Vertical stratification of the air microbiome in the lower troposphere

Significance Large-scale meteorological and biological data demonstrate the vertical stratification of airborne biomass. The previously described diel cycle of airborne microorganisms is shown to disappear at height. Atmospheric turbulence and stratification are shown to be defining factors for the scale and boundaries, dynamics, and natural variability of airborne biomass, resulting in the biological organization of the planetary air ecosystem. The atmosphere above the mixing layer height is proposed to act as a sink for microorganisms. With atmospheric processes being temperature dependent, rising global temperatures will result in major disruptions of the currently observed airborne microbial community structures. Increased abundances of radio-tolerant bacteria at height will allow investigation of these microorganisms’ life cycle in the planetary atmosphere.


1) Sample Collection & Storage
All air samples were collected with bioaerosol electret filters (6cm diameter, particle retention efficiency of 50% for 0.5μm particles; Research International), mounted on SASS3100 dry air samplers (Research International). The air samplers were operated at a high flow rate of 300L/min. Prior to installing a new filter, the air samplers were decontaminated by placing a Kimwipe wetted with lab cleaner (10% (v/v) bleach + 1% (v/v) NaOH + 0.1% (v/v) SDS + 80mM NaHCO3) onto the filter installation mount for one minute, followed by an additional round of decontamination with a Kimwipe wetted with 70% EtOH for another minute.
Upon completion of each sampling cycle, the SASS filters were returned to their original pouches, packed in Ziploc bags and transported back to the laboratory at room temperature where they were stored at −20°C until further processing.

2) Filter Processing & DNA Extraction
Biomass removal from the filters and subsequent DNA extraction were performed as previously described. In short, sterile forceps were used to remove each filter membrane from its plastic ring and transferred to a sterile 5mL tube. A wash buffer consisting of PBS and Triton X-100 (PBS-T) was then added to the tube and the filter membrane moved up and down inside the tube to dislodge the biomass. The wash buffer was subsequently squeezed out of the filter and the biomass removal repeated for two additional times with fresh aliquots of the wash buffer. For air samples collected at the meteorological tower, the combined buffer volume from the 3 wash steps for each individual filter was then filtered through a 0.02-μm Anodisc (Whatman) using a vacuum manifold (DHI), followed by DNA extraction with the DNeasy PowerWater Kit (Qiagen). Obtained DNA yields were determined with the Qubit dsDNA HS Assay kit (Thermo Fisher Scientific).
For samples collected with the research aircraft during flight 1, 6 filters collected at each height were used for DNA extraction testing to determine approximate DNA yield per filter and to inform the sampling regime for subsequent flights. As the test extractions were conducted in the field with slight variations from our standard extraction procedure, these test samples were not further processed for sequencing. For the remaining filters from flight 1 as well as all filters from flights 2-5 and the corresponding ground samples, the wash volumes from 3 replicate filters were combined and filtered through one Anodisc to ensure sufficient DNA yield for subsequent library preparation and sequencing.

3) Library Preparation & Metagenomic Sequencing
Quantification of the extracted DNA samples was repeated with Promega's QuantiFluor DNA assay on a Promega plate-reader fluorometer immediately prior to library preparation. Whenever possible, DNA input for library preparation was normalized to 5ng based on the QuantiFluor DNA concentration and the final libraries were amplified with 8 PCR cycles. For samples that did not meet the 5ng input DNA requirement, the PCR amplification was adjusted as follows to ensure sufficient library yield for sequencing:
Average library size was determined on an Agilent Tapestation 4200 and library quantitation was performed using Promega's QuantiFluor DNA assay on a Promega plate-reader fluorometer. Libraries were then diluted to 4nM and the concentrations were validated by qPCR on a QuantStudio-3 real-time thermocycler (Applied Biosystems) using the Library Quantification Kit for Illumina Platforms (Roche). Sequencing was subsequently performed on the Illumina HiSeq2500 platform in rapid mode at a read length of 251-bp paired end (Illumina HiSeq 2500 V2 rapid sequencing chemistry).
Filter and reagent blanks were processed with the same protocols as the air samples. All blank libraries were amplified with 8 PCR cycles.

4) Raw-Read Processing & Human Genome Reads Removal Analysis Pipeline
Adapter sequences and bases with Phred quality score less than 20 (Q < 20) were removed from the metagenomic reads during the quality trimming process using Cutadapt. The trimmed reads were then used for downstream analysis. In order to screen the obtained metagenomic datasets for potential human contamination, the following pipeline was implemented: Quality-trimmed sequencing data were mapped against the human genome reference (version GRCh38) using the ultrafast and memory efficient aligner Bowtie2 (version 2.4.1) with default parameters. The reference human genome was downloaded from NCBI on 9 September 2020. The Sequence Alignment Map (SAM)-formatted file generated by Bowtie2 was then converted to Binary Alignment Map (BAM) format before further analysis with Samtools, version 1.10. Mapped (human genome-associated) and unmapped reads were separated from the BAM file and the unmapped reads were further analyzed for the presence of correctly paired sequencing reads (both, paired forward and reverse reads are present) using Samtools. Singletons (only one of the paired reads is present, i.e., either forward or reverse) were discarded. Overall statistics showing the total read count for each sample and the number of unmapped reads is given below in Table S2. Unmapped, paired reads (after removal of human and human-associated microbial contamination) were aligned against the NCBI non-redundant (nr) protein database downloaded on 22 November 2019 using the taxonomic classification tool Kaiju, version 1.7.2. An overview of the entire analysis pipeline is shown below in Figure S1.
Resulting alignments were imported into MEGAN v6.18, which assigns taxon IDs based on NCBI taxonomy. To achieve the desired taxonomic specificity, we used the following filtering parameters: min score 100 (bit score), max expected 0.01 (e value), top percent 10 (top 10% of the highest bit score), min support 25 (minimum number of reads required for taxonomic assignment), lowest common ancestry (LCA) percent 100 (naive), and min complexity 0.33 (sequence complexity). LCA for each read on the NCBI taxonomy is assigned using MEGAN's LCA algorithm. In instances where all of the filtering criteria were fulfilled, reads were assigned to various levels of taxonomic classification ranging from domain to species. In our study, species-level classification was only reached if at least 25 reads uniquely aligned to a single species in the database with a bit score equivalent to a 100% match on the protein level over at least 50% of the 250-bp read. Due to limits of existing public sequence databases, some sequencing reads did not result in meaningful alignments and were classified in the nohits category. Unassigned reads were categorized in instances when alignments were made but did not fulfill one of the filtering parameters, namely the bit score.

6) Collection and Processing of Filter Blanks
Filter blanks were collected daily immediately prior to sampling. The air samplers were first decontaminated as described above. After decontamination, a brand-new, clean filter was removed from its pouch and installed on the metal grid of a randomly selected air sampler. Without running the air sampler, the filter was then collected back and placed into its original pouch for transportation to the laboratory. Two filter blanks were collected per sampling dat. All filter blanks were processed alongside the air samples with the same protocols as described above.
The analyses of the filter blanks did not reveal any significant contamination introduced by the air sampling equipment and downstream laboratory processing. Differences in DNA yields as well as taxonomic composition between the blanks and the MT and RA air samples are shown in Figure S2. After filter processing, the DNA concentrations of 6 of the 10 filter blank samples were below the detection limit of the Qubit 2.0 fluorometer. The DNA yields of the remaining four filter blanks were on average 29-and 20-fold lower than those of the MT and RA air samples, respectively (Fig. S2).
Extracted DNA from the filter blanks was subsequently processed with the default library enrichment conditions as for the air samples (8 PCR cycles). The number of sequencing reads generated for the filter blanks was very low (average of 7,524 paired reads per filter blank), compared to the MT (average of 2,804,028 paired reads per sample) and RA (average of 3,810,256 paired reads per sample) air samples. Further, only a small number of reads from the filter blank samples resulted in meaningful assignments at phylum level (332 -1,474 assigned reads) with the majority of them belonging to human-associated Proteobacteria and Actinobacteria taxa (Fig. S3). As described in detail in the method section, reads from our air samples that belong to these human associated taxa were removed prior to taxonomic assignment and subsequent downstream analyses.

7) Vertical Air Sampling in the Near-Ground Atmosphere (Meteorological Tower)
Near-ground air samples (200m height) were collected at the top of a meteorological tower at the Karlsruhe Institute for Technology (KIT) in Karlsruhe, Germany (Coordinates 49° 5' 33'' N, 8° 25' 33'' E, 110,4m a.s.l). The tower, which has been in operation since 1972, is located in a small forest clearing, surrounded by pine trees, and operated by the Department of Tropospheric Research of the Institute for Meteorology & Climate Research at KIT Campus North (Fig. S4). Sensors at different altitudes of the tower measure a variety of meteorological parameters, including: • Wind Speed (2, 20, 30, 40, 50, 60, 80, 100, 130, 160, 200m       Samples collected during night hours are shaded in grey. Bubble size is proportional to the number of sequencing reads assigned to each taxon.

7.2) Pilot 2: Air sampling at the meteorological tower in July 2018
A second pilot experiment was conducted in July 2018 during the summer period in Germany to investigate seasonal influences on the air microbiome composition. For this study, air samples were also collected at the top of the tower (200m) and near its base (~0.75m) at noon and midnight for a sampling duration of 2 hours with a flowrate of 300L/min. Samples were collected for two consecutive days on July 9 and 10 (Figs. S10 -S13).

7.4) Large-scale study to investigate temporal and vertical distribution of airborne microbial organisms in the near-surface atmosphere (October 2018)
A total of 16 SASS air samplers (8 air samplers per location) were placed on the uppermost platform of the tower at 200m height as well as near the base of the tower, approx. 0.75m above the ground. The instruments were programmed with delays so that always 2 instruments at the top and 2 instruments at the bottom were running simultaneously at the following time points without the need for human supervision: 10am-12pm, 2-4pm, 6-8pm 10pm-12am, 2-4am, 6-8am. Sample collection therefore occurred in duplicates for a duration of 2 hours per time point every 4 hours (6 time points per day) and was repeated for 5 consecutive days (8 Oct 2018, 10am until 13 Oct 2018, 8am). The sampling days in October were chosen as large high pressure system with stable weather conditions was predicted for most of Central Europe for the entire week, enabling sampling conditions as consistent as possible across all sampling days (Figs. S17 -S31).
In addition to the air samples, one blank sample from a randomly selected SASS instrument was collected at the tower top and at the ground every day, resulting in a total of 130 collected filters for the entire experiment (120 samples and 10 blanks).

7.7) Using horizontal wind speed measurements as proxy for vertical mixing processes within the atmospheric boundary layer at the meteorological tower for the experiment conducted in October 2018
At the meteorological tower the measuring height is restricted to 200m a.g.l. This height limitation prevents the detection of the daytime mixing layer height (MLH), however, the development of the night-time stable layer with suppression on vertical mixing can be observed. Horizontal wind speed measurements at the tower were used to calculate vertical Figure S29. Horizontal wind speed measurements as proxy for vertical mixing processes within the atmospheric boundary layer. Each subfigure represents three profiles of potential temperature (blue, red and black lines) with 10-minute mean values at the full hour as indicated in the legend. Ten-minute mean data of the horizontal wind speed for each hour are depicted as dots for each measured height (30m, 40m, 50m, 60m, 80m, 100m, 130m, 160m, 200m), color-coded with the corresponding profile of the potential temperature.
wind speed to provide better insights into the level of vertical mass transport during the day and night at the tower location. As vertical wind speed measurements are not recorded at the tower, this lack of information can be overcome by using horizontal wind speed and it's variability throughout the sampling period in the different measuring heights at the tower as a proxy for vertical motion within the atmospheric boundary layer (ABL). The measured 10-minute averages of horizontal wind speed are a valuable indicator for the degree of turbulence within the ABL. The higher the mean horizontal wind speed, the higher the three dimensional turbulence and vertical motion (positive or negative), and as a result, higher vertical mixing within the ABL is expected.

7.8) Distribution of specific fungal species at different heights in the MT experiment
The higher abundance of fungi in MT air samples allowed us to study the distribution of different classes of fungi at different times of the day as well as heights. Basidiomycota abundance, particularly for the listed species in Figure S30 that belong to the polyporales order, is observed to be the highest at ground level, with strong increases during night-time while declining during daytime. A strongly diminished signal was recorded for this taxonomic group at 200m height, indicating the suitability of this group of species to be considered as a proxy for ground-based taxa. This finding is expected as Basidiomycota species are largely known to be wood-rotting fungi (Fig. S30a). Inversely, the listed species from its fungal sister clade Ascomycota peaked during solar noon on ground level and showed a low abundance during the night. At 200m height, a constant high level of Ascomycota taxa was observed, largely independent of the diel cycle (Fig. S30b). This distribution pattern is more similar to that of the bacterial taxa which are mostly associated with higher heights. While airborne Ascomycetes spores are widely regarded to originate from ground sources, this finding indicates that they may possess higher tendency to remain airborne for a longer time than its Basidiomycota counterparts.

7.10) Comparison of diel variations in air microbiome composition for samples collected in the temperate climate zone (Karlsruhe, Germany) and in the tropics (Singapore)
Germany samples were collected at the ground and 200m top of the meteorological tower in Karlsruhe whereas Singapore samples were collected at an open balcony at Nanyang Technological University. This dataset was analysed with the same analysis pipeline as described in Gusareva et al. 2019 (18) to allow for comparison with the Singapore samples ( Figure S33).

8) Air sampling at higher altitudes with the help of a research aircraft (October 2018)
Air sample collection at altitudes of 300m to 3,500m was conducted in Brunswick, Germany, with the help of a research aircraft. The research aircraft, a Dornier 128-6 high-wing aircraft with two propeller-turbine engines, is owned and operated by the Institute of Flight Guidance at the Technische Universität Braunschweig. The aircraft is equipped with scientific instrumentation for meteorological measurements in the lower atmosphere. In addition, for this study, two interior racks inside the aircraft were modified to accommodate twelve SASS3100 air samplers (Research International). Furthermore, a custom-made air inlet pipe was installed to allow outside air to be funneled into the sampling chambers inside the aircraft cabin where the racks with the SASS air samplers were located. This set-up allowed for outside air to be samples within the sampling chambers inside the aircraft (Fig. S34, Table S3). Figure S33. Diel cycle of airborne microbial communities in the temperate climate zone (Karlsruhe, Germany) and in the tropics (Singapore). Samples in Germany were collected for 2 hours every 4 hours for 5 consecutive days whereas samples in Singapore were collected for 2 hours every 2 hours for 5 consecutive days. Replicates for each time point were averaged.

8.1) Construction of aerodynamically-shaped custom-made air inlet pipes
A custom-made air inlet was installed on top of the cabin, consisting of two metal pipes with 45-degree bevels. The bevels were placed perpendicular to the outside airflow. Each inlet pipe had a diameter of 2 inches and an approximate cross-sectional area of 20cm 2 . Inside the cabin, the pipes connected to plastic tubing that each led to one of two pull-out rack inserts, each holding six samplers. For each flight, live in-flight analysis of meteorological data (e.g. temperature and absolute humidity) collected during ascent to 3,500m determined which two additional sampling heights were to be sampled, one below the inversion layer and one above (altitude 1: 3,500m, altitude 2: 2,000-2,300m, altitude 3: 1,300-1,500m and altitude 4: 300-330m; Fig. S35). The experiments were carried out with a research aircraft (Dornier 128-6), operated by two pilots, one meteorologist and one air sampling expert. A total of three day flights and two night flights were undertaken. C) Air sampling was performed in two open chambers inside the aircraft cabin, consisting of custom-made tubing and modified hoods, flooded by the dynamic pressure of the outside air that allowed for sampling of external air within the chambers. Each chamber was large enough to accommodate six SASS air samplers for a total of twelve air samplers that could be operated simultaneously on the aircraft. The air samplers were secured on retractable trays which facilitated filter retrieval/change as well as cleaning of the air samplers in between sampling cycles.
Twelve SASS air samplers were operated simultaneously at each height (12 air filters collected per height). Matching ground samples were simultaneously collected at the Brunswick Wolfsburg Airport (BWE). The airport is located approx. 200km SE of the North Sea, 200km SW of the Baltic Sea and 400km NNE of the meteorological tower in Karlsruhe. For the ground samples, 6 SASS air samplers were placed on a trolley (~1m height) outside the hangar for the research aircraft at the airport. The air samplers at the ground were operated simultaneously to the air sampling that was performed on the aircraft at the different heights. At each height, air sampling was conducted for a total duration of 35-50 minutes. In between the sampling cycles at different heights, the air samplers were decontaminated with lab cleaner and 70% ethanol as described above, before installing a new filter. In total, five flights were conducted, three during the day and two at night time: A) Flight path. Each flight lasted for approx. 4 hours and the aircraft was following a repeated triangular pattern of approx. 40km x 40km for all 5 flights to control sampling consistency as much as possible between the different altitudes and also between different flights. B) Air microbiome composition was investigated as a function of meteorological factors such as humidity, temperature, vertical mixing and light intensity as well sampling height. Live inflight analysis of meteorological data was used to determine optimal sampling altitudes. For each flight, 12 SASS air samples were simultaneously collected at each of the four altitudes mentioned above. In addition, 6 matching ground samples were collected for each altitude, resulting in 72 air samples being collected per flight or 360 filters in total across the 5 flights ( Figure S36). The biomass from multiple filters from the same sampling height was then pooled to obtain sufficient DNA for metagenomic sequencing (see details in the methods section above).

8.2) Vertical transport estimations within the atmospheric boundary layer
The atmospheric boundary layer (ABL) is the lowermost layer above the ground which is directly influenced by its contact with the earth's surface. When investigating vertical transport in the atmosphere within the ABL, the state of the ABL is important, e.g., whether it is stable, neutral, or unstable stratified. When the ABL is unstable stratified, vertical motion and transport of gaseous or solid components (e.g., water, ozone, dust, microorganisms) is possible. In case of a stable stratified ABL, these substances remain suspended at the same height as when stability was developed, with the exception of larger particles which may still fall back to the ground very slowly due to gravity. The layer which enables vertical motion is called the mixing layer and its uppermost boundary is the mixing layer height (MLH). After sunrise, the ABL rises and the MLH increases. After noon, the MLH reaches its peak value and vertical transport is most effective within the neutral or slightly unstable layer. Around sunset, a new stable layer, which stops vertical mixing, develops near the ground and the mixing layer of the daytime hours (above the stable layer) is now called the residual layer (RL). Its stability is neutral or slightly stable during the night and vertical motion is suppressed.
To determine the height of the ABL and the state of the atmosphere (stable or unstable stratified), two approaches can be used which are either based on thermal or dynamic parameters. In many instances, both ABL calculation methods result in similar heights:

Thermal-driven ABL calculation approach:
The ABL can be determined using thermal measures, e.g., potential temperature. Here, temperature and pressure measurements are used to calculate vertical profiles of the potential temperature. Increasing potential temperature with height indicates a stable stratified ABL with no vertical motion. Decreasing potential temperature with height, on the other hand, is a sign for an unstable stratified ABL with the possibility of vertical motion.

Dynamics-driven ABL calculation approach:
Alternatively, direct measurements of vertical wind speed can be used to determine the dynamics of the ABL. In this case, the height of the vertical wind speed within the ABL is a direct measure for vertical transport. The effect of vertical transport can be seen when looking at the vertical profiles of trace substances such as water vapor within the mixed layer. For example, mixing is high (effective) when the humidity profile is constant with height, whereas mixing is less effective when humidity decreases with height.
Estimation of the vertical mixing processes within the ABL during the five flights based on thermal-and dynamics-driven ABL calculations ( Figure S39; Table S4):

Flight 1, 09 Oct 2018 (daytime)
The MLH estimation using potential temperature and vertical wind speed data results in the same height. At noon on 9 October 2018, the MLH was about 600m a.g.l. Within this layer, the stability was neutral, allowing for vertical transport of microorganisms between the ground and the MLH. Above the MHL, up to 3,500m a.g.l., the stratification was stable, and no vertical mass transport is expected to have occurred. For Flight 2, the MLH estimation using potential temperature resulted in a height of 800m whereas the dynamics-driven calculations based on vertical wind speed data indicated that the MLH was at 950m. At noon on 10 October 2018, the uppermost 150m of the ABL showed a wavy structure which may have been caused by high horizontal wind speed of up to 10m/s within that layer. Between the ground and the MHL at 800m/950m, the stability was neutral, enabling vertical transport of biomass between the ground and the MLH. Above the MLH, up to 3,500m a.g.l., the stratification was stable, likely preventing vertical mass transport.

Flight 3, 10 Oct 2018 (nighttime)
During the 7-hour interval between Flight 2 at noon and Flight 3 at night on the same day, a strongly stable layer developed between the ground and 500m a.g.l. Above that layer, up to 1,050m, the stratification based on potential temperature measurements was still stable, but less than in the layer below. Vertical wind speed measurements indicate that vertical motion occurred between the ground and 800m a.g.l. Although stable stratified, this indicates that some vertical transport within the stable layer that is driven by turbulence may have still occurred. In the layer around 500m, the horizontal wind speed was rather high (up to 15m/s). This increases turbulence and results in vertical transport. Above that layer, up to 3,500m a.g.l., the stratification was stable, likely preventing vertical mass transport.

Flight 4, 12 Oct 2018 (daytime)
On this day, MLH estimation based on potential temperature results in a height of 600m whereas vertical wind speed-based calculations predict a height of 700m. At noon on 12 October 2018, the uppermost 150m of the ABL again shows a wavy pattern caused by turbulence and rising air bubbles. Within the layer between the ground and 600m/700m, the stability was neutral, allowing for vertical transport of microorganisms between the ground and the MLH. Above the MLH, up to 3,500m a.g.l., the stratification was stable, and no significant vertical mass transport may have occurred.

Flight 5, 12 Oct 2018 (nighttime)
During the 7-hour interval between Flight 4 at noon and Flight 5 at night on the same day, a very stable layer developed between the ground and 100m a.g.l. (7K/100m). Above that layer, up to 500m, the stratification based on potential temperature was still stable but less pronounced than in the very stable layer below (0.5K/100m). The upper boundary of the residual layer (RL) lied at 1,400m a.g.l., which is the same height as the upper boundary of the mixed layer in the afternoon of the same day (Flight 4). Vertical wind speed measurements also indicate vertical motion between the ground and the RL height at 1,400m a.g.l. Although slightly stable stratification was present, this indicates that some vertical transport within the stable layer occurred, driven by turbulence due to vertical and horizontal motion within this layer. In the layer around 1,400m, the horizontal wind speed was up to 8 m/s. This promotes turbulence and results in vertical transport. Above this layer, up to 3,500m a.g.l., the stratification was stable, likely preventing vertical mass transport. Figure S39. Live in-flight analysis of meteorological parameters. Vertical profiles for potential temperature (Tpot), absolute humidity, vertical wind speed (w), and horizontal wind speed (v) are shown between the ground and 3,500m height for each flight. Meteorological parameters were recorded at the beginning of each flight during the initial ascent to 3,500m. Flights conducted during night hours are shaded in grey.  Figure S40. Top 40 all species. Ground samples followed by flight samples, grouped together by flight. Flight samples are ordered by altitude with the lowest altitude (300m) displayed first, followed by the higher altitudes. Ground samples are ordered according to their corresponding flight samples, e.g., ground samples that were taken simultaneously to the 300m flight samples are show first. Flights that were undertaken during night hours are shaded in grey.

Figure S41. Top 40 bacteria.
Ground samples followed by flight samples, grouped together by flight. Flight samples are ordered by altitude with the lowest altitude (300m) displayed first, followed by the higher altitudes. Ground samples are ordered according to their corresponding flight samples, e.g., ground samples that were taken simultaneously to the 300m flight samples are show first. Flights that were undertaken during night hours are shaded in grey.

Figure S42. Top 40 fungi.
Ground samples followed by flight samples, grouped together by flight. Flight samples are ordered by altitude with the lowest altitude (300m) displayed first, followed by the higher altitudes. Ground samples are ordered according to their corresponding flight samples, e.g., ground samples that were taken simultaneously to the 300m flight samples are show first. Flights that were undertaken during night hours are shaded in grey.

Figure S43. Top 40 basidiomycota.
Ground samples followed by flight samples, grouped together by flight. Flight samples are ordered by altitude with the lowest altitude (300m) displayed first, followed by the higher altitudes. Ground samples are ordered according to their corresponding flight samples, e.g., ground samples that were taken simultaneously to the 300m flight samples are show first. Flights that were undertaken during night hours are shaded in grey.

Figure S44. Top 40 ascomycota.
Ground samples followed by flight samples, grouped together by flight. Flight samples are ordered by altitude with the lowest altitude (300m) displayed first, followed by the higher altitudes. Ground samples are ordered according to their corresponding flight samples, e.g., ground samples that were taken simultaneously to the 300m flight samples are show first. Flights that were undertaken during night hours are shaded in grey.          Bubble sizes are proportional to the number of sequencing reads assigned to each taxon.

9) Rarefaction curves of different sequencing depths
A representative set of samples from the MT and RA experiments were independently analyzed with different sequencing depths to assess the saturation of the number of identified species. For the MT experiment, ground-level air samples showed a lower diversity than tower-top air samples. This difference was the most prominent during night-time when minimal vertical mixing occurs, compounded by active release of Basidiomycota taxa which dominated the ground-level air. During daytime, the high degree of vertical mixing allowed height-associated taxa to enrich the ground-level air, which resulted in minimized differences in diversity between the two air layers (Fig. S63a). The ground air samples of the RA experiment were found to be significantly more diverse. This is likely due to the difference in land-type between the two sampling campaigns (MT = forested, RA = grassland). Despite such differences, the higher altitude (3,500m) air samples were still found to have higher diversity than the ground air samples, albeit at a lesser extent (Fig. S63b). Most importantly, the ultra-deep sequencing of the two samples with the highest diversity indicated that saturation in species identification is not yet reached, even at the scale of 30 million reads per sample (Fig. S63c).

10) Presence/absence analysis of bacterial and fungal taxa in the vertical air column
Using a presence/absence analysis, bacterial and fungal taxa were binned based on their temporal (day/night) and vertical distribution (0 and 200m MT, 0 and 1,000 -3,500m RA) (Fig.  S64). While the results are categorized at the taxonomic levels of Domain, Kingdom and Phylum, the underlying taxonomic resolution for each taxon is at species level. Fungal diversity was found to largely overlap between all height levels in both experiments (MT and RA). In detail, during daytime, the fungal taxa overlap between the ground and the height reached 85% (RA) and 88% (MT). During nighttime the overlap between ground and the various heights reduced to 73% (RA) and 84% (MT). Consequently, only a small number of fungal taxa were specific to either ground (6 -23%) or height (4% @1,000-3500m RA, 10% @200m MT). The small fraction of fungal taxa specific to ground and height indicates the ground as a likely source for both settings. In contrast, overlap in bacterial diversity noticeably differs between ground and heights, in a diel-specific manner. Daytime overlap of bacterial OTUs between ground and height was 77% (RA) and 76% (MT). Night-time stratification of bacteria in the RA and MT experiment was more prominent, resulting in 61% and 49% overlap, respectively. The observed patterns in presence/absence are a consequence of the specific behavior of each of the bacterial phyla Actinobacteria, Proteobacteria, Firmicutes and Bacteroides. The higher fraction of height-specific bacterial taxa during night-time indicates these to originate from sources other than local ground sources.

11) Air Sampling at Mount Säntis, Switzerland
Air samples were collected on a roof top at the summit (2,502m) and near the car park at the base of the mountain on 4 Nov 2017 at 12pm and 3pm, respectively (Fig. S65). Mount Säntis is the highest mountain in the area and a popular tourist site. The buildings on the summit house several restaurants as well as a weather station.
With the exception of some bacterial species that may originate from a nearby wastewater treatment facility on the summit, no major differences were observed in the microbial composition of air samples collected at the summit and the base of the mountain ( Figures  S66 -S69). This may be explained by local wind systems and atmospheric turbulence caused by temperature differences between the mountain slopes and adjacent valleys. During daytime when the samples were collected, the air along mountain slopes is heated up more than air at the same elevation over an adjacent valley, resulting in the warm air rising up (upslope winds) and drawing of cooler air up from nearby valley ground layers. This results in mixing of air layers from different altitudes and allows for airborne microbial organisms from lower altitudes to reach air layers at the summit of the mountain. As such, we deemed mountains as unsuitable sites for accurate characterization of bioaerosols associated with different layers in the vertical column of the atmosphere.

12) Vertical Air Sampling in the Built Environment (Residential High-rise Building)
Our very first vertical air sampling attempts were conducted on 4 Nov 2016 at the Pinnacle@Duxton residential housing complex in Singapore (Fig. S70). From 9:15-11:15am in the morning, air samples were simultaneously collected on open corridors of different floors at the Pinnacle@Duxton which is the tallest public housing (HDB) project in Singapore at a total height of 156m. Air samples were collected in triplicates at the ground floor (1.3m), Level 3 (~10m), Level 5 (~20m), Level 10 (~40m), Level 21 (~80m) and Level 49 (~140m). Replicates were later pooled to obtain sufficient DNA amounts for metagenomic sequencing.
Metagenomic analysis did not reveal any significant differences in the microbial composition of the air samples collected at different levels of the building (Figures S71 -S74). As the sampling was conducted during the day when the surface of the planet heats up and causes ground-level air to rise up to higher altitudes, this most likely caused ground-level air to rise up to the highest sampling point and beyond, resulting in identical microbial composition of the air samples collected at the different levels of the building. In addition, solid, concrete structures such as this high-rise building may also alter the natural airflow of air masses near the ground. As such, we ruled out high-rise buildings as suitable sampling sites for future vertical air sampling experiments.      Table S5. Overall statistics of total raw sequencing reads and remaining unmapped paired reads for Kaiju analysis after removal of human and human-associated microbial contamination for all additional samples presented here in the supplement. VRTV: Samples collected at the Pinnacle@Duxton; APCH: Samples collected at Mount Säntis; APG57-APG62: Samples collected at the meteorological tower in February 2018; APG63-APG72: Samples collected at the meteorological tower in July 2018.

14) Presence of Cyanobacteria in the meteorological tower (MT) and research aircraft (RA) datasets
Cyanobacteria are photoautotrophic microorganisms and the only oxygenic photosynthetic prokaryotes that can be found in a wide range of diverse and extreme environments (1). Screening of the MT and RA metagenomic datasets for the presence of this unique group of bacteria revealed that Cyanobacteria can be found at the ground as well as greater heights at both locations (Figs. S75 and S76). Interestingly, similar to radio-tolerant bacteria, cyanobacterial species are more abundant at the tower top (200m) and at heights of 300m and above in the aircraft experiment as compared to the respective ground samples. In addition, while Cyanobacteria can be found throughout the day at greater heights, their abundance diminishes during the night hours at ground level. Furthermore, approximately 5 times more sequencing reads are being assigned to the most abundant cyanobacterial species in the RA experiment as compared to the MT dataset (e.g., up to 20,000 reads per sample assigned to a cyanobacterial species in RA (up to 0.54% of the total reads of a sample) as compared to up to 4,000 reads in MT (up to 0.09% of the total reads of a sample). . Figure S76. Top 40 Cyanobacteria in the research aircraft (RA) dataset. Ground samples followed by flight samples, grouped together by flight. Flight samples are ordered by altitude with the lowest altitude (300m) displayed first, followed by the higher altitudes. Ground samples are ordered according to their corresponding flight samples, e.g., ground samples that were taken simultaneously to the 300m flight samples are show first. Flights that were undertaken during night hours are shaded in grey. Bubble sizes are proportional to the number of sequencing reads assigned to each taxon.

15) Analysis into the presence of protists in the meteorological tower (MT) and research aircraft (RA) datasets
Protists are a diverse collection of eukaryotic organisms such as algae, amoebas and ciliates (e.g., paramecium). With some exceptions, protists are primarily microscopic in size, unicellular and form a paraphyletic assemblage that consists of diverse taxa with similar appearance (2). We screened the metagenomic datasets from the MT and RA experiments for the presence of protists (Rhodophyta, Glaucophyta, Stramenopiles, Apicomplexa, Ciliophora, Dinoflagella, Cercozoa, Foraminifera, Radiolaria, Euglenozoa, Percolozoa, Metamonada, Amoebozoa, Hemimastigophora, Apusozoa and Choanozoa). Some species belonging to the Stramenopiles clade (Fig. S77), Amoebozoa and Apicomplexa (MT dataset only, Fig. S78) were identified as shown in the following figures. Ground samples followed by flight samples, grouped together by flight. Flight samples are ordered by altitude with the lowest altitude (300m) displayed first, followed by the higher altitudes. Ground samples are ordered according to their corresponding flight samples, e.g., ground samples that were taken simultaneously to the 300m flight samples are shown first. Samples collected during night hours are shaded in grey. Bubble size is proportional to the number of sequencing reads assigned to each taxon.
At the meteorological tower, the identified Stramenopiles species appear to have slightly higher relative abundances at the tower top (200m) as compared to the ground. In addition, Stramenopiles are consistently present throughout the day at 200m whereas their abundances decrease at the ground during the nighttime hours (Fig. S77A). For the RA experiment, Stramenopiles were also observed at greater altitudes >300m throughout the day with little fluctuation in their relative abundance over time (Fig. S77B). In contrast to MT, however, the relative abundances of Stramenopiles species at the ground in RA was up to 100x higher than at greater altitudes of 300-3,500m. This may be caused by the different land types at the two sampling sites (e.g., MT = forested, RA = grass lands).
Protist species belonging to the Amoebozoa and Apicomplexa clades, identified only in the MT dataset:

16) Day/night differences in abundance of cyanobacterial species and groundassociated bacteria at various altitudes
Additional species-level analyses show that other airborne microbial taxa besides radiotolerant bacterial species with specific physiological properties are also associated with either ground-or height-based settings (Fig. S79). Members of the cyanobacterial phylum (Figs. S79a, b), for example, which adhere to an oxygenic photoautotrophic lifestyle, show similar height distribution patterns as observed for radio-tolerant taxa, with greater abundance at altitudes >1,000m throughout the day and strong depletion below the mixing layer during the night-time hours. For ground-based taxa, which include members of the Actinobacteria phylum as well as other soil-associated microbiota, the dispersal pattern is reversed (Fig.  S79c, d).

17) Abundance of UV repair genes
The stratification of specific bacterial groups in the vertical air column is likely driven by a combination of their physiological capabilities and their resilience to various environmental stresses, such as desiccation, temperature, air pressure and radiation. Resistance to UV radiation is one such example.
Trimmed metagenomic reads were subjected to the humann3 pipeline with default parameters for functional analysis. Uniref90 database was used. Reads associated with the UvrABC system protein were then selected to assess the abundance of UV repair-related