|Home | About | Journals | Submit | Contact Us | Français|
The presence of bacteria in aerosols has been known for centuries, but information on their identity and role in dispersing microbial traits is still limited. This study monitored the airborne bacterial community during an annual survey using samples collected from a 25-m tower near the Baltic Sea coast. The number of CFU was estimated using agar plates while the most probable number (MPN) of viable bacteria was estimated using dilution-to-extinction culturing assays (DCAs). The MPN and CFU values produced quantitatively similar results, displaying a pronounced seasonal pattern, with the highest numbers in winter. The most dominant bacteria growing in the DCAs all formed colonies on agar plates, were mostly pigmented (80%), and closely resembled (>97%) previously cultured bacteria based on their 16S rRNA gene sequences. 16S rRNA gene clone libraries were constructed on eight occasions during the survey; these revealed a highly diverse community with a few abundant operational taxonomic units (OTUs) and a long tail of rare OTUs. A majority of the cloned sequences (60%) were also most closely related to previously “cultured” bacteria. Thus, both culture-dependent and culture-independent techniques indicated that bacteria able to form colonies on agar plates predominate in the atmosphere. Both the DCAs and clone libraries indicated the dominance of bacteria belonging to the genera Sphingomonas sp. and Pseudomonas sp. on several sampling occasions. Potentially pathogenic strains as well as sequences closely resembling bacteria known to act as ice nuclei were found using both approaches. The origin of the sampled air mass was estimated using backward trajectories, indicating a predominant marine source.
It has long been known that microorganisms are present in the atmosphere, primarily associated with aerosols (5). These so-called bioaerosols contain bacteria as well as viruses, fungi, algae, protists, and plant material such as pollen. As estimated by microscopy, the total number of bacteria in the lower 2,000 m of the troposphere typically ranges from 1 × 103 to 1 × 105 cells m−3 (6, 25, 33). Air, however, provides a hostile environment for microorganisms, exposing them to UV light, small amounts of water, and, with increasing height, additional stress factors such as decreasing temperature, oxygen concentration, and pressure. Long-term microbial survival in the air is therefore likely limited, and a fraction of the bacteria counted with the microscope likely comprise dead cells. The number of airborne bacteria able to form colonies on solid medium typically varies from a few cells up to 1 × 104 m−3, depending on the sampling location (22, 28-30, 49). This, on the other hand, provides a minimum estimate of live bacteria since in environmental samples generally <1% of bacteria counted by microscopy form colonies on the agar plates used to determine the number of CFU (46). Although CFU may comprise only a small proportion of total bacteria, the presence of up to 104 CFU m−3 in the lower troposphere verifies that air can transfer viable bacteria. Exploring airborne microbial life underlines the versatility of microorganisms and addresses several current issues—epidemiology of zoonotic vectors, bioweapon threats, bacterial biogeography, and fundamental meteorological processes such as ice nucleation—all of which are prominent reasons to determine the viability and composition of atmospheric microorganisms.
While the number of bacteria present in the atmosphere has been investigated numerous times, the variety of methods used and the relative lack of long-term monitoring make it difficult to compile a unified background estimate. In addition, sampling location characteristics (e.g., altitude and urban versus rural sites) and temporal aspects have a large influence on the number and identity of collected bacteria. Nevertheless, some general trends can be outlined based on published sources: bacterial CFU counts are often higher in urban than in rural sites when samples are taken in the same geographic region (11, 18), bacterial CFU counts seem generally lower in coastal than inland areas (11, 43), and both fungi and bacteria occur in higher numbers in the air in summer and autumn in both rural and urban environments (11, 18, 20, 27). Bacterial abundance also varies diurnally: in urban and rural locations, the highest CFU counts have been found in the morning and evening (20, 29, 43).
Overall, available information indicates that air harbors an omnipresent bacterial community although the bacteria are low in abundance compared with, for example, bacteria in seawater and soil. This raises obvious questions as to the origin and composition of this community. The identity of bacteria collected from air has been determined in several studies using both cultivation-dependent (i.e., isolation of bacterial CFU) and cultivation-independent (e.g., cloning and sequencing the 16S rRNA gene) techniques. In general, both techniques indicate the presence of a diverse mixture of bacterial groups and species in the air. Some differences can, however, be noted. Cultivation-dependent techniques typically display a high proportion of Gram-positive bacteria (5), and species belonging to the genera Micrococcus, Bacillus, and Staphylococcus are frequently isolated from disparate sources, such as urban and rural localities (20, 30), African dust (28), and cloud and fog (4, 22). Since the year 2000, several cultivation-independent studies have been published indicating a comparatively larger proportion of Gram-negative bacteria. The dominant bacterial phyla, however, differ between sampling locations: in rural environments in northern France as well as in rural and urban environments in southern Germany, Proteobacteria dominated the community DNA (17, 33); in Boulder, CO, bacteria belonging to the Cytophaga-Flavobacteria-Bacteroidetes group were dominant (21); in an urban environment surrounding Austin and San Antonio, TX, Actinobacteria, Bacteroidetes, Firmicutes, Proteobacteria, and Verrucomicrobia were regularly detected during a 4-month sampling period (12); and at Salt Lake City, UT, bacteria belonging to Firmicutes along with Proteobacteria were prevalent in sampled air (40). The broad and partly divergent results of published studies indicate that microbial origin as well as survival may be important factors determining the composition of the airborne community. The differences between culture-dependent and culture-independent studies also indicate that the airborne community may include a fraction of bacteria that are unable to form colonies on agar plates and are thus overlooked in many culture-dependent studies. Alternatively, a fraction of bacterial DNA in the air stems from dead bacteria, resulting in cultivation-independent studies that provide a skewed profile of the airborne community. In addition, long-term time series of bacteria identified at specific locations are lacking, and, overall, surprisingly little is known about the origin and characteristics of the airborne bacterial community, for example, whether it is a community with specific life strategies or a mix of bacteria reflecting their various sources.
By using culture-dependent and culture-independent techniques as well as trajectory models, the present study aimed to determine the diversity, viability, and sources of the airborne bacterial community at a single sampling site near the Baltic Sea.
Samples were collected 25 m above ground from a tower located approximately 200 m from the Baltic Sea shoreline in the city of Kalmar, on the east coast of Sweden (56°39.576′N, 16°21.687′E). The sampling location was thus situated in the vicinity of marine, terrestrial, and urban environments. The sampling campaign extended from April 2007 to August 2008. Air was filtered using a vacuum pump. Samples for further analysis of CFU counts and dilution-to-extinction culturing assays (DCAs) were collected on gelatin filters (80-mm diameter and 3-μm pore size; Sartorius, Gottingen, Germany), while samples for DNA analyses were collected on polytetrafluoroethylene (PTFE) filters (47-mm diameter and 0.45-μm pore size; Pall Corporation, East Hills, NY). All samplings started at 06:00 a.m., and each sample was filtered for 2 h at a flow rate of 21 liters min−1, corresponding to 2.5 m3 of collected air. Samples for CFU analyses and DCAs were collected every 2 weeks, and DNA for the construction of clone libraries was collected on eight occasions during the sampling campaign. After sampling, the gelatin filter was dissolved in 15 ml of sterile Baltic seawater for 15 min at room temperature (RT). The suspension was used for estimating CFU counts and as inoculum for the DCAs. The PTFE filters used for DNA extraction were stored at −80°C until extraction.
Zobell agar plates based on Baltic seawater were used as growth medium for CFU enumerations (54). The Zobell/Baltic plates were prepared from 5 g of peptone (BD Diagnostics, Franklin Lakes, NJ), 1 g of yeast extract (BD Diagnostics), 800 ml of filtered Baltic seawater, 200 ml of ultrapure water (Milli-Q plus 185; Millipore, Billerica, MA), and 15 g of Bacto agar (BD Diagnostics). One hundred microliters of the dissolved gelatin filter was plated in triplicate and incubated at RT for 10 days. Bacterial colonies were counted every third day, and bacterial pigmentation and morphotypes were visually recorded. In the initial phase of the sampling campaign, plates based on R2A agar (Difco) as well as tryptic soy agar (TSA) and tryptic soy broth (TSB) (all from BD Diagnostics) were also tested. R2A and TSA plates were prepared according to the manufacturer's instructions, and the latter plates were labeled as 100% TSA. In addition, tryptic soy-based plates were prepared by mixing 3 g of TSB (i.e., 1/10 of original manufacturer's protocol) with 1 liter of water and 15 g of agar; these plates were labeled 10% TSA.
The DCA was modified using a protocol developed by Button and coworkers (13). Baltic seawater was used as medium to allow the growth of bacteria that are not culturable on nutrient-rich agar plates. The salinity of the water at the collection site was 7 practical salinity units (psu). The water was filtered (47-mm-diameter, 0.2-μm-pore-size filter; Supor 200; Pall Corp.), autoclaved, and refiltered. Twofold dilutions of the inoculum were made in 12 successive rows in duplicate 96-microwell plates (Nunc, Rochester, NY). Plates were incubated at RT in the dark for 3 weeks. Growth in the 96-well plates was detected by microscopy, essentially as described previously (16, 45). Briefly, 100 μl of cell suspension from each well was transferred to a 96-well filter manifold, and the cells were stained with 10 μl of 10× SYBR Gold (Invitrogen, Carlsbad, CA). Cells were collected on a black polycarbonate filter sheet of 0.2-μm pore size (Nuclepore, Millipore). The filter was cut in half and mounted on a 76- by 100-mm microscopy slide (Menzel-Glaser, Braunschweig, Germany). Growth was scored as either positive or negative, and the most probable number (MPN) of living bacteria was calculated according to the U.S. Food and Drug Administration's Bacteriological Analytical Manual (appendix 2) (10). The MPN is assumed to correspond to the total number of living bacteria, both non-colony forming and colony forming.
To estimate the proportion of colony-forming bacteria relative to non-colony-forming bacteria in the DCA, 5 μl of cell suspension from each well was spotted on a Zobell/Baltic agar plate (45). The plates were incubated for 1 week at RT. Growth in “spots” was scored as positive for colony-forming bacteria.
The most dominant bacteria in the DCAs, i.e., those able to grow at the highest dilution, were isolated on 12 occasions during the sampling campaign. A fraction of the cell suspension from the culture plate was transferred to both Zobell/Baltic liquid medium and sterile seawater. If the isolate was capable of growth in Zobell medium, this culture was chosen for further cell harvesting by centrifugation at 10,000 × g for 5 min. The seawater culture was further processed only when the isolate was incapable of growth in the rich medium. DNA from the pure isolates was extracted using an E.Z.N.A. Tissue DNA Kit (Omega Bio-Tek, Norcross, GA) according to the manufacturer's instructions. Extracted DNA was amplified using 5 μM (each) universal 16S rRNA gene primers 27F (5′-AGAGTTTGATCMTGGCTCAG-3′) and 1492R (5′-TACGGYTACCTTGTTACGACTT-3′) (Invitrogen) (53). The PCR was performed using Illustra PuReTaq Ready-To-Go PCR Beads (GE Healthcare/Life Sciences, Uppsala, Sweden) according to the manufacturer's instructions. The following cycling conditions were used: initial denaturation at 95°C for 2 min, followed by 30 cycles consisting of 95°C for 30 s, 50°C for 30 s, 72°C for 45 s, and a final elongation at 72°C for 7 min. Products were run on an agarose gel to verify the size of the amplified fragment. The PCR products were purified using an E.Z.N.A. Cycle-Pure Kit (Omega Bio-Tek) and sequenced using the chain termination method (42) with the universal bacterial primer 27F at Macrogen Corporation, Seoul, South Korea.
The DNA from cells collected on PTFE membranes was extracted using a hexadecyltrimethylammonium bromide (CTAB)-based protocol. In the protocol, 700 μl of CTAB buffer, comprising 0.1 M Tris base, 0.02 M EDTA, 1.4 M NaCl, and 0.055 M CTAB (H9151; Sigma-Aldrich, St. Louis, MO), was added to the PTFE filter and incubated for 1 h in a rotating hybridization oven at 65°C. An equal volume of water-saturated chloroform-isoamyl alcohol (24:1) was added, and the sample was mixed vigorously. Samples were then centrifuged at 9,000 × g for 40 min at RT. The top phase was transferred to new tubes, and DNA was precipitated using 1 volume of isopropanol and a 1:10 volume of 3 M sodium acetate [NaAc], pH 5.2. To increase the yield of DNA, 1 μl of a coprecipitant (nonfluorescent pellet paint, no. 70748; Novagen, Madison, WI) was added. Samples were precipitated at −20°C for a minimum of 1 h and then centrifuged at 4°C and 12,000 × g for 20 min. The supernatant was removed, and the pellet was washed with ice-cold 70% ethanol (EtOH) and centrifuged at 4°C at 12,000 × g for 20 min. The supernatant was then removed, and the pellet was dried and resuspended in 20 μl of 1× TE (Tris-EDTA) buffer. Extracted DNA was amplified using universal bacterial primers 27F and 1492R and purified using an E.Z.N.A. Cycle-Pure Kit (Omega Bio-Tek) as described above for the isolates. A blank filter was always run in parallel to the samples through both the extraction protocol and PCR to check for possible contamination. Amplified products were cloned using a TOPO TA Cloning Kit (Invitrogen). Clones were transferred to a 96-well plate containing 0.1× TE buffer. All sequencing work was done by Swegene, Lund, Sweden, using the chain termination method (42) and the universal bacterial primer 27F (Invitrogen).
The electropherograms were analyzed using the freeware program 4Peaks, version 1.7.2 (developed by A. Griekspoor and T. Groothuis; Mekentosj, Amsterdam, Netherlands). The sequences of the DCA dominants and the clone libraries were separately assembled into contigs based on ≥97% similarity using Seqman II (Lasergene, version 7; DNAstar, Madison, WI) (see Tables S1 and S2 in the supplemental material). The consensus sequences varied in length between 200 and 740 bp, with a majority being approximately 400 bp long. Each consensus sequence was considered an operational taxonomic unit (OTU) and assigned a taxonomic affiliation according to its closest relative based on a BLAST search (2, 38). The assembled contigs were aligned using the freeware program MEGA, version 4.0 (available at http://www.megasoftware.net), also used to construct the phylogenetic trees (48). The clone coverage was calculated based on Good's coverage (23). Shannon's diversity index (H) was determined for each clone library as follows: , where n is the total number of OTUs and pi is the number of sequences in each OTU divided by the total number of sequences in the clone library (44).
A standard on-site weather station provided data on air temperatures, relative humidity, wind direction, and wind speeds for each sampling occasion.
Backward trajectories are a common tool in atmospheric science that are used to derive the origin of a given air parcel. The atmospheric origin of the air studied here was calculated using the HYbrid single-particle Lagrangian integrated trajectory (HYSPLIT4) model (19). Ten-day back trajectories were calculated for each date on which a clone library was constructed. From the physical point of view, each atmospheric trajectory connects a point in space that moves through the atmosphere following the wind in a wind field that changes with weather conditions, time, and location. The HYSPLIT4 model uses three-dimensional (3-D) wind fields in which all wind data originate from the same sort of data fields used by weather forecast models. The horizontal wind data are therefore based on weather observations while the vertical wind data are derived from the initialization of the data fields for the forecast models. Statistically, 3-D trajectories can be trusted to resolve large marine area sources and vertical descents from higher altitude as far back as 5 days or more, as demonstrated by Nilsson for the marine tracer dimethylsulfide (DMS) and the stratosphere tracer ozone (O3) (35). While 3-D trajectories can resolve large-scale vertical motion well, they cannot resolve turbulent vertical motion. Air that arrives at the same geographical position but at a different height will be of a different origin due to the vertical gradients of horizontal wind in the atmosphere. Because air is mixed by turbulence within the atmospheric boundary layer (i.e., the air near the ground influenced by surface and turbulence, typically 100 to 1,000 m altitude), one must consider trajectories that arrive at more than one height within the boundary layer when analyzing the origin of an air parcel. Furthermore, air can be entrained into the boundary layer from the free troposphere. For this reason, we analyzed model trajectories for three vertical levels, arriving at 25, 500, and 1,500 m above sea level (a.s.l.).
The nucleotide sequences were deposited in GenBank under accession numbers GQ484842 to GQ484953 for the DCA dominants and GQ484363 to GQ484841 for the clones.
CFU counts were initially estimated using Zobell plates based on Baltic seawater as well as R2A and 10% and 100% TSA agar plates. The number of CFU based on growth on Zobell agar plates consistently yielded the highest number of CFU, and this was the only medium used after an initial 2-month test period (data not shown). The number of CFU ranged from 3.0 × 101 to 4.7 × 103 m−3 and displayed a clear temporal pattern, peaking in winter, i.e., the least biologically productive season (Fig. (Fig.11).
The MPN, calculated from DCAs using Baltic seawater as the growth medium, varied between 5.0 × 101 and 2.2 × 103 m−3 and generally followed the CFU count closely (Fig. (Fig.1).1). Of 41 sampling occasions, the MPN was significantly higher than the CFU count on two occasions, while the CFU count was significantly higher than the MPN on seven occasions (at the 0.95 confidence level). The CFU count was markedly higher on two occasions, on 27 June and 12 September 2007, when the CFU count exceeded the MPN by 12-fold and 8-fold, respectively.
The dominant bacteria in the DCA, i.e., those growing at the highest dilution, were isolated on 12 sampling occasions and identified based on their 16S rRNA gene sequences. In total, 112 bacteria were isolated, sequenced, and assembled into 47 contigs based on ≥97% similarity (see Table S1 in the supplemental material). Each contig was considered to represent one OTU. Based on the consensus sequences, the OTUs were assigned taxonomic affiliations using BLAST searches (2). The consensus sequences belonged to 30 genera and indicated a diverse mixture of both Gram-negative (72%) and Gram-positive (28%) bacteria dispersed among major bacterial phyla (Fig. (Fig.2).2). All DCA dominants were also able to grow on nutrient-rich solid medium, and the BLAST search revealed that all isolates except one were already represented by a cultured bacterium in GenBank. Four genera appeared on more than three sampling occasions, namely, Sphingomonas, Pseudomonas, Chryseobacterium, and Sejongia. On one sampling occasion, Xanthomonas sp. clearly dominated, making up 67% of the isolated dominants (data not shown). Visual inspection of the DCA dominants revealed diverse colors and morphotypes, with 81% of the isolates being highly pigmented. The identities of the DCA dominants were compared with two large datasets of bacteria isolated from the Baltic Sea, revealing that 80% of the DCA dominants had previously been isolated from the Baltic (41). Several bacterial isolates, based on 16S rRNA gene similarity, displayed high degrees of resemblance to genera known to include human pathogens such as Stenotrophomonas sp., Serratia sp., and Burkholderia sp.
Along with the cultivation studies, eight 16S rRNA gene clone libraries were constructed, generating a total of 479 clones that were sequenced and assembled into 169 contigs based on >97% similarity. Of the contigs, 65% were singletons, i.e., containing only one sequence (see Table S2 in the supplemental material). As for the DCA dominants, the consensus sequences were assigned taxonomic affiliations based on BLAST searches. The identified taxa were distributed among the major bacterial phyla, with on average 33% belonging to Gram-positive and 67% to Gram-negative phyla (Fig. (Fig.3).3). Of the sequences, 40% were most closely related to uncultured clones while 60% were most closely related to previously cultured bacteria. In some of the clone libraries, a large proportion of sequences belonged to plant plastids, and 5% of the assembled contigs were identified as belonging to Streptophyta (data not shown). Consequently, some of the clone libraries consisted of only approximately 30 bacterial sequences although 96 clones were initially isolated and sequenced for each library.
The relative abundance of various OTUs was determined by analyzing the number of sequences belonging to each contig. The most abundant sequences belonged to the Alpha-, Beta-, and Gammaproteobacteria, Actinobacteria, Bacteroidetes, and Firmicutes while other phyla were represented by fewer sequences and only appeared occasionally (Fig. (Fig.4).4). When individual clone libraries were analyzed, it was clear that a few OTUs dominated the airborne community on certain dates (Table (Table1;1; see also Table S2 in the supplemental material). Sphingomonas sp. made up 23% and 44% of the sequenced clones on two sampling occasions. Pseudomonas sp. appeared in six of eight samples and made up 61% of the sequenced clones on one occasion (Table (Table1).1). Propionibacterium and Janthinobacterium were less prevalent but nevertheless made up 18% and 22% of the total number of sequences, respectively, on two occasions. The OTUs that dominated on certain sampling dates also appeared repeatedly over the course of the year (see Table S2). Several sequences with high resemblance to human pathogens were also found, for example, Finegoldia magna, and several OTUs belonging to Clostridium sp. (Fig. (Fig.33).
A rank abundance curve based on all eight clone libraries was plotted with the log abundance of each identified OTU on the y axis and the OTUs ranked in order of decreasing abundance on the x axis (Fig. (Fig.5).5). The rank abundance curve confirms the presence of a few abundant OTUs and a long tail of rare OTUs over the sampling period. In addition, 87% of the OTUs were found in only one of the eight samples. However, on a single occasion, i.e., 21 November, the clone library consisted of only six OTUs, indicating that the airborne community can temporarily display low diversity (Fig. (Fig.4).4). The clone coverage for the individual clone libraries varied between 50% and 98%, and the Shannon diversity index varied between 1.2 and 3.9 (Table (Table11).
The clone libraries and DCAs were not sampled on the same dates, and different types of filters were used to collect the samples; thus, a straightforward sample-by-sample comparison between these two diversity assessments cannot be made. Still, when the identities of the DCA dominants were compared with the sequences retrieved from clone libraries, several common genera were found, for example, Sphingomonas, Pseudomonas, Janthinobacterium, Sejongia, Acinetobacter, Psychrobacter, and Pedobacter.
The CFU and MPN values were generally higher when the wind came from the east to southeast and from the southwest to west, corresponding to a possible marine origin from the Baltic Sea and the from Skagerrak/Kattegat/North Sea/Atlantic Ocean, respectively (analysis not shown). Values were mostly lower when the wind was from the south and north, corresponding to the European continent and the Scandinavian Peninsula, respectively.
The trajectories calculated for each date on which clone libraries were sampled indicated that the most common air origin was marine (Fig. (Fig.66 and Table Table1).1). For example, the air mass sampled on 11 July 2007 was of remote marine origin (i.e., the North Sea 24 h previously), followed by considerable precipitation and washout over Denmark/Sweden, while the trajectory suggests that the air mass moved above the boundary layer, thereby avoiding continental influence. The second most common air composition was a mixture of air from marine and continental sources. This can be exemplified by the sample of 4 August 2008, which evidenced a fresh marine source contribution that was very brief (only 6 h without and 6 h with rain) as well as contributions from a more distant North Atlantic source and continental passages over Denmark, the North Sea coast, and France, with several rain events. The sample collected on 20 April 2007 was the only one in which a continental origin clearly dominated, with the air having spent the last 36 h over Scandinavia and the last 12 h without rain. The nearest marine source was the North Atlantic more than 36 h before sampling, after which there were large amounts of rain.
In this paper, we have presented the results of an 18-month study of the airborne bacterial community at a small-city location near the Baltic Sea coast in southern Sweden. The discussion is centered on four results we find particularly interesting: the dominance of typical colony-forming bacteria, diversity of the airborne community, characteristics of dominant clones and isolates, and bacterial origin. Finally, the implications of our results for understanding the life strategy of airborne bacteria are discussed.
Dilution culturing assays (DCAs) as used here are presumed to provide an estimate of the most probable number (MPN) of live bacteria in a sample, including those bacteria unable to form colonies on agar plates. Interestingly, the CFU and MPN values in the air samples followed each other closely, with MPN exceeding CFU values significantly on only two occasions. In addition, all the bacteria isolated from the most diluted culture wells (here termed DCA dominants) were able to form colonies on agar plates. Thus, our results indicate that live airborne bacteria were predominantly what are referred to as colony-forming bacteria. This contrasts to previous findings from, for example, the marine environment, where MPN estimates based on seawater cultures usually result in numerical estimates that are much higher than CFU counts and where dominants in the DCAs are usually difficult to isolate due to their inability to grow in rich and/or solid medium (13, 45). A possible limitation of the DCAs presented here is that only one cultivation medium was used, i.e., Baltic seawater. This could have resulted in biased growth of bacteria originating from the marine environment. However, since the Baltic seawater used for the DCA has a salinity of 7 psu and since human tissue has a salinity equivalent to 9 psu and phosphate-buffered saline at approximately 10 psu is a common diluent for soil samples (52), the medium had a salt content well suited to support the growth of bacteria from all three environments. In addition, of the tested solid media, the Zobell plates, based on Baltic seawater, consistently produced the highest number of CFU. The suitability of the medium was further supported by the taxonomic affiliation of the DCA dominants: many of the isolates (e.g., spore-forming bacteria) are common in the terrestrial environment, and several human pathogens were also found.
The finding that CFU counts greatly exceeded MPN values on two occasions (Fig. (Fig.1)1) is somewhat disturbing. A possible but unsubstantiated explanation is that the agar plates used in the CFU estimation were contaminated on these two dates, thus overestimating the number of CFU. However, we cannot exclude the possibility that these air samples may have included bacteria unable to reproduce in the DCA, thus leading to an underestimation of the MPN. On the other hand, of the 169 assembled contigs from the clone libraries, only 40% were most closely related to what are referred to as uncultured clones while in clone libraries constructed from seawater or soils, this fraction is typically 60 to 80% (26, 39). Thus, the taxonomic affiliation of the sequences retrieved from the clone libraries supports the conclusion that typically “culturable” bacteria dominate the airborne community.
Taxonomic comparison of DCA dominants and sequenced clones gives important clues to the composition and diversity of the airborne community. The DCAs and clone libraries were sampled on different dates using different types of filters and thus cannot be directly compared. Different filters were used for cell collection because gelatin filters, which dissolve completely before further processing, allow the quantitative estimation of bacterial numbers, which was desirable for estimating CFU and MPN values. However, these filters cannot be used for DNA studies since the gelatin contains DNA of bacterial origin (tested in our lab [results not shown]). On the other hand, the PTFE filters used for the DNA studies cannot provide a quantitative estimate of cell numbers since the bacteria would have to be detached from the membrane, entailing the loss of an unknown number of cells. However, despite the different filters used for collecting cells, it is obvious that isolates/clones belonging to the genera Sphingomonas, Pseudomonas Janthinobacterium, and Psychrobacter appeared in relatively great abundance repeatedly over the year, as detected using both types of filters. This implies that these genera are the most dominant in the sampled environment. Other than these genera, the DCA dominants and clone sequences displayed little similarity to each other. Some of the discrepancy may have stemmed from the high diversity of the airborne bacteria: of the 169 OTUs identified from the clone libraries, 87% appeared on only one of the eight sampling occasions. A high degree of diversity was also indicated by the rank abundance curve, based on all eight clone libraries, which displayed a long tail of bacteria appearing in low abundances. Given the large proportion of rare bacteria, the same species can be expected to appear only occasionally or when dominant, as found in this study. The high diversity of the airborne bacterial community found here is in agreement with previous studies using a clone library (21, 41) or phylochip assay (12) approach. The relatively large proportion of clones identified as chloroplast sequences were also observed before when 16S rRNA gene clone libraries were constructed from air samples (33, 40). The lack of similarity between sampling occasions and the lack of seasonal succession observed in the present study have also been reported in several previous studies that found that the airborne community displays high temporal variation with a considerable change in the bacterial community composition over weeks and even days (12, 21).
Several of the OTUs included in the clone libraries were closely related to bacteria associated with interesting properties (see Table S1 in the supplemental material). Several facultative and obligate anaerobic bacteria were found, such as Jeotgalicoccus halotolerans and Clostridium glycolicum. The sequences of obligate anaerobic bacteria found among the clones may stem from dead bacteria but could also represent bacteria living inside anaerobic aggregates or persisting in a spore state. Several opportunistic human pathogens were found, such as F. magna and close relatives to Clostridium botulinum. Some 16S rRNA gene sequences closely resembling bacteria that carry resistance to multiple antibiotics and behave as opportunistic human pathogens were detected in both the clone libraries and DCAs, such as Citrobacter sp. and Stenotrophomonas maltophilia. Pseudomonas fluorescens and Pantoea agglomerans, two strains known to act as airborne ice-nucleating bacteria (INA) (31), were also found. INA bacteria can catalyze ice crystal formation at temperatures as high as −2°C, initiating precipitation and subsequent water deposition (rain) (1, 24, 34).
To interpret the variation in an airborne community, it is crucial to know the source of the sampled air mass. The origin of the sampled air mass, as described here by the backward trajectories, gives important information. The composition of the bacterial community gives further clues as to the air mass origin since the composition generally differs between terrestrial and aquatic environments: the Shannon diversity index is typically >2.5 in soil (47) and <2.5 in marine and fresh waters (26); soil typically harbors about 50% Gram-negative bacteria (7) while in marine waters >90% of the bacteria are Gram negative (38); finally, some species are also more common in one type of environment than in another. When we examined the data from the clone libraries and from the trajectories (Table (Table1),1), we could distinguish some systematic correlation among these eight samples. The most common air mass origin was marine from the Baltic Sea; the influence of the North Sea and North Atlantic was observed on only a few occasions. The clearly marine cases displayed Shannon indexes near or below 2.5, a high percentage of Gram-negative bacteria, and dominant species commonly found in aquatic environments. On 11 July 2007, for example, the trajectory analysis indicated an air mass of marine origin, and the community analysis displayed a relatively low Shannon index (1.3), a high proportion of Gram-negative bacteria (86%), and dominance by the genus Pseudomonas, which is common in both water and soil.
An air mass sample of mixed origin is exemplified by the sample collected on 4 August 2008, which displayed the highest diversity of all samples (3.9), with 67% Gram-positive and 33% Gram-negative sequences; the most dominant sequence made up less than 10% of the clones, and both Gram-positive and -negative OTUs were among the dominant clones (Table (Table1).1). The complex source background of this sample may explain the high diversity and mixed composition of taxa in this sample. Finally, a sample clearly of continental origin was collected on 20 April 2007. This sample displayed a high Shannon index (3.2), had a relatively high proportion of Gram-positive bacteria (39%), and was dominated by Propionibacterium acnes, a species commonly associated with human skin but that has also been detected in soil (19) and desert dust (37). Thus, this clone library displayed a clear terrestrial profile, and a correlation between origin of the air and the composition of the bacterial community seems apparent.
A predominantly marine origin was indirectly supported by the seasonal CFU count and MPN pattern. Somewhat surprisingly, these values were higher in winter than in summer months when biological activity is the highest. However, marine bacteria are ejected into the air along with sea spray aerosol particles, and this is favored by high wind speeds (36, 37). In winter months, it is reasonable to expect the wind speed over the sea source regions, upwind of the sampling location, to be higher since cyclone activity is generally greater in winter than in summer.
The reasoning underlying the community composition and air mass origin posited in the above paragraph assumes that airborne communities mirror their sources. It is, however, possible that the airborne bacterial community could result from in situ development and a particular life strategy. In fact, the finding of a close relationship between CFU and MPN values and the high frequency of sequences related to cultured bacteria in the clone libraries show that the proportion of culturable bacteria is greater in air than in either the terrestrial or aquatic environment (26, 39), which in turn indicates that the air harbors a distinct bacterial community. We can envision three forces that could influence the composition of airborne bacteria. First, considering the atmosphere as an extreme environment, bacterial cells are likely to employ various strategies to survive in atmospheric conditions, such as UV exposure or the risk of dehydration. In this study, we demonstrate that 81% of the DCA dominants were pigmented. UV radiation is known to have a markedly lethal effect on bacteria (50), and pigmentation is likewise known to protect bacterial cells from UV damage (51). Pigmentation could therefore be a selective trait that favors the survival of certain bacterial species in the atmosphere.
Second, if bacteria can grow in the atmosphere, this could possibly underlie the changed community composition compared with that of possible sources. Air is generally considered a nutrient-poor environment, and active growth in aerosols is difficult to demonstrate (47). However, atmospheric aerosol can harbor reduced carbon and mineral nutrients on the order of 0.5 μmol of C m−3 (3), 100 nmol of N m−3 (3), and 1 nmol of P m−3 (15). In terms of the C/N/P requirements of bacteria, phosphorus is clearly the limiting factor for bacterial growth; still, based on an average bacterial P content of 3.8 × 10−17 mol of P, the atmosphere can sustain growth of about 2.5 × 107 cells m−3. This is hypothetical since the nutrients are dispersed and not readily available to the bacterial cells. Still, compared with the typical number of bacteria in air, i.e., 101 to 105 m−3 depending on the determination method (22, 28-30), the amount of nutrients is sufficient to allow bacteria to multiply many times.
As a third alternative, we speculate that perhaps only a subset of bacteria is ejected into the air. The major source of primary marine aerosol particles is bursting bubbles from breaking waves caused by wind drag on the ocean surface (8). The material reaching the atmosphere as aerosol particles consists of sea salt, organic matter, and marine microorganisms (9, 14, 32). In this process, it is easy to envisage that selective factors such as surface charge and particle size may determine the bacterial composition in the aerosol.
In conclusion, we have demonstrated that the airborne bacterial community harbors a highly diverse community including members of all major bacterial phyla and dominated by typical culturable bacteria. At the studied coastal sampling site, we demonstrated that the bioaerosol was predominantly of marine origin, as indicated by both the atmospheric trajectories and the composition of bacteria in the sampled air. These findings provide important clues for understanding the global distribution of microorganisms in general and the ecology of airborne bacteria in particular. In addition, we have demonstrated that the airborne community harbors bacteria that are interesting from a health perspective and for science related to meteorological processes.
Sabina Arnautovic is acknowledged for her skillful technical assistance.
This work was financed by EU grant SEC6-PR-214400 (Aerobactics) and FORMAS (The Swedish Research Council for Environment, Agricultural Science and Spatial Planning) grant 217-2005-1626. We also acknowledge the Swedish Research Council (grant 622-2003-1120) and FORMAS grants 2004-2047 (project Ocean—atmosphere exchange of organic, biological and toxic aerosols) and 2008-1113 (project Aerobiology: a factor to consider in environmental change).
Published ahead of print on 12 March 2010.
†Supplemental material for this article may be found at http://aem.asm.org/.