Dynamic oropharyngeal and faecal microbiota during treatment in infants hospitalized for bronchiolitis compared with age-matched healthy subjects

Bronchiolitis is one of the most severe diseases affecting infants worldwide. An imbalanced oropharynx (OP) microbiota has been reported in infants hospitalized with bronchiolitis; however, the microbiota dynamics in the OP and faeces during therapy remain unexplored. In total, 27 infants who were hospitalized with bronchiolitis were selected for this study, and sampling was conducted before therapy and after clinical recovery. We also recruited 22 age-matched healthy infants for this study. The faecal and OP microbiota diversity in the patients was lower than that in the healthy children. The faecal microbiota (FM) in the diseased children significantly differed from that in the healthy subjects and contained accumulated Bacteroides and Streptococcus. The OP microbiota in both the healthy and diseased infants was dominated by Streptococcus. After the treatment, the FM and OP microbiota in the patients was comparable to that before the treatment. This study may serve as an additional reference for future bronchiolitis studies, and the “risk microbiota model” of clinically recovered infants suggests an increased susceptibility to pathogen intrusion.


Results
Participants and data output. We selected 27 infants who were hospitalized for mild bronchiolitis from Shenzhen Children's Hospital (Table 1, Supplementary Table); all infants were diagnosed with a human RSV infection. The primary medical treatment during hospitalization was nebulized budesonide combined with salbutamol (Supplementary Table). Two bacterial pathogens, i.e., Streptococcus pneumonia and Haemophilus parainfluenzae, were identified in bacterial cultures of sputum from 5 hospitalized infants (Supplementary Table). Based on clinical experience or diagnoses of bacterial pathogens, 11 diseased infants received antibiotic treatment (Supplementary Table). No inpatient was admitted to the paediatric intensive care unit (PICU) or given mechanical ventilation during hospitalization. Most hospitalized infants remained in the hospital for 3-10 days, except for the following two infants: one infant stayed in the hospital for 13 days, and another infant stayed in the hospital for 26 days (Supplementary Table). The longer hospital stays of these two inpatients were due to secondary infections with rotavirus after clinical remission of bronchiolitis. In addition, 22 age-matched healthy infants were recruited in Shenzhen, China (Table 1, Supplementary Table). In total, 4,898,610 tags were obtained using 16S rDNA amplicon sequencing, ranging from 10,617 to 55,487 tags per sample. Confounder analysis. Bronchiolitis onset, age, body weight, gender, delivery mode, feeding pattern, history of eczema, maternal asthma and smoking status were selected for the analysis of the main contributing factors to the inter-group discrepancies. According to the association analysis, the bronchiolitis onset significantly contributed to the differences in the FM/OP microbiota between the healthy and diseased children (q-value < 0.001, Supplementary Table).
FM and OP microbiota of the infants with bronchiolitis differed from that of the healthy infants. The FM structure in the hospitalized infants differed from that in the age-matched healthy infants ( Fig. 1-A). The FM in the diseased infants exhibited a lower diversity than that in the healthy infants ( Fig. 1-B). Firmicutes Bacteroidetes, Proteobacteria and Actinobacteria accounted for >99% of the FM in both the hospitalized and healthy infants (Supplementary Table), and Proteobacteria (42.14%) and Bacteroidetes (17.31%) were enriched in the FM in the hospitalized infants (Supplementary Table). At the genus level, the diseased infants harboured more Bacteroides (16.44% vs 4.26% in the healthy infants, q-value < 0.05) and Streptococcus (6.54% vs 2.60% in the healthy infants, q-value < 0.05) in the FM ( Fig. 1-C, Supplementary Table). Klebsiella (20.80% vs 0.86% in the healthy infants), Clostridium (7.32% vs 1.27% in the healthy infants) and Enterococcus (11.20% vs 3.19% in the healthy infants) were also more prevalent in the FM in the children with bronchiolitis, but this difference was not statistically significant ( Fig. 1-C, Supplementary Table). By contrast, Collinsella (0.03% vs 13.09% in the healthy infants, q-value < 0.01), Veillonella (1.46% vs 16.51% in the healthy infants, q-value < 0.001), Blautia (0.08% vs 1.78% in the healthy infants, q-value < 0.01), and Erysipelatoclostridium (0.77% vs 2.04% in the healthy infants, q-value < 0.05) accumulated in the FM in the hospitalized infants ( Fig. 1-C, Supplementary Table). Similar to the FM, the OP microbial structure also differed between the healthy and diseased infants ( Fig. 2-A), and the OP microbiota diversity in the hospitalized infants was lower than that in the healthy infants ( Fig. 2-B). Firmicutes (84.79%, q-value < 0.01) was dominant in the diseased infants, while Bacteroidetes (13.94%, q-value < 0.01) and Proteobacteria (20.46%, q-value < 0.05) accumulated dramatically in the healthy  Table). The OP microbiota in both the healthy and hospitalized infants was dominated by Streptococcus ( Fig. 2 Table). Several genera accounted for the lowered abundance of the OP microbiota in the hospitalized infants, including Neisseria (1.57% vs 12.40% in the healthy infants, q-value < 0.001), Bacteroides (0.71% vs 4.43% in the healthy infants, q-value < 0.05), Haemophilus (0.28% vs 4.72% in the healthy infants, q-value < 0.001), Granulicatella (0.40% vs 2.19% in the healthy infants, q-value < 0.001), Leptotrichia (0.07% vs 1.46% in the healthy infants, q-value < 0.001), and Porphyromonas (0.29% vs 2.78% in the healthy infants, q-value < 0.01) ( Fig. 2-C, Supplementary Table). By contrast, Bacillus (14.90% vs 6.86% in the healthy infants, q-value < 0.05), Pseudomonas (1.44% vs 0.19% in the healthy infants, q-value < 0.001), and Raoultella (1.28% vs 0.16% in the healthy infants, q-value < 0.01) accumulated in the OP microbiota in the diseased infants ( Fig. 2-C, Supplementary Table).

Treatment-induced changes in the FM/OP microbiota were few and individual-specific. The
FM and OP microbiota structures of clinically recovered patients were similar to that before therapy (Figs 3-4). A comparative analysis of each case was also conducted to understand the microbiota changes in each hospitalized infant ( Supplementary Figures 1-2).
Haemophilus influenzae or Streptococcus pneumoniae were identified in sputum cultures from 5 patients, whose FM diversity was similar to that of the other patients (Supplementary Figure 1A, Supplementary Table). Only two diseased children (patients 6 and 7) received oral antibiotics (Supplementary Table), and their FM diversity significantly decreased after therapy (Supplementary Figure 1B). By contrast, a comparable or higher FM diversity was identified in the other patients (Supplementary Figure 1A) who received intravenous or no antibiotic therapy (Supplementary Table). Klebsiella, Enterococcus, Bacteroides, Cronobacter and/or Clostridium dominated the FM in the diseased infants and did not change after the different therapies (Supplementary Figure 1B). After clinical therapy, the OP microbiota diversity in the hospitalized infants increased or remained the same,  Figure 2B). Bacillus, Veillonella, Prevotella and Raoultella also dominated after treatment (Supplementary Figure 2B).

Discussion
Bronchiolitis is mainly caused by viral infections and co-infections with bacterial pathogens 2, 3 . Viral and bacterial infections have been shown to induce an immune response and microbiota imbalance 21 . This study revealed an imbalanced FM/OP microbiota in children with viral bronchiolitis, which is consistent with that identified in other respiratory diseases, such as asthma and cystic fibrosis 22,23 .
The distal effect of the gut microbiota (GM) on respiratory health is profoundly affected by the regulation of the host immune system 23 . The GM plays a crucial role in the immune response to and protection from pulmonary infection [24][25][26][27][28] . Bifidobacterium and Lactobacillus tend to accumulate in healthy infants but not in significant amounts. Bifidobacterium spp. was found to be protective against both bacterial and viral infections of the respiratory tract 27,29,30 . Probiotics composed of Bifidobacterium and Lactobacillus are also promising for the control of respiratory infections [31][32][33][34] . In the healthy children, enriched Veillonella was positively associated with Th17-mediated immunity in the lungs 35 and negatively associated with the risk of asthma 36 . Bacteroides and Streptococcus accumulated in significant amounts and represented a high proportion of FM in diseased children. Several members of Streptococcus spp. could induce inflammation in the epithelial mucus 37 , including the insignificantly enriched Klebsiella in diseased infants 38 . By contrast, various Bacteroides spp. have the potential to relieve inflammation by expanding the T reg cell population and suppressing inflammatory responses 23 . The aforementioned prior study could partially explain the GM imbalance observed in the diseased children. In all patients, except for patients 6 and 7, who took oral antibiotics, the FM diversity either remained unchanged or decreased after the treatment. This finding indicated the need to decrease exposure to oral antibiotics, which could block GM recovery for several months 39 .
Numerous reports have described the OP microbiota in healthy children, which was primarily dominated by Streptococcus, Rothia, Prevotella, Gemella, Veillonella, Fusobacteria, Haemophilus, and Neisseria 4 . Nader Shaikh et al. also reported prevalent streptococcal carriage in children without pharyngitis symptoms 40 . We found microbial carriage in the OP in the healthy infants that was identical to that described in these reports 4,40 . However, the OP microbial structures in the hospitalized infants significantly differed from those in the healthy infants, which potentially indicates transmission to the lung 19,20 . Several reports have identified four microbiota profiles in the NP in infants hospitalized for bronchiolitis, and the Haemophilus-dominated profile was associated with the highest clinical severity 14,15 . We established a Streptococcus-dominant OP microbiota profile in both the hospitalized and healthy infants, which could be attributed to the differing microbiota components between the NP and OP 19,20 . Moreover, the OP microbiota remained unchanged in most patients who recovered clinically. This finding may explain the widespread repeated respiratory infections in children after therapy because high incidences of respiratory diseases have been reported in the presence of "unstable" URT microbiota 8,41 . Therefore, the OP microbiota has promising potential in bronchiolitis prognosis, therapy optimization, and evaluation of recovery.
This study also had some limitations. A total of 47 infants was not sufficient for partitioning information according to discrepant clinical symptoms, such as that performed by Carlos A and Camargo Jr. et al. 14,15 . The detailed clinical symptoms and host responses, including immune cytokines, should be considered to assess the contributions of microbiota. Long-term investigations exploring the incidence of acute respiratory infection in infants who recovered clinically should also be performed. In conclusion, this study reviewed the imbalanced FM and OP microbiota in Chinese children with bronchiolitis and provided additional reference data for associated studies. More importantly, we identified comparable OP and FM microbiota among patients before and after treatment. This finding suggested a risk model of OP microbiota in children who recover clinically 42 .

Materials and Methods
Ethics statement. This study was approved by the Ethical Committee of Shenzhen Children's Hospital under the registration number 2015020. All procedures were performed in accordance with the relevant guidelines and regulations stipulated by the Ethical Committee of Shenzhen Children's Hospital. We obtained written informed consent from the parents of all participants, who approved their children's participation in the study.
Sample preparation. In total, 27 inpatients (aged ≤1 year of age) were sampled within 24 h of hospitalization (before therapy) in Shenzhen Children Hospital, and the second sampling was conducted after clinical recovery from bronchiolitis (average 7-10 days after treatment). In addition, 22 age-matched healthy infants were recruited according to the following inclusion criteria: no wheezing, fever, cough, or other respiratory/allergic symptoms at the time of sampling and for 2 weeks prior to the study and no respiratory symptoms for 1 week after sampling. None of the infants was exposed to antibiotics for two weeks before sampling. Sterile oropharyngeal swabs (155 C, COPAN, Murrieta, California, USA) were used for sampling the OP and faeces. For the diseased infants, we performed the sample collection during the following two time points: 24 hours after hospitalization and upon clinical recovery (average 7-10 days after treatment). The collected samples were immediately stored at −80 °C, and the DNA extraction was performed within one week. The sputum samples were collected by performing endotracheal suctioning (AARC (American Association for Respiratory Care) Clinical Practice Guidelines) 43 and cultured for several bacterial pathogens.
DNA extraction, sequencing, and analysis. DNA was extracted using a PowerSoil ® DNA Isolation Kit (Mo Bio Laboratories) according to the manufacturer's protocol. The hypervariable V3-V4 region of the 16S rRNA gene was amplified as previously reported 44 and was sequenced using an Illumina MiSeq platform. The paired-end reads were filtered using Mothur's Miseq SOP 45 and connected to tags using FLASH (v1.2.11, http:// ccb.jhu.edu/software/FLASH/index.shtml). The qualified tags were clustered into operational taxonomic units (OTUs), and OTUs from chimaeras were removed using USEARCH (v7.0.1090). The OTUs were assigned a taxonomic classification by aligning to the RDP 16S rRNA database (201408).
The multivariable analysis was conducted via a PERMANOVA (permutations = 9999, P-value ≤0.05) using the R package "vegan" (version 2.4-3) 46 to identify the important factors that could be associated with the microbiota structure. The Wilcoxon rank-sum test was used to compare the inpatients to the healthy infants, and a time-series comparison of the hospitalized infants was performed using the Wilcoxon signed-rank test. P-values were adjusted according to the false discovery rate for multiple tests. *, ** and *** represent q-values ≤0.05, ≤0.01 and ≤0.001, respectively. Increases/decreases in the Shannon index of FM and OP microbiota diversity >50% were considered significant. R (version 3.2.3) and SVG (version 1.1) software packages were used for visualization.
Accession number. Clean reads were deposited in the GenBank database under accession number PRJNA362484.