While much effort has been made to characterize influenza A pdm09 virus (pH1N1), the flu that was responsible for the fourth influenza pandemic, there is a lack of study on the composition of bacteria that lead to secondary infection. In this study, we recruited pneumonia patients with and without pH1N1 infection and characterized their oropharyngeal microbiota by the unbiased high-throughput sequencing method. While there were no significant differences in common bacterial pneumonia-causative agents (Acinetobacter and Streptococcus species), previously unreported Pseudomonas species equipped with chemotaxis and flagellar assembly genes significantly increased (>20-fold) in the pH1N1-infected group. Bacillus and Ralstonia species that also increased significantly (5–10-fold) were also found to possess similar signaling and motility genes. In contrast, no such genes were found in oral commensal Prevotella, Veillonella and Neisseria species, which decreased significantly, or in either Acinetobacter or 10 out of 21 Streptococcus species, including Streptococcus pneumoniae. Our results support the notion that pH1N1 infection provides a niche for previously unnoticed potential respiratory pathogens that were able to access the lower respiratory tract with weakened immunity.
Pneumonia, the sixth most common deadly disease, is usually caused by bacteria, viruses or fungi. Among those infectious agents identified, Streptococcus (S.) pneumoniae and Haemophilus (H.) influenzae type b are the first and second most common causes of bacterial pneumonia, respectively. Respiratory syncytial virus, human metapneumovirus and the parainfluenza viruses are the leading viral causes of pneumonia in infants and children.
In March 2009, there was an outbreak of respiratory illness caused by influenza A pdm09 virus (pH1N1) . Although most people infected by pH1N1 recovered, a significant number of patients were at risk of severe outcomes . Compared with seasonal influenza, pH1N1 was able to replicate in a respiratory tract region lower than the nasal cavity and was more likely to cause viral pneumonia . In addition, pH1N1, but not seasonal influenza, was implicated in gastrointestinal problems . Influenza A virus can lead to reduced levels of adrenocorticotropic hormone and cortisol, and a decreased response of Toll-like receptor, which might in turn inactivate the immune response . Epithelium destruction by immune cells allows increased adhesion of bacteria to the tracheal wall and thus better retention and growth of pneumococci . Excessive bacterial, and not viral, growth is associated with increased lethality only in the presence of prior flu infection [6, 7].
There is only a single study that characterized the composition of flu-infected nasopharyngeal bacteria and the study focus was the viral metagenome , and there is a lack of study on the differences between the microbiota of pneumonia patients with and without pH1N1 infection. To investigate whether there is any modulation of the composition of microbiota by pH1N1 in pneumonia patients, we recruited throat swab samples from pneumonia patients with and without pH1N1 infection and examined the differences in the microbiota. Our results revealed the differential microbiota in the two groups and discovered potential respiratory pathogens previously overlooked. The enriched bacteria in the pH1N1 patients were equipped with chemotaxis and flagellar assembly genes, while the decreased ones were not.
Materials and methods
Diagnosis criteria, microbiological tests, participant enrollment and sample collection
Criteria for pneumonia include: (i) emerging cough and sputum, (ii) purulent sputum accompanying worsening symptoms of pre-existing respiratory disease, (iii) fever (>37.4°C), (iv) signs of consolidation in the lung and/or moist rales, (v) number of white blood cell leukocytes >10 × 109/L or < 4 × 109/L, and (vi) chest X-ray examination showing sheets, patchy infiltration shadows or interstitial changes. Patients who met two or more criteria including criterion (v) were diagnosed as having pneumonia. Nonetheless, tuberculosis, pulmonary neoplasms, non-infectious interstitial lung disease, pulmonary oedema, atelectasis, pulmonary embolism, pulmonary eosinophil infiltration syndrome and pulmonary vasculitis were excluded. Viral pneumonia due to influenza virus infection was confirmed when a positive laboratory test (real-time PCR or cell culture) and criteria for pneumonia were simultaneously met. Potential pathogens in patients with influenza virus pneumonia were inspected as follows. The bacteria isolated from induced sputum were considered to be definitive causative pathogens and identified by VITEK (BioMérieux, Craponne, France). For serological examinations, a four-fold increase in the antibody titre level between paired sera was considered definitive evidence. Based on the examination results, the pneumonia patients who consulted the Division of Respiratory Disease, the First Affiliated Hospital of Guangzhou Medical College (Guangzhou, China), from October 2009 to March 2010 were divided into two groups: pneumonia without any influenza A infection (N) (n = 11) and pneumonia with pH1N1 infection (pH1N1) (n = 11). Their clinical features are summarized in Table 1. This study was approved by the ethics committee of the First Affiliated Hospital of Guangzhou Medical College and informed consent was obtained for all subjects.
Table 1. Clinical information and diagnosis of participating pneumonia patients
Data are shown as number (%) unless otherwise specified.
The Mann–Whitney two-sample test and chi-square test were used as appropriate.
Lung disease was defined as chronic obstructive pulmonary disease, asthma, bronchiectasis, interstitial lung disease and cancer.
Other chronic diseases include kidney diseases, endocrine disease (excluding diabetes), haematologic disease, chronic disease of nervous system (excluding cerebrovascular disease), and gastrointestinal disease.
Denotes the highest body temperature of feverish patients before and on admission to hospital.
Gastrointestinal symptoms included nausea, vomiting, diarrhoea, abdominal pain and abdominal distention.
Genomic DNA was extracted from throat swab samples with the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) and amplified with the Illustra GenomiPhi V2 DNA Amplification Kit (GE Healthcare, Waukesha, USA). Briefly, 10 ng of genomic DNA was mixed with 9 μL of sample buffer. The mixture was heated to 95°C for 3 min and then cooled to 4°C on ice. It was then transferred to the amplification reaction buffer and incubated at 30°C for 90 min and 65°C for 10 min. The amplified genomic DNA samples were pooled together as two groups (N and pH1N1) for sequencing using the Illumina Solexa Sequencing Platform of the Beijing Genome Institute Shenzhen.
All sequence reads obtained from each group were searched against human, bacterial and viral genomes by the BLAST algorithm . To speed up the analysis, we first evaluated our data by assembling the reads into contiguous blocks (contigs) and subjected them to a BLAST search against the nucleotide databases in the National Center for Biotechnology Information (NCBI) of the National Institute of Health, USA. To minimize errors generated from mis-assembly, we then searched all sequence reads directly against all completed or draft genome sequences of humans (version 19), bacteria and viruses using similar methods reported in our previous study . An e-value of 1e−20 was chosen to trade-off between sensitivity and specificity. Orthologs of representative bacterial genomes were identified by OrthoMCL . Protein functionality analyses were carried out on COG and KEGG annotation [12, 13]. The statistical significance of fold-change was estimated by DEGseq .
Clinical characteristics of pneumonia patients
The recruited subjects were generally young in both groups (as seen in Table 1) and gender balanced. All of them are of the Han race. Five and four pH1N1 infected individuals complained of headache and fatigue, respectively, but none was found in the uninfected individuals. Four individuals (36.4%, comparable to 38% mentioned in ) had gastrointestinal symptoms compared with one in the non-infected group. U-tests were performed and the p-values for headache, fatigue and gastrointestinal symptoms are 0.035, 0.074 and 0.139, respectively. Leucocytes and neutrophils but not lymphocytes increased in the pH1N1-infected group. The history of exposure and other clinical characteristics displayed no significant differences.
Differential distribution of microbiota
The oropharyngeal microbiota (Fig. 1a) of both groups of samples demonstrated phylum-level distribution, which most resembled that of the oral cavity . At family level, Pseudomonadaceae, Moraxellacease and Bacillaceae in the pH1N1 group (pH1N1) were significantly more abundant than the non-flu infected pneumonia (N) patients. In contrast, Prevotellacease, Veillonellaceae and Neisseriaceae decreased significantly (Fig. 1b, Table S1). At the genus level (Fig. 1c, Table S1), the relative abundance of Pseudomonas, Ralstonia and Bacillus recorded a significant increase in the pH1N1 group. Also in accordance with the results at the family level, Prevotella, Veillonella and Neisseria species decreased significantly in the pH1N1 group as expected.
We also searched the reads for the nearest matched species at the nucleotide sequence level (Table S1). The relative abundance of some previously unnoticed Pseudomonas species, such as Pseudomonas (P.) amygdali, P. fluorescens and Pseudomonas sp. UK4, increased significantly. Interestingly, Mycoplasma pneumoniae (954 reads), a type of bacteria that often causes slowly developing infection, was detected exclusively in the N group, while Mycoplasma hyorhinis (231 reads), a primary agent of enzootic pig pneumonia, was primarily found in the pH1N1 group (the N group only had two reads). However, other common bacterial pneumonia causative agents, including H. influenzae, S. pneumoniae, Moraxella catarrhalis and Neisseria meningitidis, only accounted for a very small proportion (<1500 reads in the two samples in total) in the oropharynx and did not vary significantly between the two groups of samples; so did the other respiratory disease causative agents such as Bordetella parapertussis and Bordetella bronchiseptica (two reads). The rarity of these common respiratory bacterial pathogens reported by high-throughput sequencing is concordant with the primary diagnosis result. HACEK organisms, which comprise normal oropharyngeal flora, were found to decrease in the pH1N1 group as expected, when estimated at the genus level (N, 1670; pH1N1, 337).
Identification of bacteria present in the pH1N1 group
To unravel the identity of bacteria present in the pH1N1 group, we assembled our pair-end reads of sequences and into contiguous blocks (contigs) of length of 500bp or above. This reduces the chance of matching homologous regions shared by different bacteria. Searching the contigs against the public nucleotide database in NCBI yielded 11 and 162 hits that had top matches of Ralstonia picketti and P. fluorescens, respectively, covering various regions of the genome. While the former was almost exactly matched (matches of 100% similarity, 7; >99.5%, 2; >80%, 2), there existed a significant number of protein coding sequences that did not have high similarity to the homologues of P. fluorescens (30–40%, 3; 40–50%, 6; 50–60%, 5; 60–70%, 8; 70–80%, 13; 80–90%, 24; 90–100%, 103). Previously unreported Pseudomonas species were likely to be present in pH1N1 infected patients.
Functional capacity of enriched microbiota in the pH1N1 group
Species from the pH1N1 and N groups containing COG annotation that showed differential abundance were examined regarding differences of functional genes. We found that only the pH1N1 group contained COGs involved in cell motility, signal transduction mechanisms, inorganic ion transport and metabolism, secondary metabolites biosynthesis/transport/catabolism, and transcription regulation (Table S2).
In particular, cell motility and signal transduction associated proteins were obviously much enriched and therefore we examined their specific functions (Table S3): bacterial chemotaxis and flagellum assembly. It is possible that these genes enable the bacteria to sense the niche created by pH1N1 infection and accordingly synthesize machinery to reach potential nutrient-rich environments.
To study whether the findings can be generalized, under categories ‘chemotaxis’ and ‘flagellar assembly’ in the KEGG database, we examined all the bacteria that are equipped with these two kinds of genes. Surprisingly, Prevotella, Veillonella, Neisseria, Acinetobacter and 10 out of 21 Streptococcus species (including S. pneumoniae) lack chemotaxis genes and all the species in these five genera lack flagellar genes.
Because amplified genomic DNA samples were pooled together for the two groups of patients, significant differences between the two groups could be over-represented in only one of the samples. Moreover, our studies could not detect important RNA viruses, such as influenza viruses A and B, respiratory syncytial virus and human parainfluenza viruses. This problem could be overcome by collecting RNA samples for unbiased high-throughput sequencing in the future.
Influenza complications are common and secondary bacterial pneumonia can be fatal. In our study, a significant amount of previously unnoticed bacteria was found to comprise the oropharyngeal microbiota of pH1N1-infected patients but not the uninfected pneumonia patients. Unexpected pathogenic bacteria were also reported in a recent study of bacteria in bronchoalveolar lavage fluid from children with cystic fibrosis . These results indicate that there is a need to develop more sensitive techniques for better potential pathogen detection.
Although the oropharynx is an environment that contains heterogeneous sources of air- and food-borne microbes, the normal oropharyngeal microbiota is relatively stable and large perturbations usually signify infections. Recent reports suggest that microbiota disturbance is associated with the development of lung diseases such as asthma and COPD [17-19]. In both asthmatic and COPD patients , the relative abundance of Bacteroidetes, particularly Prevotella species, was found to decrease significantly. The aforementioned findings are consistent with our finding and pH1N1 might be responsible for the modulation.
As microflora can establish cooperation, competition or a hybrid inter-relationship, we cannot conclude whether the observed increase in potential opportunistic pathogens and decrease in oral commensal species simply reflected solely the increase of the former, decrease of the latter, or both. Despite constant challenges from microorganisms that originate from the naso- and oropharynx, healthy individuals are equipped with immunity and physical barriers that may purge, keep out or destroy potential respiratory pathogens that may affect the lower respiratory tract. However, when our mucociliary escalator is damaged by pH1N1 infection, more bacteria are then able to enter the lower respiratory tract. COG enrichment and KEGG analysis results support the idea that enriched bacteria in the pH1N1 group were better equipped with proteins that participate in cell motility and signal transduction.
Chemotaxis, a commonly observed phenomenon in living organisms, is involved in a wide array of biological processes, including development, defence, nutrient acquisition and damage avoidance . In this study, we have identified that significantly increased bacterial species in the pH1N1 group are equipped with chemotaxis and flagellar assembly genes, while the decreased bacterial species are not. The significance of chemotaxis and flagellar genes biosynthesis has also been reported in other bacteria such as Escherichia coli and Rhizobium leguminosarum [21, 22].
Contig analysis suggests the presence of previously unreported species, in particular Pseudomonas species. It is noted that a lot of reads were mapped to P. fluorescens and this bacterium had been associated with ventilator-associated pneumonia , damage to nerve and epithelial intestinal cells and bacteraemia [24-26]. The possibility of transmission through faecal viral shedding and subsequent faecal-oral contact cannot be ruled out, because more pH1N1-infected patients had gastrointestinal problems and P. fluorescens is associated with Crohn's disease . Nonetheless, careful interpretation is needed as P. fluorescens genomes are highly diverse and our results also showed large variation from the reference genome (P. fluorescens Pfl-5) . Conventional culturing techniques and polymerase chain reaction using design primers specific for the Pseudomonas community can help elucidate the composition of these potentially pathogenic species.
In conclusion, our study revealed previously unnoticed potential respiratory pathogens and suggested that prior pH1N1 infection may facilitate the colonization of opportunistic pathogens that are able to gain better access to the lower respiratory tract by chemotaxis and cell motility genes. In contrast, commensal microflora that lack these genes may be at a disadvantage and as a result decrease in relative abundance. Further efforts should be made to identify the potential pathogens so that more appropriate and prompt treatment can be received. Such results would help plan for the next wave of pandemic influenza , as well as boost the chance of survival for those who are prone to aspiration pneumonia .
We would like to thank Sihua Pan, Shiguan Wu and Yangqing Zhan for their assistance in collecting samples and clinical information. This work was supported by the Focused Investment on the Centre for Microbial Genomics and Proteomics of the Chinese University of Hong Kong and the Scientific Research Fund from the State Key Laboratory of Respiratory Disease of the Guangzhou Medical University.
None of the authors have any conflicts of interest.