Analysis of lung microbiota in pediatric pneumonia patients using BALF metagenomic next-generation sequencing: A retrospective observational study

Abstract

The contribution of the lung microbiota to pneumonia in children of varying severity remains poorly understood. This study utilized metagenomic next-generation sequencing (mNGS) technology to elucidate the characteristics of lung microbiota and their association with disease severity. This retrospective study analyzed bronchoalveolar lavage fluid (BALF) mNGS data of 92 children diagnosed with pneumonia between January 2021 and July 2022. A comparative analysis of the lung microbiota was conducted between the severe pneumonia (SP) (n = 44) and non-severe pneumonia (NSP) (n = 48) groups. Compared to conventional microbiological tests (CMT), mNGS had a higher positivity rate in etiology detection (68% vs 100%). In the NSP group, the predominant type of infection was Mycoplasma pneumoniae single infection, whereas in the SP group, the main type involved a combination of M pneumoniae and bacterial infection. The top 3 identified microbial taxa in both the groups were M pneumoniae, Rothia mucilaginosa, and Schaalia odontolyticus. Although there were no significant differences in the α and β diversity of the lung microbiota between the SP and NSP groups, the abundance of M pneumoniae was higher in the SP group (P = .053). Spearman analysis indicated a highly significant positive correlation between the abundance of Prevotella melaninogenica and M pneumoniae (P < .001). Our analysis identified an association between M pneumoniae infections and disease severity. This study provides a foundation for a better understanding of the pathogenesis of pediatric pneumonia and the relationship between microorganisms.

1. Introduction

Pneumonia is 1 of the leading causes of death in children under 5 years of age.^[1] In 2015, approximately 700,000 children under 5 years of age died of pneumonia in the world.^[2] Pneumonia caused by viruses, bacteria, parasites, and fungi may vary due to differences in environmental exposure, age, patient condition, and diagnosis methods.^[3] Changes in lung microbiota are closely related to the progression or prognosis of pneumonia.^[4,5] In pediatric pneumonia, Streptococcus pneumoniae, Haemophilus influenzae, Mycoplasma pneumoniae, Staphylococcus aureus and Klebsiella pneumoniae are widely considered the main pathogens.^[6] Among these, S pneumoniae infection is the main cause of death. The incidence of SP caused by M pneumoniae has also increased in recent years. The survival rate of patients with SP is 30% to 60%.^[7]

With the rapid development of sequencing technology, metagenomic next-generation sequencing (mNGS) provides a novel method for the rapid diagnosis of infectious diseases. Traditionally, healthy lungs have been thought to be sterile.^[8] With the progress in respiratory microbiome studies, many studies have shown that lung microbiota plays an important role in lung diseases (such as pneumonia, septicemia, and bronchiectasis).^[9,10] However, clinical studies on mNGS have mainly focused on pathogen detection ability, with limited attention to the role of other respiratory commensal microbes detected in patients. Previous studies have found that bronchoalveolar are rich in oral commensal microbes in half of healthy people, which is related to the increased lymphocyte and neutrophil counts in bronchoalveolar lavage fluid (BALF).^[11] Researchers have found that the microbiota changes significantly during pneumonia and is strongly associated with the prognosis or course of the disease.^[12]

Some normal microbes in the upper respiratory tract (URT) can cause disease under certain conditions; however, the complex interactions between microbes and hosts are influenced by many factors.^[13] Recent studies have also shown that microbial-microbial interactions significantly influence the etiology, pathogenesis, and frequency of respiratory system diseases.^[14,15] The characteristics of the pediatric lung microbiota are significantly associated with disease severity.^[14,15] Moreover, there are close interactions between microbes, including commensal and competitive interactions.^[16] Ecological imbalance of the respiratory microbiota can lead to increased colonization by opportunistic pathogens and respiratory tract infections.^[17] Compared with the rapid development of the intestinal microbiome, lung microbiota studies are still in their infancy, especially for studies in children with lower respiratory tract infections.^[8] Therefore, we conducted a retrospective study to investigate the characteristics of lung microbiota between severe pneumonia (SP) and non-severe pneumonia (NSP) in children, hoping to discover the potential mechanism of microbial interaction in children’s pneumonia and provide new insights into the etiology of children who did not find pathogens through mNGS in clinical settings.

2. Methods

2.1. Patient enrollment and study design

This retrospective cohort study was conducted at Hebei Children’s Hospital, where hospitalized patients with pneumonia in the respiratory department between January 2021 and July 2022 were enrolled. Records of a total 112 patients were investigated, and 92 were finally included in the study (Fig. 1). The diagnosis of pneumonia in children was based on the Chinese Guidelines for Management of Community-Acquired Pneumonia in Children (revised edition of 2013).^[18] The criteria included: the children presented with typical clinical signs of pneumonia, such as fever, cough, sputum, and dyspnea; the children had signs such as rapid breathing and moist rales; and the syndrome of pulmonary infection was supported by radiological evidence (e.g., chest computed tomography scan). SP was diagnosed in children with poor general condition, altered consciousness, hypoxemia, high fever, dehydration, or refusal to eat, as well as imaging manifestations such as lobed lung, pleural effusion, lung abscess, lung necrosis, and extrapulmonary complications. Children without any of these conditions were included in the NSP group. All patients underwent simultaneous mNGS and conventional microbiological testing (CMT), and follow-up analysis was conducted based on the above classification. The exclusion criteria were as follows: age > 18 years; no pathogenic microorganism detected by mNGS; and incomplete clinical data. The research program was approved by the Ethics Review Committee of Hebei Children’s Hospital (No. 202139). Due to the retrospective nature of the study, the Ethics Committee waived the requirement for patient consents. The patients were anonymized, and their information was nonidentifiable.

Figure 1.

Schematic diagram of experimental grouping in this study. Out of a total of 112 confirmed patients, 92 met the inclusion criteria for pneumonia and were included in this study. These patients were divided into 2 groups: individuals with NSP and those with SP. NSP = non-severe pneumonia, SP = severe pneumonia.

2.2. Data collection

The clinical data collected for the records extracted included age, sex, total hospitalization time, clinical characteristics, underlying medical conditions, and infection type. All patients underwent simultaneous examination using BALF mNGS and CMT. CMT include BALF culture, sputum culture, and multiplex reverse-transcription polymerase chain reaction (RT-PCR). Identification of pathogenic microorganisms was based on diagnoses made by 2 experienced clinicians, considering the patients’ clinical presentations, laboratory tests, mNGS results, and radiological evidence (Blauwkamp et al, 2019).

2.3. Bronchoalveolar lavage fluid collection and inspection

BALF collection was accomplished by strictly adhering to operating procedures. All surgeries were performed by experienced and skilled respiratory intervention physicians. The patients’ nasal or oral cavities were cleaned with normal saline solution 10 minutes before the bronchoscopy examination, and 2% lidocaine (Kelun, China) was administered for local anesthesia. Conscious intravenous sedation with midazolam was then administered. During the operation, the patients received oxygen therapy (2–3 L/min) while monitoring blood oxygen saturation. Topical anesthesia of the larynx, trachea, and carina was achieved with 2% lidocaine, the bronchoscope was wedged in the lesion’s segment or lobe, and lavage was performed with 10 mL of sterile saline, with a suction pressure of 100 to 120 mm Hg. All BALF samples were immediately processed and stored according to the laboratory requirements.

2.4. Metagenomic next-generation sequencing process

The collected BALF samples were immediately transported to WillingMed Technology (Beijing) Co. Ltd. for mNGS testing. The mNGS test included the following steps: DNA was extracted using the PathoXtract^® Basic Pathogen Nucleic Acid Kit (WYXM03211S; WillingMed Company, Beijing, China).^[19] Illumina^® DNA Prep, (M) Tagmentation (20,018,705, Illumina, San Diego) was used to build the DNA libraries, followed by sequencing of the libraries on NextSeq™ 550Dx (Illumina) with a length of 75 bp (base pair). At least 20 million sequencing reads were obtained for each sample. The original FASTQ format data were quality-controlled and evaluated to filter out low-quality or undetected sequences, splice-contaminated sequences, high-coverage repetitive sequences, and short-read long sequences. Bowtie 2 was used to compare the high-quality sequencing data with the human reference genome GRCh37 (HG19), enabling the removal of human host sequences.^[20] The remaining sequences were aligned with a reference database previously constructed by Kraken2.^[21] The genomes of the microorganisms were obtained from the National Biotechnology Information Center (ftp://ftp.ncbi.nlm.nih.gov/genomes/), and the clinical application-level reference genome database was constructed through genome screening and verification. The identified species-specific reads number was normalized to reads per ten million (RPTM) value to determine the positive results.^[19] Positive microbes were defined as potentially pathogenic or typically nonpathogenic based on a previous report that summarized lists of potentially pathogenic bacterial, fungal, and viral genera that have been associated with human pulmonary disease.^[22]

2.5. Statistical analysis

To analyze the sequencing abundance of microorganisms, we processed the RPTM values with log10 and analyzed them using Graphpad Prism8 (USA) and SPSS 25.0 (USA). Statistical significance was set at P < .05. The α diversity within a group was calculated using filtering and normalized counting to determine the Shannon, Simpson, Chao 1, and ACE indices. Principal component analysis (PCA) was performed using the Tutools platform (http://www.cloudtutu.com) to analyze the differences between groups. Linear discriminant analysis (LDA) effect size (LEfSe) was used to identify significant differences between groups. Correlation analysis and visualization were performed using the R software (v4.2.2).

3. Results

3.1. Baseline characteristics of participants

Between January 2021 and July 2022, 112 patients with pneumonia were diagnosed at the Respiratory Department of Hebei Children’s Hospital. Ninety-two patients who met the eligibility criteria were included in the study and classified into 2 groups based on clinical diagnosis: NSP (n = 48) and SP (n = 44) (Fig. 1). Although there were no significant differences in the number of patients, age, or sex between the 2 groups, the SP group had a significantly longer hospital stay than the NSP group (P < .001). Fever symptoms were significantly more prevalent in the SP group than in the NSP group (P = .032). Additionally, the occurrence of pleural effusion was significantly higher in the SP group than in the NSP group (P < .001). However, there was no significant difference in the number of patients experiencing other symptoms such as cough or dyspnea. The primary underlying medical conditions, pneumonia/bronchopneumonia, and bronchial asthma did not differ significantly between the 2 groups (Table 1). M pneumoniae pneumonia was the main clinical diagnosis in both groups, but the SP group had a higher proportion of patients diagnosed with plastic bronchitis and necrotizing pneumonia, while the NSP group had a higher proportion of patients diagnosed with bronchopneumonia (Table 1).

Table 1.

Demographics and clinical characteristics of study patients.

Variable	NSP (n = 48)	SP (n = 44)	P value
Gender, Male (%)	23 (47.92%)	24 (54.55%)	.298
Age, yr (mean ± SD)	5.51 ± 3.37	6.54 ± 3.52	.155
Hospital, d (mean ± SD)	8.42 ± 4.38	14.25 ± 8.52	<.001***
Clinical presentations
Fever (%)	37 (77.08%)	41 (93.18%)	.032*
Cough (%)	48 (100.00%)	44 (100.00%)	.000
Pleural effusion (%)	7 (14.58%)	30 (68.18%)	<.001***
Respiratory distress (%)	0 (0.00%)	1 (2.27%)	.299
Underlying medical condition
Pneumonia/bronchopneumonia (%)	6 (12.50%)	1 (2.27%)	.113
Bronchial asthma (%)	1 (2.08%)	2 (4.55%)	.605
Epilepsy (%)	0 (0%)	2 (4.55%)	.226
Acute myeloid leukemia (%)	1 (2.08%)	0 (0.00%)	>.999
Hepatic transplantation (%)	1 (2.08%)	0 (0.00%)	>.999
Other† (%)	0 (0.00%)	4 (9.09%)
Type of infection
Bacterial infection (%)	6 (12.5%)	2 (4.55%)	.180
Mycoplasma infection (%)	23 (47.92%)	11 (25.00%)	.023*
Bacterial and fungal co-infection (%)	0 (0.00%)	1 (2.27%)	.299
Bacterial and Mycoplasma co-infection (%)	5 (10.42%)	24 (54.55%)	<.001***
Viral and Mycoplasma co-infection (%)	0 (0.00%)	2 (4.55%)	.138
Bacterial, fungal and Mycoplasma co-infection (%)	0 (0.00%)	1 (2.27%)	.299
Clinical diagnosis
Mycoplasma pneumoniae pneumonia (%)	28 (58.33%)	38 (86.36%)	.0048**
Plastic bronchitis (%)	1 (2.08%)	13 (29.55%)	.0002***
Necrotizing pneumonia (%)	1 (2.08%)	11 (25.00%)	.0001***
Bronchopneumonia (%)	7 (14.58%)	0 (0.00%)	.0128*
Urticaria (%)	1 (2.08%)	1 (2.27%)	>.9999
Tuberculosis (%)	1 (2.08%)	1 (2.27%)	>.9999
Bronchitis (%)	1 (2.08%)	0 (0.00%)	>.9999
Other‡ (%)	0 (0.00%)	4 (9.09%)

Abbreviations: NSP = non-severe pneumonia, SP = severe pneumonia.

^† Other underlying medical conditions included chronic granulomatous disease, encephalitis, neonatal asphyxia, postoperative congenital heart disease, with 1 patient in each condition.

^‡ Other clinical diagnosis included bronchiolitis obliterans, pulmonary abscess, thrush, and bronchiectasis, with 1 patient in each diagnosis.

^* Represents P value < .05.

^** P value < .01.

^*** P value < .001.

Based on the etiological diagnosis by professional clinical doctors, the number of patients with M pneumoniae infection in the NSP group was significantly higher than that in the SP group (P = .023). Conversely, the number of patients with bacterial and Mycoplasma co-infection was higher in the SP group (P < .001). There was no statistically significant difference in the number of patients with other types of infections (bacterial infections and other co-infections; Table 1).

3.2. Comparison of mNGS and CMT in pathogen detection

All 92 patients underwent mNGS of BALF and CMT. In 68% (63/92) of the patients, pathogens were detected using both mNGS and CMT, whereas in the remaining 32% (29/92) of the patients, pathogens were only detected using mNGS. Furthermore, among the 63 patients who tested positive using both methods, the results were completely consistent in 33 (36%) patients, partially consistent in 26 (28%) patients, and completely inconsistent in only 4 (4%) patients (Fig. 2A).

Figure 2.

Comparison of mNGS results with CMT. (A) The consistency comparison between mNGS and CMT. (B) Pathogen distribution comparison between mNGS and CMT. CMT = conventional microbiological tests, mNGS = metagenomic next-generation sequencing.

Among the 92 patients, the most frequently detected pathogen was M pneumoniae, accounting for 77.17% (71/92) of the total number of patients (Fig. 2B). A total of 34 species were identified, with bacteria being the most predominant pathogen. This included 9 species of Gram-positive bacteria (9/34, 26.47%), with S pneumoniae being the most prevalent, and 14 species of Gram-negative bacteria (14/34, 41.18%), with H influenzae being the most common. Additionally, there were 5 types of fungi (5/34, 14.71%).

3.3. Characteristics and differences of lung microbiota between the SP and NSP group

To investigate the characteristics and correlations of the lung microbiota in children with pneumonia, we analyzed the results of the BALF mNGS. A total of 104 microorganisms were identified in 92 children (Fig. S1, Supplemental Digital Content, http://links.lww.com/MD/O130). The 5 most abundant species were M pneumoniae, Rothia mucilaginosa, Schaalia odontolyticus, Veillonella parvula, and Granulicatella adiacens, with M pneumoniae being the most common species in both the SP and NSP groups. The 5 most abundant species in the SP group were M pneumoniae, R mucilaginosa, V parvula, Actinomyces graevenitzii, and S odontolyticus. In the NSP group, M pneumoniae, R mucilaginosa, S odontolyticus were also among the 5 most abundant species, with G adiacens and Granulicatella elegans differing from the SP group.

To assess the diversity and richness of the lung microbiome community, we used the Shannon index, Simpson index, Chao1 index, and ACE index. At the species level, there were no significant differences in Shannon and Simpson indices between the SP and NSP groups (Fig. S2A and B, Supplemental Digital Content, http://links.lww.com/MD/O130). In terms of community richness, the Chao1 and ACE indices of the SP group were higher than those of the NSP group; however, the difference was not statistically significant (Fig. S2C and D, Supplemental Digital Content, http://links.lww.com/MD/O130). PCA reflected the β-diversity of the different groups. Samples from different groups partially overlapped, but the intragroup differences in the SP group were smaller than those in the NSP group, and there was no significant difference in β-diversity (Fig. S2E, Supplemental Digital Content, http://links.lww.com/MD/O130).

We further investigated whether there were significant differences in species between the 2 groups using LEfSe analysis and found that the LDA score of M pneumoniae in the SP group was higher than 5 points and that of S aureus was higher than 4 points (Fig. 3A). Moreover, the sequence number of M pneumoniae in the SP group was higher than that in the NSP group (P = .053), but this difference was not statistically significant (Fig. 3B).

Figure 3.

The species differences between the SP and NSP groups. (A) The taxonomic composition differences using LEfSe analysis (LDA Score > 4, P < .05). (B) The differences of Mycoplasma pneumoniae between the 2 groups. The sequence number (RPTM) was calculated with the value of Log 10. LDA = linear discriminant analysis, LEfSe = linear discriminant analysis effect size, NSP = non-severe pneumonia, RPTM = reads per ten million, SP = severe pneumonia.

3.4. Correlation analysis of microbial species

To investigate the potential correlations among microbial species in the lungs of children, we selected the top 30 species based on their abundance > 0.1% and conducted Spearman correlation analysis (Fig. 4). In our study, M pneumoniae was the most abundant species and was a microbial marker of SP. There was a highly significant positive correlation with Prevotella melaninogenica (P < .001). H influenzae and Bacillus cereus were negatively correlated with M pneumoniae but the difference was not statistically significant.

Figure 4.

Spearman correlation among the top 30 microbial species in terms of total richness. * represents P value < .05; **P value < .01; ***P value < .001.

4. Discussion

In this study, we investigated the lung microbiota in BALF samples of 92 children with pneumonia using mNGS. Although there have been numerous studies on the microbiota of patients with pneumonia, there is a lack of research specifically focusing on children. This may be attributed to the challenges in collecting samples from the lower respiratory tract of the children. Recent studies have revealed disparities in pathogen load and microbial composition between the lungs and URT in children.^[15] Our study aimed to explore the interaction between microorganisms in the lungs and their correlation with disease severity. Additionally, we aimed to investigate the relationship between commensal and pathogenic microorganisms in an area that remains contentious within the field. Previous research has shown that intricate microbial communities in the lower airway can either enhance or impede the ability of microbes to cause infection. These commensal bacteria may lead to pathogenic processes by inducing inflammation.^[23]

Our findings indicated that BALF mNGS was more effective in identifying the etiology of pediatric pneumonia than CMT. In a previous nationwide surveillance study conducted in China over an eleven year period, the top 5 bacterial pathogens among children with acute respiratory infections were S pneumoniae, H influenzae, M pneumoniae, S aureus, and K pneumoniae.^[6] Among school-aged children, M pneumoniae was the most prevalent, while others were slightly altered. In the present study, the top 5 bacterial pathogens were M pneumoniae, S pneumoniae, H influenzae, Streptococcus pseudopneumoniae, and S aureus (Fig. 2). The pathogen distribution was similar to that in a previous report, except for the detection of a different pathogen. However, the most notable disparity between our study and the previous report is the relatively high detection rate of M pneumoniae. This could be attributed to the composition of our study population, which mainly consisted of children with pneumonia, and thus had a higher proportion of severe cases. In a study conducted on children with community-acquired pneumonia, M pneumoniae was the most frequently detected pathogen using BALF mNGS, with a detection rate of 73.55%, which is similar to that in our study (77.17%).^[24] These findings suggest that the distribution of pathogens may vary among patients with different disease severity.

Our findings indicate that there was no significant difference in microbial diversity between children with severe and NSP. However, we observed greater similarity in the microbiota of children with SP. LEfSe analysis revealed that M pneumoniae and S aureus as microbial markers in the SP group. M pneumoniae was the most abundant bacterium in the SP group. Moreover, it exhibited a positive correlation with P melaninogenica, a bacterium predominantly found in the human oral cavity with a known influence on airway microbiota dysbiosis and interaction with opportunistic pathogens.^[25] These findings suggest that P melaninogenica and M pneumoniae may coexist or interact in children with pneumonia, potentially influencing the severity of the disease. Recent research further supports the consistent relationship between the abundance of M pneumoniae abundance and disease severity, particularly in cases involving respiratory diseases compounded by co-infection.^[15] These findings align with our initial demographic assessment, which indicated that SP in patients primarily results from co-infection with M pneumoniae.

Using BALF mNGS, we identified the pathogens responsible for the pulmonary infections in children. Additionally, we investigated the differences in microbial composition among the various groups based on the information provided by mNGS. Despite the potential interference from colonizers, BALF has significant advantages over other respiratory samples for pathogen detection. However, obtaining BALF samples from healthy children as controls is challenging. This limitation hinders our investigation of the pathogenic and commensal species in the lung microbiota of children with pneumonia. Moreover, the relatively small number of enrolled patients may have introduced a bias into the results. Another limitation is the relatively high detection rate of M pneumoniae, which reduces the diversity of patient cohorts. Further large-scale research is necessary to gain a more comprehensive understanding of the association between microbiome and disease severity. This will facilitate the utilization of mNGS to determine the etiology and control disease progression.

5. Conclusions

This study used mNGS to characterize the composition of the lung microbiota in children with pneumonia. The analysis revealed M pneumoniae as the predominant microorganism in the lung microbiota of both the SP and NSP groups. Notably, M pneumoniae was more abundant in the SP group and served as a biomarker of SP. Furthermore, a strong correlation was observed between P melaninogenica and M pneumoniae in children with pneumonia. These findings offer new insights into the relationship between microbial communities and the severity of pneumonia, shedding light on the potential involvement of specific microorganisms in disease pathogenesis.

Acknowledgments

We thank all clinicians who provided the diagnostic data of the patients for this study. We thank WillingMed Technology (Beijing) Co. for providing technical support for mNGS.

Author contributions

Conceptualization: Wei Gai, Xiaowei Cui.

Data curation: Sukun Lu, Ling Sun, Lijie Cao, Mengchuan Zhao, Yuxin Guo, Mei Li, Suxia Duan, Yu Zhai, Xiaoqing Zhang, Yakun Wang.

Formal analysis: Lijie Cao.

Funding acquisition: Xiaowei Cui.

Investigation: Ling Sun, Yakun Wang.

Methodology: Sukun Lu, Ling Sun, Xiaoqing Zhang.

Project administration: Wei Gai, Xiaowei Cui.

Resources: Suxia Duan.

Software: Yuxin Guo.

Visualization: Yuxin Guo.

Writing – original draft: Sukun Lu, Mengchuan Zhao.

Writing – review & editing: Yuxin Guo, Wei Gai, Xiaowei Cui.