcodes  and Roche 454 adaptor sequences (Roche Diagnostics, Oakland). To minimize PCR nucleotide insertion mistakes, a high fidelity Taq polymerase was used, and samples were amplified in quadruplicate reactions with 20 cycles each and then pooled.
Pyrosequencing and Data Analysis
The DNA amplicons were pyrosequenced using a GS Junior Titanium 454 (Roche Diagnostics, Oakland) following manufacturer’s protocols (http://www.gsjunior.com/454-gs-junior.php). Data analysis was performed using the software “Quantitative Insights into Microbial Ecology” (QIIME) . Reads were removed if they were <200 and >800 nucleotides (nt) in length, if there were mismatches in the barcodes or primers, if ambiguous nucleotides were present or if the read quality score was <25. The denoiser algorithm version 1.2.1  was used to avoid overestimation of diversity and chimeras were removed using ChimeraSlayer . Phylogenetic classification was assigned using the Ribosomal Database Project database (RDP).  Sequences were clustered in Operational Taxonomic Units (OTUs) using UCLUST version 2.1 at 97% sequence identity . Any sequences present once (singletons) or in only one sample were filtered out.
Phylogenetic analysis of the 16S rRNA sequences of the OTUs assigned taxonomically to Haemophilus genus (OTUs 32, 108, 162 and 190, shown in Red) together with reference Haemophilus sequences from the SILVA database, using the ARB alignment editor. The scale bar indicates 10% sequence divergence, and NCBI accession numbers are included. The tree was rooted with a near neighbour outgroup constructed with sequences from Morganella morganii, Proteus mirabilis and Providencia stuartii.
Remaining sequences were aligned using PyNast . Sequences were rarefied (to remove the heterogeneity of the number of sequences per sample) prior to calculation of alpha and beta diversity statistics. Alpha diversity indexes were calculated in QIIME from rarefied samples using for diversity the Shannon index , for richness the Chao1 index , and evenness by the equitability index . Beta diversity was calculated using weighted and unweighted UniFrac  and principal coordinate analysis (PCoA) performed. Neighbour joining with nearest neighbour interchange phylogenetic trees were created using the representative sequences of each OTU and FastTree version 2.1.3 . The heatmap to represent the abundance of sequences was constructed in iTOL . Microbial community comparisons were performed using parametric statistics in METASTATS, with the P values corrected by multiple hypothesis testing using the false discovery rate (FDR) . Sequences are available in EBI (European Bioinformatics Institute) with the accession number ERP001558.
A) Scatter dot plot comparing cases versus controls values of chao1 richness index. B) Scatter dot plot comparing values of Shannon diversity index. C) Scatter dot plot comparing equitability evenness index. D) Unweighted UNIFRAC Principal Coordinate Analysis (PCoA) plot comparing presence/absence metrics. E) Weighted UNIFRAC Principal Coordinate Analysis (PCoA) plot comparing presence/absence metrics and abundance.
Representative sequences from significantly different OTUs of further interest were investigated using more intensive phylogenetic approaches in order to maximize the quality of the identification. These test sequences were aligned using the online SINA aligner (http://www.arb-silva.de/aligner/version 1.29 ) and this was imported into the ARB phylogenetic software (version 5.1, http://www.arb-home.de/) running on Biolinux 6.0 (http://nebc.nerc.ac.uk/tools/bio-linux/bio-linux-6.0, ref 3 ). The aligned SILVA reference database SSU_REF108 of 618,442 high quality 16S rRNA gene sequences was downloaded and merged with the aligned test sequences. All Haemophilus spp. (or Streptococcus spp.) sequences within the database were selected and the SINA alignment individually checked for each test sequence in the ARB alignment editor. The length of the alignments used depended on the length of the available reference reads for each OTU. Thus, for the Haemophilus spp. alignment, the 522 bp region corresponding to the region between positions 384 and 908 of the Escherichia coli reference were selected, and for Streptococcus spp. 470 bp between positions 470 and 908. Columns of the alignment containing uninformative positions (gaps) were masked from the phylogenetic analysis. Three trees were constructed for each of the genera, an ARB neighbor joining (NJ) tree with 1000 bootstraps, a Maximum Parsimony (MP) tree with 500 bootstraps and a RAxML Maximum Likelihood (ML) tree (version 7.0.3, ) with GTR substitution model in rapid hill-climbing mode. Trees were rooted with sequences from near neighbours outside the genus of interest. Tree topology was compared between the three methods and bootstrap values for the NJ and MP trees were used to determine stability of the phylogeny. Accession numbers for the reference sequences used are recorded in the tree labels and those for the outgroup were DQ358146, AJ301681, AF008582, AJ301682, AF008581, AM040491, and AM040495.
Examination of the clinical data showed no significant differences between the 24 cases of non-infectious wheezing and the 24 controls in age, sex and antibiotic use (Table 1). The controls appeared to have higher parental income ($238 per month compared to $186) and slightly fewer individuals per room in their houses (3.1 compared to 3.6), although only 4% of their mothers had finished high school compared to 21% of cases.
An initial 108,042 raw sequences were obtained from all subjects. After denoising, singleton exclusion, chimera checking and removal of OTUs present in only one sample a total of 76,627 sequences remained (37,235 in cases and 39,392 in controls). Between 969 and 6269 sequences were obtained per sample. In order to control sample heterogeneity, sequences were rarified to the same minimum of 969 for all subjects. Consequently, we identified 182 operational taxonomic units (OTUs) at 97% sequence identity level that were assembled into a phylogenetic tree using FastTree and iTOL (Figure 1).
By extrapolation of collectors’ curves we estimated that the samples contained an average of 289.8 OTUs (95% CI 218.39–361.30). Multiple rarefaction curves using the Shannon index (Figure 2) showed that a plateau of diversity was achieved in around 360 sequences per sample. This value would be considered as the minimum sampling depth to capture diversity. Our rarefaction was performed at a minimum of 969 sequences per sample, which therefore is a realistic panorama of each sample’s diversity. Phylogenetic classification of the ungrouped sequences showed a high prevalence of the phylum Firmicutes (72% of the total number of sequences obtained) followed by Proteobacteria (12%), Actinobacteria (8%), Bacteroidetes (7%) and Fusobacteria (1%) (Figure 1). Firmicutes was the most diverse phylum containing 93 distinct OTUs (51% of the total OTUs), followed by Actinobacteria with 30 OTUs (16%), Proteobacteria with 20 OTUs (11%), then Fusobacteria and Bacteroidetes with 18 and 19 OTUs respectively (10%). Streptococcus was the most common genus (49.72% of the total) followed by Veillonella (14.5%), Atopobium (5.37%) and Prevotella (4.72%).
We performed analysis of OTU taxonomy assignments summarized at genus level (when it was not possible to define at genus level, the next best determined taxonomy assignment was used), and the number of samples from where differences were detected at greater than 1% was included (Table 2). We found that members of the groups Actinomyces (P = 1.89×10−02, OR 1.10), Atopobium (P = 8.99×10−20, OR 2.27), Corynebacterium (P = 1.37×10−129, OR 24.99), Flavobacteriaceae (P = 4.02×10−31, OR 12.07), Prevotella (P = 3.24×10−13, OR 1.38), Staphylococcus (P = 1.87×10−241, OR 124.11), Neisseriaceae (P = 5.84×10−05, OR 1.19), and Haemophilus (P = 5.46×10−23, OR 2.12) occurred highly significantly more often in the cases of infantile wheeze compared to non-wheezing controls.
By contrast in controls there was a significantly higher prevalence of Bacteroidales (P = 9.57×10−08, OR 0.55), Porphyromonas (P = 2.81×10−32, OR 0.20), Gemella (P = 4.29×10−21, OR 0.40), Lachnospiraceae (P = 7.79×10−14, OR 0.39), Veillonella (P = 8.06×10−86, OR 0.59), Leptotrichia (P = 9.37×10−14, OR 0.42), Pasteurellaceae (P = 1.13×10−20, OR 0.20) and Moraxella (P = 4.54×10−06, OR 0.79) (Table 2). We repeated the statistical analysis excluding the 6 children that had a minimal use of antibiotics and obtained similar results to the full sample set.
Inspection of the data for individual subjects (Figure 1) showed that the high abundance of Moraxella sp. in a single sample resulted in the whole genus being significantly more prevalent in controls. If this particular sample was excluded from the analysis, Moraxella became more prevalent in cases.
The biological interpretation of the differences in the frequencies of individual OTUs is limited by the imprecision of OTU assignments in identifying individual species. In particular, this problem was obvious in our data with OTUs assigned to Streptococcus and Haemophilus spp., which were likely to contain a mixture of pathogenic and non-pathogenic strains. We therefore attempted to improve the classification of these OTUs by including them in phylogenetic tress constructed from reference sequences.
Using three independent phylogenetic treeing methods it was not possible to increase the specificity of the identification of the Streptococcus spp. OTUs beyond that of the basic Ribosomal Database Project (RDP) classifier. Tree topology between the three methods of treeing was not conserved and significance of assignment to major nodes was low.
By contrast, the tree topology for Haemophilus spp. was conserved in all three methods of phylogenetic inference (Figure 3), with robust significance for assignment for each Haemophilus spp. OTU. This enabled confident assignment of OTU 162 to Haemophilus influenzae, OTU 32 and 38 to Haemophilus haemolyticus, and OTU 190 to Haemophilus parainfluenzae. OTU 162 which was assigned to the known pathogen Haemophilus influenzae was significantly more common in cases (P = 1.66×10−49, OR 3.45) compared with controls, whilst OTU 190 (assigned to Haemophilus parainfluenzae) was significantly more prevalent in controls (P = 2.74×10−48, OR 0.05).
We next tested if changes in the abundance of individual OTUs were associated with alterations in the overall structure of the microbial populations between cases and controls. We did not detect any differences between cases and controls in species richness, taxa abundances, or evenness (Figure 4 A, B and C). We found no large-scale differences in microbial community cluster patterns between cases and controls (beta diversity) using the principal coordinate analysis (PCoA) of the UniFrac distance matrix (weighted and un-weighted) (Figure 4, D and E).
This study has shown significant perturbations of the airways microbiota in infants with early onset non-infectious wheeze from a rural district in the tropics of Ecuador. Previous studies conducted in a geographically adjacent area with similar climatic, population, and geographic characteristics estimated the prevalence of asthma in school age children to be just 2.2%  and our subjects were characterized by minimal use of antibiotics and a complete absence of use of corticosteroids and anti-Streptococcal vaccination. Our samples may therefore exemplify a naturally developing microbial community in the early months of life. Bacterial diversity in these infants (estimated mean 290 OTUs) may be higher than that in seen previously in European children (mean 85 OTUs) , although differences in methodology mean that this inference should be treated with caution.
It is of interest that Firmicutes was the most common and most diverse phylum with Streptococcus the most prevalent genus, consistent with observations of the airways microbiota in European children . We were not able to differentiate well between members of the important Streptococcus group on the basis of the 16S sequences, and the typing of alternative chronometers (such as pheS, rpoA, and gyrB ) and the incorporation of mutilocus sequence analysis (MLS)  will be important in future studies of the airway microbiome.
We defined our phenotype in infants by non-febrile episodic wheezing, according to the GINA guidelines (http://www.ginasthma.org/). These guidelines recognize that asthma diagnosis before the age of 6 years is complicated by the difficulty in performing accurate lung function tests. The recognition of recurrent wheezing not related to infections is as a consequence the most important diagnostic indicator of asthma in this age group . Because not all children who experience wheezing events will develop asthma and not all children with asthma wheeze , we have followed the recommendation that the phenotype of recurrent non-infectious wheezing syndrome be applied to children less than 24 months old , .
We used culture-independent molecular techniques to characterize microbial communities and to quantitatively investigate dissimilarities. We, and others have shown previously that the microbiome of the oropharnyx correlates with that of the bronchial tree assessed by brushings or by lavage , , and therefore oropharyngeal swabs were used for this epidemiological survey. Pyrosequencing robustly determines the diversity and abundance of microbial communities in a quantitative and qualitative form. The assignment of approximately 1,600 individual sequences for each subject has conferred substantial statistical power to our study. We have reduced the possibility of bias by stringent removal of chimeric sequences, sequences present only once in the dataset and OTUs that were present in only one sample.
We were aware that a high abundance of sequences in few samples could drive the overall prevalence of bacteria in cases or controls, as occurred with Moraxella spp. in this study. We have therefore confined our tests for significant differences to OTUs present in three or more subjects with a total of more than 100 sequences per sample.
Our study identified a higher frequency of potential pathogens (Neisseriaceae, Prevotella, Corynebacterium, Staphylococcus, Actinomyces and Haemophilus) in wheezy infants compared to healthy controls. This finding is consistent with substantial epidemiological studies of European neonates that used standard bacterial cultures to show carriage of pathogens in neonates predicted later asthma risk . The finding is also consistent with earlier studies in older children and adults with asthma that used 16S rRNA gene sequencing for bacterial characterisation .
The incorporation of reference strain sequences into our phylogenetic trees allowed us to discriminate between Haemophilus spp. OTUs at species level, and to show that the pathogen H. influenzae was more prevalent in wheezing infants, in concordance with previous studies . H. parainfluenzae was more abundant in healthy children which might be associated with wheezing protection. The 16S rRNA gene has previously been suggested for use as a marker in MLST of H. influenzae  and our results confirm that its phylogeny is well matched to species and strain identification in airway samples.
In this study we found a lack of potentially ‘protective’ bacterial genera in non-infectious wheezing infants compared with controls, particularly Veillonella, Pasteurellaceae and Gemella. Alterations in the normal microbiota may alter the host resistance to pathogen colonization in the gut  and in the airways . Possible mechanisms for this include direct inhibition of pathogen growth by commensal secreted factors . Commensal bacteria may also elicit tonic signals in the gut epithelium that prevent activation of innate and adaptive immune responses , . It may be relevant that manipulating the airway microbiome in gnotobiotic mice may produce major changes in airway responsiveness and immunity , .
Our results suggest that the upper airways microbiota of early onset wheezing infants from the tropics exhibit an increase in the frequency of pathogenic bacterial OTUs that is not confounded by antibiotic or steroid medications. These specific differences do not appear to have been accompanied by detectable community changes in the numbers of species and their overall relative distribution in our samples. Larger studies and direct measurements of the communities in the lower airways may change this perception. The results provide further support for a hypothesis that pathogenic bacteria may contribute to a wheezy diathesis in infants .
Conceived and designed the experiments: PAC PJC MFM WOC. Performed the experiments: PAC. Analyzed the data: PAC MJC. Contributed reagents/materials/analysis tools: PJC MFM WOC. Wrote the paper: PAC PJC MFM WOC. Samples collection: PCA MC CA.
- 1. Cookson W (2004) The immunogenetics of asthma and eczema: a new focus on the epithelium. Nat Rev Immunol 4: 978–988.
- 2. Eder W, Ege MJ, von Mutius E (2006) The asthma epidemic. N Engl J Med 355: 2226–2235.
- 3. Ege MJ, Mayer M, Normand AC, Genuneit J, Cookson WO, et al. (2011) Exposure to environmental microorganisms and childhood asthma. N Engl J Med 364: 701–709.
- 4. Hilty M, Burke C, Pedro H, Cardenas P, Bush A, et al. (2010) Disordered microbial communities in asthmatic airways. PLoS One 5: e8578.
- 5. Huang YJ, Nelson CE, Brodie EL, Desantis TZ, Baek MS, et al.. (2011) Airway microbiota and bronchial hyperresponsiveness in patients with suboptimally controlled asthma. J Allergy Clin Immunol 127: 372–381 e373.
- 6. Bisgaard H, Hermansen MN, Buchvald F, Loland L, Halkjaer LB, et al. (2007) Childhood asthma after bacterial colonization of the airway in neonates. N Engl J Med 357: 1487–1495.
- 7. Kraft M (2000) The role of bacterial infections in asthma. Clin Chest Med 21: 301–313.
- 8. Maslowski KM, Vieira AT, Ng A, Kranich J, Sierro F, et al. (2009) Regulation of inflammatory responses by gut microbiota and chemoattractant receptor GPR43. Nature 461: 1282–1286.
- 9. Noverr MC, Huffnagle GB (2005) The ‘microflora hypothesis’ of allergic diseases. Clin Exp Allergy 35: 1511–1520.
- 10. Herbst T, Sichelstiel A, Schar C, Yadava K, Burki K, et al.. (2011) Dysregulation of Allergic Airway Inflammation in the Absence of Microbial Colonization. Am J Respir Crit Care Med.
- 11. Nembrini C, Sichelstiel A, Kisielow J, Kurrer M, Kopf M, et al.. (2011) Bacterial-induced protection against allergic inflammation through a multicomponent immunoregulatory mechanism. Thorax.
- 12. Noverr MC, Huffnagle GB (2004) Does the microbiota regulate immune responses outside the gut? Trends Microbiol 12: 562–568.
- 13. Savage DC (1977) Microbial ecology of the gastrointestinal tract. Annu Rev Microbiol 31: 107–133.
- 14. Staley JT, Konopka A (1985) Measurement of in situ activities of nonphotosynthetic microorganisms in aquatic and terrestrial habitats. Annu Rev Microbiol 39: 321–346.
- 15. Turnbaugh PJ, Ley RE, Hamady M, Fraser-Liggett CM, Knight R, et al. (2007) The human microbiome project. Nature 449: 804–810.
- 16. Ahmed S, Macfarlane GT, Fite A, McBain AJ, Gilbert P, et al. (2007) Mucosa-associated bacterial diversity in relation to human terminal ileum and colonic biopsy samples. Appl Environ Microbiol 73: 7435–7442.
- 17. Mallol J, Garcia-Marcos L, Sole D, Brand P (2010) International prevalence of recurrent wheezing during the first year of life: variability, treatment patterns and use of health resources. Thorax 65: 1004–1009.
- 18. Weinmayr G, Weiland SK, Bjorksten B, Brunekreef B, Buchele G, et al. (2007) Atopic sensitization and the international variation of asthma symptom prevalence in children. Am J Respir Crit Care Med 176: 565–574.
- 19. Cooper PJ, Rodrigues LC, Cruz AA, Barreto ML (2009) Asthma in Latin America: a public heath challenge and research opportunity. Allergy 64: 5–17.
- 20. White JR, Nagarajan N, Pop M (2009) Statistical methods for detecting differentially abundant features in clinical metagenomic samples. PLoS Comput Biol 5: e1000352.
- 21. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, et al. (2010) QIIME allows analysis of high-throughput community sequencing data. Nat Meth 7: 335–336.
- 22. Reeder J, Knight R (2010) Rapidly denoising pyrosequencing amplicon reads by exploiting rank-abundance distributions. Nat Meth 7: 668–669.
- 23. Haas BJ, Gevers D, Earl AM, Feldgarden M, Ward DV, et al. (2011) Chimeric 16S rRNA sequence formation and detection in Sanger and 454-pyrosequenced PCR amplicons. Genome Res 21: 494–504.
- 24. Cole JR, Wang Q, Cardenas E, Fish J, Chai B, et al. (2009) The Ribosomal Database Project: improved alignments and new tools for rRNA analysis. Nucleic Acids Res 37: D141–145.
- 25. Edgar RC (2010) Search and clustering orders of magnitude faster than BLAST. Bioinformatics.
- 26. Caporaso JG, Bittinger K, Bushman FD, DeSantis TZ, Andersen GL, et al. (2010) PyNAST: a flexible tool for aligning sequences to a template alignment. Bioinformatics 26: 266–267.
- 27. Shannon CE (1948) A mathematical theory of communication. Bell System Technical Journal 27: 379–423 and 623–656.
- 28. Chao A (1984) Nonparametric Estimation of the Number of Classes in a Population. Scandinavian Journal of Statistics 11: 265–270.
- 29. Magurran AE (1988) Ecological Diversity and its Measurement. Princeton, NJ: Princeton University Press.
- 30. Lozupone C, Hamady M, Knight R (2006) UniFrac–an online tool for comparing microbial community diversity in a phylogenetic context. BMC Bioinformatics 7: 371.
- 31. Price MN, Dehal PS, Arkin AP (2010) FastTree 2– Approximately Maximum-Likelihood Trees for Large Alignments. PLoS One 5: e9490.
- 32. Letunic I, Bork P (2011) Interactive Tree Of Life v2: online annotation and display of phylogenetic trees made easy. Nucleic Acids Res.
- 33. Pruesse E, Peplies J, Glöckner FO (2012) SINA: accurate high throughput multiple sequence alignment of ribosomal RNA genes. Bioinformatics.
- 34. Ludwig W, Strunk O, Westram R, Richter L, Meier H, et al. (2004) ARB: a software environment for sequence data. Nucleic Acids Res 32: 1363–1371.
- 35. Field D, Tiwari B, Booth T, Houten S, Swan D, et al. (2006) Open software for biologists: from famine to feast. Nat Biotech 24: 801–803.
- 36. Stamatakis A, Ludwig T, Meier H (2005) RAxML-III: a fast program for maximum likelihood-based inference of large phylogenetic trees. Bioinformatics 21: 456–463.
- 37. Bishop CJ, Aanensen DM, Jordan GE, Kilian M, Hanage WP, et al. (2009) Assigning strains to bacterial species via the internet. BMC Biol 7: 3.
- 38. Wechsler ME (2009) Managing asthma in primary care: putting new guideline recommendations into context. Mayo Clin Proc 84: 707–717.
- 39. Just J, Belfar S, Wanin S, Pribil C, Grimfeld A, et al. (2010) Impact of innate and environmental factors on wheezing persistence during childhood. J Asthma 47: 412–416.
- 40. Charlson ES, Bittinger K, Haas AR, Fitzgerald AS, Frank I, et al. Topographical Continuity of Bacterial Populations in the Healthy Human Respiratory Tract. Am J Respir Crit Care Med.
- 41. Sacchi CT, Alber D, Dull P, Mothershed EA, Whitney AM, et al. (2005) High Level of Sequence Diversity in the 16S rRNA Genes of Haemophilus influenzae Isolates Is Useful for Molecular Subtyping. Journal of Clinical Microbiology 43: 3734–3742.
- 42. Artis D (2008) Epithelial-cell recognition of commensal bacteria and maintenance of immune homeostasis in the gut. Nat Rev Immunol 8: 411–420.
- 43. Crowe CC, Sanders WE Jr, Longley S (1973) Bacterial interference. II. Role of the normal throat flora in prevention of colonization by group A Streptococcus. J Infect Dis 128: 527–532.
- 44. Tano K, Grahn Hakansson E, Wallbrandt P, Ronnqvist D, Holm SE, et al. (2003) Is hydrogen peroxide responsible for the inhibitory activity of alpha-haemolytic streptococci sampled from the nasopharynx? Acta Otolaryngol 123: 724–729.
- 45. Macpherson AJ, Harris NL (2004) Interactions between commensal intestinal bacteria and the immune system. Nat Rev Immunol 4: 478–485.
- 46. Mason KL, Huffnagle GB (2009) Control of mucosal polymicrobial populations by innate immunity. Cell Microbiol 11: 1297–1305.