OUP user menu

Microbial diversity of eolian dust sources from saline lake sediments and biological soil crusts in arid Southern Australia

Raeid M.M. Abed, Alban Ramette, Vera Hübner, Patrick De Deckker, Dirk de Beer
DOI: http://dx.doi.org/10.1111/j.1574-6941.2011.01289.x 294-304 First published online: 1 May 2012

Abstract

While microbial communities of aerosols have been examined, little is known about their sources. Nutrient composition and microbial communities of potential dust sources, saline lake sediments (SLS) and adjacent biological soil crusts (BSC), from Southern Australia were determined and compared with a previously analyzed dust sample. Multivariate analyses of fingerprinting profiles indicated that the bacterial communities of SLS and BSC were different, and these differences were mainly explained by salinity. Nutrient concentrations varied among the sites but could not explain the differences in microbial diversity patterns. Comparison of microbial communities with dust samples showed that deflation selects against filamentous cyanobacteria, such as the Nostocales group. This could be attributed to the firm attachment of cyanobacterial filaments to soil particles and/or because deflation occurs mainly in disturbed BSC, where cyanobacterial diversity is often low. Other bacterial groups, such as Actinobacteria and the spore-forming Firmicutes, were found in both dust and its sources. While Firmicutes-related sequences were mostly detected in the SLS bacterial communities (10% of total sequences), the actinobacterial sequences were retrieved from both (11–13%). In conclusion, the potential dust sources examined here show highly diverse bacterial communities and contain nutrients that can be transported with aerosols. The obtained fingerprinting and sequencing data may enable back tracking of dust plumes and their microorganisms.

Keywords
  • aerosols
  • ARISA
  • cyanobacteria
  • Firmicutes
  • nutrients
  • pyrosequencing

Introduction

Airborne dust can affect global climate, human health as well as ocean and terrestrial environments (Stoorvogel et al., 1997; Griffin et al., 2001; McTainsh et al., 2005). Dust plumes from the Sahara in Northern Africa, known as the largest source of soil to Earth's atmosphere, are known to spread over long distances across the Atlantic Ocean and as far as the Caribbean, SE United States and the Amazon basin (Swap et al., 1992; Prospero, 1999; Prospero et al., 2005), whereas Chinese dust, in one occasion, was known to have traveled one full circuit around the globe in about 13 days (Uno et al., 2009). Recently, it became evident that aerosols are rich in viable bacterial populations and can act as a carrier for a variety of microorganisms that may impact human health and the environment (Griffin et al., 2001; Prospero et al., 2005; Brodie et al., 2007). In the Southern Hemisphere, Australia is the main source of dust, which exits the continent frequently and deposits into the Tasman Sea and can go as far as New Zealand as well into the eastern Indian Ocean (Bowler, 1976; Hesse, 1994; McGowan et al., 2000; De Deckker et al., 2010). Numerous dust storms have been documented in Australia (Liversidge, 1902; Raupach et al., 1994), including the dust plume that engulfed Canberra in October 2002 (McTainsh et al., 2005; De Deckker et al., 2008) and more recently Sydney in September 2009. A detailed characterization of dust samples from the former event showed that the dust was generated from wind erosion of the regolith in arid and semi-arid regions and harbored a large diversity of soil-related microorganisms (De Deckker et al., 2008). Using satellite imagery and geochemical fingerprinting, the primary source of this dust was identified to be from inland saline lake sediments (SLS) and adjacent biological soil crusts (BSC) of southeastern Australia. Although the microbial composition of dusts from Australia has been studied using cultivation and molecular tools (De Deckker et al., 2008; Munday C & Allison G, unpublished), little is known about the bacterial community structure of their sources. It is essential to create a microbiological database of deflating areas, which erode as a consequence of sand and dust being removed from SLS and BSC by the wind, to gain a better understanding of the origin, bacterial composition and transport pathways of eolian dust.

Inland saline lakes are numerous in Australia, encompassing a wide range of salinities (up to 35%) and differ in their morphometry, geology and degree of permanence (Williams, 1981b; De Deckker, 1988; De Deckker & Williams, 1988). The Southern Australian lakes are characterized by their semi-arid to arid environments, with a maximum temperature of > 45 °C reached in some summers and an annual evaporation rate of well over 1000 mm (Williams, 1981b). These lakes have been studied for their geology, chemical limnology and biota (Williams, 1981a, b; De Deckker & Williams, 1988 and references therein); however, little is known about their microbial composition. Likewise, the BSC surrounding these lakes have not been investigated for the diversity of their microorganisms, although much is known about the Australian crusts’ eukaryotic diversity (i.e. mosses and lichen) and their role in soil stability and fertilization (Eldridge, 2001 and references therein; Eldridge et al., 2006; Read et al., 2008). We will consider the possibility that the BSC community near a lake is influenced by the adjacent lake community because of the deposition of lake bottom sediment.

The aim of this study is threefold: (1) to create a database of DNA fingerprints and 16S rRNA sequences of microbial communities in SLS and their adjacent BSC, identified previously as potential sources of the Canberra 2002 dust plume, (2) to find out the effects of different environmental parameters on their structure, and (3) to compare bacterial communities of the dust sample with those potential sources.

Materials and methods

Sampling sites and nutrient analysis

A total of 34 samples from SLS and BSC were collected from 18 dry lake beds in Southern Australia and their surrounding sand dunes (Fig. , Table ). The sampled lakes differed in their salinity, geological surroundings, soil type and vegetation, whereas the BSC were dry and had different surface color. From each locality, 3–8 replicates were collected and immediately stored in liquid nitrogen for further molecular analysis. The samples were shipped frozen to the Max-Planck Institute in Bremen where they were analyzed for nutrient concentrations and microbial community composition.

A map of Southern Australia and New South Wales, separated by the dotted line, showing the sampling locations. Samples labeled with S and C refer to those collected from SLS and BSC, respectively.

View this table:

Geographical location and nutrient composition of SLS and nearby BSC identified as potential dust sources

Lake name/localitySample NumberSample typeSalinity (%)Latitude SLongitude ENutrient concentration (μmol g−1)
FeSilicatePhosphateAmmoniumNitriteNitrite and nitrate
DuttonSISLS4.231°59.090′137°50.899′NANANANANANA
C4BSC0.531°47.909′137°11.162′39.10.770.010.270.010.09
PernattyS5SLS6.031°39.739′137°16.424′76.30.540.200.120.000.03
C6BSC0.131°40.084′137°16.629′14.30.750.010.070.010.05
FinnissC9BSC0.131°42.447′136°49.657′3.50.310.030.100.010.06
Island LagoonS10SLS5.331°24.049′136°52.016′247.60.362.690.840.000.04
CllBSC0.231°23.916′136°52.214′3.80.060.010.040.000.06
TorrensS12SLS5.030°28.273′137°21.399′47.80.540.240.290.000.02
C13BSC0.330°28.129′137°21.485′14.80.410.000.070.010.02
PanS14SLS0.230°07.420′137°02.133′20.80.820.010.030.010.10
C15BSC0.130°07.460′137°02.313′5.30.710.000.180.010.02
HartS17SLS4.431°13.526′136°24.246′35.40.510.010.090.000.02
C18BSC0.531°13.648′136°24.435′4.40.340.000.070.000.01
GairdnerS19SLS5.031°13.249′135°41.324′92.01.280.010.160.000.01
C20BSC0.231°13.101′135°41.445′18.90.620.000.020.000.00
HarrisS21SLS3.731°01.644′135°21.877′42.30.890.010.140.000.01
C22BSC0.331°01.653′135°21.989′46.01.120.010.060.020.10
EdwardC24BSC0.631°35.773′135°25.887′78.21.190.010.090.020.03
C26BSC0.432°18.738′134°59.390′2.60.230.000.020.000.00
Gairdner SouthC27BSC0.232°23.166′135°29.386′2.11.000.000.250.010.02
S29SLS6.632°07.469′135°53.877′25.10.120.000.040.000.01
C30BSC0.532°07.399′135°53.810′5.40.940.010.200.050.16
S31SLS2.232°07.399′135°53.810′17.50.140.000.100.000.00
Scrubby PeakC34BSC0.132°36.657′135°11.346′14.30.910.020.070.030.05
YanineeC36BSC0.332°57.995′135°16.236′13.60.800.000.040.010.03
RoundS38SLS3.033°56.470′135°16.498′1.63.370.013.580.010.02
GreenlyC41BSC0.134°09.324′135°26.833′15.21.270.001.100.010.01
GillesS42SLS5.433°01.905′136°06.165′23.90.690.031.000.000.01
C43BSC1.533°01.913′136°36.127′52.40.040.442.380.000.04
Mundi Mundi PlainsC44BSC1.531°47.346′141°11.141′112.10.030.361.250.000.07
MenindeeC45BSC1.532°03.058′141°37.044′203.50.031.093.170.000.04
S46SLS3.232°17.750′142°22.260′17.90.051.341.580.000.04
S48SLS3.532°21.284′142°24.053′74.00.041.733.230.000.00
C49BSC1.232°30.401′143°00.700′46.20.030.990.700.000.09
  • NA, data not available.

For nutrient analysis (silicate, phosphate, ammonium, nitrite and nitrate), dry samples were placed in 15 mL falcon tubes and 10 mL of distilled water was added. The tubes were shaken at a speed of 150 rpm for 24 h in a cold room (T = 4 °C). The water phase was collected by centrifugation and was then filtered through 2-μm filter syringes. The nutrients were measured in the water samples using a nutrient analyzer (San Plus Analyzer; Skalar analytical GmbH, Erkelenz, Germany). For the iron (Fe) extraction, 10 mL of 0.5 M hydrochloric acid (HCl) was added to 0.2 g of sediment in a plastic tube. The tubes were shaken for 1 h and then centrifuged at 4000 rpm for 5 min at room temperature. The supernatant was collected in a separate tube and kept at 4 °C until analysis could be performed. A solution of Ferrozin [3-(2-pyridye)-5,6-diphenyl-1,2,4-triazine-4′, 4″-disulforic acid monosodium salt/HEPES (N-2-hydroxyethylpiperazine-N-2-ethanesulfonic acid)] buffer was freshly prepared (12 g HEPES + 1 g Ferrozin in 1 L distilled water), and 200 mL of this solution was used to dissolve 2 g of Hydroxlaminchloride (Ferrozin buffer). Different volumes of the sediment extracts were mixed with 2 mL of the Ferrozin buffer, and the mixture was incubated for 15 min. The developed violet color was measured at 562 nm using a spectrophotometer (Shimadzu UV-160A Spectrophotometer, Duisburg, Germany).

Microbial community analysis by automated rRNA intergenic spacer analysis (ARISA)

The upper 1–2 mm of the samples (ca. 300–500 mg each) were subjected to DNA extraction using the UltraClean soil DNA isolation kit (MOBIO laboratories, Inc., Carlsbad, CA) according to the manufacturer's instructions. Polymerase chain reaction (PCR) amplification was performed in triplicate with the same amounts of DNA (as determined by Nanodrop, Thermo Scientific, Germany) using the universal primer ITSF and the FAM-labeled eubacterial ITSReub primer (Cardinale et al., 2004) at an annealing temperature of 55 °C. The PCR products were purified using Sephadex G-50 Superfine (Sigma-Aldrich, Munich, Germany). 150 ng of DNA was then mixed with 0.5 μL of internal size standard MapMarker 1000 ROX (50–1000 bp; BioVentures Inc., Washington, DC), and the amplified fragments were discriminated by capillary electrophoresis on an ABI PRISM 3130xl Genetic Analyzer (Applied Biosystems, Carlsbad, CA). The ARISA profiles were analyzed using the GeneMapper software v 3.7 (Applied Biosystems). The total peak area per sample was normalized to one, and only fragments between 100 and 1000 bp were considered. A ‘fixed window’ binning strategy with a bin size of 2 bp was applied to the ARISA data (Ramette, 2009), and the binning frame that offered the highest pairwise similarities among samples was subjected to multivariate analyses. An operational taxonomic unit (OTU) was considered present in a given sample only if it was detected at least twice among the three replicated PCRs from the DNA extracts of that particular sample (Ramette, 2009).

Statistical analysis of ARISA fingerprints was carried out using the PAST program (Paleontological Statistics, ver. 1.47, http://folk.uio.no\ohammer\past). A multivariate analysis of all sites was performed using multidimensional scaling (MDS) based on Bray–Curtis similarities as described in (Ramette, 2007). Ordination of the Bray–Curtis similarities was performed using non-metric MDS, with 100 random restarts, and the results were plotted in two-dimensions. The consensus ARISA table containing samples by OTUs was used to calculate pairwise similarities among samples based on the Bray–Curtis similarity index. Analysis of similarities (ANOSIM) was carried out to test for significant differences between SLS and BSC microbial communities. ANOSIM produces a sample statistic (R), which represents the degree of separation between test groups (Clarke, 1993; Ramette, 2007).

Canonical redundancy analysis (RDA) was used to investigate the significance of salinity, sample type (SLS vs. BSC), geographical location, vegetation, soil type, and rainfall on the shape of microbial communities in SLS and BSC samples. ARISA profiles were first Hellinger transformed (Legendre & Gallagher, 2001; Ramette, 2007), and the effect of different parameters was investigated by canonical variation partitioning (Borcard et al., 1992; Ramette & Tiedje, 2007), where the variation and covariation of these parameters were partitioned into pure and covarying fractions. For each response data model, the most significant variables were selected by RDA using stepwise selection and by minimizing the Akaike Information Criterion. Statistical significances were assessed by 1000 permutation of the reduced models. All statistical calculations were performed with the R statistical platform using the vegan and mass packages.

Pyrosequencing and sequence analyses

Purified DNA extracts of 20 representative SLS and BSC samples (10 each) from different geological provinces were submitted to the Research and Testing Laboratory (RTL, Lubbock, TX) for tag pyrosequencing. Bacterial tag-encoded FLX amplicon pyrosequencing (bTEFAP) was performed as described before (Dowd et al.,2008a, b) using the GS FLX titanium sequencing kit XLR70. One-step PCR was performed using a mixture of hot start and hot start high fidelity tag polymerases resulting in amplicons that extend 350–450 bp from the 27F region (Escherichia coli rRNA numbering). The bTEFAP sequencing was performed according to the RTL protocols (www.researchandtesting.com). Obtained sequences were trimmed; low quality ends and tags were removed and were checked for chimeras using custom software (Dowd et al., 2008b) and the Black Box Chimera Check software b2c2 (freely available at http://www.researchandtesting.com/B2C2.html). Sequences < 300 bp were excluded from further analysis, and the rest were checked for high quality based on criteria utilized by rdp ver 9 (Cole et al., 2009). These sequences were analyzed and taxonomically classified using a distributed BLASTn.NET algorithm (Dowd et al., 2005) against a database of high quality 16S rRNA gene sequences obtained from NCBI. The outputs were compiled and validated using taxonomic distance methods (Dowd et al., 2008a, b). Rarefaction curves, OTU number, and diversity estimates (Chao 1, ACE) were calculated using the software esprit (Sun et al., 2009) available freely at http://www.biotech.ufl.edu/people/sun/esprit.html. For large sequence datasets, the average diversity indices were calculated after performing 3–5 randomized selection of 4500 and 3000 sequences for SLS and BSC samples, using a custom script.

Results

Iron and nutrients composition

The analyzed suite of nutrients was selected because they have been shown to disperse with dust over large distances (McTainsh & Strong, 2007). Clear variations were observed in Fe and nutrient concentrations among the studied sites (Table ). Fe concentration ranged from 1.6 to 248 μmol g−1 soil with the lowest values detected in Round Lake and Lake Gairdner South and highest in Island Lagoon and the Menindee region. Most samples collected from near Menindee had higher concentrations of Fe, phosphate, and ammonium than the other samples, whereas the concentrations of silicate were lower. Phosphate concentration was < 0.2 μmol g−1 in all samples except in the Menindee samples where it ranged between 0.36 and 1.73 μmol g−1. The SLS had higher phosphate concentrations than the surrounding BSC. Nitrite and nitrate concentrations were < 0.16 μmol g−1 in all samples, while ammonium concentrations ranged from 0.02 to 3.58 μmol g−1. Both SLS and BSC samples from the Menindee-Broken Hill area showed the highest total nitrogen concentrations of all samples. The N:P ratio showed clear variations between sites and ranged from 1 : 1 to 600 : 1, thus some sites were nitrogen-limited while others were phosphate-limited.

Fingerprinting analysis

ARISA fingerprinting of bacterial communities from samples of SLS and their adjacent BSC showed significant variations in community composition as well as OTU richness. A total pool of 419 distinct OTUs (i.e. binned ARISA peaks) distributed among all samples were identified. The average number of OTUs for the SLS samples was 82 with the highest detected in Pan and Round Lake sediments (i.e. 136 and 138, respectively) and the lowest in Lake Harris and Lake Gairdner sediments (i.e. 20 and 28, respectively; Fig. left panel). The highest OTU number of BSC was detected in Finniss Lake and Island Lagoon sites with 156 and 160 OTUs, respectively, and the lowest in Lakes Hart and Greenly with 53 and 77 OTUs, respectively. Multivariate analyses of ARISA profiles of SLS identified specific bacterial signatures when compared to the surrounding BSC, and the bacterial communities of the two groups clearly clustered separately, yet with a certain degree of overlap (Fig. right panel, ANOSIM R = 0.45, P < 0.0001). Pairwise comparison of presence/absence of OTUs in the SLS and BSC of 10 selected sites showed only < 18% shared OTUs among all samples with the highest average shared OTUs detected in Lake Harris and lowest in Lake Gairdner (Fig. left panel).

Non-metric multidimensional scaling (NMDS) ordination (based on a Bray–Curtis distance matrix) of ARISA fingerprints (left) from SLS and BSC samples, identified as potential sources of eolian dust. The right-hand figure shows the percentage of shared and not shared OTUs in 20 representative SLS and BSC from 10 sampling locations. Per, Pernatty; Is-La, Island Lagoon; Gaird, Gairdner; Gair-S, Gairdner South; Menin, Menindee. The white and black circles represent the number of OTUs obtained from ARISA profiles of SLS and BSC at each location, respectively.

To obtain additional information on the environmental parameters that shaped the microbial communities in SLS and BSC, the relationships between ARISA fingerprints and salinity, sample type (SLS vs. BSC), geographical location, rainfall, vegetation, soil/regolith type, nutrients and Fe concentrations were statistically evaluated (Table ). Among all parameters, salinity was the most important structuring factor (10.7% of variance explained, P = 0.001) followed by the SLS or BSC sample type (Table ). Some covariation among factors was detected, and pure effects (i.e. effects because of a single factor taking all other factors into account) were generally not significant. All other factors accounted for < 1% of the variations in ARISA patterns.

View this table:

Effects of contextual parameters on variation in bacterial community structure

Explanatory factorsNR2 (%)F-ratioP value
Chemistry
Total effects521.61.210.017
Pure effects0.81.03ns
Salinity
Total effects110.73.10.001
Pure effects0.021.04ns
Sample type
Total effects28.82.520.001
Pure effects0.11.01ns
Vegetation
Total effects20.81.12ns
Pure effects00ns
Precipitation (rain)
Total effects14.61.260.083-
Pure effects00ns
Soil type
Total effects213.40.89ns
Pure effects00ns
Sample type controlling for space252.480.001
Space controlling for sample type211.220.017
  • Total and the pure effects (i.e. when controlling for all other factors of the analysis) of explanatory factors were calculated by using canonical RDA models on the subset of data associated with complete contextual parameters (n = 28 samples). The proportion of explained community variation is expressed as R2 values and was adjusted for the given number of parameters considered (N) for calculating pure effects. Significances of the respective F-ratios were tested by performing 1000 Monte Carlo permutation tests and are indicated by – marginally significant (P < 0.1),

  • significant (P < 0.05),

  • highly significant (P < 0.001), and ns when not significant (P > 0.05).

Pyrosequencing and sequence analysis

A total of 155 779 sequences of 16S rRNA gene were generated after discarding low quality reads and primer sequences (out of 209 507 sequences before trimming) from BSC (25 965) and SLS (129 814), respectively. The number of sequences ranged from 522 to 8607 sequences in BSC samples C14 and C48, and 3374 to 44 701 in the SLS sites S11 and S15, respectively (Table ). The number of detected OTUs in BSC was relatively lower than in SLS, which could be as a result of variation in sequencing depth. Analysis of rarefaction curves revealed that the curves did not level off at the 97% cut-off level, even for the samples with the greatest number of sequences (Fig. ). OTU richness, as determined by Chao1 index based on a 97% sequence similarity threshold to define OTU, ranged from 97 to 2532 in BSC samples and 1004 to 3944 in SLS samples (Table ).

Rarefaction curves of bacterial communities (left panel) and taxonomic affiliation and relative abundance (right panel) of the most common bacterial groups encountered by pyrosequencing of 20 samples of SLS and BSC. Per, Pernatty; Is-La, Island Lagoon; Gaird, Gairdner; Gair-S, Gairdner South; Menin, Menindee.

View this table:

Pyrosequencing and bacterial diversity estimators for 20 samples from SLS and BSC

Sample IDNumber of sequencesAt 97% cut-offNo. of ARISA OTUs
Before cleaningAfter cleaningNo. of OTUsChao 1 (lower/upper)ACE
SLS samples
S691116607859*1667 (1487/1899)*1742*141
S114484337411061929 (1761/2140)1951160
S1311 0468349994*1987 (1780/2244)*2077*129
S1557 98444 7011363*3219 (2973/3740)*3638*116
S1849 59136 556879*1777 (1580/2029)*1952*53
S20907469501719*3944 (3620/4328)*4116*102
S2275515835640*1004 (908/1136)*1047*140
S304567365910462051 (1848/2305)2114116
S43802760731078*2172 (1956/2440)*2375*124
S4910 23377101422*3085 (2801/3427)*3391*89
BSC samples
C517661208172208 (189/249)20836
C109525486497 (75/159)9044
C1211496875842250 (1813/2843)238182
C14760522140168 (152/200)170136
C1775895367325*429 (388/496)*428*44
C1936302186226283 (254/339)27428
C2133791923255363 (315/450)33220
C3143483293387*705 (604/855)*752*102
C4230071624195261 (228/327)24599
C4811 25986071018*2532 (2231/2908)*2817*90
  • The diversity indices marked with an asterisk represent the average values calculated after performing three randomized sampling of the sequences to enable comparison among different samples. The number of sequences after cleaning indicates the number after removing short (< 300 bp) and low quality sequences.

Sequences were affiliated to 21 phyla (Fig. ), among which Cyanobacteria, Proteobacteria, and Gram-positive bacteria (Actinobacteria and Firmicutes) encompassed the majority of sequences (> 77%) in most samples (only in samples S19, S31, and S42 they encompassed < 58% of the total sequences). Cyanobacteria constituted the major fraction of BSC microbial communities with the highest abundance detected in Pan (sample C15; 89%). Surprisingly, cyanobacterial sequences constituted only 0.1% of the total sequences in the Menindee BSC (sample C49), which could be the result of the heavy disturbance by cattle. Most of detected cyanobacterial sequences in BSC belonged to the nitrogen-fixing Nostocales group (> 22%), while the rest were affiliated with sequences of uncultured cyanobacteria. In SLS, cyanobacterial sequences constituted between 2% and 42% of the total number, being the highest in the Menindee area (i.e. 42%) and lowest in Lake Torrens (i.e. 2%), and the sequences were mainly affiliated to uncultured cyanobacteria. Among the proteobacterial classes, Alpha-, Beta- and Gammaproteobacteria were most frequently encountered with few sequences from Delta- and Epsilonproteobacteria (≤ 3%). Alphaproteobacteria constituted between 5–56% and 4–35% of the total sequences in BSC and SLS, respectively. The highest occurrence of Alphaproteobacteria in BSC was detected in Island Lagoon and the Menindee area (i.e. 29% and 56%, respectively) and most of these sequences were phylogenetically affiliated to Microvirga flocculans, Skermanella parooensis, Microvirga subterranea, and Rubellimicrobium thermophilum. In SLS, the sequences were related to the genera Rhodovibrio, Azospirillum, Defluviicoccus, Roseomonas, Microvirga, Magnetospirillum, and Rubellimicrobium. The Beta- and Gammaproteobacteria were more dominant in SLS samples than in BSC (3–47% vs. 0.2–3%) and contained sequences related to Delftia acidovorans and Pseudomonas spp., respectively. Actinobacteria-related sequences were detected in all studied SLS and BSC with an abundance of < 20% of total sequences in some samples (e.g. samples 5, 11, 12, 14, and 49). Sequences from this high GC group included the radiation-resistant species Rubrobacter radiotolerans, Thermoleophilum album, and Patulibacter minatonensis. The spore-forming Firmicutes group was much more abundant in SLS than in BSC, where they only formed ≤ 2% of the total sequences. This group accounted for 70% of the total sequences detected in Gairdner (sample 19) and contained sequences related to different species of Bacillus, Clostridium, and Staphylococcus. The remaining less dominant groups were distributed among Bacteriodetes, Deinococcus, Sphingobacteria, Planctomycetes, Chloroflexi, Gemmatimonadetes, Thermomicrobia, Chlorobia, Spartobacteria, Caldilineae, Solibacteres, and Ktedonobacteria.

Discussion

The combination of ARISA and pyrosequencing provided detailed information on the microbial communities occurring at the sites identified as potential sources of eolian dust in Australia. SLS and BSC represent tremendous sources of unconsolidated sediments available for deflation and eolian transport (Blank et al., 1999; Reynolds et al., 2006; Neff et al., 2008). The investigated samples from SLS and the surrounding BSC harbored a wide variety of microorganisms, with a large proportion of novel types that had no close match in publicly available 16S rRNA gene databases. The microbiological data obtained in this study may eventually help back tracking of regional dust sources.

The differences in microbial communities between SLS and BSC, shown by ARISA fingerprinting, were further supported by pyrosequencing. Pyrosequencing provided greater resolution and a more complete picture of bacterial diversity than ARISA, as indicated by the higher number of OTUs and the rarefaction curves generated from pyrosequencing data. The differences of bacterial communities of SLS and BSC could not be explained by variation in nutrients, soil type, vegetation, precipitation, or geology of the region but were mainly attributed to significant differences in salinity and sample type. Salinity has been shown to control the diversity and functions of microbial communities (Casamayor et al., 2002; Abed et al., 2007; Lozupone & Knight, 2007). The higher water content, steep chemical gradients of SLS, and limited exchange between these adjacent biota can also account for these differences. Our nutrient analysis showed that SLS and BSC varied in their Fe, P, and N composition, with some soils limited in P while others were limited in N. However, it is remarkable that these ambient differences in these nutrients were not associated with observable differences in community structure. Bacterial diversity has been shown to follow a chaotic pattern when species compete for three or more resources (e.g. nutrients), which means that diversity of bacterial communities is not explained only by external factors but also potentially by other interactions, such as competition among species (Huisman & Weissing, 1999; Winter et al., 2010). Even in a stable ecosystem with controlled environmental parameters, bacterial diversity has been shown to change dramatically (Fernandez et al., 1999).

The microbial communities of dust sources studied here were compared with those of a previously determined Australian dust sample (De Deckker et al., 2008) to find common bacterial groups. Although the samples were analyzed using two different techniques (i.e. pyrosequencing and Sanger sequencing), comparison was still possible because both techniques were previously shown to yield similar results in spite of their methodological biases (Tedersoo et al., 2010). Both techniques are similarly based on PCR amplification but differ only in terms of sequencing depth. The comparison of dust and dust sources bacterial communities suggests that deflation of bacteria with dust is a selective process. Although cyanobacteria constituted a major fraction of the dust sources (16% and 56% of the total sequences of SLS and BSC, respectively), their contribution to the dust sample was only 2% (De Deckker et al., 2008). Cyanobacterial fraction was also shown to be minor (only 2.7%) in the bacterial communities of African dust (Polymenakou et al., 2008). This low representation of cyanobacteria in dust could be attributed to their filamentous structure and to the production of extrapolymeric substances that allow them to stick firmly to substrate, thus preventing their dispersal during deflation. It could also be owing to the fact that disturbed BSC, which normally contain low abundance of cyanobacterial filaments (Belnap & Eldridge, 2001), are more likely to deflate by wind and thus may be primary contributors to aerosols. The low detection of cyanobacteria in the disturbed BSC of Menindee, a known deflating area in Australia (McTainsh et al., 2005), supports this assumption. Cyanobacteria constituted the major component (20–76%) of the prokaryotic community in the studied BSC, most of which belonged to the known heterocystous nitrogen-fixing Nostocales. This ecologically important group of cyanobacteria was found to be dominant in many BSC in arid and semi-arid environments worldwide, as their nitrogen fixation is essential for the surrounding vegetation (Dodds et al., 1995; Potts, 2000; Yeager et al., 2007; Abed et al., 2010). Although Nostocales includes species that are desiccation-tolerant and highly adapted to life under strong UV radiation through the production of the UV sunscreen pigment scytonemin (Dillon & Castenholz, 2003; Fleming & Castenholz, 2007), which are advantageous features for the survival of these microorganisms at higher altitudes, their detection in dust samples was not common.

Other bacterial groups, on the other hand, identified in the studied SLS and BSC were similar to those found in the Canberra dust sample from 2002 (De Deckker et al., 2008), implying that microorganisms from these ecosystems can be readily transported with dust. For example, Proteobacteria, Actinobacteria, and Firmicutes were among the most common groups in dusts and their sources. Gram-positive bacteria (i.e. Actinobacteria and Firmicutes) were detected in all the source sediments examined here. This widespread distribution and the ability of these bacteria to form spores and tolerate desiccation and high UV levels explain their prevalence in dust samples (Griffin et al., 2001; De Deckker et al., 2008). Gram-positive bacteria accounted for 23% of the total microbial communities in the Australian dust (De Deckker et al., 2008). This group was also found dominant in African dusts (Polymenakou et al., 2008). Actinobacteria inhabit extreme environments like hypersaline lakes, thermal springs, and arid soils and also include a large number of spore-forming plant and animal pathogens (Stackebrandt & Schumann, 2006). Our Actinobacteria sequences were related to thermal and UV-resistant species like Rubrobacter radiotolerans, Thermoleophilum album, but not to known pathogens. The presence of Firmicutes mainly in SLS but not in BSC indicates that saline lakes are the major source of this group in dust samples. This group of microorganisms constituted 23% of the total 16S rRNA sequences obtained from the Australian dust (De Deckker et al., 2008) and related strains were readily cultured from the same sample (Chris Munday, person. commun.). While our Firmicutes sequences were related to 15 different Bacillus species and 2 Clostridium species, cultured species of this group were mainly affiliated to the genera Arthrobacter, Bacillus, Planomicrobium, Paenibacillus, Curtobacterium, Macrococcus, and Staphylococcus (Chris Munday, pers. commun.). The ability to cultivate spore-forming Firmicutes-related species from aerosols shows their resilience against the extreme atmospheric conditions of water scarcity and high levels of UV radiation, with dust probably the important distribution vector for these microorganisms. Many spores are few micrometers in diameter and could survive within cracks of soil particles and clay platelets, which would protect them from UV radiation, or they may also attach to large particles or clumps (Griffin et al., 2001; Prospero et al., 2005). Indeed, it was shown that 50% of the UV radiation is attenuated within large dust plumes through back scatter (Herman et al., 1999). In our study, no pathogenic microorganisms were found, although Actinobacteria and Firmicutes are known to include pathogens (Stackebrandt & Schumann, 2006; Pakarinen et al., 2008). Interestingly, a correlation between the predominance of Gram-positive bacteria in Russian dusts and the occurrence of atopy (i.e. allergic reactions) was found (Pakarinen et al., 2008). Conversely, fungi, which were not studied here, are a major constituent of BSC and can be a major cause of asthma and fever. Consequently, it is possible that aerosols can transfer microorganisums that may be of concern to human health (Griffin, 2007).

The highest concentrations of Fe, P, and N were detected in soils from Menindee, an area previously identified as a major dust source in Australia (McTainsh et al., 2005; De Deckker et al., 2008). It is likely that ecosystems downwind from these sites receive nutrients by dust emanating from SLS and BSC of the region as proposed previously (McTainsh & Strong, 2007). The direct transfer of N and P from source soils may affect soil formation and fertility (Cattle et al., 2002; Hesse & McTainsh, 2003; McTainsh & Strong, 2007). For instance, dust was shown to deposit at an average of 31.4 t km−2 year−1 onto the upper Namoi River Valley in New South Wales, which is known to be a very productive agricultural region (Cattle et al., 2002). P is commonly a limiting nutrient in desert soils and can govern plant growth and productivity, as well as affect carbon and nitrogen mineralization rates in deserts (Reynolds et al., 2006). In theory, the nitrogen-fixing cyanobacteria in BSC, once transported with dust, could lead to increased fertility of the sink soil. However, we found low numbers of these organisms in dust, and it is debatable whether these organisms, adapted to a specific environment, will proliferate in the sink. Dust may also act as an agent for the spread of soluble salts, because the studied saline lakes with salinities up to 22% have been shown to be highly erodible dust source areas in Australia (McTainsh et al., 1999). Indeed, soluble salt content of Australian dusts was found to be between 2.7% and 9.3% (McTainsh et al., 1999). Iron was also found in high concentrations in some of our dust source samples. Large-scale iron fertilization experiments significantly boosted primary productivity in oceans (Griffin & Kellogg, 2004 and reference therein). However, rarely a clear relation between the occurrence of phytoplankton blooms and dust storms has been reported and most dust deposition events did not result in blooms (McTainsh & Strong, 2007).

In conclusion, our study shows that dust sources are highly diverse in UV-resistant and spore-forming bacterial communities and also contains nutrients that can be transported with the aerosols, thus rendering dust not only as a medium for long-range transport of viable microorganisms, but also a periodic source of nutrients for nutrient-depleted terrestrial and marine environments. Our work, however, does not preclude the possibility that dust microorganisms could have been derived from sources other than those presented in this study, as microorganisms could be recruited from other ecosystems into dust plumes during their journey prior to being deposited. Nonetheless, our study provides a comprehensive molecular baseline that would help in back tracking microorganisms in dust samples.

Acknowledgements

The authors would like to thank Jan-Berend Stuut, Gwen Allison, Chris Munday, Nigel Tapper, and Marlene Bausch for their help during sampling and discussions. Martina Alisch and Anna Lichtschlag from the Max-Planck Institute are thanked for their help in the nutrient analysis. This research was funded by the Australian Research Council (ARC grant DP0772180) and the Max-Planck Society.

Footnotes

  • Editor: Christoph Tebbe

References

View Abstract