The settling velocities of larger particles would exceed the upward speed of air entering the filter cups. Two vacuumed samples of floor dust from the chamber carpet were collected at the beginning and end of the study period, December 2013. The Dustream Collector was used to isolate the vacuumed material in the head of the sampling wand, and the vacuum cleaner was run for a 1-min duration while moving across the full extent of the chamber floor. Total particle number concentrations were measured at a frequency of once per minute with the Met One GT 526 optical particle counters , which measures particle number concentrations in six bins according to optical diameter: 0.3–0.5, 0.5– 0.7, 0.7–1.0, 1.0–2.0, 2.0–5, and >5 μm, respectively.DNA extraction protocols followed those used previously for indoor bioaerosols and are detailed in S1 File. Starting material was half of the filter from the filter cup or 200 mg of unprocessed dust from the floor dust sample. To determine the composition of the microbial communities, we used the approach of sequencing a universal “barcode,” i.e., a region of DNA targeted to a specific group of organisms that can be used for identification in samples containing a mixture of many taxa. Specifically, we targeted the V4-V5 region of the bacterial 16S rRNA gene and the ITS1 region of the fungal rRNA gene. Bacterial primers were those adopted for the Earth Microbiome Project while fungal primers were those recently described by Smith and Peay . Samples were split across two Illumina MiSeq lanes for 250 base pair paired-end sequencing at the Stanford Functional Genomics Facility. The raw sequence data were deposited into NCBI’s Sequence Read Archive under study accession SRP049464.For fungi both the forward and reverse reads for each sequence could be paired before downstream analysis. For bacteria the quality of the R2 sequencing reads was low; consequently we only proceeded with the R1 reads. The general processing approach involved quality filtering, pairing reads ,dutch bucket for tomatoes clustering reads into operational taxonomic units at 97% similarity, checking for chimeric sequences, and identifying taxonomy against a reference database.
To implement these steps, we utilized cutadapt, Trimmomatic, UPARSE scripts, homerTools, and the UNITE and Greengenes databases for fungi and bacteria, respectively. Specific program settings are detailed in S1 File.We adopted several OTU quality filtering and bench marking steps. First, we removed the OTUs that were unclassified after taxonomic identification. Second, we analyzed a fungal mock community of 18 fungal taxa whose abundance of extracted DNA was skewed to mimic a natural community and then pooled. Examining the sequences of the two mock community samples, we found that only taxa with 10 or more sequences should be included in further analyses. That is, the mock community was recovered only when OTUs with greater than 10 reads were considered; otherwise, the mock community had much higher richness than initially pooled. The specific threshold value is likely to be run-specific, as a similar approach using a mock community in another study informed a lower threshold value of 3 reads. As a comparison, other studies have taken the approach of excluding sequences that do not surpass a certain percentage of all reads . Third, we processed negative extraction controls with our samples. Even though visualization of the amplicons on an agarose gel showed no amplification, sequencing yielded reads in these negative samples. We subtracted the number of sequence reads in the negative samples from the environmental samples. We note that doing analysis without these quality-filtering steps produced qualitatively similar results as those reported here. After bio-informatics processing, 3.6 million and 4.2 million fungal and bacterial sequencing reads, respectively, were retained for analysis. Analyses were executed in R, utilizing the vegan and labdsv packages as needed. Phylogenetic analysis of the bacteria samples was conducted in QIIME. To ensure even representation of sequences per samples, samples were rarified to 5,000 sequences per sample for both fungi and bacteria.Initial exploration of the results suggested a higher than expected contribution of the ventilation supply air to the indoor bioaerosol composition. To assess whether the ventilation system itself might be contributing, we conducted two additional experiments in June, 2014, one with 2 people walking and one with the chamber unoccupied .
In addition to two samples of vacuumed floor dust collected at the start and the end of the day and the outdoor and indoor air samples, we included a third air sampler deploying an analytical filter cup within the sub-floor plenum, from which the supply air enters the chamber. These eight samples were processed as detailed above.Looking broadly at the composition of the identified microbes in aerosol particles, many of the common fungal taxa were familiar from culture and microscopy-based work, including species of Cladosporium, Aureobasidium, Phoma, Alternaria, Rhodotorula, and Penicillium. Other abundant taxa included yeasts , plant pathogens , and wood rot fungi . The dominant bacterial phylum was Proteobacteria, followed by Firmicutes and Actinobacteria. In addition to the common inhabitants of human skin, gut, and oral cavities such as Actinomycetales, Lactobacillales , and Enterobacteriales, we also observed high abundance of the outdoor-associated taxa such as Burkholderiales, Pseudomonadales, Flavobacteriales, and Streptophyta . Like other environmental surveys, most taxa appeared sporadically: over 50% of the taxa were present in only one sampling period, and over 80% of the taxa appeared in less than 10% of the samples. Considering both frequency and abundance, there was large overlap in microbial taxa between indoor and outdoor air samples. Table 1 shows the most frequently encountered taxa in the chamber air , and their corresponding frequency in outdoor samples. There were no frequently observed taxa in indoor air that were entirely absent from outdoor air. Two of the 40 taxa in this table had a greater than 2× frequency of occurrence in indoor air compared to outdoor air: the fungus Sordaria sp. and the bacterium Streptococcus sp. Fig 1 shows the abundant taxa in chamber air, outdoor air, and floor dust. Only five of the 15 fungal taxa in this set were shared across the abundant indoor and outdoor air samples, whereas 12 bacterial taxa were shared. The most abundant fungal taxon indoors was Battarea steveni, a puffball that is discussed later in the context of human-mediated transport. To determine which measured factors predict indoor microbial composition we applied an analysis of variance statistical model based on distance matrices. As shown using the Canberra community distance, fungal and bacterial communities shared similar patterns with some notable differences .
Indoor and outdoor air samples from the main experiments were more similar to each other and distinct from samples collected during the secondary experiments, which included supply-air sampling. Within the main study period, the indoor and outdoor fungal and bacterial aerosols were significantly different from each other, both when the carpet was exposed and when it was covered , although the percentage of the variation in composition explained by location was marginal . Bacterial community relationships across samples based on UniFrac, a distance index that considers phylogenetic relationships,blueberry grow pot yielded similar results to those based on taxonomic relationships . Due to the highly variable nature of the ITS marker, phylogenetic analysis was not applied to the fungal communities.We applied a statistical model to determine which measured factors predict indoor microbial composition. The sampling date and occupancy level were statistically significant predictors, explaining approximately 36% and 13% of the variation in composition, respectively . The influence of sampling date and occupancy level can be interpreted as follows: bioaerosols tend to be more similar in composition if they were collected on the same day or during the same occupancy level. After explaining variation by date and occupancy level, the effect of flooring and time of day were not significant predictors. Overall, 50% of the variation in microbial composition was unexplained by the measured variables. Higher occupancy periods were associated with ~ 2× greater taxon richness for both fungi and bacteria, and, as would be expected, this trend was unmatched in the outdoor samples . This increase in richness did not appear to be due to the addition of human-associated taxa, as those taxa did not dominate the occupied periods when looking at the entire microbial community. Considering sequence read abundance , the sum of five human skin taxa—Propionibacterineae, Staphylococcus, Enterobacteriaceae, Corynebacterineae, and Streptomycetaceae—comprise 4.3% of the indoor air sequences and 3.7% of the outdoor air samples, indicating only a modest enrichment of these species as contributors to indoor air microbial composition. Considering the frequency of taxon occurrence , there were 26 fungal taxa found in at least five of the six 8-person experiments, and only two showed increased frequency with increasing occupancy.
Rhodotorula mucilaginosa is a likely human commensal, and Aureobasidium pullulans was abundant in the floor dust samples. Likewise, of the 134 bacterial taxa found in most of the 8 person experiments, eight showed increased frequency with increased occupancy: Streptophyta, Solibacterales, Corynebacterium, Arcobacter cryaerophilus, Actinomyces, Chroococcidiopsi, Oxalobacteraceae, and Chlorophyta. Given that most of the detected bacterial sequences are environmental rather than skin-associated, this evidence suggests that resuspension of outdoor-derived microbes from indoor surfaces and/or from occupants’ clothing was a stronger source than direct shedding from human bodies. We explored patterns of ecological distance between indoor and outdoor pairs; however, few patterns emerged. The absolute values of the distances between indoor and outdoor pairs were significantly less for bacteria than fungi . Contrary to expectations, indoor air was not observed to be more compositionally similar to outdoor air in low occupancy periods than during higher occupancy periods . Moreover, high occupancy periods were not found to be more compositionally similar to each other than low occupancy periods were to each other .The most abundant fungus detected in the aggregate indoor air samples was Battarrea stevenii. This puffball appeared in high read abundance in two 8-person experiments and was not found in the paired outdoor air samples. The likely explanation is that a member of our research group acted as an inadvertent vector for the transport of these spores into the chamber. This research group member, who was also one of our study subjects, had previously handled specimens of Battarrea wearing the same sweater later worn in the chamber. In those two sampling periods, Battarea comprised 5% and 2% of all fungal sequences in indoor air.Floor dust can serve as a source of bioaerosols, so we collected vacuumed floor samples of dust from the carpet. The floor dust samples, despite being collected months apart, were similar in composition to each other, and were more similar to the composition of outdoor air than of indoor air . Considering the fifteen most abundant taxa in the respective samples, floor dust shared only one fungus and three bacteria with the indoor and outdoor air samples. The mean richness of microbes in the floor dust samples was 2× or 3× higher than that of indoor air samples for fungi and bacteria, respectively, and approximately 20% of the taxa detected were specific to the floor dust samples. The fungi in the air of walking experiments were slightly more similar to the floor dust when the carpet was exposed than when it was covered , while there was no difference for bacteria. The ventilation system itself could be a source of microbes, and we included a secondary set of experiments that included sampling the supply air. This effort yielded only a few data points, so we simply note the patterns. The air samples from the secondary study were similar to each other and distinct from the main study samples to varying degrees: for fungi, the secondary samples were quite different compositionally from the main study samples but for bacteria less so . For neither bacteria nor fungi was the ventilation system itself an obvious source of indoor microbes. That is, the taxa that were abundant in the supply air were also abundant outdoors. Those taxa that are present in the supply and indoor air but absent from outdoor air have relatively low read abundance .