Chemical Composition and Bacterial Community in Size-Resolved Cloud Water at the Summit of Mt. Tai, China

Received: November 21, 2016
Revised: March 21, 2017
Accepted: March 25, 2017

Download Citation: RIS|BibTeX|https://doi.org/10.4209/aaqr.2016.11.0493  

Cite this article:

Zhu, C., Chen, J., Wang, X., Li, J., Wei, M., Xu, C., Xu, X., Ding, A. and Collett, Jr., J.L. (2018). Chemical Composition and Bacterial Community in Size-Resolved Cloud Water at the Summit of Mt. Tai, China. Aerosol Air Qual. Res. 18: 1-14. https://doi.org/10.4209/aaqr.2016.11.0493

HIGHLIGHTS

Keywords: Cloud sample; Three-stage collector; Chemical composition; Bacterial community.

INTRODUCTION

Clouds play a vital role because they serve as efficient receivers and reactors that change the physical or chemical properties of aerosol particles and trace gases (Biswas et al., 2008; Li et al., 2011; Boone et al., 2015). Clouds are both the source and the sink of atmospheric particles. Particles in air serve as nuclei for water condensation and for the subsequent cloud droplets formation (Hao et al., 2016). By contrast, droplets evaporate when the cloud dissipates, leaving particles in air (Peng, 2015). Cloud droplets are also beneficial to heterogeneous and aqueous reactions (Notholt et al., 1992; Huijnen et al., 2014; Hoyle et al., 2015). Trace gases and radicals react on the surface or inside of the cloud droplets, leading to changes in the composition of secondary aerosol particles (Weller and Herrmann, 2015; Whalley et al., 2015).

Cloud droplets have been observed in many locations, and their chemical compositions vary depending on the droplet size (Bator and Collett, 1997; Bower et al., 2000; Straub and Collett, 2002). The chemical properties of cloud droplets depend on the source, the supersaturation state, and the size and chemical composition of cloud condensation nuclei (CCN) (Van Pinxteren et al., 2016). Relative studies applying two-fractions collectors demonstrated that non-volatile solute concentrations tend to be relatively high in larger drops (Menon et al., 2000), and crustal species tend to be relatively low in smaller drops (Hoag and Pandis, 1999). However, observations from three droplet size fractions are limited.

Bacteria are ubiquitous in the atmosphere, with the mean concentration of approximately 1 × 10⁴ cells m^–3 (Elbert et al., 2007; Burrows et al., 2009). The bacterial communities contained in clouds and fogs can markedly influence processes that affect terrestrial or aquatic ecosystems (Kourtev et al., 2011). Example, these bacterial communities can affect global clouds distribution and precipitation by serving as CCN and potential ice nuclei (IN) (Ariya et al., 2009; Pratt et al., 2009; Joly et al., 2014); bacterial communities are major allergens and can be plant or animal pathogens (Peccia and Hernandez, 2006; Cao et al., 2014), constituting a public concern. Abundant evidence has confirmed the diversity of the bacterial community in cloud water (Bauer et al., 2003), and some studies have characterized the bacterial community (Kourtev et al., 2011; Xu et al., 2016) in cloud water samples. However, the influence of different cloud droplet size ranges on the bacterial community is sparse.

The North China Plain, from which many sources of anthropogenic pollutants originate and around which large tracts of farmland are located, is one of the fastest industrializing and developing areas in China. In the present study, a three-stage collector was used to collect cloud samples at Mt. Tai. The pH value and electrical conductivity (EC) were measured in situ while inorganic ions and trace metals were analyzed in the laboratory. The tendencies of size-resolved pH and size-resolved major ions in cloud water samples were investigated, and their causes were analyzed. MiSeq amplicon sequencing, redundancy analysis, and linear discriminant analysis effect size (LEfSe) were applied to investigate the bacterial diversity of the collected samples. The relationship between chemical composition and bacterial diversity was also evaluated. 

EXPERIMENTAL

Site and Sampling

Mt. Tai, located in the north of Tai’ai City, Shandong Province, is the highest point in the North China Plain. Size-resolved cloud water samples were collected at a meteorological station situated at the summit of Mt. Tai (36°18N, 117°13E, 1534 m a.s.l.) (Wang et al., 2011) from June 24 to August 08, 2015. In the south of Mt. Tai, plenty of farmland and many factories sited. The farmland was mainly planted with peanuts and corns in summer. And the factories contained power plants, chemical plants, and coking plants, which emitted particulate matter, acidic gases, carbon monoxide, and coal ash to the local atmospheric environment.

Size-resolved cloud water samples were collected using a three-stage Caltech Active Strand Cloud water Collector (CASCC) (Demoz et al., 1996), with 50% collection efficiency cut-off diameters (D₅₀) of 22, 16, and 4 µm for stage 1, stage 2 and stage 3 respectively (Raja et al., 2008; Spiegel et al., 2012). Cloud droplets entering the collector were separated to different size fractions and captured by the given impaction obstacles. A 500-mL precleaned closed polyethylene collection bottle was used in each stage to accumulate the size-resolved cloud water samples. Generally, only one set of samples could be collected during one cloud event, and the lowest sample volume was obtained in stage 2. If rain was expected, the collector was deployed with a downward facing inlet to prevent rain water from entering the three-stage CASCC. Cloud events were distinguished by a 6 h non-cloudy period. After sampling, the three-stage collector was cleaned using high-purity deionized water (> 18.2 MΩ). Field blanks were collected and analyzed following the same processes applied for cloud water samples.

The cloud droplet number concentration and the liquid water content were measured using a droplet size spectrometer (FM-120) with a time resolution of 1s. The PM_2.5 concentration was obtained using a Beta attenuation and optical analyzer (Thermo Scientific, model 5030 SHARP monitor).

The Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) developed by National Oceanic and Atmospheric Administration (NOAA) was applied to investigate the origins of the cloud events. The 72-h backward trajectories of cloud events were calculated every 12 h arriving at Mt. Tai, with the height of 1534m a.s.l. and four clusters were classified.

Chemical Analysis of Cloud Samples

The pH and the EC values of the cloud samples were measured immediately after sampling by a pH/conductivity analyzer (pH/COND/TEMP METER, 6350). And the measure accuracy and error were presented in supplement. The formaldehyde (HCHO), hydrogen peroxide (H₂O₂) and sulfur(IV) [S(IV)] concentrations were measured on-site (Collett et al., 1998). HCHO was determined using the acetylacetone spectrophotometric method. H₂O₂ was determined using the fluorescence standard curve method. S(IV) was measured using the pararosaniline UV-visible spectrophotometric method.

The inorganic ions of sodium (Na⁺), ammonium (NH₄⁺), potassium (K⁺), magnesium (Mg²⁺), calcium (Ca²⁺), fluorine (F^–), chlorine (Cl^–), nitrite (NO₂^–), sulfate (SO₄^2–) and nitrate (NO₃^–), were determined using ion chromatography (Dionex, ICS-90) after 1:10 dilution of the samples and subsequent filtration. Fig. 1 presents the balance of inorganic anions versus cations, including [H⁺]. Most samples exhibited a favorable ion balance, indicating the high quality of measurement.

Fig. 1. Ion balance on an equivalent basis for cations and anions. Dashed line corresponds to 1:1.

Trace metals, namely aluminum (Al), vanadium (V), chromium (Cr), manganese (Mn), iron (Fe), cobalt (Co), nickel (Ni), copper (Cu), zinc (Zn), arsenic (As), selenium (Se), cadmium (Cd) and barium (Ba) were measured by inductively coupled plasma mass spectrometry (ICP-MS, Agilent 7500a). Samples were pretreated by acidification with 1% (v/v) nitric acid to reduce the adsorption from the inner wall of container. Then the samples were stored at 4 degrees until analysis. Li, Sc, Ge, Rh, Tb and Lu were used as internal standard and the recovery rates were all in the range of 90%–110%. Black samples were also prepared as above and analyzed. The detection limit of ICP-MS analysis for the trace metals were calculated as 3 times the standard deviation of the 6 blank values (Table 3) (Li et al., 2015).

DNA Extraction and Polymerase Chain Reaction Amplification

Microbial DNA was extracted from cloud water samples using the E.Z.N.A^.® Soil DNA Kit (Omega Bio-tek, Norcross, GA, USA) according to the manufacturer’s protocols. The V4-V5 region of the bacterial 16S ribosomal RNA gene were amplified by using polymerase chain reaction (PCR) (95°C for 3 min, followed by 27 cycles at 95°C for 30 s, 55°°C for 30 s, 72°C for 45 s and a final extension at 72°C for 10 min) with the primers 515F 5’-GTGCCAGCMGCCGCGG-3’ and 907R 5’-CCGTCAATTCMTTTRAGTTT-3’; 8-bp barcodes unique to each sample were used. The PCR reactions were performed in triplicate in a 20 µL mixture containing 4 µL of 5 × FastPfu Buffer, 2 µL of 2.5 mM dNTPs, 0.8 µL of each primer (5 µM), 0.4 µL of FastPfu Polymerase, and 10 ng of template DNA. Finally, dd H₂O was added to fill out the final volume to 20 µL.

Illumina MiSeq Sequencing and Processing of Sequencing Data

Amplicons were extracted from 2% agarose gels and purified using an AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA) according to the manufacturer’s instructions and quantified using QuantiFluor™ -ST (Promega, USA). The purified amplicons were pooled in equimolar amounts and paired-end sequenced (2 × 250) on an Illumina MiSeq platform according to standard protocols. The raw reads were deposited to the NCBI Sequence Read Archive (SRA) database (Accession Number: SRX2191488).

Raw fastq files were demultiplexed and quality-filtered using QIIME (version 1.17) with the following criteria: (i) the 300-bp reads were truncated at any site receiving an average quality score of less than 20 over a 50-bp sliding window, and truncated reads shorter than 50-bp were discarded; (ii) the reads with exact barcode matching or two-nucleotide mismatch in primer matching and those containing ambiguous characters were removed; and (iii) only sequences with overlap longer than 10-bp were assembled according to their overlap sequence. Reads that could not be assembled were discarded.

Operational Taxonomic Units (OTUs) with 97% similarity were clustered using UPARSE (version 7.1 http://drive5.com/uparse/), and chimeric sequences were identified and removed using UCHIME. The taxonomy of each 16S rRNA gene sequence was analyzed using the RDP Classifier (http://rdp.cme.msu.edu/) against the silva (SSU115) 16S rRNA database by applying a confidence threshold of 70% (Amato et al., 2013). The relationships between environmental variables and the cloud water bacterial community were analyzed using redundancy analysis (RDA) with CANOCO software (version 4.56, Microcomputer Power, Inc., Ithaca, NY, USA). LEfSe analysis (Segata et al., 2011) was applied to identify the phylogenetically coherent members of the bacterial community. LEfSe analysis was used to detect the differential abundant features in the three cloud droplet size fractions (n = 10 per category), then to test the phylogenetic consistency, and finally to estimate the effect size of each differentially abundant feature by using linear discriminant analysis (LDA). Moreover, the LDA score of 3.0 was chosen as the threshold for discriminative features.

RESULTS AND DISCUSSION

The pH and Chemical Composition of Cloud Water Samples at Mt. Tai

Table 1 summarizes the basic conditions of all 17 sets of cloud samples collected during 10 cloud events at Mt. Tai. Most of the cloud events formed in late night and dissipated during the morning hours. Inside, some long events lasted for almost 2 days (43.8 h). The number concentration and the liquid water content of each cloud event ranged from 293 to 1396 cm^–3 and from 0.065 to 0.420 g m^–3, respectively. In the process of sampling, once one collection bottle of any stage reached to the maximum volume, all the collecting bottles were unslung and one set of size-resolved cloud water samples were finished. And if the cloud event continued, another set of empty collection bottles would be applied to the following sampling within 5 min. During cloud events E-1 to E-3, multiple sets of samples were collected. During cloud events E-4 to E-10, only one set of samples was collected. Table 2 provides a summary of different chemical compositions. The average values and volume weighted mean (VWM) values of pH and inorganic ions were derived for all cloud samples. 

Fig. 2 illustrates all pH values for the 10 cloud events. The pH values of all cloud samples ranged between 3.6 and 6.6, with the lowest value in stage 3, E-5, and the highest value in the stage 3, E-2.3. The maximum pH difference in one set of samples reached up to 0.97 in E-2.3. 39.2% of samples had the pH value lower than 5.6 and 9.8% of samples had the pH value lower than 4.5. The VWM pH of all collected samples was 4.74, which was extremely close to the value at Hokkaido, Japan (Yamaguchi et al., 2015), and was higher than preceding cloud observation results (3.86 in the period of 2007–2008) at the same site (Guo et al., 2012). These results indicating an increased pH tendency in the Mt. Tai cloud water. Ten sets of samples exhibited the lowest pH value in stage 3 whereas 13 sets of samples exhibited the highest pH value in stage 1, indicating that small droplets tended to be more acidic than large droplets. Hence, the dominant tendency found for size-dependent pH values was that pH values decreased with decreasing droplet size.

Fig. 2. Size-resolved pH value. In cloud event E-1 to E-3, multiple sets of samples were distinguished by numbers, example E-1.1 and E-1.2 respectively represented the first and second set of samples in E-1.

The EC values of cloud water collected during this study ranged from 26 to 547 µS cm^–1. The VWM values for stage 1, stage 2, and stage 3, were 157, 183, and 157 µS cm^–1 respectively. Regarding the ionic compositions, this study found that cloud water collected at Mt. Tai mainly contained NH₄⁺, SO₄^2– , NO₃^– and Ca²⁺, accounting for approximately 90% of the total determined ionic compositions. On average, NH₄⁺ accounted for 26.1%, 31.5% and 38.6% of the total ions in stage 1, stage 2 and stage 3. Given that Mt. Tai is surrounded by large tracts of farmland and livestock, the finding that NH₄⁺ was the most abundant cation in Mt. Tai cloud water sample was reasonable. Similar to the concentrations found in previous Mt. Tai campaigns (Guo et al., 2012), Ca²⁺ was detected at relatively high concentrations (424, 236, and 92 µeq L^–1 for stage 1, stage 2 and stage 3, respectively) in cloud water samples. SO₄^2– was the most abundant anion in the cloud water samples, with VWM concentrations of 603, 773, and 585 µeq L^–1 for stage 1, stage 2 and stage 3, respectively. NO₃^– was the second most abundant anion in the cloud water samples, with a VWM concentration no more than 51% of the SO₄^2– concentration. In previous Mt. Tai campaigns (Wang et al., 2011; Guo et al., 2012), the VWM concentrations of SO₄^2– were measured as more than 1000 µeq L^–1. However, in our study, much lower concentrations were detected. The reason was mainly because the China government donated lots of efforts to SO₂ emission reduction so the precursor of sulfate decreased. The data published on the Ministry of Environment Protection of the People’s Republic of China confirmed that 24.5% (6.09 million tons) of SO₂ was reduced from 2007 to 2015. And the data published on the Ministry of Environment Protection of Shandong Province also confirmed that 16.5% (0.3 million tons) of SO₂ was reduced from 2011 to 2015 (Fig. S1). Meanwhile, these low SO₄^2– concentrations may also be caused by our sampling time, which was from late June to the early August in summer, during when the residential heating had not yet been initiated. Thus, SO₄^2– coming from the conversion of SO₂ from the coal combustion was not accounted for in our data. Moreover, rain and cloud events are more frequent in summer than in other seasons; thus, the scavenging effect would favor for the lower SO₄^2– concentration.

The concentrations of the major ions SO₄^2–, NO₃^– and NH₄⁺ in the three droplet size fractions varied significantly between different cloud events. By contrast, these concentrations were similar in the same cloud event, particularly in E-3 samples. Mostly, the ratio between the highest and lowest concentration of major ion in one set of samples were within a factor of 3. As shown in Fig. 3, NH₄⁺ had a higher concentration in smaller (stage 2 and stage 3) droplets in almost all samples (16 of 17 sets of samples). In most cases, SO₄^2– had the highest concentrations in stage 2 (12 of 17 sets of samples) but the lowest concentration in stage 3 (13 of 17 sets of samples). NO₃^– had a higher concentration in larger (stage 1 and stage 2) droplets, with the highest concentrations in 9 sets in stage 2 and 6 sets in stage 1. The Ca²⁺ concentration was widely different in the three size fractions, and VWM concentration in stage 2 and stage 1 were twice and four times greater than that in stage 3 (Fig. 3). Although Mg²⁺ accounted for only less than 2%, the concentration of Mg²⁺ exhibited a tendency similar to Ca²⁺; the concentration in stage 3 was less than a quarter of that in other stages and some concentrations even below detection limit. 

Fig. 3. Size-resolved concentrations of sulfate, nitrate, ammonium and calcium. Stages in the same set of samples were distinguished by numbers, example E-1.1_1, E-1.1_2 and E-1.1_3 respectively represented samples of stage 1, stage 2 and stage 3 from E-1.1.

Table 3 summarizes the average concentrations of trace metals in cloud water samples. Zn was the dominant trace metal with an average concentration higher than 500 µg L^–1 for each stage. The other three trace metals in the first four followed the order: Al > Fe > Mn. Cu and Ba showed similar concentrations (approximately 25 µg L^–1 for each stage). By contrast, V, Cr, Ni, As, Se and Cd were all between 1 µg L^–1 and 10 µg L^–1. The trace metal Co was detected at the lowest concentration (less than 1 µg L^–1). The concentrations of the trace metals were the same order of magnitude as the previous data in Mt. Tai (Liu et al., 2012). But the concentration levels were much higher than in Mt. Schmucke (Fomba et al., 2015) and Mt. Brocken (Plessow et al., 2001) in Europe. Especially, V, Mn, Se and As in this study were one order of magnitude higher than in Mt. Schumcke. 

At Mt. Tai, more fine particles existed in summer and the main compounds of these fine particles were SO₄^2–, NH₄⁺ and NO₃^– (Deng et al., 2011). When these fine particles served as CCN and formed cloud events, major ions would concentrate in smaller size droplets. Moreover, smaller droplets had more specific surface area for facilitating the mass transfer of soluble gases, such as SO₂, NO_x and NH₃. Thus the major ions were concentrated in smaller size droplets. In the literature, the droplet size increased with decreasing concentrations of major ions or minimum major ions in stage 2, as observed in Mt. Schmucke (Van Pinxteren et al., 2016). Raja and co-workers (Raja et al., 2008) found that major ions were concentrated in small droplets. The tendencies of Ca²⁺ and Mg²⁺ may explained by their source, soil dust. Soil dust has a much larger size distribution (Bi et al., 2013) than the droplets collected in stage 2 and stage 3, which meant if the soil dust participated in the cloud process, the larger droplets were more possible to contain soil dust components, and if the droplets were too small, it was less likely to find the soil dust elements inside. These results strongly suggested that the size distribution and chemical composition of CCN may be the causes of the chemical heterogeneity of droplets at Mt. Tai. Other minor ions did not exhibit regular size distribution in our study. But in available literature data (Collett et al., 1995; Raja et al., 2009; Van Pinxteren et al., 2016), minor ions were found diverse size distribution profiles and this may depend on sampling sites and compound species.

The pH of cloud droplets was mainly determined by the combined effect of acidic and acid-neutralizing components. The VWM pH values for stage 1, stage 2 and stage 3 were 4.86, 4.70 and 4.67 respectively. Stage 3 samples tended to have lower pH value, whereas stage 1 samples tended to have higher pH value. This result was caused by the more acidic components in smaller droplets and the more acid-neutralizing components in larger droplets.

SO₄^2– and NO₃^– were found to be the main acidic components. By contrast, NH₄⁺ and Ca²⁺ played a crucial role in cloud water neutralization (Hutchings et al., 2009). Comparison of the pH values between stage 1 and stage 3 by using VWM concentrations revealed that although acidic components in stage 1 exceeded those in stage 3 with VWM concentrations of 18 µeq L^–1 for SO₄^2– and 62 µeq L^–1 for NO₃^–, stage 1 samples (VWM pH = 4.86) still had higher pH value than stage 3 samples (VWM pH = 4.67). This finding is because the concentration of Ca²⁺ in stage 1 (424 µeq L^–1) was higher than that in stage 3 (92 µeq L^–1). Although NH₄⁺ concentration in stage 3 (688 µeq L^–1) exceeded that in stage 1 (548 µeq L^–1), overall, acid-neutralizing components in stage 1 exceeded those in stage 3 with the value of 192 µeq L^–1. Moreover, this was also far large than the deficient 80 µeq L^–1 of acidic components.

The Ca²⁺ concentration in stage 1 exceeded that in stage 2 with the value of 188 µeq L^–1. However, NH₄⁺ concentration in stage 2 surpassed that in stage 1 with the value of 191 µeq L^–1. Thus the neutralizing effects were almost the same in the two size fractions. However, the effect from acidic components was much greater in stage 2, which had 217 µeq L^–1 more of acidic compounds (170 µeq L^–1 of SO₄^2– and 47 µeq L^–1 of NO₃^–) than stage 1. Therefore, stage 1 (VWM pH = 4.86) had higher pH values contrast to stage 2 (VWM pH = 4.70).

Three-day air mass backward trajectories were applied to estimate the potential source regions and pathway of air pollutants (Fig. 4). According to the source direction, air masses were classified into four clusters. Cluster Southwest (contained E-2, 3, 9, and 10) mainly originated from continent area Hunan and Jiangxi Province. Cluster East (contained E-4, 6, 7, and 8) mainly came from the East China Sea and traveled through the north Jiangsu Province. Samples in Cluster East had higher Cl- concentration (Fig. 4(c)) than others, which consisted well with the marine source. Cluster East and Cluster Southwest were similar that they all arrived at the Mt. Tai from the south where factories and plants located. But on average, the major ions concentrations in Cluster East were much higher than in Cluster Southwest (Fig. 4(b)). Partly or all of the following reasons may explained the phenomenon: i) the air mass from the east received pollutants in Jiangsu Province; ii) the air mass from the east contained more moisture, which availed to the conversion of the secondary inorganic ions. 

Fig. 4. (a) Three-day backward trajectories arriving at Mt. Tai; (b) Averaged major ions (including ammonium, sulfate, and nitrate) concentration of four clusters; (c) Averaged chloride concentration of four clusters.

Air mass that generated near the ground in E-5 was Cluster Local because it was traveling and revolving around Mt. Tai. However, the cloud water samples in E-5 had the lowest pH values in each stage (3.75 in E-5.1, 3.70 in E-5.2 and 3.60 in E-3) for it passed through the frequent acidic precipitation area, Beijing-Tianjin-Hebei region (Wang et al., 2012). Air mass in E-1 was Cluster Northwest, for it came from the northwest part of China. Usually, the air mass from this direction would contain crustal particles and lead higher pH values in samples (Wang et al., 2011). But in our study, E-1 samples did not show these features as the PM_2.5 concentration only 5.1 µg m^–3, the reason could be the original altitude the air mass formed was too high (more than 6 km).

Bacterial Community in Cloud Water Samples at Mt. Tai

Given the low volume of stage 2 samples, it was difficult to obtain sufficient quantity for bacterial community analysis. Therefore, bacterial community analysis was conducted for only 10 of 17 sets of samples. Fig. 4 presents the bacterial diversity in terms of phylum and genus levels. In 30 cloud water samples from the 10 sets, the mean coverage was 0.9988, indicating adequate sequencing. Overall, 30065–43674 sequences representing 157–299 OTUs were obtained. The average number of OTUs was 210 per sample. Estimators of alpha diversity were used to assess the differences in bacterial community structure in the cloud water samples. At 97% similarity, the chao1 index ranged from 185 to 339, and the Shannon diversity index ranged from 1.83 to 2.98. Compared with the chao1 index in particulate matter (2585–12067) (Gou et al., 2016) and Shannon diversity index in aerosol (2.3–3.6) (Fahlgren et al., 2011), the diversity of the bacterial community in cloud water samples was relatively low.

The relative abundance exhibited a considerably parallel distribution in the cloud water samples. The dominant microbial groups were phyla Firmicutes, Proteobacteria, Bacteroidetes, Actinobacteria, and Fusobacteria. These five phyla accounted for 99.8% of the total OTUs. The averaged relative abundance of the dominant phylum, Firmicutes, was 80.5%, and those of Proteobacteria, Bacteroidetes, Actinobacteria, and Fusobacteria were 17.4%, 1.2%, 0.46% and 0.23%, respectively. At the genus level, the dominant bacterial groups and the mean relative abundance were: Lactococcus (34.9%), Bacillus (34.0%), Pseudomonas (5.2%), Streptococcus (4.7%), Massilia (3.7%), Methylobacterium (2.9%) and Leuconostoc (2.1%), accounting for 87.5% of the total OTUs. We compared the bacterial phyla identified in this study with those in previous studies. These five phyla were also found in aerosol particles (Bowers et al., 2016) and other aquatic samples, such as lake water samples (Zeng et al., 2014), rain water samples (Evans et al., 2009; Kaushik et al., 2013; Peter et al., 2014) and river water samples (Kovatch and Schultz, 2013). But in the different kinds of samples, the dominant bacteria phylum was diverse and the relative abundance of the same bacteria phylum differed greatly (Table S4). In other cloud water studies, similar genera were revealed, but the relative abundance varied a lot. For example: the genus Pseudomonas dominated the cloud water bacterial community in puy de Dome (Vaïtilingom et al., 2012) and Hebrides (Ahern et al., 2007), but the relative abundance of this genus was only 5.2% in our study.

Bacteria from the dominant phylum Firmicutes, which affiliated to the genus Bacillus can form resistance spores (Vaïtilingom et al., 2012) that enhance survival rates under conditions such as low nutrient levels or chemical damage. Thus, expectedly, this phylum was dominant in the bacteria communities in the cloud water samples. The genus Lactococcus in the phylum Firmicutes and the genera Pseudomonas and Methylobacterium in the phylum Proteroidetes were commonly found on leaf-surface environment (Casalta and Montel, 2008; Morris et al., 2008; Redford et al., 2010), which was the type of environment provided by vegetation growing in the summer on Mt. Tai. Some taxa had potential ecological roles were identified. Pseudomonas (5.2%) had the potential of Fe(II) oxidization and Fe oxide precipitation (Liu et al., 2013). Acinetobacter (0.73%) and Comamonas (0.02%), belong to phylum Proteobacteria, played a key process for the Sb(III) oxidation (Lee, 2015). Stains in the genus Sphingomonas (0.69%) and Pseudomonas (5.2%) were correlated to the degradation of atmospheric organic matter (formate, acetate, etc.) (Amato et al., 2007). Beyond that, Acinetobacter ursingii (0.26%) was confirmed as an opportunistic human pathogen for causing bacteremia (Loubinoux et al., 2003).

Thirteen phylogenetic consistent units were identified as significant discriminators for stage 1, stage 2 or stage 3 (n = 10 per stage) (Fig. 5). Although bacterial community analysis exhibited general agreement for the dominant species, phylogenetic groups, and overall diversity, some low content OTUs differed considerably in size-resolved cloud water samples. The genus Bacteroides, which was the only genus from the family Bacteroidaceae in our samples, was the characteristic group in stage 1, the large droplet size fraction. The phylum Actinobacteria was the characteristic group in stage 3. And Pedobacter was distinctive feature of the stage 2. 

Fig. 5 Relative abundance of bacterial community in the cloud samples. A:Phylum level, B: genus level.

The reported bacterial sizes (W × L) for Bacteroides were 0.5 × 2–6 µm (Buchanan et al., 1994) for stage 1. For stage 3, six genera were found in the phylum Actinobacteria, including Arthrobacter and Micrococcus. Typical bacterial species of Arthrobacter and Micrococcus and the dimensions from the literature were: A.globiformis [φ = 0.6–0.8 µm (for coccus) or 0.5–0.8×1–4 µm (for bacillus)], A.simplex (0.4–0.5 × 0.5–0.8 µm), M.luteus (φ = 1.0–2.0 µm), M.roseus (φ = 1.0–2.5 µm), and M.varians (φ = 1.0–1.5 µm) (Buchanan et al., 1994). As shown above, the bacteria that were the significant group in stage 1 had a larger size than those in stage 3, and the bacillus with a longer length in stage 1 may serve as larger CCN compared with the coccus and bacillus in stage 3. Therefore, the significant groups in the stages were probably caused by the size of bacteria serving as CCN. 

Fig. 6 LEFSe analysis results. A: Output of the LEFSe cladogram, which identified taxonomically consistent differences between stage 1, stage 2 and stage 3 community members. Taxa with nonsignificant differences are presented as yellow circles and the diameter of the circles are proportional to relative abundance. B: LDA score. 

Influence of Environmental Factors on Bacterial Community

RDA was used to evaluate the relationships between environmental factors and cloud droplet bacterial community, and the result was illustrated in Fig. S2. The cloud water samples from the same cloud event were clustered much closer than the samples from different cloud events, especially for samples in E-3 and E-4.

Given the limited samples (10 sets) and volumes collected in the study, we did not find markedly correlation between chemical composition and bacterial community. But several minor correlations were still seen. All environmental factors explained 74.1% of the total variance. The two principal components, PC1 and PC2 that were extracted by the algorithm significantly explained 54.0% and 18.6% of the result respectively. The NO₂^– concentration positive correlated with PC1, with the coefficient of 0.194. By contrast, H₂O₂ concentration showed negative correlation, with a coefficient of –0.327. Trace metals are essential micro-nutrients for microorganisms and participate in biogeochemical cycles, which also affected the microbial community. In this study, Cr showed positive correlation with PC1, with the factor of 0.198. Concurrently, Cd and Co showed negative correlation with PC1, with the factors of –0.192 and –0.119, respectively.

The species in the genus Bacillus (B. weihenstephanensis) can express nitrate reductase enzyme for reducing NO₃^– to NO₂^– (Seenivasagan et al., 2016). Some other species in the genus Bacillus exhibited chromium-reducing ability, enabling them to survive and resist toxicity from Cr (Wani et al., 2007; Ahemad, 2015), which was consistent with the correlation factor (0.357) between PC1 and Bacillus. Firmicutes was not tolerant toward Cd (Lorenz et al., 2006), together with the fact that cobalt is toxic to microorganisms (Norberg and Molin, 1983) and the bactericidal effect of H₂O₂ (Forstrom and Wardle, 1980; Shintani, 2009), these factors exhibited a negative correlation with PC1.

CONCLUSIONS

In this study, a three-stage collector was used to collect size-resolved cloud water samples at Mt. Tai in the summer of 2015 to investigate chemical composition and bacterial diversity. The pH values of cloud water and the soluble ions exhibited size dependence. The pH value decreased with decreasing droplet size, with VWM average of 4.86 for stage 1, 4.70 for stage 2, and 4.67 for stage 3. The existences of more acidic and less acid-neutralizing substances led to lower pH value in smaller cloud droplets. The NH₄⁺ was more abundant in smaller droplets, whereas SO₄^2– and NO₃^– were more abundant in larger droplets. The Ca²⁺ and Mg²⁺ concentrations also decreased with the decreasing droplet size, and the concentrations in small droplets were extremely low.

The dominant taxa and the relative abundances were similar [Lactococcus (17.2%–41.0%) and Bacillus (17.5%–40.1%)] in each sample, but LEfSe analysis revealed disparity in bacterial community in size-resolved cloud water samples. The phylum Actinobacteria was the characteristic group for stage 3 samples. The families Bacteroidaceae and Lachnospiraceae characterized stage 1 samples. Moreover, the genus Pedobacter was distinctive feature of stage 2 samples. The size of these bacteria might affect the size-dependence distribution of bacterial community.

ACKNOWLEDGMENTS

This work was supported by Taishan Scholar Grant (ts20120552), the National Natural Science Foundation of China (Nos. 41375126, 21190053, 21177025), Cyrus Tang Foundation (No. CTF-FD2014001), Ministry of Science and Technology of China (2016YFC0202701, 2014BAC22B01), Strategic Priority Research Program of the Chinese Academy of Sciences (Grant No. XDB05010200).

Aerosol Air Qual. Res. 18 :1 -14 . https://doi.org/10.4209/aaqr.2016.11.0493