Evaluating the Health Risks of Pneumonia from Airborne Bacterial Communities Using 16S rDNA Sequences of Pneumonia-related Pathogens

Pneumonia affects approximately 7% of the world's population resulting in approximately 4 million deaths per year^{[1, 2]}. The 2010 Global Burden of Disease Study reported that lower respiratory tract infections, including pneumonia, are the fourth most common cause of death globally, and they are the second most frequent reason for reduced years of life^[3]. Bacteria are essential infectious agents that cause pneumonia^{[4, 5]}. However, the etiology of the disease is identified in only approximately 50% of community-acquired pneumonia cases^[6] and in only approximately 36% of nosocomial pneumonia cases^[7]. This unidentified etiology may be associated with unidentified pneumonia-causing microbes that cannot be isolated by traditional culture methods used routinely in clinical diagnostic laboratories^[8].

The environment plays an essential role in the development and spread of pneumonia. Residents exchange microbes with air through breathing that allows large numbers of microbes to enter the respiratory tract. Indoor dust and outdoor particulate matter entrained in air and purification equipment inhabited by microorganisms in enclosed spaces may be a source of pneumonia-causing bacteria. The evaluation of the environmental risk of pneumonia is of great significance for the prevention of pneumonia and improving quality of life. The development of high-throughput sequencing has allowed the detection of an astonishing number of microorganisms. Perhaps as many as 90% of environmental microbes cannot be cultured^{[9, 10]}, and only a small fraction can be cultured with routine methods^[10]. Several previous studies report that microbial concentrations measured in bioaerosols by cultivation represent only one in 1,000 microbes present^[11]. In this study, 16S rDNA sequences of common pathogenic bacteria that cause pneumonia were used to evaluate risks by calculating evolutionary relationships between these pathogens and bacterial species in environmental samples. The validity of the method was verified with simulated microflora and using data on community-acquired pneumonia (CAP) registered in hospitals.

We aimed to evaluate the risks of pneumonia from airborne bacterial communities. Genomic sequences of closely related bacteria are very similar^[12]. We used the evolutionary distances between known pneumonia-related pathogens and environment species to evaluate health risks. That is, the closer the evolutionary distance, the greater the health risk. An operational taxonomic unit (OTU) is a designation used to classify groups of closely related individuals. OTUs are the most commonly used units of diversity when analyzing small subunit 16S of prokaryotes or 18S rRNA for eukaryotes as marker sequences. OTUs are pragmatic proxies for 'species' in environmental research. The pathogenicity of unknown species increases sharply as the evolutionary distance between known pathogens and unknown species diminishes. Therefore, we used an exponential function. Next, we calculated the risk of each OTU in the sample compared with all pathogens selected. Finally, we calculated the risk of all OTUs in the samples combined with their relative abundance to yield an estimate of health risk of implied by samples using the G-Qin formula (Equation 1):

where Rbr is the risk of bacterial pneumonia, Dn is the evolutionary distance between OTU sequences in samples and each pathogen, and Am is the relative abundance of each OTU.

We identified ten common pneumonia-related bacterial pathogens from three clinical trials of tigecycline for the treatment of community-acquired bacterial pneumonia^[5]. These choices are representative of common pneumonia-causing bacteria. We obtained appropriate 16S rDNA sequences from NCBI (Table 1).

We selected a student dormitory at approximately 10 m above the ground as a sampling location (116°26′35.86″E, 39°52′22.30″N). We collected airborne bacteria twice a day using natural sedimentation from five sites—three different rooms, one corridor, and one balcony—from November 23 to December 27, 2016 (35 days). The balcony was located outside of the rooms. We placed four 90 mm uncovered, disposable, sterile cell culture dishes at each site at a height of 1.5 m and retrieved them 12 h later. Samples were collected by wiping the inner surface of dishes with a sterile swab dipped in 100 µL of sterile normal saline. One swab used to wipe two dishes was considered a single sample. Two samples were thus collected from each site every time. The swabs were individually stored in sterile 2 mL tubes at −80 ℃. Cell culture dishes for the next 12 h sampling period were placed at the time of sample retrieval. We placed the dishes and collected samples at 0800 and 2,000 h. Masks and gloves were worn during the sampling process. We used one sample of each site at each sampling time in subsequent research. A total of 215 samples were used for next-generation sequencing (NGS) of 16S rDNA, including all 70 samples from the balcony from November 23 to December 27 and 145 samples from December 9 to 27 from the other sites. All the bacterial 16S rDNA sequences were deposited in the National Center for Biotechnology Information Sequence Read Archive (http://www.ncbi.nlm.nih.gov/sra) under accession number PRJNA503826.

The sampling period was in winter, haze occurred frequently, and average temperature and relative humidity were 1 ℃ and 54% (Supplementary Table S1, available in www.besjournal.com). The average value of wind speed was 0.3–1.5 km/h indicating relative stable conditions (Supplementary Table S1). The average values of temperature and relative humidity indoors were 21.5 ℃ and 29.3%, respectively (data not shown).

Genomic DNA was extracted from each sample using a PowerSoil® DNA Isolation Kit (MO BIO Laboratories, Inc., Carlsbad, CA, USA) following the manufacturer's recommendations. Extracted DNA was diluted to a concentration of 1 ng/µL and stored at −20 ℃ until further processing. The V3-V4 variable region of the 16S rRNA genes was amplified (26 cycles) with universal primers (343F 5′-TACGGRAGGCAGCAG-3′, 798R 5′-AGGGTATCTAATCCT-3′)^[13]. A negative control in the same amplification system used sterile deionized water as a template to monitor contamination. Amplicon quality was visualized by gel electrophoresis. Amplicons were then purified with AMPure XP beads, amplified for another round of PCR (seven cycles), and purified again with AMPure XP beads. Final amplicon concentration was quantified using a Qubit^TM dsDNA Assay Kit. Equal amounts of the purified amplicons were then pooled for subsequent sequencing using a MiSeq Sequencing System (Illumina, Inc., San Diego, CA, USA). The PE300 sequencing model was used and paired ends were applied. Raw sequencing data were obtained in the FASTQ format. Paired-end reads were pre-processed using the Trimmomatic software to detect and cut ambiguous bases (N)^[14]. Low-quality sequences were filtered as previously described^[15]. Clean reads were subjected to primer sequence removal and clustering to generate OTUs using UPARSE software with a 97% similarity cutoff^[16].

We selected Beijing Chuiyangliu Hospital (1.6 km from the sampling sites), where patients experience no delay in registering to see a doctor, and actual disease occurrence is apparent. Data were collected from the fever clinic during the experiments to represent daily disease cases in the area. Records include information on sex, age, date of the hospital visit, and diagnosis (Supplementary Tables S1 and S2 available in www.besjournal.com). The hospital mainly serves the people in the surrounding area, and records show that patients live within 3 km of the hospital. Thus, we considered bacterial flora collected in our sampling to represent air quality within a 3 km radius. We screened cases from the infectious disease clinic between November 23 and December 27, 2016. Cases of viral infection were excluded. Remaining cases were classified as pneumonia, amygdalitis, pharyngitis, or bacterial infection. The daily pneumonia data were considered in the study (Supplementary Table S1).

We used the muscle function to align sequences and the dist.dna function to calculate the phylogenetic distance using model K81^[17] in the ape package in the R environment. The correlation between Rbr and daily pneumonia cases used the cor.test function with the Pearson method. Mann-Whitney U tests were conducted for comparisons between groups, and P-values of < 0.05 were considered statistically significantly different. A phylogenetic tree of artificial flora was inferred using the maximum likelihood method based on the Tamura-Nei model^[18]. Evolutionary analyses were conducted in MEGA5^[19]. We have uploaded the R script for Rbr calculation to 'https://www.researchgate.net/publication/344189028_Methods_for_assessing_the_pneumonia-associated_risk_of_the_airborne_bacterial_community_using_16S_rDNA_sequences_of_pneumonia-related_pathogens?showFulltext=1&linkId=5f59e5e0299bf1d43cf92085'.

We artificially formed 50 bacterial communities of representative pathogens (Supplementary Table S3, available in www.besjournal.com). Ten conformed to a Gaussian distribution (A1–A10), 10 to a uniform distribution (A11–A20), 10 to an exponential distribution (A21–A30), 10 to a Poisson distribution (A31–A40), and 10 to a Binomial distribution (A41–A50) (Supplementary Table S3). Probiotics benefit human health^[20-22]. Thus, we selected seven common probiotics to form additional artificial bacterial communities (Supplementary Table S4, available in www.besjournal.com). A total of 50 bacterial communities of the representative probiotics were formed, 10 for each distribution described above. These communities were numbers as above, but with 'B' designations (Supplementary Table S5, available in www.besjournal.com). Phylogenetic tree analysis of pathogens and probiotics showed no obvious aggregation of selected species (Figure 1A). This property excludes the possibility of similar phylogenetic distances in pathogen and probiotic groups. We then calculated Rbr for artificial pathogens and probiotic flora and compared values. Rbr was significantly less in probiotic than pathogen flora (P < 2.2 x 10⁻¹⁶) (Figure 1B). No report was found for selected probiotics as a causative agent for pneumonia was found, which is consistent with the Rbr result.

Figure 1.  Phylogenetic analysis of artificial flora (A) and comparison of Rbr values in probiotic and pathogen groups (B). The Mann-Whitney U-test was performed.

Species richness of airborne bacteria was substantial. The advent of high-throughput nucleotide sequencing has greatly enhanced our understanding of the diversity of airborne microorganisms^[23], yet we still could not identify all species in most cases. We collected 215 airborne microbe samples from indoor and outdoor environments by natural sedimentation. We analyzed the total abundance of OTUs greater than a threshold of relative abundance. A value of 0.8 for total abundance was located between the threshold 0.005 and 0.001 (Supplementary Figure S1, available in www.besjournal.com). We assumed that 80% of total abundance would include most dominant OTUs. Thus, 80% of OTUs were used to calculate Rbr indices.

Figure S1.  Total abundance of bacterial OTUs greater than various thresholds of relative abundance in 215 airborne samples.

We collected outdoor airborne samples between November 23, 2016, to December 27, 2016. Each sample was rarefied to 19,363 sequences and converted to relative abundance. We then calculated the Rbr of each sample using 80% of OTUs. Samples from the outdoor location could represent the bacterial community experienced by residents near the hospital, especially in the absence of a home air purifier^[24]. Correlation analysis revealed that the Rbr80 of airborne bacteria significantly and positively correlated with daily cases of pneumonia (P < 0.05) (Figure 2A). Animal model experiments show a delay between exposure and development of acute pneumonia^[25]. Additionally, pneumonia occurring within 48 h of hospital admission was likely CAP^[26]. Thus, we used a delay for the onset of disease for the correlation analysis. We found that Rbr80 positively correlated with the number of daily pneumonia cases using delays of 1 and 2 days (Figure 2B and 2C). A significant correlation was not observed with a delay of 3 days (Figure 2D). These findings support the conclusion that Rbr80 predicted an onset of pneumonia within approximately 3 days.

Figure 2.  Correlation analyses of Rbr80 and daily pneumonia cases after (A) 0, (B) 1, (C) 2, and (D) 3 days. Pearson correlation analysis was used.

The etiology of a large portion of pneumonia cases is unclear^[8]. Bacteria are important pathogens for airway inflammation, possibly leading to a cyclical disease^[27]. Thus, we reasonably propose a causal relationship between airborne bacteria and pneumonia (Figure 2). Airborne bacterial communities may be essential contributors to human illness, and Rbr80 may be a useful index to evaluate health risks from airborne bacteria.

We analyzed the distribution of the daily pneumonia cases in the sampling period. The numbers of days with four cases/day, five cases/day, and seven cases/day reached five during the sampling period (Figure 3A). Twenty days with daily cases less than eight were observed. Therefore, we assigned days with daily cases less than eight as a background group and days with daily cases more than seven as the Risk group (Figure 3A). We compared the Rbr80 between these two groups and found the values of Rbr80 in the Risk group were significantly higher (P = 0.0015) (Figure 3B). Thus, we assumed index values of 14.44 and 15.40 could be considered thresholds for airborne samples. Airborne samples with Rbr80 less than 14.44 might be considered safe (background pneumonia incidence), and Rbr80 greater than 15.40 might be considered high risk. Values between these values are likely low risk (Figure 3C). The Rbr80 in 93.8% of samples (15/16) in the Risk group were greater than 14.44, indicating accurate prediction for most airborne bacterial communities.

Figure 3.  The proposed threshold of Rbr80 to predict airborne bacterial risk related to pneumonia. (A) Distribution of the number of days with each value of daily pneumonia cases, (B) comparison of values of Rbr80 in Background and Risk groups, and (C) suggestion of Rbr80 value for evaluation. The Mann-Whitney U-test was used.

Numerous microorganisms exist in the air. They enter the body via respiration and might be beneficial, benign, or pathological. The identification of health risks posed by airborne microorganisms is typically based on concentrations of cultured microorganisms^[28]. However, numerous microorganisms cannot be cultured using standard laboratory methods^[29], limiting our understanding of microbial communities^[30] and the ability to evaluate their risks to health. Evidence indicates that exposure to high levels of airborne, noninfectious microorganisms may cause respiratory symptoms and disease among workers, such as farmers and sawmill workers^[31]. Numerous nonculturable microorganisms may also cause disease^[32]. New evidence show previously described inoperative Chlamydia abortusis can induce pneumonia^[33]. Thus, airborne bacterial community structure is also essential.

The advent of high-throughput nucleotide sequencing has greatly enhanced our understanding of airborne microorganisms^[23]. However, to our knowledge, no effective index exists for evaluating the health risks of airborne bacteria that considers the complexity of bacterial abundance and diversity. As indicated above, the use of the total number of culturable colonies determined in a contained (e.g., indoor) environment is not capable of evaluating the entire bacterial community. Thus, the Rbr index proves valuable for evaluating the health risk of airborne bacteria and is verified using clinical data. In our previous study, we confirmed that indoor airborne bacterial communities in homes without air purifiers closely tracked outdoor airborne bacterial communities^[24]. We did not survey the presence of air purifiers in the patients' homes, but patients were from the same population with a high proportion of rooms without air purifiers with indoor airborne bacterial communities similar to outdoor communities, and fewer rooms with air purifiers where indoor communities might reflect outdoor communities only during day time. Thus, the patient data likely reflects exposure to outdoor airborne bacterial.

People are exposed to various microbes in indoor environments affected by human occupancy, occupant behavior, and pets, as well as outdoor air^[34-37]. New ecological niches accompany living habits and architectural pattern changes. Therefore, Rbr may be successfully applied to the evaluation of bacterial communities in such niches, such as air purifiers, ventilation devices, and medical equipment. The application of Rbr80 will facilitate the prevention of bacterial pneumonia.

The value of the Rbr is limited because fungi, archaea, and viruses are not considered. Furthermore, the calculation of Rbr considers only evolutionary distances among species; horizontal transfer of genes encoding drug resistance and virulence were also not considered. Additionally, studies of airborne bacterial communities should include information about bacterial diversity and the effects of microbial metabolites, toxins, and microbial debris. Thus, the calculation of Rbr could be optimized by future studies.

Rbr index to evaluate the risk of airborne bacteria was developed based on the principle that closer evolutionary relationships reflect similar biology, using 16S rDNA sequences of pneumonia-related pathogens. The feasibility of Rbr80 was verified by both artificial flora communities and a case study. We proposed 14.44 and 15.40 values of Rbr80 as thresholds of safe, low risk, and high risk for pneumonia produced by airborne bacteria.

The authors declare no competing interests.