Insights into the Profile of the Human Expiratory Microbiota and Its Associations with Indoor Microbiotas.

Microorganisms residing in the human respiratory tract can be exhaled, and they constitute a part of environmental microbiotas. However, the expiratory microbiota community and its associations with environmental microbiotas remain poorly understood. Here, expiratory bacteria and fungi and the corresponding microbiotas from the living environments were characterized by DNA amplicon sequencing of residents' exhaled breath condensate (EBC) and environmental samples collected from 14 residences in Nanjing, China. The microbiotas of EBC samples, with a substantial heterogeneity, were found to be as diverse as those of skin, floor dust, and airborne microbiotas. Model fitting results demonstrated the role of stochastic processes in the assembly of the expiratory microbiota. Using a fast expectation-maximization algorithm, microbial community analysis revealed that expiratory microbiotas were differentially associated with other types of microbiotas in a type-dependent and residence-specific manner. Importantly, the expiratory bacteria showed a composition similarity with airborne bacteria in the bathroom and kitchen environments with an average of 12.60%, while the expiratory fungi showed a 53.99% composition similarity with the floor dust fungi. These differential patterns indicate different relationships between expiratory microbiotas and the airborne microbiotas and floor dust microbiotas. The results here illustrated for the first time the associations between expiratory microbiotas and indoor microbiotas, showing a potential microbial exchange between the respiratory tract and indoor environment. Thus, improved hygiene and ventilation practices can be implemented to optimize the indoor microbial exposome, especially in indoor bathrooms and kitchens.

Introduction

As important players in the mucosal immunity, microorganisms residing in the respiratory tract (respiratory microbiota, RM) are critical to human respiratory health and disease.[1−4] The formation and homeostasis of these RMs are modulated by various processes, including microbial immigration, elimination, growth, and death.[5] These processes indicate a continuous microbial communication between the respiratory tract and environment along with the exchange of carbon dioxide and oxygen.[6,7] Exhalation of the pathogenic RM from patients has a substantial role in spreading respiratory infectious diseases, for example, tuberculosis, measles, and the coronavirus disease 2019 (COVID-19).[8−11] In the context of COVID-19, even asymptomatic subjects could breathe out the pathogenic SARS-CoV-2 (severe acute respiratory syndrome coronavirus).[12] Apart from these pathogens, advanced high-throughput sequencing techniques (HSTs) have expanded our understanding of the RM compositions to include a diversified complex microbial community.[13−18] Accompanied by this expansion, our view on the role of the microbial exposome on human health has also shifted and gone beyond the germ theory of disease, which mainly focused on the pathogenic exposomes.[19] Exposure to the microbiota as a whole has been suggested to play a role in sustaining human health.[20,21] However, how will the exhaled RM affect the environmental microbial exposure is poorly understood, which might be important for the microbiological air quality and public health.

Respiratory activities including coughing, sneezing, singing, speaking, and tidal breathing release particles through various mechanisms, for instance, vocalization and breakup of film filaments or films during the airway closing and reopening processes.[22−25] The expiratory microbiota has been studied by collecting samples from these respiratory activities, followed by various analysis approaches. Among the various collection approaches, cooling and condensing-based exhaled breath condensate (EBC) collection is widely applied.[26] Specifically, respiratory viruses have been detected in EBC with the G-II sampler and the PKU Bioscreen sampler.[27−30] Moreover, emissions of millions of SARS-CoV-2 RNA copies per hour by COVID-19 patients have been revealed using the exhaled breath sampler.[31] Additionally, bacterial pathogens, for example, Haemophilus influenzae, Pseudomonas aeruginosa, and methicillin-resistant Staphylococcus aureus, have been detected by combining the breath sampler with cultivation, loop-mediated isothermal amplification, and silicon nanoware biosensors.[32−35] In addition, the fungal species Aspergillus sydowii and Cladosporium spp. were detected in EBC samples with the combination of traditional culture and DNA sequencing approaches.[36] Despite these known specific exhaled pathogens, knowledge on the microbial community diversity in the expiratory particles is still limited.

Convergence of expiratory microbiotas with environmental microbiotas can be related to the occupants’ microbial exposome, especially in indoor environments where people spend most of their time.[37,38] Various approaches have been applied to deciphering indoor microbiotas, which is critical to their control and potential engineering approaches.[38,39] By collecting various surface and air samples in indoor environments with and without occupants, the role of occupants in shaping indoor microbiotas has been revealed.[37,40−44] Millions of bacteria and fungi were released from one human occupant to the environment per hour.[45] In addition, statistical-based methods have been applied to quantitatively determine the contribution of humans to indoor microbiotas, for example, the Bayesian approach SourceTracker and the fast expectation–maximization microbial source tracking (FEAST), based on the similarity of microbial structures between the sink and the possible source samples.[39,46−50] Although the causality cannot be directly revealed, the application of these methods helps to fill the gaps regarding the relationships between the human microbiota and indoor environmental microbiotas. However, most of the studies highlighted the importance of human skin or human activity disturbance. Relationships between expiratory microbiotas and indoor microbiotas are poorly understood.

The major objective of this study was thus to explore the expiratory bacterial and fungal communities and investigate their associations with indoor microbiotas. First, expiratory bacterial and fungal communities were investigated by integrating the noninvasive exhaled breath sampler, HST technique, and statistical modeling tools. Second, human skin samples and various environmental samples, including air and surface samples, were microbially studied. Third, relationships between expiratory microbiotas and other microbiotas were assessed using FEAST. The results expand our understanding of the expiratory microorganisms from pathogens to more diversified unknown bacteria and fungi and further associate these expiratory microbiotas with indoor microbiotas, providing new insights into the human microbial exposome.

Materials and Methods

Sample Collection

The campaign was conducted in Nanjing, China, in the winter season between December 2018 and January 2019. The locations of these homes are shown in Figure S1. Information regarding the resident volunteers and housing conditions is summarized in Table S1. One EBC sample and one corresponding forearm skin sample were collected from one volunteer at each residence. Environmental samples from the residences were collected simultaneously, including indoor air of the living room (IAL), indoor air of the bathroom and kitchen (IABK), floor dust of the living room (DTL), floor dust of the bathroom (DTB), biofilm in the bathroom sink (BFB), biofilm in the kitchen sink (BFK), and outdoor air (OA). Fourteen residential homes were selected after removing three homes with incomplete samples. Altogether, a total of 120 samples were collected and analyzed. The detailed sampling procedures are described in the Supporting Information. This study was approved by the Biological and Medical Ethics Committee at Beihang University (No: BM20180042).

High-Throughput Sequencing of Bacteria and Fungi

DNA was extracted with the DNeasy PowerSoil Kit (Qiagen 12888-100, Germany). The purity and concentration of extracted DNA were quantified with a Qubit dsDNA assay kit (Life Technologies, Q328520, USA). Then, the quantified DNA samples were diluted to 1 ng/μL for subsequent PCR amplification. Specifically, the V3–V4 region of bacteria was amplified with the following primers with barcodes: 343F (5’-AGGGTATCTAATCCT-3′) and 798R (5′-TACGGRAGGCAGCAG-3′).[51] The ITS region of fungi was amplified with the following barcoded primers: ITS1F (5’-CTTGGTCATTTAGAGGAAGTAA-3′) and ITS2 (5′-GCTGCGTTCTTCATCGATGC-3′).[52−54] Sterile deionized water used for EBC collection and sample extraction was used as a blank control for DNA isolation and PCR. No gel band was observed for the blank control from the gel electrophoresis result. Then, the products from gel electrophoresis were purified with AMPure XP beads (Agencourt). The purified amplicons were amplified in a second round of PCR and purified again. After two rounds of purification, these amplicons were quantified again and pooled in equal amounts for sequencing.

High-throughput sequencing was performed with an Illumina MiSeq sequencing platform. The obtained raw paired-end reads were processed to cut off ambiguous bases using Trimmomatic (v0.35).[55] Then, the average quality score was evaluated using the sliding widow trimming approach, and sequences with scores less than 20 were excluded. Sequentially, FLASH (v1.2.11) was used to assemble paired-end reads with the following parameters: 10 bp of minimal overlapping, 200 bp of maximum overlapping, and 20% of maximum mismatch rate.[56] Clean reads were obtained from these assembled sequences using QIIME software (v1.8.0).[57] Amplicon sequence variants (ASVs) were produced with DADA2 within QIIME. For bacteria, annotation and BLAST searches were performed using the Silva database version 123 (or Greengens) and RDP classifier (the confidence threshold was 70%).[58] The Unite database was applied for fungal annotation.[59] Rarefication was performed with multiple_rarefactions.py within QIIME for all samples for subsequent analysis and comparison (Figures S2 and S3). The rDNA sequences were uploaded into the National Center for Biotechnology Information with the accession number PRJNA793288.

Statistical Analysis

The Sloan neutral model (SNM) was used to assess the role of stochastic processes in shaping the microbial structures in exhaled breath.[60,61] Specifically, communities in all EBC samples were pooled as a metacommunity, in which the relative abundance (RA) of each ASV was determined. Then, relationships between the RAs of all ASVs and the corresponding predicted occurrence frequency in the metacommunity were fitted. The rationale for this model is that the abundant taxa in the metacommunity have a high dispersal probability, resulting in a high occurrence frequency. In addition, a quantitative framework iCAMP (infer community assembly mechanisms by phylogenetic bin-based null model analysis) was applied to unravel the drivers in bacterial community assembly.[62] Due to the limitation of ITS primers used for fungi, iCAMP analysis was not applied for fungal community analysis.

The FEAST model was applied for association analysis between the expiratory microbiota and indoor microbiotas.[50] Compared to the commonly used Markov chain Monte Carlo (MCMC)-based Bayesian SourceTracker approach,[49] FEAST is 30–300 times faster and fits amplicon data. This advantage facilitates the usage of FEAST in fractioning the proportion of each source in the sink. A bidirectional analysis was performed to investigate the associations between different types of microbiotas by assuming either the microbiota in EBC or indoor microbiota as a sink and others as sources. This method was based on the structural similarities between sink samples (either EBC, or IABK, or IAL, or DTL, or DTB) and source samples (the remaining samples except the sink sample).

A nonparameterized Kruskal–Wallis test (KW test) was applied to assess the differences among multiple groups, and the Kolmogorov–Smirnov test (KS test) was applied to compare the differences between two groups (GraphPad Prism, USA). The Bray–Curtis dissimilarity index was applied to compare community structures between the expiratory microbiotas and the indoor microbiotas. A value of 1 for the Bray–Curtis dissimilarity index indicates that the two communities are completely different, while 0 indicates that the two communities are exactly the same. Principal coordinates analysis (PCoA) was used to visualize the relationships among the communities of different samples using the Bray–Curtis index (vegan and ape package in R 4.0.3). ADONIS (permutational multivariate analysis of variance using distance matrices) and ANOSIM (analysis of similarities) analyses were performed to evaluate the difference between two or more types of microbial communities using the Bray–Curtis index (vegan package). The relationships between the environmental factors and the microbial associations of EBC and others were assessed with redundancy analysis (RDA) (vegan package).

Results and Discussion

Profiles of EBC Microbiotas

Abundant and diversified microbiotas were observed in the EBC samples. Figure S4 shows the comparison between the diversity indices of microbial communities of EBC and other types of samples after rarefaction. The abundance and evenness of bacterial and fungal species in EBC were comparable to those in human skin samples (p > 0.05), and bacterial diversity was higher than fungal diversity. Distinct trends were observed when comparing microbial diversities in EBC with those in environmental samples. Nevertheless, it is clear that microbiota diversity in human exhaled breath was no less than in skin and environmental microbiotas and indicates that expiratory microbiotas could have strong relationships to indoor environmental microbiotas.

Five bacterial phyla accounted for more than 90% of the total taxa, including Firmicutes, Proteobacteria, Bacteroidetes, Actinobacteria, and Fusobacteria (Figure A). On average, the distribution of the top 16 bacterial genera in EBC was significantly different from those in skin and environmental samples (Figure B,C), as shown by the statistical tests with ANOSIM and ADONIS (Table S2).

Figure 1

Microbial compositions in EBCs of different subjects and other types of samples at the phylum and genus levels based on the sequences of the bacterial 16S rRNA gene and the fungal ITS1–2 region. Box-whisker plots show the RA of bacterial taxa at the phylum (A) and genus levels (B) and the RA of fungal taxa at the phylum (D) and genus levels (E) in the 14 EBC samples. Bar plots show the mean relative RA of the top 16 bacterial (C) and the top 15 fungal (F) genera in each type of microbial community. The color in panel C (F) corresponds to the color in panels B and A (D and E), representing the phylum information. BP: bacterial phylum; BG: bacterial genus; FP: fungal phylum; FG: fungal genus. EBC (sample size N = 14), skin (N = 13), IAL (indoor air of the living room, N = 14), IABK (IA of the bathroom and kitchen (BK), N = 14), DTB (floor dust of the bathroom, N = 13), DTL (DT of the living room, N = 14), BFB (biofilm of the bathroom sink, N = 14), BFK (BF of the kitchen sink, N = 14), and OA (outdoor air, N = 10).

Bacterial community structures in EBC at the phylum and genus levels are consistent with those reported in the respiratory microbiota.[13,14,63−65] Firmicutes has been found to be more abundant in the upper airway tract (URT) and is mostly driven by the genus Streptococcus.[64] In addition, the genera Neisseria (Proteobacteria) and Rothia (Actinobacteria) are common oropharynx bacteria.[66] In contrast, Bacteroides, Prevetella (Bacteroidetes), Haemophilus (Proteobacteria), and Fusobacterium (Fusobacteria) are common in the lower respiratory tract (LRT).[6] These results implied that the EBC bacterial community is a complex mixture of bacteria from the URT and LRT.

The bacterial genera Streptococcus, Bacteroides, and Haemophilus usually include more pathogenic species.[67] Due to the low accuracy of using 16S rDNA data for bacterial annotation at the species level, classification data for bacteria at the species level are not discussed here. However, the relatively high RAs of these potential pathogenic genera in the EBC suggest the spread potential of these pathogens via the aerosol route.

Fungi in EBC samples were mostly dominated by the phylum Ascomycota (>80%) and genus Fusarium (Figure D,E). The distribution of the top 15 fungal genera in EBC was significantly different from those in skin and environmental samples (Figure F), as shown in Table S2. High percentages of Fusarium in personal air samples (immediately close to the human subjects) and in occupied indoor environments have been reported.[68] This coincidence between fungi in exhaled breath and personal air suggests a potential microbial exchange between the respiratory tract and surrounding air. The other genera, that is, Schizothecium, Pyrenochaetopsis, and Monographella, have also been found in human respiratory tract samples.[69−71]

The dominant fungal genera recovered in this study were completely distinct from those reported in another study, which documented the dominance of Aspergillus sydowii and Cladosporium spp.[36] This distinction is likely due to the differences caused by culture-dependent and culture-independent sequencing methods.[72] In addition, some common fungi in the sputum and oral cavity, including Candida, Aspergillus, and Saccharomyces, were less frequent in the EBC.[73−75] This is possibly related to the respiratory pattern differences, and the oral cavity microbiota is prone to entrainment in speech-produced particles.[24] To the best of our knowledge, the present study is the first to investigate the fungal taxa in healthy human EBC using a culture-independent HTS method.

The dominant genus Fusarium serves as both a respiratory commensal and an opportunistic pathogen and has been reported in the BAL fluid samples from healthy subjects and diseased subjects.[17,76,77] Other less abundant fungal genera, including Mortierella, Chaetomium, Curvelaria, Cryptococcus, Candida, Guehomyces, Phaeoacremonium, Myrothecium, Aspergillus, and Penicillium, have been found in the respiratory samples and are opportunistic pathogens for respiratory fungal infections.[77,78]

Assembly and Heterogeneity of EBC Microbiotas

Stochastic processes, for example, dispersal, birth, death, extinction, and immigration, play a role in the assembly of expiratory bacterial and fungal communities.[79] The SNM fitting performance (Figure ) for the fungal microbiota was better than that for bacteria (R2 = 0.353 for bacteria and R2 = 0.683 for fungi). This difference indicates that stochastic processes may be relatively more important for the assembly of the EBC fungal community and less for the EBC bacterial community. This phenomenon is likely related to the size difference between bacteria and fungi, and smaller bacteria are less affected by dispersal limitation and more affected by deterministic processes.[80] In addition to SNM results, the results from iCAMP also showed the great role of stochastic processes. According to Figure S5, the relative importance of stochastic processes in the EBC bacterial community assembly was 66.25%, which was higher than that for selective deterministic processes (33.75%). In addition to the expiratory microbiota, this pattern has been reported in various communities of human body niches[60] and of other types of environments, including indoor and outdoor air.[81−83] Furthermore, the role of stochastic processes suggested a possible relationship between microbiota outside the respiratory tract and those inside the respiratory tract.

Figure 2

Stochastic processes played a role in assembling expiratory bacterial (A) and fungal (B) communities based on Sloan neutral model fitting. The black solid line represents the best fit, and the dotted lines represent the 95% CI (confidence interval) around the model fit. The blue dots refer to taxa that occur more frequently than predicted, and the red dots refer to taxa that occur less frequently than predicted. The green dots refer to taxa that occur in a manner consistent with predicted values.

Information regarding the specific taxa that fell within the below and the above partitions is provided in the supporting Excel file. Compared to the expiratory fungi, more expiratory bacteria taxa fell within the below prediction (Figure ), which is partly due to the large numbers of bacterial ASVs (6076 (bacteria) vs 2193 (fungi), supporting Excel file). The proportions of ASVs in the below partition for the bacteria and fungi were on the same level (1.79 vs 1.78%). For expiratory bacteria, there were no specific ASVs in the below partition at the genus level. However, at the species level, some specific species occurred in the below partition, for example, Haemophilus influenzae, Haemophilus parainfluenzae, Prevotella heparinolytica, Rothia mucilaginosa, and Fusobacterium nucleatum, which mostly belong to pathogens, opportunistic pathogens, or anaerobes.[84−87] This is likely related to the role of deterministic processes. For instance, the existence of these taxa was probably constrained by the healthy commensal microbiota in the respiratory tract.[88,89] Furthermore, different types of bacteria were found to have differential aerosolization capacities.[90] Collectively, these factors finally resulted in their lower than predicted occurrence in the exhaled air. It would be interesting to investigate the distribution of these taxa from the EBC of patients with respiratory infections in the future.

Variations in the EBC taxa among different subjects were likely partly accounted for by the stochastic processes (Figures S6 and S7), consistent with the previous findings.[91] Nevertheless, the role of deterministic processes is not excluded by the stochastic processes. Considering the respiratory microbiota is embedded in the lining fluid of the respiratory tract, the aerosolization ability of different microorganisms from the respiratory tract lining fluid could be possibly different depending on the taxa.[92] In addition, because the inhaled particles cannot completely settle into the respiratory tract, residuals that did not settle could be breathed out following the exhaled airflow.[93,94] Thus, compositions of EBC microbiotas likely have some residuals from the inhaled microorganisms. Together with the intrinsic heterogeneity of the respiratory microbiota, a heterogeneous expiratory microbial community could be formed as a mixture of those aerosolized from the respiratory tract and those not-settled inhaled microbiotas.

However, the disparities within EBCs were significantly less than those between EBC and other types of samples, as shown by the Bray–Curtis dissimilarity index (Figures A,B and S8, KS test p < 0.05) and by the visualized PCoA plots. Microbial communities in the EBC clustered together (green circles) and were statistically separated from other types of samples (Figure C,D, Table S2). Community similarities within the similar ecological niches have also been reported in a previous meta-analysis, for example, within the same human body sites, toilets, and surfaces in kitchens and restrooms,[39] which is likely due to the similarities in the environmental features that surround the microorganisms. This result suggests that each type of sample serves as a consistent habitat for microorganisms[39] and that the associations of one community with the others could vary depending on the sample types.

Figure 3

Diversity of bacteria and fungi between EBC microbiotas and other microbiotas. Box-whisker plots show the community dissimilarities obtained from pairwise comparisons between EBC and other samples for bacteria (A) and fungi (B). A value of 1 for the Bray–Curtis index indicates that the two communities are completely different, and 0 indicates that they are exactly the same. PCoA plots visualize the dissimilarities between the communities across all samples based on the Bray–Curtis dissimilarity index: bacteria (C) and fungi (D). The R2 and p values of ADONIS indicate that the differences among different types of samples were statistically significant.

Associations between EBC and Indoor Microbiotas

Strong associations existed between bacterial communities of the EBC and indoor air from bathrooms and kitchens. When EBC bacteria were assigned as a sink, high levels of associations were seen between the EBC and IABK, followed by the skin (Figure A, Table S3). Likewise, when the IABK was assigned as a sink, high levels of association were seen between the IABK and EBC, followed by skin (Figure B, Table S4). The EBC-IABK bacterial community similarity levels were comparable to those of the skin-IABK (p > 0.05). Generally, the role of skin in shaping indoor microbiota has been well appreciated.[95−97] Based on this, it is possible to infer that EBC bacteria played a role in shaping the indoor airborne bacterial community, especially in the bathroom and kitchen environments. Additionally, inspired by the counterfactual method,[98] the similarity patterns of the IABK were further evaluated by estimating the similar fractions of the IABK with each type of sample by excluding EBC samples from the source samples (Table S4). We found that when EBC was excluded, the fractions of unknowns have increased significantly (Wilcoxon matched-pairs signed rank test, p < 0.05), suggesting an association between the microbiotas of EBC and IABK.

Figure 4

Different associations between the EBC and indoor airborne microbiotas. Box-whisker plots show the association fractions between sink bacterial (A–C) and fungal (D–F) communities when the EBC (A, D) or IABK (B, E) or IAL (D, F) was assigned as a sink, and the remaining samples were assigned as sources. The inset pie charts show the mean values of associations between the sink and the sources.

Unlike EBC bacteria, EBC fungal communities showed strong associations with floor dust fungal communities. When EBC fungi were assigned as the sink, high levels of associations were seen between the fungal communities of DTL-EBC and DTB-EBC (Figure D, Table S3). This was also verified when the DTB or DTL was assigned as a sink. High levels of associations between dust-borne fungi and EBC fungi were observed, especially between DTL fungi and EBC fungi (Figure C,D). Similar to EBC bacteria, when the EBC was excluded as the possible source for DTL fungi, the fractions of unknowns also increased significantly (Wilcoxon matched-pairs signed rank test, p < 0.05), suggesting the association of expiratory fungi with DTL fungi (Tables S9 and S11). In contrast, associations between EBC fungi and indoor airborne fungi (both IAL and IABK) were relatively low (Figure E,F and Tables S5 and S7). The comparison suggested that expiratory fungi were more correlated with those in the floor dust and less correlated with airborne fungi. This difference is likely related to the large size of fungi, which results in the rapid settling of fungi,[99] which may thus partly account for the strong fungal associations of EBC-dust. The differences between EBC-DTB fungi and EBC-DTL fungi (Figure C,D) are likely related to the fact that people spend different amounts of time in living rooms and bathrooms/kitchens. Associations between fungi in floor dust and children’s asthma severity have been observed.[100,101] Based on this, it is probable that floor dust fungi were partly from the human exhaled breath, and there is a potential relationship between exhaled fungi and asthma severity.[36]

Figure 5

Different associations between the EBC and indoor floor dust-borne microbiotas by assuming the latter as the sink. Box-whisker plots show the association fractions between sink bacterial (A, B) and fungal (C, D) communities when the DTB (A, C) or DTL (B, D) was assigned as a sink, and the remaining samples were assigned as sources. The inset pie charts show the mean values of associations between the sink and the sources.

In addition, strong associations between EBC fungi and skin fungi were also seen (Figure D). Previously, skin keratins have been detected in EBC.[102] Analogous to the interactions between the skin microbiotas of residents and cohabitants in the literature,[103,104] interactions between the skin and expiratory microbiota also likely occurred.

A notable point is that EBC bacteria were associated more with airborne bacteria, while EBC fungi were associated more with dust-borne fungi. Moreover, airborne bacteria and fungi were less related than those within floor dust (Figure ). This could be caused by the varied microbial dynamics in the air depending on their size, which directly resulted in fate variations between bacteria and fungi. For exhaled bacteria, their small size facilitated their long-term suspension in the air, in which they would be finally degraded in the air. For the exhaled fungi, fast settling onto the floor may extend the lifetime of exhaled fungi, which was longer than those bacteria suspended in the air. One limitation in this study is that the analysis with DNA could not differentiate between DNA from intact cells and transient DNA released from dead cells. Therefore, it would be necessary and interesting to track the fate of exhaled microorganisms from their exhalation and final degradation.

Another point of concern is the distribution heterogeneity of EBC microbiotas in different types of indoor environments. Specifically, IABK-EBC microbial association patterns differed from those of IAL-EBC. IABK microbiotas were associated more with EBC microbiotas than IAL microbiotas. In contrast, IABK microbiotas showed relatively low associations with OA microbiotas, while IAL microbiotas showed a relatively strong relationship with OA microbiotas (Figure B,C,E,F). This phenomenon could be attributed to different physical conditions between the living room and bathroom and kitchen environments. Generally, bathroom and kitchen environments were relatively small, closed, and featured moisture production and occupancy presence, which were conducive to the accumulation and even growth of the expiratory microbiota.[105] Although the installation and operation of fans/hoods in bathrooms and kitchens were not recorded in this study, high installation rates have been reported in a previous CCHH population study in six major cities in China.[106] Moreover, hood operation was not enough to completely remove the particle pollution in a residence kitchen.[107] In contrast, the living room is more connected to the outdoor ambient air through windows and doors, which is conducive to the rapid dilution of the expiratory microbiota. Based on this, it is likely that associations between EBC microbiotas and indoor airborne microbiotas would be weakened when the outdoor airborne microbiota accounted for a large proportion.

However, the above results should be interpreted cautiously since the association patterns obtained with FEAST could not support the causal source–sink relationships. The proportion of each source community in the sink community was evaluated by leveraging the structural similarity between sink and possible source samples.[50] The high similarities between the two samples were mostly driven by the shared dominant taxa. For instance, a 61.08% similarity of bacterial communities between EBC16 and BFB16 was found, which is largely accounted for by the domination of certain genera in both communities (Streptococcus (12.18% (EBC16) and 17.14% (BFB16)), Neisseria (23.87 and 17.55%), and Haemophilus (9.50 and 5.32%)) (Table S3). In addition, when two “source” samples share a high similarity, exclusion of one of the two might directly affect the similarity pattern between the other of the two and the “sink” sample, as exemplified by the relationships of IABK06-EBC06 and IABK06-skin06. Specifically, a 22.73% similarity for IABK06-EBC06 and 2.59% for IABK06-skin06 were observed when IABK was assigned as a sink, but IABK06-skin06 bacterial community similarity changed from 2.59 to 20.03% when EBC was excluded as a source (Table S4). Additionally, due to the different survival capacities of each microorganism in the air, structural changes could occur after they entered into the indoor air. This change in the air could directly affect the reliability of associations between expiratory bacteria and bacteria in other microecosystems to be colonized. Hence, overlooking potential process changes further limits the usage of FEAST results in accurately inferring the source–sink relationships.

The differential associations between EBC microbiotas and indoor microbiotas were also likely related to the characteristics among individuals and among households at each residence. Different individuals have distinct particle release capacities, with some subjects classified as high emitters and some as low emitters.[108,109] Furthermore, the residence time of exhaled bacteria and fungi in the air might vary due to different indoor environmental conditions. As shown by the RDA (Figure S9), strong positive correlations were observed between CO2 and bacterial community similarities of EBC-IABK, EBC-skin, and EBC-DTB, as well as between CO2 and fungal community similarities of EBC-DTL and EBC-DTB. CO2 is generally assumed to be an indicator for indoor ventilation conditions, and thus, the results here indicated that low ventilation was positively associated with the above relationships. In addition, strong correlations were also observed between the indoor ambient PM2.5 mass concentrations and bacterial community similarities of EBC-IABK, EBC-skin, and EBC-DTL fungal community similarities. This is likely related to the carrier feature of PM2.5, which is a complex mixture of chemicals and biologicals. However, these results should be explained with caution since the environmental factors were not collected at each sampling site, while varied spatial distributions of gas and particle pollutants, for example, CO2 and PM2.5, have been reported in the residence house.[107,110] Nevertheless, the results here suggest the importance of ventilation in reducing the relationships between human expiratory microbiotas and indoor microbiotas. In addition, sampling in this work was performed in one season and one city. Different associations could be expected when the season and locations changed depending on the outdoor and indoor ventilation and air quality conditions.

Implications

As our understanding of the role of indoor microbial exposure in human health and disease continues to deepen, growing evidence shows that humans themselves play a vital role in shaping the indoor microbiota. Historically, human breath is generally not assumed to be a significant particle source because particle concentrations in human exhaled breath are low compared to those in the indoor and outdoor air.[109] It is time to reinforce the significance of the expiratory microbiota. In the context of the current COVID-19 epidemic, the implications of this study are possibly universally applicable in indoor environments.

On the one hand, the results indicate a diversified microbial diversity in the human expiratory microbiota. On the other hand, the results signify a necessity to balance the contribution of the expiratory microbiota to the indoor microbiota for human health. Moreover, considering that only one occupant per residence was recruited in this study, we could expect an increasing release of the human expiratory microbiota with more occupants or in the more densely populated offices and crowded public areas. The newly released expiratory microbiota could accumulate if not diluted in time. Due to the difficulty in modulating the airborne microbiota and a lack of a recipe for a healthy airborne microbiota, the introduction of more outdoor fresh air could be an optimal route.

The strong associations between EBC bacteria and those in bathrooms and kitchens suggest that these environments could serve as meeting and exchange spots for expiratory microbiotas. This matters particularly when potential pathogens exist in the expiratory microbiota. Bathroom and kitchen environments are ideal environments for microbial survival and growth due to poor water, poor hygiene, and less ventilation conditions.[111,112] In addition, activities in these environments could further increase this risk, for instance, the operation of hand dryers and toilet flushing.[113−116] In addition, these environments at residential homes in multistory buildings are connected among different homes due to the ventilation systems in these two places. Once pathogens appear in either the bathroom or kitchen of one home, they will likely be transported to bathrooms or kitchens of other homes in the same building, which might subsequently pose health risks to the exposed residents.

The differential associations between expiratory fungi and bacteria and indoor airborne and floor dust-borne microbiotas indicate that the further impacts of the exhaled microbiota should be differentially treated. While the associations between exhaled bacteria and indoor airborne bacteria could be attenuated with improved ventilation, the associations between exhaled fungi and dust-borne fungi could be weakened with improved floor hygiene. Likewise, exhaled fungi deposited on the other surfaces should also be taken seriously to reduce their potential health concern. On the other hand, the differential associations also suggest that there was likely a certain difference between airborne and dust-borne microbiota in indoor environments, which, although was not the theme of this investigation, is worthy of future study. Overall, the results from this work provide important information regarding the microbial relationships between humans and the environment, and improvements in the microbial structure in indoor environments could help improve human health.