2014 Archer, McDonald et al. FEMS Microbiology Ecology

14 Pages·2014·0.74 MB·English

RESEARCH ARTICLE
Characterisation of bacterioplankton communities in the
meltwater ponds of Bratina Island, Victoria Land, Antarctica
Stephen D.J. Archer, Ian R. McDonald, Craig W. Herbold & Stephen C. Cary
InternationalCentreforTerrestrialAntarcticResearch,SchoolofScience,UniversityofWaikato,Hamilton,NewZealand
Correspondence:StephenC.Cary,School Abstract
ofScience,UniversityofWaikato,PrivateBag
A unique collection of Antarctic aquatic environments (meltwater ponds) lies
3105,Hamilton3240,NewZealand.
Tel.:[phone]; in close proximity on the rock and sediment-covered undulating surface of the
fax:[phone]; McMurdo Ice Shelf, near Bratina Island (Victoria Land, Antarctica). During
e-mail:[email protected] the 2009–10 mid-austral summer, sets of discrete water samples were collected
across the vertical geochemical gradients of five meltwater ponds (Egg, P70E,
Received12November2013;revised12May
Legin, Salt and Orange) for geochemical and microbial community structure
2014;accepted17May2014.
analysis. Bacterial DNA fingerprints (using Automated Ribosomal Intergenic
DOI:10.1111/1574-6941.12358 Spacer Analysis) statistically clustered communities within ponds based on ANO-
SIM (R = 0.766, P = 0.001); however, one highly stratified pond (Egg) had two
Editor:JohannaLaybourn-Parry distinct depth-related bacterial communities (R = 0.975, P = 0.008). 454 py-
rosequencing at three depths within Egg also identified phylum level shifts and
Keywords increased diversity with depth, Bacteroidetes being the dominant phyla in the
Antarctic;planktonic;microorganism; surface sample and Proteobacteria being dominant in the bottom two depths.
community;meltwater. BEST analysis, which attempts to link community structure and the geochemis-
try of a pond, identified conductivity and pH individually, and to a lesser
Y extent Ag109, NO and V51 as dominant influences to the microbial community
2
G structure in these ponds. Increasing abundances of major halo-tolerant OTUs
O across the strong conductivity gradient reinforce it as the primary driver of
community structure in this study.
L
O
C ponds can be more susceptible to seasonal (and global)
Introduction
E climatic changes, while offering easy access to a broad
The Antarctic continent offers access to one of the most range of unique pond-specific geochemical conditions.
Y
physically and chemically demanding environments on During the peak of the Antarctic summer up to 20%
G
earth. Antarctic limnetic ecosystems (lakes and ponds), in of the McMurdo Ice Shelf (MIS) in southern Victoria
O particular, experience large fluctuations in temperature Land is covered with meltwater ponds of various sizes,
and light regimes and are characterised by steep chemical making them a dominant environment in this area (De
L
O gradients, which greatly impact physiological adaptations Mora et al., 1994). The ponds surrounding Bratina Island
and life cycle strategies (Wait et al., 2006). These ecosys- on the MIS (Fig. 1) provide a cross-section of the many
I
B tems, scattered throughout the Antarctic continent and geochemistries found in aquatic systems throughout Ant-
O on the surrounding sea ice, provide an extraordinary and arctica (Cowan & Tow, 2004). Their existence is depen-
tractable opportunity to examine microbial distribution dent on the short period, at the peak of summer, when
R
in distinct and environmentally extreme microenviron- liquid water from local ice melt collects in the depressions
C
ments. Of the thousands of lakes and ponds throughout of the rugged landscape of the MIS replenishing pre-exist-
I
M Antarctica, most research has focused primarily on lakes ing ponds and forming new ones (Howard-Williams &
due to their size, stability and influence on the surround- Hawes, 2005). Marine-derived surface sediments and
ing terrestrial system (Craig et al., 1992; Bell & Laybourn- aerosols, as well as seawater trapped in tide cracks across
Parry, 1999; Bowman et al., 2000; Pearce et al., 2003; the ice shelf, provide multiple variable inputs of dissolved
Taton et al., 2003). However, due to their smaller size, salts into ponds. The fine balance between water and salt
FEMSMicrobiolEcol&&(2014)1–14 ª2014FederationofEuropeanMicrobiologicalSocieties.
PublishedbyJohnWiley&SonsLtd.Allrightsreserved
2 S.D.J.Archeretal.
0 50 100 150 200 m
Broline George
Hour
Glass
Tidal
Extra Bratina
Smith Rotifer Lagoon
Upper
Kraak Tidal
Fresh
Salt Pond NZField
Brack Retro Camp
Conophyton
Lunch Long
Ice Helo
Ridge Pad
K081 Pond
Legin Duet 1stTidal
SubSkua Nicholas Crack
Bambi Ribbon Permanent JASCkua
P70E P70 IcePond Casten
20 km
Galore
Mostoc 2ndTidal
McMurdo Sound
Foam Crack
Ross Island Huey Orange
Wendy Duey
Lucy
Scott Base Egg
Bratina Island
Fig.1. MapoftheMISPonds,basedonanoriginalmapproducedbyNewZealandDepartmentofLandandSurveys,1991.Pondscircledwere
usedinthisstudy.
input and evaporation leads to both inter and intrasea- reducing carbonic acid (Hendriks et al., 2014), both
sonal variation of physicochemistry within a pond (Lay- parameters are constrained by the conductivity density
bourn-Parry et al., 2002) and variable chemistries gradients, illustrating the interplay between biotic and
between ponds. Ponds in close proximity (metres apart) abiotic variables along physicochemical gradients that
can vary greatly in size (one to several hundred square occur with depth in these ponds (Wait et al., 2006).
metres), depth (< 1 m to around 4 m) and age (1 to at PaststudiesofAntarcticponds,andlakes,haveprovided
least 30 years old) (Gibson et al., 2006), as well as in some insight into how microbial communities are struc-
metal concentrations, salinity (Schmidt et al., 1991; Mat- turedintheseextremeaquaticenvironments.Thesestudies
sumoto et al., 1992), pH and dissolved oxygen (Koob & primarily focused on the dense microbial mats and sedi-
Leister, 1972). The water column within some ponds can ments lining the ponds (Fernandez-Valiente et al., 2001;
be highly stratified, with a unique brine geochemistry Griffin & Tiedje, 2007) and on the geochemistry of the
derived from the freeze concentration of dissolved salts water column (Matsumoto et al., 1992; De Mora et al.,
layering the bottom and newer meltwater input layering 1994; Wait et al., 2006), with little or no work on the
the surface. Oxygen, derived from Cyanobacteria domi- microbiology of the water column itself. Previous research
nated mats lining the ponds, also exhibits a steep gradient hasalsobeenlimitedbyarelianceonmicroscopicidentifi-
increasing with depth, frequently resulting in oxygen cation of microorganisms, (Vincent & James, 1996; de los
supersaturation in the lower depths of certain highly Rioset al.,2004;Sabbeet al.,2004),whichsuggestedthese
stratified ponds (James et al., 1995; Wait et al., 2006). pondecosystemssupportedlowbiodiversity,probablydue
Photosynthesis, which causes the oxygen supersaturation, tothepresenceofcrypticspecies.However,morerecently,
increases pH via the removal of CO from the water theuseofculture-independent moleculargenetictoolshas
2
ª2014FederationofEuropeanMicrobiologicalSocieties. FEMSMicrobiolEcol&&(2014)1–14
PublishedbyJohnWiley&SonsLtd.Allrightsreserved
BacterioplanktoninBratinaIslandmeltwaterponds 3
allowed microbiologists to identify and investigate a more data were collected using a Hobo temperature data logger
extensive biodiversity in these systems (Vincent & James, (H09-003-08, Onset Computer Corporation). Finally, a
1996; Brambilla et al., 2001; Sjoling & Cowan, 2003; 30 mL unfiltered sample was collected in a 50-mL falcon
Hugheset al.,2006;Caryet al.,2010). tube for immediate pH, and conductivity measurements
This project is the first in-depth study to characterise performedinthefieldusingaSPERScientificwaterquality
the microbial communities in the water column of Brati- meter(SperScientific,Arizona).
na Island ponds and to describe their relationship to
pond geochemistry. Utilising molecular genetic tools (a
Nutrient and elemental analysis
combination of DNA fingerprinting and high throughput
next-gen sequencing) coupled with geochemistry, five NH , NO and PO measurements on the filtrate from
4 2 4
ponds were characterised, resulting in the most stratified each sample were carried out at the University of Waika-
pond being selected for higher resolution analysis. These to using an Aquakem 200CD following the manufac-
data allowed for the identification of possible geochemical turer’s instructions (Thermo Fisher Scientific, Waltham).
drivers, and the specific bacterial groups they control, Elemental analysis was performed on each sample by
within a highly stratified water column. inductively coupled plasma mass spectrometry (ICP-MS)
(cid:2)
using a Mass Spectrometer ELAN DRC II (PerkinElmer
Inc., Mu€nster, Germany). To prepare samples for ICP-
Materials and methods
MS, 0.22 lm prefiltered pond water was diluted 1 : 50
with Milli-Q water (Millipore, Billerica, MA). Six samples
Field sampling strategy
were diluted 1 : 1000 due to excessive salt concentrations.
Water column samples were collected from five separate Once diluted, samples were acidified with 2% HN0
3
ponds near Bratina Island 78°010S latitude, 165°320E lon- (Extra pure Nitric Acid, Ajax Finechem, NSW, Australia).
gitude on the MIS (Fig. 1) shortly after most ponds Elements analysed were the following: Li7, B10, Na23,
became ice-free at the beginning of the austral summer in Mg24, Al27, S34, K39, Ca43, V51, Cr52, Fe54, Mn55, Co59,
December 2009. Sampled pond depths ranged from 0.3 to Ni60, Cu65, Zn68, As75, Se82, Sr88, Ag109, Cd111, In115,
0.5 m (Supporting Information, Table S1) and sizes from Ba137, Tl205, Pb207, Bi209 and U238 (results of all measur-
3 to 20 m across. Four ponds were ice-free, but a 1-cm able variables are in Table S1).
surface layer and the middle 25% of Legin pond was still
frozen at this time. A customised micromanipulator/sam-
DNA extraction of samples
pling apparatus was constructed (Fig. S1) and used to
simultaneously collect physicochemical data (in situ data DNA was extracted from the filtered biomass using a
used to determine sampling frequency) and water samples modified CTAB extraction protocol (Campbell et al.,
from the approximate centre of each pond at 2–10 cm 2001). Briefly, the frozen, sealed 0.22-lm sample filters
increments ((cid:2) 1 mm) through a sterilised sampling tube were first thawed on ice and then connected to a syringe
(cid:2)
(TYGON R-3603)thatwasinsulatedandcapableofmild containing 1 mL of fresh CTAB. The exposed filter nipple
heating with a nicrome wire heating tape controlled via a was parafilmed, and the entire assembly incubated in a
rheostat. The tube was flushed with two void volumes (at Ratek Orbital mixer at 150 r.p.m. and 65 °C for 30 min.
each depth) before each sample was collected. 8–100 mL The filter assembly was allowed to cool then 0.5 mL of
of sample was immediately filtered through syringe- the CTAB in the syringe was pushed through the filter
mounted 0.22-lm filters (Whatman International Ltd, assembly to evacuate the lysate into an eppendorf tube.
Kent, UK) until the filter clogged. A small amount of air An equal volume of chloroform/isoamyl alcohol (24 : 1)
waspushedthroughthefiltertoremoveexcesswater,then was then added to the lysate and mixed on the orbital
gently flooded with a nucleic acid preservative/lysis buffer mixer at 150 r.p.m. and 65 °C for a further 30 min. The
(CTAB – cetyltrimethylammonium bromide-polyvinyl- eppendorf tube was then centrifuged for 15 min at
pyrrolidone-b-mecaptoethanol) (Coyne et al., 2001) and 16 900 g, and the aqueous phase transferred to a new
frozen fortransportbackto thelaboratory. Thirteenmilli- tube. Nucleic acids were precipitated by addition of 1 vol-
litreofthefiltratewasalsocollectedin15-mLfalcontubes ume of isopropanol and 0.5 volume of 5 M NaCl and
and frozen immediately for later geochemical analysis. then incubated at (cid:3)80 °C for at least 1 h. The tube was
Oxygen concentration and temperature were measured in then centrifuged at 26 000 g for 30 min, and the super-
situ using probes attached near the tip of the sampling natant discarded, and the DNA pellet washed with
tube.OxygendatawerecollectedusingaFibox3LCDtrace 0.5 mL of 70% EtOH and then centrifuged at
minisensor oxygen metre with data-logger (PreSens 15 500 r.p.m. for 5 s. The pellet was dried and resus-
Precisionsensing,Regensburg,Germany),andtemperature pended in 10–50 lL of sterile milliQ H O. The extracted
2
FEMSMicrobiolEcol&&(2014)1–14 ª2014FederationofEuropeanMicrobiologicalSocieties.
PublishedbyJohnWiley&SonsLtd.Allrightsreserved
4 S.D.J.Archeretal.
DNA was quantified using a Nanodrop ND-1000 at dendrogram. Nonmetric multidimensional scaling (MDS)
260 nm (NanoDrop Technologies, Montchanin, DE). was performed on the resemblance matrix, which displays
relative similarities between communities as distance (i.e.
the closer two samples are the more similar community).
Automated ribosomal intergenicspacer
2-D MDS plots with a stress value of < 0.2 were used as
analysis (ARISA)
they were considered to have accurate information. Infor-
ARISA DNA fingerprinting (Fisher & Triplett, 1999) was mation from the dendrogram of the communities was
utilised to resolve bacterial community structure and rela- overlaid onto the 2-D MDS plots, which provide percent-
tive diversity within all ponds, and a comparative subset age similarity levels at 20, 40, 60 and 80%, assisting in
(Egg, Legin and P70E) was selected to resolve cyanobacte- the evaluation of community structure between samples.
rial community structure in Egg pond. From each sample, ANOSIM analyses and Mantel tests were performed on the
the bacterial intergenic spacer region (ISR) in the rRNA resemblance matrix to test specific hypotheses formed
operon was amplified using PCR. Triplicate 30 lL reac- from interpretation of MDS plots.
tions [to reduce stochastic PCR bias (Wintzingerode The influence of pond geochemistry on the microbial
et al., 1997)] were run for each sample. Bacterial primers community was determined using a BEST analysis to find
used were ITSReub-Hex (50-GCCAAGGCATCCACC-30) the ‘best match’ in PRIMER 6. Geochemical data were nor-
and ITSF [50-GTCGTAACAAGGTAGCCGTA-30) (Cardi- malised and a Draftsman plot created to view interrela-
nale et al., 2004)] and the cyanobacterial primers were tions between the different variables, so the data could be
CY-ARISA-F (50-GYCAYRCCCGAAGTCRTTAC-30) and simplified. Spearman’s rank correlation coefficients were
23S30R (50-CHTCGCCTCTGTGTGCCWAGGT-30) then calculated, resulting in a probability (P) of the com-
(Wood et al., 2008)]. Full PCR components and condi- munity differences being explained by the differences in
tions and quality control procedures are described in geochemistry. The combination of geochemical variables
Supporting Information. whose Euclidean distance matrix gave the highest P-value
To visualise, compare and interpret the ARISA finger- were considered the most likely drivers of community
prints from different samples, the output was processed dynamics.
with an informatics pipeline (modified from (Abdo et al.,
2006) First, all peaks with heights exceeding 250 fluores-
454 pyrosequencing
cence units were accepted as true peaks. The remaining
peaks were used to calculate model parameters for a log- PCR amplicons containing V5–V6 hypervariable regions
normal distribution. Iteratively, peaks with an area of the 16S rRNA gene were utilised to identify specific
exceeding the 99.9% cumulative distribution of the calcu- bacterial community differences in the water column of
lated log-normal distribution for noise were accepted as Egg pond using 454 pyrosequencing. Triplicate 30 lL
true peaks. Fragment lengths < 100 bp for bacteria and reactions were run for each sample, each reaction con-
150 bp for cyanobacteria were removed from analysis as taining 0.4 lM of each unadapted primer Tx9F
these were considered to be too small to be ITS fragments (50-GGATTAGAWACCCBGGTAGTC-30) and 1391R
(Cardinale et al., 2004; Wood et al., 2008). Peaks were (50-GACGGGCRGTGWGTRCA-30) (Ashby et al., 2007),
binned into ARISA Fragment Lengths (AFLs) with 1x PCR buffer, 0.2 mM dNTPs, 0.02 mg mL(cid:3)1 BSA,
width = 2 nt, and total peak area of each bin was used to 0.02 U Platinum Taq, 2 mM MgCl , 20 ng bp genomic
2
calculate relative adundance of each AFL in a given data DNA, and the reaction was made up to 30 lL with mil-
set. The resulting data matrix was imported into PRIMER 6 liQ H2O. Thermal cycling conditions were the following:
for statistical analysis (Clarke & Gorley, 2006). 94 °C for 2 min, then 30 cycles of 94 °C for 20 s, 55 °C
In PRIMER 6, peaks were converted to presence/absence for 10 s ((cid:3)0.2 °C per cycle), 72 °C for 20 s and a final
data for further analysis, and the number of peaks was extension of 72 °C for 3 min. Once amplified, all tripli-
totalled from each sample to be used as a proxy for rela- cate PCRs were combined, run on a 2% TAE agarose gel
tive alpha diversity (the number of species in a sample). stained with ‘SYBR Safe’ (Invitrogen Ltd) at 80 V, the
These were subjected to various statistical tests (Spear- bands excised and DNA retrieved using the MO BIO
man’s coefficient and Tukey’s honest significant difference UltraClean 15 DNA Purification Kit (MO BIO Laborato-
test) to elucidate relationships between communities. Beta ries, Carlsbad, CA) as per manufacturer’s instructions,
diversity (the comparison of diversity between sites) was but running the final spin step twice. A second round of
investigated using a resemblance matrix created based on triplicate PCR was run as above but with only 10 cycles
the Bray Curtis similarity index (Bray & Curtis, 1957). and using 25 ng of the purified DNA from the previous
From the resemblance matrix, a hierarchical clustering step per reaction (milli-Q H O volume adjusted accord-
2
analysis was performed, producing a relational ingly). The primers used were adapted for one-way reads
ª2014FederationofEuropeanMicrobiologicalSocieties. FEMSMicrobiolEcol&&(2014)1–14
PublishedbyJohnWiley&SonsLtd.Allrightsreserved
BacterioplanktoninBratinaIslandmeltwaterponds 5
according to the Roche GS Junior System Guidelines for sequences of all identified OTU were analysed using
0.03
Amplicon Experimental Design Manual (August 2010), the Classifier function provided by the Ribosomal Data-
including unique MID identifiers for each sample base Project (RDP) Release 10, Update 15 (Wang et al.,
[BacX-Tx9F (50-CCATCTCATCCCTGCGTGTCTCCGA 2007). Taxonomic assignment threshold was set at 80%.
CTCAG-MID-GGATTAGAWACCCBGGTAGTC-30) and Phylogenetic analysis of sequences was performed with
BacB-1391R (50-CCTATCCCCTGTGTGCCTTGGCAGTC representative sequences of the top 101 OTUs (> 0.1%
0.03
TCAG-GACGGGCR GTGWGTRCA-30)]. A second gel overall abundance in pyrosequencing data sets, represent-
extraction was performed as above. Samples went through ing 12 376 reads). Taxonomic trees were constructed
a final cleanup step using the Agencourt AMPure XP sys- using ARB (Ludwig et al., 2004), with DNADIST and neigh-
tem (Beckman Coulter Genomics, Danvers, MA) as per bour joining analysis. Sequence length ranged from 244
the manufacturer’s instructions. A verification gel was to 392 bp, with an average of 384.9 bp.
run on a 2% TAE gel to confirm a lack of unwanted
bands in the sample. SampleDNAcontentwasquantified
Results
using a Qubit Flurometer (Invitrogen Ltd) and was then
diluted to 1 9 109 molecules lL(cid:3)1 as per the Roche
Physicochemical analysis
AmpliconLibraryPreparationMethodManual[GSJunior
TitaniumSeries,May2010(Rev.June2010)].QPCRusing Inter and intrapond physicochemical properties were
a KAPA Library Quantification Kit for Roche 454 Tita- found to be extremely variable, with dissolved oxygen
nium/Universal(KapaBiosystems,Woburn,MA)wasused concentrations from 4.4 to 27.3 ppm (Fig. 3a), conduc-
tocheckthe1 9 109dilutionandwasadjustedaccordingly tivity ranging from 0.43 to 110.9 mS cm(cid:3)1 (Fig. 3b), pH
for making the amplicon library. The diluted amplicons from 7.3 to 10.1 (Fig. 3c), and temperature from (cid:3)3.5 to
were mixed together in the desired proportions to create 5.1 °C (Fig. 3d). A notable geochemical gradient was
the 1 9 109 amplicon pool. Sequencing was performed identified throughout the water column of most ponds.
using the GS Junior Titanium emPCR Kit (Lib-L), the GS Oxygen concentration varied through the water column,
JuniorTitaniumSequencingKit,PicoTiterPlateKitandGS but generally it was highest at the deepest point (increas-
JuniorSystemaccordingtothemanufacturer’sinstructions ing from 9.7 at the top to 17.0 ppm in the bottom of
(Roche454LifeSciences,Branford,CT). Legin). Conductivity was also found to increase with
depth in each pond (from 8.2 mS cm(cid:3)1 at the top to
93.6 mS cm(cid:3)1 at the bottom of Egg). Although there are
454 pyrosequencing data processing
significant intrapond variations, the geochemistry MDS
454 PCR amplicon pyrosequencing data were first pro- analysis (Fig. 2) clearly shows that each pond is geochem-
cessed using AmpliconNoise v1.0 (Quince et al., 2011). ically distinct, with Egg forming two distinct clusters.
Briefly, raw flowgrams (sff files) with perfectly matching Egg pond exhibits the most extreme physicochemical
primer and barcode sequences were filtered for a mini- stratification observed in this study (Fig. 3a–d). The pH
mum flowgram length of 360 cycles (including primer was stable between 9.4 and 9.5 in the upper 38 cm, drop-
and barcode sequences) before the first noisy signal (i.e. ping to 8.8 at 40 cm and to 8.1 at the bottom (48 cm).
0.5–0.7 or no signal in all four nucleotides). All flow- The drop in pH corresponds with an increase in conduc-
grams were then truncated at 360 bases and clustered to tivity, which rose modestly from 8.2 mS cm(cid:3)1 at 0 cm to
remove sequencing noise using PyroNoise (Quince et al., 10.4 mS cm(cid:3)1 at 38 cm, with a steep increase between 40
2009, 2011). Noise introduced by PCR was removed and 46 cm, up to 93.6 mS cm(cid:3)1 at 48 cm deep. Dis-
using SeqNoise (Quince et al., 2011), and PCR chimeras solved oxygen within Egg increased from 11.5 ppm at the
were removed using Perseus (Quince et al., 2011). The surface, to 17.1 ppm at 36 cm, and 27.3 ppm at 44 cm.
resulting de-replicated sequences were processed using The temperature profile was relatively stable, decreasing
Mothur 1.17.0 (Schloss et al., 2009) to create a unique from 1.6 °C at the surface to (cid:3)1.1 °C at 46 cm, with a
sequence and names file. Pairwise alignments and dis- notable increase at the bottom (48 cm) to 0.6 °C.
tance were calculated using Espirit (Sun et al., 2009). Mo- NH was only detected throughout the water column
4
thur was then used to cluster sequences into operational in Egg and Salt pond (which also have the greatest con-
taxonomic units (OTUs) defined at the furthest neigh- ductivities) and at the top of Orange and at the bottom
bour Jukes–Cantor distance of 0.03 (OTU ). Rankabun- of P70E. No NH was detected in Legin. NO was found
0.03 4 2
dance data were generated for each sample as well as at extremely low concentrations in all ponds, except P70E
Venn diagrams and Good’s nonparametric coverage esti- where it was 10 times higher in the upper region
mator was used to evaluate sequencing coverage for each (0.032 ppm compared with 0.002 ppm in most other
sample. For phylogenetic assignments, representative ponds). ICPMS analsis revealed that most resolved
FEMSMicrobiolEcol&&(2014)1–14 ª2014FederationofEuropeanMicrobiologicalSocieties.
PublishedbyJohnWiley&SonsLtd.Allrightsreserved
6 S.D.J.Archeretal.
Fig.2. MDSplotofgeochemistrycomparing
allpondsstudied,numbersrepresentsample
depth.Thegreaterthedistancebetweentwo
points,thegreaterthedifferencebetween
theirgeochemistries.Circlessurrounding
samplesrepresentrelativeconductivity
concentration(thelargerthecirclethehigher
theconductivity).MDSplot2-Dstress:0.08.
Outliningcircleusedtoidentifypondclusters.
(a) (b)
(c) (d)
Fig.3. Geochemistrydepthprofileshowingchangesinoxygenconcentration(ppm)(a)conductivity(mScm(cid:3)1)(b),pH(c)andtemperature(°C)(d).
elements increased along conductivity gradients, notable
DNA fingerprinting
observations were that Co59 was only detected in the bot-
tom of Egg (46–48 cm) and throughout Salt pond. Ni60 The number of AFLs was used as a proxy for alpha diver-
was below the detection limit in Legin and Orange and sity (Fig. S2). There are no significant differences between
the upper 38 cm of Egg. Legin also had a significantly alpha diversity of Bacteria from different ponds; however,
reduced amount of Cu65 and Mn55 compared with other P70E has significantly higher Cyanobacteria diversity (29–
ponds, and Orange had a reduced level of Mn55. 39 AFLs per sample compared with a mean of 16.6 over
ª2014FederationofEuropeanMicrobiologicalSocieties. FEMSMicrobiolEcol&&(2014)1–14
PublishedbyJohnWiley&SonsLtd.Allrightsreserved
BacterioplanktoninBratinaIslandmeltwaterponds 7
all ponds) than Legin and Egg according to Tukey’s Hon- similarity: cyanobacterial ANOSIM R = 0.8605, P = 0.001;
est Significant Difference Test (P ≤ 0.001 for Egg and bacterial ANOSIM R = 0.766, P = 0.001 (Fig. 4 a and b).
P ≤ 0.001 for Legin). A Spearman’s rank correlation coef- P70E and Egg ponds were exceptions to this general
ficient showed no significant correlations between alpha observation. In P70E, the bacterial community observed
diversity and depth for Bacteria or Cyanobacteria. Lowest at 0 cm depth differs from the bacterial communities
diversity occurred in Salt pond (12–25 AFLs) containing observed at all other depths (40% similarity). A Mantel
the highest overall conductivity of the sample set; how- test confirmed that bacterial communities at similar
ever, when removed as an outlier, there was a positive depths in Egg pond were more comparable than those
correlation between average conductivity and average bac- further apart (R = 0.8934, P = 0.002). Bacterial commu-
terial ARISA AFLs but with only a marginal significance nities from Egg pond formed two major clusters based on
(P = 0.04). depth; Egg (1), the upper 0–38 cm and Egg (2), the lower
Beta diversity comparisons were visualised using a 2-D depths, 40–48 cm (Fig. 4a), while a similar clustering
MDS plot for bacterial communities from five different structure is apparent in cyanobacterial communities at
ponds (Fig. 4a) and cyanobacterial communities from 60% similarity (Fig. 4b). These observations were
three ponds (Fig. 4b). Microbial communities were iden- reflected in an ANOSIM test for bacterial (R = 0.975,
tified as significantly dissimilar between ponds at 40% P = 0.008) and cyanobacterial communities (R = 0.4438,
P = 0.026). This combined data resulted in the selection
of three samples from Egg pond, which represented the
(a) greatest variation in community structure down the strat-
ified water column. Bacterial communities from Salt pond
stood out as being unique from all other ponds (20%
similarity).
BEST analysis was used to correlate bacterial and cyano-
bacterial community structure to differences in geochemi-
cal profiles (Table 1), with both appearing to be well
correlated to conductivity as a major contributing factor.
Conductivity and Ag109 were the best explanatory vari-
ables for cyanobacterial communities (Spearman’s
q = 0.662), and temperature, conductivity and V51 were
the major explanatory variables for bacterial communities
(Spearman’s q = 0.698). When Egg was analysed on its
own, higher probabilities were achieved for community
(b)
Table1. Geochemistry linked to community structure using BEST
analysis for cyanobacterial and bacterial DNA fingerprints (ARISA). A
Spearman’s rank correlation of 1.0 would indicate a 100% linkage
betweenthevariable/sselectedandthedifferencesincommunities
CyanobacterialARISA BacterialARISA
Correlation Variable Correlation Variable
Allponds 0.662 4,11 0.698 2,4,10
0.639 2,4,11 0.697 2,4
0.635 4,7,11 0.675 4,10
0.623 2,4,7,11 0.654 4
0.591 4,5,11 0.652 2–4,10
Onlyegg 0.734 3,11 0.927 5
0.732 5 0.927 5,7
0.732 5,7 0.926 4
Fig.4. Bacterial and Cyanobacterial ARISA MDS Plots showing
0.730 4,11 0.926 4,7
percentage similarities of pond communities (a) bacterial ARISA, data
0.714 3,5 0.925 4,5
normalised, converted to presence or absence points, and subjected
to a Bray Curtis similarity matrix, MDS plot 2-D stress 0.15. (b) Variable designation: 1=depth, 2=temperature (C+10), 3=dis-
cyanobacterial ARISA, data normalised, converted to presence or solved oxygen (ppm), 4=conductivity (mScm(cid:3)1), 5=pH, 6=total
absence points,andsubjectedtoaBrayCurtis similarity matrix,MDS phosphate (ppm), 7=NO2 (ppm), 8=NH4 (ppm), 9=Al27,
plot2Dstress0.09. 10=V51,11=Ag109.
FEMSMicrobiolEcol&&(2014)1–14 ª2014FederationofEuropeanMicrobiologicalSocieties.
PublishedbyJohnWiley&SonsLtd.Allrightsreserved
8 S.D.J.Archeretal.
linkages to geochemical variables within Egg. Cyanobacte- water column of Egg (Fig. 6). At the surface (0 cm), the
rial communities of Egg appear moderately influenced by community was least diverse, with six phyla well repre-
pH, which had the highest probability as a single driver sented. This sample was dominated by Bacteroidetes
of diversity (Spearman’s q of 0.732). Both pH and con- (68%), with one OTU contributing 64.8% [the sequence
ductivity were determined to be strongly correlated to of this OTU has 100% identity with Algoriphagus yeomje-
bacterial diversity differences within Egg (Spearman’s oni, isolated from a marine solar saltern (Yoon et al.,
q = 0.927 for pH and q = 0.926 for conductivity). 2005)], and b-Proteobacteria (18.9%), with one OTU con-
tributing 18.6% [which has 100% sequence identity to
Hydrogenophaga taeniospiralis isolated in the Arctic, and
454 pyrosequencing
to a number of Antarctic lake isolates (VanTrappen et al.,
454 pyrosequencing of three representative samples from 2002)]. At 40 cm, community diversity increased, with 12
Egg pond created a data set of 12 995 reads that passed phyla including the six seen at the surface. At this depth,
quality scoring, which clustered into 583 OTUs at a 97% the community had also shifted to a c-Proteobacteria
sequence identity (OTU ; Fig. 5). Good’s nonparamet- (66.1%) dominated community with the reads from one
0.03
ric coverage estimator indicates that sampling coverage OTU contributing 60.8% [this OTU has 99% identity to
was 98.6% for 0 cm, 97.4% for 40 cm and 93.5% for Psychromonas sp. isolated from Arctic sea ice (Auman
48 cm. Each sample contained a high number of unique et al., 2006)], the Bacteroidetes were again a major phyla
OTUs, the number increasing with depth; however, the but only 10.9% of the community. In the deepest sample
percentage of reads represented by sample-specific unique (48 cm), diversity increased dramatically, with 18 phyla
OTUs was < 7% in the surface and middle (40 cm) including all those seen in the 40 cm sample. Again the
depths and 14.2% in the bottom (48 cm) depth. A c-Proteobacteria were the dominant phyla (50.8%), with
majority of reads present in the surface (93.1%) were rep- one OTU contributing 46.1% (the same OTU dominant
resented by OTUs present at all depths. A majority of at 40 cm), the only other major group were uncharacter-
reads in the middle and bottom depths (83.6% and ised, although this contained only one OTU (16.8%) with
84.2%, respectively) were represented by OTUs shared a [96–98% identity to environmental sequences from a
between those depths, but absent in the surface. hypersaline microbial mat (Isenbarger et al., 2008)] The
Phylum/class level investigation revealed dramatic shifts Bacteroidetes (5.6%) were the third most represented
in the bacterial communities with depth through the phyla.
Fig.5. Summaryof454sequencingdata(at
adistanceof0.03),andsimilaritiesbetween
depthsofEggpond.ThetotalOTUrichnessof
allthegroupsis583,andtotalreadnumberis
12995.
ª2014FederationofEuropeanMicrobiologicalSocieties. FEMSMicrobiolEcol&&(2014)1–14
PublishedbyJohnWiley&SonsLtd.Allrightsreserved
BacterioplanktoninBratinaIslandmeltwaterponds 9
(a) (b)
Bacteroidetes (68%) Chlamydia (0.3%) Bacteroidetes (10.9%)
Cyanobacteria
(0.7%)
Verrucomicrobia (2.9%)
Unknown (0.2%)
Chloroflexi (0.1%)
Cyanobacteria (0.7%)
Actinobacteria (1.3%)
Actinobacteria (9.4%) Firmicutes (0.1%)
Beta− Alpha−Proteobacteria (1.4%)
Gamma−Proteobacteria (0.1%) Proteobacteria
(1.3%)
Epsilon−Proteobacteria (1.2%)
Alpha−Proteobacteria (0.3%)
Beta−Proteobacteria (18.9%)
Gamma−Proteobacteria (66.1%)
10%
10%
Planctomycetes (0.1%) SR1 (0.2%)
(c)
OD1 (0.1%)
Chlamydia (0.3%)
Spirochetes (0.7%) Cyanobacteria
(0.1%)
Verrucomicrobia (4.3%)
Bacteroidetes (5.6%)
Chloroflexi (0.2%)
Unknown (16.8%)
Actinobacteria (0.3%)
Delta−Proteobacteria (0.5%)
Alpha− Firmicutes (4.1%)
Epsilon−Proteobacteria (1.9%) Proteobacteria
0.3%)
Beta−
Proteobacteria
(0.4%)
Deinococcus/Thermus (0.1%)
Gamma−Proteobacteria
(50.8%)
10%
Fig.6. Phylum-leveldiversityofbacterial 16SrRNAgenesequencesfromdifferentdepthsofEggpond.Percentage abundanceofphyla within
eachsiteareshown(inparentheses)forthe(a)surfacesample,(b)40cmdeepsampleand(c)48cmdeepsample.454pyrosequencingreads
were denoised and clustered at 97% (farthest neighbour) to obtain 583 OTUs in total, of which 120 OTUs were shared. The top 101
0.03 0.03
OTUs with>0.1%abundanceinanysampleareusedinthisanalysis.
0.03
communities (Matsumoto et al., 1992; Wait et al., 2006;
Discussion
Webster-Brown et al., 2012). Early studies of pond sedi-
The focus of this study was to determine whether the ments from Bratina meltwater ponds have identified
unique geochemistry previously reported in the water col- highly diverse (Sjoling & Cowan, 2003) and variable
umns of Bratina ponds supports unique bacterioplankton microbial communities between ponds (Kemp & Aller,
FEMSMicrobiolEcol&&(2014)1–14 ª2014FederationofEuropeanMicrobiologicalSocieties.
PublishedbyJohnWiley&SonsLtd.Allrightsreserved
10 S.D.J.Archeretal.
2004); however; the environmental factors driving Jiang et al., 2010; Laque et al., 2010). Examination of
these differences were not determined, and overlying major OTU abundances between depths reinforces con-
waters were never investigated. Analysis of ARISA frag- ductivity as the dominant driver to community structure.
ments in this study has revealed a statistically significant The Genus Hydrogenophaga was identified as a major
difference (ANOSIM) between pond microbial communities OTU in the surface sample, decreasing significantly with
for Bacteria (R = 0.766, P = 0.001) and Cyanobacteria depth along the increasing conductivity gradient. Several
(R = 0.8605, P = 0.001). The highly geochemically strati- species of this genus have been previously isolated from a
fied Egg pond formed two distinct bacterial communities number of Arctic and Antarctic environments (Newdell &
(R = 0.975, P = 0.008), with BEST analysis indicating pH Rutter, 1994; VanTrappen et al., 2002) and found to have
and conductivity as major contributors to community a maximum NaCl tolerance of 1% (Yoon et al., 2008)
differences along this steep gradient. Pyrosequencing of explaining this observation. Algoriphagus yeomjeoni, previ-
representative samples from Egg pond identified a diverse ously identified in a cold marine environment and in
community of 583 unique OTUs, which exhibit notably Antarctic lake mats (Bowman et al., 2003; Van Trappen
different community structures with depth reinforcing the et al., 2004), has a broader salt tolerance of up to 9%
conclusions based on ARISA results. Although sequencing NaCl and an optimum of 1% (Yoon et al., 2005). Its
coverage was not 100% in any sample for this investiga- abundance is highest at the surface then decreases signifi-
tion, this would not affect conclusions formed from this cantly with depth; however, it remains at higher concen-
data, only resulting in a slight underestimation of diver- trations than Hydrogenophaga in the 40 and 48 cm
sity. samples likely due to its greater salt tolerance. The genus
The relative beta diversity of Cyanobacteria and Bacte- Psychromonas, previously isolated from pond sediments at
ria (ARISA) was generally pond-specific (visualised using Bratina Island (Mountfort et al., 1998) and Arctic ice
a MDS plot, confirmed statistically by ANOSIM). This (Auman et al., 2006), was dominant at both 40 and
observation is consistent with microscopic analysis of 48 cm and was barely detected at 0 cm, reflective of its
planktonic communities (James et al., 1995), which classi- ability to grow well between 1 and 10% NaCl (Auman
fied three groups of ponds based on size, conductivity, et al., 2006). The second most abundant OTU at 48 cm
and the presence and abundance of various planktonic further reinforces halophilic dominance with depth, shar-
groups. Each of the three groups included ponds sampled ing a high identity to sequences gained from a hypersa-
in this study (P70E, Legin and Salt). This suggests that line solar saltern (Isenbarger et al., 2008).
these ponds consistently harbour different geochemical Combinations of pH and conductivity with tempera-
environments and biological communities. The exception ture, V, NO , Ag and dissolved oxygen are all highly cor-
2
to this observation was Egg pond, which formed two related to microbial community structure. Each identified
depth-related microbial groupings (Fig. 4a, confirmed by variable has previously reported to have some influence
ANOSIM) consistent with a steep geochemical gradient of on planktonic microorganisms. Vanadium can interfere
conductivity and pH, both found to be drivers of with iron uptake (Baysse et al., 2000); Ag has well-known
community shifts in other studies (Sabbe et al., 2004; de antimicrobial properties (Throback et al., 2007); nitrogen
Figueiredo et al., 2010). limitation alters primary productivity in phytoplankton
BEST analysis has previously been successfully utilised to population in these ponds (Sorrell et al., 2013); hyper-
link geochemical variability to variation in biological variation in oxygen concentration with depth selects for
community profiles from fingerprinting data (Wood zones of aerobic, anaerobic (Bell & Laybourn-Parry,
et al., 2008; Soo et al., 2009; Smith et al., 2010). In this 1999) and hyperoxic tolerant microorganisms (Fridovich,
study, BEST analysis indicated a strong correlation between 1998) and due to variable optimum growth temperatures
differences in conductivity contributing to cyanobacterial for microorganisms minor variations in temperature,
and bacterial community structure for all five ponds stud- especially at near-freezing temperatures can dramatically
ied. In addition, independent analysis of Egg indicated influence community structure (Mountfort et al., 2003).
that both bacterial and cyanobacterial community struc- Regardless of biological mechanism, the extent of differ-
ture was significantly correlated to differences in pH and ences in V, Ag and NO concentration are well below
2
conductivity. The pH and conductivity in Egg show an those found to have biological significance (Yamanaka
inverse relationship to one another, so neither can be sin- et al., 2005; Sorrell et al., 2013), and their low correlation
gled out as the primary variable responsible for commu- scores individually indicate that it is unlikely that any of
nity structure changes; however, both have previously these variables will strongly influence microbial commu-
been identified as drivers of community structure in nity structure compared with pH and conductivity. DO
aquatic environments (James et al., 1995; Sabbe et al., and temperature are capable of influencing community
2004; Kaartokallio et al., 2005; de Figueiredo et al., 2010; structure; however, they too provide weak correlations
ª2014FederationofEuropeanMicrobiologicalSocieties. FEMSMicrobiolEcol&&(2014)1–14
PublishedbyJohnWiley&SonsLtd.Allrightsreserved