Microbial richness and air chemistry in aerosols above the PBL confirm 2,000-km long-distance transport of potential human pathogens

Significance

Atmospheric aircraft monitoring with 10 tropospheric flights over the planetary boundary layer in Japan (between 1,000 m and 3,000 m above sea-level) demonstrate the presence of viable bacteria and fungi harmful to humans. Long-distance transport for over 2,000 km is possible in the free troposphere for air masses originating in agricultural regions enriched in fertilizers and pesticides.

Abstract

The existence of viable human pathogens in bioaerosols which can cause infection or affect human health has been the subject of little research. In this study, data provided by 10 tropospheric aircraft surveys over Japan in 2014 confirm the existence of a vast diversity of microbial species up to 3,000 m height, which can be dispersed above the planetary boundary layer over distances of up to 2,000 km, thanks to strong winds from an area covered with massive cereal croplands in Northeast (NE) Asia. Microbes attached to aerosols reveal the presence of diverse bacterial and fungal taxa, including potential human pathogens, originating from sewage, pesticides, or fertilizers. Over 266 different fungal and 305 bacterial genera appeared in the 10 aircraft transects. Actinobacteria, Bacillota, Proteobacteria, and Bacteroidetes phyla dominated the bacteria composition and, for fungi, Ascomycota prevailed over Basidiomycota. Among the pathogenic species identified, human pathogens include bacteria such as Escherichia coli, Serratia marcescens, Prevotella melaninogenica, Staphylococcus epidermidis, Staphylococcus haemolyticus, Staphylococcus saprophyticus, Cutibacterium acnes, Clostridium difficile, Clostridium botulinum, Stenotrophomonas maltophilia, Shigella sonnei, Haemophillus parainfluenzae and Acinetobacter baumannii and health-relevant fungi such as Malassezia restricta, Malassezia globosa, Candida parapsilosis and Candida zeylanoides, Sarocladium kiliense, Cladosporium halotolerans, and Cladosporium herbarum. Diversity estimates were similar at heights and surface when entrainment of air from high altitudes occurred. Natural antimicrobial-resistant bacteria (ARB) cultured from air samples were found indicating long-distance spread of ARB and microbial viability. This would represent a novel way to disperse both viable human pathogens and resistance genes among distant geographical regions.

It has been established that living organisms rarely reach high altitudes and survive for long in the troposphere (1–3). However, there is limited knowledge regarding the microbial species richness and diversity in this environment. Low levels of moisture and nutrients combined with high ultraviolet (UV) are known to restrict microbial proliferation in mid-to-high altitudes in the troposphere. It is noteworthy that certain extremophiles, like Deinococcus radiodurans can survive in the harsh conditions found at high altitudes with high radiation levels (4). The opposite is true for surface air, as demonstrated by several studies, which possesses a rich microbial biodiversity. The reported estimates for bacteria (e.g., annual mean of 2.5 × 10⁴– 5 × 10⁴ cells m⁻³) and fungi (e.g., annual mean of 2.5 × 10⁴–1 × 10⁵ cells m⁻³) arise from terrestrial biomes (5) or during dust events (6) or intense rains (7). Overall, 33 to 68% of these microorganisms can be transported, at times, over short distances of a few km (8) but quantitative studies are lacking (9). Whether or not these organisms stay viable and may then trigger the initiation of human pathology remains, however, an open and elusive question.

Limited data exist to date on the magnitude of microorganism propagation in the free troposphere (10), above the planetary boundary layer (PBL), whether attached to soil dust (11) or to organic aggregates (12). For a review, see ref. 13. The atmospheric PBL delimits the area of the lower troposphere most influenced by surface processes (1). The height of the PBL varies as a function of atmospheric conditions, with vertical mixing occurring below this layer. In contrast, above the PBL, materials uplifted can be subject to long-range transport due to the reduction in friction from both the surface and air density and reach the ground again through dry deposition in subsiding air masses or by being washed out by rain (14). Yet, the implications for human and plant health of this transport, as well as the viability of microbes in bioaerosols that can later cause infection have not been adequately explored.

Although little investigated, long-distance transport of viable airborne fungi and bacteria with soil dust has been shown to occur [e.g., from Africa to the Caribbean (15–17)] thereby causing adverse effects on the survival of corals and plants (18, 19). A consensus has been established that dust particles may pose serious risks to the environment and human health in countries in dust source regions and surrounding areas (20), though, for example, pathogens such as nontuberculosis Mycobacterium (8). Furthermore, tularemia outbreaks in Spain and Iran were linked, among other causes, to the inhalation of contaminated aerosols, and dust storms have been connected to regional outbreaks of meningococcal meningitis caused by Neisseria meningitidis, although the exact mechanism is still unknown (21). Cardiovascular, respiratory, and lung diseases can be caused by the inhalation of submicron radius particles because these can be ingested deep into the human body and are associated with respiratory morbidity and mortality (22). Cases of eye infections and diseases such as meningitis and valley fever have been recorded during and after significant dust events in some regions (23). Despite its significant relevance to human health, a mechanistic understanding of their airborne epidemiology and links to bioaerosols still remains elusive, therefore making a definitive attribution impossible.

Bacterial resistance to antimicrobial agents is becoming increasingly common and poses a serious threat to health security, as animal pathogens are among the main reservoirs of various drug resistance genes later transmitted to humans via the food chain. Antimicrobial resistance genes can occur through horizontal gene transfer (HGT) in any environment, particularly where bacterial loads are high, for example, in soil, in wastewater treatment plants (24), and in the gut microbiome of humans and animals, according to the transfer-related genes carried on plasmids (25). In addition, the natural environment and its microorganisms provide a pool of natural drug resistance genes that, when influenced by human activities, climate change, and environmental disruption, may all affect the evolution of bacteria and give rise to new genes that promote drug resistance (24). Antibiotics have also become one of the most frequently detected new pollutants in the environment, with antimicrobial resistance spreading from livestock and contaminated meat products to people.

In the present study, we sought to confirm that air masses above the PBL contain and disperse both fully viable human pathogens and bacterial species promoting drug resistance thousands of km away from their sources and that, despite their low concentrations, they may still have the potential to affect human health. Next, we briefly describe the airborne sampling protocol, the chemical analysis, DNA extraction and sequencing approaches followed. Afterward, how the cultures of samples were performed and the resistance profiles obtained. After the overall data analysis section, we end with a summarizing discussion on the conceptual progress obtained.

Methods and Analysis

Aircraft Surveys and Aerosol Sampling.

Ten air surveys with a Cessna 172 aircraft were conducted over Japan in February (22nd-26th and 28th) and April (1st and 5th-8th) 2014, departing from Chofu airport (Fig. 1 and SI Appendix, Fig. S1 and Table S1). These days were selected for their optimal conditions in terms of flight paths because they were against the NW direction of prevailing winds over Japan (Fig. 2 A–C and SI Appendix, Figs. S1–S3). The collection of aerosol samples in the lower troposphere was automated by opening an internal manifold when the plane was above the PBL (over 1,000 m a.s.l. and ascending to around 3,000 m a.s.l.; Fig. 1 and SI Appendix, Fig. S1). On the same day, after each flight, aerosol samples were also collected at the surface level at the Chofu airport near Tokyo, following the same experimental workflow (Fig. 1). Air sampling was carried out as described by Rodó et al. (14). A total of 22 aerosol samples were obtained during the two sampling periods, as detailed in SI Appendix, Table S1. In addition, for each sample period each day, both in flight and on the ground, a field blank filter was collected, exposed, and handled in an identical manner except that the sampler was not in operation.

Fig. 1.

Sampling and analysis pipeline for collected air samples. Flight and ground campaigns were conducted in parallel according to the established sampling schedule in parallel. The flight trajectory was documented. One-eighth of each collected filter was used for DNA extraction, bacterial culture, and major and trace elemental analysis. Comprehensive chemical and biological analysis together with source tracking of air masses was performed.

Fig. 2.

Time-height distribution of particles and the PBL over the days of sampling. (A) Average FLEXPART time-altitude 120-h backward simulation for the composite of all 10 flights in 2014 (The X axis denotes hours since 10.000 particles—log10 scale—were thrown at 3 h steps from Tokyo; colors denote concentration of particles in the total column from the surface up to 5 km; see Methods and Analysis). (B) Same as (A) but showing the average lat-lon source dispersion plume simulated from Tokyo. (C) Path covered by the centroid of air mass particles at each time step. Colors denote days. (D) Heights of both the centroid of particles (solid line) and the PBL (dashed line) during the days of sampling. White boxes indicate the times of flights and the color intensity scale, vertical wind speed. Note the downward direction of winds and high speeds lowering the height of the PBL.

Air mass trajectories were simulated 5 d back in time for each aircraft sampling date using the Lagrangian particle transport and dispersion model FLEXible PARTicle (FLEXPART v10.4) (26). Light detection and ranging (LIDAR) measurements for the same days and times were obtained from the Asian Dust and Aerosol Lidar Observation Network (https://www-lidar.nies.go.jp/AD-Net/).

Aerosol filter samples (PM_2.5) were analyzed for the determination of major (µg/m³) and trace elements (ng/m³) by Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP-AES) and Mass Spectrometry (ICP-MS), respectively, as described in detail elsewhere (27). One eighth of the filters (approx. 22 cm²) was used for DNA extraction. 16S rRNA gene amplicons were obtained following the 16S rRNA gene Metagenomic Sequencing Library Preparation Illumina protocol (Cod. 15044223 Rev. A). Raw FASTQ data were filtered to facilitate high-quality reads. To account for potential contamination, the processed high-quality reads were compared to a negative control included in the sequencing run.

Considering the heterogeneity of microorganisms collected in the filters, different culturing methodologies were applied and prior to the MALDI-TOF MS analysis, individual colonies were incubated for 24 h at 37 °C. If the strain was not identified by MALDI-TOF MS, 16S rRNA sequencing was performed. The extraction of genetic material from pure bacterial cultures was carried out and DNA quantification performed. The 16S rRNA gene amplification and detection in real time was performed. Antimicrobial susceptibility was assessed with Sensititre™ Custom Antimicrobial Susceptibility Testing Plates. All the isolates were categorized as multidrug-resistant (MDR), extensively drug-resistant, or pandrug-resistant (PDR; see definitions in SI Appendix). All methodologies are fully detailed in the Methods section of SI Appendix.

Results

Analysis of Air Mass Trajectories.

All flights were planned to track the wind currents coming from the mainland Asian continent (Fig. 2 and SI Appendix, Figs. S1 and S2). According to the backtrajectory analysis, over all the different aircraft flights in 2014 (SI Appendix, Figs. S2 and S3), the air took 2 d on average to travel from NE China to Tokyo (Fig. 2). This meant that the air traveled at a speed of around 12 m/s above the PBL (at a height between 1,500 m and 3,000 m a.s.l.) when over the Sea of Japan. This northwest (NW) atmospheric bridge (Fig. 2 A–C) did not allow the entrainment of new air from the maritime surface at any time during the transect collection (Fig. 2A, and PBL time-height distribution in Fig. 2D and SI Appendix, Fig. S4B). Afterward, the air was normally channeled down again across the PBL over central Japan, when strong downward winds push all the air mass from up to 5,000 m a.s.l. down to the surface (Fig. 2 and SI Appendix, Fig. S4). This movement down the air column is facilitated by the low-pressure system that establishes itself over northern Japan in winter. Before reaching this system, the air is uplifted by the strong Siberian High, the semipermanent cell of high pressure centered poleward of 45°N in NE Siberia (28). The centroids of the mass of particles (solid lines in Fig. 2D) in the backward simulations are located at heights from 1,000 m to nearly 3,000 m and always well above the height of the PBL (e.g., Fig. 2D; dotted lines). As seen by the white boxes in the simulations (Fig. 2D) of the flight durations each day, the sampling took place always collecting air above the PBL, without any contamination from air entrainment from the surface below the PBL (Fig. 2D).

LIDAR records for the same two groups of days in February and April 2014 confirm a massive downward flux of particles from very high altitudes, as marked by the temporal evolution in the downward movement of particles (SI Appendix, Fig. S4A). Total homogenization of the air column up to 6 km may occur on days when strong air intrusions take place, with at times high cloud cover conditions also appearing (red areas in SI Appendix, Fig. S4A), therefore yielding out-of-range measurements.

Overall Abundance and Diversity of Airborne Bacteria and Fungi.

Aerosols collected on quartz filters were analyzed for their chemical and biological composition. The detailed description of the experimental workflow is presented in Fig. 1. The amount of collected biomass was assessed by the quantification of extracted DNA. Although the DNA extracted from the 2 pools of blanks (one containing 10 flight blanks portions and another 11 ground blanks) was undetectable, those blanks were also included and processed in the downstream sequencing analysis, alongside the flight and ground samples. With regard to the total DNA yield, or the rates of unidentified reads per taxonomic levels, there were no significant differences between flight and ground samples (e.g., mean value of 0.1 ng m⁻³; SI Appendix, Fig. S5 A and B and Tables S1 and S2).

The mean counts of raw sequencing reads were 109 K and 91 K for bacteria and 199 K and 172 K for fungi in flight and ground samples, respectively; from which 88 K and 72 K (for 16S flight and ground) and 141 K and 123 K (for ITS flight and ground) quality-filtered nonchimeric sequences were used for further analysis (Fig. 3A and SI Appendix, Table S2). The detailed information on the distribution of reads along the quality check protocol is presented in SI Appendix, Table S2. In-depth taxonomical analysis resulted in the identification of 98% and 94% of all reads from bacteria and fungi, up to genus level (SI Appendix, Fig. S5 B and C). In total, 18 different bacterial phyla and 2 fungal phyla were identified, distributed over 305 and 266 genera, respectively (SI Appendix, Fig. S5 B and C). The airborne microbiome associated with PM_2.5 particles was largely dominated by Proteobacteria and Bacteroidota, while the most abundant fungal populations detected were from Ascomycota (Fig. 3A).

Fig. 3.

Microbial diversity and richness above the PBL and in the near-surface atmosphere. (A) The taxonomic profile of bacterial and fungal phyla represented as the total number of reads (Top) and relative abundance per sample (Bottom). (B) Two-dimensional nonmetric multidimensional scaling projection depicting the distances between microbial communities at the genus level in various samples. The scatter plot differentiates sample types with color coding (blue for ground, orange for flight) and collection months with symbols (triangles for February, circles for April). Ellipses on the Left panel group flight and ground samples, while those on the Right panel differentiate between February and April collections. PERMANOVA test results for group differences are displayed atop each panel. (C) Variations in alpha-diversity estimators for bacteria and fungi across altitude (Left) and month (Right). Significant differences, identified using an FDR-corrected Wilcoxon Rank-Sum test, are indicated with brackets (*denotes P < 0.05). (D) Number of unique bacterial and fungal genera exclusive to flight or ground samples, or found on both (Left), and unique to February or April samples, or found in both (Right).

The global comparison of the number of sequenced reads from flight and ground samples revealed a very high temporal similarity, confirming the massive homogenization of the air column (Figs. 3A and 4A). The number of unique genera present in samples followed a similar distribution in both bacteria and fungi (SI Appendix, Fig. S6), with the distribution across phyla and genera shown in SI Appendix, Fig. S6B. These results were stable after a sensitivity analysis. The similarity is clearly evident in the inspection of the most and the least similar samples (SI Appendix, Fig. S7), as the closest samples in time belong to the flight-to-surface and the surface-to-surface classes and also have the larger similarity scores (SI Appendix, Fig. S8). Applying a permutational multivariate ANOVA (PERMANOVA), no statistically significant differences were discerned between flight and ground samples, however, samples collected in February and April were significantly different (Fig. 3B).

Fig. 4.

Composition of air microbiome at high altitude and ground atmosphere. (A) Total number of read counts per sample, categorized by collection date. The top 10 most common bacterial and fungal genera are individually color-coded, with the remaining grouped in the “Other” category (blue) and unassigned reads in gray. (B) Relative abundances of the top 30 bacterial and fungal genera recovered from flight and ground samples. The size of each dot reflects the relative abundance of the genus in each sample with color indicating its corresponding class.

Alpha and beta diversity metrics for high-altitude air masses were similar to those of near-ground air, suggesting that high altitude air masses, at specific times, are very diverse and their species richness values are comparable to those of near ground air (Fig. 3 B and C and SI Appendix, Fig. S9). A substantial proportion of taxa (43.1% out of all taxa) appear in both flight and surface samples, while 32.6% out of all detected taxa were only observed in the flight samples, underscoring the distinct origin of those organisms overlapping with the different sources of air masses (Fig. 3D and SI Appendix, Figs. S2 and S3). Surprisingly, 41.9% of all taxa were only detected in February, due to the higher predominance of fungi during this temporal interval (Fig. 3D).

Taxonomic Assignment of Airborne Microbiome.

The elevated prevalence of Sphingomonas (Fig. 4A) in PM_2.5 dominated other bacterial genera, but upon meticulous examination of the remaining, less-abundant bacterial genera, a rich bacterial diversity unfolded, marked by the dominance of Methylobacterium, Bacteroides, Enterococcus, Streptococcus, Delftia, Escherichia, Nevskia, and Serratia (Fig. 4 A and B). The consistent pattern in relative abundances among the principal 30 bacterial and fungal genera appears in both airborne and surface samples across the entire temporal interval (Fig. 4B), once again underscoring the consistent distribution of similar microbial populations in high altitude and ground air masses.

The fungal air microbiome was mainly dominated by Exophiala, however, a closer examination of fungal diversity revealed Cladosporium, Alternaria, Malassezia, Penicillium, Hypoxylon, Sarocladium, and Pyrenophora as the most abundant, in both altitude and ground samples (Fig. 4B).

Temporal shifts of airborne microbes, influenced by nearby sources or by environmental factors in the near surface atmosphere, are well-studied phenomena, but less is known about similar changes in high altitude air. Despite the small sample size, clear changes between samples from February and April were observed in both DNA yield and alpha-diversity (SI Appendix, Fig. S5A and Table S2). The bacterial diversity observed was higher in April than in February while the converse was shown for fungi (Figs. 3C and 4B). When comparing the compositions of each domain level group (bacteria and fungi) between the 2-mo in the flight samples, both bacteria and fungi exhibited substantial changes (SI Appendix, Fig. S10). Attribution to seasonal changes only was not possible due to the 2-mo span of the sampling. The declining proportion of Proteobacteria in April was accompanied by an increase in Firmicutes and Bacteroidota (Figs. 3C and 4B). At the genus level, despite Sphingomonas standing as the most prevalent bacteria at high altitude, its relative abundance dropped in April, coinciding with a corresponding reduction in the relative abundance of Methylobacterium (Fig. 4B). Significant shifts were evident across a multitude of bacterial genera including Bacteroides, Streptococcus, Enterococcus, Clostridium, Sutterella, Faecalibacterium, Alistipes, Prevotella, Roseburia, Blautia, Lachnoclostridium, Oscillibater, and Subdoligranulum, transitioning from nearly undetectable levels in February to a markedly abundant presence in April. Conversely, Serratia displayed a distinct pattern, as it was detectable only in February, while its presence was largely undetectable in April (Fig. 4B).

No detectable shifts in fungal phylum distribution between February and April were observed, the fungal community remained predominantly composed of Ascomycota with a notably lower representation of Basidiomycota (Figs. 3A and 4B). The dominant fungal genera in both months were Exophiala, Alternaria, Cladosporium, Malassezia, and Penicillium. In February, highly diverse fungal taxa were also detected, including Hypoxylon, Phaeosphaeria, Peniophora, Taphrina, Periconia, and Beauveria (Fig. 4B).

Aerosol Chemical Composition and Relationships with Airborne Microbial Communities.

The average composition of all PM_2.5 samples revealed S and Na (>1 µg m⁻³) as the most prominent elements. Less prevalent were Al, K, Fe, and Ca (>100 ng m⁻³), followed by Mg, Zn, Zr, P, Ni, Cr, Ba, and Pb (>10 ng m⁻³). A range of other trace elements (Mn, Ti, Cu, Sn, As, Se, V, etc.) were also detected but at even lower concentrations (<10 ng m⁻³). However, only S, K, Zn, Ni, and Sn were present in all 21 samples, with average relative abundances of 52%, 7.8%, 2.2%, 1.8%, and 0.1%, respectively (SI Appendix, Fig. S11).

The mean element content at the ground level (3,821 ng m⁻³) nearly doubled that at the flight level (1,906 ng m⁻³), which may indicate additional local intense anthropogenic activity nearby the sampling site (Fig. 5A). Despite the detection of elevated concentrations for the majority of elements at ground level, these values were statistically not significantly different from those in the flights. On average though, the mean element concentration in February (3,935 ng m⁻³) was higher than in April samples (1,780 ng m⁻³) (Fig. 5A). Peaks of S were accompanied by increased levels of Pb, Zn, and Zr, and a range of other potentially toxic elements (Ba, Cr, Cu, Mn, and Ni; SI Appendix, Fig. S11).

Fig. 5.

Chemical composition of air masses and associations with airborne microorganisms. (A) Line-graph comparing daily total elemental concentrations in Flight (red) and Ground (blue) samples over the 11 d sampled. (B) Element-Genus Correlation Matrix: Displays Spearman’s r values for the 40 bacterial and fungal genera most correlated (or anticorrelated) with chemical element/compound concentrations. Blue shades represent negative correlations, red shades indicate positive correlations. Cells with a “+” signify P < 0.05, and those with a “*” denote P < 0.01 for the Spearman rank correlation test.

We used Spearman’s rank-order correlation analysis to search for potential relations between the bacterial genera in flights and aerosol’s chemistry (the top 40 most correlated genera present in at least two different flight samples are shown in Fig. 5B). With the exception of Deinococcus and Delftia, bacteria appear more strongly correlated to the chemical composition than fungi, indicating a close adherence of the former to the fine particles aerosolized. A majority of the highest correlated taxa appear strongly linked to K first and then, by groups, to Zn, Al, B, and Fe. Noticeably, Staphylococcus, Dialister, Actinomyces, Campylobacter, and Rothia are the ones more related to Zn than to K, denoting a different origin. Acinetobacter displays a strong relationship mostly to Fe, but also to Al, Ca, Cd, Cr, Mn, and Th, while Bryocella and UCG-05 only to Fe. Pseudomonas appears strongly correlated to Hf and Zr. In the case of fungi, the correlations are not so strong or nonexistent with K, in contrast to bacteria, with only Alternaria showing strong negative associations with SB, Pb, Rb, and As.

Indeed, a strong association between bacterial communities and B was detected (SI Appendix, Fig. S12). Namely, Streptococcus, Parabacteroides, Veillonella, Enterococcus, Faecalibacterium, Clostridium, and Bacteroides had a significant positive correlation (P < 0.01) with B (SI Appendix, Fig. S12), while Serratia had, instead, a significant negative correlation (P < 0.05). Shigella positively (P < 0.05) correlated with Al, B, and Cd, Streptococcus showed a significant negative correlation with Hf, S, and Zr (P < 0.01) and positively with B (P < 0.01). Parabacteroides displayed a significant negative correlation with Hf (P < 0.05), Delftia with As and Pb (P < 0.05), Anaerobacillus with Pb and Sb (P < 0.05), and Porphyrobacter negatively correlated with a group of metals (Pb, Rb, Sb, Sn, V, and As), as well as with S.

The broad analysis of the correlation between the composite microbial community detected in our air samples and the total concentration of chemical elements revealed a group of organisms that have a correlation accounting for at least 60% of the variability. Intriguingly, whereas B may have a very strong positive correlation (P < 0.01), Hf and Zr, typically found to be both associated in nature, showed a strong significant negative correlation (P < 0.05) with the majority of the detected organisms (SI Appendix, Fig. S12). These correlations between specific microbial communities and chemical elements may serve as indicators of the distinct geological sources contributing to the collected PM_2.5.

Characterization of Bacteria and Fungi from a Human Health Perspective.

Several genera classified as the top 10 genera hosting the majority of pathogenic species, including Streptococcus, Prevotella, Staphylococcus, Clostridium, and Bacteroides (29), were observed among the top 30 most abundant genera in our samples (Fig. 4B). Around 35% and 39% of bacterial and fungal species detected in PM_2.5 collected from the lower troposphere or near-surface, respectively, have the potential to pose a risk to human health since those taxa may act as opportunistic pathogens (Fig. 6A). Species-level bioinformatic identification was conducted and among those identified species, it is worth highlighting Serratia marcescens, Delftia acidovorans, Bacteroides vulgatus, Bacteroides dorei, Bacteroides uniformis, Bacteroides thetaiotaomicron, Bacteroides fragilis, Bacteroides eggerthii, and Alistipes onderdonkii (Fig. 6A). It is also noteworthy that the species appearing in populations larger in the lower troposphere than in the surface level belong to Bacteroides genera. B. thetaiotaomicron, B. fragilis, and B. eggerthii were predominantly detected in flight samples (Fig. 6A), while the rest of pathogenic species were present in equal amounts at both altitudes. Another intriguing observation is that a noticeable proportion of the remaining identified species constitute normal taxa of the human intestinal or oral microbiota. The distribution of those species demonstrated a notable shift between months, and a higher abundance of the opportunistic pathogens was detected in April than in February (Fig. 6B). Among the fungi, several human opportunistic pathogens were detected: The most prominent were Exophiala oligosperma, Malassezia globosa and Malassezia restricta, Cladosporium herbarum, Sarocladium kiliense, and Aureobasidium pullulans (Fig. 6C). Notably, all the identified fungal pathogens were found in both February and April samples, highlighting a high degree of similarity in their compositions (Fig. 6B).

Fig. 6.

Pathogenic species distribution across blanks and sample types. On (A) and (B), Average relative abundance of potentially pathogenic bacterial and fungal species identified across samples. Cells highlighted in red showcase the absence of a species in the given sample type, whereas blue cells denote a confirmed presence. On (C) and (D) Dot-plots representing the variability in total pathogen counts on a per-sample basis, distinguishing between bacterial and fungal pathogens, with blue dots for Ground samples and yellow dots for Flight samples. Only species with an average relative abundance of 0.01% or more in at least one of the sample groups are shown.

The reliability of these species assignments as pathogens or opportunistic pathogens in the filters is high and clearly not a result of human contamination during filter processing. This is supported by both their predominant place in the samples and their absence in the corresponding blank controls (i.e., corresponding to flight and ground monitoring; Fig. 6 A and C and SI Appendix, Fig. S7).

Viability and Antibiotic Susceptibility Tests.

To confirm whether bacteria collected from the air remained viable, we cultured a portion of the collected filters. The identification of isolated colonies was performed by MALDI-TOF MS, and when the strain could not be determined, the identification was made by sequencing the 16S rRNA gene (Methods and Analysis). The cultured bacteria mainly belonged to three genera: Bacillus sp., Micrococcus sp., and Lactobacillus. (SI Appendix, Table S3), organisms that were rarely identified by sequencing from the DNA extracted directly from the filters.

Several Bacillus species were isolated from flight samples: Bacillus pumilus (4 strains), Bacillus cereus (2 strains), Bacillus firmus (2 strains), Bacillus aerius (2 strains), Bacillus subtilis (1 strain), Bacillus aryabhattai strain 2 (1 strain), Bacillus wiedmannii (1 strain), Bacillus muralis (1 strain), and Bacillus megaterium (1 strain). In addition, Micrococcus luteus (1 strain), Micrococcus aloeverae (1 strain) and Lactobacillus kitasatonis (1 strain) were cultured from samples collected in air. Furthermore, Staphylococcus hominis was also detected.

All isolates belonging to B. pumilus were susceptible to penicillin, ampicillin, amoxicillin/clavulanic acid, clindamycin, rifampicin, tetracycline, teicoplanin, imipenem, meropenem, vancomycin, ciprofloxacin, moxifloxacin, and levofloxacin, but resistant to cephalosporins. One out of four isolates were resistant to erythromycin, one strain to chloramphenicol and one to trimethoprim-sulfamethoxazole (Table 1). Both strains of B. cereus were resistant to penicillin, ampicillin, cephalosporins, and clindamycin and susceptible to the remaining antibacterial agents, with the exception of trimethoprim-sulfamethoxazole, to which one of the strains was susceptible and the other one resistant. The antimicrobial susceptibility of the two isolates belonging to B. firmus was very similar, as both were resistant to cephalosporins, rifampicin, clindamycin, and chloramphenicol, meanwhile one of the strains was also resistant to penicillin and ampicillin but susceptible to amoxicillin-clavulanic acid and the other one resistant to erythromycin and trimethoprim-sulfamethoxazole. The other three species of Bacillus (B. subtilis, B. muralis, and B. megaterium), and M. aloeverae were susceptible to all tested antimicrobial agents. In addition, S. hominis was also susceptible to all tested antibiotics. The L. kitasatonis strain was resistant to penicillin, ampicillin, and trimethoprim-sulfamethoxazole and was intermediate to ciprofloxacin, while it could be considered susceptible to levofloxacin and resistant to moxifloxacin. The M. luteus strain was resistant to carbapenems, cephalosporins, glycopeptides, ciprofloxacin, and trimethoprim-sulfamethoxazole (Table 1). Thus, the antibiotic susceptibility of cultured organisms revealed that there were several MDR strains isolated from the samples.

Table 1.

Antibiotic resistance profile of the cultured taxa from air samples

	M. luteus (from S15)	B. aryabhattai (from S17)	B. wiedmannii (from S19)	B. aerius (from S21) (2 Strains)*	B. firmus (from S15 /S3) (2 strains)†	B. pumilus (from S5/S7/S17/S19) (4 strains)†	S. hominis (from S1)	L. kitasatonis (from S19)	B. cereus (from S15, S17) (2 strains)†
Penicillin	≤0.03 S	>4 R	>4 R	<0.03 S	>4/2 R/S	<0.03 S		>4 R
Ampicillin	0.25 S	>4 R	>4 R	<0.12 S	>4/1 R/ S	<0.12 S		>4 R	>4 R
Amox+clav	<1 S	<0.5 S	>4 R	<0.5 S	<0.5 S	≤1 S		≤0.5 S	>4 R
Cefepime	>2 R	<0.25 S	>2 R	>2 R	>2 R	>2 R		>2 R	>2 R
Cefotaxime	>2 R	<0.06 S	>2 R	>2 R	>2 R	>2 R		>2 R	>2 R
Imipenem	>4 R	<0.12 S	<0.12 S	<0.12 S	0.25 S	<0.25 S	<1 S	<0.12 S	≤0.5 S
Meropenem	>2 R	<0.25 S	<0.25 S	<0.25 S	<0.25 S	<0.25 S		<0.25 S	0.25 S
Vancomycin	>2 R	<0.5 S	<0.5 S	<0.5 S	<0.5 S	<0.5 S	<0.5 S	<0.5 S	<0.5 S
Teicoplanin	2	<1 S	<1 S	<1 S	<1 S	≤1 S	<0.5 S	<1 S	<1 S
Erithromycin	>4 R	4 I	<0.25 S	0.5 S	0.5 S/16 R	<0.25 S/16R	≤0.25 S	<0.25 S	≤0.25 S
Ciprofloxacin	>2 R	0.12 S	0.06 S	0.25 S	1	0.5 S		2	≤0.5 S
Moxifloxacin	2	<0.5 S	<0.5 S	<0.5 S	<0.5 S	<0.5 S	<025 S	>2	<0.5 S
Levofloxacin	>4 R	<0.5 S	<0.5 S	<0.5 S	2 S	<0.5 S		<0.5 S	<0.5 S
Co-trimoxazole	>5/76 R	<0.5 S	<0.5 S	<0.5 S	<0.5 S/>4-76 R	<0.5 S/>4-76 R		>4-76 R	<0.5 S/>4-76 R
Clindamycin	0.12 S	>0.5 R	0.5 S	>0.5 R	>0.5 R	≤0.25 S	0.25 S	0.5 S	>0.5 R
Chloramphenicol	4 S	4 S	2 S	4 S	>8 R	4 S/>8 R		8	2 S
Rifampicin	<0.25 S	<0.25 S	<0.25 S	<0.25 S	>2 R	<0.25 S		–	<0.25 S
Tetracycline	<1 S	<1 S	<1 S	<1 S	<1 S	<1 S		<1 S	<1 S
Quinupristin/dalfopristin	2 I	2 I	2 I	<1 S	2 I	2 I	<0.5 S	<1 S	<1 S
Daptomycin							0.25 S
Gentamicin							≤1 S
Tobramicin							≤1 S
Linezolid							1 S
Tigecycline							<0.125 S

The specified value represents the minimum inhibitory concentration (MIC), measured in µg/mL, for each antibiotic individually. “R” denotes resistance, indicating that the studied strain is resistant to the tested antibiotic at a concentration higher than the indicated MIC value. “S” signifies susceptibility, meaning that the strain is susceptible to the tested antibiotic at a concentration lower than the specified MIC value.

^*Both strains showed the same antibiotic susceptibility profile.

^†Differences in the MIC depending on the strain.

Discussion

The identification of pathogenic organisms above the PBL indicates that large portions of the troposphere can become potential reservoirs and act as long-distance spreaders of a rich variety of microbes. These microbes are uplifted by strong air currents linked to the Siberian High and, when crossing the PBL, can travel long distances before subsiding again to the surface. Among these microbes, a vast majority of human and plant pathogens were identified, along with bacteria exhibiting antibiotic resistance profiles. In the current study, we revealed how microbial diversity values above the PBL (and up to 3,000 m a.s.l.) were comparable to those at the near-surface level, without entrainment from surface air below. This was further confirmed by both FLEXPART simulations and the daily LIDAR profiles.

Our data pointed to the air masses’ mixed agricultural origin in the NE China region, given that no dust event was detected during any of the sampling days. It has been suggested that organic matrices of transparent exopolymeric particles that strongly absorb UV radiation and may contribute to the prevention of severe dehydration are responsible for the persistence and viability of both viruses and bacteria in the upper troposphere and above and may enable viable long-distance transport (30). Our study also confirms this possibility and results in Rodó et al. (14) showing a similar direction of winds from NE China and indicates that the main source of bioaerosol provenance was the vast cereal croplands in the region (31, 32).

Worth mentioning that there is, a very limited number of studies (and consequently, few datasets) on the microbial characterization at high altitude in the troposphere, as the majority of studies were performed at only a few meters above the ground level over either terrestrial or oceanic locations (10, 33) with only few exceptions (1, 34–36). Those studies implemented pumps for air collection with very low flow rates (8.5 L/min, 14,15 L/min, and 17 L/min) and a short time of sampling (30 to 40 min, 5 min, and 60 min) (34–36). One of the reasons for this lack of studies is possibly the extremely low biomass content of air samples, which presents a challenge for the extraction of sufficient genetic material of high enough quality for efficient sequencing analysis (37). The main airborne bacterial genera detected both in flight and surface samples are Sphingomonas and Methylobacterium, which echoes other studies where they were reported as dominant airborne taxa in outdoor environments, particularly in the lower troposphere across different geographic regions (3, 38, 39) particularly in Japan (40). Those phyla ubiquitously reported in air samples include spore-forming bacteria with known UV and desiccation tolerance traits considered advantageous during atmospheric transport (41, 42). Methylobacterium can use C1-C4 carbon compounds that are ubiquitously present in the atmosphere, showcasing its adaptability to thrive in complex atmospheric conditions (43).

Other bacterial genera detected in our study are commonly distributed in natural environments (water and soil), namely: Serratia (44), Bacteroides (45), Streptococcus (46), Enterococcus (47), all of which contain species that play a role as human opportunistic pathogens.

Although the majority of Exophiala sp. are common outdoor and indoor environmental fungi, some of them are also involved in human infection (48). Of note, some Exophiala species are able to decompose toxic chemicals (49) and Cladosporium, Alternaria, and Penicillium were previously detected in air samples and considered the most prevalent fungi genera in ambient air (50, 51). Alternaria is a common fungus producing a variety of mycotoxins and a known plant pathogen (52). In grain facilities, Penicillium, Alternaria, Cladosporium were detected as predominant fungal genera, which also include plant and animal pathogenic and allergenic species (53). Other detected fungal communities include Trametes, a widespread wood decomposer that plays a role in urban wastewater purification by removing micropollutants (54) and Sporisorium, a known plant pathogen that causes sugarcane smut and is widespread in all the major sugarcane production areas in China (55).

While changes in bacterial taxa distribution in near-surface air are a well-known phenomenon (33, 56, 57), such alterations in the lower troposphere have not been widely reported. Here, we present seminal evidence demonstrating temporal changes in the bacterial and fungal diversity at high altitudes. Temperature, humidity, plant crop seasonality, and common agricultural practices significantly influence the seasonal release and diffusion of airborne microbial populations (58). Therefore, not surprisingly, the primary identified taxa at high altitudes were those that had developed an adaptation enabling them to thrive in the challenging conditions, among others Gram-positive endospore-forming Bacillus sp, as well as Exophiala, Penicillium, and Alternaria sp, which were even detected in Mars 2020 spacecraft, highlighting their high radiation resistance (59).

Interestingly, among the identified genera collected well above the PBL, there were not only environmental organisms but also known potential human pathogens. Among the principal health effects are those initially associated with respiratory complications that, in turn, may affect downstream inflammatory signaling (60), also inducing cell and organ damage (61) Among those species identified from the filters, B. fragilis is of particular interest. Despite constituting part of a normal human gut microbiome, once it leaves the intestine due to the disruption of the intestinal wall integrity, B. fragilis can cause severe infections (i.e., to name a few: sepsis; peritonitis; soft tissue infections; pelvic, lung, and brain abscesses; and a toxin-associated diarrhea) (62). Less is known about the environmental presence of B. fragilis, despite a recent study demonstrating its presence in wastewater (63). Another organism, S. marcescens, is a ubiquitous bacterium that exists in multiple ecological niches, including the air (64). It has the potential to induce a wide range of infections such as pneumonia, sepsis, meningitis, peritonitis, endocarditis, arthritis, osteomyelitis, keratitis, and urinary tract and skin infections, and its involvement in various outbreaks worldwide has been confirmed (65). The most predominant fungal species in our samples was E. oligosperma, frequently detected in the environment but that, in immunocompromised patients, it may induce soft-tissue infection, and it has been identified as a potential trigger of ocular and periocular sarcoidosis (66). Malassezia species are part of the normal skin flora of humans and animals (67) and its higher abundance in the near surface area is expected. Though M. restricta is a common skin-resident fungus, its enhanced abundance in the gut is associated with high inflammatory cytokines and recently was reported to be involved in inflammatory bowel disease, particularly Crohn’s disease, a pathology associated with changes in mycobiota (68). M. restricta has been implicated in various diseases of the skin, including dandruff, pityriasis versicolor, seborrheic dermatitis, folliculitis, psoriasis, and atopic dermatitis, thanks to its ability, under appropriate conditions, to invade the stratum corneum and interact with the host’s immune system, both directly and through chemical mediators (69). Although the link between the inhalation of airborne particles and a suite of diseases and ailments has long been suspected, not much is known at the population level about the pathogenic effects of the microbial portion of aerosols and the chemistry of dust particles.

Due to biases generated during sampling preparation, identification at the genus level has been deemed more reliable and is frequently used (70). However, in our study, to also inform at the species level, we conducted a thorough and meticulous in-depth sequential analysis by employing the RDP Naive Bayesian Classifier together with BLASTN. Following this, we only accepted taxonomic assignments when both methods, RDP Naive Bayesian Classifier and BLASTN, yielded coinciding species identities, therefore ensuring a more robust and accurate classification process.

The analysis of elemental composition of the collected PM_2.5 particles revealed high concentrations of aluminum (Al) in both ground and flight samples. Al typically occurs as silicates in clays, feldspars, kyanite, and many other minerals as the primary terrestrial sources. This observation, along with the strong correlation with K of the majority of the microbial taxa, suggests a potential agricultural origin, in line with the vast local presence of intensively farmed croplands in NE China. The appearance of strong correlations of K and Zn with the main taxa from flight samples, together with their association to clays, Cd, and Mn further reinforces the alleged agricultural origin and the potential link to areas with high loads of fertilizers and pesticides. Indeed, it is well known that the combined application of zinc sulfate nanoparticles and potassium fertilizers effectively stimulates the growth and quality of maize (Zea mays L.) in soils contaminated with metals, particularly with Cd (71). These nutrients synergistically enhance maize grain quality, ensure food safety, and alleviate plant’s stress by improving growth. This example illustrates the combined application of fertilizers, like zinc sulfate nanoparticles and sulfate of potash, in Cd-contaminated soils to boost maize biomass, alleviate abiotic stresses, and enhance the crop’s nutritional value. In turn, plants readily absorb and accumulate Cd in their tissues and in order to combat its toxicity in soils, various zinc fertilizers, including nano-fertilizers, have proven effective in enhancing zinc efficacy in such soils (72).

Interestingly, Zr and Hf are trace elements with very limited global availability and are not mined anywhere in Japan (73), despite appearing in all our samples. Instead, China being the third-largest global producer of these elements and the backward simulation paths crossing the main producing regions like Inner Mongolia (which accounts for 70% of the national total) (74), provide additional support for the findings of our study. These also coincide well with the backtrajectory simulations included indicating this area in NE China and beyond as the likely source. Additionally, B -another element highly correlated with the microorganisms found in our study- does not appear to be mined in Japan either and, again, both the coastal and inland regions of China emerge as primary mine source regions (75). The negative correlations between these elements originate from the observation of Hf and Zr predominance in February samples, while B was predominant in April samples. Finally, given the range of nanogram concentrations found in our aerosol samples, they most likely originated from terrestrial sources in one of these regions.

Metals are considered one of the most harmful components in fine particles with reported effects on pulmonary health and allergic responses (56, 76). Additionally, bioaerosols may play a role in modifying PM toxicity by modulating the oxidative potential of toxic chemicals presented within PM, thus representing a potential hazard to human health (61). Additionally, certain microbial species detected in our sampled bioaerosols, including those known to cause human allergies and respiratory diseases, further compound this health risk (77). Although some reports highlight the antimicrobial properties of B against pathogens like Candida, Aspergillus, and Staphylococcus (78), the significant positive correlations between the Acinetobacter and metals such as Cd, Fe, Cr, Mn, and Th, identified in the current study, underscore the need for a reassessment of bioaerosols’ health impact (79).

From a health perspective, cultured organisms indicate potential threats to human health, despite the fact that low dose concentrations obviously limit their impacts in healthy individuals (albeit possibly not for those susceptible or immunocompromised individuals). For instance, the potential risks of Bacillus sp. to public health have been well documented (80). B. pumilus is linked to sepsis in both immunocompromised and immunocompetent patients (81, 82). Some strains of this species can produce a heat-stable enterotoxin (83), classifying them as food-borne pathogens causing gastroenteritis. Moreover, B. pumilus can cause skin infections similar to cutaneous anthracis lesions (84). Our airborne B. pumilus strain seems susceptible to most tested antibiotics, and it has been reported to exhibit MDR activity in some cases (85, 86). B. firmus, although multidrug resistant, does not show infectivity and pathogenicity, and instead it is used as a seed treatment resulting in increased plant and root biomass (87). B. cereus is recognized as potentially pathogenic in humans. Also, food poisoning caused by B. cereus results in acute foodborne intoxication when this microorganism produces toxins. B. cereus, whose spores survive in dried raw rice, is considered a relatively common cause of gastroenteritis worldwide, and, although normally only resulting in mild symptoms, it has been associated with four fatal cases. It has also been shown to cause endocarditis mainly in intravenous drug users and patients with valvular heart disease, those with pacemakers, prosthetic mitral valves, and other underlying conditions (88). In addition, for B. cereus, bacteremia, brain abscess, and meningoencephalitis, as well as eye infections such as pan ophthalmitis and endophthalmitis, have been reported. Neonates are also at particular risk for hospital-acquired B. cereus infections (89). Different studies have also shown that B. cereus is highly resistant to penicillin, ampicillin, and ceftriaxone, which our results largely confirm. M. luteus is an environmental organism found in soil, dust, water, and in human skin microbiota as well. It is an opportunistic pathogen, which can cause infections such as general bacteremia, and endocarditis. Despite that M. luteus is normally susceptible to most antibiotics, the strain isolated in this study showed multidrug resistance, displaying resistance to carbapenems, glycopeptides, erythromycin, fluoroquinolones, and trimethoprim-sulfamethoxazole. The mechanisms underlying this resistant phenotype have not been investigated yet. In the scientific literature, only a plasmid-borne macrolide resistance has been shown, which is the mechanism of resistance at 23S rRNA adenine methyltransferase encoded in the Erm36 gene (90). If so, it could potentially be transmitted to other microorganisms through mobile genetic elements, a matter that requires further investigation.

Our study has several limitations. First, the 10 flights do not cover all year-round variability, despite already representing a significant effort to characterize the tropospheric microbiome. Second, despite our two-stage bioinformatic analysis reinforcing the presentation of species taxonomic assignments, other sequencing approaches may offer better performances.

Conclusions

Our study uncovered a rich diversity of microbial taxa being dispersed by wind currents thousands of kilometers away from their sources. The transport takes place above the PBL, far from the surface, which enables long-distance connections among geographic sites, thereby opening the door to even much longer propagations. The number of such flights is unprecedented and should further promote others to follow exploring the presence of microbial life in the free troposphere.

The isolation of harmful species to humans (i.e., also others to animals and plants), had never been reported before for such long distances. The link of aerosol particles to remote agricultural areas, leaves it open the potential role of pesticides, fertilizers, or other components derived from anthropogenic activities ultimately capable of affecting human health, far from their sources.

Also, the fact that some of the cultured bacteria showed antimicrobial resistance capacity adds further novelty to our study as such route of propagation was never reported.

While our study does not necessarily prove causality between the presence of known human pathogens in bioaerosols and health effects, it does pave the way for further research along these lines. This future avenue could examine the exchange of microbial pathogens across very long distances and of very different sorts of pathogens (bacteria, fungi, and viruses as well, although not specifically addressed in this study). Our results introduce a novel perspective on the potential transmission routes of bioaerosol particles and their association with human, plant, and animal diseases, suggesting a need for further research to explore potential mechanistic connections in more detail.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Sequencing results are available in ref. 91. Data, code, and outputs of the analysis can be accessed in ref. 92 in an open repository at GitHub with the following link: https://github.com/AirLabBcn/microbial-richness-troposphere. All other data are included in the manuscript and/or SI Appendix.